by
H. L. Royden
Contents
1 Set Theory 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Unions, intersections and complements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4 Algebras of sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 The axiom of choice and inﬁnite direct products . . . . . . . . . . . . . . . . . . . . . . . 2
1.6 Countable sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.7 Relations and equivalences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.8 Partial orderings and the maximal principle . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.9 Well ordering and the countable ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 The Real Number System 5
2.1 Axioms for the real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 The natural and rational numbers as subsets of R . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 The extended real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Sequences of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Open and closed sets of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.6 Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.7 Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3 Lebesgue Measure 13
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Outer measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Measurable sets and Lebesgue measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 A nonmeasurable set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Measurable functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.6 Littlewood’s three principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 The Lebesgue Integral 18
4.1 The Riemann integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 The Lebesgue integral of a bounded function over a set of ﬁnite measure . . . . . . . . . . 18
4.3 The integral of a nonnegative function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.4 The general Lebesgue integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.5 Convergence in measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5 Diﬀerentiation and Integration 22
5.1 Diﬀerentiation of monotone functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.2 Functions of bounded variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3 Diﬀerentiation of an integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.4 Absolute continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.5 Convex functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6 The Classical Banach Spaces 27
6.1 The L
p
spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.2 The Minkowski and H¨older inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.3 Convergence and completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.4 Approximation in L
p
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.5 Bounded linear functionals on the L
p
spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7 Metric Spaces 30
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.2 Open and closed sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.3 Continuous functions and homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.4 Convergence and completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.5 Uniform continuity and uniformity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.6 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.7 Compact metric spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.8 Baire category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.9 Absolute G
δ
’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7.10 The AscoliArzel´a Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
8 Topological Spaces 41
8.1 Fundamental notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
8.2 Bases and countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8.3 The separation axioms and continuous realvalued functions . . . . . . . . . . . . . . . . . 44
8.4 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
8.5 Products and direct unions of topological spaces . . . . . . . . . . . . . . . . . . . . . . . 48
8.6 Topological and uniform properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8.7 Nets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
9 Compact and Locally Compact Spaces 51
9.1 Compact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
9.2 Countable compactness and the BolzanoWeierstrass property . . . . . . . . . . . . . . . . 52
9.3 Products of compact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
9.4 Locally compact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
9.5 σcompact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
9.6 Paracompact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
9.7 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
9.8 The Stone
˘
Cech compactiﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
9.9 The StoneWeierstrass Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
10 Banach Spaces 60
ii
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
10.2 Linear operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
10.3 Linear functionals and the HahnBanach Theorem . . . . . . . . . . . . . . . . . . . . . . 62
10.4 The Closed Graph Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.5 Topological vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
10.6 Weak topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
10.7 Convexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
10.8 Hilbert space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
11 Measure and Integration 73
11.1 Measure spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
11.2 Measurable functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
11.3 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
11.4 General convergence theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
11.5 Signed measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
11.6 The RadonNikodym Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
11.7 The L
p
spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
12 Measure and Outer Measure 83
12.1 Outer measure and measurability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
12.2 The extension theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
12.3 The LebesgueStieltjes integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
12.4 Product measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
12.5 Integral operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
12.6 Inner measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
12.7 Extension by sets of measure zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
12.8 Carath´eodory outer measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
12.9 Hausdorﬀ measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
13 Measure and Topology 92
13.1 Baire sets and Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
iii
1 Set Theory
1.1 Introduction
1. If ¦x : x = x¦ = ∅, then there exists x such that x = x. Contradiction.
2. ∅ ⊂ ¦greeneyed lions¦.
3. X (Y Z) = ¦'x, 'y, z``¦, (X Y ) Z = ¦''x, y`, z`¦; 'x, 'y, z`` ↔ ''x, y`, z` ↔ 'x, y, z`.
4. Suppose P(1) is true and P(n) ⇒ P(n +1) for all n. Suppose that ¦n ∈ N : P(n) is false¦ = ∅. Then
it has a smallest element m. In particular, m > 1 and P(m) is false. But P(1) ⇒ P(2) ⇒ ⇒ P(m).
Contradiction.
5. Given a nonempty subset S of natural numbers, let P(n) be the proposition that if there exists m ∈ S
with m ≤ n, then S has a smallest element. P(1) is true since 1 will then be the smallest element of
S. Suppose that P(n) is true and that there exists m ∈ S with m ≤ n + 1. If m ≤ n, then S has a
smallest element by the induction hypothesis. If m = n+1, then either m is the smallest element of S or
there exists m
t
∈ S with m
t
< m = n + 1, in which case the induction hypothesis again gives a smallest
element.
1.2 Functions
6. (⇒) Suppose f is onetoone. For each y ∈ f[X], there exists a unique x
y
∈ X such that f(x
y
) = y.
Fix x
0
∈ X. Deﬁne g : Y → X such that g(y) = x
y
if y ∈ f[X] and g(y) = x
0
if y ∈ Y ` f[X]. Then g is
a welldeﬁned function and g ◦ f = id
X
.
(⇐) Suppose there exists g : Y → X such that g ◦ f = id
X
. If f(x
1
) = f(x
2
), then g(f(x
1
)) = g(f(x
2
)).
i.e. x
1
= x
2
. Thus f is onetoone.
7. (⇒) Suppose f is onto. For each y ∈ Y , there exists x
y
∈ X such that f(x
y
) = y. Deﬁne g : Y → X
such that g(y) = x
y
for all y ∈ Y . Then g is a welldeﬁned function and f ◦ g = id
Y
.
(⇐) Suppose there exists g : Y → X such that f ◦ g = id
Y
. Given y ∈ Y , g(y) ∈ X and f(g(y)) = y.
Thus f is onto.
8. Let P(n) be the proposition that for each n there is a unique ﬁnite sequence 'x
(n)
1
, . . . , s
(n)
n
` with
x
(n)
1
= a and x
(n)
i+1
= f
i
(x
(n)
1
, . . . , x
(n)
i
). Clearly P(1) is true. Given P(n), we see that P(n + 1) is true
by letting x
(n+1)
i
= x
(n)
i
for 1 ≤ i ≤ n and letting x
(n+1)
n+1
= f
n
(x
(n+1)
1
, . . . , x
(n+1)
n
). By letting x
n
= x
(n)
n
for each n, we get a unique sequence 'x
i
` from X such that x
1
= a and x
i+1
= f
i
(x
1
, . . . , x
i
).
1.3 Unions, intersections and complements
9. A ⊂ B ⇒ A ⊂ A ∩ B ⇒ A ∩ B = A ⇒ A ∪ B = (A ` B) ∪ (A ∩ B) ∪ (B ` A) = (A ∩ B) ∪ (B ` A) =
B ⇒ A ⊂ A∪ B = B.
10. x ∈ A∩(B∪C) ⇔ x ∈ A and x ∈ B or C ⇔ x ∈ A and B or x ∈ A and C ⇔ x ∈ (A∩B) ∪(A∩C).
Thus A∩(B∪C) = (A∩B) ∪(A∩C). x ∈ A∪(B∩C) ⇔ x ∈ A or x ∈ B and C ⇔ x ∈ A or B and x ∈
A or C ⇔ x ∈ (A∪ B) ∩ (A∪ C). Thus A∪ (B ∩ C) = (A∪ B) ∩ (A∪ C).
11. Suppose A ⊂ B. If x / ∈ B, then x / ∈ A. Thus B
c
⊂ A
c
. Conversely, if B
c
⊂ A
c
, then A = (A
c
)
c
⊂
(B
c
)
c
= B.
12a. A∆B = (A` B) ∪ (B ` A) = (B ` A) ∪ (A` B) = B∆A.
A∆(B∆C) = [A ` ((B ` C) ∪ (C ` B))] ∪ [((B ` C) ∪ (C ` B)) ` A] = [A ∩ ((B ` C) ∪ (C ` B))
c
] ∪ [((B `
C) ∪(C ` B)) ∩A
c
] = [A∩((B
c
∪C) ∩(C
c
∪B))] ∪[((B ∩C
c
) ∪(C ∩B
c
)) ∩A
c
] = [A∩(B
c
∪C) ∩(B ∪
C
c
)] ∪ (A
c
∩ B ∩ C
c
) ∪ (A
c
∩ B
c
∩ C) = (A∩ B
c
∩ C
c
) ∪ (A∩ B ∩ C) ∪ (A
c
∩ B ∩ C
c
) ∪ (A
c
∩ B
c
∩ C).
(A∆B)∆C = [((A` B) ∪(B` A)) ` C] ∪[C ` ((A` B) ∪(B` A))] = [((A` B) ∪(B` A)) ∩C
c
] ∪[C ∩((A`
B) ∪ (B ` A))
c
] = [((A ∩ B
c
) ∪ (B ∩ A
c
)) ∩ C
c
] ∪ [C ∩ ((A
c
∪ B) ∩ (B
c
∪ A))] = (A ∩ B
c
∩ C
c
) ∪ (A
c
∩
B ∩ C
c
) ∪ (A
c
∩ B
c
∩ C) ∪ (A∩ B ∩ C).
Hence A∆(B∆C) = (A∆B)∆C.
12b. A∆B = ∅ ⇔ (A` B) ∪ (B ` A) = ∅ ⇔ A` B = ∅ and B ` A = ∅ ⇔ A ⊂ B andB ⊂ A ⇔ A = B.
1
12c. A∆B = X ⇔ (A` B) ∪ (B ` A) = X ⇔ A∩ B = ∅ and A∪ B = X ⇔ A = B
c
.
12d. A∆∅ = (A` ∅) ∪ (∅ ` A) = A∪ ∅ = A; A∆X = (A` X) ∪ (X ` A) = ∅ ∪ A
c
= A
c
.
12e. (A∆B) ∩E = ((A` B) ∪(B` A)) ∩E = ((A` B) ∩E) ∪((B` A) ∩E) = [(A∩E) ` (B∩E)] ∪[(B∩
E) ` (A∩ E)] = (A∩ E)∆(B ∩ E).
13. x ∈ (
¸
A∈c
A)
c
⇔ x / ∈ A for any A ∈ ( ⇔ x ∈ A
c
for all A ∈ ( ⇔ x ∈
¸
A∈c
A
c
.
x ∈ (
¸
A∈c
A)
c
⇔ x / ∈
¸
A∈c
A ⇔ x ∈ A
c
for some A ∈ ( ⇔ x ∈
¸
A∈c
A
c
.
14. x ∈ B ∩ (
¸
A∈c
A) ⇔ x ∈ B and x ∈ A for some A ∈ ( ⇔ x ∈ B ∩ A for some A ∈ ( ⇔ x ∈
¸
A∈c
(B ∩ A). Thus B ∩ (
¸
A∈c
A) =
¸
A∈c
(B ∩ A).
15. (
¸
A∈A
A) ∩ (
¸
B∈B
B) =
¸
B∈B
((
¸
A∈A
A) ∩ B) =
¸
B∈B
(
¸
A∈A
(A∩ B)) =
¸
A∈A
¸
B∈B
(A∩ B).
16a. If x ∈
¸
A
λ
, then x ∈ A
λ
0
for some λ
0
and f(x) ∈ f[A
λ
0
] ⊂
¸
f[A
λ
]. Thus f[
¸
A
λ
] ⊂
¸
f[A
λ
].
Conversely, if y ∈
¸
f[A
λ
], then y ∈ f[A
λ
0
] for some λ
0
so y ∈ f[
¸
A
λ
]. Thus
¸
f[A
λ
] ⊂ f[
¸
A
λ
].
16b. If x ∈
¸
A
λ
, then x ∈ A
λ
for all λ and f(x) ∈ f[A
λ
] for all λ. Thus f(x) ∈
¸
f[A
λ
] and
f[
¸
A
λ
] ⊂
¸
f[A
λ
].
16c. Consider f : ¦1, 2, 3¦ → ¦1, 3¦ with f(1) = f(2) = 1 and f(3) = 3. Let A
1
= ¦1, 3¦ and A
2
= ¦2, 3¦.
Then f[A
1
∩ A
2
] = f[¦3¦] = ¦3¦ but f[A
1
] ∩ f[A
2
] = ¦1, 3¦.
17a. If x ∈ f
−1
[
¸
B
λ
], then f(x) ∈ B
λ
0
for some λ
0
so x ∈ f
−1
[B
λ
0
] ⊂
¸
f
−1
[B
λ
]. Thus f
−1
[
¸
B
λ
] ⊂
¸
f
−1
[B
λ
]. Conversely, if x ∈
¸
f
−1
[B
λ
], then x ∈ f
−1
[B
λ
0
] for some λ
0
so f(x) ∈ B
λ
0
⊂
¸
B
λ
and
x ∈ f
−1
[
¸
B
λ
].
17b. If x ∈ f
−1
[
¸
B
λ
], then f(x) ∈ B
λ
for all λ and x ∈ f
−1
[B
λ
] for all λ so x ∈
¸
f
−1
[B
λ
]. Thus
f
−1
[
¸
B
λ
] ⊂
¸
f
−1
[B
λ
]. Conversely, if x ∈
¸
f
−1
[B
λ
], then x ∈ f
−1
[B
λ
] for all λ and f(x) ∈
¸
B
λ
so
x ∈ f
−1
[
¸
B
λ
].
17c. If x ∈ f
−1
[B
c
], then f(x) / ∈ B so x / ∈ f
−1
[B]. i.e. x ∈ (f
−1
[B])
c
. Thus f
−1
[B
c
] ⊂ (f
−1
[B])
c
.
Conversely, if x ∈ (f
−1
[B])
c
, then x / ∈ f
−1
[B] so f(x) ∈ B
c
. i.e. x ∈ f
−1
[B
c
]. Thus (f
−1
[B])
c
⊂ f
−1
[B
c
].
18a. If y ∈ f[f
−1
[B]], then y = f(x) for some x ∈ f
−1
[B]. Since x ∈ f
−1
[B], f(x) ∈ B. i.e. y ∈ B.
Thus f[f
−1
[B]] ⊂ B. If x ∈ A, then f(x) ∈ f[A] so x ∈ f
−1
[f[A]]. Thus f
−1
[f[A]] ⊃ A.
18b. Consider f : ¦1, 2¦ → ¦1, 2¦ with f(1) = f(2) = 1. Let B = ¦1, 2¦. Then f[f
−1
[B]] = f[B] =
¦1¦ B. Let A = ¦1¦. Then f
−1
[f[A]] = f
−1
[¦1¦] = ¦1, 2¦ A.
18c. If y ∈ B, then there exists x ∈ X such that f(x) = y. In particular, x ∈ f
−1
[B] and y ∈ f[f
−1
[B]].
Thus B ⊂ f[f
−1
[B]]. This, together with the inequality in Q18a, gives equality.
1.4 Algebras of sets
19a. {(X) is a σalgebra containing (. Let T be the family of all σalgebras containing ( and let
/ =
¸
¦B : B ∈ T¦. Then / is a σalgebra containing (. Furthermore, by deﬁnition, if B is a σalgebra
containing (, then B ⊃ /.
19b. Let B
1
be the σalgebra generated by ( and let B
2
be the σalgebra generated by /. B
1
, being
a σalgebra, is also an algebra so B
1
⊃ /. Thus B
1
⊃ B
2
. Conversely, since ( ⊂ /, B
1
⊂ B
2
. Hence
B
1
= B
2
.
20. Let /
t
be the union of all σalgebras generated by countable subsets of (. If E ∈ (, then E is in the
σalgebra generated by ¦E¦. Thus ( ⊂ /
t
. If E
1
, E
2
∈ /
t
, then E
1
is in some σalgebra generated by
some countable subset (
1
of ( and E
2
is in some σalgebra generated by some countable subset (
2
of (.
Then E
1
∪ E
2
is in the σalgebra generated by the countable subset (
1
∪ (
2
so E
1
∪ E
2
∈ /
t
. If F ∈ /
t
,
then F is in some σalgebra generated by some countable set and so is F
c
. Thus F
c
∈ /
t
. Furthermore,
if 'E
i
` is a sequence in /
t
, then each E
i
is in some σalgebra generated by some countable subset (
i
of (. Then
¸
E
i
is in the σalgebra generated by the countable subset
¸
(
i
. Hence /
t
is a σalgebra
containing ( and it contains /.
1.5 The axiom of choice and inﬁnite direct products
21. For each y ∈ Y , let A
y
= f
−1
[¦y¦]. Consider the collection A = ¦A
y
: y ∈ Y ¦. Since f is onto,
A
y
= ∅ for all y. By the axiom of choice, there is a function F on A such that F(A
y
) ∈ A
y
for all y ∈ Y .
2
i.e. F(A
y
) ∈ f
−1
[¦y¦] so f(F(A
y
)) = y. Deﬁne g : Y → X, y → F(A
y
). Then f ◦ g = id
Y
.
1.6 Countable sets
22. Let E = ¦x
1
, . . . , x
n
¦ be a ﬁnite set and let A ⊂ E. If A = ∅, then A is ﬁnite by deﬁnition. If A = ∅,
choose x ∈ A. Deﬁne a new sequence 'y
1
, . . . , y
n
` by setting y
i
= x
i
if x
i
∈ A and y
i
= x if x
i
/ ∈ A. Then
A is the range of 'y
1
, . . . , y
n
` and is therefore ﬁnite.
23. Consider the mapping 'p, q, 1` → p/q, 'p, q, 2` → −p/q, '1, 1, 3` → 0. Its domain is a subset of the
set of ﬁnite sequences from N, which is countable by Propositions 4 and 5. Thus its range, the set of
rational numbers, is countable.
24. Let f be a function from N to E. Then f(v) = 'a
vn
`
∞
n=1
with a
vn
= 0 or 1 for each n. Let
b
v
= 1 −a
vv
for each v. Then 'b
n
` ∈ E but 'b
n
` = 'a
vn
` for any v. Thus E cannot be the range of any
function from N and E is uncountable.
25. Let E = ¦x : x / ∈ f(x)¦ ⊂ X. If E is in the range of f, then E = f(x
0
) for some x
0
∈ X. Now if
x
0
/ ∈ E, then x
0
∈ f(x
0
) = E. Contradiction. Similarly when x
0
∈ E. Hence E is not in the range of f.
26. Let X be an inﬁnite set. By the axiom of choice, there is a choice function F : {(X) ` ¦∅¦ → X.
Pick a ∈ X. For each n ∈ N, let f
n
: X
n
→ X be deﬁned by f
n
(x
1
, . . . , x
n
) = F(X ` ¦x
1
, . . . , x
k
¦). By
the generalised principle of recursive deﬁnition, there exists a unique sequence 'x
i
` from X such that
x
1
= a, x
i+1
= f
i
(x
1
, . . . , x
i
). In particular, x
i+1
= F(X ` ¦x
1
, . . . , x
i
¦) ∈ X ` ¦x
1
, . . . , x
i
¦ so x
i
= x
j
if
i = j and the range of the sequence 'x
i
` is a countably inﬁnite subset of X.
1.7 Relations and equivalences
27. Let F, G ∈ Q = X/ ≡. Choose x
1
, x
2
∈ F and y
1
, y
2
∈ G. Then x
1
≡ x
2
and y
1
≡ y
2
so
x
1
+x
2
≡ y
1
+y
2
. Thus y
1
+y
2
∈ E
x
1
+x
2
and E
x
1
+x
2
= E
y
1
+y
2
.
28. Suppose ≡ is compatible with +. Then x ≡ x
t
implies x + y ≡ x
t
+ y since y ≡ y. Conversely,
suppose x ≡ x
t
implies x + y ≡ x
t
+ y. Now suppose x ≡ x
t
and y ≡ y
t
. Then x + y ≡ x + y
t
and
x +y
t
≡ x
t
+y
t
so x +y ≡ x
t
+y
t
.
Let E
1
, E
2
, E
3
∈ Q = X/ ≡. Choose x
i
∈ E
i
for i = 1, 2, 3. Then (E
1
+ E
2
) + E
3
= E
x
1
+x
2
+
E
3
= E
(x
1
+x
2
)+x
3
and E
1
+ (E
2
+ E
2
) = E
1
+ E
x
2
+x
3
= E
x
1
+(x
2
+x
3
)
. Since X is a group under +,
(x
1
+ x
2
) + x
3
= x
1
+ (x
2
+ x
3
) so + is associative on Q. Let 0 be the identity of X. For any F ∈ Q,
choose x ∈ F. Then F + E
0
= E
x+0
= E
x
= F. Similarly for E
0
+ F. Thus E
0
is the identity of Q.
Let F ∈ Q and choose x ∈ F and let −x be the inverse of x in X. Then F + E
−x
= E
x+(−x)
= E
0
.
Similarly for E
−x
+ F. Thus E
−x
is the inverse of F in Q. Hence the induced operation + makes the
quotient space Q into a group.
1.8 Partial orderings and the maximal principle
29. Given a partial order ≺ on X, deﬁne x < y if x ≺ y and x = y. Also deﬁne x ≤ y if x ≺ y or
x = y. Then < is a strict partial order on X and ≤ is a reﬂexive partial order on X. Furthermore,
x ≺ y ⇔ x < y ⇔ x ≤ y for x = y. Uniqueness follows from the deﬁnitions.
30. Consider the set (0, 1] ∪ [2, 3) with the ordering ≺ given by x ≺ y if and only if either x, y ∈ (0, 1]
and x < y, or x, y ∈ [2, 3) and x < y. Then ≺ is a partial ordering on the set, 2 is the unique minimal
element and there is no smallest element.
1.9 Well ordering and the countable ordinals
31a. Let X be a wellordered set and let A ⊂ X. Any strict linear ordering on X is also a strict linear
ordering on A and every nonempty subset of A is also a nonempty subset of X. Thus any nonempty
subset of A has a smallest element.
31b. Let < be a partial order on X with the property that every nonempty subset has a least element.
For any two elements x, y ∈ X, the set ¦x, y¦ has a least element. If the least element is x, then x < y.
If the least element is y, then y < x. Thus < is a linear ordering on X with the property that every
3
nonempty subset has a least element and consequently a well ordering.
32. Let Y = ¦x : x < Ω¦ and let E be a countable subset of Y . Y is uncountable so Y ` E is nonempty.
If for every y ∈ Y ` E there exists x
y
∈ E such that y < x
y
, then y → x
y
deﬁnes a mapping from Y ` E
into the countable set E so Y ` E is countable. Contradiction. Thus there exists y ∈ Y ` E such that
x < y for all x ∈ E. i.e. E has an upper bound in Y . Consider the set of upper bounds of E in Y . This
is a nonempty subset of Y so it has a least element, which is then a least upper bound of E.
33. Let ¦S
λ
: λ ∈ Λ¦ be a collection of segments. Suppose
¸
S
λ
= X. Then each S
λ
is of the form
¦x ∈ X : x < y
λ
¦ for some y
λ
∈ X. X `
¸
S
λ
is a nonempty subset of the wellordered set X so it
has a least element y
0
. If y
0
< y
λ
for some λ, then y
0
∈ S
λ
. Contradiction. Thus y
0
≥ y
λ
for all λ.
Clearly,
¸
S
λ
⊂ ¦x ∈ X : x < y
0
¦. Conversely, if x < y
0
, then x / ∈ X `
¸
S
λ
so x ∈
¸
S
λ
. Hence
¸
S
λ
= ¦x ∈ X : x < y
0
¦.
34a. Suppose f and g are distinct successorpreserving maps from X into Y . Then ¦x ∈ X : f(x) = g(x)¦
is a nonempty subset of X and has a least element x
0
. Now g
x
0
is the ﬁrst element of Y not in
g[¦z : z < x
0
¦] = f[¦z : z < x
0
¦] and so is f(x
0
). Thus f(x
0
) = g(x
0
). Contradiction. Hence there is at
most one successorpreserving map from X into Y .
34b. Let f be a successorpreserving map from X into Y . Suppose f[X] = Y . Then ¦y : y / ∈ f[X]¦ is a
nonempty subset of Y and has a least element y
0
. Clearly, ¦y : y < y
0
¦ ⊂ f[X]. Conversely, if y ∈ f[X],
then y = f(x) is the ﬁrst element not in f[¦z : z < x¦]. Since y
0
/ ∈ f[X] ⊃ f[¦z : z < x¦], y < y
0
. Thus
f[X] ⊂ ¦y : y < y
0
¦. Hence f[X] = ¦y : y < y
0
¦.
34c. Suppose f is successorpreserving. Let x, y ∈ X with x < y. f(y) is the ﬁrst element of Y
not in f[¦z : z < y¦] so f(y) = f(x). Furthermore, if f(y) < f(x), then f(y) ∈ f[¦z : z < x¦] and
y < x. Contradiction. Thus f(x) < f(y). Hence f is a onetoone order preserving map. Since f is
onetoone, f
−1
is deﬁned on f[X]. If f(x
1
) < f(x
2
), then x
1
< x
2
since f is order preserving. i.e.
f
−1
(f(x
1
)) < f
−1
(f(x
2
)). Thus f
−1
is order preserving.
34d. Let S = ¦z : z < x¦ be a segment of X and let z ∈ S. f[
S
(z) = f(z) is the ﬁrst element of Y not
in f[¦w : w < z¦]. Thus f[
S
(z) / ∈ f[
S
[¦w : w < z¦]. Suppose y < f[
S
(z). Then y ∈ f[¦w : w < z¦] so
y = f(w) for some w < z. Thus w ∈ S and y = f[
S
(w) ∈ f[
S
[¦w : w < z¦]. Hence f[
S
(z) is the ﬁrst
element of Y not in f[
S
[¦w : w < z¦].
34e. Consider the collection of all segments of X on which there is a successorpreserving map into
Y . Let S be the union of all such segments S
λ
and let f
λ
be the corresponding map on S
λ
. Given
s ∈ S, deﬁne f(s) = f
λ
(s) if s ∈ S
λ
. f is a welldeﬁned map because if s belongs to two segments S
λ
and S
µ
, we may assume S
λ
⊂ S
µ
so that f
µ
[
S
λ
= f
λ
by parts (d) and (a). Also, f(s) = f
λ
(s) is the
ﬁrst element of Y not in f
λ
[¦z : z < s¦] = f[¦z : z < s¦]. Thus f is a successorpreserving map from
S into Y . By Q33, S is a segment so either S = X or S = ¦x : x < x
0
¦ for some x
0
∈ X. In the
second case, suppose f[S] = Y . Then there is a least y
0
/ ∈ Y ` f[S]. Consider S ∪ ¦x
0
¦. If there is no
x ∈ X such that x > x
0
, then S ∪ ¦x
0
¦ = X. Otherwise, ¦x ∈ X : x > x
0
¦ has a least element x
t
and
S ∪ ¦x
0
¦ = ¦x : x < x
t
¦. Thus S ∪ ¦x
0
¦ is a segment of X. Deﬁne f(x
0
) = y
0
. Then f(x
0
) is the ﬁrst
element of Y not in f[¦x : x < x
0
¦]. Thus f is a successorpreserving map on the segment S ∪ ¦x
0
¦
so S ∪ ¦x
0
¦ ⊂ S. Contradiction. Hence f[S] = Y . Now if S = X, then by part (b), f[X] = f[S] is
a segment of Y . On the other hand, if f[S] = Y , then f
−1
is a successorpreserving map on Y and
f
−1
[Y ] = S is a segment of X.
34f. Let X be the wellordered set in Proposition 8 and let Y be an uncountable set wellordered by ≺
such that there is a last element Ω
Y
in Y and if y ∈ Y and y = Ω
Y
, then ¦z ∈ Y : z < y¦ is countable.
By part (e), we may assume there is a successorpreserving map f from X onto a segment of Y . If
f[X] = Y , then we are done. Otherwise, f[X] = ¦z : z < y
0
¦ for some y
0
∈ Y . If y
0
= Ω
Y
, then f[X] is
countable and since f is onetoone, X is countable. Contradiction. If y
0
= Ω
Y
, then f(Ω
X
) < Ω
Y
and
f(Ω
X
) is the largest element in f[X]. i.e. f[X] = ¦z : z ≤ f(Ω
X
)¦, which is countable. Contradiction.
Hence f is an order preserving bijection from X onto Y .
4
2 The Real Number System
2.1 Axioms for the real numbers
1. Suppose 1 / ∈ P. Then −1 ∈ P. Now take x ∈ P. Then −x = (−1)x ∈ P so 0 = x + (−x) ∈ P.
Contradiction.
2. Let S be a nonempty set of real numbers with a lower bound. Then the set −S = ¦−s : s ∈ S¦ has
an upper bound and by Axiom C, it has a least upper bound b. −b is a lower bound for S and if a is
another lower bound for S, then −a is an upper bound for −S and b ≤ −a so −b ≥ a. Thus −b is a
greatest lower bound for S.
3. Let L and U be nonempty subsets of R with R = L ∪ U and such that for each l ∈ L and u ∈ U we
have l < u. L is bounded above so it has a least upper bound l
0
. Similarly, U is bounded below so it
has a greatest lower bound u
0
. If l
0
∈ L, then it is the greatest element in L. Otherwise, l
0
∈ U and
u
0
≤ l
0
. If u
0
< l
0
, then there exists l ∈ L with u
0
< l. Thus l is a lower bound for U and is greater
than u
0
. Contradiction. Hence u
0
= l
0
so u
0
∈ U and it is the least element in U.
4a.
x ≤ y ≤ z x ≤ z ≤ y y ≤ x ≤ z y ≤ z ≤ x z ≤ x ≤ y z ≤ y ≤ x
x ∧ y x x y y x y
y ∧ z y z y y z z
(x ∧ y) ∧ z x ∧ z = x x ∧ z = x y ∧ z = y y ∧ z = y x ∧ z = z y ∧ z = z
x ∧ (y ∧ z) x ∧ y = x x ∧ z = x x ∧ y = y x ∧ y = y x ∧ z = z x ∧ z = z
4b. Suppose x ≤ y. Then x ∧ y = x and x ∨ y = y so x ∧ y +x ∨ y = x +y. Similarly for y ≤ x.
4c. Suppose x ≤ y. Then −y ≤ −x so (−x) ∧ (−y) = −y = −(x ∨ y). Similarly for y ≤ x.
4d. Suppose x ≤ y. Then x +z ≤ y +z so x ∨ y +z = y +z = (x +z) ∨ (y +z). Similarly for y ≤ x.
4e. Suppose x ≤ y. Then zx ≤ zy if z ≥ 0 so z(x ∨ y) = zy = (zx) ∨ (zy). Similarly for y ≤ x.
5a. If x, y ≥ 0, then xy ≥ 0 so [xy[ = xy = [x[[y[. If x, y < 0, then xy > 0 so [xy[ = xy = (−x)(−y) =
[x[[y[. If x ≥ 0 and y < 0, then xy ≤ 0 so [xy[ = −xy = x(−y) = [x[[y[. Similarly when x < 0 and y ≥ 0.
5b. If x, y ≥ 0, then x + y ≥ 0 so [x + y[ = x + y = [x[ + [y[. If x, y < 0, then x + y < 0 so
[x + y[ = −x −y = [x[ +[y[. If x ≥ 0, y < 0 and x + y ≥ 0, then [x + y[ = x + y ≤ x −y = [x[ +[y[. If
x ≥ 0, y < 0 and x +y < 0, then [x +y[ = −x −y ≤ x −y = [x[ +[y[. Similarly when x < 0 and y ≥ 0.
5c. If x ≥ 0, then x ≥ −x so [x[ = x = x ∨ (−x). Similarly for x < 0.
5d. If x ≥ y, then x −y ≥ 0 so x ∨ y = x =
1
2
(x +y +x −y) =
1
2
(x +y +[x −y[). Similarly for y ≥ x.
5e. Suppose −y ≤ x ≤ y. If x ≥ 0, then [x[ = x ≤ y. If x < 0, then [x[ = −x ≤ y since −y ≤ x.
2.2 The natural and rational numbers as subsets of R
No problems
2.3 The extended real numbers
6. If E = ∅, then sup E = −∞ and inf E = ∞ so sup E < inf E. If E = ∅, say x ∈ E, then
inf E ≤ x ≤ sup E.
2.4 Sequences of real numbers
7. Suppose l
1
and l
2
are (ﬁnite) limits of a sequence 'x
n
`. Given ε > 0, there exist N
1
and N
2
such that
[x
n
−l
1
[ < ε for n ≥ N
1
and [x
n
−l
2
[ < ε for n ≥ N
2
. Then [l
1
−l
2
[ < 2ε for n ≥ max(N
1
, N
2
). Since ε
is arbitrary, [l
1
−l
2
[ = 0 and l
1
= l
2
. The cases where ∞ or −∞ is a limit are clear.
8. Suppose there is a subsequence 'x
n
j
` that converges to l ∈ R. Then given ε > 0, there exists N
t
such
that [x
n
j
−l[ < ε for j ≥ N
t
. Given N, choose j ≥ N
t
such that n
j
≥ N. Then [x
n
j
−l[ < ε and l is a
cluster point of 'x
n
`. Conversely, suppose l ∈ R is a cluster point of 'x
n
`. Given ε > 0, there exists n
1
≥ 1
5
such that [x
n
1
− l[ < ε. Suppose x
n
1
, . . . , x
n
k
have been chosen. Choose n
k+1
≥ 1 + max(x
n
1
, . . . , x
n
k
)
such that [x
n
k+1
−l[ < ε. Then the subsequence 'x
n
j
` converges to l. The cases where l is ∞ or −∞ are
similarly dealt with.
9a. Let l = limx
n
∈ R. Given ε > 0, there exists n
1
such that x
k
< l + ε for k ≥ n
1
. Also, there
exists k
1
≥ n
1
such that x
k
1
> l −ε. Thus [x
k
1
−l[ < ε. Suppose n
1
, . . . , n
j
and x
k
1
, . . . , x
k
j
have been
chosen. Choose n
j+1
> max(k
1
, . . . , k
j
). Then x
k
< l + ε for k ≥ n
j+1
. Also, there exists k
j+1
≥ n
j+1
such that x
k
j+1
> l − ε. Thus [x
k
j+1
− l[ < ε. Then the subsequence 'x
k
j
` converges to l and l is a
cluster point of 'x
n
`. If l is ∞ or −∞, it follows from the deﬁnitions that l is a cluster point of 'x
n
`.
If l
t
is a cluster point of 'x
n
` and l
t
> l, we may assume that l ∈ R. By deﬁnition, there exists N such
that sup
k≥N
x
k
< (l + l
t
)/2. If l
t
∈ R, then since it is a cluster point, there exists k ≥ N such that
[x
k
− l
t
[ < (l
t
− l)/2 so x
k
> (l + l
t
)/2. Contradiction. By deﬁnition again, there exists N
t
such that
sup
k≥N
x
k
< l + 1. If l
t
= ∞, then there exists k ≥ N
t
such that x
k
≥ l + 1. Contradiction. Hence l is
the largest cluster point of 'x
n
`. Similarly for limx
n
.
9b. Let 'x
n
` be a bounded sequence. Then limx
n
≤ sup x
n
< ∞. Thus limx
n
is a ﬁnite real number
and by part (a), there is a subsequence converging to it.
10. If 'x
n
` converges to l, then l is a cluster point. If l
t
= l and l
t
is a cluster point, then there is
a subsequence of 'x
n
` converging to l
t
. Contradiction. Thus l is the only cluster point. Conversely,
suppose there is exactly one extended real number x that is a cluster point of 'x
n
`. If x ∈ R, then
limx
n
= limx
n
= x. Given ε > 0, there exists N such that sup
k≥N
x
k
< x +ε and there exists N
t
such
that inf
k≥N
x
k
> x − ε. Thus [x
k
− x[ < ε for k ≥ max(N, N
t
) and 'x
n
` converges to x. If x = ∞,
then limx
n
= ∞ so for any ∆ there exists N such that inf
k≥N
x
k
> ∆. i.e. x
k
> ∆ for k ≥ N. Thus
limx
n
= ∞. Similarly when x = −∞.
The statement does not hold when the word “extended” is removed. The sequence '1, 1, 1, 2, 1, 3, 1, 4, . . .`
has exactly one real number 1 that is a cluster point but it does not converge.
11a. Suppose 'x
n
` converges to l ∈ R. Given ε > 0, there exists N such that [x
n
−l[ < ε/2 for n ≥ N.
Now if m, n ≥ N, [x
n
−x
m
[ ≤ [x
n
−l[ +[x
m
−l[ < ε. Thus 'x
n
` is a Cauchy sequence.
11b. Let 'x
n
` be a Cauchy sequence. Then there exists N such that [[x
n
[ −[x
m
[[ ≤ [x
n
− x
m
[ < 1 for
m, n ≥ N. In particular, [[x
n
[ −[x
N
[[ < 1 for n ≥ N. Thus the sequence '[x
n
[` is bounded above by
max([x
1
[, . . . , [x
N−1
[, [x
N
[ + 1) so the sequence 'x
n
` is bounded.
11c. Suppose 'x
n
` is a Cauchy sequence with a subsequence 'x
n
k
` that converges to l. Given ε > 0,
there exists N such that [x
n
− x
m
[ < ε/2 for m, n ≥ N and [x
n
k
− l[ < ε/2 for n
k
≥ N. Now choose k
such that n
k
≥ N. Then [x
n
−l[ ≤ [x
n
−x
n
k
[ +[x
n
k
−l[ < ε for n ≥ N. Thus 'x
n
` converges to l.
11d. If 'x
n
` is a Cauchy sequence, then it is bounded by part (b) so it has a subsequence that converges
to a real number l by Q9b. By part (c), 'x
n
` converges to l. The converse holds by part (a).
12. If x = limx
n
, then every subsequence of 'x
n
` also converges to x. Conversely, suppose every
subsequence of 'x
n
` has in turn a subsequence that converges to x. If x ∈ R and 'x
n
` does not converge
to x, then there exists ε > 0 such that for all N, there exists n ≥ N with [x
n
−x[ ≥ ε. This gives rise to
a subsequence 'x
n
k
` with [x
n
k
−x[ ≥ ε for all k. This subsequence will not have a further subsequence
that converges to x. If x = ∞ and limx
n
= ∞, then there exists ∆ such that for all N, there exists
n ≥ N with x
n
< ∆. This gives rise to a subsequence 'x
n
k
` with x
n
k
< ∆ for all k. This subsequence
will not have a further subsequence with limit ∞. Similarly when x = −∞.
13. Suppose l = limx
n
. Given ε, there exists n such that sup
k≥n
x
k
< l +ε so x
k
< l +ε for all k ≥ n.
Furthermore, given ε and n, sup
k≥n
x
k
> l − ε so there exists k ≥ n such that x
k
> l − ε. Conversely,
suppose conditions (i) and (ii) hold. By (ii), for any ε and n, sup
k≥n
x
k
≥ l −ε. Thus sup
k≥n
x
k
≥ l for
all n. Furthermore by (i), if l
t
> l, then there exists n such that x
k
< l
t
for all k ≥ n. i.e. sup
k≥n
x
k
≤ l
t
.
Hence l = inf
n
sup
k≥n
x
k
= limx
n
.
14. Suppose limx
n
= ∞. Then given ∆ and n, sup
k≥n
x
k
> ∆. Thus there exists k ≥ n such that
x
k
> ∆. Conversely, suppose that given ∆ and n, there exists k ≥ n such that x
k
> ∆. Let x
n
1
= x
1
and let n
k+1
> n
k
be chosen such that x
n
k+1
> x
n
k
. Then limx
n
k
= ∞ so limx
n
= ∞.
15. For all m < n, inf
k≥m
x
k
≤ inf
k≥n
x
k
≤ sup
k≥n
x
k
. Also, inf
k≥n
x
k
≤ sup
k≥n
x
k
≤ sup
k≥m
x
k
. Thus
inf
k≥m
x
k
≤ sup
k≥n
x
k
whenever m = n. Hence limx
n
= sup
n
inf
k≥n
x
k
≤ inf
n
sup
k≥n
x
k
= limx
n
.
If limx
n
= limx
n
= l, then the sequence has exactly one extended real number that is a cluster point
6
so it converges to l by Q10. Conversely, if l = limx
n
and limx
n
< limx
n
, then the sequence has a
subsequence converging to limx
n
and another subsequence converging to limx
n
. Contradiction. Thus
limx
n
= limx
n
= l.
16. For all n, x
k
+y
k
≤ sup
k≥n
x
k
+sup
k≥n
y
k
for k ≥ n. Thus sup
k≥n
(x
k
+y
k
) ≤ sup
k≥n
x
k
+sup
k≥n
y
k
for all n. Then inf
n
sup
k≥n
(x
k
+y
k
) ≤ inf
n
sup
k≥n
x
k
+inf
n
sup
k≥n
y
k
. i.e. lim(x
n
+y
n
) ≤ limx
n
+limy
n
.
Now limx
n
≤ lim(x
n
+y
n
) + lim(−y
n
) = lim(x
n
+y
n
) −limy
n
. Thus limx
n
+ limy
n
≤ lim(x
n
+y
n
).
17. For any n and any k ≥ n, x
k
y
k
≤ sup
k≥n
x
k
sup
k≥n
y
k
. Thus for all n, sup
k≥n
x
k
> 0 and
sup
k≥n
x
k
y
k
≤ sup
k≥n
x
k
sup
k≥n
y
k
. Now, taking limits, we get limx
n
y
n
≤ limx
n
limy
n
.
18. Since each x
v
≥ 0, the sequence 's
n
` is monotone increasing. If the sequence 's
n
` is bounded,
then lims
n
= sup s
n
∈ R so sup s
n
=
¸
∞
v=1
x
v
. If the sequence 's
n
` is unbounded, then lims
n
= ∞ so
∞ =
¸
∞
v=1
x
v
.
19. For each n, let t
n
=
¸
n
v=1
[x
v
[. Since
¸
∞
v=1
[x
v
[ < ∞, the sequence 't
n
` is a Cauchy sequence so
given ε, there exists N such that [t
n
− t
m
[ < ε for n > m ≥ N. Then [s
n
− s
m
[ = [x
m+1
+ + x
n
[ ≤
[x
m+1
[ + + [x
n
[ = [t
n
− t
m
[ < ε for n > m ≥ N. Thus the sequence 's
n
` is a Cauchy sequence so it
converges and 'x
n
` has a sum.
20. x
1
+
¸
∞
v=1
(x
v+1
−x
v
) = x
1
+ lim
n
¸
n
v=1
(x
v+1
−x
v
) = lim
n
[x
1
+
¸
n
v=1
(x
v+1
−x
v
)] = lim
n
x
n+1
=
lim
n
x
n
. Thus x = lim
n
x
n
if and only if x = x
1
+
¸
∞
v=1
(x
v+1
−x
v
).
21a. Suppose
¸
x∈E
x < ∞. For each n, let E
n
= ¦x ∈ E : x ≥ 1/n¦. Then each E
n
is a ﬁnite subset of
E. Otherwise, if E
n
0
is an inﬁnite set for some n
0
, then letting F
k
be a subset of E
n
0
with kn
0
elements
for each k ∈ N, s
F
k
≥ k. Then
¸
x∈E
x ≥ s
F
k
≥ k for each k. Contradiction. Now E =
¸
E
n
so E is
countable.
21b. Clearly, ¦x
1
, . . . , x
n
¦ ∈ T for all n. Thus sup s
n
≤ sup
F∈J
s
F
. On the other hand, given
F ∈ T, there exists n such that F ⊂ ¦x
1
, . . . , x
n
¦ so s
F
≤ s
n
and sup
F∈J
s
F
≤ sup s
n
. Hence
¸
x∈E
x = sup
F∈J
s
F
= sup s
n
=
¸
∞
n=1
x
n
.
22. Given x ∈ R, let a
1
be the largest integer such that 0 ≤ a
1
< p and a
1
/p ≤ x. Suppose a
1
, . . . , a
n
have
been chosen. Let a
n+1
be the largest integer such that 0 ≤ a
n+1
< p and a
n+1
/p
n+1
≤ x −
¸
n
k=1
a
k
/p
k
.
This gives rise to a sequence 'a
n
` of integers with 0 ≤ a
n
< p and x −
¸
n
k=1
a
k
/p
k
< 1/p
n
for all n.
Now given ε, there exists N such that 1/p
N
< ε. Then [x −
¸
n
k=1
a
n
/p
n
[ < 1/p
N
< ε for n ≥ N. Thus
x =
¸
∞
n=1
a
n
/p
n
. This sequence is unique by construction. When x = q/p
n
with q ∈ ¦1, . . . , p −1¦, the
sequence 'a
n
` obtained in this way is such that a
n
= q and a
m
= 0 for m = n. However the sequence
'b
n
` with b
m
= 0 for m < n, b
n
= q −1 and b
m
= p −1 for m > n also satisﬁes x =
¸
∞
n=1
b
n
.
Conversely, if 'a
n
` is a sequence of integers with 0 ≤ a
n
< p, let s
n
=
¸
n
k=1
a
k
/p
k
. Then 0 ≤ s
n
≤
(p − 1)
¸
∞
k=1
1/p
k
= 1 for all n. Thus 's
n
` is a bounded monotone increasing sequence so it converges.
Furthermore, since 0 ≤ s
n
≤ 1 for all n, the sequence converges to a real number x with 0 ≤ x ≤ 1.
23. Given a real number x with 0 ≤ x ≤ 1, form its binary expansion (by taking p = 2 in Q22), which
we may regard as unique by ﬁxing a way of representing those numbers of the form q/2
n
. By Q22, this
gives a bijection from [0, 1] to the set of inﬁnite sequences from ¦0, 1¦, which is uncountable by Q1.24.
Thus [0, 1] is uncountable and since R ⊃ [0, 1], R is uncountable.
2.5 Open and closed sets of real numbers
24. The set of rational numbers is neither open nor closed. For each x ∈ Q
c
and for each δ > 0, there is
a rational number r with x < r < x + δ so Q
c
is not open and Q is not closed. On the other hand, for
each x ∈ Q and for each δ > 0, there is an irrational number s with x < s < x +δ so Q is not open.
(*) Given x, y ∈ R with x < y, there exists r ∈ Q such that x/
√
2 < r < y/
√
2. We may assume 0 / ∈ (x, y)
by taking x = 0 if necessary. Then r = 0 so r
√
2 is irrational and x < r
√
2 < y.
25. ∅ and R are both open and closed. Suppose X is a nonempty subset of R that is both open and
closed. Take x ∈ X. Since X is open, there exists δ > 0 such that (x − δ, x + δ) ⊂ X. Thus the set
S = ¦y : [x, y) ⊂ X¦ is nonempty. Suppose [x, y) ⊂ X for some y > x. Then S is bounded above. Let
b = sup S. Then b ∈ S and b is a point of closure of X so b ∈ X since X is closed. But since X is open,
there exists δ
t
> 0 such that (b −δ
t
, b +δ
t
) ⊂ X. Then [x, b +δ
t
) = [x, b) ∪[b, b +δ
t
) ⊂ X. Contradiction.
Thus [x, y) ⊂ X for all y > x. Similarly, (z, x] ⊂ X for all z < x. Hence X = R.
7
26. Let A = (−1, 0) and B = (0, 1). Then A∩ B = ∅ but
¯
A∩
¯
B = [−1, 0] ∩ [0, 1] = ¦0¦.
27. Suppose x is a point of closure of E. Then for every n, there exists y
n
∈ E with [y
n
− x[ < 1/n.
Given ε > 0, choose N such that 1/N < ε. Then [y
n
− x[ < 1/N < ε for n ≥ N so 'y
n
` is a sequence
in E with x = limy
n
. Conversely, suppose there is a sequence 'y
n
` in E with x = limy
n
. Then for any
δ > 0, there exists N such that [x −y
n
[ < δ for n ≥ N. In particular, [x −y
N
[ < δ. Thus x is a point of
closure of E.
28. Let x be a point of closure of E
t
. Given δ > 0, there exists y ∈ E
t
such that [x − y[ < δ. If y = x,
then x ∈ E
t
and we are done. Suppose y = x. We may assume y > x. Let δ
t
= min(y − x, x + δ − y).
Then x / ∈ (y −δ
t
, y +δ
t
) and (y −δ
t
, y +δ
t
) ⊂ (x −δ, x +δ). Since y ∈ E
t
, there exists z ∈ E ` ¦y¦ such
that z ∈ (y −δ
t
, y +δ
t
). In particular, z ∈ E ` ¦x¦ and [z −x[ < δ. Thus x ∈ E
t
. Hence E
t
is closed.
29. Clearly E ⊂
¯
E and E
t
⊂
¯
E so E ∪ E
t
⊂
¯
E. Conversely, let x ∈
¯
E and suppose x / ∈ E. Then given
δ > 0, there exists y ∈ E such that [y −x[ < δ. Since x / ∈ E, y ∈ E ` ¦x¦. Thus x ∈ E
t
and
¯
E ⊂ E ∪E
t
.
30. Let E be an isolated set of real numbers. For any x ∈ E, there exists δ
x
> 0 such that [y −x[ ≥ δ
x
for all y ∈ E`¦x¦. We may take each δ
x
to be rational and let I
x
= ¦y : [y −x[ < δ
x
¦. Then ¦I
x
: x ∈ E¦
is a countable collection of open intervals, each I
x
contains only one element of E, namely x, and
E ⊂
¸
x∈E
I
x
. If E is uncountable, then I
x
0
will contain two elements of E for some x
0
. Contradiction.
Thus E is countable.
31. Let x ∈ R. Given δ > 0, there exists r ∈ Q such that x < r < x + δ. Thus x is an accumulation
point of Q. Hence Q
t
= R and
¯
Q = R.
32. Let F
1
and F
2
be closed sets. Then F
c
1
and F
c
2
are open so F
c
1
∩F
c
2
is open. i.e. (F
1
∪F
2
)
c
is open.
Thus F
1
∪ F
2
is closed. Let ( be a collection of closed sets. Then F
c
is open for any F ∈ ( so
¸
F∈J
F
c
is open. i.e. (
¸
F∈c
F)
c
is open. Thus
¸
F∈c
F is closed.
33. Let O
1
and O
2
be open sets. Then O
c
1
and O
c
2
are closed so O
c
1
∪ O
c
2
is closed. i.e. (O
1
∩ O
2
)
c
is
closed. Thus O
1
∩ O
2
is open. Let ( be a collection of open sets. Then O
c
is closed for any O ∈ ( so
¸
O∈c
O
c
is closed. i.e. (
¸
O∈c
O)
c
is closed. Thus
¸
O∈c
O is open.
34a. Clearly A
◦
⊂ A for any set A. A is open if and only if for any x ∈ A, there exists δ > 0 such that
¦y : [y −x[ < δ¦ ⊂ A if and only if every point of A is an interior point of A. Thus A is open if and only
if A = A
◦
.
34b. Suppose x ∈ A
◦
. Then there exists δ > 0 such that (x −δ, x + δ) ⊂ A. In particular, x / ∈ A
c
and
x / ∈ (A
c
)
t
. Thus x ∈ (A
c
)
c
. Conversely, suppose x ∈ (A
c
)
c
. Then x ∈ A and x is not an accumulation
point of A
c
. Thus there exists δ > 0 such that [y−x[ ≥ δ for all y ∈ A
c
`¦x¦ = A
c
. Hence (x−δ, x+δ) ⊂ A
so x ∈ A
◦
.
35. Let ( be a collection of closed sets of real numbers such that every ﬁnite subcollection of ( has
nonempty intersection and suppose one of the sets F ∈ ( is bounded. Suppose
¸
F∈c
F = ∅. Then
¸
F∈c
F
c
= R ⊃ F. By the HeineBorel Theorem, there is a ﬁnite subcollection ¦F
1
, . . . , F
n
¦ ⊂ ( such
that F ⊂
¸
n
i=1
F
c
i
. Then F ∩
¸
n
i=1
F
i
= ∅. Contradiction. Hence
¸
F∈c
F = ∅.
36. Let 'F
n
` be a sequence of nonempty closed sets of real numbers with F
n+1
⊂ F
n
. Then for any ﬁnite
subcollection ¦F
n
1
, . . . , F
n
k
¦ with n
1
< < n
k
,
¸
k
i=1
F
n
i
= F
n
k
= ∅. If one of the sets F
n
is bounded,
then by Proposition 16,
¸
∞
i=1
F
i
= ∅.
For each n, let F
n
= [n, ∞). Then 'F
n
` is a sequence of nonempty closed sets of real numbers with
F
n+1
⊂ F
n
but none of the sets F
n
is bounded. Now
¸
∞
n=1
F
n
= ¦x : x ≥ n for all n¦ = ∅.
37. Removing the middle third (1/3, 2/3) corresponds to removing all numbers in [0, 1] with unique
ternary expansion 'a
n
` such that a
1
= 1. Removing the middle third (1/9, 2/9) of [0, 1/3] corresponds
to removing all numbers with unique ternary expansion such that a
1
= 0 and a
2
= 1 and removing the
middle third (7/9, 8/9) of [2/3, 1] corresponds to removing all numbers with unique ternary expansion
such that a
1
= 2 and a
2
= 1. Suppose the middle thirds have been removed up to the nth stage, then
removing the middle thirds of the remaining intervals corresponds to removing all numbers with unique
ternary expansion such that a
i
= 0 or 2 for i ≤ n and a
n+1
= 1. Each stage of removing middle thirds
results in a closed set and C is the intersection of all these closed sets so C is closed.
38. Given an element in the Cantor ternary set with (unique) ternary expansion 'a
n
` such that a
n
= 1
for all n, let 'b
n
` be the sequence obtained by replacing all 2’s in the ternary expansion by 1’s. Then
'b
n
` may be regarded as the binary expansion of a number in [0, 1]. This gives a onetoone mapping
8
from the Cantor ternary set into [0, 1]. This mapping is also onto since given a number in [0, 1], we can
take its binary expansion and replace all 1’s by 2’s to get a sequence consisting of only 0’s and 2’s, which
we may then regard as the ternary expansion of a number in the Cantor ternary set.
39. Since the Cantor ternary set C is closed, C
t
⊂ C. Conversely, given x ∈ C, let 'a
n
` be its ternary
expansion with a
n
= 1 for all n. Given δ > 0, choose N such that 1/3
N
< δ. Now let 'b
n
` be the sequence
with b
N+1
= [a
N+1
−2[ and b
n
= a
n
for n = N + 1. Let y be the number with ternary expansion 'b
n
`.
Then y ∈ C ` ¦x¦ and [x −y[ = 2/3
N+1
< 1/3
N
< δ. Thus x ∈ C
t
. Hence C = C
t
.
2.6 Continuous functions
40. Since F is closed, F
c
is open and it is the union of a countable collection of disjoint open intervals.
Take g to be linear on each of these open intervals and take g(x) = f(x) for x ∈ F. Then g is deﬁned
and continuous on R and g(x) = f(x) for all x ∈ F.
41. Suppose f is continuous on E. Let O be an open set and let x ∈ f
−1
[O]. Then f(x) ∈ O so
there exists ε
x
> 0 such that (f(x) − ε
x
, f(x) + ε
x
) ⊂ O. Since f is continuous, there exists δ
x
> 0
such that [f(y) − f(x)[ < ε
x
when y ∈ E and [y − x[ < δ
x
. Hence (x − δ
x
, x + δ
x
) ∩ E ⊂ f
−1
[O]. Let
U =
¸
x∈f
−1
[O]
(x − δ
x
, x + δ
x
). Then U is open and f
−1
[O] = E ∩ U. Conversely, suppose that for
each open set O, there is an open set U such that f
−1
[O] = E ∩ U. Let x ∈ E and let ε > 0. Then
O = (f(x) −ε, f(x) +ε) is open so there is an open set U such that f
−1
[O] = E ∩U. Now x ∈ U and U
is open so there exists δ > 0 such that (x −δ, x +δ) ⊂ U. Thus E ∩ (x −δ, x +δ) ⊂ f
−1
[O]. i.e. for any
y ∈ E with [y −x[ < δ, [f(y) −f(x)[ < ε. Thus f is continuous on E.
42. Suppose 'f
n
` is a sequence of continuous functions on E and that 'f
n
` converges uniformly to f on
E. Given ε > 0, there exists N such that for all x ∈ E and n ≥ N, [f
n
(x) − f(x)[ < ε/3. Also, there
exists δ > 0 such that [f
N
(y) − f
N
(x)[ < ε/3 if y ∈ E and [y − x[ < δ. Now if y ∈ E and [y − x[ < δ,
then [f(y) −f(x)[ ≤ [f(y) −f
N
(y)[ +[f
N
(y) −f
N
(x)[ +[f
N
(x) −f(x)[ < ε. Thus f is continuous on E.
*43. f is discontinuous at the nonzero rationals:
Given a nonzero rational q, let ε = [f(q) − q[ > 0. If q > 0, given any δ > 0, pick an irrational
x ∈ (q, q +δ). Then [f(x) −f(q)[ = x −f(q) > q −f(q) = ε. If q < 0, given any δ > 0, pick an irrational
x ∈ (q −δ, q). Then [f(x) −f(q)[ = f(q) −x > f(q) −q = ε.
f is continuous at 0:
Given ε > 0, let δ = ε. Then when [x[ < δ, [f(x)[ ≤ [x[ < ε.
f is continuous at each irrational:
Let x be irrational. First we show that for any M, there exists δ > 0 such that q ≥ M for any
rational p/q ∈ (x − δ, x + δ). Otherwise, there exists M such that for any n, there exists a rational
p
n
/q
n
∈ (x −1/n, x + 1/n) with q
n
< M. Then [p
n
[ ≤ max([x −1[, [x + 1[)q
n
< M max([x −1[, [x + 1[)
for all n. Thus there are only ﬁnitely many choices of p
n
and q
n
for each n. This implies that there
exists a rational p/q in (x −1/n, x + 1/n) for inﬁnitely many n. Contradiction.
Given ε > 0, choose M such that M
2
> max([x + 1[, [x − 1[)/6ε. Then choose δ > 0 such that
δ < min(1, ε) and q ≥ M for any rational p/q ∈ (x − δ, x + δ). Suppose [x − y[ < δ. If y is irrational,
then [f(x) − f(y)[ = [x − y[ < δ < ε. If y = p/q is rational, then [f(y) − f(x)[ ≤ [f(y) − y[ + [y − x[ =
[p[[1/q−sin(1/q)[+[y−x[ < [p[/6q
3
+δ < max([x+1[, [x−1[)/6q
2
+δ ≤ max([x+1[, [x−1[)/6M
2
+δ < 2ε.
(*) Note that sin x < x for all x > 0 and x < sin x for all x < 0. Also, [x − sin x[ < x
3
/6 for all x = 0
(by Taylor’s Theorem for example).
44a. Let f and g be continuous functions. Given ε > 0, choose δ > 0 such that [f(x) −f(y)[ < ε/2 and
[g(x) −g(y)[ < ε/2 whenever [x −y[ < δ. Then [(f +g)(x) −(f +g)(y)[ = [f(x) −f(y) +g(x) −g(y)[ ≤
[f(x) −f(y)[ +[g(x) −g(y)[ < ε whenever [x −y[ < δ. Thus f +g is continuous at x. Now choose δ
t
> 0
such that [g(x) − g(y)[ < ε/2[f(x)[ and [f(x) − f(y)[ < ε/2 max([g(x) − ε/2[f(x)[[, [g(x) + ε/2[f(x)[[)
whenever [x − y[ < δ
t
. Then [(fg)(x) − (fg)(y)[ = [f(x)g(x) − f(x)g(y) + f(x)g(y) − f(y)g(y)[ ≤
[f(x)[[g(x) −g(y)[ +[f(x) −f(y)[[g(y)[ < ε whenever [x −y[ < δ
t
. Thus fg is continuous at x.
44b. Let f and g be continuous functions. Given ε > 0, there exists δ > 0 such that [f(a) − f(b)[ < ε
whenever [a − b[ < δ. There also exists δ
t
> 0 such that [g(x) − g(y)[ < δ whenever [x − y[ < δ
t
. Thus
[(f ◦ g)(x) −(f ◦ g)(y)[ = [f(g(x)) −f(g(y))[ < ε when [x −y[ < δ
t
. Thus f ◦ g is continuous at x.
9
44c. Let f and g be continuous functions. Given ε > 0, choose δ > 0 such that [f(x) −f(y)[ < ε/2 and
[g(x) −g(y)[ < ε/2 whenever [x−y[ < δ. Now [(f ∨g)(x) −(f ∨g)(y)[ ≤ [f(x) −f(y)[ +[g(x) −g(y)[ < ε.
Thus f ∨ g is continuous at x. Furthermore, f ∧ g = f +g −(f ∨ g) so f ∧ g is continuous at x.
44d. Let f be a continuous function. Then [f[ = (f ∨ 0) −(f ∧ 0) so f is continuous.
45. Let f be a continuous realvalued function on [a, b] and suppose that f(a) ≤ γ ≤ f(b). Let
S = ¦x ∈ [a, b] : f(x) ≤ γ¦. Then S = ∅ since a ∈ S and S is bounded. Let c = sup S. Then c ∈ [a, b]. If
f(c) < γ, there exists δ > 0 such that δ < b − c and [f(x) − f(c)[ < γ − f(c) whenever [x − c[ < δ. In
particular, [f(c + δ/2) − f(c)[ < γ − f(c). Then f(c + δ/2) < γ so c + δ/2 ∈ S. Contradiction. On the
other hand, if f(c) > γ, there exists δ
t
> 0 such that δ
t
< c − a and [f(x) − f(c)[ < f(c) − γ whenever
[x − c[ < δ
t
. Then [f(x) − f(c)[ < f(c) − γ for all x ∈ (c − δ
t
, c] so f(x) > γ and x / ∈ S for all such x.
Contradiction. Hence f(c) = γ.
46. Let f be a continuous function on [a, b]. Suppose f is strictly monotone. We may assume f
is strictly monotone increasing. Then f is onetoone. Also, by the Intermediate Value Theorem, f
maps [a, b] onto [f(a), f(b)]. Let g = f
−1
: [f(a), f(b)] → [a, b]. Then g(f(x)) = x for all x ∈ [a, b].
Let y ∈ [f(a), f(b)]. Then y = f(x) for some x ∈ [a, b]. Given ε > 0, choose δ > 0 such that
δ < min(f(x) − f(x − ε), f(x + ε) − f(x)). When [y − z[ < δ, z = f(x
t
) for some x
t
∈ [a, b] with
[g(y) − g(z)[ = [g(f(x)) − g(f(x
t
))[ = [x
t
− x[ < ε. Thus g is continuous on [f(a), f(b)]. Conversely,
suppose there is a continuous function g such that g(f(x)) = x for all x ∈ [a, b]. If x, y ∈ [a, b] with x < y,
then g(f(x)) < g(f(y)) so f(x) = f(y). We may assume x = a and f(a) < f(b). If f(x) < f(a), then
by the Intermediate Value Theorem, f(a) = f(x
t
) for some x
t
∈ [x, b] and a = g(f(a)) = g(f(x
t
)) = x
t
.
Contradiction. Thus f(a) < f(x). Now if f(x) ≥ f(y), then f(a) < f(y) ≤ f(x) so f(y) = f(x
tt
) for
some x
tt
∈ [a, x] and y = g(f(y)) = g(f(x
tt
)) = x
tt
. Contradiction. Thus f(x) < f(y). Hence f is strictly
monotone increasing.
47. Let f be a continuous function on [a, b] and ε a positive number. Then f is uniformly continuous
so there exists δ > 0 such that [f(x) − f(y)[ < ε/2 whenever x, y ∈ [a, b] and [x − y[ < δ. Choose N
such that (b − a)/N < δ and let x
i
= a + i(b − a)/N for i = 1, . . . , N. Now deﬁne ϕ on [a, b] to be
linear on each [x
i
, x
i+1
] with ϕ(x
i
) = f(x
i
) for each i so that ϕ is polygonal on [a, b]. Let x ∈ [a, b].
Then x ∈ [x
i
, x
i+1
] for some i. We may assume f(x
i
) ≤ f(x
i+1
). Then ϕ(x
i
) ≤ ϕ(x) ≤ ϕ(x
i+1
) and
[ϕ(x) −f(x)[ ≤ [ϕ(x) −ϕ(x
i
)[ +[ϕ(x
i
) −f(x)[ ≤ [ϕ(x
i+1
) −ϕ(x
i
)[ +[f(x
i
) −f(x)[ < ε.
48. Suppose x ∈ [0, 1] is of the form q/3
n
0
with q = 1 or 2. Then x has two ternary expansions
'a
n
` and 'a
t
n
` where a
n
0
= q, a
n
= 0 for n = n
0
, a
t
n
0
= q − 1, a
t
n
= 0 for n < n
0
and a
t
n
= 2 for
n > n
0
. If q = 1, then from the ﬁrst expansion we get N = n
0
, b
N
= 1 and b
n
= 0 for n < N so
¸
N
n=1
b
n
/2
n
= 1/2
n
0
. From the second expansion we get N = ∞, b
n
= 0 for n < n
0
and b
n
= 1 for
n > n
0
so
¸
N
n=1
b
n
/2
n
=
¸
∞
n=n
0
+1
1/2
n
= 1/2
n
0
. If q = 2, then from the ﬁrst expansion we get N = ∞,
b
n
0
= 1 and b
n
= 0 for n = n
0
so
¸
N
n=1
b
n
/2
n
=
¸
∞
n=1
b
n
/2
n
= 1/2
n
0
. From the second expansion we
get N = n
0
, b
N
= 1 and b
n
= 0 for n < N so
¸
N
n=1
b
n
/2
n
= 1/2
n
0
. Hence the sum is independent of
the ternary expansion of x.
Let f(x) =
¸
N
n=1
b
n
/2
n
. Given x, y ∈ [0, 1] with ternary expansions 'a
n
` and 'a
t
n
` respectively, suppose
x < y and let n
0
be the smallest value of n such that a
n
0
= a
t
n
0
. Then a
n
0
< a
t
n
0
and b
n
0
≤ b
t
n
0
. Thus
f(x) =
¸
N
x
n=1
b
n
/2
n
≤
¸
N
y
n=1
b
t
n
/2
n
= f(y). Given x ∈ [0, 1] and ε > 0, choose M such that 1/2
M
< ε.
Now choose δ > 0 such that δ < 1/3
M+2
. Then when [x−y[ < δ, [f(x) −f(y)[ < ε. Hence f is monotone
and continuous on [0, 1].
Each interval contained in the complement of the Cantor ternary set consists of elements with ternary
expansions containing 1. Furthermore, for any x, y in the same interval and having ternary expansions
'a
n
` and 'a
t
n
` respectively, the smallest n such that a
n
= 1 is also the smallest such that a
t
n
= 1. Hence f
is constant on each such interval. For any y ∈ [0, 1], let 'b
n
` be its binary expansion. Take x ∈ [0, 1] with
ternary expansion 'a
n
` such that a
n
= 2b
n
for all n. Then x is in the Cantor ternary set and f(x) = y.
Thus f maps the Cantor ternary set onto [0, 1].
49a. Suppose lim
x→y
f(x) ≤ A. Given ε > 0, there exists δ > 0 such that sup
0<]x−y]<δ
f(x) < A + ε.
Thus for all x with 0 < [x − y[ < δ, f(x) ≤ A + ε. Conversely, suppose there exists for any ε > 0
some δ > 0 such that f(x) ≤ A + ε for all x with 0 < [x − y[ < δ. Then for each n, there exists
10
δ
n
> 0 such that f(x) ≤ A + 1/n for all x with 0 < [x − y[ < δ
n
so sup
0<]x−y]<δ
n
f(x) ≤ A + 1/n. Thus
lim
x→y
f(x) = inf
δ>0
sup
0<]x−y]<δ
f(x) ≤ inf
n
sup
0<]x−y]<δ
n
f(x) ≤ A+ 1/n for all n so lim
x→y
f(x) ≤ A.
49b. Suppose lim
x→y
f(x) ≥ A. Given ε > 0 and δ > 0, sup
0<]x−y]<δ
f(x) > A −ε so there exists x such that
0 < [x − y[ < δ and f(x) ≥ A − ε. Conversely, suppose that given ε > 0 and δ > 0, there exists x such
that 0 < [x −y[ < δ and f(x) ≥ A−ε. Then for each n, there exists x
n
such that 0 < [x
n
−y[ < δ and
f(x
n
) ≥ A−1/n. Thus for each δ > 0, sup
0<]x−y]<δ
f(x) ≥ A−1/n for all n so sup
0<]x−y]<δ
f(x) ≥ A. Hence
lim
x→y
f(x) = inf
δ>0
sup
0<]x−y]<δ
f(x) ≥ A.
49c. For any δ
1
, δ
2
> 0, if δ
1
< δ
2
, then inf
0<]x−y]<δ
2
f(x) ≤ inf
0<]x−y]<δ
1
f(x) ≤ sup
0<]x−y]<δ
1
f(x) ≤
sup
0<]x−y]<δ
2
f(x). In particular, inf
0<]x−y]<δ
1
f(x) ≤ sup
0<]x−y]<δ
2
f(x) and inf
0<]x−y]<δ
2
f(x) ≤ sup
0<]x−y]<δ
1
f(x).
Hence sup
δ>0
inf
0<]x−y]<δ
f(x) ≤ sup
0<]x−y]<δ
0
f(x) for any δ
0
> 0 so sup
δ>0
inf
0<]x−y]<δ
f(x) ≤ inf
δ>0
sup
0<]x−y]<δ
f(x). i.e.
lim
x→y
f(x) ≤ lim
x→y
f(x).
Suppose lim
x→y
f(x) = lim
x→y
f(x) with limf = ±∞. Let L be the common value. Given ε > 0, there exists
δ
1
> 0 such that sup
0<]x−y]<δ
1
f(x) < L + ε. i.e. f(x) < L + ε whenever 0 < [x − y[ < δ
1
. There also
exists δ
2
> 0 such that inf
0<]x−y]<δ
2
f(x) > L − ε. i.e. f(x) > L − ε whenever 0 < [x − y[ < δ
2
. Let
δ = min(δ
2
, δ
2
). Then when 0 < [x − y[ < δ, [f(x) − L[ < ε so lim
x→y
f(x) exists. Conversely, suppose
lim
x→y
f(x) exists and let L be its value. Given ε, there exists δ > 0 such that [f(x) − L[ < ε whenever
0 < [x − y[ < δ. By part (a), lim
x→y
f(x) ≤ L. Similarly, lim
x→y
f(x) ≥ L. i.e. lim
x→y
f(x) ≤ lim
x→y
f(x). Thus
equality holds.
49d. Suppose lim
x→y
f(x) = A and 'x
n
` is a sequence with x
n
= y such that y = limx
n
. For any δ > 0,
there exists N
δ
such that 0 < [x
n
−y[ < δ for n ≥ N
δ
. Thus for any δ > 0, inf
N
sup
n≥N
f(x
n
) ≤ sup
n≥N
δ
f(x
n
) ≤
sup
0<]x−y]<δ
f(x) so inf
N
sup
n≥N
f(x
n
) ≤ inf
δ>0
sup
0<]x−y]<δ
f(x). i.e. limf(x
n
) ≤ lim
x→y
f(x) = A.
49e. Suppose lim
x→y
f(x) = A. By part (a), for each n, there exists δ
n
> 0 such that f(x) < A + 1/n
whenever 0 < [x − y[ < δ
n
. By part (b), there exists x
n
such that 0 < [x
n
− y[ < min(δ
n
, 1/n) and
f(x
n
) > A − 1/n. Thus 0 < [x
n
− y[ < 1/n and [f(x
n
) − A[ < 1/n for each n. i.e. 'x
n
` is a sequence
with x
n
= y such that y = limx
n
and A = limf(x
n
).
49f. Suppose l = lim
x→y
f(x) and let 'x
n
` be a sequence with x
n
= y and y = limx
n
. Given ε > 0,
there exists δ > 0 such that [f(x) − l[ < ε whenever 0 < [x − y[ < δ. Also there exists N such that
0 < [x
n
− y[ < δ for n ≥ N. Thus for n ≥ N, [f(x
n
) − l[ < ε. i.e. l = limf(x
n
). Conversely, suppose
l = lim
x→y
f(x). Then there exists ε > 0 such that for each n there exists x
n
with 0 < [x
n
−y[ < 1/n and
[f(x
n
) −l[ ≥ ε. Thus 'x
n
` is a sequence with x
n
= y and y = limx
n
but l = limf(x
n
).
50a. Let f(y) be ﬁnite. Then f is lower semicontinuous at y if and only if −f is upper semicontinuous
at y if and only if −f(y) ≥ lim
x→y
(−f(x)) if and only if given ε > 0, there exists δ > 0 such that
−f(x) ≤ −f(y) + ε whenever 0 < [x − y[ < δ if and only if given ε > 0, there exists δ > 0 such that
−f(x) ≤ −f(y) + ε whenever [x − y[ < δ if and only if given ε > 0, there exists δ > 0 such that
f(y) ≤ f(x) +ε whenever [x −y[ < δ.
50b. Suppose f is both upper and lower semicontinuous at y. Then lim
x→y
f(x) ≤ f(y) ≤ lim
x→y
f(x). Thus
lim
x→y
f(x) exists and equals f(y) so f is continuous at y. Conversely, if f is continuous at y, then lim
x→y
f(x)
exists and equals f(y). Thus lim
x→y
f(x) = lim
x→y
f(x) = f(y) so f is both upper and lower semicontinuous
11
at y. The result for intervals follows from the result for points.
50c. Let f be a realvalued function. Suppose f is lower semicontinuous on [a, b]. For λ ∈ R, consider
the set S = ¦x ∈ [a, b] : f(x) ≤ λ¦. Let y be a point of closure of S. Then there is a sequence 'x
n
` with
x
n
∈ S and y = limx
n
. Thus f(y) ≤ limf(x
n
) ≤ λ so y ∈ S. Hence S is closed so ¦x ∈ [a, b] : f(x) > λ¦
is open (in [a, b]). Conversely, suppose ¦x ∈ [a, b] : f(x) > λ¦ is open (in [a, b]) for each λ ∈ R. Let
y ∈ [a, b]. Given ε > 0, the set ¦x ∈ [a, b] : f(x) > f(y) − ε¦ is open so there exists δ > 0 such that
f(x) > f(y) −ε whenever [x −y[ < δ. i.e. f(y) < f(x) +ε whenever [x −y[ < δ. By part (a), f is lower
semicontinuous at y. Since y is arbitrarily chosen from [a, b], f is lower semicontinuous on [a, b].
50d. Let f and g be lower semicontinuous functions. Let y be in the domain of f and g. Given ε > 0,
there exists δ > 0 such that f(y) ≤ f(x) + ε/2 and g(y) ≤ g(x) + ε/2 whenever [x − y[ < δ. Thus
(f ∨g)(y) ≤ (f ∨g)(x) +ε and (f +g)(y) ≤ (f +g)(x) +ε whenever [x −y[ < δ. By part (a), f ∨g and
f +g are lower semicontinuous.
50e. Let 'f
n
` be a sequence of lower semicontinuous functions and deﬁne f(x) = sup
n
f
n
(x). Given
ε > 0, f(y) − ε/2 < f
n
(y) for some n. There also exists δ > 0 such that f
n
(y) ≤ f
n
(x) + ε/2 whenever
[x −y[ < δ. Thus f(y) ≤ f
n
(x) +ε ≤ f(x) +ε whenever [x −y[ < δ. Hence f is lower semicontinuous.
50f. Let ϕ : [a, b] → R be a step function. Suppose ϕ is lower semicontinuous. Let c
i
and c
i+1
be the
values assumed by ϕ in (x
i−1
, x
i
) and (x
i
, x
i+1
) respectively. For each n, there exists δ > 0 such that
ϕ(x
i
) ≤ ϕ(x)+1/n whenever [x−x
i
[ < δ. In particular, ϕ(x
i
) ≤ c
i
+1/n and ϕ(x
i
) ≤ c
i+1
+1/n for each
n. Thus ϕ(x
i
) ≤ min(c
i
, c
i+1
). Conversely, suppose ϕ(x
i
) ≤ min(c
i
, c
i+1
) for each i. Let ε > 0 and let
y ∈ [a, b]. If y = x
i
for some i, let δ = min(x
i
−x
i−1
, x
i+1
−x
i
). Then f(y) = f(x
i
) ≤ f(x) +ε whenever
[x −y[ < δ. If y ∈ (x
i
, x
i+1
) for some i, let δ = min(y −x
i
, x
i+1
−y). Then f(y) < f(y) +ε = f(x) +ε
whenever [x −y[ < δ. Thus ϕ is lower semicontinuous.
*50g. Let f be a function deﬁned on [a, b]. Suppose there is a monotone increasing sequence 'ϕ
n
` of
lower semicontinuous step functions on [a, b] such that for each x ∈ [a, b] we have f(x) = limϕ
n
(x).
Since 'ϕ
n
` is monotone increasing, for each x ∈ [a, b] we have f(x) = sup
n
ϕ
n
(x). By part (e), f is lower
semicontinuous. Conversely, suppose that f is lower semicontinuous. The sets ¦x ∈ [a, b] : f(x) > c¦
with c ∈ Z form an open covering of [a, b] so by the HeineBorel Theorem, there is a ﬁnite subcovering.
Thus there exists c ∈ Z such that f(x) > c for all x ∈ [a, b]. Now for each n, let x
(n)
k
= a+k(b−a)/2
n
and
let I
(n)
k
= (x
(n)
k−1
, x
(n)
k
) for k = 0, 1, 2, . . . , 2
n
. Deﬁne ϕ
n
(x) = inf
x∈I
(n)
k
f(x) if x ∈ I
(n)
k
and ϕ
n
(x
(n)
k
) =
min(c
(n)
k
, c
(n)
k+1
, f(x
(n)
k
)) where c
(n)
k
= inf
x∈I
(n)
k
f(x) and c
(n)
k+1
= inf
x∈I
(n)
k+1
f(x). Then each ϕ
n
is a lower
semicontinuous step function on [a, b] by part (f) and f ≥ ϕ
n+1
≥ ϕ
n
for each n. Let y ∈ [a, b]. Given
ε > 0, there exists δ > 0 such that f(y) ≤ f(x) +ε whenever [x −y[ < δ. Choose N such that 1/2
N
< δ.
For n ≥ N, if y ∈ I
(n)
k
for some k, then ϕ
n
(y) = inf
x∈I
(n)
k
f(x) ≥ f(y) − ε since I
(n)
k
⊂ (y − δ, y + δ).
Thus f(y) −ϕ
n
(y) ≤ ε. If y = x
(n)
k
for some k, then ϕ
n
(y) =≥ f(y) −ε since I
(n)
k
∪I
(n)
k+1
⊂ (y −δ, y +δ).
Thus f(y) −ϕ
n
(y) ≤ ε. Hence f(y) = limϕ
n
(y).
*50h. Let f be a function deﬁned on [a, b]. Suppose there is a monotone increasing sequence 'ψ
n
`
of continuous functions on [a, b] such that for each x ∈ [a, b] we have f(x) = limψ
n
(x). Then each
ψ
n
is lower semicontinuous by part (b) and f(x) = sup
n
ψ
n
(x). By part (e), f is lower semicontinuous.
Conversely, suppose that f is lower semicontinuous. By part (g), there is a monotone increasing sequence
'ϕ
n
` of lower semicontinuous step functions on [a, b] with f(x) = limϕ
n
(x) for each x ∈ [a, b]. For each
n, deﬁne ψ
n
by linearising ϕ
n
at a neighbourhood of each partition point such that ψ
n
≤ ψ
n+1
and
0 ≤ ϕ
n
(x) −ψ
n
(x) < ε/2 for each x ∈ [a, b]. Then 'ψ
n
` is a monotone increasing sequence of continuous
functions on [a, b] and f(x) = limψ
n
(x) for each x ∈ [a, b].
(*) More rigorous proof: Deﬁne ψ
n
by ψ
n
(x) = inf¦f(t) +n[t −x[ : t ∈ [a, b]¦. Then ψ
n
(x) ≤ inf¦f(t) +
n[t − y[ + n[y − x[ : t ∈ [a, b]¦ = ψ
n
(y) + n[y − x[. Thus ψ
n
is (uniformly) continuous on [a, b]. Also
ψ
n
≤ ψ
n+1
≤ f for all n. In particular, f(x) is an upper bound for ¦ψ
n
(x) : n ∈ N¦. Now if α < f(x),
then there exists δ > 0 such that α ≤ f(y) ≤ f(y) +n[y −x[ whenever [y −x[ < δ. On the other hand,
when [y −x[ ≥ δ, we have α ≤ f(y) +nδ ≤ f(y) +n[y −x[ for suﬃciently large n. Thus α ≤ ψ
n
(x) for
suﬃciently large n. Hence f(x) = sup ψ
n
(x) = limψ
n
(x).
50i. Let f be a lower semicontinuous function on [a, b]. The sets ¦x ∈ [a, b] : f(x) > c¦ with c ∈ Z form
an open covering of [a, b] so by the HeineBorel Theorem, there is a ﬁnite subcovering. Thus there exists
c ∈ Z such that f(x) > c for all x ∈ [a, b]. Hence f is bounded from below. Let m = inf
x∈[a,b]
f(x), which
12
is ﬁnite since f is bounded from below. Suppose f(x) > m for all x ∈ [a, b]. For each x ∈ [a, b], there exists
δ
x
> 0 such that f(x) ≤ f(y) +m−f(x) for y ∈ I
x
= (x−δ
x
, x+δ
x
). The open intervals ¦I
x
: x ∈ [a, b]¦
form an open covering of [a, b] so by the HeineBorel Theorem, there is a ﬁnite subcovering ¦I
x
1
, . . . , I
x
n
¦.
Let c = min(f(x
1
), . . . , f(x
n
)). Each y ∈ [a, b] belongs to some I
x
k
so f(y) ≥ 2f(x
k
) −m ≥ 2c−m. Thus
2c − m is a lower bound for f on [a, b] but 2c − m > m. Contradiction. Hence there exists x
0
∈ [a, b]
such that m = f(x
0
).
51a. Let x ∈ [a, b]. Then inf
]y−x]<δ
f(y) ≤ f(x) ≤ sup
]y−x]<δ
f(y) for any δ > 0. Hence we have g(x) =
sup
δ>0
inf
]y−x]<δ
f(y) ≤ f(x) ≤ inf
δ>0
sup
]y−x]<δ
f(y) = h(x). Suppose g(x) = f(x). Then given ε > 0, there exists
δ > 0 such that f(x) − ε = g(x) − ε < inf
]y−x]<δ
f(y). Thus f(x) − ε < f(y) whenever [y − x[ < δ and
f is lower semicontinuous at x. Conversely, suppose f is lower semicontinuous at x. f(x) is an upper
bound for ¦ inf
]x−y]<δ
f(y) : δ > 0¦ and given ε > 0, there exists δ > 0 such that f(x) −ε ≤ f(y) whenever
[x − y[ < δ. Thus f(x) − ε ≤ inf
]x−y]<δ
f(y) so f(x) = sup
δ>0
inf
]x−y]<δ
f(y) = g(x). By a similar argument,
f(x) = h(x) if and only if f is upper semicontinuous at x. Thus f is continuous at x if and only if f is
both upper semicontinuous and lower semicontinuous at x if and only if f(x) = h(x) and g(x) = f(x) if
and only if g(x) = h(x).
51b. Let λ ∈ R. Suppose g(x) > λ. Then there exists δ > 0 such that f(y) > λ whenever [x − y[ < δ.
Hence ¦x : g(x) > λ¦ is open in [a, b] and g is lower semicontinuous. Suppose h(x) < λ. Then there
exists δ > 0 such that f(y) < λ whenever [x − y[ < δ. Hence ¦x : h(x) < λ¦ is open in [a, b] and h is
upper semicontinuous.
51c. Let ϕ be a lower semicontinuous function such that ϕ(x) ≤ f(x) for all x ∈ [a, b]. Suppose
ϕ(x) > g(x) for some x ∈ [a, b]. Then there exists δ > 0 such that ϕ(x) ≤ ϕ(y) + ϕ(x) −g(x) whenever
[x − y[ < δ. i.e. g(x) ≤ ϕ(y) whenever [x − y[ < δ. In particular, g(x) ≤ ϕ(x). Contradiction. Hence
ϕ(x) ≤ g(x) for all x ∈ [a, b].
2.7 Borel sets
52. Let f be a lower semicontinuous function on R. Then ¦x : f(x) > α¦ is open. ¦x : f(x) ≥
α¦ =
¸
n
¦x : f(x) > α − 1/n¦ so it is a G
δ
set. ¦x : f(x) ≤ α¦ = ¦x : f(x) > α¦
c
is closed.
¦x : f(x) < α¦ = ¦x : f(x) ≥ α¦
c
so it is an F
σ
set. ¦x : f(x) = α¦ = ¦x : f(x) ≥ α¦ ∩ ¦x : f(x) ≤ α¦ is
the intersection of a G
δ
set with a closed set so it is a G
δ
set.
53. Let f be a realvalued function deﬁned on R. Let S be the set of points at which f is continuous.
If f is continuous at x, then for each n, there exists δ
n,x
> 0 such that [f(x) − f(y)[ < 1/n whenever
[x − y[ < δ
n,x
. Consider G
n
=
¸
x∈S
(x − δ
n,x
/2, x + δ
n,x
/2) and G =
¸
n
G
n
. Then x ∈ G for every
x ∈ S. Conversely, suppose x
0
∈ G for some x
0
/ ∈ S. There exists ε > 0 such that for every δ > 0, there
exists y with [y −x
0
[ < δ and [f(x
0
) −f(y)[ ≥ ε. Choose N such that 1/N < ε/2. There exists y with
[y −x
0
[ < δ
N,x
/2 and [f(y) −f(x
0
)[ ≥ ε. On the other hand, x
0
∈ G
N
=
¸
x∈S
(x−δ
N,x
/2, x+δ
N,x
/2) so
x
0
∈ (x−δ
N,x
/2, x+δ
N,x
/2) for some x ∈ S. Thus [f(x) −f(x
0
)[ < 1/N < ε/2. Also, [y −x[ ≤ [y −x
0
[ +
[x
0
−x[ < δ
N,x
so [f(y) −f(x)[ < 1/N < ε/2. Thus [f(y) −f(x
0
)[ ≤ [f(y) −f(x)[ +[f(x) −f(x
0
)[ < ε.
Contradiction. Hence G = S and S is a G
δ
set.
54. Let 'f
n
` be a sequence of continuous functions on R and let C be the set of points where the
sequence converges. If x ∈ C, then for any m, there exists n such that [f
k
(x) − f
n
(x)[ ≤ 1/m for all
k ≥ n. Consider F
n,m
= ¦x : [f
k
(x) − f
n
(x)[ ≤ 1/m for k ≥ n¦. Now C ⊂
¸
m
¸
n
F
n,m
. Conversely,
if x ∈
¸
m
¸
n
F
n,m
, then given ε > 0, choose M such that 1/M < ε. Now x ∈
¸
n
F
n,M
so there exists
N such that [f
k
(x) − f
N
(x)[ ≤ 1/M < ε for k ≥ N. Thus 'f
n
(x)` converges so
¸
m
¸
n
F
n,m
⊂ C. By
continuity of each f
k
, each F
n,m
is closed. Hence C is an F
σδ
set.
3 Lebesgue Measure
3.1 Introduction
1. Let A and B be two sets in M with A ⊂ B. Then mA ≤ mA+m(B ` A) = mB.
13
2. Let 'E
n
` be a sequence of sets in M. Let F
1
= E
1
and let F
n+1
= E
n+1
`
¸
n
k=1
E
k
. Then F
m
∩F
n
= ∅
for m = n, F
n
⊂ E
n
for each n and
¸
E
n
=
¸
F
n
. Thus m(
¸
E
n
) = m(
¸
F
n
) =
¸
mF
n
≤
¸
mE
n
.
3. Suppose there is a set A in M with mA < ∞. Then mA = m(A∪ ∅) = mA+m∅ so m∅ = 0.
4. Clearly n is translation invariant and deﬁned for all sets of real numbers. Let 'E
k
` be a sequence of
disjoint sets of real numbers. We may assume E
k
= ∅ for all k since n∅ = 0. If some E
k
is an inﬁnite
set, then so is
¸
E
k
. Thus n(
¸
E
k
) = ∞ =
¸
nE
k
. If all E
k
’s are ﬁnite sets and ¦E
k
: k ∈ N¦ is a ﬁnite
set, then
¸
E
k
is a ﬁnite set and since the E
k
’s are disjoint, n(
¸
E
k
) =
¸
nE
k
. On the other hand,
if all E
k
’s are ﬁnite sets and ¦E
k
: k ∈ N¦ is an inﬁnite set, then
¸
E
k
is a countably inﬁnite set so
n(
¸
E
k
) = ∞ and since nE
k
≥ 1 for all k,
¸
nE
k
= ∞.
3.2 Outer measure
5. Let A be the set of rational numbers between 0 and 1. Also let ¦I
n
¦ be a ﬁnite collection of open
intervals covering A. Then 1 = m
∗
([0, 1]) = m
∗
A ≤ m
∗
(
¸
I
n
) = m
∗
(
¸
I
n
) ≤
¸
m
∗
I
n
=
¸
l(I
n
) =
¸
l(I
n
).
6. Given any set A and any ε > 0, there is a countable collection ¦I
n
¦ of open intervals covering
A such that
¸
l(I
n
) ≤ m
∗
A + ε. Let O =
¸
I
n
. Then O is an open set such that A ⊂ O and
m
∗
O ≤
¸
m
∗
I
n
=
¸
l(I
n
) ≤ m
∗
A +ε. Now for each n, there is an open set O
n
such that A ⊂ O
n
and
m
∗
O
n
≤ m
∗
A+ 1/n. Let G =
¸
O
n
. Then G is a G
δ
set such that A ⊂ G and m
∗
G = m
∗
A.
7. Let E be a set of real numbers and let y ∈ R. If ¦I
n
¦ is a countable collection of open intervals such
that E ⊂
¸
I
n
, then E+y ⊂
¸
(I
n
+y) so m
∗
(E+y) ≤
¸
l(I
n
+y) =
¸
l(I
n
). Thus m
∗
(E+y) ≤ m
∗
E.
Conversely, by a similar argument, m
∗
E ≤ m
∗
(E +y). Hence m
∗
(E +y) = m
∗
E.
8. Suppose m
∗
A = 0. Then m
∗
(A ∪ B) ≤ m
∗
A + m
∗
B = m
∗
B. Conversely, since B ⊂ A ∪ B,
m
∗
B ≤ m
∗
(A∪ B). Hence m
∗
(A∪ B) = m
∗
B.
3.3 Measurable sets and Lebesgue measure
9. Let E be a measurable set, let A be any set and let y ∈ R. Then m
∗
A = m
∗
(A−y) = m
∗
((A−y)∩E)+
m
∗
((A−y)∩E
c
) = m
∗
(((A−y)∩E)+y)+m
∗
(((A−y)∩E
c
)+y) = m
∗
(A∩(E+y))+m
∗
(A∩(E
c
+y)) =
m
∗
(A∩ (E +y)) +m
∗
(A∩ (E +y)
c
). Thus E +y is a measurable set.
10. Suppose E
1
and E
2
are measurable. Then m(E
1
∪E
2
) +m(E
1
∩E
2
) = mE
1
+m(E
2
` E
1
) +m(E
1
∩
E
2
) = mE
1
+mE
2
.
11. For each n, let E
n
= (n, ∞). Then E
n+1
⊂ E
n
for each n,
¸
E
n
= ∅ and mE
n
= ∞ for each n.
12. Let 'E
i
` be a sequence of disjoint measurable sets and let A be any set. Then m
∗
(A ∩
¸
∞
i=1
E
i
) =
m
∗
(
¸
∞
i=1
(A ∩ E
i
)) ≤
¸
∞
i=1
m
∗
(A ∩ E
i
) by countable subadditivity. Conversely, m
∗
(A ∩
¸
∞
i=1
E
i
) ≥
m
∗
(A∩
¸
n
i=1
E
i
) =
¸
n
i=1
m
∗
(A∩E
i
) for all n by Lemma 9. Thus m
∗
(A∩
¸
∞
i=1
E
i
) ≥
¸
∞
i=1
m
∗
(A∩E
i
).
Hence m
∗
(A∩
¸
∞
i=1
E
i
) =
¸
∞
i=1
m
∗
(A∩ E
i
).
13a. Suppose m
∗
E < ∞.
(i) ⇒ (ii). Suppose E is measurable. Given ε > 0, there is a countable collection ¦I
n
¦ of open intervals
such that E ⊂
¸
I
n
and
¸
l(I
n
) < m
∗
E + ε. Let O =
¸
I
n
. Then O is an open set, E ⊂ O and
m
∗
(O ` E) = m(O ` E) = m(
¸
I
n
) −mE = m
∗
(
¸
I
n
) −m
∗
E ≤
¸
l(I
n
) −m
∗
E < ε.
(ii) ⇒ (vi). Given ε > 0, there is an open set O such that E ⊂ O and m
∗
(O ` E) < ε/2. O is the union
of a countable collection of disjoint open intervals ¦I
n
¦ so
¸
l(I
n
) = m(
¸
I
n
) < mE + ε/2. Thus there
exists N such that
¸
∞
n=N+1
l(I
n
) < ε/2. Let U =
¸
N
n=1
I
n
. Then m
∗
(U∆E) = m
∗
(U `E)+m
∗
(E`U) ≤
m
∗
(O ` E) +m
∗
(O ` U) < ε/2 +ε/2 = ε.
(vi) ⇒ (ii). Given ε > 0, there is a ﬁnite union U of open intervals such that m
∗
(U∆E) < ε/3. Also
there is an open set O such that E ` U ⊂ O and m
∗
O ≤ m
∗
(E ` U) + ε/3. Then E ⊂ U ∪ O and
m
∗
((U ∪ O) ` E) = m
∗
((U ` E) ∪ (O ` E)) ≤ m
∗
((O ` (E ` U)) ∪ (E ` U) ∪ (U ` E)) < ε.
13b. (i) ⇒ (ii). Suppose E is measurable. The case where m
∗
E < ∞ was proven in part (a). Suppose
m
∗
E = ∞. For each n, let E
n
= [−n, n] ∩ E. By part (a), for each n, there exists an open set O
n
⊃ E
n
such that m
∗
(O
n
` E
n
) < ε/2
n
. Let O =
¸
O
n
. Then E ⊂ O and m
∗
(O ` E) = m
∗
(
¸
O
n
`
¸
E
n
) ≤
m
∗
(
¸
(O
n
` E
n
)) < ε.
14
(ii) ⇒ (iv). For each n, there exists an open set O
n
such that E ⊂ O
n
and m
∗
(O
n
` E) < 1/n. Let
G =
¸
O
n
. Then E ⊂ G and m
∗
(G` E) ≤ m
∗
(O
n
` E) < 1/n for all n. Thus m
∗
(G` E) = 0.
(iv) ⇒ (i). There exists a G
δ
set G such that E ⊂ G and m
∗
(G ` E) = 0. Now G and G ` E are
measurable sets so E = G` (G` E) is a measurable set.
13c. (i) ⇒ (iii). Suppose E is measurable. Then E
c
is measurable. By part (b), given ε > 0, there is
an open set O such that E
c
⊂ O and m
∗
(O ` E
c
) < ε. i.e. m
∗
(O ∩ E) < ε. Let F = O
c
. Then F is
closed, F ⊂ E and m
∗
(E ` F) = m
∗
(E ` O
c
) = m
∗
(E ∩ O) < ε.
(iii) ⇒ (v). For each n, there is a closed set F
n
such that F
n
⊂ E and m
∗
(E`F
n
) < 1/n. Let F =
¸
F
n
.
Then F ⊂ E and m
∗
(E ` F) ≤ m
∗
(E ` F
n
) < 1/n for all n. Thus m
∗
(E ` F) = 0.
(v) ⇒ (i). There exists an F
σ
set F such that F ⊂ E and m
∗
(E ` F) = 0. Now F and E ` F are
measurable sets so E = F ∪ (E ` F) is a measurable set.
14a. For each n, let E
n
be the union of the intervals removed in the nth step. Then mE
n
= 2
n−1
/3
n
so
m(
¸
E
n
) =
¸
mE
n
= 1. Thus mC = m([0, 1]) −m(
¸
E
n
) = 0.
14b. Each step of removing open intervals results in a closed set and F is the intersection of all these
closed sets so F is closed. Let x ∈ [0, 1] ` F
c
. Note that removing intervals in the nth step results in
2
n
disjoint intervals, each of length less than 1/2
n
. Given δ > 0, choose N such that 1/2
N
< 2δ. Then
(x−δ, x+δ) must intersect one of the intervals removed in step N. i.e. there exists y ∈ (x−δ, x+δ)∩F
c
.
Thus F
c
is dense in [0, 1]. Finally, mF = m([0, 1]) −
¸
α2
n−1
/3
n
= 1 −α.
3.4 A nonmeasurable set
15. Let E be a measurable set with E ⊂ P. For each i, let E
i
= E
◦
+r
i
. Since E
i
⊂ P
i
for each i, 'E
i
` is
a disjoint sequence of measurable sets with mE
i
= mE. Thus
¸
mE
i
= m(
¸
E
i
) ≤ m([0, 1)) = 1. Since
¸
mE
i
=
¸
mE, we must have mE = 0.
16. Let A be any set with m
∗
A > 0. If A ⊂ [0, 1), let 'r
i
`
∞
i=0
be an enumeration of Q ∩ [−1, 1). For
each i, let P
i
= P +r
i
. Then [0, 1) ⊂
¸
P
i
since for any x ∈ [0, 1), there exists y ∈ P such that x and y
diﬀer by some rational r
i
. Also, P
i
∩ P
j
= ∅ if i = j since if p
i
+ r
i
= p
j
+ r
j
, then p
i
∼ p
j
and since
P contains exactly one element from each equivalence class, p
i
= p
j
and r
i
= r
j
. Now let E
i
= A ∩ P
i
for each i. If each E
i
is measurable, then mE
i
= m(A ∩ P
i
) = m((A − r
i
) ∩ P) = 0 for each i by Q15.
On the other hand,
¸
m
∗
E
i
≥ m
∗
(
¸
E
i
) ≥ m
∗
(A ∩ [0, 1)) = m
∗
A > 0. Hence E
i
0
is nonmeasurable for
some i
0
and E
i
0
⊂ A. Similarly, if A ⊂ [n, n+1) where n ∈ Z, then there is a nonmeasurable set E ⊂ A.
In general, A = A ∩
¸
n∈Z
[n, n + 1) and 0 < m
∗
A ≤
¸
n∈Z
m
∗
(A ∩ [n, n + 1)) so m
∗
(A ∩ [n, n + 1)) > 0
for some n ∈ Z and there is a nonmeasurable set E ⊂ A∩ [n, n + 1) ⊂ A.
17a. Let 'r
i
`
∞
i=0
be an enumeration of Q∩[−1, 1) and let P
i
= P +r
i
for each i. Then 'P
i
` is a disjoint
sequence of sets with m
∗
(
¸
P
i
) ≤ m
∗
[−1, 2) = 3 and
¸
m
∗
P
i
=
¸
m
∗
P = ∞. Thus m
∗
(
¸
P
i
) <
¸
m
∗
P
i
.
17b. For each i, let P
i
be as deﬁned in part (a) and let E
i
=
¸
n≥i
P
n
. Then 'E
i
` is a sequence with
E
i
⊃ E
i+1
for each i and m
∗
E
i
≤ m
∗
(
¸
P
n
) < ∞ for each i. Furthermore,
¸
E
i
= ∅ since if x ∈ P
k
, then
x / ∈
¸
n≥k+1
P
n
so m
∗
(
¸
E
i
) = 0. On the other hand, P
i
⊂ E
i
for each i so 0 < m
∗
P = m
∗
P
i
≤ m
∗
E
i
for each i and limm
∗
E
i
≥ m
∗
P > 0 = m
∗
(
¸
E
i
).
3.5 Measurable functions
18. Let E be the nonmeasurable set deﬁned in Section 3.4. Let f be deﬁned on [0, 1] with f(x) = x +1
if x ∈ E and f(x) = −x if x / ∈ E. Then f takes each value at most once so ¦x : f(x) = α¦ has at most
one element for each α ∈ R and each of these sets is measurable. However, ¦x : f(x) > 0¦ = E, which is
nonmeasurable.
19. Let D be a dense set of real numbers and let f be an extended realvalued function on R such that
¦x : f(x) > α¦ is measurable for each α ∈ D. Let β ∈ R. For each n, there exists α
n
∈ D such that
β < α
n
< β +1/n. Now ¦x : f(x) > β¦ =
¸
¦x : f(x) ≥ β +1/n¦ =
¸
¦x : f(x) > α
n
¦ so ¦x : f(x) > β¦
is measurable and f is measurable.
20. Let ϕ
1
=
¸
n
1
i=1
α
i
χ
A
i
and let ϕ
2
=
¸
n
2
i=1
β
i
χ
B
i
. Then ϕ
1
+ ϕ
2
=
¸
n
1
i=1
α
i
χ
A
i
+
¸
n
2
i=1
β
i
χ
B
i
is a
15
simple function. Also ϕ
1
ϕ
2
=
¸
i,j
α
i
β
j
χ
A
i
χ
B
j
is a simple function. χ
A∩B
(x) = 1 if and only if x ∈ A
and x ∈ B if and only if χ
A
(x) = 1 = χ
B
(x). Thus χ
A∩B
= χ
A
χ
B
. If χ
A∪B
(x) = 1, then x ∈ A ∪ B. If
x ∈ A∩B, then χ
A
(x) +χ
B
(x) +χ
A
χ
B
(x) = 1 +1 −1 = 1. If x / ∈ A∩B, then x ∈ A` B or x ∈ B` A so
χ
A
(x)+χ
B
(x) = 1 and χ
A
χ
B
(x) = 0. If χ
A∪B
(x) = 0, then x / ∈ A∪B so χ
A
(x) = χ
B
(x) = χ
A
χ
B
(x) = 0.
Hence χ
A∪B
= χ
A
+χ
B
+χ
A
χ
B
. If χ
A
c (x) = 1, then x / ∈ A so χ
A
(x) = 0. If χ
A
c (x) = 0, then x ∈ A so
χ
A
(x) = 1. Hence χ
A
c = 1 −χ
A
.
21a. Let D and E be measurable sets and f a function with domain D ∪ E. Suppose f is measurable.
Since D and E are measurable subsets of D∪E, f[
D
and f[
E
are measurable. Conversely, suppose f[
D
and
f[
E
are measurable. Then for any α ∈ R, ¦x : f(x) > α¦ = ¦x ∈ D : f[
D
(x) > α¦∪¦x ∈ E : f[
E
(x) > α¦.
Each set on the right is measurable so ¦x : f(x) > α¦ is measurable and f is measurable.
21b. Let f be a function with measurable domain D. Let g be deﬁned by g(x) = f(x) if x ∈ D and
g(x) = 0 if x / ∈ D. Suppose f is measurable. If α ≥ 0, then ¦x : g(x) > α¦ = ¦x : f(x) > α¦, which
is measurable. If α < 0, then ¦x : g(x) > α¦ = ¦x : f(x) > α¦ ∪ D
c
, which is measurable. Hence
g is measurable. Conversely, suppose g is measurable. Then f = g[
D
and since D is measurable, f is
measurable.
22a. Let f be an extended realvalued function with measurable domain D and let D
1
= ¦x : f(x) =
∞¦, D
2
= ¦x : f(x) = −∞¦. Suppose f is measurable. Then D
1
and D
2
are measurable by Proposition
18. Now D`(D
1
∪D
2
) is a measurable subset of D so the restriction of f to D`(D
1
∪D
2
) is measurable.
Conversely, suppose D
1
and D
2
are measurable and the restriction of f to D` (D
1
∪D
2
) is measurable.
For α ∈ R, ¦x : f(x) > α¦ = D
1
∪¦x : f[
D\(D
1
∪D
2
)
(x) > α¦, which is measurable. Hence f is measurable.
22b. Let f and g be measurable extended realvalued functions deﬁned on D. D
1
= ¦fg = ∞¦ =
¦f = ∞, g > 0¦ ∪ ¦f = −∞, g < 0¦ ∪ ¦f > 0, g = ∞¦ ∪ ¦f < 0, g = −∞¦, which is measurable.
D
2
= ¦fg = −∞¦ = ¦f = ∞, g < 0¦ ∪ ¦f = −∞, g > 0¦ ∪ ¦f > 0, g = −∞¦ ∪ ¦f < 0, g = ∞¦,
which is measurable. Let h = fg[
D\(D
1
∪D
2
)
and let α ∈ R. If α ≥ 0, then ¦x : h(x) > α¦ = ¦x :
f[
D\¦x:f(x)=¬∞¦
(x) g[
D\¦x:g(x)=¬∞¦
(x) > α¦, which is measurable. If α < 0, then ¦x : h(x) > α¦ = ¦x :
f(x) = 0¦ ∪ ¦x : g(x) = 0¦ ∪ ¦x : f[
D\¦f=¬∞¦
(x) g[
D\¦g=¬∞¦
(x) > α¦, which is measurable. Hence fg
is measurable.
22c. Let f and g be measurable extended realvalued functions deﬁned on D and α a ﬁxed number.
Deﬁne f +g to be α whenever it is of the form ∞−∞or −∞+∞. D
1
= ¦f +g = ∞¦ = ¦f ∈ R, g = ∞¦∪
¦f = g = ∞¦ ∪ ¦f = ∞, g ∈ R¦, which is measurable. D
2
= ¦f +g = −∞¦ = ¦f ∈ R, g = −∞¦ ∪ ¦f =
g = −∞¦ ∪ ¦f = −∞, g ∈ R¦, which is measurable. Let h = (f +g)[
D\(D
1
∪D
2
)
and let β ∈ R. If β ≥ α,
then ¦x : h(x) > β¦ = ¦x : f[
D\¦f=¬∞¦
(x) + g[
D\¦g=¬∞¦
(x) > β¦, which is measurable. If β < α, then
¦x : h(x) > β¦ = ¦f = ∞, g = −∞¦ ∪ ¦f = −∞, g = ∞¦ ∪ ¦x : f[
D\¦f=¬∞¦
(x) + g[
D\¦g=¬∞¦
(x) > β¦,
which is measurable. Hence f +g is measurable.
22d. Let f and g be measurable extended realvalued functions that are ﬁnite a.e. Then the sets
D
1
, D
2
, ¦x : h(x) > β¦ can be written as unions of sets as in part (c), possibly with an additional set of
measure zero. Thus these sets are measurable and f +g is measurable.
23a. Let f be a measurable function on [a, b] that takes the values ±∞ only on a set of measure zero
and let ε > 0. For each n, let E
n
= ¦x ∈ [a, b] : [f(x)[ > n¦. Each E
n
is measurable and E
n+1
⊂ E
n
for each n. Also, mE
1
≤ b − a < ∞. Thus limmE
n
= m(
¸
E
n
) = 0. Thus there exists M such that
mE
M
< ε/3. i.e. [f[ ≤ M except on a set of measure less than ε/3.
23b. Let f be a measurable function on [a, b]. Let ε > 0 and M be given. Choose N such that M/N < ε.
For each k ∈ ¦−N, −N + 1, . . . , N − 1¦, let E
k
= ¦x ∈ [a, b] : kM/N ≤ f(x) < (k + 1)M/N¦. Each E
k
is measurable. Deﬁne ϕ by ϕ =
¸
N−1
k=−N
(kM/N)χ
E
k
. Then ϕ is a simple function. If x ∈ [a, b] such
that [f(x)[ < M, then x ∈ E
k
for some k. i.e. kM/N ≤ f(x) < (k + 1)M/N and ϕ(x) = kM/N. Thus
[f(x) − ϕ(x)[ < M/N < ε. If m ≤ f ≤ M, we may take ϕ so that m ≤ ϕ ≤ M by replacing M/N by
(M −m)/N < ε in the preceding argument.
23c. Let ϕ be a simple function on [a, b] and let ε > 0 be given. Let ϕ =
¸
n
i=1
a
i
χ
A
i
. For each i, there
is a ﬁnite union U
i
=
¸
N
i
k=1
I
i,k
of (disjoint) open intervals such that m(U
i
∆A
i
) < ε/3n. Deﬁne g by
g =
¸
n
i=1
a
i
χ
U
i
\
i−1
m=1
U
m
. Then g is a step function on [a, b]. If g(x) = ϕ(x), then either g(x) = a
i
= ϕ(x),
or g(x) = 0 and ϕ(x) = a
i
. In the ﬁrst case, x ∈ U
i
` A
i
. In the second case, x ∈ A
i
` U
i
. Thus
¦x ∈ [a, b] : ϕ(x) = g(x)¦ ⊂
¸
n
i=1
((U
i
` A
i
) ∪ (A
i
` U
i
)) =
¸
n
i=1
(U
i
∆A
i
) so it has measure less than ε/3.
If m ≤ ϕ ≤ M, we may take g so that m ≤ g ≤ M since both g and ϕ take values in ¦a
1
, . . . , a
n
¦.
16
23d. Let g be a step function on [a, b] and let ε > 0 be given. Let x
0
, . . . , x
n
be the partition points
corresponding to g. Let d = min¦x
i
−x
i−1
: i = 1, . . . , n¦. For each i, let I
i
be an open interval of length
less than min(ε/3(n + 1), d/2) centred at x
i
. Deﬁne h by linearising g in each I
i
. Then h is continuous
and ¦x ∈ [a, b] : g(x) = h(x)¦ ⊂
¸
n
i=0
I
i
, which has measure less than ε/3. If m ≤ g ≤ M, we may take
h so that m ≤ h ≤ M by construction.
24. Let f be measurable and B a Borel set. Let ( be the collection of sets E such that f
−1
[E] is
measurable. Suppose E ∈ (. Then f
−1
[E
c
] = (f
−1
[E])
c
, which is measurable, so E
c
∈ (. Suppose 'E
i
`
is a sequence of sets in (. Then f
−1
[
¸
E
i
] =
¸
f
−1
[E
i
], which is measurable, so
¸
E
i
∈ (. Thus ( is
a σalgebra. Now for any a, b ∈ R with a < b, ¦x : f(x) > a¦ and ¦x : f(x) < b¦ are measurable. i.e.
(a, ∞) and (−∞, b) are in (. Thus (a, b) ∈ ( and ( is a σalgebra containing all the open intervals so it
contains all the Borel sets. Hence f
−1
[B] is measurable.
25. Let f be a measurable realvalued function and g a continuous function deﬁned on (−∞, ∞). Then
g is also measurable. For any α ∈ R, ¦x : (g ◦ f)(x) > α¦ = (g ◦ f)
−1
[(α, ∞)] = f
−1
[g
−1
[(α, ∞)]], which
is measurable by Q24. Hence g ◦ f is measurable.
26. Propositions 18 and 19 and Theorem 20 follow from arguments similar to those in the original proofs
and the fact that the collection of Borel sets is a σalgebra. If f is a Borel measurable function, then
for any α ∈ R, the set ¦x : f(x) > α¦ is a Borel set so it is Lebesgue measurable. Thus f is Lebesgue
measurable. If f is Borel measurable and B is a Borel set, then consider the collection ( of sets E such
that f
−1
[E] is a Borel set. By a similar argument to that in Q24, ( is a σalgebra containing all the
open intervals. Thus ( contains all the Borel sets. Hence f
−1
[B] is a Borel set. If f and g are Borel
measurable, then for α ∈ R, ¦x : (f ◦g)(x) > α¦ = (f ◦g)
−1
[(α, ∞)] = g
−1
[f
−1
[(α, ∞)]], which is a Borel
set. Thus f ◦ g is Borel measurable. If f is Borel measurable and g is Lebesgue measurable, then for
any α ∈ R, f
−1
[(α, ∞)] is a Borel set and g
−1
[f
−1
[(α, ∞)]] is Lebesgue measurable by Q24. Thus f ◦ g
is Lebesgue measurable.
27. Call a function Ameasurable if for each α ∈ R the set ¦x : f(x) > α¦ is in A. Propositions 18
and 19 and Theorem 20 still hold. An Ameasurable function need not be Lebesgue measurable. For
example, let A be the σalgebra generated by the nonmeasurable set P deﬁned in Section 4. Then χ
P
is A
measurable but not Lebesgue measurable. There exists a Lebesgue measurable function g and a Lebesgue
measurable set A such that g
−1
[A] is nonmeasurable (see Q28). If f and g are Lebesgue measurable, f ◦g
may not be Lebesgue measurable. For example, take g and A to be Lebesgue measurable with g
−1
[A]
nonmeasurable. Let f = χ
A
so that f is Lebesgue measurable. Then ¦x : (f ◦ g)(x) > 1/2¦ = g
−1
[A],
which is nonmeasurable. This is also a counterexample for the last statement.
28a. Let f be deﬁned by f(x) = f
1
(x) + x for x ∈ [0, 1]. By Q2.48, f
1
is continuous and monotone on
[0, 1] so f is continuous and strictly monotone on [0, 1] and f maps [0, 1] onto [0, 2]. By Q2.46, f has a
continuous inverse so it is a homeomorphism of [0, 1] onto [0, 2].
28b. By Q2.48, f
1
is constant on each interval contained in the complement of the Cantor set. Thus f
maps each of these intervals onto an interval of the same length. Thus m(f[[0, 1] `C]) = m([0, 1] `C) = 1
and since f is a bijection of [0, 1] onto [0, 2], mF = m(f[C]) = m([0, 2]) −1 = 1.
28c. Let g = f
−1
: [0, 2] → [0, 1]. Then g is measurable. Since mF = 1 > 0, there is a nonmeasurable set
E ⊂ F. Let A = f
−1
[E]. Then A ⊂ C so it has outer measure zero and is measurable but g
−1
[A] = E
so it is nonmeasurable.
28d. The function g = f
−1
is continuous and the function h = χ
A
is measurable, where A is as deﬁned
in part (c). However the set ¦x : (h ◦ g)(x) > 1/2¦ = g
−1
[A] is nonmeasurable. Thus h ◦ g is not
measurable.
28e. The set A in part (c) is measurable but by Q24, it is not a Borel set since g
−1
[A] is nonmeasurable.
3.6 Littlewood’s three principles
29. Let E = R and let f
n
= χ
[n,∞)
for each n. Then f
n
(x) → 0 for each x ∈ E. For any measurable set
A ⊂ E with mA < 1 and any integer N, pick x ≥ N such that x / ∈ A. Then [f
N
(x) −0[ ≥ 1.
30. Egoroﬀ’s Theorem: Let 'f
n
` be a sequence of measurable functions that converges to a realvalued
function f a.e. on a measurable set E of ﬁnite measure. Let η > 0 be given. For each n, there exists
A
n
⊂ E with mA
n
< η/2
n
and there exists N
n
such that for all x / ∈ A
n
and k ≥ N
n
, [f
k
(x)−f(x)[ < 1/n.
17
Let A =
¸
A
n
. Then A ⊂ E and mA <
¸
η/2
n
= η. Choose n
0
such that 1/n
0
< η. If x / ∈ A and
k ≥ N
n
0
, we have [f
k
(x) −f(x)[ < 1/n
0
< η. Thus f
n
converges to f uniformly on E ` A.
31. Lusin’s Theorem: Let f be a measurable realvalued function on [a, b] and let δ > 0 be given. For
each n, there is a continuous function h
n
on [a, b] such that m¦x : [h
n
(x) − f(x)[ ≥ δ/2
n+2
¦ < δ/2
n+2
.
Let E
n
= ¦x : [h
n
(x)−f(x)[ ≥ δ/2
n+2
¦. Then [h
n
(x)−f(x)[ < δ/2
n+2
for x ∈ [a, b] `E
n
. Let E =
¸
E
n
.
Then mE < δ/4 and 'h
n
` is a sequence of continuous, thus measurable, functions that converges to f on
[a, b] ` E. By Egoroﬀ’s Theorem, there is a subset A ⊂ [a, b] ` E such that mA < δ/4 and h
n
converges
uniformly to f on [a, b] `(E∪A). Thus f is continuous on [a, b] `(E∪A) with m(E∪A) < δ/2. Now there
is an open set O such that O ⊃ (E ∪ A) and m(O ` (E ∪ A)) < δ/2. Then f is continuous on [a, b] ` O,
which is closed, and mO < δ. By Q2.40, there is a function ϕ that is continuous on (−∞, ∞) such that
f = ϕ on [a, b] ` O. In particular, ϕ is continuous on [a, b] and m¦x ∈ [a, b] : f(x) = ϕ(x)¦ = mO < δ.
If f is deﬁned on (−∞, ∞), let δ
t
= min(δ/2
n+3
, 1/2). Then for each n, there is a continuous function
ϕ
n
on [n + δ
t
, n + 1 − δ
t
] such that m¦x ∈ [n + δ
t
, n + 1 − δ
t
] : f(x) = ϕ
n
(x)¦ < δ/2
n+2
. Similarly for
[−n−1 +δ
t
, −n−δ
t
]. Linearise in each interval [n−δ
t
, n+δ
t
]. Similarly for intervals [−n−δ
t
, −n+δ
t
].
Then we have a continuous function ϕ deﬁned on (−∞, ∞) with m¦x : f(x) = ϕ(x)¦ < 4
¸
δ/2
n+2
= δ.
*32. For t ∈ [0, 1) with 1/2
i+1
≤ t < 1/2
i
, deﬁne f
t
: [0, 1) → R by f
t
(x) = 1 if x ∈ P
i
= P + r
i
and x = 2
i+1
t − 1, and f
t
(x) = 0 otherwise. For each t, there is at most one x such that f
t
(x) = 1
so each f
t
is measurable. For each x, x ∈ P
i(x)
for some i(x). Let t(x) = (x + 1)/2
i(x)+1
. Then
1/2
i(x)+1
≤ t(x) < 1/2
i(x)
and f
t(x)
(x) = 1. This is the only t such that f
t
(x) = 1. Thus for each x,
f
t
(x) → 0 as t → 0. Note that any measurable subset of [0, 1) with positive measure intersects inﬁnitely
many of the sets P
i
. Thus for any measurable set A ⊂ [0, 1) with mA < 1/2, m([0, 1) ` A) ≥ 1/2 so
there exists x ∈ [0, 1) ` A with i(x) arbitrarily large and so with t(x) arbitrarily small. i.e. there exist
an x ∈ [0, 1) ` A and arbitrarily small t such that f
t
(x) ≥ 1/2.
(*) Any measurable set A ⊂ [0, 1) with positive measure intersects inﬁnitely many of the sets P
i
:
Suppose A intersects only ﬁnitely many of the sets P
i
. i.e. A ⊂
¸
n
i=1
P
q
i
, where P
q
i
= P + q
i
. Choose
r
1
∈ Q∩[−1, 1] such that r
1
= q
i
−q
j
for all i, j. Suppose r
1
, . . . , r
n
have been chosen. Choose r
n+1
such
that r
n+1
= q
i
−q
j
+r
k
for all i, j and k ≤ n. Now the measurable sets A+r
i
are disjoint by the deﬁnition
of P and the construction of the sequence 'r
i
`. Then m(
¸
n
i=1
(A + r
i
)) =
¸
n
i=1
m(A + r
i
) = nmA for
each n. Since
¸
n
i=1
(A+r
i
) ⊂ [−1, 2], nmA ≤ 3 for all n and mA = 0.
4 The Lebesgue Integral
4.1 The Riemann integral
1a. Let f be deﬁned by f(x) = 0 if x is irrational and f(x) = 1 if x is rational. For any subdivision
a = ξ
0
< ξ
1
< < ξ
n
= b of [a, b], M
i
= sup
ξ
i−1
<x≤ξ
i
f(x) = 1 and m
i
= inf
ξ
i−1
<x≤ξ
i
f(x) = 0 for each
i. Thus S =
¸
n
i=1
(ξ
i
− ξ
i−1
)M
i
= b − a and s =
¸
n
i=1
(ξ
i
− ξ
i−1
)m
i
= 0 for any subdivision of [a, b].
Hence R
b
a
f(x) = b −a and R
b
a
f(x) = 0.
1b. Let 'r
n
` be an enumeration of Q∩[a, b]. For each n, let f
n
= χ
¦r
1
,...,r
n
¦
. Then 'f
n
` is a sequence of
nonnegative Riemann integrable functions increasing monotonically to f. Thus the limit of a sequence
of Riemann integrable functions may not be Riemann integrable so in general we cannot interchange the
order of integration and the limiting process.
4.2 The Lebesgue integral of a bounded function over a set of ﬁnite measure
2a. Let f be a bounded function on [a, b] and let h be the upper envelope of f. i.e. h(y) = inf
δ>0
sup
]y−x]<δ
f(x).
Since f is bounded, h is upper semicontinuous by Q2.51b and h is bounded. For any α ∈ R, the set
¦x : h(x) < α¦ is open. Thus h is measurable. Let ϕ be a step function on [a, b] such that ϕ ≥ f.
Then ϕ ≥ h except at a ﬁnite number of points (the partition points). Thus R
b
a
f = inf
ϕ≥f
b
a
ϕ ≥
b
a
h.
Conversely, there exists a monotone decreasing sequence 'ϕ
n
` of step functions such that h(x) = limϕ
n
(x)
for each x ∈ [a, b]. Thus
b
a
h = lim
b
a
ϕ
n
≥ R
b
a
f. Hence R
b
a
f =
b
a
h.
18
2b. Let f be a bounded function on [a, b] and let E be the set of discontinuities of f. Let g be the
lower envelope of f. By a similar argument as in part (a), R
b
a
f =
b
a
g. Suppose E has measure zero.
Then g = h a.e. so R
b
a
f =
b
a
g =
b
a
h = R
b
a
f so f is Riemann integrable. Conversely, suppose f is
Riemann integrable. Then
b
a
g =
b
a
h. Thus
b
a
[g −h[ = 0 so g = h a.e. since
b
a
[g −h[ ≥ (1/n)m¦x :
[g(x) −h(x)[ > 1/n¦ for all n. Hence f is continuous a.e. and mE = 0.
4.3 The integral of a nonnegative function
3. Let f be a nonnegative measurable function and suppose inf f = 0. Let E = ¦x : f(x) > 0¦. Then
E =
¸
E
n
where E
n
= ¦x : f(x) ≥ 1/n¦. Now
f ≥ (1/n)mE
n
for each n so mE
n
= 0 for each n and
mE = 0. Thus f = 0 a.e.
4a. Let f be a nonnegative measurable function. For n = 1, 2, . . ., let E
n,i
= f
−1
[(i −1)2
−n
, i2
−n
) where
i = 1, . . . , n2
n
and let E
n,0
= f
−1
[n, ∞). Deﬁne ϕ
n
=
¸
n2
n
i=1
(i − 1)2
−n
χ
(E
n,i
∩[−n,n])
+ nχ
(E
n,0
∩[−n,n])
.
Then each ϕ
n
is a nonnegative simple function vanishing outside a set of ﬁnite measure and ϕ
n
≤ ϕ
n+1
for each n. Furthermore, for suﬃciently large n, f
n
(x) −ϕ
n
(x) ≤ 2
−n
if x ∈ [−n, n] and f(x) < n. Thus
ϕ
n
(x) → f(x) when f(x) < ∞. Also, for suﬃciently large n, ϕ
n
(x) = n → ∞ if f(x) = ∞.
4b. Let f be a nonnegative measurable function. Then by part (a), there is an increasing sequence 'ϕ
n
`
of simple functions such that f = limϕ
n
. By the Monotone Convergence Theorem,
f = lim
ϕ
n
=
sup
ϕ
n
. Thus
f ≤ sup
ϕ over all simple functions ϕ ≤ f. On the other hand,
f ≥
ϕ for all
simple functions ϕ ≤ f. Thus
f ≥ sup
ϕ over all simple functions ϕ ≤ f. Hence
f = sup
ϕ over
all simple functions ϕ ≤ f.
5. Let f be a nonnegative integrable function and let F(x) =
x
−∞
f. For each n, let f
n
= fχ
(−∞,x−1/n]
.
Then 'f
n
` is an increasing sequence of nonnegative measurable functions with fχ
(−∞,x]
= limf
n
. By
the Monotone Convergence Theorem, limF(x − 1/n) = lim
f
n
=
fχ
(−∞,x]
= F(x). Now for each
n, let g
n
= fχ
(x+1/n,∞)
. Then 'g
n
` is an increasing sequence of nonnegative measurable functions with
fχ
(x,∞)
= limg
n
. By the Monotone Convergence Theorem, lim
g
n
=
fχ
(x,∞)
. i.e. lim
∞
x+1/n
f =
∞
x
f. Since f is integrable, we have lim(
f −
x+1/n
−∞
f) =
f −
x
−∞
f so lim
x+1/n
−∞
f =
x
−∞
f.
i.e. limF(x + 1/n) = F(x). Now given ε > 0, there exists N such that F(x) − F(x − 1/n) < ε and
F(x+1/n) −F(x) < ε whenever n ≥ N. Choose δ < 1/N. Then [F(y) −F(x)[ < ε whenever [x−y[ < δ.
Hence F is continuous.
6. Let 'f
n
` be a sequence of nonnegative measurable functions converging to f and suppose f
n
≤ f for
each n. By Fatou’s Lemma,
f ≤ lim
f
n
. On the other hand,
f ≥ lim
f
n
since f ≥ f
n
for each n.
Hence
f = lim
f
n
.
7a. For each n, let f
n
= χ
[n,n+1)
. Then 'f
n
` is a sequence of nonnegative measurable functions with
limf
n
= 0. Now
0 = 0 < 1 = lim
f
n
and we have strict inequality in Fatou’s Lemma.
7b. For each n, let f
n
= χ
[n,∞)
. Then 'f
n
` is a decreasing sequence of nonnegative measurable functions
with limf
n
= 0. Now
0 = 0 < ∞ = lim
f
n
so the Monotone Convergence Theorem does not hold.
8. Let 'f
n
` be a sequence of nonnegative measurable functions. For each n, let h
n
= inf
k≥n
f
k
. Then
each h
n
is a nonnegative measurable function with h
n
≤ f
n
. By Fatou’s Lemma,
limf
n
=
limh
n
≤
lim
h
n
≤ lim
f
n
.
9. Let 'f
n
` be a sequence of nonnegative measurable functions such that f
n
→ f a.e. and suppose
that
f
n
→
f < ∞. Then for any measurable set E, 'f
n
χ
E
` is a sequence of nonnegative measurable
functions with f
n
χ
E
→ fχ
E
a.e. By Fatou’s Lemma,
E
f ≤ lim
E
f
n
. Now fχ
E
is integrable since
fχ
E
≤ f. Also, f
n
is integrable for suﬃciently large n so f
n
χ
E
is integrable for suﬃciently large n. By
Fatou’s Lemma,
(f −fχ
E
) ≤ lim
(f
n
−f
n
χ
E
). i.e.
f −
E
f ≤ lim
f
n
−lim
E
f
n
=
f −lim
E
f
n
.
Thus lim
E
f
n
≤
E
f and we have
E
f
n
→
E
f.
4.4 The general Lebesgue integral
10a. If f is integrable over E, then so are f
+
and f
−
. Thus [f[ = f
+
+ f
−
is integrable over E and
[
E
f[ = [
E
f
+
−
E
f
−
[ ≤ [
E
f
+
[ +[
E
f
−
[ =
E
f
+
+
E
f
−
=
E
[f[. Conversely, if [f[ is integrable
19
over E, then
E
f
+
≤
E
[f[ < ∞ and
E
f
−
≤
E
[f[ < ∞ so f
+
and f
−
are integrable over E and f is
integrable over E.
10b. f(x) = sin x/x is not Lebesgue integrable on [0, ∞] although R
∞
0
f(x) = π/2 (by contour
integration for example). In general, suppose f is Lebesgue integrable and the Riemann integral R
b
a
f
exists with improper lower limit a. If a is ﬁnite, let f
n
= fχ
[a+1/n,b]
. Then f
n
→ f on [a, b] and [f
n
[ ≤ [f[
so R
b
a
f = limR
b
a+1/n
f = lim
f
n
=
b
a
f. If a = −∞, let g
n
= fχ
[−n,b]
. Then g
n
→ f on [a, b]
and [g
n
[ ≤ [f[ so R
b
a
f = limR
b
−n
f = lim
g
n
=
b
a
f. The cases where the Riemann integral has
improper upper limit are similar.
11. Let ϕ =
¸
n
i=1
a
i
χ
A
i
be a simple function with canonical representation. Let S+ = ¦i : a
i
≥ 0¦
and let S− = ¦i : a
i
< 0¦. Then ϕ
+
=
¸
i∈S+
a
i
χ
A
i
and ϕ
−
= −
¸
i∈S−
a
i
χ
A
i
. Clearly ϕ
+
and ϕ
−
are simple functions. Then
ϕ
+
=
¸
i∈S+
a
i
mA
i
and
ϕ
−
= −
¸
i∈S−
a
i
mA
i
so
ϕ =
ϕ
+
−
ϕ
−
=
¸
n
i=1
a
i
mA
i
.
12. Let g be an integrable function on a set E and suppose that 'f
n
` is a sequence of measurable
functions with [f
n
[ ≤ g a.e. on E. Then 'f
n
+ g` is a sequence of nonnegative measurable functions on
E. Thus
limf
n
+
g ≤
lim(f
n
+ g) ≤ lim
(f
n
+ g) ≤ lim
f
n
+
g so
limf
n
≤ lim
f
n
. Also,
'g−f
n
` is a sequence of nonnegative measurable functions on E. Thus
g+
lim(−f
n
) ≤
lim(g−f
n
) ≤
lim
(g −f
n
) ≤
g + lim(−
f
n
) so
lim(−f
n
) ≤ lim(−
f
n
). i.e. lim
f
n
≤
limf
n
. Hence we have
limf
n
≤ lim
f
n
≤ lim
f
n
≤
limf
n
.
13. Let h be an integrable function and 'f
n
` a sequence of measurable functions with f
n
≥ −h and
limf
n
= f. For each n, f
n
+ h is a nonnegative measurable function. Since h is integrable,
f
n
=
(f
n
+ h) −
h. Similarly,
f =
(f + h) −
h. Now
f +
h ≤
lim(f
n
+ h) ≤ lim
f
n
+
h so
f ≤ lim
f
n
.
14a. Let 'g
n
` be a sequence of integrable functions which converges a.e. to an integrable function
g and let 'f
n
` be a sequence of measurable functions such that [f
n
[ ≤ g
n
and 'f
n
` converges to f
a.e. Suppose
g = lim
g
n
. Since [f
n
[ ≤ g
n
, [f[ ≤ g. Thus [f
n
− f[ ≤ [f
n
[ + [f[ ≤ g
n
+ g and
'g
n
+ g − [f
n
− f[` is a sequence of nonnegative measurable functions. By Fatou’s Lemma,
lim(g
n
+
g − [f
n
− f[) ≤ lim
(g
n
+ g − [f
n
− f[). i.e.
2g ≤
2g + lim(−
[f
n
− f[) =
2g − lim
[f
n
− f[.
Hence lim
[f
n
−f[ ≤ 0 ≤ lim
[f
n
−f[ and we have
[f
n
−f[ → 0.
14b. Let 'f
n
` be a sequence of integrable functions such that f
n
→ f a.e. with f integrable. If
[f
n
− f[ → 0, then
[f
n
[ −
[f[
≤
[[f
n
[ −[f[[ ≤
[f
n
− f[ → 0. Thus
[f
n
[ →
[f[. Conversely,
suppose
[f
n
[ →
[f[. By part (a), with [f
n
[ in place of g
n
and [f[ in place of g, we have
[f
n
−f[ → 0.
15a. Let f be integrable over E and let ε > 0 be given. By Q4, there is a simple function ψ ≤ f
+
such
that
E
f
+
− ε/2 <
E
ψ. Also, there is a simple function ψ
t
≤ f
−
such that
E
f
−
− ε/2 <
E
ψ
t
. Let
ϕ = ψ−ψ
t
. Then ϕ is a simple function and
E
[f−ϕ[ =
E
[f
+
−ψ−f
−
+ψ
t
[ ≤
E
[f
+
−ψ[+
E
[f
−
−ψ
t
[ =
E
(f
+
−ψ) +
E
(f
−
−ψ
t
) < ε.
15b. Let f
n
= fχ
[−n,n]
. Then f
n
→ f and [f
n
[ ≤ [f[. By Lebesgue’s Dominated Convergence Theorem,
E
[f − fχ
[−n,n]
[ → 0. i.e.
E∩[−n,n]
c
[f[ → 0. Thus there exists N such that
E∩[−N,N]
c
[f[ < ε/3.
By part (a), there is a simple function ϕ such that
E
[f − ϕ[ < ε/3. By Proposition 3.22, there
is a step function ψ on [−N, N] such that [ϕ − ψ[ < ε/12NM except on a set of measure less than
ε/12NM, where M ≥ max([ϕ[, [ψ[) + 1. We may regard ψ as a function on R taking the value 0
outside [−N, N]. Then
N
−N
[ϕ − ψ[ < ε/3 so
E
[f − ψ[ =
E∩[−N,N]
[f − ψ[ +
E∩[−N,N]
c
[f − ψ[ ≤
E∩[−N,N]
[f −ϕ[ +
E∩[−N,N]
[ϕ−ψ[ +
E∩[−N,N]
c
[f −ψ[ ≤
E
[f −ϕ[ +
N
−N
[ϕ−ψ[ +
E∩[−N,N]
c
[f[ < ε.
15c. By part (b), there is a step function ψ such that
E
[f −ψ[ < ε/2. Suppose ψ is deﬁned on [a, b].
We may regard ψ as a function on R taking the value 0 outside [a, b]. By linearising ψ at each partition
point, we get a continuous function g vanishing outside a ﬁnite interval such that ψ = g except on a set
of measure less than ε/4M, where M ≥ [ψ[. Then
E
[f −g[ ≤
E
[f −ψ[ +
E
[ψ −g[ < ε.
16. RiemannLebesgue Theorem: Suppose f is integrable on (−∞, ∞). By Q15, given ε > 0,
there is a step function ψ such that
[f − ψ[ < ε/2. Now [
f(x) cos nxdx[ ≤
[f(x) cos nx[ dx ≤
[(f(x) −ψ(x)) cos nx[ dx +
[ψ(x) cos nx[ dx < ε/2 +
[ψ(x) cos nx[ dx. Integrating [ψ(x) cos nx[ over
each interval on which ψ is constant, we see that
[ψ(x) cos nx[ dx → 0 as n → ∞. Thus there
exists N such that
[ψ(x) cos nx[ dx < ε/2 for n ≥ N so [
f(x) cos nxdx[ < ε for n ≥ N. i.e.
20
lim
n→∞
f(x) cos nxdx = 0.
17a. Let f be integrable over (−∞, ∞). Then f
+
and f
−
are nonnegative integrable functions. There
exists an increasing sequence 'ϕ
n
` of nonnegative simple functions such that f
+
= limϕ
n
. Now since
χ
E
(x) dx = mE = m(E − t) =
χ
E
(x + t) dx for any measurable set E, we have
ϕ
n
(x) dx =
ϕ
n
(x+t) dx for all n. By the Monotone Convergence Theorem,
f
+
(x) dx =
f
+
(x+t) dx. Similarly
for f
−
. Thus
f(x) dx =
f(x +t) dx.
17b. Let g be a bounded measurable function and let M be such that [g[ ≤ M. Since f is integrable, given
ε > 0, there is a continuous function h vanishing outside a ﬁnite interval [a, b] such that
[f −h[ < ε/4M.
Now
[g(x)[f(x) −f(x +t)][ ≤
[g(x)[h(x) −h(x +t)][ +
[g(x)[(f −h)(x) −(f −h)(x +t)][. Now h is
uniformly continuous on [a, b] so there exists δ > 0 such that [h(x)−h(x+t)[ < ε/2M(b−a) whenever [t[ <
δ. Then
[g(x)[f(x)−f(x+t)][ ≤ ε/2+M
[(f −h)(x)[ +
[(f −h)(x +t)[
= ε/2+2M
[f −h[ < ε
whenever [t[ < δ. i.e. lim
t→0
[g(x)[f(x) −f(x +t)][ = 0.
18. Let f be a function of 2 variables 'x, t` deﬁned on the square Q = [0, 1] [0, 1] and which is a
measurable function of x for each ﬁxed t. Suppose that lim
t→0
f(x, t) = f(x) and that for all t we have
[f(x, t)[ ≤ g(x) where g is an integrable function on [0, 1]. Let 't
n
` be a sequence such that t
n
= 0 for all
n and lim
n
t
n
= 0. Then lim
n
f(x, t
n
) = f(x). For each n, h
n
(x) = f(x, t
n
) is measurable and [h
n
[ ≤ g.
By Lebesgue’s Dominated Convergence Theorem, lim
n
h
n
=
f. i.e. lim
t→0
f(x, t) dx =
f(x) dx.
Suppose further that f(x, t) is continuous in t for each x and let h(t) =
f(x, t) dx. Let 't
n
` be a
sequence converging to t. Then limf(x, t
n
) = f(x, t) for each x. By Lebesgue’s Dominated Convergence
Theorem, lim
f(x, t
n
) dx =
f(x, t) dx. i.e. limh(t
n
) = h(t). Hence h is a continuous function of t.
19. Let f be a function deﬁned and bounded in the square Q = [0, 1] [0, 1] and suppose that for each
ﬁxed t the function f is a measurable function of x. For each 'x, t` in Q, let the partial derivative ∂f/∂t
exist. Suppose that ∂f/∂t is bounded in Q. Let 's
n
` be a sequence such that s
n
= 0 for all n and
lim
n
s
n
= 0. Then lim
n
[f(x, t + s
n
) − f(x, t)]/s
n
→ ∂f/∂t. Since ∂f/∂t is bounded, there exists M
such that [[f(x, t + s
n
) − f(x, t)]/s
n
[ ≤ M + 1 for suﬃciently large n. For each ﬁxed t, f is a bounded
measurable function of x so [
1
0
f(x, t+s
n
) dx−
1
0
f(x, t) dx]/s
n
=
1
0
([f(x, t+s
n
)−f(x, t)]/s
n
) dx. Thus
d
dt
1
0
f(x, t) dx = lim
n
[
1
0
f(x, t + s
n
) dx −
1
0
f(x, t) dx]/s
n
= lim
n
1
0
([f(x, t + s
n
) − f(x, t)]/s
n
) dx =
1
0
∂f
∂t
dx, the last equality following from Lebesgue’s Dominated Convergence Theorem.
4.5 Convergence in measure
20. Let 'f
n
` be a sequence that converges to f in measure. Then given ε > 0, there exists N such that
m¦x : [f
n
(x) − f(x)[ ≥ ε¦ < ε for n ≥ N. For any subsequence 'f
n
k
`, choose M such that n
k
≥ N for
k ≥ M. Then m¦x : [f
n
k
(x) −f(x)[ ≥ ε¦ < ε for k ≥ M. Thus 'f
n
k
` converges to f in measure.
21. Fatou’s Lemma: Let 'f
n
` be a sequence of nonnegative measurable functions that converges in
measure to f on E. Then there is a subsequence 'f
n
k
` such that lim
E
f
n
k
= lim
E
f
n
. By Q20, 'f
n
k
`
converges in measure to f on E so it in turn has a subsequence 'f
n
k
j
` that converges to f a.e. Thus
E
f ≤ lim
E
f
n
k
j
= lim
E
f
n
k
= lim
E
f
n
.
Monotone Convergence Theorem: Let 'f
n
` be an increasing sequence of nonnegative measurable functions
that converges in measure to f. Any subsequence 'f
n
k
` also converges in measure to f so it in turn has
a subsequence 'f
n
k
j
` that converges to f a.e. Thus
f = lim
f
n
k
j
. By Q2.12,
f = lim
f
n
.
Lebesgue’s Dominated Convergence Theorem: Let g be integrable over E and let 'f
n
` be a sequence of
measurable functions such that [f
n
[ ≤ g on E and converges in measure to f on E. Any subsequence
'f
n
k
` also converges in measure to f so it in turn has a subsequence 'f
n
k
j
` that converges to f a.e. Thus
f = lim
f
n
k
j
. By Q2.12,
f = lim
f
n
.
22. Let 'f
n
` be a sequence of measurable functions on a set E of ﬁnite measure. If 'f
n
` converges to
f in measure, then so does any subsequence 'f
n
k
`. Thus any subsequence of 'f
n
k
` also converges to f
in measure. Conversely, if 'f
n
` does not converge in measure to f, then there exists ε > 0 such that for
any N there exists n ≥ N with m¦x : [f
n
(x) − f(x)[ ≥ ε¦ ≥ ε. This gives rise to a subsequence 'f
n
k
`
such that m¦x : [f
n
k
(x) −f(x) ≥ ε¦ ≥ ε for all k. This subsequence will not have a further subsequence
that converges in measure to f.
23. Let 'f
n
` be a sequence of measurable functions on a set E of ﬁnite measure. If 'f
n
` converges to f
21
in measure, then so does any subsequence 'f
n
k
`. Thus 'f
n
k
` has in turn a subsequence that converges to
f a.e. Conversely, if every subsequence 'f
n
k
` has in turn a subsequence 'f
n
k
j
` that converges to f a.e.,
then 'f
n
k
j
` converges to f in measure so by Q22, 'f
n
` converges to f in measure.
24. Suppose that f
n
→ f in measure and that there is an integrable function g such that [f
n
[ ≤ g for
all n. Let ε > 0 be given. Now [f
n
−f[ is integrable for each n and [f
n
−f[χ
[−k,k]
converges to [f
n
−f[.
By Lebesgue’s Dominated Convergence Theorem,
k
−k
[f
n
−f[ converges to
[f
n
−f[. Thus there exists
N such that
]x]≥N
[f
n
− f[ < ε/3. By Proposition 14, for each n, given ε > 0, there exists δ > 0 such
that for any set A with mA < δ,
A
[f
n
− f[ < ε/3. We may assume δ < ε/6N. There also exists N
t
such that m¦x : [f
n
(x) − f(x)[ ≥ δ¦ < δ for all n ≥ N
t
. Let A = ¦x : [f
n
(x) − f(x)[ ≥ δ¦. Then
[f
n
− f[ =
]x]≥N
[f
n
− f[ +
A∩[−N,N]
[f
n
− f[ +
A
c
∩[−N,N]
[f
n
− f[ < ε/3 + ε/3 + 2Nδ < ε for all
n ≥ N
t
. i.e.
[f
n
−f[ → 0.
25. Let 'f
n
` be a Cauchy sequence in measure. Then we may choose n
v+1
> n
v
such that m¦x :
[f
n
v+1
(x) − f
n
v
(x)[ ≥ 1/2
v
¦ < 1/2
v
. Let E
v
= ¦x : [f
n
v+1
(x) − f
n
v
(x)[ ≥ 1/2
v
¦ and let F
k
=
¸
v≥k
E
v
.
Then m(
¸
k
F
k
) ≤ m(
¸
v≥k
E
v
) ≤ 1/2
k−1
for all k so m(
¸
k
F
k
) = 0. If x / ∈
¸
k
F
k
, then x / ∈ F
k
for
some k so [f
n
v+1
(x) − f
n
v
(x)[ < 1/2
v
for all v ≥ k and [f
n
w
(x) − f
n
v
(x)[ < 1/2
v−1
for w ≥ v ≥ k.
Thus the series
¸
(f
n
v+1
− f
n
v
) converges a.e. to a function g. Let f = g + f
n
1
. Then f
n
v
→ f in
measure since the partial sums of the series are of the form f
n
v
−f
n
1
. Given ε > 0, choose N such that
m¦x : [f
n
(x) − f
r
(x)[ ≥ ε/2¦ < ε/2 for all n, r ≥ N and m¦x : [f
n
v
(x) − f(x)[ ≥ ε/2¦ < ε/2 for all
v ≥ N. Now ¦x : [f
n
(x) −f(x)[ ≥ ε¦ ⊂ ¦x : [f
n
(x) −f
n
v
(x)[ ≥ ε/2¦ ∪ ¦x : [f
n
v
(x) −f(x)[ ≥ ε/2¦ for all
n, v ≥ N. Thus m¦x : [f
n
(x) −f(x)[ ≥ ε¦ < ε for all n ≥ N. i.e. f
n
→ f in measure.
5 Diﬀerentiation and Integration
5.1 Diﬀerentiation of monotone functions
1. Let f be deﬁned by f(0) = 0 and f(x) = xsin(1/x) for x = 0. Then D
+
f(0) = lim
h→0
+
f(0+h)−f(0)
h
=
lim
h→0
+ sin(1/h) = 1. Similarly, D
−
f(0) = 1, D
+
f(0) = D
−
f(0) = −1.
2a. D
+
[−f(x)] = lim
h→0
+
[−f(x+h)]−[−f(x)]
h
= −lim
h→0
+
f(x+h)−f(x)
h
= −D
+
f(x).
2b. Let g(x) = f(−x). Then D
+
g(x) = lim
h→0
+
g(x+h)−g(x)
h
= lim
h→0
+
f(−x−h)−f(−x)
h
= lim
h→0
+ −
f(−x)−f(−x−h)
h
= −lim
h→0
+
f(−x)−f(−x−h)
h
= −D
−
f(−x).
3a. Suppose f is continuous on [a, b] and assumes a local maximum at c ∈ (a, b). Now there exists
δ > 0 such that f(c + h) < f(c) for 0 < h < δ. Then
f(c+h)−f(c)
h
< 0 for 0 < h < δ. Thus
D
+
f(c) = lim
h→0
+
f(c+h)−f(c)
h
≤ 0. Similarly, there exists δ
t
> 0 such that f(c) > f(c − h) for 0 <
h < δ
t
. Then
f(c)−f(c−h)
h
> 0 for 0 < h < δ
t
. Thus D
−
f(c) = lim
h→0
+
f(c)−f(c−h)
h
≥ 0. Hence
D
+
f(c) ≤ D
+
f(c) ≤ 0 ≤ D
−
f(c) ≤ D
−
f(c).
(*) Note error in book.
3b. If f has a local maximum at a, then D
+
f(a) ≤ D
+
f(a) ≤ 0. If f has a local maximum at b, then
0 ≤ D
−
f(b) ≤ D
−
f(b).
4. Suppose f is continuous on [a, b] and one of its derivates, say D
+
f, is everywhere nonnegative on
(a, b). First consider a function g such that D
+
g(x) ≥ ε > 0 for all x ∈ (a, b). Suppose there exist
x, y ∈ [a, b] with x < y and g(x) > g(y). Since D
+
g(x) > 0 for all x ∈ (a, b), g has no local maximum in
(a, b) by Q3. Thus g is decreasing on (a, y] and D
+
g(c) ≤ 0 for all c ∈ (a, y). Contradiction. Hence g is
nondecreasing on [a, b]. Now for any ε > 0, D
+
(f(x) +εx) ≥ ε on (a, b) so f(x) +εx is nondecreasing on
[a, b]. Let x < y. Then f(x)+εx ≤ f(y)+εy. Suppose f(x) > f(y). Then 0 < f(x)−f(y) ≤ ε(y −x). In
particular, choosing ε = (f(x) −f(y))/(2(y −x)), we have f(x) −f(y) ≤ (f(x) −f(y))/2. Contradiction.
Hence f is nondecreasing on [a, b].
The case where D
−
f is everywhere nonnegative on (a, b) follows from a similar argument and the cases
where D
+
f or D
−
f is everywhere nonnegative on (a, b) follow from the previous cases.
5a. For any x, D
+
(f + g)(x) = lim
h→0
+
(f+g)(x+h)−(f+g)(x)
h
= lim
h→0
+(
f(x+h)−f(x)
h
+
g(x+h)−g(x)
h
) ≤
lim
h→0
+
f(x+h)−f(x)
h
+ lim
h→0
+
g(x+h)−g(x)
h
= D
+
f(x) +D
+
g(x).
22
5b. For any x, D
+
(f + g)(x) = lim
h→0
+
(f+g)(x+h)−(f+g)(x)
h
= lim
h→0
+(
f(x+h)−f(x)
h
+
g(x+h)−g(x)
h
) ≤
lim
h→0
+
f(x+h)−f(x)
h
+ lim
h→0
+
g(x+h)−g(x)
h
= D
+
f(x) +D
+
g(x).
Similarly, D
−
(f +g) ≤ D
−
f +D
−
g and D
−
(f +g) ≤ D
−
f +D
−
g.
5c. Let f and g be nonnegative and continuous at c. Then D
+
(fg)(c) = lim
h→0
+
(fg)(c+h)−(fg)(c)
h
=
lim
h→0
+
f(c+h)g(c+h)−f(c)g(c)
h
= lim
h→0
+
(f(c+h)−f(c))g(c+h)+f(c)(g(c+h)−g(c))
h
≤ g(c)lim
h→0
+
f(c+h)−f(c)
h
+
f(c)lim
h→0
+
g(c+h)−g(c)
h
= f(c)D
+
g(c) +g(c)D
+
f(c).
*6a. Let f be deﬁned on [a, b] and g a continuous function on [α, β] that is diﬀerentiable at γ with
g(γ) = c ∈ (a, b). Suppose g
t
(γ) > 0. Note that if D
+
(f ◦ g)(γ) = ±∞, then D
+
f(c) = ±∞. Now
suppose D
+
f(c) < ∞. Let ε > 0 be given and let ε
1
= min(1,
ε
g
(γ)+1+D
+
f(c)
). There exists δ
1
> 0 such
that [
g(γ+h)−g(γ)
h
−g
t
(γ)[ < ε
1
for 0 < h < δ
1
. There exists δ
2
> 0 such that
f(c+h
)−f(c)
h
−D
+
f(c) < ε
1
for 0 < h
t
< δ
2
so that f(c+h
t
)−f(c)−h
t
D
+
f(c) < ε
1
h
t
. By continuity of g, there exists δ
3
> 0 such that
g(γ+h
tt
)−g(γ) < δ
2
for 0 < h
tt
< δ
3
. Now let δ = min(δ
1
, δ
3
). When 0 < h < δ, [
g(γ+h)−g(γ)
h
−g
t
(γ)[ < ε
1
and f(g(γ + h)) −f(g(γ)) −(g(γ + h) −g(γ))D
+
f(c) < ε
1
(g(γ + h) −g(γ)). Hence
f(g(γ+h))−f(g(γ))
h
−
D
+
f(c)g
t
(γ) =
f(g(γ+h))−f(g(γ))−D
+
f(c)(g(γ+h)−g(γ))
h
+
D
+
f(c)(g(γ+h)−g(γ))−hD
+
f(c)g
(γ)
h
< ε
1
g(γ+h)−g(γ)
h
+
D
+
f(c)(
g(γ+h)−g(γ)
h
− g
t
(γ)) < ε
1
(g
t
(γ) + 1) + ε
1
D
+
f(c) < ε. Thus D
+
(f ◦ g)(γ) ≤ D
+
f(c)g
t
(γ) and
similarly, it can be shown that D
+
(f ◦ g)(γ) ≥ D
+
f(c)g
t
(γ). Hence D
+
(f ◦ g)(γ) = D
+
f(c)g
t
(γ).
*6b. Suppose g
t
(γ) < 0. Note that if D
+
(f ◦ g)(γ) = ±∞, then D
−
f(c)g
t
(γ) = ∓∞. Also note that
there exists δ > 0 such that g(γ + h) − g(γ) < 0 for 0 < h < δ. By a similar argument to that in part
(a), D
+
(f ◦ g)(γ) = D
−
f(c)g
t
(γ).
*6c. Suppose g
t
(γ) = 0 and all the derivates of f at c are ﬁnite. By a similar argument to that in part
(a), D
+
(f ◦ g)(γ) = 0.
5.2 Functions of bounded variation
7a. Let f be of bounded variation on [a, b]. Then f = g − h where g and h are monotone increasing
functions on [a, b]. Let c ∈ (a, b). Also let A = sup
x∈[a,c)
g(x) and let B = sup
x∈[a,c)
h(x). Note that
A, B < ∞. Given ε > 0, there exists δ > 0 such that A−ε/2 < g(c−δ) ≤ A and B−ε/2 < h(c−δ) ≤ B.
Then for x ∈ (c − δ, c), A − ε/2 < g(x) ≤ A and B − ε/2 < h(x) ≤ B. i.e. 0 ≤ A − g(x) < ε/2 and
0 ≤ B − h(x) < ε/2. Now 0 ≤ [A − B − f(x)[ ≤ (A − g(x)) + (B − h(x)) < ε for x ∈ (c − δ, c). Hence
f(c−) exists. Similarly f(c+) exists. Let g be a monotone function and let E be the set of discontinuities
of g. Now for c ∈ E, g(c−) < g(c+) so there is a rational r
c
such that g(c−) < r
c
< g(c+). Note that
if x
1
< x
2
, then g(x
1
+) ≤ g(x
2
−) so r
x
1
= r
x
2
. Thus we have a bijection between E and a subset of Q
so E is countable. Since a function f of bounded variation is a diﬀerence of two monotone functions, f
also has only a countable number of discontinuities.
7b. Let 'x
n
` be an enumeration of Q ∩ [0, 1]. Deﬁne f on [0, 1] by f(x) =
¸
x
n
<x
2
−n
. Then f is
monotone. Also, at each x
n
, for any δ > 0, there exists x ∈ (x
n
, x
n
+δ) such that f(x) −f(x
n
) > 2
−n−1
so f is discontinuous at each x
n
.
8a. Suppose a ≤ c ≤ b. Let a = x
0
< x
1
< < x
n
= b be a subdivision of [a, b]. If c = x
k
for
some k, then
¸
n
1
[f(x
i
) − f(x
i−1
)[ =
¸
k
1
[f(x
i
) − f(x
i−1
)[ +
¸
n
k+1
[f(x
i
) − f(x
i−1
)[ ≤ T
c
a
(f) + T
b
c
(f).
Thus T
b
a
(f) ≤ T
c
a
(f) + T
b
c
(f). The case where c ∈ (x
k
, x
k+1
) for some k is similar. Conversely, let
a = x
0
< x
1
< < x
n
= c be a subdivision of [a, c] and let c = y
0
< y
1
< < y
m
= b be a
subdivision of [c, b]. Then a = x
0
< x
1
< < x
n
= c < y
1
< < y
m
= b is a subdivision of [a, b] and
¸
n
1
[f(x
i
) − f(x
i−1
)[ +
¸
m
1
[f(y
i
) − f(y
i−1
)[ ≤ T
b
a
(f). It follows that T
c
a
(f) + T
b
c
(f) ≤ T
b
a
(f). Hence
T
b
a
(f) = T
c
a
(f) +T
b
c
(f) and T
c
a
(f) ≤ T
b
a
(f).
8b. Let a = x
0
< x
1
< < x
n
= b be a subdivision of [a, b]. Then
¸
n
1
[(f +g)(x
i
) −(f +g)(x
i−1
)[ ≤
¸
n
1
[f(x
i
)−f(x
i−1
)[+
¸
n
1
[g(x
i
)−g(x
i−1
)[ ≤ T
b
a
(f)+T
b
a
(g). Hence T
b
a
(f +g) ≤ T
b
a
(f)+T
b
a
(g). Let c ∈ R.
If c = 0, then T
b
a
(cf) = 0 = [c[T
b
a
(f). If c = 0, then
¸
n
1
[cf(x
i
) −cf(x
i−1
)[ = [c[
¸
n
1
[f(x
i
) −f(x
i−1
)[ ≤
[c[T
b
a
(f). Thus T
b
a
(cf) ≤ [c[T
b
a
(f). On the other hand,
¸
n
1
[f(x
i
) − f(x
i−1
)[ = [c[
−1
¸
n
1
[cf(x
i
) −
cf(x
i−1
)[ ≤ [c[
−1
T
b
a
(cf). Thus T
b
a
(f) ≤ [c[
−1
T
b
a
(cf) so [c[T
b
a
(f) ≤ T
b
a
(cf). Hence T
b
a
(cf) = [c[T
b
a
(f).
9. Let 'f
n
` be a sequence of functions on [a, b] that converges at each point of [a, b] to f. Let a =
x
0
< x
1
< < x
n
= b be a subdivision of [a, b] and let ε > 0. Then there exists N such that
23
¸
n
1
[f(x
i
)−f(x
i−1
)[ ≤
¸
n
1
[f(x
i
)−f
n
(x
i
)[+
¸
n
1
[f(x
i−1
)−f
n
(x
i−1
)[+
¸
n
1
[f
n
(x
i
)−f
n
(x
i−1
)[ < ε+T
b
a
(f
n
)
for n ≥ N. Thus T
b
a
(f) ≤ ε + T
b
a
(f
n
) for n ≥ N so T
b
a
(f) ≤ ε + limT
b
a
(f
n
). Since ε is arbitrary,
T
b
a
(f) ≤ limT
b
a
(f
n
).
10a. Let f be deﬁned by f(0) = 0 and f(x) = x
2
sin(1/x
2
) for x = 0. Consider the subdivision
−1 <
2/nπ <
2/(n −1)π < <
2/π < 1 of [−1, 1]. Note that t
b
a
(f) → ∞ as n → ∞. Thus f is
not of bounded variation on [−1, 1].
10b. Let g be deﬁned by g(0) = 0 and g(x) = x
2
sin(1/x) for x = 0. Note that g is diﬀerentiable
on [−1, 1] and [g
t
(x)[ ≤ 3 on [−1, 1]. Thus for any subdivision a = x
0
< x
1
< < x
n
= b of [a, b],
¸
n
1
[g(x
i
−g(x
i−1
)[ ≤ 4
¸
n
1
[x
i
−x
i−1
[ = 3(g(1) −g(−1)). Hence T
1
−1
(g) ≤ 3(g(−1) −g(1)) < ∞.
11. Let f be of bounded variation on [a, b]. For any x ∈ [a, b], f(x) = P
x
a
(f) − N
x
a
(f) − f(a) so
f
t
(x) =
d
dx
P
x
a
(f) −
d
dx
N
x
a
(f) a.e. in [a, b]. Thus [f
t
[ ≤
d
dx
P
x
a
(f) +
d
dx
N
x
a
(f) =
d
dx
T
x
a
(f) a.e. in [a, b] and
b
a
[f
t
[ ≤
b
a
d
dx
T
x
a
(f) ≤ T
b
a
(f) −T
a
a
(f) = T
b
a
(f).
5.3 Diﬀerentiation of an integral
No problems
5.4 Absolute continuity
*12. The function f deﬁned by f(0) = 0 and f(x) = x
2
sin(1/x
2
) for x = 0 is absolutely continuous on
[ε, 1] for ε > 0, continuous at 0 but not of bounded variation on [0, 1], thus not absolutely continuous on
[0, 1].
Suppose f is absolutely continuous on [η, 1] for η > 0, continuous at 0 and of bounded variation on [0, 1].
For η ∈ (0, 1], let 0 = x
0
< x
1
< < x
n
= η be a subdivision of [0, η]. Then since f is continuous at 0,
¸
n
1
[f(x
i
)−f(x
i−1
)[ → 0 as η → 0
+
. Thus T
η
0
(f) → 0 as η → 0
+
. Given ε > 0, there exists η ∈ (0, 1] such
that T
η
0
(f) < ε/2. Since f is absolutely continuous on [η, 1], there exists δ > 0 such that for any ﬁnite
collection ¦(x
i
, x
t
i
)¦
n
1
of disjoint intervals in [η, 1] with
¸
n
1
[x
t
i
−x
i
[ < δ, we have
¸
n
1
[f(x
t
i
)−f(x
i
)[ < ε/2.
Now let ¦(y
i
, y
t
i
)¦
n
1
be a ﬁnite collection of disjoint intervals in [0, 1] with
¸
n
1
[y
t
i
−y
i
[ < δ. If η ∈ [y
t
k
, y
k+1
]
for some k, then
¸
n
1
[f(y
t
i
) − f(y
i
)[ ≤
¸
k
1
[f(y
t
i
) − f(y
i
)[ +
¸
n
k+1
[f(y
t
i
) − f(y
i
)[ < T
η
0
(f) + ε/2 < ε. If
η ∈ (y
k
, y
t
k
) for some k, then
¸
n
1
[f(y
t
i
) −f(y
i
)[ ≤
¸
k−1
1
[f(y
t
i
) −f(y
i
)[ +[f(η) −f(y
k
)[ +[f(y
t
k
) −f(η)[ +
¸
n
k+1
[f(y
t
i
) −f(y
i
)[ < T
η
0
(f) +ε/2 < ε. Hence f is absolutely continuous on [0, 1].
13. Let f be absolutely continuous on [a, b]. Then f is of bounded variation on [a, b] so by Q11,
b
a
[f
t
[ ≤ T
b
a
(f). Conversely, since f is absolutely continuous on [a, b], for any subdivision a = x
0
<
x
1
< < x
n
= b of [a, b],
¸
n
1
[f(x
i
) − f(x
i−1
)[ =
¸
n
1
[
x
i
x
i−1
f
t
[ ≤
¸
n
1
x
i
x
i−1
[f
t
[ =
b
a
[f
t
[. Thus
T
b
a
(f) ≤
b
a
[f
t
[. Hence T
b
a
(f) =
b
a
[f
t
[.
For any x ∈ [a, b], f(x) = P
x
a
(f) − N
x
a
(f) − f(a) so f
t
(x) =
d
dx
P
x
a
(f) −
d
dx
N
x
a
(f) a.e. in [a, b]. Thus
(f
t
)
+
≤ (
d
dx
P
x
a
(f))
+
+(
d
dx
N
x
a
(f))
−
=
d
dx
P
x
a
(f). Thus
b
a
(f
t
)
+
≤
b
a
d
dx
P
x
a
(f) ≤ P
b
a
(f) −P
a
a
(f) = P
b
a
(f).
Conversely, for any subdivision a = x
0
< x
1
< < x
n
= b of [a, b],
¸
n
1
(f(x
i
) − f(x
i−1
))
+
=
¸
n
1
(
x
i
x
i−1
f
t
)
+
≤
¸
n
1
x
i
x
i−1
(f
t
)
+
=
b
a
(f
t
)
+
. Thus P
b
a
(f) ≤
b
a
(f
t
)
+
. Hence P
b
a
(f) =
b
a
(f
t
)
+
.
(*) If
g < 0, then (
g)
+
= 0 ≤
g
+
. If
g ≥ 0, then (
g)
+
=
g ≤
g
+
.
14a. Let f and g be two absolutely continuous functions on [a, b]. Given ε > 0, there exists δ > 0 such
that
¸
n
1
[f(x
t
i
) −f(x
i
)[ < ε/2 and
¸
n
1
[g(x
t
i
) −g(x
i
)[ < ε/2 for any ﬁnite collection ¦(x
i
, x
t
i
)¦
n
1
of disjoint
intervals in [a, b] with
¸
n
1
[x
t
i
− x
i
[ < δ. Then
¸
n
1
[(f ± g)(x
t
i
) − (f ± g)(x
i
)[ ≤
¸
n
1
[f(x
t
i
) − f(x
i
)[ +
¸
n
1
[g(x
t
i
) −g(x
i
)[ < ε. Thus f +g and f −g are absolutely continuous.
14b. Let f and g be two absolutely continuous functions on [a, b]. There exists M such that [f(x)[ ≤ M
and [g(x)[ ≤ M for any x ∈ [a, b]. Given ε > 0, there exists δ > 0 such that
¸
n
1
[f(x
t
i
) −f(x
i
)[ < ε/2M
and
¸
n
1
[g(x
t
i
) − g(x
i
)[ < ε/2M for any ﬁnite collection ¦(x
i
, x
t
i
)¦ of disjoint intervals in [a, b] with
¸
n
1
[x
t
i
−x
i
[ < δ. Then
¸
n
1
[(fg)(x
t
i
)−(fg)(x
i
)[ ≤
¸
n
1
[f(x
t
i
)[[g(x
t
i
)−g(x
i
)[+
¸
n
1
[g(x
i
)[[f(x
t
i
)−f(x
i
)[ < ε.
Thus fg is absolutely continuous.
14c. Suppose f is absolutely continuous on [a, b] and is never zero there. Let g = 1/f. There exists M
such that [f(x)[ ≥ M for x ∈ [a, b]. Given ε > 0, there exists δ > 0 such that
¸
n
1
[f(x
t
i
) −f(x
i
)[ < εM
2
24
for any ﬁnite collection ¦(x
i
, x
t
i
)¦ of disjoint intervals in [a, b] with
¸
n
1
[x
t
i
−x
i
[ < δ. Then [g(x
t
i
)−g(x
i
)[ =
]f(x
i
)−f(x
i
)]
]f(x
i
)f(x
i
)]
< ε. Thus g is absolutely continuous.
15. Let f be the Cantor ternary function. By Q2.48, f is continuous and monotone on [0, 1]. Note
that f
t
= 0 a.e. on [0, 1] since f is constant on each interval in the complement of the Cantor set and
the Cantor set has measure zero. If f is absolutely continuous, then 1 = f(1) =
1
0
f
t
+ f(0) = 0.
Contradiction. Thus f is not absolutely continuous.
16a. Let f be a monotone increasing function on [a, b]. Let g be deﬁned by g(x) =
x
a
f
t
and let
h = f − g. Then g is absolutely continuous, h
t
(x) = f
t
(x) −
d
dx
x
a
f
t
= 0 a.e. so h is singular, and
f = g +h.
16b. Let f be a nondecreasing singular function on [a, b]. Let ε, δ > 0 be given. Since f is singular on
[a, b], for each x ∈ [a, b], there is an arbitrarily small interval [x, x+h] ⊂ [a, b] such that [f(x+h)−f(x)[ <
εh/(b −a). Then there exists a ﬁnite collection ¦[x
k
, y
k
]¦ of nonoverlapping intervals of this sort which
cover all of [a, b] except for a set of measure less than δ. Labelling x
k
such that x
k
≤ x
k+1
, we have
y
0
= a ≤ x
1
< y
1
≤ x
2
< ≤ y
n
≤ b = x
n+1
. Then
¸
n
0
[x
k+1
− y
k
[ < δ and
¸
n
1
[f(y
k
) − f(x
k
)[ < ε.
Since f is nondecreasing,
¸
n
0
[f(x
k+1
) −f(y
k
)[ > f(b) −f(a) −ε.
16c. Let f be a nondecreasing function on [a, b] with property (S). i.e. Given ε, δ > 0, there is a ﬁnite
collection ¦[y
k
, x
k
]¦ of nonoverlapping intervals in [a, b] such that
¸
n
1
[x
k
− y
k
[ < δ and
¸
n
1
(f(x
k
) −
f(y
k
)) > f(b) − f(a) − ε. By part (a), f = g + h where g =
x
a
f
t
and h is singular. It suﬃces to show
that g = 0 a.e. Letting x
0
= a and y
n+1
= b, we have
¸
n
1
(f(y
k+1
) −f(x
k
)) < ε. We may choose δ such
that
n
1
[y
k
,x
k
]
f
t
< ε. Then
b
a
f
t
< 2ε so
b
a
f
t
= 0 and g = 0.
16d. Let 'f
n
` be a sequence of nondecreasing singular functions on [a, b] such that the function f(x) =
¸
f
n
(x) is everywhere ﬁnite. Let ε, δ > 0 be given. Now f(b) − f(a) =
¸
(f
n
(b) − f
n
(a)) < ∞ so
there exists N such that
¸
∞
N+1
(f
n
(b) −f
n
(a)) < ε/2. Let F(x) =
¸
N
1
f
n
(x). Then F is nondecreasing
and singular. By part (b), there exists a ﬁnite collection ¦[y
k
, x
k
]¦ of nonoverlapping intervals such that
¸
[x
k
−y
k
[ < δ and
¸
(F(y
k
)−F(x
k
)) > F(b)−F(a)−ε/2. Now
¸
(f(y
k
)−f(x
k
)) ≥
¸
(F(y
k
)−F(x
k
)) >
F(b) − F(a) − ε/2 = f(b) − f(a) −
¸
∞
N+1
(f
n
(b) − f
n
(a)) − ε/2 > f(b) − f(a) − ε. By part (c), f is
singular.
*16e. Let C be the Cantor ternary function on [0, 1]. Extend C to R by deﬁning C(x) = 0 for x < 0 and
C(x) = 1 for x > 1. For each n, deﬁne f
n
by f
n
(x) = 2
−n
C(
x−a
n
b
n
−a
n
) where ¦[a
n
, b
n
]¦ is an enumeration of
the intervals with rational endpoints in [0, 1]. Then each f
n
is a nondecreasing singular function on [0, 1].
Deﬁne f(x) =
¸
f
n
(x). Then f is everywhere ﬁnite, strictly increasing and by part (d), f is singular.
17a. Let F be absolutely continuous on [c, d]. Let g be monotone and absolutely continuous on [a, b]
with c ≤ g ≤ d. Given ε > 0, there exists δ > 0 such that for any ﬁnite collection ¦(y
i
, y
t
i
)¦ of disjoint
intervals with
¸
n
1
[y
t
i
− y
i
[ < δ, we have
¸
n
1
[F(y
t
i
) − F(y
i
)[ < ε. Now there exists δ
t
> 0 such that for
any ﬁnite collection ¦(x
i
, x
t
i
)¦ of disjoint intervals with
¸
n
1
[x
t
i
−x
i
[ < δ
t
, we have
¸
n
1
[g(x
t
i
) −g(x
i
)[ < δ.
Now ¦(g(x
i
), g(x
t
i
))¦ is a ﬁnite collection of disjoint intervals so
¸
n
1
[F(g(x
t
i
)) − F(g(x
i
))[ < ε. Hence
F ◦ g is absolutely continuous.
(*) Additional assumption that g is monotone. Counterexample: Consider f(x) =
√
x for x ∈ [0, 1] and
g(0) = 0, g(x) = (xsin x
−1
)
2
for x ∈ (0, 1]. Then (f ◦ g)(0) = 0 and (f ◦ g)(x) = xsin x
−1
for x ∈ (0, 1].
f and g are absolutely continuous but not f ◦ g.
*17b. Let E = ¦x : g
t
(x) = 0¦. Note that [g(x) − g(a)[ = [
x
a
g
t
[ ≤
x
a
[g
t
[ for all x ∈ [a, b]. Let ε > 0.
There exists δ > 0 such that
A
[g
t
[ < ε/2 whenever mA < δ. Let η = ε/4(b − a). For any x ∈ E,
there exists h
x
> 0 such that [g(x + h) − g(x)[ < ηh for 0 < h ≤ h
x
. Deﬁne V = ¦[x, x + h
x
] : x ∈
E, [g(y) −g(x)[ < η(y −x) for y ∈ (x, x +h
x
]¦. Then V is a Vitali covering for E so there exists a ﬁnite
disjoint collection ¦I
1
, . . . , I
N
¦ of intervals in V such that m(E`
¸
N
n=1
I
n
) < δ. Now let O be an open set
such that O ⊃ E `
¸
N
n=1
I
n
and mO < δ. Then O is a countable union of disjoint open intervals J
m
and
g[E `
¸
N
n=1
I
n
] ⊂
¸
g[J
m
]. Thus m(g[E `
¸
N
n=1
I
n
)] ≤
¸
m(g[J
m
]) ≤
¸
J
m
[g
t
[ =
O
[g
t
[ < ε/2. Also,
g[E ∩
¸
N
n=1
I
n
] ⊂
¸
N
n=1
g[I
n
] so m(g[E ∩
¸
N
n=1
I
n
)] ≤
¸
2h
x
n
η < 2(b − a)η = ε/2. Hence m(g[E]) < ε.
Since ε is arbitrary, m(g[E]) = 0.
18. Let g be an absolutely continuous monotone function on [0, 1] and E a set of measure zero. Let
ε > 0. There is an open set O ⊃ E such that mO = m(O`E) < δ where δ is given by absolute continuity
25
of g. Now O is a countable union
¸
I
n
of disjoint open intervals so
¸
l(I
n
) < δ and
¸
l(g[I
n
∩[0, 1]]) < ε.
Now g[E] ⊂
¸
g[I
n
∩ [0, 1]] so m(g[E]) < ε. Since ε is arbitrary, m(g[E]) = 0.
19a. Let G be the complement of a generalised Cantor set of positive measure and let g =
x
0
χ
G
. Then
g is absolutely continuous and strictly monotone on [0, 1]. Also, g
t
= χ
G
= 0 on G
c
.
19b. Since ¦x : g
t
(x) = 0¦ has positive measure, it has a nonmeasurable subset F. By Q17(b),
m(g[F]) = 0. Also, g
−1
[g[F]] = F is nonmeasurable.
20a. Suppose f is Lipschitz. There exists M such that [f(x) − f(y)[ ≤ M[x − y[ for all x, y. Given
ε > 0, let δ = ε/M. For any ﬁnite collection ¦(x
i
, x
t
i
)¦ of nonoverlapping intervals with
¸
n
1
[x
t
i
−x
i
[ < δ,
we have
¸
n
1
[f(x
t
i
) −f(x
i
)[ ≤ M
¸
n
1
[x
t
i
−x
i
[ < ε. Thus f is absolutely continuous.
20b. Let f be absolutely continuous. Suppose f is Lipschitz. Now f
t
(x) = lim
y→x
f(y)−f(x)
y−x
so [f
t
(x)[ =
lim
y→x
[
f(y)−f(x)
y−x
[ ≤ M for all x. Conversely, if f is not Lipschitz, then for any M, there exist x and y
such that [f(x) −f(y)[ > M[x −y[. Then [f
t
(c)[ > M for some c ∈ (x, y) by the Mean Value Theorem.
Thus for any M, there exists c such that [f
t
(c)[ > M so [f
t
[ is unbounded.
*20c.
21a. Let O be an open set in [c, d]. Then O is a countable union
¸
I
n
of disjoint open intervals. Now
for each n, I
n
= (g(c
n
), g(d
n
)) for some c
n
, d
n
∈ [c, d]. Also, g
−1
[O] =
¸
g
−1
[I
n
] =
¸
(c
n
, d
n
). Thus
mO =
¸
l(I
n
) =
¸
(g(d
n
) −g(c
n
)) =
¸
d
n
c
n
g
t
=
g
−1
[O]
g
t
.
21b. Let H = ¦x : g
t
(x) = 0¦. Let E ⊂ [c, d] with mE = 0 and let δ > 0. Then there exists an open
set O ⊃ E with mO < δ. By part (a),
g
−1
[O]
g
t
< δ. Thus
g
−1
[E]∩H
g
t
=
g
−1
[E]
g
t
< δ. Since δ > 0 is
arbitrary,
g
−1
[E]∩H
g
t
= 0. Since g
t
> 0 on g
−1
[E] ∩ H, the set g
−1
[E] ∩ H has measure zero.
21c. Let E be a measurable subset of [c, d] and let F = g
−1
[E] ∩H. Since g is absolutely continuous, it
is continuous and thus measurable so g
−1
[E] is measurable. Also, g
t
is measurable so H is measurable.
Thus F = g
−1
[E] ∩ H is measurable.
There exists a G
δ
set G ⊃ E with m(G` E) = 0. We may assume G ⊂ [c, d]. By part (b), m((g
−1
[G] `
g
−1
[E]) ∩ H) = 0 so
g
−1
[G]∩H
g
t
=
g
−1
[E]∩H
g
t
. Now G is a countable intersection
¸
O
n
of open sets.
Let G
k
=
¸
k
n=1
O
n
. Then G
1
⊃ G
2
⊃ so limmG
k
= m(
¸
G
k
) = mG. Now mE = mG = limmG
k
=
lim
g
−1
[G
k
]∩H
g
t
= lim
k
n=1
g
−1
[O
n
]∩H
g
t
=
g
−1
[O
n
]∩H
g
t
=
g
−1
[G]∩H
g
t
=
g
−1
[E]∩H
g
t
=
F
g
t
. Also,
F
g
t
=
g
−1
[E]
g
t
=
b
a
χ
E
(g(x))g
t
(x) dx.
21d. Let f be a nonnegative measurable function on [c, d]. Then there is an increasing sequence
'ϕ
n
` of simple functions on [c, d] with limϕ
n
= f so limϕ
n
(g(x))g
t
(x) = f(g(x))g
t
(x). Since each
(ϕ
n
◦g)g
t
is measurable, (f ◦g)g
t
is measurable. Now
b
a
ϕ
n
(g(x))g
t
(x) dx =
¸
b
a
c
k
χ
E
k
(g(x))g
t
(x) dx =
¸
c
k
mE
k
=
d
c
ϕ
n
(y) dy. By the Monotone Convergence Theorem, lim
ϕ
n
=
f. Thus
d
c
f(y) dy =
lim
d
c
ϕ
n
(y) dy = lim
b
a
ϕ
n
(g(x))g
t
(x) dx =
b
a
f(g(x))g
t
(x) dx.
22a. F is absolutely continuous on [c, d], g is monotone and absolutely continuous on [a, b] with c ≤ g ≤ d.
By Q17(a), H = F ◦ g is absolutely continuous. Whenever H
t
and g
t
exist with g
t
(x) = 0, we have
D
+
F(g(x)) = D
+
F(g(x)) = D
−
F(g(x)) = D
−
F(g(x)) = H
t
(x)/g
t
(x) by Q6a so F
t
(g(x)) exists. Now
H
t
and g
t
exist a.e. so H
t
(x) = F
t
(g(x))g
t
(x) a.e. except on E = ¦x : g
t
(x) = 0¦.
22b. Let f
0
be deﬁned by f
0
(y) = f(y) if y / ∈ g[E] and f
0
(y) = 0 if y ∈ g[E]. By Q17b, m(g[E]) = 0 so
f
0
= f a.e. Hence H
t
(x) = f(g(x))g
t
(x) = f
0
(g(x))g
t
(x) a.e.
*22c.
*22d.
5.5 Convex functions
23a. Let ϕ be convex on a ﬁnite interval [a, b). Let x
0
∈ (a, b) and let f(x) = m(x −x
0
) +ϕ(x
0
) be the
equation of a supporting line at x
0
. Then ϕ(x) ≥ f(x) for all x ∈ (a, b). Since ϕ is continuous at a, we
have ϕ(x) ≥ f(x) ≥ min(f(a), f(b)) for all x ∈ [a, b). Hence ϕ is bounded from below.
23b. Suppose ϕ is convex on (a, b). If ϕ is monotone on (a, b), then ϕ(x) has limits (possibly inﬁnite)
as it approaches a and b respectively from within (a, b). If ϕ is not monotone, then there exists c ∈ (a, b)
26
such that D
+
ϕ(x) ≤ 0 on (a, c] and D
+
ϕ(x) ≥ 0 on [c, b) since the righthand derivative of ϕ is increasing
on (a, b). Thus ϕ is monotone on (a, c] and on [c, b) and it follows that the righthand and lefthand
limits exist at a and b respectively. If a (or b) is ﬁnite, then by part (a), the limits at a (or b) cannot be
−∞.
23c. Let ϕ be continuous on an interval I (open, closed, halfopen) and convex in the interior of I.
Then ϕ(tx + (1 − t)y) ≤ tϕ(x) + (1 − t)ϕ(y) for all x, y in the interior of I and all t ∈ [0, 1]. Since ϕ is
continuous, the inequality also holds at the included endpoints.
24. Let ϕ have a second derivative at each point of (a, b). If ϕ
tt
(x) ≥ 0 for all x ∈ (a, b), then ϕ
t
is
increasing on (a, b). Also, ϕ is continuous on (a, b). Hence ϕ is convex on (a, b). Conversely, if ϕ is convex
on (a, b), then its left and righthand derivatives are monotone increasing on (a, b) so ϕ
t
is monotone
increasing on (a, b). Hence ϕ
tt
(x) ≥ 0 for all x ∈ (a, b).
25a. Suppose a ≥ 0 and b > 0. Let ϕ(t) = (a + bt)
p
. Then ϕ is continuous on [0, ∞) for all p. For
p = 1, ϕ(t) = a + bt so ϕ
tt
(t) = 0. For 1 < p < ∞, ϕ
tt
(t) = b
2
p(p − 1)(a + bt)
p−2
> 0. For 0 < p < 1,
−ϕ
tt
(t) = −b
2
p(p − 1)(a + bt)
p−2
> 0. Hence, by Q24, ϕ is convex for 1 ≤ p < ∞ and concave for
0 < p ≤ 1.
25b. For p > 1, ϕ
tt
(t) > 0 for all t ∈ (0, ∞) so ϕ
t
is strictly increasing on (0, ∞). Now for x < y,
ϕ(λx+(1−λ)y)−ϕ(x)
(1−λ)(y−x)
= ϕ
t
(ξ
1
) for some ξ
1
∈ (x, λx+(1−λ)y). Also,
ϕ(y)−ϕ(λ(x)+(1−λ)(y))
λ(y−x)
= ϕ
t
(ξ
2
) for some
ξ
2
∈ (λx + (1 − λ)y, y). Since ϕ
t
(ξ
1
) < ϕ
t
(ξ
2
),
ϕ(λx+(1−λ)y)−ϕ(x)
(1−λ)(y−x)
<
ϕ(y)−ϕ(λ(x)+(1−λ)(y))
λ(y−x)
. Equivalently,
ϕ(λx + (1 − λ)y) < λϕ(x) + (1 − λ)ϕ(y). Hence ϕ is strictly convex for p > 1. Similarly, ϕ is strictly
concave for 0 < p < 1.
*26. Let α =
f(t) dt and let g(x) = m(x − α) + exp(α) be the equation of a supporting line at α.
Equality holds when
exp(f(t)) dt = exp(α). Now
exp(f(t)) dt − exp(α) =
exp(f(t)) dt − g(α) =
exp(f(t)) dt − g(
f(t) dt) =
(exp(f(t)) − g(f(t))) dt. Since exp(f(t)) − g(f(t)) ≥ 0, the integral is
zero only when exp(f(t)) −g(f(t)) = 0 a.e. and this can happen only when f(t) = α a.e.
27. Let 'α
n
` be a sequence of nonnegative numbers whose sum is 1 and let 'ξ
n
` be a sequence of positive
numbers. Deﬁne f on [0, 1] by f(x) = log ξ
n
if x ∈ [
¸
k−1
n=1
α
n
,
¸
k
n=1
α
n
). For each k,
¸
k
n=1
ξ
α
n
n
=
exp(
¸
k
n=1
α
n
log ξ
n
) = exp(
k
n=1
α
n
0
f(t) dt). Thus
¸
k
n=1
ξ
α
n
n
≤
k
n=1
α
n
0
exp(f(t)) dt =
¸
k
n=1
α
n
ξ
n
.
Letting k → ∞, we have
¸
∞
n=1
ξ
α
n
n
≤
¸
∞
n=1
α
n
ξ
n
.
28. Let g be a nonnegative measurable function on [0, 1]. Since log is concave,
−log(g(t)) dt ≥
−log(
g(t) dt) by Jensen’s inequality. Hence log(
g(t) dt) ≥
log(g(t)) dt.
6 The Classical Banach Spaces
6.1 The L
p
spaces
1. If [f(t)[ ≤ M
1
a.e. and [g(t)[ ≤ M
2
a.e., then [f(t) +g(t)[ ≤ M
1
+M
2
a.e. so [[f +g[[
∞
≤ M
1
+M
2
.
Note that [f(t)[ ≤ [[f[[
∞
a.e. and [g(t)[ ≤ [[g[[
∞
a.e. Thus [[f +g[[
∞
≤ [[f[[
∞
+[[g[[
∞
.
2. Let f be a bounded measurable function on [0, 1]. Now [[f[[
p
= (
1
0
[f[
p
)
1/p
≤ (
1
0
[[f[[
p
∞
)
1/p
= [[f[[
∞
.
Thus lim
p→∞
[[f[[
p
≤ [[f[[
∞
. Let ε > 0 and let E = ¦x ∈ [0, 1] : [f(x)[ > [[f[[
∞
− ε¦. Then [[f[[
p
=
(
1
0
[f[
p
)
1/p
≥ (
E
[f[
p
)
1/p
≥ ([[f[[
∞
− ε)(mE)
1/p
. If mE = 0, then [[f[[
∞
≤ [[f[[
∞
− ε. Contradiction.
Thus mE > 0 and lim
p→∞
[[f[[
p
≥ [[f[[
∞
− ε. Since ε > 0 is arbitrary, lim
p→∞
[[f[[
p
≥ [[f[[
∞
. Hence
lim
p→0
[[f[[
p
= [[f[[
∞
.
3. [[f +g[[
1
=
1
0
[f +g[ ≤
1
0
([f[ +[g[) =
1
0
[f[ +
1
0
[g[ = [[f[[
1
+[[g[[
1
.
4. Suppose f ∈ L
1
and g ∈ L
∞
. Then
[fg[ ≤
[f[[[g[[
∞
= [[g[[
∞
[f[ = [[f[[
1
[[g[[
∞
.
6.2 The Minkowski and H¨ older inequalities
5a. Let f and g be two nonnegative functions in L
p
with 0 < p < 1. We may assume [[f[[
p
> 0 and
[[g[[
p
> 0. Let α = [[f[[
p
and β = [[g[[
p
so f = αf
0
and g = βg
0
where [[f
0
[[
p
= [[g
0
[[
p
= 1. Set
λ = α/(α + β). Then 1 − λ = β/(α + β) and [f + g[
p
= (f + g)
p
= (αf
0
+ βg
0
)
p
= (α + β)
p
(λf
0
+
(1 − λ)g
0
)
p
≥ (α + β)
p
(λf
p
0
+ (1 − λ)g
p
0
) by concavity of the function ϕ(t) = t
p
for 0 < p < 1. Thus
27
[[f +g[[
p
p
≥ (α+β)
p
(λ[[f
0
[[
p
p
+(1−λ)[[g
0
[[
p
p
) = (α+β)
p
= ([[f[[
p
+[[g[[
p
)
p
. Hence [[f +g[[
p
≥ [[f[[
p
+[[g[[
p
.
5b. Suppose f ∈ L
p
and g ∈ L
p
. For 1 ≤ p ≤ ∞, f +g ∈ L
p
by the Minkowski inequality. For 0 < p < 1,
[[f +g[[
p
p
≤ [[2 max(f, g)[[
p
p
= 2
p
[[ max(f, g)[[
p
p
≤ 2
p
([[f[[
p
p
+[[g[[
p
p
). Thus f +g ∈ L
p
.
*6. Suppose 0 < p < 1. Let p
t
= 1/p so p
t
> 1. Let q be such that 1/p + 1/q = 1. Note that
q = pp
t
/(1 − p
t
). Let u = (fg)
p
and let v = g
−p
. Then fg = u
p
, f
p
= uv and g
q
= v
p
/(p
−1)
.
Let q
t
be such that 1/p
t
+ 1/q
t
= 1. By the H¨older inequality,
[uv[ ≤ [[u[[
p
[[v[[
q
. i.e.
[f[
p
≤
(
[fg[
pp
)
1/p
(
[g[
pp
/(1−p
)
)
(p
−1)/p
= (
[fg[)
p
(
[g[
q
)
1−p
. Hence
[fg[ ≥ (
[f[
p
)
1/p
(
[g[
q
)
1/q
=
[[f[[
p
[[g[[
q
.
7a. For p = ∞, [['ξ
v
+η
v
`[[
∞
= sup [ξ
v
+η
v
[ ≤ sup([ξ
v
[ +[η
v
[) ≤ sup [ξ
v
[ +sup [η
v
[ = [['ξ
v
`[[
∞
+[['η
v
`[[
∞
.
For 1 ≤ p < ∞, let α = [['ξ
v
`[[
p
and β = [['η
v
`[[
p
so 'ξ
v
` = α'ξ
t
v
` and 'η
v
` = β'η
t
v
` where [[ξ
t
v
[[
p
=
[[η
t
v
[[
p
= 1. Set λ = α/(α + β). Then 1 − λ = β/(α + β) and [['ξ
v
+ η
v
`[[
p
= (
¸
[ξ
v
+ η
v
[
p
)
1/p
≤
(
¸
([ξ
v
[ +[η
v
[)
p
)
1/p
= (
¸
(α[ξ
t
v
[ +β[η
t
v
[)
p
)
1/p
= (
¸
(α+β)
p
[λ[ξ
t
v
[ +(1−λ)[η
t
v
[]
p
)
1/p
≤ (
¸
(α+β)
p
[λ[ξ
t
v
[
p
+
(1 −λ)[η
t
v
[
p
])
1/p
= α +β = [['ξ
v
`[[
p
+[['η
v
`[[
p
.
7b. For p = 1, q = ∞,
¸
[ξ
v
η
v
[ ≤ sup [η
v
[
¸
[ξ
v
[ = [['ξ
v
`[[
1
[['η
v
`[[
∞
.
For 1 < p < ∞, let α
v
= [η
v
[
q/p
. Then [η
v
[ = α
p−1
v
and pt[ξ
v
[[η
v
[ = pt[ξ
v
[α
p−1
v
≤ (α
v
+t[ξ
v
[)
p
−α
p
v
. Thus
¸
pt[ξ
v
[[η
v
[ ≤
¸
(α
v
+ t[ξ
v
[)
p
−
¸
α
p
v
= [['α
v
+ t[ξ
v
[`[[
p
p
− [['α
v
`[[
p
p
≤ ([['α
v
`[[
p
+ t[['ξ
v
`[[
p
)
p
− [['α
v
`[[
p
p
.
Diﬀerentiating with respect to t at t = 0, we get p
¸
[ξ
v
[[η
v
[ ≤ p[['ξ
v
`[[
p
[['α
v
`[[
p−1
p
= p[['ξ
v
`[[
p
[['η
v
`[[
q
.
Hence
¸
[ξ
v
[[η
v
[ ≤ [['ξ
v
`[[
p
[['η
v
`[[
q
.
8a. Let a, b be nonnegative, 1 < p < ∞, 1/p +1/q = 1. Consider the graph of the function f(x) = x
p−1
.
The area of the rectangle bounded by the xaxis, the yaxis, x = a and y = b is ab. The area of the
region bounded by f(x), the xaxis and x = a is
a
0
x
p−1
dx =
a
p
p
. The area of the region bounded by
f(x), the yaxis and y = b is
b
0
y
1/(p−1)
dy =
b
p/(p−q)
p/(p−q)
=
b
q
q
. By comparing these areas, we see that
ab ≤
a
p
p
+
b
q
q
.
8b. We may assume that [[f[[
p
> 0 and [[g[[
q
> 0. By part (a),
]f]
]]f]]
p
]g]
]]g]]
q
≤
]f]
p
p]]f]]
p
p
+
]g]
q
q]]g]]
q
q
=
1
p
+
1
q
= 1.
Hence
[fg[ ≤ [[f[[
p
[[g[[
q
. Equality holds in part (a) if and only if b = a
p−1
. Thus equality holds here if
and only if [[f[[
p−1
p
[g[ = [[g[[
q
[f[
p−1
. Equivalently, [[f[[
p
p
[g[
q
= [[g[[
q
q
[f[
p
.
8c. Suppose 0 < p < 1 and 1/p + 1/q = 1. Let p
t
= 1/p and q
t
= −q/p. Then p
t
> 1, q
t
> 1 and
1/p
t
+ 1/q
t
= 1. Thus (ab)
p
b
−p
≤
(ab)
pp
p
+
b
−pq
q
= pab −
pb
q
q
so a
p
≤ pab −
pb
q
q
and ab ≥
a
p
p
+
b
q
q
.
8d. By a similar argument as part (b),
[fg[ ≥ [[f[[
p
[[g[[
q
.
6.3 Convergence and completeness
9. Let 'f
n
` be a convergent sequence in L
p
. There exists f ∈ L
p
such that for any ε > 0, there exists N
such that [[f
n
− f[[
p
< ε/2 for n ≥ N. Now for n, m ≥ N, [[f
n
− f
m
[[
p
≤ [[f
n
− f[[
p
+ [[f
m
− f[[
p
< ε.
Thus 'f
n
` is a Cauchy sequence.
10. Let 'f
n
` be a sequence of functions in L
∞
. Suppose [[f
n
−f[[
∞
→ 0. Given ε > 0, there exists N such
that inf¦M : m¦t : [f
n
(t)−f(t)[ > M¦ = 0¦ < ε for n ≥ N. Thus m¦t : [f
n
(t)−f(t)[ ≥ ε¦ = 0 for n ≥ N.
Let E = ¦t : [f
n
(t) − f(t)[ ≥ ε¦. Then mE = 0 and 'f
n
` converges uniformly to f on E
c
. Conversely,
suppose there exists a set E with mE = 0 and 'f
n
` converges uniformly to f on E
c
. Given ε > 0, there
exists N such that [f
n
(t) − f(t)[ < ε/2 for n ≥ N and t ∈ E
c
. Thus ¦t : [f
n
(t) − f(t)[ > ε/2¦ ⊂ E for
n ≥ N. Hence inf¦M : m¦t : [f
n
(t) −f(t)[ > M¦ = 0¦ < ε for n ≥ N. i.e. [[f
n
−f[[
∞
→ 0.
11. Let 'f
n
` be a Cauchy sequence in L
∞
. Given ε > 0, there exists N such that inf¦M : m¦t :
[f
n
(t) −f
m
(t)[ > M¦ = 0¦ = [[f
n
−f
m
[[
∞
< ε/2 for n, m ≥ N. Thus for n, m ≥ N, there exists M < ε/2
such that m¦t : [f
n
(t) −f
m
(t)[ > M¦ = 0 so m¦t : [f
n
(t) −f
m
(t)[ > ε/2¦ = 0. Then 'f
n
` converges a.e.
to a function f and [f
n
− f[ < ε/2 a.e. for n ≥ N so [f[ ≤ [f
N
[ + ε/2 a.e. and f ∈ L
∞
. Furthermore,
inf¦M : m¦t : [f
n
(t) −f(t)[ > M¦ = 0¦ < ε for n ≥ N. i.e. [[f
n
−f[[
∞
→ 0.
12. Let 1 ≤ p < ∞ and let 'ξ
(n)
v
` be a Cauchy sequence in
p
. Given ε > 0, there exists N such that
¸
[ξ
(n)
v
−ξ
(m)
v
[
p
< ε
p
for n, m ≥ N. In particular, [ξ
(n)
v
−ξ
(m)
v
[
p
< ε
p
for n, m ≥ N and each v. Thus for
each v, 'ξ
(n)
v
` is Cauchy in R so it converges to some ξ
v
. Consider 'ξ
v
`. Then
¸
k
v=1
[ξ
(n)
v
−ξ
v
[
p
< ε
p
for
each k and each n ≥ N so
¸
[ξ
(n)
v
−ξ
v
[
p
< ε
p
for n ≥ N. Thus 'ξ
(n)
v
−ξ
v
` ∈
p
for n ≥ N so 'ξ
v
` ∈
p
28
and [['ξ
(n)
v
` −'ξ
v
`[[
p
→ 0.
13. Let C = C[0, 1] be the space of all continuous functions on [0, 1] and deﬁne [[f[[ = max [f(x)[ for
f ∈ C. It is straightforward to check that [[ [[ is a norm on C. Let 'f
n
` be a Cauchy sequence in C.
Given ε > 0, there exists N such that max [f
n
(x) − f
m
(x)[ < ε for n, m ≥ N so [f
n
(x) − f
m
(x)[ < ε
for n, m ≥ N and x ∈ [0, 1]. Thus 'f
n
(x)` converges to some f(x) for each x ∈ [0, 1]. Furthermore, the
convergence is uniform. Thus f ∈ C. Also, max [f
n
(x) −f(x)[ < ε for n ≥ N. i.e. [[f
n
−f[[ → 0.
14. It is straightforward to check that [[ [[
∞
is a norm on
∞
. Let 'ξ
(n)
v
` be a Cauchy sequence in
∞
.
Given ε > 0, there exists N such that sup [ξ
(n)
v
− ξ
(m)
v
[ < ε for n, m ≥ N. Then [ξ
(n)
v
− ξ
(m)
v
[ < ε for
each v and n, m ≥ N. Thus 'ξ
(n)
v
` converges to some ξ
v
for each v and [ξ
(n)
v
−ξ
v
[ < ε for n ≥ N. Then
[ξ
v
[ ≤ [ξ
(N)
v
[ +ε for each v and 'ξ
v
` ∈
∞
. Also, sup [ξ
(n)
v
−ξ
v
[ < ε for n ≥ N. i.e. [['ξ
(n)
v
` −'ξ
v
`[[
∞
→ 0.
15. Let c be the space of all convergent sequences of real numbers and let c
0
be the space of all sequences
which converge to 0. It is straightforward to check that [[ [[
∞
is a norm on c and c
0
. Let 'ξ
(n)
v
` be a
Cauchy sequence in c. Given ε > 0, there exists N such that sup [ξ
(n)
v
− ξ
(m)
v
[ < ε for n, m ≥ N. Then
[ξ
(n)
v
−ξ
(m)
v
[ < ε for each v and n, m ≥ N. Thus 'ξ
(n)
v
` converges to some ξ
v
for each v. Now for each v
and v
t
, there exists N
t
such that [ξ
v
−ξ
v
[ ≤ [ξ
v
−ξ
(N
)
v
[ +[ξ
(N
)
v
−ξ
(N
)
v
[ +[ξ
v
−ξ
(N
)
v
[ < ε. Thus 'ξ
v
`
is Cauchy in R. Hence 'ξ
v
` ∈ c and sup [ξ
(n)
v
−ξ
v
[ < ε. i.e. [['ξ
(n)
v
` −'ξ
v
`[[
∞
→ 0. If 'ξ
(n)
v
` is a Cauchy
sequence in c
0
, then 'ξ
v
` converges to 0 since [ξ
v
[ ≤ [ξ
(n)
v
−ξ
v
[ +[ξ
(n)
v
[.
16. Let 'f
n
` be a sequence in L
p
, 1 ≤ p < ∞, which converges a.e. to a function f in L
p
. Suppose
[[f
n
− f[[
p
→ 0. Then [[f
n
[[
p
→ [[f[[
p
since [ [[f
n
[[
p
− [[f[[
p
[ ≤ [[f
n
− f[[
p
. Conversely, suppose
[[f
n
[[
p
→ [[f[[
p
. Now 2
p
([f
n
[
p
+ [f[
p
) − [f
n
− f[
p
≥ 0 for each n so by Fatou’s Lemma,
2
p+1
[f[
p
≤
lim
2
p
([f
n
[
p
+[f[
p
) −[f
n
−f[
p
=
2
p+1
[f[
p
−lim
[f
n
−f[
p
. Thus lim
[f
n
−f[
p
≤ 0 ≤ lim
[f
n
−f[
p
.
Hence [[f
n
−f[[
p
→ 0.
17. Let 'f
n
` be a sequence in L
p
, 1 < p < ∞, which converges a.e. to a function f in L
p
. Suppose there
is a constant M such that [[f
n
[[
p
≤ M for all n. Let g ∈ L
q
. Given ε > 0, there exists δ > 0 such that
E
[g[
q
< (ε/4M)
q
whenever mE < δ. By Egoroﬀ’s Theorem, there exists E such that mE < δ and 'f
n
`
converges uniformly to f on E
c
. Thus there exists N such that [f
n
(x) − f(x)[ < ε/(2(mE
c
)
1/p
[[g[[
q
)
for n ≥ N and x ∈ E
c
. Now [
f
n
g −
fg[ ≤
[f
n
− f[[g[ ≤ (
E
[f
n
− f[
p
)
1/p
(
E
[g[
q
)
1/q
+ (
E
c
[f
n
−
f[
p
)
1/p
(
E
c
[g[
q
)
1/q
≤ 2M(ε/4m) + (ε/(2(mE
c
)
1/p
[[g[[
q
))(mE
c
)
1/p
[[g[[
q
= ε for n ≥ N. i.e.
fg =
lim
f
n
g.
For p = 1, it is not true. Let f
n
= nχ
[0,1/n]
for each n. Then f
n
→ 0 and [[f
n
[[
1
= 1 for each n. Let
g = χ
[0,1]
∈ L
∞
. Then
fg = 0 but
f
n
g =
f
n
= 1 for each n.
18. Let f
n
→ f in L
p
, 1 ≤ p < ∞ and let 'g
n
` be a sequence of measurable functions such that [g
n
[ ≤ M
for all n and g
n
→ g a.e. Given ε > 0, there exists δ > 0 such that
E
[f[
p
< (ε/8M)
p
whenever
mE < δ. By Egoroﬀ’s Theorem, there exists E such that mE < δ and 'g
n
` converges uniformly to
g on E
c
. Thus there exists N such that [[f
n
− f[[
p
< ε/2M and [g
n
(x) − g(x)[ < ε/(4(mE
c
)
1/p
[[f[[
p
)
for n ≥ N and x ∈ E
c
. Now [[g
n
f
n
− gf[[
p
≤ [[g
n
f
n
− g
n
f[[
p
+ [[g
n
f − gf[[
p
= (
[g
n
[
p
[f
n
− f[
p
)
1/p
+
(
[g
n
−g[
p
[f[
p
)
1/p
≤ M[[f
n
−f[[
p
+(
E
[g
n
−g[
p
[f[
p
)
1/p
+(
E
c
[g
n
−g[
p
[f[
p
)
1/p
< ε/2 +(ε/8M)(2M) +
ε/(4(mE
c
)
1/p
[[f[[
p
)(mE
c
)
1/p
[[f[[
p
= ε for n ≥ N. Thus [[g
n
f
n
−gf[[
p
→ 0.
6.4 Approximation in L
p
*19. [[T
∆
(f)[[
p
p
= [[
¸
m−1
k=0
1
ξ
k+1
−ξ
k
(
ξ
k+1
ξ
k
f)χ
[ξ
k
,ξ
k+1
)
[[
p
p
≤
¸
m−1
k=0
[[
1
ξ
k+1
−ξ
k
(
ξ
k+1
ξ
k
f)χ
[ξ
k
,ξ
k+1
)
[[
p
p
. Now
[[
1
ξ
k+1
−ξ
k
(
ξ
k+1
ξ
k
f)χ
[ξ
k
,ξ
k+1
)
[[
p
p
=
b
a
[
1
ξ
k+1
−ξ
k
(
ξ
k+1
ξ
k
f)χ
[ξ
k
,ξ
k+1
)
[
p
=
ξ
k+1
ξ
k
1
(ξ
k+1
−ξ
k
)
p
[
ξ
k+1
ξ
k
f[
p
. By the
H¨older inequality, [
ξ
k+1
ξ
k
f[
p
≤
ξ
k+1
ξ
k
[f[
p
(
ξ
k+1
ξ
k
1)
p/q
=
ξ
k+1
ξ
k
[f[
p
(
ξ
k+1
ξ
k
1)
p−1
. Thus [[T
∆
(f)[[
p
p
≤
¸
m−1
k=0
ξ
k+1
ξ
k
1
(ξ
k+1
−ξ
k
)
p
ξ
k+1
ξ
k
[f[
p
(
ξ
k+1
ξ
k
1)
p−1
=
¸
m−1
k=0
1
(ξ
k+1
−ξ
k
)
p−1
ξ
k+1
ξ
k
[f[
p
(ξ
k+1
−ξ
k
)
p−1
=
b
a
[f[
p
.
Hence [[T
∆
(f)[[
p
p
≤ [[f[[
p
p
and [[T
∆
(f)[[
p
≤ [[f[[
p
.
*20. By Chebyshev’s inequality, for any ε > 0,
[ϕ
∆
− f[
p
≥ ε
p
m¦x : [ϕ
∆
(x) − f(x)[
p
> ε
p
¦. Thus
m¦x : [ϕ
∆
(x) −f(x)[ > ε¦ = m¦x : [ϕ
∆
(x) −f(x)[
p
> ε
p
¦ ≤ ε
−p
[[ϕ
∆
−f[[
p
p
< ε for suﬃciently small δ.
29
6.5 Bounded linear functionals on the L
p
spaces
21a. Let g be an integrable function on [0, 1]. If [[g[[
1
= 0, let f = sgn(g). Then f is a bounded
measurable function, [[f[[
∞
= 1 and
fg =
[g[ = [[g[[
1
[[f[[
∞
. If [[g[[
1
= 0, then g = 0 a.e. Let f = 1.
Then f is a bounded measurable function, [[f[[
∞
= 1 and
fg =
g = 0 = [[g[[
1
[[f[[
∞
.
21b. Let g be a bounded measurable function. Given ε > 0, let E = ¦x : g(x) > [[g[[
∞
− ε¦ and let
f = χ
E
. Then
fg =
E
g ≥ ([[g[[
∞
−ε)mE = ([[g[[
∞
−ε)[[f[[
1
.
22. Let F be a bounded linear functional on
p
. For each v, let e
v
be the sequence with 1 in the
vth entry and 0 elsewhere. For p = 1, note that [['ξ
v
` −
¸
n
v=1
ξ
v
e
v
[[
1
→ 0 for each 'ξ
v
` ∈
1
so
F('ξ
v
`) =
¸
ξ
v
F(e
v
) by linearity and continuity of F. Now [F(e
v
)[ = [F(e
v
)[/[[e
v
[[
1
≤ [[F[[ for all v so
'F(e
v
)` ∈
∞
and [['F(e
v
)`[[
∞
≤ [[F[[. Conversely, [F('ξ
v
`)[ = [
¸
ξ
v
F(e
v
)[ ≤ [['ξ
v
`[[
1
[['F(e
v
)`[[
∞
so
[['F(e
v
)`[[
∞
≥ [F('ξ
v
`)[/[['ξ
v
`[[
1
for all nonzero 'ξ
v
` ∈
1
. Thus [['F(e
v
)`[[
∞
≥ [[F[[.
For 1 < p < ∞, note that [['ξ
v
` −
¸
n
v=1
ξ
v
e
v
[[
p
→ 0 for each 'ξ
v
` ∈
p
so F('ξ
v
`) =
¸
ξ
v
F(e
v
)
by linearity and continuity of F. For each v, let x
v
= [F(e
v
)[
q
/F(e
v
) = [F(e
v
)[
q−1
sgn(F(e
v
)) and let
x = 'x
1
, . . . , x
n
, 0, 0, . . .`. Then F(x) =
¸
n
v=1
x
v
F(e
v
) =
¸
n
v=1
[F(e
v
)[
q
. Now [[x[[
p
= (
¸
n
v=1
[x
v
[
p
)
1/p
=
(
¸
n
v=1
[F(e
v
)[
p(q−1)
)
1/p
= (
¸
n
v=1
[F(e
v
)[
q
)
1/p
so (
¸
n
v=1
[F(e
v
)[
q
)
1/q
=
n
v=1
]F(e
v
)]
q
(
n
v=1
]F(e
v
)]
q
)
1/p
=
]F(x)]
]]x]]
p
≤
[[F[[ for all n. Thus 'F(e
v
)` ∈
q
and [['F(e
v
)`[[
q
≤ [[F[[. Conversely, [F('ξ
v
`)[ = [
¸
ξ
v
F(e
v
)[ ≤
[['ξ
v
`[[
p
[['F(e
v
)`[[
q
so [['F(e
v
)`[[
q
≥ [F('ξ
v
`)[/[['ξ
v
`[[
p
for all nonzero 'ξ
v
` ∈
p
. Thus [['F(e
v
)`[[
q
≥ [[F[[.
23. Note that [['ξ
v
` −
¸
n
v=1
ξ
v
e
v
−
¸
∞
v=n+1
ξe
v
[[
∞
→ 0 for each 'ξ
v
` ∈ c with limξ
v
= ξ so F('ξ
v
`) =
¸
ξ
v
F(e
v
) by linearity and continuity of F. For each v, let x
v
= sgn(F(e
v
)) and let x be deﬁned as
before. Then F(x) =
¸
n
v=1
x
v
F(e
v
) =
¸
n
v=1
[F(e
v
)[ so
¸
n
v=1
[F(e
v
)[ = [F(x)[/[[x[[
∞
≤ [[F[[ for all n.
Thus 'F(e
v
)` ∈
1
and [['F(e
v
)`[[
1
≤ [[F[[. Conversely, [F('ξ
v
`)[ = [
¸
ξ
v
F(e
v
)[ ≤ [['ξ
v
`[[
∞
[['F(e
v
)`[[
1
so [['F(e
v
)`[[
1
≥ [F('ξ
v
`)[/[['ξ
v
`[[
∞
for all nonzero 'ξ
v
` ∈ c. Thus [['F(e
v
)`[[
1
≥ [[F[[. Similarly for c
0
.
*24. Let F be a bounded linear functional on L
p
, 1 ≤ p < ∞, and suppose there exist functions g and
h in L
q
such that F(f) =
fg =
fh for all f ∈ L
p
. For p > 1, choose f = [g − h[
q−2
(g − h). Then
[f[
p
= [g −h[
p(q−1)
= [g −h[
q
so f ∈ L
p
. Now
[g −h[
q−2
(g −h)g =
[g −h[
q−2
(g −h)h so
[g −h[
q
= 0.
Thus g = h a.e. For p = 1, choose f = sgn(g −h). Then f ∈ L
1
and f(g −h) = [g −h[. Now
fg =
fh
so
[g −h[ =
f(g −h) = 0. Thus g = h a.e.
7 Metric Spaces
7.1 Introduction
1a. Clearly, ρ
∗
(x, y) ≥ 0 for all x, y. Now ρ
∗
(x, y) = 0 if and only if [x
i
−y
i
[ = 0 for all i if and only if
x
i
= y
i
for all i if and only if x = y. Since [x
i
− y
i
[ = [y
i
− x
i
[ for each i, ρ
∗
(x, y) = ρ
∗
(y, x). Finally,
ρ
∗
(x, y) =
¸
n
i=1
[x
i
−y
i
[ ≤
¸
n
i=1
([x
i
−z
i
[ +[z
i
−y
i
[) =
¸
n
i=1
[x
i
−z
i
[ +
¸
n
i=1
[z
i
−y
i
[ = ρ
∗
(x, z)+ρ
∗
(z, y).
The argument for ρ
+
is similar except for the last property. For any x
j
, y
j
, z
j
, [x
j
− y
j
[ ≤ [x
j
− z
j
[ +
[z
j
−y
j
[ ≤ max
i
[x
i
−z
i
[ + max
i
[z
i
−y
i
[ = ρ
+
(x, z) +ρ
+
(z, y). Thus ρ
+
(x, y) ≤ ρ
+
(x, z) +ρ
+
(z, y).
1b. For n = 2 (resp. n = 3), ¦x : ρ(x, y) < 1¦ is the interior of the circle (resp. sphere) with center
y and radius 1. ¦x : ρ
∗
(x, y) < 1¦ is the interior of the diamond (resp. bipyramid) with center y with
height and width 2. ¦x : ρ
+
(x, y) < 1¦ is the interior of the square (resp. cube) with center y and sides
of length 1.
2. Suppose 0 < ε < δ − ρ(x, z) and y ∈ S
z,ε
. Then ρ(z, y) < ε < δ − ρ(x, z). Hence ρ(x, y) ≤
ρ(x, z) +ρ(z, y) < δ so y ∈ S
x,δ
.
3a. For any x, ρ(x, x) = 0. If ρ(x, y) = 0, then ρ(y, x) = ρ(x, y) = 0. If ρ(x, z) = 0 and ρ(z, y) = 0,
then 0 ≤ ρ(x, y) ≤ ρ(x, z) + ρ(z, y) = 0 so ρ(x, y) = 0. Thus ρ(x, y) = 0 is an equivalence relation. Let
X
∗
be the set of equivalence classes under this relation. Suppose x and x
t
are in the same equivalence
class. Also suppose that y and y
t
are in the same equivalence class. Then ρ(x, y) ≤ ρ(x, x
t
) +ρ(x
t
, y) =
ρ(x
t
, y) ≤ ρ(x
t
, y
t
) + ρ(y
t
, y) = ρ(x
t
, y
t
). If ρ(x, y) = 0, then x and y are in the same equivalence class.
Thus ρ deﬁnes a metric on X
∗
.
3b. Let ρ be an extended metric on X. For any x, ρ(x, x) = 0 < ∞. If ρ(x, y) < ∞, then ρ(y, x) =
ρ(x, y) < ∞. If ρ(x, z) < ∞and ρ(z, y) < ∞, then ρ(x, y) ≤ ρ(x, z)+ρ(z, y) < ∞. Thus ρ(x, y) < ∞is an
equivalence relation. Let X
α
0
be a part of the extended metric space (X, ρ) and let x be a representative
30
of X
α
0
. If y ∈ X
α
0
, then for any z ∈ S
y,δ
with δ > 0, ρ(x, z) ≤ ρ(x, y) + ρ(y, z) < ∞ so z ∈ X
α
0
. i.e.
S
y,δ
⊂ X
α
0
. Thus X
α
0
is open. Now X =
¸
X
α
so X
α
0
= X `
¸
α,=α
0
X
α
. Since the union is open, X
α
0
is closed.
7.2 Open and closed sets
4a. Since continuous functions on [0, 1] are bounded, C ⊂ L
∞
. By Q6.3.13, C is complete. Let g be
a point of closure of C. For every n, there exists f
n
∈ C such that [[f
n
− g[[
∞
< 1/n. Then 'f
n
` is a
Cauchy sequence in C and it converges to g so g ∈ C. Thus C is a closed subset of L
∞
.
4b. Let g be a point of closure of the set of integrable functions that vanish for 0 ≤ t < 1/2. For each
n, there exists f
n
in the set such that [[f
n
− g[[
1
< 1/n. Then
1/2
0
[g[ ≤
1/2
0
[f
n
− g[ +
1/2
0
[f
n
[ ≤
1
0
[f
n
− g[ < 1/n for all n. Hence g vanishes a.e. and the set of integrable functions that vanish for
0 ≤ t < 1/2 is closed in L
1
.
4c. Let x(t) be a measurable function with
x < 1. Let δ = 1 −
x. For any y ∈ S
x,δ
,
[y −x[ < δ so
y =
(y −x) +
x < δ +
x < 1. Hence the set of measurable functions x(t) with
x < 1 is open in
L
1
.
5. If E ⊂ F and F is closed, then
¯
E ⊂
¯
F = F. Thus
¯
E ⊂
¸
E⊂F
F for closed sets F. Conversely, since
¯
E is a closed set containing E,
¸
E⊂F
F ⊂
¯
E.
5a. From the deﬁnition, E
◦
is an open subset of E so E
◦
⊂
¸
O⊂E
O. Conversely, for any open subset
O ⊂ E and y ∈ O, there exists δ > 0 such that x ∈ O ⊂ E for all x with ρ(x, y) < δ. Thus O ⊂ E
◦
for
any open subset O ⊂ E so
¸
O⊂E
⊂ E
◦
.
5b. (
¯
E)
c
= (
¸
E⊂F
F)
c
=
¸
E⊂F
F
c
=
¸
F
c
⊂E
c
F
c
= (E
c
)
◦
.
6a. Consider the ball S
y,δ
= ¦x : ρ(x, y) < δ¦. For any x ∈ S
y,δ
, let 0 < ε < δ − ρ(x, y). By Q2,
S
x,ε
⊂ S
y,δ
so S
y,δ
is open.
6b. Consider the set S = ¦x : ρ(x, y) ≤ δ¦. Take x ∈ S
c
. Then ρ(x, y) > δ. Let δ
t
= ρ(x, y) − δ. For
any z ∈ S
x,δ
, ρ(z, y) ≥ ρ(x, y) −ρ(x, z) > ρ(x, y) −δ
t
= δ so z ∈ S
c
. Thus S
c
is open and S is closed.
6c. The set in part (b) is not always the closure of the ball ¦x : ρ(x, y) < δ¦. For example, let X be any
metric space with [X[ > 1 and ρ being the discrete metric. i.e. ρ(x, y) = 1 if x = y and ρ(x, y) = 0 if
x = y. Then ¦x : ρ(x, y) < 1¦ = ¦y¦, ¦x : ρ(x, y) < 1¦ = ¦y¦ and ¦x : ρ(x, y) ≤ 1¦ = X.
*7. R
n
is separable with the set of ntuples of rational numbers being a countable dense subset. C is
separable with the set of polynomials on [0, 1] with rational coeﬃcients being a countable dense subset
(Weierstrass approximation theorem).
L
∞
is not separable. Consider χ
[0,x]
and χ
[0,y]
where x, y ∈ [0, 1]. Then [[χ
[0,x]
−χ
[0,y]
[[
∞
= 1 if x = y.
If there exists a countable dense subset D, then there exists d ∈ D and x, y ∈ [0, 1] with x = y such that
[[χ
[0,x]
−d[[
∞
< 1/2 and [[χ
[0,y]
−d[[
∞
< 1/2. Then [[χ
[0,x]
−χ
[0,y]
[[
∞
< 1. Contradiction.
L
1
is separable. By Proposition 6.8, given f ∈ L
1
and ε > 0, there exists a step function ϕ such that
[[f − ϕ[[
1
< ε. We may further approximate ϕ by a step function where the partition intervals have
rational endpoints and the coeﬃcients are rational to get a countable dense subset.
7.3 Continuous functions and homeomorphisms
8. Let h be the function on [0, 1) given by h(x) = x/(1 −x). The function h, being a rational function,
is continuous on [0, 1). If h(x) = h(y), then x(1 −y) = y(1 −x) so x = y. Thus h is onetoone. For any
y ∈ [0, ∞), let x = y/(1 + y). Then x ∈ [0, 1) and h(x) = y. Thus h is onto. The inverse function h
−1
is given by h
−1
(x) = x/(1 + x), which is continuous. Hence h is a homeomorphism between [0, 1) and
[0, ∞).
9a. For a ﬁxed set E, let f(x) = ρ(x, E) = inf
y∈E
ρ(x, y). Given ε > 0, let δ = ε. When ρ(x, z) < δ,
take any y ∈ E. Then f(x) = ρ(x, E) ≤ ρ(x, y) ≤ ρ(x, z) + ρ(z, y) < δ + ρ(z, E) ≤ ε + f(z). Thus
f(x) −f(z) < ε. Similarly, by interchanging x and z, f(z) −f(x) < ε. Thus [f(x) −f(z)[ < ε and f is
continuous.
9b. If ρ(x, E) = 0, then for any δ > 0, there exists y ∈ E such that ρ(x, y) < δ so x ∈
¯
E. Conversely, if
ρ(x, E) > 0, say ρ(x, E) = α, then ρ(x, y) > α/2 for all y ∈ E so x / ∈
¯
E. Hence ¦x : ρ(x, E) = 0¦ =
¯
E.
31
10a. Suppose ρ and σ are equivalent metrics on X. The identity mapping is a homeomorphism between
(X, ρ) and (X, σ). Thus given x ∈ X and ε > 0, there exists δ > 0 such that if ρ(x, y) < δ, then
σ(x, y) < ε. By considering the inverse function, we see that if σ(x, y) < δ, then ρ(x, y) < ε. Conversely,
the two implications show that the identity mapping is continuous from (X, ρ) to (X, σ) as well as from
(X, σ) to (X, ρ). Since the identity mapping is clearly bijective, it is a homeomorphism so ρ and σ are
equivalent metrics.
10b. Given ε > 0, let δ = ε/n. When ρ
+
(x, y) < δ, ρ
∗
(x, y) < nδ = ε. When ρ
∗
(x, y) < δ, ρ
+
(x, y) <
δ < ε. Thus ρ
+
and ρ
∗
are equivalent metrics. When ρ
+
(x, y) < δ, ρ(x, y) < (nδ
2
)
1/2
= (ε
2
/n)
1/2
< ε.
When ρ(x, y) < δ, (ρ
+
(x, y))
2
< δ
2
so ρ
+
(x, y) < δ < ε. Thus ρ
+
and ρ are equivalent metrics. There
exists δ
t
> 0 with δ
t
< ε/
√
n such that ρ
∗
(x, y) < δ
t
implies ρ
+
(x, y) < δ. Then when ρ
∗
(x, y) < δ
t
,
ρ(x, y) < ε. When ρ(x, y) < δ
t
, ρ
∗
(x, y) ≤ ρ(x, y)
√
n < ε. Thus ρ and ρ
∗
are equivalent metrics.
10c. Consider the discrete metric ψ. Let x = (0, . . . , 0). For any δ > 0, we can choose y = x such that
ρ(x, y) < δ but ψ(x, y) = 1. Similarly for ρ
∗
and ρ
+
. Thus ψ is not equivalent to the metrics in part (b).
11a. Let ρ be a metric on a set X and let σ = ρ/(1 + ρ). Clearly, σ(x, y) ≥ 0 with σ(x, y) = 0 if and
only if x = y. Also, σ(x, y) = σ(y, x). Now σ(x, y) =
ρ(x,y)
1+ρ(x,y)
= 1 −
1
1+ρ(x,y)
≤ 1 −
1
1+ρ(x,z)+ρ(z,y)
=
ρ(x,z)+ρ(z,y)
1+ρ(x,z)+ρ(z,y)
≤ σ(x, z) + σ(z, y). Hence σ is a metric on X. Note that ρ(x, y) =
σ(x,y)
1−σ(x,y)
= h(σ(x, y))
where h is the function in Q8. Since h is continuous at 0, given ε > 0, there exists δ
t
> 0 such
that h(σ(x, y)) < ε when σ(x, y) < δ
t
. Now given ε > 0, let δ < min(δ
t
, ε). When ρ(x, y) < δ,
σ(x, y) < ρ(x, y) < δ < ε. When σ(x, y) < δ, ρ(x, y) = h(σ(x, y)) < ε. Hence σ and ρ are equivalent
metrics for X. Furthermore, σ(x, y) ≤ 1 for all x, y ∈ X so (X, σ) is a bounded metric space.
11b. If ρ is an extended metric (resp. pseudometric), then σ is an extended metric (resp. pseudometric).
The rest of the argument in part (a) follows.
7.4 Convergence and completeness
12. Suppose the sequence 'x
n
` has x as a cluster point. There exists n
1
≥ 1 such that ρ(x, x
n
1
) < 1.
Suppose x
n
1
, . . . , x
n
k
have been chosen. There exists n
k+1
≥ n
k
such that ρ(x, x
n
k+1
) < 1/(k + 1). The
subsequence 'x
n
k
` converges to x. Conversely, suppose there is a subsequence 'x
n
k
` that converges to
x. Given ε > 0 and given N, there exists N
t
such that ρ(x, x
n
k
) < ε for k ≥ N
t
. Pick k ≥ max(N, N
t
).
Then n
k
≥ k ≥ N and ρ(x, x
n
k
) < ε. Thus x is a cluster point of the sequence 'x
n
`.
13. Suppose the sequence 'x
n
` converges to x. Then every subsequence of 'x
n
` also converges to x
and so has x as a cluster point. Conversely, suppose 'x
n
` does not converge to x. There exists ε > 0
such that for each N, there exists n ≥ N with ρ(x, x
n
) ≥ ε. Pick n
1
such that ρ(x, x
n
1
) ≥ ε. Suppose
x
n
1
, . . . , x
n
k
have been chosen. Then pick n
k+1
≥ n
k
such that ρ(x, x
n
k+1
) ≥ ε. The subsequence 'x
n
k
`
does not have x as a cluster point.
If every subsequence of 'x
n
` has in turn a subsequence that converges to x, then every subsequence of
'x
n
` has x as a cluster point by Q12. Hence the sequence 'x
n
` converges to x.
14. Let E be a set in a metric space X. If x is a cluster point of a sequence from E, then given ε > 0,
there exists n such that ρ(x, x
n
) < ε. Since x
n
∈ E, x ∈
¯
E. On the other hand, if x ∈
¯
E, then for each
n, there exists x
n
∈ E with ρ(x, x
n
) < 1/n. The sequence 'x
n
` converges to x.
15. Suppose a Cauchy sequence 'x
n
` in a metric space has a cluster point x. By Q12, there is a
subsequence 'x
n
k
` that converges to x. Given ε > 0, there exists N such that ρ(x, x
n
k
) < ε/2 for k ≥ N
and ρ(x
n
, x
m
) < ε/2 for n, m ≥ N. Now for k ≥ N, ρ(x, x
k
) ≤ ρ(x, x
n
k
) + ρ(x
n
k
, x
k
) < ε. Thus 'x
n
`
converges to x.
16. Let X and Y be metric spaces and f a mapping from X to Y . Suppose f is continuous at x and let
'x
n
` be a sequence in X that converges to x. Given ε > 0, there exists δ > 0 such that σ(f(x), f(y)) < ε
if ρ(x, y) < δ. There also exists N such that ρ(x, x
n
) < δ for n ≥ N. Thus σ(f(x), f(x
n
)) < ε for n ≥ N
so the sequence 'f(x
n
)` converges to f(x) in Y . Conversely, suppose f is not continuous at x. Then
there exists ε > 0 such that for every n, there exists x
n
with ρ(x, x
n
) < 1/n but σ(f(x), f(x
n
)) ≥ ε. The
sequence 'x
n
` converges to x but 'f(x
n
)` does not converge to f(x).
17a. Let 'x
n
` and 'y
n
` be Cauchy sequences from a metric space X. Given ε > 0, there exists N such that
ρ(x
n
, x
m
) < ε/2 and ρ(y
n
, y
m
) < ε/2 for n, m ≥ N. Now ρ(x
n
, y
n
) ≤ ρ(x
n
, x
m
)+ρ(x
m
, y
m
)+ρ(y
m
, y
n
) so
32
ρ(x
n
, y
n
) −ρ(x
m
, y
m
) ≤ ρ(x
n
, x
m
) +ρ(y
n
, y
m
). Similarly, ρ(x
m
, y
m
) −ρ(x
n
, y
n
) ≤ ρ(x
n
, x
m
) +ρ(y
n
, y
m
).
Thus [ρ(x
n
, y
n
) −ρ(x
m
, y
m
)[ ≤ ρ(x
n
, x
m
) +ρ(y
n
, y
m
) < ε for n, m ≥ N. Hence the sequence 'ρ(x
n
, y
n
)`
is Cauchy in R and thus converges.
17b. Deﬁne ρ
∗
('x
n
`, 'y
n
`) = limρ(x
n
, y
n
) for Cauchy sequences 'x
n
` and 'y
n
`. Then ρ
∗
('x
n
`, 'y
n
`) ≥ 0
since ρ(x
n
, y
n
) ≥ 0 for each n. Also, ρ
∗
('x
n
`, 'x
n
`) = limρ(x
n
, x
n
) = 0. Furthermore, ρ
∗
('x
n
`, 'y
n
`) =
limρ(x
n
, y
n
) = limρ(y
n
, x
n
) = ρ
∗
('y
n
`, 'x
n
`). Finally, ρ
∗
('x
n
`, 'y
n
`) = limρ(x
n
, y
n
) ≤ limρ(x
n
, z
n
) +
limρ(z
n
, y
n
) = ρ
∗
('x
n
`, 'z
n
`) +ρ
∗
('z
n
`, 'y
n
`). Hence the set of all Cauchy sequences from a metric space
becomes a pseudometric space under ρ
∗
.
17c. Deﬁne 'x
n
` to be equivalent to 'y
n
` (written as 'x
n
` ∼ 'y
n
`) if ρ
∗
('x
n
`, 'y
n
`) = 0. If 'x
n
` ∼
'x
t
n
` and 'y
n
` ∼ 'y
t
n
`, then [ρ
∗
('x
n
`, 'y
n
`) − ρ
∗
('x
t
n
`, 'y
t
n
`)[ ≤ ρ
∗
('x
n
`, 'x
t
n
`) + ρ
∗
('y
n
`, 'y
t
n
`) = 0 so
ρ
∗
('x
n
`, 'y
n
`) = ρ
∗
('x
t
n
`, 'y
t
n
`). If ρ
∗
('x
n
`, 'y
n
`) = 0, then 'x
n
` ∼ 'y
n
` so they are equal in X
∗
. Thus
the pseudometric space becomes a metric space.
Associate each x ∈ X with the equivalence class in X
∗
containing the constant sequence 'x, x, . . .`. This
deﬁnes a mapping T from X onto T[X]. Since ρ
∗
(Tx, Ty) = limρ(x, y) = ρ(x, y), if Tx = Ty, then
ρ(x, y) = 0 so x = y. Thus T is onetoone. Also, T is continuous on X and its inverse is continuous on
T[X]. Hence T is an isometry between X and T[X] ⊂ X
∗
. Furthermore, T[X] is dense in X
∗
.
17d. If 'x
n
` is a Cauchy sequence from X, we may assume (by taking a subsequence) that ρ(x
n
, x
n+1
) <
2
−n
. Let ''x
n,m
`
∞
n=1
`
∞
m=1
be a sequence of such Cauchy sequences which represents a Cauchy se
quence in X
∗
. Given ε > 0, there exists N such that for m, m
t
≥ N, ρ
∗
('x
n,m
`, 'x
n,m
`) < ε/2.
i.e. lim
n
ρ(x
n,m
, x
n,m
) < ε/2. We may assume that for n ≥ N, ρ(x
n,m
, x
n,m
) < ε/2. In particular,
ρ(x
m,m
, x
m,m
) < ε/2. We may also assume that ρ(x
m,m
, x
m
,m
) < ε/2 since the sequence 'x
n,m
`
∞
n=1
is Cauchy in X. Thus ρ(x
m,m
, x
m
,m
) < ε for m, m
t
≥ N so the sequence 'x
n,n
` is Cauchy in X and
represents the limit of the Cauchy sequence in X
∗
.
*17e. T is an isometry from X onto T[X] and T
−1
is an isometry from T[X] onto X. Thus there is a
unique isometry T
t
from
¯
X ∩ Y onto X
∗
that extends T. Similarly, there is a unique isometry T
tt
from
X
∗
onto
¯
X ∩ Y that extends T
−1
. Then T
t
[
X
= T and T
tt
[
T[X]
= T
−1
so (T
t
◦ T
tt
)[
T[X]
= id
T[X]
and
(T
tt
◦ T
t
)[
X
= id
X
. Since T[X] is dense in X
∗
and X is dense in
¯
X ∩ Y , we have T
t
◦ T
tt
= id
X
∗ and
T
tt
◦ T
t
= id ¯
X∩Y
so (T
t
= T
tt
)
−1
. Hence X
∗
is isometric with the closure of X in Y .
18. Let (X, ρ) and (Y, σ) be two complete metric spaces. Let '(x
n
, y
n
)` be a Cauchy sequence in XY .
Since ρ(x
n
, x
m
) ≤ (ρ(x
n
, x
m
)
2
+ σ(y
n
, y
m
)
2
)
1/2
= τ((x
n
, y
n
), (x
m
, y
m
)), the sequence 'x
n
` is Cauchy in
X. Similarly, the sequence 'y
n
` is Cauchy in Y . Since X is complete, 'x
n
` converges to some x ∈ X.
Similarly, 'y
n
` converges to some y ∈ Y . Given ε > 0, there exists N such that ρ(x
n
, x) < ε/2 and
σ(y
n
, y) < ε/2 for n ≥ N. Then τ((x
n
, y
n
), (x, y)) = (ρ(x
n
, x)
2
+σ(y
n
, y)
2
)
1/2
< ε for n ≥ N. Hence the
sequence '(x
n
, y
n
)` converges to (x, y) ∈ X Y and X Y is complete.
7.5 Uniform continuity and uniformity
19. ρ
1
('x, y`, 'x
t
, y
t
`) = ρ(x, x
t
) + σ(y, y
t
) ≥ 0. ρ
1
('x, y`, 'x
t
, y
t
`) = 0 if and only if ρ(x, x
t
) = 0 and
σ(y, y
t
) = 0 if and only if x = x
t
and y = y
t
if and only if 'x, y` = 'x
t
, y
t
`. ρ
1
('x, y`, 'x
t
, y
t
`) =
ρ(x, x
t
) + σ(y, y
t
) = ρ(x
t
, x) + σ(y
t
, y) = ρ
1
('x
t
, y
t
`, 'x, y`). ρ
1
('x, y`, 'x
t
, y
t
`) = ρ(x, x
t
) + σ(y, y
t
) ≤
ρ(x, x
tt
) + ρ(x
tt
, x
t
) + σ(y, y
tt
) + σ(y
tt
, y
t
) = ρ
1
('x, y`, 'x
tt
, y
tt
`) + ρ
1
('x
tt
, y
tt
`, 'x
t
, y
t
`). Hence ρ
1
is a
metric. Similarly, ρ
∞
is a metric.
Given ε > 0, let δ = ε/2. When τ('x, y`, 'x
t
, y
t
`) < δ, ρ(x, x
t
)
2
< ε
2
/4 and σ(y, y
t
)
2
< ε
2
/4 so
ρ
1
('x, y`, 'x
t
, y
t
`) = ρ(x, x
t
) + σ(y, y
t
) < ε. Also, ρ
∞
('x, y`, 'x
t
, y
t
`) = max(ρ(x, x
t
), σ(y, y
t
)) < ε. When
ρ
1
('x, y`, 'x
t
, y
t
`) < δ, ρ(x, x
t
) < ε/2 and σ(y, y
t
) < ε/2 so τ('x, y`, 'x
t
, y
t
`) <
ε
2
/4 +ε
2
/4 < ε. When
ρ
∞
< δ, ρ(x, x
t
) < ε/2 and σ(y, y
t
) < ε/2 so τ('x, y`, 'x
t
, y
t
`) <
ε
2
/4 +ε
2
/4 < ε. Hence ρ
1
and ρ
∞
are uniformly equivalent to the usual product metric τ.
20. Let f be a uniformly continuous mapping of a metric space X into a metric space Y and let 'x
n
` be
a Cauchy sequence in X. Given ε > 0, there exists δ > 0 such that ρ(x, y) < δ implies σ(f(x), f(y)) < ε.
There exists N such that ρ(x
n
, x
m
) < δ for n, m ≥ N. Thus σ(f(x
n
), f(x
m
)) < ε for n, m ≥ N. Hence
'f(x
n
)` is a Cauchy sequence in Y .
21a. Let 'x
n
` be a sequence from E that converges to a point x ∈
¯
E. Then 'x
n
` is Cauchy in X so
'f(x
n
)` is Cauchy in Y . Since Y is complete, 'f(x
n
)` converges to some y ∈ Y .
33
21b. Suppose 'x
n
` and 'x
t
n
` both converge to x. Suppose 'f(x
n
)` converges to y and 'f(x
t
n
)` converges to
y
t
with y = y
t
. Let ε = σ(y, y
t
)/4 > 0. There exists δ > 0 such that ρ(x, x
t
) < δ implies σ(f(x), f(x
t
)) <
ε. There also exists N such that ρ(x
n
, x) < δ/2 and ρ(x
t
n
, x) < δ/2 for n ≥ N. Thus ρ(x
n
, x
t
m
) < δ
for n, m ≥ N so σ(f(x
n
), f(x
t
m
)) < ε for n, m ≥ N. We may assume that σ(f(x
n
), y) < ε and
σ(f(x
t
n
), y
t
) < ε for n ≥ N. Then σ(y, y
t
) ≤ σ(y, f(x
n
)) + σ(f(x
n
), f(x
t
n
)) + σ(f(x
t
n
), y
t
) < 3ε =
3σ(y, y
t
)/4. Contradiction. Hence 'f(x
n
)` and 'f(x
t
n
)` converge to the same point. By deﬁning y = g(x),
we get a function on
¯
E extending f.
21c. Given ε > 0, there exists δ > 0 such that ρ(x, x
t
) < δ implies σ(f(x), f(x
t
)) < ε/3 for x, x
t
∈ E.
Suppose ρ(¯ x, ¯ x
t
) < δ/3 with ¯ x, ¯ x
t
∈
¯
E. Let 'x
n
` be a sequence in E converging to ¯ x and let 'x
t
n
` be a
sequence in E converging to ¯ x
t
. There exists N such that ρ(x
n
, ¯ x) < δ/3 and ρ(x
t
n
, ¯ x
t
) < δ/3 for n ≥ N.
Then ρ(x
n
, x
t
n
) < δ for n ≥ N so σ(f(x
n
), f(x
t
n
)) < ε/3 for n ≥ N. Also, 'f(x
n
)` converges to some
y = g(¯ x) ∈ Y and 'f(x
t
n
)` converges to some y
t
= g(¯ x
t
) ∈ Y . We may assume σ(f(x
n
), g(¯ x)) < ε/3 and
σ(f(x
t
n
), g(¯ x
t
)) < ε/3 for n ≥ N. Thus σ(g(¯ x), g(¯ x
t
)) < ε. Hence g is uniformly continuous on
¯
E.
21d. Let h be a continuous function from
¯
E to Y that agrees with f on E. Let x ∈
¯
E and let 'x
n
` be
a sequence in E converging to x. Then 'h(x
n
)` converges to h(x) and 'g(x
n
)` converges to g(x). Since
h(x
n
) = g(x
n
) for all n, g(x) = h(x). Hence h ≡ g.
22a. Given ε > 0, let δ = ε/n. When ρ
+
(x, y) < δ, ρ
∗
(x, y) < nδ = ε. When ρ
∗
(x, y) < δ, ρ
+
(x, y) <
δ < ε. Thus ρ
+
and ρ
∗
are uniformly equivalent metrics. When ρ
+
(x, y) < δ, ρ(x, y) < (nδ
2
)
1/2
=
(ε
2
/n)
1/2
< ε. When ρ(x, y) < δ, (ρ
+
(x, y))
2
< δ
2
so ρ
+
(x, y) < δ < ε. Thus ρ
+
and ρ are uniformly
equivalent metrics. There exists δ
t
> 0 with δ
t
< ε/sqrtn such that ρ
∗
(x, y) < δ
t
implies ρ
+
(x, y) < δ.
Then when ρ
∗
(x, y) < δ
t
, ρ(x, y) < ε. When ρ(x, y) < δ
t
, ρ
∗
(x, y) ≤ ρ(x, y)
√
n < ε. Thus ρ and ρ
∗
are
uniformly equivalent metrics.
*22b. Deﬁne ρ
t
(x, y) = [x
3
1
− y
3
1
[ +
¸
n
i=2
[x
i
− y
i
[. Then ρ
t
is a metric on the set of ntuples of real
numbers. Given x ∈ R
n
and ε > 0, choose δ < min(1, [x
1
[, ε/3n[x
1
+ 1[
2
, ε/3n[x
1
− 1[
2
, ε/n, 3ε[x
1
−
1[
2
/n, 3ε[x
1
+ 1[
2
/n). When ρ
∗
(x, y) < δ, [x
i
− y
i
[ < δ for each i and [x
3
1
− y
3
1
[ = [3ξ
2
[[x
1
− y
1
[ for
some ξ ∈ (x
1
− δ, x
1
+ δ). Thus [x
3
1
− y
3
1
[ < 3 max([x
1
+ 1[, [x
1
− 1[)
2
δ < ε/n and ρ
t
(x, y) < ε. When
ρ
t
(x, y) < δ, [x
i
− y
i
[ < ε/n for i = 2, . . . , n and [x
3
1
− y
3
1
[ < δ. Then [x
1
− y
1
[ = [x
3
1
− y
3
1
[/[3ξ
2
[ < ε/n.
Thus ρ
∗
(x, y) < ε. Hence ρ
t
and ρ
∗
are equivalent metrics and since ρ
∗
and ρ are equivalent metrics, ρ
t
and ρ are equivalent metrics.
However ρ and ρ
t
are not uniformly equivalent metrics. Let ε = 1. Given δ > 0, choose n large
enough so that 3n
2
δ > 1. Let x = 'n, 0, . . . , 0` and let y = 'n + δ, 0, . . . , 0`. Then ρ(x, y) = δ and
ρ
t
(x, y) = (n +δ)
3
−n
3
> 3n
2
δ > 1.
22c. Note that ρ(x, y) =
σ(x,y)
1−σ(x,y)
= h(σ(x, y)) where h is the function in Q8. Since h is continuous at
0, given ε > 0, there exists δ
t
> 0 such that h(σ(x, y)) < ε when σ(x, y) < δ
t
. Now given ε > 0, let
δ < min(δ
t
, ε). When ρ(x, y) < δ, σ(x, y) < ρ(x, y) < δ < ε. When σ(x, y) < δ, ρ(x, y) = h(σ(x, y)) < ε.
Hence σ and ρ are uniformly equivalent metrics for X. (c.f. Q11a)
23a. [0, ∞) with the usual metric ρ is an unbounded metric space. By Q22c, the metric σ = ρ/(1 +ρ) is
uniformly equivalent to ρ so there is a uniform homeomorphism between ([0, ∞), ρ) and ([0, ∞), σ). By
Q11a, ([0, ∞), σ) is a bounded metric space. Hence boundedness is not a uniform property.
23b. Let (X, ρ) and (Y, σ) be metric spaces with X totally bounded. Let f : (X, ρ) → (Y, σ) be a uniform
homeomorphism. Given ε > 0, there exists δ > 0 such that ρ(x, x
t
) < δ implies σ(f(x), f(x
t
)) < ε. There
exist ﬁnitely many balls S
x
n
,δ
that cover X. i.e. X =
¸
k
n=1
S
x
n
,δ
. Take y ∈ Y . Then y = f(x) for some
x ∈ X. Now x ∈ S
x
n
,δ
for some n so ρ(x, x
n
) < δ and σ(f(x), f(x
n
)) < ε. Hence Y =
¸
k
n=1
S
f(x
n
),ε
so
Y is totally bounded. Thus total boundedness is a uniform property.
23c. By Q8, the function h(x) = x/(1 − x) is a homeomorphism between [0, 1) and [0, ∞). Let ε > 0
be given. Choose N such that N > 2/ε and let x
n
= (n − 1)/N for n = 1, . . . , N. The intervals
(x
n
− 1/N, x
n
+ 1/N) ∩ [0, 1) are balls of radius ε that cover [0, 1). Thus [0, 1) is totally bounded.
Suppose [0, ∞) is totally bounded. Then there are a ﬁnite number of balls of radius 1 that cover [0, ∞),
say [0, ∞) =
¸
k
n=1
S
x
n
,1
. We may assume that x
1
, . . . , x
k
are arranged in increasing order. But then
x
k
+ 2 is not in any of the balls S
x
n
,1
. Contradiction. Hence [0, ∞) is not totally bounded. Thus total
boundedness is not a topological property.
23d. Let (X, ρ) be a totally bounded metric space. For each n, there are a ﬁnite number of balls of
34
radius 1/n that cover X. Let S
n
be the set of the centres of these balls. Then each S
n
is a ﬁnite
set and S =
¸
S
n
is a countable set. Given ε > 0, choose N such that N > 1/ε. For any x ∈ X,
ρ(x, x
t
) < 1/N < ε for some x
t
∈ S
N
⊂ S. Thus S is a dense subset of X. Hence X is separable.
24a. Let 'X
k
, ρ
k
` be a sequence of metric spaces and deﬁne their direct product Z =
¸
∞
k=1
X
k
. Deﬁne
τ(x, y) =
¸
∞
k=1
2
−k
ρ
∗
k
(x
k
, y
k
) where ρ
∗
k
= ρ
k
/(1 + ρ
k
). Then τ(x, y) ≥ 0 since ρ
∗
k
(x
k
, y
k
) ≥ 0 for all k.
Also, τ(x, y) = 0 if and only if ρ
∗
k
(x
k
, y
k
) = 0 for all k if and only if x
k
= y
k
for all k if and only if
x = y. Furthermore, τ(x, y) = τ(y, x) since ρ
∗
k
(x
k
, y
k
) = ρ
∗
k
(y
k
, x
k
) for all k. Now for each n, τ(x, y) =
¸
n
k=1
2
−k
ρ
∗
k
(x, y) ≤
¸
n
k=1
2
−k
ρ
∗
k
(x, z) +
¸
n
k=1
2
−k
ρ
∗
k
(z, y) ≤
¸
∞
k=1
2
−k
ρ
∗
k
(x, z) +
¸
∞
k=1
2
−k
ρ
∗
k
(z, y) =
τ(x, z) +τ(z, y). Hence τ(x, y) ≤ τ(x, z) +τ(z, y). Thus τ is a metric on Z.
Suppose a sequence 'x
(n)
` in Z converges to x ∈ Z. Given ε > 0, let ε
t
= min(2
−k−1
ε, 2
−k−2
). There
exists N such that τ(x
(n)
, x) < ε
t
for n ≥ N. i.e.
¸
∞
k=1
2
−k
ρ
∗
k
(x
(n)
k
, x
k
) < ε
t
for n ≥ N. Then
ρ
k
0
(x
(n)
k
0
, x
k
0
) < 2
k
0
ε
t
(1 +ρ
k
0
(x
(n)
k
0
, x
k
0
)) for n ≥ N so (1 −2
k
0
ε
t
)ρ
k
0
(x
(n)
k
0
, x
k
0
) < 2
k
0
ε
t
for n ≥ N. Since
2
k
0
ε
t
= min(ε/2, 1/4), we have 3ρ
k
0
(x
(n)
k
0
, x
k
0
)/4 < 2
k
0
ε
t
< ε/2 for n ≥ N. Hence ρ
k
0
(x
(n)
k
0
, x
k
0
) < ε for
n ≥ N and 'x
(n)
k
` converges to x
k
for each k.
Conversely, suppose 'x
(n)
k
` converges to x
k
for each k. Then given ε > 0, there exists N such that
¸
∞
k=N+1
2
−k
ρ
∗
k
(x
k
, x
k
) < ε/2. For k = 1, . . . , N, there exists M such that ρ
k
(x
(n)
k
, x
k
) < ε/2 for n ≥ M.
Thus τ(x
(n)
, x) =
¸
N
k=1
2
−k
ρ
∗
k
(x
k
, x
k
) +
¸
∞
k=N+1
2
−k
ρ
∗
k
(x
k
, x
k
) <
¸
N
k=1
ε/2
k+1
+ ε/2 < ε for n ≥ M.
Hence 'x
(n)
` converges to x.
24b. Suppose each (X
k
, ρ
k
) is complete. Let 'x
(n)
` be a Cauchy sequence in (Z, τ). Then for each k,
'x
(n)
k
` is a Cauchy sequence in (X
k
, ρ
k
) so it converges to x
k
∈ X
k
. By part (a), 'x
(n)
` converges to
'x
k
` ∈ Z. Hence (Z, τ) is complete.
24c. Suppose that for each k, the spaces (X
k
, ρ
k
) and (Y
k
, σ
k
) are homeomorphic with f
k
: X
k
→ Y
k
a
homeomorphism. Deﬁne f :
¸
∞
k=1
X
k
→
¸
∞
k=1
Y
k
by f('x
k
`) = 'f(x
k
)`. Then f is bijective since each f
k
is. Also, note that f is continuous if and only if p
Y
k
◦ f is continuous for each k, where p
Y
k
:
¸
∞
k=1
→ Y
k
is the projection map. Now p
Y
k
◦f = f
k
◦p
X
k
so it is continuous. Thus f is continuous. Similarly, f
−1
is
continuous since p
X
k
◦ f
−1
= f
−1
k
◦ p
Y
k
is continuous for each k. Hence f is a homeomorphism between
¸
∞
k=1
X
k
and
¸
∞
k=1
Y
k
.
24d. If for each k, the spaces (X
k
, ρ
k
) and (Y
k
, σ
k
) are uniformly homeomorphic, then by a similar
argument as part (c), the spaces
¸
∞
k=1
X
k
and
¸
∞
k=1
Y
k
are uniformly homeomorphic.
7.6 Subspaces
25. Let A be a complete subset of a metric space X. Let x ∈
¯
A. Then by Q14, there is a sequence from
A that converges to x. Since A is complete, x ∈ A. Thus A is closed. Now suppose B is a closed subset
of a complete metric space Y . Let 'y
n
` be a Cauchy sequence in B. Then it is a Cauchy sequence in Y
so it converges to some y ∈ Y . Now y ∈
¯
B = B so B is complete.
*26. Let O be an open subset of a complete metric space (X, ρ). Let ϕ(x) = ρ(x, O
c
) for each x ∈ O.
Then S = ¦'x, y` : x ∈ O, y = ϕ(x)¦ is closed in X R. Since X R is complete, S is complete.
Consider f : (O, ρ) → (S, ρ
∞
) with f(x) = 'x, ϕ(x)`. Since ρ
∞
is uniformly equivalent to the usual
product metric by Q19, S is complete under the metric ρ
∞
. Now f is bijective. Given ε > 0, there
exists δ > 0 such that [ϕ(x) − ϕ(y)[ < ε/2 when ρ(x, y) < δ. Let δ
t
= min(δ, ε/2). When ρ(x, y) < δ
t
,
ρ
∞
(f(x), f(y)) = ρ(x, y) +[ϕ(x) −ϕ(y)[ < ε. When ρ
∞
('x, ϕ(x)`, 'y, ϕ(y)`) < δ
t
, ρ(x, y) < δ
t
< ε. Thus
f is a uniform homeomorphism and (O, ρ) is complete. Let σ = ρ/(1 + ρ). By Q22c, σ is uniformly
equivalent to ρ. Hence σ is a bounded metric for which (O, σ) is a complete metric space.
7.7 Compact metric spaces
27. Let X be a metric space, K a compact subset and F a closed subset. Consider the function
f(x) = ρ(x, F) = inf
y∈F
ρ(x, y). By Q9b, ¦x : ρ(x, F) = 0¦ =
¯
F = F. Thus if F ∩ K = ∅, then
ρ(x, F) > 0 for all x ∈ K. The function f[
K
is continuous on a compact set so it attains a minimum
δ > 0. Thus ρ(x, y) > δ for all x ∈ K and all y ∈ F. Conversely, if F ∩ K = ∅, then there exists x ∈ K
and y ∈ F such that ρ(x, y) = 0 so ρ(F, K) = 0.
35
28a. Let X be a totally bounded metric space and let f : X → Y be a uniformly continuous map onto
Y . Given ε > 0, there exists δ > 0 such that ρ(x, x
t
) < δ implies σ(f(x), f(x
t
)) < ε. There exist ﬁnitely
many balls ¦B
x
n
,δ
¦
k
n=2
that cover X. Take y ∈ Y . Then y = f(x) for some x ∈ X. Now x ∈ B
x
n
,δ
for
some n so ρ(x, x
n
) < δ and σ(f(x), f(x
n
)) < ε. Hence the balls ¦B
f(x
n
),ε
¦
k
n=1
cover Y and Y is totally
bounded.
28b. The function h(x) = x/(1−x) is a continuous map from [0, 1) onto [0, ∞). [0, 1) is totally bounded
but [0, ∞) is not.
29a. We may assume X / ∈ U. Set ϕ(x) = sup¦r : ∃O ∈ U with B
x,r
⊂ O¦. For each x ∈ X, there exists
O ∈ U such that x ∈ O. Since O is open, there exists r such that B
x,r
⊂ O. Thus ϕ(x) > 0. Since X is
compact, it is bounded so ϕ(x) < ∞.
29b. Suppose B
x,r
⊂ O for some O ∈ U. If 0 < r
t
< r − ρ(x, y), then B
y,r
⊂ B
x,r
⊂ O. If
ϕ(y) < ϕ(x) − ρ(x, y), then there exists r > ϕ(y) + ρ(x, y) such that B
x,r
⊂ O for some O ∈ U. Now
ϕ(y) < r − ρ(x, y) so taking r
t
= ϕ(y) + (r − ρ(x, y) − ϕ(y))/2, we have ϕ(y) < r
t
< r − ρ(x, y) and
B
y,r
⊂ O. Contradiction. Hence ϕ(y) ≥ ϕ(x) −ρ(x, y).
29c. Given ε > 0, let δ = ε. When ρ(x, y) < δ, [ϕ(x) − ϕ(y)[ ≤ ρ(x, y) < ε by part (b). Thus ϕ is
continuous on X.
29d. If X is sequentially compact, ϕ attains its minimum on X. Let ε = inf ϕ. Then ε > 0 since
ϕ(x) > 0 for all x.
29e. Let ε be as deﬁned in part (d). For any x ∈ X and δ < ε, δ < ϕ(x) so there exists r > δ such that
B
x,r
⊂ O for some O ∈ U. Then B
x,δ
⊂ B
x,r
⊂ O.
*30a. Let Z =
¸
∞
k=1
X
k
. Suppose each X
k
is totally bounded. Given ε > 0, choose N such that
2
N
> 2/ε. For k > N, pick p
k
∈ X
k
. For each k ≤ N, there exists A
k
= ¦x
(k)
1
, . . . , x
(k)
M
k
¦ such that for
any x ∈ X
k
, there exists x
(k)
j
∈ A
k
with ρ
k
(x, x
(k)
j
) < ε/2. Let A = ¦'x
n
` : x
k
∈ A
k
for k ≤ N, x
k
=
p
k
for k > N¦. Then [A[ = M
1
M
N
. If 'x
n
` ∈ Z, for each k ≤ N, there exists x
(k)
j
∈ A
k
such
that ρ
k
(x
k
, x
(k)
j
) < ε/2. Let y
k
= x
(k)
j
for k ≤ N and let y
k
= p
k
for k > N. Then 'y
n
` ∈ A and
τ(x, y) =
¸
n
k=1
2
−k
ρ
k
(x
k
,y
k
)
1+ρ
k
(x
k
,y
k
)
<
¸
N
k=1
2
−k ε
2
+
¸
∞
k=N+1
2
−k
<
ε
2
+
1
2
N
< ε.
30b. Suppose each X
k
is compact. Then each X
k
is complete and totally bounded. By part (a), Z is
totally bounded and by Q24b, Z is complete. Hence Z is compact.
7.8 Baire category
31a. Suppose a closed set F is nowhere dense. Then F
c
is dense. Thus any open set contains a point
in F
c
so any open set cannot be contained in F. Conversely, suppose a closed set contains no open set.
Then any open set contains a point in F
c
so F
c
is dense and F is nowhere dense.
31b. Suppose E is nowhere dense and let O be a nonempty open set. Then O contains a point x in
¯
E
c
.
Now O `
¯
E is open so there is a ball centred at x and contained in O `
¯
E ⊂ O ` E. Conversely, suppose
that for any nonempty open set O there is a ball contained in O ` E. Let the centre of the ball be x. If
x ∈
¯
E, then the ball contains a point y ∈ E. Contradiction. Thus x ∈
¯
E
c
. Hence any nonempty open
set contains a point in
¯
E
c
and E is nowhere dense.
32a. Suppose that E is of the ﬁrst category and A ⊂ E. Then E =
¸
E
n
where each E
n
is nowhere
dense and A =
¸
(A ∩ E
n
). Now (A∩ E
n
)
c
⊃ (
¯
A ∩
¯
E
n
)
c
=
¯
A
c
∪
¯
E
n
c
for each n. Since
¯
E
n
c
is dense for
each n, (A∩ E
n
)
c
is dense for each n. Hence A is of the ﬁrst category.
32b. Suppose that 'E
n
` is a sequence of sets of the ﬁrst category. Then E
n
=
¸
k
E
n,k
for each n,
where each E
n,k
is nowhere dense. Now
¸
E
n
=
¸
n
¸
k
E
n,k
=
¸
n,k
E
n,k
, which is a countable union of
nowhere dense sets. Hence
¸
E
n
is of the ﬁrst category.
33a. By Q3.14b, the generalised Cantor set constructed from [0, 1] by removing intervals of length
1/(n3
n
) at the nth step is closed, nowhere dense, and has Lebesgue measure 1 −1/n.
33b. By part (a), for each n, there is a nowhere dense closed set F
n
⊂ [0, 1] with mF
n
= 1 − 1/n. Let
E =
¸
F
n
⊂ [0, 1]. Then E is of the ﬁrst category and 1 −1/n ≤ mE ≤ 1 for each n so mE = 1.
34. Suppose O is open and F is closed. If
¯
O` O contains an open set O
t
, then O
t
contains a point in O.
Contradiction. Thus
¯
O ` O is a closed set containing no open sets. By Q31a,
¯
O ` O is nowhere dense. If
36
F ` F
◦
contains an open set O
tt
, then for any point x ∈ O
tt
, there is a ball centred at x and contained
in O
tt
so x ∈ F
◦
. Contradiction. Thus F ` F
◦
is a closed set containing no open sets. By Q31a, F ` F
◦
is nowhere dense.
If F is closed and of the ﬁrst category in a complete metric space, then F
◦
is of the ﬁrst category. By
the Baire category theorem, F
circ
is empty. Thus F = F ` F
◦
is nowhere dense.
35. Let E be a subset of a complete metric space. If E is residual, then E
c
is of the ﬁrst category so
E
c
=
¸
E
n
where each E
n
is nowhere dense. Thus E =
¸
E
c
n
⊃
¸
(
¯
E
n
)
c
. Since each (
¯
E
n
)
c
is dense and
the metric space is complete,
¸
(
¯
E
n
)
c
is dense so E contains a dense G
δ
. Conversely, suppose E contains
a dense G
δ
, say E ⊃
¸
O
n
. Then E
c
⊂
¸
O
c
n
. Since
¸
O
n
is dense, each O
n
is dense so O
c
n
is nowhere
dense and E
c
is of the ﬁrst category. Hence E is residual.
A subset E of a complete metric space is of the ﬁrst category if and only if E
c
is a residual subset of a
complete metric space if and only if E
c
contains a dense G
δ
if and only if E is contained in an F
σ
whose
complement is dense.
36a. Let X be a complete metric space without isolated points. Suppose X has a countable number of
points x
1
, x
2
, . . .. Then X =
¸
¦x
n
¦ where each ¦x
n
¦ is nowhere dense and X is of the ﬁrst category,
contradicting the Baire category theorem. Hence X has an uncountable number of points.
36b. [0, 1] is complete since it is a closed subspace of the complete metric space R. Also, [0, 1] has no
isolated points. By part (a), [0, 1] is uncountable.
37a. Let E be a subset of a complete metric space. Suppose E
c
is dense and F is a closed set contained
in E. Then F
c
⊃ E
c
so F
c
is dense and F is nowhere dense.
37b. Suppose E and E
c
are both dense in a complete metric space X. Also suppose that E and E
c
are
both F
σ
’s, say E =
¸
F
n
and E
c
=
¸
F
t
n
where each F
n
and F
t
n
is closed. By part (a), each F
n
and F
t
n
is nowhere dense. Then X =
¸
F
n
∪
¸
F
t
n
so X is of the ﬁrst category, contradicting the Baire category
theorem. Hence at most one of the sets E and E
c
is an F
σ
.
37c. The set of rational numbers in [0, 1] is an F
σ
. Since the set of rational numbers in [0, 1] and the
set of irrational numbers in [0, 1] are both dense in the complete metric space [0, 1], the set of irrational
numbers in [0, 1] is not an F
σ
by part (b). Hence its complement, the set of rational numbers in [0, 1],
is not a G
δ
.
37d. If f is a realvalued function on [0, 1] which is continuous on the rationals and discontinuous on
the irrationals, then the set of points of continuity of f is a G
δ
. i.e. the set of rationals in [0, 1] is a G
δ
,
contradicting part (c). Hence there is no such function.
*38a. Let C = C[0, 1] and set F
n
= ¦f : ∃x
0
with 0 ≤ x
0
≤ 1 − 1/n and [f(x) − f(x
0
)[ ≤ n(x −
x
0
) for all x, x
0
≤ x < 1¦. Suppose [[f
k
− f[[ → 0 where f
k
∈ F
n
. For each k, there exists x
k
with
0 ≤ x
k
≤ 1 − 1/n and [f
k
(x) − f
k
(x
k
)[ ≤ n(x − x
k
) for x
k
≤ x < 1. Since x
k
∈ [0, 1 − 1/n] for all
k and [0, 1 − 1/n] is compact, we may assume that 'x
k
` converges to some x
0
∈ [0, 1 − 1/n]. Then
[f(x) −f(x
0
)[ ≤ n(x −x
0
) so f ∈ F
n
.
*38b. By Q2.47, for any g ∈ C and any ε > 0, there exists a polygonal function ϕ such that [g(x) −
ϕ(x)[ < ε/2 for all x ∈ [0, 1]. There also exists a polygonal function ψ whose righthand derivative is
everywhere greater than n in absolute value and [ϕ(x) −ψ(x)[ < ε/2 for all x ∈ [0, 1]. Then ψ ∈ F
c
n
and
[g(x) −ψ(x)[ < ε for all x ∈ [0, 1]. Hence F
n
is nowhere dense.
38c. The set D of continuous functions which have a ﬁnite derivative on the right for at least one point
of [0, 1] is the union of the F
n
’s so D is of the ﬁrst category in C.
38d. Since D is of the ﬁrst category in the complete metric space C, D = C so there is a function in
C ` D, that is, a nowhere diﬀerentiable continuous function on [0, 1].
39. Let F be a family of realvalued continuous functions on a complete metric space X, and suppose
that for each x ∈ X there is a number M
x
such that [f(x)[ ≤ M
x
for all f ∈ F. For each m, let
E
m,f
= ¦x : [f(x)[ ≤ m¦, and let E
m
=
¸
F
E
m,f
. Since each f is continuous, E
m,f
is closed and so E
m
is closed. For each x ∈ X, there exists m such that [f(x)[ ≤ m for all f ∈ F. Hence X =
¸
E
m
. Then
O =
¸
E
◦
m
⊂ X is a dense open set and for each x ∈ O, x ∈ E
◦
m
for some m so there is a neighbourhood
U of x such that U ⊂ E
◦
m
. In particular, F is uniformly bounded on U.
40a. Suppose that given ε > 0, there exist N and a neighbourhood U of x such that σ(f
n
(x
t
), f(x)) < ε
for n ≥ N and all x
t
∈ U. Let 'x
n
` be a sequence with x = limx
n
. We may assume that x
n
∈ U for
37
all n ≥ N so σ(f
n
(x
n
), f(x)) < ε for n ≥ N and 'f
n
` converges continuously to f at x. Conversely,
suppose there exists ε > 0 such that for any N and any neighbourhood U of x, there exists n ≥ N and
x
t
∈ U with σ(f
n
(x
t
), f(x)) ≥ ε. For each n, let U
n
= B
x,1/n
. There exists n
1
≥ 1 and x
n
1
∈ U
1
with
σ(f
n
1
(x
n
1
), f(x)) ≥ ε. Suppose x
n
1
, . . . , x
n
k
have been chosen. There exists n
k+1
≥ n
k
and x
n
k
∈ U
k
with σ(f
n
k+1
(x
n
k+1
), f(x)) ≥ ε. Then x = limx
n
k
by construction but f(x) = limf
n
k
(x
n
k
) so 'f
n
` does
not converge continuously to f.
40b. Let Z = ¦1/n¦ ∪ ¦0¦. Deﬁne g : X Z → Y by g(x, 1/n) = f
n
(x) and g(x, 0) = f(x). Suppose
g is continuous at 'x
0
, 0` in the product metric. Let 'x
n
` be a sequence with x
0
= limx
n
. Then
'x
0
, 0` = lim'x
n
, 1/n` and f(x) = g(x
0
, 0) = limg(x
n
, 1/n) = limf
n
(x
n
) so 'f
n
` converges continuously
to f at x
0
. Conversely, suppose 'f
n
` converges continuously to f at x
0
. Let '(x
n
, z
n
)` be a sequence
in X Z converging to 'x
0
, 0`. Then x
0
= limx
n
, 0 = limz
n
and f(x
0
) = limf
n
(x
n
). i.e. g(x
0
, 0) =
limg(x
n
, 1/n). Since 0 = limz
n
, it follows that g(x
0
, 0) = limg(x
n
, z
n
) so g is continuous at 'x
0
, 0`.
40c. Let 'f
n
` converge continuously to f at x. By part (b), the function g is continuous at 'x, 0`. If 'x
n
`
is a sequence converging to x, then f(x) = g(x, 0) = limg(x
n
, 0) = limf(x
n
). Hence f is continuous at
x.
*40d. Let 'f
n
` be a sequence of continuous maps. Suppose 'f
n
` converges continuously to f at x. By part
(a), given ε > 0, there exists N and a neighbourhood U of x such that σ(f
n
(x
t
), f(x)) < ε/2 for n ≥ N
and x
t
∈ U. By part (c), f is continuous at x so we may assume that σ(f(x
t
), f(x)) < ε/2 for x
t
∈ U.
Thus σ(f
n
(x
t
), f(x
t
)) < ε for n ≥ N and x
t
∈ U. Conversely, suppose that given ε > 0, there exist N
and a neighbourhood U of x such that σ(f
n
(x
t
), f(x
t
)) < ε/4 for all n ≥ N and all x
t
∈ U. In particular,
σ(f
N
(x), f(x)) < ε/4. Let 'x
k
` be a sequence with x = limx
k
. We may assume that x
k
∈ U for all k
and σ(f
N
(x
k
), f
N
(x)) < ε/4 for k ≥ N. Then σ(f(x
k
), f(x)) ≤ σ(f(x
k
), f
N
(x
k
)) + σ(f
N
(x
k
), f
N
(x)) +
σ(f
N
(x), f(x)) < 3ε/4 for k ≥ N. Thus σ(f
k
(x
k
), f(x)) ≤ σ(f
k
(x
k
), f(x
k
)) + σ(f(x
k
), f(x)) < ε for
k ≥ N so f(x) = limf
k
(x
k
) and 'f
n
` converges continuously to f at x.
40e. Let 'f
n
` be a sequence of continous maps. Suppose 'f
n
` converges continuously to f on X. For
each x ∈ X and each ε > 0, there exists N
x
and a neighbourhood U
x
of x such that σ(f
n
(x
t
), f(x
t
)) < ε
for each n ≥ N
x
and x
t
∈ U
x
. Then X =
¸
U
x
so for any compact subset K ⊂ X, K ⊂
¸
m
i=1
U
x
i
for
some x
1
, . . . , x
m
. Thus for n ≥ max N
x
i
and x
t
∈ K, we have σ(f
n
(x
t
), f(x
t
)) < ε so 'f
n
` converges
uniformly to f on K. Conversely, suppose 'f
n
` converges uniformly to f on each compact subset of
X. Then 'f
n
` converges uniformly to f on ¦x¦ for each x ∈ X. Given ε > 0, there exists N such that
σ(f
n
(x), f(x)) < ε/3 for n ≥ N. Since each f
n
is continuous, f is also continuous. Then there exists
a neighbourhood U of x such that σ(f
n
(x
t
), f
n
(x)) < ε/3 and σ(f(x
t
), f(x)) < ε/3 for x
t
∈ U. Thus
σ(f
n
(x
t
), f(x
t
)) < ε for n ≥ N and x
t
∈ U. Hence 'f
n
` converges continuously to f on X.
40f. Let X be a complete metric space and 'f
n
` a sequence of continuous maps of X into a met
ric space Y such that f(x) = limf
n
(x) for each x ∈ X. For m, n ∈ N, deﬁne F
m,n
= ¦x ∈ X :
σ(f
k
(x), f
l
(x)) ≤ 1/m for all k, l ≥ n¦. Let x
t
be a point of closure of F
m,n
. There is a sequence 'x
i
`
in F
m,n
converging to x
t
. For any ε > 0 and k, l ≥ n, there exists δ > 0 such that ρ(x, x
t
) < δ implies
σ(f
k
(x), f
k
(x
t
)) < ε and σ(f
l
(x), f
l
(x
t
)) < ε. Then there exists N such that ρ(x
i
, x
t
) < δ for i ≥ N.
Thus σ(f
k
(x
t
), f
l
(x
t
)) ≤ σ(f
k
(x
t
), f
k
(x
N
)) + σ(f
k
(x
N
), f
l
(x
N
)) + σ(f
l
(x
N
), f
l
(x
t
)) < 2ε + 1/m. Since ε
is arbitrary, σ(f
k
(x
t
), f
l
(x
t
)) ≤ 1/m and x
t
∈ F
m,n
so F
m,n
is closed.
For any x ∈ X, there exists n such that σ(f
k
(x), f(x)) < 1/2m for k ≥ n. Then for k, l ≥ n,
σ(f
k
(x), f
l
(x)) ≤ σ(f
k
(x), f(x)) +σ(f(x), f
l
(x)) < 1/m so x ∈ F
m,n
. Hence X =
¸
n
F
m,n
. By Proposi
tion 31, O
m
=
¸
n
F
◦
m,n
is open and since X is complete, O
m
is also dense.
40g. Let x ∈ O
m
. Then x ∈ F
◦
m,n
for some n. Thus there exists a neighbourhood U of x such
that U ⊂ F
m,n
. It follows that for any k, l ≥ n and any x
t
∈ U, σ(f
k
(x
t
), f
l
(x
t
)) ≤ 1/m. Since
f(x
t
) = limf
l
(x
t
), we have σ(f
k
(x
t
), f(x
t
)) ≤ 1/m for any k ≥ n and x
t
∈ U.
40h. Let E =
¸
O
m
. Since each O
m
is open and dense, E is a dense G
δ
by Baire’s theorem. If x ∈ E,
then x ∈ O
m
for all m. Given ε > 0, choose m > 1/ε. There exists a neighbourhood U of x and an n
such that σ(f
k
(x
t
), f(x
t
)) ≤ 1/m < ε for any k ≥ n and x
t
∈ U.
40i. Let X be a complete metric space and 'f
n
` a sequence of continuous functions of X into a metric
space Y . Suppose 'f
n
` converges pointwise to f. By parts (f), (g), (h) and (a), there exists a dense G
δ
in X on which 'f
n
` converges continuously to f.
41a. Let (X, ρ) and (Y, σ) be complete metric spaces and f : X Y → Z be a mapping into a metric
38
space (Z, τ) that is continuous in each variable. Fix y
0
∈ Y . Set F
m,n
= ¦x ∈ X : τ[f(x, y), f(x, y
0
)] ≤
1/m for all y with σ(y, y
0
) ≤ 1/n¦. Let x
t
be a point of closure of F
m,n
. There is a sequence 'x
k
` in F
m,n
converging to x
t
. For each k and any y with σ(y, y
0
) ≤ 1/n, τ[f(x
k
, y), f(x
k
, y
0
)] ≤ 1/m. Given ε > 0,
there exists δ > 0 such that ρ(x, x
t
) < δ implies τ[f(x, y), f(x
t
, y)] < ε and τ[f(x, y
0
), f(x
t
, y
0
)] < ε. Then
there exists N such that ρ(x
k
, x
t
) < δ for k ≥ N. Thus τ[f(x
t
, y), f(x
t
, y
0
)] ≤ τ[f(x
t
, y), f(x
N
, y)] +
τ[f(x
N
, y), f(x
N
, y
0
)]+τ[f(x
N
, y
0
), f(x
t
, y
0
)] < 2ε+1/m. Since ε is arbitrary, τ[f(x
t
, y), f(x
t
, y
0
)] ≤ 1/m
and x
t
∈ F
m,n
so F
m,n
is closed.
For any x ∈ X and m, there exists δ > 0 such that σ(y, y
0
) < δ implies τ[f(x, y), f(x, y
0
)] < 1/m.
Choose n > 1/δ. Then for any y with σ(y, y
0
) < 1/n, τ[f(x, y), f(x, y
0
)] < 1/m so x ∈ F
m,n
. Hence
X =
¸
n
F
m,n
.
41b. Let O
m
=
¸
n
F
◦
m,n
. By Proposition 31, O
m
is open and since X is complete, O
m
is also dense. Let
x ∈ O
m
. Then x ∈ F
◦
m,n
for some n. Thus there exists a neighbourhood U
t
of x such that U
t
⊂ F
m,n
.
For any y with σ(y, y
0
) ≤ 1/n and any x
t
∈ U
t
, τ[f(x
t
, y), f(x
t
, y
0
)] ≤ 1/m. We may assume that
τ[f(x
t
, y
0
), f(x, y
0
)] ≤ 1/m for x
t
∈ U
t
by continuity. Then τ[f(x
t
, y), f(x, y
0
)] ≤ 2/m for x
t
∈ U
t
.
Hence there is a neighbourhood U of 'x, y
0
` in X Y such that τ[f(p), f(x, y
0
)] ≤ 2/m for all p ∈ U.
41c. Let G =
¸
O
m
. Since each O
m
is open and dense, G is a dense G
δ
by Baire’s theorem. If x ∈ G,
then x ∈ O
m
for each m. Given ε > 0, choose m > 2/ε. There exists a neighbourhood U of 'x, y
0
` such
that τ[f(p), f(x, y
0
)] ≤ 2/m < ε for p ∈ U. Hence f is continuous at 'x, y
0
` for each x ∈ G.
41d. Let E ⊂ X Y be the set of points at which f is continuous. If f is continuous at z ∈ X Y ,
then for each n, there exists δ
n,z
> 0 such that f[B
z,δ
n,z
] ⊂ B
f(z),1/n
. Let E
n
=
¸
z∈S
B
z,δ
z,1/n
/2
. Then
E =
¸
n
E
n
(c.f. Q2.7.53) and E is a G
δ
set. Take 'x
0
, y
0
` ∈ X Y and let ε > 0 be given. By part (c),
there exists a dense G
δ
set G ⊂ X such that f is continuous at 'x, y
0
` for each x ∈ G. Since G is dense,
there exists x
1
∈ G such that ρ(x
0
, x
1
) < ε. Then 'x
1
, y
0
` ∈ E and τ
t
('x
1
, y
0
`, 'x
0
, y
0
`) = ρ(x
1
, x
0
) < ε.
Thus E is dense in X Y .
41e. Given a ﬁnite product X
1
X
p
of complete metric spaces, we may regard the product as
(X
1
X
p−1
) X
n
where X
1
X
p−1
is a complete metric space. Let f : X
1
X
p
→ Z be
a mapping into a metric space Z that is continuous in each variable. By a similar argument as in part
(d), there exists a dense G
δ
⊂ X
1
X
p
on which f is continuous.
42a. Let X and Y be complete metric spaces. Also let G ⊂ X and H ⊂ Y be dense G
δ
’s. Then
G =
¸
G
n
and H =
¸
H
n
where each G
n
and H
n
is open. Thus GH =
¸
G
n
¸
H
n
=
¸
(G
n
H
n
)
where each G
n
H
n
is open in XY . Hence GH is a G
δ
in XY . Given ε > 0 and 'x, y` ∈ XY ,
there exist x
0
∈ G and y
0
∈ H such that ρ(x, x
0
) < ε/2 and σ(y, y
0
) < ε/2. Then τ('x, y`, 'x
0
, y
0
`) < ε.
Hence GH is dense in X Y .
*42b.
42c. Let E be a dense G
δ
in X Y . Then E =
¸
O
n
where each O
n
is a dense open set in X Y .
By part (b), for each n, there exists a dense G
δ
set G
n
⊂ X such that ¦y : 'x, y` ∈ O
n
¦ is a dense open
subset of Y for each x ∈ G
n
. Let G =
¸
G
n
. Then G is still a countable intersection of dense open sets
in X. Thus G is a dense G
δ
set in X such that ¦y : 'x, y` ∈ E¦ =
¸
¦y : 'x, y` ∈ O
n
¦ is a dense G
δ
for
each x ∈ G.
*42d.
*43. Let (X, ρ) be a complete metric space and f : X →R be an upper semicontinuous function. Let E
be the set of points at which f is continuous. Then E is a G
δ
. Suppose f is discontinuous at x. Then f
is not lower semicontinuous at x so f(x) > lim
y→x
f(y). There exists q ∈ Q such that f(x) ≥ q > lim
y→x
f(y).
Let F
q
= ¦x : f(x) ≥ q¦. Then x ∈ F
q
`F
◦
q
. Conversely, if x ∈ F
q
`F
◦
q
for some q ∈ Q, then f(x) ≥ q and
for any δ > 0, there exists y with ρ(x, y) < δ and f(y) < q so f(x) > lim
y→x
f(y). Thus E
c
=
¸
q∈Q
F
q
` F
◦
q
.
Since each F
q
` F
◦
q
is nowhere dense, E
c
is of the ﬁrst category and E is residual. Hence E is a dense
G
δ
in X
7.9 Absolute G
δ
’s
44. Let σ =
¸
2
−n
σ
n
. Then σ(x, y) =
¸
2
−n
σ
n
(x, y) ≤
¸
2
−n
= 1 for all x, y ∈ E. Since σ
n
≥ 0
for all n, σ ≥ 0. Also, σ(x, y) = 0 if and only if σ
n
(x, y) = 0 for all n if and only if x = y. Since
39
σ
n
(x, y) = σ
n
(y, x) for all n, σ(x, y) = σ(y, x). Also, σ(x, y) =
¸
2
−n
σ
n
(x, y) ≤
¸
2
−n
σ
n
(x, z) +
¸
2
−n
σ
n
(z, y) = σ(x, z) +σ(z, y). Thus σ is a metric on E.
Let ε > 0 and ﬁx x ∈ E. For each n, there exists δ
n
> 0 such that σ
n
(x, y) < δ
n
implies ρ(x, y) < ε/2
and ρ(x, y) < δ
n
implies σ
n
(x, y) < ε/2. There exists N such that
¸
∞
n=N+1
2
−n
< ε/2. Let δ =
min(δ
1
/2, . . . , δ
N
/2
N
). If ρ(x, y) < δ, then σ
n
(x, y) < ε/2 for n = 1, . . . , N so
¸
N
n=1
2
−n
σ
n
(x, y) < ε/2
and σ(x, y) < ε. If σ(x, y) < δ, then σ
1
(x, y) < 2δ ≤ δ
1
so ρ(x, y) < ε. Hence σ is equivalent to ρ on E.
If each σ
n
is uniformly equivalent to ρ, then σ is uniformly equivalent to ρ.
45. Let A ⊂ B ⊂
¯
A be subsets of a metric space and let g and h be continuous maps of B into a metric
space X. Suppose g(u) = h(u) for all u ∈ A. Let u ∈ B. Then u ∈
¯
A so there is a sequence 'u
n
` in
A converging to u. Since g and h are continuous on B, 'g(u
n
)` and 'h(u
n
)` converge to f(u) and g(u)
respectively. But g(u
n
) = h(u
n
) for all n so g(u) = h(u). Hence g ≡ h.
*46a.
46b. Starting from E, let I
1
, I
2
, . . . be disjoint open intervals satisfying the conditions of part (a) with
ε = 1/2. For E
I
j
= E ∩ I
j
, repeat the process with ε = 1/4, getting intervals I
j,1
, I
j,2
, . . .. Continuing,
at the nth stage getting intervals I
j,k,...,l
with ε = 2
−n
. Given x ∈ E, there is a unique integer k
1
such
that x ∈ I
k
1
. Then x ∈ E
I
k
1
and there is a unique integer k
2
such that x ∈ I
k
1
,k
2
and so on. Thus there
is a unique sequence of integers k
1
, k
2
, . . . such that for each n we have x ∈ I
k
1
,...,k
n
.
46c. Given a sequence of integers k
1
, k
2
, . . ., suppose there are x, y ∈ E such that x, y ∈ I
k
1
,...,k
n
for all
n. By construction, the diameter of I
k
1
,...,k
n
is less than 2
−n
for each n. Thus σ(x, y) < 2
−n
for all n so
σ(x, y) = 0 and x = y.
46d. Let N
ω
be the space of inﬁnite sequences of integers and make N into a metric space by set
ting ρ(i, j) = δ
ij
. Let τ be the product metric on N
ω
. Given ε > 0, choose N such that 2
−N
<
ε. Let δ < 2
−N
. When σ(x, y) < δ, k
n
(x) = k
n
(y) for n = 1, . . . , N so τ('k
n
(x)`, 'k
n
(y)`) ≤
¸
∞
i=N+1
2
−i
ρ
∗
(k
n
(x), k
n
(y)) ≤ 2
−N
< ε. When τ('k
n
(x)`, 'k
n
(y)`) < δ, k
n
(x) = k
n
(y) for n = 1, . . . , N.
Then x, y ∈ I
k
1
(x),...,k
N
(x)
so σ(x, y) < 2
−N
< ε. Hence the correspondence between N
ω
and E given by
parts (b) and (c) is a uniform homeomorphism between (N
ω
, τ) and (E, σ).
*46e. By parts (a)(d), any dense G
δ
E in (0, 1) whose complement is dense is uniformly homeomorphic
to N
ω
, which is in turn homeomorphic to the set of irrationals in (0, 1) by the continued fraction expansion.
Hence E is homeomorphic to the set of irrationals in (0, 1).
7.10 The AscoliArzel´a Theorem
47. Let X be a metric space. Let 'f
n
` be a sequence of continuous functions from X to a metric
space Y which converge to a function f uniformly on each compact subset K of X. Let x ∈ X and let
'x
k
` be a sequence in X converging to x. Then K = ¦x
k
¦
∞
k=1
∪ ¦x¦ is a compact subset of X. Given
ε > 0, there exists N such that σ(f
N
(x
t
), f(x
t
)) < ε/3 for any x
t
∈ K. Also, there exists N
t
such
that σ(f
N
(x
k
), f
N
(x)) < ε/3 for k ≥ N
t
. Thus σ(f(x
k
), f(x)) ≤ σ(f(x
k
), f
N
(x
k
)) +σ(f
N
(x
k
), f
N
(x)) +
σ(f
N
(x), f(x)) < ε for k ≥ N
t
. Hence f is continuous on X.
48a. Let X be a separable, locally compact metric space, and (Y, σ) any metric space. Let ¦x
n
¦ be a
countable dense subset of X. For each n, there is an open set U
n
such that x
n
∈ U
n
and U
n
is compact.
Then X = ¦x
n
¦ ⊂
¸
U
n
so X =
¸
U
n
. For each n and any x ∈ U
n
, there is an open set V
x
containing
x with V
x
compact. Then U
n
⊂
¸
x∈U
n
V
x
and since U
n
is compact, there is a ﬁnite number of V
x
’s
covering U
n
. Let O
n
be the union of the ﬁnite number of V
x
’s. Then X =
¸
O
n
.
48b. Let σ
∗
(f, g) =
¸
2
−n
σ
∗
n
(f, g) where σ
∗
n
(f, g) = sup
O
n
σ(f(x),g(x))
1+σ(f(x),g(x))
. Since σ
∗
n
(f, g) ≤ 1 for all
n, σ
∗
(f, g) ≤ 1 < ∞ and since σ
∗
n
(f, g) ≥ 0 for all n, σ
∗
(f, g) ≥ 0. Since σ
∗
n
(f, g) = σ
∗
n
(g, f) for
all n, σ
∗
(f, g) = σ
∗
(g, f). For each n, σ
∗
n
(f, g) = sup
O
n
σ(f(x),g(x))
1+σ(f(x),g(x))
= sup
O
n
[1 −
1
1+σ(f(x),g(x))
] ≤
sup
O
n
[1 −
1
1+σ(f(x),h(x))+σ(h(x),g(x))
] = sup
O
n
σ(f(x),h(x))+σ(h(x),g(x))
1+σ(f(x),h(x))+σ(h(x),g(x))
≤ σ
∗
n
(f, h) + σ
∗
n
(h, g). Thus
σ
∗
(f, g) ≤ σ
∗
(f, h) + σ
∗
(h, g). Hence the set of functions from X into Y becomes a metric space
under σ
∗
.
49. Let F be an equicontinuous family of functions from X to Y , and let F
+
be the family of all
pointwise limits of functions in F, that is of f for which there is a sequence 'f
n
` from F such that
40
f(x) = limf
n
(x) for each x ∈ X. Given x ∈ X and ε > 0, there is an open set O containing x such that
σ(f(x), f(y)) < ε/3 for all y ∈ O and f ∈ F. Now if f ∈ F
+
and y ∈ O, then there is a sequence 'f
n
` from
F such that f(x) = limf
n
(x) so there is an N such that σ(f
N
(x), f(x)) < ε/3 and σ(f
N
(y), f(y)) < ε/3.
Then σ(f(x), f(y)) ≤ σ(f(x), f
N
(x)) + σ(f
N
(x), f
N
(y)) + σ(f
N
(y), f(y)) < ε. Hence F
+
is also an
equicontinuous family of functions.
50. Let 0 < α ≤ 1 and let F = ¦f : [[f[[
α
≤ 1¦. If f ∈ F, then max [f(x)[ +sup [f(x)−f(y)[/[x−y[
α
≤ 1.
In particular, [f(x) −f(y)[ ≤ [x −y[
α
for all x, y. Thus f ∈ C[0, 1]. Also, F is an equicontinuous family
of functions on the separable space [0, 1] and each f ∈ F is bounded. By the AscoliArzel´a Theorem,
each sequence 'f
n
` in F has a subsequence that converges pointwise to a continuous function. Hence F
is a sequentially compact, and thus compact, subset of C[0, 1].
*51a. Let F be the family of functions that are holomorphic on the unit disk ∆ = ¦z : [z[ < 1¦ with
[f(z)[ ≤ 1. Let f ∈ F and ﬁx z
t
∈ ∆. Let U be an open ball of radius min([z
t
[, 1 − [z
t
[) centred
at z
t
. Then U ⊂ ∆. Let C be the circle of radius r/2 centred at z
t
. For every z within r/4 of z
t
,
f(z) − f(z
t
) =
1
2πi
C
[
f(w) dw
w−z
−
f(w) dw
w−z
] =
z−z
2πi
C
f(w) dw
(w−z)(w−z
)
by Cauchy’s integral formula. Since
[f(z)[ ≤ 1 for all f ∈ F and all z ∈ ∆, [f(z) − f(z
t
)[ ≤ 4r
−1
[z − z
t
[ for all f ∈ F. It follows that F is
equicontinuous.
*51b. By the AscoliArzel´a Theorem, any sequence 'f
n
` in F has a subsequence that converges uniformly
to a function f on each compact subset of ∆. Furthermore f is holomorphic on ∆.
*51c. Let 'f
n
` be a sequence of holomorphic functions on ∆ such that f
n
(z) → f(z) for all z ∈ ∆. For
each z ∈ ∆, there is an M
z
such that [f
n
(z)[ ≤ M
z
for all n. Let E
m
=
¸
n
E
m,n
where E
m,n
= ¦z :
[f
n
(z)[ ≤ m¦. Then E
m
is closed and ∆ =
¸
m
E
m
. There is an E
m
that is not nowhere dense so it has
nonempty interior. Then O =
¸
m
E
◦
m
is a dense open subset of ∆ and 'f
n
` is locally bounded on O.
By an argument similar to that in part (a), 'f
n
` is equicontinuous on O so there is a subsequence that
converges uniformly to a function f on each compact subset of O. Furthermore, f is holomorphic on O.
8 Topological Spaces
8.1 Fundamental notions
1a. Given a set X, deﬁne ρ on X X by ρ(x, y) = 0 if x = y and ρ(x, y) = 1 otherwise. Then for any
set A ⊂ X, A =
¸
x∈A
B
x,1/2
so A is open. Thus the associated topological space is discrete. If X has
more than one point, then there is no metric on X such that the associated topological space is trivial
because in a metric space, any two distinct points can be enclosed in disjoint open sets.
1b. Let X be a space with a trivial topology. If f is a continuous mapping of X into R, then f
−1
[I] is
either ∅ or X for any open interval I. Take c ∈ R. Then f
−1
[c] = f
−1
[(c − ε, c + ε)] is either ∅ or X.
Thus f must be a constant function. Conversely, if f(x) = c for all x ∈ X, then for any open interval
I, f
−1
[I] = ∅ if c / ∈ I and f
−1
[I] = X if c ∈ I. Since any open set of real numbers is a countable union
of disjoint open intervals, it follows that f is continuous. Hence the only continuous mappings from X
into R are the constant functions.
1c. Let X be a space with a discrete topology. Let f be a mapping of X into R. Since f
−1
[O] ⊂ X for
any open set O ⊂ R, f is continuous. Hence all mappings of X into R are continuous.
2.
¯
E is the intersection of all closed sets containing E so E ⊂
¯
E. Also,
¯
E ⊂
¯
¯
E. On the other hand,
¯
¯
E is
the intersection of all closed sets containing
¯
E, one of which is
¯
E itself, so
¯
¯
E ⊂
¯
E. Thus
¯
¯
E =
¯
E.
¯
A∪
¯
B is a closed set containing A∪B so A∪ B ⊂
¯
A∪
¯
B. On the other hand, A ⊂ A∪B and B ⊂ A∪B
so
¯
A ⊂ A∪ B and
¯
B ⊂ A∪ B. Thus
¯
A∪
¯
B ⊂ A∪ B. Hence A∪ B ⊂
¯
A∪
¯
B.
If F is closed, then F is a closed set containing F so
¯
F ⊂ F. Since we always have F ⊂
¯
F, F =
¯
F.
Conversely, if F =
¯
F, then since
¯
F is closed, F is closed.
If x ∈
¯
E, then x is in every closed set containing E. Let O be an open set containing x. If O ∩ E = ∅,
then E ⊂ O
c
. Since O
c
is closed, x ∈ O
c
. Contradiction. Hence O ∩ E = ∅ and x is a point of closure
of E. Conversely, suppose x is a point of closure of E. If x / ∈
¯
E, then x is not in some closed set F
containing E so x ∈ F
c
. Since F
c
is an open set containing x, F
c
∩E = ∅. Contradiction. Hence x ∈
¯
E.
E
◦
is the union of all open sets contained in E so E
◦
⊂ E. Also, E
◦◦
⊂ E
◦
. On the other hand, E
◦◦
is
41
the union of all open sets contained in E
◦
, one of which is E
◦
itself, so E
◦
⊂ E
◦◦
. Thus E
◦◦
= E
◦
.
A
◦
∩ B
◦
is an open set contained in A∩ B so A
◦
∩ B
◦
⊂ (A∩ B)
◦
. On the other hand, A∩ B ⊂ A and
A∩B ⊂ B so (A∩B)
◦
⊂ A
◦
and (A∩B)
◦
⊂ B
◦
. Thus (A∩B)
◦
⊂ A
◦
∩B
◦
. Hence (A∩B)
◦
= A
◦
∩B
◦
.
If x ∈ E
◦
, then x is in some open set O contained in E so x is an interior point of E. Conversely, if x is
an interior point of E, then x ∈ O ⊂ E for some open set O so x ∈ E
◦
.
If x ∈ (E
c
)
◦
, then x is in some open set O ⊂ E
c
so x / ∈ O
c
⊃ E and thus x ∈
¯
E
c
. Conversely, if x ∈
¯
E
c
,
then x is not in some closed set F ⊃ E so x ∈ F
c
⊂ E
c
and thus x ∈ (E
c
)
◦
. Hence (E
c
)
◦
=
¯
E
c
.
3. Suppose A ⊂ X is open. Then given x ∈ A, x ∈ O ⊂ A with O = A. Conversely, suppose that given
x ∈ A, there is an open set O
x
such that x ∈ O
x
⊂ A. Then A =
¸
x∈A
O
x
, which is open.
4. Let f be a mapping of X into Y . Suppose f is continuous. For any closed set F, f
−1
[F
c
] is open.
But f
−1
[F
c
] = (f
−1
[F])
c
so f
−1
[F] is closed. Conversely, suppose the inverse image of every closed set
is closed. For any open set O, f
−1
[O
c
] is closed. But f
−1
[O
c
] = (f
−1
[O])
c
so f
−1
[O] is open. Thus f is
continuous.
5. Suppose f is a continuous mapping of X into Y and g is a continuous mapping of Y into Z. For
any open set O ⊂ Z, g
−1
[O] is open in Y so f
−1
[g
−1
[O]] is open in X. But f
−1
[g
−1
[O]] = (g ◦ f)
−1
[O].
Hence g ◦ f is a continuous mapping of X into Z.
6. Let f and g be two realvalued continuous functions on X and let x ∈ X. Given ε > 0, there is an
open set O containing x such that [f(x) − f(y)[ < ε/2 and [g(x) − g(y)[ < ε/2 whenever y ∈ O. Then
[(f + g)(x) − (f + g)(y)[ = [f(x) − f(y) + g(x) − g(y)[ ≤ [f(x) − f(y)[ + [g(x) − g(y)[ < ε whenever
y ∈ O. Thus f + g is continuous at x. There is an open set O
t
such that [g(x) − g(y)[ < ε/2[f(x)[ and
[f(x)−f(y)[ < ε/2 max([g(x)−ε/2[f(x)[[, [g(x)+ε/2[f(x)[[) whenever y ∈ O
t
. Then [(fg)(x)−(fg)(y)[ =
[f(x)g(x)−f(x)g(y)+f(x)g(y)−f(y)g(y)[ ≤ [f(x)[[g(x)−g(y)[+[f(x)−f(y)[[g(y)[ < ε whenever y ∈ O
t
.
Thus fg is continuous at x.
7a. Let F be a closed subset of a topological space and 'x
n
` a sequence of points from F. If x is a
cluster point of 'x
n
`, then for any open set O containing x and any N, there exists n ≥ N such that
x
n
∈ O. Suppose x / ∈ F. Then x is in the open set F
c
so there exists x
n
∈ F
c
. Contradiction. Hence
x ∈ F.
7b. Suppose f is continuous and x = limx
n
. For any open set U containing f(x), there is an open set
O containing x such that f[O] ⊂ U. There also exists N such that x
n
∈ O for n ≥ N so f(x
n
) ∈ U for
n ≥ N. Hence f(x) = limf(x
n
).
7c. Suppose f is continuous and x is a cluster point of 'x
n
`. For any open set U containing f(x), there
is an open set O containing x such that f[O] ⊂ U. For any N, there exists n ≥ N such that x
n
∈ O so
f(x
n
) ∈ U. Hence f(x) is a cluster point of 'f(x
n
)`.
*8a. Let E be an arbitrary set in a topological space X. (E
◦
)
◦
is the intersection of all closed sets
containing (E
◦
)
◦
, one of which is E
◦
, so (E
◦
)
◦
⊂ E
◦
. On the other hand, E
◦
is the intersection of
all closed sets containing E
◦
, one of which is (E
◦
)
◦
since E
◦
= E
◦◦
⊂ (E
◦
)
◦
, so E
◦
⊂ (E
◦
)
◦
. Hence
(E
◦
)
◦
= E
◦
.
Now since
¯
¯
E =
¯
E and (E
c
)
c
= E, new sets can only be obtained if the operations are performed
alternately. Also, (
¯
E)
c
= (E
c
)
◦
so
(E)
c
c
c
=
(E)
c
◦
=
(E
c
)
◦
◦
= (E
c
)
◦
= (E)
c
. Hence the
distinct sets that can be obtained are at most E,
¯
E, (E)
c
, (E)
c
,
(E)
c
c
,
(E)
c
c
,
(E)
c
c
c
, E
c
,
E
c
, (E
c
)
c
, (E
c
)
c
,
(E
c
)
c
c
,
(E
c
)
c
c
,
(E
c
)
c
c
c
.
*8b. Let E = ¦(0, 0)¦ ∪¦z : 1 < [z[ < 2¦ ∪¦z : 2 < [z[ < 3¦ ∪¦z : 4 < [z[ < 5 and z = (p, q) where p, q ∈
Q¦ ⊂ R
2
. Then E gives 14 diﬀerent sets by repeated use of complementation and closure.
9. Let f be a function from a topological space X to a topological space Y . Suppose f is continuous
and let x ∈ X. For any open set U containing f(x), f
−1
[U] is open and x ∈ f
−1
[U]. If y ∈ f
−1
[U], then
f(y) ∈ U. Thus f is continuous at x. Conversely, suppose f is not continuous. There exists an open set
O such that f
−1
[O] is not open. By Q3, there exists x ∈ f
−1
[O] such that U is not a subset of f
−1
[O]
for any open set U containing x. Then O is an open set containing f(x) and f[U] is not a subset of O
42
for any open set U containing x. Hence f is not continuous at x.
10a. Let the subset A of a topological space X be the union of two sets A
1
and A
2
both of which
are closed (resp. both of which are open). Let f be a map of A into a topological space Y such
that the restrictions f[
A
1
and f[
A
2
are each continuous. For any closed (resp. open) set E ⊂ Y ,
f
−1
[E] = (f
−1
[E] ∩ A
1
) ∪ (f
−1
[E] ∩ A
2
) = f[
−1
A
1
[E] ∪ f[
−1
A
2
[E] = (A
1
∩ F) ∪ (A
2
∩ F
t
) for some closed
(resp. open) sets F, F
t
⊂ A. Hence f
−1
[E] is closed (resp. open) and f is continuous.
10b. Consider the subset (0, 2] of R. Then (0, 2] is the union of (0, 1), which is open, and [1, 2], which
is closed. Let f be the map deﬁned by f(x) = x on (0, 1) and f(x) = x + 1 on [1, 2]. Then f[
(0,1)
and
f[
[1,2]
are each continuous but f is not continuous (at 1).
*10c. Suppose A = A
1
∪ A
2
, A
1
` A
2
∩ (A
2
` A
1
) = ∅ and A
2
` A
1
∩ (A
1
` A
2
) = ∅. Let f be a map
of A into a topological space Y such that the restrictions f[
A
1
and f[
A
2
are each continuous. Let F
be a closed subset of Y and let F
t
= f
−1
[F] ⊂ A. Then F
t
= F
t
∩ (A
1
` A
2
∪ (A
1
∩ A
2
) ∪ A
2
` A
1
) =
F
t
∩ (A
1
` A
2
)∪F
t
∩ A
1
∩ A
2
∪F
t
∩ (A
2
` A
1
). Note that F
t
∩ (A
1
` A
2
) ⊂ A
1
and F
t
∩ (A
2
` A
1
) ⊂ A
2
.
Thus f[F
t
] = f[F
t
∩ (A
1
` A
2
)] ∪f[F
t
∩ A
1
∩ A
2
] ∪f[F
t
∩ (A
2
` A
1
)] = f[F
t
] ⊂ F and F
t
⊂ f
−1
[F] = F
t
.
Hence F
t
= f
−1
[F] is closed and f is continuous.
8.2 Bases and countability
11a. Let B be a base for the topological space (X, T ). If x ∈
¯
E, then for every B ∈ B with x ∈ B,
since B is open, B ∩ E = ∅. i.e. there is a y ∈ B ∩ E. Conversely, suppose that for every B ∈ B with
x ∈ B there is a y ∈ B ∩ E. For each open set O in X with x ∈ O, there is a B ∈ B with x ∈ B ⊂ O.
Then there is a y ∈ B ∩ E ⊂ O ∩ E. Thus x ∈
¯
E.
11b. Let X satisfy the ﬁrst axiom of countability. If there is a sequence 'x
n
` in E converging to x, then
for any open set O containing x, there exists N such that x
n
∈ O for n ≥ N. Thus O∩E = ∅ so x ∈
¯
E.
Conversely, suppose x ∈
¯
E. There is a countable base 'B
n
` at x. By considering B
1
, B
1
∩B
2
, B
1
∩B
2
∩
B
3
, . . ., we may assume B
1
⊃ B
2
⊃ B
3
. Now each B
n
contains an x
n
∈ E. For any open set O
containing x, O ⊃ B
N
⊃ B
N+1
⊃ for some N so x
n
∈ O for n ≥ N and 'x
n
` is a sequence in E
converging to x.
11c. Let X satisfy the ﬁrst axiom of countability and let 'x
n
` be a sequence from X. If 'x
n
` has a
subsequence converging to x, it follows from the deﬁnition that x is a cluster point of 'x
n
`. Conversely,
suppose x is a cluster point of 'x
n
`. There is a countable base 'B
n
` at x. We may assume B
1
⊃ B
2
⊃ .
Since B
1
is open and contains x, x
n
1
∈ B
1
for some n
1
≥ 1. Suppose x
n
1
, . . . , x
n
k
have been chosen. Then
there exists n
k+1
≥ n
k
such that x
n
k+1
∈ B
k+1
. For any open set O containing x, O ⊃ B
N
⊃ B
N+1
for some N so x
n
k
∈ O for k ≥ N and the subsequence 'x
n
k
` converges to x.
12a. See Q9.
12b. Let B
x
be a base at x and C
y
be a base at y = f(x). Suppose f is continuous at x. For each
C ∈ C
y
, since C is open and contains f(x), there is an open set U containing x such that U ⊂ f
−1
[C].
Then there exists B ∈ B
x
such that B ⊂ U ⊂ f
−1
[C]. Conversely, suppose that for each C ∈ C
y
there
exists B ∈ B
x
such that B ⊂ f
−1
[C]. For any open set O containing f(x), there exists C ∈ C
y
such that
C ⊂ O. Then there exists B ∈ B
x
such that B ⊂ f
−1
[C] ⊂ f
−1
[O]. Hence f is continuous at x.
13. Let C be any collection of subsets of X. Let B consist of X and all ﬁnite intersections of sets in C.
By Proposition 5, B is a base for some topology o on X. For any topology T on X containing C, we
have B ⊂ T . Since any set in o is a union of sets in B, o ⊂ T . Hence o is the weakest topology on X
containing C.
14. Let X be an uncountable set of points and let T consist of the empty set and all subsets of X
whose complement is ﬁnite. Then X ∈ T since X
c
= ∅, which is ﬁnite. If O
1
, O
2
∈ T , then we may
assume O
1
, O
2
= ∅ so (O
1
∩ O
2
)
c
= O
c
1
∪ O
c
2
, which is ﬁnite. Thus O
1
∩ O
2
∈ T . If O
α
∈ T , then
(
¸
O
α
)
c
=
¸
O
c
α
, which is ﬁnite. Thus
¸
O
α
∈ T . Hence T is a topology for X.
Suppose (X, T ) satisﬁes the ﬁrst axiom of countability. Take x ∈ X. There is a countable base 'B
n
` at
x. Note that B
c
n
is ﬁnite for all n so
¸
B
c
n
is countable. Thus there exists y ∈ (
¸
B
c
n
)
c
=
¸
B
n
with
y = x since (
¸
B
c
n
)
c
is uncountable. Now X ` ¦y¦ is open and contains x so there exists B
N
⊂ X ` ¦y¦.
But y ∈ B
N
⊂ X ` ¦y¦. Contradiction. Hence (X, T ) does not satisfy the ﬁrst axiom of countability.
43
15. Let X be the set of real numbers and let B be the set of all intervals of the form [a, b). For any
x ∈ X, we have x ∈ [x, x + 1). If x ∈ [a, b) ∩ [c, d), then x ∈ [c, b) (or [a, d)). Hence B is a base of a
topology T (the halfopen interval topology) for X.
For any x ∈ X, let B
x,q
= [x, x + q) where q is a positive rational number. Then ¦B
x,q
¦ is a countable
base at x. Thus (X, T) satisﬁes the ﬁrst axiom of countability. Let B be a base for (X, T). For any
x ∈ X, there exist B
x
∈ B such that x ∈ B
x
⊂ [x, x + 1). If x = y, then B
x
= B
y
. Thus ¦B
x
¦
is uncountable and (X, T) does not satisfy the second axiom of countability. For any x ∈ X and any
open set O containing x, there is an interval [a, b) ⊂ O containing x. Furthermore, there is a rational
q ∈ [a, b). Thus the rationals are dense in (X, T ). Now (X, T ) is not metrizable because it is separable
and a separable metric space satisﬁes the second axiom of countability.
16. Suppose X is second countable. Let B be a countable base for the topology on X and let  be an
open cover of X. For each x ∈ X, there exists U
x
∈ , such that x ∈ U
x
. Then there exists B
x
∈ B such
that x ∈ B
x
⊂ U
x
. Thus X =
¸
x∈X
B
x
. Since B is countable, we may choose countably many U
x
’s to
form a countable subcover of . Hence X is Lindel¨of.
17a. Let X
1
= N N and take X = X
1
∪ ¦ω¦. For each sequence s = 'm
k
` of natural numbers deﬁne
B
s,n
= ¦ω¦∪¦'j, k` : j ≥ m
k
if k ≥ n¦. Clearly, any x ∈ X is in some B
s,n
or ¦'j, k`¦. If x ∈ B
s,n
∩B
s
,n
,
then either x = ω or x = 'j, k`. If x = 'j, k`, then ¦'j, k`¦ ⊂ B
s,n
∩B
s
n
. If x = ω, let s
tt
= 'm
tt
k
` be the
sequence where m
tt
k
= max(m
k
, m
t
k
) for all k and let n
tt
= min(n, n
t
). Then ω ∈ B
s
,n
⊂ B
s,n
∩ B
s
,n
.
Other possible intersections are similarly dealt with. Hence the sets B
s,n
together with the sets ¦'j, k`¦
form a base for a topology on X.
*17b. For any open set O containing ω, there is some B
s,n
such that ω ∈ B
s,n
⊂ O. Since B
s,n
contains
some 'j, k` ∈ X
1
, so does O. Thus ω is a point of closure of X
1
.
17c. X is countable so it is separable. If it satisﬁes the ﬁrst axiom of countability, then since ω is a
point of closure of X
1
, by Q11b, there is a sequence from X
1
converging to ω and by Q11c, ω is a cluster
point of that sequence, contradicting part (b). Hence X does not satisfy the ﬁrst axiom of countability
and thus it also does not satisfy the second axiom of countability.
17d. X is countable so it is Lindel¨of.
8.3 The separation axioms and continuous realvalued functions
18a. Given two distinct points x, y in a metric space (X, ρ), let r = ρ(x, y)/2. Then the open balls B
x,r
and B
y,r
are disjoint open sets with x ∈ B
x,r
and y ∈ B
y,r
. Hence every metric space is Hausdorﬀ.
18b. Given two disjoint closed sets F
1
, F
2
in a metric space (X, ρ), let O
1
= ¦x : ρ(x, F
1
) < ρ(x, F
2
)¦
and let O
2
= ¦x : ρ(x, F
2
) < ρ(x, F
1
)¦. The sets O
1
and O
2
are open since the functions f(x) = ρ(x, F
1
)
and g(x) = ρ(x, F
2
) are continuous. Thus O
1
and O
2
are disjoint open sets with F
1
⊂ O
1
and F
2
⊂ O
2
.
Hence every metric space is normal.
19. Let X consist of [0, 1] and an element 0
t
. Consider the sets (α, β), [0, β), (α, 1] and ¦0
t
¦∪(0, β) where
α, β ∈ [0, 1]. Clearly, any x ∈ [0, 1] belongs to one of the sets. Note that (α, β) ⊂ [0, β) ∩ (α, 1]. If x ∈
(α, β)∩[0, β
t
), then x ∈ (α, min(β, β
t
)) ⊂ (α, β)∩[0, β
t
). If x ∈ (α, β)∩(α
t
, 1], then x ∈ (max(α, α
t
), β) ⊂
(α, β) ∩ (α
t
, 1]. If x ∈ (α, β) ∩ (¦0
t
¦ ∪ (0, β
t
)), then x ∈ (α, min(β, β
t
)) ⊂ (α, β) ∩ (¦0
t
¦ ∪ (0, β
t
)). If
x ∈ [0, β) ∩(¦0
t
¦∪(0, β
t
)), then x ∈ (0, min(β, β
t
)) ⊂ [0, β) ∩(¦0
t
¦∪(0, β
t
)). If x ∈ (α, 1] ∩(¦0
t
¦∪(0, β)),
then x ∈ (α, β) ⊂ (α, 1] ∩ (¦0
t
¦ ∪ (0, β)). Hence the sets form a base for a topology on X.
Given two distinct points x, y ∈ X, if neither of them is 0’, then [0, (x +y)/2) is an open set containing
one but not the other. If y = 0
t
, then ¦0
t
¦ ∪ (0, x/2) is an open set containing y but not x. Hence X
is T
1
. Consider the distinct points 0 and 0’. If O is an open set containing 0 and O
t
is an open set
containing 0’, then 0 ∈ [0, β) ⊂ O and 0
t
∈ ¦0
t
¦ ∪ (0, β
t
) ⊂ O
t
for some β, β
t
. Since [0, β) ∩ (0, β
t
) = ∅,
O ∩ O
t
= ∅. Hence X is not Hausdorﬀ.
20. Let f be a realvalued function on a topological space X. Suppose f is continuous. Then for any
real number a, ¦x : f(x) < a¦ = f
−1
[(−∞, a)] and ¦x : f(x) > a¦ = f
−1
[(a, ∞)]. Since (−∞, a) and
(a, ∞) are open, the sets ¦x : f(x) < a¦ and ¦x : f(x) > a¦ are open. Conversely, suppose the sets
¦x : f(x) < a¦ and ¦x : f(x) > a¦ are open for any real number a. Since any open set O ⊂ R is a union
of open intervals and any open interval is an intersection of at most two of the sets, f
−1
[O] is open so f
is continuous.
44
Since ¦x : f(x) ≥ a¦
c
= ¦x : f(x) < a¦, we also have f is continuous if and only if for any real number
a, the set ¦x : f(x) > a¦ is open and the set ¦x : f(x) ≥ a¦ is closed.
21. Suppose f and g are continuous realvalued functions on a topological space X. For any real
number a, ¦x : f(x) + g(x) < a¦ =
¸
q∈Q
(¦x : g(x) < q¦ ∩ ¦x : f(x) < a − q¦), which is open, and
¦x : f(x) +g(x) > a¦ =
¸
q∈Q
(¦x : g(x) > q¦ ∩¦x : f(x) > a −q¦), which is open, so f +g is continuous.
¦x : f(x)g(x) < a¦ =
¸
q∈Q
(¦x : g(x) < q¦) ∩ ¦x : f(x) < a/q¦, which is open, and ¦x : f(x)g(x) > a¦ =
¸
q∈Q
(¦x : g(x) > q¦) ∩ ¦x : f(x) > a/q¦, which is open, so fg is continuous. ¦x : (f ∨ g)(x) < a¦ = ¦x :
f(x) < a¦ ∩ ¦x : g(x) < a¦, which is open, and ¦x : (f ∨ g)(x) > a¦ = ¦x : f(x) > a¦ ∪ ¦x : g(x) > a¦,
which is open, so f ∨ g is continuous. ¦x : (f ∧ g)(x) < a¦ = ¦x : f(x) < a¦ ∪ ¦x : g(x) < a¦, which is
open, and ¦x : (f ∧ g)(x) > a¦ = ¦x : f(x) > a¦ ∩ ¦x : g(x) > a¦, which is open, so f ∧ g is continuous.
22. Let 'f
n
` be a sequence of continuous functions from a topological space X to a metric space Y .
Suppose 'f
n
` converges uniformly to a function f. Let ε > 0 and let x ∈ X. There exists N such
that ρ(f
n
(x), f(x)) < ε/3 for all n ≥ N and all x ∈ X. Since f
N
is continuous, there is an open
set O containing x such that f
N
[O] ⊂ B
f
N
(x),ε/3
. If y ∈ O, then ρ(f(y), f(x)) ≤ ρ(f(y), f
N
(y)) +
ρ(f
N
(y), f
N
(x)) + ρ(f
N
(x), f(x)) < ε. Thus f[O] ⊂ B
f(x),ε
. Since any open set in Y is a union of open
balls, f is continuous.
23a. Let X be a Hausdorﬀ space. Suppose X is normal. Given a closed set F and an open set O
containing F, there exist disjoint open sets U and V such that F ⊂ U and O
c
⊂ V . Now
¯
U is disjoint
from O
c
since if y ∈ O
c
, then V is an open set containing y that is disjoint from U. Thus
¯
U ⊂ O.
Conversely, suppose that given a closed set F and an open set O containing F, there is an open set U
such that F ⊂ U and
¯
U ⊂ O. Let F and G be disjoint closed sets. Then F ⊂ G
c
and there is an open
set U such that F ⊂ U and
¯
U ⊂ G
c
. Equivalently, F ⊂ U and G ⊂
¯
U
c
. Since U and
¯
U
c
are disjoint
open sets, X is normal.
*23b. Let F be a closed subset of a normal space contained in an open set O. Arrange the rationals
in (0, 1) of the form r = p2
−n
in a sequence 'r
n
`. Let U
1
= O. By part (a), there exists an open set
U
0
such that F ⊂ U
0
and U
0
⊂ O = U
1
. Let P
n
be the set containing the ﬁrst n terms of the sequence.
Since U
0
⊂ U
1
, there exists an open set U
r
1
such that U
0
⊂ U
r
1
and U
r
1
⊂ U
1
. Suppose open sets U
r
have been deﬁned for rationals r in P
n
such that U
p
⊂ U
q
whenever p < q. Now r
n+1
has an immediate
predecessor r
i
and an immediate successor r
j
(under the usual order relation) in P
n+1
∪ ¦0, 1¦. Note
that U
r
i
⊂ U
r
j
. Thus there is an open set U
r
n+1
such that U
r
i
⊂ U
r
n+1
and U
r
n+1
⊂ U
r
j
. Now if r ∈ P
n
,
then either r ≤ r
i
, in which case U
r
⊂ U
r
i
⊂ U
r
n+1
, or r ≥ r
j
, in which case U
r
n+1
⊂ U
r
j
⊂ U
r
. By
induction, we have a sequence ¦U
r
¦ of open sets, one corresponding to each rational in (0, 1) of the form
p = r2
−n
, such that F ⊂ U
r
⊂ O and U
r
⊂ U
s
for r < s.
23c. Let ¦U
r
¦ be the family constructed in part (b) with U
1
= X. Let f be the realvalued function on
X deﬁned by f(x) = inf¦r : x ∈ U
r
¦. Clearly, 0 ≤ f ≤ 1. If x ∈ F, then x ∈ U
r
for all r so f(x) = 0.
If x ∈ O
c
, then x / ∈ U
r
for any r < 1 so f(x) = 1. Given x ∈ X and an open interval (c, d) containing
f(x), choose rationals r
1
and r
2
of the form p2
−n
such that c < r
1
< f(x) < r
2
< d. Consider the open
set U = U
r
2
` U
r
1
. Since f(x) < r
2
, x ∈ U
r
⊂ U
r
2
for some r < r
2
. If x ∈ U
r
1
, then x ∈ U
r
for all r > r
1
so f(x) ≤ r
1
. Since f(x) > r
1
, x / ∈ U
r
1
. Thus x ∈ U. If y ∈ U, then y ∈ U
r
2
so f(y) ≤ r
2
< d. Also,
y / ∈ U
r
1
so f(y) ≥ r
1
> c. Thus f[U] ⊂ (c, d). Hence f is continuous.
23d. Let X be a Hausdorﬀ space. Suppose X is normal. For any pair of disjoint closed sets A and B
on X, B
c
is an open set containing A. By the constructions in parts (b) and (c), there is a continuous
realvalued function f on X such that 0 ≤ f ≤ 1, f ≡ 0 on A and f ≡ 1 on (B
c
)
c
= B.
(*) Proof of Urysohn’s Lemma.
24a. Let A be a closed subset of a normal topological space X and let f be a continuous realvalued
function on A. Let h = f/(1 +[f[). Then [h[ = [f[/(1 +[f[) < 1.
24b. Let B = ¦x : h(x) ≤ −1/3¦ and let C = ¦x : h(x) ≥ 1/3¦. By Urysohn’s Lemma, there is a
continuous function h
1
which is −1/3 on B and 1/3 on C while [h
1
(x)[ ≤ 1/3 for all x ∈ X. Then
[h(x) −h
1
(x)[ ≤ 2/3 for all x ∈ A.
24c. Suppose we have continuous functions h
n
on X such that [h
n
(x)[ < 2
n−1
/3
n
for all x ∈ X and
[h(x) −
¸
n
i=1
h
i
(x)[ < 2
n
/3
n
for all x ∈ A. Let B
t
= ¦x : h(x) −
¸
n
i=1
h
i
(x) ≤ −2
n
/3
n+1
¦ and
let C
t
= ¦x : h(x) −
¸
n
i=1
h
i
(x) ≥ 2
n
/3
n+1
¦. By Urysohn’s Lemma, there is a continuous function
h
n+1
which is −2
n
/3
n+1
on B
t
and 2
n
/3
n+1
on C
t
while [h
n+1
(x)[ ≤ 2
n
/3
n+1
for all x ∈ X. Then
45
[h(x) −
¸
n+1
i=1
h
i
(x)[ < 2
n+1
/3
n+1
for all x ∈ A.
24d. Let k
n
=
¸
n
i=1
h
i
. Then each k
n
is continuous and [h(x) − k
n
(x)[ < 2
n
/3
n
for all x ∈ A. Also,
[k
n
(x) −k
n−1
(x)[ < 2
n−1
/3
n
for all n and all x ∈ X. If m > n, then [k
m
(x) −k
n
(x)[ = [
¸
m
i=n+1
h
i
(x)[ ≤
¸
m
i=n+1
2
i−1
/3
i
< (2/3)
n
. Let k(x) =
¸
∞
i=1
h
i
(x). Then [k(x) − k
n
(x)[ ≤ (2/3)
n
for all x ∈ X. Thus
'k
n
` converges uniformly to k. i.e. 'h
n
` is uniformly summable to k and since each k
n
is continuous, k is
continuous. Also, since [k
n
[ ≤ 1 for each n, [k[ ≤ 1. Now [h(x) −k
n
(x)[ < 2
n
/3
n
for all n and all x ∈ A
so letting n → ∞, [h(x) −k(x)[ = 0 for all x ∈ A. i.e. k = h on A.
24e. Since [k[ = [h[ < 1 on A, A and ¦x : k(x) = 1¦ are disjoint closed sets. By Urysohn’s Lemma,
there is a continuous function ϕ on X which is 1 on A and 0 on ¦x : k(x) = 1¦.
24f. Set g = ϕk/(1 −[ϕk[). Then g is continuous and g = k/(1 −[k[) = h/(1 −[h[) = f on A.
(*) Proof of Tietze’s Extension Theorem.
25. Let F be a family of realvalued functions on a set X. Consider the sets of the form ¦x : [f
i
(x) −
f
i
(y)[ < ε for some ε > 0, some y ∈ X, and some ﬁnite set f
1
, . . . , f
n
of functions in F¦. The weak
topology on X generated by F has ¦f
−1
[O] : f ∈ F, O open in R¦ as a base. Now if x
t
∈ f
−1
[O] for
some f ∈ F and O open in R, we may assume O is an open interval (c, d). If x ∈ ¦x : [f(x) − f(x
t
)[ <
min(f(x
t
) − c, d − f(x
t
))¦, then f(x) ∈ (c, d). i.e. x ∈ f
−1
[O]. Thus x
t
∈ ¦x : [f(x) − f(x
t
)[ <
min(f(x
t
) − c, d − f(x
t
))¦ ⊂ f
−1
[O]. Hence the sets of the form ¦x : [f
i
(x) − f
i
(y)[ < ε for some ε >
0, some y ∈ X, and some ﬁnite set f
1
, . . . , f
n
of functions in F¦ is a base for the weak topology on X
generated by F.
Suppose this topology is Hausdorﬀ. For any pair ¦x, y¦ of distinct points in X, there are disjoint open sets
O
x
and O
y
such that x ∈ O
x
and y ∈ O
y
. Then there are sets B
x
and B
y
of the form ¦x : [f
i
(x)−f
i
(z)[ <
ε for some ε > 0, some z ∈ X, and some ﬁnite set f
1
, . . . , f
n
of functions in F¦ such that x ∈ B
x
⊂ O
x
and y ∈ B
y
⊂ O
y
. Suppose f(x) = f(y) for all f ∈ F. Then x, y ∈ B
x
∩ B
y
. Contradiction. Hence
there is a function f ∈ F such that f(x) = f(y). Conversely, suppose that for each pair ¦x, y¦ of distinct
points in X there exists f ∈ F such that f(x) = f(y). Then x ∈ ¦x
t
: [f(x
t
) −f(x)[ < [f(y) − f(x)[/2¦
and y ∈ ¦x
t
: [f(x
t
) − f(y)[ < [f(y) − f(x)[/2¦ with the two sets being disjoint open sets. Hence the
topology is Hausdorﬀ.
26. Let F be a family of realvalued continuous functions on a topological space (X, T ). Since f is
continuous for each f ∈ F, the weak topology generated by F is contained in T . Suppose that for each
closed set F and each x / ∈ F there is an f ∈ F such that f(x) = 1 and f ≡ 0 on F. Then for each O ∈ T
and each x ∈ O, there is an f ∈ F such that f(x) = 1 and f ≡ 0 on O
c
. i.e. x ∈ f
−1
[(1/2, 3/2)] ⊂ O.
Thus O is in the weak topology generated by F.
27. Let X be a completely regular space. Given a closed set F and x / ∈ F, there is a function f ∈ C(X)
such that f(x) = 1 and f ≡ 0 on F. Then F ⊂ f
−1
[(−∞, 1/2)], x ∈ f
−1
[(1/2, ∞)], and the sets
f
−1
[(−∞, 1/2)] and f
−1
[(1/2, ∞)] are disjoint open sets. Hence X is regular.
28. Let X be a Hausdorﬀ space and let Y be a subset of X. Given two distinct points x and y in Y ,
there are disjoint open sets O
1
and O
2
in X such that x ∈ O
1
and y ∈ O
2
. Then x ∈ O
1
∩ Y and
y ∈ O
2
∩ Y , with O
1
∩ Y and O
2
∩ Y being disjoint open sets in Y . Hence Y is Hausdorﬀ.
*29. In R
n
let B be the family of sets ¦x : p(x) = 0¦ where p is a polynomial in n variables. Let T be
the family of all ﬁnite intersections O = B
1
∩ ∩B
k
from B. By considering the constant polynomials,
we see that ∅ and X are in T . It is also clear that if O
1
, O
2
∈ T , then O
1
∩ O
2
∈ T . Now if O
α
∈ T ,
Given two distinct points x, y ∈ R
n
, say x = 'x
1
, . . . , x
n
` and y = 'y
1
, . . . , y
n
`, let p be the polynomial
¸
n
i=1
(X
i
−x
i
)
2
. Then p(x) = 0 and p(y) = 0. Thus there is an open set containing y but not x. Hence
T is T
1
.
Any two open sets are not disjoint since for any ﬁnite collection of polynomials there is an x that is not
a root of any of the polynomials. Hence T is not T
2
.
30. Let A ⊂ B ⊂
¯
A be subsets of a topological space, and let f and g be two continuous maps of B into a
Hausdorﬀ space X with f(u) = g(u) for all u ∈ A. For any v ∈ B, we have v ∈
¯
A. Suppose f(v) = g(v).
Then there are disjoint open sets O
1
and O
2
such that f(v) ∈ O
1
and g(v) ∈ O
2
. Since f and g are
continuous, there are open sets U
1
and U
2
such that v ∈ U
1
, f[U
1
] ⊂ O
1
, v ∈ U
2
, g[U
2
] ⊂ O
2
. Now
there exists u ∈ A with u ∈ U
1
∩ U
2
. Then f(u) ∈ O
1
and g(u) ∈ O
2
but f(u) = g(u) so O
1
∩ O
2
= ∅.
Contradiction. Hence f(v) = g(v). i.e. f ≡ g on B.
46
31. Omitted.
8.4 Connectedness
32. Let ¦C
α
¦ be a collection of connected sets and suppose that any two of them have a point in
common. Let G =
¸
C
α
. Suppose O
1
and O
2
is a separation of G. For any pair C
α
and C
α
, there is
a p ∈ C
α
∩ C
α
. Then p ∈ O
1
or p ∈ O
2
. Suppose p ∈ O
1
. Since C
α
is connected and p ∈ C
α
, we have
C
α
⊂ O
1
. Now any other C
α
has a point in common with C
α
so by the same argument, C
α
⊂ O
1
.
Thus G ⊂ O
1
, contradicting O
2
= ∅. Hence G is connected.
33. Let A be a connected subset of a topological space and suppose that A ⊂ B ⊂
¯
A. Suppose O
1
and
O
2
is a separation of B. Since A ⊂ B = O
1
∪O
2
and A is connected, we may assume that A ⊂ O
1
. Then
B = B ∩
¯
A ⊂ B ∩ O
1
= O
1
, contradicting O
2
= ∅. Hence B is connected.
34a. Let E be a connected subset of R having more than one point. Suppose x, y ∈ E with x < y.
Suppose there exists z ∈ (x, y) such that z / ∈ E. Then E∩(−∞, z) and E∩(z, ∞) are a separation of E,
contradicting the connectedness of E. Thus [x, y] ⊂ E. Let a = inf E and b = sup E. Clearly, E ⊂ [a, b].
If z ∈ (a, b), then there exists z
t
∈ E such that a ≤ z
t
< z and there exists z
tt
∈ E such that z < z
tt
≤ b
so that z ∈ E. Thus (a, b) ⊂ E ⊂ [a, b]. Hence E is an interval.
34b. Let I = (a, b) and let O be a subset of I that is both open and closed in I. Clearly sup¦y : (a, y) ⊂
O¦ ≤ b. Suppose d = sup¦y : (a, y) ⊂ O¦ < b. If d ∈ O, then (d −ε, d +ε) ⊂ O ⊂ I for some ε > 0 since
O is open. There exists z > d − ε such that (a, z) ⊂ O. Then (a, d + ε) = (a, z) ∪ (d − ε, d + ε) ⊂ O.
Contradiction. If d / ∈ O, then (d − ε, d + ε) ⊂ O
c
for some ε > 0 since O
c
is open. But there exists
z > d −ε such that (d −ε, z) ⊂ (a, z) ⊂ O. Contradiction. Hence d = b. Thus for any c < b, there exists
c
t
> c such that (a, c
t
) ⊂ O. i.e. c ∈ O. Hence O = I and I is connected. It follows from Q33 that
intervals of the form (a, b], [a, b), [a, b] are also connected.
35a. Let X be an arcwise connected space. Suppose O
1
and O
2
is a separation of X. Take x ∈ O
1
and y ∈ O
2
. There is a continuous function f : [0, 1] → X with f(0) = x and f(1) = y. Since [0, 1] is
connected, f[0, 1] is connected so we may assume f[0, 1] ⊂ O
1
. Then y ∈ O
1
∩O
2
. Contradiction. Hence
X is connected.
*35b. Consider X = ¦'x, y` : x = 0, −1 ≤ y ≤ 1¦ ∪ ¦'x, y` : y = sin(1/x), 0 < x ≤ 1¦. Let
I = ¦'x, y` : x = 0, −1 ≤ y ≤ 1¦ and S = ¦'x, y` : y = sin(1/x), 0 < x ≤ 1¦. Since S is the image of (0, 1]
under a continuous map, S is connected. Let '0, y` ∈ I. Given ε > 0, choose n such that 1/2nπ < ε.
Since sin
1
2
(4n + 1)π = 1 and sin
1
2
(4n + 3)π = −1, sin(1/x) takes on every value between 1 and 1 in
the interval [2/(4n + 3)π, 2/(4n + 1)π]. In particular, sin(1/x
0
) = y for some x
0
in the interval. Then
ρ['0, y`, 'x
0
, y`] = x
0
< ε so 'x
0
, y` ∈ B
¹0,y),ε
∩ S. Hence I ⊂
¯
S and by Q33, X = I ∪ S is connected.
Suppose X is arcwise connected. Then there is a path f from '0, 0` to some point of S. The set of t such
that f(t) ∈ I is closed so it has a largest element. We may assume that t = 0. Let f(t) = 'x(t), y(t)`.
Then x(0) = 0 while x(t) > 0 and y(t) = sin(1/x(t)) for t > 0. For each n, choose u with 0 < u < x(1/n)
such that sin(1/u) = (−1)
n
. Then there exists t
n
∈ (0, 1/n) such that x(t
n
) = u. Now 't
n
` converges to
0 but 'y(t
n
)` does not converge, contradicting the continuity of f. Hence X is not arcwise connected.
(*) (Closed) topologist’s sine curve
35c. Let G be a connected open set in R
n
. Let x ∈ G and let H be the set of points in G that can be
connected to x by a path. There exists δ > 0 such that B
x,δ
⊂ G. For y ∈ B
x,δ
, f(t) = (1 −t)x +ty is a
path connecting x and y. Thus H = ∅. For each y ∈ H, there exists δ
t
> 0 such that B
y,δ
⊂ G. There
is a path f connecting y to x. For each z ∈ B
y,δ
, there is a path g connecting z to y. Then h given
by h(t) = g(2t) for t ∈ [0, 1/2] and h(t) = f(2t − 1) for t ∈ [1/2, 1] is a path connecting z to x. Thus
B
y,δ
⊂ H and H is open. If y ∈
¯
H, then y ∈ B
y,δ
⊂ G for some δ > 0 and B
y,δ
contains a point z of
H. Now there is a path connecting z to x and there is a path connecting y to z. Thus there is a path
connecting y to x so y ∈ H and H is closed. Hence H = G. Since x is arbitrary, G is arcwise connected.
36. Let X be a locally connected space and let C be a component of X. If x ∈ C, then there is a
connected basic set B such that x ∈ B. Since B is connected, B ⊂ C. Thus C is open.
*37. Let X be as in Q35b. Suﬃciently small balls centred at '0, 0` do not contain connected open sets
so X is not locally connected.
47
8.5 Products and direct unions of topological spaces
38a. Let Z =
◦
¸
X
α
. Suppose f : Z → Y is continuous. For any open set O ⊂ Y , f
−1
[O] is open in Z
so f
−1
[O] ∩ X
α
is open in X
α
for each α. i.e. f[
−1
X
α
[O] is open in X
α
for each α. Thus each restriction
f[
X
α
is continuous. Conversely, suppose each restriction f[
X
α
is continuous. For any open set O ⊂ Y ,
f
−1
[O] =
¸
f[
−1
X
α
[O]. Each of the sets f[
−1
X
α
[O] is open so f
−1
[O] is open and f is continuous.
38b. F ⊂ Z is closed if and only if F
c
is open if and only if F
c
∩ X
α
is open in X
α
for each α if and
only if F ∩ X
α
is closed in X
α
for each α.
38c. Suppose Z is Hausdorﬀ. Given x, y ∈ X
α
, x, y ∈ Z so there are disjoint open sets O
1
, O
2
⊂ Z
such that x ∈ O
1
and y ∈ O
2
. Then O
1
∩ X
α
and O
2
∩ X
α
are disjoint open sets in X
α
containing x
and y respectively. Thus X
α
is Hausdorﬀ. Conversely, suppose each X
α
is Hausdorﬀ. Given x, y ∈ Z,
x ∈ X
α
and y ∈ X
β
for some α and β. If α = β, then since X
α
is Hausdorﬀ, there are disjoint open sets
O
1
, O
2
⊂ X
α
such that x ∈ O
1
and y ∈ O
2
. The sets O
1
and O
2
are also open in Z so it follows that Z
is Hausdorﬀ. If α = β, then since X
α
∩ X
β
= ∅ and the sets X
α
and X
β
are open in Z, it follows that
Z is Hausdorﬀ.
38d. Suppose Z is normal. Given disjoint closed sets F
1
, F
2
⊂ X
α
, F
1
and F
2
are closed in Z so there are
disjoint open sets O
1
, O
2
⊂ Z such that F
1
⊂ O
1
and F
2
⊂ O
2
. Then F
1
⊂ O
1
∩ X
α
and F
2
⊂ O
2
∩ X
α
with O
1
∩X
α
and O
2
∩X
α
being disjoint open sets in X
α
. Thus X
α
is normal. Conversely, suppose each
X
α
is normal. Given disjoint closed sets F
1
, F
2
⊂ Z, the sets F
1
∩ X
α
and F
2
∩ X
α
are closed in X
α
for
each α so there are disjoint open sets O
1,α
, O
2,α
⊂ X
α
such that F
1
∩ X
α
⊂ O
1,α
and F
2
∩ X
α
⊂ O
2,α
.
Then F
1
⊂
¸
O
1,α
and F
2
⊂
¸
O
2,α
with
¸
O
1,α
and
¸
O
2,α
being disjoint open sets in Z. Hence Z is
normal.
39a. Let X
1
⊂ X be a direct summand. Then it follows from the deﬁnition that X
1
is open. Let X
2
be
another direct summand. Then X
2
is open and X
1
⊂ X
c
2
. Now X
c
2
is closed so X
1
= X
c
2
∩ X
1
is closed.
39b. Let X
1
be a subset of X that is both open and closed. Let X
2
= X ` X
1
. Then X
2
is open,
X
1
∩ X
2
= ∅ and X = X
1
∪ X
2
. If O is open in X, then O ∩ X
i
is open in X
i
for i = 1, 2. On the
other hand, if O is open in X
1
◦
∪ X
2
, then O ∩ X
i
is open in X
i
for i = 1, 2 and thus open in X. Then
O = (O ∩ X
1
) ∪ (O ∩ X
2
) is open in X. Hence X = X
1
◦
∪ X
2
.
*40a. If X has a base, each element being Tychonoﬀ, let x, y be distinct points in X. There is a basic
element B containing x. Since B is Tychonoﬀ, there exists O open in B such that x ∈ O but y / ∈ O.
Then O is also open in X so X is Tychonoﬀ.
Consider E = R¦0, 1¦/ ∼ where ∼ is the smallest equivalence relation with 'x, 0` 'x, 1` for x ∈ R`¦0¦.
Let q : R¦0, 1¦ → E be the quotient map and consider the base q[(a, b) ¦e¦] for a < b and e ∈ ¦0, 1¦.
Then E is not Hausdorﬀ and thus not regular and not completely regular but the elements in the base
are.
(*) Error in book.
*40b.
*40c.
41. Let (X, ρ) be a metric space with an extended realvalued metric and X
α
its parts. i.e. equivalence
classes under the equivalence relation ρ(x, y) < ∞. Then X is the disjoint union of its parts and by
Q7.3b, each part is open (and closed). Thus X is the direct union X =
◦
¸
X
α
.
42. Let 'X
α
, T
α
` be a family of Hausdorﬀ spaces. Given distinct points x, y ∈
¸
α
X
α
, we have x
α
= y
α
for some α. Since X
α
is Hausdorﬀ, there are disjoint open sets O
1
, O
2
⊂ X
α
such that x ∈ O
1
and
y ∈ O
2
. Then π
−1
α
[O
1
] and π
−1
α
[O
2
] are disjoint open sets in
¸
α
X
α
containing x and y respectively.
43. Note that
¸
α
X
α
is a basic element. If x ∈ B
1
∩B
2
where B
i
=
¸
α
O
α,i
, then x
α
∈ O
α,1
∩O
α,2
for
each α. Now each O
t
α
= O
α,1
∩ O
α,2
is open in X
α
and only ﬁnitely many of them are not X
α
. Thus
x ∈
¸
α
O
t
α
⊂ B
1
∩ B
2
.
Let (X, ρ) and (Y, σ) be two metric spaces. Since ρ
∞
is equivalent to the usual product metric, we
consider (X Y, ρ
∞
). The open ball B
¹x,y),ε
is the same as B
x,ε
B
y,ε
so B
¹x,y),ε
is open in the
product topology. For any U V with U open in X and V open in Y , given 'x, y` ∈ U V , there
are open balls B
x,ε
and B
y,δ
such that x ∈ B
x,ε
⊂ U and y ∈ B
y,δ
⊂ V . Let η = min(ε, δ). Then
48
B
¹x,y),η
⊂ B
x,ε
B
y,δ
⊂ U V . Thus U V is open in the metric topology. Hence the product topology
on X Y is the same as the topology induced by the product metric.
44. By deﬁnition, X
A
=
¸
α∈A
X
α
. We may identify each x ∈
¸
α∈A
X
α
with the function f : A →
X given by f(α) = x
α
. We may then also identify each basic element
¸
α∈A
O
α
with ¦f : f(α) ∈
O
α
for each α¦. But since all but ﬁnitely many of the sets O
α
are X
α
, we only need to consider the sets
¦f : f(α
1
) ∈ O
1
, . . . , f(α
n
) ∈ O
n
¦ where ¦α
1
, . . . , α
n
¦ is some ﬁnite subset of A and ¦O
1
, . . . , O
n
¦ is a
ﬁnite collection of open subsets of X.
Suppose a sequence 'f
n
` converges to f in X
A
. Then we may regard this as a sequence 'x
n
` converging
to x under the identiﬁcation described above. Now since π
α
is continuous, 'π
α
(x
n
)` converges to π
α
(x).
i.e. 'x
n,α
` converges to x
α
for each α. Under the identiﬁcation again, 'f
n
(α)` converges to f(α) for each
α. Conversely, suppose 'f
n
(α)` converges to f(α) for each α. Let B be a basic element containing f.
Then f(α
i
) ∈ O
i
for some ﬁnite subset ¦α
1
, . . . , α
m
¦ ⊂ A and some ﬁnite collection ¦O
1
, . . . , O
m
¦ of
open subsets of X. Then there exists N such that f
n
(α
i
) ∈ O
i
for n ≥ N and i = 1, . . . , m. Thus f
n
∈ B
for n ≥ N and 'f
n
` converges to f in X
A
.
45. Suppose X is metrizable and A is countable. We may enumerate A as ¦α
1
, α
2
, . . .¦. Then by
Q7.11a, X can be given an equivalent metric ρ that is bounded by 1. Deﬁne σ on X
A
by σ(x, y) =
¸
n
2
−n
ρ(x
α
n
, y
α
n
). Then σ is a metric on X
A
so X
A
is metrizable.
46. Given an open set O
α
0
⊂ X
α
0
, π
−1
α
[O
α
0
] = ¦x ∈
¸
α
X
α
: x
α
0
∈ O
α
0
¦, which is open in
¸
α
X
α
.
Hence each π
α
is continuous. If T is a topology on X
A
such that each π
α
is continuous, then each basic
element in the product topology is a ﬁnite intersection of preimages under some π
α
of open sets in X.
Thus each basic element in the product topology is in T and thus the product topology is contained in
T . Hence the product topology is the weakest topology such that each π
α
is continuous.
47. The ternary expansion of numbers gives a homeomorphism between 2
ω
and the Cantor ternary set.
48a. Each x ∈ X can be identiﬁed with the element in I
F
with fth coordinate f(x). Let F be the
mapping of X onto its image in I
F
. Suppose each f ∈ F is continuous. Given a basic element
¸
f
O
f
⊂ I
F
,
F
−1
[
¸
f
O
f
] = ¦x : F(x) ∈
¸
f
O
f
¦ = ¦x : f(x) ∈ O
f
for each f¦ =
¸
f
−1
[O
f
]. The intersection is in
fact a ﬁnite intersection since all but ﬁnitely many of the sets O
f
are I. Thus F
−1
[
¸
f
O
f
] is open and
F is continuous. Further suppose that given a closed set F and x / ∈ F there is f ∈ F such that f[F] = 0
and f(x) = 1. Let U be open in X and let y ∈ F[U]. Now y = F(x) for some x ∈ U and there is f ∈ F
such that f[U
c
] = 0 and f(x) = 1. Let V = π
−1
f
[(0, ∞)] and let W = V ∩ F[X]. Then W is open. Also,
π
f
(y) = f(x) = 1 so y ∈ W. If z ∈ W, then z = F(x) for some x ∈ X with f(x) > 0 so x ∈ U. Thus
W ⊂ F[U]. Thus F[U] is open in I
F
. Hence F is a homeomorphism.
*48b. Suppose X is a normal space satisfying the second axiom of countability. Let ¦B
n
¦ be a countable
base for X. For each pair of indices n, m such that B
n
⊂ B
m
, by Urysohn’s Lemma, there exists a
continuous function g
n,m
on X such that g
n,m
≡ 1 on B
n
and g
n,m
≡ 0 on B
c
m
. Given a closed set F
and x / ∈ F, choose a basic element B
m
such that x ∈ B
m
⊂ F
c
. By Q23a, there exists B
n
such that
x ∈ B
n
and B
n
⊂ B
m
. Then g
n,m
is deﬁned with g
n,m
(x) = 1 and g
n,m
[F] = 0. Furthermore the family
¦g
n,m
¦ is countable.
48c. Suppose X is a normal space satisfying the second axiom of countability. By part (b), there is a
countable family F of continuous functions with the property in part (a). Then there is a homeomorphism
between X and I
F
. By Q45, I
F
is metrizable and thus X is metrizable.
*49. First we consider ﬁnite products of connected spaces. Suppose X and Y are connected and choose
'a, b` ∈ X Y . The subspaces X ¦b¦ and ¦x¦ Y are connected, being homeomorphic to X and Y
respectively. Thus T
x
= (X ¦b¦) ∪(¦x¦ Y ) is connected for each x ∈ X. Then
¸
x∈X
T
x
is the union
of a collection of connected spaces having the point 'a, b` in common so it is connected. But this union
is X Y so X Y is connected. The result for any ﬁnite product follows by induction.
Let ¦X
α
¦
α∈A
be a collection of connected spaces and let X =
¸
α
X
α
. Fix a point a ∈ X. For any ﬁnite
subset K ⊂ A, let X
K
be the subspace consisting of points x such that x
α
= a
α
for α / ∈ K. Then X
K
is homeomorphic to the ﬁnite product
¸
α∈K
X
α
so it is connected. Let Y be the union of the sets X
K
.
Since any two of them have a point in common, by Q32, Y is connected. Let x ∈ X and let U =
¸
α
O
α
be a basic element containing x. Now O
α
= X
α
for all but ﬁnitely many α so let K be that ﬁnite set.
Let y be the element with y
α
= x
α
for α ∈ K and y
α
= a
α
for α / ∈ K. Then y ∈ X
K
⊂ Y and y ∈ U.
Hence X =
¯
Y so X is connected.
49
8.6 Topological and uniform properties
50a. Let 'f
n
` be a sequence of continuous maps from a topological space X to a metric space (Y, σ) that
converges uniformly to a map f. Given ε > 0, there exists N such that σ(f
n
(x), f(x)) < ε/3 for n ≥ N
and x ∈ X. Given x ∈ X, there is an open set O containing x such that σ(f
N
(x), f
N
(y)) < ε/3 for
y ∈ O. Then σ(f(x), f(y)) ≤ σ(f(x), f
N
(x)) +σ(f
N
(x), f
N
(y)) +σ(f
N
(y), f(y)) < ε for y ∈ O. Hence f
is continuous.
50b. Let 'f
n
` be a sequence of continuous maps from a topological space X to a metric space (Y, σ)
that is uniformly Cauchy. Suppose Y is complete. For each x ∈ X, the sequence 'f
n
(x)` is Cauchy
so it converges since Y is complete. Let f(x) be the limit of the sequence. Then 'f
n
` converges to
f. Given ε > 0, there exists N such that σ(f
n
(x), f
m
(x)) < ε/2 for n, m ≥ N and x ∈ X. For each
x ∈ X, there exists N
x
≥ N such that σ(f
N
x
(x), f(x)) < ε/2. Then for n ≥ N and x ∈ X, we have
σ(f
n
(x), f(x)) ≤ σ(f
n
(x), f
N
x
(x)) + σ(f
N
x
(x), f(x)) < ε. Thus 'f
n
` converges uniformly to f and by
part (a), f is continuous.
51. The proofs of Lemmas 7.3739 remain valid when X is a separable topological space. Thus the
AscoliArzel´ a Theorem and its corollary are still true.
8.7 Nets
52. Suppose X is Hausdorﬀ and suppose a net 'x
α
` in X has two limits x, y. Then there are disjoint
open sets U and V such that x ∈ U and y ∈ V . Since x and y are limits, there exist α
0
and α
1
such
that x
α
∈ U for α ~ α
0
and x
α
∈ V for α ~ α
1
. Choose α
t
such that α
t
~ α
0
and α
t
~ α
1
. Then
x
α
∈ U ∩V for α ~ α
t
. Contradiction. Hence every net in X has at most one limit. Conversely, suppose
X is not Hausdorﬀ. Let x, y be two points that cannot be separated and let the directed system be the
collection of all pairs 'A, B` of open sets with x ∈ A, y ∈ B. Choose x
¹A,B)
∈ A ∩ B. Let O be an
open set containing x and let O
t
be an open set containing y. For 'A, B` ~ 'O, O
t
`, we have A ⊂ O and
B ⊂ O
t
so x
¹A,B)
∈ A∩ B ⊂ O ∩ O
t
. Thus both x and y are limits.
53. Suppose f is continuous. Let 'x
α
` be a net that converges to x. For any open set O containing f(x),
we have x ∈ f
−1
[O], which is open. There exists α
0
such that x
α
∈ f
−1
[O] for α ~ α
0
. Then f(x
α
) ∈ O
for α ~ α
0
. Hence 'f(x
α
)` converges to f(x). Conversely, suppose that for each net 'x
α
` converging to
x the net 'f(x
α
)` converges to f(x). Then in particular the statement holds for all sequences. It follows
that f is continuous.
*54. Let X be any set and f a realvalued function on X. Let A be the system consisting of all ﬁnite
subsets of X, with F ≺ G meaning F ⊂ G. For each F ∈ A, let y
F
=
¸
x∈F
f(x). Suppose that f(x) = 0
except for x in a countable subset ¦x
n
¦ and
¸
[f(x
n
)[ < ∞. Given ε > 0, there exists N such that
¸
∞
n=N+1
[f(x
n
)[ < ε. Let y =
¸
f(x
n
). For any open interval (y −ε, y +ε), let F
0
= ¦x
1
, . . . , x
N
¦. For
F ~ F
0
, [y −y
F
[ ≤
¸
∞
n=N+1
[f(x
n
)[ < ε so y
F
∈ (y −ε, y +ε). Thus limy
F
= y. Conversely, if f(x) = 0
on an uncountable set G, then for some n, [f(x)[ > 1/n for uncountably many x. Thus by considering
arbitrarily large ﬁnite subsets of G, we see that 'y
F
` does not converge. Hence f(x) = 0 except on a
countable set. Now if
¸
[f(x
n
)[ = ∞, then we only have
¸
f(x
n
) < ∞ and it follows that the limit is
not unique. Hence
¸
[f(x
n
)[ < ∞.
55. Let X =
¸
α
X
α
. Suppose a net 'x
β
` in X converges to x. Since each projection is continuous, each
coordinate of x
β
converges to the corresponding coordinate of x. Conversely, suppose each coordinate of
x
β
converges to the corresponding coordinate of x. Let
¸
α
O
α
be a basic element containing x. Then
π
α
(x) ∈ O
α
for each α. For each α, there exists β
α
such that π
α
(x
β
) ∈ O
α
for β ~ β
α
. Since all but
ﬁnitely many of the O
α
are X
α
, we only need to consider a ﬁnite set β
α
1
, . . . , β
α
n
. In particular, we may
choose β
0
such that β
0
~ β
α
i
for all i. For β ~ β
0
, we have π
α
i
(x
β
) ∈ O
α
i
. For α = α
i
, we also have
π
α
(x
β
) ∈ O
α
= X
α
. Thus x
β
∈
¸
α
O
α
for β ~ β
0
. Hence the net 'x
β
` converges to x.
50
9 Compact and Locally Compact Spaces
9.1 Compact spaces
1. Suppose X is compact. Then every open cover has a ﬁnite subcovering. In particular, it has a ﬁnite
reﬁnement. Conversely, suppose every open cover has a ﬁnite reﬁnement. Since every element in the
reﬁnement is a subset of an element in the open cover, the open cover has a ﬁnite subcovering so X is
compact.
*2. Let 'K
n
` be a decreasing sequence of compact sets with K
0
Hausdorﬀ. Let O be an open set
with
¸
K
n
⊂ O. Suppose K
n
is not a subset of O for any n. Then K
n
` O is nonempty and closed
in K
n
. Since K
n
is a compact subset of the Hausdorﬀ space K
0
, K
n
is closed in K
0
so K
n
` O is
closed in K
0
. Also, 'K
n
` O` is a decreasing sequence so it has the ﬁnite intersection property. Now
∅ = (
¸
K
n
) ` O =
¸
(K
n
` O) ⊂
¸
K
n
⊂ O. Contradiction. Hence K
n
⊂ O for some n.
(*) Assume K
0
is Hausdorﬀ.
3. Suppose X is a compact Hausdorﬀ space. Let F be a closed subset and let x / ∈ F. For each y ∈ F,
there are disjoint open sets U
y
and V
y
with x ∈ U
y
and y ∈ V
y
. Now F is compact and ¦V
y
: y ∈ F¦ is an
open cover for F. Thus there is a ﬁnite subcovering ¦V
y
1
, . . . , V
y
n
¦. Let U =
¸
n
i=1
U
y
i
and V =
¸
n
i=1
V
y
i
.
Then U and V are disjoint open sets with x ∈ U and F ⊂ V . Hence X is regular.
4. Suppose X is a compact Hausdorﬀ space. Let F and G be disjoint closed subsets. By Q3, for each
y ∈ G, there are disjoint open sets U
y
and V
y
such that F ⊂ U
y
and y ∈ V
y
. Now G is compact and
¦V
y
: y ∈ G¦ is an open cover for G. Thus there is a ﬁnite subcovering ¦V
y
1
, . . . , V
y
n
¦. Let U
t
=
¸
n
i=1
U
y
i
and V
t
=
¸
n
i=1
V
y
i
. Then U
t
and V
t
are disjoint open sets with F ⊂ U
t
and G ⊂ V
t
. Hence X is normal.
5a. If (X, T ) is a compact space, then for T
∞
weaker than T , any open cover from T
1
is an open cover
from T so it has a ﬁnite subcovering. Thus (X, T
1
) is compact.
5b. If (X, T ) is a Hausdorﬀ space, then for T
∈
stronger than T , any two distinct points in X can be
separated by disjoint sets in T , which are also sets in T
∈
. Thus (X, T
2
) is Hausdorﬀ.
5c. Suppose (X, T ) is a compact Hausdorﬀ space. If T
1
is a weaker topology, then id : (X, T ) → (X, T
1
)
is a continuous bijection. If (X, T
1
) is Hausdorﬀ, then id is a homeomorphism. Contradiction. Hence
(X, T
1
) is not Hausdorﬀ. If T
2
is a stronger topology, then id : (X, T
2
) → (X, T ) is a continuous bijection.
If (X, T
2
) is compact, then id is a homeomorphism. Contradiction. Hence (X, T
2
) is not compact.
6. Let X be a compact space and F an equicontinuous family of maps from X to a metric space (Y, σ).
Let 'f
n
` be a sequence from F such that f
n
(x) → f(x) for all x ∈ X. For each x ∈ X, given ε > 0,
there exists N
x
such that σ(f
n
(x), f(x)) < ε/3 for n ≥ N
x
. Also, there exists an open set O
x
containing
x such that σ(f
n
(x), f
n
(y)) < ε/3 for y ∈ O
x
and all n. Then we also have σ(f(x), f(y)) < ε/3 for
y ∈ O
x
. Now ¦O
x
: x ∈ X¦ is an open cover for X so there is a ﬁnite subcovering ¦O
x
1
, . . . , O
x
k
¦. Let
N = max
1≤i≤k
N
x
i
. For n ≥ N, σ(f
n
(x
i
), f(x
i
)) < ε/3 for 1 ≤ i ≤ k. For each x ∈ X, x ∈ O
x
i
for
some i so σ(f
n
(x), f(x)) ≤ σ(f
n
(x), f
n
(x
i
)) + σ(f
n
(x
i
), f(x
i
)) + σ(f(x
i
), f(x)) < ε for n ≥ N. Hence
'f
n
` converges uniformly to f on X.
*7. Let X be a Hausdorﬀ space and 'C
n
` a decreasing sequence of compact and connected sets. Let
C =
¸
C
n
. For any open cover  of C,  is an open cover of some C
n
by Q2. Thus it has a ﬁnite
subcovering, which also covers C. Hence C is compact.
Suppose C is disconnected with A and B being a separation for C. Then A and B are nonempty disjoint
closed subsets of C. Since C is an intersection of closed sets, C is closed. Thus A and B are closed in
C
0
. Since C
0
is compact and Hausdorﬀ, it is normal. Thus there are disjoint open sets U, V ⊂ C
0
such
that A ⊂ U and B ⊂ V . Then C = A ∪ B ⊂ U ∪ V . By Q2, C
n
⊂ U ∪ V for some n. Hence C
n
is
disconnected. Contradiction. Hence C is connected.
(*) Assume X is Hausdorﬀ
8a. Let 'f
n
` be a sequence of maps from X to Y that converge in the compactopen topology to f.
For x ∈ X, let O be an open set containing f(x). Then N
¦x¦,O
is open in the compactopen topology
and contains f. There exists N
t
such that f
n
∈ N
¦x¦,O
for n ≥ N
t
. i.e. f
n
(x) ∈ O for n ≥ N
t
. Hence
f(x) = limf
n
(x).
*8b. Let 'f
n
` be a sequence of continuous maps from a topological space X to a metric space (Y, σ).
Suppose 'f
n
` converges to f uniformly on each compact subset C of X. Let N
K,O
be a subbasic element
51
containing f. Then f[K] is a compact set disjoint from O
c
so σ(f[K], O
c
) > 0. Let ε = σ(f[K], O
c
).
There exists N
t
such that σ(f
n
(x), f(x)) < ε for n ≥ N
t
and x ∈ K. Then f
n
(x) ∈ O for n ≥ N
t
and
x ∈ K. i.e. f
n
∈ N
K,O
for n ≥ N
t
. Hence 'f
n
` converges to f in the compactopen topology.
Conversely, suppose 'f
n
` converges to f in the compactopen topology. Let C be a compact subset
of X and let ε > 0 be given. Since C is compact and f is continuous, f[C] is compact. Thus there
exist x
1
, . . . , x
n
∈ C such that the open balls B
f(x
1
),ε/4
, . . . , B
f(x
n
),ε/4
cover f[C]. For each i, let
C
i
= C∩f
−1
[B
f(x
i
),ε/4
] and O
i
= B
f(x
i
),ε/4
. Then C
i
is compact and f[C
i
] ⊂ O
i
. Thus f ∈
¸
n
i=1
N
C
i
,O
i
.
There exists N
t
such that f
n
∈
¸
n
i=1
N
C
i
,O
i
for n ≥ N
t
. For any x ∈ C, x ∈ C
i
for some i so
σ(f(x
i
), f(x)) < ε/4. Also, f
n
(x) ∈ O
i
for n ≥ N
t
so σ(f
n
(x), f(x
i
)) < ε/4 for n ≥ N
t
. Thus
σ(f
n
(x), f(x)) < ε for n ≥ N
t
and x ∈ C. Hence 'f
n
` converges uniformly to f on C.
(*) Assume f to be continuous.
9.2 Countable compactness and the BolzanoWeierstrass property
9a. A realvalued function f on X is continuous if and only if ¦x : f(x) < α¦ and ¦x : f(x) > α¦ are
open for any real number α if and only if f is both upper semicontinuous and lower semicontinuous.
9b. Suppose f and g are upper semicontinuous. Then ¦x : f(x) < α¦ and ¦x : g(x) < α¦ are open for
any real number α. Now for any real number α, ¦x : f(x) + g(x) < α¦ =
¸
q∈Q
[¦x : g(x) < q¦ ∩ ¦x :
f(x) < α −q¦], which is open. Hence f +g is upper semicontinuous.
9c. Let 'f
n
` be a decreasing sequence of upper semicontinuous functions which converge pointwise to a
realvalued function f. If f(x) < α, then there exists N such that f
n
(x) − f(x) < α − f(x) for n ≥ N.
i.e. f
n
(x) < α for n ≥ N. On the other hand, if f
n
(x) < α for some n, then f(x) ≤ f
n
(x) < α. Thus
¦x : f(x) < α¦ =
¸
n
¦x : f
n
(x) < α¦, which is open. Hence f is upper semicontinuous.
9d. Let 'f
n
` be a decreasing sequence of upper semicontinuous functions on a countably compact space,
and suppose that limf
n
(x) = f(x) where f is a lower semicontinuous realvalued function. By part (c),
f is also upper semicontinuous so f is continuous. Now 'f
n
−f` is a sequence of upper semicontinuous
functions on a countably compact space such that for each x, 'f
n
(x) −f(x)` decreases to zero. Thus by
Proposition 11 (Dini), 'f
n
−f` converges to zero uniformly. i.e. 'f
n
` converges to f uniformly.
9e. Suppose a sequence 'f
n
` of upper semicontinuous functions converges uniformly to a function f.
Fix y ∈ X. Given ε > 0, there exists N such that [f
N
(x) − f(x)[ < ε/3 for all x ∈ X. Since
¦x : f
N
(x) < f
N
(y) +ε/3¦ is open, there exists δ > 0 such that [x−y[ < δ implies f
N
(x) −f
N
(y) < ε/3.
Now if [x − y[ < δ, then f(x) − f(y) = [f(x) − f
N
(x)] + [f
N
(x) − f
N
(y)] + [f
N
(y) − f(y)] < ε. Given
α ∈ R, pick y ∈ ¦x : f(x) < α¦. There exists δ > 0 such that [x −y[ < δ implies f(x) −f(y) < α−f(y).
i.e. f(x) < α. Hence ¦x : f(x) < α¦ is open and f is upper semicontinuous.
*10 (i) ⇒ (iii) by Proposition 9. Suppose that every bounded continuous realvalued function on X
assumes its maximum. Let f be a continuous function and suppose it is unbounded. We may assume,
by taking max(1, f), that f ≥ 1. Then −1/f is a bounded continuous function with no maximum.
Thus (iii) ⇒ (ii). Suppose X is no countably compact. Then it does not have the BolzanoWeierstrass
property. There is a sequence 'x
n
` in X with no cluster point. Thus the sequence has no limit points.
Let A = ¦x
n
¦. Then A is closed and since all subsets of A are also closed, A is discrete. Deﬁne f(x
n
) = n
for x
n
∈ A. Then f is continuous on the closed set A and since X is normal, by Tietze’s Extension
Theorem, there is a continuous function g on X such that g[
A
= f. Then g is an unbounded continuous
function on X. Thus (ii) ⇒ (i).
11a. Let X be the set of ordinals less than the ﬁrst uncountable ordinal and let B be the collection of
sets of the form ¦x : x < a¦, ¦x : a < x < b¦, ¦x : a < x¦. For any x
0
∈ X, x
0
∈ ¦x : x < x
0
+ 1¦. Now
¦x : x < b¦ ∩ ¦x : a < x¦ = ¦x : a < x < b¦, ¦x : x < c¦ ∩ ¦x : a < x < b¦ = ¦x : a < x < min(b, c)¦,
¦x : max(a, c) < x < b¦ = ¦x : a < x < b¦ ∩ ¦x : c < x¦. Hence B is a base for a topology on X.
*11b. For any sequence 'x
n
` in X, let x
0
= sup x
n
. Then x
n
< x
0
+1 for all n. i.e. x
n
∈ ¦x : x < x
0
+1¦
for all n so the sequence converges to x
0
. Hence X is sequentially compact. The sets ¦x : x < a¦ form
an open cover for X. If it has a ﬁnite subcovering, then there exists a ∈ X such that x < a for all x ∈ X.
Contradiction. Hence X is not compact.
*11c. Let f be a continuous realvalued function on X. We ﬁrst show the existence of a sequence 'x
n
`
in X such that [f(y) −f(x
n
)[ < 1/n for y > x
n
. If no such sequence exists, we may construct for some
52
N an increasing sequence 'z
n
` such that [f(z
n
) −f(z
n−1
)[ ≥ 1/N for each n. Then 'z
n
` converges to its
supremum but 'f(z
n
)` does not converge, contradicting the continuity of f. Now let x
0
= sup x
n
. Then
f(x) = f(x
0
) for x ≥ x
0
.
12a. Similar argument as Q11a.
*12b. Let  be an open cover for Y consisting of basic sets. Let L
a
= ¦x : x < a¦ and U
b
= ¦x : b < x¦.
Then ω
1
∈ U
b
for some U
b
∈ . Also, 0 ∈ L
a
for some L
a
∈ . Let a
0
= sup¦a : L
a
∈ ¦. Then a
0
∈ U
b
0
for some U
b
0
∈  so b
0
< a
0
. Then there exists a
1
> b
0
such that L
a
1
∈ . Since y > b
0
or y < a
1
for
any y ∈ Y , ¦L
a
1
, U
b
0
¦ is a ﬁnite subcovering of  that covers Y . Hence Y is compact.
Let D be a countable subset of Y . Then y = sup D exists and y ∈ Y . Now ¦x : y < x¦ is an open set
in Y that does not intersect D. Thus D is not dense in Y and Y is not separable. Suppose there is a
countable base ¦U
n
¦ at ω
1
. Then each U
n
is of the form ¦x : x > a
n
¦. There exists a < ω
1
such that
a > a
n
for all n. Then the open set ¦x : x > a¦ does not contain any U
n
. Contradiction. Thus Y is not
ﬁrst countable.
9.3 Products of compact spaces
13. Each closed and bounded subset X of R
n
is contained in a cube I
n
where I = [a, b]. Each I is
compact in R so I
n
is compact in R
n
by Tychonoﬀ’s Theorem. Now X is closed in I
n
so X is compact.
*14. Suppose X is compact and I is a closed interval. Let  be an open cover of X I and let t ∈ I.
Since X ¦t¦ is homeomorphic to X, X ¦t¦ is compact so there is a ﬁnite subcovering ¦U
1
, . . . , U
n
¦
of  such that U =
¸
n
i=1
U
i
⊃ X ¦t¦. Now X ¦t¦ can be covered by ﬁnitely many basic sets
A
1
B
1
, . . . , A
k
B
k
⊂ U. Then B = B
1
∩ ∩ B
k
is an open set containing t. If 'x, y` ∈ X B,
then 'x, t` ∈ A
j
B
j
for some j so x ∈ A
j
and y ∈
¸
n
i=1
B
i
⊂ B
j
. Thus 'x, y` ∈ A
j
B
j
and
XB ⊂
¸
n
i=1
(A
i
B
i
) ⊂ U. Thus for each t ∈ I, there is an open set B
t
containing t such that XB
t
can be covered by ﬁnitely many elements of . The collection of the sets B
t
forms an open cover of X so
there is a ﬁnite subcollection ¦B
t
1
, . . . , B
t
m
¦ covering X. Now X I =
¸
m
i=1
(B
t
i
I), which can then
be covered by ﬁnitely many elements of .
15. Let 'X
n
` be a countable collection of sequentially compact spaces and let X =
¸
X
n
. Given a
sequence 'x
n
` in X, there is a subsequence 'x
(1)
n
` whose ﬁrst coordinate converges. Then there is a
subsequence 'x
(2)
n
` of 'x
(1)
n
` whose second coordinate converges. Consider the diagonal sequence 'x
(n)
n
`.
Each coordinate of this sequence converges so the sequence converges in X. Hence X is sequentially
compact.
16. Let X be a compact Hausdorﬀ space. Let F be the family of continuous realvalued functions on
X with values in [0, 1]. Let Q =
¸
f∈F
I
f
. Consider the mapping g of X into Q mapping x to the
point whose fth coordinate is f(x). If x = y, then since X is compact Hausdorﬀ, and thus normal, by
Urysohn’s Lemma, there exists f ∈ F such that f(x) = 0 and f(y) = 1. Thus g is onetoone. Since
f is continuous for each f ∈ F, g is continuous. Now g[X] is a compact subset of the Hausdorﬀ space
Q so g[X] is closed in Q. Furthermore, g is a continuous bijection from the compact space X onto the
Hausdorﬀ space g[X] so X is homeomorphic to g[X].
*17. Let Q = I
A
be a cube and let f be a continuous realvalued function on Q. Given ε > 0, cover
f[Q] by ﬁnitely many open intervals I
1
, . . . , I
n
of length ε. Consider f
−1
[I
j
] for j = 1, . . . , n. These sets
cover Q and we may assume each of them is a basic set, that is, f
−1
[I
j
] =
¸
α
U
(j)
α
where U
(j)
α
is open in
I and all but ﬁnitely many of the U
(j)
α
are I. For each j, let F
j
= ¦α : U
(j)
α
= I¦ and let F =
¸
n
j=1
F
j
,
which is a ﬁnite set. Deﬁne h : Q → Q by π
α
h(x) = x
α
for α ∈ F and 0 otherwise. Deﬁne g = f ◦ h.
Then g depends only on F and [f −g[ < ε.
9.4 Locally compact spaces
18. Let X be a locally compact space and K a compact subset of X. For each x ∈ K, there is an open
set O
x
containing x with O
x
compact. Since K is compact, there is a ﬁnite subcollection ¦O
x
1
, . . . , O
x
n
¦
that covers K. Let O =
¸
n
i=1
O
x
i
. Then O ⊃ K and
¯
O =
¸
n
i=1
O
x
i
is compact.
*19a. Let X be a locally compact Hausdorﬀ space and K a compact subset. Then K is closed in
53
X
∗
. By Q8.23a, there exists a closed set D containing ω with D ∩ K = ∅. By Urysohn’s Lemma,
there is a continuous function g on X
∗
with 0 ≤ g ≤ 1 that is 1 on K and 0 on D. Deﬁne f(x) =
min(2(g(x) −1/2), 0). Then ¦x : f(x) > 0¦ = ¦x : g(x) > 1/2¦, which is compact because g is continuous
on X
∗
.
(*) Alternatively, use Q16 to regard K as a closed subset of the compact Hausdorﬀ space Q.
*19b. Let K be a compact subset of a locally compact Hausdorﬀ space X. Then by Q18, there is an
open set O ⊃ K with
¯
O compact. Now X
∗
` O and K are disjoint closed subsets of X
∗
so by Urysohn’s
Lemma, there is a continuous function g on X
∗
with 0 ≤ g ≤ 1 that is 1 on K and 0 on X
∗
` O. Then
f = g[
X
is the required function.
20a. Let X
∗
be the Alexandroﬀ onepoint compactiﬁcation of a locally compact Hausdorﬀ space X.
Consider the collection of open sets of X and complements of compact subsets of X. Then ∅ and X
∗
are
in the collection. If U
1
and U
2
are open in X, then so is U
1
∩ U
2
. If K
1
and K
2
are compact subsets of
X, then (X
∗
` K
1
) ∩ (X
∗
` K
2
) = X
∗
` (K
1
∪ K
2
) where K
1
∪ K
2
is compact. Also, U
1
∩ (X
∗
` K
1
) =
U
1
∩(X` K
1
), which is open in X. Thus the collection of sets is closed under ﬁnite intersection. If ¦U
α
¦
is a collection of open sets in X, then
¸
U
α
is open in X. If ¦K
β
¦ is a collection of compact subsets
of X, then
¸
(X
∗
` K
β
) = X
∗
`
¸
K
β
. Since each K
β
is compact in the Hausdorﬀ space X, each K
β
is closed in X and
¸
K
β
is closed in X. Thus
¸
K
β
is closed in each K
β
so
¸
K
β
is compact. Finally,
¸
U
α
∪
¸
(X
∗
` K
β
) = U ∪ (X
∗
` K) = X
∗
` (K ` U), where U =
¸
U
α
and K =
¸
K
β
. Since K ` U
is closed in the compact set K, K ` U is compact. Thus the collection of sets is closed under arbitrary
union. Hence the collection of sets forms a topology for X
∗
.
20b. Let id be the identity mapping from X to X
∗
` ¦ω¦. Clearly id is a bijection. If U is open in
X
∗
` ¦ω¦, then U = (X
∗
` ¦ω¦) ∩ U
t
for some U
t
open in X
∗
. If U
t
is open in X, then U = U
t
so
U is open in X. If U
t
= X
∗
` K for some compact K ⊂ X, then U = X ` K, which is open in X.
Thus id is continuous. Also, any open set in X is open in X
∗
so id is an open mapping. Hence id is a
homeomorphism.
20c. Let  be an open cover of X
∗
. Then  contains a set of the form X
∗
`K for some compact K ⊂ X.
Take the other elements of  and intersect each of them with X to get an open cover of K. Then there
is a ﬁnite subcollection that covers K. The corresponding ﬁnite subcollection of  together with X
∗
` K
then covers X
∗
. Hence X
∗
is compact.
Let x, y be distinct points in X
∗
. If x, y ∈ X, then there are disjoint open sets in X, and thus in X
∗
,
that separate x and y. If x ∈ X and y = ω, then there is an open set O containing x with
¯
O compact.
The sets O and X
∗
`
¯
O are disjoint open sets in X
∗
separating x and y. Hence X
∗
is Hausdorﬀ.
*21a. Let S
n
denote the unit sphere in R
n+1
. Let p = '0, . . . , 0, 1` ∈ R
n+1
. Deﬁne f : S
n
− p → R
n
by f(x) =
1
1−x
n+1
'x
1
, . . . , x
n
`. The map g : R
n
→ S
n
−p deﬁned by g(y) = 't(y)y
1
, . . . , t(y)y
n
, 1 −t(y)`
where t(y) = 2/(1+[[y[[
2
) is the inverse of f. Thus R
n
is homeomorphic to S
n
−p and the Alexandroﬀ one
point compactiﬁcation of R
n
is homeomorphic to the Alexandroﬀ onepoint compactiﬁcation of S
n
−p,
which is S
n
.
21b. Let X be the space in Q11 and Y be the space in Q12. Deﬁne f : X
∗
→ Y by f(x) = x for x ∈ X
and f(ω) = ω
1
. Then f is clearly a bijection. Consider the basic sets in Y . Now f
−1
[¦x : x < a¦] =
¦x : x < a¦, which is open in X and thus open in X
∗
. Similarly for sets of the form ¦x : a < x < b¦.
Also, f
−1
[¦x : a < x¦] = ¦x ∈ X : a < x¦ ∪ ¦ω¦, whose complement ¦x : x ≤ a¦ is compact. Thus
f
−1
[¦x : a < x¦] is open in X
∗
. Since f is a continuous bijection from the compact space X
∗
to the
Hausdorﬀ space Y , f is a homeomorphism. Hence the onepoint compactiﬁcation of X is Y .
22a. Let O be an open subset of a compact Hausdorﬀ space X. By Q8.23a, for any x ∈ O, there is an
open set U such that x ∈ U and
¯
U ⊂ O. Then the closure of U in X is the same as the closure of U in
O and
¯
U is compact, being closed in a compact space. Hence O is locally compact.
22b. Let O be an open subset of a compact Hausdorﬀ space X. Consider the mapping f of X to the
onepoint compactiﬁcation of O which is identity on O and takes each point in X ` O to ω. If U is open
in O, then f
−1
[U] = U is open in X. If K ⊂ O is compact, then f
−1
[O
∗
` K] = X ` K, which is open in
X. Hence f is continuous.
23. Let X and Y be locally compact Hausdorﬀ spaces, and f a continuous mapping of X into Y . Let
X
∗
and Y
∗
be the onepoint compactiﬁcations of X and Y , and f
∗
the mapping of X
∗
into Y
∗
whose
restriction to X is f and which takes the point at inﬁnity in X
∗
to the point at inﬁnity in Y
∗
. Suppose
54
f is proper. If U is open in Y , then (f
∗
)
−1
[U] = f
−1
[U], which is open in X and thus open in X
∗
. If
K ⊂ Y is compact, then (f
∗
)
−1
[Y
∗
` K] = f
−1
[Y ` K] ∪¦ω
X
¦ = X
∗
` f
−1
[K], which is open in X
∗
since
f
−1
[K] ⊂ X is compact. Hence f
∗
is continuous. Conversely, suppose f
∗
is continuous. Then f = f
∗
[
X
is continuous. Also, for any compact set K ⊂ Y , (f
∗
)
−1
[Y
∗
` K] = X
∗
` f
−1
[K] is open in X
∗
. Hence
f
−1
[K] ⊂ X is compact.
*24a. Let X be a locally compact Hausdorﬀ space. Suppose F is a closed subset of X. For each closed
compact set K, F ∩ K is closed. Conversely, suppose F is not closed. Take x ∈
¯
F ` F. There is an
open set O containing x with
¯
O compact. For any open set U containing x, (O ∩ U) ∩ F = ∅. Thus
U ∩ (F ∩
¯
O) = ∅. Then x ∈ F ∩
¯
O ` (F ∩
¯
O) so F ∩
¯
O is not closed.
*24b. Let X be a Hausdorﬀ space satisfying the ﬁrst axiom of countability. Suppose F is a closed subset
of X. For each closed compact set K, F∩K is closed. Conversely, suppose F is not closed. Take x ∈
¯
F`F.
Since X is ﬁrst countable, there is a sequence 'x
n
` in F converging to x. Let K = ¦x
n
: n ∈ N¦ ∪ ¦x¦.
Then K is compact in the Hausdorﬀ space X and thus closed. Now F ∩K = ¦x
n
: n ∈ N¦ is not closed.
25. Let F be a family of realvalued continuous functions on a locally compact Hausdorﬀ space X, and
suppose that F has the following properties: (i) If f, g ∈ F, then f +g ∈ F. (ii) If f, g ∈ F, then f/g ∈ F
provided that supportf ⊂ ¦x ∈ X : g(x) = 0¦. (iii) Given an open set O ⊂ X and x
0
∈ O, there is an
f ∈ F with f(x
0
) = 1, 0 ≤ f ≤ 1 and supportf ⊂ O.
Let ¦O
λ
¦ be an open covering of a compact subset K of a locally compact Hausdorﬀ space X. Let O be an
open set with K ⊂ O and
¯
O compact. For each x
0
∈ K, there is an f
x
0
∈ F with f
x
0
(x
0
) = 1, 0 ≤ f
x
0
≤ 1
and supportf
x
0
⊂ O∩O
λ
for some λ. For each x
0
∈
¯
O`K, there is a g
x
0
∈ F with g
x
0
(x
0
) = 1, 0 ≤ g
x
0
≤ 1
and supportg
x
0
⊂ K
c
. By compactness of
¯
O, we may choose a ﬁnite number f
1
, . . . , f
n
, g
1
, . . . , g
m
of
these functions such that the sets where they are positive cover
¯
O. Set f =
¸
n
i=1
f
i
and g =
¸
m
i=1
g
i
.
Then f, g ∈ F, f > 0 on K, supportf ⊂ O, f + g > 0 on
¯
O and g ≡ 0 on K. Thus f/(f + g) ∈ F is
continuous and ≡ 1 on K. The functions ϕ
i
= f
i
/(f + g) ∈ F, i = 1, . . . , n form a ﬁnite collection of
functions subordinate to the collection ¦O
λ
¦ and such that ϕ
1
+ ϕ
n
≡ 1 on K.
*26. Lemma: Let X be a locally compact Hausdorﬀ space and U be an open set containing x ∈ X.
Then there is an open set V containing x such that
¯
V is compact and
¯
V ⊂ U.
Proof: There is an open set O ⊂ X containing x such that
¯
O is compact. Then U ∩ O is open in
¯
O
and contains x. Thus there is an open set O
t
⊂
¯
O such that x ∈ O
t
and
¯
O
t
¯
O
⊂ U ∩ O. Note that
¯
O
t
¯
O
=
¯
O ∩
¯
O
t
=
¯
O
t
. Thus
¯
O
t
⊂ U ∩ O. Let V = O ∩ O
t
. Then x ∈ V . Furthermore, since O
t
is open in
¯
O, it is open in O and thus open in X so V is open in X. Also,
¯
V ⊂
¯
O
t
⊂ U ∩O ⊂ U. Since
¯
V is closed
in
¯
O,
¯
V is compact.
Let X be a locally compact Hausdorﬀ space and ¦O
n
¦ a countable collection of dense open sets. Given an
open set U, let x
1
be a point in O
1
∩U. Let V
1
be an open set containing x
1
such that V
1
is compact and
V
1
⊂ O
1
∩ U. Suppose x
1
, . . . , x
n
and V
1
, . . . , V
n
have been chosen. Let x
n+1
∈ O
n+1
∩ V
n
and let V
n+1
be an open set containing x
n+1
such that V
n+1
is compact and V
n+1
⊂ O
n+1
∩ V
n
. Then V
1
⊃ V
2
⊃
is a decreasing sequence of closed sets in the compact set V
1
. This collection of closed sets has the ﬁnite
intersection property so
¸
V
n
= ∅. Let y ∈
¸
V
n
. Then y ∈
¸
O
n
∩ U. Hence
¸
O
n
is dense in X.
(*) Assume that X is Hausdorﬀ.
*27. Let X be a locally compact Hausdorﬀ space and let O be an open subset contained in a countable
union
¸
F
n
of closed sets. Note that O is a locally compact Hausdorﬀ space (see Q29b). Also, O =
¸
(O ∩ F
n
), which is a union of sets closed in O. If
¸
(O ∩ F
n
)
◦
= ∅, then (O ∩ F
n
)
◦
= ∅ for all n (with
the interior taken in O) so O ` F
n
is dense and open in O for all n. By Q26,
¸
(O ` F
n
) is dense in O.
But
¸
(O` F
n
) = O` (
¸
F
n
)
c
= ∅. Contradiction. Hence
¸
(O∩F
n
)
◦
= ∅. Now O∩
¸
F
◦
n
⊃
¸
(O∩F
n
)
◦
so O ∩
¸
F
◦
n
= ∅ and
¸
F
◦
n
is an open set dense in O.
(*) Assume that X is Hausdorﬀ.
*28. Let Y be a dense subset of a Hausdorﬀ space X, and suppose that Y with its subspace topology is
locally compact. Given y ∈ Y , there is an open set U ⊂ Y containing y with
¯
U
Y
= Y ∩
¯
U compact. Since
X is Hausdorﬀ, Y ∩
¯
U is closed. Then since U ⊂ Y ∩
¯
U, we have
¯
U ⊂ Y ∩
¯
U ⊂ Y . Now U = V ∩Y for some
open set V ⊂ X. Note that since Y is dense,
¯
V = V ∩ Y . Then x ∈ V and V ⊂
¯
V = V ∩ Y =
¯
U ⊂ Y .
Hence Y is open in X.
29a. Suppose F is closed in a locally compact space X. Given x ∈ F, there is an open set O ⊂ X
55
containing x with
¯
O compact. Then O ∩ F is an open set in F containing y and O ∩ F
F
= F ∩ O ∩ F
is closed in
¯
O and thus compact. Hence F is locally compact.
*29b. Suppose O is open in a locally compact Hausdorﬀ space X. Take x ∈ O. By the lemma in Q26,
there is an open set U containing x such that
¯
U is compact and
¯
U ⊂ O. Now U ∩O is an open set in O
containing x with U ∩ O
O
= O ∩ U ∩ O = U ∩ O being compact since it is closed in the compact set
¯
U.
Hence O is locally compact.
29c. Suppose a subset Y of a locally compact Hausdorﬀ space X is locally compact in its subspace
topology. Then Y is dense in the Hausdorﬀ space
¯
Y . By Q28, Y is open in
¯
Y . Conversely, suppose Y is
open in
¯
Y . By part (a),
¯
Y is locally compact since
¯
Y is closed in X. By part (b), Y is locally compact
since Y is open in
¯
Y .
9.5 σcompact spaces
30. Let X be a locally compact Hausdorﬀ space. Suppose there is a sequence 'O
n
` of open sets with
O
n
compact, O
n
⊂ O
n+1
and X =
¸
O
n
. For each n, let ϕ
n
be a continuous realvalued function with
ϕ
n
≡ 1 on O
n−1
and supportϕ
n
⊂ O
n
. Deﬁne ϕ : X → [0, ∞) by ϕ =
¸
(1 −ϕ
n
).
Given y ∈ X and ε > 0, y ∈ O
N
for some N and thus y ∈ O
n
for n ≥ N. There exists N
t
such that
ϕ(y) −
¸
N
n=1
(1 − ϕ
n
(y)) < ε/2. Let N
tt
= max(N, N
t
). Take x ∈ O
N
. Then ϕ
N
+1
(x) = 1. In fact,
ϕ
n
(x) = 1 for n ≥ N
tt
+1 so ϕ(x) =
¸
N
n=1
(1 −ϕ
n
(x)). For each n = 1, . . . , N
tt
, there is an open set U
n
containing y such that [ϕ
n
(y)−ϕ
n
(x)[ < ε/2N
tt
for x ∈ U
n
. Let U =
¸
N
n=1
U
n
∩O
N
. Then U is an open
set containing y and for x ∈ U, [ϕ(y) −ϕ(x)[ ≤ [ϕ(y) −
¸
N
n=1
(1 −ϕ
n
(y))] +[
¸
N
n=1
(1 −ϕ
n
(y)) −ϕ(x)[ ≤
[ϕ(y) −
¸
N
n=1
(1 −ϕ
n
(y))] +
¸
N
n=1
[ϕ
n
(y) −ϕ
n
(x)[ < ε. Hence ϕ is continuous.
To show that ϕ is proper, we consider closed bounded intervals in [0, ∞). By considering O
N
−1
and
open sets V
n
with V
n
compact and V
n
⊂ U
n
, we then apply a similar argument as above.
31a. Let (X, ρ) be a proper locally compact metric space. If K is a compact subset, then K is closed
and bounded by Proposition 7.22. Conversely, suppose a subset K is closed and bounded. Since X is
proper, the closed balls ¦x : ρ(x, x
0
) ≤ a¦ are compact for some x
0
and all a ∈ (0, ∞). Since K is
bounded, there exist x
1
and b such that ρ(x, x
1
) ≤ b for all x ∈ K. Then ρ(x, x
0
) ≤ b +ρ(x
1
, x
0
) for all
x ∈ K. Thus K is a closed subset of the compact set ¦x : ρ(x, x
0
) ≤ b +ρ(x
1
, x
0
)¦ so K is compact.
31b. Let (X, ρ) be a proper locally compact metric space. Then the closed balls ¦x : ρ(x, x
0
) ≤ a¦ are
compact for some x
0
and all x ∈ (0, ∞). Note that a compact subset K ⊂ [0, ∞) is closed and bounded.
Also, the function f(x) = ρ(x, x
0
) is continuous from X to [0, ∞). Now f
−1
[K] is bounded since K is
bounded and closed since f is continuous and K is closed. Thus by part (a), f
−1
[K] is compact. Hence
f : X → [0, ∞) is a proper continuous map and X is σcompact.
31c. Let (X, ρ) be a σcompact and locally compact metric space. Then there is a proper continuous
map ϕ : X → [0, ∞). Deﬁne ρ
∗
(x, y) = ρ(x, y) +[ϕ(x) −ϕ(y)[. Then ρ
∗
is a metric on X. Given x ∈ X
and ε > 0, there exists δ
t
> 0 such that ρ(x, y) < δ
t
implies [ϕ(x)−ϕ(y)[ < ε/2. Choose δ < min(δ
t
, ε/2).
When ρ(x, y) < δ, ρ
∗
(x, y) < ε/2 + ε/2 = ε and when ρ
∗
(x, y) < δ, ρ(x, y) ≤ ρ
∗
(x, y) < ε. Thus ρ and
ρ
∗
are equivalent metrics. For a ﬁxed x
0
∈ X and any a ∈ (0, ∞), ¦x : ρ
∗
(x, x
0
) ≤ a¦ ⊂ ¦x : ϕ(x) ≤
a +ϕ(x
0
)¦, which is compact. Thus ¦x : ρ(x, x
0
) ≤ a¦ is compact. Hence ρ
∗
is a proper metric.
9.6 Paracompact spaces
32. Let ¦E
λ
¦ be a locally ﬁnite collection of subsets of a topological space X, and set E =
¸
E
λ
. Since
E
λ
⊂ E for all λ, E
λ
⊂
¯
E for all λ. Thus
¸
E
λ
⊂
¯
E. If x ∈
¯
E, then there is an open set containing x
that meets only a ﬁnite number of sets E
λ
1
, . . . , E
λ
n
so x ∈
¸
n
i=1
E
λ
i
⊂
¸
E
λ
. Thus
¯
E ⊂
¸
E
λ
. Hence
¯
E =
¸
E
λ
.
(*) Proof of Lemma 22.
33. Let ¦E
λ
¦ be a locally ﬁnite collection of subsets of X and K a compact subset of X. For each
x ∈ X, there is an open set O
x
containing x that meets only a ﬁnite number of sets in ¦E
λ
¦. Now
K ⊂
¸
x∈K
O
x
so K ⊂
¸
n
i=1
O
x
i
for some x
1
, . . . , x
n
∈ K. Since each O
x
i
meets only a ﬁnite number of
sets in ¦E
λ
¦, so does K.
56
(*) Proof of Lemma 23.
34a. Let X be a paracompact Hausdorﬀ space. Let F be a closed subset and let x / ∈ F. For each y ∈ F,
there are disjoint open sets U
y
and V
y
with x ∈ U
y
and y ∈ V
y
. Now X ` F ∪ ¦V
y
: y ∈ F¦ is an open
cover for X so it has a locally ﬁnite open reﬁnement ¦E
λ
¦. Let E =
¸
¦E
λ
: E
λ
∩F = ∅¦. Then E is an
open set containing F. Also, there is an open set U containing x that meets only a ﬁnite number of sets
E
λ
1
, . . . , E
λ
n
in ¦E
λ
¦. Each of these sets must lie in some V
y
i
where y
i
∈ F. Consider O = U ∩
¸
n
i=1
U
y
i
.
Then O is an open set containing x and O ∩ E = ∅. Hence X is regular.
34b. Let X be a paracompact Hausdorﬀ space. Let F and G be disjoint closed subsets. By part (a), for
each y ∈ G, there are disjoint open sets U
y
and V
y
such that F ⊂ U
y
and y ∈ V
y
. Now X`G∪¦V
y
: y ∈ G¦
is an open cover for X so it has a locally ﬁnite open reﬁnement ¦E
λ
¦. Let E =
¸
¦E
λ
: E
λ
∩ F = ∅¦.
Then E is an open set containing F. For each y ∈ G, there is an open set O
y
that meets only a ﬁnite
number of sets E
λ
1
, . . . , E
λ
n
in ¦E
λ
¦. Each of these sets must lie in some V
y
i
where y
i
∈ G. Let
O
t
y
= O
y
∩
¸
n
i=1
U
y
i
. Then O
t
y
is an open set containing y and O
t
y
∩ E = ∅. Let O =
¸
y∈G
O
t
y
. Then O
is an open set containing G and O ∩ E = ∅.
35. Let (X, ρ) be a locally compact metrizable space. Suppose it can be metrized by a proper extended
metric ρ
∗
. By Q8.41, X is the direct union of its parts X
α
. Now (X
α
, ρ
∗
[
X
α
) is a proper locally compact
metric space for each α so by Q31b, each X
α
is σcompact. Hence X is the direct union of σcompact
spaces so it is paracompact. Conversely, suppose X is paracompact. Then X is the direct union of
σcompact spaces X
β
. Each (X
β
, ρ[
X
β
) is a σcompact and locally compact metric space so by Q31c,
each X
β
can be metrized by a proper metric ρ
β
. Now deﬁne ρ
∗∗
(x, y) = ρ
β
(x, y) if x, y ∈ X
β
and
ρ
∗∗
(x, y) = ∞ if x ∈ X
β
1
and y ∈ X
β
2
with β
1
= β
2
. Then ρ
∗∗
is a proper extended metric on X.
9.7 Manifolds
36. Let X = (−1, 1) ∪ [2, 3), and make X into a topological space by taking as a base all open intervals
(a, b) ⊂ X and all sets of the form (−ε, 0) ∪[2, 2 +ε) for 0 < ε < 1. Clearly all open intervals (a, b) ⊂ X
are open balls in R. Also, sets of the form (−ε, 0) ∪ [2, 2 +ε) are homeomorphic to (−ε, ε). Hence X is
locally Euclidean. There are no disjoint basic sets that separate 0 and 2 so X is not Hausdorﬀ.
37a. A not necessarily connected manifold X is the disjoint union of its components. By Q8.36, since
X is locally connected, each component of X is open. Thus X is the direct union of its components.
Also, its components are Hausdorﬀ and locally Euclidean. Hence X is the direct union of (connected)
manifolds.
37b. For a not necessarily connected manifold, statements (ii), (iii), (v), (vi), (vii) are equivalent. Also,
statements (i) and (iv) are equivalent. The ﬁrst set of statements imply the second set of statements.
*38a.
9.8 The Stone
˘
Cech compactiﬁcation
39a. Let f be a bounded continuous realvalued function on X with [f[ ≤ 1. The restriction to X of
the projection π
f
on β(X) is f and π
f
is continuous on β(X) since β(X) is a subspace of the product
space I
F
.
*39b. Suppose X is a dense open subset of a compact Hausdorﬀ space Y . Then Y is a subset of I
G
where G is the space of continuous g on Y with [g[ ≤ 1. The inclusion i : X → Y induces a continuous
function F : I
F
→ I
G
as follows. If g ∈ G, then g◦i ∈ F. so deﬁne F('t
f
`) = '(t
g◦i
)
g
`. Since π
g
◦F = π
g◦i
is continuous for each g, F is continuous. Now F[β(X)] ⊂ F[E] ⊂ Y . Let ϕ = F[
β(X)
. Then ϕ is a
continuous mapping of β(X) onto Y with ϕ(x) = x for all x ∈ X. Also, ϕ is unique as a map to a
Hausdorﬀ space is determined by its values on a dense subset.
39c. Suppose Z is a space with the same properties. By part (b), there is a unique continuous mapping
ψ of β(X) onto Z with ψ(x) = x for all x ∈ X. Also, there is a unique continuous mapping ϕ of Z onto
β(X) with ϕ(x) = x for all x ∈ X. Thus ϕ ◦ ψ = id[
X
so ϕ and ψ are homeomorphisms.
*40. Let X be the set of ordinals less than the ﬁrst uncountable ordinal and let Y be the set of ordinals
less than or equal to the ﬁrst uncountable ordinal. Then X is dense in the compact Hausdorﬀ space Y .
By Q11, every continuous realvalued function on X is eventually constant so it extends to a continuous
57
function on Y .
*41. If A ⊂ N, deﬁne f : N → [0, 1] by f(x) = 0 if x ∈ A and f(x) = 1 if x / ∈ A. Now f is continuous
so it extends to a continuous function
ˆ
f on β(N). Then
¯
A∪ N ` A = β(N). It follows that
ˆ
f
−1
[¦1¦] =
¯
A
and
ˆ
f
−1
[¦0¦] = N ` A. Thus
¯
A∩ N ` A = ∅ and
¯
A is open.
If B ⊂ N and A∩ B = ∅, then B ⊂ N ` A and
¯
A∩
¯
B = ∅.
If V is an open subset of β(N), then V ∩ N is an open subset of β(N). Now if x ∈
¯
V and W is an open
neighbourhood of x, then W ∩ V ∩ N = ∅ so x ∈ V ∩ N. Thus
¯
V = V ∩ N and
¯
V is open (β(N) is
extremally disconnected).
Let Y be a subset of β(N) with at least two distinct points x and y. Since β(N) is Hausdorﬀ, there is an
open set U ⊂ β(N) such that x ∈ U and y / ∈
¯
U. Then Y = (Y ∩
¯
U) ∪ (Y `
¯
U) is a separation of Y so Y
is not connected. Hence β(N) is totally disconnected.
Clearly, if a sequence in N converges in N, then it converges in β(N). Conversely, if it converges in
β(N) ` N, then consider a function f : N → [0, 1] with f(x
2n
) = 0 and f(x
2n+1
) = 1 for all n. Now f is
continuous since N is discrete so it has a continuous extension to β(N). This is a contradiction as the
sequence 'g(x
n
)` does not converge.
Hence β(N) is compact but not sequentially compact as the sequence x
n
= n does not have a convergent
subsequence.
9.9 The StoneWeierstrass Theorem
42. Let A be the set of ﬁnite Fourier series ϕ given by ϕ(x) = a
0
+
¸
N
n=1
(a
n
cos nx +b
n
sin nx), N ∈ N.
Then A is a linear space of functions in C(X) where X is taken to be the unit circle in C. From the
trigonometric identities cos mxcos nx =
1
2
[cos(m−n)x +cos(m+n)x], sin mxcos nx =
1
2
[sin(m+n)x +
sin(m−n)x] and sin mxsin nx =
1
2
[cos(m−n)x −cos(m+n)x], we see that A is a subalgebra of C(X).
Furthermore A separates the points of X and contains the constant functions. By the StoneWeierstrass
Theorem, given any continuous periodic realvalued function f on R with period 2π and any ε > 0, there
is a ﬁnite Fourier series ϕ such that [ϕ(x) −f(x)[ < ε for all x.
*43. Let A be an algebra of continuous realvalued functions on a compact space X, and assume that
A separates the points of X. If for each x ∈ X there is an f
x
∈ A with f
x
(x) = 0, then by continuity,
there is an open neighbourhood O
x
of x such that f
x
(y) = 0 for y ∈ O
x
. The sets ¦O
x
¦ cover X so by
compactness, ﬁnitely many of them cover X, say ¦O
x
1
, . . . , O
x
n
¦. Let g = f
2
x
1
+ + f
2
x
n
. Then g ∈ A
and g = 0 everywhere. The closure of the range of g is a compact set K not containing 0. The function h
given by h(t) = 1/t for t ∈ K and h(0) = 0 is continuous on K∪¦0¦ so it can be uniformly approximated
by polynomials h
n
so that h
n
◦ g ∈ A and h
n
◦ g → 1/g. Note that if h
n
is uniformly within ε/2 of h,
then [h
n
(0)[ < ε/2 but h
n
(0) is the constant term of h
n
so subtracting the constant term results in a
polynomial still within ε of h. Thus we may assume that the polynomials h
n
have no constant term.
Thus 1/g ∈
¯
A so 1 ∈
¯
A and
¯
A contains the constant functions. Hence
¯
A = C(X).
44. Let F be a family of continuous realvalued functions on a compact Hausdorﬀ space X, and suppose
that F separates the points of X. Let A be the set of polynomials in a ﬁnite number of functions of F.
Then A is a subalgebra of C(X) that separates the points of X and contains the constant functions. By
the StoneWeierstrass Theorem, A is dense in C(X). Hence every continuous realvalued function on X
can be uniformly approximated by a polynomial in a ﬁnite number of functions of F.
45a. Let X be a topological space and A a set of realvalued functions on X. Deﬁne x ≡ y if f(x) = f(y)
for all f ∈ A. Clearly, x ≡ x for all x ∈ X and y ≡ x if x ≡ y. If x ≡ y and y ≡ z, then f(x) = f(y) = f(z)
for all f ∈ A so x ≡ z. Hence ≡ is an equivalence relation.
45b. Let
¯
X be the set of equivalence classes of ≡ and ϕ the natural map of X into
¯
X. Given f ∈ A,
deﬁne
˜
f on
¯
X by
˜
f(˜ x) = f(x). If ˜ x = ˜ y, then x ≡ y so f(x) = f(y) and
˜
f(˜ x) =
˜
f(˜ y). Thus
˜
f is
welldeﬁned and it is the unique function such that f =
˜
f ◦ ϕ.
45c. Let
¯
X have the weak topology generated by the functions
˜
f in part (b). Consider a basic set
˜
f
−1
[O]. Now ϕ
−1
[
˜
f
−1
[O]] = ¦x :
˜
f(ϕ(x)) ∈ O¦ = ¦x : f(x) ∈ O¦ = f
−1
[O], which is open since f is
continuous. Hence ϕ is continuous.
45d. Since ϕ is continuous and maps X onto
¯
X, if X is compact, then so is
¯
X. By deﬁnition of the
58
weak topology on
¯
X, the functions
˜
f are continuous.
*45e. Let X be a compact space and A a closed subalgebra of C(X) containing the constant functions.
Deﬁne
¯
X and ϕ as above. If ˜ x and ˜ y are distinct points in
¯
X, then f(x) = f(y) for some f ∈ A so
there are disjoint open sets O
x
and O
y
in R with f(x) ∈ O
x
and f(y) ∈ O
y
. Then ˜ x ∈
˜
f
−1
[O
x
] and
˜ y ∈
˜
f
−1
[O
y
]. Thus
¯
X is a compact Hausdorﬀ space. Furthermore, ϕ induces a (continuous) mapping
of ϕ
∗
: A → C(
¯
X), f →
˜
f, where f =
˜
f ◦ ϕ. The image of A in C(
¯
X) is a subalgebra containing the
constant functions and separating the points of
¯
X. Suppose ' ˜ g
n
` is a sequence in ϕ
∗
[A] that converges
to ˜ g in C(
¯
X). Then the sequence 'g
n
`, where g
n
= ˜ g
n
◦ϕ, converges to ˜ g ◦ϕ. Since each g
n
∈ A and A is
closed, ˜ g ◦ ϕ ∈ A so ˜ g ∈ ϕ
∗
[A] by uniqueness. Thus ϕ
∗
[A] is closed. By the StoneWeierstrass Theorem,
the image of A is C(
¯
X). Hence A is the set of all functions of the form
˜
f ◦ ϕ with
˜
f ∈ C(
¯
X).
46. Let X and Y be compact spaces. The set of ﬁnite sums of functions of the form f(x)g(y) where
f ∈ C(X) and g ∈ C(Y ) is an algebra of continuous realvalued functions on X Y that contains the
constant functions and separates points in X Y . By the StoneWeierstrass Theorem, this set is dense
in C(X Y ). Thus for each continuous realvalued function f on X Y and each ε > 0, there exist
continuous functions g
1
, . . . , g
n
on X and h
1
, . . . , h
n
on Y such that [f(x, y) −
¸
n
i=1
g
i
(x)h
i
(y)[ < ε for
all 'x, y` ∈ X Y .
47. The functions of norm 1 in the algebra A give a mapping of X into the inﬁnitedimensional cube
¸
¦I
f
: f ∈ A, [[f[[ = 1¦. By the Tietze Extension Theorem, each continuous function f on the image
of X can be extended to a continuous function g on the cube and by Q17, g can be approximated by a
continuous function h of only a ﬁnite number of coordinates. Then h can be regarded as a continuous
function on a cube in R
n
, which can be uniformly approximated by a polynomial in (a ﬁnite number of)
the coordinate functions.
48a. Let ϕ be the polynomial deﬁned by ϕ(x) = x + x(1 −2x)(1 −x). Then ϕ
t
(x) = 6x
2
−6x + 2 > 0
for all x. Thus ϕ is monotone increasing and its ﬁxed points are 0,
1
2
, 1.
*48b. Choose ε > 0. Note that ϕ(x) > x on (0,
1
2
) and ϕ(x) < x on (
1
2
, 1). Let [a
n
, b
n
] = ϕ
n
[ε, 1 −ε] for
each n where ϕ
n
is an iterate of ϕ. Then a
n
= ϕ
n
(ε) increases to some a and b
n
= ϕ
n
(1 −ε) decreases
to some b. Furthermore, ε ≤ a ≤ b ≤ 1 −ε and a, b are ﬁxed points of ϕ. Thus a = b =
1
2
. Hence some
iterate ϕ
n
is a polynomial with integral coeﬃcients that is monotone increasing on [0, 1] and such that
[ϕ
n
(x) −
1
2
[ < ε for x ∈ [ε, 1 −ε].
*48c. Given α with 0 < α < 1 and any ε > 0, it suﬃces to consider the case where α is a rational
number
a
b
. Deﬁne ϕ(x) = x +x(a −bx)(1 −x). Then α is a ﬁxed point of ϕ. By parts (a) and (b), some
iterate ψ = ϕ
n
is a polynomial with integral coeﬃcients (and no constant term) such that 0 ≤ ψ(x) ≤ 1
in [0, 1] and [ψ(x) −α[ < ε for all x ∈ [ε, 1 −ε].
*48d. Let P be a polynomial with integral coeﬃcients, and suppose that P(−1) = P(0) = P(1) = 0.
Let β be any real number. We may assume that 0 < β < 1. For any ε > 0, there exists δ > 0 such that
[P(x)[ < ε/2 for x ∈ (−δ, δ), x ∈ (1 − δ, 1] and x ∈ [−1, −1 + δ). We may assume that δ < ε/[[P[[. By
part (c), there is a polynomial ψ with integral coeﬃcients and no constant term such that [ψ(x
2
)−β[ < δ
for all x ∈ [δ, 1 −δ]. Then [P(x)ψ(x
2
) −βP(x)[ < δ[[P[[ < ε for all x ∈ [−1, 1].
*48e. Let I = [−1, 1] and f a continuous realvalued function on I such that f(−1), f(0), f(1) are
integers and f(1) ≡ f(−1) mod 2. Let f(−1) = a, f(0) = b, f(1) = a+2c, where a, b, c are integers. Let
Q(x) = (a − b + c)x
2
+ cx + b. Replacing f by f − Q, we may assume that a = b = c = 0. Then use
the StoneWeierstrass Theorem to approximate f by a polynomial R with rational coeﬃcients such that
R(−1) = R(0) = R(1) = 0. Let N be the least common multiple of the denominators of the coeﬃcients
of R so that NR has integral coeﬃcients and vanishes at 1,0,1. Let β = 1/N and apply part (d) to the
polynomial NR so that R, and thus f, can be approximated by a polynomial P with integral coeﬃcients.
*49a.
*49b.
*50a.
*50b.
59
10 Banach Spaces
10.1 Introduction
1. Suppose x
n
→ x. Then [ [[x
n
[[ −[[x[[ [ ≤ [[x
n
−x[[ → 0. Hence [[x
n
[[ → [[x[[.
2. The metric ρ(x, y) = [
¸
n
i=1
(x
i
− y
i
)
2
]
1/2
is derived from the norm [[x[[ = (
¸
n
i=1
x
2
i
)
1/2
. The metric
ρ
∗
(x, y) =
¸
n
i=1
[x
i
−y
i
[ is derived from the norm [[x[[
∗
=
¸
n
i=1
[x
i
[. The metric ρ
+
(x, y) = max [x
i
−y
i
[
is derived from the norm [[x[[
+
= max [x
i
[. Now n
−1
[[x[[
+
≤ [[x[[
+
≤ [[x[[
∗
≤ n[[x[[
+
so [[x[[
∗
and [[x[[
+
are equivalent. Also, (
√
n)
−1
[[x[[
+
≤ [[x[[
+
≤ [[x[[ ≤ (n([[x[[
+
)
2
)
1/2
=
√
n[[x[[
+
so [[x[[ and [[x[[
+
are
equivalent. Thus [[x[[ and [[x[[
∗
are also equivalent.
3. Consider + as a function from XX into X. Since [[(x
1
+y
1
)−(x
2
+y
2
)[[ ≤ [[x
1
−x
2
[[ +[[y
1
−y
2
[[, +
is continuous. Consider as a function from RX into X. Since [[cx−cy[[ = [c[ [[x−y[[, is continuous.
4. Let M be a nonempty set. Then M ⊂ M +M since m = m+θ for all m, where θ is the zero vector.
Also, M ⊂ λM since m = 1 m for all m. If M is a linear manifold, then M +M ⊂ M and λM ⊂ M for
each λ so M +M = M and λM = M. Conversely, suppose M +M = M and λM = M. Then λx ∈ M
for each λ ∈ R and x ∈ M. Thus also λ
1
x
1
+ λ
2
x
2
∈ M for λ
1
, λ
2
∈ R and x
1
, x
2
∈ M. Hence M is a
linear manifold.
5a. Let ¦M
i
: i ∈ I¦ be a family of linear manifolds and let M =
¸
M
i
. For any λ
1
, λ
2
∈ R and
x
1
, x
2
∈ M, we have x
1
, x
2
∈ M
i
for all i. Since each M
i
is a linear manifold, λ
1
x
1
+λ
2
x
2
∈ M
i
for each
i. i.e. λ
1
x
1
+λ
2
x
2
∈ M. Hence M is a linear manifold.
5b. Given a set A in a vector space X, X is a linear manifold containing A. Consider the family of
linear manifolds containing A. The intersection ¦A¦ of this family is a linear manifold containing A and
it is the smallest such linear manifold.
5c. Consider the set M of all ﬁnite linear combinations of the form λ
1
x
1
+ +λ
n
x
n
with x
i
∈ A. Then
M is a linear manifold containing A. Also, any linear manifold containing A will contain M. Hence M
is the smallest linear manifold containing A. i.e. ¦A¦ = M.
6a. Let M and N be linear manifolds. For λ
1
, λ
2
∈ R, m
1
, m
2
∈ M and n
1
, n
2
∈ N, λ
1
(m
1
+ n
1
) +
λ
2
(m
2
+n
2
) = (λ
1
m
1
+λ
2
m
2
) +(λ
1
n
1
+λ
2
n
2
) ∈ M +N. Hence M +N is a linear manifold. Note that
M ⊂ M + N and N ⊂ M + N so M + N contains M ∪ N. Also, any manifold containing M ∪ N will
also contain M +N. Hence M +N = ¦M ∪ N¦.
6b. Let M be a linear manifold. For λ
1
, λ
2
∈ R, x
1
, x
2
∈
¯
M and δ > 0, there exist y
1
, y
2
∈ M such that
[[x
1
−y
1
[[ < δ/2λ
1
and [[x
2
−y
2
[[ < δ/2λ
2
(We may assume λ
1
, λ
2
= 0). Then [[(λ
1
x
1
+λ
2
x
2
) −(λ
1
y
1
+
λ
2
y
2
)[[ ≤ [λ
1
[ [[x
1
−y
1
[[ +[λ
2
[ [[x
2
−y
2
[[ < δ. Thus λ
1
x
1
+λ
2
x
2
∈
¯
M so
¯
M is a linear manifold.
7. Let P be the set of all polynomials on [0, 1]. Then P ⊂ C[0, 1]. For λ
1
, λ
2
∈ R and p
1
, p
2
∈ P,
λ
1
p
1
+ λ
2
p
2
is still a polynomial on [0, 1] so λ
1
p
1
+ λ
2
p
2
∈ P. Thus P is a linear manifold in C[0, 1].
The set P is not closed in C[0, 1] because by the Weierstrass Approximation Theorem, every continuous
function on [0, 1] can be uniformly approximated by polynomials on [0, 1]. i.e.
¯
P contains a continuous
function that is not a polynomial.
The set of continuous functions f with f(0) = 0 is a closed linear manifold in C[0, 1].
8. Let M be a ﬁnitedimensional linear manifold in a normed vector space X with M = ¦x
1
, . . . , x
n
¦.
Each x ∈ X can be written as a unique linear combination λ
1
x
1
+ + λ
n
x
n
. We may deﬁne a norm
[[x[[
1
=
¸
n
i=1
[λ
i
[ and see that [[ [[
1
is equivalent to the original norm on X. Thus convergence under the
original norm on X is equivalent to convergence of each sequence of coeﬃcients in R. Let '
¸
n
i=1
λ
(k)
i
x
i
`
k
be a sequence in M converging to x ∈ X. Let λ
i
= lim
k
λ
(k)
i
for each i. By continuity of addition and
scalar multiplication, x = lim
k
¸
n
i=1
λ
(k)
i
x
i
=
¸
n
i=1
λ
i
x
i
∈ M. Hence M is closed.
9. Let S = ¦x : [[x[[ < 1¦. Given x ∈ S, let δ = (1 − [[x[[)/2. When [[y − x[[ < δ, we have
[[y[[ ≤ [[y − x[[ + [[x[[ < (1 − [[x[[)/2 + [[x[[ < 1 so y ∈ S. Hence S is open. For any sequence 'x
n
` in
S that converges to some x, we have [[x
n
[[ → [[x[[ so [[x[[ ≤ 1. Thus
¯
S ⊂ ¦x : [[x[[ ≤ 1¦. On the other
hand, if [[x[[ = 1 and δ > 0, let λ = max(1 −δ/2, 0). Then [[λx[[ ∈ S and [[x −λx[[ = [1 −λ[ < δ. Thus
x ∈
¯
S and ¦x : [[x[[ ≤ 1¦ ⊂
¯
S. Hence
¯
S = ¦x : [[x[[ ≤ 1¦.
10. Deﬁne x ≡ y if [[x − y[[ = 0. Then x ≡ x since [[0x[[ = 0[[x[[ = 0. Also, x ≡ y implies
y ≡ x since [[y − x[[ = [[ − (x − y)[[ = [ − 1[ [[x − y[[ = [[x − y[[. Finally, if x ≡ y and y ≡ z, then
60
[[x −z[[ ≤ [[x −y[[ +[[y −z[[ = 0 so x ≡ z. Thus ≡ is an equivalence relation. If x
1
≡ y
1
and x
2
≡ y
2
,
then [[(x
1
+ x
2
) − (y
1
+ y
2
)[[ ≤ [[x
1
− y
1
[[ + [[x
2
− y
2
[[ = 0 so x
1
+ x
2
≡ y
1
+ y
2
. If x ≡ y, then
[[cx − cy[[ = [c[ [[x − y[[ = 0 so cx ≡ cy for c ∈ R. Hence ≡ is compatible with addition and scalar
multiplication. If x ≡ y, then [ [[x[[ −[[y[[ [ ≤ [[x −y[[ = 0 so [[x[[ = [[y[[.
Let X
t
be the set of equivalence classes under ≡. Deﬁne αx
t
+βy
t
as the (unique) equivalence class which
contains αx + βy for x ∈ x
t
and y ∈ y
t
and deﬁne [[x
t
[[ = [[x[[ for x ∈ x
t
. Then X
t
becomes a normed
vector space. The mapping ϕ of X onto X
t
that takes each element of X into the equivalence class to
which it belongs is a homomorphism of X onto X
t
since ϕ(αx +βy) = αx
t
+βy
t
= αϕ(x) +βϕ(y). The
kernel of ϕ consists of the elements of X that belong to the equivalence class containing the zero vector
θ. These are the elements x with [[x[[ = 0.
On the L
p
spaces on [0, 1] we have the pseudonorm [[f[[
p
=
1
0
[f[
p
1/2
. Then f ≡ g if
1
0
[f −g[
p
= 0.
i.e. f = g a.e. The kernel of the mapping ϕ consists of the functions that are 0 a.e.
11. Let X be a normed linear space (with norm [[ [[) and M a linear manifold in X. Let [[x[[
1
=
inf
m∈M
[[x − m[[. Since [[x − m[[ ≥ 0 for all x ∈ X and m ∈ M, [[x[[
1
≥ 0 for all x ∈ X. For x, y ∈ X
and ε > 0, there exist m, n ∈ M such that [[x − m[[ < [[x[[
1
+ ε/2 and [[y − m[[ < [[y[[
1
+ ε/2. Then
[[x + y[[
1
≤ [[(x + y) −(m + n)[[ < [[x[[
1
+[[y[[
1
+ ε. Since ε > 0 is arbitrary, [[x + y[[
1
≤ [[x[[
1
+[[y[[
1
.
Also, for x ∈ X and α ∈ R, [[αx[[
1
= inf
m∈M
[[αx −m[[ = [α[ inf
m∈M
[[x −m[[ = [α[ [[x[[
1
. Hence [[ [[
1
is a pseudonorm on X.
Let X
t
be the normed linear space derived from X and the pseudonorm [[ [[
1
using the process in
Q10. The natural mapping ϕ of X onto X
t
has kernel
¯
M since it consists of the elements x with
[[x[[
1
= inf
m∈M
[[x − m[[ = 0. Let O be an open set in X. Take x ∈ O. Then there exists δ > 0 such
that y ∈ O if [[y −x[[
1
< δ. Now if [[z −ϕ(x)[[ < δ, where z = ϕ(y) for some y ∈ X, then [[y −x[[
1
< δ
so y ∈ O and z ∈ ϕ[O]. Hence ϕ[O] is open. i.e. ϕ maps open sets into open sets.
12. Suppose X is complete and M is a closed linear manifold in X. Let 'ϕ(x
n
)` be an absolutely
summable sequence in X/M. Then
¸
[[ϕ(x
n
)[[ < ∞ so
¸
[[x
n
[[
1
< ∞. Given ε > 0, for each n, there
exists m
n
∈ M such that [[x
n
−m
n
[[ < [[x
n
[[
1
+2
−n
. Then
¸
[[x
n
−m
n
[[ ≤
¸
[[x
n
[[
1
+1 < ∞. Since X
is complete, the sequence 'x
n
−m
n
` is summable in X, say
¸
(x
n
−m
n
) = x. Now ϕ is continuous since
[[ϕ(x)[[ = [[x[[
1
≤ [[x[[. Also, M is the kernel of ϕ. Thus
¸
ϕ(x
n
) =
¸
ϕ(x
n
−m
n
) = ϕ(
¸
(x
n
−m
n
)) =
ϕ(x) ∈ X/M. Since any absolutely summable sequence in X/M is summable, X/M is complete.
10.2 Linear operators
13. Suppose A
n
→ A and x
n
→ x. Then [[A
n
− A[[ → 0 and [[x
n
− x[[ → 0. Since [[A
n
x
n
− Ax[[ ≤
[[A
n
x
n
−Ax
n
[[ +[[Ax
n
−Ax[[ ≤ [[A
n
−A[[ [[x
n
[[ +[[A[[ [[x
n
−x[[ and [[x
n
[[ is bounded, [[A
n
x
n
−Ax[[ → 0.
i.e. A
n
x
n
→ Ax.
14. Let A be a linear operator and ker A = ¦x : Ax = θ¦. If x, y ∈ ker A and α, β ∈ R, then
A(αx +βy) = αAx +βAy = θ so αx +βy ∈ ker A. Thus ker A is a linear manifold.
Suppose A is continuous. Let 'x
n
` be a sequence in ker A converging to some x. Since A is continuous,
'Ax
n
` converges to Ax. Now Ax
n
= θ for all n so Ax = θ and x ∈ ker A. Thus ker A is closed.
15a. Let X be a normed linear space and M a closed linear manifold. Let ϕ be the natural homomor
phism of X onto X/M. Now [[ϕ(x)[[ = [[x[[
1
≤ [[x[[, where [[[[
1
is the pseudonorm in Q11. Thus [[ϕ[[ ≤ 1.
Given ε > 0 and x ∈ X, there exists m ∈ M such that [[x−m[[ < [[x[[
1
+ε = [[ϕ(x)[[+ε = [[ϕ(x−m)[[+ε.
Let y = (x−m)/[[x−m[[. Then 1 = [[y[[ < [[ϕ(y)[[+ε. i.e. [[ϕ(y)[[ > 1−ε. Since [[ϕ[[ = sup
]]x]]=1
[[ϕ(x)[[,
we have [[ϕ[[ > 1 −ε for all ε > 0. Thus [[ϕ[[ ≥ 1.
15b. Let X and Y be normed linear spaces and A a bounded linear operator from X into Y whose kernel
is M. Deﬁne a mapping B from X/M into Y by Bx
t
= Ax where x
t
is the equivalence class containing
x. If x
t
= y
t
, then [[x − y[[
1
= 0. Thus for any ε > 0, there exists m ∈ M such that [[x − y − m[[ < ε.
Then [[Ax − Ay[[ = [[A(x − y − m)[[ ≤ [[A[[ε. Since ε > 0 is arbitrary, [[Ax − Ay[[ = 0. i.e. Ax = Ay.
Thus B is welldeﬁned and A = B ◦ ϕ. Furthermore, it is the unique such mapping. If x
t
, y
t
∈ X/M
and α, β ∈ R, then B(αx
t
+βy
t
) = A(αx +βy) = αAx +βAy = αBx
t
+βBy
t
so B is a linear operator.
Also, [[Bx
t
[[ = [[Ax[[ ≤ [[A[[ [[x[[ = [[A[[ [[x − m[[ for all m ∈ M. Thus [[Bx
t
[[ ≤ [[A[[ [[x[[
1
= [[A[[ [[x
t
[[
so [[B[[ ≤ [[A[[ and B is bounded. For any ε > 0, there exists x ∈ X with [[x[[ = 1 and [[Ax[[ > [[A[[ −ε.
61
Then [[x
t
[[ ≤ 1 and [[Bx
t
[[ > [[A[[ − ε. Since [[B[[ = sup
]]x
]]≤1
[[Bx
t
[[, we have [[B[[ ≥ [[A[[. Hence
[[A[[ = [[B[[.
16. Let X be a metric space and Y the space of realvalued functions f on X vanishing at a ﬁxed
point x
0
∈ X and satisfying [f(x) − f(y)[ ≤ Mρ(x, y) for some M (depending on f). Deﬁne [[f[[ =
sup
]f(x)−f(y)]
ρ(x,y)
. Clearly [[f[[ ≥ 0. Also, [[f[[ = 0 if and only if f(x) = f(y) for all x, y ∈ X if and only if f is
the zero function. Furthermore, [[f+g[[ ≤ [[f[[+[[g[[ since [(f+g)x−(f+g)y[ ≤ [f(x)−f(y)[+[g(x)−g(y)[
and sup A+B ≤ sup A+ sup B. Similarly, [[αf[[ = [α[ [[f[[. Thus [[ [[ deﬁnes a norm on Y .
For each x ∈ X, deﬁne the functional F
x
by F
x
(f) = f(x). Then F
x
(αf + βg) = αf(x) + βg(x) =
αF
x
(f)+βF
x
(g) so F
x
is a linear functional on Y . Also, [[F
x
(f)[[ = [f(x)[ = [f(x)−f(x
0
)[ ≤ ρ(x, x
0
)[[f[[
so F
x
is bounded. Furthermore, [[F
x
[[ ≤ ρ(x, x
0
) so [[F
x
−F
y
[[ ≤ ρ(x, x
0
) +ρ(y, x
0
) ≤ ρ(x, y). If [[f[[ = 1
and ε > 0, then there exist x, y ∈ X such that
]f(x)−f(y)]
ρ(x,y)
> 1 − ε. i.e. [f(x) − f(y)[ > (1 − ε)ρ(x, y).
Since [[F
x
−F
y
[[ = sup
]]f]]=1
[(F
x
−F
y
)f[ = sup
]]f]]=1
[f(x) −f(y)[ > (1 −ε)ρ(x, y) for all ε > 0, we have
[[F
x
−F
y
[[ ≥ ρ(x, y). Hence [[F
x
−F
y
[[ = ρ(x, y).
Thus X is isometric to a subset of the space Y
∗
of bounded linear operators from Y to R. Since Y
∗
is
complete, the closure of this subset gives a completion of X.
10.3 Linear functionals and the HahnBanach Theorem
17. Let f be a linear functional on a normed linear space. If f is bounded, then it is uniformly
continuous and by Q14, its kernel is closed. Conversely, if f is unbounded, then there is a sequence 'x
n
`
with [[x
n
[[ ≤ 1 for all n and f(x
n
) → ∞. Take x / ∈ ker f and consider y
n
= x − (f(x)/f(x
n
))x
n
. Each
y
n
is in ker f and y
n
→ x. Thus ker f is not closed.
18. Let T be a linear subspace of a normed linear space X and y a given element of X. If y ∈ T, then
inf
t∈T
[[y − t[[ = 0 = sup¦f(y) : [[f[[ = 1, f(t) = 0 for all t ∈ T¦. Thus we may assume y / ∈ T. Let
δ = inf
t∈T
[[y − t[[. Then [[y − t[[ ≥ δ for all t ∈ T. There is a bounded linear functional f on X such
that [[f[[ = 1, f(y) = δ and f(t) = 0 for all t ∈ T. Thus δ ≤ sup¦f(y) : [[f[[ = 1, f(t) = 0 for all t ∈ T¦.
If δ < sup¦f(y) : [[f[[ = 1, f(t) = 0 for all t ∈ T¦, then there exists f with [[f[[ = 1 and f(t) = 0 for t ∈ T
such that f(y) > δ so there exists t ∈ T such that f(y) > [[y − t[[. But then [[y − t[[ = [[f[[ [[y − t[[ ≥
f(y − t) = f(y) > [[y − t[[. Contradiction. Thus δ ≥ sup¦f(y) : [[f[[ = 1, f(t) = 0 for all t ∈ T¦. Hence
inf
t∈T
[[y −t[[ = sup¦f(y) : [[f[[ = 1, f(t) = 0 for all t ∈ T¦.
19. Let T be a linear subspace of a normed linear space X and y an element of X whose distance to T
is at least δ. Let S be the subspace consisting of multiples of y. Deﬁne f(λy) = λδ. Then f is a linear
functional on S. Let p(x) = inf
t∈T
[[x − t[[. Then f(λy) = λδ ≤ λp(y) ≤ p(λy). By the HahnBanach
Theorem, we may extend f to all of X so that f(x) ≤ p(x) for all x ∈ X. In particular, f(y) = δ and
f(t) = 0 for all t ∈ T. Also, f(x) ≤ p(x) = inf
t∈T
[[x −t[[ ≤ [[x[[ so [[f[[ ≤ 1.
20. Let
∞
be the space of all bounded sequences and let S be the subspace consisting of the constant
sequences. Let G be the Abelian semigroup of operators generated by the shift operator A given by
A['ξ
n
`] = 'ξ
n+1
`. If 'ξ
n
` ∈ S, say ξ
n
= ξ for all n, deﬁne f['ξ
n
`] = ξ. Then f is a linear functional on
S. Deﬁne p['ξ
n
`] = limξ
n
. Then f['ξ
n
`] = p['ξ
n
`] on S. Also, p(A
n
x) = p(x) for all x ∈ X. If 'ξ
n
` ∈ S,
then A
n
['ξ
n
`] = 'ξ
n
` ∈ S and f(A
n
['ξ
n
`]) = f['ξ
n
`]. By Proposition 5, there is an extension F of f to a
linear functional on X such that F(x) ≤ p(x) and F(Ax) = F(x) for all x ∈ X.
In particular, F['ξ
n
`] ≤ limξ
n
. Also, −F['ξ
n
`] = F['−ξ
n
`] ≤ lim(−ξ
n
) = −limξ
n
so limξ
n
≤ F['ξ
n
`]. By
linearity, F['ξ
n
+η
n
`] = F['ξ
n
` +'η
n
`] = F['ξ
n
`] +F['η
n
`] and F['αξ
n
`] = F[α'ξ
n
`] = αF['ξ
n
`]. Finally,
if η
n
= ξ
n+1
, then F['η
n
`] = F[A['ξ
n
`]] = F['ξ
n
`].
(*) The functional F is called a Banach limit and is often denoted by Lim.
*21. Let X be the space of bounded realvalued functions on the unit circle and let S be the subspace
of bounded Lebesgue measurable functions on the unit circle. For s ∈ S, deﬁne f(s) =
s. Also deﬁne
p(x) = inf
x≤s
f(s). Then f is a linear functional on S with f ≤ p on S. Let G consist of the rotations
so that it is an Abelian semigroup of operators on X such that for every A ∈ G we have p(Ax) ≤ p(x)
for x ∈ X while for s ∈ S we have As ∈ S and f(As) = f(s). Then there is an extension of f to a linear
functional F on X such that F(x) ≤ p(x) and F(Ax) = F(x) for x ∈ X. For a subset P of the unit
circle, let µ(P) = F(χ
P
). This will be a rotationally invariant measure on [0, 2π]. Then extend it to the
62
bounded subsets of R to get the required set function.
22. Let X be a Banach space. Suppose X
∗
is reﬂexive. If X is not reﬂexive, then there is a nonzero
function y ∈ X
∗∗∗
such that y(x
t
) = 0 for all x
t
∈ ϕ[X]. But there exists x
∗
∈ X
∗
such that y = ϕ
∗
(x
∗
).
If x ∈ X, then 0 = y(ϕ(x)) = (ϕ
∗
(x
∗
))(ϕ(x)) = (ϕ(x))(x
∗
) = x
∗
(x). Thus x
∗
= 0 and so y = 0.
Contradiction. Thus X is reﬂexive. Conversely, suppose X is reﬂexive. Let x
∗∗∗
∈ X
∗∗∗
. Deﬁne x
∗
∈ X
∗
by x
∗
(x) = x
∗∗∗
(ϕ(x)). Then ϕ
∗
(x
∗
)(x
∗∗
) = x
∗∗
(x
∗
) = (ϕ(x))(x
∗
) = x
∗
(x) = x
∗∗∗
(ϕ(x)) = x
∗∗∗
(x
∗∗
).
Thus X
∗
is reﬂexive.
23a. If x, y ∈ S
◦
and α, β ∈ R, then (αx + βy)(s) = αx(s) + βy(s) = 0 for all s ∈ S so S
◦
is a linear
subspace of X
∗
. Let 'y
n
` be a sequence in S
◦
that converges to some y ∈ X
∗
. By Q13, for each s ∈ S,
y
n
(s) → y(s). Since y
n
(s) = 0 for all n, we have y(s) = 0. Thus y ∈ S
◦
and S
◦
is closed.
*23b. If x ∈
¯
S, then there is a sequence 's
n
` in S converging to x. Let y ∈ S
◦
. Then y(s
n
) = 0 for
all n so y(x) = 0. Thus
¯
S ⊂ S
◦◦
. Suppose there exists x ∈ S
◦◦
`
¯
S. Then there is a linear functional
f with [[f[[ ≤ 1, f(x) = inf
t∈
¯
S
[[x − t[[ > 0 and f(t) = 0 for t ∈
¯
S. Thus f ∈ (
¯
S)
◦
= S
◦
so f(x) = 0.
Contradiction. Hence S
◦◦
=
¯
S.
23c. Let S be a closed subspace of X and let ϕ : X
∗
→ X
∗
/S
◦
be the natural homomorphism. Deﬁne
A : X
∗
→ S
∗
by Ay = y[
S
. Then A is a bounded linear operator with kernel S
◦
. By Q15b, there is a
unique bounded linear operator B : X
∗
/S
◦
→ S
∗
such that A = B ◦ ϕ. By the HahnBanach Theorem,
A is onto. Thus so is B. If y[
S
= z[
S
, then y −z ∈ S
◦
so ϕ(y) = ϕ(z) and B is onetoone. Hence B is
an isomorphism between S
∗
and X
∗
/S
◦
.
23d. Let S be a closed subspace of a reﬂexive Banach space X. Let ϕ : X → X
∗∗
be the natural
isomorphism and deﬁne A : X
∗
→ S
∗
by Ay = y[
S
. Let s
∗∗
∈ S
∗∗
. Then s
∗∗
◦A ∈ X
∗∗
so s
∗∗
◦A = ϕ(x)
for some x ∈ X. If x / ∈ S, then there exists x
∗
∈ X
∗
such that x
∗
(x) > 0 and x
∗
(s) = 0 for s ∈ S.
Then A(x
∗
) = 0 so x
∗
(x) = (ϕ(x))(x
∗
) = (s
∗∗
◦ A)(x
∗
) = 0. Contradiction. Thus x ∈ S. Now for any
s
∗
∈ S
∗
, there exists x
∗
∈ X
∗
such that A(x
∗
) = s
∗
. Then s
∗∗
(s
∗
) = (s
∗∗
◦ A)(x
∗
) = (ϕ(x))(x
∗
) =
x
∗
(x) = s
∗
(x) = (ϕ
S
(x))(s
∗
). i.e. s
∗∗
= ϕ
S
(x). Hence S is reﬂexive.
24. Let X be a vector space and P a subset of X such that x, y ∈ P implies x +y ∈ P and αx ∈ P for
α > 0. Deﬁne a partial order in X by deﬁning x ≤ y to mean y −x ∈ P. A linear functional f on X is
said to be positive (with respect to P) if f(x) ≥ 0 for all x ∈ P. Let S be any subspace of X with the
property that for each x ∈ X there is an s ∈ S with x ≤ s. Let f be a positive linear functional on S.
The family of positive linear functionals on S is partially ordered by setting f ≺ g if g is an extension of
f. By the Hausdorﬀ Maximal Principle, there is a maximal linearly ordered subfamily ¦g
α
¦ containing
f. Deﬁne a functional F on the union of the domains of the g
α
by setting F(x) = g
α
(x) if x is in the
domain of g
α
. Since the subfamily is linearly ordered, F is welldeﬁned. Also, F is a positive linear
functional extending f. Furthermore, F is a maximal extension since if G is any extension of F, then
g
α
≺ F ≺ G implies that G must belong to ¦g
α
¦ by maximality of ¦g
α
¦. Thus G ≺ F so G = F.
Let T be a proper subspace of X with the property that for each x ∈ X there is a t ∈ T with x ≤ t. We
show that each positive linear functional g on T has a proper extension h. Let y ∈ X ` T and let U be
the subspace spanned by T and y. If h is an extension of g, then h(λy +t) = λh(y) +h(t) = λh(y) +g(t).
There exists t
t
∈ T with y ≤ t
t
. i.e. t
t
−y ∈ T. Then λ(t
t
−y) + t ∈ T and g(λ(t
t
−y) + t) ≥ 0. Deﬁne
h(y) = g(t
t
−y). Then h(λy +t) = λh(y) +g(t) = g(λ(t
t
−y)) +g(t) = g(λ(t
t
−y) +t) ≥ 0. Thus h is a
proper extension of g. Since F is a maximal extension, it follows that F is deﬁned on X.
*25. Let f be a mapping of the unit ball S = ¦x : [[x[[ ≤ 1¦ into R such that f(αx+βy) = αf(x)+βf(y)
whenever x, y and αx + βy are in S. Deﬁne g(x) = [[x[[f(
x
]]x]]
). If [[x[[ ≤ 1, then g(x) = [[x[[
1
]]x]]
f(x) =
f(x). If x, y ∈ X, then g(x + y) = [[x + y[[f(
x+y
]]x+y]]
) = [[x + y[[f(
]]x]]
]]x+y]]
x
]]x]]
+
]]y]]
]]x+y]]
y
]]y]]
) = [[x +
y[[[
]]x]]
]]x+y]]
f(
x
]]x]]
) +
]]y]]
]]x+y]]
f(
y
]]y]]
)] = [[x[[f(
x
]]x]]
) + [[y[[f(
y
]]y]]
) = g(x) + g(y). If α ∈ R and x ∈ X, then
g(αx) = [[αx[[f(
αx
]]αx]]
) = [α[ [[x[[f(
αx
]α] ]]x]]
) = [α[ [[x[[
α
]α]
f(
x
]]x]]
) = αg(x). Thus g is a linear functional on
X extending f.
10.4 The Closed Graph Theorem
26. Let 'T
n
` be a sequence of continuous linear operators from a Banach space X to a normed vector
space Y . Suppose that for each x ∈ X the sequence 'T
n
x` converges to a value Tx. Now for each
63
x ∈ X there exists M
x
such that [[T
n
x[[ ≤ M
x
for all n. Thus there exists M such that [[T
n
[[ ≤ M
for all n. Given ε > 0, for each x ∈ X, there exists N such that [[T
N
x − Tx[[ < ε. Then [[Tx[[ ≤
[[T
N
x − Tx[[ + [[T
N
x[[ < ε + M[[x[[. Thus [[Tx[[ ≤ M[[x[[ for all x ∈ X. i.e. T is a bounded linear
operator.
27. Let A be a bounded linear transformation from a Banach space X to a Banach space Y , and let M
be the kernel and S the range of A. Suppose S is isomorphic to X/M. Since X is complete and M is
closed, X/M is complete by Q12. Thus X/M is closed and since S is isomorphic to X/M, S is also closed.
Conversely, suppose S is closed. Then since Y is complete, so is S. Let ϕ be the natural homomorphism
from X to X/M. There is a unique bounded linear operator B : X/M → S such that A = B◦ϕ. Since A
and ϕ are onto, so is B. Thus B is an open mapping. It remains to show that B is onetoone. Suppose
Bx
t
= By
t
. Then x
t
= ϕ(x) and y
t
= ϕ(y) for some x, y ∈ X so B(ϕ(x)) = B(ϕ(y)). Then Ax = Ay so
x −y ∈ M and x
t
= y
t
. Hence B is onetoone and is thus an isomorphism.
28a. Let S be a linear subspace of C[0, 1] that is closed as a subspace of L
2
[0, 1]. Let 'f
n
` be a sequence
in S converging to f in C[0, 1]. i.e. [[f
n
− f[[
∞
→ 0. Then since [[f
n
− f[[
2
≤ [[f
n
− f[[
∞
, we have
[[f
n
−f[[
2
→ 0. Thus f ∈ S. Hence S is closed as a subspace of C[0, 1].
28b. For any f ∈ S, we have [[f[[
2
= (
f
2
)
1/2
≤ (
[[f[[
2
∞
)
1/2
= [[f[[
∞
. Since S is closed in both C[0, 1]
and L
2
[0, 1], it is complete in both norms. Thus there exists M such that [[f[[
∞
≤ M[[f[[
2
.
*28c. Let y ∈ [0, 1] and deﬁne F(f) = f(y). Then F is a linear functional on L
2
[0, 1]. Also, [F(f)[ =
[f(y)[ ≤ [[f[[
∞
≤ M[[f[[
2
so F is bounded. By the Riesz Representation Theorem, there exists k
y
∈ L
2
such that f(y) = F(f) =
k
y
(x)f(x) dx.
*29a. Let Y = C[0, 1] and let X be the subspace of functions which have a continuous derivative. Let A
be the diﬀerential operator. Let x
n
(t) = t
n
. Then [[x
n
[[ = 1 and Ax
n
(t) = nt
n−1
so [[Ax
n
[[ = n. Thus
A is unbounded and thus discontinuous. Let x
n
∈ X such that x
n
→ x and x
t
n
= Ax
n
→ y. Since we
have uniform convergence,
y =
limx
t
n
= lim
x
t
n
= x(t) − x(0) so x(t) = x(0) +
y. Thus x ∈ X
and Ax = x
t
= y. Hence A has a closed graph.
*29b. Consider A : R → R given by A(x) = 1/x if x = 0 and A(0) = 0. Then A is a discontinuous
operator from a Banach space to a normed linear space with a closed graph.
10.5 Topological vector spaces
30a. Let B be a collection of subsets containing θ. Suppose B is a base at θ for a translation invariant
topology. By deﬁnition of a base, if U, V ∈ B, there exists W ∈ B such that W ⊂ U ∩ V so (i) holds. If
U ∈ B and x ∈ U, then U −x is open so there exists V ∈ B such that V ⊂ U −x. Then x + V ⊂ U so
(ii) holds.
Conversely, suppose (i) and (ii) hold. Let T = ¦O : x ∈ O ⇒ ∃y ∈ X and U ∈ B such that x ∈ y+U ⊂ O¦.
It follows that T contains ∅ and X, and is closed under union. If x ∈ O
1
∩O
2
, then there exist y
1
, y
2
∈ X
and U
1
, U
2
∈ B such that x ∈ y
i
+U
i
∈ O
i
, i = 1, 2. Now x −y
i
∈ U
i
so by (ii), there exists V
i
∈ B such
that x −y
i
+V
i
⊂ U
i
. i.e. x +V
i
⊂ y
i
+U
i
⊂ O
i
. Now by (i), there exists W ∈ B such that W ⊂ V
1
∩V
2
so x ∈ x + W ⊂ O
1
∩ O
2
. Thus T is closed under ﬁnite intersection. If O ∈ T and y ∈ x + O, then
y −x ∈ O so there exists z ∈ X and U ∈ B such that y −x ∈ z +U ⊂ O. Thus y ∈ x+z +U ⊂ x+O so
x + O ∈ T . Hence T is a translation invariant topology. Furthermore, if θ ∈ O, then there exist x ∈ X
and U ∈ B such that θ ∈ x + U ⊂ O. Thus −x ∈ U so there exists V ∈ B such that −x + V ⊂ U. i.e.
V ⊂ x +U. Then θ ∈ V ⊂ O. Thus B is a base at θ.
30b. Let B be a base at θ for a translation invariant topology. Suppose addition is continuous from
XX to X. In particular, addition is continuous at 'θ, θ`. Thus for each U ∈ B, there exists V
1
, V
2
∈ B
such that V
1
+V
2
⊂ U. Take V ∈ B with V ⊂ V
1
∩V
2
. Then V +V ⊂ U. Conversely, suppose (iii) holds.
For x
0
, y
0
∈ X, ¦x
0
+ y
0
+ U : U ∈ B¦ is a base at x
0
+ y
0
. Now for each U ∈ B, pick V ∈ B such that
V +V ⊂ U. If x ∈ x
0
+V and y ∈ y
0
+V , then x +y ∈ x
0
+y
0
+U. Thus addition is continuous from
X X to X.
30c. Suppose scalar multiplication is continuous (at '0, θ`) from RX to X. Given U ∈ B and x ∈ X,
there exist ε > 0 and V ∈ B such that β(x + V ) ⊂ U for [β[ < ε. Let α = 2/ε. Then
1
α
(x + V ) ⊂ U so
x +V ⊂ αU. In particular, x ∈ αU.
30d. Let X be a topological vector space and let B be the family of all open sets U that contain θ and
64
such that αU ⊂ U for all α with [α[ < 1. If O is an open set containing θ, then continuity of scalar
multiplication implies that there is an open set V containing θ and an ε > 0 such that λV ⊂ O for all
[λ[ < ε. Let U =
¸
]λ]<ε
λV . Then V is open, θ ∈ U ⊂ O, and αU ⊂ U for α with [α[ < 1. Thus B is a
local base for the topology and it satisﬁes (v) by its deﬁnition.
*30e. Suppose B satisﬁes the conditions of the proposition. By part (a), B is a base at θ for a translation
invariant topology and by part (b), addition is continuous. Given U ∈ B, there exists V ∈ B such that
V + V ⊂ U. Now given x ∈ X, there exists α ∈ R such that x ∈ αV . Let ε = 1/[α[. If [λ[ < min(ε, 1),
then λ(x + V ) ⊂ U since λx ∈ λαV ⊂ V and λV ⊂ V . Thus scalar multiplication is continuous at
'0, x`. Given U ∈ B and α ∈ R, let β be such that 0 < [β[ ≤ 1 and choose n such that [β[
−n
> [α[. Let
ε = [β[
n
−[α[. Now when [λ −α[ < ε, we have [λ[ < [β[
−n
since [α[ = max(α, −α). Then β
n
U ∈ B and
λβ
n
U ⊂ U. Thus scalar multiplication is continuous at 'α, θ`. Now given U ∈ B, there exists V ∈ B
such that V +V ⊂ U.
Since scalar multiplication is continuous at '0, x` and 'α, θ`, given α
0
∈ R and x
0
∈ X, there exist ε > 0
and W, W
t
∈ B such that α(x
0
+ W) ⊂ V when [α[ < ε and αW
t
⊂ V when [α − α
0
[ < ε. Then
αx − α
0
x
0
= α(x − x
0
) + (α − α
0
)x
0
∈ V + V ⊂ U when [α − α
0
[ < ε and x ∈ x
0
+ W
t
. Hence scalar
multiplication is continuous from R X to X.
30f. Suppose X is T
1
. If x = θ and x ∈
¸
¦U ∈ B¦, then any open set containing θ will also contain x.
Contradiction. Hence (vi) holds. Conversely, suppose (vi) holds. Given two distinct points x and y, there
exists U ∈ B such that x−y / ∈ U. Also, there exists V ∈ B such that V +V ⊂ U. If (x+V )∩(y+V ) = ∅,
then x −y ∈ V −V . By (v), −V ⊂ V so x −y ∈ V +V ⊂ U. Contradiction. Thus x +V and y +V are
disjoint open sets separating x and y so X is Hausdorﬀ.
(*) Proof of Proposition 14
31a. Suppose a linear transformation f from one topological vector space X to a topological vector
space Y is continuous at one point. We may assume f is continuous at the origin. Let O be an open set
containing the origin in Y . There exists an open set U containing the origin in X such that f[U] ⊂ O.
Since f is linear, for any x ∈ X, f[x +U] = f(x) +f[U] ⊂ f(x) +O. Hence f is uniformly continuous.
*31b. Let f be a linear functional on a topological vector space X. Suppose f is continuous. Let I be
a bounded open interval containing 0. There exists an open set O containing the origin in X such that
f[O] ⊂ I. Thus f[O] = R. Conversely, suppose there is a nonempty open set O such that f[O] = R.
Take x ∈ O. Then O −x is an open neighbourhood of θ so there is an open neighbourhood U of θ such
that U ⊂ O −x and αU ⊂ U for α with [α[ < 1. Now f[U] = R and αf[U] = f[αU] ⊂ f[U] if [α[ < 1 so
f[U] is a bounded interval. Thus f is continuous at θ and thus continuous everywhere.
32. Let X be a topological vector space and M a closed linear subspace. Let ϕ be the natural homo
morphism of X onto X/M, and deﬁne a topology on X/M by taking O to be open if and only if ϕ
−1
[O]
is open in X. Clearly, ϕ is continuous. If U is open in X, then ϕ
−1
[ϕ[U]] =
¸
m∈M
(m + U), which is
open so ϕ[U] is open. Now let O be an open set containing x
t
+ y
t
∈ X/M. Then ϕ
−1
[O] is an open
set containing x + y ∈ X. There exist open sets U and V containing x and y respectively such that
U + V ⊂ ϕ
−1
[O]. Since ϕ is open, ϕ[U] and ϕ[V ] are open sets containing x
t
and y
t
respectively. Then
ϕ[U] +ϕ[V ] ⊂ ϕ[ϕ
−1
[O]] = O. Thus addition is continuous from X/MX/M to X/M. Now let O be an
open set containing cx
t
∈ X/M. Then ϕ
−1
[O] is an open set containing cx ∈ X. There exist open sets
U and V containing c and x respectively such that UV ⊂ ϕ
−1
[O]. Since ϕ is open, ϕ[V ] is an open set
containing x
t
. Then Uϕ[V ] ⊂ ϕ[ϕ
−1
[O]] = O. Thus scalar multiplication is continuous from R X/M
to X/M. Hence X/M is a topological vector space.
33a. Suppose X is ﬁnite dimensional topological vector space. Let x
1
, . . . , x
n
be a vector space basis of X
and let e
1
, . . . , e
n
be the standard basis of R
n
. Deﬁne a linear map ϕ of R
n
to X so that ϕ(
¸
n
i=1
a
i
e
i
) =
¸
n
i=1
a
i
x
i
. Then ϕ is onetoone and thus onto. If a sequence '
¸
a
(j)
i
e
i
` in R
n
converges to
¸
a
i
e
i
,
then 'a
(j)
i
` converges to a
i
for each i. Since addition and scalar multiplication are continuous on X, the
sequence '
¸
a
(j)
i
x
i
` converges to
¸
a
i
x
i
. Thus ϕ is continuous.
33b. Suppose X is Hausdorﬀ. Let S and B be the subsets of R
n
deﬁned by S = ¦y : [[y[[ = 1¦ and
B = ¦y : [[y[[ < 1¦. Since S is compact and ϕ is continuous, ϕ[S] is compact in X and thus closed. Then
X ` ϕ[S] is open.
33c. Let B be a base at θ satisfying the conditions of Proposition 14. Since X ` ϕ[S] is an open set
65
containing θ, there exists U ∈ B such that U ⊂ X ` ϕ[S]. Furthermore, αU ⊂ U for each [α[ < 1 by
condition (v) of Proposition 14.
33d. Suppose u = ϕ(y) ∈ U for some y with [[y[[ > 1. Then
1
]]y]]
< 1 so
1
]]y]]
u ∈ U but
1
]]y]]
u =
1
]]y]]
ϕ(y) = ϕ(
y
]]y]]
) where [[
y
]]y]]
[[ = 1 so
1
]]y]]
u ∈ ϕ[S]. Contradiction. Hence U ⊂ ϕ[B]. Thus ϕ[B] is
open and ϕ
−1
is continuous.
(*) Proof of Proposition 15.
34. Let M be a ﬁnite dimensional subspace of a Hausdorﬀ topological vector space X. If x / ∈ M, let N
be the ﬁnite dimensional subspace spanned by x and M. Then N has the usual topology so x is not a
point of closure of M. Hence M is closed.
35. Let A be a linear mapping of a ﬁnite dimensional Hausdorﬀ vector space X into a topological vector
space Y . The range of A is a ﬁnite dimensional subspace of Y so it has the usual topology. If x
n
→ x in
X where x
n
=
¸
a
(n)
i
e
i
and x =
¸
a
i
e
i
, then a
(n)
i
→ a
i
for each i. Since the range of A has the usual
topology, Ax
n
=
¸
a
(n)
i
Ae
i
→
¸
a
i
Ae
i
= Ax. Hence A is continuous.
*36. Let A be a linear mapping from a topological vector space X to a ﬁnite dimensional topological
space Y . If A is continuous, then its kernel M is closed by Q14. Conversely, suppose M is closed. There
is a unique linear mapping B : X/M → Y such that A = B ◦ ϕ where ϕ : X → X/M is the natural
homomorphism, which is a continuous open map by Q32. Then ker ϕ is closed and ϕ is an open map so
X/M is a Hausdorﬀ topological vector space. Furthermore, B is onetoone and Y is ﬁnite dimensional
so X/M is ﬁnite dimensional. By Q35, B is continuous. Hence A is continuous.
*37. Let X be a locally compact Hausdorﬀ vector space. Let V be a neighbourhood of θ with
¯
V compact
and αV ⊂ V for each α with [α[ < 1. The set ¦x+
1
3
V : x ∈
¯
V ¦ is an open cover of
¯
V so
¯
V can be covered
by a ﬁnite number of translates x
1
+
1
3
V, . . . , x
n
+
1
3
V . To show that x
1
, . . . , x
n
span X, it suﬃces to
show that they span V . Let x ∈ V . Then x = x
k
1
+
1
3
u
1
for some k
1
∈ ¦1, . . . , n¦ and u
1
∈ U. Now
1
3
U
is covered by
1
3
(x
1
+
1
3
V ), . . . ,
1
3
(x
n
+
1
3
V ) so x = x
k
1
+
1
3
x
k
2
+
1
9
u
2
. Continuing in this way, we have
x ∈ x
k
1
+
1
3
x
k
2
+ +
1
3
r−2
x
k
r
+
1
3
r
U for each r. Let y
r
= x
k
1
+
1
3
x
k
2
+ +
1
3
r−2
x
k
r
. Then y
r
is in the
span of S, which is ﬁnite dimensional and thus closed by Q34.
Now for each y ∈
¯
V , there exists ε
y
> 0 and an open neighbourhood U
y
of y such that δU
y
⊂ V whenever
[δ[ < ε
y
since scalar multiplication is continuous at '0, y`. The open sets U
y
cover
¯
V so
¯
V ⊂ U
y
1
∪ ∪U
y
n
.
Then δ
¯
V ⊂ V whenever [δ[ < min
1≤k≤n
ε
y
k
.
Choose N such that δ
0
¯
V ⊂ V whenever [δ
0
[ < 3
−N
. Let δ > 0 be given. Choose M such that
3
−M
δ
−1
< 3
−N
. When m ≥ M, we have 3
−m
δ
−1
< 3
−N
so 3
−m
δ
−1
¯
V ⊂ V . i.e. 3
−m
¯
V ⊂ δV . Thus
x − y
m
∈ δV for m ≥ M. Thus x − y
m
→ θ and y
m
→ x. Hence x is in the span of S. It follows that
x
1
, . . . , x
n
span X so X is ﬁnite dimensional.
10.6 Weak topologies
38a. Suppose x
n
→ x weakly. Then f(x
n
) → f(x) for each f ∈ X
∗
. Thus [f(x
n
)[ ≤ C
f
for each f ∈ X
∗
and each n. Now let ϕ : X → X
∗∗
be the natural homomorphism so that ϕ(x
n
)(f) = f(x
n
). Then
[ϕ(x
n
)(f)[ ≤ C
f
for each f ∈ X
∗
and each n. Since X
∗
is a Banach space, '[[ϕ(x
n
)[[` is bounded but
[[ϕ(x
n
)[[ = [[x
n
[[ so '[[x
n
[[` is bounded.
*38b. Let 'x
n
` be a sequence in
p
, 1 < p < ∞, and let x
n
= 'ξ
m,n
`
∞
m=1
. Suppose 'x
n
` converges
weakly to x = 'ξ
m
`. By part (a), '[[x
n
[[` is bounded. Each bounded linear functional F on
p
is given by
F(x
n
) =
¸
m
ξ
m,n
η
m
for some 'η
m
` ∈
q
. Conversely, taking the mth term to be 1 and the remaining
terms to be 0 gives a sequence 'η
m
` ∈
q
so that F(x
n
) =
¸
ξ
m,n
η
m
= ξ
m,n
is a bounded linear functional
on
p
. Thus 'F(x
n
)` converges to F(x). i.e. 'ξ
m,n
`
∞
m=1
converges to ξ
m
.
Conversely, suppose '[[x
n
[[` is bounded and for each m we have ξ
m,n
→ ξ
m
. Then [[x
n
[[ ≤ C and
[[x[[ ≤ C for some C. Let F ∈ (
p
)
∗
=
q
. Let e
n
be the sequence having 1 as the nth term and
0 elsewhere. Then span¦e
n
¦ is dense in
q
so there exists 'F
n
` with F
n
∈ span¦e
n
¦ and F
n
→ F.
Given ε > 0, there exists N such that [[F
N
− F[[ < ε/3C. Since F
N
∈ span¦e
n
¦ and ξ
m,n
→ ξ
m
for each n, there is an M such that [F
N
(x
n
) − F
N
(x)[ < ε/3 for n ≥ M. Thus for n ≥ M, we have
[F(x
n
) − F(x)[ ≤ [F(x
n
) − F
N
(x
n
)[ + [F
N
(x
n
) − F
N
(x)[ + [F
N
(x) − F(x)[ < ε. Hence 'x
n
` converges
weakly to x.
66
38c. Let 'x
n
` be a sequence in L
p
[0, 1], 1 < p < ∞. Suppose '[[x
n
[[` is bounded and 'x
n
` converges to
x in measure. By Corollary 4.19, every subsequence of 'x
n
` has in turn a subsequence that converges
a.e. to x. By Q6.17, every subsequence of 'x
n
` has in turn a subsequence 'x
n
k
j
` such that for each
y ∈ L
q
[0, 1] we have
x
n
k
j
y →
xy. Now for each bounded linear functional F on L
p
[0, 1], we have
F(x) =
xy for some y ∈ L
q
[0, 1]. Thus every subsequence of 'x
n
` has in turn a subsequence 'x
n
k
j
`
such that F(x
n
k
j
) → F(x). By Q2.12, F(x
n
) → F(x). Hence 'x
n
` converges weakly to x.
38d. Let x
n
= nχ
[0,1/n]
for each n. Then x
n
→ 0 and [[x
n
[[
1
= 1 for all n. In particular, 'x
n
` is
a sequence in L
1
[0, 1] converging to 0 in measure. Let y = χ
[0,1]
∈ L
∞
[0, 1]. Then F(x) =
xy is a
bounded linear functional on L
1
[0, 1] and F(0) = 0 while F(x
n
) = [[x
n
[[
1
= 1 for each n so F(x
n
) does
not converge to F(0). Hence 'x
n
` does not converge weakly to 0.
(*) See Q6.17
38e. In
p
, 1 < p < ∞, let x
n
be the sequence whose nth term is one and whose remaining terms are
zero. For any bounded linear functional F on
p
, there exists 'y
n
` ∈
q
such that F(x
n
) = y
n
. Then
F(x
n
) = y
n
→ 0 since 'y
n
` ∈
q
. Thus x
n
→ 0 in the weak topology. If 'x
n
` converges in the strong
topology, then it must converge to 0 but [[x
n
[[
p
= 1 for all n so it does not converge to 0 and thus does
not converge in the strong topology.
38f. Let x
n
be as in part (e), and deﬁne y
n,m
= x
n
+ nx
m
. Let F = ¦y
n,m
: m > n¦. Note that
the distance between any two points in F is at least 1 so there are no nonconstant sequences in F that
converge in the strong topology. Any sequence in F that converges must be a constant sequence so its
limit is in F. Hence F is strongly closed.
38g. Let F be as in part (f). The sets ¦x : [f
i
(x)[ < ε, i = 1, . . . , n¦ where ε > 0 and f
1
, . . . , f
n
∈ (
p
)
∗
form a base at θ for the weak topology. Given ε > 0 and f
1
, . . . , f
n
∈ (
p
)
∗
, f
i
(y
m,n
) is of the form
ξ
i
n
+ nξ
i
m
where 'ξ
i
n
` ∈
q
. Choose n such that [ξ
i
n
[ < ε/2 for all i. Then choose m > n such that
[ξ
i
m
[ < ε/2n for all i. Then [f
i
(y
m,n
)[ ≤ [ξ
i
n
[ + n[ξ
i
m
[ < ε. Thus F ∩ ¦x : [f
i
(x)[ < ε, i = 1, . . . , n¦ = ∅
and θ is a weak closure point of F.
Suppose 'z
k
` = 'y
m
k
,n
k
` = 'x
n
k
+ n
k
x
m
k
` is a sequence from F that converges weakly to zero. Given
ε > 0 and 'ξ
n
` ∈
q
, there exists N such that [ξ
n
k
+ n
k
ξ
m
k
[ < ε for k ≥ N. Suppose ¦m
k
¦ is bounded
above. Then some m is repeated inﬁnitely many times. Let ξ
m
= 1 and ξ
n
= 0 otherwise. For each N
there exists k ≥ N such that m
k
= m so [ξ
n
k
+ n
k
ξ
m
k
[ = [n
k
[ ≥ 1. Thus ¦m
k
¦ is not bounded above
and we may assume the sequence 'm
k
` is strictly increasing. Now suppose ¦n
k
¦ is bounded above. Then
some n is repeated inﬁnitely many times. Let ξ
n
= 1 and ξ
m
= 0 otherwise. For each N there exists
k ≥ N such that n
k
= n so [ξ
n
k
+ n
k
ξ
m
k
[ = 1. Thus ¦n
k
¦ is not bounded above and we may assume
the sequence 'n
k
` is strictly increasing. Now let ξ
m
k
= 1/n
k
for each k and ξ
m
= 0 if m = m
k
for any
k. Then 'ξ
n
` ∈
q
and [ξ
n
k
+n
k
ξ
m
k
[ ≥ 1 for all k. Contradiction. Hence there is no sequence 'z
k
` from
F that converges weakly to zero.
38h. The weak topology of
1
is the weakest topology such that all functionals in (
1
)
∗
=
∞
are
continuous. A base at θ is given by the sets ¦x ∈
1
: [f
i
(x)[ < ε, i = 1, . . . , n¦ where ε > 0 and
f
1
, . . . , f
n
∈
∞
. A net '(x
(α)
n
)` in
1
converges weakly to (x
n
) ∈
1
if and only if
¸
n
x
(α)
n
y
n
converges
to
¸
n
x
n
y
n
for each (y
n
) ∈
∞
.
If a net '(x
(α)
n
)` in
1
converges weakly to (x
n
) ∈
1
, then for each n, taking (y
n
) ∈
∞
where y
n
= 1 and
y
m
= 0 for m = n, we have x
(α)
n
converging to x
n
for each n.
If the net '(x
(α)
n
)` in
1
is bounded, say by M, and x
(α)
n
converges to x
n
for each n, then
¸
n
[x
n
[ ≤
¸
n
[x
(α)
n
−x
n
[+
¸
n
[x
(α)
n
[ ≤ M so (x
n
) ∈
1
. If (y
n
) ∈
∞
, then [
¸
n
(x
(α)
n
−x
n
)y
n
[ ≤ [[(y
n
)[[
∞
¸
n
[x
(α)
n
−
x
n
[ → 0 so
¸
n
x
(α)
n
y
n
converges to
¸
n
x
n
y
n
and '(x
(α)
n
)` converges weakly to (x
n
).
For k ∈ N, let (x
(k)
n
) ∈
1
where x
(k)
k
= k and x
(k)
n
= 0 if n = k. Then the sequence '(x
(k)
n
)` is not
bounded and x
(k)
n
converges to 0 for each n. However, taking (y
n
) ∈
∞
where y
n
= 1 for all n, we have
¸
n
x
(k)
n
y
n
= k, which does not converge to 0 so '(x
(k)
n
)` does not converge weakly to θ.
The weak* topology on
1
as the dual of c
0
is the weakest topology such that all functionals in ϕ[c
0
] ⊂
∞
are continuous. A base at θ is given by the sets ¦f ∈
1
: [f(x
i
)[ < ε, i = 1, . . . , n¦ where ε > 0 and
x
1
, . . . , x
n
∈ c
0
. A net '(x
(α)
n
)` in
1
is weak* convergent to (x
n
) ∈
1
if and only if
¸
n
x
(α)
n
y
n
converges
67
to
¸
n
x
n
y
n
for all (y
n
) ∈ c
0
.
Using the same arguments as above and replacing
∞
by c
0
, we see that if a net '(x
(α)
n
)` in
1
is weak*
convergent to (x
n
) ∈
1
, then x
(α)
n
converges to x
n
for each n. We also see that if the net is bounded
and x
(α)
n
converges to x
n
, then the net is weak* convergent to (x
n
).
For k ∈ N, let (x
(k)
n
) ∈
1
where x
(k)
k
= k and x
(k)
n
= 0 if n = k. Then the sequence '(x
(k)
n
)` is not
bounded and x
(k)
n
converges to 0 for each n. However, taking (y
n
) ∈ c
0
where y
n
= 1/n for all n, we
have
¸
n
x
(k)
n
y
n
= 1, which does not converge to 0 so '(x
(k)
n
)` is not weak* convergent to θ.
39a. Let X = c
0
, T =
1
and T
0
the set of sequences with ﬁnitely many nonzero terms, which is dense
in
1
. Consider the sequence '(x
(k)
n
)` in c
0
where x
(k)
k
= k
2
and x
(k)
n
= 0 if n = k. For any sequence
(y
n
) ∈ T
0
, we have
¸
n
x
(k)
n
y
n
= k
2
y
k
→ 0. Thus the sequence '(x
(k)
n
)` converges to zero in the weak
topology generated by T
0
. Now if the sequence '(x
(k)
n
)` converges in the weak topology generated by T,
the weak limit must then be zero. Let (z
n
) be the sequence in T where z
n
= 1/n
2
for each n. Then
¸
n
x
(k)
n
z
n
= 1. Thus the sequence '(x
(k)
n
)` does not converge in the weak topology generated by T.
Hence T and T
0
generate diﬀerent weak topologies for X.
Now suppose S is a bounded subset of X. We may assume that S contains θ. Let T be a set of functionals
in X
∗
and let T
0
be a dense subset of T (in the norm topology on X
∗
). Note that in general, the weak
topology generated by T
0
is weaker than the weak topology generated by T. A base at θ for the weak
topology on S generated by T is given by the sets ¦x ∈ S : [f
i
(x)[ < ε, i = 1, . . . , n¦ where ε > 0 and
f
1
, . . . , f
n
∈ T. A set in a base at θ for the weak topology on S generated by T
0
is also in a base at θ
for the weak topology generated by T. Suppose x ∈ S so that [[x[[ ≤ M and [f
i
(x)[ < ε for some ε > 0
and f
1
, . . . , f
n
∈ T. For each i, there exists g
i
∈ T
0
such that [[f
i
− g
i
[[ < ε/2M. If [g
i
(x)[ < ε/2 for
i = 1, . . . , n, then [f
i
(x)[ ≤ [f
i
(x) − g
i
(x)[ + [g
i
(x)[ ≤ [[f
i
− g
i
[[ [[x[[ + [g
i
(x)[ < ε for i = 1, . . . , n. Thus
any set in a base at θ for the weak topology on S generated by T contains a set in a base at θ for the
weak topology generated by T
0
. Hence the two weak topologies are the same on S.
*39b. Let S
∗
be the unit sphere in the dual X
∗
of a separable Banach space X. Let ¦x
n
¦ be a countable
dense subset of X. Then ¦ϕ(x
n
)¦ is a countable dense subset of ϕ[X]. By part (a), ¦ϕ(x
n
)¦ generates
the same weak topology on S
∗
as ϕ[X]. i.e. ¦ϕ(x
n
)¦ generates the weak* topology on S
∗
. Now deﬁne
ρ(f, g) =
¸
2
−n
]f(x
n
)−g(x
n
)]
1+]f(x
n
)−g(x
n
)]
. Then ρ is a metric on S
∗
. Furthermore ρ(f
n
, f) → 0 if and only if
[f
n
(x
k
) −f(x
k
)[ → 0 for each k (see Q7.24a) if and only if [ϕ(x
k
)(f
n
) −ϕ(x
k
)(f)[ → 0 for each k if and
only if f
n
→ f in the weak* topology. Hence S
∗
is metrizable.
40. Suppose X is a weakly compact set. Every x
∗
∈ X
∗
is continuous so x
∗
[X] is compact in R, and
thus bounded, for each x
∗
∈ X
∗
. For each x ∈ X and x
∗
∈ X
∗
, there is a constant M
x
∗ such that
[ϕ(x)(x
∗
)[ = [x
∗
(x)[ ≤ M
x
∗. Thus ¦[[ϕ(x)[[ : x ∈ X¦ is bounded. Since [[ϕ(x)[[ = [[x[[ for each x, we
have ¦[[x[[ : x ∈ X¦ is bounded.
41a. Let S be the linear subspace of C[0, 1] given in Q28 (S is closed as a subspace of L
2
[0, 1]. Suppose
'f
n
` is a sequence in S such that f
n
→ f weakly in L
2
. By Q28c, for each y ∈ [0, 1], there exists
k
y
∈ L
2
such that for all f ∈ S we have f(y) =
k
y
f. Now
f
n
k
y
→
fk
y
for each y ∈ [0, 1] since
k
y
∈ L
2
= (L
2
)
∗
. Thus f
n
(y) → f(y) for each y ∈ [0, 1].
41b. Suppose 'f
n
` is a sequence in S such that f
n
→ f weakly in L
2
. By Q38a, '[[f
n
[[
2
` is bounded.
By Q28b, there exists M such that [[f[[
∞
≤ M[[f[[
2
for all f ∈ S. In particular, [[f
n
[[
∞
≤ M[[f
n
[[
2
for
all n. Hence '[[f
n
[[
∞
` is bounded. Now 'f
2
n
` is a sequence of measurable functions with [f
n
[
2
≤ M
t
on
[0, 1] and f
2
n
(y) → f
2
(y) for each y ∈ [0, 1] as a consequence of part (a). By the Lebesgue Convergence
Theorem, [[f
n
[[
2
2
→ [[f[[
2
2
and so [[f
n
[[
2
→ [[f[[
2
. By Q6.16, f
n
→ f strongly in L
2
.
*41c. Since L
2
is reﬂexive and S is a closed linear subspace, S is a reﬂexive Banach space by Q23d. By
Alaoglu’s Theorem, the unit ball of S
∗∗
is weak* compact. Then the unit ball of S is weakly compact
since the weak* topology on S
∗∗
induces the weak topology on S when S is regarded as a subspace of
S
∗∗
. By part (b), the unit ball of S is compact. Thus S is locally compact Hausdorﬀ so by Q37, S is
ﬁnite dimensional.
68
10.7 Convexity
42. Let A be a linear operator from the vector space X to the vector space Y . Let K be a convex set in
X. If x, y ∈ X and 0 ≤ λ ≤ 1, then λAx+(1 −λ)Ay = A(λx) +A((1 −λ)y) = A(λx+(1 −λ)y) ∈ A[X].
Thus A[X] is a convex set in Y . Let K
t
be a convex set in Y . If x, y ∈ A
−1
[K
t
], then Ax, Ay ∈ K
t
so if
0 ≤ λ ≤ 1, then λAx+(1−λ)Ay ∈ K
t
. Now λAx+(1−λ)Ay = A(λx+(1−λ)y) so λx+(1−λ)y ∈ A
−1
[K
t
].
Thus A
−1
[K
t
] is a convex set in X. By using the linearity of A, it can be shown that a similar result
holds when “convex set” is replaced by “linear manifold”.
Deﬁne the linear operator A : R
2
→R by A('x, y`) = x+y. Take the nonconvex set ¦'0, 1`, '1, 0`¦ ⊂ R
2
.
Its image under A is the conves set ¦1¦ ⊂ R.
43. Let K be a convex set in a topological vector space. Let x, y ∈
¯
K and 0 ≤ λ ≤ 1. Let O be an open
set containing λx+(1−λ)y. Since addition and scalar multiplication are continuous, there are open sets
U and V containing x and y respectively, as well as ε > 0, such that µU +ηV ⊂ O whenever [µ−λ[ < ε
and [η −(1 −λ)[ < ε. Now there exist x
t
∈ K∩U and y
t
∈ K∩V . Then λx
t
+(1 −λ)y
t
∈ K∩O. Hence
λx + (1 −λ)y ∈
¯
K and
¯
K is convex.
44a. Let x
0
be an interior point of a subset K of a topological vector space X. There is an open set
O such that x
0
∈ O ⊂ K. Let x ∈ X. By continuity of addition at 'x
0
, θ`, there exists an open set U
containing θ such that x
0
+ U ⊂ O. By continuity of scalar multiplication at '0, x`, there exists ε > 0
such that λx ∈ U whenever [λ[ < ε. Thus x
0
+λx ∈ O ⊂ K whenever [λ[ < ε and x
0
is an internal point
of K.
44b. In R, a convex set must be an interval and an internal point must not be an endpoint of the
interval so it is an interior point. In R
n
for n ≥ 2, let ¦e
i
: i = 1, . . . , n¦ be the standard basis.
Suppose x = 'x
1
, . . . , x
n
` is an internal point of a convex set K. For each i, there exists ε
i
> 0 such
that x + λe
i
∈ K if [λ[ < ε
i
. Let ε = min
1≤i≤n
ε
i
. If [λ[ < ε, then x + λe
i
∈ K for all i. Now
suppose [[y − x[[ < ε/n. Then [y
i
− x
i
[ < ε/n for all i so x + n(y
i
− x
i
)e
i
∈ K for all i. Note that
y =
¸
n
i=1
1
n
(x +n(y
i
−x
i
)e
i
), which belongs to K since K is convex. Thus there is an open ball centred
at x and contained in K so x is an interior point of K.
*44c. Consider K = B
¹0,1),1
∪ B
¹0,−1),1
∪ ¦'x, 0` : [x[ < 1¦ ⊂ R
2
. Then '0, 0` is an internal point of K
but not an interior point.
44d. Suppose a convex set K in a topological vector space has an interior point x. Let y be an internal
point of K. There exists ε > 0 such that y+
ε
2
(y−x) ∈ K. Now there is an open set O such that x ∈ O ⊂
K. Let λ = ε/2. Then y =
λ
1+λ
x+
1
1+λ
[y +λ(y −x)] ∈
λ
1+λ
O+
1
1+λ
[y +λ(y −x)] ⊂
λ
1+λ
K +
1
1+λ
K ⊂ K.
Since
λ
1+λ
O +
1
1+λ
[y +λ(y −x)] is an open neighbourhood of y contained in K, y is an interior point of
K.
*44e. Let X be a topological vector space that is of second Baire category with respect to itself. Suppose
a closed convex subset K of X has an internal point y. Let X
n
= ¦x ∈ X : y+tx ∈ C for all t ∈ [0, 1/n]¦.
Each X
n
is a closed subset of X since addition and scalar multiplication are continuous and C is closed.
Now X =
¸
X
n
so some X
n
has an interior point x. Then x ∈ O ⊂ X
n
for some open set O. Thus
y +
1
n
x ∈ y +
1
n
O ⊂ C so y +
1
n
x is an interior point of C.
(*) Assumption of convexity not necessary?
45. Let K be a convex set containing θ and suppose that x is an internal point of K. Since x is an
internal point, there exists ε > 0 such that x + µx ∈ K for [µ[ < ε. Choose λ such that 0 < λ < 1 and
(1 −λ)
−1
x ∈ K. Note that 0 < 1 −λ < 1. For y ∈ K, x +λy = (1 −λ)[(1 −λ)
−1
x] +λy ∈ K since K is
convex.
46. Let ^ be a family of convex sets (containing θ) in a vector space X. Suppose ^ satisﬁes (i), (ii)
and (iii). To show that the translates of sets in ^ form a base for a topology that makes X into a locally
convex topological vector space, it suﬃces to show that ^ is a base at θ. If N
1
, N
2
∈ ^, there is an
N
3
∈ ^ with N
3
⊂ N
1
∩ N
2
. Thus condition (i) of Proposition 14 is satisﬁed. If N ∈ ^ and x ∈ N,
then x is internal. By Q45, there exists λ with 0 < λ < 1 and x + λN ⊂ N. Note that λN ∈ ^.
Thus condition (ii) of Proposition 14 is satisﬁed. If N ∈ ^, then
1
2
N +
1
2
N ⊂ N. Note that
1
2
N ∈ ^.
Thus condition (iii) of Proposition 14 is satisﬁed. If N ∈ ^ and x ∈ X, then λx ∈ N for some λ ∈ R
since θ ∈ N and θ is internal. Now x ∈
1
λ
N so condition (iv) of Proposition 14 is satisﬁed. If N ∈ ^
and 0 < [α[ < 1, then αN ⊂ N since θinN and N is convex. Also, αN ∈ ^. Thus condition (v) of
69
Proposition 14 is satisﬁed. Hence ^ is a base at θ for a topology that makes X into a (locally convex)
topological vector space.
Conversely, if X is a locally convex topological vector space, then there is a base for the topology
consisting of convex sets. Consider the family ^ consisting of all sets N in the base containing θ
together with αN for 0 < [α[ < 1. Then ^ is a base at θ. If N ∈ ^, then each point of N is interior
and thus internal by Q. If N
1
, N
2
∈ ^, there is an N
3
∈ ^ with N
3
⊂ N
1
∩N
2
since N
1
∩N
2
is an open
set containing θ. Condition (iii) is satisﬁed by the deﬁnition of ^.
(*) Proof of Proposition 21.
*47.
48a. In L
p
[0, 1], 1 < p < ∞, suppose [[x[[ = 1. If x = λy + (1 − λ)z with [[y[[ ≤ 1 and [[z[[ ≤ 1, then
1 = [[x[[ = [[λy +(1−λ)z[[ ≤ λ[[y[[ +(1−λ)[[z[[ ≤ 1 so equality holds in the Minkowski inequality. Thus
λy and (1 −λ)z are collinear. i.e. λy = α(1 −λ)z for some α. Also, [[y[[ = [[z[[ = 1. Then y = z. Hence
x is an extreme point of the unit sphere S = ¦x : [[x[[ ≤ 1¦.
(*) Note that an extreme point of the unit sphere of any normed space must have norm one. Otherwise
if 0 < [[x[[ < 1, then x = [[x[[
x
]]x]]
+ (1 −[[x[[)θ so x cannot be an extreme point of the unit sphere.
*48b.
*48c. Let x ∈ L
1
[0, 1] with [[x[[ = 1. Choose t ∈ [0, 1] such that
t
0
[x(t)[ dt =
1
2
. Deﬁne y(s) = 2x(s)
if s ≤ t and y(s) = 0 otherwise. Also deﬁne z(s) = 2x(s) if s ≥ t and z(s) = 0 otherwise. Now
y, z ∈ L
1
[0, 1], [[y[[ = [[z[[ = 1 and x =
1
2
(y + z) so x is not an extreme point of the unit sphere. Since
any point with norm one in L
1
[0, 1] is not an extreme point of the unit sphere, the unit sphere has no
extreme points.
48d. If L
1
[0, 1] is the dual of some normed space, then its unit sphere is weak* compact and convex so
it has extreme points, contradicting part (c). Hence L
1
[0, 1] is not the dual of any normed space.
48e. For 1 < p < ∞, every 'x
i
` with [['x
i
`[[ = 1 is an extreme point of the unit sphere of
p
by a similar
argument as in part (a).
Let 'x
i
` ∈
1
with [['x
i
`[[ = 1. We may assume [x
1
[ ≥ [x
2
[ > 0. Let y
1
= (sgn x
1
)([x
1
[ − [x
2
[) and
y
2
= 2x
2
. Also, let z
1
= x
1
+ (sgn x
1
)[x
2
[ and z
2
= 0. Then y
1
+ z
1
= 2x
1
and y
2
+ z
2
= 2x
2
. Also,
[y
1
[+[y
2
[ = [x
1
[+[x
2
[ and [z
1
[+[z
2
[ ≤ [x
1
[+[x
2
[. For i > 2, deﬁne y
i
= z
i
= x
i
. Then 'x
i
` =
1
2
('y
i
`+'z
i
`)
with 'y
i
`, 'z
i
` ∈ S. Thus the unit sphere in
1
has no extreme points.
Let 'x
i
` ∈
∞
with [x
i
[ = 1 for all i. If 1 = x
i
= λy
i
+ (1 − λ)z
i
with [y
i
[ ≤ 1, [z
i
[ ≤ 1 and 0 < λ < 1,
then y
i
= z
i
= 1. Similarly, if −1 = x
i
= λy
i
+ (1 − λ)z
i
with [y
i
[ ≤ 1, [z
i
[ ≤ 1 and 0 < λ < 1, then
y
i
= z
i
= −1. Thus 'x
i
` is an extreme point of the unit sphere in
∞
. If [x
N
[ < 1 for some N, then deﬁne
y
i
= x
i
+
1−]x
i
]
2
and z
i
= x
i
−
1−]x
i
]
2
for all i so that 'x
i
` =
1
2
('y
i
` + 'z
i
`). Furthermore, 'y
i
`, 'z
i
` ∈ S.
Hence the extreme points of the unit sphere in
∞
are those 'x
i
` with [x
i
[ = 1 for all i.
*48f. The constant functions ±1 are extreme points of the unit sphere in C(X) where X is a compact
Hausdorﬀ space. Let f ∈ C(X) with [[f[[ = 1 and suppose [f(x
0
)[ < 1 for some x
0
∈ X. Fix ε > 0
such that 0 < [f(x
0
)[ − ε and [f(x
0
)[ + ε < 1. Let A = ¦x : [f(x)[ = [f(x
0
)[¦ and B = ¦x : [f(x)[ ∈
[0, f(x
0
) −ε] ∪ [f(x
0
) +ε, 1]¦. Then A and B are disjoint closed subsets of the compact Hausdorﬀ (and
thus normal) space X so by Urysohn’s Lemma, there exists g ∈ C(X) such that 0 ≤ g ≤ ε, g ≡ 0 on
B and g ≡ ε on A. Now [[f + g[[ ≤ 1, [[f − g[[ ≤ 1 and f =
1
2
[(f + g) + (f − g)]. Thus f is not an
extreme point of the unit sphere in C(X). Hence the extreme points of the unit sphere are those f where
[f(x)[ = 1 for all x ∈ X and by continuity, these are the constant functions ±1.
The only extreme points of the unit sphere in C[0, 1] are the constant functions ±1. If C[0, 1] is the dual of
some normed space, then its unit sphere is a weak* compact convex set so by the KreinMilman Theorem,
it is the closed convex hull of its extreme points and contains only constant functions. Contradiction.
Hence C[0, 1] is not the dual of any normed space.
49a. Let X be the vector space of all measurable realvalued functions on [0, 1] with addition and
scalar multiplication deﬁned in the usual way. Deﬁne σ(x) =
1
0
]x(t)]
1+]x(t)]
dt. Since
]x(t)+y(t)]
1+]x(t)+y(t)]
=
1 −
1
1+]x(t)+y(t)]
≤ 1 −
1
1+]x(t)]+]y(t)]
=
]x(t)]+]y(t)]
1+]x(t)]+]y(t)]
≤
]x(t)]
1+]x(t)]
+
]y(t)]
1+]y(t)]
, we have σ(x +y) ≤ σ(x) +σ(y).
By deﬁning ρ(x, y) = σ(x −y), we have a metric for X.
49b. Suppose x
n
→ x in measure. Given ε > 0, there exists N such that for all n ≥ N we have m¦t :
70
[x
n
(t) −x(t)[ ≥ ε/2¦ < ε/2. Then for n ≥ N, ρ(x
n
, x) =
1
0
]x
n
(t)−x(t)]
1+]x
n
(t)−x(t)]
dt ≤
¦t:]x
n
(t)−x(t)]≥ε/2¦
1 dt +
¦t:]x
n
(t)−x(t)]<ε/2¦
[x
n
(t) − x(t)[ dt < ε. Thus x
n
→ x in the metric ρ. Conversely, suppose 'x
n
`
does not converge to x in measure. There exists ε > 0 such that for all N there is n ≥ N with
m¦t : [x
n
(t) −x(t)[ ≥ ε¦ ≥ ε. Thus there is a subsequence 'x
n
k
` such that m¦t : [x
n
k
(t) −x(t)[ ≥ ε¦ ≥ ε
for all k. Now ρ(x
n
k
, x) =
1
0
]x
n
k
(t)−x(t)]
1+]x
n
k
(t)−x(t)]
dt ≥
¦]x
n
k
−x]≥ε¦
]x
n
k
(t)−x(t)]
1+]x
n
k
(t)−x(t)]
dt ≥
ε
2
1+ε
for all k so the
subsequence 'x
n
k
` does not converge to x in the metric ρ. Hence 'x
n
` does not converge to x in the
metric ρ.
49c. Let 'x
n
` be Cauchy in the metric ρ. Suppose 'x
n
` is not Cauchy in measure. There exists
ε > 0 such that for all N there is n, m ≥ N with m¦t : [x
n
(t) − x
m
(t)[ ≥ ε¦ ≥ ε. Thus there is
a subsequence 'x
n
k
` such that m¦t : [x
n
2k−1
(t) − x
n
2k
(t)[ ≥ ε¦ ≥ ε for all k. Now ρ(x
n
2k−1
, x
n
2k
) =
1
0
]x
n
2k−1
(t)−x
n
2k
(t)]
1+]x
n
2k−1
(t)−x
n
2k
(t)]
dt ≥
¦]x
n
2k−1
−x
n
2k
]≥ε¦
]x
n
2k−1
(t)−x
n
2k
(t)]
1+]x
n
2k−1
(t)−x
n
2k
(t)]
dt ≥
ε
2
1+ε
for all k so the subsequence
'x
n
k
` is not Cauchy in the metric ρ. Contradiction. Hence 'x
n
` is Cauchy in measure. By Q4.25, 'x
n
`
converges in measure and by part (b), it converges in the metric ρ. Hence X is a complete metric space.
49d. Given ε > 0, let δ = ε/2. When τ('x, y`, 'x
t
, y
t
`) < δ, we have ρ(x, x
t
) < δ and ρ(y, y
t
) < δ. Now
ρ(x +y, x
t
+y
t
) = σ(x −x
t
+y −y
t
) ≤ σ(x −x
t
) +σ(y −y
t
) = ρ(x, x
t
) +ρ(y, y
t
) < ε. Hence addition is
a continuous mapping of X X into X.
49e. Given a ∈ R, x ∈ X and ε > 0, by Q3.23a, there exists M such that [x[ ≤ M except on a set of
measure less than ε/3. Let δ = min(1, ε/3([a[ +1), ε/3M). When τ('a, x`, 'c, x
t
`) < δ, we have [c−a[ < δ
and ρ(x, x
t
) < δ. Now ρ(ax, cx
t
) =
1
0
]ax(t)−cx
(t)]
1+]ax(t)−cx
(t)]
dt ≤
1
0
]a−c] ]x(t)]
1+]a−c] ]x(t)]
dt +
1
0
]c] ]x(t)−x
(t)]
1+]c] ]x(t)−x
(t)]
dt <
Mδ +ε/3 + ([a[ +δ)δ < ε. Hence scalar multiplication is a continuous mapping of R X to X.
49f. Given x ∈ X and ε > 0, there is a step function s such that [x − s[ < ε/2 except on a set of
measure less than ε/2. i.e. m¦t : [x(t) − s(t)[ ≥ ε/2¦ < ε/2. Now ρ(x, s) =
1
0
]x(t)−s(t)]
1+]x(t)−s(t)]
dt =
¦]x−s]<ε/2¦
]x(t)−s(t)]
1+]x(t)−s(t)]
dt +
¦]x−s]≥ε/2¦
]x(t)−s(t)]
1+]x(t)−s(t)]
dt <
1
0
ε/2 dt +
¦]x−s]≥ε/2¦
1 dt < ε. Hence the
set of step functions is dense in X.
*49g.
49h. By parts (d) and (e), X is a topological vector space. Since there are no nonzero continuous linear
functionals on X by part (g), X cannot be locally convex.
49i. Let s be the space of all sequences of real numbers and deﬁne σ('ξ
v
`) =
¸
2
−v
]ξ
v
]
1+]ξ
v
]
.
Analogues of parts (a) and (c) follow from Q7.24. The analogue of part (d) follows from the same
argument as above. For the analogue of part (e), suppose a ∈ R, 'ξ
v
` ∈ s and ε > 0 are given.
Let δ = min(1, ε/2[a[, ε/2(σ('ξ
v
`) + 1)). When τ('a, 'ξ
v
``, 'c, 'η
v
``) < δ, we have [c − a[ < δ and
σ('ξ
v
` − 'η
v
`) < δ. Now σ(a'ξ
v
` − c'η
v
`) =
¸
2
−v
]aξ
v
−cη
v
]
1+]aξ
v
−cη
v
]
≤
¸
2
−v
]a] ]ξ
v
−η
v
]
1+]a] ]ξ
v
−η
v
]
+
¸
2
−v
]a−c] ]η
v
]
1+]a−c] ]η
v
]
=
[a[σ('ξ
v
` −'η
v
`) +[a −c[σ('η
v
`) < [a[δ +(σ('ξ
v
`) +δ)δ < ε. Hence scalar multiplication is a continuous
mapping of R s into s.
Let f be a continuous linear functional on s. For each v, let e
v
be the sequence where the vth entry
is 1 and all other entries are 0. Now any sequence 'ξ
v
` ∈ s can be expressed as
¸
ξ
v
e
v
so f('ξ
v
`) =
f(
¸
ξ
v
e
v
) =
¸
ξ
v
f(e
v
) since f is continuous and linear. Since the series converges for each sequence
'ξ
v
`, there exists N such that f(e
v
) = 0 for v > N. Thus f('ξ
v
`) =
¸
N
v=1
ξ
v
f(e
v
).
10.8 Hilbert space
50. Suppose x
n
→ x and y
n
→ y. Then there exists M such that [[x
n
[[ ≤ M for all n. Given ε > 0,
choose N such that [[y
n
− y[[ < ε/2M and [[x
n
− x[[ < ε/2[[y[[ for n ≥ N. Now [(x
n
, y
n
) − (x, y)[ ≤
[(x
n
, y
n
−y)[ +[(x
n
−x, y)[ ≤ [[x
n
[[ [[y
n
−y[[ +[[x
n
−x[[ [[y[[ < ε for n ≥ N. Hence (x
n
, y
n
) → (x, y).
*51a. The use of trigonometric identities shows that ¦
1
√
2π
,
cos vt
√
π
,
sin vt
√
π
¦ is an orthonormal system for
L
2
[0, 2π]. Suppose x ∈ L
2
[0, 2π] such that (x,
cos vt
√
π
) = 0 and (x,
sin vt
√
π
) = 0 for all v. Let ε > 0. By Propo
sition 6.8 and Q9.42, there is a ﬁnite Fourier series ϕ = a
0
+
¸
N
n=1
a
n
cos nt +
¸
N
n=1
b
v
sin nt such that
[[x−ϕ[[ < ε. Now [[ϕ[[
2
= (ϕ, ϕ) = 2πa
2
0
+π
¸
N
n=1
(a
2
n
+b
2
n
) and
¸
v
[(ϕ, ϕ
v
)[
2
= 2πa
2
0
+
¸
N
n=1
((
a
n
√
π
π)
2
+
(
b
n
√
π
π)
2
) so [[ϕ[[
2
=
¸
v
[(ϕ, ϕ
v
)[
2
. Thus there exists M such that [ [[ϕ[[
2
−
¸
n
v=−n
[(ϕ, ϕ
v
)[
2
[ < ε
2
71
for n ≥ M. Now [[ϕ −
¸
n
v=−n
(ϕ, ϕ
v
)ϕ
v
[[
2
= [[ϕ[[
2
−
¸
n
v=−n
[(ϕ, ϕ
v
)[
2
< ε
2
for n ≥ M. Also,
[[
¸
n
v=−n
(ϕ, ϕ
v
)ϕ
v
−
¸
n
v=−n
(x, ϕ
v
)ϕ
v
[[
2
=
¸
n
v=−n
[(ϕ − x, ϕ
v
)[
2
≤ [[ϕ − x[[
2
< ε
2
for n ≥ M. Thus
[[x −
¸
n
v=−n
(x, ϕ
v
)ϕ
v
[[ ≤ [[x −ϕ[[ +[[ϕ−
¸
n
v=−n
(ϕ, ϕ
v
)ϕ
v
[[ +[[
¸
n
v=−n
(ϕ−x, ϕ
v
)ϕ
v
[[ < 3ε for n ≥ M.
Hence x =
¸
v
(x, ϕ
v
)ϕ
v
and [[x[[
2
=
¸
v
[(x, ϕ
v
)[
2
= 0 so x = 0. Hence ¦
1
√
2π
,
cos vt
√
π
,
sin vt
√
π
¦ is a complete
orthonormal system for L
2
[0, 2π].
51b. By part (a) and Proposition 27, every function in L
2
[0, 2π] is the limit in mean (of order 2) of its
Fourier series. i.e. the Fourier series converges to the function in L
2
[0, 2π].
52a. Let x ∈ H and let ¦ϕ
v
¦ be an orthonormal system. Note that ¦v : a
v
= 0¦ =
¸
n
¦v : a
v
≥ 1/n¦.
By Bessel’s inequality, we have
¸
v
[a
v
[
2
≤ [[x[[
2
< ∞ so ¦v : a
v
≥ 1/n¦ is ﬁnite for each n. Hence
¦v : a
v
= 0¦ is countable.
*52b. Let H be a Hilbert space and ¦ϕ
v
¦ a complete orthonormal system. By part (a), only countably
many Fourier coeﬃcients a
v
are nonzero so we may list them as a sequence 'a
v
`. By Bessel’s inequality,
¸
v
[a
v
[
2
≤ [[x[[
2
< ∞ so given ε > 0, there exists N such that
¸
∞
v=N+1
[a
v
[
2
< ε
2
. For m > n ≥ N, we
have [[
¸
m
v=n
a
v
ϕ
v
[[
2
=
¸
m
v=n
[a
2
v
< ε
2
so the sequence of partial sums is Cauchy in H and thus converges
in H. i.e.
¸
v
a
v
ϕ
v
∈ H. Now (x −
¸
v
a
v
ϕ
v
, ϕ
v
) = (x, ϕ
v
) − a
v
= 0 for all v so x −
¸
v
a
v
ϕ
v
= 0. i.e.
x =
¸
v
a
v
ϕ
v
.
If a complete orthonormal system in H is countable, say 'ϕ
v
`, then x =
¸
v
a
v
ϕ
v
and x is a cluster point
of the set of linear combinations of ϕ
v
, which contains the countable dense set of linear combinations of
ϕ
v
with rational coeﬃcients so H is separable. Thus a complete orthonormal system in a nonseparable
Hilbert space is uncountable.
*52c. Let f be a bounded linear functional on H. Let K = ker f. Since f is continuous, K is a closed
linear subspace of H. We may assume K
⊥
= ¦0¦ so there exists x
0
∈ K
⊥
with f(x
0
) = 1. Deﬁne
y = x
0
/[[x
0
[[
2
. Then 0 = (x −f(x)x
0
, x
0
) = (x, x
0
) −f(x)[[x
0
[[
2
for all x ∈ H. i.e. f(x) = (x, y) for all
x ∈ H. If y
t
∈ H such that (x, y) = (x, y
t
) for all x ∈ H, then y − y
t
∈ H
⊥
. In particular, [[y − y
t
[[
2
=
(y −y
t
, y −y
t
) = 0 so y = y
t
. This proves the uniqueness of y. Since [f(x)[ = [(x, y)[ ≤ [[x[[ [[y[[, we have
[[f[[ ≤ [[y[[. Furthermore, since f(y/[[y[[) = (y/[[y[[, y) = [[y[[, we have [[f[[ = [[y[[.
52d. Let H be an inﬁnite dimensional Hilbert space. If ¦ϕ
v
¦ is a complete orthonormal system in
H, then the set of ﬁnite linear combinations of ϕ
v
is a dense subset of H. Now if [¦ϕ
v
¦[ = n, then
[¦ﬁnite linear combinations of ϕ
v
¦[ = nℵ
0
= n so there is a dense subset of H with cardinality n. If
S is a dense subset of H, then for each ϕ
v
, there exists x
v
∈ S with [[x
v
− ϕ
v
[[ < 1/
√
2. If v = u,
then [[x
v
− x
u
[[ ≥ [[ϕ
v
− ϕ
u
[[ − [[x
v
− ϕ
v
[[ − [[x
u
− ϕ
u
[[ >
√
2 − 1/
√
2 − 1/
√
2 = 0 so x
v
= x
u
. Thus
[S[ ≥ [¦ϕ
v
¦[. Hence the number of elements in a complete orthonormal system in H is the smallest
cardinal n such that there is a dense subset of H with n elements. Furthermore, this proves that every
complete orthonormal system in H has the same number of elements.
*52e. Suppose two Hilbert spaces H and H
t
are isomorphic with an isomorphism Φ : H → H
t
. If
¦ϕ
v
¦ is a complete orthonormal system in H, then ¦Φ(ϕ
v
)¦ is a complete orthonormal system in H
t
.
Similarly for Φ
−1
. Thus dimH = dimH
t
. Conversely, suppose dimH = dimH
t
. Let c be a complete
orthonormal system in H. Consider the Hilbert space
2
(c) = ¦(f : c → R) :
¸
e∈L
[f(e)[
2
< ∞¦. If
x ∈ H, deﬁne ˆ x : c →R by ˆ x(e) = (x, e). Then
¸
e∈L
[ˆ x(e)[
2
=
¸
e∈L
[(x, e)[
2
= [[x[[
2
< ∞ so ˆ x ∈
2
(c).
Furthermore, [[x[[ = [[ˆ x[[. Deﬁne Φ : H →
2
(c) by Φ(x) = ˆ x. Then Φ is a linear isometry. Now the
range of Φ contains functions f such that f(e) = 0 for all but ﬁnitely many e ∈ c and is closed since Φ
is an isometry. Hence Φ is an isomorphism. If T is a complete orthonormal system in H
t
, then [c[ = [T[
so
2
(c) and
2
(T) must be isomorphic. Hence H and H
t
are isomorphic.
*52f. Let A be a nonempty set and let
2
(A) = ¦(f : A → R) :
¸
a∈A
[f(a)[
2
< ∞¦. Then
2
(A) is a
Hilbert space with (f, g) =
¸
a∈A
f(a)g(a). For each a ∈ A, let χ
a
be the characteristic function of ¦a¦.
Then χ
a
∈
2
(A) for each a ∈ A. Furthermore ¦χ
a
: a ∈ A¦ is a complete orthonormal system in
2
(A).
Thus dim
2
(A) = [¦χ
a
: a ∈ A¦[ = [A[. Hence there is a Hilbert space of each dimension.
53a. Let P be a subset of H. Suppose y
n
∈ P
⊥
and y
n
→ y. Then for each x ∈ P, we have
(x, y
n
) → (x, y). Since (x, y
n
) = 0 for each n, we have (x, y) = 0 so y ∈ P
⊥
. Thus P
⊥
is closed. If
a, b ∈ R and y, z ∈ P
⊥
, then for each x ∈ P, we have (ay +bz, x) = a(y, x) +b(z, x) = 0 so ay +bz ∈ P
⊥
and P
⊥
is a linear manifold.
*53b. If x ∈ P, then (x, y) = 0 for all y ∈ P
⊥
. Thus P
⊥⊥
is a closed linear manifold containing
P. Suppose P ⊂ Q where Q is a closed linear manifold. Then Q
⊥
⊂ P
⊥
and P
⊥⊥
⊂ Q
⊥⊥
. Now
72
if Q is a proper subset of Q
⊥⊥
, pick x ∈ Q
⊥⊥
` Q. Then there exists y
0
∈ Q such that [[x − y
0
[[ =
inf¦[[x − y[[ : y ∈ Q¦ = δ. Let z = x − y
0
. For any a ∈ R and y
1
∈ Q, we have δ
2
≤ [[z − ay
1
[[
2
=
[[z[[
2
− 2a(z, y
1
) + a
2
[[y
1
[[
2
= δ
2
− 2a(z, y
1
) + a
2
[[y
1
[[
2
so a
2
[[y
1
[[
2
− 2a(z, y
1
) ≥ 0. Thus 4(z, y
1
)
2
≤ 0
and it follows that (z, y
1
) = 0. i.e. z⊥Q. Thus z ∈ Q
⊥⊥
∩ Q
⊥
so z = 0 i.e. x = y
0
∈ Q. Contradiction.
Hence Q = Q
⊥⊥
so P
⊥⊥
⊂ Q.
53c. Let M be a closed linear manifold. Given x ∈ H, there exists y ∈ M such that x − y⊥M. Then
x = y + z where z = x − y ∈ M
⊥
. If x = y + z = y
t
+ z
t
where y, y
t
∈ M and z, z
t
∈ M
⊥
, then
y −y
t
= z −z
t
so y −y
t
, z −z
t
∈ M∩M
⊥
. Hence y = y
t
and z = z
t
. Furthermore [[x[[
2
= (y +z, y +z) =
[[y[[
2
+[[z[[
2
+ 2(y, z) = [[y[[
2
+[[z[[
2
since z⊥y.
54. Let 'x
n
` be a bounded sequence of elements in a separable Hilbert space H. Suppose [[x
n
[[ ≤ M for
all n and let 'ϕ
n
` be a complete orthonormal system in H. Now [(x
n
, ϕ
1
)[ ≤ M for all n so there is a sub
sequence '(x
n
k
, ϕ
1
)` that converges. Then [(x
n
k
, ϕ
2
)[ ≤ M for all k so there is a subsequence '(x
n
k
l
, ϕ
2
)`
that converges. Furthermore, '(x
n
k
l
, ϕ
1
)` also converges. Continuing the process, we obtain the diagonal
sequence 'x
n
n
` such that '(x
n
n
, ϕ
k
)` converges for any k. For any bounded linear functional f on H,
there is a unique y ∈ H such that f(x) = (x, y) for all x ∈ H. Furthermore, y =
¸
k
(y, ϕ
k
)ϕ
k
. Now
(x
n
n
, y) = (x
n
n
,
¸
k
(y, ϕ
k
)ϕ
k
) =
¸
k
(x
n
n
, ϕ
k
)(y, ϕ
k
). Since '(x
n
n
, ϕ
k
)` converges for each k, '(x
n
n
, y)`
converges. i.e. 'x
n
` has a subsequence which converges weakly.
55. Let S be a subspace of L
2
[0, 1] and suppose that there is a constant K such that [f(x)[ ≤ K[[f[[
for all x ∈ [0, 1]. If 'f
1
, . . . , f
n
` is any ﬁnite orthonormal sequence in S, then for any a
1
, . . . , a
n
∈ R,
we have (
¸
n
i=1
a
i
f
i
(x))
2
≤ K
2
[[
¸
n
i=1
a
i
f
i
[[
2
= K
2
(
¸
n
i=1
a
i
f
i
,
¸
n
i=1
a
i
f
i
) = K
2
¸
n
i=1
a
2
i
for all x ∈ [0, 1].
Fix x ∈ [0, 1]. For each i, let a
i
=
f
i
(x)
√
n
i=1
f
i
(x)
2
. Then (
¸
n
i=1
a
i
f
i
(x))
2
=
¸
n
i=1
f
i
(x)
2
and
¸
n
i=1
a
2
i
= 1.
Thus
¸
n
i=1
f
i
(x)
2
≤ K
2
for all x ∈ [0, 1] and n =
¸
n
i=1
[[f
i
[[
2
=
1
0
(
¸
n
i=1
f
i
(x)
2
) ≤
1
0
K
2
= K
2
. Hence
the dimension of S is at most K
2
.
11 Measure and Integration
11.1 Measure spaces
1. Let ¦A
n
¦ be a collection of measurable sets. Let B
1
= A
1
and B
k
= A
k
`
¸
k−1
n=1
A
n
for k > 1.
Then ¦B
n
¦ is a collection of pairwise disjoint measurable sets such that
¸
A
n
=
¸
B
n
. Now µ(
¸
A
k
) =
µ(
¸
B
k
) =
¸
µ(B
k
) = lim
n
¸
n
k=1
µ(B
k
) = lim
n
µ(
¸
n
k=1
B
k
) = lim
n
µ(
¸
n
k=1
A
k
).
2a. Let ¦(X
α
, B
α
, µ
α
)¦ be a collection of measure spaces, and suppose that the sets ¦X
α
¦ are disjoint.
Deﬁne X =
¸
X
α
, B = ¦B : ∀α B ∩ X
α
∈ B
α
¦ and µ(B) =
¸
µ
α
(B ∩ X
α
). Now ∅ ∈ B since
∅ ∩ X
α
= ∅ ∈ B
α
for all α. If B ∈ B, then B ∩ X
α
∈ B
α
for all α so B
c
∩ X
α
= X
α
` (B ∩ X
α
) ∈ B
α
for all α. Thus B
c
∈ B. If 'B
n
` is a sequence in B, then for each n, B
n
∩ X
α
∈ B
α
for all α so
¸
B
n
∩ X
α
=
¸
(B
n
∩ X
α
) ∈ B
α
for each α. Thus
¸
B
n
∈ B. Hence B is a σalgebra.
2b. µ(∅) =
¸
µ
α
(∅ ∩ X
α
) =
¸
µ
α
(∅) = 0. For any sequence of disjoint sets B
i
∈ B, we have
µ(
¸
B
i
) =
¸
µ
α
(
¸
B
i
∩ X
α
) =
¸
α
¸
i
µ
α
(B
i
∩ X
α
) =
¸
i
¸
α
µ
α
(B
i
∩ X
α
) =
¸
i
µ(B
i
). Hence µ is a
measure.
2c. Suppose that all but a countable number of the µ
α
are zero and the remainder are σﬁnite. Let
'µ
n
` be the countably many µ
α
that are nonzero and σﬁnite. Then for each n, X
n
=
¸
k
X
n,k
where
X
n,k
∈ B
n
and µ
n
(X
n,k
) < ∞. In particular, X
n,k
∈ B for all n, k. Let A = ¦α : µ
α
is zero¦. Then
X =
¸
n
X
n
∪
¸
α∈A
X
α
=
¸
n,k
X
n,k
∪
¸
α∈A
X
α
. Since
¸
α∈A
X
α
∩ X
α
= X
α
∈ B
α
if α ∈ A and
¸
α∈A
X
α
∩ X
α
= ∅ ∈ B
α
if α / ∈ A, we have
¸
α∈A
X
α
∈ B. Also, µ(
¸
α∈A
X
α
) =
¸
µ
α
(
¸
α∈A
∩X
α
) =
¸
α∈A
µ
α
X
α
= 0 < ∞. Hence µ is σﬁnite.
Conversely, suppose µ is σﬁnite. Then X =
¸
n
Y
n
where Y
n
∈ B and µ(Y
n
) < ∞ for each n. We
may assume that the Y
n
are disjoint. Now for each n,
¸
µ
α
(Y
n
∩ X
α
) = µ(Y
n
∩
¸
α
X
α
) = µ(Y
n
) < ∞
so ¦α : µ
α
(Y
n
∩ X
α
) > 0¦ is countable. Thus
¸
n
¦α : µ
α
(Y
n
∩ X
α
) > 0¦ is countable. For each α,
µ
α
(X
α
) = µ
α
(X
α
∩
¸
n
Y
n
) =
¸
n
µ
α
(X
α
∩ Y
n
). If µ
α
(X
α
) > 0, then µ
α
(X
α
∩ Y
n
) > 0 for some n.
Hence ¦α : µ
α
(X
α
) > 0¦ is countable. i.e. all but a countable number of the µ
α
are zero. Furthermore,
X
α
=
¸
n
(X
α
∩ Y
n
) where X
α
∩ Y
n
∈ B
α
and µ
α
(X
α
∩ Y
n
) ≤
¸
µ
α
(X
α
∩ Y
n
) = µ(Y
n
) < ∞ for each n.
Hence the remaining µ
α
are σﬁnite.
73
(*) Union of measure spaces
3a. Suppose E
1
, E
2
∈ B and µ(E
1
∆E
2
) = 0. Then µ(E
1
` E
2
) = µ(E
2
` E
1
) = 0. Hence µ(E
1
) =
µ(E
1
` E
2
) +µ(E
1
∩ E
2
) = µ(E
2
` E
1
) +µ(E
1
∩ E
2
) = µ(E
2
).
3b. Suppose µ is complete, E
1
∈ B and µ(E
1
∆E
2
) = 0. Then E
2
` E
1
∈ B since E
2
` E
1
⊂ E
1
∆E
2
.
Also, E
1
∩ E
2
= E
1
` (E
1
` E
2
) ∈ B. Hence E
2
= (E
2
` E
1
) ∪ (E
1
∩ E
2
) ∈ B.
4. Let (X, B, µ) be a measure space and Y ∈ B. Let B
Y
consist of those sets of B that are contained
in Y . Set µ
Y
E = µE if E ∈ B
Y
. Clearly ∅ ∈ B
Y
. If B ∈ B
Y
, then B ∈ B and B ⊂ Y . Thus Y ` B ∈ B
and Y ` B ⊂ Y so Y ` B ∈ B
Y
. If 'B
n
` is a sequence of sets in B
Y
, then B
n
∈ B and B
n
⊂ Y for all n.
Thus
¸
B
n
∈ B and
¸
B
n
⊂ Y so
¸
B
n
∈ B
Y
. Furthermore, µ
Y
(∅) = µ(∅) = 0 and if 'B
n
` is a sequence
of disjoint sets in B
Y
, then µ
Y
(
¸
B
n
) = µ(
¸
B
n
) =
¸
µ(B
n
) =
¸
µ
Y
(B
n
). Hence (Y, B
Y
, µ
Y
) is a
measure space.
(*) Restriction of measure to measurable subset
5a. Let (X, B) be a measurable space. Suppose µ and ν are measures deﬁned on B and deﬁne the set
function λE = µE +νE on B. Then λ(∅) = µ(∅) +ν(∅) = 0. Also, if 'E
n
` is a sequence of disjoint sets
in B, then λ(
¸
E
n
) = µ(
¸
E
n
) +ν(
¸
E
n
) =
¸
µE
n
+
¸
νE
n
=
¸
λE
n
. Hence λ is also a measure.
*5b. Suppose µ and ν are measures on B and µ ≥ ν. Note that µ −ν is a measure when restricted to
measurable sets with ﬁnite νmeasure. Deﬁne λ(E) = sup
F⊂E,ν(F)<∞
(µ(F)−ν(F)). Clearly λ(∅) = 0. If
E
1
, E
2
∈ Bwith E
1
∩E
2
= ∅, then for any F ⊂ E
1
∪E
2
with ν(F) < ∞, we have µ(F)−ν(F) = (µ−ν)(F∩
E
1
)+(µ−ν)(F ∩E
2
) ≤ λ(E
1
)+λ(E
2
). Thus λ(E
1
∪E
2
) ≤ λ(E
1
)+λ(E
2
). On the other hand, if F
1
⊂ E
1
and F
2
⊂ E
2
with ν(F
1
), ν(F
2
) < ∞, we have (µ−ν)(F
1
) +(µ−ν)(F
2
) = (µ−ν)(F
1
∪F
2
) ≤ λ(E
1
∪E
2
).
Thus λ(E
1
) +λ(E
2
) ≤ λ(E
1
∪ E
2
). Hence λ is ﬁnitely additive.
Now suppose 'E
n
` is a sequence of disjoint sets in B. If F ⊂
¸
E
n
with ν(F) < ∞, then (µ −ν)(F) =
¸
(µ − ν)(F ∩ E
n
) ≤
¸
λ(E
n
). Thus λ(
¸
E
n
) ≤
¸
λ(E
n
). On the other hand, for any N, we have
¸
N
n=1
λ(E
n
) = λ(
¸
N
n=1
E
n
) ≤ λ(
¸
E
n
). Thus
¸
λE
n
≤ λ(
¸
E
n
). Hence λ is countably additive.
If ν(E) = ∞, then µ(E) = ∞ so ν(E) + λ(E) = µ(E) and if ν(E) < ∞, then ν(E) + λ(E) = ν(E) +
(µ(E) −ν(E)) = µ(E). Hence µ = ν +λ.
5c. Suppose µ = ν + η for some measure η. If ν is ﬁnite, then η = µ − ν = λ. Suppose ν is σ
ﬁnite. Then X =
¸
X
n
where X
n
∈ B and ν(X
n
) < ∞ for each n. We may assume that the X
n
are disjoint. Now for any set E ∈ B, we have η(E ∩ X
n
) = λ(E ∩ X
n
) since ν(E ∩ X
n
) < ∞. Then
η(E) = η(
¸
(E ∩ X
n
)) =
¸
η(E ∩ X
n
) =
¸
λ(E ∩ X
n
) = λ(
¸
(E ∩ X
n
)) = λ(E). Hence λ in part (b) is
the unique such measure.
*5d. If µ = ν + λ = ν + λ
t
, then λ(F) = λ
t
(F) for any set F with ν(F) < ∞. Given E ∈ B, for any
F ⊂ E with ν(F) < ∞, we have µ(F) − ν(F) = λ(F) = λ
t
(F) ≤ λ
t
(E). Hence λ(E) ≤ λ
t
(E) and λ in
part (b) is the smallest such measure.
6a. Let µ be a σﬁnite measure so X =
¸
X
n
where µ(X
n
) < ∞ for each n. For any measurable set E
of inﬁnite measure, we have ∞ = µ(
¸
(E ∩ X
n
)) = lim
k
µ(
¸
k
n=1
(E ∩ X
n
)). Thus for any N, there exists
k such that N < µ(
¸
k
n=1
(E ∩ X
n
)) < ∞. Hence µ is semiﬁnite.
6b. Given a measure µ, deﬁne µ
1
(E) = sup
F⊂E,µ(F)<∞
µ(F). Then µ
1
(∅) = µ(∅) = 0. If 'E
n
` is
a sequence of disjoint measurable sets, then for any F ⊂
¸
E
n
with µ(F) < ∞, we have µ(F) =
¸
µ(F ∩ E
n
) ≤
¸
µ
1
(E
n
). Thus µ
1
(
¸
E
n
) ≤
¸
µ
1
(E
n
). Conversely, if F
1
⊂ E
1
and F
2
⊂ E
2
with
µ(F
1
), µ(F
2
) < ∞, then µ(F
1
) +µ(F
2
) = µ(F
1
∪F
2
) ≤ µ
1
(E
1
∪E
2
) so µ
1
(E
1
) +µ
1
(E
2
) ≤ µ
1
(E
1
∪E
2
). It
follows that for any k,
¸
k
n=1
µ
1
(E
n
) ≤ µ
1
(
¸
k
n=1
E
n
) ≤ µ
1
(
¸
E
n
). Thus
¸
µ
1
(E
n
) ≤ µ
1
(
¸
E
n
). Hence
µ
1
is a measure. Furthermore, if µ
1
(E) = ∞, then for any N, there exists F ⊂ E with N < µ(F) < ∞.
Then N < µ
1
(F) < ∞. Hence µ
1
is semiﬁnite.
Now deﬁne µ
2
(E) = sup
F⊂E,µ
1
(F)<∞
(µ(F) − µ
1
(F)). Then µ
2
is a measure (c.f. Q5b). If µ(F) < ∞
for all F ⊂ E with µ
1
(F) < ∞, then µ(F) = µ
1
(F) for all F ⊂ E with µ
1
(F) < ∞ so µ
2
(E) = 0. If
µ(F) = ∞ for some F ⊂ E with µ
1
(F) < ∞, then µ(F) − µ
1
(F) = ∞ so µ
2
(E) = ∞. Hence µ
2
only
assumes the values 0 and ∞.
*6c.
7. Given a measure space (X, B, µ), let B
0
be the collection of sets A∪ B where A ∈ B and B ⊂ C for
some C ∈ B with µ(C) = 0. Clearly ∅ ∈ B
0
. If 'E
n
` is a sequence in B
0
, then E
n
= A
n
∪ B
n
for each
74
n where A
n
∈ B and B
n
⊂ C
n
for some C
n
∈ B with µ(C
n
) = 0. Now
¸
E
n
=
¸
A
n
∪
¸
B
n
where
¸
A
n
∈ B and
¸
B
n
⊂
¸
C
n
with µ(
¸
C
n
) = 0. If E ∈ B
0
, then E = A ∪ B where A ∈ B and B ⊂ C
for some C ∈ B with µ(C) = 0. Then E
c
= A
c
∩ B
c
= A
c
∩ (C ` (C ` B))
c
= A
c
∩ (C
c
∪ (C ` B)) =
(A
c
∩ C
c
) ∪ (A
c
∩ (C ` B)) where A
c
∩ C
c
∈ B and A
c
∩ (C ` B) ⊂ C with µ(C) = 0. Hence B
0
is a
σalgebra. Furthermore B ⊂ B
0
.
If E ∈ B
0
with E = A ∪ B = A
t
∪ B
t
with A, A
t
∈ B, B ⊂ C and B
t
⊂ C
t
for some C, C
t
∈ B
with µ(C) = µ(C
t
) = 0, then A ∪ B ∪ C ∪ C
t
= A
t
∪ B
t
∪ C ∪ C
t
so A ∪ C ∪ C
t
= A
t
∪ C ∪ C
t
.
Now µ(A) ≤ µ(A ∪ C ∪ C
t
) ≤ µ(A) + µ(C) + µ(C
t
) = µ(A). Thus µ(A) = µ(A ∪ C ∪ C
t
). Similarly,
µ(A
t
) = µ(A
t
∪ C ∪ C
t
). Hence µ(A) = µ(A
t
).
For E ∈ B
0
, deﬁne µ
0
(E) = µ(A). Then µ
0
is welldeﬁned. Clearly, µ
0
(∅) = 0. If 'E
n
` is a sequence of
disjoint sets in B
0
, then
¸
E
n
=
¸
A
n
∪
¸
B
n
so µ
0
(
¸
E
n
) = µ(
¸
A
n
) =
¸
µ(A
n
) =
¸
µ
0
(E
n
). Thus
µ
0
is a measure. Furthermore, if E ∈ B, then µ
0
(E) = µ(E).
If µ
0
(A∪B) = 0 and E ⊂ A∪B, then E = A∪(E∩A) where E∩A ⊂ B ⊂ C with µ(C) = 0 so E ∈ B
0
.
Hence (X, B
0
, µ
0
) is a complete measure space extending (X, B, µ).
8a. Let µ be a σﬁnite measure. If E is locally measurable, then E ∩ B ∈ B for each B ∈ B with
µ(B) < ∞. Now X =
¸
X
n
with X
n
∈ B and µ(X
n
) < ∞ for all n. Thus E =
¸
(E ∩ X
n
) where
E ∩ X
n
∈ B for all n. Thus E ∈ B. i.e. E is measurable. Hence µ is saturated.
8b. Let C be the collection of locally measurable sets. Clearly ∅ ∈ C. If C ∈ C, then C ∩B ∈ B for each
B ∈ B with µ(B) < ∞. Now for any such set B, C
c
∩ B = B ` (C ∩ B) ∈ B. Thus C
c
∈ C. If 'C
n
`
is a sequence of sets in C, then C
n
∩ B ∈ B for each B ∈ B with µ(B) < ∞. Now for any such set B,
¸
C
n
∩ B =
¸
(C
n
∩ B) ∈ B. Thus
¸
C
n
∈ C. Hence C is a σalgebra.
8c. Let (X, B, µ) be a measure space and C the σalgebra of locally measurable sets. For E ∈ C, set
¯ µE = µE if E ∈ B and ¯ µE = ∞ if E / ∈ B. Clearly ¯ µ(∅) = 0. If 'E
n
` is a sequence of disjoint sets in
C, then we consider a few cases. If
¸
E
n
/ ∈ B, then ¯ µ(
¸
E
n
) = ∞. Furthermore, E
n
/ ∈ B for some n so
¸
¯ µ(E
n
) = ∞ = ¯ µ(
¸
E
n
). Now suppose
¸
E
n
∈ B. If µ(
¸
E
n
) < ∞, then E
n
= E
n
∩
¸
E
n
∈ B for
each n so ¯ µ(
¸
E
n
) = µ(
¸
E
n
) =
¸
µ(E
n
) =
¸
¯ µ(E
n
). If µ(
¸
E
n
) = ∞, then either E
n
∈ B for all n, so
¸
¯ µ(E
n
) =
¸
µ(E
n
) = µ(
¸
E
n
) = ∞ = ¯ µ(
¸
E
n
), or E
n
/ ∈ B for some n, so
¸
¯ µ(E
n
) = ∞ = ¯ µ(
¸
E
n
).
Hence ¯ µ is a measure on C.
Let E be a locally measurable set in (X, C, ¯ µ). Then E ∩ C ∈ C for any C ∈ C with ¯ µ(C) < ∞. i.e.
E ∩ C ∈ C for any C ∈ B with µ(C) < ∞. Thus E ∩ C ∈ B for any C ∈ B with µ(C) < ∞ and E is a
locally measurable set in (X, B, µ) so E ∈ C. Hence (X, C, ¯ µ) is a saturated measure space.
*8d. Suppose µ is semiﬁnite and E ∈ C. Set µ(E) = sup¦µ(B) : B ∈ B, B ⊂ E¦. Clearly ¯ µ(∅) = 0.
If 'E
n
` is a sequence of disjoint sets in C, for any B ∈ B with B ⊂
¸
E
n
, provided µ(B) < ∞, we
have µ(B) = µ(
¸
(B ∩ E
n
)) =
¸
µ(B ∩ E
n
) ≤
¸
µ(E
n
). On the other hand, if µ(B) = ∞, then
µ(
¸
E
n
) = ∞. Thus µ(
¸
E
n
) ≤
¸
µ(E
n
). If E
1
, E
2
∈ C with E
1
∩ E
2
= ∅ and B
1
, B
2
∈ B with
B
1
⊂ E
1
and B
2
⊂ E
2
, then µ(E
1
) +µ(E
2
) ≤ µ(B
1
) +µ(B
2
) = µ(B
1
∪ B
2
) ≤ µ(E
1
∪ E
2
). Now for any
N, we have
¸
N
n=1
µ(E
n
) ≤ µ(
¸
N
n=1
E
n
) ≤ µ(
¸
E
n
). Thus
¸
µ(E
n
) ≤ µ(
¸
E
n
). Hence µ is a measure
on C.
Let E be a locally measurable set in (X, C, µ). Then E∩C ∈ C for any C ∈ C with µ(C) < ∞. If B ∈ B
with µ(B) < ∞, then B ∈ C with µ(B) = µ(B) < ∞. Thus E ∩ B ∈ C. It follows that E ∩ B ∈ B so
E ∈ C. Hence (X, C, µ) is a saturated measure space. Furthermore, µ is an extension of µ.
9a. Let R be a σring that is not a σalgebra. Let B be the smallest σalgebra containing R and set
R
t
= ¦E : E
c
∈ R¦. Note that ∅ ∈ R ⊂ R∪ R
t
. If A ∈ R∪ R
t
, then either A ∈ R so A
c
∈ R
t
⊂ R∪ R
t
or A ∈ R
t
so A
c
∈ R ⊂ R∪R
t
. If 'A
n
` is a sequence in R∪R
t
, then
¸
A
n
=
¸
¦A
n
: A
n
∈ R¦ ∪
¸
¦A
n
:
A
n
∈ R
t
¦ ∈ R∪R
t
. If A is a σalgebra containing R, then A also contains R
t
. Thus A ⊃ R∪R
t
. Hence
R∪ R
t
= B. Furthermore, if E ∈ R∩ R
t
, then E, E
c
∈ R so X = E ∪ E
c
∈ R and R is a σalgebra.
Contradiction. Hence R∩ R
t
= ∅.
9b. If µ is a measure on R, deﬁne ¯ µ on B by ¯ µ(E) = µ(E) if E ∈ R and ¯ µ(E) = ∞ if E ∈ R
t
. Clearly
¯ µ(∅) = µ(∅) = 0. Let 'E
n
` be a sequence of disjoint sets in B. If E
1
∈ R and E
2
∈ R
t
, then (E
2
)
c
∈ R
so (E
1
)
c
∩ (E
2
)
c
= (E
2
)
c
` E
1
∈ R. Thus E
1
∪ E
2
∈ R
t
so ¯ µ(E
1
∪ E
2
) = ∞ = ¯ µ(E
1
) + ¯ µ(E
2
). If E
n
∈ R
for all n, then
¸
E
n
∈ R. Similarly, if E
n
∈ R
t
for all n, then
¸
E
n
∈ R
t
. In both cases we have
¯ µ(
¸
E
n
) =
¸
¯ µ(E
n
). Thus in general, ¯ µ(
¸
E
n
) = ¯ µ(
¸
¦E
n
: E
n
∈ R¦ ∪
¸
¦E
n
: E
n
∈ R
t
¦) = ¯ µ(
¸
¦E
n
:
E
n
∈ R¦) + ¯ µ(
¸
¦E
n
: E
n
∈ R
t
¦) =
¸
E
n
∈R
¯ µ(E
n
) +
¸
E
n
∈R
¯ µ(E
n
) =
¸
¯ µ(E
n
). Hence ¯ µ is a measure
75
on B.
9c. Deﬁne µ on B by µ(E) = µ(E) if E ∈ R and µ(E) = sup¦µ(A) : A ⊂ E, A ∈ R¦ if E ∈ R
t
. Clearly
µ(∅) = µ(∅) = 0. Let 'E
n
` be a sequence of disjoint sets in B. Note that if E
n
∈ R for all n or E
n
∈ R
t
for all n, then µ(
¸
E
n
) =
¸
µ(E
n
). It follows that in general µ(
¸
E
n
) =
¸
µ(E
n
). Hence µ is a measure
on B.
*9d.
*9e.
11.2 Measurable functions
10. Let f be a nonnegative measurable function. For each pair 'n, k` of integers, set E
n,k
= ¦x : k2
−n
≤
f(x) < (k + 1)2
−n
¦ and set ϕ
n
= 2
−n
¸
2
2n
k=0
kχ
E
n,k
+ (2
2n
+ 1)2
−n
χ
¦f≥(2
2n
+1)2
−n
¦
. Then each E
n,k
is measurable and each ϕ
n
is a simple function. Note that E
n,k
= E
n+1,2k
∪ E
n+1,2k+1
for all 'n, k`.
Suppose x ∈ E
n,k
. Then ϕ
n
(x) = k2
−n
. If x ∈ E
n+1,2k
, then ϕ
n+1
(x) = (2k)2
−(n+1)
= k2
−n
= ϕ
n
(x).
If x ∈ E
n+1,2k+1
, then ϕ
n+1
(x) = (2k + 1)2
−(n+1)
> (2k)2
−(n+1)
ϕ
n
(x). Also, if f(x) ≥ (2
2n
+ 1)2
−n
,
then ϕ
n
(x) = (2
2n
+ 1)2
−n
≤ ϕ
n+1
(x). Thus ϕ
n
≤ ϕ
n+1
. For x ∈ X, if f(x) < ∞, then for suﬃciently
large n, f(x) −ϕ
n
(x) ≤ 2
−n
. If f(x) = ∞, then ϕ
n
(x) = (2
2n
+ 1)2
−n
> 2
n
→ ∞. Thus f = limϕ
n
at
each point of X.
If f is deﬁned on a σﬁnite measure space (X, B, µ), then X =
¸
X
n
where µ(X
n
) < ∞ for each n.
Deﬁne ϕ
n
== 2
−n
¸
2
2n
k=0
kχ
E
n,k
∩
n
m=1
X
m
+ (2
2n
+ 1)2
−n
χ
¦f≥(2
2n
+1)2
−n
¦∩
n
m=1
X
m
. Then ϕ
n+1
≥ ϕ
n
and f = limϕ
n
. Furthermore, each ϕ
n
vanishes outside
¸
n
m=1
X
m
, a set of ﬁnite measure.
(*) Proof of Proposition 7
11. Suppose µ is a complete measure and f be a measurable function. Suppose f = g a.e. Then
¦x : g(x) < α¦ = ¦x : f = g, f(x) < α¦ ∪ ¦x : f = g, g(x) < α¦ = (¦x : f(x) < α¦ ` ¦x : f = g, f(x) <
α¦) ∪ ¦x : f = g, g(x) < α¦. Now ¦x : f = g, f(x) < α¦ and ¦x : f = g, g(x) < α¦ are subsets of a set of
measure zero so they are measurable. Also, ¦x : f(x) < α¦ is measurable since f is measurable. Hence
¦x : g(x) < α¦ is measurable for any α and g is measurable.
Q3.28 gives an example of a set A of Lebesgue measure zero that is not a Borel set so Lebesgue measures
restricted to the σalgebra of Borel sets is not complete. Now χ
A
is not measurable since ¦x : χ
A
(x) >
1/2¦ = A but m(A) = 0 so χ
A
= 0 a.e. and the constant zero function is measurable.
(*) Proof of Proposition 8
12. Let 'f
n
` be a sequence of measurable functions that converge to a function f except at the points of a
set E of measure zero. Suppose µ is complete. Given α, ¦x : f(x) > α¦ = ¦x / ∈ E : limf
n
(x) > α¦∪¦x ∈
E : f(x) > α¦. Since µ(E) = 0, ¦x ∈ E : f(x) > α¦ is measurable. Also, ¦x / ∈ E : limf
n
(x) > α¦ =
E
c
∩ ¦x : limf
n
(x) > α¦ so it is measurable. Hence ¦x : f(x) > α¦ is measurable and f is measurable.
Note: If f
n
(x) → f(x) for all x, then completeness is not required.
13a. Let 'f
n
` be a sequence of measurable realvalued functions that converge to f in measure. For
any k, there exists n
k
and a measurable set E
k
with µ(E
k
) < 2
−k
such that [f
n
k
(x) − f(x)[ < 2
−k
for x / ∈ E
k
. If x / ∈
¸
∞
k=m
E
k
, then [f
n
k
(x) − f(x)[ < 2
−k
for k ≥ m. Thus if x / ∈
¸
m
¸
∞
k=m
E
k
,
then for some m, we have [f
n
k
(x) − f(x)[ < 2
−k
for k ≥ m so 'f
n
k
` converges to f. Furthermore,
µ(
¸
m
¸
∞
k=m
E
k
) ≤ µ(
¸
∞
k=m
E
k
) ≤
¸
∞
k=m
µ(E
k
) <
¸
∞
k=m
2
−k
= 2
−m+1
for all m so µ(
¸
m
¸
∞
k=m
E
k
) =
0 and 'f
n
k
` converges to f a.e.
(c.f. Proposition 4.18)
13b. Let 'f
n
` be a sequence of measurable functions each of which vanishes outside a ﬁxed measurable set
A with µ(A) < ∞. Suppose that f
n
(x) → f(x) for almost all x, say except on a set B of measure zero. If
x ∈ A
c
`B, then f
n
(x) = 0 for all n so f(x) = 0. Given ε > 0, let G
n
= ¦x ∈ A`B : [f
n
(x)−f(x)[ ≥ ε¦ and
let E
N
=
¸
∞
n=N
G
n
. Note that E
N+1
⊂ E
N
. If x ∈ A`B, then there exists N such that [f
n
(x)−f(x)[ < ε
for n ≥ N. i.e. x / ∈ E
N
for some N. Thus
¸
E
N
= ∅ and limµ(E
N
) = 0 so there exists N such that
µ(E
N
) < ε. Furthermore, µ(E
N
∪ B) < ε and if x / ∈ E
N
∪ B, then either x ∈ A and x / ∈ G
c
n
for all
n ≥ N so [f
n
(x) −f(x)[ < ε for all n ≥ N or x / ∈ A so [f
n
(x) −f(x)[ = 0 for all n. Hence 'f
n
` converges
to f in measure.
76
13c. Let 'f
n
` be a sequence that is Cauchy in measure. We may choose n
k+1
> n
k
such that µ¦x :
[f
n
k+1
(x) − f
n
k
(x)[ ≥ 2
−k
¦ < 2
−k
. Let E
k
= ¦x : [f
n
k+1
(x) − f
n
k
(x)[ ≥ 2
−k
¦ and let F
m
=
¸
∞
k=m
E
k
.
Then µ(
¸
F
m
) ≤ µ(
¸
∞
k=m
E
k
) ≤ 2
−k+1
for all k so µ(
¸
F
m
) = 0. If x / ∈
¸
F
m
, then x / ∈ F
m
for some m
so [f
n
k+1
(x) −f
n
k
(x)[ < 2
−k
for all k ≥ m and [f
n
l
(x) −f
n
k
(x)[ < 2
−k+1
for l ≥ k ≥ m. Thus the series
¸
(f
n
k+1
−f
n
k
) converges a.e. to a function g. Let f = g+f
n
1
. Then f
n
k
→ f in measure since the partial
sums are of the form f
n
k
− f
n
1
. Given ε > 0, choose N such that µ¦x : [f
n
(x) − f
m
(x)[ ≥ ε/2¦ < ε/2
for n > m ≥ N and µ¦x : [f
n
k
(x) −f(x)[ ≥ ε/2¦ < ε/2 for k ≥ N. Then ¦x : [f
n
(x) −f(x)[ ≥ ε¦ ⊂ ¦x :
[f
n
(x) −f
n
k
(x)[ ≥ ε/2¦∪¦x : [f
n
k
(x) −f(x)[ ≥ ε/2¦ for all n, k ≥ N. Thus µ¦x : [f
n
(x) −f(x)[ ≥ ε¦ < ε
for all n ≥ N and 'f
n
` converges to f in measure.
(c.f. Q4.25)
14. Let (X, B, µ) be a measure space and (X, B
0
, µ
0
) its completion. Suppose g is measurable with
respect to B and there is a set E ∈ B with µ(E) = 0 and f = g on X`E. For any α, ¦x : g(x) < α¦ ∈ B.
Now ¦x : f(x) < α¦ = ¦x ∈ X ` E : g(x) < α¦ ∪ ¦x ∈ E : f(x) < α¦ where ¦x ∈ X ` E : g(x) < α¦ ∈ B
and ¦x ∈ E : f(x) < α¦ ⊂ E. Thus ¦x : f(x) < α¦ ∈ B
0
and f is measurable with respect to B
0
.
For the converse, ﬁrst consider the case of characteristic functions. Suppose χ
A
is measurable with
respect to B
0
. Then A ∈ B
0
so A = A
t
∪ B
t
where A
t
∈ B and B
t
⊂ C
t
with C
t
∈ B and µ(C
t
) = 0.
Deﬁne f(x) = χ
A
(x) if x / ∈ C
t
and f(x) = 1 if x ∈ C
t
. If α < 0, then ¦x : f(x) > α¦ = X. If α ≥ 1,
then ¦x : f(x) > α¦ = ∅. If 0 ≤ α < 1, then ¦x : f(x) > α¦ = ¦x : f(x) = 1¦ = A ∪ C
t
= A
t
∪ C
t
.
Thus ¦x : f(x) > α¦ ∈ B for all α and f is measurable with respect to B. Next consider the case
of simple functions. Suppose g =
¸
n
i=1
c
i
χ
A
i
is measurable with respect to B
0
. Then each χ
A
i
is
measurable with respect to B
0
. For each i, there is a function h
i
measurable with respect to B and
a set E
i
∈ B with µ(E
i
) = 0 and χ
A
i
= h
i
on X ` E
i
. Then g =
¸
n
i=1
c
i
h
i
on X `
¸
n
i=1
E
i
where
¸
n
i=1
E
i
∈ B and µ(
¸
n
i=1
E
i
) = 0. Furthermore,
¸
n
i=1
c
i
h
i
is measurable with respect to B. Now for
a nonnegative measurable function f, there is a sequence of simple functions 'ϕ
n
` converging pointwise
to f on X. For each n, there is a function ψ
n
measurable with respect to B and a set E
n
∈ B with
µ(E
n
) = 0 and ϕ
n
= ψ
n
on X ` E
n
. Then f = limψ
n
on X `
¸
E
n
where
¸
E
n
∈ B and µ(
¸
E
n
) = 0.
Furthermore, limψ
n
is measurable with respect to B. Finally, for general a measurable function f, we
have f = f
+
− f
−
where f
+
and f
−
are nonnegative measurable functions and the result follows from
the previous case.
15. Let D be the rationals. Suppose that to each β ∈ D there is assigned a B
β
∈ B such that
B
α
⊂ B
β
for α < β. Then there is a unique measurable function f on X such that f ≤ β on B
β
and
f ≥ β on X ` B
β
. Now ¦x : f(x) < α¦ =
¸
β<α
B
β
and ¦x : f(x) ≤ α¦ =
¸
γ>α
¸
β<γ
B
β
. Similarly,
¦x : f(x) > α¦ = X `
¸
β>α
B
β
and ¦x : f(x) ≥ α¦ =
¸
γ<α
(X `
¸
β>γ
B
β
). Also, ¦x : f(x) = α¦ =
¸
γ>α
¸
β<γ
B
β
∩
¸
γ<α
(X `
¸
β>γ
B
β
).
16. Egoroﬀ’s Theorem: Let 'f
n
` be a sequence of measurable functions each of which vanishes outside
a ﬁxed measurable set A of ﬁnite measure. Suppose that f
n
(x) → f(x) for almost all x. By Q13b, 'f
n
`
converges to f in measure. Given ε > 0 and n, there exist N
n
and a measurable set E
n
with µ(E
n
) < ε2
−n
such that [f
n
(x) −f(x)[ < 1/n for n ≥ N
n
and x / ∈ E
n
. Let E =
¸
E
n
. Then µ(
¸
E
n
) ≤
¸
µ(E
n
) < ε.
Choose n
0
such that 1/n
0
< ε. If x / ∈ E and n ≥ N
n
0
, we have [f
n
(x) − f(x)[ < 1/n
0
< ε. Thus f
n
converges uniformly to f on X ` E.
11.3 Integration
17. Let f and g be measurable functions and E a measurable set.
(i) For a constant c
1
, note that if c
1
≥ 0, then (c
1
f)
+
= c
1
f
+
and (c
1
f)
−
= c
1
f
−
so
E
c
1
f =
E
c
1
f
+
−
E
c
1
f
−
= c
1
E
f
+
−c
1
E
f
−
= c
1
E
f. On the other hand, if c
1
< 0, then (c
1
f)
+
= −c
1
f
−
and (c
1
f)
−
= −c
1
f
+
so
E
c
1
f =
E
(−c
1
f
−
) −
E
(−c
1
f
+
) = −c
1
E
f
−
−(−c
1
)
E
f
+
= c
1
E
f. Note
that if f = f
1
− f
2
where f
1
, f
2
are nonnegative integrable functions, then f
+
+ f
2
= f
−
+ f
1
so
f
+
+
f
2
=
f
−
+
f
1
and
f =
f
+
−
f
−
=
f
1
−
f
2
. Now since f and g are integrable, so
are f
+
+g
+
and f
−
+g
−
. Furthermore, f +g = (f
+
+g
+
) −(f
−
+g
−
). Thus
(f +g) =
(f
+
+g
+
) −
(f
−
+g
−
) =
f
+
+
g
+
−
f
−
−
g
−
=
f +
g. Together, we have
E
(c
1
f +c
2
g) = c
1
E
f +c
2
E
g.
(ii) Suppose [h[ ≤ [f[ and h is measurable. Since f is integrable, so are f
+
and f
−
. Thus [f[ = f
+
+f
−
is integrable. By (iii), we have
h ≤
[h[ ≤
[f[ < ∞. Thus h is integrable.
77
(iii) Suppose f ≥ g a.e. Then f −g ≥ 0 a.e. and
(f −g) ≥ 0. By (i), we have
(f −g) =
f −
g so
f −
g ≥ 0. i.e.
f ≥
g.
(*) Proof of Proposition 15
18. Suppose that µ is not complete. Deﬁne a bounded function f to be integrable over a set E of ﬁnite
measure if sup
ϕ≤f
E
ϕ dµ = inf
ψ≥f
E
ψ dµ for all simple functions ϕ and ψ. If f is integrable, then
given n, there are simple functions ϕ
n
and ψ
n
such that ϕ
n
≤ f ≤ ψ
n
and
ψ
n
dµ −
ϕ
n
dµ < 1/n.
Then the functions ψ
∗
= inf ψ
n
and ϕ
∗
= sup ϕ
n
are measurable and ϕ
∗
≤ f ≤ ψ
∗
. Now the set
∆ = ¦x : ϕ
∗
(x) < ψ
∗
(x)¦ is the union of the sets ∆
v
= ¦x : ϕ
∗
(x) < ψ
∗
(x) −1/v¦. Each ∆
v
is contained
in the set ¦x : ϕ
n
(x) < ψ
n
(x) −1/v¦, which has measure less than v/n. Since n is arbitrary, µ(∆
v
) = 0
so µ(∆) = 0. Thus ϕ
∗
= f = ψ
∗
except on a set of measure zero. Hence f is measurable with respect to
the completion of µ by Q14.
Conversely, if f is measurable with respect to the completion of µ, then by Q14, f = g on X ` E
where g is measurable with respect to µ and µ(E) = 0. We may assume that g is bounded by M. By
considering the sets E
k
= ¦x : kM/n ≥ g(x) ≥ (k − 1)M/n¦, −n ≤ k ≤ n, and the simple functions
ψ
n
= (M/n)
¸
n
k=−n
kχ
E
k
and ϕ
n
= (M/n)
¸
n
k=−n
(k − 1)χ
E
k
, we see that g is integrable. Note that
sup
ϕ≤g
E
ϕ dµ = sup
ϕ≤f
E
ϕ dµ and inf
ψ≥g
E
ψ dµ = inf
ψ≥f
E
ψ dµ. It follows that f is integrable.
19. Let f be an integrable function on the measure space (X, B, µ). Let ε > 0 be given. Suppose f is
nonnegative and bounded by M. If µ(E) < ε/M, then
E
f < ε. In particular, the result holds for all
simple functions. If f is nonnegative, there is an increasing sequence 'ϕ
n
` of nonnegative simple functions
converging to f on X. By the Monotone Convergence Theorem, we have
f = lim
ϕ
n
so there exists
N such that
(f −f
N
) < ε/2. Choose δ < ε/2 sup [ϕ
N
[. If µ(E) < δ, then
E
f =
E
(f −f
N
) +
E
f
N
<
ε/2 + ε/2 = ε so the result holds for nonnegative measurable functions. For an integrable function f,
there exists δ > 0 such that if µ(E) < δ, then
E
[f[ < ε. Thus [
E
f[ ≤
E
[f[ < ε.
20. Fatou’s Lemma: Let 'f
n
` be a sequence of nonnegative measurable functions that converge in
measure on a set E to a function f. There is a subsequence 'f
n
k
` such that lim
E
f
n
k
= lim
E
f
n
. Now
'f
n
k
` converges to f in measure on E so by Q13a it has a subsequence 'f
n
k
j
` that converges to f almost
everywhere on E. Thus
E
f ≤ lim
E
f
n
k
j
= lim
E
f
n
k
= lim
E
f
n
.
Monotone Convergence Theorem: Let 'f
n
` be a sequence of nonnegative measurable functions which
converge in measure to a function f and suppose that f
n
≤ f for all n. Since f
n
≤ f, we have
f
n
≤
f.
By Fatou’s Lemma, we have
f ≤ lim
f
n
≤ lim
f
n
≤
f so equality holds and
f = lim
f
n
.
Lebesgue Convergence Theorem: Let g be integrable over E and suppose that 'f
n
` is a sequence of
measurable functions such that on E we have [f
n
(x)[ ≤ g(x) and such that 'f
n
` converges in measure
to f almost everywhere on E. The functions g −f
n
are nonnegative so by Fatou’s Lemma,
E
(g −f) ≤
lim
E
(g −f
n
). There is a subsequence 'f
n
k
` that converges to f almost everywhere on E so [f[ ≤ [g[ on
E and f integrable. Thus
E
g −
E
f ≤
E
g −lim
E
f
n
and lim
E
f
n
≤
E
f. Similarly, by considering
the functions g +f
n
, we have
E
f ≤ lim
E
f
n
. Thus equality holds and
E
f = lim
E
f
n
.
21a. Suppose f is integrable. Then [f[ is integrable. Now ¦x : f(x) = 0¦ = ¦x : [f(x)[ > 0¦ =
¸
¦x : [f(x)[ ≥ 1/n¦. Each of the sets ¦x : [f(x)[ ≥ 1/n¦ is measurable. If µ¦x : [f(x)[ ≥ 1/n¦ = ∞
for some n, then
[f[ ≥
¦x:]f(x)]≥1/n¦
[f[ ≥ (1/n)µ¦x : [f(x)[ ≥ 1/n¦ = ∞. Contradiction. Thus
µ¦x : [f(x)[ ≥ 1/n¦ < ∞ for all n and ¦x : f(x) = 0¦ is of σﬁnite measure.
21b. Suppose f is integrable and f ≥ 0. There is an increasing sequence 'ϕ
n
` of simple functions such
that f = limϕ
n
. We may redeﬁne each ϕ
n
to be zero on ¦x : f(x) = 0¦. Since ¦x : f(x) = 0¦ is σﬁnite,
we may further redeﬁne each ϕ
n
to vanish outside a set of ﬁnite measure by Proposition 7 (c.f. Q10).
21c. Suppose f is integrable with respect to µ. Then f
+
and f
−
are nonnegative integrable functions
so by part (b), there are increasing sequences 'ϕ
n
` and 'ψ
n
` of simple functions each of which vanishes
outside a set of ﬁnite measure such that limϕ
n
= f
+
and limψ
n
= f
−
. By the Monotone Convergence
Theorem,
f
+
dµ = lim
ϕ
n
dµ and
f
−
dµ = lim
ψ
n
dµ. Given ε > 0, there is a simple function ϕ
N
such that
f
+
dµ−
ϕ
n
dµ < ε/2 and there is a simple function ψ
N
such that
f
−
dµ−
ψ
N
dµ < ε/2.
Let ϕ = ϕ
N
− ψ
N
. Then ϕ is a simple function and
[f − ϕ[ dµ =
[f
+
− ϕ
N
− f
−
+ ψ
N
[ dµ ≤
[f
+
−ϕ
N
[ dµ +
[f
−
−ψ
N
[ dµ = (
f
+
dµ −
ϕ
N
dµ) + (
f
−
dµ −
ψ
N
dµ) < ε.
22a. Let (X, B, µ) be a measure space and g a nonnegative measurable function on X. Set ν(E) =
E
g dµ. Clearly ν(∅) = 0. Let 'E
n
` be a sequence of disjoint sets in B. Then ν(
¸
E
n
) =
E
n
g dµ =
78
gχ
E
n
dµ =
¸
gχ
E
n
dµ =
¸
gχ
E
n
dµ =
¸
E
n
g dµ =
¸
ν(E
n
). Hence ν is a measure on B.
22b. Let f be a nonnegative measurable function on X. If ϕ is a simple function given by
¸
n
i=1
c
i
χ
E
i
.
Then
ϕ dν =
¸
n
i=1
c
i
ν(E
i
) =
¸
n
i=1
c
i
E
i
g dµ =
¸
n
i=1
c
i
gχ
E
i
dµ =
¸
n
i=1
c
i
gχ
E
i
dµ =
ϕg dµ.
Now if f is a nonnegative measurable function, there is an increasing sequence 'ϕ
n
` of simple functions
such that f = limϕ
n
. Then 'ϕ
n
g` is an increasing sequence of nonnegative measurable functions such
that fg = limϕ
n
g. By the Monotone Convergence Theorem, we have
fg dµ = lim
ϕ
n
g dµ =
lim
ϕ
n
dν =
f dν.
23a. Let (X, B, µ) be a measure space. Suppose f is locally measurable. For any α and any E ∈ B
with µ(E) < ∞, ¦x : f(x) > α¦ ∩ E = ¦x : fχ
E
(x) > α¦ if α ≥ 0 and ¦x : f(x) > α¦ ∩ E = ¦x :
fχ
E
(x) > 0¦ ∪ (¦x : fχ
E
(x) = 0¦ ∩ E) ∪ ¦x : 0 > fχ
E
(x) > α¦ if α < 0. Thus ¦x : f(x) > α¦ ∩ E is
measurable so ¦x : f(x) > α¦ is locally measurable and f is measurable with respect to the σalgebra of
locally measurable sets.
Conversely, suppose f is measurable with respect to the σalgebra of locally measurable sets. For
any α and any E ∈ B with µ(E) < ∞, ¦x : fχ
E
(x) > α¦ = ¦x : f(x) > α¦ ∩ E if α ≥ 0 and
¦x : fχ
E
(x) > α¦ = (¦x : f(x) > α¦ ∩ E) ∪ E
c
if α < 0. Thus ¦x : fχ
E
(x) > α¦ is measurable and f is
locally measurable.
23b. Let µ be a σﬁnite measure. Deﬁne integration for nonnegative locally measurable functions f by
taking
f to be the supremum of
ϕ as ϕ ranges over all simple functions less than f. For a simple
function ϕ =
¸
n
i=1
c
i
χ
E
i
, we have
ϕ =
¸
n
i=1
c
i
µ(E
i
) =
¸
n
i=1
c
i
µ(E
i
) =
ϕ dµ.
Let X =
¸
X
n
where each X
n
is measurable and µ(X
n
) < ∞. Then
f =
fχ
X
n
=
¸
n
fχ
X
n
.
Now fχ
X
n
is measurable for each n so there is an increasing sequence 'ϕ
(n)
k
` of simple functions con
verging to fχ
X
n
. Thus
f =
¸
n
fχ
X
n
=
¸
n
fχ
X
n
=
¸
n
lim
k
ϕ
(n)
k
=
¸
n
lim
k
ϕ
(n)
k
dµ =
¸
n
fχ
X
n
dµ =
¸
n
fχ
X
n
dµ =
fχ
X
n
dµ =
f dµ.
11.4 General convergence theorems
24. Let (X, B) be a measurable space and 'µ
n
` a sequence of measures on B such that for each E ∈ B,
µ
n+1
(E) ≥ µ
n
(E). Let µ(E) = limµ
n
(E). Clearly µ(∅) = 0. Let 'E
k
` be a sequence of disjoint sets
in B. Then µ(
¸
k
E
k
) = lim
n
µ
n
(
¸
k
E
k
) = lim
n
¸
k
µ
n
(E
k
) =
¸
k
lim
n
µ
n
(E
k
) =
¸
k
µ(E
k
) where the
interchanging of the limit and the summation is valid because µ
n
(E
k
) is increasing in n for each k. Hence
µ is a measure.
*25. Let m be Lebesgue measure. For each n, deﬁne µ
n
by µ
n
(E) = m(E)2
−n
. Then 'µ
n
` is a
decreasing sequence of measures. Note that R =
¸
k∈Z
[k, k +1). Now µ(
¸
k
[k, k +1) = lim
n
µ
n
(
¸
k
[k, k +
1)) = lim
n
m(R)2
−n
= ∞ but
¸
k
µ([k, k + 1)) =
¸
k
lim
n
µ
n
([k, k + 1)) =
¸
k
lim
n
m([k, k + 1))2
−n
=
¸
k
lim
n
2
−n
= 0. Hence µ is not a measure.
*26. Let (X, B) be a measurable space and 'µ
n
` a sequence of measures on Bthat converge setwise to to
a set function µ. Clearly µ(∅) = 0. If E
1
, E
2
∈ B and E
1
∩E
2
= ∅, then µ(E
1
∪E
2
) = limµ
n
(E
1
∪E
2
) =
lim(µ
n
(E
1
)+µ
n
(E
2
)) = limµ
n
(E
1
)+limµ
n
(E
2
) = µ(E
1
)+µ(E
2
). Thus µ is ﬁnitely additive. If µ is not
a measure, then it is not ∅continuous. i.e. there is a decreasing sequence 'E
n
` of set in B with
¸
E
n
= ∅
and limµ(E
n
) = ε > 0. Deﬁne α
1
= β
1
= 1. If α
j
and β
j
have been deﬁned for j ≤ k, let α
k+1
> α
k
such
that µ
α
k+1
(E
β
k
) ≥ 7ε/8. Then let β
k+1
> β
k
such that ε/8 ≥ µ
α
k+1
(E
β
k+1
). Deﬁne F
n
= E
β
n
` E
β
n+1
.
Then µ
α
n+1
(F
n
) ≥ 3ε/4. Now for j odd and 1 ≤ k < j, we have µ
α
j
(
¸
n even,n≥k
F
n
) ≥ 3ε/4 so for
k ≥ 1, we have µ(
¸
n even,n≥k
F
n
) ≥ 3ε/4. This inequality is also true for odd n. Thus for k ≥ 1, we
have µ(E
β
k
) = µ(
¸
F
n
) ≥ 3ε/2. Contradiction. Hence µ is a measure.
11.5 Signed measures
27a. Let ν be a signed measure on a measurable space (X, B) and let ¦A, B¦ be a Hahn decomposition
for ν. Let N be a null set and consider ¦A ∪ N, B ` N¦. Note that (A ∪ N) ∪ (B ` N) = A ∪ B =
X and (A ∪ N) ∩ (B ` N) = A ∩ (B ` N) = ∅. For any measurable subset E of A ∪ N, we have
ν(E) = ν(E ∩ A) + ν(E ∩ (N ` A)) = ν(E ∩ A) ≥ 0. For any measurable subset E
t
of B ` N, we have
ν(E
t
) = ν(E
t
∩B) −ν(E
t
∩B∩N) = ν(E
t
∩B) ≤ 0. Thus ¦A∪N, B` N¦ is also a Hahn decomposition
for ν.
79
Consider a set ¦a, b, c¦ with the σalgebra being its power set. Deﬁne ν(¦a¦) = −1, ν(¦b¦) = 0, ν(¦c¦) = 1
and extend it to a signed measure on ¦a, b, c¦ in the natural way. Then ¦¦a, b¦, ¦c¦¦ and ¦¦a¦, ¦b, c¦¦
are both Hahn decompositions for ν.
27b. Suppose ¦A, B¦ and ¦A
t
, B
t
¦ are Hahn decompositions for ν. Then A` A
t
= A∩ B
t
and A
t
` A =
A
t
∩B. Thus A` A
t
and A
t
` A are null sets. Since A∆A
t
= (A` A
t
) ∪(A
t
` A), A∆A
t
is also a null set.
Similarly, B∆B
t
is a null set. Hence the Hahn decomposition is unique except for null sets.
28. Let ¦A, B¦ be a Hahn decomposition for ν. Suppose ν = ν
+
− ν
−
= µ
1
− µ
2
where µ
1
⊥µ
2
. There
are disjoint measurable sets A
t
, B
t
such that X = A
t
∪ B
t
and µ
1
(B
t
) = µ
2
(A
t
) = 0. If E ⊂ A
t
, then
ν(E) = µ
1
(E) −µ
2
(E) = µ
1
(E) ≥ 0. If E ⊂ B
t
, then ν(E) = −µ
2
(E) ≤ 0. Thus ¦A
t
, B
t
¦ is also a Hahn
decomposition for ν. Now for any measurable set E, we have ν
+
(E) = ν(E∩A) = ν(E∩A∩A
t
)+ν(E∩(A`
A
t
)) = ν(E∩A∩A
t
)+ν(E∩(A
t
`A)) = µ
1
(E∩A∩A
t
)+µ
1
(E∩(A
t
`A)) = µ
1
(E∩A
t
)+µ
1
(E∩B
t
) = µ
1
(E).
Similarly, ν
−
(E) = µ
2
(E).
29. Let ¦A, B¦ be a Hahn decomposition for ν and let E be any measurable set. Since ν
2
(E) ≥ 0, we
have ν(E) = ν
+
(E) − ν
−
(E) ≤ ν
+
(E). Since ν
1
(E) ≥ 0, we have −ν
−
(E) ≤ ν
+
(E) − ν
−
(E) = ν(E).
Thus −ν
−
(E) ≤ ν(E) ≤ ν
+
(E).
Now ν(E) ≤ ν
+
(E) ≤ ν
+
(E) +ν
−
(E) = [ν[(E) and −ν(E) ≤ ν
−
(E) ≤ ν
+
(E) +ν
−
(E) = [ν[(E). Hence
[ν(E)[ ≤ [ν[(E).
30. Let ν
1
and ν
2
be ﬁnite signed measures and let α, β ∈ R. Then αν
1
+βν
2
is a signed measure. Then
αν
1
+βν
2
is a ﬁnite signed measure since (αν
1
+βν
2
)(
¸
E
n
) = αν
1
(
¸
E
n
) +βν
2
(
¸
E
n
) = α
¸
ν
1
(E
n
) +
β
¸
ν
2
(E
n
) =
¸
(αν
1
(E
n
) + βν
2
(E
n
)) for any sequence 'E
n
` of disjoint measurable sets where the last
series converges absolutely.
Let ¦A, B¦ be a Hahn decomposition for ν. Suppose α ≥ 0. Then [αν[(E) = (αν)
+
(E) + (αν)
−
(E) =
αν
+
(E) + αν
−
(E) = [α[ [ν[(E) since αν = αν
+
− αν
−
and αν
+
⊥αν
−
as αν
+
(B) = 0 = αν
−
(A).
Suppose α < 0. Then [αν[(E) = (αν)
+
(E) + (αν)
−
(E) = −αν
−
(E) − αν
+
(E) = [α[ [ν[(E) since
αν = αν
+
− αν
−
= −αν
−
− (−αν
+
) and −αν
+
⊥ − αν
−
as −αν
+
(B) = 0 = −αν
−
(A). Hence
[αν[ = [α[ [ν[.
Also, [ν
1
+ν
2
[(E) = [ν
1
(E) +ν
2
(E)[ ≤ [ν
1
(E)[ +[ν
2
(E)[ = [ν
1
[(E) +[ν
2
[(E). Hence [ν
1
+ν
2
[ ≤ [ν
1
[ +[ν
2
[.
31. Deﬁne integration with respect to a signed measure ν by deﬁning
f dν =
f dν
+
−
f dν
−
.
Suppose [f[ ≤ M. Then [
E
f dν[ = [
E
f dν
+
−
E
f dν
−
[ ≤ [
E
f dν
+
[ + [
E
f dν
−
[ ≤
E
[f[ dν
+
+
E
[f[ dν
−
≤ Mν
+
(E) +Mν
−
(E) = M[ν[(E).
Let ¦A, B¦ be a Hahn decomposition for ν. Then
E
(χ
E∩A
− χ
E∩B
) dν =
E∩A
1 dν −
E∩B
1 dν =
[ν
+
(E∩A)−ν
−
(E∩A)] −[ν
+
(E∩B)−ν
−
(E∩B)] = ν
+
(E∩A)+ν
−
(E∩A)+ν
+
(E∩B)+ν
−
(E∩B) =
ν
+
(E) +ν
−
(E) = [ν[(E) where χ
E∩A
−χ
E∩B
is measurable and [χ
E∩A
−χ
E∩B
[ ≤ 1.
32a. Let µ and ν be ﬁnite signed measures. Deﬁne µ ∧ ν by (µ ∧ ν)(E) = min(µ(E), ν(E)). Note that
µ ∧ ν =
1
2
(µ + ν − [µ − ν[) so µ ∧ ν is a ﬁnite signed measure by Q30 and it is smaller than µ and ν.
Furthermore, if η is a signed measure smaller than µ and ν, then η ≤ µ ∧ ν.
32b. Deﬁne µ ∨ ν by (µ ∨ ν)(E) = max(µ(E), ν(E)). Note that µ ∨ ν =
1
2
(µ +ν +[µ −ν[) so µ ∨ ν is a
ﬁnite signed measure by Q30 and it is larger than µ and ν. Furthermore, if η is a signed measure larger
than µ and ν, then η ≥ µ ∨ ν. Also, µ ∨ ν +µ ∧ ν = max(µ, ν) + min(µ, ν) = µ +ν.
32c. Suppose µ and ν are positive measures. If µ⊥ν, then there are disjoint measurable sets A and
B such that X = A ∪ B and µ(B) = 0 = ν(A). For any measurable set E, we have (µ ∧ ν)(E) =
(µ ∧ν(E ∩A) +(µ ∧ν)(E ∩B) = min(µ(E ∩A), ν(E ∩A)) +min(µ(E ∩B), ν(E ∩B)) = 0. Conversely,
suppose µ ∧ ν = 0. If µ(E) = ν(E) = 0 for all measurable sets, then µ = ν = 0 and µ⊥ν. Thus we may
assume that µ(E) = 0 < ν(E) for some E. If ν(E
c
) = 0, it follows that µ⊥ν. On the other hand, if
ν(E
c
) > 0, then µ(E
c
) = 0 so µ(X) = µ(E) +µ(E
c
) = 0. Thus µ = 0 and we still have µ⊥ν.
11.6 The RadonNikodym Theorem
33a. Let (X, B, µ) be a σﬁnite measure space and let ν be a measure on B which is absolutely
continuous with respect to µ. Let X =
¸
X
i
with µ(X
i
) < ∞ for each i. We may assume the X
i
are
pairwise disjoint. For each i, let B
i
= ¦E ∈ B : E ⊂ E
i
¦, µ
i
= µ[
B
i
and ν
i
= ν[
B
i
. Then (X
i
, B
i
, µ
i
) is a
ﬁnite measure space and ν
i
<< µ
i
. Thus for each i there is a nonnegative µ
i
measurable function f
i
such
80
that ν
i
(E) =
E
f
i
dµ
i
for all E ∈ B
i
. Deﬁne f by f(x) = f
i
(x) if x ∈ X
i
. If E ⊂ X, then E ∩ X
i
∈ B
i
for each i. Thus ν(E) =
¸
ν(E ∩ X
i
) =
¸
ν
i
(E ∩ X
i
) =
¸
E∩X
i
f
i
dµ
i
=
¸
E∩X
i
f dµ =
E
f dµ.
33b. If f and g are nonnegative measurable functions such that ν(E) =
E
f dµ =
E
g dµ for any
measurable set E, then
f dµ =
g dµ so f = g a.e.
34a. RadonNikodym derivatives: Suppose ν << µ and f is a nonnegative measurable function. For
a nonnegative simple function ϕ =
¸
n
i=1
c
i
χ
E
i
, we have
ϕ dν =
¸
n
i=1
c
i
ν(E
i
) =
¸
n
i=1
c
i
(
E
i
dν
dµ
dµ) =
¸
n
i=1
c
i
χ
E
i
dν
dµ
dµ =
ϕ
dν
dµ
dµ. For a nonnegative measurable function f, let 'ϕ
i
` be an increas
ing sequence of nonnegative simple functions converging pointwise to f. Then
f dν = lim
ϕ
i
dν =
lim
ϕ
i
dν
dµ
dµ =
f
dν
dµ
dµ.
34b. If ν
1
<< µ and ν
2
<< µ, then ν
1
+ ν
2
<< µ. For any measurable set E, we have ν
1
(E) =
E
dν
1
dµ
dµ and ν
2
(E) =
E
dν
2
dµ
dµ. Thus (ν
1
+ν
2
)(E) =
E
dν
1
dµ
+
dν
2
dµ
dµ. By uniqueness of the
RadonNikodym derivative, we have
d(ν
1
+ν
2
)
dµ
=
dν
1
dµ
+
dν
2
dµ
.
34c. Suppose ν << µ << λ. For any measurable set E, we have ν(E) =
E
dν
dµ
dµ =
E
dν
dµ
dµ
dλ
dλ
where the last equality follows from part (a). Hence
dν
dλ
=
dν
dµ
dµ
dλ
.
34d. Suppose ν << µ and µ << ν. Then
dν
dν
=
dν
dµ
dµ
dν
by part (c). But
dν
dν
≡ 1 so
dν
dµ
=
dµ
dν
−1
.
35a. Suppose ν is a signed measure such that ν⊥µ and ν << µ. There are disjoint measurable sets A
and B such that X = A∪ B and [ν[(B) = 0 = [µ[(A). Then [ν(A)[ = 0 so [ν[(X) = [ν[(A) +[ν[(B) = 0.
Hence ν
+
= 0 = ν
−
so ν = 0.
35b. Suppose ν
1
and ν
2
are singular with respect to µ. There are disjoint measurable sets A
1
and B
1
such that X = A
1
∪B
1
and µ(B
1
) = 0 = ν
1
(A
1
). Similarly, there are disjoint measurable sets A
2
and B
2
such that X = A
2
∪B
2
and µ(B
2
) = 0 = ν
2
(A
2
). Now X = (A
1
∩A
2
)∪(B
1
∪B
2
), (A
1
∩A
2
)∩(B
1
∪B
2
) = ∅
and (c
1
ν
1
+c
2
ν
2
)(A
1
∩ A
2
) = 0 = µ(B
1
∪ B
2
). Hence c
1
ν
1
+c
2
ν
2
⊥µ.
35c. Suppose ν
1
<< µ and ν
2
<< µ. If µ(E) = 0, then ν
1
(E) = 0 = ν
2
(E). Thus (c
1
ν
1
+c
2
ν
2
)(E) = 0
and c
1
ν
1
+c
2
ν
2
<< µ.
35d. Let (X, B, µ) be a σﬁnite measure space and let ν be a σﬁnite measure on B. Suppose ν =
ν
0
+ν
1
= ν
t
0
+ν
t
1
where ν
0
⊥µ, ν
t
0
⊥µ, ν
1
<< µ and ν
2
<< µ. Now ν
0
−ν
t
0
= ν
t
1
−ν
1
where ν
0
−ν
t
0
and
ν
t
1
−ν
1
are signed measures such that ν
0
−ν
t
0
⊥µ and ν
t
1
−ν
1
<< µ by parts (b) and (c). Then by part
(a), we have ν
0
−ν
t
0
= 0 = ν
t
1
−ν
1
so ν
0
= ν
t
0
and ν
1
= ν
t
1
.
36. Let (X, B, µ) be a σﬁnite measure space and suppose ν is a signed measure on Bwhich is absolutely
continuous with respect to µ. Consider the Jordan decomposition ν = ν
1
−ν
2
where either ν
+
or ν
−
must
be ﬁnite. There are measurable functions f and g such that ν
+
(E) =
E
f dµ and ν
−
(E) =
E
g dµ.
Then ν(E) = ν
+
(E) −ν
−
(E) =
E
(f −g) dµ.
37a. Complex measures: Given a complex measure ν, its real and imaginary parts are ﬁnite signed
measures and so have Jordan decompositions µ
1
−µ
2
and µ
3
−µ
4
respectively. Hence ν = µ
1
−µ
2
+iµ
3
−iµ
4
where µ
1
, µ
2
, µ
3
, µ
4
are ﬁnite measures.
*37b.
*37c.
*37d.
*37e.
(*) Polar decomposition for complex measures
38a. Let µ and ν be ﬁnite measures on a measurable space (X, B) and set λ = µ + ν. Deﬁne F(f) =
f dµ. Note that L
2
(λ) ⊂ L
1
(λ) ⊂ L
1
(µ) since
[f[ dλ ≤ (λ(X))
1/2
(
[f[
2
dλ)
1/2
and
[f[ dµ ≤
[f[ dλ. Thus F is welldeﬁned. Clearly F is linear. Furthermore [F(f)[ ≤ (µ(X))
1/2
[[f[[
2
. Hence F is
a bounded linear functional on L
2
(λ).
38b. There exists a unique function g ∈ L
2
(λ) such that F(f) = (f, g). Note that ¦g > 1¦ =
¸
¦g ≥
1 + 1/n¦. Let E
n
= ¦g ≥ 1 + 1/n¦. Then µ(E
n
) = F(χ
E
n
) =
E
n
g dλ ≥ (1 + 1/n)λ(E
n
) so
81
(1/n)µ(E
n
) + ν(E
n
) ≤ 0. It follows that µ(E
n
) = 0 = ν(E
n
) and λ(E
n
) = 0. Thus λ(¦g > 1¦) = 0.
Similarly, λ(¦g < 0¦=0. Now µ(E) = F(χ
E
) = (χ
E
, g) =
E
g dλ and ν(E) = λ(E) − µ(E) =
E
(1 −g) dλ.
38c. Suppose ν << µ. If µ(E) = 0, then ν(E) = 0 so λ(E) = 0. Thus λ << µ. Let E = ¦x : g(x) = 0¦.
Then µ(E) =
E
g dλ = 0. Thus g = 0 only on a set of µmeasure zero. In this case, we have
λ(E) =
E
1 dλ =
E
gg
−1
dλ =
E
g
−1
dµ by Q22.
38d. Suppose ν << µ. Then
E
(1 − g)g
−1
dµ =
E
g
−1
dµ −
E
gg
−1
dµ = λ(E) − µ(E) = ν(E). In
particular, (1 −g)g
−1
is integrable with respect to µ.
(*) Alternate proof of the RadonNikodym Theorem
39. Let X = [0, 1], B the class of Lebesgue measurable subsets of [0, 1] and take ν to be Lebesgue
measure and µ to be the counting measure on B. Then ν is ﬁnite and absolutely continuous with respect
to µ. Suppose there is a function f such that ν(E) =
E
f dµ for all E ∈ B. Since ν is ﬁnite, f is
integrable with respect to µ. Thus E
0
= ¦x : f(x) = 0¦ is countable. Now 0 = ν(E
0
) =
E
0
f dµ.
Contradiction. Hence there is no such function f.
40a. Decomposable measures: Let ¦X
α
¦ be a decomposition for a measure µ and E a measurable set.
Note that µ(
¸
¦X
α
∩E : µ(X
α
∩E) = 0¦) = 0 and µ(E`
¸
X
α
) = 0. Thus µ(E) = µ(
¸
¦X
α
∩E : µ(X
α
∩
E) > 0¦). If µ(X
α
∩ E) > 0 for countably many α, then we have a countable union and it follows that
µ(E) =
¸
µ(X
α
∩E). If µ(X
α
∩E) > 0 for uncountably many α, then
¸
µ(X
α
∩E) = ∞. If µ(E) < ∞,
then for any ﬁnite union
¸
n
k=1
(X
k
∩ E), we have µ(
¸
n
k=1
(X
k
∩ E)) ≤ µ(
¸
(X
α
∩ E)) ≤ µ(
¸
¦X
α
∩ E :
µ(X
α
∩E) > 0¦) = µ(E) so
¸
µ(X
α
∩E) ≤ µ(E) < ∞. Contradiction. Thus µ(E) = ∞ =
¸
µ(X
α
∩E).
40b. Let ¦X
α
¦ be a decomposition for a complete measure µ. If f is locally measurable, then since
µ(X
α
) < ∞ for each α, the restriction of f to each X
α
is measurable. Conversely, suppose the restriction
of f to each X
α
is measurable. Given a measurable set E with µ(E) < ∞, let A =
¸
¦X
α
∩ E :
µ(X
α
∩ E) > 0¦, B =
¸
¦X
α
∩ E : µ(X
α
∩ E) = 0¦ and C = E `
¸
X
α
). For β ≥ 0, ¦x : fχ
E
(x) >
β¦ = ¦x ∈ A : f(x) > β¦ ∪ ¦x ∈ B : f(x) > β¦ ∪ ¦x ∈ C : f(x) > β¦. The last two sets are
measurable since they are subsets of sets of measure zero. The ﬁrst set is measurable since it is simply
¦x : fχ
X
α
(x) > β¦ ∩ E for some α. For β < 0, ¦x : fχ
E
(x) > β¦ = E
c
∪ (E ∩ ¦x : f(x) > β¦). By a
similar argument as before, ¦x : fχ
E
(x) > β¦ is measurable. Hence f is locally measurable.
Let f be a nonnegative locally measurable function of X. Now
X
f dµ ≥
n
i=1
X
i
f dµ =
¸
n
i=1
X
i
f dµ
for any ﬁnite set ¦X
1
, . . . , X
n
¦. Thus
X
f dµ ≥
¸
α
X
α
f dµ. On the other hand, for any simple function
ϕ =
¸
n
i=1
c
i
χ
E
i
with ϕ ≤ f, we have
X
ϕ dµ =
¸
n
i=1
c
i
µ(E
i
∩
¸
X
α
) =
¸
n
i=1
c
i
(
¸
α
µ(E
i
∩ X
α
)) =
¸
α
(
¸
n
i=1
c
i
µ(E
i
∩ X
α
)) =
¸
α
X
α
ϕ dµ. Thus
X
f dµ ≤
¸
α
X
α
f dµ.
*40c. Let ν be absolutely continuous with respect to µ and suppose that there is a collection ¦X
α
¦ which
is a decomposition for both µ and ν. Let ν
α
be deﬁned by ν
α
(E) = ν(X
α
∩E) for each α. Then there is
a nonnegative measurable function f
α
such that ν
α
(E) =
E
f
α
dµ. The function f =
¸
α
f
α
is locally
measurable and ν(E) =
¸
α
ν(X
α
∩ E) =
¸
α
ν
α
(E) =
¸
α
E
f
α
dµ =
¸
α
X
α
∩E
f dµ =
E
f dµ.
40d. If instead of assuming ¦X
α
¦ to be a decomposition for ν, we merely assume that if E ∈ B and
ν(E ∩ X
α
) = 0 for all α, then ν(E) = 0, the reverse implication in part (b) remains valid although
the forward implication may not be true. Thus the argument in part (c) remains valid and so does the
conclusion.
11.7 The L
p
spaces
41. Let f ∈ L
p
(µ), 1 ≤ p < ∞, and ε > 0. First assume that f ≥ 0. Let 'ϕ
n
` be an increasing sequence
of nonnegative simple functions converging pointwise to f, each of which vanishes outside a set of ﬁnite
measure. Then the sequence '[f − ϕ
n
[
p
` converges pointwise to zero and is bounded by 2f
p
. By the
Lebesgue Convergence Theorem, lim
[f −ϕ
n
[
p
dµ = 0. i.e. lim[[f −ϕ
n
[[
p
= 0. Thus [[f −ϕ[[
p
< ε for
some ϕ.
Now for a general f, there are simple functions ϕ and ψ vanishing outside sets of ﬁnite measure such
that [[f
+
−ϕ[[
p
< ε/2 and [[f
−
−ψ[[
p
< ε/2. Then [[f −(ϕ +ψ)[[
p
< ε.
(*) Proof of Proposition 26
42. Let (X, B, µ) be a ﬁnite measure space and g an integrable function such that for some constant
82
M, [
gϕ dµ[ ≤ M[[ϕ[[
1
for all simple functions ϕ. Let E = ¦x : [g(x)[ ≥ M +ε¦ and let ϕ = (sgng)χ
E
.
Then (M +ε)µ(E) ≤
E
[g[ dµ = [
gϕ dµ[ ≤ M[[ϕ[[
1
= Mµ(E). Thus µ(E) = 0 and g ∈ L
∞
.
(*) Proof of Lemma 27 for p = 1
43. Let (X, B, µ) be a σﬁnite measure space and g an integrable function such that for some constant
M, [
gϕ dµ[ ≤ M[[ϕ[[
p
for all simple functions ϕ. Now X =
¸
X
n
where µ(X
n
) < ∞ for each n. Let
g
n
= gχ
n
i=1
X
i
. Then g
n
∈ L
q
, g
n
→ g and [g
n
[ ≤ [g[ for each n. Thus lim[g − g
n
[
q
dµ = 0. It follows
that lim[[g −g
n
[[
q
= 0 and g ∈ L
q
.
44. Let 'E
n
` be a sequence of disjoint measurable sets and for each n let f
n
be a function in L
p
, 1 ≤
p < ∞, that vanishes outside E
n
. Set f =
¸
f
n
. Then
[f[
p
=
[
¸
f
n
[
p
=
¸
[f
n
[
p
=
¸
[f
n
[
p
=
¸
[[f
n
[[
p
. Hence f ∈ L
p
if and only if
¸
[[f
n
[[
p
< ∞. In this case, [[f[[
p
=
¸
[[f
n
[[
p
. Also, since the
norm is continuous, we have [[
¸
n
i=1
f
i
[[
p
→ [[f[[
p
so [[f −
¸
n
i=1
f
i
[[
p
→ 0 (c.f. Q6.16).
45. For g ∈ L
q
, let F be the linear functional on L
p
deﬁned by F(f) =
fg dµ. By the H¨older
inequality, we have [F(f)[ = [
fg dµ[ ≤
[fg[ dµ ≤ [[f[[
p
[[g[[
q
so [[F[[ ≤ [[g[[
q
. For 1 < p < ∞, let
f = [g[
q/p
(sgng). Then [f[
p
= [g[
q
= fg so f ∈ L
p
and [[f[[
p
= [[g[[
p/q
q
. Now [F(f)[ = [
fg dµ[ =
[g[
q
dµ = [[g[[
q
q
= [[g[[
q
[[f[[
p
. Thus [[F[[ ≥ [[g[[
q
. If q = 1 and p = ∞, we may assume [[g[[
1
> 0.
Let f = sgng. Then f ∈ L
∞
, [[f[[
∞
= 1 and [F(f)[ = [
fg dµ[ =
[g[ dµ = [[g[[
1
= [[f[[
∞
[[g[[
1
so
[[F[[ ≥ [[g[[
1
. If q = ∞ and p = 1, given ε > 0, let E = ¦x : g(x) > [[g[[
∞
− ε¦ and let f = χ
E
. Then
f ∈ L
1
, [[f[[
1
=
[f[ dµ = µ(E) and [F(f)[ = [
fg dµ[ = [
E
g dµ[ ≥ ([[g[[
∞
−ε)[[f[[
1
so [[F[[ ≥ [[g[[
∞
.
46a. Let µ be the counting measure on a countable set X. We may enumerate the elements of X by
'x
n
`. By considering simple functions, we see that [f[
p
is integrable if and only if
¸
[f(x
n
)[
p
< ∞.
Hence L
p
(µ) =
p
.
*46b.
*47a.
*47b.
*48. Let A and B be uncountable sets with diﬀerent numbers of elements and let X = AB. Let B be
the collection of subsets E of X such that for every horizontal or vertical line L either E ∩ L or E
c
∩ L
is countable. Clearly ∅ ∈ B. Suppose E ∈ B. By the symmetry in the deﬁnition, E
c
∈ B. Suppose
'E
n
` is a sequence of sets in B. For every horizontal or vertical line L, (
¸
E
n
) ∩ L =
¸
(E
n
∩ L) and
(
¸
E
n
)
c
∩ L =
¸
E
c
n
∩ L. If E
n
∩ L is countable for all n, then (
¸
E
n
) ∩ L is countable. Otherwise,
E
c
n
∩ L is countable for some n and (
¸
E
n
)
c
∩ L is countable. Thus
¸
E
n
∈ B. Hence B is a σalgebra.
Let µ(E) be the number of horizontal and vertical lines L for which E
c
∩ L is countable and ν(E)
be the number of horizontal lines with E
c
∩ L countable. Clearly µ(∅) = 0 = ν(∅) since A and B are
uncountable. Suppose 'E
n
` is a sequence of disjoint sets in B. Then µ(
¸
E
n
) is the number of horizontal
and vertical lines L for which
¸
E
c
n
∩ L is countable. Note that if E
c
n
∩ L is countable, then E
m
∩ L is
countable for m = n since E
m
⊂ E
c
n
. If E
c
n
∩ L is countable, then
¸
E
c
n
∩ L is countable. On the other
hand, if
¸
E
c
n
∩ L is countable, then E
c
n
∩ L is countable for some n, for otherwise we have (
¸
E
n
) ∩ L
being countable. It follows that µ(
¸
E
n
) =
¸
µ(E
n
). Thus µ is a measure on B and similarly, ν is a
measure on B.
Deﬁne a bounded linear functional F on L
1
(µ) by setting F(f) =
f dν.
12 Measure and Outer Measure
12.1 Outer measure and measurability
1. Suppose ¯ µ(E) = µ
∗
(E) = 0. For any set A, we have µ
∗
(A∩E) = 0 since A∩E ⊂ E. Also, A∩E
c
⊂ E
so µ
∗
(E) ≥ µ
∗
(A∩ E
c
) = µ
∗
(A∩ E) +µ
∗
(A∩ E
c
). Thus E is measurable. If F ⊂ E, then µ
∗
(F) = 0 so
F is measurable. Hence ¯ µ is complete.
2. Suppose that 'E
i
` is a sequence of disjoint measurable sets and E =
¸
E
i
. For any set A, we have
µ
∗
(A ∩ E) ≤
¸
µ
∗
(A ∩ E
i
) by countable subadditivity. Now µ
∗
(A ∩ (E
1
∪ E
2
)) ≥ µ
∗
(A ∩ (E
1
∪ E
2
) ∩
E
1
) +µ
∗
(A∩(E
1
∪E
2
) ∩E
c
1
) = µ
∗
(A∩E
1
) +µ
∗
(A∩E
2
) by measurability of E
1
. By induction we have
µ
∗
(A ∩
¸
n
i=1
E
i
) ≥
¸
n
i=1
µ
∗
(A ∩ E
i
) for all n. Thus µ
∗
(A ∩ E) ≥ µ
∗
(A ∩
¸
n
i=1
E
i
) ≥
¸
n
i=1
µ
∗
(A ∩ E
i
)
for all n so µ
∗
(A∩ E) ≥
¸
µ
∗
(A∩ E
i
). Hence µ
∗
(A∩ E) =
¸
µ
∗
(A∩ E
i
).
83
12.2 The extension theorem
*3. Let X be the set of rational numbers and A be the algebra of ﬁnite unions of intervals of the
form (a, b] with µ(a, b] = ∞ and µ(∅) = 0. Note that the smallest σalgebra containing A will contain
onepoint sets. Let k be a positive number. Deﬁne µ
k
(A) = k[A[. Then µ
k
is a measure on the smallest
σalgebra containing A and extends µ.
4a. If A is the union of each of two ﬁnite disjoint collections ¦C
i
¦ and ¦D
j
¦ of sets in (, then for
each i, we have µ(C
i
) =
¸
j
µ(C
i
∩ D
j
). Similarly, for each j, we have µ(D
j
) =
¸
i
µ(C
i
∩ D
j
). Thus
¸
i
µ(C
i
) =
¸
i
¸
j
µ(C
i
∩ D
j
) =
¸
j
¸
i
µ(C
i
∩ D
j
) =
¸
j
µ(D
j
).
4b. Deﬁning µ(A) =
¸
n
i=1
µ(C
i
) whenever A is the disjoint union of the sets C
i
∈ (, we have a ﬁnitely
additive set function on A. Thus µ(
¸
C
i
) ≥ µ(
¸
n
i=1
C
i
) =
¸
n
i=1
µ(C
i
) for all n so µ(
¸
C
i
) ≥
¸
µ(C
i
).
Condition (ii) gives the reverse inequality µ(
¸
C
i
) ≤
¸
µ(C
i
) so µ is countably additive.
(*) Proof of Proposition 9
5a. Let ( be a semialgebra of sets and A the smallest algebra of sets containing (. The union of two ﬁnite
unions of sets in ( is still a ﬁnite union of sets in (. Also, (
¸
n
i=1
C
i
)
c
=
¸
n
i=1
C
c
i
is a ﬁnite intersection
of ﬁnite unions of sets in (, which is then a ﬁnite union of sets in (. Thus the collection of ﬁnite unions
of sets in ( is an algebra containing (. Furthermore, if A
t
is an algebra containing (, then it contains all
ﬁnite unions of sets in (. Hence A is comprised of sets of the form A =
¸
n
i=1
C
i
.
5b. Clearly (
σ
⊂ A
σ
. On the other hand, since each set in A is a ﬁnite union of sets in (, we have
A
σ
⊂ (
σ
. Hence A
σ
= (
σ
.
6a. Let A be a collection of sets which is closed under ﬁnite unions and ﬁnite intersections. Countable
unions of sets in A
σ
are still countable unions of sets in A. Also, if A
i
, B
j
∈ A, then (
¸
i
A
i
) ∩(
¸
j
B
j
) =
¸
i,j
(A
i
∩ B
j
), which is a countable union of sets in A. Hence A
σ
is closed under countable unions and
ﬁnite intersections.
6b. If
¸
B
i
∈ A
σδ
, then
¸
B
i
=
¸
n
¸
n
i=1
B
i
where
¸
n
i=1
B
i
∈ A
σ
and
¸
n
i=1
B
i
⊃
¸
n+1
i=1
B
i
for each n.
Hence each set in A
σδ
is the intersection of a decreasing sequence of sets in A
σ
.
7. Let µ be a ﬁnite measure on an algebra A, and µ
∗
the induced outer measure. Suppose that for each
ε > 0 there is a set A ∈ A
σδ
, A ⊂ E, such that µ
∗
(E ` A) < ε. Note that A is measurable. Thus for
any set B, we have µ
∗
(B) = µ
∗
(B ∩ A) + µ
∗
(B ∩ A
c
) ≥ µ
∗
(B ∩ E) − µ
∗
(B ∩ (E ` A)) + µ
∗
(B ∩ E
c
) >
µ
∗
(B ∩ E) +µ
∗
(B ∩ E
c
) −ε. Thus µ
∗
(B) ≥ µ
∗
(B ∩ E) +µ
∗
(B ∩ E
c
) and E is measurable.
Conversely, suppose E is measurable. Given ε > 0, there is a set B ∈ A
σ
such that E
c
⊂ B and
µ
∗
(B) ≤ µ
∗
(E
c
) + ε. Let A = B
c
. Then A ∈ A
δ
and A ⊂ E. Furthermore, µ
∗
(E ` A) = µ
∗
(E ∩ B) =
µ
∗
(B) −µ
∗
(B ∩ E
c
) = µ
∗
(B) −µ
∗
(E
c
) ≤ ε.
8a. If we start with an outer measure µ
∗
on X and form the induced measure ¯ µ on the µ
∗
measurable
sets, we can use ¯ µ to induce an outer measure µ
+
. For each set E, we have µ
∗
(E) ≤
¸
µ
∗
(A
i
) for any
sequence of µ
∗
measurable sets A
i
with E ⊂
¸
A
i
. Taking the inﬁmum over all such sequences, we have
µ
∗
(E) ≤ inf
¸
µ
∗
(A
i
) = inf
¸
¯ µ(A
i
) = µ
+
(E).
8b. Suppose there is a µ
∗
measurable set A ⊃ E with µ
∗
(A) = µ
∗
(E). Then µ
+
(E) ≥ ¯ µ(A) = µ
∗
(A) =
µ
∗
(E). Thus by part (a), we have µ
+
(E) = µ
∗
(E).
Conversely, suppose µ
+
(E) = µ
∗
(E). For each n, there is a µ
∗
measurable set (a countable union of µ
∗

measurable sets) A
n
with E ⊂ A
n
and µ
∗
(A
n
) = ¯ µ(A
n
) ≤ µ
+
(E) + 1/n = µ
∗
(E) + 1/n. Let A =
¸
A
n
.
Then A is µ
∗
measurable, E ⊂ A and µ
∗
(A) ≤ µ
∗
(E) + 1/n for all n so µ
∗
(A) = µ
∗
(E).
8c. If µ
+
(E) = µ
∗
(E) for each set E, then by part (b), for each set E, there is a µ
∗
measurable set A
with A ⊃ E and µ
∗
(A) = µ
∗
(E). In particular, µ
∗
(A) ≤ µ
∗
(E) +ε so µ
∗
is regular.
Conversely, if µ
∗
is regular, then for each set E and each n, there is a µ
∗
measurable set A
n
with A
n
⊃ E
and µ
∗
(A
n
) ≤ µ
∗
(E) + 1/n. Let A =
¸
A
n
. Then A is µ
∗
measurable, A ⊃ E and µ
∗
(A) = µ
∗
(E). By
part (b), µ
+
(E) = µ
∗
(E) for each set E.
8d. If µ
∗
is regular, then µ
+
(E) = µ
∗
(E) for every E by part (c). In particular, µ
∗
is induced by the
measure ¯ µ on the σalgebra of µ
∗
measurable sets.
Conversely, suppose µ
∗
is induced by a measure µ on an algebra A. For each set E and any ε > 0, there
is a sequence 'A
i
` of sets in A with E ⊂
¸
A
i
and µ
∗
(
¸
A
i
) ≤ µ
∗
(E) + ε. Each A
i
is µ
∗
measurable so
¸
A
i
is µ
∗
measurable. Hence µ
∗
is regular.
84
8e. Let X be a set consisting of two points a and b. Deﬁne µ
∗
(∅) = µ
∗
(¦a¦) = 0 and µ
∗
(¦b¦) = µ
∗
(X) =
1. Then µ
∗
is an outer measure on X. The set X is the only µ
∗
measurable set containing ¦a¦ and
µ
∗
(X) = 1 > µ
∗
(¦a¦) + 1/2. Hence µ
∗
is not regular.
9a. Let µ
∗
be a regular outer measure. The measure ¯ µ induced by µ
∗
is complete by Q1. Let E be locally
¯ µmeasurable. Then E ∩B is µ
∗
measurable for any µ
∗
measurable set B with µ
∗
(B) = ¯ µ(B) < ∞. We
may assume µ
∗
(E) < ∞. Since µ
∗
is regular, it is induced by a measure on an algebra A so there is a
set B ∈ A
σ
with E ⊂ B and µ
∗
(B) ≤ µ
∗
(E) + 1 < ∞. Thus E = E ∩ B is µ
∗
measurable.
*9b.
10. Let µ be a measure on an algebra A and ¯ µ the extension of it given by the Carath´eodory process.
Let E be measurable with respect to ¯ µ and ¯ µ(E) < ∞. Given ε > 0, there is a countable collection
¦A
n
¦ of sets in A such that E ⊂
¸
A
n
and
¸
µ(A
n
) ≤ µ
∗
(E) + ε/2. There exists N such that
¸
∞
n=N+1
µ(A
n
) < ε/2. Let A =
¸
N
n=1
µ(A
n
). Then A ∈ A and ¯ µ(A∆E) = ¯ µ(A ` E) + ¯ µ(E ` A) ≤
¸
µ(A
n
) −µ
∗
(E) +
¸
∞
n=N+1
µ(A
n
) < ε.
11a. Let µ be a measure on A and ¯ µ its extension. Let ε > 0. If f is ¯ µintegrable, then there is
a simple function
¸
n
i=1
c
i
χ
E
i
where each E
i
is µ
∗
measurable and
[f −
¸
n
i=1
c
i
χ
E
i
[ d¯ µ < ε/2. The
simple function may be taken to vanish outside a set of ﬁnite measure so we may assume each E
i
has
ﬁnite µ
∗
measure. For each E
i
, there exists A
i
∈ A such that ¯ µ(A
i
∆E
i
) < ε/2n. Consider the Asimple
function ϕ =
¸
n
i=1
c
i
χ
A
i
. Then
[f −ϕ[ d¯ µ < ε.
*11b.
12.3 The LebesgueStieltjes integral
12. Let F be a monotone increasing function continuous on the right. Suppose (a, b] ⊂
¸
∞
i=1
(a
i
, b
i
].
Let ε > 0. There exists η
i
> 0 such that F(b
i
+ η
i
) < F(b
i
) + ε2
−i
. There exists δ > 0 such that
F(a + δ) < F(a) + ε. Then the open intervals (a
i
, b
i
+ η
i
) cover the closed interval [a + δ, b]. By the
HeineBorel Theorem, a ﬁnite subcollection of the open intervals covers [a +δ, b]. Pick an open interval
(a
1
, b
1
+ η
1
) containing a + δ. If b
1
+ η
1
≤ b, then there is an interval (a
2
, b
2
+ η
2
) containing b
1
+ η
1
.
Continuing in this fashion, we obtain a sequence (a
1
, b
1
+η
1
), . . . , (a
k
, b
k
+η
k
) from the ﬁnite subcollection
such that a
i
< b
i−1
+η
i−1
< b
i
+eta
i
. The process must terminate with some interval (a
k
, b
k
+η
k
but it
terminates only if b ∈ (a
k
, b
k
+η
k
). Thus
¸
∞
i=1
F(b
i
) −F(a
i
) ≥ (F(b
k
+η
k
) −ε2
−i
−F(a
k
)) +(F(b
k−1
+
η
k−1
) −ε2
−i+1
−F(a
k−1
)) + +(F(b
1
+η
1
) −ε2
−1
−F(a
1
)) > F(b
k
+η
k
) −F(a
1
) > F(b) −F(a +δ).
Now F(b)−F(a) = (F(b)−F(a+δ))+(F(a+δ)−F(a)) <
¸
∞
i=1
F(b
i
)−F(a
i
)+ε. Hence F(b)−F(a) ≤
¸
∞
i=1
F(b
i
) −F(a
i
). If (a, b] is unbounded, we may approximate it by a bounded interval by considering
limits.
(*) Proof of Lemma 11
13. Let F be a monotone increasing function and deﬁne F
∗
(x) = lim
y→x
+ F(y). Clearly F
∗
is monotone
increasing since F is. Given ε > 0, there exists δ > 0 such that F(y) −F
∗
(x) < ε whenever y ∈ (x, x+δ).
Now when z ∈ (y, y + δ
t
) where 0 < δ
t
< x + δ −y, we have F(z) −F
∗
(x) < ε. Thus F
∗
(y) −F
∗
(x) ≤
F(z) − F
∗
(x) < ε. Hence F
∗
is continuous on the right. If F is continuous on the right at x, then
F
∗
(x) = lim
y→x
+ F(y) = F(x). Thus since F
∗
is continuous on the right, we have (F
∗
)
∗
= F
∗
. Suppose
F and G are monotone increasing functions which agree wherever they are both continuous. Then F
∗
and G
∗
agree wherever both F and G are continuous. Since they are monotone, their points of continuity
are dense. It follows that F
∗
= G
∗
since F
∗
and G
∗
are continuous on the right.
14a. Let F be a bounded function of bounded variation. Then F = G−H where G and H are monotone
increasing functions. There are unique Baire measures µ
G
and µ
H
such that µ
G
(a, b] = G
∗
(b) − G
∗
(a)
and µ
H
= H
∗
(b) −H
∗
(a). Let ν = µ
G
−µ
H
. Then ν is a signed Baire measure and ν(a, b] = µ
G
(a, b] −
µ
H
(a, b] = (G
∗
(b)−G
∗
(a))−(H
∗
(b)−H
∗
(a)) = (G(b+)−G(a+))−(H(b+)−H(a+)) = F(b+)−F(a+).
14b. The signed Baire measure in part (a) has a Jordan decomposition ν = ν
+
−ν
−
. Now F = G−H
where G corresponds to the positive variation of F and H corresponds to the negative variation of F.
Then G and H give rise to Baire measures µ
G
and µ
H
with ν = µ
G
− µ
H
. By the uniqueness of the
Jordan decomposition, ν
+
= µ
G
and ν
−
= µ
H
so ν
+
and ν
−
correspond to the positive and negative
variations of F.
14c. If F is of bounded variation, deﬁne
ϕ dF =
ϕ dν =
ϕ dν
+
−
ϕ dν
−
where ν is the signed
85
Baire measure in part (a).
14d. Suppose [ϕ[ ≤ M and the total variation of F is T. Then [
ϕ dF[ = [
ϕ dν[ ≤ MT since ν
+
and ν
−
correspond to the positive and negative variations of F and T = P +N.
15a. Let F be the cumulative distribution function of the Baire measure ν and assume that F is
continuous. Suppose the interval (a, b) is in the range of F. Since F is monotone, F
−1
[(a, b)] = (c, d)
where F(c) = a and F(d) = b. Thus m(a, b) = b − a = F(d) − F(c) = ν[F
−1
[(a, b)]]. Since the class
of Borel sets is the smallest σalgebra containing the algebra of open intervals, the uniqueness of the
extension in Theorem 8 gives the result for general Borel sets.
15b. For a discontinuous cumulative distribution function F, note that the set C of points at which F
is continuous is a G
δ
and thus a Borel set. Similarly, the set D of points at which F is discontinuous
is a Borel set. Furthermore, D is at most countable since F is monotone. Thus F
−1
[D] is also at most
countable. Now for a Borel set E, we have m(E) = m(E∩C)+m(E∩D) = m(E∩C) = ν[F
−1
[E∩C]] =
ν[F
−1
[E ∩ C]] +ν[F
−1
[E ∩ D]] = ν[F
−1
[E]].
16. Let F be a continuous increasing function on [a, b] with F(a) = c, F(b) = d and let ϕ be a nonnegative
Borel measurable function on [c, d]. Now F is the cumulative distribution function of a ﬁnite Baire
measure ν. First assume that ϕ is a characteristic function χ
E
of a Borel set. Then
b
a
ϕ(F(x)) dF(x) =
b
a
χ
E
(F(x)) dF(x) =
b
a
χ
E
(F(x)) dν = ν[F
−1
[E]] = m(E) =
d
c
χ
E
(y) dy =
d
c
ϕ(y) dy. Since simple
functions are ﬁnite linear combinations of characteristic functions of Borel sets, the result follows from
the linearity of the integrals. For a general ϕ, there is an increasing sequence of nonnegative simple
functions converging pointwise to ϕ and the result follows from the Monotone Convergence Theorem.
*17a. Suppose a measure µ is absolutely continuous with respect to Lebesgue measure and let F be
its cumulative distribution function. Given ε > 0, there exists δ > 0 such that µ(E) < ε whenever
m(E) < δ. For any ﬁnite collection ¦(x
i
, x
t
i
)¦ of nonoverlapping intervals with
¸
n
i=1
[x
t
i
− x
i
[ < δ, we
have µ(
¸
n
i=1
(x
i
, x
t
i
)) < ε. i.e.
¸
n
i=1
[F(x
t
i
) −F(x
i
)[ < ε. Thus F is absolutely continuous.
Conversely, suppose F is absolutely continuous. Given ε > 0, there exists δ > 0 such that
¸
n
i=1
[f(x
t
i
) −
f(x
i
)[ < ε for any ﬁnite collection ¦(x
i
, x
t
i
)¦ of nonoverlapping intervals with
¸
n
i=1
[x
t
i
−x
i
[ < δ. Let E
be a measurable set with m(E) < δ/2. There is a sequence of open intervals 'I
n
` such that E ⊂
¸
I
n
and
¸
m(I
n
) < δ. We may assume the intervals are nonoverlapping. Now µ(E) ≤ µ(
¸
I
n
) =
¸
µ(I
n
) =
¸
(F(b
n
) − F(a
n
)) where I
n
= (a
n
, b
n
). Since
¸
k
n=1
(b
n
− a
n
) =
¸
k
n=1
m(I
n
) < δ for each n, we have
¸
k
n=1
(F(b
n
) −F(a
n
)) < ε for each n. Thus µ(E) ≤
¸
(F(b
n
) −F(a
n
)) < ε and µ << m.
17b. If µ is absolutely continuous with respect to Lebesgue measure, then its cumulative distribution
function F is absolutely continuous so by Theorem 5.14, we have F(x) =
x
a
F
t
(t) dt + F(a). i.e.
µ(a, x] =
x
a
F
t
(t) dt. It follows that µ(E) =
E
F
t
dt for any measurable set E. By uniqueness of the
RadonNikodym derivative, [
dµ
dm
] = F
t
a.e.
17c. If F is absolutely continuous, then the Baire measure for which µ(a, b] = F(b) −F(a) is absolutely
continuous with respect to Lebesgue measure and F
t
is its RadonNikodym derivative. By Q11.34a, we
have
f dF =
f dµ =
fF
t
dx.
*18. Riemann’s Convergence Criterion: Let f be a nonnegative monotone decreasing function on
(0, ∞), g a nonnegative increasing function on (0, ∞) and 'a
n
` a nonnegative sequence. Suppose that
for each x ∈ (0, ∞) the number of n such that a
n
≥ f(x) is at most g(x).
12.4 Product measures
19. Let X = Y be the set of positive integers, A = B = {(X), and let ν = µ be the counting measure.
Fubini’s Theorem: Let f be a function on X Y such that
¸
x,y
[f(x, y)[ < ∞.
(i) For (almost) all x, the function f
x
(y) = f(x, y) satisﬁes
¸
y
[f
x
(y)[ < ∞.
(i’) For (almost) all y, the function f
y
(x) = f(x, y) satisﬁes
¸
x
[f
y
(x)[ < ∞.
(ii)
¸
x
¸
y
f(x, y) < ∞
(ii’)
¸
y
¸
x
f(x, y) < ∞
(iii)
¸
x
¸
y
f(x, y) =
¸
x,y
f(x, y) =
¸
y
¸
x
f(x, y)
Tonelli’s Theorem: Let f be a nonnegative function on X Y .
Parts (i), (i’), (ii) and (ii’) are trivial.
86
(iii)
¸
x
¸
y
f(x, y) =
¸
x,y
f(x, y) =
¸
y
¸
x
f(x, y)
20. Let (X, B, µ) be any σﬁnite measure space and Y the set of positive integers with ν the counting
measure. Let f be a nonnegative measurable function on X Y . For each n ∈ Y , the function
f
n
(x) = f(x, n) is a nonnegative measurable function on X. Both Theorem 20 and Corollary 11.14 give
the result that
X
¸
n
f(x, n) =
¸
n
X
f(x, n). Note that Corollary 11.14 is valid even if µ is not σﬁnite
so the Tonelli Theorem is true without σﬁniteness if (Y, B, ν) is this special measure space.
21. Let X = Y = [0, 1] and let µ = ν be the Lebesgue measure. Note that X Y satisﬁes the second
axiom of countability and has a countable basis consisting of measurable rectangles. Thus each open set
in X Y is a countable union of measurable rectangles and is itself measurable. Since each open set is
measurable and the collection of Borel sets is the smallest σalgebra containing all the open sets, each
Borel set in X Y is measurable.
22. Let h and g be integrable functions on X and Y , and deﬁne f(x, y) = h(x)g(y). If h = χ
A
and g = χ
B
where A ⊂ X and B ⊂ Y are measurable sets, then f = χ
A·B
where A B is a measurable rectangle.
Thus f is integrable on X Y and
X·Y
f d(µ ν) = (µ ν)(A B) = µ(A)ν(B) =
X
h dµ
Y
g dν.
It follows that the result holds for simple functions and thus nonnegative integrable functions. For
general integrable functions h and g, note that f
+
= h
+
g
+
+h
−
g
−
and f
−
= h
+
g
−
+h
−
g
+
. Thus f is
integrable on XY and
X·Y
f d(µν) =
X·Y
f
+
d(µν)−
X·Y
f
−
d(µν) =
X
h
+
dµ
Y
g
+
dν+
X
h
−
dµ
Y
g
−
dν −
X
h
+
dµ
Y
g
−
dν −
X
h
−
dµ
Y
g
+
dν =
X
h
+
dµ
Y
g dν −
X
h
−
dµ
Y
g dν =
X
h dµ
Y
g dν.
23. Suppose that instead of assuming µ and ν to be σﬁnite, we merely assume that ¦'x, y` : f(x, y) = 0¦
is a set of σﬁnite measure. Then there is still an increasing sequence of simple functions each vanishing
outside a set of ﬁnite measure and converging pointwise to f. Thus the proof of Tonelli’s Theorem is
still valid and the theorem is still true.
24. Let X = Y be the positive integers and µ = ν be the counting measure. Let
f(x, y) =
2 −2
−x
if x = y,
−2 + 2
−x
if x = y + 1,
0 otherwise.
Then
¸
x
¸
y
f(x, y) = f(1, 1) =
3
2
but
¸
y
¸
x
f(x, y) =
¸
y
(2−2
−y
)+(−2+2
−y−1
) =
¸
y
2
−y−1
−2
−y
=
−
¸
y
2
−y−1
= −
1
2
.
(*) We cannot remove the hypothesis that f be nonnegative from Tonelli’s Theorem or that f be inte
grable from Fubini’s Theorem.
25. Let X = Y be the interval [0, 1], with A = B the class of Borel sets. Let µ be Lebesgue measure
and ν the counting measure. Let ∆ = ¦'x, y` ∈ X Y : x = y¦ be the diagonal. Let I
j,n
= [
j−1
n
,
j
n
]
for j = 1, . . . , n and let I
n
=
¸
n
j=1
I
j,n
. Note that ∆ =
¸
n
I
n
. Hence ∆ is measurable (and in fact an
R
σδ
). Now
X
[
Y
χ
∆
dν] dµ =
X
ν(¦y : y = x¦) dµ =
X
1 dµ = 1 but
Y
[
X
χ
∆
dµ] dν =
Y
µ(¦x :
x = y¦) dν =
Y
0 dν = 0. Let ∆ ⊂
¸
(A
n
B
n
) where A
n
∈ A and B
n
∈ B. Then some B
n
must be
inﬁnite so that (µ ν)(A
n
B
n
) = ∞. Thus
¸
(µ ν)(A
n
B
n
) = ∞. By deﬁnition of outer measure,
it follows that
X·Y
χ
∆
d(µ ν) = (µ ν)(∆) = ∞.
(*) We cannot remove the hypothesis that f be integrable from Fubini’s Theorem or that µ and ν be
σﬁnite from Tonelli’s Theorem.
26. Let X = Y be the set of ordinals less than or equal to the ﬁrst uncountable ordinal Ω. Let A = B be
the σalgebra consisting of all countable sets and their complements. Deﬁne µ = ν by setting µ(E) = 0
if E is countable and µ(E) = 1 otherwise. Deﬁne a subset S of X Y by S = ¦'x, y` : x < y¦.
Now S
x
= ¦y : y > x¦ and S
y
= ¦x : x < y¦ are measurable for each x and y. Let f be the
characteristic function of S. Then
Y
[
X
f dµ(x)] dν(y) =
Y
µ(S
y
) dν(Y ) =
¦Ω¦
1 dν(y) = 0 and
X
[
Y
f dν(y)] dµ(x) =
X
ν(S
x
) dµ(x) =
X
1 dµ(x) = 1.
If we assume the continuum hypothesis, i.e. that X can be put in oneone correspondence with [0, 1],
then we can take f to be a function on the unit square such that f
x
and f
y
are bounded and measurable
for each x and y but such that the conclusions of the Fubini and Tonelli Theorems do not hold.
(*) The hypothesis that f be measurable with respect to the product measure cannot be omitted from
the Fubini and Tonelli Theorems even if we assume the measurability of f
y
and f
x
and the integrability
87
of
f(x, y) dν(y) and
f(x, y) dµ(x).
*27. Suppose (X, A, µ) and (Y, B, ν) are two σﬁnite measure spaces. Then the extension of the
nonnegative set function λ(A B) = µ(A)ν(B) on the semialgebra of measurable rectangles to A B
is σﬁnite. It follows from the Carath´eodory extension theorem that the product measure is the only
measure on A B which assigns the value µ(A)ν(B) to each measurable rectangle AB.
28a. Suppose E ∈ AB. If µ and ν are σﬁnite, then so is µν so E =
¸
(E∩F
i
) where (µν)(F
i
) < ∞
for each i. By Proposition 18, (E ∩ F
i
)
x
is measurable for almost all x so E
x
=
¸
(E ∩ F
i
)
x
∈ B for
almost all x.
28b. Suppose f is measurable with respect to A B. For any α, we have E = ¦'x, y` : f(x, y) > α¦ ∈
AB so E
x
∈ B for almost all x. Now E
x
= ¦y : f(x, y) > α¦ = ¦y : f
x
(y) > α¦. Thus f
x
is measurable
with respect to B for almost all x.
29a. Let X = Y = R and let µ = ν be Lebesgue measure. Then µ ν is twodimensional Lebesgue
measure on X Y = R
2
. For each measurable subset E of R, let σ(E) = ¦'x, y` : x −y ∈ E¦. If E is an
open set, then σ(E) is open and thus measurable. If E is a G
δ
with E =
¸
E
i
where each E
i
is open,
then σ(E) =
¸
σ(E
i
), which is measurable. If E is a set of measure zero, then σ(E) is a set of measure
zero and is thus measurable. A general measurable set E is thhe diﬀerence of a G
δ
set A and a set B of
measure zero and σ(E) = σ(A) ` σ(B) so σ(E) is measurable.
29b. Let f be a measurable function on R and deﬁne the function F by F(x, y) = f(x−y). For any α, we
have ¦'x, y` : F(x, y) > α¦ = ¦'x, y` : f(x −y) > α¦ = ¦'x, y` : x −y ∈ f
−1
[(α, ∞)]¦ = σ(f
−1
[(α, ∞)]).
The interval (α, ∞) is a Borel set so f
−1
[(α, ∞)] is measurable. It follows from part (a) that ¦'x, y` :
F(x, y) > α¦ is measurable. Hence F is a measurable function on R
2
.
29c. Let f and g be integrable functions on R and deﬁne the function ϕ by ϕ(y) = f(x − y)g(y).
By Tonelli’s Theorem,
X·Y
[f(x − y)g(y)[ dxdy =
Y
[
X
[f(x − y)g(y)[ dx] dy =
Y
[g(y)[[
X
[f(x −
y)[ dx] dy =
Y
[g(y)[[
X
[f(x)[ dx] dy =
[f[
[g[. Thus the function [f(x − y)g(y)[ is integrable. By
Fubini’s Theorem, for almost all x, the function ϕ is integrable. Let h =
Y
ϕ. Then
[h[ =
X
[
Y
ϕ[ ≤
X
Y
[ϕ[ =
X·Y
[ϕ[ ≤
[f[
[g[.
30a. Let f and g be functions in L
1
(−∞, ∞) and deﬁne f ∗ g to be the function deﬁned by
f(y −
x)g(x) dx. If f(y − x)g(x) is integrable at y, then deﬁne F(x) = f(y − x)g(x) and G(x) = F(y − x).
Then G is integrable and
G(x) dx =
F(x) dx. i.e.
f(x)g(y − x) dx =
f(y − x)g(x) dx. Thus
for y ∈ R, f(y −x)g(x) is integrable if and only if f(x)g(y −x) is integrable and their integrals are the
same in this case. When f(y −x)g(x) is not integrable, (f ∗ g)(y) = (g ∗ f)(y) = 0 since the function is
integrable for almost all y. Hence f ∗ g = g ∗ f.
30b. For x, y ∈ R such that f(y − x − u)g(u) is integrable, deﬁne F(u) = f(y − x − u)g(u). Consider
G(u) = F(u−x). Then G is integrable and
f(y−u)g(u−x) du =
G(u) du =
F(u) du = (f∗g)(y−x).
The function H(u, x) = f(y−u)g(u−x)h(x) is integrable. Then ((f ∗g)∗h)(y) =
(f ∗g)(y−x)h(x) dx =
[
f(y − u)g(u − x) du]h(x) dx =
f(y − u)g(u − x)h(x) dudx =
[
f(y − u)g(u − x)h(x) dx] du =
f(y −u)(g ∗ h)(u) du = (f ∗ (g ∗ h))(y).
30c. For f ∈ L
1
, deﬁne
ˆ
f by
ˆ
f(s) =
e
ist
f(t) dt. Then [
ˆ
f[ ≤
[f[ so
ˆ
f is a bounded complex function.
Furthermore, for any s ∈ R, we have
f ∗ g(s) =
e
ist
(f∗g)(t) dt =
e
ist
[
f(t−x)g(x) dx] dt =
[
f(t−
x)e
is(t−x)
g(x)e
isx
dx] dt =
[
f(t − x)e
is(t−x)
g(x)e
isx
dt] dx =
[
f(t − x)e
is(t−x)
dt]g(x)e
isx
dx =
[
f(u)e
isu
du]g(x)e
isx
dx =
ˆ
f(s)ˆ g(s).
31. Let f be a nonnegative integrable function on (−∞, ∞) and let m
2
be twodimensional Lebesgue
measure on R
2
. There is an increasing sequence of nonnegative simple functions 'ϕ
n
` converging point
wise to f. Let E
n
= ¦'x, y` : 0 < y < ϕ
n
(x)¦. Then ¦'x, y` : 0 < y < f(x)¦ =
¸
E
n
. Write
ϕ
n
=
¸
k
n
i=1
a
(n)
i
χ
K
(n)
i
. We may assume the K
(n)
i
are disjoint. Also, a
(n)
i
> 0 for each i. Then
E
n
= (F
(n)
1
(0, a
(n)
1
)) ∪ ∪ (F
(n)
k
n
(0, a
(n)
k
n
)). Hence E
n
is measurable so ¦'x, y` : 0 < y <
f(x)¦ is measurable. Furthermore, m
2
(E
n
) → m
2
¦'x, y` : 0 < y < f(x)¦. On the other hand,
m
2
(E
n
) =
¸
k
n
i=1
m(F
(n)
i
)m(0, a
(n)
i
) =
¸
n
i=1
a
(n)
i
m(F
(n)
i
) =
ϕ
n
dx. By Monotone Convergence The
orem,
ϕ
n
dx →
f dx so m
2
¦'x, y` : 0 < y < f(x)¦ =
f dx. Now ¦'x, y` : 0 ≤ y ≤ f(x)¦ =
¦'x, y` : 0 < y < f(x)¦ ∪ ¦'x, 0` : x ∈ R¦ ∪ ¦'x, f(x)` : x ∈ R¦. Note that m
2
¦'x, 0` : x ∈ R¦ =
m(R)m¦0¦ = ∞ 0 = 0. Also, m
2
¦'x, f(x)` : x ∈ R¦ = 0 by considering a covering of the set by
88
A
n
∪ [(n − 1)ε, nε) where A
n
= ¦x ∈ R : f(x) ∈ [(n − 1)ε, nε)¦. Thus ¦'x, y` : 0 ≤ y ≤ f(x)¦ is
measurable and m
2
¦'x, y` : 0 ≤ y ≤ f(x)¦ =
f dx.
Let ϕ(t) = m¦x : f(x) ≥ t¦. Since ¦x : f(x) ≥ t¦ ⊂ ¦x : f(x) ≥ t
t
¦ when t ≥ t
t
, ϕ is a decreas
ing function. Now
∞
0
ϕ(t) dt =
∞
0
[
χ
¦x:f(x)≥t¦
(x) dx] dt =
[
χ
¦x:f(x)≥t¦
(x)χ
[0,∞)
(t) dx] dt =
χ
¦¹x,y):0≤t≤f(x)¦
= m
2
¦'x, y` : 0 ≤ t ≤ f(x)¦ =
f dx.
*32. Let '(X
i
, A
i
, µ
i
)`
n
i=1
be a ﬁnite collection of measure spaces. We can form the product measure
µ
1
µ
n
on the space X
1
X
n
by starting with the semialgebra of rectangles of the form
R = A
1
A
n
and µ(R) =
¸
µ
i
(A
i
), and using the Carath´eodory extension procedure. If E ⊂
X
1
X
n
is covered by a sequence of measurable rectangles R
k
⊂ X
1
X
n
, then R
k
=
R
1
k
∪ R
2
k
where R
1
k
⊂ X
1
X
p
and R
2
k
⊂ X
p+1
X
n
are measurable rectangles. Then
((µ
1
µ
p
)(µ
p+1
µ
n
))
∗
(E) ≤
¸
(µ
1
µ
p
)(R
1
k
)(µ
p+1
µ
n
)(R
2
k
) =
¸
(µ
1
µ
n
)(R
k
)
so ((µ
1
µ
p
) (µ
p+1
µ
n
))
∗
(E) ≤ (µ
1
µ
n
)
∗
(E).
On the other hand, if R = F G is a measurable rectangle in (X
1
X
p
) (X
p+1
X
n
),
then F ⊂
¸
F
k
and G ⊂
¸
G
j
where F
k
is a measurable rectangle in X
1
X
p
and G
j
is a
measurable rectangle in X
p+1
X
n
. Now R ⊂
¸
k,j
(F
k
G
j
) and ((µ
1
µ
p
) (µ
p+1
µ
n
))(R) + ε = (µ
1
µ
p
)(F)(µ
p+1
µ
n
)(G) + ε >
¸
k,j
(µ
1
µ
n
)(F
k
G
j
). Now if
E ⊂ (X
1
X
p
)(X
p+1
X
n
), then E ⊂
¸
R
i
and ((µ
1
µ
p
)(µ
p+1
µ
n
))
∗
(E)+ε >
¸
i
((µ
1
µ
p
)(µ
p+1
µ
n
))(R
i
) >
¸
i
¸
k,j
(µ
1
µ
n
)(F
k
G
j
)−ε
t
≥ (µ
1
µ
n
)
∗
(E)−ε.
Thus ((µ
1
µ
p
) (µ
p+1
µ
n
))
∗
(E) ≥ (µ
1
µ
n
)
∗
(E).
Hence ((µ
1
µ
p
) (µ
p+1
µ
n
))
∗
= µ
1
µ
n
∗
. It then follows that (µ
1
µ
p
)
(µ
p+1
µ
n
) = µ
1
µ
n
.
*33. Let ¦(X
λ
, A
λ
, µ
λ
)¦ be a collection of probability measure spaces.
12.5 Integral operators
*34. By Proposition 21, we have [[T[[ ≤ M
α,β
.
35. Let k(x, y) be a measurable function on XY of absolute operator type (p, q) and g ∈ L
q
(ν). Then
[k[ is of operator type (p, q) so by Proposition 21, for almost all x, the integral
Y
[k(x, y)[g(y) dν exists.
Thus for almost all x, the integral f(x) =
Y
k(x, y)g(y) dν exists. Furthermore, the function f belongs
to L
p
(µ) an [[f[[
p
≤ [[
Y
[k(x, y)[g(y) dν[[
p
≤ [[ [k[ [[
p,q
[[g[[
q
.
(*) Proof of Corollary 22
36. Let g, h and k be functions on R
n
of class L
q
, L
p
and L
r
respectively, with 1/p + 1/q + 1/r = 2.
We may write 1/p = 1 − (1 − λ)/r and 1/q = 1 − λ/r for some 0 ≤ λ ≤ 1. Then k is of covariant type
(p, q) so the integral f(x) =
R
n
k(x, y)g(y) dy exists for almost all x and the function f belongs to L
p
∗
with [[f[[
p
∗ ≤ [[k[[
r
[[g[[
q
by Proposition 21. Now
R
2n
[h(x)k(x − y)g(y)[ dxdy =
R
n
[h(x)f(x)[ dx ≤
[[h[[
p
[[k[[
r
[[g[[
q
.
(*) Proof of Proposition 25
37. Let g ∈ L
q
and k ∈ L
r
, with 1/q + 1/r > 1. Let 1/p = 1/q + 1/r −1. We may write 1/q = 1 −λ/r
and 1/p = 1 − (1 − λ)/r where 0 ≤ λ ≤ 1. Then k is of covariant type (p, q) so the function f(x) =
R
n
k(x −y)g(y) dy is deﬁned for almost all x and [[f[[
p
≤ [[k[[
r
[[g[[
q
by Proposition 21.
(*) Proof of Proposition 26
*38. Let g, h and k be functions on R
n
of class L
q
, L
p
and L
r
with 1/p + 1/q + 1/r ≤ 2.
12.6 Inner measure
39a. Suppose µ(X) < ∞. By deﬁnition, µ
∗
(E) ≥ µ(X) − µ
∗
(E
c
). Conversely, since E and E
c
are
disjoint, we have µ
∗
(E) +µ
∗
(E
c
) ≤ µ
∗
(E ∪ E
c
) = µ(X). Hence µ
∗
(E) = µ(X) −µ
∗
(E
c
).
39b. Suppose A is a σalgebra. If A ∈ A and E ⊂ A, then µ
∗
(E) ≤ µ(A). Thus µ
∗
(E) ≤ inf¦µ(A) :
E ⊂ A, A ∈ A¦. Conversely, for any sequence 'A
i
` of sets in A covering E, we have
¸
A
i
∈ A so
inf¦µ(A) : E ⊂ A, A ∈ A¦ ≤ µ(
¸
A
i
) ≤
¸
µ(A
i
). Thus inf¦µ(A) : E ⊂ A, A ∈ A¦ ≤ µ
∗
(E). Hence
µ
∗
(E) = inf¦µ(A) : E ⊂ A, A ∈ A¦.
If A ∈ A and A ⊂ E, then µ(A) = µ(A) − µ
∗
(A ` E). Thus sup¦µ(A) : A ⊂ E, A ∈ A¦ ≤ µ
∗
(E).
89
Conversely, if A ∈ A and µ
∗
(A`E) < ∞, then for any ε > 0, there is a sequence 'A
i
` of sets in A such that
A`E ⊂
¸
A
i
and
¸
µ(A
i
) < µ
∗
(A`E)+ε. Let B =
¸
A
i
. Then B ∈ A and µ(B) < µ
∗
(A`E)+ε. Also,
A`B ∈ A and A`B ⊂ E. Thus µ(A)−µ
∗
(A`E)−ε = µ(A)−µ(B) = µ(A`B) ≤ sup¦µ(A) : A ⊂ E, A ∈
A¦. It follows that µ
∗
(E) ≤ sup¦µ(A) : A ⊂ E, A ∈ A¦. Hence µ
∗
(E) = sup¦µ(A) : A ⊂ E, A ∈ A¦.
39c. Let µ be Lebesgue measure on R. By part (c), we have µ
∗
(E) = sup¦µ(A) : A ⊂ E, A measurable¦.
If A is measurable and A ⊂ E, then given ε > 0, there is a closed set F ⊂ A with µ
∗
(A ` F) < ε. Thus
µ(A) = µ(F) +µ(A` F) < µ(F) +ε. It follows that µ
∗
(E) ≤ sup¦µ(F) : F ⊂ E, F closed¦. Conversely,
if F ⊂ E and F is closed, then F is measurable so µ(F) ≤ µ
∗
(E). Thus sup¦µ(F) : F ⊂ E, F closed¦ ≤
µ
∗
(E). Hence µ
∗
(E) = sup¦µ(F) : F ⊂ E, F closed¦.
40. Let 'B
i
` be a sequence of disjoint sets in B. Then ¯ µ(
¸
B
i
) = µ
∗
(
¸
B
i
∩ E) + µ
∗
(
¸
B
i
∩ E
c
) =
¸
µ
∗
(B
i
∩E) +
¸
µ
∗
(B
i
∩E
c
) =
¸
¯ µ(B
i
). Also, µ(
¸
B
i
) = µ
∗
(
¸
B
i
∩E) +µ
∗
(
¸
B
i
∩E
c
) =
¸
µ
∗
(B
i
∩
E) +
¸
µ
∗
(B
i
∩ E
c
) =
¸
µ(B
i
). Hence the measures ¯ µ and µ in Theorem 38 are countably additive on
B.
41. Let µ be a measure on an algebra A, and let E be a µ
∗
measurable set. If B ∈ B, then B
is of the form (A ∩ E) ∪ (A
t
∩ E
c
) where A, A
t
∈ A. Thus ¯ µ(B) = µ
∗
(A ∩ E) + µ
∗
(A
t
∩ E
c
) and
µ
∗
(B) = µ
∗
(A ∩ E) + µ
∗
(A
t
∩ E
c
). If µ
∗
(A
t
∩ E
c
) = ∞, then µ
∗
(A
t
∩ E
c
) = ∞ and ¯ µ(B) = µ
∗
(B). If
µ
∗
(A
t
∩ E
c
) < ∞, then since A
t
∩ E
c
is µ
∗
measurable, µ
∗
(A
t
∩ E
c
) =
¯
¯ µ(A
t
∩ E
c
) = µ
∗
(A
t
∩ E
c
) where
¯
¯ µ is the measure induced by µ
∗
. Thus ¯ µ(B) = µ
∗
(B).
42a. Let G and H be two measurable kernels for E so G ⊂ E, H ⊂ E and µ
∗
(E ` G) = 0 = µ
∗
(E ` H).
Then µ
∗
(G ` H) = 0 = µ
∗
(H ` G) since G ` H ⊂ E ` H and H ` G ⊂ E ` G. In particular, G∆H
is measurable and µ(G∆H) = 0. Let G
t
and H
t
be two measurable covers for E so G
t
⊃ E, H
t
⊃ E
and µ
∗
(G
t
` E) = 0 = µ
∗
(H
t
` E). Then µ
∗
(G
t
` H
t
) = 0 = µ
∗
(H
t
` G
t
) since G
t
` H
t
⊂ G
t
` E and
H
t
` G
t
⊂ H
t
` E. In particular, G
t
∆H
t
is measurable and µ(G
t
∆H
t
) = 0.
42b. Suppose E is a set of σﬁnite outer measure. Then E =
¸
E
n
where each E
n
is µ
∗
measurable and
µ
∗
(E
n
) < ∞. Then µ
∗
(E
n
) < ∞ so there exist G
n
∈ A
δσ
such that G
n
⊂ E
n
and ¯ µ(G
n
) = µ
∗
(E
n
) =
µ
∗
(E
n
). Let G =
¸
G
n
. Then G is measurable, G ⊂ E and µ
∗
(E`G) ≤ µ
∗
(E`G) ≤
¸
µ
∗
(E
n
`H
n
) = 0.
Hence µ
∗
(E ` G) = 0 and G is a measurable kernel for E.
Also, there exist H
n
∈ A
σδ
such that E
n
⊂ H
n
and µ
∗
(H
n
) = µ
∗
(E
n
). Let H =
¸
H
n
. Then H is
measurable, E ⊂ H and µ
∗
(H ` E) ≤ µ
∗
(H ` E) ≤
¸
µ
∗
(H
n
` E
n
) = 0. Hence µ
∗
(H ` E) = 0 and H is
a measurable cover for E.
43a. Let P be the nonmeasurable set in Section 3.4. Note that m
∗
(P) ≤ m
∗
(P) ≤ 1. By Q3.15,
for any measurable set E with E ⊂ P, we have m(E) = 0. Hence m
∗
(P) = sup¦m(E) : E ⊂
P, E measurable, m(E) < ∞¦ = 0.
43b. Let E = [0, 1] ` P. Let 'I
n
` be a sequence of open intervals such that E ⊂
¸
I
n
. We may
assume that I
n
⊂ [0, 1] for all n. Then [0, 1] `
¸
I
n
⊂ P so m([0, 1] `
¸
I
n
) = 0. Thus m(
¸
I
n
) = 1.
Hence m
∗
(E) = 1. Suppose A ∩ [0, 1] is a measurable set. Note that m
∗
(A ∩ E) ≤ m(A ∩ [0, 1]) and
m
∗
(E` A) ≤ m([0, 1] ` A). Thus 1 = m
∗
(E) = m
∗
(A∩E) +m
∗
(E` A) ≤ m(A∩[0, 1]) +m([0, 1] ` A) = 1
so m
∗
(A∩ E) = m(A∩ [0, 1]).
*43c.
*44. Let µ be a measure on an algebra A and E a set with µ
∗
(E) < ∞. Let β be a real number with
µ
∗
(E) ≤ β ≤ µ
∗
(E).
*45a.
*45b.
46a. Let A be the algebra of ﬁnite unions of halfopen intervals of R and let µ(∅) = 0 and µ(A) = ∞
for A = ∅. If E = ∅ and 'A
i
` is a sequence of sets in A with E ⊂
¸
A
i
, then A
i
= ∅ for some i and
µ(A
i
) = ∞. Thus
¸
µ(A
i
) = ∞ so µ
∗
(E) = ∞.
46b. If E contains no interval, then the only A ∈ A with m
∗
(A ` E) < ∞ is ∅. Thus µ
∗
(E) = 0. If E
contains an interval I, then µ
∗
(E) ≥ µ(I) −µ
∗
(I ` E) = µ(I) = ∞.
46c. Note that R = Q∪Q
c
, µ
∗
(R) = ∞, µ
∗
(Q) = µ
∗
(Q
c
) = 0. Thus µ
∗
restricted to B is not a measure
and there is no smallest extension of µ to B.
46d. The counting measure is a measure on B and counting measure restricted to A equals µ. Hence
the counting measure on B is an extension of µ to B.
90
46e. Let E be an interval. Then µ
∗
(E) = ∞ but µ
∗
(E ∩ Q) = 0 and µ
∗
(E ∩ Q
c
) = 0. Hence Lemma 37
fails if we replace “sets in A” by “measurable sets”.
47a. Let X = ¦a, b, c¦ and set µ
∗
(X) = 2, µ
∗
(∅) = 0 and µ
∗
(E) = 1 if E is not X or ∅. Setting
µ
∗
(E) = µ
∗
(X) −µ
∗
(E
c
), we have µ
∗
(X) = 2, µ
∗
(∅) = 0 and µ
∗
(E) = 1 if E is not X or ∅.
47b. ∅ and X are the only measurable subsets of X.
47c. µ
∗
(E) = µ
∗
(E) for all subsets E of X but all subsets except ∅ and X are nonmeasurable.
47d. By taking E = ¦a¦ and F = ¦b¦, we have µ
∗
(E) + µ
∗
(F) = µ
∗
(E) + µ
∗
(F) = 2 but µ
∗
(E ∪ F) =
µ
∗
(E ∪ F) = 1. Hence the ﬁrst and third inequalities of Theorem 35 fail.
(*) If µ
∗
is not a regular outer measure (i.e. it does not come from a measure on an algebra), then we
do not get a reasonable theory of inner measure by setting µ
∗
(E) = µ
∗
(X) −µ
∗
(E
c
).
48a. Let X = R
2
and A the algebra consisting of all disjoint unions of vertical intervals of the form
I = ¦'x, y` : a < y ≤ b¦. Let µ(A) be the sum of the lengths of the intervals of which A is composed.
Then µ is a measure on A. Let E = ¦'x, y` : y = 0¦. If E ⊂
¸
A
n
where 'A
n
` is a sequence in A, then
some A
n
must be an uncountable union of vertical intervals so
¸
µ(A
n
) = ∞. Thus µ
∗
(E) = ∞. If
A ∈ A with µ
∗
(A` E) < ∞, then µ
∗
(A` E) = µ(A) so µ
∗
(E) = 0.
48b. Let E
t
⊂ E and let A
n
=
¸
¦x:¹x,0)∈E
¦
¦'x, y` : −1/n < y ≤ 1/n¦. Then E
t
=
¸
A
n
so E
t
is an A
δ
.
*48c.
12.7 Extension by sets of measure zero
49. Let A be a σalgebra on X and Ma collection of subsets of X which is closed under countable unions
and which has the property that each subset of a set in M is in M. Consider B = ¦B : B = A∆M, A ∈
A, M ∈ M¦. Clearly ∅ ∈ B. If B = A∆M ∈ B, then B
c
= (A`M)
c
∩(M`A)
c
= (A
c
∪M) ∩(M
c
∪A) =
(A ∪ M)
c
∪ (A ∩ M) = (A
c
` M) ∪ (M ` A
c
) = A
c
∆M ∈ B. Suppose 'B
i
` is a sequence in B with
B
i
= A
i
∆M
i
. Then
¸
B
i
=
¸
(A
i
`M
i
)∪
¸
(M
i
`A
i
) = (
¸
A
i
`
¸
M
i
)∪(
¸
M
i
`
¸
A
i
) =
¸
A
i
∆
¸
M
i
∈ B.
Hence B is a σalgebra.
*50.
12.8 Carath´eodory outer measure
51. Let (X, ρ) be a metric space and let µ
∗
be an outer measure on X with the property that µ
∗
(A∪B) =
µ
∗
(A)+µ
∗
(B) whenever ρ(A, B) > 0. Let Γ be the set of functions ϕ of the form ϕ(x) = ρ(x, E). Suppose
A and B are separated by some ϕ ∈ Γ. Then there are numbers a and b with a > b, ρ(x, E) > a on A
and ρ(x, E) < b on B. Thus ρ(A, B) > 0 so µ
∗
(A∪B) = µ
∗
(A) +µ
∗
(B) and µ
∗
is a Carath´eodory outer
measure with respect to Γ. Now for a closed set F, we have F = ¦x : ρ(x, F) ≤ 0¦, which is measurable
since ϕ(x) = ρ(x, F) is µ
∗
measurable. Thus every closed set (and hence every Borel set) is measurable
with respect to µ
∗
.
(*) Proof of Proposition 41
12.9 Hausdorﬀ measures
52. Suppose E ⊂
¸
E
n
. If ε > 0 and 'B
i,n
`
i
is a sequence of balls covering E
n
with radii r
i,n
< ε,
then 'B
i,n
`
i,n
is a sequence of balls covering E so λ
(ε)
α
(E) ≤
¸
i,n
r
α
i,n
=
¸
n
¸
i
r
α
i,n
. Thus λ
(ε)
α
(E) ≤
¸
n
λ
(ε)
α
(E
n
). Letting ε → 0, we have m
∗
α
(E) ≤
¸
n
m
∗
α
(E
n
) so m
∗
α
is countably subadditive.
53a. If E is a Borel set and 'B
i
` is a sequence of balls covering E with radii r
i
< ε, then 'B
i
+ y`
is a sequence of balls covering E + y with radii r
i
. Conversely, if 'B
i
` is a sequence of balls covering
E + y with radii r
i
< ε, then 'B
i
− y` is a sequence of balls covering E with radii r
i
. It follows that
λ
(ε)
α
(E +y) = λ
(ε)
α
(E). Letting ε → 0, we have m
α
(E +y) = m
α
(E).
53b. Since m
α
is invariant under translations, it suﬃces to consider rotations about 0. Let T denote
rotation about 0. If 'B
i
` is a sequence of balls covering E with radii r
i
< ε, then 'T(B
i
)` is a sequence
of balls covering T(E) with radii r
i
. It follows that λ
(ε)
α
(E) = λ
(ε)
α
(T(E)) for all ε > 0. Letting ε → 0,
91
we have m
α
(E) = m
α
(T(E)).
*54.
*55a. Let E be a Borel subset of some metric space X. Suppose m
α
(E) is ﬁnite for some α.
55b. Suppose m
α
(E) > 0 for some α. If m
β
(E) < ∞ for some β > α, then by part (a), m
α
(E) = 0.
Contradiction. Hence m
β
(E) = ∞ for all β > α.
55c. Let I = inf¦α : m
α
(E) = ∞¦. If α > I, then m
α
(E) = ∞ by part (a). Thus I is an upper bound
for ¦β : m
β
(E) = 0¦. If U < I, then there exists β such that U < β < I and m
β
(E) = 0 by part (b).
Hence I = sup¦β : m
β
(E) = 0¦.
*55d. Let α = log 2/ log 3. Given ε > 0, choose n such that 3
−n
< ε. Then since the Cantor ternary set
C can be covered by 2
n
intervals of length 3
−n
, we have λ
(ε)
α
(C) ≤ 2
n
(3
−n
)
α
= 1. Thus m
∗
α
(C) ≤ 1.
Conversely, given ε > 0, if 'I
n
` is any sequence of open intervals covering C and with lengths less than
ε, then we may enlarge each interval slightly and use compactness of C to reduce to the case of a ﬁnite
collection of closed intervals. We may further take each I to be the smallest interval containing some pair
of intervals J, J
t
from the construction of C. If J, J
t
are the largest such intervals, then there is an interval
K ⊂ C
c
between them. Now l(I)
s
≥ (l(J)+l(K)+l(J
t
))
s
≥ (
3
2
(l(J)+l(J
t
)))
s
= 2(
1
2
(l(J))
s
+
1
2
(l(J
t
))
s
) ≥
(l(J))
s
+ (l(J
t
))
s
. Proceed in this way until, after a ﬁnite number of steps, we reach a covering of C by
equal intervals of length 3
−j
. This must include all intervals of length 3
−j
in the construction of C. It
follows that
¸
l(I
n
) ≥ 1. Thus λ
(ε)
α
(C) ≥ 1 for all ε > 0 and m
∗
α
(C) ≥ 1.
Since m
∗
α
(C) = 1, by parts (a) and (b), we have m
β
(C) = 0 for 0 < β < α and m
β
(C) = ∞ for β > α.
Hence the Hausdorﬀ dimension of C is α = log 2/ log 3.
13 Measure and Topology
13.1 Baire sets and Borel sets
1.
[Incomplete: Q5.20c, Q5.22c, d, Q7.42b, d, Q7.46a, Q8.17b, Q8.29, Q8.40b, c, Q9.38, Q9.49, Q9.50,
Q10.47, Q10.48b, Q10.49g, Q11.5d, Q11.6c, Q11.8d, Q11.9d, e, Q11.37b, c, d, e, Q11.46b, Q11.47a,
b, Q11.48, Q12.9b, Q12.11b, Q12.18, Q12.27, Q12.33, Q12.34, Q12.38, Q12.43c, Q12.44, Q12.45a, b,
Q12.48c, Q12.50, Q12.54, Q12.55a]
92
5 Diﬀerentiation and Integration 5.1 Diﬀerentiation of monotone functions . 5.2 Functions of bounded variation . . . . 5.3 Diﬀerentiation of an integral . . . . . . 5.4 Absolute continuity . . . . . . . . . . . 5.5 Convex functions . . . . . . . . . . . . 6 The 6.1 6.2 6.3 6.4 6.5
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
22 22 23 24 24 26 27 27 27 28 29 30 30 30 31 31 32 33 35 35 36 39 40 41 41 43 44 47 48 50 50 51 51 52 53 53 56 56 57 57 58 60
Classical Banach Spaces The Lp spaces . . . . . . . . . . . . . . . . . The Minkowski and H¨lder inequalities . . . o Convergence and completeness . . . . . . . Approximation in Lp . . . . . . . . . . . . . Bounded linear functionals on the Lp spaces
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
7 Metric Spaces 7.1 Introduction . . . . . . . . . . . . . . . . . . 7.2 Open and closed sets . . . . . . . . . . . . . 7.3 Continuous functions and homeomorphisms 7.4 Convergence and completeness . . . . . . . 7.5 Uniform continuity and uniformity . . . . . 7.6 Subspaces . . . . . . . . . . . . . . . . . . . 7.7 Compact metric spaces . . . . . . . . . . . . 7.8 Baire category . . . . . . . . . . . . . . . . 7.9 Absolute Gδ ’s . . . . . . . . . . . . . . . . . 7.10 The AscoliArzel´ Theorem . . . . . . . . . a
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
8 Topological Spaces 8.1 Fundamental notions . . . . . . . . . . . . . . . . . . . . . . 8.2 Bases and countability . . . . . . . . . . . . . . . . . . . . . 8.3 The separation axioms and continuous realvalued functions 8.4 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . 8.5 Products and direct unions of topological spaces . . . . . . 8.6 Topological and uniform properties . . . . . . . . . . . . . . 8.7 Nets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
9 Compact and Locally Compact Spaces 9.1 Compact spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 Countable compactness and the BolzanoWeierstrass property 9.3 Products of compact spaces . . . . . . . . . . . . . . . . . . . 9.4 Locally compact spaces . . . . . . . . . . . . . . . . . . . . . 9.5 σcompact spaces . . . . . . . . . . . . . . . . . . . . . . . . . 9.6 Paracompact spaces . . . . . . . . . . . . . . . . . . . . . . . 9.7 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ˘ 9.8 The StoneCech compactiﬁcation . . . . . . . . . . . . . . . . 9.9 The StoneWeierstrass Theorem . . . . . . . . . . . . . . . . . 10 Banach Spaces
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
ii
10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8
Introduction . . . . . . . . . . . . . . . . Linear operators . . . . . . . . . . . . . Linear functionals and the HahnBanach The Closed Graph Theorem . . . . . . . Topological vector spaces . . . . . . . . Weak topologies . . . . . . . . . . . . . Convexity . . . . . . . . . . . . . . . . . Hilbert space . . . . . . . . . . . . . . .
. . . . . . . . . . . . Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
60 61 62 63 64 66 69 71 73 73 76 77 79 79 80 82 83 83 84 85 86 89 89 91 91 91
11 Measure and Integration 11.1 Measure spaces . . . . . . . . . 11.2 Measurable functions . . . . . . 11.3 Integration . . . . . . . . . . . 11.4 General convergence theorems . 11.5 Signed measures . . . . . . . . 11.6 The RadonNikodym Theorem 11.7 The Lp spaces . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
12 Measure and Outer Measure 12.1 Outer measure and measurability . 12.2 The extension theorem . . . . . . . 12.3 The LebesgueStieltjes integral . . 12.4 Product measures . . . . . . . . . . 12.5 Integral operators . . . . . . . . . . 12.6 Inner measure . . . . . . . . . . . . 12.7 Extension by sets of measure zero . 12.8 Carath´odory outer measure . . . . e 12.9 Hausdorﬀ measures . . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
13 Measure and Topology 92 13.1 Baire sets and Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
iii
z ↔ x. Thus A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C). 10. A∆B = (A \ B) ∪ (B \ A) = (B \ A) ∪ (A \ B) = B∆A. A∆B = ∅ ⇔ (A \ B) ∪ (B \ A) = ∅ ⇔ A \ B = ∅ and B \ A = ∅ ⇔ A ⊂ B andB ⊂ A ⇔ A = B. . By letting xn = xn by letting xi = xi for 1 ≤ i ≤ n and letting xn+1 = fn (x1 . (n+1) (n) (n+1) (n+1) (n+1) (n) (n) (n) 1. Suppose P (1) is true and P (n) ⇒ P (n + 1) for all n. 2. x1 = x2 . X × (Y × Z) = { x.3 Unions. Suppose A ⊂ B. If x ∈ B. y. (⇐) Suppose there exists g : Y → X such that g ◦ f = idX . xi ). Thus A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C). (⇒) Suppose f is onetoone. . 7. Fix x0 ∈ X. . then either m is the smallest element of S or there exists m ∈ S with m < m = n + 1. 1 . Let P (n) be the proposition that for each n there is a unique ﬁnite sequence x1 . Deﬁne g : Y → X such that g(y) = xy if y ∈ f [X] and g(y) = x0 if y ∈ Y \ f [X]. y. Contradiction. Given a nonempty subset S of natural numbers. Contradiction. . (X × Y ) × Z = { x. If m ≤ n. g(y) ∈ X and f (g(y)) = y. Thus f is onto. Suppose that {n ∈ N : P (n) is false} = ∅. In particular. z ↔ x. . 8. Given P (n). z }. . P (1) is true since 1 will then be the smallest element of S. x ∈ A ∩ (B ∪ C) ⇔ x ∈ A and x ∈ B or C ⇔ x ∈ A and B or x ∈ A and C ⇔ x ∈ (A ∩ B) ∪ (A ∩ C). Thus B c ⊂ Ac . . let P (n) be the proposition that if there exists m ∈ S with m ≤ n. we get a unique sequence xi from X such that x1 = a and xi+1 = fi (x1 .1 1. in which case the induction hypothesis again gives a smallest element. . then g(f (x1 )) = g(f (x2 )). A∆(B∆C) = [A \ ((B \ C) ∪ (C \ B))] ∪ [((B \ C) ∪ (C \ B)) \ A] = [A ∩ ((B \ C) ∪ (C \ B))c ] ∪ [((B \ C) ∪ (C \ B)) ∩ Ac ] = [A ∩ ((B c ∪ C) ∩ (C c ∪ B))] ∪ [((B ∩ C c ) ∪ (C ∩ B c )) ∩ Ac ] = [A ∩ (B c ∪ C) ∩ (B ∪ C c )] ∪ (Ac ∩ B ∩ C c ) ∪ (Ac ∩ B c ∩ C) = (A ∩ B c ∩ C c ) ∪ (A ∩ B ∩ C) ∪ (Ac ∩ B ∩ C c ) ∪ (Ac ∩ B c ∩ C). But P (1) ⇒ P (2) ⇒ · · · ⇒ P (m). If m = n + 1. .1 Set Theory Introduction 1. there exists a unique xy ∈ X such that f (xy ) = y. Then g is a welldeﬁned function and f ◦ g = idY . 12b. For each y ∈ Y . x ∈ A ∪ (B ∩ C) ⇔ x ∈ A or x ∈ B and C ⇔ x ∈ A or B and x ∈ A or C ⇔ x ∈ (A ∪ B) ∩ (A ∪ C). . Then g is a welldeﬁned function and g ◦ f = idX . xi ).e. Suppose that P (n) is true and that there exists m ∈ S with m ≤ n + 1. If f (x1 ) = f (x2 ). 1. y. 12a. y . if B c ⊂ Ac . we see that P (n + 1) is true ). . . Then it has a smallest element m. A ⊂ B ⇒ A ⊂ A ∩ B ⇒ A ∩ B = A ⇒ A ∪ B = (A \ B) ∪ (A ∩ B) ∪ (B \ A) = (A ∩ B) ∪ (B \ A) = B ⇒ A ⊂ A ∪ B = B. Given y ∈ Y . intersections and complements 9. ∅ ⊂ {greeneyed lions}. y . Thus f is onetoone. then A = (Ac )c ⊂ / / (B c )c = B. . Conversely. there exists xy ∈ X such that f (xy ) = y. (⇐) Suppose there exists g : Y → X such that f ◦ g = idY . then S has a smallest element by the induction hypothesis. . . 4. 11. i. m > 1 and P (m) is false.2 Functions 6. (A∆B)∆C = [((A \ B) ∪ (B \ A)) \ C] ∪ [C \ ((A \ B) ∪ (B \ A))] = [((A \ B) ∪ (B \ A)) ∩ C c ] ∪ [C ∩ ((A \ B) ∪ (B \ A))c ] = [((A ∩ B c ) ∪ (B ∩ Ac )) ∩ C c ] ∪ [C ∩ ((Ac ∪ B) ∩ (B c ∪ A))] = (A ∩ B c ∩ C c ) ∪ (Ac ∩ B ∩ C c ) ∪ (Ac ∩ B c ∩ C) ∪ (A ∩ B ∩ C). then there exists x such that x = x. z }. Deﬁne g : Y → X such that g(y) = xy for all y ∈ Y . then x ∈ A. If {x : x = x} = ∅. x. xn for each n. For each y ∈ f [X]. 3. (⇒) Suppose f is onto. Clearly P (1) is true. . sn with (n) (n) (n) (n) x1 = a and xi+1 = fi (x1 . z . 5. then S has a smallest element. Hence A∆(B∆C) = (A∆B)∆C.
f (x) ∈ B. 17b. Since f is onto. then each Ei is in some σalgebra generated by some countable subset Ci of C. (A∆B) ∩ E = ((A \ B) ∪ (B \ A)) ∩ E = ((A \ B) ∩ E) ∪ ((B \ A) ∩ E) = [(A ∩ E) \ (B ∩ E)] ∪ [(B ∩ E) \ (A ∩ E)] = (A ∩ E)∆(B ∩ E). then y = f (x) for some x ∈ f −1 [B]. 18b. 2} → {1. Consider the collection A = {Ay : y ∈ Y }. there is a function F on A such that F (Ay ) ∈ Ay for all y ∈ Y .e. Let B = {1. 3} and A2 = {2. Conversely. Then f −1 [f [A]] = f −1 [{1}] = {1. Let A = {1}. ( A∈A A) ∩ ( B∈B B) = B∈B (( A∈A A) ∩ B) = B∈B ( A∈A (A ∩ B)) = A∈A B∈B (A ∩ B). then y ∈ f [Aλ0 ] for some λ0 so y ∈ f [ Aλ ]. If E1 . By the axiom of choice. 2} with f (1) = f (2) = 1. then F is in some σalgebra generated by some countable set and so is F c . then x ∈ Aλ0 for some λ0 and f (x) ∈ f [Aλ0 ] ⊂ f [Aλ ]. If F ∈ A . If y ∈ f [f −1 [B]]. 3} with f (1) = f (2) = 1 and f (3) = 3. x ∈ ( A∈C A)c ⇔ x ∈ A for any A ∈ C ⇔ x ∈ Ac for all A ∈ C ⇔ x ∈ A∈C Ac . 2 . 12d. Then f [A1 ∩ A2 ] = f [{3}] = {3} but f [A1 ] ∩ f [A2 ] = {1. A∆B = X ⇔ (A \ B) ∪ (B \ A) = X ⇔ A ∩ B = ∅ and A ∪ B = X ⇔ A = B c . Thus C ⊂ A . then x ∈ f −1 [B] so f (x) ∈ B c . 2}. since C ⊂ A. x ∈ (f −1 [B])c . If y ∈ B. if y ∈ f [Aλ ]. if Ei is a sequence in A . Furthermore. If x ∈ Aλ . If E ∈ C. if x ∈ f −1 [Bλ ]. Thus f [ Aλ ] ⊂ f [Aλ ]. then x ∈ f −1 [Bλ0 ] for some λ0 so f (x) ∈ Bλ0 ⊂ Bλ and x ∈ f −1 [ Bλ ]. 16a. being a σalgebra. 3}. 2} A. 1.4 Algebras of sets 19a. A∆∅ = (A \ ∅) ∪ (∅ \ A) = A ∪ ∅ = A. 3} → {1. Since x ∈ f −1 [B]. If x ∈ f −1 [ Bλ ]. Conversely. Conversely. then f (x) ∈ B so x ∈ f −1 [B]. Thus f −1 [f [A]] ⊃ A. then x ∈ f −1 [Bλ ] for all λ and f (x) ∈ Bλ so x ∈ f −1 [ Bλ ]. then f (x) ∈ Bλ0 for some λ0 so x ∈ f −1 [Bλ0 ] ⊂ f −1 [Bλ ]. 18c. Let A be the union of all σalgebras generated by countable subsets of C. Consider f : {1. Thus B ⊂ f [f −1 [B]]. Then f [f −1 [B]] = f [B] = {1} B. B1 . B1 ⊂ B2 .e. Thus f −1 [ Bλ ] ⊂ f −1 [Bλ ]. together with the inequality in Q18a. Ay = ∅ for all y. then there exists x ∈ X such that f (x) = y. y ∈ B. Conversely. Thus B ∩ ( A∈C A) = A∈C (B ∩ A). by deﬁnition. i. then E1 is in some σalgebra generated by some countable subset C1 of C and E2 is in some σalgebra generated by some countable subset C2 of C. Hence A is a σalgebra containing C and it contains A. 15. P(X) is a σalgebra containing C. 1. Thus F c ∈ A . Thus B1 ⊃ B2 . x ∈ f −1 [B c ]. then E is in the σalgebra generated by {E}. gives equality. Thus f [Aλ ] ⊂ f [ Aλ ]. then f (x) ∈ f [A] so x ∈ f −1 [f [A]]. Thus f −1 [B c ] ⊂ (f −1 [B])c . 19b. / / Conversely. Thus (f −1 [B])c ⊂ f −1 [B c ]. / x ∈ ( A∈C A)c ⇔ x ∈ A∈C A ⇔ x ∈ Ac for some A ∈ C ⇔ x ∈ A∈C Ac . Hence B1 = B2 .5 The axiom of choice and inﬁnite direct products 21. 12e. x ∈ f −1 [B] and y ∈ f [f −1 [B]]. then x ∈ Aλ for all λ and f (x) ∈ f [Aλ ] for all λ. 20. 2. Furthermore. If x ∈ f −1 [B c ]. 17a. A∆X = (A \ X) ∪ (X \ A) = ∅ ∪ Ac = Ac . 17c. Thus f −1 [ Bλ ] ⊂ f −1 [Bλ ]. Thus f [f −1 [B]] ⊂ B. / 14. Let F be the family of all σalgebras containing C and let A = {B : B ∈ F }. 13. Then A is a σalgebra containing C. i. then B ⊃ A. Then E1 ∪ E2 is in the σalgebra generated by the countable subset C1 ∪ C2 so E1 ∪ E2 ∈ A . For each y ∈ Y .12c.e. If x ∈ f −1 [ Bλ ]. x ∈ B ∩ ( A∈C A) ⇔ x ∈ B and x ∈ A for some A ∈ C ⇔ x ∈ B ∩ A for some A ∈ C ⇔ x ∈ A∈C (B ∩ A). if x ∈ (f −1 [B])c . Thus f (x) ∈ f [Aλ ] and f [ Aλ ] ⊂ f [Aλ ]. If x ∈ A. 3}. let Ay = f −1 [{y}]. i. / 18a. then f (x) ∈ Bλ for all λ and x ∈ f −1 [Bλ ] for all λ so x ∈ f −1 [Bλ ]. If x ∈ Aλ . Then Ei is in the σalgebra generated by the countable subset Ci . if x ∈ f −1 [Bλ ]. if B is a σalgebra containing C. 16c. In particular. 16b. E2 ∈ A . Let A1 = {1. Let B1 be the σalgebra generated by C and let B2 be the σalgebra generated by A. Consider f : {1. This. is also an algebra so B1 ⊃ A.
Then f (v) = avn ∞ with avn = 0 or 1 for each n. Let X be a wellordered set and let A ⊂ X. Any strict linear ordering on X is also a strict linear ordering on A and every nonempty subset of A is also a nonempty subset of X. Pick a ∈ X. . 2 → −p/q. 1. / 26. xk }). (x1 + x2 ) + x3 = x1 + (x2 + x3 ) so + is associative on Q. . Also deﬁne x ≤ y if x y or x = y. Thus E0 is the identity of Q. . p. xn } be a ﬁnite set and let A ⊂ E. Let E = {x1 . Choose xi ∈ Ei for i = 1. 3) and x < y. Let E = {x : x ∈ f (x)} ⊂ X. . Conversely. Contradiction. . By the axiom of choice. . . . Hence E is not in the range of f . 1. Hence the induced operation + makes the quotient space Q into a group. . 1. . xi }) ∈ X \ {x1 . In particular. y ∈ [2. Then (E1 + E2 ) + E3 = Ex1 +x2 + E3 = E(x1 +x2 )+x3 and E1 + (E2 + E2 ) = E1 + Ex2 +x3 = Ex1 +(x2 +x3 ) . 25. If A = ∅. Consider the mapping p. . Then x + y ≡ x + y and x + y ≡ x + y so x + y ≡ x + y .i. y ∈ X. Then is a partial ordering on the set. or x. Let F ∈ Q and choose x ∈ F and let −x be the inverse of x in X. choose x ∈ A. 30. . Then bn ∈ E but bn = avn for any v. let fn : X n → X be deﬁned by fn (x1 . Suppose ≡ is compatible with +. 1. Thus its range. Furthermore. . . 1. yn and is therefore ﬁnite. . which is countable by Propositions 4 and 5. 1] and x < y. Let X be an inﬁnite set. Similarly for E0 + F . . Let f be a function from N to E. Deﬁne g : Y → X. . 28. y ∈ (0. For any two elements x. Since X is a group under +. Then x ≡ x implies x + y ≡ x + y since y ≡ y. Let < be a partial order on X with the property that every nonempty subset has a least element. . . y} has a least element. Thus E cannot be the range of any function from N and E is uncountable. xi ). Now if / x0 ∈ E. If E is in the range of f . 1 → p/q. Then < is a strict partial order on X and ≤ is a reﬂexive partial order on X. G ∈ Q = X/ ≡. there exists a unique sequence xi from X such that x1 = a. Deﬁne a new sequence y1 . xi+1 = F (X \ {x1 .e. . Similarly when x0 ∈ E. xi+1 = fi (x1 . Choose x1 . q. y → F (Ay ). 24. q. 2. . Consider the set (0.6 Countable sets 22. 3. deﬁne x < y if x y and x = y. there is a choice function F : P(X) \ {∅} → X. choose x ∈ F . If the least element is y. For each n ∈ N. then E = f (x0 ) for some x0 ∈ X. Let E1 . Then f ◦ g = idY . then y < x. E3 ∈ Q = X/ ≡. x2 ∈ F and y1 . 31b. Now suppose x ≡ x and y ≡ y . .9 Well ordering and the countable ordinals 31a. 3 → 0. . Then F + E−x = Ex+(−x) = E0 . Similarly for E−x + F . suppose x ≡ x implies x + y ≡ x + y. 1] ∪ [2. the set of rational numbers. . then x0 ∈ f (x0 ) = E. .8 Partial orderings and the maximal principle 29. y2 ∈ G. Then / A is the range of y1 . then x < y.7 Relations and equivalences 27. . 23. . 1. F (Ay ) ∈ f −1 [{y}] so f (F (Ay )) = y. Its domain is a subset of the set of ﬁnite sequences from N. Thus any nonempty subset of A has a smallest element. . xn ) = F (X \ {x1 . the set {x. Thus < is a linear ordering on X with the property that every 3 . If the least element is x. Thus E−x is the inverse of F in Q. 3) with the ordering given by x y if and only if either x. By the generalised principle of recursive deﬁnition. . If A = ∅. E2 . Let n=1 bv = 1 − avv for each v. Let 0 be the identity of X. Thus y1 + y2 ∈ Ex1 +x2 and Ex1 +x2 = Ey1 +y2 . Let F. Uniqueness follows from the deﬁnitions. yn by setting yi = xi if xi ∈ A and yi = x if xi ∈ A. is countable. . 2 is the unique minimal element and there is no smallest element. . xi } so xi = xj if i = j and the range of the sequence xi is a countably inﬁnite subset of X. Then x1 ≡ x2 and y1 ≡ y2 so x1 + x2 ≡ y1 + y2 . . Then F + E0 = Ex+0 = Ex = F . Given a partial order on X. x y ⇔ x < y ⇔ x ≤ y for x = y. For any F ∈ Q. then A is ﬁnite by deﬁnition.
34b. Thus S ∪ {x0 } is a segment of X. Contradiction. If y0 < yλ for some λ. 33. Clearly. Conversely. Then y ∈ f [{w : w < z}] so / y = f (w) for some w < z. 34c. If f (x1 ) < f (x2 ). f −1 (f (x1 )) < f −1 (f (x2 )). Thus there exists y ∈ Y \ E such that x < y for all x ∈ E. Let f be a successorpreserving map from X into Y . Also. Contradiction. Let Y = {x : x < Ω} and let E be a countable subset of Y . f (s) = fλ (s) is the ﬁrst element of Y not in fλ [{z : z < s}] = f [{z : z < s}]. Hence / Sλ = {x ∈ X : x < y0 }. y < y0 . This is a nonempty subset of Y so it has a least element. Hence f is an order preserving bijection from X onto Y . Let S = {z : z < x} be a segment of X and let z ∈ S. then S ∪ {x0 } = X. i. then y0 ∈ Sλ . i. Now gx0 is the ﬁrst element of Y not in g[{z : z < x0 }] = f [{z : z < x0 }] and so is f (x0 ). f (y) is the ﬁrst element of Y not in f [{z : z < y}] so f (y) = f (x). Then {x ∈ X : f (x) = g(x)} is a nonempty subset of X and has a least element x0 . then by part (b). f [X] = {z : z < y0 } for some y0 ∈ Y . then f −1 is a successorpreserving map on Y and f −1 [Y ] = S is a segment of X. Sλ ⊂ {x ∈ X : x < y0 }. Suppose f is successorpreserving. if y ∈ f [X]. Let X be the wellordered set in Proposition 8 and let Y be an uncountable set wellordered by such that there is a last element ΩY in Y and if y ∈ Y and y = ΩY . Thus f (x0 ) = g(x0 ). Y is uncountable so Y \ E is nonempty. then f (y) ∈ f [{z : z < x}] and y < x. 34f. Thus w ∈ S and y = f S (w) ∈ f S [{w : w < z}]. then x1 < x2 since f is order preserving. Hence there is at most one successorpreserving map from X into Y . Then there is a least y0 ∈ Y \ f [S]. Suppose y < f S (z). If f [X] = Y . X \ Sλ is a nonempty subset of the wellordered set X so it has a least element y0 . Otherwise. E has an upper bound in Y . Then each Sλ is of the form {x ∈ X : x < yλ } for some yλ ∈ X. Clearly. Hence f [X] = {y : y < y0 }. In the second case. f is a welldeﬁned map because if s belongs to two segments Sλ and Sµ . Contradiction. 34d. S is a segment so either S = X or S = {x : x < x0 } for some x0 ∈ X. If there is no / x ∈ X such that x > x0 . If for every y ∈ Y \ E there exists xy ∈ E such that y < xy . Thus f is a successorpreserving map from S into Y .e. On the other hand. we may assume there is a successorpreserving map f from X onto a segment of Y . {x ∈ X : x > x0 } has a least element x and S ∪ {x0 } = {x : x < x }. Otherwise. we may assume Sλ ⊂ Sµ so that fµ Sλ = fλ by parts (d) and (a). Thus y0 ≥ yλ for all λ. Thus f −1 is order preserving. Consider the set of upper bounds of E in Y . Contradiction. Conversely. Consider S ∪ {x0 }. Furthermore. then we are done. Contradiction. Now if S = X. f −1 is deﬁned on f [X]. Thus f (x) < f (y). Thus f S (z) ∈ f S [{w : w < z}]. Contradiction. then {z ∈ Y : z < y} is countable. Deﬁne f (x0 ) = y0 . 34a. if f [S] = Y . X is countable. then f [X] is countable and since f is onetoone. If y0 = ΩY . then f (ΩX ) < ΩY and f (ΩX ) is the largest element in f [X]. Let {Sλ : λ ∈ Λ} be a collection of segments.nonempty subset has a least element and consequently a well ordering. {y : y < y0 } ⊂ f [X]. Hence f [S] = Y . 32. Hence f S (z) is the ﬁrst element of Y not in f S [{w : w < z}]. Suppose f and g are distinct successorpreserving maps from X into Y . if f (y) < f (x). Suppose f [X] = Y . Suppose Sλ = X. i. f [X] = {z : z ≤ f (ΩX )}. f S (z) = f (z) is the ﬁrst element of Y not in f [{w : w < z}]. Consider the collection of all segments of X on which there is a successorpreserving map into Y . Contradiction. Thus / f [X] ⊂ {y : y < y0 }.e. suppose f [S] = Y . Given s ∈ S. then y = f (x) is the ﬁrst element not in f [{z : z < x}]. if x < y0 . 4 .e. By part (e). y ∈ X with x < y. which is then a least upper bound of E. Let x. then y → xy deﬁnes a mapping from Y \ E into the countable set E so Y \ E is countable. Since y0 ∈ f [X] ⊃ f [{z : z < x}]. then x ∈ X \ Sλ so x ∈ Sλ . Let S be the union of all such segments Sλ and let fλ be the corresponding map on Sλ . Then {y : y ∈ f [X]} is a / nonempty subset of Y and has a least element y0 . which is countable. Then f (x0 ) is the ﬁrst element of Y not in f [{x : x < x0 }]. f [X] = f [S] is a segment of Y . Since f is onetoone. Thus f is a successorpreserving map on the segment S ∪ {x0 } so S ∪ {x0 } ⊂ S. Hence f is a onetoone order preserving map. deﬁne f (s) = fλ (s) if s ∈ Sλ . If y0 = ΩY . 34e. By Q33.
4d. then x = −x ≤ y since −y ≤ x. y < 0 and x + y ≥ 0. Hence u0 = l0 so u0 ∈ U and it is the least element in U . there exists n1 ≥ 1 5 . If x ≥ y. there exist N1 and N2 such that xn − l1  < ε for n ≥ N1 and xn − l2  < ε for n ≥ N2 . then x + y < 0 so x + y = −x − y = x + y. 4a. then x + y ≥ 0 so x + y = x + y = x + y. If x. If u0 < l0 . Similarly for y ≤ x. Thus l is a lower bound for U and is greater than u0 . If x < 0. Then x + z ≤ y + z so x ∨ y + z = y + z = (x + z) ∨ (y + z). Then −1 ∈ P . 2. Suppose x ≤ y. Conversely. Now take x ∈ P . Then given ε > 0. Then zx ≤ zy if z ≥ 0 so z(x ∨ y) = zy = (zx) ∨ (zy). Thus −b is a greatest lower bound for S. Similarly when x < 0 and y ≥ 0. Similarly for y ≤ x. 4c. If E = ∅. 4e. If E = ∅.4 Sequences of real numbers 7. there exists N such that xnj − l < ε for j ≥ N . 5a. 5c. choose j ≥ N such that nj ≥ N . If x. Suppose x ≤ y. Similarly. 5b. If x. Let S be a nonempty set of real numbers with a lower bound. Similarly for y ≥ x. 3. then it is the greatest element in L. Then xnj − l < ε and l is a cluster point of xn .1 The Real Number System Axioms for the real numbers 1. then −a is an upper bound for −S and b ≤ −a so −b ≥ a. If x ≥ 0. it has a least upper bound b. y < 0 and x + y < 0. Similarly for y ≤ x.3 The extended real numbers 6. then x + y = −x − y ≤ x − y = x + y. Given N . Then l1 − l2  < 2ε for n ≥ max(N1 . If x ≥ 0 and y < 0. 1 5d. / Contradiction. Similarly when x < 0 and y ≥ 0. Then −x = (−1)x ∈ P so 0 = x + (−x) ∈ P . Suppose x ≤ y. Then x ∧ y = x and x ∨ y = y so x ∧ y + x ∨ y = x + y. then xy ≤ 0 so xy = −xy = x(−y) = xy. Given ε > 0. suppose l ∈ R is a cluster point of xn . then x + y = x + y ≤ x − y = x + y. 2. Suppose −y ≤ x ≤ y. then there exists l ∈ L with u0 < l. say x ∈ E. 8. l0 ∈ U and u0 ≤ l0 . N2 ). Since ε is arbitrary. If x. x∧y y∧z (x ∧ y) ∧ z x ∧ (y ∧ z) x≤y≤z x y x∧z =x x∧y =x x≤z≤y x z x∧z =x x∧z =x y≤x≤z y y y∧z =y x∧y =y y≤z≤x y y y∧z =y x∧y =y z≤x≤y x z x∧z =z x∧z =z z≤y≤x y z y∧z =z x∧z =z 4b. If x ≥ 0. 2 5e. Similarly for x < 0. Similarly for y ≤ x. 2. y ≥ 0. Let L and U be nonempty subsets of R with R = L ∪ U and such that for each l ∈ L and u ∈ U we have l < u. If x ≥ 0. y < 0. Then −y ≤ −x so (−x) ∧ (−y) = −y = −(x ∨ y). Suppose there is a subsequence xnj that converges to l ∈ R. then inf E ≤ x ≤ sup E. Given ε > 0. l1 − l2  = 0 and l1 = l2 . y ≥ 0. Otherwise. If l0 ∈ L. then xy ≥ 0 so xy = xy = xy. y < 0. −b is a lower bound for S and if a is another lower bound for S. Contradiction. L is bounded above so it has a least upper bound l0 .2 2. Suppose 1 ∈ P . If x ≥ 0. Suppose l1 and l2 are (ﬁnite) limits of a sequence xn . Then the set −S = {−s : s ∈ S} has an upper bound and by Axiom C. Suppose x ≤ y. then x = x ≤ y. then xy > 0 so xy = xy = (−x)(−y) = xy. U is bounded below so it has a greatest lower bound u0 .2 The natural and rational numbers as subsets of R No problems 2. The cases where ∞ or −∞ is a limit are clear. then sup E = −∞ and inf E = ∞ so sup E < inf E. then x − y ≥ 0 so x ∨ y = x = 2 (x + y + x − y) = 1 (x + y + x − y). then x ≥ −x so x = x = x ∨ (−x).
Given ε. Furthermore by (i). then since it is a cluster point. Then there exists N such that xn  − xm  ≤ xn − xm  < 1 for m. Then lim xnk = ∞ so lim xn = ∞. Let xn be a Cauchy sequence. xk > ∆ for k ≥ N . . Choose nk+1 ≥ 1 + max(xn1 . 11c. By (ii). By deﬁnition. The converse holds by part (a). . The statement does not hold when the word “extended” is removed. inf k≥n xk ≤ supk≥n xk ≤ supk≥m xk . . nj and xk1 . Suppose l = lim xn . Conversely. Thus inf k≥m xk ≤ supk≥n xk whenever m = n. 15. kj ). there exists n ≥ N with xn < ∆. xnk ) such that xnk+1 − l < ε. Then lim xn ≤ sup xn < ∞. Hence l = inf n supk≥n xk = lim xn . This gives rise to a subsequence xnk with xnk − x ≥ ε for all k. Thus l is the only cluster point. Thus the sequence xn  is bounded above by max(x1 . 13. then lim xn = lim xn = x. Suppose xn is a Cauchy sequence with a subsequence xnk that converges to l. If x = ∞. If l is ∞ or −∞. 11b. xnk have been chosen. Suppose n1 . n ≥ N . If l = l and l is a cluster point. suppose every subsequence of xn has in turn a subsequence that converges to x. Let l = lim xn ∈ R. 3. . Thus xn is a Cauchy sequence. supk≥n xk ≤ l . . Furthermore. Thus lim xn = ∞. Similarly for lim xn . there exists n such that supk≥n xk < l + ε so xk < l + ε for all k ≥ n. 1. xn  − xN  < 1 for n ≥ N . there exists n ≥ N with xn − x ≥ ε. Then xn − l ≤ xn − xnk  + xnk − l < ε for n ≥ N . If l ∈ R. This gives rise to a subsequence xnk with xnk < ∆ for all k. Contradiction. suppose that given ∆ and n. there exists k ≥ n such that xk > ∆. Given ε > 0. . supk≥n xk > ∆. . 14. Suppose lim xn = ∞. The sequence 1. Then the subsequence xnj converges to l. Also. . 9a. we may assume that l ∈ R. Then the subsequence xkj converges to l and l is a cluster point of xn . . 11a. 10. If xn is a Cauchy sequence. . xn converges to l. Thus there exists k ≥ n such that xk > ∆. suppose there is exactly one extended real number x that is a cluster point of xn . This subsequence will not have a further subsequence that converges to x. . then the sequence has exactly one extended real number that is a cluster point 6 . 9b. Thus xn converges to l. The cases where l is ∞ or −∞ are similarly dealt with. given ε and n. Choose nj+1 > max(k1 . supk≥n xk > l − ε so there exists k ≥ n such that xk > l − ε. Given ε > 0. Then given ∆ and n. there exists N such that supk≥N xk < l + 1. Conversely. there exists N such that xn − xm  < ε/2 for m. Also. there exists n1 such that xk < l + ε for k ≥ n1 . 12.such that xn1 − l < ε. then there exists k ≥ N such that xk ≥ l + 1. . . 1. then every subsequence of xn also converges to x. Given ε > 0.e. . . If x ∈ R. By part (c). supk≥n xk ≥ l − ε. it follows from the deﬁnitions that l is a cluster point of xn . If l = ∞. If x = ∞ and lim xn = ∞. for any ε and n. suppose conditions (i) and (ii) hold. Similarly when x = −∞. xkj have been chosen. 4. Now if m. If xn converges to l. there exists N such that supk≥N xk < (l + l )/2. Also. there exists k1 ≥ n1 such that xk1 > l − ε. there is a subsequence converging to it. then it is bounded by part (b) so it has a subsequence that converges to a real number l by Q9b. then there is a subsequence of xn converging to l . 1. xN  + 1) so the sequence xn is bounded. Suppose xn1 . inf k≥m xk ≤ inf k≥n xk ≤ supk≥n xk . xN −1 . Thus xkj+1 − l < ε. . xn − xm  ≤ xn − l + xm − l < ε. i. Similarly when x = −∞. N ) and xn converges to x. . then l is a cluster point. . Let xn be a bounded sequence. then there exists n such that xk < l for all k ≥ n. . . Thus xk − x < ε for k ≥ max(N. 2. . Thus supk≥n xk ≥ l for all n. Conversely. . For all m < n. By deﬁnition again. This subsequence will not have a further subsequence with limit ∞. Let xn1 = x1 and let nk+1 > nk be chosen such that xnk+1 > xnk . has exactly one real number 1 that is a cluster point but it does not converge. If x = lim xn . then lim xn = ∞ so for any ∆ there exists N such that inf k≥N xk > ∆. then there exists ∆ such that for all N . Then xk < l + ε for k ≥ nj+1 . Given ε > 0. . i. If x ∈ R and xn does not converge to x. Thus lim xn is a ﬁnite real number and by part (a). there exists N such that xn − l < ε/2 for n ≥ N . then there exists ε > 0 such that for all N . Hence lim xn = supn inf k≥n xk ≤ inf n supk≥n xk = lim xn . there exists k ≥ N such that xk − l  < (l − l)/2 so xk > (l + l )/2. 11d. Hence l is the largest cluster point of xn . Thus xk1 − l < ε. 1. n ≥ N and xnk − l < ε/2 for nk ≥ N . In particular.e. there exists N such that supk≥N xk < x + ε and there exists N such that inf k≥N xk > x − ε. If l is a cluster point of xn and l > l. Contradiction. n ≥ N . . . Now choose k such that nk ≥ N . Conversely. . Suppose xn converges to l ∈ R. if l > l. there exists kj+1 ≥ nj+1 such that xkj+1 > l − ε. If lim xn = lim xn = l. . Contradiction.
x1 + v=1 (xv+1 − xv ) = x1 + limn v=1 (xv+1 − xv ) = limn [x1 + v=1 (xv+1 − xv )] = limn xn+1 = ∞ limn xn . let sn = k=1 ak /pk . Furthermore.e. This sequence is unique by construction. which is uncountable by Q1. On the other hand. Then each En is a ﬁnite subset of E. Let b = sup S. If the sequence sn is unbounded. Thus lim xn + lim yn ≤ lim (xn + yn ). Thus lim xn = lim xn = l. x + δ) ⊂ X. xk +yk ≤ supk≥n xk +supk≥n yk for k ≥ n. . Then x∈E x ≥ sFk ≥ k for each k. Then b ∈ S and b is a point of closure of X so b ∈ X since X is closed. the sequence tn is a Cauchy sequence so given ε. 23. b) ∪ [b. Given x ∈ R. 21a. . y) ⊂ X for some y > x. . On the other hand. 1}. there exists N such that tn − tm  < ε for n > m ≥ N . Hence X = R. n ∞ 19.5 Open and closed sets of real numbers 24. then the sequence has a subsequence converging to lim xn and another subsequence converging to lim xn . let tn = v=1 xv . R is uncountable. n Conversely. For all n. Then inf n supk≥n (xk +yk ) ≤ inf n supk≥n xk +inf n supk≥n yk . (z. let En = {x ∈ E : x ≥ 1/n}. Similarly. For each x ∈ Qc and for each δ > 0. ∞ n n 20. 2. Contradiction. if an is a sequence of integers with 0 ≤ an < p. Now E = En so E is countable. given F ∈ F. Since v=1 xv  < ∞. . Given a real number x with 0 ≤ x ≤ 1. for each x ∈ Q and for each δ > 0. Suppose X is a nonempty subset of R that is both open and closed.so it converges to l by Q10. However the sequence ∞ bn with bm = 0 for m < n. . Since X is open. Thus sn is a bounded monotone increasing sequence so it converges. which we may regard as unique by ﬁxing a way of representing those numbers of the form q/2n . Thus [x. √ √ / (*) Given x. . the sequence sn is monotone increasing. 1] is uncountable and since R ⊃ [0.24. When x = q/pn with q ∈ {1. . . 1]. b + δ ) ⊂ X. Since each xv ≥ 0. Then sn − sm  = xm+1 + · · · + xn  ≤ xm+1  + · · · + xn  = tn − tm  < ε for n > m ≥ N . Let an+1 be the largest integer such that 0 ≤ an+1 < p and an+1 /pn+1 ≤ x − k=1 ak /pk . if En0 is an inﬁnite set for some n0 . . there exists δ > 0 such that (b − δ . lim (xn +yn ) ≤ lim xn +lim yn . Contradiction. b + δ ) = [x. there exists N such that 1/pN < ε. Hence ∞ x∈E x = supF ∈F sF = sup sn = n=1 xn . ∅ and R are both open and closed. ∞ then lim sn = sup sn ∈ R so sup sn = v=1 xv . . the sequence converges to a real number x with 0 ≤ x ≤ 1. bn = q − 1 and bm = p − 1 for m > n also satisﬁes x = n=1 bn . let a1 be the largest integer such that 0 ≤ a1 < p and a1 /p ≤ x. there is an irrational number s with x < s < x + δ so Q is not open. . . The set of rational numbers is neither open nor closed. if l = lim xn and lim xn < lim xn . i. Then x − k=1 an /pn  < 1/pN < ε for n ≥ N . there is a rational number r with x < r < x + δ so Qc is not open and Q is not closed. 7 . {x1 . xn } so sF ≤ sn and supF ∈F sF ≤ sup sn . 21b. Contradiction. p − 1}. Thus x = limn xn if and only if x = x1 + v=1 (xv+1 − xv ). Then S is bounded above. For each n. the sequence an obtained in this way is such that an = q and am = 0 for m = n. 18. n k n This gives rise to a sequence an of integers with 0 ≤ an < p and x − k=1 ak /p < 1/p for all n. Thus supk≥n (xk +yk ) ≤ supk≥n xk +supk≥n yk for all n. sFk ≥ k. we get lim xn yn ≤ lim xn lim yn . taking limits. Suppose x∈E x < ∞. y) ⊂ X for all y > x. form its binary expansion (by taking p = 2 in Q22). 22. then letting Fk be a subset of En0 with kn0 elements for each k ∈ N. . xk yk ≤ supk≥n xk supk≥n yk . By Q22. Take x ∈ X. 1] to the set of inﬁnite sequences from {0. 17. . this gives a bijection from [0. x] ⊂ X for all z < x. Clearly. 16. there exists n such that F ⊂ {x1 . Thus the sequence sn is a Cauchy sequence so it converges and xn has a sum. Suppose [x. Conversely. We may assume 0 ∈ (x. y) ⊂ X} is nonempty. 25. Then [x. Thus ∞ x = n=1 an /pn . . supk≥n xk > 0 and supk≥n xk yk ≤ supk≥n xk supk≥n yk . For each n. Now lim xn ≤ lim (xn + yn ) + lim (−yn ) = lim (xn + yn ) − lim yn . Thus [0. there exists δ > 0 such that (x − δ. . Thus the set S = {y : [x. y) Q √ by taking x = 0 if necessary. Otherwise. xn } ∈ F for all n. n Now given ε. b + δ ) ⊂ X. Then 0 ≤ sn ≤ ∞ k (p − 1) k=1 1/p = 1 for all n. For any n and any k ≥ n. then lim sn = ∞ so ∞ ∞ = v=1 xv . there exists r ∈√ such that x/ 2 < r < y/ 2. Now. y ∈ R with x < y. since 0 ≤ sn ≤ 1 for all n. If the sequence sn is bounded. Thus sup sn ≤ supF ∈F sF . Suppose a1 . But since X is open. an have n been chosen. Thus for all n. Then r = 0 so r 2 is irrational and x < r 2 < y.
Let δ = min(y − x. Thus x ∈ (Ac )c . namely x.e. 1). This gives a onetoone mapping 8 . suppose x ∈ (Ac )c . Then A ∩ B = ∅ but A ∩ B = [−1. y + δ ). Thus F ∈C F is closed. i. Then for any δ > 0.e. z ∈ E \ {x} and z − x < δ. Thus there exists δ > 0 such that y−x ≥ δ for all y ∈ Ac \{x} = Ac . x + δ). there is a ﬁnite subcollection {F1 . . x ∈ Ac and / x ∈ (Ac ) . Then Oc is closed for any O ∈ C so c c O∈C O is closed. there exists δ > 0 such that {y : y − x < δ} ⊂ A if and only if every point of A is an interior point of A. Thus x is an accumulation ¯ point of Q. y + δ ) and (y − δ . A is open if and only if for any x ∈ A. y ∈ E \ {x}. 34b. Hence (x−δ. Suppose y = x. Let E be an isolated set of real numbers. then Ix0 will contain two elements of E for some x0 . 8/9) of [2/3. Then x ∈ A and x is not an accumulation / point of Ac . ◦ 34a. If y = x. 37. We may assume y > x. Then c F ∈C F = R ⊃ F . / 30. ( F ∈C F )c is open. then x ∈ E and we are done. i=1 Fni = Fnk = ∅. Since x ∈ E. 27.e. (F1 ∪ F2 )c is open. Thus O∈C O is open. Contradiction. Let O1 and O2 be open sets. i=1 Fi = ∅. Since y ∈ E . Let C be a collection of closed sets. c c c c 33. Then {Ix : x ∈ E} is a countable collection of open intervals. Conversely. For any x ∈ E. there exists z ∈ E \ {y} such / that z ∈ (y − δ . x + δ − y). . Then there exists δ > 0 such that (x − δ. Given δ > 0. and E ⊂ x∈E Ix . Given an element in the Cantor ternary set with (unique) ternary expansion an such that an = 1 for all n. Then x ∈ (y − δ . If E is uncountable. 2/3) corresponds to removing all numbers in [0. y + δ ) ⊂ (x − δ. 28. ∞). ( O∈C O) is closed. Fnk } with n1 < · · · < nk . Thus E is countable. each Ix contains only one element of E. c Thus F1 ∪ F2 is closed. Then for any ﬁnite k subcollection {Fn1 . Hence F ∈C F = ∅. Conversely. Then bn may be regarded as the binary expansion of a number in [0. Thus x ∈ E . 1] = {0}. x+δ) ⊂ A so x ∈ A◦ . there exists y ∈ E such that x − y < δ. choose N such that 1/N < ε. x − yN  < δ. ∞ then by Proposition 16. i. In particular. Then F1 and F2 are open so F1 ∩ F2 is open. 36. Conversely. Thus x is a point of closure of E. c c c c 32. Thus A is open if and only if A = A◦ . Let Fn be a sequence of nonempty closed sets of real numbers with Fn+1 ⊂ Fn . Clearly A ⊂ A for any set A. there exists yn ∈ E with yn − x < 1/n. 0] ∩ [0. Hence Q = R and Q = R. there exists r ∈ Q such that x < r < x + δ. 31. Suppose the middle thirds have been removed up to the nth stage. Then F ∩ i=1 Fi = ∅. Then Fn is a sequence of nonempty closed sets of real numbers with ∞ Fn+1 ⊂ Fn but none of the sets Fn is bounded. Then F is open for any F ∈ C so F ∈F F c is open. ¯ ¯ ¯ ¯ 29. Now n=1 Fn = {x : x ≥ n for all n} = ∅. 1] with unique ternary expansion an such that a1 = 1. Contradiction. Suppose F ∈C F = ∅. In particular. there exists δx > 0 such that y − x ≥ δx for all y ∈ E \ {x}. For each n. there exists y ∈ E such that y − x < δ. . . Let C be a collection of open sets. Removing the middle third (1/3. let Fn = [n. Each stage of removing middle thirds results in a closed set and C is the intersection of all these closed sets so C is closed. let bn be the sequence obtained by replacing all 2’s in the ternary expansion by 1’s. . In particular. Let F1 and F2 be closed sets. Removing the middle third (1/9. Then O1 and O2 are closed so O1 ∪ O2 is closed. i. Given δ > 0. 1] corresponds to removing all numbers with unique ternary expansion such that a1 = 2 and a2 = 1. If one of the sets Fn is bounded. Given ε > 0. Fn } ⊂ C such n n that F ⊂ i=1 Fic . 1]. We may take each δx to be rational and let Ix = {y : y − x < δx }. . Let C be a collection of closed sets of real numbers such that every ﬁnite subcollection of C has nonempty intersection and suppose one of the sets F ∈ C is bounded. 1/3] corresponds to removing all numbers with unique ternary expansion such that a1 = 0 and a2 = 1 and removing the middle third (7/9. x + δ) ⊂ A. . Then given / ¯ δ > 0. Let x be a point of closure of E . let x ∈ E and suppose x ∈ E. Suppose x is a point of closure of E. Then for every n. (O1 ∩ O2 )c is closed. suppose there is a sequence yn in E with x = lim yn . . Clearly E ⊂ E and E ⊂ E so E ∪ E ⊂ E. then removing the middle thirds of the remaining intervals corresponds to removing all numbers with unique ternary expansion such that ai = 0 or 2 for i ≤ n and an+1 = 1. Let x ∈ R. 2/9) of [0. Suppose x ∈ A◦ . 38.e. Thus x ∈ E and E ⊂ E ∪ E . 35. Thus O1 ∩ O2 is open. Hence E is closed. 0) and B = (0.¯ ¯ 26. Then yn − x < 1/N < ε for n ≥ N so yn is a sequence in E with x = lim yn . Let A = (−1. By the HeineBorel Theorem. i. there exists N such that x − yn  < δ for n ≥ N .
Now x ∈ U and U is open so there exists δ > 0 such that (x − δ. This mapping is also onto since given a number in [0. there exists M such that for any n. x−1)/6q 2 +δ ≤ max(x+1. x + 1/n) with qn < M . (*) Note that sin x < x for all x > 0 and x < sin x for all x < 0. If y is irrational. Thus x ∈ C . Hence C = C . Let f and g be continuous functions. Also. choose N such that 1/3N < δ. Let O be an open set and let x ∈ f −1 [O]. Now if y ∈ E and y − x < δ. Now let bn be the sequence with bN +1 = aN +1 − 2 and bn = an for n = N + 1. C ⊂ C. suppose that for each open set O. choose δ > 0 such that f (x) − f (y) < ε/2 and g(x) − g(y) < ε/2 whenever x − y < δ. First we show that for any M . Suppose fn is a sequence of continuous functions on E and that fn converges uniformly to f on E. q + δ). 9 . x + δx ). x + 1)qn < M max(x − 1. x + 1) for all n. Contradiction. There also exists δ > 0 such that g(x) − g(y) < δ whenever x − y < δ . then f (x) − f (y) = x − y < δ < ε. If y = p/q is rational. choose M such that M 2 > max(x + 1. This implies that there exists a rational p/q in (x − 1/n. x + δ). Then g is deﬁned and continuous on R and g(x) = f (x) for all x ∈ F . Then (f + g)(x) − (f + g)(y) = f (x) − f (y) + g(x) − g(y) ≤ f (x) − f (y) + g(x) − g(y) < ε whenever x − y < δ. 2. Then y ∈ C \ {x} and x − y = 2/3N +1 < 1/3N < δ. Thus f ◦ g is continuous at x. 44a. F c is open and it is the union of a countable collection of disjoint open intervals. 39. 41. x − sin x < x3 /6 for all x = 0 (by Taylor’s Theorem for example). Then pn  ≤ max(x − 1. f is continuous at each irrational: Let x be irrational. Conversely. Also. there is an open set U such that f −1 [O] = E ∩ U . let ε = f (q) − q > 0. Hence (x − δx . there exists δ > 0 such that q ≥ M for any rational p/q ∈ (x − δ. Thus f is continuous on E. Given ε > 0. let an be its ternary expansion with an = 1 for all n. x + δ) ⊂ U . which we may then regard as the ternary expansion of a number in the Cantor ternary set. 42. fn (x) − f (x) < ε/3.from the Cantor ternary set into [0. pick an irrational x ∈ (q − δ. f (x) + ε) is open so there is an open set U such that f −1 [O] = E ∩ U . Thus (f ◦ g)(x) − (f ◦ g)(y) = f (g(x)) − f (g(y)) < ε when x − y < δ . there exists δ > 0 such that fN (y) − fN (x) < ε/3 if y ∈ E and y − x < δ.6 Continuous functions 40. pick an irrational x ∈ (q. there exists δx > 0 such that f (y) − f (x) < εx when y ∈ E and y − x < δx .e. f is discontinuous at the nonzero rationals: Given a nonzero rational q. If q > 0. Since the Cantor ternary set C is closed. Let f and g be continuous functions. x + δ). Then f (x) ∈ O so there exists εx > 0 such that (f (x) − εx . Then U is open and f −1 [O] = E ∩ U . Then choose δ > 0 such that δ < min(1. f (y) − f (x) < ε. Let x ∈ E and let ε > 0. If q < 0. Let y be the number with ternary expansion bn . Then f (x) − f (q) = f (q) − x > f (q) − q = ε. f (x) ≤ x < ε. there exists a rational pn /qn ∈ (x − 1/n. *43. Since F is closed. Then (f g)(x) − (f g)(y) = f (x)g(x) − f (x)g(y) + f (x)g(y) − f (y)g(y) ≤ f (x)g(x) − g(y) + f (x) − f (y)g(y) < ε whenever x − y < δ . there exists N such that for all x ∈ E and n ≥ N . Given δ > 0. Thus f is continuous on E. x + 1/n) for inﬁnitely many n. Since f is continuous. Otherwise. then f (y) − f (x) ≤ f (y) − y + y − x = p1/q−sin(1/q)+y−x < p/6q 3 +δ < max(x+1. x−1)/6M 2 +δ < 2ε. ε) and q ≥ M for any rational p/q ∈ (x − δ. x + δx ) ∩ E ⊂ f −1 [O]. Then f (x) − f (q) = x − f (q) > q − f (q) = ε. Given ε > 0. given x ∈ C. 44b. Then O = (f (x) − ε. Thus there are only ﬁnitely many choices of pn and qn for each n. Conversely. Now choose δ > 0 such that g(x) − g(y) < ε/2f (x) and f (x) − f (y) < ε/2 max(g(x) − ε/2f (x). q). Suppose f is continuous on E. given any δ > 0. x + δ) ⊂ f −1 [O]. Take g to be linear on each of these open intervals and take g(x) = f (x) for x ∈ F . g(x) + ε/2f (x)) whenever x − y < δ . Thus f g is continuous at x. Thus f + g is continuous at x. Thus E ∩ (x − δ. x − 1)/6ε. Then when x < δ. 1]. there exists δ > 0 such that f (a) − f (b) < ε whenever a − b < δ. let δ = ε. f (x) + εx ) ⊂ O. Let U = x∈f −1 [O] (x − δx . f is continuous at 0: Given ε > 0. 1]. Suppose x − y < δ. given any δ > 0. Given ε > 0. then f (y) − f (x) ≤ f (y) − fN (y) + fN (y) − fN (x) + fN (x) − f (x) < ε. for any y ∈ E with y − x < δ. Given ε > 0. i. we can take its binary expansion and replace all 1’s by 2’s to get a sequence consisting of only 0’s and 2’s.
Then x is in the Cantor ternary set and f (x) = y. Thus f (a) < f (x). Let S = {x ∈ [a. Now (f ∨ g)(x) − (f ∨ g)(y) ≤ f (x) − f (y) + g(x) − g(y) < ε. N Let f (x) = n=1 bn /2n . Given x. then f (a) < f (y) ≤ f (x) so f (y) = f (x ) for some x ∈ [a. Then f  = (f ∨ 0) − (f ∧ 0) so f is continuous. f (x + ε) − f (x)). Hence f is constant on each such interval. Thus f (x) < f (y).44c. Then c ∈ [a. Choose N such that (b − a)/N < δ and let xi = a + i(b − a)/N for i = 1. bN = 1 and bn = 0 for n < N so n=1 bn /2n = 1/2n0 . Given x ∈ [0. Let f be a continuous function on [a. choose δ > 0 such that f (x) − f (y) < ε/2 and g(x) − g(y) < ε/2 whenever x − y < δ. 46. the smallest n such that an = 1 is also the smallest such that an = 1. let bn be its binary expansion. b] and ε a positive number. there exists δ > 0 such that δ < b − c and f (x) − f (c) < γ − f (c) whenever x − c < δ. Then an0 < an0 and bn0 ≤ bn0 . Then f is uniformly continuous so there exists δ > 0 such that f (x) − f (y) < ε/2 whenever x. there exists δ > 0 such that x→y 0<x−y<δ Thus for all x with 0 < x − y < δ. Then S = ∅ since a ∈ S and S is bounded. 1] with ternary expansions an and an respectively. b] and suppose that f (a) ≤ γ ≤ f (b). 1]. Let f be a continuous realvalued function on [a. Let x ∈ [a. choose M such that 1/2M < ε. f (c + δ/2) − f (c) < γ − f (c). Then g(f (x)) = x for all x ∈ [a. 1] and ε > 0. Take x ∈ [0. sup f (x) < A + ε. Hence f (c) = γ. Now choose δ > 0 such that δ < 1/3M +2 . 1] is of the form q/3n0 with q = 1 or 2. Let y ∈ [f (a). . Now if f (x) ≥ f (y). for any x. suppose x < y and let n0 be the smallest value of n such that an0 = an0 . Then ϕ(xi ) ≤ ϕ(x) ≤ ϕ(xi+1 ) and ϕ(x) − f (x) ≤ ϕ(x) − ϕ(xi ) + ϕ(xi ) − f (x) ≤ ϕ(xi+1 ) − ϕ(xi ) + f (xi ) − f (x) < ε. Thus f ∨ g is continuous at x. f ∧ g = f + g − (f ∨ g) so f ∧ g is continuous at x. Thus g is continuous on [f (a). then from the ﬁrst expansion we get N = n0 . then from the ﬁrst expansion we get N = ∞. Hence f is monotone and continuous on [0. then g(f (x)) < g(f (y)) so f (x) = f (y). . f maps [a. We may assume x = a and f (a) < f (b). b] to be linear on each [xi . Suppose lim f (x) ≤ A. N . b]. f (b)]. Given ε > 0. bn = 0 for n < n0 and bn = 1 for N ∞ n > n0 so n=1 bn /2n = n=n0 +1 1/2n = 1/2n0 . 47. Given ε > 0. then by the Intermediate Value Theorem. suppose there exists for any ε > 0 some δ > 0 such that f (x) ≤ A + ε for all x with 0 < x − y < δ. Let g = f −1 : [f (a). Conversely. b]. Then f (c + δ/2) < γ so c + δ/2 ∈ S. Then for each n. b]. From the second expansion we N get N = n0 . Suppose f is strictly monotone. an = 0 for n < n0 and an = 2 for n > n0 . xi+1 ] for some i. We may assume f (xi ) ≤ f (xi+1 ). 49a. Thus Ny Nx f (x) = n=1 bn /2n ≤ n=1 bn /2n = f (y). there exists δ > 0 such that δ < c − a and f (x) − f (c) < f (c) − γ whenever x − c < δ . For any y ∈ [0. if f (c) > γ. b]. f (x) ≤ A + ε. b] with x < y. Then x has two ternary expansions an and an where an0 = q. Furthermore. Suppose x ∈ [0. f (a) = f (x ) for some x ∈ [x. b]. b] : f (x) ≤ γ}. If f (c) < γ. If x. 45. bN = 1 and bn = 0 for n < N so N n n0 n=1 bn /2 = 1/2 . Then x ∈ [xi . Conversely. y ∈ [a. Given ε > 0. 1] with ternary expansion an such that an = 2bn for all n. xi+1 ] with ϕ(xi ) = f (xi ) for each i so that ϕ is polygonal on [a. Then y = f (x) for some x ∈ [a. 1]. If q = 1. y ∈ [a. suppose there is a continuous function g such that g(f (x)) = x for all x ∈ [a. there exists 10 . an0 = q − 1. f (b)]. . y in the same interval and having ternary expansions an and an respectively. b] and x − y < δ. choose δ > 0 such that δ < min(f (x) − f (x − ε). Then f is onetoone. c] so f (x) > γ and x ∈ S for all such x. Each interval contained in the complement of the Cantor ternary set consists of elements with ternary expansions containing 1. In particular. f (b)] → [a. If f (x) < f (a). Hence the sum is independent of the ternary expansion of x. N ∞ bn0 = 1 and bn = 0 for n = n0 so n=1 bn /2n = n=1 bn /2n = 1/2n0 . y ∈ [0. When y − z < δ. Contradiction. Let f be a continuous function. On the other hand. f (b)]. Let c = sup S. Let f be a continuous function on [a. We may assume f is strictly monotone increasing. by the Intermediate Value Theorem. Hence f is strictly monotone increasing. b]. Contradiction. Now deﬁne ϕ on [a. 48. . b]. Contradiction. From the second expansion we get N = ∞. Then when x − y < δ. If q = 2. / Contradiction. an = 0 for n = n0 . z = f (x ) for some x ∈ [a. 44d. Then f (x) − f (c) < f (c) − γ for all x ∈ (c − δ . f (x) − f (y) < ε. b] with g(y) − g(z) = g(f (x)) − g(f (x )) = x − x < ε. 1]. b]. Also. b] and a = g(f (a)) = g(f (x )) = x . Thus f maps the Cantor ternary set onto [0. Let f and g be continuous functions. x] and y = g(f (y)) = g(f (x )) = x . b] onto [f (a). Furthermore.
In particular. Thus xn is a sequence with xn = y and y = lim xn but l = lim f (xn ). sup f (x) ≥ A − 1/n for all n so sup f (x) ≥ A. there exists δ > 0 such that f (y) ≤ f (x) + ε whenever x − y < δ. then lim f (x) x→y x→y exists and equals f (y). lim f (x) ≥ L. Thus for any δ > 0. suppose that given ε > 0 and δ > 0. Suppose lim f (x) = A.e. By part (a). Given ε > 0. Then there exists ε > 0 such that for each n there exists xn with 0 < xn − y < 1/n and x→y f (xn ) − l ≥ ε.e. for each n. For any δ > 0. i. 50a. then sup 0<x−y<δ2 f (x) ≤ f (x) and inf 0<x−y<δ1 f (x) ≤ inf f (x) ≤ sup 0<x−y<δ1 f (x) ≤ f (x). Let L be the common value. f (x) < L + ε whenever 0 < x − y < δ1 . suppose x→y lim f (x) exists and let L be its value. x→y there exists δ > 0 such that f (x) − l < ε whenever 0 < x − y < δ.e. inf x→y inf 0<x−y<δ1 f (x) ≤ sup 0<x−y<δ2 sup 0<x−y<δ1 0<x−y<δ2 Hence sup x→y f (x) ≤ sup 0<x−y<δ0 δ>0 0<x−y<δ f (x) for any δ0 > 0 so sup inf f (x) ≤ inf δ>0 0<x−y<δ δ>0 0<x−y<δ sup f (x). lim f (x) ≤ L. Suppose f is both upper and lower semicontinuous at y. Let f (y) be ﬁnite. Thus for each δ > 0. if δ1 < δ2 . there exists x→y x→y δ1 > 0 such that sup 0<x−y<δ1 f (x) < L + ε. Conversely. f (x) > L − ε whenever 0 < x − y < δ2 . Thus 0 < xn − y < 1/n and f (xn ) − A < 1/n for each n. Given ε. Conversely.δn > 0 such that f (x) ≤ A + 1/n for all x with 0 < x − y < δn so x→y sup 0<x−y<δn x→y f (x) ≤ A + 1/n. there exists δ > 0 such that −f (x) ≤ −f (y) + ε whenever x − y < δ if and only if given ε > 0. 50b. Suppose lim f (x) = lim f (x) with limf = ±∞. l = lim f (xn ). Thus for n ≥ N . sup 0<x−y<δ 49b. By part (b). Similarly. xn is a sequence with xn = y such that y = lim xn and A = lim f (xn ). Thus lim f (x) = inf δ>0 0<x−y<δ x→y sup f (x) ≤ inf n 0<x−y<δ n sup f (x) ≤ A + 1/n for all n so lim f (x) ≤ A. Suppose lim f (x) ≥ A. f (x) − L < ε so lim f (x) exists. Thus lim f (x) = lim f (x) = f (y) so f is both upper and lower semicontinuous 11 . 49d. there exists δ > 0 such that f (x) − L < ε whenever x→y x→y x→y x→y 0 < x − y < δ. Conversely. Suppose l = lim f (x) and let xn be a sequence with xn = y and y = lim xn . i. if f is continuous at y. 49f. there exists xn such that 0 < xn − y < δ and f (xn ) ≥ A − 1/n. δ2 > 0. limf (xn ) ≤ lim f (x) = A.e. For any δ1 . i. x→y 49e. there exists x such that 0 < x − y < δ and f (x) ≥ A − ε. f (x). lim f (x) ≤ lim f (x). x→y there exists Nδ such that 0 < xn − y < δ for n ≥ Nδ . i.e. there exists δ > 0 such that x→y −f (x) ≤ −f (y) + ε whenever 0 < x − y < δ if and only if given ε > 0. Let x→y exists δ2 > 0 such that 0<x−y<δ2 δ = min(δ2 . inf sup f (xn ) ≤ sup f (xn ) ≤ N n≥N n≥Nδ sup 0<x−y<δ f (x) so inf sup f (xn ) ≤ inf N n≥N x→y δ>0 0<x−y<δ sup f (x). 1/n) and f (xn ) > A − 1/n. i. f (xn ) − l < ε. f (x) > A − ε so there exists x such that 0 < x − y < δ and f (x) ≥ A − ε. Given ε > 0. there exists δn > 0 such that f (x) < A + 1/n whenever 0 < x − y < δn . Thus equality holds. i. There also inf f (x) > L − ε. Thus x→y x→y x→y x→y lim f (x) exists and equals f (y) so f is continuous at y. Also there exists N such that 0 < xn − y < δ for n ≥ N .e. lim f (x) ≤ lim f (x). i. By part (a). suppose l = lim f (x). Then lim f (x) ≤ f (y) ≤ lim f (x). Conversely. Then f is lower semicontinuous at y if and only if −f is upper semicontinuous at y if and only if −f (y) ≥ lim (−f (x)) if and only if given ε > 0.e. Given ε > 0 and δ > 0. δ2 ). Then when 0 < x − y < δ. Hence 0<x−y<δ x→y 0<x−y<δ lim f (x) = inf δ>0 0<x−y<δ sup f (x) ≥ A. inf 0<x−y<δ2 49c. there exists xn such that 0 < xn − y < min(δn . Suppose lim f (x) = A and xn is a sequence with xn = y such that y = lim xn . Then for each n.
Conversely. 2. Given ε > 0. Conversely. . then ϕn (y) =≥ f (y) − ε since Ik ∪ Ik+1 ⊂ (y − δ. suppose {x ∈ [a. *50h. f is lower semicontinuous. By part (a). b]. By part (e). then ϕn (y) = inf x∈I (n) f (x) ≥ f (y) − ε since Ik ⊂ (y − δ. Deﬁne ϕn (x) = inf x∈I (n) f (x) if x ∈ Ik and ϕn (xk ) = k min(ck . For each n. Hence f (x) = sup ψn (x) = lim ψn (x). Now if α < f (x). suppose that f is lower semicontinuous. b]. for each x ∈ [a. there exists δ > 0 such that f (y) ≤ f (x) + ε/2 and g(y) ≤ g(x) + ε/2 whenever x − y < δ. b]). xi+1 ) for some i. 50i. b]. Choose N such that 1/2N < δ. i. suppose ϕ(xi ) ≤ min(ci .e. Given ε > 0. b] we have f (x) = lim ψn (x). f (x) is an upper bound for {ψn (x) : n ∈ N}. If y = xk for some k. . there exists δ > 0 such that f (y) ≤ f (x) + ε whenever x − y < δ. Thus f (y) ≤ limf (xn ) ≤ λ so y ∈ S. *50g. y + δ). f is lower semicontinuous. For each n. b] such that for each x ∈ [a. The result for intervals follows from the result for points. b] : f (x) > f (y) − ε} is open so there exists δ > 0 such that f (x) > f (y) − ε whenever x − y < δ. y + δ). xi+1 ) respectively. deﬁne ψn by linearising ϕn at a neighbourhood of each partition point such that ψn ≤ ψn+1 and 0 ≤ ϕn (x) − ψn (x) < ε/2 for each x ∈ [a. let δ = min(xi − xi−1 . 50c. b]. b]. consider the set S = {x ∈ [a. which 12 (n) (n) (n) . Hence S is closed so {x ∈ [a. there is a ﬁnite subcovering. (n) Thus there exists c ∈ Z such that f (x) > c for all x ∈ [a. Let ε > 0 and let y ∈ [a. Hence f is bounded from below. Then each ψn is lower semicontinuous by part (b) and f (x) = supn ψn (x). Thus there exists c ∈ Z such that f (x) > c for all x ∈ [a. ϕ(xi ) ≤ ci + 1/n and ϕ(xi ) ≤ ci+1 + 1/n for each n. On the other hand. b] → R be a step function. Thus f (y) − ϕn (y) ≤ ε. b] so by the HeineBorel Theorem. Then ψn is a monotone increasing sequence of continuous functions on [a. Let y be in the domain of f and g. ck+1 . b]. Thus f (y) ≤ fn (x) + ε ≤ f (x) + ε whenever x − y < δ. In particular. The sets {x ∈ [a. Given ε > 0. Then f (y) = f (xi ) ≤ f (x) + ε whenever x − y < δ. b]. Thus (f ∨ g)(y) ≤ (f ∨ g)(x) + ε and (f + g)(y) ≤ (f + g)(x) + ε whenever x − y < δ. 1. Thus ϕ(xi ) ≤ min(ci . the set {x ∈ [a. b] : f (x) > λ} is open (in [a. Let f be a lower semicontinuous function on [a. then there exists δ > 0 such that α ≤ f (y) ≤ f (y) + ny − x whenever y − x < δ. b] we have f (x) = supn ϕn (x). Let f be a realvalued function. ci+1 ). 2n . there is a ﬁnite subcovering. Since ϕn is monotone increasing. By part (a). Let y be a point of closure of S. k Thus f (y) − ϕn (y) ≤ ε. Thus α ≤ ψn (x) for suﬃciently large n. let δ = min(y − xi . f (y) − ε/2 < fn (y) for some n. f (xk )) where ck (n) (n) (n) (n) = inf x∈I (n) f (x) and ck+1 = inf x∈I (n) f (x). when y − x ≥ δ. b]. ci+1 ) for each i. 50e. there is a monotone increasing sequence ϕn of lower semicontinuous step functions on [a. b]. let xk = a+k(b−a)/2n and (n) (n) (n) (n) (n) let Ik = (xk−1 . b] and f (x) = lim ψn (x) for each x ∈ [a. b] : f (x) > λ} is open (in [a. The sets {x ∈ [a. Suppose f is lower semicontinuous on [a. Then there is a sequence xn with xn ∈ S and y = lim xn . b] we have f (x) = lim ϕn (x). b] with f (x) = lim ϕn (x) for each x ∈ [a. b]. For λ ∈ R. xi+1 − xi ). Let f be a function deﬁned on [a. suppose that f is lower semicontinuous. Then each ϕn is a lower k k+1 (n) semicontinuous step function on [a. In particular. b] : f (x) > c} with c ∈ Z form an open covering of [a. Let f and g be lower semicontinuous functions. b]. b]) for each λ ∈ R. 50d. (*) More rigorous proof: Deﬁne ψn by ψn (x) = inf{f (t) + nt − x : t ∈ [a. If y = xi for some i. Suppose there is a monotone increasing sequence ϕn of lower semicontinuous step functions on [a. b] : f (x) ≤ λ}. Let ci and ci+1 be the values assumed by ϕ in (xi−1 . f is lower semicontinuous at y. f is lower semicontinuous on [a. Then ψn (x) ≤ inf{f (t) + nt − y + ny − x : t ∈ [a. . Conversely. we have α ≤ f (y) + nδ ≤ f (y) + ny − x for suﬃciently large n. b]. Let ϕ : [a. b] : f (x) > c} with c ∈ Z form an open covering of [a. Also ψn ≤ ψn+1 ≤ f for all n. Let y ∈ [a. Then f (y) < f (y) + ε = f (x) + ε whenever x − y < δ. Hence f (y) = lim ϕn (y). b]. b]} = ψn (y) + ny − x. There also exists δ > 0 such that fn (y) ≤ fn (x) + ε/2 whenever x − y < δ. Suppose ϕ is lower semicontinuous. xi+1 − y). (n) (n) For n ≥ N . By part (g). f (y) < f (x) + ε whenever x − y < δ. Thus ϕ is lower semicontinuous. By part (e). there exists δ > 0 such that ϕ(xi ) ≤ ϕ(x) + 1/n whenever x − xi  < δ. Suppose there is a monotone increasing sequence ψn of continuous functions on [a. b] by part (f) and f ≥ ϕn+1 ≥ ϕn for each n. b] such that for each x ∈ [a. Conversely. Since y is arbitrarily chosen from [a. Hence f is lower semicontinuous. b]. xk ) for k = 0. Now for each n.b] f (x). Let m = inf x∈[a. Let f be a function deﬁned on [a. f ∨ g and f + g are lower semicontinuous. . if y ∈ Ik for some k. If y ∈ (xi . Given ε > 0. b] so by the HeineBorel Theorem. b]. Let fn be a sequence of lower semicontinuous functions and deﬁne f (x) = supn fn (x). 50f. b]}. xi ) and (xi . Let y ∈ [a.at y. Thus ψn is (uniformly) continuous on [a.
g(x) ≤ ϕ(x). On the other hand. each Fn.x /2) so x0 ∈ (x − δN. 54. Then x ∈ G for every x ∈ S. Let x ∈ [a. Let f be a lower semicontinuous function on R.x > 0 such that f (x) − f (y) < 1/n whenever x − y < δn. Then there exists δ > 0 such that f (y) > λ whenever x − y < δ. Thus f is continuous at x if and only if f is both upper semicontinuous and lower semicontinuous at x if and only if f (x) = h(x) and g(x) = f (x) if and only if g(x) = h(x). then for any m. Thus f (x) − ε < f (y) whenever y − x < δ and f is lower semicontinuous at x. Thus f (y) − f (x0 ) ≤ f (y) − f (x) + f (x) − f (x0 ) < ε. Then inf f (y) ≤ f (x) ≤ sup f (y) for any δ > 0. b] but 2c − m > m.x /2) for some x ∈ S. Suppose ϕ(x) > g(x) for some x ∈ [a. x0 ∈ GN = x∈S (x − δN. Then there exists δ > 0 such that ϕ(x) ≤ ϕ(y) + ϕ(x) − g(x) whenever x − y < δ. Also. Conversely. Then mA ≤ mA + m(B \ A) = mB. b] belongs to some Ixk so f (y) ≥ 2f (xk ) − m ≥ 2c − m. x + δx ). Thus f (x) − f (x0 ) < 1/N < ε/2. .x /2. then for each n.m is closed.m . Let A and B be two sets in M with A ⊂ B. b] and h is upper semicontinuous. . 51c. Let f be a realvalued function deﬁned on R. Conversely. There exists ε > 0 such that for every δ > 0. g(x) ≤ ϕ(y) whenever x − y < δ. {x : f (x) = α} = {x : f (x) ≥ α} ∩ {x : f (x) ≤ α} is the intersection of a Gδ set with a closed set so it is a Gδ set. b]. b]. x + δN.m ⊂ C. 53.x so f (y) − f (x) < 1/N < ε/2. Contradiction.7 Borel sets 52. Contradiction. The open intervals {Ix : x ∈ [a. Suppose f (x) > m for all x ∈ [a. Let S be the set of points at which f is continuous.e. {x : f (x) ≥ α} = n {x : f (x) > α − 1/n} so it is a Gδ set. Thus 2c − m is a lower bound for f on [a. Consider Fn. there exists y−x<δ δ > 0 such that f (x) − ε = g(x) − ε < inf f (y).x /2 and f (y) − f (x0 ) ≥ ε. there exists δn.is ﬁnite since f is bounded from below. Hence ϕ(x) ≤ g(x) for all x ∈ [a. Hence there exists x0 ∈ [a. suppose f is lower semicontinuous at x.x /2) and G = n Gn . {x : f (x) ≤ α} = {x : f (x) > α}c is closed. Contradiction. b] and g is lower semicontinuous. Let λ ∈ R. Choose N such that 1/N < ε/2. then given ε > 0. If x ∈ C. . Now C ⊂ m n Fn.m = {x : fk (x) − fn (x) ≤ 1/m for k ≥ n}. Suppose h(x) < λ. By a similar argument. Hence {x : h(x) < λ} is open in [a. there exists δ > 0 such that f (x) − ε ≤ f (y) whenever x−y<δ x − y < δ. By continuity of each fk . .m . suppose x0 ∈ G for some x0 ∈ S. Hence G = S and S is a Gδ set. there exists n such that fk (x) − fn (x) ≤ 1/m for all k ≥ n.x /2. if x ∈ m n Fn. b] so by the HeineBorel Theorem. There exists y with y − x0  < δN. f (x) = h(x) if and only if f is upper semicontinuous at x. Suppose g(x) > λ. b] such that m = f (x0 ). x + δN. 51b. b]. i. b]. 13 . Conversely. If f is continuous at x. Hence C is an Fσδ set. there is a ﬁnite subcovering {Ix1 . choose M such that 1/M < ε. there exists δx > 0 such that f (x) ≤ f (y) + m − f (x) for y ∈ Ix = (x − δx . f (x) is an upper bound for { inf f (y) : δ > 0} and given ε > 0. {x : f (x) < α} = {x : f (x) ≥ α}c so it is an Fσ set. f (xn )). Ixn }. 2. Let c = min(f (x1 ). x + δn. Let fn be a sequence of continuous functions on R and let C be the set of points where the sequence converges. . Hence {x : g(x) > λ} is open in [a. In particular. y − x ≤ y − x0  + x0 − x < δN. . Let ϕ be a lower semicontinuous function such that ϕ(x) ≤ f (x) for all x ∈ [a.1 Lebesgue Measure Introduction 1. Then {x : f (x) > α} is open. . b]} form an open covering of [a.x . Now x ∈ n Fn. there / exists y with y − x0  < δ and f (x0 ) − f (y) ≥ ε. Thus fn (x) converges so m n Fn.x /2. Then there exists δ > 0 such that f (y) < λ whenever x − y < δ. b].M so there exists N such that fk (x) − fN (x) ≤ 1/M < ε for k ≥ N . For each x ∈ [a. Suppose g(x) = f (x). Then given ε > 0. 51a. . Each y ∈ [a. 3 3. Thus f (x) − ε ≤ x−y<δ inf f (y) so f (x) = sup δ>0 x−y<δ inf f (y) = g(x). Hence we have g(x) = y−x<δ y−x<δ sup δ>0 y−x<δ inf f (y) ≤ f (x) ≤ inf δ>0 y−x<δ sup f (y) = h(x). Consider Gn = x∈S (x − δn. b].
then Ek is a countably inﬁnite set so n( Ek ) = ∞ and since nEk ≥ 1 for all k. ∞ 12. Then O is an open set such that A ⊂ O and ∗ ∗ m O≤ m In = l(In ) ≤ m∗ A + ε. Let O = In . Then mA = m(A ∪ ∅) = mA + m∅ so m∅ = 0. Then G is a Gδ set such that A ⊂ G and m∗ G = m∗ A. Then m∗ (A ∩ i=1 Ei ) = ∞ ∞ ∞ ∗ ∗ ∗ m ( i=1 (A ∩ Ei )) ≤ i=1 m (A ∩ Ei ) by countable subadditivity. Conversely. nEk = ∞. there is a countable collection {In } of open intervals covering A such that l(In ) ≤ m∗ A + ε. (vi) ⇒ (ii). let En = [−n. 10. Then 1 = m∗ ([0. Now for each n.3 Measurable sets and Lebesgue measure 9. there exists an open set On ⊃ En such that m∗ (On \ En ) < ε/2n . If {In } is a countable collection of open intervals such that E ⊂ In . Then Fm ∩ Fn = ∅ for m = n. 3. if all Ek ’s are ﬁnite sets and {Ek : k ∈ N} is an inﬁnite set. Thus there ∞ N exists N such that n=N +1 l(In ) < ε/2.2. 8. For each n. Conversely. Thus m( En ) = m( Fn ) = mFn ≤ mEn . 1]) = m∗ A ≤ m∗ ( In ) = m∗ ( In ) ≤ l(In ). Given ε > 0. Clearly n is translation invariant and deﬁned for all sets of real numbers. 3. Let A be the set of rational numbers between 0 and 1. Let En be a sequence of sets in M. Then m∗ A = m∗ (A−y) = m∗ ((A−y)∩E)+ m∗ ((A−y)∩E c ) = m∗ (((A−y)∩E)+y)+m∗ (((A−y)∩E c )+y) = m∗ (A∩(E +y))+m∗ (A∩(E c +y)) = m∗ (A ∩ (E + y)) + m∗ (A ∩ (E + y)c ). For each n. Then m(E1 ∪ E2 ) + m(E1 ∩ E2 ) = mE1 + m(E2 \ E1 ) + m(E1 ∩ E2 ) = mE1 + mE2 . there is a ﬁnite union U of open intervals such that m∗ (U ∆E) < ε/3. m∗ E ≤ m∗ (E + y). m∗ B ≤ m∗ (A ∪ B). Let O = In . n( Ek ) = nEk . Hence m∗ (A ∪ B) = m∗ B. Suppose there is a set A in M with mA < ∞. there is a countable collection {In } of open intervals such that E ⊂ In and l(In ) < m∗ E + ε. Then m∗ (U ∆E) = m∗ (U \ E) + m∗ (E \ U ) ≤ m∗ (O \ E) + m∗ (O \ U ) < ε/2 + ε/2 = ε. Then m∗ (A ∪ B) ≤ m∗ A + m∗ B = m∗ B. Hence m∗ (E + y) = m∗ E. Suppose m∗ E = ∞. Suppose m∗ A = 0. since B ⊂ A ∪ B. Then E ⊂ U ∪ O and m∗ ((U ∪ O) \ E) = m∗ ((U \ E) ∪ (O \ E)) ≤ m∗ ((O \ (E \ U )) ∪ (E \ U ) ∪ (U \ E)) < ε. ∞ ∞ Hence m∗ (A ∩ i=1 Ei ) = i=1 m∗ (A ∩ Ei ). Also let {In } be a ﬁnite collection of open m∗ In = l(In ) = intervals covering A. then Ek is a ﬁnite set and since the Ek ’s are disjoint. Let E be a measurable set. n 3. Then E ⊂ O and m∗ (O \ E) = m∗ ( On \ En ) ≤ m∗ ( (On \ En )) < ε. then so is Ek . then E + y ⊂ (In + y) so m∗ (E + y) ≤ l(In + y) = l(In ). 6. Suppose m∗ E < ∞.2 Outer measure 5. 11. ∞). The case where m∗ E < ∞ was proven in part (a). m (A ∩ i=1 Ei ) ≥ n n ∞ ∞ m∗ (A ∩ i=1 Ei ) = i=1 m∗ (A ∩ Ei ) for all n by Lemma 9. Then En+1 ⊂ En for each n. let A be any set and let y ∈ R. By part (a). Let U = n=1 In . Given ε > 0. there is an open set On such that A ⊂ On and m∗ On ≤ m∗ A + 1/n. En = ∅ and mEn = ∞ for each n. 7. O is the union of a countable collection of disjoint open intervals {In } so l(In ) = m( In ) < mE + ε/2. Let Ek be a sequence of disjoint sets of real numbers. If some Ek is an inﬁnite set. Let Ei be a sequence of disjoint measurable sets and let A be any set. 14 . If all Ek ’s are ﬁnite sets and {Ek : k ∈ N} is a ﬁnite set. Suppose E is measurable. Thus m∗ (E + y) ≤ m∗ E. (i) ⇒ (ii). Thus E + y is a measurable set. Let E be a set of real numbers and let y ∈ R. Conversely. Fn ⊂ En for each n and En = Fn . We may assume Ek = ∅ for all k since n∅ = 0. Thus m∗ (A ∩ i=1 Ei ) ≥ i=1 m∗ (A ∩ Ei ). Let O = On . 13b. for each n. E ⊂ O and ∗ m (O \ E) = m(O \ E) = m( In ) − mE = m∗ ( In ) − m∗ E ≤ l(In ) − m∗ E < ε. Thus n( Ek ) = ∞ = nEk . there is an open set O such that E ⊂ O and m∗ (O \ E) < ε/2. (i) ⇒ (ii). Also there is an open set O such that E \ U ⊂ O and m∗ O ≤ m∗ (E \ U ) + ε/3. Suppose E is measurable. On the other hand. Suppose E1 and E2 are measurable. Let G = On . let En = (n. Let F1 = E1 and let Fn+1 = En+1 \ k=1 Ek . by a similar argument. 4. Given any set A and any ε > 0. 13a. Then O is an open set. n] ∩ E. Given ε > 0. (ii) ⇒ (vi).
Suppose E is measurable. 17a. Furthermore. Finally. 1]. For each n. Now let Ei = A ∩ Pi for each i. given ε > 0. which is nonmeasurable. let ri ∞ be an enumeration of Q ∩ [−1. Thus m∗ (G \ E) = 0. {x : f (x) > 0} = E. Let E be a measurable set with E ⊂ P . Let D be a dense set of real numbers and let f be an extended realvalued function on R such that {x : f (x) > α} is measurable for each α ∈ D. 1). Since Ei ⊂ Pi for each i. Let G = On . (v) ⇒ (i). Then f takes each value at most once so {x : f (x) = α} has at most / one element for each α ∈ R and each of these sets is measurable. x + δ) ∩ F c . For each i. Thus m∗ (E \ F ) = 0. 14a. 2) = 3 and m∗ Pi = m∗ P = ∞.5 Measurable functions 18. there is a closed set Fn such that Fn ⊂ E and m∗ (E \ Fn ) < 1/n. Then F ⊂ E and m∗ (E \ F ) ≤ m∗ (E \ Fn ) < 1/n for all n. 3. There exists an Fσ set F such that F ⊂ E and m∗ (E \ F ) = 0. A = A ∩ n∈Z [n. (i) ⇒ (iii). 19. Then [0. 1]) − m( En ) = 0. 1] with f (x) = x + 1 if x ∈ E and f (x) = −x if x ∈ E. Let A be any set with m∗ A > 0. let Pi = P + ri . 1)) = m∗ A > 0. On the other hand. Ei is a disjoint sequence of measurable sets with mEi = mE. then pi ∼ pj and since P contains exactly one element from each equivalence class. let Pi be as deﬁned in part (a) and let Ei = n≥i Pn . Pi ⊂ Ei for each i so 0 < m∗ P = m∗ Pi ≤ m∗ Ei / for each i and lim m∗ Ei ≥ m∗ P > 0 = m∗ ( Ei ). let Ei = E + ri . there exists y ∈ P such that x and y diﬀer by some rational ri . Similarly. For each n. then mEi = m(A ∩ Pi ) = m((A − ri ) ∩ P ) = 0 for each i by Q15. 16. Also. Pi ∩ Pj = ∅ if i = j since if pi + ri = pj + rj . 14b. then there is a nonmeasurable set E ⊂ A. n + 1) and 0 < m∗ A ≤ n∈Z m∗ (A ∩ [n. 13c. there exists y ∈ (x − δ. By part (b). F ⊂ E and m∗ (E \ F ) = m∗ (E \ Oc ) = m∗ (E ∩ O) < ε.(ii) ⇒ (iv). Then E c is measurable. For each n. If A ⊂ [0. 1) and let Pi = P + ri for each i. there exists αn ∈ D such that β < αn < β + 1/n. Then F is closed. Thus mC = m([0. 1]) − α2n−1 /3n = 1 − α. (iv) ⇒ (i). i. Let E be the nonmeasurable set deﬁned in Section 3. each of length less than 1/2n . Let x ∈ [0. x + δ) must intersect one of the intervals removed in step N . i. For each i. Given δ > 0. Now F and E \ F are measurable sets so E = F ∪ (E \ F ) is a measurable set. If each Ei is measurable. Now G and G \ E are measurable sets so E = G \ (G \ E) is a measurable set. 1) ⊂ Pi since for any x ∈ [0. Then ϕ1 + ϕ2 = i=1 αi χAi + i=1 βi χBi is a 15 . 17b. Then mEn = 2n−1 /3n so m( En ) = mEn = 1. On the other hand.4. However. For i=0 each i. we must have mE = 0. Thus F c is dense in [0. then x ∈ n≥k+1 Pn so m∗ ( Ei ) = 0. Then Ei is a sequence with Ei ⊃ Ei+1 for each i and m∗ Ei ≤ m∗ ( Pn ) < ∞ for each i. n + 1) ⊂ A. pi = pj and ri = rj . Let F = Oc . m∗ (O ∩ E) < ε. 1)) = 1. m∗ Ei ≥ m∗ ( Ei ) ≥ m∗ (A ∩ [0. n1 n2 n1 n2 20. 1). Now {x : f (x) > β} = {x : f (x) ≥ β + 1/n} = {x : f (x) > αn } so {x : f (x) > β} is measurable and f is measurable. there is an open set O such that E c ⊂ O and m∗ (O \ E c ) < ε. n + 1) where n ∈ Z. Ei = ∅ since if x ∈ Pk . if A ⊂ [n.e. Let β ∈ R. 1] \ F c . Let F = Fn . let En be the union of the intervals removed in the nth step. Each step of removing open intervals results in a closed set and F is the intersection of all these closed sets so F is closed. Let ϕ1 = i=1 αi χAi and let ϕ2 = i=1 βi χBi .4 A nonmeasurable set ◦ 15. mF = m([0. For each n. Thus mEi = m( Ei ) ≤ m([0. Note that removing intervals in the nth step results in 2n disjoint intervals. Let ri ∞ be an enumeration of Q ∩ [−1. n + 1)) > 0 for some n ∈ Z and there is a nonmeasurable set E ⊂ A ∩ [n. Then E ⊂ G and m∗ (G \ E) ≤ m∗ (On \ E) < 1/n for all n. Since mEi = mE. Let f be deﬁned on [0. There exists a Gδ set G such that E ⊂ G and m∗ (G \ E) = 0. Then (x − δ. 1). Hence Ei0 is nonmeasurable for some i0 and Ei0 ⊂ A. 3. In general.e. there exists an open set On such that E ⊂ On and m∗ (On \ E) < 1/n. choose N such that 1/2N < 2δ. (iii) ⇒ (v). Then Pi is a disjoint i=0 sequence of sets with m∗ ( Pi ) ≤ m∗ [−1. Thus m∗ ( Pi ) < ∗ m Pi . n + 1)) so m∗ (A ∩ [n.
If m ≤ ϕ ≤ M . then x ∈ A so χA (x) = 0. If χAc (x) = 1. {x : f (x) > α} = D1 ∪{x : f D\(D1 ∪D2 ) (x) > α}. then either g(x) = ai = ϕ(x). 22a. D1 = {f g = ∞} = {f = ∞. Let f and g be measurable extended realvalued functions that are ﬁnite a. which is measurable. m=1 or g(x) = 0 and ϕ(x) = ai . then x ∈ A ∪ B. then {x : h(x) > α} = {x : f D\{x:f (x)=±∞} (x) · gD\{x:g(x)=±∞} (x) > α}. {x : h(x) > β} can be written as unions of sets as in part (c). 16 . Now D \ (D1 ∪ D2 ) is a measurable subset of D so the restriction of f to D \ (D1 ∪ D2 ) is measurable. If x ∈ [a. Let ϕ = i=1 ai χAi . D2 = {f g = −∞} = {f = ∞. then x ∈ A \ B or x ∈ B \ A so / χA (x)+χB (x) = 1 and χA χB (x) = 0. let Ek = {x ∈ [a. suppose D1 and D2 are measurable and the restriction of f to D \ (D1 ∪ D2 ) is measurable. . Then g is a step function on [a. Deﬁne ϕ by ϕ = k=−N (kM/N )χEk . we may take g so that m ≤ g ≤ M since both g and ϕ take values in {a1 . g = −∞} ∪ {f = g = −∞} ∪ {f = −∞. Conversely. {x : f (x) > α} = {x ∈ D : f D (x) > α}∪{x ∈ E : f E (x) > α}. Suppose f is measurable. χA∩B (x) = 1 if and only if x ∈ A and x ∈ B if and only if χA (x) = 1 = χB (x). Each En is measurable and En+1 ⊂ En for each n. Since D and E are measurable subsets of D∪E. If m ≤ f ≤ M . b]. / Hence χA∪B = χA + χB + χA χB . Let f and g be measurable extended realvalued functions deﬁned on D. If α ≥ 0. then {x : g(x) > α} = {x : f (x) > α}. If g(x) = ϕ(x). 23b. there Ni is a ﬁnite union Ui = k=1 Ii. possibly with an additional set of measure zero. If χAc (x) = 0. For each i. . . which is measurable. Deﬁne g by n g = i=1 ai χUi \ i−1 Um . suppose g is measurable.e. f D and f E are measurable. −N + 1. f is measurable. then {x : h(x) > α} = {x : f (x) = 0} ∪ {x : g(x) = 0} ∪ {x : f D\{f =±∞} (x) · gD\{g=±∞} (x) > α}. x ∈ Ui \ Ai . an }. g = ∞} ∪ {x : f D\{f =±∞} (x) + gD\{g=±∞} (x) > β}. Thus χA∩B = χA χB . which is measurable. then {x : h(x) > β} = {x : f D\{f =±∞} (x) + gD\{g=±∞} (x) > β}.k of (disjoint) open intervals such that m(Ui ∆Ai ) < ε/3n. g = ∞} ∪ {f < 0. g ∈ R}. g > 0} ∪ {f > 0. i. If χA∪B (x) = 1. Let D and E be measurable sets and f a function with domain D ∪ E. b] that takes the values ±∞ only on a set of measure zero and let ε > 0. Suppose f is measurable. b] such that f (x) < M . Then D1 and D2 are measurable by Proposition 18. g = −∞} ∪ {f = −∞. which is measurable. Thus there exists M such that mEM < ε/3. then {x : g(x) > α} = {x : f (x) > α} ∪ Dc . If α < 0. Then for any α ∈ R. then x ∈ Ek for some k. 23a. Thus n n {x ∈ [a. Hence f + g is measurable. Hence f g is measurable. Let f be a measurable function on [a. If x ∈ A ∩ B. then {x : h(x) > β} = {f = ∞. If β ≥ α. Each set on the right is measurable so {x : f (x) > α} is measurable and f is measurable. . n 23c. then x ∈ A so / χA (x) = 1. we may take ϕ so that m ≤ ϕ ≤ M by replacing M/N by (M − m)/N < ε in the preceding argument. D2 . Then ϕ is a simple function. . which / is measurable. g = −∞} ∪ {f < 0. For α ∈ R. Conversely. If α < 0. then x ∈ A∪B so χA (x) = χB (x) = χA χB (x) = 0. which is measurable. b] and let ε > 0 be given. b]. Choose N such that M/N < ε. If χA∪B (x) = 0. g < 0} ∪ {f > 0. f  ≤ M except on a set of measure less than ε/3. i. which is measurable. g = ∞}∪ {f = g = ∞} ∪ {f = ∞. . 22b. Each Ek N −1 is measurable. D1 = {f +g = ∞} = {f ∈ R. For each n. let En = {x ∈ [a. Let g be deﬁned by g(x) = f (x) if x ∈ D and g(x) = 0 if x ∈ D. g = −∞}. Let f be an extended realvalued function with measurable domain D and let D1 = {x : f (x) = ∞}. g > 0} ∪ {f = −∞. Thus lim mEn = m( En ) = 0. D2 = {f + g = −∞} = {f ∈ R.e. Also ϕ1 ϕ2 = i. 22c. Thus f (x) − ϕ(x) < M/N < ε. g ∈ R}. Let f and g be measurable extended realvalued functions deﬁned on D and α a ﬁxed number. If α ≥ 0. Thus these sets are measurable and f + g is measurable. In the ﬁrst case. which is measurable. then χA (x) + χB (x) + χA χB (x) = 1 + 1 − 1 = 1. x ∈ Ai \ Ui . Suppose f is measurable. .j αi βj χAi χBj is a simple function. Let f be a function with measurable domain D. N − 1}.simple function. 22d. Hence f is measurable. mE1 ≤ b − a < ∞. If β < α. which is measurable. In the second case. suppose f D and f E are measurable. Hence g is measurable. b] : ϕ(x) = g(x)} ⊂ i=1 ((Ui \ Ai ) ∪ (Ai \ Ui )) = i=1 (Ui ∆Ai ) so it has measure less than ε/3. Also. . b] : kM/N ≤ f (x) < (k + 1)M/N }. Let h = f gD\(D1 ∪D2 ) and let α ∈ R. Let f be a measurable function on [a.e. kM/N ≤ f (x) < (k + 1)M/N and ϕ(x) = kM/N . Let ε > 0 and M be given. D2 = {x : f (x) = −∞}. 21b. Deﬁne f +g to be α whenever it is of the form ∞−∞ or −∞+∞. which is measurable. Then the sets D1 . Let ϕ be a simple function on [a. For each k ∈ {−N. Conversely. 21a. which is measurable. g < 0} ∪ {f = −∞. If x ∈ A ∩ B. Let h = (f + g)D\(D1 ∪D2 ) and let β ∈ R. g = ∞}. b] : f (x) > n}. Hence χAc = 1 − χA . Then f = gD and since D is measurable.
Let η > 0 be given. ∞)] is a Borel set and g −1 [f −1 [(α. 2]. If f is Borel measurable and B is a Borel set. the set {x : f (x) > α} is a Borel set so it is Lebesgue measurable. Then A ⊂ C so it has outer measure zero and is measurable but g −1 [A] = E so it is nonmeasurable. Let f be deﬁned by f (x) = f1 (x) + x for x ∈ [0. Then f −1 [ Ei ] = f −1 [Ei ]. ∞)]]. 28e. If m ≤ g ≤ M . 2]. fk (x)−f (x) < 1/n.∞) for each n. then for any α ∈ R. Let d = min{xi − xi−1 : i = 1. An Ameasurable function need not be Lebesgue measurable. Now for any a. Then h is continuous n and {x ∈ [a.48. Egoroﬀ ’s Theorem: Let fn be a sequence of measurable functions that converges to a realvalued function f a. Since mF = 1 > 0. 26. so E c ∈ C. By a similar argument to that in Q24. . so Ei ∈ C. on a measurable set E of ﬁnite measure. Hence f −1 [B] is a Borel set. f1 is continuous and monotone on [0. take g and A to be Lebesgue measurable with g −1 [A] nonmeasurable. Thus f ◦ g is Borel measurable. {x : (g ◦ f )(x) > α} = (g ◦ f )−1 [(α. If f and g are Lebesgue measurable. which is measurable. Let E = R and let fn = χ[n. 2]) − 1 = 1. xn be the partition points corresponding to g. d/2) centred at xi . . 1] onto [0. Propositions 18 and 19 and Theorem 20 still hold. Thus (a. {x : (f ◦ g)(x) > α} = (f ◦ g)−1 [(α. Suppose Ei is a sequence of sets in C. Let f be measurable and B a Borel set. pick x ≥ N such that x ∈ A. C is a σalgebra containing all the open intervals. Thus f is Lebesgue measurable. 28a. Thus f ◦ g is Lebesgue measurable. The function g = f −1 is continuous and the function h = χA is measurable. Deﬁne h by linearising g in each Ii . 1]. The set A in part (c) is measurable but by Q24. There exists a Lebesgue measurable function g and a Lebesgue measurable set A such that g −1 [A] is nonmeasurable (see Q28).46. If f and g are Borel measurable. let Ii be an open interval of length less than min(ε/3(n + 1). For each i. 1] \ C) = 1 and since f is a bijection of [0. 25. ∞)] = g −1 [f −1 [(α. If f is Borel measurable and g is Lebesgue measurable. Thus C is a σalgebra. Hence f −1 [B] is measurable. Then χP is Ameasurable but not Lebesgue measurable. 28d. .e. which is measurable by Q24. {x : f (x) > a} and {x : f (x) < b} are measurable. it is not a Borel set since g −1 [A] is nonmeasurable. By Q2. which is a Borel set. 27.48. For example. b) are in C. Let g = f −1 : [0. . Let f be a measurable realvalued function and g a continuous function deﬁned on (−∞. Then f −1 [E c ] = (f −1 [E])c . Let f = χA so that f is Lebesgue measurable. Then fN (x) − 0 ≥ 1. Let g be a step function on [a. 1] onto [0. b) ∈ C and C is a σalgebra containing all the open intervals so it contains all the Borel sets. By Q2. Suppose E ∈ C.e. f −1 [(α. . b ∈ R with a < b.23d. let A be the σalgebra generated by the nonmeasurable set P deﬁned in Section 4. For any α ∈ R. then for any α ∈ R. Let A = f −1 [E]. / 17 . b] and let ε > 0 be given. . 1]. which is nonmeasurable. Let x0 . Hence g ◦ f is measurable. ∞) and (−∞. we may take h so that m ≤ h ≤ M by construction. By Q2. mF = m(f [C]) = m([0. n}. i.6 Littlewood’s three principles 29. For example. 1] \ C]) = m([0. ∞)]]. Thus f maps each of these intervals onto an interval of the same length. However the set {x : (h ◦ g)(x) > 1/2} = g −1 [A] is nonmeasurable. f1 is constant on each interval contained in the complement of the Cantor set. Then g is measurable. . where A is as deﬁned in part (c). f ◦g may not be Lebesgue measurable. Let C be the collection of sets E such that f −1 [E] is measurable. ∞)]] is Lebesgue measurable by Q24. f has a continuous inverse so it is a homeomorphism of [0. Then g is also measurable. 3. If f is a Borel measurable function. / 30. 24. ∞)] = f −1 [g −1 [(α. Thus m(f [[0. For any measurable set A ⊂ E with mA < 1 and any integer N . 1] onto [0. Propositions 18 and 19 and Theorem 20 follow from arguments similar to those in the original proofs and the fact that the collection of Borel sets is a σalgebra. 28c. (a. which is measurable. Then fn (x) → 0 for each x ∈ E. This is also a counterexample for the last statement. there is a nonmeasurable set E ⊂ F . b] : g(x) = h(x)} ⊂ i=0 Ii . 2]. Then {x : (f ◦ g)(x) > 1/2} = g −1 [A]. then consider the collection C of sets E such that f −1 [E] is a Borel set. 28b. Thus C contains all the Borel sets. ∞). Thus h ◦ g is not measurable. 1] and f maps [0. Call a function Ameasurable if for each α ∈ R the set {x : f (x) > α} is in A. 2] → [0. which has measure less than ε/3. there exists An ⊂ E with mAn < η/2n and there exists Nn such that for all x ∈ An and k ≥ Nn . then for α ∈ R. For each n. 1] so f is continuous and strictly monotone on [0. .
the set {x : h(x) < α} is open. i. ∞). Let E = En . Since i=1 (A + ri ) ⊂ [−1. b] such that m{x : hn (x) − f (x) ≥ δ/2n+2 } < δ/2n+2 . thus measurable. Then f is continuous on [a. Now the measurable sets A+ri are disjoint by the deﬁnition n n of P and the construction of the sequence ri . ϕ is continuous on [a. For each x. For any subdivision a = ξ0 < ξ1 < · · · < ξn = b of [a. Then ϕ ≥ h except at a ﬁnite number of points (the partition points). Hence R b a f (x) = b − a and R b a f (x) = 0. b] such that ϕ ≥ f . Now there is an open set O such that O ⊃ (E ∪ A) and m(O \ (E ∪ A)) < δ/2. there is a continuous function ϕn on [n + δ . Lusin’s Theorem: Let f be a measurable realvalued function on [a. . ft (x) → 0 as t → 0. Then we have a continuous function ϕ deﬁned on (−∞. 2]. . which is closed. n + δ ]. x ∈ Pi(x) for some i(x). Thus R a f = inf ϕ≥f a ϕ ≥ a h. h is upper semicontinuous by Q2. there exists a monotone decreasing sequence ϕn of step functions such that h(x) = lim ϕn (x) for each x ∈ [a.. Then hn (x) − f (x) < δ/2n+2 for x ∈ [a. there is a function ϕ that is continuous on (−∞. Thus S = i=1 (ξi − ξi−1 )Mi = b − a and s = i=1 (ξi − ξi−1 )mi = 0 for any subdivision of [a. there is at most one x such that ft (x) = 1 so each ft is measurable. Mi = supξi−1 <x≤ξi f (x) = 1 and mi = inf ξi−1 <x≤ξi f (x) = 0 for each n n i. For t ∈ [0.40. ∞) such that f = ϕ on [a. Thus f is continuous on [a. Similarly for intervals [−n − δ .51b and h is bounded. b]\(E ∪A) with m(E ∪A) < δ/2. b]. n + 1 − δ ] such that m{x ∈ [n + δ . 1) \ A) ≥ 1/2 so there exists x ∈ [0. 1) with mA < 1/2. Linearise in each interval [n − δ . Thus for each x. Similarly for [−n − 1 + δ .e. For each t. . there is a subset A ⊂ [a. there is a continuous function hn on [a. b] and let δ > 0 be given. 18 . i.e. 1/2). Let ϕ be a step function on [a. 1) with positive measure intersects inﬁnitely many of the sets Pi : n Suppose A intersects only ﬁnitely many of the sets Pi . Thus b a b b b h = lim b a ϕn ≥ R b a f . i. *32. 1] such that r1 = qi − qj for all i. If x ∈ A and / k ≥ Nn0 . Let En = {x : hn (x) − f (x) ≥ δ/2n+2 }. we have fk (x) − f (x) < 1/n0 < η. . b] \ E such that mA < δ/4 and hn converges uniformly to f on [a.1 The Lebesgue Integral The Riemann integral 1a. Let f be deﬁned by f (x) = 0 if x is irrational and f (x) = 1 if x is rational. where Pqi = P + qi . Then m( i=1 (A + ri )) = i=1 m(A + ri ) = nmA for n each n. b] and let h be the upper envelope of f . Since f is bounded. Then A ⊂ E and mA < η/2n = η. b] \ En ... Let f be a bounded function on [a. rn have been chosen. −n − δ ]. −n + δ ]. b]. For each n. A ⊂ i=1 Pqi . deﬁne ft : [0. 4. Thus h is measurable. n + 1 − δ ] : f (x) = ϕn (x)} < δ/2n+2 . 1) with 1/2i+1 ≤ t < 1/2i . let δ = min(δ/2n+3 . 1) \ A and arbitrarily small t such that ft (x) ≥ 1/2. This is the only t such that ft (x) = 1. functions that converges to f on [a. b] \ E. there exist an x ∈ [0. Hence R b a f= b a h. Thus the limit of a sequence of Riemann integrable functions may not be Riemann integrable so in general we cannot interchange the order of integration and the limiting process. Thus for any measurable set A ⊂ [0.Let A = An . Suppose r1 . b] and m{x ∈ [a. Let t(x) = (x + 1)/2i(x)+1 .2 The Lebesgue integral of a bounded function over a set of ﬁnite measure δ>0 y−x<δ 2a. For each n. Thus fn converges to f uniformly on E \ A. Choose r1 ∈ Q ∩ [−1. If f is deﬁned on (−∞. b] : f (x) = ϕ(x)} = mO < δ. Note that any measurable subset of [0. and ft (x) = 0 otherwise. Conversely. Choose rn+1 such that rn+1 = qi −qj +rk for all i. 31. and mO < δ.e. b] \ O. 4 4. 1b. 1) \ A with i(x) arbitrarily large and so with t(x) arbitrarily small. For any α ∈ R. Then 1/2i(x)+1 ≤ t(x) < 1/2i(x) and ft(x) (x) = 1. h(y) = inf sup f (x). b]\(E ∪A).rn } . nmA ≤ 3 for all n and mA = 0.. m([0. Then fn is a sequence of nonnegative Riemann integrable functions increasing monotonically to f . Then mE < δ/4 and hn is a sequence of continuous. Choose n0 such that 1/n0 < η. 1) → R by ft (x) = 1 if x ∈ Pi = P + ri and x = 2i+1 t − 1. 1) with positive measure intersects inﬁnitely many of the sets Pi . let fn = χ{r1 . By Q2. j and k ≤ n. j. b] \ O. b]. By Egoroﬀ’s Theorem. b]. Let rn be an enumeration of Q ∩ [a. Then for each n. ∞) with m{x : f (x) = ϕ(x)} < 4 δ/2n+2 = δ. (*) Any measurable set A ⊂ [0. In particular.
for suﬃciently large n. f ≥ lim fn since f ≥ fn for each n. . lim gn = f χ(x. Hence f = lim fn . Hence f = sup ϕ over all simple functions ϕ ≤ f . . n] and f (x) < n. f ≤ lim fn .∞) . Thus f  = f + + f − is integrable over E and  E f  =  E f + − E f −  ≤  E f +  +  E f −  = E f + + E f − = E f . if f  is integrable 19 . Then fn is an increasing sequence of nonnegative measurable functions with f χ(−∞. Let f be a nonnegative measurable function. 6. fn χE is a sequence of nonnegative measurable functions with fn χE → f χE a. Also. let En. Now f ≥ (1/n)mEn for each n so mEn = 0 for each n and mE = 0.x−1/n] . 4a. By Fatou’s Lemma.e.. Hence F is continuous. and mE = 0. then so are f + and f − .e. On the other hand. ∞). Let f be a nonnegative integrable function and let F (x) = −∞ f . so R b a f = b b a g= b b a h=R b a f so f is Riemann integrable.e. b] and let E be the set of discontinuities of f . we have lim( f − −∞ f ) = f − −∞ f so lim −∞ f = −∞ f . there exists N such that F (x) − F (x − 1/n) < ε and F (x + 1/n) − F (x) < ε whenever n ≥ N . For each n. Then by part (a). (f − f χE ) ≤ lim (fn − fn χE ). 7b. Furthermore. Let g be the b b lower envelope of f . By a similar argument as in part (a).n]) + nχ(En. Choose δ < 1/N .4 The general Lebesgue integral 10a. let fn = f χ(−∞. ϕn (x) = n → ∞ if f (x) = ∞. 4b.x] = F (x).e. Then gn is an increasing sequence of nonnegative measurable functions with ∞ f χ(x. . Then fn is a decreasing sequence of nonnegative measurable functions with lim fn = 0. Let fn be a sequence of nonnegative measurable functions such that fn → f a. i. let fn = χ[n.2b. let hn = inf k≥n fk . Then g = h a. i. If f is integrable over E.e. Conversely. let gn = f χ(x+1/n. g − h ≥ (1/n)m{x : 4. By the Monotone Convergence Theorem. . Then each hn is a nonnegative measurable function with hn ≤ fn . Let f be a nonnegative measurable function. Also. Then E = En where En = {x : f (x) ≥ 1/n}. 7a. By the Monotone Convergence Theorem. Now 0 = 0 < ∞ = lim fn so the Monotone Convergence Theorem does not hold. Thus a g − h = 0 so g = h a. Thus ϕn (x) → f (x) when f (x) < ∞. n2n and let En. f − E f ≤ lim fn − lim E fn = f − lim E fn . Thus lim E fn ≤ E f and we have E fn → E f .0 = f −1 [n.n]) . Then for any measurable set E. For each n. Now f χE is integrable since f χE ≤ f . . Let fn be a sequence of nonnegative measurable functions. Thus f = 0 a. lim F (x + 1/n) = F (x).e. let fn = χ[n. suppose f is a b b a Riemann integrable. there is an increasing sequence ϕn of simple functions such that f = lim ϕn . Let f be a bounded function on [a. . E f ≤ lim E fn . Now 0 = 0 < 1 = lim fn and we have strict inequality in Fatou’s Lemma. limfn = lim hn ≤ lim hn ≤ lim fn .0 ∩[−n. Let fn be a sequence of nonnegative measurable functions converging to f and suppose fn ≤ f for each n. Hence f is continuous a. i2−n ) where n2n i = 1. . Now for each n. 2. Thus f ≥ sup ϕ over all simple functions ϕ ≤ f . Thus f ≤ sup ϕ over all simple functions ϕ ≤ f .∞) . f = lim ϕn = sup ϕn . Now given ε > 0. By Fatou’s Lemma. By the Monotone Convergence Theorem. Conversely. lim F (x − 1/n) = lim fn = f χ(−∞. x 5.i = f −1 [(i − 1)2−n . and suppose that fn → f < ∞. i. Since f is integrable. lim x+1/n f = f . Deﬁne ϕn = i=1 (i − 1)2−n χ(En.∞) = lim gn . fn (x) − ϕn (x) ≤ 2−n if x ∈ [−n. fn is integrable for suﬃciently large n so fn χE is integrable for suﬃciently large n.e. By Fatou’s Lemma. Let E = {x : f (x) > 0}.∞) .3 The integral of a nonnegative function 3.e. f ≥ ϕ for all simple functions ϕ ≤ f . since g(x) − h(x) > 1/n} for all n. For each n. 8. For n = 1. By Fatou’s Lemma. for suﬃciently large n.e.n+1) . Then fn is a sequence of nonnegative measurable functions with lim fn = 0. Then each ϕn is a nonnegative simple function vanishing outside a set of ﬁnite measure and ϕn ≤ ϕn+1 for each n. Then a g = a h.i ∩[−n.x] = lim fn . ∞ x x+1/n x x+1/n x 4. On the other hand. Suppose E has measure zero. Let f be a nonnegative measurable function and suppose inf f = 0. R f = a g. For each n. 9. Then F (y) − F (x) < ε whenever x − y < δ.
N ].N ] b f a+1/n b R af = = lim fn = b −n b a f . Let gn be a sequence of integrable functions which converges a. In general. where M ≥ max(ϕ. fn − f  → 0. then E f + ≤ E f  < ∞ and E f − ≤ E f  < ∞ so f + and f − are integrable over E and f is integrable over E.e. Thus fn − f  ≤ fn  + f  ≤ gn + g and gn + g − fn − f  is a sequence of nonnegative measurable functions. Also. there is a simple function ϕ such that E f − ϕ < ε/3. i. 15c. b] b f − ϕ + E∩[−N.e. 13. E 15b. there is a step function ψ on [−N. Then ϕ+ = i∈S+ ai χAi and ϕ− = − i∈S− ai χAi . Conversely. By Q4. ∞] although R 0 f (x) = π/2 (by contour integration for example). Let S+ = {i : ai ≥ 0} and let S− = {i : ai < 0}. g−fn is a sequence of nonnegative measurable functions on E. 14a. with fn  in place of gn and f  in place of g. Suppose g = lim gn . We may regard ψ as a function on R taking the value 0 outside [a. Hence lim fn − f  ≤ 0 ≤ lim fn − f  and we have fn − f  → 0. The cases where the Riemann integral has improper upper limit are similar. Let ϕ = ψ−ψ . ψ) + 1. Then fn + g is a sequence of nonnegative measurable functions on E. By part (b).22. We may regard ψ as a function on R taking the value 0 N outside [−N. By Proposition 3. Thus fn  → f . b]. Then E f − g ≤ E f − ψ + E ψ − g < ε.N ] f − ψ + E∩[−N. Thus g+ lim(−fn ) ≤ lim(g−fn ) ≤ lim (g − fn ) ≤ g + lim(− fn ) so lim(−fn ) ≤ lim(− fn ). b] and fn  ≤ f  so R b a b f = lim R and gn  ≤ f  so lim R f = lim gn = a f . there is a step function ψ such that E f − ψ < ε/2.N ]c f  < ε/3. lim fn ≤ limfn . E∩[−n. 15a. By Fatou’s Lemma. 20 . By linearising ψ at each partition point. f  ≤ g.e. let fn = f χ[a+1/n. f − f χ[−n. Let fn be a sequence of integrable functions such that fn → f a. RiemannLebesgue Theorem: Suppose f is integrable on (−∞. Similarly. there is a step function ψ such that f − ψ < ε/2. By part (a). Thus limfn + g ≤ lim(fn + g) ≤ lim (fn + g) ≤ lim fn + g so limfn ≤ lim fn . b].e. then suppose fn  → f . Since fn  ≤ gn .b] . If fn  − f  ≤ fn  − f  ≤ fn − f  → 0. N ] such that ϕ − ψ < ε/12N M except on a set of measure less than ε/12N M . Then ϕ is a simple function and E f −ϕ = E f + −ψ−f − +ψ  ≤ E f + −ψ+ E f − −ψ  = (f + − ψ) + E (f − − ψ ) < ε.N ]c f − ψ ≤ E∩[−N.b] . i. Clearly ϕ+ and ϕ− are simple functions. Also. Thus there exists N such that ψ(x) cos nx dx < ε/2 for n ≥ N so  f (x) cos nx dx < ε for n ≥ N . lim(gn + g − fn − f ) ≤ lim (gn + g − fn − f ).e. we get a continuous function g vanishing outside a ﬁnite interval such that ψ = g except on a set of measure less than ε/4M .e.e.n] . By Q15. ∞ 10b. E By part (a). we see that ψ(x) cos nx dx → 0 as n → ∞. we have fn − f  → 0. there is a simple function ψ ≤ f − such that E f − − ε/2 < E ψ . to an integrable function g and let fn be a sequence of measurable functions such that fn  ≤ gn and fn converges to f a. i. let gn = f χ[−n. Now f + h ≤ lim(fn + h) ≤ lim fn + h so f ≤ lim fn . on E. Let h be an integrable function and fn a sequence of measurable functions with fn ≥ −h and lim fn = f . Suppose ψ is deﬁned on [a.N ]c f  < ε. 14b. 16. Then gn → f on [a. 2g ≤ 2g + lim(− fn − f ) = 2g − lim fn − f .over E. Integrating ψ(x) cos nx over each interval on which ψ is constant. fn + h is a nonnegative measurable function. there is a simple function ψ ≤ f + such that E f + − ε/2 < E ψ. where M ≥ ψ. with f integrable.n]c f  → 0. If a is ﬁnite. Let ϕ = i=1 ai χAi be a simple function with canonical representation. i. Let f be integrable over E and let ε > 0 be given. f = (f + h) − h.N ]c f − ψ ≤ E f − ϕ + N −N ϕ − ψ + E∩[−N. Thus there exists N such that E∩[−N. ∞). Then ϕ+ = i∈S+ ai mAi and ϕ− = − i∈S− ai mAi so ϕ = ϕ+ − ϕ− = n i=1 ai mAi . Let g be an integrable function on a set E and suppose that fn is a sequence of measurable functions with fn  ≤ g a. If a = −∞. Then −N ϕ − ψ < ε/3 so E f − ψ = E∩[−N. Then fn → f on [a. Now  f (x) cos nx dx ≤ f (x) cos nx dx ≤ (f (x) − ψ(x)) cos nx dx + ψ(x) cos nx dx < ε/2 + ψ(x) cos nx dx. f (x) = sin x/x is not Lebesgue integrable on [0.e. given ε > 0. Then fn → f and fn  ≤ f . Let fn = f χ[−n. For each n. suppose f is Lebesgue integrable and the Riemann integral R a f exists with improper lower limit a. 12. Hence we have limfn ≤ lim fn ≤ lim fn ≤ limfn . By Lebesgue’s Dominated Convergence Theorem. n 11.n]  → 0.N ] ϕ − ψ + E∩[−N. fn = (fn + h) − h. Since h is integrable.
Then there is a subsequence fnk such that lim E fnk = lim E fn . Suppose that ∂f /∂t is bounded in Q. t)]/sn ) dx = 1 ∂f dx. t) ≤ g(x) where g is an integrable function on [0. there exists M such that [f (x. b] so there exists δ > 0 such that h(x)−h(x+t) < ε/2M (b−a) whenever t < (f − h)(x) + (f − h)(x + t) = ε/2 + 2M f − h < ε δ. hn (x) = f (x. f = lim fn . For each x. Thus f (x) dx = f (x + t) dx. then so does any subsequence fnk . t) = f (x) and that for all t we have f (x. Since ∂f /∂t is bounded. i. 1] × [0.e.e. t) dx]/sn = limn 0 ([f (x. the last equality following from Lebesgue’s Dominated Convergence Theorem. Then given ε > 0. t+sn ) dx− 0 f (x. t deﬁned on the square Q = [0. If fn converges to f 21 . i. choose M such that nk ≥ N for k ≥ M .12. t)]/sn → ∂f /∂t. Any subsequence fnk also converges in measure to f so it in turn has a subsequence fnkj that converges to f a. By Lebesgue’s Dominated Convergence Theorem.e. t+sn )−f (x. lim f (x. Now since χE (x) dx = mE = m(E − t) = χE (x + t) dx for any measurable set E. t)]/sn ) dx. E Monotone Convergence Theorem: Let fn be an increasing sequence of nonnegative measurable functions that converges in measure to f .e. 17b. Thus any subsequence of fnk also converges to f in measure. Thus 1 1 1 1 d dt 0 f (x.5 Convergence in measure 20. Hence h is a continuous function of t. t + sn ) − f (x. Then lim f (x. 1] and which is a measurable function of x for each ﬁxed t. This gives rise to a subsequence fnk such that m{x : fnk (x) − f (x) ≥ ε} ≥ ε for all k. i. Then limn [f (x. then there exists ε > 0 such that for any N there exists n ≥ N with m{x : fn (x) − f (x) ≥ ε} ≥ ε. t)]/sn  ≤ M + 1 for suﬃciently large n. there exists N such that m{x : fn (x) − f (x) ≥ ε} < ε for n ≥ N . f is a bounded 1 1 1 measurable function of x so [ 0 f (x. there is a continuous function h vanishing outside a ﬁnite interval [a. Now g(x)[f (x) − f (x + t)] ≤ g(x)[h(x) − h(x + t)] + g(x)[(f − h)(x) − (f − h)(x + t)]. tn ) = f (x). Lebesgue’s Dominated Convergence Theorem: Let g be integrable over E and let fn be a sequence of measurable functions such that fn  ≤ g on E and converges in measure to f on E. Since f is integrable. For any subsequence fnk . 19. Let tn be a sequence converging to t. By Lebesgue’s Dominated Convergence Theorem. let the partial derivative ∂f /∂t exist. 1]. if fn does not converge in measure to f . 22. By Q2.limn→∞ f (x) cos nx dx = 0. For each n. Let fn be a sequence that converges to f in measure. limn hn = f . Suppose further that f (x. f + (x) dx = f + (x + t) dx.12. t in Q. tn ) = f (x. we have ϕn (x) dx = ϕn (x + t) dx for all n. Any subsequence fnk also converges in measure to f so it in turn has a subsequence fnkj that converges to f a. Now h is uniformly continuous on [a. t + sn ) dx − 0 f (x.e. limt→0 g(x)[f (x) − f (x + t)] = 0. Let fn be a sequence of measurable functions on a set E of ﬁnite measure. 23.e. t + sn ) − f (x. tn ) is measurable and hn  ≤ g. Conversely. f = lim fn . 21. If fn converges to f in measure. Let g be a bounded measurable function and let M be such that g ≤ M . Let tn be a sequence such that tn = 0 for all n and limn tn = 0. Let sn be a sequence such that sn = 0 for all n and limn sn = 0. tn ) dx = f (x. 18. Then g(x)[f (x) − f (x + t)] ≤ ε/2 + M whenever t < δ. t) dx]/sn = 0 ([f (x. t) dx. Thus f = lim fnkj . given ε > 0. Suppose that limt→0 f (x. b] such that f −h < ε/4M . t) for each x. Thus f ≤ lim E fnkj = lim E fnk = lim E fn . t) dx = f (x) dx. t) dx. t) dx = limn [ 0 f (x. Let f be a function of 2 variables x. Let fn be a sequence of measurable functions on a set E of ﬁnite measure. Then limn f (x. By Q2. 1] and suppose that for each ﬁxed t the function f is a measurable function of x. This subsequence will not have a further subsequence that converges in measure to f . lim h(tn ) = h(t). Let f be integrable over (−∞. ∞). By the Monotone Convergence Theorem. Thus f = lim fnkj . By Q20. 0 ∂t 4. Fatou’s Lemma: Let fn be a sequence of nonnegative measurable functions that converges in measure to f on E. limt→0 f (x. Thus fnk converges to f in measure. fnk converges in measure to f on E so it in turn has a subsequence fnkj that converges to f a. Then m{x : fnk (x) − f (x) ≥ ε} < ε for k ≥ M . Similarly for f − . t + sn ) − f (x. 17a. 1] × [0. Let f be a function deﬁned and bounded in the square Q = [0. t) is continuous in t for each x and let h(t) = f (x. For each ﬁxed t. Then f + and f − are nonnegative integrable functions. There exists an increasing sequence ϕn of nonnegative simple functions such that f + = lim ϕn .
Then 0 < f (x) − f (y) ≤ ε(y − x). b) so f (x) + εx is nondecreasing on [a. Let Ev = {x : fnv+1 (x) − fnv (x) ≥ 1/2v } and let Fk = v≥k Ev . In particular. b]. g has no local maximum in (a. b] and assumes a local maximum at c ∈ (a. Then fn − f  = x≥N fn − f  + A∩[−N.N ] fn − f  < ε/3 + ε/3 + 2N δ < ε for all n ≥ N . is everywhere nonnegative on (a. By Proposition 14. b].e. −k fn − f  converges to fn − f . Let A = {x : fn (x) − f (x) ≥ δ}. Suppose f is continuous on [a. Now {x : fn (x) − f (x) ≥ ε} ⊂ {x : fn (x) − fnv (x) ≥ ε/2} ∪ {x : fnv (x) − f (x) ≥ ε/2} for all n. D+ (f + g)(x) = limh→0+ (f +g)(x+h)−(f +g)(x) = limh→0+ ( f (x+h)−f (x) + h h limh→0+ f (x+h)−f (x) + limh→0+ g(x+h)−g(x) = D+ f (x) + D+ g(x). Thus g is decreasing on (a. Thus the series (fnv+1 − fnv ) converges a. r ≥ N and m{x : fnv (x) − f (x) ≥ ε/2} < ε/2 for all v ≥ N . Thus there exists N such that x≥N fn − f  < ε/3. Thus h D+ f (c) = limh→0+ f (c+h)−f (c) ≤ 0. b). Then fnv → f in measure since the partial sums of the series are of the form fnv − fn1 . to a function g. y ∈ [a. 24. A fn − f  < ε/3.in measure. D+ f (0) = D− f (0) = −1. If f has a local maximum at b. Let ε > 0 be given. fn → f in measure. b) follow from the previous cases. If x ∈ k Fk . b) follows from a similar argument and the cases where D+ f or D− f is everywhere nonnegative on (a. b] and one of its derivates. Then f (c)−f (c−h) > 0 for 0 < h < δ . b). (*) Note error in book. h h 22 g(x+h)−g(x) ) h ≤ .1 Diﬀerentiation and Integration Diﬀerentiation of monotone functions 1.e. Now there exists δ > 0 such that f (c + h) < f (c) for 0 < h < δ. The case where D− f is everywhere nonnegative on (a. Then f (x) + εx ≤ f (y) + εy. b) by Q3. Contradiction. then so does any subsequence fnk . Contradiction. Thus m{x : fn (x) − f (x) ≥ ε} < ε for all n ≥ N . then x ∈ Fk for / / v v−1 some k so fnv+1 (x) − fnv (x) < 1/2 for all v ≥ k and fnw (x) − fnv (x) < 1/2 for w ≥ v ≥ k. Hence g is nondecreasing on [a. h h 3a.e. Now for any ε > 0. 5a.e. there exists δ > 0 such that f (c) > f (c − h) for 0 < h h < δ . Then f (c+h)−f (c) < 0 for 0 < h < δ. For any x. Suppose there exist x. 2a. Hence h h D+ f (c) ≤ D+ f (c) ≤ 0 ≤ D− f (c) ≤ D− f (c). then fnkj converges to f in measure so by Q22. Let fn be a Cauchy sequence in measure. Since D+ g(x) > 0 for all x ∈ (a. Then D+ g(x) = limh→0+ g(x+h)−g(x) = limh→0+ f (−x−h)−f (−x) = limh→0+ − h h f (−x)−f (−x−h) = −limh→0+ f (−x)−f (−x−h) = −D− f (−x). b). Suppose f (x) > f (y). fn converges to f in measure. i. Similarly. Then m( k Fk ) ≤ m( v≥k Ev ) ≤ 1/2k−1 for all k so m( k Fk ) = 0. D+ [−f (x)] = limh→0+ [−f (x+h)]−[−f (x)] = −limh→0+ f (x+h)−f (x) = −D+ f (x). Let g(x) = f (−x). D− f (0) = 1. Thus D− f (c) = limh→0+ f (c)−f (c−h) ≥ 0. If f has a local maximum at a. Let x < y. Then we may choose nv+1 > nv such that m{x : fnv+1 (x) − fnv (x) ≥ 1/2v } < 1/2v . By Lebesgue’s Dominated Convergence Theorem. b]. we have f (x) − f (y) ≤ (f (x) − f (y))/2. Then D+ f (0) = limh→0+ f (0+h)−f (0) = h limh→0+ sin(1/h) = 1. fn − f  → 0. choosing ε = (f (x) − f (y))/(2(y − x)). v ≥ N . Let f = g + fn1 . h h 2b.N ] fn − f  + Ac ∩[−N. b] with x < y and g(x) > g(y).. Given ε > 0. for each n. b). say D+ f . There also exists N such that m{x : fn (x) − f (x) ≥ δ} < δ for all n ≥ N .k] converges to fn − f . Similarly. then D+ f (a) ≤ D+ f (a) ≤ 0. there exists δ > 0 such that for any set A with mA < δ. given ε > 0. y). 25. Conversely. k 5 5. D+ (f (x) + εx) ≥ ε on (a. y] and D+ g(c) ≤ 0 for all c ∈ (a. Let f be deﬁned by f (0) = 0 and f (x) = x sin(1/x) for x = 0. choose N such that m{x : fn (x) − fr (x) ≥ ε/2} < ε/2 for all n. Suppose that fn → f in measure and that there is an integrable function g such that fn  ≤ g for all n. if every subsequence fnk has in turn a subsequence fnkj that converges to f a. 4. then 0 ≤ D− f (b) ≤ D− f (b). 3b. We may assume δ < ε/6N . First consider a function g such that D+ g(x) ≥ ε > 0 for all x ∈ (a. Thus fnk has in turn a subsequence that converges to f a. i. Suppose f is continuous on [a. Now fn − f  is integrable for each n and fn − f χ[−k.e. Hence f is nondecreasing on [a.
5b. For any x, D+ (f + g)(x) = limh→0+ (f +g)(x+h)−(f +g)(x) = limh→0+ ( f (x+h)−f (x) + h h limh→0+ f (x+h)−f (x) + limh→0+ g(x+h)−g(x) = D+ f (x) + D+ g(x). h h Similarly, D− (f + g) ≤ D− f + D− g and D− (f + g) ≤ D− f + D− g.
g(x+h)−g(x) ) h
≤
5c. Let f and g be nonnegative and continuous at c. Then D+ (f g)(c) = limh→0+ (f g)(c+h)−(f g)(c) = h limh→0+ f (c+h)g(c+h)−f (c)g(c) = limh→0+ (f (c+h)−f (c))g(c+h)+f (c)(g(c+h)−g(c)) ≤ g(c)limh→0+ f (c+h)−f (c) + h h h f (c)limh→0+ g(c+h)−g(c) = f (c)D+ g(c) + g(c)D+ f (c). h *6a. Let f be deﬁned on [a, b] and g a continuous function on [α, β] that is diﬀerentiable at γ with g(γ) = c ∈ (a, b). Suppose g (γ) > 0. Note that if D+ (f ◦ g)(γ) = ±∞, then D+ f (c) = ±∞. Now ε suppose D+ f (c) < ∞. Let ε > 0 be given and let ε1 = min(1, g (γ)+1+D+ f (c) ). There exists δ1 > 0 such that  g(γ+h)−g(γ) − g (γ) < ε1 for 0 < h < δ1 . There exists δ2 > 0 such that f (c+hh)−f (c) − D+ f (c) < ε1 h for 0 < h < δ2 so that f (c+h )−f (c)−h D+ f (c) < ε1 h . By continuity of g, there exists δ3 > 0 such that g(γ+h )−g(γ) < δ2 for 0 < h < δ3 . Now let δ = min(δ1 , δ3 ). When 0 < h < δ,  g(γ+h)−g(γ) −g (γ) < ε1 h and f (g(γ + h)) − f (g(γ)) − (g(γ + h) − g(γ))D+ f (c) < ε1 (g(γ + h) − g(γ)). Hence f (g(γ+h))−f (g(γ)) − h D+ f (c)g (γ) = f (g(γ+h))−f (g(γ))−D f (c)(g(γ+h)−g(γ)) + D f (c)(g(γ+h)−g(γ))−hD f (c)g (γ) < ε1 g(γ+h)−g(γ) + h h h D+ f (c)( g(γ+h)−g(γ) − g (γ)) < ε1 (g (γ) + 1) + ε1 D+ f (c) < ε. Thus D+ (f ◦ g)(γ) ≤ D+ f (c)g (γ) and h similarly, it can be shown that D+ (f ◦ g)(γ) ≥ D+ f (c)g (γ). Hence D+ (f ◦ g)(γ) = D+ f (c)g (γ). *6b. Suppose g (γ) < 0. Note that if D+ (f ◦ g)(γ) = ±∞, then D− f (c)g (γ) = ∞. Also note that there exists δ > 0 such that g(γ + h) − g(γ) < 0 for 0 < h < δ. By a similar argument to that in part (a), D+ (f ◦ g)(γ) = D− f (c)g (γ). *6c. Suppose g (γ) = 0 and all the derivates of f at c are ﬁnite. By a similar argument to that in part (a), D+ (f ◦ g)(γ) = 0.
+ + +
5.2
Functions of bounded variation
7a. Let f be of bounded variation on [a, b]. Then f = g − h where g and h are monotone increasing functions on [a, b]. Let c ∈ (a, b). Also let A = supx∈[a,c) g(x) and let B = supx∈[a,c) h(x). Note that A, B < ∞. Given ε > 0, there exists δ > 0 such that A − ε/2 < g(c − δ) ≤ A and B − ε/2 < h(c − δ) ≤ B. Then for x ∈ (c − δ, c), A − ε/2 < g(x) ≤ A and B − ε/2 < h(x) ≤ B. i.e. 0 ≤ A − g(x) < ε/2 and 0 ≤ B − h(x) < ε/2. Now 0 ≤ A − B − f (x) ≤ (A − g(x)) + (B − h(x)) < ε for x ∈ (c − δ, c). Hence f (c−) exists. Similarly f (c+) exists. Let g be a monotone function and let E be the set of discontinuities of g. Now for c ∈ E, g(c−) < g(c+) so there is a rational rc such that g(c−) < rc < g(c+). Note that if x1 < x2 , then g(x1 +) ≤ g(x2 −) so rx1 = rx2 . Thus we have a bijection between E and a subset of Q so E is countable. Since a function f of bounded variation is a diﬀerence of two monotone functions, f also has only a countable number of discontinuities. −n 7b. Let xn be an enumeration of Q ∩ [0, 1]. Deﬁne f on [0, 1] by f (x) = . Then f is xn <x 2 monotone. Also, at each xn , for any δ > 0, there exists x ∈ (xn , xn + δ) such that f (x) − f (xn ) > 2−n−1 so f is discontinuous at each xn . 8a. Suppose a ≤ c ≤ b. Let a = x0 < x1 < · · · < xn = b be a subdivision of [a, b]. If c = xk for n k n c b some k, then 1 f (xi ) − f (xi−1 ) = 1 f (xi ) − f (xi−1 ) + k+1 f (xi ) − f (xi−1 ) ≤ Ta (f ) + Tc (f ). b c b Thus Ta (f ) ≤ Ta (f ) + Tc (f ). The case where c ∈ (xk , xk+1 ) for some k is similar. Conversely, let a = x0 < x1 < · · · < xn = c be a subdivision of [a, c] and let c = y0 < y1 < · · · < ym = b be a subdivision of [c, b]. Then a = x0 < x1 < · · · < xn = c < y1 < · · · < ym = b is a subdivision of [a, b] and n m b c b b 1 f (xi ) − f (xi−1 ) + 1 f (yi ) − f (yi−1 ) ≤ Ta (f ). It follows that Ta (f ) + Tc (f ) ≤ Ta (f ). Hence b c b c b Ta (f ) = Ta (f ) + Tc (f ) and Ta (f ) ≤ Ta (f ). n 8b. Let a = x0 < x1 < · · · < xn = b be a subdivision of [a, b]. Then 1 (f + g)(xi ) − (f + g)(xi−1 ) ≤ n n b b b b b 1 f (xi )−f (xi−1 )+ 1 g(xi )−g(xi−1 ) ≤ Ta (f )+Ta (g). Hence Ta (f +g) ≤ Ta (f )+Ta (g). Let c ∈ R. n n b b If c = 0, then Ta (cf ) = 0 = cTa (f ). If c = 0, then 1 cf (xi ) − cf (xi−1 ) = c 1 f (xi ) − f (xi−1 ) ≤ n n b b b −1 cTa (f ). Thus Ta (cf ) ≤ cTa (f ). On the other hand, 1 f (xi ) − f (xi−1 ) = c 1 cf (xi ) − b b b b b b b cf (xi−1 ) ≤ c−1 Ta (cf ). Thus Ta (f ) ≤ c−1 Ta (cf ) so cTa (f ) ≤ Ta (cf ). Hence Ta (cf ) = cTa (f ). 9. Let fn be a sequence of functions on [a, b] that converges at each point of [a, b] to f . Let a = x0 < x1 < · · · < xn = b be a subdivision of [a, b] and let ε > 0. Then there exists N such that 23
b f (xi )−f (xi−1 ) ≤ 1 f (xi )−fn (xi )+ 1 f (xi−1 )−fn (xi−1 )+ 1 fn (xi )−fn (xi−1 ) < ε+Ta (fn ) b b b b for n ≥ N . Thus Ta (f ) ≤ ε + Ta (fn ) for n ≥ N so Ta (f ) ≤ ε + limTa (fn ). Since ε is arbitrary, b b Ta (f ) ≤ limTa (fn ). 10a. Let f be deﬁned by f (0) = 0 and f (x) = x2 sin(1/x2 ) for x = 0. Consider the subdivision −1 < 2/nπ < 2/(n − 1)π < · · · < 2/π < 1 of [−1, 1]. Note that tb (f ) → ∞ as n → ∞. Thus f is a not of bounded variation on [−1, 1]. 10b. Let g be deﬁned by g(0) = 0 and g(x) = x2 sin(1/x) for x = 0. Note that g is diﬀerentiable on [−1, 1] and g (x) ≤ 3 on [−1, 1]. Thus for any subdivision a = x0 < x1 < · · · < xn = b of [a, b], n n 1 1 g(xi − g(xi−1 ) ≤ 4 1 xi − xi−1  = 3(g(1) − g(−1)). Hence T−1 (g) ≤ 3(g(−1) − g(1)) < ∞. x x 11. Let f be of bounded variation on [a, b]. For any x ∈ [a, b], f (x) = Pa (f ) − Na (f ) − f (a) so d d d d d x x x x x f (x) = dx Pa (f ) − dx Na (f ) a.e. in [a, b]. Thus f  ≤ dx Pa (f ) + dx Na (f ) = dx Ta (f ) a.e. in [a, b] and b b d x b a b f  ≤ a dx Ta (f ) ≤ Ta (f ) − Ta (f ) = Ta (f ). a
n 1
n
n
n
5.3
Diﬀerentiation of an integral
No problems
5.4
Absolute continuity
xi b n n n xi 1 f (xi ) − f (xi−1 ) = 1  xi−1 f  ≤ 1 xi−1 f  = a f . Thus b b b b Ta (f ) ≤ a f . Hence Ta (f ) = a f . d d x x x x For any x ∈ [a, b], f (x) = Pa (f ) − Na (f ) − f (a) so f (x) = dx Pa (f ) − dx Na (f ) a.e. in [a, b]. Thus b b d d d d x x x x b a b (f )+ ≤ ( dx Pa (f ))+ + ( dx Na (f ))− = dx Pa (f ). Thus a (f )+ ≤ a dx Pa (f ) ≤ Pa (f ) − Pa (f ) = Pa (f ). n + Conversely, for any subdivision a = x0 < x1 < · · · < xn = b of [a, b], = 1 (f (xi ) − f (xi−1 )) xi b b b n n xi + b + b + + + 1 ( xi−1 f ) ≤ 1 xi−1 (f ) = a (f ) . Thus Pa (f ) ≤ a (f ) . Hence Pa (f ) = a (f ) . + + + +
*12. The function f deﬁned by f (0) = 0 and f (x) = x2 sin(1/x2 ) for x = 0 is absolutely continuous on [ε, 1] for ε > 0, continuous at 0 but not of bounded variation on [0, 1], thus not absolutely continuous on [0, 1]. Suppose f is absolutely continuous on [η, 1] for η > 0, continuous at 0 and of bounded variation on [0, 1]. For η ∈ (0, 1], let 0 = x0 < x1 < · · · < xn = η be a subdivision of [0, η]. Then since f is continuous at 0, n η + + 1 f (xi )−f (xi−1 ) → 0 as η → 0 . Thus T0 (f ) → 0 as η → 0 . Given ε > 0, there exists η ∈ (0, 1] such η that T0 (f ) < ε/2. Since f is absolutely continuous on [η, 1], there exists δ > 0 such that for any ﬁnite n n collection {(xi , xi )}n of disjoint intervals in [η, 1] with 1 xi −xi  < δ, we have 1 f (xi )−f (xi ) < ε/2. 1 n n Now let {(yi , yi )}1 be a ﬁnite collection of disjoint intervals in [0, 1] with 1 yi −yi  < δ. If η ∈ [yk , yk+1 ] n k n η for some k, then 1 f (yi ) − f (yi ) ≤ 1 f (yi ) − f (yi ) + k+1 f (yi ) − f (yi ) < T0 (f ) + ε/2 < ε. If n k−1 η ∈ (yk , yk ) for some k, then 1 f (yi ) − f (yi ) ≤ 1 f (yi ) − f (yi ) + f (η) − f (yk ) + f (yk ) − f (η) + n η k+1 f (yi ) − f (yi ) < T0 (f ) + ε/2 < ε. Hence f is absolutely continuous on [0, 1]. 13. Let f be absolutely continuous on [a, b]. Then f is of bounded variation on [a, b] so by Q11, b b f  ≤ Ta (f ). Conversely, since f is absolutely continuous on [a, b], for any subdivision a = x0 < a x1 < · · · < xn = b of [a, b],
(*) If g < 0, then ( g) = 0 ≤ g . If g ≥ 0, then ( g) = g ≤ g . 14a. Let f and g be two absolutely continuous functions on [a, b]. Given ε > 0, there exists δ > 0 such n n that 1 f (xi ) − f (xi ) < ε/2 and 1 g(xi ) − g(xi ) < ε/2 for any ﬁnite collection {(xi , xi )}n of disjoint 1 n n n intervals in [a, b] with 1 xi − xi  < δ. Then 1 (f ± g)(xi ) − (f ± g)(xi ) ≤ 1 f (xi ) − f (xi ) + n 1 g(xi ) − g(xi ) < ε. Thus f + g and f − g are absolutely continuous. 14b. Let f and g be two absolutely continuous functions on [a, b]. There exists M such that f (x) ≤ M n and g(x) ≤ M for any x ∈ [a, b]. Given ε > 0, there exists δ > 0 such that 1 f (xi ) − f (xi ) < ε/2M n and 1 g(xi ) − g(xi ) < ε/2M for any ﬁnite collection {(xi , xi )} of disjoint intervals in [a, b] with n n n n xi −xi  < δ. Then 1 (f g)(xi )−(f g)(xi ) ≤ 1 f (xi )g(xi )−g(xi )+ 1 g(xi )f (xi )−f (xi ) < ε. 1 Thus f g is absolutely continuous. 14c. Suppose f is absolutely continuous on [a, b] and is never zero there. Let g = 1/f . There exists M n such that f (x) ≥ M for x ∈ [a, b]. Given ε > 0, there exists δ > 0 such that 1 f (xi ) − f (xi ) < εM 2 24
for any ﬁnite collection {(xi , xi )} of disjoint intervals in [a, b] with f (xi )−f (xi ) f (xi )f (x ) < ε. Thus g is absolutely continuous.
i
n 1
xi −xi  < δ. Then g(xi )−g(xi ) =
15. Let f be the Cantor ternary function. By Q2.48, f is continuous and monotone on [0, 1]. Note that f = 0 a.e. on [0, 1] since f is constant on each interval in the complement of the Cantor set and 1 the Cantor set has measure zero. If f is absolutely continuous, then 1 = f (1) = 0 f + f (0) = 0. Contradiction. Thus f is not absolutely continuous. x 16a. Let f be a monotone increasing function on [a, b]. Let g be deﬁned by g(x) = a f and let x d h = f − g. Then g is absolutely continuous, h (x) = f (x) − dx a f = 0 a.e. so h is singular, and f = g + h. 16b. Let f be a nondecreasing singular function on [a, b]. Let ε, δ > 0 be given. Since f is singular on [a, b], for each x ∈ [a, b], there is an arbitrarily small interval [x, x+h] ⊂ [a, b] such that f (x+h)−f (x) < εh/(b − a). Then there exists a ﬁnite collection {[xk , yk ]} of nonoverlapping intervals of this sort which cover all of [a, b] except for a set of measure less than δ. Labelling xk such that xk ≤ xk+1 , we have n n y0 = a ≤ x1 < y1 ≤ x2 < · · · ≤ yn ≤ b = xn+1 . Then 0 xk+1 − yk  < δ and 1 f (yk ) − f (xk ) < ε. n Since f is nondecreasing, 0 f (xk+1 ) − f (yk ) > f (b) − f (a) − ε. 16c. Let f be a nondecreasing function on [a, b] with property (S). i.e. Given ε, δ > 0, there is a ﬁnite n n collection {[yk , xk ]} of nonoverlapping intervals in [a, b] such that 1 xk − yk  < δ and 1 (f (xk ) − x f (yk )) > f (b) − f (a) − ε. By part (a), f = g + h where g = a f and h is singular. It suﬃces to show n that g = 0 a.e. Letting x0 = a and yn+1 = b, we have 1 (f (yk+1 ) − f (xk )) < ε. We may choose δ such b b that n [yk ,xk ] f < ε. Then a f < 2ε so a f = 0 and g = 0.
1
16d. Let fn be a sequence of nondecreasing singular functions on [a, b] such that the function f (x) = fn (x) is everywhere ﬁnite. Let ε, δ > 0 be given. Now f (b) − f (a) = (fn (b) − fn (a)) < ∞ so ∞ N there exists N such that N +1 (fn (b) − fn (a)) < ε/2. Let F (x) = 1 fn (x). Then F is nondecreasing and singular. By part (b), there exists a ﬁnite collection {[yk , xk ]} of nonoverlapping intervals such that xk −yk  < δ and (F (yk )−F (xk )) > F (b)−F (a)−ε/2. Now (f (yk )−f (xk )) ≥ (F (yk )−F (xk )) > ∞ F (b) − F (a) − ε/2 = f (b) − f (a) − N +1 (fn (b) − fn (a)) − ε/2 > f (b) − f (a) − ε. By part (c), f is singular. *16e. Let C be the Cantor ternary function on [0, 1]. Extend C to R by deﬁning C(x) = 0 for x < 0 and C(x) = 1 for x > 1. For each n, deﬁne fn by fn (x) = 2−n C( bx−an ) where {[an , bn ]} is an enumeration of n −an the intervals with rational endpoints in [0, 1]. Then each fn is a nondecreasing singular function on [0, 1]. Deﬁne f (x) = fn (x). Then f is everywhere ﬁnite, strictly increasing and by part (d), f is singular. 17a. Let F be absolutely continuous on [c, d]. Let g be monotone and absolutely continuous on [a, b] with c ≤ g ≤ d. Given ε > 0, there exists δ > 0 such that for any ﬁnite collection {(yi , yi )} of disjoint n n intervals with 1 yi − yi  < δ, we have 1 F (yi ) − F (yi ) < ε. Now there exists δ > 0 such that for n n any ﬁnite collection {(xi , xi )} of disjoint intervals with 1 xi − xi  < δ , we have 1 g(xi ) − g(xi ) < δ. n Now {(g(xi ), g(xi ))} is a ﬁnite collection of disjoint intervals so 1 F (g(xi )) − F (g(xi )) < ε. Hence F ◦ g is absolutely continuous. √ (*) Additional assumption that g is monotone. Counterexample: Consider f (x) = x for x ∈ [0, 1] and g(0) = 0, g(x) = (x sin x−1 )2 for x ∈ (0, 1]. Then (f ◦ g)(0) = 0 and (f ◦ g)(x) = x sin x−1 for x ∈ (0, 1]. f and g are absolutely continuous but not f ◦ g. x x *17b. Let E = {x : g (x) = 0}. Note that g(x) − g(a) =  a g  ≤ a g  for all x ∈ [a, b]. Let ε > 0. There exists δ > 0 such that A g  < ε/2 whenever mA < δ. Let η = ε/4(b − a). For any x ∈ E, there exists hx > 0 such that g(x + h) − g(x) < ηh for 0 < h ≤ hx . Deﬁne V = {[x, x + hx ] : x ∈ E, g(y) − g(x) < η(y − x) for y ∈ (x, x + hx ]}. Then V is a Vitali covering for E so there exists a ﬁnite N disjoint collection {I1 , . . . , IN } of intervals in V such that m(E \ n=1 In ) < δ. Now let O be an open set N such that O ⊃ E \ n=1 In and mO < δ. Then O is a countable union of disjoint open intervals Jm and N N g[E \ n=1 In ] ⊂ g[Jm ]. Thus m(g[E \ n=1 In )] ≤ m(g[Jm ]) ≤ g  = O g  < ε/2. Also, Jm 2hxn η < 2(b − a)η = ε/2. Hence m(g[E]) < ε. g[E ∩ n=1 In ] ⊂ n=1 g[In ] so m(g[E ∩ n=1 In )] ≤ Since ε is arbitrary, m(g[E]) = 0. 18. Let g be an absolutely continuous monotone function on [0, 1] and E a set of measure zero. Let ε > 0. There is an open set O ⊃ E such that mO = m(O \ E) < δ where δ is given by absolute continuity
N N N
25
Thus F = g −1 [E] ∩ H is measurable. Let f be absolutely continuous. Now for each n. b). Since g > 0 on g −1 [E] ∩ H.e. Now f (x) = limy→x limy→x  f (y)−f (x)  y−x f (y)−f (x) y−x so f (x) = ≤ M for all x. Suppose f is Lipschitz. m(g[E]) = 0. d] with mE = 0 and let δ > 0. b). For any ﬁnite collection {(xi . Then there is an increasing sequence ϕn of simple functions on [c. If ϕ is not monotone. By part (a). 20a. Also. Then c g is absolutely continuous and strictly monotone on [0. (f ◦ g)g is measurable. 22b. 21d.5 Convex functions 23a. Since each b b (ϕn ◦ g)g is measurable. b) and let f (x) = m(x − x0 ) + ϕ(x0 ) be the equation of a supporting line at x0 . Now a ϕn (g(x))g (x) dx = c χ (g(x))g (x) dx = a k Ek ck mEk = d d c ϕn (y) dy. There exists a Gδ set G ⊃ E with m(G \ E) = 0. Since δ > 0 is arbitrary. except on E = {x : g (x) = 0}. we have D+ F (g(x)) = D+ F (g(x)) = D− F (g(x)) = D− F (g(x)) = H (x)/g (x) by Q6a so F (g(x)) exists. so H (x) = F (g(x))g (x) a. *20c. g−1 [E]∩H g = 0. y) by the Mean Value Theorem. we have ϕ(x) ≥ f (x) ≥ min(f (a). g −1 [g[F ]] = F is nonmeasurable. m(g[E]) = 0 so / f0 = f a. By the Monotone Convergence Theorem. if f is not Lipschitz. Let G be the complement of a generalised Cantor set of positive measure and let g = 0 χG . Now G is a countable intersection On of open sets. Now g[E] ⊂ g[In ∩ [0. xi )} of nonoverlapping intervals with 1 xi − xi  < δ. d]. there exist x and y such that f (x) − f (y) > M x − y. We may assume G ⊂ [c. *22d. Now O is a countable union In of disjoint open intervals so l(In ) < δ and l(g[In ∩[0. mO = l(In ) = (g(dn ) − g(cn )) = cn 21b. Given n ε > 0. 1]] so m(g[E]) < ε.e. 21a. b) 26 . d]. then for any M . g is monotone and absolutely continuous on [a. 19b. By part (b). Since ϕ is continuous at a. lim ϕn = b b f . Let ϕ be convex on a ﬁnite interval [a. *22c. Conversely. Let E ⊂ [c. dn ∈ [c. there exists c such that f (c) > M so f  is unbounded. Thus g−1 [E]∩H g = g−1 [E] g < δ. Thus d c f (y) dy = lim c ϕn (y) dy = lim a ϕn (g(x))g (x) dx = a f (g(x))g (x) dx. m(g[F ]) = 0. Also. dn ). it is continuous and thus measurable so g −1 [E] is measurable. b). g−1 [O] g < δ. Also. Since g is absolutely continuous. 22a. By Q17b. b] with c ≤ g ≤ d. There exists M such that f (x) − f (y) ≤ M x − y for all x. g −1 [On ]∩H n=1 k F g = g −1 [E] g = b a χE (g(x))g (x) dx. d]. b). Then G1 ⊃ G2 ⊃ · · · so lim mGk = m( Gk ) = mG. Let Gk = n=1 On . d] with lim ϕn = f so lim ϕn (g(x))g (x) = f (g(x))g (x). m((g −1 [G] \ g −1 [E]) ∩ H) = 0 so g−1 [G]∩H g = g−1 [E]∩H g . g is measurable so H is measurable. Suppose ϕ is convex on (a. Hence ϕ is bounded from below. Now H and g exist a. Then ϕ(x) ≥ f (x) for all x ∈ (a. Then there exists an open set O ⊃ E with mO < δ. Thus for any M . By Q17(a). g −1 [O] = g −1 [In ] = (cn . If ϕ is monotone on (a. n n we have 1 f (xi ) − f (xi ) ≤ M 1 xi − xi  < ε. Whenever H and g exist with g (x) = 0. 1]]) < ε. In = (g(cn ). Let O be an open set in [c. d]. b). Let H = {x : g (x) = 0}.e. 23b. Thus f is absolutely continuous. H = F ◦ g is absolutely continuous. Also. f (b)) for all x ∈ [a. Suppose f is Lipschitz. Then O is a countable union In of disjoint open intervals.of g. 5. g = χG = 0 on G .e. b). let δ = ε/M . Thus dn g = g−1 [O] g . the set g −1 [E] ∩ H has measure zero. d]. F is absolutely continuous on [c. Now mE = mG = lim mGk = lim g−1 [Gk ]∩H g = lim k g−1 [On ]∩H g = g = g−1 [G]∩H g = g−1 [E]∩H g = F g . it has a nonmeasurable subset F . Let f be a nonnegative measurable function on [c. Let x0 ∈ (a. g(dn )) for some cn . Since {x : g (x) = 0} has positive measure. Hence H (x) = f (g(x))g (x) = f0 (g(x))g (x) a. 1]. Also. then ϕ(x) has limits (possibly inﬁnite) as it approaches a and b respectively from within (a. Then f (c) > M for some c ∈ (x. By Q17(b). Let E be a measurable subset of [c. then there exists c ∈ (a. Since ε is arbitrary. Let f0 be deﬁned by f0 (y) = f (y) if y ∈ g[E] and f0 (y) = 0 if y ∈ g[E]. x 19a. d] and let F = g −1 [E] ∩ H. 20b. y. 21c.
Thus ϕ is monotone on (a. Then 1 − λ = β/(α + β) and f + gp = (f + g)p = (αf0 + βg0 )p = (α + β)p (λf0 + p p (1 − λ)g0 )p ≥ (α + β)p (λf0 + (1 − λ)g0 ) by concavity of the function ϕ(t) = tp for 0 < p < 1. Since exp(f (t)) − g(f (t)) ≥ 0. Hence. 4. Now exp(f (t)) dt − exp(α) = exp(f (t)) dt − g(α) = exp(f (t)) dt − g( f (t) dt) = (exp(f (t)) − g(f (t))) dt. For 1 < p < ∞. Hence ϕ is strictly convex for p > 1. For p = 1. If a (or b) is ﬁnite. Suppose a ≥ 0 and b > 0. If mE = 0. 28. Then f g ≤ f g∞ = g∞ f  = f 1 g∞ . Let ϕ be continuous on an interval I (open. b). halfopen) and convex in the interior of I. 1 1 1 1 1 1 6. Then ϕ(tx + (1 − t)y) ≤ tϕ(x) + (1 − t)ϕ(y) for all x. − log(g(t)) dt ≥ − log( g(t) dt) by Jensen’s inequality. Since ϕ is continuous. ϕ(λx+(1−λ)y)−ϕ(x) < ϕ(y)−ϕ(λ(x)+(1−λ)(y)) . k k α k k α k 6 6.e.e. Similarly.and righthand derivatives are monotone increasing on (a. the integral is zero only when exp(f (t)) − g(f (t)) = 0 a. the inequality also holds at the included endpoints. For 0 < p < 1. b) so ϕ is monotone increasing on (a.e. Let ϕ(t) = (a + bt)p . ∞) for all p. and g(t) ≤ M2 a.2 The Minkowski and H¨lder inequalities o 5a. c] and on [c. 27. and this can happen only when f (t) = α a. Let ε > 0 and let E = {x ∈ [0. ϕ(y)−ϕ(λ(x)+(1−λ)(y)) = ϕ (ξ2 ) for some (1−λ)(y−x) λ(y−x) ξ2 ∈ (λx + (1 − λ)y.e. Thus n=1 ξn n ≤ 0 n=1 n exp(f (t)) dt = n=1 αn ξn .. Deﬁne f on [0. limp→∞ f p ≥ f ∞ . ∞ ∞ α Letting k → ∞. Thus f + g∞ ≤ f ∞ + g∞ . Then f p = 1 ( 0 f p )1/p ≥ ( E f p )1/p ≥ (f ∞ − ε)(mE)1/p . Now f p = ( 0 f p )1/p ≤ ( 0 f p )1/p = f ∞ .e. ∞) so ϕ is strictly increasing on (0. Note that f (t) ≤ f ∞ a. Let ϕ have a second derivative at each point of (a. 25a. by Q24. Then ϕ is continuous on [0.such that D+ ϕ(x) ≤ 0 on (a. Also. Equivalently. b) since the righthand derivative of ϕ is increasing on (a. *26. Thus mE > 0 and limp→∞ f p ≥ f ∞ − ε. b). Now for x < y. 23c. the limits at a (or b) cannot be −∞. b). Conversely. 1]. For each k. 1] by f (x) = log ξn if x ∈ [ n=1 αn . c] and D+ ϕ(x) ≥ 0 on [c. Hence log( g(t) dt) ≥ log(g(t)) dt. ϕ(λx+(1−λ)y)−ϕ(x) = ϕ (ξ1 ) for some ξ1 ∈ (x. b). n=1 ξn n = α exp( n=1 αn log ξn ) = exp( 0 n=1 n f (t) dt). ϕ is convex for 1 ≤ p < ∞ and concave for 0 < p ≤ 1. ϕ (t) = b2 p(p − 1)(a + bt)p−2 > 0. then f (t) + g(t) ≤ M1 + M2 a. We may assume f p > 0 and gp > 0.e. Equality holds when exp(f (t)) dt = exp(α). then by part (a). (1−λ)(y−x) λ(y−x) ϕ(λx + (1 − λ)y) < λϕ(x) + (1 − λ)ϕ(y). b). y). If f (t) ≤ M1 a. ϕ is continuous on (a. For p > 1. y in the interior of I and all t ∈ [0.e. closed. Suppose f ∈ L1 and g ∈ L∞ . Let f and g be two nonnegative functions in Lp with 0 < p < 1. 3. −ϕ (t) = −b2 p(p − 1)(a + bt)p−2 > 0. then its left. 1]. f + g1 = 0 f + g ≤ 0 (f  + g) = 0 f  + 0 g = f 1 + g1 . 25b. then ϕ is increasing on (a. Let f be a bounded measurable function on [0. ϕ(t) = a + bt so ϕ (t) = 0. b). Since ϕ (ξ1 ) < ϕ (ξ2 ). 1]. n=1 αn ). b). ϕ (t) > 0 for all t ∈ (0. Also. 1] : f (x) > f ∞ − ε}. λx + (1 − λ)y). so f + g∞ ≤ M1 + M2 . If ϕ (x) ≥ 0 for all x ∈ (a. if ϕ is convex on (a. b) and it follows that the righthand and lefthand limits exist at a and b respectively. Let αn be a sequence of nonnegative numbers whose sum is 1 and let ξn be a sequence of positive k−1 k k α numbers. 24.1 The Classical Banach Spaces The Lp spaces 1. Hence limp→0 f p = f ∞ . and g(t) ≤ g∞ a. Thus 27 . Let α = f (t) dt and let g(x) = m(x − α) + exp(α) be the equation of a supporting line at α. we have n=1 ξn n ≤ n=1 αn ξn . then f ∞ ≤ f ∞ − ε. ∞ Thus limp→∞ f p ≤ f ∞ . b). Contradiction. Since ε > 0 is arbitrary. Since log is concave. ∞). b). Hence ϕ (x) ≥ 0 for all x ∈ (a. ϕ is strictly concave for 0 < p < 1. Hence ϕ is convex on (a. Set λ = α/(α + β). Let g be a nonnegative measurable function on [0. 2. Let α = f p and β = gp so f = αf0 and g = βg0 where f0 p = g0 p = 1.
Let fn be a convergent sequence in Lp . b be nonnegative. 7a. let α =  ξv p and β =  ηv p so ξv = α ξv and ηv = β ηv where ξv p = ηv p = 1. Suppose 0 < p < 1.e. Let E = {t : fn (t) − f (t) ≥ ε}. p−1 p−1 p For 1 < p < ∞. Given ε > 0. Thus fn is a Cauchy sequence. the yaxis. Then fn converges a. Thus m{t : fn (t)−f (t) ≥ ε} = 0 for n ≥ N . Given ε > 0. Let q be such that 1/p + 1/q = 1. fn − fm p ≤ fn − f p + fm − f p < ε. q = ∞. x = a and y = b is ab. Consider the graph of the function f (x) = xp−1 . 8d. there exists N such that inf{M : m{t : fn (t) − fm (t) > M } = 0} = fn − fm ∞ < ε/2 for n. g)p = 2p  max(f. q p q Hence f g ≤ f p gq . For 1 ≤ p ≤ ∞.3 Convergence and completeness 9. In particular. 11. g)p ≤ 2p (f p + gp ).  ξv + ηv ∞ = sup ξv + ηv  ≤ sup(ξv  + ηv ) ≤ sup ξv  + sup ηv  =  ξv ∞ +  ηv ∞ . f g ≥ f p gq . Let p = 1/p and q = −q/p. f p gq ≤ pfp + qgq = p + 1 = 1. Let u = (f g)p and let v = g −p . Thus p p p p p ptξv ηv  ≤ (αv + tξv ) − αv =  αv + tξv  p −  αv p ≤ ( αv p + t ξv p ) −  αv p . q > 1 and 1/p + 1/q = 1. there exists M < ε/2 such that m{t : fn (t) − fm (t) > M } = 0 so m{t : fn (t) − fm (t) > ε/2} = 0. 8a. there exists N such that (n) (m) p (n) (m) p ξv − ξv  < ε for n. uv ≤ up vq . We may assume that f p > 0 and gq > 0. Hence f +gp ≥ f p +gp .e. f + g ∈ Lp by the Minkowski inequality. Set λ = α/(α + β). Suppose 0 < p < 1 and 1/p + 1/q = 1.e. Let 1 ≤ p < ∞ and let ξv be a Cauchy sequence in p . the xaxis and x = a is 0 xp−1 dx = a .e. fn − f ∞ → 0. The area of the rectangle bounded by the xaxis. There exists f ∈ Lp such that for any ε > 0. b 0 y 1/(p−1) dy = bp/(p−q) p/(p−q) = bq q . By the H¨lder inequality. The area of the p a region bounded by f (x). Thus f + g ∈ Lp . Equality holds in part (a) if and only if b = ap−1 . 1 < p < ∞. for n ≥ N so f  ≤ fN  + ε/2 a. p p−1 Diﬀerentiating with respect to t at t = 0. the yaxis and y = b is ab ≤ ap p + bq q . ξv ηv  ≤ sup ηv  ξv  =  ξv 1  ηv ∞ . Hence f g ≥ ( f  ) ( g ) = f p gq . For p = 1. Suppose fn −f ∞ → 0. there exists N such that fn (t) − f (t) < ε/2 for n ≥ N and t ∈ E c . 10. f + gp ≤ 2 max(f. Thus for (n) (n) k each v. we get p ξv ηv  ≤ p ξv p  αv p = p ξv p  ηv q . For 0 < p < 1. ξv − ξv p < εp for n. suppose there exists a set E with mE = 0 and fn converges uniformly to f on E c . there exists N such that fn − f p < ε/2 for n ≥ N . 7b. Note that q = pp /(1 − p ).e. Suppose f ∈ Lp and g ∈ Lp . Equivalently. Then 1 − λ = β/(α + β) and  ξv + ηv p = ( ξv + ηv p )1/p ≤ ( (ξv +ηv )p )1/p = ( (αξv +βηv )p )1/p = ( (α+β)p [λξv +(1−λ)ηv ]p )1/p ≤ ( (α+β)p [λξv p + (1 − λ)ηv p ])1/p = α + β =  ξv p +  ηv p . 1/p + 1/q = 1. Then ηv  = αv and ptξv ηv  = ptξv αv ≤ (αv + tξv )p − αv . For p = ∞. By a similar argument as part (b). Then mE = 0 and fn converges uniformly to f on E c . By part (a). 12. Then f g = up . Then v=1 ξv − ξv p < εp for (n) (n) each k and each n ≥ N so ξv − ξv p < εp for n ≥ N . For 1 ≤ p < ∞. q p p 8c. Let fn be a sequence of functions in L∞ . By comparing these areas. to a function f and fn − f  < ε/2 a. there exists N such that inf{M : m{t : fn (t)−f (t) > M } = 0} < ε for n ≥ N . ξv is Cauchy in R so it converges to some ξv . i. Then p > 1. Hence inf{M : m{t : fn (t) − f (t) > M } = 0} < ε for n ≥ N . Thus ξv − ξv ∈ p for n ≥ N so ξv ∈ p 28 (n) . Conversely.f +gp ≥ (α+β)p (λf0 p +(1−λ)g0 p ) = (α+β)p = (f p +gp )p . m ≥ N . m ≥ N and each v. Given ε > 0. Furthermore. Thus (ab)p b−p ≤ (ab)pp p + b−pq q = pab − pbq q so ap ≤ pab − pbq q and ab ≥ ap p + bq q . m ≥ N . Let a. Consider ξv . i. Thus {t : fn (t) − f (t) > ε/2} ⊂ E for n ≥ N . Now for n. we see that p q g f f  g 1 8b. p p p 5b. 6. m ≥ N . The area of the region bounded by p f (x). o f p ≤ pp 1/p pp /(1−p ) (p −1)/p p q 1−p p 1/p q 1/q ( f g ) ( g ) = ( f g) ( g ) . Thus equality holds here if and only if f p−1 g = gq f p−1 . i. Let fn be a Cauchy sequence in L∞ . p p p p p *6. f p = uv and g q = v p /(p −1) . let αv = ηv q/p . Let p = 1/p so p > 1. Let q be such that 1/p + 1/q = 1. fn − f ∞ → 0. Hence ξv ηv  ≤  ξv p  ηv q . Given ε > 0. f p gq = gq f p . Thus for n. and f ∈ L∞ . m ≥ N . inf{M : m{t : fn (t) − f (t) > M } = 0} < ε for n ≥ N .e.
4 Approximation in Lp ξk+1 ξk+1 m−1 m−1 1 1 f )χ[ξk . 1 ≤ p < ∞. the convergence is uniform. m ≥ N . fg = lim fn g. Thus m{x : ϕ∆ (x) − f (x) > ε} = m{x : ϕ∆ (x) − f (x)p > εp } ≤ ε−p ϕ∆ − f p < ε for suﬃciently small δ. 1 ≤ p < ∞ and let gn be a sequence of measurable functions such that gn  ≤ M for all n and gn → g a. It is straightforward to check that  ·  is a norm on C. i. suppose fn p → f p . to a function f in Lp . 2p+1 f p ≤ lim 2p (fn p + f p ) − fn − f p = 2p+1 f p − lim fn − f p . Then ξv − ξv  < ε for (n) (n) each v and n. For p = 1. Also. Thus ξv (n) (n) (n) is Cauchy in R. 17. Now p p k=0 ξk+1 −ξk ( ξk k=0  ξk+1 −ξk ( ξk b ξk+1 ξk+1 ξk+1 ξk+1 1 1 1 p p p  ξk+1 −ξk ( ξk f )χ[ξk . Conversely. p *19. Let fn → f in Lp .e.e. Thus ξv converges to some ξv for each v and ξv − ξv  < ε for n ≥ N . Let ξv be a (n) (m) Cauchy sequence in c. By Egoroﬀ’s Theorem. Now gn fn − gf p ≤ gn fn − gn f p + gn f − gf p = ( gn p fn − f p )1/p + ( gn − gp f p )1/p ≤ M fn − f p + ( E gn − gp f p )1/p + ( E c gn − gp f p )1/p < ε/2 + (ε/8M )(2M ) + ε/(4(mE c )1/p f p )(mE c )1/p f p = ε for n ≥ N . Let fn be a Cauchy sequence in C. 1 < p < ∞. Let C = C[0. By Egoroﬀ’s Theorem. Thus f ∈ C. there exists N such that max fn (x) − fm (x) < ε for n. 18. there exists δ > 0 such that gq < (ε/4M )q whenever mE < δ.  ξv − ξv ∞ → 0. Now 2p (fn p + f p ) − fn − f p ≥ 0 for each n so by Fatou’s Lemma. (n) (n) 6. ϕ∆ − f p ≥ εp m{x : ϕ∆ (x) − f (x)p > εp }.e. Thus there exists N such that fn − f p < ε/2M and gn (x) − g(x) < ε/(4(mE c )1/p f p ) for n ≥ N and x ∈ E c . If ξv is a Cauchy (n) (n) sequence in c0 . By the ξ ξ ξ ξ ξ H¨lder inequality. then ξv converges to 0 since ξv  ≤ ξv − ξv  + ξv . fn − f  → 0.1/n] for each n.1] ∈ L∞ .e. Let c be the space of all convergent sequences of real numbers and let c0 be the space of all sequences (n) which converge to 0.  ξkk+1 f p ≤ ξkk+1 f p ( ξkk+1 1)p/q = ξkk+1 f p ( ξkk+1 1)p−1 . Given ε > 0. to a function f in Lp . i. 15. Given ε > 0. max fn (x) − f (x) < ε for n ≥ N . m ≥ N . T∆ (f )p =  p 29 . for any ε > 0. Thus there exists N such that fn (x) − f (x) < ε/(2(mE c )1/p gq ) for n ≥ N and x ∈ E c . Hence fn − f p → 0. Then (N ) (n) (n) ξv  ≤ ξv  + ε for each v and ξv ∈ ∞ . 1] and deﬁne f  = max f (x) for f ∈ C. 16. p p *20. Thus fn (x) converges to some f (x) for each x ∈ [0.ξk+1 )  = ξk (ξk+1 −ξk )p  ξk f  . it is not true. m ≥ N . Let ξv be a Cauchy sequence in ∞ . Then (n) (m) (n) ξv − ξv  < ε for each v and n. there exists δ > 0 such that E f p < (ε/8M )p whenever mE < δ. Thus gn fn − gf p → 0. By Chebyshev’s inequality. there exists N such that sup ξv − ξv  < ε for n. sup ξv − ξv  < ε for n ≥ N . Then fn p → f p since  fn p − f p  ≤ fn − f p .ξk+1 ) p ≤ f )χ[ξk .  ξv − ξv ∞ → 0. Also. (n) (m) (n) (m) Given ε > 0. 1]. Suppose fn − f p → 0. k=0 ξk (ξk+1 −ξk )p ξk Hence T∆ (f )p ≤ f p and T∆ (f )p ≤ f p . Let g = χ[0.ξk+1 ) p . 13. Let g ∈ Lq .and  ξv − ξv p → 0. 1]. Then fn → 0 and fn 1 = 1 for each n. there exists E such that mE < δ and fn E converges uniformly to f on E c . Let fn be a sequence in Lp . there exists N such that ξv − ξv  ≤ ξv − ξv  + ξv − ξv  + ξv − ξv  < ε. It is straightforward to check that  · ∞ is a norm on c and c0 .e.e.ξk+1 ) p = a  ξk+1 −ξk ( ξk f )χ[ξk . 1] be the space of all continuous functions on [0. which converges a. which converges a. Hence ξv ∈ c and sup ξv − ξv  < ε. m ≥ N . Given ε > 0. Thus ξv converges to some ξv for each v. m ≥ N and x ∈ [0. i. Thus lim fn − f p ≤ 0 ≤ lim fn − f p . Now for each v (N ) (N ) (N ) (N ) and v . 14. It is straightforward to check that  · ∞ is a norm on ∞ . there exists E such that mE < δ and gn converges uniformly to g on E c . Let fn be a sequence in Lp . i. Furthermore. Suppose there is a constant M such that fn p ≤ M for all n. Now  fn g − f g ≤ fn − f g ≤ ( E fn − f p )1/p ( E gq )1/q + ( E c fn − f p )1/p ( E c gq )1/q ≤ 2M (ε/4m) + (ε/(2(mE c )1/p gq ))(mE c )1/p gq = ε for n ≥ N . there exists N such that sup ξv − ξv  < ε for n. Let fn = nχ[0. m ≥ N so fn (x) − fm (x) < ε for n. Given ε > 0.e. Thus T∆ (f )p ≤ o p b ξk+1 ξ ξ m−1 ξk+1 m−1 1 1 f p ( ξkk+1 1)p−1 = k=0 (ξk+1 −ξk )p−1 ξkk+1 f p (ξk+1 − ξk )p−1 = a f p . Then f g = 0 but fn g = fn = 1 for each n.
ρ) and let x be a representative 30 . ρ∗ (x. . Let f = 1. z) = 0 and ρ(z. Then f is a bounded measurable function. Conversely. If g1 = 0. If ρ(x. then g = 0 a. Thus  F (ev ) ∞ ≥ F . If ρ(x. y). Let F be a bounded linear functional on p . {x : ρ(x. Conversely. . then ρ(x. For p = 1. 0. ρ∗ (x.1 Metric Spaces Introduction 1a. x) = ρ(x. y. For p > 1. z) + ρ∗ (z. n = 3). y) = 0. Thus ρ+ (x. Given ε > 0. y) < ε < δ − ρ(x. choose f = g − hq−2 (g − h). y) < ∞. 1]. x) = ρ(x. F ( ξv ) =  ξv F (ev ) ≤  ξv p  F (ev ) q so  F (ev ) q ≥ F ( ξv )/ ξv p for all nonzero ξv ∈ p . Let X ∗ be the set of equivalence classes under this relation. Let ρ be an extended metric on X. Thus  F (ev ) 1 ≥ F . n ∞ 23. Then f p = g − hp(q−1) = g − hq so f ∈ Lp . then ρ(y. let xv = sgn(F (ev )) and let x be deﬁned as n n n before. F ( ξv ) =  ξv F (ev ) ≤  ξv ∞  F (ev ) 1 so  F (ev ) 1 ≥ F ( ξv )/ ξv ∞ for all nonzero ξv ∈ c. *24. Then f ∈ L1 and f (g − h) = g − h. Thus g = h a. 7 7. Now xp = ( v=1 xv p )1/p = n q F (ev ) n n n ( v=1 F (ev )p(q−1) )1/p = ( v=1 F (ev )q )1/p so ( v=1 F (ev )q )1/q = ( n v=1 (ev )q )1/p = F (x) ≤ xp v=1 F F  for all n. Let g be an integrable function on [0. Thus ρ deﬁnes a metric on X ∗ . z) and y ∈ Sz. If ρ(x. y) = 0. Then F (x) = v=1 xv F (ev ) = v=1 F (ev ) so v=1 F (ev ) = F (x)/x∞ ≤ F  for all n. For any x. . z)+ρ(z.e. then x and y are in the same equivalence class. . y) ≤ ρ(x . y) ≤ ρ(x. y) ≤ ρ(x. Thus F (ev ) ∈ 1 and  F (ev ) 1 ≤ F . Now f g = f h so g − h = f (g − h) = 0. y) = 0 is an equivalence relation. zj . y) < ∞. z) < ∞ and ρ(z. . 1b. y) = ρ(x . Then f is a bounded measurable function. F ( ξv ) =  ξv F (ev ) ≤  ξv 1  F (ev ) ∞ so  F (ev ) ∞ ≥ F ( ξv )/ ξv 1 for all nonzero ξv ∈ 1 . Now g − hq−2 (g − h)g = g − hq−2 (g − h)h so g − hq = 0. For n = 2 (resp. y) = 0. For any x. 1 ≤ p < ∞. Then ρ(z. x) = 0. y) ≤ ρ(x. For each v. y). y) < 1} is the interior of the circle (resp. 22. yj . y) = i=1 xi − yi  ≤ i=1 (xi − zi  + zi − yi ) = i=1 xi − zi  + i=1 zi − yi  = ρ∗ (x. let xv = F (ev )q /F (ev ) = F (ev )q−1 sgn(F (ev )) and let n n n x = x1 . y) = 0 if and only if xi − yi  = 0 for all i if and only if xi = yi for all i if and only if x = y. The argument for ρ+ is similar except for the last property. y) < 1} is the interior of the square (resp. 2. Hence ρ(x. y) ≤ ρ(x. Thus ρ(x. Then f g = E g ≥ (g∞ − ε)mE = (g∞ − ε)f 1 . 21b. xj − yj  ≤ xj − zj  + zj − yj  ≤ maxi xi − zi  + maxi zi − yi  = ρ+ (x. y) ≤ ρ+ (x. y ). For each v. f ∞ = 1 and f g = g = 0 = g1 f ∞ . ρ(x. Thus  F (ev ) q ≥ F . Thus F (ev ) ∈ q and  F (ev ) q ≤ F . Let g be a bounded measurable function. Also suppose that y and y are in the same equivalence class. let ev be the sequence with 1 in the n vth entry and 0 elsewhere. n n n n ρ∗ (x.δ . x) = 0 < ∞. If ρ(x. y). let E = {x : g(x) > g∞ − ε} and let f = χE . For each v. Since xi − yi  = yi − xi  for each i. Suppose 0 < ε < δ − ρ(x. If g1 = 0. 0. Conversely.6. Now ρ∗ (x. xn . y) < ∞. y) = ρ∗ (y. y) < ∞. Let Xα0 be a part of the extended metric space (X. 3b. {x : ρ∗ (x. x). y) < ∞ is an equivalence relation. Then F (x) = v=1 xv F (ev ) = v=1 F (ev )q . note that  ξv − v=1 ξv ev p → 0 for each ξv ∈ p so F ( ξv ) = ξv F (ev ) by linearity and continuity of F . cube) with center y and sides of length 1. {x : ρ+ (x. z). y) ≥ 0 for all x. Let F be a bounded linear functional on Lp . sphere) with center y and radius 1.ε . and suppose there exist functions g and h in Lq such that F (f ) = f g = f h for all f ∈ Lp . Note that  ξv − v=1 ξv ev − v=n+1 ξev ∞ → 0 for each ξv ∈ c with lim ξv = ξ so F ( ξv ) = ξv F (ev ) by linearity and continuity of F . z) + ρ+ (z. ρ(x. n For 1 < p < ∞. 3a. note that  ξv − v=1 ξv ev 1 → 0 for each ξv ∈ 1 so F ( ξv ) = ξv F (ev ) by linearity and continuity of F . f ∞ = 1 and f g = g = g1 f ∞ . z) + ρ+ (z. Then ρ(x. y ) + ρ(y . bipyramid) with center y with height and width 2. y) < 1} is the interior of the diamond (resp. then ρ(y. .e. z) + ρ(z. x ) + ρ(x . . Thus ρ(x. Similarly for c0 . y) = 0 so ρ(x. y) = ρ(x . Thus g = h a. y) < δ so y ∈ Sx. . Finally. y) = 0.e. Clearly.5 Bounded linear functionals on the Lp spaces 21a. For any xj . Suppose x and x are in the same equivalence class. z) + ρ(z. then 0 ≤ ρ(x. y) = 0. Now F (ev ) = F (ev )/ev 1 ≤ F  for all v so F (ev ) ∈ ∞ and  F (ev ) ∞ ≤ F . choose f = sgn(g − h). For p = 1. let f = sgn(g). If ρ(x.
let f (x) = ρ(x. For example. let x = y/(1 + y). 9a. for any open subset O ⊂ E and y ∈ O. Then fn is a Cauchy sequence in C and it converges to g so g ∈ C. there exists y ∈ E such that ρ(x. Thus f (x) − f (z) < ε. / 31 . By Q2.δ so Sy. y ∈ [0.y] ∞ = 1 if x = y. Thus Xα0 is open.y] − d∞ < 1/2.δ . z) < δ. If there exists a countable dense subset D. z) < ∞ so z ∈ Xα0 . 6b. E) ≤ ε + f (z). then ρ(x. is continuous on [0. ∞). E) = 0. given f ∈ L1 and ε > 0. then x(1 − y) = y(1 − x) so x = y. Then {x : ρ(x. For a ﬁxed set E. y) ≥ ρ(x. let X be any metric space with X > 1 and ρ being the discrete metric. Thus S c is open and S is closed. Thus h is onto. Sy. f (z) − f (x) < ε. y) ≤ δ}.x] − d∞ < 1/2 and χ[0. 1). ρ(x. E) = inf y∈E ρ(x. let δ = ε. If E ⊂ F and F is closed. Conversely. Since continuous functions on [0. ∞). y) − ρ(x. 7.2 Open and closed sets 4a. y) ≤ 1} = X. The function h. ¯ 5b. 6a. there exists fn ∈ C such that fn − g∞ < 1/n. Thus O ⊂ E ◦ for any open subset O ⊂ E so O⊂E ⊂ E ◦ . Let h be the function on [0. y) < δ}. E) = α. L∞ is not separable. y) − δ.δ . y) < δ}.ε ⊂ Sy. C ⊂ L∞ .δ = {x : ρ(x. Consider χ[0. Similarly.3. Rn is separable with the set of ntuples of rational numbers being a countable dense subset.x] − χ[0. Then f (x) = ρ(x. y) = 0 if x = y. Sx. Thus C is a closed subset of L∞ . *7. The inverse function h−1 is given by h−1 (x) = x/(1 + x). y). Consider the set S = {x : ρ(x. 6c. Conversely. say ρ(x. which is continuous. ¯ ¯ ¯ 5. then E ⊂ F = F . ¯ 9b.δ is open. For each 1/2 1/2 1/2 n. Let g be a point of closure of C. then for any z ∈ Sy.e. C is complete. there exists δ > 0 such that x ∈ O ⊂ E for all x with ρ(x. ¯ E E⊂F F ⊂ E. let 0 < ε < δ − ρ(x. We may further approximate ϕ by a step function where the partition intervals have rational endpoints and the coeﬃcients are rational to get a countable dense subset. y) > δ. ◦ 5a.y] where x. y) ≤ ρ(x. For every n. y) + ρ(y. y) − δ = δ so z ∈ S c . Then x ∈ [0. For any x ∈ Sy. For any y ∈ [0. y) < δ so x ∈ E. 1) and h(x) = y. then for any δ > 0. Let g be a point of closure of the set of integrable functions that vanish for 0 ≤ t < 1/2.e. Conversely.δ . Xα0 is closed. Let x(t) be a measurable function with x < 1.8. E) = 0} = E. there exists a step function ϕ such that f − ϕ1 < ε. since ¯ is a closed set containing E. y) < 1} = {y} and {x : ρ(x. If h(x) = h(y). then there exists d ∈ D and x. When ρ(x. 1] are bounded.13. 1] with x = y such that χ[0. {x : ρ(x. y) < δ + ρ(z. Thus f (x) − f (z) < ε and f is continuous. y) > α/2 for all y ∈ E so x ∈ E. Then χ[0. Then χ[0. ρ(z.δ ⊂ Xα0 . Consider the ball Sy. 1) and [0. Then ρ(x. by interchanging x and z. y) < 1} = {y}. Let δ = 1 − x.y] ∞ < 1. 7. take any y ∈ E. For any z ∈ Sx. z) > ρ(x. E) > 0. Hence h is a homeomorphism between [0. Now X = Xα so Xα0 = X \ α=α0 Xα . E is an open subset of E so E ◦ ⊂ O⊂E O. 1) given by h(x) = x/(1 − x). By Proposition 6. ρ(x.e. y − x < δ so y = (y − x) + x < δ + x < 1. Contradiction. y ∈ [0. By Q6. 4b. and the set of integrable functions that vanish for 0 0 ≤ t < 1/2 is closed in L1 . For any y ∈ Sx.of Xα0 . Then 0 g ≤ 0 fn − g + 0 fn  ≤ 1 fn − g < 1/n for all n. i. y) < δ. z) + ρ(z. y). there exists fn in the set such that fn − g1 < 1/n. If ρ(x. Take x ∈ S c .x] and χ[0. (E)c = ( E⊂F F )c = E⊂F F c = F c ⊂E c F c = (E c )◦ . C is separable with the set of polynomials on [0. Given ε > 0. y) = 1 if x = y and ρ(x.δ with δ > 0. ¯ ρ(x. From the deﬁnition. Hence g vanishes a. 1]. L1 is separable. If y ∈ Xα0 . Hence the set of measurable functions x(t) with x < 1 is open in L1 . E) ≤ ρ(x. being a rational function. The set in part (b) is not always the closure of the ball {x : ρ(x. z) ≤ ρ(x. 1] with rational coeﬃcients being a countable dense subset (Weierstrass approximation theorem).x] − χ[0. if ¯ Hence {x : ρ(x. Since the union is open. 4c. Thus h is onetoone. Let δ = ρ(x.3 Continuous functions and homeomorphisms 8. i. Thus E ⊂ E⊂F F for closed sets F .
it is a homeomorphism so ρ and σ are equivalent metrics. 10c. f (xn )) ≥ ε. suppose xn does not converge to x. Thus ψ is not equivalent to the metrics in part (b). there exists n ≥ N with ρ(x. ρ(x. Suppose the sequence xn converges to x. xnk ) < ε/2 for k ≥ N and ρ(xn . Given ε > 0 and given N . y) < δ. Also. y) < δ. there exists δ > 0 such that if ρ(x. Conversely. y) ≥ 0 with σ(x.z)+ρ(z. then σ is an extended metric (resp. xm ) < ε/2 and ρ(yn . N ). xn ) < ε. ym ) < ε/2 for n. y) = σ(y. The sequence xn converges to x but f (xn ) does not converge to f (x). If every subsequence of xn has in turn a subsequence that converges to x. For any δ > 0. . xnk+1 ) ≥ ε. Suppose a Cauchy sequence xn in a metric space has a cluster point x. we see that if σ(x. xnk ) + ρ(xnk . (ρ+ (x. yn ) ≤ ρ(xn . Let X and Y be metric spaces and f a mapping from X to Y . y ∈ X so (X. there exists δ > 0 such that h(σ(x. y) < δ. . Suppose the sequence xn has x as a cluster point. Let ρ be a metric on a set X and let σ = ρ/(1 + ρ). y) < nδ = ε. Thus ρ and ρ∗ are equivalent metrics. f (y)) < ε if ρ(x. When ρ(x. ρ∗ (x. then σ(x. σ) as well as from (X. ε). Conversely. 13. . xnk have been chosen. xn ) ≥ ε. y) = h(σ(x.y) ≤ 1 − 1+ρ(x. suppose f is not continuous at x. y)) < ε when σ(x. ¯ ¯ there exists n such that ρ(x. y)) where h is the function in Q8. Now σ(x. xnk ) < ε.y) = h(σ(x. . Suppose xn1 . y) < δ . 11b. Suppose ρ and σ are equivalent metrics on X. Let xn and yn be Cauchy sequences from a metric space X. 14. Thus σ(f (x). Now given ε > 0. . m ≥ N . By considering the inverse function. Since the identity mapping is clearly bijective. xn ) < 1/n. Similarly for ρ∗ and ρ+ . The identity mapping is a homeomorphism between (X. then ρ(x. there exists N such that ρ(x. ρ+ (x. x ∈ E.y) = σ(x.z)+ρ(z. y) < (nδ 2 )1/2 = (ε2 /n)1/2 < ε. σ(x. there exists δ > 0 such that σ(f (x). Note that ρ(x. y) = 1−σ(x. 15. . then for each n. y) < ε. 10b. σ(x. ρ(x. y)) < ε. Thus given x ∈ X and ε > 0. Pick k ≥ max(N. y) = 1.y) 1+ρ(x. Since xn ∈ E. Given ε > 0. Now for k ≥ N . y). y) < δ. then given ε > 0. ρ(x. suppose there is a subsequence xnk that converges to x. There also exists N such that ρ(x. y) = 0 if and ρ(x. xn1 ) ≥ ε. Since h is continuous at 0. y) < δ. Then every subsequence of xn also converges to x and so has x as a cluster point. 17a. ρ(x. y) = 1+ρ(x. xn ) < δ for n ≥ N . . there exists N such that ρ(x. When ρ+ (x. Thus ρ+ and ρ∗ are equivalent metrics. Furthermore. Then there exists ε > 0 such that for every n. xn ) < 1/n but σ(f (x). xk ) < ε. The sequence xn converges to x. there exists xn with ρ(x. xk ) ≤ ρ(x. there exists xn ∈ E with ρ(x. . There exists nk+1 ≥ nk such that ρ(x. 0). When ρ(x. xnk ) < ε for k ≥ N . y) < δ. There exists n1 ≥ 1 such that ρ(x. xm )+ρ(xm . Pick n1 such that ρ(x. If x is a cluster point of a sequence from E. there exists N such that ρ(xn . 16. y) < δ. given ε > 0. y) < δ. y) < δ. Hence σ and ρ are equivalent metrics for X. y) ≤ 1 for all x. y) < δ < ε.4 Convergence and completeness 12. . ρ) and (X. yn ) so 32 .10a. When ρ+ (x. Now ρ(xn . y) < δ. y) < δ . If ρ is an extended metric (resp. m ≥ N . xnk+1 ) < 1/(k + 1). The rest of the argument in part (a) follows. Suppose xn1 . σ) to (X. The subsequence xnk converges to x. There √ < exists δ > 0 with δ < ε/ n such that ρ∗ (x. y) < δ . σ). When ρ∗ (x. Given ε > 0. By Q12. y) n < ε. y) < ρ(x. ρ). Given ε > 0. Consider the discrete metric ψ. The subsequence xnk does not have x as a cluster point. y) < δ < ε. There exists ε > 0 such that for each N . Thus ρ+ and ρ are equivalent metrics. ρ(x. Suppose f is continuous at x and let xn be a sequence in X that converges to x. Thus xn converges to x. the two implications show that the identity mapping is continuous from (X. Thus x is a cluster point of the sequence xn . Given ε > 0. y) < δ < ε. Let x = (0. ρ∗ (x. we can choose y = x such that ρ(x. x). y) < ε. then every subsequence of xn has x as a cluster point by Q12. y)√ δ implies ρ+ (x.y) 7.y) ≤ σ(x. Clearly. ρ) to (X. xm ) < ε/2 for n. On the other hand. Then nk ≥ k ≥ N and ρ(x.y) = 1 − 1+ρ(x. Let E be a set in a metric space X. if x ∈ E. . z) + σ(z. pseudometric). Hence the sequence xn converges to x. When ρ(x. Conversely. pseudometric). Then pick nk+1 ≥ nk such that ρ(x.y) 1 1 only if x = y. f (xn )) < ε for n ≥ N so the sequence f (xn ) converges to f (x) in Y . y) ≤ ρ(x. y) < δ but ψ(x. xnk have been chosen. ym )+ρ(ym . there is a subsequence xnk that converges to x. . 11a. σ) is a bounded metric space. let δ < min(δ . y) < ε. σ(x. let δ = ε/n. Conversely. σ(x.z)+ρ(z. . Then when ρ∗ (x. When σ(x. Hence σ is a metric on X. y))2 < δ 2 so ρ+ (x. xn1 ) < 1.
y . there exists N such that for m. y ) ≤ ρ(x. ym )2 )1/2 = τ ((xn . y) < ε/2 for n ≥ N . let δ = ε/2. We may also assume that ρ(xm. there exists δ > 0 such that ρ(x. ρ1 ( x. y . Furthermore. Deﬁne xn to be equivalent to yn (written as xn ∼ yn ) if ρ∗ ( xn . σ) be two complete metric spaces. m ≥ N . y ) = 0 if and only if ρ(x. x ) < ε/2 and σ(y. T is an isometry from X onto T [X] and T −1 is an isometry from T [X] onto X. ym ) < ε for n. y) ∈ X × Y and X × Y is complete. 20. Let xn be a sequence from E that converges to a point x ∈ E. Since ρ(xn . yn ) = 0 so ρ∗ ( xn . 17d. Hence T is an isometry between X and T [X] ⊂ X ∗ . Thus ρ(xm. If xn is a Cauchy sequence from X. ym ) ≤ ρ(xn . yn ) for Cauchy sequences xn and yn . 33 . the sequence xn is Cauchy in X.m . y . Hence X is isometric with the closure of X in Y . x .m ∞ n=1 is Cauchy in X. xn ) = lim ρ(xn . (xm . y )2 < ε2 /4 so ρ1 ( x. Similarly. y)) = (ρ(xn . f (xm )) < ε for n. y ) < ε/2 so τ ( x. xm ) + ρ(yn . then ρ∗ ( xn . Then τ ((xn . y ). y . x) < ε/2 and σ(yn . x . y). x . we have T ◦ T = idX ∗ and −1 ∗ T ◦ T = idX∩Y so (T = T ) . ρ∗ ( xn . yn ). x) + σ(y . y . yn ) = ρ∗ ( xn . y = x . We may assume that for n ≥ N . xn ). x .m ) < ε for m. yn converges to some y ∈ Y . y) = 0 so x = y. y)2 )1/2 < ε for n ≥ N . Deﬁne ρ∗ ( xn . m ≥ N . if T x = T y. If xn ∼ xn and yn ∼ yn . 7. xm ) + ρ(yn . y ) < ε2 /4 + ε2 /4 < ε. zn ) + lim ρ(zn . limn ρ(xn. y . Hence ρ1 and ρ∞ are uniformly equivalent to the usual product metric τ . xm )2 + σ(yn . yn ) is Cauchy in R and thus converges. x. f (xn ) converges to some y ∈ Y . xm ) + ρ(yn .m ) < ε/2.m ) < ε/2 since the sequence xn. ym ) − ρ(xn . yn ) be a Cauchy sequence in X × Y . Hence the set of all Cauchy sequences from a metric space becomes a pseudometric space under ρ∗ . y . ρ(xm .m . yn ) = lim ρ(xn . yn ) − ρ∗ ( xn . y )) < ε. yn ) − ρ(xm . yn ) ≤ lim ρ(xn . ρ∞ ( x. there is a unique isometry T from ∗ ¯ ∩ Y that extends T −1 . y ) = ρ(x.m ) < ε/2. Let f be a uniformly continuous mapping of a metric space X into a metric space Y and let xn be a Cauchy sequence in X. y ) < δ. xn ) = 0. xm . ρ(xm. x . Then T X = T and T T [X] = T −1 so (T ◦ T )T [X] = idT [X] and X onto X ¯ (T ◦ T )X = idX . y ) = ρ(x. . y . Furthermore. yn ) = lim ρ(xn . Also. yn ) ≤ ρ∗ ( xn . f (y)) < ε. the sequence yn is Cauchy in Y . Hence the sequence (xn . yn ) = 0. yn ) ≤ ρ(xn . xm . Associate each x ∈ X with the equivalence class in X ∗ containing the constant sequence x. Since T [X] is dense in X ∗ and X is dense in X ∩ Y .m ) < ε/2. yn ). Thus there is a ¯ unique isometry T from X ∩ Y onto X ∗ that extends T . y) < δ implies σ(f (x). Given ε > 0. x ) + σ(y. Similarly. x . y ) < ε/2 so τ ( x. y . y ) = max(ρ(x. There exists N such that ρ(xn . y ) + ρ1 ( x . xn+1 ) < 2−n . x ) + σ(y. m ≥ N . Finally. xm.5 Uniform continuity and uniformity 19. there exists N such that ρ(xn . xm ) ≤ (ρ(xn . y ) ≥ 0.m . y ) = ρ(x.m ∞ ∞ be a sequence of such Cauchy sequences which represents a Cauchy sen=1 m=1 quence in X ∗ . Also. Given ε > 0. . ¯ 18. *17e. In particular. yn ). 17c. we may assume (by taking a subsequence) that ρ(xn . x . T is continuous on X and its inverse is continuous on T [X]. Given ε > 0. y ) = ρ(x . σ(y. y .e. Thus T is onetoone. Similarly. ρ) and (Y. yn ) − ρ(xm . Hence ρ1 is a metric. . x. then ρ(x. Thus ρ(xn . then xn ∼ yn so they are equal in X ∗ . xn ) + ρ∗ ( yn . yn ) = 0. ρ1 ( x. x ) + σ(y. When τ ( x. x ) + σ(y. Then ρ∗ ( xn . y) = ρ1 ( x . T [X] is dense in X ∗ . ym ).m ) < ε/2. yn ) = lim ρ(xn . ρ∗ ( xn . Similarly. Since ρ∗ (T x. xn converges to some x ∈ X.m . xn.m . ρ(x. y ) = ρ(x. . x ) + ρ(x . x . (x. ρ∗ ( xn. x)2 + σ(yn . Since X is complete. x . This deﬁnes a mapping T from X onto T [X]. x . ρ(xn. T y) = lim ρ(x. m ≥ N . y) = ρ(x. Hence the sequence ρ(xn . Thus the pseudometric space becomes a metric space. x ) = 0 and σ(y. y ) < δ. Since Y is complete. y . Then xn is Cauchy in X so f (xn ) is Cauchy in Y . yn ) ≥ 0 since ρ(xn . y ) < ε. Similarly. y . If ρ∗ ( xn .m . yn ) ≥ 0 for each n. y ). m ≥ N so the sequence xn. ρ∗ ( xn . When ρ∞ < δ. y ) = ρ1 ( x. ym ). y ) = 0 if and only if x = x and y = y if and only if x. x )2 < ε2 /4 and σ(y. i. y . yn ) converges to (x. Let (X. y ) + σ(y . ρ1 ( x. y . ρ1 ( x. ym ) ≤ ρ(xn . ¯ 21a. Also. x ) + σ(y. ym )). Given ε > 0. ρ(x. When ρ1 ( x. yn ) = lim ρ(yn . x . yn ). ρ(x. x ) < ε/2 and σ(y. Hence f (xn ) is a Cauchy sequence in Y . x ). xn ) = ρ∗ ( yn . Thus σ(f (xn ). Let xn. zn ) + ρ∗ ( zn . xm ) < δ for n.ρ(xn . xn. ρ∞ is a metric. xn. x . y ) < ε2 /4 + ε2 /4 < ε. Let (xn .n is Cauchy in X and represents the limit of the Cauchy sequence in X ∗ . yn ) = ρ∗ ( xn . 17b.
Then σ(y. Thus ρ and ρ∗ are uniformly equivalent metrics. y) < δ. . choose n large enough so that 3n2 δ > 1. y) = δ and ρ (x. . f (xn ) converges to some y = g(¯) ∈ Y and f (xn ) converges to some y = g(¯ ) ∈ Y . But then xk + 2 is not in any of the balls Sxn . f (xn )) + σ(f (xn ). ρ) and ([0. n 3 *22b. ∞). x) < δ/2 for n ≥ N . the function h(x) = x/(1 − x) is a homeomorphism between [0. xn ) < δ and σ(f (x). ε/n. Let ε > 0 be given. ¯ Suppose ρ(¯. ρ+ (x. Hence Y = n=1 Sf (xn ). f (x )) < ε/3 for x. x ) < δ implies σ(f (x). σ(x. given ε > 0. x ∈ E. y) < ε. By deﬁning y = g(x). 1 1 ∗ ∗ ∗ Thus ρ (x. Now given ε > 0. Thus ρ+ and ρ are uniformly equivalent metrics. Let f : (X. √ Then when ρ∗ (x. ∞) is totally bounded. g(¯ )) < ε/3 for n ≥ N . y) < δ . Thus [0. σ(x. X = n=1 Sxn . ρ) and (Y. . x) < δ/2 and ρ(xn . y ) < ε for n ≥ N . N . Now x ∈ Sxn . ρ) be a totally bounded metric space. x1 + δ). y) = x3 − y1  + i=2 xi − yi . σ) be a uniform homeomorphism. ∞). . When ρ(x.ε so Y is totally bounded. Let xn be a sequence in E converging to x and let xn be a x ¯ ¯ ¯ ¯ sequence in E converging to x . 0. Let h be a continuous function from E a sequence in E converging to x. Choose N such that N > 2/ε and let xn = (n − 1)/N for n = 1. y) < δ . ε/3nx1 − 12 . Also. let δ < min(δ . y) n < ε. . ∞) with the usual metric ρ is an unbounded metric space. Given x ∈ R and ε > 0. g(¯)) < ε/3 and x x x ¯ σ(f (xn ). y) < δ. m ≥ N . Then there are a ﬁnite number of balls of radius 1 that cover [0. . . . Q11a) 23a. x1 − 1) δ < ε/n and ρ (x. y)) where h is the function in Q8. y) < δ < ε. Hence g is uniformly continuous on E. Since h(xn ) = g(xn ) for all n. there exists δ > 0 such that ρ(x. Suppose [0. σ) be metric spaces with X totally bounded. . Given ε > 0. x) < δ/3 and ρ(xn . Then y = f (x) for some k x ∈ X. ρ and ρ are equivalent metrics. Let (X. Contradiction.1 . Then x1 − y1  = x3 − y1 /3ξ 2  < ε/n. ¯ ¯ ¯ Then ρ(xn . . Given ε > 0. 0 . ρ) → (Y. Then ρ is a metric on the set of ntuples of real 1 n numbers. ∞). . When σ(x.1 . Since h is continuous at 0. y)) < ε. . y )/4 > 0. ∞) = n=1 Sxn . Thus total boundedness is a uniform property. Then ρ(x. . When ρ(x. k say [0.y) = h(σ(x. σ) is a bounded metric space. choose δ < min(1. Given ε > 0. x ) < δ implies σ(f (x). ([0. f (xm )) < ε for n. xi − yi  < ε/n for i = 2. y) = h(σ(x. 21c. ε). Let ε = 1. g(x) = h(x). σ). there are a ﬁnite number of balls of 34 . ¯ we get a function on E extending f . the metric σ = ρ/(1 + ρ) is uniformly equivalent to ρ so there is a uniform homeomorphism between ([0. Let (X. Thus x1 − y1  < 3 max(x1 + 1. xi − yi  < δ for each i and x3 − y1  = 3ξ 2 x1 − y1  for 1 3 3 2 some ξ ∈ (x1 − δ. . . f (x )) < ε.f. We may assume that σ(f (xn ). When ρ(x. Let x = n. Thus ρ(xn . y) < δ. 3εx1 − 3 12 /n. x1 . y ) ≤ σ(y. y )/4. x x x ¯ to Y that agrees with f on E. Deﬁne ρ (x. By Q22c. x ) < δ/3 with x. y) < δ. f (x )) < ε. n and x3 − y1  < δ.δ . 1). Hence f (xn ) and f (xn ) converge to the same point. . . There also exists N such that ρ(xn . there exists δ > 0 such that h(σ(x. ∞) is not totally bounded. y) = (n + δ)3 − n3 > 3n2 δ > 1. ρ(x.δ that cover X. Thus σ(g(¯). y ) < 3ε = 3σ(y. ∞). ρ∗ (x. Let ε = σ(y. For each n. The intervals (xn − 1/N. f (xn )) + σ(f (xn ). y) < δ . x ) < δ implies σ(f (x). y) < ε and σ(f (xn ). y) < (nδ 2 )1/2 = (ε2 /n)1/2 < ε. 23c. Note that ρ(x. Suppose xn and xn both converge to x. Let x ∈ E and let xn be ¯ 21d. y) < δ. (ρ+ (x. g(¯ )) < ε. xm ) < δ for n. 22a. y) < nδ = ε. ∞). there exists δ > 0 such that ρ(x. Then h(xn ) converges to h(x) and g(xn ) converges to g(x). y)) < ε when σ(x. There k exist ﬁnitely many balls Sxn . Hence ρ and ρ are equivalent metrics and since ρ and ρ are equivalent metrics. 1) are balls of radius ε that cover [0. When 3 3 ρ (x. .e. f (xn )) < ε. ρ∗ (x. Suppose f (xn ) converges to y and f (xn ) converges to y with y = y . xn + 1/N ) ∩ [0. x ) < δ/3 for n ≥ N . ρ(x. There exists δ > 0 such that ρ(x.δ for some n so ρ(x. ε/3nx1 + 12 . y) = 1−σ(x. Given δ > 0. There exists N such that ρ(xn . Hence boundedness is not a uniform property. 3εx1 + 12 /n). y) < ρ(x. y) < δ. 23b. When ρ+ (x. . xk are arranged in increasing order. y) < δ < ε. Thus total boundedness is not a topological property. ρ(x. When ρ+ (x. i. By Q11a. x ∈ E. Thus ρ+ and ρ∗ are uniformly equivalent metrics.y) 22c. 0 and let y = n + δ. Take y ∈ Y . 0. y) ≤ ρ(x. 23d. y) < δ. Hence σ and ρ are uniformly equivalent metrics for X. f (xn )) < ε/3 for n ≥ N . m ≥ N so σ(f (xn ). When ρ∗ (x. y) < δ < ε. y) < ε. However ρ and ρ are not uniformly equivalent metrics. Hence h ≡ g. y) < δ. xn ) < δ for n ≥ N so σ(f (xn ). We may assume that x1 . There exists δ > 0 with δ < ε/sqrtn such that ρ∗ (x. (c. y) < δ. 1) is totally bounded. Contradiction.21b. [0. y) < ε. y))2 < δ 2 so ρ+ (x. By Q8. y) < δ implies ρ+ (x. When ρ∗ (x. 1) and [0. . Hence [0. let δ = ε/n. We may assume σ(f (xn ).
xk ) < ε for n ≥ N . *26. f (y)) = ρ(x. Hence σ is a bounded metric for which (O. Now suppose B is a closed subset of a complete metric space Y . Also. . Since A is complete. σk ) are uniformly homeomorphic. Let yn be a Cauchy sequence in B. Then given ε > 0. xk ) < k=1 ε/2k+1 + ε/2 < ε for n ≥ M . Let x(n) be a Cauchy sequence in (Z. τ (x. When ρ∞ ( x. Thus f is a uniform homeomorphism and (O. Then each Sn is a ﬁnite set and S = Sn is a countable set. let ε = min(2−k−1 ε. By part (a). xk0 )) for n ≥ N so (1 − 2 ε )ρk0 (xk0 . there exists N such that (n) ∞ −k ∗ ρk (xk . Given ε > 0. Thus ρ(x. Since X × R is complete. Thus S is a dense subset of X. xk0 ) < 2k0 ε for n ≥ N . N . y) = k=1 2 ρk (xk . F ) > 0 for all x ∈ K. yk ) where ρk = ρk /(1 + ρk ). x) < ε for n ≥ N . Deﬁne f : k=1 Xk → k=1 Yk by f ( xk ) = f (xk ) . x(n) converges to xk ∈ Z. z) + τ (z. y. y). Then S = { x. Oc ) for each x ∈ O. there is a sequence from A that converges to x. x) since ρ∗ (xk . Let A be a complete subset of a metric space X. ϕ(y) ) < δ . Suppose a sequence x(n) in Z converges to x ∈ Z. σk ) are homeomorphic with fk : Xk → Yk a ∞ ∞ homeomorphism. For any x ∈ X. xk0 )/4 < 2k0 ε < ε/2 for n ≥ N . 24d. By Q22c. Since 2k0 ε = min(ε/2. note that f is continuous if and only if pYk ◦ f is continuous for each k. Then f is bijective since each fk ∞ is. Let σ = ρ/(1 + ρ). y) + ϕ(x) − ϕ(y) < ε. Let ϕ(x) = ρ(x. the spaces (Xk .6 Subspaces ¯ 25. Furthermore. Then for each k. S is complete. Hence f is a homeomorphism between ∞ ∞ k=1 Xk and k=1 Yk . there exists δ > 0 such that ϕ(x) − ϕ(y) < ε/2 when ρ(x. ρk ) so it converges to xk ∈ Xk . (n) Conversely. y) = 0 if and only if ρ∗ (xk . yk ) = 0 for all k if and only if xk = yk for all k if and only if k x = y. Let X be a metric space. Consider the function ¯ f (x) = ρ(x. ϕ(x) . τ (x. then ρ(x. y) = 0 so ρ(F. If for each k. y = ϕ(x)} is closed in X × R. ρ∞ (f (x). y) ≤ τ (x. y) < δ. the spaces (Xk . {x : ρ(x. ρ(x. Let Sn be the set of the centres of these balls. σ is uniformly equivalent to ρ. Deﬁne ∞ −k ∗ ∗ ∗ τ (x. S is complete under the metric ρ∞ . xk0 ) < 2 ε (1 + ρk0 (xk0 . 7. Hence (Z. xk0 ) < ε for n ≥ N and (n) xk (n) (n) converges to xk for each k. . 24c. τ ).7 Compact metric spaces 27. y) = k k k k k=1 2 τ (x. y) = τ (y. there exists M such that ρk (xk . 2−k−2 ). y) ≤ k=1 2−k ρ∗ (x. yk ) = ρ∗ (yk . τ ) is complete. xk ) < ε/2 for n ≥ M . F ) = inf y∈F ρ(x. Since ρ∞ is uniformly equivalent to the usual product metric by Q19. the spaces k=1 Xk and k=1 Yk are uniformly homeomorphic. Let x ∈ A. ρ) → (S. xk ) + k=N +1 2−k ρ∗ (xk . 1/4). y). Thus if F ∩ K = ∅. y) < δ . ρk be a sequence of metric spaces and deﬁne their direct product Z = k=1 Xk . ρk ) and (Yk . choose N such that N > 1/ε. Let Xk . Then it is a Cauchy sequence in Y ¯ so it converges to some y ∈ Y . Hence τ (x. 35 . By Q9b. x ∈ A. Now f is bijective. xk ) for all k. 7. ρ(x. . x) = k=1 2−k ρ∗ (xk . Similarly. z) + k=1 2−k ρ∗ (z. y : x ∈ O. Consider f : (O. Now y ∈ B = B so B is complete. τ (x. Hence ρk0 (xk0 .radius 1/n that cover X. K) = 0. ρk ) and (Yk . Let O be an open subset of a complete metric space (X. ∞ 24a. 24b. Then exists N such that τ (x(n) . x ) < 1/N < ε for some x ∈ SN ⊂ S. y). y) ≥ 0 since ρk (xk . k k Hence x(n) converges to x. (n) xk is a Cauchy sequence in (Xk . we have 3ρk0 (xk0 . z) + τ (z. Conversely. if F ∩ K = ∅. Thus τ is a metric on Z. σ) is a complete metric space. The function f K is continuous on a compact set so it attains a minimum δ > 0. ρk ) is complete. . Thus A is closed. Thus f is continuous. F ) = 0} = F = F . ρ∞ ) with f (x) = x. k=1 2 (n) (n) (n) k0 k0 ρk0 (xk0 . ρ) is complete. y) = k k n n n ∞ ∞ −k ∗ ρk (x. Then τ (x. Suppose that for each k. suppose xk converges to xk for each k. i. Suppose each (Xk . xk ) < ε/2. then there exists x ∈ K and y ∈ F such that ρ(x. Let δ = min(δ.e. then by a similar ∞ ∞ argument as part (c). There ∞ −k ∗ (n) ρk (xk . Given ε > 0. ε/2). Also. z) + k=1 2−k ρ∗ (z. y) < δ < ε. Given ε > 0. Now for each n. y) > δ for all x ∈ K and all y ∈ F . For k = 1. ρ). yk ) ≥ 0 for all k. Hence X is separable. ϕ(x) . K a compact subset and F a closed subset. When ρ(x. f −1 is −1 continuous since pXk ◦ f −1 = fk ◦ pYk is continuous for each k. where pYk : k=1 → Yk is the projection map. Now pYk ◦ f = fk ◦ pXk so it is continuous. k=N +1 2 N ∞ N Thus τ (x(n) . Then by Q14. y) ≤ k=1 2−k ρ∗ (x.
y) and By. xk = (k) (k) pk for k > N }. Suppose Bx. Suppose that En is a sequence of sets of the ﬁrst category. Conversely. If 0 < r < r − ρ(x. For k > N . y) − ϕ(y))/2. ∞ *30a. Thus any open set contains a point in F c so any open set cannot be contained in F . Now (A ∩ En )c ⊃ (A ∩ En )c = Ac ∪ En for each n. 1) is totally bounded but [0. suppose ¯ Now O \ E that for any nonempty open set O there is a ball contained in O \ E. Given ε > 0. ¯ \ O is a closed set containing no open sets.k . 7. Z is complete. For any x ∈ X and δ < ε. Given ε > 0.δ }k that cover X.ε }k cover Y and Y is totally n=1 bounded. ∞). Thus x ∈ E c .8 Baire category 31a. then there exists r > ϕ(y) + ρ(x. y). Thus ϕ(x) > 0. 29c. Let ε be as deﬁned in part (d). xn ) < δ and σ(f (x). it is bounded so ϕ(x) < ∞. ∞) is not. There exist ﬁnitely many balls {Bxn . let δ = ε. we have ϕ(y) < r < r − ρ(x. Conversely. there exists xj (k) that ρk (xk . 1] by removing intervals of length 1/(n3n ) at the nth step is closed. 32b. f (xn )) < ε. We may assume X ∈ U . choose N such that (k) (k) 2N > 2/ε. for each n. 33a. Thus O 36 . δ < ϕ(x) so there exists r > δ such that Bx. which is a countable union of nowhere dense sets. Then any open set contains a point in F c so F c is dense and F is nowhere dense. For each k ≤ N . Let A = { xn : xk ∈ Ak for k ≤ N.r ⊂ O. set contains a point in E 32a. Let E = Fn ⊂ [0. Contradiction. . there exists δ > 0 such that ρ(x. the generalised Cantor set constructed from [0.k En. Z is totally bounded and by Q24b. Suppose that E is of the ﬁrst category and A ⊂ E. for each k ≤ N . . If ¯ Contradiction. Suppose O is open and F is closed. xMk } such that for any x ∈ Xk . xj ) < ε/2. there exists xj (k) ∈ Ak with ρk (x. Then F c is dense. 1] with mFn = 1 − 1/n.y. Now x ∈ Bxn . pick pk ∈ Xk . y) < ε by part (b). .r ⊂ O. 30b. f (x )) < ε. Suppose each Xk is compact. Let yk n ρk k τ (x.14b. 1]. Contradiction. where each En. 1) onto [0. 29e. Since En is dense for c each n.δ for n=2 some n so ρ(x. Then En = k En.r ⊂ O}. 28b. If ϕ(y) < ϕ(x) − ρ(x. Then y = f (x) for some x ∈ X. Since X is compact. Given ε > 0. y). Suppose each Xk is totally bounded. If xn ∈ Z. there exists r such that Bx. there exists Ak = {x1 . nowhere dense. ϕ(x) − ϕ(y) ≤ ρ(x. then By. then O contains a point in O. Thus ϕ is continuous on X. 29a. 29d. Hence ϕ(y) ≥ ϕ(x) − ρ(x. Suppose a closed set F is nowhere dense. (A ∩ En ) is dense for each n. Let ε = inf ϕ. Then Bx. By part (a). By Q31a. Now En = n k En.r ⊂ O. ϕ attains its minimum on X.r ⊂ O for some O ∈ U . Hence the balls {Bf (xn ).28a. xj ) < ε/2. Since O is open. Hence Z is compact. For each x ∈ X. If X is sequentially compact.k = n.δ ⊂ Bx.k is nowhere dense. Let the centre of the ball be x. Then each Xk is complete and totally bounded. Set ϕ(x) = sup{r : ∃O ∈ U with Bx. If O \ O contains an open set O . By part (a). O \ O is nowhere dense. Let X be a totally bounded metric space and let f : X → Y be a uniformly continuous map onto Y .r ⊂ O for some O ∈ U . Then ε > 0 since ϕ(x) > 0 for all x. y) < δ. [0. y). ¯ 31b. Then yn ∈ A and < ε 2 + 1 2N < ε. Then E = En where each En is nowhere ¯ ¯ ¯ ¯c ¯c dense and A = (A ∩ En ). . there exists / O ∈ U such that x ∈ O. Let Z = k=1 Xk . Suppose E is nowhere dense and let O be a nonempty open set.k for each n. Then E is of the ﬁrst category and 1 − 1/n ≤ mE ≤ 1 for each n so mE = 1. ¯ is open so there is a ball centred at x and contained in O \ E ⊂ O \ E. If ¯ ¯ x ∈ E. By Q3. suppose a closed set contains no open set. ¯ 34. Take y ∈ Y . y) so taking r = ϕ(y) + (r − ρ(x.r ⊂ O for some O ∈ U . Hence A is of the ﬁrst category. Hence any nonempty open ¯ c and E is nowhere dense. Now ϕ(y) < r − ρ(x. Then O contains a point x in E c .r ⊂ Bx. Then A = M1 · · · MN . The function h(x) = x/(1 − x) is a continuous map from [0. x ) < δ implies σ(f (x). 33b. Hence En is of the ﬁrst category. there is a nowhere dense closed set Fn ⊂ [0. and has Lebesgue measure 1 − 1/n.r ⊂ O.y) ) < k (xk k ∈ Ak such = (k) xj for k ≤ N and let N ∞ −k ε −k k=1 2 k=N +1 2 2 + yk = pk for k > N . then the ball contains a point y ∈ E. y) = k=1 2−k 1+ρ(xk . When ρ(x. y) such that Bx. 29b.
Since D is of the ﬁrst category in the complete metric space C. 1 − 1/n]. F \ F ◦ is nowhere dense. . Hence E is residual. 1] and set Fn = {f : ∃x0 with 0 ≤ x0 ≤ 1 − 1/n and f (x) − f (x0 ) ≤ n(x − x0 ) for all x. the set of rationals in [0. We may assume that xn ∈ U for 37 . Hence there is no such function. 37c. i. and suppose that for each x ∈ X there is a number Mx such that f (x) ≤ Mx for all f ∈ F. Thus F = F \ F ◦ is nowhere dense. 1] is an Fσ . the set of irrational numbers in [0. 1]. [0. Since each f is continuous. Conversely. Then X = Fn ∪ Fn so X is of the ﬁrst category. a nowhere diﬀerentiable continuous function on [0. contradicting the Baire category theorem. 1]. 1 − 1/n] is compact. Let C = C[0. Since On is dense. contradicting the Baire category theorem. If E is residual. Since the set of rational numbers in [0. Contradiction. the set of rational numbers in [0. 1] is a Gδ . Hence its complement. [0. The set D of continuous functions which have a ﬁnite derivative on the right for at least one point of [0. 1]. Also suppose that E and E c are both Fσ ’s. There also exists a polygonal function ψ whose righthand derivative is c everywhere greater than n in absolute value and ϕ(x) − ψ(x) < ε/2 for all x ∈ [0. each Fn and Fn is nowhere dense. D = C so there is a function in C \ D. Let xn be a sequence with x = lim xn . By the Baire category theorem. x0 ≤ x < 1}. there exists xk with 0 ≤ xk ≤ 1 − 1/n and fk (x) − fk (xk ) ≤ n(x − xk ) for xk ≤ x < 1. *38b. By part (a). x ∈ Em for some m so there is a neighbourhood ◦ U of x such that U ⊂ Em . For each m. The set of rational numbers in [0. there is a ball centred at x and contained in O so x ∈ F ◦ . F circ is empty.F \ F ◦ contains an open set O . [0. By part (a). By Q31a. Hence at most one of the sets E and E c is an Fσ . If f is a realvalued function on [0. Also. Then F c ⊃ E c so F c is dense and F is nowhere dense. Suppose fk − f  → 0 where fk ∈ Fn . By Q2. then for any point x ∈ O . that is. 1 − 1/n] for all k and [0.e. 1]. is not a Gδ . . A subset E of a complete metric space is of the ﬁrst category if and only if E c is a residual subset of a complete metric space if and only if E c contains a dense Gδ if and only if E is contained in an Fσ whose complement is dense. then the set of points of continuity of f is a Gδ . 1] are both dense in the complete metric space [0. In particular. Em. 36a. If F is closed and of the ﬁrst category in a complete metric space. then E c is of the ﬁrst category so c ¯ ¯ E c = En where each En is nowhere dense. For each x ∈ X. 1] is uncountable. Suppose E c is dense and F is a closed set contained in E. 37b. Then X = {xn } where each {xn } is nowhere dense and X is of the ﬁrst category.f = {x : f (x) ≤ m}. 1] is complete since it is a closed subspace of the complete metric space R. . Then E c ⊂ On . suppose E contains the metric space is complete. f (x)) < ε for n ≥ N and all x ∈ U . Since each (En )c is dense and ¯n )c is dense so E contains a dense Gδ . x2 . 1]. *38a.f . 1] which is continuous on the rationals and discontinuous on the irrationals. Suppose X has a countable number of points x1 . 1] is not an Fσ by part (b). Suppose E and E c are both dense in a complete metric space X. 38c. Let X be a complete metric space without isolated points. 37d. Let E be a subset of a complete metric space. 38d. let Em. each On is dense so On is nowhere c dense and E is of the ﬁrst category.f is closed and so Em is closed. Then ψ ∈ Fn and g(x) − ψ(x) < ε for all x ∈ [0. 1] is the union of the Fn ’s so D is of the ﬁrst category in C. (E c c a dense Gδ . Then f (x) − f (x0 ) ≤ n(x − x0 ) so f ∈ Fn . there exists m such that f (x) ≤ m for all f ∈ F. Since xk ∈ [0. there exist N and a neighbourhood U of x such that σ(fn (x ). F is uniformly bounded on U . for any g ∈ C and any ε > 0. Then ◦ ◦ O = Em ⊂ X is a dense open set and for each x ∈ O. say E = Fn and E c = Fn where each Fn and Fn is closed. For each k. 35. say E ⊃ On . Thus E = En ⊃ (En )c . 40a. there exists a polygonal function ϕ such that g(x) − ϕ(x) < ε/2 for all x ∈ [0. Thus F \ F ◦ is a closed set containing no open sets. Hence X has an uncountable number of points. we may assume that xk converges to some x0 ∈ [0. Hence X = Em . Let E be a subset of a complete metric space.. 1] and the set of irrational numbers in [0. then F ◦ is of the ﬁrst category. Let F be a family of realvalued continuous functions on a complete metric space X. and let Em = F Em. Suppose that given ε > 0. 1]. Hence Fn is nowhere dense. 37a. 1] has no isolated points.47. 36b. contradicting part (c). 39.
there exists Nx and a neighbourhood Ux of x such that σ(fn (x ). ◦ 40g. f (x)) < ε/2 for n ≥ N and x ∈ U .all n ≥ N so σ(fn (xn ). f is continuous at x so we may assume that σ(f (x ). Hence fn converges continuously to f on X. . 0) = lim g(xn .n converging to x . 0 = lim zn and f (x0 ) = lim fn (xn ). x ) < δ implies σ(fk (x). . fl (x )) < ε. 0) = lim g(xn . f (x)) ≥ ε. fk (x )) < ε and σ(fl (x). g(x0 . f (x )) ≤ 1/m for any k ≥ n and x ∈ U . 40b. f (x)) ≥ ε. we have σ(fn (x ). By parts (f). f (x)) < ε for k ≥ N so f (x) = lim fk (xk ) and fn converges continuously to f at x. . fl (x )) ≤ 1/m. For each x ∈ X and each ε > 0. zn ) so g is continuous at x0 . 40i. f (x)) < ε for n ≥ N and fn converges continuously to f at x. Conversely. given ε > 0. f (x )) ≤ 1/m < ε for any k ≥ n and x ∈ U . .n is closed. By part (b). f (x)) < ε/3 for x ∈ U . Conversely. By part (c). For each n. Since ε is arbitrary. f (x )) < ε/4 for all n ≥ N and all x ∈ U . σ(fk (x). σ(fk (x ). fl (x )) ≤ 1/m and x ∈ Fm. there exists N and a neighbourhood U of x such that σ(fn (x ). . fl (x)) ≤ σ(fk (x).n is open and since X is complete. then x ∈ Om for all m. l ≥ n. 1/n) = fn (x) and g(x. Suppose fn converges continuously to f at x. suppose fn converges uniformly to f on each compact subset of X. If x ∈ E. fl (x )) ≤ σ(fk (x ). Let (xn . there exists n ≥ N and x ∈ U with σ(fn (x ). It follows that for any k. In particular.n . Then x = lim xnk by construction but f (x) = lim fnk (xnk ) so fn does not converge continuously to f . 1/n and f (x) = g(x0 . Let Z = {1/n} ∪ {0}. Since each fn is continuous. If xn is a sequence converging to x. Then fn converges uniformly to f on {x} for each x ∈ X. Suppose fn converges continuously to f on X. 41a. Then there exists a neighbourhood U of x such that σ(fn (x ). Hence X = n Fm. Let xk be a sequence with x = lim xk . Let fn be a sequence of continuous maps. f (x )) < ε m for each n ≥ Nx and x ∈ Ux . (g). Given ε > 0.n = {x ∈ X : σ(fk (x). there exists δ > 0 such that ρ(x.e. there exists n such that σ(fk (x).1/n . 40f. f (x)) ≤ σ(fk (xk ). it follows that g(x0 . Conversely. f (x)) ≥ ε. Suppose xn1 . Then x0 = lim xn . For any x ∈ X. choose m > 1/ε. f (x)) ≤ σ(f (xk ). i. suppose fn converges continuously to f at x0 . suppose that given ε > 0. Thus σ(fn (x ). f (x)) < 3ε/4 for k ≥ N . we have σ(fk (x ). Then σ(f (xk ). 0 . K ⊂ i=1 Uxi for some x1 . fN (x)) < ε/4 for k ≥ N . f (x )) < ε for n ≥ N and x ∈ U . There exists n1 ≥ 1 and xn1 ∈ U1 with σ(fn1 (xn1 ). Thus for n ≥ max Nxi and x ∈ K. fn (x)) < ε/3 and σ(f (x ). σ) be complete metric spaces and f : X × Y → Z be a mapping into a metric 38 . l ≥ n}. Then there exists N such that ρ(xi . 0 . x ) < δ for i ≥ N . Thus σ(fk (x ). Hence f is continuous at x. Suppose fn converges pointwise to f . Let X be a complete metric space and fn a sequence of continuous maps of X into a metric space Y such that f (x) = lim fn (x) for each x ∈ X. fN (xk )) + σ(fN (xk ). 0) = lim g(xn . There exists a neighbourhood U of x and an n such that σ(fk (x ). Thus σ(fk (xk ). 40c. fl (x )) < 2ε + 1/m. 0 in the product metric. there exists N such that σ(fn (x). E is a dense Gδ by Baire’s theorem. fl (xN )) + σ(fl (xN ). ρ) and (Y. fl (x)) < 1/m so x ∈ Fm.n so Fm. Then x0 . f (x)) < ε/4. fk (xN )) + σ(fk (xN ). let Un = Bx. . Thus there exists a neighbourhood U of x such that U ⊂ Fm. Deﬁne g : X × Z → Y by g(x. n ∈ N. f is also continuous. Om is also dense. 0 . By Proposi◦ tion 31. then f (x) = g(x. fN (x)) + σ(fN (x). Then for k. . Since 0 = lim zn . Suppose g is continuous at x0 .n . For m. 40e. Let x be a point of closure of Fm. f (x)) < ε/2 for x ∈ U . xm . fl (x)) ≤ 1/m for all k. *40d. f (x)) < ε/3 for n ≥ N . xnk have been chosen.n . There is a sequence xi in Fm. l ≥ n. f (x )) < ε for n ≥ N and x ∈ U . f (x)) < 1/2m for k ≥ n. l ≥ n and any x ∈ U . Then x ∈ Fm. There exists nk+1 ≥ nk and xnk ∈ Uk with σ(fnk+1 (xnk+1 ). 1/n) = lim fn (xn ) so fn converges continuously to f at x0 . Since each Om is open and dense. there exists a dense Gδ in X on which fn converges continuously to f . Conversely. 0) = f (x). f (x)) + σ(f (x).n for some n. Let xn be a sequence with x0 = lim xn . 0) = lim g(xn . 1/n). suppose there exists ε > 0 such that for any N and any neighbourhood U of x. σ(fN (x). deﬁne Fm. 40h. Let fn converge continuously to f at x. Let (X. f (x )) < ε so fn converges uniformly to f on K.n . Then X = Ux so for any compact subset K ⊂ X. Om = n Fm. zn ) be a sequence in X × Z converging to x0 . We may assume that xk ∈ U for all k and σ(fN (xk ). Since f (x ) = lim fl (x ). Let fn be a sequence of continous maps. Thus σ(fn (x ). the function g is continuous at x. there exist N and a neighbourhood U of x such that σ(fn (x ). For any ε > 0 and k. Let X be a complete metric space and fn a sequence of continuous functions of X into a metric space Y . . σ(fk (x ). Let x ∈ Om . 0) = lim f (xn ). f (xk )) + σ(f (xk ). By part (a). (h) and (a). Given ε > 0. 0 = lim xn . Let E = Om .
Given ε > 0. Om is open and since X is complete. There exists a neighbourhood U of x. f (x . y). then for each n. Hence f is continuous at x. f (x. y) = 2−n σn (x. f (x. y)] < ε and τ [f (x. Let E be the set of points at which f is continuous. y0 ) < δ implies τ [f (x. f (x. τ ) that is continuous in each variable. y0 ) < 1/n. Thus τ [f (x . Hence X = n Fm. Also. f (x.1/n . x1 ) < ε. Let E be a dense Gδ in X × Y . y0 )] < 2ε+1/m. G is a dense Gδ by Baire’s theorem. y0 ) ≤ 1/n and any x ∈ U . if x ∈ Fq \ Fq for some q ∈ Q. y0 )] ≤ 1/m for x ∈ U by continuity. y ∈ On } is a dense open subset of Y for each x ∈ Gn . f (x . x ) < δ for k ≥ N . x ) < δ implies τ [f (x. 42c. y) < δ and f (y) < q so f (x) > lim f (y). Let E ⊂ X × Y be the set of points at which f is continuous. E c is of the ﬁrst category and E is residual. y . Let (X. y0 )] < ε. Then x1 . y). y). *43. By Proposition 31. Let En = z∈S Bz. Since each Om is open and dense. y ∈ X × Y . y0 ) = ρ(x1 . y0 )] ≤ 2/m for x ∈ U . f (xN .δz. Thus G × H = Gn × Hn = (Gn × Hn ) where each Gn × Hn is open in X × Y . Choose n > 1/δ.z > 0 such that f [Bz. We may assume that τ [f (x . Then τ ( x. there exists δn. f (x . If x ∈ G. Then G is still a countable intersection of dense open sets in X. y0 ∈ E and τ ( x1 . y0 ).53) and E is a Gδ set. σ ≥ 0.f. y). f (x. Let x be a point of closure of Fm. y ∈ E} = {y : x. y ∈ E. Hence E is a dense Gδ in X 7. For each k and any y with σ(y. x0 ) < ε. Then E = n En (c. Let G = Om . Also let G ⊂ X and H ⊂ Y be dense Gδ ’s. y0 )] ≤ 2/m < ε for p ∈ U . Let σ = 2−n σn . Then σ(x.n . y→x y→x ◦ ◦ Let Fq = {x : f (x) ≥ q}. Let Om = n Fm. f (x . Let X and Y be complete metric spaces. there exists δ > 0 such that ρ(x. Suppose f is discontinuous at x. Then E = On where each On is a dense open set in X × Y . there exists a dense Gδ set Gn ⊂ X such that {y : x. Since G is dense. y) = 0 for all n if and only if x = y.n for some n. y) = 0 if and only if σn (x. y0 )] ≤ 1/m.n is closed. x0 . y ∈ On } is a dense Gδ for each x ∈ G. y). f (x . y0 ). f (xN . y0 ) ≤ 1/n. y0 )] ≤ 2/m for all p ∈ U . x0 ) < ε/2 and σ(y. Let G = Gn . Hence G × H is a Gδ in X × Y . y0 for each x ∈ G. there exists a dense Gδ ⊂ X1 × · · · × Xp on which f is continuous. y). Om is also dense. then f (x) ≥ q and ◦ for any δ > 0. Since ε is arbitrary. By part (c). there exists y with ρ(x. there exists a dense Gδ set G ⊂ X such that f is continuous at x. By part (b). Then for any y with σ(y. Conversely. Then E is a Gδ . f (x. By a similar argument as in part (d). y0 )]+τ [f (xN . Then there exists N such that ρ(xk . y0 )] < 1/m. f (x .space (Z. 42a. τ [f (xk . If f is continuous at z ∈ X × Y . ◦ 41b. τ [f (x. Hence there is a neighbourhood U of x.n . Then x ∈ Fm. there exists x1 ∈ G such that ρ(x0 . 41e.n = {x ∈ X : τ [f (x. choose m > 2/ε. y0 ). y). Since 39 . σ(x. y→x ◦ Since each Fq \ Fq is nowhere dense.7. Q2.n so Fm. Thus E c = q∈Q Fq \ Fq . Then f is not lower semicontinuous at x so f (x) > lim f (y). *42d. y0 )] ≤ τ [f (x . Take x0 . y). Then G = Gn and H = Hn where each Gn and Hn is open. y0 )] < 1/m so x ∈ Fm.n . we may regard the product as (X1 × · · · × Xp−1 ) × Xn where X1 × · · · × Xp−1 is a complete metric space. Then τ [f (x . y) ≤ 2−n = 1 for all x.1/n /2 . y0 )] ≤ 1/m. y0 ) < ε/2. y0 such that τ [f (p). For any x ∈ X and m. y0 in X × Y such that τ [f (p). There exists q ∈ Q such that f (x) ≥ q > lim f (y).δn. Fix y0 ∈ Y . Given a ﬁnite product X1 × · · · × Xp of complete metric spaces.n . y). Let ◦ x ∈ Om . Given ε > 0 and x. Let f : X1 × · · · × Xp → Z be a mapping into a metric space Z that is continuous in each variable. x0 . Then x ∈ Fq \ Fq . There is a sequence xk in Fm. ρ) be a complete metric space and f : X → R be an upper semicontinuous function. y0 ) < ε. For any y with σ(y. Hence G × H is dense in X × Y . Thus there exists a neighbourhood U of x such that U ⊂ Fm. y). Thus G is a dense Gδ set in X such that {y : x. y). Thus E is dense in X × Y . for each n.n . Given ε > 0. 41c.n converging to x . y0 . τ [f (x . 41d. τ [f (x . y0 )] ≤ 1/m for all y with σ(y. *42b. there exist x0 ∈ G and y0 ∈ H such that ρ(x. Since σn ≥ 0 for all n. y)] + τ [f (xN .z ] ⊂ Bf (z). y0 ∈ X × Y and let ε > 0 be given. Set Fm. y0 for each x ∈ G.9 Absolute Gδ ’s 44. f (xk . then x ∈ Om for each m. f (x. there exists δ > 0 such that σ(y. y0 ) ≤ 1/n}. y0 )] ≤ 1/m and x ∈ Fm.
Also. k2 . repeat the process with ε = 1/4.kN (x) so σ(x. Then X = {xn } ⊂ Un so X = Un . kn (y) ) < δ... Hence σ is equivalent to ρ on E. 1) whose complement is dense is uniformly homeomorphic to Nω . . getting intervals Ij. y) = 0 and x = y.. any dense Gδ E in (0. there exists δn > 0 such that σn (x.. Starting from E. the diameter of Ik1 . and (Y. locally compact metric space. Given k=1 ε > 0. Thus σ(f (xk ). g). Hence E is homeomorphic to the set of irrationals in (0.kn is less than 2−n for each n. there exists N such that σ(fN (xk ). Let fn be a sequence of continuous functions from X to a metric space Y which converge to a function f uniformly on each compact subset K of X. g(un ) and h(un ) converge to f (u) and g(u) respectively. let I1 . y) = σ(x. Hence g ≡ h. If each σn is uniformly equivalent to ρ... y) < δn implies ρ(x. Then x ∈ EIk1 and there is a unique integer k2 such that x ∈ Ik1 . Let {xn } be a countable dense subset of X.kn for all n... Given a sequence of integers k1 . x) for all n.. y) < ε. . . σ(f (x). Let X be a separable. y) < δ. h) + σn (h. Thus ∗ ∗ ∗ σ (f. y) < ε. Hence the correspondence between Nω and E given by parts (b) and (c) is a uniform homeomorphism between (Nω . . Let X be a metric space. f ). g) ≥ 0. . Thus σ is a metric on E. Given ε > 0. which is in turn homeomorphic to the set of irrationals in (0.. When σ(x. . Let τ be the product metric on Nω . 46b. Then u ∈ A so there is a sequence un in A converging to u. . 49. . Since σn (f. g) ≥ 0 for all n.10 The AscoliArzel´ Theorem a 47.h(x))+σ(h(x). g) = 2−n σn (f.h(x))+σ(h(x). . y) < ε/2 for n = 1. *46a. Since g and h are continuous on B. . . . y ∈ Ik1 (x). Let u ∈ B. choose N such that 2−N < ε. N . If σ(x.g(x)) . Let x ∈ X and let xk be a sequence in X converging to x.g(x)) ∗ ∗ ∗ 48b. Hence the set of functions from X into Y becomes a metric space under σ ∗ . Let F be an equicontinuous family of functions from X to Y . g) where σn (f. . kn (x) = kn (y) for n = 1. Then Un ⊂ x∈Un Vx and since Un is compact. Thus σ(x. then σ1 (x. σ ∗ (f.g(x)) ] ≤ σ(f (x). ¯ 45. y) < ε/2 and σ(x. . 46c. Suppose g(u) = h(u) for all u ∈ A.g(x)) 1 ∗ all n. y) < δn implies σn (x. there is a ﬁnite number of Vx ’s covering Un . 1). g) = supOn 1+σ(f (x). g) = σ ∗ (g. 46d. Let A ⊂ B ⊂ A be subsets of a metric space and let g and h be continuous maps of B into a metric ¯ space X. 1) by the continued fraction expansion.kn . τ ) and (E. σ) any metric space. σn (f. kn (y) ) ≤ ∞ −i ∗ −N < ε.g(x)) = supOn [1 − 1+σ(f (x). I2 . y ∈ Ik1 . g) = σn (g. 7. . Continuing. there is an open set Vx containing x with Vx compact. there exists N such that σ(fN (x ). y) = σn (y. ... z) + σ(z. f (x)) ≤ σ(f (xk ). Since σn (f. σ(x. f (x)) < ε for k ≥ N .1 . g). . . δN /2 ). By parts (a)(d). fN (x)) + σ(fN (x). and let F+ be the family of all pointwise limits of functions in F. fN (x)) < ε/3 for k ≥ N . Also.g(x)) ] = supOn 1+σ(f (x). y ∈ E such that x. then σ is uniformly equivalent to ρ. x). Given x ∈ E. .h(x))+σ(h(x). If ρ(x. For EIj = E ∩ Ij .σn (x. suppose there are x. k2 . 48a. f (x )) < ε/3 for any x ∈ K. z) + −n 2 σn (z. fN (xk )) + σ(fN (xk ). y) < 2−n for all n so σ(x. .g(x)) 1 ∗ ∗ supOn [1 − 1+σ(f (x). σ(x. For each n. be disjoint open intervals satisfying the conditions of part (a) with ε = 1/2.g(x)) ≤ σn (f. Let ε > 0 and ﬁx x ∈ E. Ij. But g(un ) = h(un ) for all n so g(u) = h(u). N so n=1 2−n σn (x. ... y) = 2−n σn (x. y) < δ. σ (f. . For each n.l with ε = 2−n . Thus there is a unique sequence of integers k1 . at the nth stage getting intervals Ij.. y) ≤ 2−n σn (x.. h) + σ (h. g) = supOn 1+σ(f (x). that is of f for which there is a sequence fn from F such that 40 . For each n. Hence f is continuous on X.. There exists N such that n=N +1 2 N N min(δ1 /2. *46e. then σn (x.. kn (y)) ≤ 2 Then x. g) ≤ 1 for all ∗ ∗ ∗ ∗ ∗ n. y) < ε/2..2 . there is an open set Un such that xn ∈ Un and Un is compact. . Let On be the union of the ﬁnite number of Vx ’s. When τ ( kn (x) . y) < δ. i=N +1 2 ρ (kn (x). j) = δij .k. By construction. . N so τ ( kn (x) . Then K = {xk }∞ ∪ {x} is a compact subset of X.. .. f ) for σ(f (x). such that for each n we have x ∈ Ik1 .k2 and so on. Let δ = and ρ(x. . y) < 2−N < ε. σ (f. σ). kn (x) = kn (y) for n = 1. . . y) < 2δ ≤ δ1 so ρ(x. y) = σ(y. there is a unique integer k1 such that x ∈ Ik1 . Let Nω be the space of inﬁnite sequences of integers and make N into a metric space by setting ρ(i. Then X = On . For each n and any x ∈ Un . g) ≤ σ (f.. Let δ < 2−N . y) < ε/2 ∞ −n < ε/2. y). g) ≤ 1 < ∞ and since σn (f. Let σ ∗ (f.
then there is a sequence fn from F such that f (x) = lim fn (x) so there is an N such that σ(fN (x). Given a set X. By an argument similar to that in part (a). A = x∈A Bx. Hence F is a sequentially compact. 50. For each z ∈ ∆. On the other hand. ¯ ¯ ¯ ¯ A ∪ B is a closed set containing A ∪ B so A ∪ B ⊂ A ∪ B. If X has more than one point. Hence the only continuous mappings from X into R are the constant functions. ¯ ¯ If F is closed. Take c ∈ R. f (y)) ≤ σ(f (x). Then σ(f (x). Let U be an open ball of radius min(z . If x ∈ E. any sequence fn in F has a subsequence that converges uniformly a to a function f on each compact subset of ∆. then f −1 [I] is either ∅ or X for any open interval I. Thus the associated topological space is discrete. F is closed. By the AscoliArzel´ Theorem. Let f ∈ F and ﬁx z ∈ ∆. y) = 0 if x = y and ρ(x. Then f −1 [c] = f −1 [(c − ε. f −1 [I] = ∅ if c ∈ I and f −1 [I] = X if c ∈ I. Then O = m Em is a dense open subset of ∆ and fn is locally bounded on O. In particular. Since F c is an open set containing x. E is the intersection of all closed sets containing E so E ⊂ E. A ⊂ A ∪ B and B ⊂ A ∪ B ¯ ¯ ¯ ¯ ¯ ¯ so A ⊂ A ∪ B and B ⊂ A ∪ B. f (x) − f (y) ≤ x − yα for all x. Thus f must be a constant function. Let 0 < α ≤ 1 and let F = {f : f α ≤ 1}. 1].1/2 so A is open. f (y)) < ε/3. 1c. Thus A ∪ B ⊂ A ∪ B. any two distinct points can be enclosed in disjoint open sets. a each sequence fn in F has a subsequence that converges pointwise to a continuous function. Let Em = n Em.1 Topological Spaces Fundamental notions 1a. 1] and each f ∈ F is bounded. Since any open set of real numbers is a countable union / of disjoint open intervals. Since w−z w−z 2πi f (z) ≤ 1 for all f ∈ F and all z ∈ ∆. it follows that f is continuous. Hence all mappings of X into R are continuous. f (w) dw 1 f (z) − f (z ) = 2πi C [ f (w) dw − f (w) dw ] = z−z C (w−z)(w−z ) by Cauchy’s integral formula. Hence O ∩ E = ∅ and x is a point of closure of E. E ◦◦ is 41 . Thus f ∈ C[0. Then U ⊂ ∆. E is ¯ ¯ ¯ one of which is E itself. *51a. Hence x ∈ E. Given x ∈ X and ε > 0. then x is not in some closed set F / ¯ ¯ containing E so x ∈ F c . Hence F+ is also an equicontinuous family of functions. ¯ ¯ ¯ ¯ ¯ 2. E ◦ is the union of all open sets contained in E so E ◦ ⊂ E. fn is equicontinuous on O so there is a subsequence that converges uniformly to a function f on each compact subset of O. *51c. Then for any set A ⊂ X. If f is a continuous mapping of X into R. ¯ ⊂ F . y) = 1 otherwise. then there is no metric on X such that the associated topological space is trivial because in a metric space. Let C be the circle of radius r/2 centred at z . For every z within r/4 of z . F = F . then x is in every closed set containing E. Hence A ∪ B ⊂ A ∪ B. F is an equicontinuous family of functions on the separable space [0. then for any open interval I. ¯ ¯ ¯ the intersection of all closed sets containing E. c + ε)] is either ∅ or X. there is an open set O containing x such that σ(f (x). f (y)) < ε/3 for all y ∈ O and f ∈ F. there is an Mz such that fn (z) ≤ Mz for all n.n where Em. Let F be the family of functions that are holomorphic on the unit disk ∆ = {z : z < 1} with f (z) ≤ 1. F c ∩ E = ∅. Conversely. It follows that F is equicontinuous. f (z) − f (z ) ≤ 4r−1 z − z  for all f ∈ F. 8 8. f (y)) < ε. and thus compact. Also. If f ∈ F. then E ⊂ Oc . fN (x)) + σ(fN (x). 1b. f is holomorphic on O. suppose x is a point of closure of E. deﬁne ρ on X × X by ρ(x. By the AscoliArzel´ Theorem. Furthermore. 1]. Let f be a mapping of X into R. Conversely. Since f −1 [O] ⊂ X for any open set O ⊂ R. Since Oc is closed. fN (y)) + σ(fN (y). Contradiction. then F is a closed set containing F so F ¯ . subset of C[0. f is continuous. *51b. Now if f ∈ F+ and y ∈ O. Let fn be a sequence of holomorphic functions on ∆ such that fn (z) → f (z) for all z ∈ ∆. Thus E = E. Furthermore f is holomorphic on ∆. Also. E ◦◦ ⊂ E ◦ . so E ⊂ E. 1 − z ) centred at z . If O ∩ E = ∅.n = {z : fn (z) ≤ m}.f (x) = lim fn (x) for each x ∈ X. Let X be a space with a discrete topology. then since F is closed. f (x)) < ε/3 and σ(fN (y). x ∈ Oc . y. ¯ Conversely. if f (x) = c for all x ∈ X. E ⊂ E. Then Em is closed and ∆ = m Em . Contradiction. if F = F ¯ If x ∈ E. Since we always have F ⊂ F . Also. then max f (x) + sup f (x) − f (y)/x − yα ≤ 1. On the other hand. Let X be a space with a trivial topology. There is an Em that is not nowhere dense so it has ◦ nonempty interior. Let O be an open set containing x. On the other hand.
the union of all open sets contained in E ◦ , one of which is E ◦ itself, so E ◦ ⊂ E ◦◦ . Thus E ◦◦ = E ◦ . A◦ ∩ B ◦ is an open set contained in A ∩ B so A◦ ∩ B ◦ ⊂ (A ∩ B)◦ . On the other hand, A ∩ B ⊂ A and A ∩ B ⊂ B so (A ∩ B)◦ ⊂ A◦ and (A ∩ B)◦ ⊂ B ◦ . Thus (A ∩ B)◦ ⊂ A◦ ∩ B ◦ . Hence (A ∩ B)◦ = A◦ ∩ B ◦ . If x ∈ E ◦ , then x is in some open set O contained in E so x is an interior point of E. Conversely, if x is an interior point of E, then x ∈ O ⊂ E for some open set O so x ∈ E ◦ . ¯ ¯ If x ∈ (E c )◦ , then x is in some open set O ⊂ E c so x ∈ Oc ⊃ E and thus x ∈ E c . Conversely, if x ∈ E c , / c c c ◦ c ◦ ¯c. then x is not in some closed set F ⊃ E so x ∈ F ⊂ E and thus x ∈ (E ) . Hence (E ) = E 3. Suppose A ⊂ X is open. Then given x ∈ A, x ∈ O ⊂ A with O = A. Conversely, suppose that given x ∈ A, there is an open set Ox such that x ∈ Ox ⊂ A. Then A = x∈A Ox , which is open. 4. Let f be a mapping of X into Y . Suppose f is continuous. For any closed set F , f −1 [F c ] is open. But f −1 [F c ] = (f −1 [F ])c so f −1 [F ] is closed. Conversely, suppose the inverse image of every closed set is closed. For any open set O, f −1 [Oc ] is closed. But f −1 [Oc ] = (f −1 [O])c so f −1 [O] is open. Thus f is continuous. 5. Suppose f is a continuous mapping of X into Y and g is a continuous mapping of Y into Z. For any open set O ⊂ Z, g −1 [O] is open in Y so f −1 [g −1 [O]] is open in X. But f −1 [g −1 [O]] = (g ◦ f )−1 [O]. Hence g ◦ f is a continuous mapping of X into Z. 6. Let f and g be two realvalued continuous functions on X and let x ∈ X. Given ε > 0, there is an open set O containing x such that f (x) − f (y) < ε/2 and g(x) − g(y) < ε/2 whenever y ∈ O. Then (f + g)(x) − (f + g)(y) = f (x) − f (y) + g(x) − g(y) ≤ f (x) − f (y) + g(x) − g(y) < ε whenever y ∈ O. Thus f + g is continuous at x. There is an open set O such that g(x) − g(y) < ε/2f (x) and f (x)−f (y) < ε/2 max(g(x)−ε/2f (x), g(x)+ε/2f (x)) whenever y ∈ O . Then (f g)(x)−(f g)(y) = f (x)g(x)−f (x)g(y)+f (x)g(y)−f (y)g(y) ≤ f (x)g(x)−g(y)+f (x)−f (y)g(y) < ε whenever y ∈ O . Thus f g is continuous at x. 7a. Let F be a closed subset of a topological space and xn a sequence of points from F . If x is a cluster point of xn , then for any open set O containing x and any N , there exists n ≥ N such that xn ∈ O. Suppose x ∈ F . Then x is in the open set F c so there exists xn ∈ F c . Contradiction. Hence / x ∈ F. 7b. Suppose f is continuous and x = lim xn . For any open set U containing f (x), there is an open set O containing x such that f [O] ⊂ U . There also exists N such that xn ∈ O for n ≥ N so f (xn ) ∈ U for n ≥ N . Hence f (x) = lim f (xn ). 7c. Suppose f is continuous and x is a cluster point of xn . For any open set U containing f (x), there is an open set O containing x such that f [O] ⊂ U . For any N , there exists n ≥ N such that xn ∈ O so f (xn ) ∈ U . Hence f (x) is a cluster point of f (xn ) . *8a. Let E be an arbitrary set in a topological space X. (E ◦ )◦ is the intersection of all closed sets containing (E ◦ )◦ , one of which is E ◦ , so (E ◦ )◦ ⊂ E ◦ . On the other hand, E ◦ is the intersection of all closed sets containing E ◦ , one of which is (E ◦ )◦ since E ◦ = E ◦◦ ⊂ (E ◦ )◦ , so E ◦ ⊂ (E ◦ )◦ . Hence (E ◦ )◦ = E ◦ . ¯ ¯ Now since E = E and (E c )c = E, new sets can only be obtained if the operations are performed ¯ alternately. Also, (E)c = (E c )◦ so (E)c
c c
=
(E)c
◦
=
(E c )◦
◦
= (E c )◦ = (E)c . Hence the
c
¯ distinct sets that can be obtained are at most E, E, (E)c , (E)c , (E)c E c , (E c )c , (E c )c , (E c )c
c
, (E)c
c
,
(E)c
c
c
, Ec,
, (E c )c
c
,
(E c )c
c
c
.
*8b. Let E = {(0, 0)} ∪ {z : 1 < z < 2} ∪ {z : 2 < z < 3} ∪ {z : 4 < z < 5 and z = (p, q) where p, q ∈ Q} ⊂ R2 . Then E gives 14 diﬀerent sets by repeated use of complementation and closure. 9. Let f be a function from a topological space X to a topological space Y . Suppose f is continuous and let x ∈ X. For any open set U containing f (x), f −1 [U ] is open and x ∈ f −1 [U ]. If y ∈ f −1 [U ], then f (y) ∈ U . Thus f is continuous at x. Conversely, suppose f is not continuous. There exists an open set O such that f −1 [O] is not open. By Q3, there exists x ∈ f −1 [O] such that U is not a subset of f −1 [O] for any open set U containing x. Then O is an open set containing f (x) and f [U ] is not a subset of O 42
for any open set U containing x. Hence f is not continuous at x. 10a. Let the subset A of a topological space X be the union of two sets A1 and A2 both of which are closed (resp. both of which are open). Let f be a map of A into a topological space Y such that the restrictions f A1 and f A2 are each continuous. For any closed (resp. open) set E ⊂ Y , f −1 [E] = (f −1 [E] ∩ A1 ) ∪ (f −1 [E] ∩ A2 ) = f −1 [E] ∪ f −1 [E] = (A1 ∩ F ) ∪ (A2 ∩ F ) for some closed A1 A2 (resp. open) sets F, F ⊂ A. Hence f −1 [E] is closed (resp. open) and f is continuous. 10b. Consider the subset (0, 2] of R. Then (0, 2] is the union of (0, 1), which is open, and [1, 2], which is closed. Let f be the map deﬁned by f (x) = x on (0, 1) and f (x) = x + 1 on [1, 2]. Then f (0,1) and f [1,2] are each continuous but f is not continuous (at 1). *10c. Suppose A = A1 ∪ A2 , A1 \ A2 ∩ (A2 \ A1 ) = ∅ and A2 \ A1 ∩ (A1 \ A2 ) = ∅. Let f be a map of A into a topological space Y such that the restrictions f A1 and f A2 are each continuous. Let F be a closed subset of Y and let F = f −1 [F ] ⊂ A. Then F = F ∩ (A1 \ A2 ∪ (A1 ∩ A2 ) ∪ A2 \ A1 ) = F ∩ (A1 \ A2 ) ∪ F ∩ A1 ∩ A2 ∪ F ∩ (A2 \ A1 ). Note that F ∩ (A1 \ A2 ) ⊂ A1 and F ∩ (A2 \ A1 ) ⊂ A2 . Thus f [F ] = f [F ∩ (A1 \ A2 )]∪f [F ∩ A1 ∩ A2 ]∪f [F ∩ (A2 \ A1 )] = f [F ] ⊂ F and F ⊂ f −1 [F ] = F . Hence F = f −1 [F ] is closed and f is continuous.
8.2
Bases and countability
¯ 11a. Let B be a base for the topological space (X, T ). If x ∈ E, then for every B ∈ B with x ∈ B, since B is open, B ∩ E = ∅. i.e. there is a y ∈ B ∩ E. Conversely, suppose that for every B ∈ B with x ∈ B there is a y ∈ B ∩ E. For each open set O in X with x ∈ O, there is a B ∈ B with x ∈ B ⊂ O. ¯ Then there is a y ∈ B ∩ E ⊂ O ∩ E. Thus x ∈ E. 11b. Let X satisfy the ﬁrst axiom of countability. If there is a sequence xn in E converging to x, then ¯ for any open set O containing x, there exists N such that xn ∈ O for n ≥ N . Thus O ∩ E = ∅ so x ∈ E. ¯ There is a countable base Bn at x. By considering B1 , B1 ∩ B2 , B1 ∩ B2 ∩ Conversely, suppose x ∈ E. B3 , . . ., we may assume B1 ⊃ B2 ⊃ B3 · · · . Now each Bn contains an xn ∈ E. For any open set O containing x, O ⊃ BN ⊃ BN +1 ⊃ · · · for some N so xn ∈ O for n ≥ N and xn is a sequence in E converging to x. 11c. Let X satisfy the ﬁrst axiom of countability and let xn be a sequence from X. If xn has a subsequence converging to x, it follows from the deﬁnition that x is a cluster point of xn . Conversely, suppose x is a cluster point of xn . There is a countable base Bn at x. We may assume B1 ⊃ B2 ⊃ · · · . Since B1 is open and contains x, xn1 ∈ B1 for some n1 ≥ 1. Suppose xn1 , . . . , xnk have been chosen. Then there exists nk+1 ≥ nk such that xnk+1 ∈ Bk+1 . For any open set O containing x, O ⊃ BN ⊃ BN +1 · · · for some N so xnk ∈ O for k ≥ N and the subsequence xnk converges to x. 12a. See Q9. 12b. Let Bx be a base at x and Cy be a base at y = f (x). Suppose f is continuous at x. For each C ∈ Cy , since C is open and contains f (x), there is an open set U containing x such that U ⊂ f −1 [C]. Then there exists B ∈ Bx such that B ⊂ U ⊂ f −1 [C]. Conversely, suppose that for each C ∈ Cy there exists B ∈ Bx such that B ⊂ f −1 [C]. For any open set O containing f (x), there exists C ∈ Cy such that C ⊂ O. Then there exists B ∈ Bx such that B ⊂ f −1 [C] ⊂ f −1 [O]. Hence f is continuous at x. 13. Let C be any collection of subsets of X. Let B consist of X and all ﬁnite intersections of sets in C. By Proposition 5, B is a base for some topology S on X. For any topology T on X containing C, we have B ⊂ T . Since any set in S is a union of sets in B, S ⊂ T . Hence S is the weakest topology on X containing C. 14. Let X be an uncountable set of points and let T consist of the empty set and all subsets of X whose complement is ﬁnite. Then X ∈ T since X c = ∅, which is ﬁnite. If O1 , O2 ∈ T , then we may c c assume O1 , O2 = ∅ so (O1 ∩ O2 )c = O1 ∪ O2 , which is ﬁnite. Thus O1 ∩ O2 ∈ T . If Oα ∈ T , then c c ( Oα ) = Oα , which is ﬁnite. Thus Oα ∈ T . Hence T is a topology for X. Suppose (X, T ) satisﬁes the ﬁrst axiom of countability. Take x ∈ X. There is a countable base Bn at c c c x. Note that Bn is ﬁnite for all n so Bn is countable. Thus there exists y ∈ ( Bn )c = Bn with c c y = x since ( Bn ) is uncountable. Now X \ {y} is open and contains x so there exists BN ⊂ X \ {y}. But y ∈ BN ⊂ X \ {y}. Contradiction. Hence (X, T ) does not satisfy the ﬁrst axiom of countability.
43
15. Let X be the set of real numbers and let B be the set of all intervals of the form [a, b). For any x ∈ X, we have x ∈ [x, x + 1). If x ∈ [a, b) ∩ [c, d), then x ∈ [c, b) (or [a, d)). Hence B is a base of a topology T (the halfopen interval topology) for X. For any x ∈ X, let Bx,q = [x, x + q) where q is a positive rational number. Then {Bx,q } is a countable base at x. Thus (X, T) satisﬁes the ﬁrst axiom of countability. Let B be a base for (X, T). For any x ∈ X, there exist Bx ∈ B such that x ∈ Bx ⊂ [x, x + 1). If x = y, then Bx = By . Thus {Bx } is uncountable and (X, T) does not satisfy the second axiom of countability. For any x ∈ X and any open set O containing x, there is an interval [a, b) ⊂ O containing x. Furthermore, there is a rational q ∈ [a, b). Thus the rationals are dense in (X, T ). Now (X, T ) is not metrizable because it is separable and a separable metric space satisﬁes the second axiom of countability. 16. Suppose X is second countable. Let B be a countable base for the topology on X and let U be an open cover of X. For each x ∈ X, there exists Ux ∈ U, such that x ∈ Ux . Then there exists Bx ∈ B such that x ∈ Bx ⊂ Ux . Thus X = x∈X Bx . Since B is countable, we may choose countably many Ux ’s to form a countable subcover of U. Hence X is Lindel¨f. o 17a. Let X1 = N × N and take X = X1 ∪ {ω}. For each sequence s = mk of natural numbers deﬁne Bs,n = {ω}∪{ j, k : j ≥ mk if k ≥ n}. Clearly, any x ∈ X is in some Bs,n or { j, k }. If x ∈ Bs,n ∩Bs ,n , then either x = ω or x = j, k . If x = j, k , then { j, k } ⊂ Bs,n ∩ Bs n . If x = ω, let s = mk be the sequence where mk = max(mk , mk ) for all k and let n = min(n, n ). Then ω ∈ Bs ,n ⊂ Bs,n ∩ Bs ,n . Other possible intersections are similarly dealt with. Hence the sets Bs,n together with the sets { j, k } form a base for a topology on X. *17b. For any open set O containing ω, there is some Bs,n such that ω ∈ Bs,n ⊂ O. Since Bs,n contains some j, k ∈ X1 , so does O. Thus ω is a point of closure of X1 . 17c. X is countable so it is separable. If it satisﬁes the ﬁrst axiom of countability, then since ω is a point of closure of X1 , by Q11b, there is a sequence from X1 converging to ω and by Q11c, ω is a cluster point of that sequence, contradicting part (b). Hence X does not satisfy the ﬁrst axiom of countability and thus it also does not satisfy the second axiom of countability. 17d. X is countable so it is Lindel¨f. o
8.3
The separation axioms and continuous realvalued functions
18a. Given two distinct points x, y in a metric space (X, ρ), let r = ρ(x, y)/2. Then the open balls Bx,r and By,r are disjoint open sets with x ∈ Bx,r and y ∈ By,r . Hence every metric space is Hausdorﬀ. 18b. Given two disjoint closed sets F1 , F2 in a metric space (X, ρ), let O1 = {x : ρ(x, F1 ) < ρ(x, F2 )} and let O2 = {x : ρ(x, F2 ) < ρ(x, F1 )}. The sets O1 and O2 are open since the functions f (x) = ρ(x, F1 ) and g(x) = ρ(x, F2 ) are continuous. Thus O1 and O2 are disjoint open sets with F1 ⊂ O1 and F2 ⊂ O2 . Hence every metric space is normal. 19. Let X consist of [0, 1] and an element 0 . Consider the sets (α, β), [0, β), (α, 1] and {0 }∪(0, β) where α, β ∈ [0, 1]. Clearly, any x ∈ [0, 1] belongs to one of the sets. Note that (α, β) ⊂ [0, β) ∩ (α, 1]. If x ∈ (α, β)∩[0, β ), then x ∈ (α, min(β, β )) ⊂ (α, β)∩[0, β ). If x ∈ (α, β)∩(α , 1], then x ∈ (max(α, α ), β) ⊂ (α, β) ∩ (α , 1]. If x ∈ (α, β) ∩ ({0 } ∪ (0, β )), then x ∈ (α, min(β, β )) ⊂ (α, β) ∩ ({0 } ∪ (0, β )). If x ∈ [0, β) ∩ ({0 } ∪ (0, β )), then x ∈ (0, min(β, β )) ⊂ [0, β) ∩ ({0 } ∪ (0, β )). If x ∈ (α, 1] ∩ ({0 } ∪ (0, β)), then x ∈ (α, β) ⊂ (α, 1] ∩ ({0 } ∪ (0, β)). Hence the sets form a base for a topology on X. Given two distinct points x, y ∈ X, if neither of them is 0’, then [0, (x + y)/2) is an open set containing one but not the other. If y = 0 , then {0 } ∪ (0, x/2) is an open set containing y but not x. Hence X is T1 . Consider the distinct points 0 and 0’. If O is an open set containing 0 and O is an open set containing 0’, then 0 ∈ [0, β) ⊂ O and 0 ∈ {0 } ∪ (0, β ) ⊂ O for some β, β . Since [0, β) ∩ (0, β ) = ∅, O ∩ O = ∅. Hence X is not Hausdorﬀ. 20. Let f be a realvalued function on a topological space X. Suppose f is continuous. Then for any real number a, {x : f (x) < a} = f −1 [(−∞, a)] and {x : f (x) > a} = f −1 [(a, ∞)]. Since (−∞, a) and (a, ∞) are open, the sets {x : f (x) < a} and {x : f (x) > a} are open. Conversely, suppose the sets {x : f (x) < a} and {x : f (x) > a} are open for any real number a. Since any open set O ⊂ R is a union of open intervals and any open interval is an intersection of at most two of the sets, f −1 [O] is open so f is continuous.
44
F ⊂ U and G ⊂ U c . 24c. then y ∈ Ur2 so f (y) ≤ r2 < d. If y ∈ O. then either r ≤ ri . Since U0 ⊂ U1 . {x : (f ∧ g)(x) < a} = {x : f (x) < a} ∪ {x : g(x) < a}. Note that Uri ⊂ Urj . such that F ⊂ Ur ⊂ O and Ur ⊂ Us for r < s.ε/3 . Since f (x) > r1 . B c is an open set containing A. For any real number a. we have a sequence {Ur } of open sets. and {x : f (x)g(x) > a} = q∈Q ({x : g(x) > q}) ∩ {x : f (x) > a/q}. Suppose f and g are continuous realvalued functions on a topological space X. By Urysohn’s Lemma. Let h = f /(1 + f ). 1) of the form r = p2−n in a sequence rn . or r ≥ rj . f (x)) < ε. Now U is disjoint ¯ from Oc since if y ∈ Oc . fN (x)) + ρ(fN (x). Let U1 = O. there exists an open set Ur1 such that U0 ⊂ Ur1 and Ur1 ⊂ U1 . there exists an open set U0 such that F ⊂ U0 and U0 ⊂ O = U1 . For any pair of disjoint closed sets A and B on X. Now rn+1 has an immediate predecessor ri and an immediate successor rj (under the usual order relation) in Pn+1 ∪ {0. Given x ∈ X and an open interval (c. Hence f is continuous. Thus f [O] ⊂ Bf (x). and {x : f (x) + g(x) > a} = q∈Q ({x : g(x) > q} ∩ {x : f (x) > a − q}). 24b. Let fn be a sequence of continuous functions from a topological space X to a metric space Y . x ∈ Ur ⊂ Ur2 for some r < r2 . which is open. Also. Since any open set in Y is a union of open balls. then x ∈ Ur for all r so f (x) = 0. fN (y)) + ρ(fN (y). Suppose fn converges uniformly to a function f . then V is an open set containing y that is disjoint from U . Let ε > 0 and let x ∈ X. f ≡ 0 on A and f ≡ 1 on (B c )c = B. There exists N such that ρ(fn (x). in which case Ur ⊂ Uri ⊂ Urn+1 . Since f (x) < r2 . and {x : (f ∨ g)(x) > a} = {x : f (x) > a} ∪ {x : g(x) > a}. Thus f [U ] ⊂ (c. suppose that given a closed set F and an open set O containing F . By induction. there is a continuous function h1 which is −1/3 on B and 1/3 on C while h1 (x) ≤ 1/3 for all x ∈ X. which is open. 1) of the form p = r2−n . and {x : (f ∧ g)(x) > a} = {x : f (x) > a} ∩ {x : g(x) > a}. choose rationals r1 and r2 of the form p2−n such that c < r1 < f (x) < r2 < d. Then 45 . Suppose X is normal. / 23d. If x ∈ Oc . Let B = {x : h(x) ≤ −1/3} and let C = {x : h(x) ≥ 1/3}. there exist disjoint open sets U and V such that F ⊂ U and Oc ⊂ V .ε . If x ∈ Ur1 . Since fN is continuous. Let F and G be disjoint closed sets. Now if r ∈ Pn . there is a continuous realvalued function f on X such that 0 ≤ f ≤ 1. {x : (f ∨ g)(x) < a} = {x : f (x) < a} ∩ {x : g(x) < a}. Consider the open set U = Ur2 \ Ur1 . {x : f (x) + g(x) < a} = q∈Q ({x : g(x) < q} ∩ {x : f (x) < a − q}). 23c. If y ∈ U . Thus there is an open set Urn+1 such that Uri ⊂ Urn+1 and Urn+1 ⊂ Urj . Equivalently. 24a. Let Pn be the set containing the ﬁrst n terms of the sequence. so f g is continuous. there is a continuous function hn+1 which is −2n /3n+1 on B and 2n /3n+1 on C while hn+1 (x) ≤ 2n /3n+1 for all x ∈ X. 22. Suppose open sets Ur have been deﬁned for rationals r in Pn such that Up ⊂ Uq whenever p < q. so f + g is continuous. Let B = {x : h(x) − i=1 hi (x) ≤ −2n /3n+1 } and n n n+1 let C = {x : h(x) − i=1 hi (x) ≥ 2 /3 }. which is open. x ∈ Ur1 . d). in which case Urn+1 ⊂ Urj ⊂ Ur . Clearly. so f ∧ g is continuous. Let {Ur } be the family constructed in part (b) with U1 = X. which is open. so f ∨ g is continuous. Thus x ∈ U . f (x)) < ε/3 for all n ≥ N and all x ∈ X. Suppose we have continuous functions hn on X such that hn (x) < 2n−1 /3n for all x ∈ X and n n h(x) − i=1 hi (x) < 2n /3n for all x ∈ A. X is normal.Since {x : f (x) ≥ a}c = {x : f (x) < a}. Let A be a closed subset of a normal topological space X and let f be a continuous realvalued function on A. Let X be a Hausdorﬀ space. Let f be the realvalued function on X deﬁned by f (x) = inf{r : x ∈ Ur }. Given a closed set F and an open set O ¯ containing F . which is open. 21. 1}. f (x)) ≤ ρ(f (y). f is continuous. which is open. If x ∈ F . Suppose X is normal. Let X be a Hausdorﬀ space. the set {x : f (x) > a} is open and the set {x : f (x) ≥ a} is closed. By Urysohn’s Lemma. 23a. then x ∈ Ur for any r < 1 so f (x) = 1. Arrange the rationals in (0. we also have f is continuous if and only if for any real number a. which is open. Since U and U c are disjoint open sets. Let F be a closed subset of a normal space contained in an open set O. Thus U ⊂ O. Conversely. then ρ(f (y). which is open. By the constructions in parts (b) and (c). / y ∈ Ur1 so f (y) ≥ r1 > c. Then h = f /(1 + f ) < 1. there is an open set O containing x such that fN [O] ⊂ BfN (x). Then h(x) − h1 (x) ≤ 2/3 for all x ∈ A. d) containing / f (x). Then F ⊂ Gc and there is an open ¯ ¯ ¯ set U such that F ⊂ U and U ⊂ Gc . By part (a). 0 ≤ f ≤ 1. one corresponding to each rational in (0. there is an open set U ¯ such that F ⊂ U and U ⊂ O. *23b. then x ∈ Ur for all r > r1 so f (x) ≤ r1 . (*) Proof of Urysohn’s Lemma. {x : f (x)g(x) < a} = q∈Q ({x : g(x) < q}) ∩ {x : f (x) < a/q}.
. Then x ∈ O1 ∩ Y and y ∈ O2 ∩ Y . m kn (x) − kn−1 (x) < 2n−1 /3n for all n and all x ∈ X. Then x. . d − f (x ))} ⊂ f −1 [O]. say x = x1 . 26. O2 ∈ T . i. Hence there is a function f ∈ F such that f (x) = f (y). y} of distinct points in X. d). d). Then f (u) ∈ O1 and g(u) ∈ O2 but f (u) = g(u) so O1 ∩ O2 = ∅. Hence Y is Hausdorﬀ. we may assume O is an open interval (c. . Hence the topology is Hausdorﬀ. there is a function f ∈ C(X) / such that f (x) = 1 and f ≡ 0 on F . . then km (x) − kn (x) =  i=n+1 hi (x) ≤ ∞ m i−1 i /3 < (2/3)n . Now h(x) − kn (x) < 2n /3n for all n and all x ∈ A so letting n → ∞. In Rn let B be the family of sets {x : p(x) = 0} where p is a polynomial in n variables. By Urysohn’s Lemma. k ≤ 1.e. since kn  ≤ 1 for each n. . 27. then f (x) ∈ (c. 25. we have v ∈ A.h(x) − i=1 hi (x) < 2n+1 /3n+1 for all x ∈ A. T ). n 24d. . Since f is continuous for each f ∈ F.e. Then each kn is continuous and h(x) − kn (x) < 2n /3n for all x ∈ A. By considering the constant polynomials. fn of functions in F}. g[U2 ] ⊂ O2 . . d − f (x ))}. Let kn = i=1 hi . there are disjoint open sets O1 and O2 in X such that x ∈ O1 and y ∈ O2 . Suppose that for each closed set F and each x ∈ F there is an f ∈ F such that f (x) = 1 and f ≡ 0 on F . . there are open sets U1 and U2 such that v ∈ U1 . Then there are disjoint open sets O1 and O2 such that f (v) ∈ O1 and g(v) ∈ O2 . (*) Proof of Tietze’s Extension Theorem. . . ∞)] are disjoint open sets. Set g = ϕk/(1 − ϕk). and some ﬁnite set f1 . Then p(x) = 0 and p(y) = 0. . Then there are sets Bx and By of the form {x : fi (x)−fi (z) < ε for some ε > 0. Let T be the family of all ﬁnite intersections O = B1 ∩ · · · ∩ Bk from B. Then for each O ∈ T / and each x ∈ O. . fn of functions in F} such that x ∈ Bx ⊂ Ox and y ∈ By ⊂ Oy . there is an f ∈ F such that f (x) = 1 and f ≡ 0 on Oc . x ∈ f −1 [O]. some y ∈ X. O open in R} as a base. . . 1/2)].e. Hence the sets of the form {x : fi (x) − fi (y) < ε for some ε > 0. x ∈ f −1 [(1/2. A and {x : k(x) = 1} are disjoint closed sets. Any two open sets are not disjoint since for any ﬁnite collection of polynomials there is an x that is not a root of any of the polynomials. h(x) − k(x) = 0 for all x ∈ A. Let X be a completely regular space. Then g is continuous and g = k/(1 − k) = h/(1 − h) = f on A. Given a closed set F and x ∈ F . Then x ∈ {x : f (x ) − f (x) < f (y) − f (x)/2} and y ∈ {x : f (x ) − f (y) < f (y) − f (x)/2} with the two sets being disjoint open sets. The weak topology on X generated by F has {f −1 [O] : f ∈ F. let p be the polynomial n 2 i=1 (Xi − xi ) . the weak topology generated by F is contained in T . Let F be a family of realvalued functions on a set X. . Now there exists u ∈ A with u ∈ U1 ∩ U2 . It is also clear that if O1 . Thus there is an open set containing y but not x. *29. . Consider the sets of the form {x : fi (x) − fi (y) < ε for some ε > 0. Contradiction. Then k(x) − kn (x) ≤ (2/3)n for all x ∈ X. y ∈ Rn . there are disjoint open sets Ox and Oy such that x ∈ Ox and y ∈ Oy . Conversely. xn and y = y1 . Hence T is not T2 . i. hn is uniformly summable to k and since each kn is continuous. 24e. Given two distinct points x. . Thus i=n+1 2 kn converges uniformly to k.e. 3/2)] ⊂ O. then O1 ∩ O2 ∈ T . Suppose f (v) = g(v). y ∈ Bx ∩ By . and some ﬁnite set f1 . For any v ∈ B. Hence T is T1 . ∞)]. Thus x ∈ {x : f (x) − f (x ) < min(f (x ) − c. and let f and g be two continuous maps of B into a ¯ Hausdorﬀ space X with f (u) = g(u) for all u ∈ A. v ∈ U2 .e. and the sets f −1 [(−∞. 24f. k is continuous. i. Let X be a Hausdorﬀ space and let Y be a subset of X. Hence X is regular. Hence f (v) = g(v). f ≡ g on B. Since k = h < 1 on A. some z ∈ X. with O1 ∩ Y and O2 ∩ Y being disjoint open sets in Y . Now if x ∈ f −1 [O] for some f ∈ F and O open in R. y} of distinct points in X there exists f ∈ F such that f (x) = f (y). x ∈ f −1 [(1/2. f [U1 ] ⊂ O1 . suppose that for each pair {x. . . . n+1 46 . k = h on A. Contradiction. If m > n. Thus O is in the weak topology generated by F. we see that ∅ and X are in T . some y ∈ X. Since f and g are continuous. 1/2)] and f −1 [(1/2. there is a continuous function ϕ on X which is 1 on A and 0 on {x : k(x) = 1}. Also. Given two distinct points x and y in Y . i. yn . Let A ⊂ B ⊂ A be subsets of a topological space. and some ﬁnite set f1 . Then F ⊂ f −1 [(−∞. If x ∈ {x : f (x) − f (x ) < min(f (x ) − c. Suppose this topology is Hausdorﬀ. Let F be a family of realvalued continuous functions on a topological space (X. Suppose f (x) = f (y) for all f ∈ F. fn of functions in F} is a base for the weak topology on X generated by F. Let k(x) = i=1 hi (x). Also. Now if Oα ∈ T . For any pair {x. i. 28. ¯ 30.
Take x ∈ O1 and y ∈ O2 . If d ∈ O. −1 ≤ y ≤ 1} and S = { x. G is arcwise connected. Now there is a path connecting z to x and there is a path connecting y to z. Let X be a locally connected space and let C be a component of X. d + ε) ⊂ O ⊂ I for some ε > 0 since O is open.y . d + ε) ⊂ O. y ∈ B 0. Hence O = I and I is connected. There exists z > d − ε such that (a. [a. 1] ⊂ O1 . *37. Clearly. then (d − ε. y) ⊂ O} < b. Then there exists tn ∈ (0. 34a. sin(1/x0 ) = y for some x0 in the interval. Hence d = b. Clearly sup{y : (a.4 Connectedness 32. y : x = 0. Then x(0) = 0 while x(t) > 0 and y(t) = sin(1/x(t)) for t > 0. Let G be a connected open set in Rn . x0 . If y ∈ H. Let f (t) = x(t).e. There is a path f connecting y to x. y : x = 0. y] ⊂ E. X = I ∪ S is connected. Suﬃciently small balls centred at 0. Let a = inf E and b = sup E. b] are also connected. there exists c > c such that (a. b) and let O be a subset of I that is both open and closed in I. contradicting the continuity of f . Let I = (a. Thus C is open. For y ∈ Bx. 1/2] and h(t) = f (2t − 1) for t ∈ [1/2. Hence E is an interval. Let X be as in Q35b. we have Cα ⊂ O1 . Now any other Cα has a point in common with Cα so by the same argument. Suppose O1 and O2 is a separation of X. y . Cα ⊂ O1 . then y ∈ By. y) such that z ∈ E. 47 . Suppose O1 and O2 is a separation of G. then (d − ε. z) ∪ (d − ε. Let 0. y ∈ E with x < y.δ ⊂ G. d + ε) = (a. Then E ∩ (−∞. Hence G is connected. E ⊂ [a. Hence I ⊂ S and by Q33. y(t) . f [0. Then p ∈ O1 or p ∈ O2 . Let X be an arcwise connected space. Since B is connected. For any pair Cα and Cα .31. y : y = sin(1/x). 1 1 Since sin 2 (4n + 1)π = 1 and sin 2 (4n + 3)π = −1. If z ∈ (a. [a. For each y ∈ H. *35b. If d ∈ O. there is a path g connecting z to y. Since A ⊂ B = O1 ∪ O2 and A is connected. Let G = Cα .δ ⊂ G for some δ > 0 and By. −1 ≤ y ≤ 1} ∪ { x. Omitted.δ . Suppose d = sup{y : (a. y : y = sin(1/x). Let I = { x. then there exists z ∈ E such that a ≤ z < z and there exists z ∈ E such that z < z ≤ b so that z ∈ E. y ∈ I. there exists δ > 0 such that By. Suppose x. b) ⊂ E ⊂ [a. Since [0. It follows from Q33 that intervals of the form (a. Thus for any c < b. z) ⊂ (a. 34b. c ) ⊂ O. Contradiction. y ] = x0 < ε so x0 . then there is a connected basic set B such that x ∈ B. y) ⊂ O} ≤ b. choose u with 0 < u < x(1/n) such that sin(1/u) = (−1)n . For each z ∈ By. We may assume that t = 0. Thus ¯ By. b]. z) ⊂ O. Now tn converges to 0 but y(tn ) does not converge. sin(1/x) takes on every value between 1 and 1 in the interval [2/(4n + 3)π. ¯ 33. Suppose p ∈ O1 . In particular. 1] is connected. Thus there is a path connecting y to x so y ∈ H and H is closed. For each n. Then there is a path f from 0. Since Cα is connected and p ∈ Cα . Thus (a. There is a continuous function f : [0. There exists δ > 0 such that Bx. 2/(4n + 1)π]. If x ∈ C. d + ε) ⊂ Oc for some ε > 0 since Oc is open. B ⊂ C. we may assume that A ⊂ O1 . But there exists / z > d − ε such that (d − ε. Hence X is connected. b]. Hence H = G. Since x is arbitrary. Suppose there exists z ∈ (x. Contradiction. c ∈ O. Hence X is not arcwise connected. z) ⊂ O. ∞) are a separation of E. Let A be a connected subset of a topological space and suppose that A ⊂ B ⊂ A. The set of t such that f (t) ∈ I is closed so it has a largest element. Contradiction. f (t) = (1 − t)x + ty is a path connecting x and y. contradicting O2 = ∅. 36. 1/n) such that x(tn ) = u. Thus H = ∅. Then y ∈ O1 ∩ O2 . Then (a. Thus [x. Thus G ⊂ O1 .δ ⊂ H and H is open. z) and E ∩ (z. b). Suppose O1 and O2 is a separation of B. there is a p ∈ Cα ∩ Cα . Then h given by h(t) = g(2t) for t ∈ [0. 0 do not contain connected open sets so X is not locally connected. 0 < x ≤ 1}. 1] under a continuous map. / contradicting the connectedness of E. 0 < x ≤ 1}. S is connected. 1] → X with f (0) = x and f (1) = y. (*) (Closed) topologist’s sine curve 35c.δ ⊂ G. Suppose X is arcwise connected. b]. 8. Let E be a connected subset of R having more than one point. Let x ∈ G and let H be the set of points in G that can be connected to x by a path. Then ¯ B = B ∩ A ⊂ B ∩ O1 = O1 . Given ε > 0.ε ∩ S. Let {Cα } be a collection of connected sets and suppose that any two of them have a point in common.δ contains a point z of H. b). Then ¯ ρ[ 0. Hence B is connected. Since S is the image of (0. 35a. 0 to some point of S. contradicting O2 = ∅.δ . choose n such that 1/2nπ < ε. Consider X = { x. 1] is connected so we may assume f [0. i. 1] is a path connecting z to x.
Then ◦ O = (O ∩ X1 ) ∪ (O ∩ X2 ) is open in X. x ∈ Xα and y ∈ Xβ for some α and β. 39a. ρ) be a metric space with an extended realvalued metric and Xα its parts. Suppose Z is Hausdorﬀ. y ∈ α Xα . Each of the sets f −1 [O] is open so f −1 [O] is open and f is continuous. then since Xα is Hausdorﬀ.2 is open in Xα and only ﬁnitely many of them are not Xα .2 for each α.ε × By. i. Given disjoint closed sets F1 . If x ∈ B1 ∩ B2 where Bi = α Oα.1 ∩ Oα.e. Hence X = X1 ∪ X2 . equivalence classes under the equivalence relation ρ(x. Conversely. Then it follows from the deﬁnition that X1 is open.5 Products and direct unions of topological spaces ◦ 38a.δ ⊂ V . *40a. it follows that Z is Hausdorﬀ. Now each Oα = Oα.1 ∩ Oα. 1}/ ∼ where ∼ is the smallest equivalence relation with x. f −1 [O] is open in Z so f −1 [O] ∩ Xα is open in Xα for each α. Conversely. x. y be distinct points in X. Then F1 ⊂ O1.α with O1. Let η = min(ε. Let X2 be c c c another direct summand. F ⊂ Z is closed if and only if F c is open if and only if F c ∩ Xα is open in Xα for each α if and only if F ∩ Xα is closed in Xα for each α. f −1 [O] = f −1 [O]. Consider E = R × {0.α and F2 ∩ Xα ⊂ O2. Since ρ∞ is equivalent to the usual product metric. 1 for x ∈ R \ {0}. Xα Xα 38b. On the ◦ other hand. Let (X. Thus each restriction Xα f Xα is continuous. Thus X is the direct union X = Xα . 1} → E be the quotient map and consider the base q[(a. Since B is Tychonoﬀ.y .α . The sets O1 and O2 are also open in Z so it follows that Z is Hausdorﬀ. Thus x ∈ α Oα ⊂ B1 ∩ B2 . there are disjoint open sets O1 . then O ∩ Xi is open in Xi for i = 1. If O is open in X. The open ball B x.y . ρ∞ ).α ⊂ Xα such that F1 ∩ Xα ⊂ O1. then O ∩ Xi is open in Xi for i = 1. 2 and thus open in X. F1 and F2 are closed in Z so there are disjoint open sets O1 . there are disjoint open sets O1 . (*) Error in book. Thus Xα is normal. 38d. For any U × V with U open in X and V open in Y . then since Xα ∩ Xβ = ∅ and the sets Xα and Xβ are open in Z. Hence Z is normal. each part is open (and closed). O2 ⊂ Z such that F1 ⊂ O1 and F2 ⊂ O2 . / Then O is also open in X so X is Tychonoﬀ. there are open balls Bx. For any open set O ⊂ Y . the sets F1 ∩ Xα and F2 ∩ Xα are closed in Xα for each α so there are disjoint open sets O1.e. F2 ⊂ Xα . Given distinct points x. y) < ∞. Note that α Xα is a basic element. given x. Let Z = Xα . Then πα [O1 ] and πα [O2 ] are disjoint open sets in α Xα containing x and y respectively. Since Xα is Hausdorﬀ. Then F1 ⊂ O1 ∩ Xα and F2 ⊂ O2 ∩ Xα with O1 ∩ Xα and O2 ∩ Xα being disjoint open sets in Xα . suppose each restriction f Xα is continuous. *40c. 2.α . Let X2 = X \ X1 .3b. suppose each Xα is Hausdorﬀ. 1}. 0 x. 42. 38c. we consider (X × Y. we have xα = yα for some α. σ) be two metric spaces. then xα ∈ Oα. Then 48 . Given x. 41.ε so B x. Then E is not Hausdorﬀ and thus not regular and not completely regular but the elements in the base are. Suppose Z is normal. There is a basic element B containing x. δ). Conversely. For any open set O ⊂ Y . X1 ∩ X2 = ∅ and X = X1 ∪ X2 . suppose each Xα is normal. *40b. there exists O open in B such that x ∈ O but y ∈ O.ε and By. f −1 [O] is open in Xα for each α. Then O1 ∩ Xα and O2 ∩ Xα are disjoint open sets in Xα containing x and y respectively. ρ) and (Y. If α = β. Suppose f : Z → Y is continuous. Given x. Then X2 is open. Then X is the disjoint union of its parts and by ◦ Q7. F2 ⊂ Z.ε is the same as Bx. O2 ⊂ Z such that x ∈ O1 and y ∈ O2 . y ∈ Z so there are disjoint open sets O1 . y ∈ Z. If X has a base. Let (X. i. Now X2 is closed so X1 = X2 ∩ X1 is closed. 43.ε is open in the product topology.i . Thus Xα is Hausdorﬀ. Let X1 ⊂ X be a direct summand. O2. Then X2 is open and X1 ⊂ X2 . y ∈ Xα . 39b.α being disjoint open sets in Z. O2 ⊂ Xα such that x ∈ O1 and y ∈ O2 .α and F2 ⊂ O2. y ∈ U × V . If α = β. let x.ε ⊂ U and y ∈ By. each element being Tychonoﬀ. Let Xα .α and O2. if O is open in X1 ∪ X2 . O2 ⊂ Xα such that x ∈ O1 and −1 −1 y ∈ O2 .8. Let X1 be a subset of X that is both open and closed. Tα be a family of Hausdorﬀ spaces. Let q : R × {0. b) × {e}] for a < b and e ∈ {0. Given disjoint closed sets F1 .δ such that x ∈ Bx.
Thus Tx = (X × {b}) ∪ ({x} × Y ) is connected for each x ∈ X. . Each x ∈ X can be identiﬁed with the element in I F with f th coordinate f (x). Suppose each f ∈ F is continuous.m } is countable. Suppose X is a normal space satisfying the second axiom of countability. Now since πα is continuous. . Let F be the mapping of X onto its image in I F . . i. αm } ⊂ A and some ﬁnite collection {O1 . X can be given an equivalent metric ρ that is bounded by 1. Then there exists N such that fn (αi ) ∈ Oi for n ≥ N and i = 1. there is a countable family F of continuous functions with the property in part (a). we only need to consider the sets {f : f (α1 ) ∈ O1 . Fix a point a ∈ X. We may then also identify each basic element α∈A Oα with {f : f (α) ∈ Oα for each α}. then each basic element in the product topology is a ﬁnite intersection of preimages under some πα of open sets in X. Suppose X is metrizable and A is countable. Let U be open in X and let y ∈ F [U ]. Thus F [U ] is open in I F . being homeomorphic to X and Y respectively. πf (y) = f (x) = 1 so y ∈ W . by Q32. Also. .m ≡ 0 on Bm . For any ﬁnite subset K ⊂ A.}.m (x) = 1 and gn. Furthermore the family {gn. Thus W ⊂ F [U ]. Suppose X is a normal space satisfying the second axiom of countability. By deﬁnition. The subspaces X × {b} and {x} × Y are connected. For each pair of indices n. πα (xn ) converges to πα (x). Hence the product topology is the weakest topology such that each πα is continuous. m such that Bn ⊂ Bm . First we consider ﬁnite products of connected spaces.δ ⊂ U × V . 48a. Now y = F (x) for some x ∈ U and there is f ∈ F −1 such that f [U c ] = 0 and f (x) = 1. Then x∈X Tx is the union of a collection of connected spaces having the point a. .m on X such that gn. .e. then z = F (x) for some x ∈ X with f (x) > 0 so x ∈ U . πα [Oα0 ] = {x ∈ α Xα : xα0 ∈ Oα0 }. By part (b). I F is metrizable and thus X is metrizable. Further suppose that given a closed set F and x ∈ F there is f ∈ F such that f [F ] = 0 / and f (x) = 1. f (αn ) ∈ On } where {α1 . . Y is connected. Then gn. But this union is X × Y so X × Y is connected. y) = −n ρ(xαn . Given an open set Oα0 ⊂ Xα0 . n2 −1 46. . The result for any ﬁnite product follows by induction. Conversely. We may identify each x ∈ α∈A Xα with the function f : A → X given by f (α) = xα .ε × By. Given a basic element f Of ⊂ I F . If z ∈ W . Since any two of them have a point in common. If T is a topology on X A such that each πα is continuous. Then f (αi ) ∈ Oi for some ﬁnite subset {α1 . F −1 [ f Of ] = {x : F (x) ∈ f Of } = {x : f (x) ∈ Of for each f } = f −1 [Of ]. ∞)] and let W = V ∩ F [X]. Let y be the element with yα = xα for α ∈ K and yα = aα for α ∈ K. Then there is a homeomorphism between X and I F . But since all but ﬁnitely many of the sets Oα are Xα . The intersection is in fact a ﬁnite intersection since all but ﬁnitely many of the sets Of are I.m [F ] = 0. Let Y be the union of the sets XK . Thus fn ∈ B for n ≥ N and fn converges to f in X A . 47. Let V = πf [(0. . The ternary expansion of numbers gives a homeomorphism between 2ω and the Cantor ternary set. suppose fn (α) converges to f (α) for each α. Suppose X and Y are connected and choose a. . choose a basic element Bm such that x ∈ Bm ⊂ F . / ¯ Hence X = Y so X is connected.m ≡ 1 on Bn and gn. Given a closed set F c and x ∈ F . . there exists a c continuous function gn. Thus each basic element in the product topology is in T and thus the product topology is contained in T . α2 . . 44. . αn } is some ﬁnite subset of A and {O1 . By Q45. Suppose a sequence fn converges to f in X A . . Now Oα = Xα for all but ﬁnitely many α so let K be that ﬁnite set. . b ∈ X × Y .11a. there exists Bn such that / x ∈ Bn and Bn ⊂ Bm . which is open in α Xα . Let {Bn } be a countable base for X. . yαn ). We may enumerate A as {α1 . . 45.B x. .y . *48b. Let B be a basic element containing f . Under the identiﬁcation again. . Let x ∈ X and let U = α Oα be a basic element containing x. Hence the product topology on X × Y is the same as the topology induced by the product metric. On } is a ﬁnite collection of open subsets of X. Thus F −1 [ f Of ] is open and F is continuous. 48c. Hence each πα is continuous. Then XK / is homeomorphic to the ﬁnite product α∈K Xα so it is connected. . *49. Then we may regard this as a sequence xn converging to x under the identiﬁcation described above.η ⊂ Bx. . by Urysohn’s Lemma.α converges to xα for each α. Then y ∈ XK ⊂ Y and y ∈ U . . .m is deﬁned with gn. . xn. Then σ is a metric on X A so X A is metrizable. . X A = α∈A Xα . m. let XK be the subspace consisting of points x such that xα = aα for α ∈ K. By Q23a. Thus U × V is open in the metric topology. 49 . fn (α) converges to f (α) for each α. Deﬁne σ on X A by σ(x. . Then W is open. b in common so it is connected. Then by Q7. Om } of open subsets of X. . Let {Xα }α∈A be a collection of connected spaces and let X = α Xα . Hence F is a homeomorphism.
The proofs of Lemmas 7. f (y)) < ε for y ∈ O. B of open sets with x ∈ A. Conversely. Choose α such that α α0 and α α1 . For each F ∈ A. fN (y)) < ε/3 for y ∈ O. Let x. f (x) > 1/n for uncountably many x. 55. Let A be the system consisting of all ﬁnite subsets of X. Then σ(f (x). . Suppose f is continuous. fN (x)) + σ(fN (x). Then fn converges to f . 50b. Then xα ∈ U ∩ V for α α . O . βαn . suppose X is not Hausdorﬀ. let F0 = {x1 . fNx (x)) + σ(fNx (x). there exist α0 and α1 such that xα ∈ U for α α0 and xα ∈ V for α α1 . then for some n. there exists βα such that πα (xβ ) ∈ Oα for β βα . Choose x A. we only need to consider a ﬁnite set βα1 . Thus by considering arbitrarily large ﬁnite subsets of G. we also have πα (xβ ) ∈ Oα = Xα . Hence f (xα ) converges to f (x). . suppose each coordinate of xβ converges to the corresponding coordinate of x. For β β0 . f (x)) < ε. f (x)) < ε/3 for n ≥ N and x ∈ X. y ∈ B. m ≥ N and x ∈ X. Hence f (xn ) < ∞. we have σ(fn (x). y − yF  ≤ n=N +1 f (xn ) < ε so yF ∈ (y − ε. For A.7 Nets 52. . the sequence fn (x) is Cauchy so it converges since Y is complete. For α = αi . there exists N such that ∞ f (xn ) < ε. For each x ∈ X. Suppose X is Hausdorﬀ and suppose a net xα in X has two limits x. we have x ∈ f −1 [O]. Then in particular the statement holds for all sequences. Conversely. Suppose that f (x) = 0 except for x in a countable subset {xn } and f (xn ) < ∞. Let X be any set and f a realvalued function on X. Given ε > 0.3739 remain valid when X is a separable topological space. y + ε). Hence the net xβ converges to x. a 8. Let X = α Xα . σ) that converges uniformly to a map f . *54. 51. Then for n ≥ N and x ∈ X. Thus both x and y are limits. Conversely. For any open interval (y − ε. f is continuous. Let fn be a sequence of continuous maps from a topological space X to a metric space (Y. there exists Nx ≥ N such that σ(fNx (x). Suppose a net xβ in X converges to x. For n=N +1 ∞ F F0 . Hence f (x) = 0 except on a countable set. y + ε).6 Topological and uniform properties 50a. Let y = f (xn ). suppose that for each net xα converging to x the net f (xα ) converges to f (x). For each α. there is an open set O containing x such that σ(fN (x). f (y)) ≤ σ(f (x). 53. It follows that f is continuous. Let O be an open set containing x and let O be an open set containing y. . Given ε > 0. Given x ∈ X. let yF = x∈F f (x). Then f (xα ) ∈ O for α α0 . Let α Oα be a basic element containing x. fm (x)) < ε/2 for n. each coordinate of xβ converges to the corresponding coordinate of x. For each x ∈ X. which is open. there exists N such that σ(fn (x). Thus fn converges uniformly to f and by part (a). 50 . y be two points that cannot be separated and let the directed system be the collection of all pairs A. f (x)) ≤ σ(fn (x). with F G meaning F ⊂ G.8. Thus lim yF = y. Since all but ﬁnitely many of the Oα are Xα . Then there are disjoint open sets U and V such that x ∈ U and y ∈ V . Let fn be a sequence of continuous maps from a topological space X to a metric space (Y. then we only have f (xn ) < ∞ and it follows that the limit is not unique. f (x)) < ε/2. we may choose β0 such that β0 βαi for all i. Then πα (x) ∈ Oα for each α. we see that yF does not converge. .B ∈ A ∩ B ⊂ O ∩ O . . fN (y)) + σ(fN (y). Contradiction. . Given ε > 0. if f (x) = 0 on an uncountable set G. Let xα be a net that converges to x. there exists N such that σ(fn (x). xN }. Let f (x) be the limit of the sequence. Thus xβ ∈ α Oα for β β0 . Thus the AscoliArzel´ Theorem and its corollary are still true. In particular. Hence f is continuous. There exists α0 such that xα ∈ f −1 [O] for α α0 . Since each projection is continuous. Since x and y are limits. For any open set O containing f (x). Suppose Y is complete. we have παi (xβ ) ∈ Oαi . . Hence every net in X has at most one limit.B ∈ A ∩ B. y. B O. Conversely. σ) that is uniformly Cauchy. Now if f (xn ) = ∞. we have A ⊂ O and B ⊂ O so x A.
Hence C is connected. Suppose X is a compact Hausdorﬀ space. Suppose X is compact. For n ≥ N . For each x ∈ X. Let F and G be disjoint closed subsets. Vyn }. 5b. f (x)) ≤ σ(fn (x). Then U and V are disjoint open sets with x ∈ U and F ⊂ V .9 9. T2 ) is compact. Let U = i=1 Uyi and V = i=1 Vyi . Then N{x}. it is normal. If T2 is a stronger topology. T ) is a Hausdorﬀ space. Suppose Kn is not a subset of O for any n. Let Kn be a decreasing sequence of compact sets with K0 Hausdorﬀ. Suppose fn converges to f uniformly on each compact subset C of X. If T1 is a weaker topology. let O be an open set containing f (x). . 3. Then we also have σ(f (x). fn (x) ∈ O for n ≥ N . If (X. T1 ) is compact. For x ∈ X. then for T∈ stronger than T . fn (xi )) + σ(fn (xi ). Since C is an intersection of closed sets. If (X. suppose every open cover has a ﬁnite reﬁnement. then id is a homeomorphism. Let NK. . Let C = Cn . f (xi )) + σ(f (xi ). *7. Since Kn is a compact subset of the Hausdorﬀ space K0 . Conversely.O for n ≥ N . Contradiction. . T2 ) → (X. Let fn be a sequence of continuous maps from a topological space X to a metric space (Y. Then U and V are disjoint open sets with F ⊂ U and G ⊂ V . any two distinct points in X can be separated by disjoint sets in T . Thus there is a ﬁnite subcovering {Vy1 .e. In particular. it has a ﬁnite reﬁnement. Thus there is a ﬁnite subcovering {Vy1 .O is open in the compactopen topology and contains f . T2 ) is Hausdorﬀ. Contradiction. Let X be a Hausdorﬀ space and Cn a decreasing sequence of compact and connected sets. . i. Since every element in the reﬁnement is a subset of an element in the open cover. 6. U is an open cover of some Cn by Q2. Hence X is regular. 4. For each x ∈ X. f (x)) < ε/3 for n ≥ Nx . . Let O be an open set with Kn ⊂ O. f (y)) < ε/3 for y ∈ Ox . Kn is closed in K0 so Kn \ O is closed in K0 . there exists Nx such that σ(fn (x). Also. x ∈ Oxi for some i so σ(fn (x). Hence f (x) = lim fn (x). there exists an open set Ox containing x such that σ(fn (x). *8b. Hence X is normal. Kn \ O is a decreasing sequence so it has the ﬁnite intersection property. If (X. Now G is compact and n {Vy : y ∈ G} is an open cover for G. which are also sets in T∈ . Hence C is compact. the open cover has a ﬁnite subcovering so X is compact. Let U = i=1 Uyi n and V = i=1 Vyi .O be a subbasic element 51 . for each y ∈ G. Vyn }. Then every open cover has a ﬁnite subcovering. σ(fn (xi ). For any open cover U of C. . σ). Then Kn \ O is nonempty and closed in Kn . Thus it has a ﬁnite subcovering. . Let fn be a sequence from F such that fn (x) → f (x) for all x ∈ X. Hence (X. 5a. T ) → (X. For each y ∈ F . . Contradiction. f (x)) < ε for n ≥ N . Suppose C is disconnected with A and B being a separation for C. T ) is a continuous bijection. (*) Assume K0 is Hausdorﬀ. . Let fn be a sequence of maps from X to Y that converge in the compactopen topology to f . then for T∞ weaker than T . . f (xi )) < ε/3 for 1 ≤ i ≤ k. V ⊂ C0 such that A ⊂ U and B ⊂ V . T2 ) is not compact. any open cover from T1 is an open cover from T so it has a ﬁnite subcovering. Now {Ox : x ∈ X} is an open cover for X so there is a ﬁnite subcovering {Ox1 . Suppose X is a compact Hausdorﬀ space. T ) is a compact Hausdorﬀ space. Now ∅ = ( Kn ) \ O = (Kn \ O) ⊂ Kn ⊂ O. / there are disjoint open sets Uy and Vy with x ∈ Uy and y ∈ Vy . T1 ) is Hausdorﬀ. Also. Thus (X. T ) is a compact space.1 Compact and Locally Compact Spaces Compact spaces 1. By Q3. Let X be a compact space and F an equicontinuous family of maps from X to a metric space (Y. Then C = A ∪ B ⊂ U ∪ V . there are disjoint open sets Uy and Vy such that F ⊂ Uy and y ∈ Vy . T1 ) is a continuous bijection. There exists N such that fn ∈ N{x}. By Q2. Hence (X. Hence fn converges uniformly to f on X. (*) Assume X is Hausdorﬀ 8a. 5c. Hence Cn is disconnected. σ). Cn ⊂ U ∪ V for some n. fn (y)) < ε/3 for y ∈ Ox and all n. Thus (X. Thus A and B are closed in C0 . Thus there are disjoint open sets U. If (X. Suppose (X. Oxk }. Then A and B are nonempty disjoint closed subsets of C. then id is a homeomorphism. . Since C0 is compact and Hausdorﬀ. Contradiction. *2. Hence Kn ⊂ O for some n. Let F be a closed subset and let x ∈ F . T1 ) is not Hausdorﬀ. Now F is compact and {Vy : y ∈ F } is an n n open cover for F . which also covers C. given ε > 0. . C is closed. Let N = max1≤i≤k Nxi . then id : (X. then id : (X.
suppose fn converges to f in the compactopen topology. Oc ). f (x)) < ε/4. Thus by Proposition 11 (Dini). xn ∈ {x : x < x0 +1} for all n so the sequence converges to x0 . Thus {x : f (x) < α} = n {x : fn (x) < α}. {x : a < x}. x ∈ Ci for some i so σ(f (xi ). Given α ∈ R. {x : a < x < b}. if fn (x) < α for some n.ε/4 . On the other hand. Now if x − y < δ. Let fn be a decreasing sequence of upper semicontinuous functions which converge pointwise to a realvalued function f . 9b. Thus there exist x1 . Fix y ∈ X. f (xi )) < ε/4 for n ≥ N . . c) < x < b} = {x : a < x < b} ∩ {x : c < x}. by Tietze’s Extension Theorem. Then xn < x0 +1 for all n. If f (x) < α. let n Ci = C ∩f −1 [Bf (xi ). .e. i. For any sequence xn in X.e. which is open. Thus the sequence has no limit points. which is open. Given ε > 0. Let X be the set of ordinals less than the ﬁrst uncountable ordinal and let B be the collection of sets of the form {x : x < a}. Let f be a continuous realvalued function on X. then there exists a ∈ X such that x < a for all x ∈ X. f is also upper semicontinuous so f is continuous. There exists δ > 0 such that x − y < δ implies f (x) − f (y) < α − f (y).Oi . Hence X is sequentially compact. Since C is compact and f is continuous. Hence B is a base for a topology on X. {x : x < c} ∩ {x : a < x < b} = {x : a < x < min(b. f (x) < α. Hence {x : f (x) < α} is open and f is upper semicontinuous.e. We ﬁrst show the existence of a sequence xn in X such that f (y) − f (xn ) < 1/n for y > xn . Contradiction. . . Now {x : x < b} ∩ {x : a < x} = {x : a < x < b}. i. xn ∈ C such that the open balls Bf (x1 ). A is discrete. then there exists N such that fn (x) − f (x) < α − f (x) for n ≥ N . Then A is closed and since all subsets of A are also closed. Now for any real number α. A realvalued function f on X is continuous if and only if {x : f (x) < α} and {x : f (x) > α} are open for any real number α if and only if f is both upper semicontinuous and lower semicontinuous. Hence X is not compact. 9. Thus (ii) ⇒ (i). f ). that f ≥ 1. 9e. Hence fn converges to f in the compactopen topology. pick y ∈ {x : f (x) < α}. there is a continuous function g on X such that gA = f .2 Countable compactness and the BolzanoWeierstrass property 9a. fn (x) < α for n ≥ N . By part (c). Since {x : fN (x) < fN (y) + ε/3} is open. Hence fn converges uniformly to f on C. Suppose that every bounded continuous realvalued function on X assumes its maximum. by taking max(1. Then {x : f (x) < α} and {x : g(x) < α} are open for any real number α. Suppose f and g are upper semicontinuous. Let f be a continuous function and suppose it is unbounded. Then it does not have the BolzanoWeierstrass property. Let ε = σ(f [K]. . *10 (i) ⇒ (iii) by Proposition 9.containing f . there exists δ > 0 such that x − y < δ implies fN (x) − fN (y) < ε/3. For any x ∈ C. fn converges to f uniformly. f [C] is compact.O for n ≥ N . fn − f converges to zero uniformly. Thus f ∈ i=1 NCi . {x : f (x) + g(x) < α} = q∈Q [{x : g(x) < q} ∩ {x : f (x) < α − q}]. There exists N such that σ(fn (x). Now fn − f is a sequence of upper semicontinuous functions on a countably compact space such that for each x.e. Let A = {xn }. 11a. For any x0 ∈ X. The sets {x : x < a} form an open cover for X. Also. i. then f (x) ≤ fn (x) < α. let x0 = sup xn . Then g is an unbounded continuous function on X. (*) Assume f to be continuous. Thus σ(fn (x). there exists N such that fN (x) − f (x) < ε/3 for all x ∈ X. . . Let fn be a decreasing sequence of upper semicontinuous functions on a countably compact space. If it has a ﬁnite subcovering. fn ∈ NK. Bf (xn ). 9d. If no such sequence exists. we may construct for some 52 .ε/4 . For each i.ε/4 cover f [C]. Then f [K] is a compact set disjoint from Oc so σ(f [K]. Then −1/f is a bounded continuous function with no maximum. We may assume. f (x)) < ε for n ≥ N and x ∈ K. c)}. Oc ) > 0. Hence f is upper semicontinuous. Suppose a sequence fn of upper semicontinuous functions converges uniformly to a function f . and suppose that lim fn (x) = f (x) where f is a lower semicontinuous realvalued function. Let C be a compact subset of X and let ε > 0 be given. fn (x) ∈ Oi for n ≥ N so σ(fn (x). i. f (x)) < ε for n ≥ N and x ∈ C. n There exists N such that fn ∈ i=1 NCi . i. Then f is continuous on the closed set A and since X is normal.e. Conversely. Hence f + g is upper semicontinuous. There is a sequence xn in X with no cluster point. *11c. {x : max(a. then f (x) − f (y) = [f (x) − fN (x)] + [fN (x) − fN (y)] + [fN (y) − f (y)] < ε. x0 ∈ {x : x < x0 + 1}. fn (x) − f (x) decreases to zero. . Then fn (x) ∈ O for n ≥ N and x ∈ K. 9c. Deﬁne f (xn ) = n for xn ∈ A. Suppose X is no countably compact. *11b.Oi for n ≥ N .ε/4 ] and Oi = Bf (xi ). Then Ci is compact and f [Ci ] ⊂ Oi . Thus (iii) ⇒ (ii).
. Ub0 } is a ﬁnite subcovering of U that covers Y . Consider f −1 [Ij ] for j = 1. Thus for each t ∈ I. If x = y. Then each Un is of the form {x : x > an }. g is continuous. Thus x. *19a. . . . Contradiction. b]. Since y > b0 or y < a1 for any y ∈ Y . there is an open set Bt containing t such that X × Bt can be covered by ﬁnitely many elements of U. If x. Let O = i=1 Oxi . Then K is closed in 53 . . Similar argument as Q11a. Let Q = f ∈F If . Given ε > 0. X × {t} is compact so there is a ﬁnite subcovering {U1 . cover f [Q] by ﬁnitely many open intervals I1 . y ∈ X × B. Now X is closed in I n so X is compact. Hence Y is compact. n then x. Then zn converges to its supremum but f (zn ) does not converge. 9. and thus normal. {La1 . Now X × {t} can be covered by ﬁnitely many basic sets A1 × B1 . there is a subsequence xn whose ﬁrst coordinate converges.3 Products of compact spaces 13. f −1 [Ij ] = α Uα where Uα is open in (j) (j) n I and all but ﬁnitely many of the Uα are I. there exists f ∈ F such that f (x) = 0 and f (y) = 1. 9. Thus D is not dense in Y and Y is not separable. Then f (x) = f (x0 ) for x ≥ x0 . contradicting the continuity of f . Let Q = I A be a cube and let f be a continuous realvalued function on Q. Let X be a locally compact space and K a compact subset of X. 1]. Each closed and bounded subset X of Rn is contained in a cube I n where I = [a. Each coordinate of this sequence converges so the sequence converges in X. Since f is continuous for each f ∈ F. Suppose X is compact and I is a closed interval. Let F be the family of continuous realvalued functions on X with values in [0. then since X is compact Hausdorﬀ. Let a0 = sup{a : La ∈ U}. Now X × I = i=1 (Bti × I). *12b. 12a. Suppose there is a countable base {Un } at ω1 . These sets (j) (j) cover Q and we may assume each of them is a basic set. Ak × Bk ⊂ U . . . Each I is compact in R so I n is compact in Rn by Tychonoﬀ’s Theorem. Also. g is a continuous bijection from the compact space X onto the Hausdorﬀ space g[X] so X is homeomorphic to g[X]. 0 ∈ La for some La ∈ U. Btm } covering X. that is. Then there is a (n) (1) (2) subsequence xn of xn whose second coordinate converges. . 15. There exists a < ω1 such that a > an for all n.N an increasing sequence zn such that f (zn ) − f (zn−1 ) ≥ 1/N for each n. . Then ω1 ∈ Ub for some Ub ∈ U. *17. . . Consider the mapping g of X into Q mapping x to the point whose f th coordinate is f (x). Deﬁne h : Q → Q by πα h(x) = xα for α ∈ F and 0 otherwise. let Fj = {α : Uα = I} and let F = j=1 Fj . there is a ﬁnite subcollection {Ox1 . For each x ∈ K. Then O ⊃ K and O = i=1 Oxi is compact. y ∈ Aj × Bj and n X × B ⊂ i=1 (Ai × Bi ) ⊂ U . . Then y = sup D exists and y ∈ Y . Let U be an open cover for Y consisting of basic sets. The collection of the sets Bt forms an open cover of X so m there is a ﬁnite subcollection {Bt1 . . *14. . Oxn } n n ¯ that covers K.4 Locally compact spaces 18. there is an open set Ox containing x with Ox compact. Then a0 ∈ Ub0 for some Ub0 ∈ U so b0 < a0 . Let La = {x : x < a} and Ub = {x : b < x}. Deﬁne g = f ◦ h. Now {x : y < x} is an open set in Y that does not intersect D. n. Since K is compact. Then the open set {x : x > a} does not contain any Un . Consider the diagonal sequence xn . . . Now let x0 = sup xn . Let Xn be a countable collection of sequentially compact spaces and let X = Xn . Let D be a countable subset of Y . . For each j. Then g depends only on F and f − g < ε. Since X × {t} is homeomorphic to X. Thus Y is not ﬁrst countable. which is a ﬁnite set. Given a (1) sequence xn in X. Let X be a locally compact Hausdorﬀ space and K a compact subset. Let X be a compact Hausdorﬀ space. Hence X is sequentially compact. Let U be an open cover of X × I and let t ∈ I. Then B = B1 ∩ · · · ∩ Bk is an open set containing t. . Now g[X] is a compact subset of the Hausdorﬀ space Q so g[X] is closed in Q. . Un } n of U such that U = i=1 Ui ⊃ X × {t}. . t ∈ Aj × Bj for some j so x ∈ Aj and y ∈ i=1 Bi ⊂ Bj . which can then be covered by ﬁnitely many elements of U. . In of length ε. Thus g is onetoone. . . Then there exists a1 > b0 such that La1 ∈ U. by Urysohn’s Lemma. Furthermore. . 16.
xn . Let X ∗ be the Alexandroﬀ onepoint compactiﬁcation of a locally compact Hausdorﬀ space X. Also. Let K be a compact subset of a locally compact Hausdorﬀ space X.X ∗ . Deﬁne f : X ∗ → Y by f (x) = x for x ∈ X and f (ω) = ω1 . Hence id is a homeomorphism. and f ∗ the mapping of X ∗ into Y ∗ whose restriction to X is f and which takes the point at inﬁnity in X ∗ to the point at inﬁnity in Y ∗ . K \ U is compact. Thus the collection of sets is closed under ﬁnite intersection. then (X ∗ \ Kβ ) = X ∗ \ Kβ . Then by Q18. which is open in X. y be distinct points in X ∗ . each Kβ is closed in X and Kβ is closed in X. Then f is clearly a bijection. Thus the collection of sets is closed under arbitrary union. Hence the onepoint compactiﬁcation of X is Y . 23. which is compact because g is continuous on X ∗ . there is an ¯ open set O ⊃ K with O compact. any open set in X is open in X ∗ so id is an open mapping. Since f is a continuous bijection from the compact space X ∗ to the Hausdorﬀ space Y . Then {x : f (x) > 0} = {x : g(x) > 1/2}. Let x. Finally. *21a. By Q8. By Urysohn’s Lemma. . O and U 22b. Thus Rn is homeomorphic to S n −p and the Alexandroﬀ onepoint compactiﬁcation of Rn is homeomorphic to the Alexandroﬀ onepoint compactiﬁcation of S n − p. Deﬁne f : S n − p → Rn 1 by f (x) = 1−xn+1 x1 . Then there is a ﬁnite subcollection that covers K. there exists a closed set D containing ω with D ∩ K = ∅. then f −1 [U ] = U is open in X. (*) Alternatively. Let O be an open subset of a compact Hausdorﬀ space X. Consider the basic sets in Y . . then Uα is open in X. Similarly for sets of the form {x : a < x < b}. By Q8. The map g : Rn → S n − p deﬁned by g(y) = t(y)y1 . Now X ∗ \ O and K are disjoint closed subsets of X ∗ so by Urysohn’s Lemma. . there is an ¯ open set U such that x ∈ U and U ⊂ O. Let S n denote the unit sphere in Rn+1 . Let O be an open subset of a compact Hausdorﬀ space X. Also. there is a continuous function g on X ∗ with 0 ≤ g ≤ 1 that is 1 on K and 0 on D. which is open in X. Hence X ∗ is compact. Hence O is locally compact. *19b. . then so is U1 ∩ U2 . then (X ∗ \ K1 ) ∩ (X ∗ \ K2 ) = X ∗ \ (K1 ∪ K2 ) where K1 ∪ K2 is compact. . Suppose 54 . Let X and Y be locally compact Hausdorﬀ spaces. whose complement {x : x ≤ a} is compact. f is a homeomorphism. f −1 [{x : a < x}] = {x ∈ X : a < x} ∪ {ω}. ¯ The sets O and X ∗ \ O are disjoint open sets in X ∗ separating x and y. If K1 and K2 are compact subsets of X. Hence X ∗ is Hausdorﬀ. . If {Uα } is a collection of open sets in X. being closed in a compact space. . U1 ∩ (X ∗ \ K1 ) = U1 ∩ (X \ K1 ). Now f −1 [{x : x < a}] = {x : x < a}. . . which is open in X and thus open in X ∗ . Consider the mapping f of X to the onepoint compactiﬁcation of O which is identity on O and takes each point in X \ O to ω. . If U is open in X. then U = (X ∗ \ {ω}) ∩ U for some U open in X ∗ . 1 ∈ Rn+1 . then f −1 [O∗ \ K] = X \ K. 1 − t(y) where t(y) = 2/(1+y2 ) is the inverse of f . Let U be an open cover of X ∗ . Let p = 0. If U is open in X ∗ \ {ω}. then there is an open set O containing x with O compact. If U1 and U2 are open in X.23a. If x ∈ X and y = ω. 0. Then ∅ and X ∗ are in the collection. ¯ that separate x and y. where U = Uα and K = Kβ . Then U contains a set of the form X ∗ \ K for some compact K ⊂ X. and f a continuous mapping of X into Y . Also. for any x ∈ O. use Q16 to regard K as a closed subset of the compact Hausdorﬀ space Q. If {Kβ } is a collection of compact subsets of X. there is a continuous function g on X ∗ with 0 ≤ g ≤ 1 that is 1 on K and 0 on X ∗ \ O. The corresponding ﬁnite subcollection of U together with X ∗ \ K then covers X ∗ . Thus id is continuous. then U = U so U is open in X. 20a. 20c. Thus Kβ is closed in each Kβ so Kβ is compact. then U = X \ K. then there are disjoint open sets in X. Let X be the space in Q11 and Y be the space in Q12. Hence f is continuous. 0). Thus f −1 [{x : a < x}] is open in X ∗ . 20b. .23a. 21b. . and thus in X ∗ . Take the other elements of U and intersect each of them with X to get an open cover of K. Since each Kβ is compact in the Hausdorﬀ space X. Consider the collection of open sets of X and complements of compact subsets of X. If K ⊂ O is compact. Since K \ U is closed in the compact set K. 22a. If U is open in O. Let id be the identity mapping from X to X ∗ \ {ω}. Let X ∗ and Y ∗ be the onepoint compactiﬁcations of X and Y . Clearly id is a bijection. t(y)yn . Hence the collection of sets forms a topology for X ∗ . which is S n . Then f = gX is the required function. Deﬁne f (x) = min(2(g(x) − 1/2). If U = X ∗ \ K for some compact K ⊂ X. which is open in X. Then the closure of U in X is the same as the closure of U in ¯ is compact. y ∈ X. Uα ∪ (X ∗ \ Kβ ) = U ∪ (X ∗ \ K) = X ∗ \ (K \ U ). If x.
Given y ∈ Y . Hence Y is open in X. Then x ∈ F ∩ O \ (F ∩ O) so F ∩ O is not closed. Let y ∈ Vn . *28. *27. (*) Assume that X is Hausdorﬀ. . Furthermore. ¯ ¯ Proof: There is an open set O ⊂ X containing x such that O is compact. Y ∩ U ¯ ¯ ¯ open set V ⊂ X. *24b. Vn have been chosen. For any open set U containing x. f > 0 on K. Given an open set U . ¯ ¯ Then there is an open set V containing x such that V is compact and V ⊂ U . If K ⊂ Y is compact. (*) Assume that X is Hausdorﬀ. Let K = {xn : n ∈ N} ∪ {x}. Thus there is an open set O ⊂ O ¯ ¯ ¯ ¯ ¯ ¯ O O = O ∩ O = O . Let X be a locally compact Hausdorﬀ space and {On } a countable collection of dense open sets. (iii) Given an open set O ⊂ X and x0 ∈ O. (f ∗ )−1 [Y ∗ \ K] = X ∗ \ f −1 [K] is open in X ∗ . By Q26. For each closed compact set K. ¯ V is compact. Let F be a family of realvalued continuous functions on a locally compact Hausdorﬀ space X. F ∩ K is closed. Then K is compact in the Hausdorﬀ space X and thus closed. there is a gx0 ∈ F with gx0 (x0 ) = 1. . Since ¯ is closed. Suppose F is a closed subset of X. ¯ in O. 29a. . Also. *26. . V = V ∩ Y . g ∈ F. For each x0 ∈ O\K. Set f = i=1 fi and g = i=1 gi . g ∈ F. we may choose a ﬁnite number f1 . Conversely. . Then y ∈ On ∩ U . Since V is closed ¯ ¯ ¯ O. Take x ∈ F \F . . . Suppose x1 . *24a. Now U = V ∩Y for some ¯ ¯ ¯ X is Hausdorﬀ. fn . Then x ∈ V and V ⊂ V = V ∩ Y = U ⊂ Y . F ∩K is closed. Also. Let O be an ¯ open set with K ⊂ O and O compact. ¯ Then f. . Take x ∈ F \ F . there is an fx0 ∈ F with fx0 (x0 ) = 1. Let X be a locally compact Hausdorﬀ space and let O be an open subset contained in a countable union Fn of closed sets. there is an open set O ⊂ X 55 . (O ∩ U ) ∩ F = ∅. Let {Oλ } be an open covering of a compact subset K of a locally compact Hausdorﬀ space X. 0 ≤ fx0 ≤ 1 ¯ and supportfx0 ⊂ O∩Oλ for some λ. Hence (O ∩ Fn )◦ = ∅. O = (O ∩ Fn ). If U is open in Y . . . 0 ≤ gx0 ≤ 1 ¯ and supportgx0 ⊂ K c . g1 . The functions ϕi = fi /(f + g) ∈ F. Suppose F is closed in a locally compact space X. Thus O ⊂ U ∩ O. . then f + g ∈ F. By compactness of O. For each closed ¯ compact set K. . Hence On is dense in X. . Then U ∩ O is open in O ¯ such that x ∈ O and O O ⊂ U ∩ O. g ∈ F. Hence f −1 [K] ⊂ X is compact. For each x0 ∈ K. . 0 ≤ f ≤ 1 and supportf ⊂ O. This collection of closed sets has the ﬁnite intersection property so Vn = ∅. Now O ∩ Fn ⊃ (O ∩ Fn )◦ ◦ ◦ so O ∩ Fn = ∅ and Fn is an open set dense in O. . (ii) If f. which is a union of sets closed in O. and suppose that Y with its subspace topology is ¯ ¯ locally compact. Note that ¯ ¯ and contains x. we have U ⊂ Y ∩ U ⊂ Y .f is proper. Also. (O \ Fn ) is dense in O. there is a sequence xn in F converging to x. Thus f /(f + g) ∈ F is continuous and ≡ 1 on K. Conversely. Note that since Y is dense. . . there is an f ∈ F with f (x0 ) = 1. ◦ But (O \ Fn ) = O \ ( Fn )c = ∅. Suppose F is a closed subset ¯ of X. . then (f ∗ )−1 [Y ∗ \ K] = f −1 [Y \ K] ∪ {ωX } = X ∗ \ f −1 [K]. f + g > 0 on O and g ≡ 0 on K. since O is open in ¯ it is open in O and thus open in X so V is open in X. There is an ¯ open set O containing x with O compact. Thus ¯ ¯ ¯ ¯ U ∩ (F ∩ O) = ∅. n form a ﬁnite collection of functions subordinate to the collection {Oλ } and such that ϕ1 + · · · ϕn ≡ 1 on K. Given x ∈ F . Let Y be a dense subset of a Hausdorﬀ space X. let x1 be a point in O1 ∩ U . xn and V1 . Lemma: Let X be a locally compact Hausdorﬀ space and U be an open set containing x ∈ X. there is an open set U ⊂ Y containing y with UY = Y ∩ U compact. Let X be a Hausdorﬀ space satisfying the ﬁrst axiom of countability. for any compact set K ⊂ Y . Let X be a locally compact Hausdorﬀ space. which is open in X ∗ since f −1 [K] ⊂ X is compact. Then f = f ∗ X is continuous. Conversely. Since X is ﬁrst countable. V ⊂ O ⊂ U ∩ O ⊂ U . . . Let V1 be an open set containing x1 such that V1 is compact and V1 ⊂ O1 ∩ U . Then since U ⊂ Y ∩ U . suppose f ∗ is continuous. i = 1. supportf ⊂ O. suppose F is not closed. gm of n m ¯ these functions such that the sets where they are positive cover O. and suppose that F has the following properties: (i) If f. then (f ∗ )−1 [U ] = f −1 [U ]. Let xn+1 ∈ On+1 ∩ Vn and let Vn+1 be an open set containing xn+1 such that Vn+1 is compact and Vn+1 ⊂ On+1 ∩ Vn . Let V = O ∩ O . 25. Hence f ∗ is continuous. then f /g ∈ F provided that supportf ⊂ {x ∈ X : g(x) = 0}. Then V1 ⊃ V2 ⊃ · · · is a decreasing sequence of closed sets in the compact set V1 . If (O ∩ Fn )◦ = ∅. Note that O is a locally compact Hausdorﬀ space (see Q29b). Contradiction. then (O ∩ Fn )◦ = ∅ for all n (with the interior taken in O) so O \ Fn is dense and open in O for all n. suppose F is not closed. which is open in X and thus open in X ∗ . Now F ∩ K = {xn : n ∈ N} is not closed. Then x ∈ V .
If K is a compact subset. y ∈ ON for some N and thus y ∈ On for n ≥ N . For a ﬁxed x0 ∈ X and any a ∈ (0. By part (b). Hence ϕ is continuous. Thus {x : ρ(x. Note that a compact subset K ⊂ [0. ϕ(y) − ϕ(x) ≤ [ϕ(y) − n=1 (1 − ϕn (y))] +  n=1 (1 − ϕn (y)) − ϕ(x) ≤ N N [ϕ(y) − n=1 (1 − ϕn (y))] + n=1 ϕn (y) − ϕn (x) < ε. x1 ) ≤ b for all x ∈ K. Eλn so x ∈ i=1 Eλi ⊂ Eλ . Given x ∈ X and ε > 0. On ⊂ On+1 and X = On . Y is locally compact since Y is closed in X. Now U ∩ O is an open set in O ¯ containing x with U ∩ OO = O ∩ U ∩ O = U ∩ O being compact since it is closed in the compact set U . x0 ) ≤ a} ⊂ {x : ϕ(x) ≤ a + ϕ(x0 )}. Hence ρ∗ is a proper metric. . Let {Eλ } be a locally ﬁnite collection of subsets of X and K a compact subset of X. Thus E ⊂ Eλ . ∞). Thus by part (a). ∞). let ϕn be a continuous realvalued function with ϕn ≡ 1 on On−1 and supportϕn ⊂ On . ∞) is closed and bounded. Eλ ⊂ E for all λ. By Q28. ∞). For each n. Since each Oxi meets only a ﬁnite number of sets in {Eλ }. x0 ) ≤ a} is compact. xn ∈ K. Choose δ < min(δ . and set E = Eλ . Thus ρ and ρ∗ are equivalent metrics. 31c. there exists δ > 0 such that ρ(x. 9. . . Y is open in Y . Since K is bounded. Deﬁne ρ∗ (x. . Then ρ∗ is a metric on X. For each n = 1. N ϕn (x) = 1 for n ≥ N + 1 so ϕ(x) = n=1 (1 − ϕn (x)).5 σcompact spaces 30. then K is closed and bounded by Proposition 7. then there is an open set containing x n ¯ that meets only a ﬁnite number of sets Eλ1 . 29c. y) ≤ ρ∗ (x. To show that ϕ is proper. 33.¯ containing x with O compact. we consider closed bounded intervals in [0. Let (X. Then U is an open N N set containing y and for x ∈ U .22. Let U = n=1 Un ∩ON . the closed balls {x : ρ(x. Hence ¯ E = Eλ . x0 ) for all x ∈ K. ρ∗ (x. Then O ∩ F is an open set in F containing y and O ∩ F F = F ∩ O ∩ F ¯ and thus compact. 56 . y) < ε. y) + ϕ(x) − ϕ(y). x0 ) is continuous from X to [0. ρ(x. ∞) by ϕ = (1 − ϕn ). ρ) be a σcompact and locally compact metric space. In fact. ε/2). Then ϕN +1 (x) = 1. 31a. suppose Y is ¯ . Given y ∈ X and ε > 0. Deﬁne ϕ : X → [0. is closed in O *29b. y) < δ. Take x ∈ ON . . For each x ∈ X. ∞). ∞). Let (X. x0 ) ≤ a} are compact for some x0 and all x ∈ (0. When ρ(x. Then there is a proper continuous map ϕ : X → [0. Suppose a subset Y of a locally compact Hausdorﬀ space X is locally compact in its subspace ¯ ¯ topology. x0 )} so K is compact. Let N = max(N. . x0 ) ≤ a} are compact for some x0 and all a ∈ (0. . . Suppose there is a sequence On of open sets with On compact. Let {Eλ } be a locally ﬁnite collection of subsets of a topological space X. y) < ε/2 + ε/2 = ε and when ρ∗ (x. Since X is proper. ρ) be a proper locally compact metric space. Let X be a locally compact Hausdorﬀ space. we then apply a similar argument as above. suppose a subset K is closed and bounded. there exist x1 and b such that ρ(x. Suppose O is open in a locally compact Hausdorﬀ space X. Then Y is dense in the Hausdorﬀ space Y . Then the closed balls {x : ρ(x. N . ρ) be a proper locally compact metric space. y) < δ. y) = ρ(x. there is an open set Un N containing y such that ϕn (y)−ϕn (x) < ε/2N for x ∈ Un . . (*) Proof of Lemma 22. ∞) is a proper continuous map and X is σcompact. x0 ) ≤ b + ρ(x1 . so does K. . x0 ) ≤ b + ρ(x1 . Thus Eλ ⊂ E. There exists N such that N ϕ(y) − n=1 (1 − ϕn (y)) < ε/2. {x : ρ∗ (x. 9. By part (a). Thus K is a closed subset of the compact set {x : ρ(x. Conversely. Also. Since ¯ ¯ ¯ Eλ ⊂ E for all λ. By the lemma in Q26. Now n K ⊂ x∈K Ox so K ⊂ i=1 Oxi for some x1 . If x ∈ E. Take x ∈ O. which is compact. the function f (x) = ρ(x. ∞). By considering ON −1 and open sets Vn with Vn compact and Vn ⊂ Un . Hence f : X → [0. Then ρ(x.6 Paracompact spaces 32. Conversely. . ¯ ¯ there is an open set U containing x such that U is compact and U ⊂ O. Y is locally compact ¯ ¯ open in Y ¯ since Y is open in Y . N ). f −1 [K] is compact. 31b. Hence O is locally compact. . there is an open set Ox containing x that meets only a ﬁnite number of sets in {Eλ }. Let (X. Hence F is locally compact. y) < δ implies ϕ(x) − ϕ(y) < ε/2. Now f −1 [K] is bounded since K is bounded and closed since f is continuous and K is closed.
8 ˘ The StoneCech compactiﬁcation 39a. The restriction to X of the projection πf on β(X) is f and πf is continuous on β(X) since β(X) is a subspace of the product space I F . *38a. 2 + ε) are homeomorphic to (−ε. 37a. 9. there is a unique continuous mapping ϕ of Z onto β(X) with ϕ(x) = x for all x ∈ X. Let n Oy = Oy ∩ i=1 Uyi . . b) ⊂ X are open balls in R. Since πg ◦F = πg◦i is continuous for each g. Let X = (−1. there is an open set U containing x that meets only a ﬁnite number of sets n Eλ1 . A not necessarily connected manifold X is the disjoint union of its components. For each y ∈ G. Suppose Z is a space with the same properties. Let ϕ = F β(X) .7 Manifolds 36. . 39c. Then E is an open set containing F . Let E = {Eλ : Eλ ∩ F = ∅}. Each (Xβ . Thus ϕ ◦ ψ = idX so ϕ and ψ are homeomorphisms. . . its components are Hausdorﬀ and locally Euclidean. every continuous realvalued function on X is eventually constant so it extends to a continuous 57 . Let f be a bounded continuous realvalued function on X with f  ≤ 1. Let E = {Eλ : Eλ ∩ F = ∅}. Consider O = U ∩ i=1 Uyi . By Q8. statements (ii). For a not necessarily connected manifold. Then ϕ is a continuous mapping of β(X) onto Y with ϕ(x) = x for all x ∈ X. Let X be the set of ordinals less than the ﬁrst uncountable ordinal and let Y be the set of ordinals less than or equal to the ﬁrst uncountable ordinal. Suppose it can be metrized by a proper extended metric ρ∗ . By part (a). . there are disjoint open sets Uy and Vy such that F ⊂ Uy and y ∈ Vy . 2 + ε) for 0 < ε < 1. For each y ∈ F . *40. Then X is the direct union of σcompact spaces Xβ . ρ) be a locally compact metrizable space. Now X \G∪{Vy : y ∈ G} is an open cover for X so it has a locally ﬁnite open reﬁnement {Eλ }. y ∈ Xβ and ρ∗∗ (x.41. there is a unique continuous mapping ψ of β(X) onto Z with ψ(x) = x for all x ∈ X. Hence X is the direct union of (connected) manifolds. 0) ∪ [2. 35. The ﬁrst set of statements imply the second set of statements. Each of these sets must lie in some Vyi where yi ∈ G. 34a. y) if x. Eλn in {Eλ }. 34b. there is an open set Oy that meets only a ﬁnite number of sets Eλ1 . Then Y is a subset of I G where G is the space of continuous g on Y with g ≤ 1. b) ⊂ X and all sets of the form (−ε. Suppose X is a dense open subset of a compact Hausdorﬀ space Y . Hence X is locally Euclidean. for each y ∈ G. By Q8. Hence X is the direct union of σcompact spaces so it is paracompact. Let (X. Also. / there are disjoint open sets Uy and Vy with x ∈ Uy and y ∈ Vy . 37b. each Xα is σcompact. Hence X is regular. Now X \ F ∪ {Vy : y ∈ F } is an open cover for X so it has a locally ﬁnite open reﬁnement {Eλ }. 1) ∪ [2. 9. (v). Now F [β(X)] ⊂ F [E] ⊂ Y . Also. (vi). .(*) Proof of Lemma 23. Thus X is the direct union of its components. Now deﬁne ρ∗∗ (x. Then O is an open set containing G and O ∩ E = ∅. Let X be a paracompact Hausdorﬀ space. Conversely. Now (Xα . Also. Then E is an open set containing F . Clearly all open intervals (a. ϕ is unique as a map to a Hausdorﬀ space is determined by its values on a dense subset. The inclusion i : X → Y induces a continuous function F : I F → I G as follows. By Q11. There are no disjoint basic sets that separate 0 and 2 so X is not Hausdorﬀ. If g ∈ G. ρ∗ Xα ) is a proper locally compact metric space for each α so by Q31b. and make X into a topological space by taking as a base all open intervals (a. (iii). Let O = y∈G Oy . Each of these sets must lie in some Vyi where yi ∈ F . Then O is an open set containing x and O ∩ E = ∅. ρXβ ) is a σcompact and locally compact metric space so by Q31c. each component of X is open. X is the direct union of its parts Xα . *39b. (vii) are equivalent. Let X be a paracompact Hausdorﬀ space. statements (i) and (iv) are equivalent. F is continuous.36. so deﬁne F ( tf ) = (tg◦i )g . Then Oy is an open set containing y and Oy ∩ E = ∅. each Xβ can be metrized by a proper metric ρβ . since X is locally connected. Also. By part (b). suppose X is paracompact. Let F be a closed subset and let x ∈ F . Then X is dense in the compact Hausdorﬀ space Y . 0) ∪ [2. y) = ∞ if x ∈ Xβ1 and y ∈ Xβ2 with β1 = β2 . Let F and G be disjoint closed subsets. Eλn in {Eλ }. Then ρ∗∗ is a proper extended metric on X. . then g◦i ∈ F. ε). Also. . Also. y) = ρβ (x. sets of the form (−ε. 3).
then x ≡ y so f (x) = f (y) and f (˜) = f (˜). 45d. Let A be the set of ﬁnite Fourier series ϕ given by ϕ(x) = a0 + n=1 (an cos nx + bn sin nx). A is dense in C(X). Given f ∈ A. . say {Ox1 . Let F be a family of continuous realvalued functions on a compact Hausdorﬀ space X. if a sequence in N converges in N. then W ∩ V ∩ N = ∅ so x ∈ V ∩ N. Hence ≡ is an equivalence relation. ˜ ˜x ˜x ˜y ˜ deﬁne f on X by f (˜) = f (x). 44. Hence A = C(X). By the StoneWeierstrass Theorem. then by continuity. Let A be an algebra of continuous realvalued functions on a compact space X. then so is X. Let X be the set of equivalence classes of ≡ and ϕ the natural map of X into X. then it converges in β(N). Thus we may assume that the polynomials hn have no constant term. 9. Now f is continuous since N is discrete so it has a continuous extension to β(N). Then g ∈ A and g = 0 everywhere. there is an ¯ ¯ open set U ⊂ β(N) such that x ∈ U and y ∈ U . Conversely. Then Y = (Y ∩ U ) ∪ (Y \ U ) is a separation of Y so Y / ¯ is not connected. *43. Hence β(N) is totally disconnected. *41. 45b. then V ∩ N is an open subset of β(N).function on Y . and suppose that F separates the points of X. Thus A ∩ N \ A = ∅ and A is open. sin mx cos nx = 2 [sin(m + n)x + 1 sin(m − n)x] and sin mx sin nx = 2 [cos(m − n)x − cos(m + n)x]. Then A ∪ N \ A = β(N). . If A ⊂ N. Since β(N) is Hausdorﬀ. we see that A is a subalgebra of C(X). The sets {Ox } cover X so by 2 2 compactness. Let A be the set of polynomials in a ﬁnite number of functions of F. Note that if hn is uniformly within ε/2 of h. 1] by f (x) = 0 if x ∈ A and f (x) = 1 if x ∈ A. Then A is a subalgebra of C(X) that separates the points of X and contains the constant functions. N ∈ N. If x = y . then hn (0) < ε/2 but hn (0) is the constant term of hn so subtracting the constant term results in a polynomial still within ε of h. Clearly. Clearly. Now if x ∈ V and W is an open ¯ ¯ neighbourhood of x. . ˜ 45c. Let X have the weak topology generated by the functions f in part (b). Thus V = V ∩ N and V is open (β(N) is extremally disconnected). then f (x) = f (y) = f (z) for all f ∈ A so x ≡ z. 1] with f (x2n ) = 0 and f (x2n+1 ) = 1 for all n. ﬁnitely many of them cover X. Then A is a linear space of functions in C(X) where X is taken to be the unit circle in C. The function h given by h(t) = 1/t for t ∈ K and h(0) = 0 is continuous on K ∪ {0} so it can be uniformly approximated by polynomials hn so that hn ◦ g ∈ A and hn ◦ g → 1/g. and assume that A separates the points of X. If for each x ∈ X there is an fx ∈ A with fx (x) = 0. Consider a basic set ˜−1 [O]. ¯ ¯ and f ¯ ¯ If B ⊂ N and A ∩ B = ∅. ¯ If V is an open subset of β(N). . Since ϕ is continuous and maps X onto X. If x ≡ y and y ≡ z. Let Y be a subset of β(N) with at least two distinct points x and y. Let X be a topological space and A a set of realvalued functions on X. Deﬁne x ≡ y if f (x) = f (y) for all f ∈ A. Hence β(N) is compact but not sequentially compact as the sequence xn = n does not have a convergent subsequence. given any continuous periodic realvalued function f on R with period 2π and any ε > 0. Now ϕ−1 [f −1 [O]] = {x : f (ϕ(x)) ∈ O} = {x : f (x) ∈ O} = f −1 [O]. By deﬁnition of the 58 . Furthermore A separates the points of X and contains the constant functions. This is a contradiction as the sequence g(xn ) does not converge. 45a. Oxn }. if X is compact. there is an open neighbourhood Ox of x such that fx (y) = 0 for y ∈ Ox . then B ⊂ N \ A and A ∩ B = ∅. deﬁne f : N → [0. From the 1 1 trigonometric identities cos mx cos nx = 2 [cos(m − n)x + cos(m + n)x]. Thus f is ˜ ˜ ˜ welldeﬁned and it is the unique function such that f = f ◦ ϕ.9 The StoneWeierstrass Theorem N 42. Now f is continuous / ˆ ˆ on β(N). x ≡ x for all x ∈ X and y ≡ x if x ≡ y. By the StoneWeierstrass Theorem. It follows that f −1 [{1}] = A ¯ ¯ so it extends to a continuous function f ˆ−1 [{0}] = N \ A. which is open since f is ˜ ˜ f continuous. The closure of the range of g is a compact set K not containing 0. there is a ﬁnite Fourier series ϕ such that ϕ(x) − f (x) < ε for all x. Let g = fx1 + · · · + fxn . then consider a function f : N → [0. ¯ ¯ ¯ ¯ Thus 1/g ∈ A so 1 ∈ A and A contains the constant functions. Hence ϕ is continuous. Hence every continuous realvalued function on X can be uniformly approximated by a polynomial in a ﬁnite number of functions of F. if it converges in β(N) \ N.
1 − δ]. The functions of norm 1 in the algebra A give a mapping of X into the inﬁnitedimensional cube {If : f ∈ A. . y ∈ X × Y . Let P be a polynomial with integral coeﬃcients. Let N be the least common multiple of the denominators of the coeﬃcients of R so that N R has integral coeﬃcients and vanishes at 1. 1. *48e. If x and y are distinct points in X. Let [an . By parts (a) and (b). ˜ ˜ ˜ ˜ the image of A is C(X). 1). 1] and f a continuous realvalued function on I such that f (−1). We may assume that 0 < β < 1. Then ϕ (x) = 6x2 − 6x + 2 > 0 1 for all x. y) − i=1 gi (x)hi (y) < ε for all x. Let ϕ be the polynomial deﬁned by ϕ(x) = x + x(1 − 2x)(1 − x). Then the sequence gn . *50a. We may assume that δ < ε/P . 59 . it suﬃces to consider the case where α is a rational number a . 2 . 1] and ψ(x) − α < ε for all x ∈ [ε. f (0) = b. and thus f . 1 − ε]. there exists δ > 0 such that P (x) < ε/2 for x ∈ (−δ. Hence some iterate ϕn is a polynomial with integral coeﬃcients that is monotone increasing on [0. c are integers. hn on Y such that f (x. 48a. 2 *48c. g can be approximated by a continuous function h of only a ﬁnite number of coordinates. Thus a = b = 2 . −1 + δ). where gn = gn ◦ ϕ. b are ﬁxed points of ϕ. δ). The set of ﬁnite sums of functions of the form f (x)g(y) where f ∈ C(X) and g ∈ C(Y ) is an algebra of continuous realvalued functions on X × Y that contains the constant functions and separates points in X × Y . f → f . each continuous function f on the image of X can be extended to a continuous function g on the cube and by Q17. For any ε > 0. *50b. gn on X and h1 . Thus ϕ is monotone increasing and its ﬁxed points are 0. 1 − ε]. ϕ induces a (continuous) mapping y ∈f ˜ ˜ ˜ of ϕ∗ : A → C(X). Furthermore. *48d. The image of A in C(X) is a subalgebra containing the constant functions and separating the points of X. . b. Then h can be regarded as a continuous function on a cube in Rn .˜ weak topology on X. Since each gn ∈ A and A is ˜ ˜ ˜ closed. some b iterate ψ = ϕn is a polynomial with integral coeﬃcients (and no constant term) such that 0 ≤ ψ(x) ≤ 1 in [0. Thus ϕ∗ [A] is closed. Furthermore. 1] and such that ϕn (x) − 1  < ε for x ∈ [ε. can be approximated by a polynomial P with integral coeﬃcients. . . Then P (x)ψ(x2 ) − βP (x) < δP  < ε for all x ∈ [−1. Deﬁne ϕ(x) = x + x(a − bx)(1 − x). . Thus X is a compact Hausdorﬀ space. . .0. Let X and Y be compact spaces. *49a. Let β = 1/N and apply part (d) to the polynomial N R so that R. then f (x) = f (y) for some f ∈ A so ˜ ˜ ˜ there are disjoint open sets Ox and Oy in R with f (x) ∈ Ox and f (y) ∈ Oy . Thus for each continuous realvalued function f on X × Y and each ε > 0. By the StoneWeierstrass Theorem. Choose ε > 0. Then α is a ﬁxed point of ϕ. 1]. which can be uniformly approximated by a polynomial in (a ﬁnite number of) the coordinate functions.1. f (1) = a + 2c. Note that ϕ(x) > x on (0. and suppose that P (−1) = P (0) = P (1) = 0. converges to g ◦ ϕ. 46. f  = 1}. Hence A is the set of all functions of the form f ◦ ϕ with f ∈ C(X). f (0). the functions f are continuous. this set is dense in C(X × Y ). By part (c). f (1) are integers and f (1) ≡ f (−1) mod 2. there is a polynomial ψ with integral coeﬃcients and no constant term such that ψ(x2 ) − β < δ for all x ∈ [δ. Then an = ϕn (ε) increases to some a and bn = ϕn (1 − ε) decreases 1 to some b. Let X be a compact space and A a closed subalgebra of C(X) containing the constant functions. 47. Suppose gn is a sequence in ϕ∗ [A] that converges ˜ to g in C(X). *45e. 1 *48b. Given α with 0 < α < 1 and any ε > 0. Let β be any real number. x ∈ (1 − δ. bn ] = ϕn [ε. By the Tietze Extension Theorem. 1] and x ∈ [−1. where a. By the StoneWeierstrass Theorem. ε ≤ a ≤ b ≤ 1 − ε and a. Deﬁne X and ϕ as above. . Replacing f by f − Q. g ◦ ϕ ∈ A so g ∈ ϕ∗ [A] by uniqueness. there exist n continuous functions g1 . Then x ∈ f −1 [Ox ] and ˜ ˜−1 [Oy ]. we may assume that a = b = c = 0. 2 ) and ϕ(x) < x on ( 1 . 1 − ε] for 2 each n where ϕn is an iterate of ϕ. Then use the StoneWeierstrass Theorem to approximate f by a polynomial R with rational coeﬃcients such that R(−1) = R(0) = R(1) = 0. Let f (−1) = a. *49b. Let Q(x) = (a − b + c)x2 + cx + b. Let I = [−1. where f = f ◦ ϕ.
· is continuous. if x = 1 and δ > 0. we have y ≤ y − x + x < (1 − x)/2 + x < 1 so y ∈ S. For any sequence xn in ¯ S that converges to some x. Let M be a ﬁnitedimensional linear manifold in a normed vector space X with M = {x1 . we have xn  → x so x ≤ 1. Hence S = {x : x ≤ 1}. we have x1 . p2 ∈ P . Consider + as a function from X × X into X. 5a. x2 ∈ M and δ > 0. Also. 10. M ⊂ λM since m = 1 · m for all m. Then M ⊂ M + M since m = m + θ for all m. xn }. any linear manifold containing A will contain M . Consider · as a function from R × X into X. then M + M ⊂ M and λM ⊂ M for each λ so M + M = M and λM = M . Then M is a linear manifold containing A. Since (x1 + y1 ) − (x2 + y2 ) ≤ x1 − x2  + y1 − y2 . For λ1 . Hence M is a linear manifold. λ1 p1 + λ2 p2 is still a polynomial on [0. Also. 4. ¯ 6b. λ2 ∈ R. 1] so λ1 p1 + λ2 p2 ∈ P . Each x ∈ X can be written as a unique linear combination λ1 x1 + · · · + λn xn . 7. 1]. Consider the set M of all ﬁnite linear combinations of the form λ1 x1 + · · · + λn xn with xi ∈ A. Then (λ1 x1 + λ2 x2 ) − (λ1 y1 + ¯ ¯ λ2 y2 ) ≤ λ1  x1 − y1  + λ2  x2 − y2  < δ. Hence M is closed. x2 ∈ Mi for all i. Thus S ⊂ {x : x ≤ 1}. x2 ∈ M . y2 ∈ M such that x1 − y1  < δ/2λ1 and x2 − y2  < δ/2λ2 (We may assume λ1 . m2 ∈ M and n1 . Thus ¯ ¯ ¯ x ∈ S and {x : x ≤ 1} ⊂ S. 3. Thus also λ1 x1 + λ2 x2 ∈ M for λ1 . Given a set A in a vector space X. When y − x < δ. Hence M is the smallest linear manifold containing A. Let M and N be linear manifolds. Since cx − cy = c x − y. X is a linear manifold containing A. suppose M + M = M and λM = M . The metric ρ(x. 1]. . n2 ∈ N . if x ≡ y and y ≡ z. Thus P is a linear manifold in C[0. Thus λ1 x1 + λ2 x2 ∈ M so M is a linear manifold. i. any manifold containing M ∪ N will also contain M + N . every continuous ¯ function on [0. For any λ1 . We may deﬁne a norm n x1 = i=1 λi  and see that  · 1 is equivalent to the original norm on X. Thus convergence under the (k) n original norm on X is equivalent to convergence of each sequence of coeﬃcients in R. y) = [ i=1 (xi − yi )2 ]1/2 is derived from the norm x = ( i=1 x2 )1/2 .e. For λ1 . Since each Mi is a linear manifold. Consider the family of linear manifolds containing A. λ2 ∈ R. If M is a linear manifold. 5b. Hence M + N = {M ∪ N }. 1] can be uniformly approximated by polynomials on [0. Then λx ∈ M for each λ ∈ R and x ∈ M . Let P be the set of all polynomials on [0. Also. where θ is the zero vector. x2 ∈ M . . ( n)−1 x+ ≤ x+ ≤ x ≤ (n(x+ )2 )1/2 = nx+ so x and x+ are equivalent. {A} = M . x = limk i=1 λi xi = i=1 λi xi ∈ M .e. Let S = {x : x < 1}. let λ = max(1 − δ/2. Also. . Then  xn  − x  ≤ xn − x → 0. λ2 ∈ R and p1 . 0). 6a. 1]. Conversely. 1]. P contains a continuous function that is not a polynomial. Hence M + N is a linear manifold. 1]. Then λx ∈ S and x − λx = 1 − λ < δ. m1 . The set P is not closed in C[0. 5c. let δ = (1 − x)/2. Then P ⊂ C[0. Finally. λ2 ∈ R and x1 . i. . Given x ∈ S. there exist y1 . n n 2. Let λi = limk λi for each i. The metric i n n ∗ ∗ ρ (x. Also. Let M be a nonempty set.1 Banach Spaces Introduction 1. y) = max xi − yi  is derived from the norm x+ = max xi . Deﬁne x ≡ y if x − y = 0. λ1 x1 + λ2 x2 ∈ M . 8. λ2 = 0). The intersection {A} of this family is a linear manifold containing A and it is the smallest such linear manifold. Then x ≡ x since 0x = 0x = 0. Now n−1 x+ ≤ x+ ≤ x∗ ≤ nx+ so x∗ and x+ √ √ are equivalent. λ1 x1 + λ2 x2 ∈ Mi for each i. λ1 (m1 +