Math 215 Lecture Notes

MATH 215
Mathematical Analysis
Lecture Notes
Anl Tas
September 2005
M.A. student at the Department of Economics, Bilkent University, Bilkent, 06800,

Ankara, Turkey. E-mail: aniltas@bilkent.edu.tr. This course was given by Prof. Mefharet
Kocatepe in the 2005 Bilkent Summer School.
Contents
1 The Real And Complex Number Systems 3
1.1 The Real Number System . . . . . . . . . . . . . . . . . . . . 3
1.2 Extended Real Numbers . . . . . . . . . . . . . . . . . . . . . 8
1.3 The Complex Field . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Sets And Functions 12
2.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Countable And Uncountable Sets . . . . . . . . . . . . . . . . 14
3 Basic Topology 18
3.1 Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Compact Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 The Cantor Set . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Connected Sets . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4 Sequences And Series 34
4.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Subsequences . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3 Cauchy Sequences . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 Upper And Lower Limits . . . . . . . . . . . . . . . . . . . . . 44
4.5 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6 Operations With Series . . . . . . . . . . . . . . . . . . . . . . 61
5 Continuity 64
5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2 Continuity And Compactness . . . . . . . . . . . . . . . . . . 67
5.3 Continuity And Connectedness . . . . . . . . . . . . . . . . . 69
5.4 Uniform Continuity . . . . . . . . . . . . . . . . . . . . . . . . 69
6 Sequences And Series Of Functions 73
6.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2 Uniform Convergence . . . . . . . . . . . . . . . . . . . . . . . 74
6.3 Uniform Convergence And Continuity . . . . . . . . . . . . . . 77
6.4 Uniform Convergence And Integration . . . . . . . . . . . . . 80
6.5 Uniform Convergence And Dierentiation . . . . . . . . . . . . 83
7 Figures 86
2
1 The Real And Complex Number Systems
1.1 The Real Number System
R : The set of real numbers
Q : The set of rational numbers (Numbers in the form
m
n
where m, n are
integers, n ,= 0)
Z : The set of integers, 0, 1, 1, 2, 2, 3, 3, . . .
N : The set of natural numbers, 1, 2, 3, 4, . . .
Proofs:
1) Direct Proofs
2) Indirect Proofs
a) Proof by Contraposition
b) Proof by Contradiction
1.1 Example (Direct Proof). If f is dierentiable at x = c then f is con-
tinuous at x = c.
Proof. Hypothesis: f is dierentiable at x = c, i.e.
lim
xc
f(x) f(c)
x c
= f
(c) exists and it is a number.

Claim: f is continuous at c, i.e. lim
xc
f(x) = f(c).
lim
xc
f(x) = lim
xc
_
f(x) f(c)
x c
(x c) +f(c)
_
= lim
xc
f(x) f(c)
x c
. .
f
(c)
lim
xc
(x c)
. .
0
+lim
xc
f(c)
. .
f(c)
= f(c)
1.2 Example (Indirect Proof: Proof by Contraposition). Let n be an integer.
If n
2
is even
. .
p
then n is even
. .
q
.
If p then q: p q. This is equivalent to: If not q then not p, i.e. q p.
If n is odd then n
2
is odd.
3
Proof. Assume n is odd then n = 2k + 1 for some integer k. Then
n
2
= 4k
2
+ 4k + 1 = 2(2k
2
+ 2k
. .
an integer
) + 1
So n
2
is odd.
1.3 Example (Indirect Proof: Proof by Contradiction). Show that
2 is
not a rational number.
Proof. If c
2
= 2
. .
p
then c is not rational
. .
q
. We assume p and q, proceed and
get a contradiction. So assume c
2
= 2 and c is rational. Then c =
m
n
where
m, n Z, n ,= 0 and m, n have no common factors. We have c
2
=
m
2
n
2
then
m
2
= 2n
2
. So m
2
is even and m is even. Then m = 2k for some integer k.
We get 4k
2
= 2n
2
or 2k
2
= n
2
. Then n
2
is even and n is even. Then n = 2
for some integer . So m and n have a common factor. A contradiction.
Some Symbols And Notation
p q : if p then q (p implies q)
p q : if p then q and if q then p (p i q)
: such that
: therefore, so
: for all, for every
: for some, there exists
1.4 Example. Consider the following two statements:
(1) x R y R, y < x
(2) y R x R, y < x not true
(1) says, given any real number x we can nd a real number y (depending
on x) such that y is smaller than x.
(2) says, there is a real number y that is smaller than every real number x.
This is false.
4
1.5 Example. Let A = p : p Q, p > 0, p
2
< 2 and B = p : p Q, p >
0, p
2
> 2.
Claim: A has no largest element and B has no smallest element.
Proof. Given any p Q, p > 0, let q = p
p
2
2
p+2
=
2p+2
p+2
. Then q Q and
q > 0. Let p A, i.e. p
2
< 2. Then show q A and p < q.
p
2
2 < 0 so q = p
p
2
2
p + 2
> p
q
2
2 =
_
2p + 2
p + 2
_
2
2 =
4p
2
+ 4 + 8p 2p
2
8 8p
p
2
+ 4 + 4p
=
2(p
2
2)
(p + 2)
2
< 0
So q
2
< 2. Then we have q A and A has no largest element.
Properties of R
1) R has two operations + and with respect to which it is a eld.
(i) x, y R, x +y R
(ii) x, y R, x +y = y +x (commutativity of +)
(iii) x, y, z R, x + (y +z) = (x +y) +z (associativity of +)
(iv) an element (0 element) such that x R, x + 0 = x
(v) x R an element x R such that x + (x) = 0
(vi) x, y R, x y R
(vii) x, y R, x y = y x
(viii) x, y, z R, x (y z) = (x y) z
(ix) an element 1 ,= 0 in R such that x R, x 1 = x
(x) x ,= 0 in R there is an element
1
x
in R such that x
1
x
= 1
(xi) x, y, z R, x (y +z) = x y +x z
1.6 Remark. Note that Q with + and is also a eld.
2) R is an ordered eld, i.e. there is a relation < with the following properties
(i) x, y R one and only one of the following is true:
x < y, x = y, y < x (Trichotomy Law)
5
(ii) x, y, z R, x < y and y < z x < z (Transitive Law)
(iii) x, y, z R, x < y x +z < y +z
(iv) 0 < x and 0 < y 0 < x y
1.7 Remark. Note that Q is also an ordered eld.
3) R is complete.
1.8 Denition. Let E R. We say E is bounded above if there is an element
b R such that x E we have x b. b is called an upper bound for E.
1.9 Example. E = p : p Q, p > 0, p
2
< 2. Then b =
3
2
, b = 2, b = 5,
b = 100, b =
2 are all upper bounds for E.

1.10 Example. E = N = 1, 2, 3, 4, . . . is not bounded above.
1.11 Remark. Bounded below and lower bound are dened analogously.
1.12 Denition. Let E R be bounded above. A number b R is called
a least upper bound (lub) or supremum (sup) of E if
(i) b is an upper bound for E
and
(ii) if b
any upper bound for E, then b b
1.13 Remark. sup E need not be a member of E. If sup E is in E then it

is called the maximum element.
Completeness Property (or Least Upper Bound Property) of R:
Every non-empty set E R that is bounded above has a least upper bound
in R.
1.14 Remark. Q does not have LUB property.
Proof. Let E = p : p Q, p > 0, p
2
< 2 is a non-empty subset of Q that
is bounded above but it has no least upper bound in Q. Assume b Q is a
least upper bound of E. Since p = 1 E and p b we have 1 b. We have
two possibilities:
6
(i) b E
(ii) b / E
If b E, then q E such that b < q. Then b cannot be an upper bound.
So b / E. Since b Q and 0 < b, we have that b
2
< 2 is not true. By
trichotomy law, we have either b
2
= 2 or 2 0, 2 0 and p
2
< 2. Also q F so q > 0 and 2 < q
2
.
p
2
< 2 and 2 < q
2
p
2
< q
2
p < q
So q is bigger than every p E, i.e. q is an upper bound for E. Then b q.
Also q < b. Contradiction.
1.15 Remark. Analogously we have greatest lower bound (glb) or inmum
(inf) and the Greatest Lower Bound Property.
1.16 Theorem.
(a) (Archimedian Property) For every x, y R, x > 0, there is n N
(depending on x, y) such that nx > y.
(b) (Q is dense in R) For every x, y R with x < y, p Q such that
x 0 for which
there is no n N such that nx > y. So for all n N we have nx y.
Let E = nx : n N = x, 2x, 3x, . . .. Then y is an upper bound
for E and E ,= . So E has a least upper bound, say R. Since
x > 0, x < . Then x is not an upper bound for E. So there
is an element of E, say mx (where m N) such that x < mx.
Then < (m + 1)x. This element of E is bigger than = sup E.
Contradiction.
7
(b) Let x, y R be such that x < y. Then y x > 0. By part (a),
n N such that n(y x) > 1, i.e. ny > 1 + nx. In (a), replace
y by nx and x by 1 > 0. So m
1
N such that m
1
> nx. Let
A = m : m Z, nx < m. Then A ,= since m
1
A. Since A
is non-empty set of integers which is bounded below, A has a smallest
element m
0
. Then nx < m
0
and m
0
Z. m
0
1 / A. So we have
nx m
0
1. So m
0
1 nx < m
0
. Then nx < m
0
, m
0
1 +nx and
1 + nx < ny. So nx < m
0
< ny. If we divide this by n > 0, we get
x <
m
0
n
< y.
m
0
n
is a rational number.
Uniqueness of Least Upper Bound: Assume E R is non-empty and
bounded above. Then E has only one least upper bound.
Proof. Assume b, b
are two least upper bounds for E.

(i) b = sup E and b
is an upper bound for E b b
(ii) b
= sup E and b is an upper bound for E b
b
Then b = b
.
1.17 Fact. Let E R be non-empty and bounded above and let R.
Then
= sup E (i) x E, x
and
(ii) Given any > 0, y E such that < y
1.2 Extended Real Numbers
In the set of extended real numbers we consider the set R , +.
Preserve the order of R and x R, set < +. This way every non-
empty subset E of R , + has a least upper bound and greatest
lower bound in R , +. For example, if E = N, then sup E = +.
Also for x R, we dene the following
x + = +, x + () = ,
x
+
=
x
= 0
If x > 0, we dene x (+) = + and x () =
If x < 0, we dene x (+) = and x () = +
0 () is undened.
8
1.3 The Complex Field
Let C denote the set of all ordered pairs (a, b) where a, b R. We say
(a, b) = (c, d) a = c and b = d
Let x = (a, b), y = (c, d) C. We dene
x +y = (a +c, b +d) and x y = (ac bd, ad +bc)
Under these operations C becomes a eld with (0, 0), (1, 0) being the zero
element and multiplicative unit.
Dene : R C by (a) = (a, 0). Then
(a +c) = (a +c, 0) = (a, 0) + (c, 0) = (a) +(c)
(ac) = (ac, 0) = (a, 0)(c, 0) = (a) (c)
is 1-1 (one-to-one), i.e. if a ,= a
then (a) ,= (a
). This way we can

identify R with the subset (a, 0) : a R by means of .
Let i = (0, 1). Then i
2
= (0, 1) (0, 1) = (1, 0) = (1). Also if (a, b) C,
then
(a) +i(b) = (a, 0) + (0, 1)(b, 0) = (a, 0) + (0, b) = (a, b)
If we identify (a) with a then we identify (a, b) with a +ib. So C is the set
of all imaginary numbers in the form a +ib where a, b R and i
2
= 1.
If z = x +iy C, x, y R we dene
z = x iy (Conjugate of z)
x = Re z (Real Part of z)
y = Im z (Imaginary Part of z)
If z, w C, we have
z +w = z +w, zw = z w, Re z =
z +z
2
, Im z =
z z
2i
Since zz = x
2
+y
2
0, we dene the modulus of z as [z[ =
zz =
_
x
2
+y
2
.
1.18 Proposition. Let z, w C. Then
(a) z ,= 0 [z[ > 0 and z = 0 [z[ = 0
(b) [z[ = [z[
(c) [zw[ = [z[[w[
9
(d) [Re z[ [z[ (i.e. [x[
_
x
2
+y
2
)
(e) [z +w[ [z[ +[w[ (Triangle Inequality)
Proof of (e).
[z +w[
2
= (z +w)(z +w) = (z +w)(z +w)
= zz +zw +wz
. .
2Re(zw)
+ww = [z[
2
+ 2 Re(zw)
. .
|zw|=|z||w|
+[w[
2
[z[
2
+ 2[z[[w[ +[w[
2
= ([z[ +[w[)
2
So [z + w[
2
([z[ + [w[)
2
. If we take positive square root of both sides, we
get [z +w[ [z[ +[w[.
1.19 Theorem (Cauchy-Schwarz Inequality). Let a
j
, b
j
C where j =
1, 2, . . . , n. Then
j=1
a
j
b
j
_
n
j=1
[a
j
[
2
__
n
j=1
[b
j
[
2
_
Proof. Let us dene
A =
n
j=1
[a
j
[
2
, B =
n
j=1
[b
j
[
2
, C =
n
j=1
a
j
b
j
We have B 0. If B = 0, then all b
j
= 0. Then inequality becomes 0 0.
So assume B > 0. Then
0
n
j=1
[Ba
j
Cb
j
[
2
=
n
j=1
(Ba
j
Cb
j
)
_
Ba
j
C b
j
_
=
n
j=1
_
B
2
[a
j
[
2
BCa
j
b
j
CBb
j
a
j
+[C[
2
[b
j
[
2
_
= B
2
A BCC
. .
B|C|
2
CBC
. .
|C|
2
B
+[C[
2
B
= B
_
AB [C[
2
_
So 0 B(AB [C[
2
) and since B > 0, we have AB [C[
2
0. Then
[C[
2
AB.
10
Replacing b
j
by b
j
and using z = z, [z[ = [z[ we get
j=1
a
j
b
j
_
n
j=1
[a
j
[
2
__
n
j=1
[b
j
[
2
_
If a
j
0, b
j
0 are real numbers then
_
n
j=1
a
j
b
j
_
2
_
n
j=1
a
2
j
__
n
j=1
b
2
j
_
11
2 Sets And Functions
2.1 General
Let X and Y be two non-empty sets. A function f : X Y is a rule which
assigns to each x X a unique element y = f(x) in Y .
2.1 Example. Given x R, consider its decimal expansion in which there
is no innite chain of 9s. x =
1
4
is represented as 0.25000 instead of
0.24999 Let y = f(x) be the fortyninth digit after the decimal point. We
have f : R 0, 1, 2, . . . , 9.
2.2 Denition. Let f : X Y , A X. We dene the image of A under
f as the set f(A) = y Y : x A such that y = f(x). Let B Y . We
dene the inverse image of B under f as the set f
1
(B) = x X : f(x)
B.
2.3 Example. f : R R, f(x) = x
2
. Then
A = (1, 2] then f(A) = (1, 4]
A = (1, 3) then f(A) = [0, 9)
B = [1, 4] then f
1
(B) = [2, 1] [1, 2]
B = [1, 4] then f
1
(B) = [2, 2]
B = [2, 1] then f
1
(B) =
2.4 Denition. Let f : X Y be a function. Then we dene
X: Domain of f and f(X): Range of f
If f(X) = Y , then f is called a surjection or onto.
If x
1
,= x
2
then f(x
1
) ,= f(x
2
) (equivalently f(x
1
) = f(x
2
) x
1
= x
2
) then
f is called one-to-one (1-1) or an injection.
If f is both an injection and a surjection, then f is called a bijection or a 1-1
correspondence.
Let f : X Y , A X, B Y . We have
(a) f(f
1
(B)) B. f(f
1
(B)) = B for all B f is onto
(b) A f(f
1
(A)). A = f(f
1
(A)) for all A f is 1-1
(c) f(A
1
A
2
) f(A
1
) f(A
2
). f(A
1
A
2
) = f(A
1
) f(A
2
) for all A
1
and A
2
f is 1-1
12
(d) f(A
1
A
2
) = f(A
1
) f(A
2
)
Proof of (b).
A f
1
(f(A))
Let x A and let y = f(x). Then y f(A), i.e. f(x) f(A). So
x f
1
(f(A)) and A f
1
(f(A)).
A = f
1
(f(A)) for all A X f is 1-1
(): Assume f is 1-1. Show for all A X, A = f
1
(f(A)), i.e. show
A f
1
(f(A))
. .
always true
and f
1
(f(A)) A. So show f
1
(f(A)) A. Let x
f
1
(f(A)). Then f(x) f(A). Then x
A such that f(x) = f(x
).
Since f is 1-1, x = x
. So x A and f
1
(f(A)) A.
(): Assume A = f
1
(f(A)) for all A X. Show f is 1-1. Assume f is
not 1-1. Then there are x
1
, x
2
X such that x
1
,= x
2
and f(x
1
) =
f(x
2
). Let y = f(x
1
) = f(x
2
). Let A = x
1
. Then f(A) = y and
f
1
(f(A)) = x
1
, x
2
, . . .. Then A ,= f
1
(f(A)). Contradiction.
Let A
i
: i I be an arbitrary class of subsets of a set X indexed by a set
I of subscripts. We dene
_
iI
A
i
= x : x A
i
for at least one i I
iI
A
i
= x : x A
i
for every i I
If I = , then we dene
i
A
i
= and
i
A
i
= X. (If we require of an
element that it belongs to each set in the class and if there are no sets in the
class, then every element x X satises this requirement.) If A X we
dene A
C
= x : x / A complement of A.
_
_
iI
A
i
_
C
=
iI
A
C
i
,
_
iI
A
i
_
C
=
_
iI
A
C
i
(De Morgans Laws)
Let f : X Y , let A
i
: i I, B
j
: j J be classes of subsets of X and
Y .
f
_
_
iI
A
i
_
=
_
iI
f(A
i
) , f
_
iI
A
i
_
iI
f(A
i
)
f
1
_
_
jJ
B
j
_
=
_
jJ
f
1
(B
j
) , f
1
_
jJ
B
j
_
=
jJ
f
1
(B
j
)
For all B Y we have
13
(a) f
1
(B
C
) = (f
1
(B))
C
(b) f(A)
C
f(A
C
) for all A X f is onto
(c) f(A
C
) f(A)
C
for all A X f is 1-1
Proof of (a). Let x f
1
(B
C
). Then f(x) B
C
. Show x (f
1
(B))
C
.
Assume it is not true, i.e. x f
1
(B) f(x) B. So f(x) B
C
B
. .
.
Contradiction. Thus, x (f
1
(B))
C
. So, f
1
(B
C
) (f
1
(B))
C
.
Let x (f
1
(B))
C
. Show x f
1
(B
C
), i.e. f(x) B
C
. Assume it is
not true, i.e. f(x) B so x f
1
(B). Then, x (f
1
(B))
C
f
1
(B)
. .
.
Contradiction. So, (f
1
(B))
C
f
1
(B
C
).
If f : X Y is 1-1 onto, then y Y unique x X such that y = f(x).
Send Y X. This way we get f
1
: Y X (the inverse function of f)
f
1
(f(x)) = x x X, f(f
1
(y)) = y y Y .
2.2 Countable And Uncountable Sets
Let C be any collection of sets. For A, B C we write A B (and say A
and B are numerically equivalent) if there is a 1-1 correspondence f : A B.
has the following properties
(i) A A
(ii) A B B A
(iii) A C A C
2.5 Example. A = N = 1, 2, 3, . . ., B = 2N = 2, 4, 6, . . .. Then A B
by f : A B, f(n) = 2n.
2.6 Denition. For n = 1, 2, 3, . . . let J
n
= 1, 2, 3, . . . , n. Let X ,= be
any set. We say
X is nite if n N such that X J
n
.
X is innite if it is not nite.
X is countable (or denumerable) if X N.
X is uncountable if X is not nite and not countable.
14
X is at most countable if X is nite or countable.
2.7 Example. N, 2N are countable.
2.8 Example. Z = . . . , 3, 2, 1, 0, 1, 2, 3, . . . is countable. Dene f :
N Z as follows
N : 1, 2, 3, 4, 5, . . .

Z : 0, 1, -1, 2, -2, . . .
So we have
f(n) =
_
n
2
if n is even
n1
2
if n is odd
2.9 Example. Q
+
= q : q Q, q > 0 is countable. Given q Q
+
, we
have m, n Z m > 0, n > 0 such that q =
m
n
. List the elements of Q
+
in
mn 1 2 3 4
1
1
1
1
2
1
3
1
4

2
2
1
2
2
2
3
2
4

3
3
1
3
2
3
3
3
4

4
4
1
4
2
4
3
4
4

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
this order omitting the ones which are already listed before. Then we get the
following sequence
1
1
,
1
2
,
2
1
,
1
3
,
3
1
,
1
4
,
2
3
,
3
2
,
4
1

Dene f : N Q
+
as follows
f(1) =
1
1
, f(2) =
1
2
, f(3) =
2
1
, f(4) =
1
3

15
Suppose X is countable, so we have 1-1, onto function f : N X. Let
f(n) = x
n
. Then we can write the elements of X as a sequence
X = x
1
, x
2
, x
3
, . . .
2.10 Example. Q is countable. (Similar to the proof of Z is countable.)
2.11 Proposition. Let I be a countable index set and assume for every
i I, A
i
is a countable set. Then
iI
A
i
is countable. (Countable union of
countable sets is countable.)
2.12 Example. X = [0, 1) is not countable. Every x X has a binary
(i.e. base 2) expansion x = 0.x
1
x
2
x
3
x
4
where each of x
1
, x
2
, x
3
, . . . is 0 or
1. Assume X = [0, 1) is countable. Then write its elements as a sequence.
X = y
1
, y
2
, y
3
, . . .. Then write each of y
1
, y
2
, y
3
, . . . in the binary expansion.
y
1
= 0.y
1
1
y
1
2
y
1
3

y
2
= 0.y
2
1
y
2
2
y
2
3

y
3
= 0.y
3
1
y
3
2
y
3
3

.
.
.
Dene
z
1
=
_
0 if y
1
1
= 1
1 if y
1
1
= 0
z
2
=
_
0 if y
2
2
= 1
1 if y
2
2
= 0
z
n
=
_
0 if y
n
n
= 1
1 if y
n
n
= 0
Let z = 0.z
1
z
2
z
3
[0, 1). But z ,= y
1
, z ,= y
2
, . . . Contradiction.
2.13 Example. [0, 1], (0, 1), R, (a, b) are all uncountable. They are all
numerically equivalent.
2.14 Example. f :
_
2
,

2
_
R and f(x) = tan x gives
_
2
,

2
_
R.
2.15 Theorem (Cantor-Schroder-Bernstein). Let X, Y be two non-empty
sets. Assume there are 1-1 functions. f : X Y and g : Y X. Then
X Y .
16
2.16 Example. Let X = (x, y) : 0 < x < 1, 0 < y < 1. Then X (0, 1).
Dene g : (0, 1) X by g(x) =
_
x,
1
2
_
.
Dene f : X (0, 1) as follows: Given (x, y) X, write X and Y in their
decimal expansion with no innite chain of 9s.
x = 0.x
1
x
2
x
3
and y = 0.y
1
y
2
y
3

Let
f(x, y) = 0.x
1
y
1
x
2
y
2
x
3
y
3

Then f, g are 1-1 so by the above theorem X (0, 1).
17
3 Basic Topology
3.1 Metric Spaces
In R, R
k
= (x
1
, . . . , x
k
) : x
1
, . . . , x
k
R we have the notion of distance.
In R, d(x, y) = [x y[
In R
k
, x = (x
1
, . . . , x
k
), y = (y
1
, . . . , y
k
). We have
d(x, y) =
_
(x
1
y
1
)
2
+ + (x
k
y
k
)
2
3.1 Denition. Let X ,= be a set. Suppose we have a function
d : X X
. .
{(x,y):x,yX}
R
with the following properties
(i) x, y X, d(x, y) 0 d(x, y) = 0 x = y
(ii) x, y X, d(x, y) = d(y, x)
(iii) x, y, z X, d(x, y) d(x, z) +d(z, y) (Triangle Inequality)
d is called a metric and the pair (X, d) is called a metric space.
Proof of (iii). X = R
k
, d(x, y) =
_
(x
1
y
1
)
2
+ + (x
k
y
k
)
2
then d sat-
ises (i) and (ii). For triangle inequality, let x, y, z R
k
be given. Then we
have
d(x, y)
2
=(x
1
y
1
)
2
+ + (x
k
y
k
)
2
=((x
1
z
1
) + (z
1
y
1
))
2
+ + ((x
k
z
k
) + (z
k
y
k
))
2
=(x
1
z
1
)
2
+ + (x
k
z
k
)
2
+ (z
1
y
1
)
2
+ + (z
k
y
k
)
2
+ 2[(x
1
z
1
)(z
1
y
1
) + + (x
k
z
k
)(z
k
y
k
)]
(d(x, z))
2
+ (d(z, y))
2
+ 2 [(x
1
z
1
)(z
1
y
1
) + + (x
k
z
k
)(z
k
y
k
)[
. .
((x
1
z
1
)
2
++(x
k
z
k
)
2
)
1/2
((z
1
y
1
)
2
++(z
k
y
k
)
2
)
1/2
(from Cauchy-Schwarz inequality)
(d(x, z))
2
+ (d(z, y))
2
+ 2d(x, z)d(z, y)
(d(x, z) +d(z, y))
2
18
3.2 Example.
1) X ,= any set. Given any x, y X let
d(x, y) =
_
1 if x ,= y
0 if x = y
(i) and (ii) are trivially true. Check (iii).
Proof of (iii). If d(x, y) = 0, then it is true as RHS 0. If d(x, y) = 1, then
we cannot have RHS = 0. If RHS = 0 we have, d(x, z) = 0 and d(z, y) = 0.
That is x = z and z = y. So x = y. Then d(x, y) = 0. Contradiction.
This metric is called the discrete metric.
2) Let X = R
k
= x = (x
1
, . . . , x
k
) : x
1
, , x
k
R
d
1
(x, y) = [x
1
y
1
[ +[x
2
y
2
[ + +[x
k
y
k
[
Proof of (iii). Given x, y, z R
k
d
1
(x, y) = [x
1
y
1
[ + +[x
k
y
k
[
= [(x
1
z
1
) + (z
1
y
1
)[ + +[(x
k
z
k
) + (z
k
y
k
)[
[x
1
z
1
[ +[z
1
y
1
[ + +[x
k
z
k
[ +[z
k
y
k
[
= d
1
(x, z) +d
1
(z, y)
d
1
is called the
1
metric on R
k
.
3) X = R
k
d
(x, y) = max[x
1
y
1
[, . . . , [x
k
y
k
[
Proof of (iii). Given x, y, z R
k
d
(x, y) = max[x
1
y
1
[, . . . , [x
k
y
k
[
= [x
i
y
i
[ (1 i k)
= [(x
i
z
i
) + (z
i
y
i
)[ [x
i
z
i
[
. .
max
+[z
i
y
i
[
. .
max
d
(x, z) +d
(z, y)
19
d
is called the
metric on R
k
.
4) Let S be any xed non-empty set. A function f : S R is called bounded
if f(S) is a bounded subset of R, i.e. there are two numbers A, B such that
s S, A f(s) B.
For example, f : R R, f(s) = arctan s is bounded, f(s) =
s
2
+ 1 is
unbounded.
Let X = B(s) = all bounded functions f : S R. Let f, g B(S), then
f g is also bounded. We dene
d(f, g) = sup[f(s) g(s)[ : s S
Geometrically, it is the supremum of the vertical distances between the two
graphs (See Figure 1). (i) and (ii) are trivially true. Check (iii). We need the
following: Let A, B be two non-empty subsets of R that are bounded above.
Then, sup(A + B) sup A + sup B. A + B = x + y : x A, y B. Let
a = sup A and b = sup B. Let z A + B be arbitrary. Then x A, y B
such that z = x+y. Then we have x+y a +b. Since z = x+y, z a +b.
So a+b is an upper bound for the set A+B. Since supremum is the smallest
upper bound, sup(A+B) a+b. In fact we have sup(A+B) = sup A+sup B.
Show sup(A + B) sup A + sup B. Given > 0, x A and y B such
that a < x and b < y. Then a +b 2 < x +y. Since x +y A+B,
we have x + y sup(A + B). Then a + b 2 < sup(A + B). Here a, b,
sup(A + B) are xed numbers, i.e. they do not depend on > 0. We have
a+b < sup(A+B)+2 true for every > 0. Then we have a+b sup(A+B).
If it is not true, then a + b > sup(A + B). Let =
a+bsup(A+B)
4
then > 0.
So we have
a +b <sup(A +B) + 2
<sup(A +B) +
a +b sup(A +B)
2
a +b
2
<
sup(A +B)
2
Contradiction.
Proof of (iii). Given f, g, h B(S), let
C = [f(s) g(s)[ : s S
A = [f(s) h(s)[ : s S
B = [h(s) g(s)[ : s S
20
Then sup C = d(f, g), sup A = d(f, h) and supB = d(h, g). Let x C. Then
s S such that x = [f(s) g(s)[. Then
x = [f(s) g(s)[
= [(f(s) h(s)) + (h(s) g(s))[
[f(s) h(s)[
. .
call y
+[h(s) g(s)[
. .
call z
= y +z A +B
So x C, there is an element u A + B such that x u. Then, sup C
sup(A +B) sup A + sup B. That is, d(f, g) d(f, h) +d(h, g).
3.3 Denition. Let (X, d) be a metric space, p X and r > 0. Then
The open ball centered at p of radius r is dened as the set
B
r
(p) = x X : d(x, p) < r
The closed ball centered at p of radius r is dened as the set
B
r
[p] = x X : d(x, p) r
3.4 Example.
1) X = R
k
, d
2
(x, y) =
_
(x
1
y
1
)
2
+ + (x
k
y
k
)
2
. (See Figure 2).
2) X ,= , d: discrete metric. Then
B
r
(p) =
_
p if r 1
X if r > 1
B
r
[p] =
_
p if r < 1
X if r 1
3) X = R
k
with
1
metric d
1
. Let X = R
2
, p = (p
1
, p
2
), x = (x
1
, x
2
).
Then [x
1
p
1
[ + [x
2
p
2
[ < r. If p
1
= p
2
= 0 we have [x
1
[ + [x
2
[ < r.
(See Figure 3).
4) X = R
k
with
metric d
. Let X = R
2
, p = (0, 0), x = (x
1
, x
2
).
Then max[x
1
[, [x
2
[ < r. (See Figure 4).
5) X = B(S) with sup metric. Let S = [a, b]. Then B
r
(f) is the set of all
functions g whose graph is in the shaded area. (See Figure 5).
21
3.5 Denition. Let (X, d) be a metric space and E a subset of X.
1) A point p E is called an interior point of E if r > 0 such that
B
r
(p) E. (See Figure 6).
2) The set of all interior points of E is called the interior of E. It is
denoted by int E or E
. We have int E E. (See Figure 7).

3) E is said to be open if int E = E, i.e. if every point of E is an interior
point of E, i.e. p E r > 0 such that B
r
(p) E. (See Figure 8).
3.6 Proposition. Every open ball B
r
(p) is an open set.
Proof. Let q B
r
(p). Show that s > 0 such that B
s
(q) B
r
(p). Since
q B
r
(p) we have d(q, p) < r. So r d(q, p)
. .
let this be s
> 0. Show B
s
(q) B
r
(p). Let
x B
s
(q), i.e. d(x, q) < s. Then
d(x, p) d(x, q) +d(q, p)
< s +d(q, p)
< r d(q, p) +d(q, p)
< r
So x B
r
(p).
4) Let p X. A subset N X is called a neighborhood of p if p int N.
(See Figure 9).
B
r
(p) is a neighborhood of p or B
r
(p) is a neighborhood of all of its
points.
5) A point p X is called a limit point (or accumulation point or cluster
point) of E if every neighborhood N of p contains a point q E with
q ,= p. (See Figure 10).
6) A point p E is called an isolated point of E if p is not a limit point
of E, i.e. if there is a neighborhood N of p such that N E = p.
3.7 Example. Let X = R, E = 1,
1
2
,
1
3
,
1
4
, . . ., d(x, y) = [x y[. Isolated
points of E are 1,
1
2
,
1
3
,
1
4
, . . . E has only one limit point that is 0. Given any
open ball B
r
(0) = (r, r), nd n N such that
1
r
< n. Then x =
1
n

B
r
(0) E and x ,= 0.
22
7) E
is the set of all limit points of E. We dene E = EE
. E is called
the closure of E.
We have p E for every neighborhood N of p we have N E ,= .
8) E is closed if every limit point of E is an element of E, i.e. E
E,
i.e. E = E. (See Figure 11).
9) E is perfect if E is closed and has no isolated points. (See Figure 12).
10) E is bounded if m > 0 such that x, y E d(x, y) m. (See Fig-
ure 13).
11) E is dense in X if E = X, i.e. p X and for all neighborhood N of
p we have N E ,= .
3.8 Example. Let X = R, E = Q. We have Q = R. Given p R and given
a neighborhood B
r
(p) = (p r, p + r), nd a rational number x such that
p r < x < p +r. So x B
r
(p) Q.
3.9 Theorem. E is open E
C
is closed.
Proof.
(): Let E be open. Show that every limit point p of E
C
is an element of E
C
.
Assume it is not true. Then E
C
has a limit point p
0
such that p
0
/ E
C
.
Then p
0
E. Since E is open r > 0 such that B
r
(p
0
) E. Also
since p
0
is a limit point of E
C
, every neighborhood N of p
0
contains a
point q E
C
such that q ,= p
0
. Since B
r
(p
0
) is also a neighborhood
of p
0
we have that B
r
(p
0
) contains a point q E
C
such that q ,= p
0
.
Then q B
r
(p
0
) E but also q E
C
. Contradiction.
(): Let E
C
be closed. Show that every point p in E is an interior point.
Let p E be arbitrary. Since p / E
C
, p is not a limit point of E
C
.
Then p has a neighborhood N such that N does not contain any point
q E
C
with q ,= p. Then N E
C
= so N E. Since N is a
neighborhood of p, r > o such that B
r
(p) N. So B
r
(p) E.
3.10 Theorem. p is a limit point of E every neighborhood N of p contains
innitely many points of E.
Proof.
23
(): Trivial.
(): Let p be a limit point of E and let N be an arbitrary neighborhood of
p. Then r > 0 such that B
r
(p) N.
q
1
B
r
(p) E such that q
1
,= p. d(q
1
, p) > 0. Let r
1
= d(q
1
, p) < r.
q
2
B
r
1
(p) E such that q
2
,= p. d(q
2
, p) < r
1
= d(q
1
, p). Then
q
2
,= q
1
. d(q
2
, p) > 0. Let r
2
= d(q
2
, p) < r
1
.
Continue this way and get a sequence of points q
1
, q
2
, q
3
, . . . , q
n
, . . . in
E such that q
i
,= q
j
for i +j and also each q
i
,= p and q
i
B
r
(p).
3.11 Corollary. If E is a nite set then E has no limit points.
3.12 Theorem. Let E X. Then
(a) E is a closed set.
(b) E = E E is a closed set.
(c) E is the smallest closed set that contains E, i.e. if F is any closed set
such that E F then E F.
Proof.
(a) Show (E)
C
is an open set. Let p (E)
C
. Then r > 0 such that
B
r
(p) E = . This means B
r
(p) E
C
. Show that actually B
r
(p)
(E)
C
. If not true, q B
r
(p) such that q / (E)
C
, i.e. q E. Then for
every neighborhood N of q we have N E ,= . This is also true for
N = B
r
(p), i.e. B
r
(p) E ,= . But B
r
(p) E = . Contradiction. So
B
r
(p) (E)
C
.
(b) (): E is closed by (a) so E is closed.
(): Let E be closed. Then E is contains all of its limit points, i.e.
E
E, E = E E
E. Since E E is always true, we have

E = E.
(c) Let E be given and F be a closed set such that E F. Show E F.
Let p E = E E
. If p E then p F. If p E
show p F.
Given any neighborhood N of p, N contains innitely many points of
E so N contains innitely many points of F, i.e. p F
F.
Basic properties: Let (X, d) be a metric space.
24
1) The union of any collection of open sets is open.
2) The intersection of a nite number of open sets is open.
3) The intersection of any collection of closed sets is closed.
4) The union of a nite number of closed sets is closed.
5) E is open E
C
is closed.
6) E is closed E = E.
7) E is the smallest closed set that contains E.
8) E is open E = int E.
9) int E is the largest open set that is contained in E.
3.13 Example. Intersection of innitely many open sets may not be open.
Let X = R, d(x, y) = [x y[. For n = 1, 2, 3, . . . let E
n
=
_
1
n
,
n+1
n
_
. All
E
n
s are open but
n=1
E
n
= [0, 1] is not open.
3.14 Proposition. Let , = E R be bounded above. Let y = sup E. Then
y E.
Proof. Let N be an arbitrary neighborhood of y. Then r > 0 such that
B
r
(y) N. y r cannot be an upper bound for E. Then x E such that
yr < x. Also, x y < y+r. So yr < x < y+r, i.e. x B
r
(y) x N.
So x E N, i.e. E N ,= .
3.2 Subspaces
3.15 Denition. Let (X, d) be a metric space and Y ,= be a subset of X.
Then Y is a metric space in its own right with the same distance function.
In this case we say (Y, d) is a subspace of (X, d).
3.16 Example. X = R
2
, Y = (x, 0) : x R. Let E = (x, 0) : 1 < x <
2. Then E Y so E X. As a subset of Y , E is an open set. As a subset
of X, E is not an open set. Let E Y X. We say E is open (closed)
relative to Y if E is open (closed) as a subset of the metric space (Y, d).
E is open relative to Y p E r > 0 such that B
Y
r
(p)
. .
B
X
r
(p)Y
E.
E is closed relative to Y Y E is open relative to Y .
25
3.17 Theorem. Let E Y X. Then
(a) E is open relative to Y there is an open set F X such that
E = F Y .
(b) E is closed relative to Y there is a closed set F X such that
E = F Y .
Proof.
(a) (): Let E be open relative to Y . Then p E r
p
> 0 such that
B
X
r
p
(p) Y E. Let F =
pE
B
X
r
p
(p). Then F is an open set in
X. Show F Y = E.
F Y =
_
_
pE
B
X
r
p
(p)
_
Y =
_
pE
_
B
X
r
p
(p) Y
_
. .
E for all p
E
Conversely, let p
0
E. Then p
0
Y since E Y . Then p
0

B
X
r
p
0
F. So p
0
Y F.
(): Assume E = F Y where F X is open in X. Let p E.
Then p Y and p F. Since F is open in X, r > 0 such that
B
X
r
(p) F. Then B
X
r
(p) Y F Y = E.
(b) (): Let E Y be closed relative to Y . Then Y E is open relative to
Y . So there exists an open set A X such that Y E = A Y .
Then E = Y (Y E) = Y (A Y ) = (A Y )
C
Y = (A
C
Y
C
) Y = (A
C
Y ) (Y
C
Y )
. .
= A
C
Y . A
C
is closed in X
and we may call it F.
(): Assume E = F Y where F X is a closed set. Then Y E =
Y (F Y ) = Y (F Y )
C
= Y (F
C
Y
C
) = Y F
C
..
open in X
is
open relative to Y .
3.3 Compact Sets
3.18 Denition. Let (X, d) be a metric space and E be a non-empty subset
of X. An open cover of X is a collection of G
i
: i I of open subsets G
i
of X such that E
iI
G
i
.
26
3.19 Example. X = R, d(x, y) = [x y[, E = (0, 1). For every x E let
G
x
= (1, x).
xE
G
x
= (1, 1) E. So G
x
: x E is an open cover of
E.
3.20 Example. X = R
2
with d
2
and E = B
1
(0). For n N let G
n
=
B
n
n+1
(0).
n=1
G
n
= E. So G
n
: n N is an open cover of E.
3.21 Denition. A subset K of a metric space (X, d) is said to be compact
if every open cover of K contains a nite subcover, i.e. given any open cover
G
i
: i I of K, we have that i
1
, . . . , i
n
I such that K G
i
1
G
i
n
.
3.22 Example. X = R, d(x, y) = [x y[. E = (0, 1) is not compact. Take
I = E = (0, 1). For x I, G
x
= (1, x).
xI
G
x
= (1, 1) E. So,
G
x
: x I is an open cover for E. This open cover does not have any
nite subcover. Assume it is not true, so assume x
1
, . . . , x
n
E such that
G
x
1
G
x
n
E. Let x
k
= maxx
1
, . . . , x
n
. Then (0, 1) (1, x
k
) but
x
k
E = (0, 1), i.e. x
k
< 1. Let x =
x
k
+1
2
. Then x E but x / (1, x
k
).
3.23 Theorem. Let K Y X. Then K is compact relative to Y K
is compact relative to Y .
Proof.
(): Assume K is compact relative to Y . Let G
i
: i I be any collection of
sets open relative to X such that
iI
G
i
K. Then Y G
i
: i I is
a collection of sets open relative to Y and K = KY
_
iI
G
i
_
Y =
iI
(Y G
i
). So Y G
i
: i I is an open (relative to Y ) cover
of K. Since K is compact relative to Y , i
1
, . . . , i
n
I such that
K (Y G
i
1
) (Y G
i
n
). Then K G
i
1
G
i
n
.
(): Let K be compact relative to X. Let G
i
: i I be an arbitrary
collection of sets open relative to Y such that K

iI
G
i
. Then we
have G
i
= Y E
i
for some open subset E
i
of X. Then K
iI
(Y
E
i
)
iI
E
i
. So E
i
: i I is an open cover of K in X. Since K is
compact relative to X, i
1
, . . . , i
n
such that K E
i
1
E
i
n
. Then
K = K Y (E
i
1
Y ) (E
i
n
Y ) = G
i
1
G
i
n
.
3.24 Theorem. If K X is compact then K is closed.
27
Proof. We will show K
C
is open. Let p K
C
. Show r > 0 such that
B
r
(p) K
C
q K (since q ,= p). d(q, p) > 0. Let
r
q
=
d(q, p)
2
V
q
= B
r
q
(p) W
q
= B
r
q
(q)
Then V
q
W
q
= . Find V
q
and W
q
q K K

qK
W
q
. The collection
W
q
: q K is an open cover of K. Since K is compact, q
1
, . . . , q
n
K
such that K W
q
1
W
q
n
. Let V = V
q
1
V
q
n
. Then V is an open
set and p V . If r
q
k
= minr
q
1
, . . . , r
q
n
then V = B
r
q
k
(p). Show V K
C
.
If it is not true, then z V but z / K
C
, i.e. z K. Then z W
q
i
for
some i 1, . . . , n. z V V
q
i
(the same i). So z W
q
i
V
q
i
= .
Contradiction.
3.25 Theorem. Closed subsets of compact sets are compact.
Proof. Let F K X where K is compact and F is closed. (relative to X
and relative to K are the same since K is closed.) Let G
i
: i I be a set
open in X such that F
iI
G
i
. Then G
i
: i IF
C
is an open cover
of K. Since K is compact, i
1
, . . . , i
n
I such that K G
i
1
G
i
n
F
C
.
Then F = F K (G
i
1
F) (G
i
n
F)(F
C
F)
. .
G
i
1
G
i
n
.
3.26 Corollary. Let F, K X where K is compact and F is closed. Then
F K is compact.
3.27 Theorem. Let K
i
: i I be a collection of compact subsets of X
such that the intersection of every nite subcollection of K
i
: i I is
non-empty.
iI
K
i
,= .
Proof. Assume the contrary.
iI
K
i
= . Fix one of these sets, K
i
0
. Then
K
i
0

_
i=i
0
K
i
_
= . Then K
i
0

_
i=i
0
K
i
_
C
, i.e. K
i
0

i=i
0
K
C
i
. So
K
C
i
: i ,= i
0
is an open cover of K
i
0
. Since K
i
0
is compact, i
1
, . . . , i
n
such
that K
i
0
K
C
i
1
K
C
i
n
. .
(K
i
1
K
i
n
)
C
. But K
i
0
K
i
1
K
i
n
= . This contradicts
the hypothesis.
3.28 Corollary. Let K
n
: n N be a sequence of non-empty compact
sets such that K
1
K
2
K
3
Then
n=1
K
n
,= .
28
3.29 Example. X = R, d(x, y) = [x y[, E
n
= [n, +) = x : x R, x
n, n = 1, 2, 3, . . . Then E
n
s are closed, E
n
,= and E
1
E
2
E
3

We have
n=1
E
n
= .
3.30 Theorem (Nested Intervals). Let I
n
: n = 1, 2, . . . be a sequence
of non-empty, closed, bounded intervals in R such that I
1
I
2
I
3

Then
n=1
I
n
,= .
Proof. Let I
n
= [a
n
, b
n
] and a
n
b
n
. That is, a
1
a
2
a
3
and
b
3
b
2
b
1
. We also have that n k a
n
b
k
. Given n, k N
(i) If n = k then a
n
b
n
= b
k
.
(ii) If n > k then I
n
I
k
. Then a
n
I
n
I
k
a
k
a
n
b
k
.
(iii) If n < k then I
k
I
n
. Then b
k
I
k
I
n
a
n
b
k
b
n
.
Let E = a
1
, a
2
, a
3
, . . .. Then E ,= and E is bounded above. Let x =
sup E. Then a
n
x for all n.
Claim: x b
n
n.
Assume it is not true. Then n such that b
n
0
< X. Then b
n
0
cannot be an
upper bound for E. So there is an element a
k
0
E such that a
k
0
> b
n
0
.
Contradiction. So n we have a
n
x b
n
, i.e. x I
n
. So x
n=1
I
n
.
3.31 Remark. If also lim
n
(b
n
a
n
) = 0 then
n=1
I
n
consists of only
one point.
3.32 Denition. Let a
1
b
1
, . . . , a
k
b
k
be real numbers. Then the set of
all points p = (x
1
, . . . , x
k
) R
k
such that a
1
x
1
b
1
, . . . , a
k
x
k
b
k
is
called a k-cell in R
k
. (See Figure 14).
3.33 Theorem. Let k be xed. Let I
n
be a sequence of non-empty k-cells
such that I
1
I
2
I
3
Then
n=1
I
n
,= .
3.34 Theorem. Let K be a compact subset of a metric space (X, d). Then
K is bounded.
Proof. Fix a point p
0
K. X =
n=1
B
n
(p
0
) and K

n=1
B
n
(p
0
).
So n
1
, . . . n
r
N such that K B
n
1
(p
0
) B
n
r
(p
0
). Let n
0
=
maxn
1
, . . . , n
r
. Let p, q K, then p B
n
i
(p
0
) and q B
n
j
(p
0
) where n
i
, n
j
are one of n
1
, . . . , n
r
. We have d(p, q) d(p, p
0
)+d(p
0
, q) n
i
+n
j
2n
0
.
29
3.35 Theorem (Heine-Borel). A subset K of R
k
is compact K is closed
and bounded.
Proof.
(): True in all metric spaces.
(): Let K R
k
be closed and bounded. Since K is bounded, there is a
k-cell I such that K I. I is compact, so K being a closed subset of
the compact set I is compact.
3.36 Theorem. Every k-cell is a compact set in R
k
.
Proof. Let E R
k
be a k-cell. Then there are real numbers
a
1
, b
1
, a
2
, b
2
, . . . , a
k
, b
k
such that a
1
b
1
, a
2
b
2
, . . . , a
k
b
k
and
E = p = (x
1
, . . . , x
k
) R
k
: a
1
x
1
b
1
, . . . , a
k
x
k
b
k
If a
1
= b
1
, a
2
= b
2
, . . . , a
k
= b
k
then E consists of one point which is compact.
So assume there is at least one j, 1 j k, such that a
j
 0. Assume E is not compact. So there is an open cover G
:
A such that no nite subcollection of G
s covers E. Let
c
i
=
a
i
+b
i
2
We divide each side of E into two parts and this way we divide E into 2
k
subcells. Call them Q
1
, Q
2
, . . . , Q
2
k. Then at least one of these Q
j
s cannot
be covered by nitely many sets, G
s. Call this Q
j
E
1
. For all p, q E we
have d
2
(p, q) and for all p, q E
1
, d
2
(p, q)

2
. Next divide E
1
into 2
k
subcells by halving each side and continue this way. This way we obtain a
sequence E
n
of k-cells such that
(a) E E
1
E
2
E
3

(b) E
n
cannot be covered by any nite subcollection of G
: A
(c) For all p, q E
n
, d
2
(p, q)

2
n
30
By (a),
n=1
E
n
,= . Let p
n=1
E
n
. Then p
E. Since G
: A is
an open cover of E, there is an
0
A such that p
0
. Since G
0
is open,
there is r > 0 such that B
r
(p
) G
0
. Find a natural number n
0
such that
r
< 2
n
0
, i.e.

2
n
0
< r. Now we show that E
n
0
G
0
. p
E
n
0
. Let p E
n
0
be an arbitrary point. By (c), d
2
(p, p
)

2
n
0
. Also,

2
n
0
< r. So d
2
(p, p
) < r.
So p B
r
(p
) G
0
. Thus, p E
n
0
p G
0
. So E
n
0
G
0
. This means
E
n
0
can be covered by nitely many sets from G
: A (indeed just by
one set). This contradicts (b).
3.37 Theorem. Let (X, d) be a metric space and K X. Then K is
compact every innite subset of K has a limit point in K.
Proof.
(): Let K be compact. Assume claim is not true. Then there is an innite
subset A K such that A has no limit point in K. So, given any
point p K, p is not a limit point of A. So there is r
p
> 0 such
that B
r
p
(p) contains no point of A dierent from p. The collection of
B
r
p
(p) : p K is an open cover of K. Since K is compact, there are
p
1
, p
2
, . . . , p
n
K such that K B
r
p
1
(p
1
) B
r
p
2
(p
2
) B
r
p
n
(p
n
).
Since A K and A is an innite set, one of the open balls on the right
hand side must contain innitely many points of A. Contradiction.
(): Omitted.
3.38 Theorem (Bolzano-Weierstrass). Every innite, bounded subset E of
R
k
has a limit point in R
k
.
Proof. Since E is bounded, there is a k-cell I such that E I. Then since
I is compact, the innite subset E of I has a limit point p I R
k
.
3.4 The Cantor Set
Let us dene
E
0
= [0, 1]
E
1
=
_
0,
1
3
_
_
2
3
, 1
_
E
2
=
_
0,
1
9
_
_
2
9
,
1
3
_
_
2
3
,
7
9
_
_
8
9
, 1
_
.
.
.
31
Continue this way by removing the open middle thirds of the remaining
intervals. This way we get a sequence E
1
E
2
E
3
E
n
such
that E
n
is the union of 2
n
disjoint closed intervals of length
1
3
n
. We dene
( =
n=1
E
n
which is called the Cantor Set.
Properties of (
(1) ( is compact
(2) ( , =
(3) int ( =
(4) ( is perfect
(5) ( is uncountable
Proof of (3). Assume int ( , = . Then there is a non-empty open interval
(, ) ( and ( does not contain intervals of the form
_
3k+1
3
m
,
3k+2
3
m
_
. Since
they are removed in the process of construction. Assume ( contains (, )
where < . Let a > 0 be a constant which will be determined later. Choose
m N such that
a
< 3
m
, i.e.
1
3
m
<

a
. Let k be the smallest integer
such that <
3k+1
3
m
, i.e.
3
m
1
3
< k. Then k 1
3
m
1
3
. Show
3k+2
3
m
< .
k 1
3
m
1
3
k
3
m
1
3
+ 1
Show
3
m
1
3
+ 1 <
3
m
2
3
1 <
3
m
2
3

3
m
1
3
1 <
3
m
( ) 1
3
Now we have
3
m
( ) 1
3
>
3
m
a
1
3
m
1
3
=
a 1
3
So let
a1
3
1, i.e. a 4. Choose a = 4. This way we have that
_
3k + 1
3
m
,
3k + 2
3
m
_
(, )
_
3k + 1
3
m
,
3k + 2
3
m
_
(
Contradiction.
32
3.5 Connected Sets
3.39 Denition. Let (X, d) be a metric space and A, B X. We say A
and B are separated if A B = and A B = .
3.40 Example. X = R, A = Q, B = RQ. We have AB = R(RQ) ,= .
Then Q, R Q in R are not separated.
3.41 Denition. A subset E X is said to be disconnected if there are two
non-empty separated sets A, B such that E = A B.
A subset E X is said to be connected if it is not disconnected, i.e. there
are no non-empty separated sets A, B such that E = A B.
3.42 Theorem. A non-empty subset E R is connected E is an interval.
Proof. An interval is dened as follows: Whenever x < z and x, z E for
all y with x < y < z we have y E.
(): Let E R be connected. Assume E is not an interval. So there are
two points x, z E with x < z, there is y with x < y < z and y / E.
Let A = E (, y), B = E (y, +). x A, z B. So A ,= ,
B ,= . Show A B = , A B = . A (, y) A (, y].
So A B (, y) (y, +) = . Similarly we have A B = .
A B = (E (, y)) (E (y, +))
= E ((, y) (y, +))
. .
R\{y}
= E since y / E
(): Omitted.
33
4 Sequences And Series
4.1 Sequences
Let (X, d) be a metric space. A sequence in X is a function f : N X. If
p
n
= f(n), we denote this sequence by (p
n
) or p
n
.
4.1 Example. X = R
2
, p
n
=
_
1n
n
,
(1)
n
n
_
where n = 1, 2, 3, . . . Then
p
1
= (0, 1)
p
2
=
_
1
2
,
1
2
_
p
3
=
_
2
3
,
1
3
_
.
.
.
4.2 Denition. We say the sequence p
n
converges to p X if for every
> 0 there is a natural number n
0
(depending on > 0 in general) such that
for all n N with n n
0
we have d(p
n
, p) < , i.e. p
n
B
(p). We write
p
n
p or lim
n
p
n
= p.
p
n
p every neighborhood of p contains all but nitely many terms p
n
.
If p
n
does not converge to any p X, we say p
n
is divergent.
4.3 Example. X = R
2
, p
n
=
_
1n
n
,
(1)
n
n
_
. p = (1, 0). Show p
n
p. Let
> 0 be given. Let n
0
be any natural number such that
< n
0
. Let n be
any natural number such that n
0
n.
d
2
(p
n
, p) =
_
1 n
n
(1)
_
2
+
_
(1)
n
n
0
_
2
=
_
1
n
_
2
+
_
(1)
n
n
_
2
=
_
2
n
2
=
2
n
2
n
0
<
34
4.4 Remark. x
n
=
_
1
n
_
converges to x = 0 in (R, [ [), but it is divergent
in ((0, 2), [ [).
4.5 Denition. We say that the sequence p
n
is bounded if the set E =
p
1
, p
2
, p
3
, . . . is a bounded subset of X, i.e. there is a constant M > 0 such
that for all p
i
, p
j
E we have d(p
i
, p
j
) M.
4.6 Theorem.
(a) Let p
n
be a sequence such that p
n
p and p
n
p
. Then p = p
.
(b) If p
n
is convergent then p
n
is bounded.
(c) Let E ,= be subset of X. Then p E there is a sequence p
n
contained in E such that p

n
p.
Proof.
(a) Let p
n
p and p
n
p
. Assume p ,= p
. Then d(p, p
) > 0. Let
0
=
d(p,p
)
3
then
0
> 0. We have p
n
p so there is n
1
N such that for all
n n
1
we have d(p
n
, p) <
0
. We also have p
n
p
so there is n
2
N
such that for all n n
2
we have d(p
n
, p
) <
0
. Let n
0
= maxn
1
, n
2
.
Then n
0
n
1
d(p
n
0
, p) <
0
and n
0
n
2
d(p
n
0
, p
) <
0
. Then
3
0
= d(p, p
) d(p, p
n
0
) + d(p
n
0
, p
) <
0
+
0
= 2
0
. So 3
0
< 2
0
.
Since
0
> 0, this cannot be true.
(b) Let p
n
p. For = 1 > 0, there is n
0
N such that for all n n
0
we
have d(p
n
, p) < 1. If i, j n
0
, then d(p
i
, p
j
) d(p
i
, p) + d(p, p
j
) < 1 +
1 = 2. Let K = max1, d(p
1
, p), . . . , d(p
n
0
1
, p). Then for all n N we
have d(p
n
, p) K. For any i, j N, d(p
i
, p
j
) d(p
i
, p)+d(p, p
j
) 2K.
(c) (): p
n
in E such that p
n
p. Show p E. Given r > 0 we have
n
0
0
, d(p
n
, p) < r. So p
n
0
B
r
(p) E.
So B
r
(p) E ,= . So p E.
(): Let p E. Then for every r > 0, B
r
(p) E ,= .
For r = 1, nd p
1
B
1
(p) E
For r =
1
2
, nd p
2
B1
2
(p) E
.
.
.
For r =
1
n
, nd p
n
B1
n
(p) E
35
Then p
n
is a sequence in E. Given > 0 let n
0
be such that
1
< n
0
. Then for all n n
0
we have d(p
n
, p) <
1
n

1
n
0
< . So
p
n
p.
4.7 Theorem. Let s
n
and t
n
be sequences in R such that s
n
s and
t
n
t where s, t R. Then
(a) s
n
+t
n
s +t
(b) For all constants c R, cs
n
c s
(c) s
n
t
n
s t
(d) If s ,= 0 then
1
s
n
1
s
Proof of (a). We have that
d(s
n
+t
n
, s +t) = [s
n
+t
n
(s +t)[
= [(s
n
s) + (t
n
t)[
[s
n
s[ +[t
n
t[
Given > 0, let
=

2
> 0. We have s
n
s so there is n
1
N such that
for all n n
1
we have [s
n
s[ <
. We also have t
n
t so there is n
2
N
2
we have [t
n
t[ <
. Let n
0
= maxn
1
, n
2
. Then
n n
0
n n
1
[s
n
s[ <
and n n
0
n n
2
[t
n
t[ <
. So
for all n n
0
we have
d(s
n
+t
n
, s +t) [s
n
s[ +[t
n
t[
<
= 2
=
Proof of (d). Let
0
=
|s|
2
then
0
> 0. So there is n
1
N such that for all
n n
1
we have [s
n
s[ <
0
. Let n n
1
, [s[ = [ss
n
+s
n
[ [ss
n
[ +[s
n
[ <
|s|
2
+ [s
n
[. So for all n n
1
,
|s|
2
< [s
n
[. In particular for all n n
1
, s
n
,= 0
so
1
s
n
is dened. And also
1
|s
n
|
<
2
|s|
. To show lim
n
1
s
n
=
1
s
, let > 0 be
given. Let
=
|s|
2
2
> 0. We have s
n
s so there is n
2
N such that for
all n n
2
we have [s
n
s[ <
. Let n
0
= maxn
1
, n
2
and n n
0
. Then
1
s
n
1
s
s s
n
s s
n
=
[s s
n
[
[s[[s
n
[
<

[s[

1
[s
n
[
<

[s[

2
[s[
=
[s[
2
2

2
[s[
2
=
36
4.8 Theorem.
(a) Let p
n
be a sequence in R
k
where p
n
= (x
n
1
, . . . , x
n
k
) and p =
(x
1
, . . . , x
k
) R
k
. Then p
n
p x
n
1
x
1
, x
n
2
x
2
, . . . , x
n
k
x
k
.
(b) Let p
n
, q
n
be two sequences in R
k
and
n
be a sequence in R.
Assume p
n
p, q
n
q in R
k
and
n
in R. Then p
n
+q
n
p +q
and
n
p
n
p.
Proof of (a). We need the following: If q = (y
1
, . . . , y
k
) R
k
then for all
i = 1, 2, . . . , k
[y
i
[
_
y
2
1
+y
2
2
+ +y
2
k
[y
1
[ +[y
2
[ + +[y
k
[
[y
i
[
2
= y
2
i
y
2
1
+y
2
2
+ +y
2
k
y
2
1
+y
2
2
+ +y
2
k
([y
1
[ +[y
2
[ + +[y
k
[)
2
Assume p
n
p. Given > 0, we have n
0
0
,
d
2
(p
n
, p) < . Let n n
0
. Then
[x
n
1
x
1
[
_
(x
n
1
x
1
)
2
+ (x
n
2
x
2
)
2
+ + (x
n
k
x
k
)
2
= d
2
(p
n
, p) <
[x
n
2
x
2
[ <
.
.
.
Conversely, assume x
n
1
x
1
, x
n
2
x
2
, . . . , x
n
k
x
k
. To show p
n
p, let
> 0 be given. Let
=

k
> 0.
x
n
1
x
1
so we have n
1
1
, [x
n
1
x
1
[ <
x
n
2
x
2
so we have n
2
2
, [x
n
2
x
2
[ <
.
.
.
x
n
k
x
k
so we have n
k
k
, [x
n
k
x
k
[ <
Let n
0
= maxn
1
, n
2
, . . . , n
k
. For all n n
0
d
2
(p
n
, p) =
_
(x
n
1
x
1
)
2
+ (x
n
2
x
2
)
2
+ + (x
n
k
x
k
)
2
[x
n
1
x
1
[ +[x
n
2
x
2
[ + +[x
n
k
x
k
[
<
+ +
= k
=
37
4.2 Subsequences
4.9 Denition. Let p
n
be a sequence in X. Let n
k
be a sequence of
natural numbers such that n
1
< n
2
< n
3
< Then the sequence p
n
k
is
called a subsequence of p
n
.
4.10 Example. p
1
, p
3
, p
7
, p
10
, p
23
, . . . is a subsequence of p
n
. n
1
= 2,
n
2
= 3, n
3
= 7, n
4
= 10, n
5
= 23 and so on.
4.11 Proposition. p
n
p every subsequence of p
n
converges to p.
Proof.
(): Since p
n
is a subsequence of itself, p
n
p.
(): Let p
n
p. Let p
n
k
be an arbitrary subsequence of p
n
. To show
p
n
k
p, let > 0 be given. Since p
n
p, we have n
0
such that for all
n n
0
, d(p
n
, p) < . If k n
0
then n
k
k n
0
so d(p
n
k
, p) < .
4.12 Remark. Limits of subsequences are limit points of the sequence.
4.13 Example. X = R
2
and p
n
=
_
n+(1)
n
n+1
n
,
1
n
_
n is even p
n
=
_
2n + 1
n
,
1
n
_
(2, 0)
n is odd p
n
=
_
1
n
,
1
n
_
(0, 0)
So the sequence p
n
has limit points (2, 0) and (0, 0).
4.14 Example. In X = R, x
n
= n + (1)
n
n +
1
n
n is even x
n
= 2n +
1
n
+
n is odd x
n
=
1
n
0
We do not accept + as a limit since + is not a member of R. So 0 is
the only limit point of the sequence x
n
but x
n
is divergent since it is not
bounded.
38
4.15 Theorem.
(a) Let (X, d) be a compact metric space and p
n
be any sequence in X.
Then p
n
has a subsequence that converges to a point p X.
(b) Every bounded sequence in R
k
has a convergent subsequence.
Proof.
(a) Case 1: p
n
has only nitely many distinct terms. Then at least one
term, say p
n
0
is repeated innitely many times, i.e. the sequence p
n
has a subsequence all of whose terms are p. Then the limit of this
subsequence is p = p
n
0
X.
Case 2: p
n
has innitely many distinct terms. Then the set E =
p
1
, p
2
, p
3
, . . . is an innite subset of the compact set X. So it has a
limit point p X. Then p is the limit of a subsequence of p
n
.
(b) Since p
n
is bounded, there is a k-cell I such that p
n
I. I is
compact, so by (a), p
n
has a subsequence that converges to a point
p I.
4.3 Cauchy Sequences
4.16 Denition. Let (X, d) be a metric space. A sequence p
n
in X is
called a Cauchy sequence if for every > 0 we have n
0
N such that for all
n, m n
0
, d(p
n
, p
m
) < .
4.17 Example. In X = R
x
n
=
_
n
1
cos t
t
2
dt
Then x
n
is a Cauchy sequence in R. Given n, m if n = m then x
n
= x
m
39
so d(x
n
, x
m
) = [x
n
x
m
[ = 0 < . If n ,= m, assume m < n.
d(x
n
x
m
) = [x
n
x
m
[
=
_
n
1
cos t
t
2
dt
_
m
1
cos t
t
2
dt
_
n
m
cos t
t
2
dt
_
n
m
cos t
t
2
dt We know
cos t
t
2
=
[ cos t[
t
2

1
t
2
_
n
m
1
t
2
dt =
1
t
n
m
=
1
n
+
1
m
1
m

1
n
0
<
Given > 0, choose n
0
N such that
1
< n
0
. Then for all n, m N with
n
0
m n we have d(x
n
, x
m
) < .
4.18 Example. X = R, x
n
=

n. If n = m + 1 then
d(x
n
, x
m
) = [x
m+1
x
m
[
= [
m + 1
m[ =
m + 1
m
= (
m + 1
m)
m + 1 +
m + 1 +
m
=
1
m + 1 +
m
<
1
m
Given > 0, choose n
0
N such that
1
2
< n
0
. Then for all m n
0
we have
d(x
m+1
, x
m
) < .
So the distance between successive terms gets smaller as the index gets larger
but this sequence x
n
is not a Cauchy sequence. For example, for = 1,
consider
d(x
m
, x
3m+1
) = [
3m + 1[
=
3m + 1
m =
m + 2m + 1
_
(
m + 1)
2
m + 1
m 1
4.19 Theorem. Let (X, d) be a metric space.
40
(a) Every convergent sequence in X is Cauchy.
(b) Every Cauchy sequence is bounded.
Proof.
(a) Assume p
n
p. Show p
n
is Cauchy. Given > 0 let
=

2
> 0.
Since p
n
p, there is n
0
N such that d(p
n
, p) <
for all n n
0
.
Let n, m n
0
. Then
d(p
n
, p
m
) d(p
n
, p) +d(p, p
m
) <
= 2
=
(b) Let p
n
be a Cauchy sequence in X. For = 1 there is n
0
N
such that for all n, m n
0
we have d(p
n
, p
m
) < 1. Let K =
max1, d(p
1
, p
n
0
), . . . , d(p
n
0
1
, p
n
0
) and M = 2k. Then we show that
for all n, m N, d(p
n
, p
m
) M.
Case 1: n, m n
0
. Then
d(p
n
, p
m
) < 1 K < M
Case 2: n, m < n
0
. Then
d(p
n
, p
m
) d(p
n
, p
n
0
) +d(p
n
0
, p
m
) K +K = M
Case 3: m < n
0
and n n
0
. Then
d(p
n
, p
m
) d(p
n
, p
n
0
)
. .
<1
+d(p
n
0
, p
m
)
. .
K
< 1 +K K +K = M
Converse of (a) is not true in general.
4.20 Example. X = (0, 1) = x R : 0 < x < 1 with d(x, x
) = [x x
[.
x
n
=
1
2n
. x
n
is a Cauchy sequence in X but x
n
has no limit in X.
4.21 Denition. A metric space (X, d) is said to be complete if every Cauchy
sequence in (X, d) is convergent to some point p X.
4.22 Theorem.
(a) Every compact metric space (X, d) is complete.
(b) (R
k
, d
2
) is complete. ((R
k
, d
1
), (R
k
, d
) are also complete.)

41
Proof.
(a) Let (X, d) be a compact metric space and let p
n
be a Cauchy sequence
in X. Then p
n
has a subsequence p
n
k
which converges to a point
p X. Show p
n
p. Let > 0 be given. Let
=

2
> 0. p
n
is Cauchy, so there is N
1
N such that for all n, m N
1
we have
d(p
n
, p
m
) <
. p
n
k
p, so there is N
2
such that for all k N
2
we
have d(p
n
k
, p) <
. Let N = maxN
1
, N
2
. Then for all n N
d(p
n
, p) d(p
n
, p
n
N
)
. .
<
+d(p
n
N
, p)
. .
<
< 2
=
(b) Let p
n
be a Cauchy sequence in R
k
. Then p
n
is bounded so there
is a k-cell I such that p
n
I. (I, d
2
) is compact. Then by (a), p
n
has a limit p I R
k
.
4.23 Remark. S ,= , B(S) all bounded functions f : S R.
d(f, g) = sup[f(s) g(s)[ : s S
(B(S), d) is complete.
4.24 Theorem. Let (X, d) be a complete metric space and Y ,= be a
subset of X. The subspace (Y, d) is complete Y is a closed subset of X.
Proof.
(): Assume (Y, d) is complete and show Y is closed, i.e. Y Y . Let p Y .
Then there is a sequence p
n
in Y such that p
n
p. Then p
n
is
convergent in X. So p
n
is Cauchy in X. Since all p
n
Y , p
n
is
Cauchy in Y . Since Y is complete, there is an element q Y such that
p
n
q. Then p = q Y . So p Y , i.e. Y Y .
(): Assume Y is closed. Show (Y, d) is complete. Let p
n
be a Cauchy
sequence in Y . Then p
n
is a Cauchy sequence in X. Since (X, d) is
complete, there is p X such that p
n
p. Then p is the limit of the
sequence p
n
in Y . Then p Y . Since Y is closed, Y = Y . So p Y .
So (Y, d) is complete.
4.25 Example. X = R
2
, Y = (x, y) : x 0, y 0. Then (Y, d
2
) is
complete.
42
Monotone Sequence Property In R
Let s
n
be a sequence in R. We say
s
n
is increasing if s
1
s
2
s
3
s
n
s
n+1

s
n
is decreasing if s
1
s
2
s
3
s
n
s
n+1

s
n
is monotone if s
n
is either increasing or decreasing.
4.26 Theorem (Monotone Sequence Property). Let s
n
be a monotone
sequence in R. Then s
n
is convergent s
n
is bounded.
Proof.
(): True for all sequences.
(): We do the proof for decreasing sequences. Let s = infs
1
, s
2
, s
3
, . . ..
Show lim
n
s
n
= s. Let > 0 be given. Then s + cannot be a
lower bound for the set s
1
, s
2
, s
3
, . . .. Then there is s
n
0
such that
s
n
0
< s +. Let n n
0
. s < s s
n
s
n
0
< s +. For all n n
0
s < s
n
< s +
< s
n
s <
[s
n
s[ <
d(s
n
, s) <
4.27 Example. Let A > 0 be xed. Start with any x
1
> 0 and dene
x
n
=
1
2
_
x
n1
+
A
x
n1
_
n = 2, 3, 4, . . .
Then lim
n
x
n
=
A. We will show x
2
x
3
x
4
For n 2
x
2
n
A =
1
4
_
x
2
n1
+
A
2
x
2
n1
+ 2A
_
A
=
1
4
_
x
2
n1
+
A
2
x
2
n1
2A
_
=
1
4
_
x
n1
A
x
n1
_
2
0
43
So x
2
n
A for all n 2. Since all x
n
> 0, x
n

A for all n 2. For n 2

x
n
x
n+1
= x
n
1
2
_
x
n
+
A
x
n
_
=
1
2
_
x
n
A
x
n
_
=
1
2
x
2
n
A
x
n
0
So x
n
x
n+1
for all n 2. So x
n
n=2
is decreasing and bounded. So
lim
n
x
n
= x exists. Then we solve for x. We have that
x
n+1
=
1
2
_
x
n
+
A
x
n
_

x =
1
2
_
x +
A
x
_
Then
2x = x +
A
x
x
2
= A
x =
A
Since all x
n
> 0, limit x cannot be negative. So x =
A.
4.4 Upper And Lower Limits
Let x
n
be a sequence in R.
We write lim
n
x
n
= + (or x
n
+) if for every M > 0 we can nd a
natural number n
0
(depending on M in general) such that for all n n
0
we
have M x
n
.
We write lim
n
x
n
= (or x
n
) if for every M < 0 we can nd a
natural number n
0
(depending on M in general) such that for all n n
0
we
have x
n
M.
In either case, we say x
n
is divergent.
4.28 Example. x
n
= n +
1
n
and lim
n
x
n
= +. x
n
is divergent.
4.29 Denition. Let x
n
be a sequence in R. Let E be the set of all
subsequential limits of x
n
. Then E R , +.
44
4.30 Example. x
n
= n + (1)
n
n +
1
n
. Subsequential limits are + and 0.
So E = 0, +.
4.31 Denition. Let x
= sup E and x
= inf E (Considered in the set of

extended real numbers.)
x
is called the upper limit (or limit superior) of x

n
and it is denoted by
x
= limsup
n
x
n
= lim
n
x
n
x
is called the lower limit (or limit inferior) of x

n
and it is denoted by
x
= liminf
n
x
n
= lim
n
x
n
4.32 Example. The function : NN N dened by (r, s) = 2
r1
(2s
1), is 1-1 and onto. Let s be xed. N
s
= 2
r1
(2s 1) : r = 1, 2, 3, . . . =
rs 1 2 3 4
1 1 3 5 7
2 2 6 10 14
3 4 12 20 28
4 8 24 40 56
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2s 1, 2(2s 1), 4(2s 1), . . . Then for s ,= s
, N
s
N
s
= . Also
s=1
N
s
= N. Dene a sequence x
n
in R as follows: Given n N, there is
a unique s such that n N
s
. Dene x
n
=
s
n
n+1
. What are the subsequential
limits of x
n
? What are limsup
n
x
n
and liminf
n
x
n
?
If n N
1
= 1, 2, 4, 8, . . . then x
n
=
n
n + 1
1
If n N
2
= 3, 6, 12, 24, . . . then x
n
=
2n
n + 1
2
.
.
.
If n N
s
= then x
n
=
s
n
n + 1
s
So all 1, 2, 3, . . . are subsequential limits. Then we have 1, 2, 3, . . . E.
Then sup E = +, i.e. limsup
n
x
n
= +.
45
Since x
n
=
s
n
n+1
, s 1 x
n

n
n+1
so all x
n

1
2
. So if x is a subsequential
limit of x
n
then x
1
2
. Can we have a subsequential limit x such that
1
2
x 1 ? If n N
s
where s 2 then x
n
=
s
n
n+1

2n
n+1
1. If n N
1
then x
n
=
n
n+1
1. So 1 is the smallest subsequential limit of x
n
. Thus
liminf
n
x
n
= 1.
Properties
(i) liminf
n
x
n
limsup
n
x
n
(ii) liminf
n
x
n
= limsup
n
x
n
. .
call this x
lim
n
x
n
= x (Here x R or x = + or
x = )
(iii) Let x R, i.e. x ,= . We have that limsup
n
x
n
= x
(a) For every > 0 there is a natural number n
0
such that for all
n n
0
, x
n
< x +
and
(b) For every > 0 there are innitely many n such that x < x
n
4.33 Theorem (Squeeze Property or Sandwich Property). Let x
n
, y
n
,
z
n
be three sequences in R such that x
n
y
n
z
n
for all n. Assume that
x
n
and z
n
are convergent and lim
n
x
n
= lim
n
z
n
. .
call this c
. Then y
n
is convergent
and lim
n
y
n
= c.
Proof. Given > 0
x
n
c, so there is n
1
1
, [x
n
c[ <
z
n
c, so there is n
2
2
, [z
n
c[ <
Let n
0
= maxn
1
, n
2
and n n
0
. Show [y
n
c[ <
If c y
n
then [y
n
c[ = y
n
c z
n
c [z
n
c[ <
If c > y
n
then [y
n
c[ = c y
n
c x
n
[c x
n
[ = [x
n
c[ <
Some Special Sequences In R
4.34 Theorem.
(a) If p > 0 constant then lim
n
1
n
p
= 0
(b) If p > 0 constant then lim
n
n
p = 1
46
(c) lim
n
n
n = 1
(d) If p > 0 and R are constants then lim
n
n
(1+p)
n
= 0
Note: This is usually expressed as polynomials tend to increase slower
than exponentials.
(e) If x is constant and [x[ < 1, i.e. 1 < x < 1 then lim
n
x
n
= 0
Proof of (b).
If p = 1 then
n
p = 1 for all n, so lim

n
n
p = 1.
If p > 1 let x
n
=
n
p 1. Then x
n
> 0 for all n.
p = (1 +x
n
)
n
= 1 +nx
n
+
n(n 1)
2
x
2
n
+ +x
n
n
. .
positive
So p 1 + nx
n
. So 0 < x
n
<
p1
n
. By sandwich property, lim
n
x
n
= 0.
Then
n
p = 1 + x
n
1.
If p < 1 then
n
_
1
p
1 by the previous case so
n
p 1.
Proof of (c). Let x
n
=
n
n 1. Then x
n
0 for all n.
n = (1 +x
n
)
n
= 1 + nx
n
+
n(n 1)
2
x
2
n
+
n(n 1)(n 2)
6
x
3
n
+ +x
n
n
. .
0
n(n 1)
2
x
2
n
So n
n(n1)
2
x
2
n
0 x
2
n

2
n1
0 x
n

_
2
n1
. By sandwich theorem,
lim
n
x
n
= 0. Then
n
n = 1 +x
n
1.
Proof of (d). If 0 we have lim
n
n
(1+p)
n
= 0. So assume > 0. Fix a
natural number k such that < k. Then for n 2k
n(n 1) (n k + 1)
. .
k terms
>
n
2

n
2

n
2
. .
k terms
=
_
n
2
_
k
47
(1 +p)
n
=
n
=0
_
n
_
p
1
n
>
_
n
k
_
p
k
=
n(n 1) (n k + 1)
k!
p
k
>
_
n
2
_
k
k!
p
k
=
n
k
2
k
k!
p
k
=
n
k
n
2
k
k!
p
k
2
k
k!
p
k
1
n
k
>
n
(1 +p)
n
> 0
By sandwich property, we have lim
n
n
(1+p)
n
= 0.
4.5 Series
4.35 Denition. Given a sequence a
n
in R, the symbol
n=1
a
n
is called
an (innite) series. Given a series
n=1
a
n
we dene the following sequence
s
n
s
1
= a
1
s
2
= a
1
+a
2
s
3
= a
1
+a
2
+a
3
.
.
.
s
n
= a
1
+a
2
+ +a
n
s
n
is called the sequence of partial sums.
If lim
n
s
n
= s exists in R (s = is not acceptable) we say the series is
convergent and has sum = s. We write
n=1
a
n
= s.
If lim
n
s
n
= or lim
n
s
n
does not exist we say that the series
n=1
a
n
is divergent.
4.36 Example.
n=1
1
(n+1)
2
1
=
1
2
2
1
+
1
3
2
1
+
1
4
2
1
+ Then
a
n
=
1
(n + 1)
2
1
=
1
n(n + 2)
=
A
n
+
B
n + 2
Then A =
1
2
and B =
1
2
. We get
1
2
n

1
2
n + 2
=
1
2
_
1
n

1
n + 2
_
=
1
2
n + 2 n
n(n + 2)
48
So a
n
=
1
2
_
1
n

1
n+2
_
. Then
s
n
= a
1
+a
2
+ +a
n
=
1
2
_
1
1

1
3
_
+
1
2
_
1
2

1
4
_
+
1
2
_
1
3

1
5
_
+ +
1
2
_
1
n

1
n + 2
_
Cancellation pattern: ( [X) + ( [ ) + (X[ )+ Then
s
n
=
1
2
_
1
1
+
1
2

1
n + 1

1
n + 2
_
=
1
2
_
3
2

1
n + 1

1
n + 2
_
Then lim
n
s
n
=
1
2

3
2
=
3
4
so s =
3
4
i.e.
1
2
2
1
+
1
3
2
1
+ =
3
4
4.37 Example. Let r R be a constant. Consider
n=0
r
n
= 1 +r +r
2
+
r
3
+ geometrical series.
s
n
= 1 + r +r
2
+ +r
n
rs
n
= r +r
2
+ +r
n
. .
s
n
1
+r
n+1
rs
n
= s
n
1 +r
n+1
1 r
n+1
= s
n
rs
n
s
n
=
1 r
n+1
1 r
if r ,= 1
If [r[ < 1, i.e. 1 < r < 1 then r
n+1
0 so lim
n
s
n
=
1
1r
If r = 1 then s
n
= n + 1 +
For any other value of r, lim
n
s
n
does not exist. So the geometrical series
is convergent only for 1 < r < 1.
n=0
r
n
= 1 +r +r
2
+r
3
+ =
1
1 r
if 1 < r < 1
4.38 Theorem (Cauchy Criterion).
n=1
a
n
is convergent for every > 0
there is n
0
N such that for all n, m n
0
with n m we have [
n
k=m
a
k
[ =
[a
m
+a
m+1
+ +a
n
[ < .
Proof.
n=1
a
n
is convergent then s
n
is convergent. So s
n
is Cauchy.
That is, for every > 0 there is n
0
0
with
n m we have [s
n
s
m1
[ < .
s
n
s
m1
= a
1
+a
2
+ +a
n
(a
1
+ +a
m1
) = a
m
+ +a
n
49
4.39 Theorem. If
a
n
is convergent then lim
n
a
n
= 0.
Proof. a
n
= s
n
s
n1
s s = 0.
4.40 Example.
n=1
(1)
n
= (1) +1 +(1) +1 +(1) + is divergent
since lim
n
(1)
n
does not exist. So lim
n
(1)
n
,= 0.
s
n
=
_
0 if n is even
1 if n is odd
4.41 Denition. A series
n=1
a
n
is said to be non-negative if there is
n
0
0
we have a
n
0.
4.42 Theorem. Let
a
n
be a non-negative series. Then
a
n
is convergent
s
n
is bounded.
Proof. There is n
0
0
we have a
n
0.
s
n
0
= s
n
0
1
+a
n
0
s
n
0
1
s
n
0
+1
= s
n
0
+a
n
0
+1
s
n
0
.
.
.
s
n+1
= s
n
+a
n+1
s
n
.
.
.
s
n
0
s
n
0
+1
s
n
s
n+1

By monotone sequence property (since s
n
n=n
0
is increasing), s
n
is con-
vergent s
n
is bounded.
4.43 Theorem (Comparison Test).
(a) Suppose there is n
0
0
[a
n
[ c
n
and
c
n
is
convergent. Then
a
n
is also convergent.
(b) Suppose there is n
0
0
a
n
d
n
0 and
d
n
is
divergent. Then
a
n
is also divergent.
Proof.
50
(a) Use Cauchy criterion. Let > 0 be given. Since
c
n
is convergent,
there is n
1
1
with m n we have
[
n
k=m
c
k
[ < . Let n
2
= maxn
0
, n
1
. Let n m n
2
. Then
k=m
a
k
= [a
m
+a
m+1
+ +a
n
[ [a
m
[
..
c
m
+[a
m+1
[
. .
c
m+1
+ + [a
n
[
..
c
n
c
m
+c
m+1
+ +c
n
= [c
m
+c
m+1
+ +c
n
[ <
(b) This follows from (a). If
a
n
were convergent, then by (a),
d
n
would be convergent.
4.44 Theorem (Cauchy Condensation Test). Suppose a
1
a
2
a
3

0. Then
n=1
a
n
is convergent

k=0
2
k
a
2
k = a
1
+ 2a
2
+ 4a
4
+ 8a
8
+
is convergent.
Proof. Let s
n
= a
1
+a
2
+ +a
n
and t
n
= a
1
+ 2a
2
+ + 2
n
a
2
n.
t
n
= a
1
+ 2a
2
+ 4a
4
+ + 2
n
a
2
n
= a
1
+ (a
2
+a
2
) + (a
4
+a
4
+a
4
+a
4
) + + (a
2
n +a
2
n + +a
2
n)
. .
2
n
terms
a
1
+a
2
+a
3
+ +a
2
n+1
1
= s
2
n+1
1
s
2
n = a
1
+a
2
+ (a
3
+a
4
) + (a
5
+a
6
+a
7
+a
8
) + + (a
2
n1
+1
+ +a
2
n)
. .
2
n1
terms
1
2
a
1
+a
2
+ 2a
4
+ 4a
8
+ + 2
n1
a
2
n
=
1
2
(a
1
+ 2a
2
+ 4a
4
+ 8a
8
+ + 2
n
a
2
n) =
1
2
t
n
Then s
2
n+1
1
t
n
2s
2
n. So
a
n
is convergent s
n
is bounded t
n
is bounded
2
k
a
2
k is convergent.
4.45 Example. Let p > 0 be constant and consider the p-series
n=1
1
n
p
.
Then a
n
=
1
n
p
> 0 and a
1
a
2
a
3

k=0
2
k
a
2
k =
k=0
2
k
1
(2
k
)
p
=
k=0
2
kkp
=
k=0
(2
1p
)
k
Geometric series with r = 2
1p
> 0. It is convergent r < 1, i.e.
2
1p
< 1 p > 1. If p 0 let q = p 0. a
n
=
1
n
p
=
1
n
q
= n
q
51
then lim
n
a
n
,= 0. So
1
n
p
is divergent.
Summary: Let p be a constant. Then
n=1
1
n
p
is convergent p > 1.
For example,
n=1
1
n
2
and
n=1
1
n
n
are convergent but
n=1
1
n
and
n=1
1
n
are divergent.
n=1
1
n
1.00001
is convergent.
n=1
1
n
1+
1
n
is not a
p-series since the exponent 1 +
1
n
is not a constant.
4.46 Remark.
n=1
1
n
2
=

2
6
The Number e
The series
n=0
1
n!
= 1 + 1 +
1
2!
+
1
3!
+
1
4!
+
is convergent. For n 2
a
n
=
1
n!
=
1
2 3 n
. .
n1 factors
1
2
n1
1
2
n1
= 2
_
1
2
_
n
is convergent. So by comparison test,
n=0
a
n
=
n=0
1
n!
is convergent.
4.47 Denition.
e =
n=0
1
n!
4.48 Theorem. e is not rational.
Proof. Suppose e is rational. Then e =
p
q
where p, q are natural numbers.
52
Since all a
n
> 0, we have s
n
< e for all n. We have 0 < e s
q
<
1
q!q
e s
q
=
1
(q + 1)!
+
1
(q + 2)!
+
1
(q + 3)!
+
=
1
(q + 1)!
_
1 +
1
q + 2
+
1
(q + 2)(q + 3)
+
_
<
1
(q + 1)!
_
1 +
1
q + 1
+
1
(q + 1)(q + 1)
+
_
=
1
(q + 1)!
n=0
_
1
q + 1
_
n
Geometrical series with r =
1
q + 1
<
1
(q + 1)!
1
1
1
q+1
=
1
(q + 1)!
q + 1
q
=
1
q!q
Then 0 < q!(e s
q
) <
1
q
. p = e q is an integer. So q! e = 1 2 (q 1) q e
is an integer. q!s
q
= q!
_
1 + 1 +
1
2!
+ +
1
q!
_
is also an integer. So q!e qs
q
is an integer. Also,
1
q
1 so 0 < q!e q!s
q
< 1. Contradiction.
4.49 Remark. e is not even an algebraic number.
A real number r is called an algebraic number if there is a polynomial P(x) =
a
n
x
n
+a
n1
x
n1
+ +a
1
x +a
0
with integer coecients a
n
, a
n1
, . . . , a
1
, a
0
such that P(r) = 0.
4.50 Example. r =
2 is algebraic. P(x) = x
2
2 then P(r) = 0.
A real number that is not algebraic is called transcendental. e, are tran-
scendental numbers.
4.51 Theorem (Root Test). Given
a
n
let = limsup
n
n
_
[a
n
[. Then
0 +.
(a) < 1
a
n
is convergent
(b) > 1
a
n
is divergent
(c) = 1 No information
Proof.
(a) Find such that < < 1. Then there is n
0
N such that for all
n n
0
we have
n
_
[a
n
[ < . That is, [a
n
[ <
n
for all n n
0
.
n
is convergent (geometrical series, 0 < < 1) so by comparison test,
a
n
is convergent.
53
(b) We have > 1. Since limsup is the largest subsequential limit, we can
nd a subsequence n
k
of natural numbers such that
n
k
_
[a
n
k
[ .
Since > 1, we have
n
k
_
[a
n
k
[ > 1 [a
n
k
[ > 1. Then lim
n
a
n
= 0
cannot be true. So
a
n
is divergent.
(c)
1
n
is divergent and = 1. (in fact lim
n
n
_
1
n
= 1)
1
n
2
is convergent and = 1. (in fact lim
n
n
_
1
n
2
= 1)
4.52 Example. Consider
n=1
(2)
n
n
Then a
n
=
(2)
n
n
. We apply root test.
n
_
[a
n
[ =
n
_
2
n
n
=
2
n
n
. We have = lim
n
2
n
n
= 2. > 1 so
(2)
n
n
is
divergent.
4.53 Theorem (Ratio Test). Let
a
n
be an arbitrary series.
(a) If limsup
n
|a
n+1
|
|a
n
|
< 1 then
a
n
is convergent.
(b) If liminf
n
|a
n+1
|
|a
n
|
> 1 then
a
n
is divergent.
(c) If liminf
n
|a
n+1
|
|a
n
|
1 limsup
n
|a
n+1
|
|a
n
|
then no information.
Proof. Omitted. Similar to the proof of the root test.
a
n
where
a
n
=
_
n
2
10
n
if n is odd
n
3
100
n
if n is even
a
n
> 0 for all n so [a
n
[ = a
n
. Then
[a
n+1
[
[a
n
[
=
a
n+1
a
n
=
_
_
(n+1)
3
100
n+1
n
2
10
n
if n is odd
(n+1)
2
10
n+1
n
3
100
n
if n is even
Then
[a
n+1
[
[a
n
[
=
_
(n+1)
3
n
2
1
10
n+2
if n is odd
(n+1)
2
n
3
10
n1
if n is even
54
So limsup
n
|a
n+1
|
|a
n
|
= + and liminf
n
|a
n+1
|
|a
n
|
= 0. So ratio test gives no
information. Then try root test.
n
_
[a
n
[ =
_
(
n
n)
2
10
if n is odd
(
n
n)
3
100
if n is even
1
10
and
1
100
are the only subsequential limits of
n
_
[a
n
[. We have
limsup
n
n
_
[a
n
[ =
1
10
< 1. So by root test, the series
a
n
is conver-
gent.
4.55 Remark. Root test has wider scope.
Ratio test shows convergence Root test shows convergence
Root test gives no information Ratio test gives no information
4.56 Theorem. Let c
n
be any sequence of positive numbers. Then
liminf
n
c
n+1
c
n
liminf
n
n
c
n
limsup
n
n
c
n
limsup
n
c
n+1
c
n
Raabes Test: Let
a
n
be a series of real numbers.
(a) liminf
n
n
_
|a
n
|
|a
n+1
|
1
_
= p If 1 < p then
a
n
is convergent.
(b) limsup
n
n
_
|a
n
|
|a
n+1
|
1
_
= q If q < 1 then
a
n
is divergent.
n=1
13(2n1)
24(2n)
[a
n+1
[
[a
n
[
=
a
n+1
a
n
=
13(2n1)(2n+1)
24(2n)(2n+2)
13(2n1)
24(2n)
=
2n + 1
2n + 2
1
We have limsup
n
|a
n+1
|
|a
n
|
= liminf
n
|a
n+1
|
|a
n
|
. So ratio test gives no infor-
mation. Then try Raabes test.
n
_
[a
n
[
[a
n+1
[
1
_
= n
_
2n + 2
2n + 1
1
_
= n
1
2n + 1

1
2
We have limsup
n
n
_
|a
n
|
|a
n+1
|
1
_
=
1
2
< 1. So the series is divergent.
55
4.58 Denition (Power Series). Let c
n
be a xed sequence of real numbers
and x R be a variable. The series
c
0
+c
1
x +c
2
x
2
+ +c
n
x
n
+ =
n=0
c
n
x
n
is called a power series.
4.59 Example. Let c
n
=
1
n!
. Consider 1 + x +
x
2
2!
+
x
3
3!
+ =
n=0
x
n
n!
General Question: Given a power series, nd the set of all x for which the
power series is convergent.
We apply root test.
limsup
n
n
_
[c
n
x
n
[ = limsup
n
[x[
n
_
[c
n
[ = [x[ limsup
n
n
_
[c
n
[
. .
call
= [x[
If [x[ < 1, i.e. [x[ <
1
then
c
n
x
n
is convergent.
If [x[ > 1, i.e. [x[ >
1
then
c
n
x
n
is divergent.
4.60 Theorem. With any power series
c
n
x
n
is associated a radius of
convergence R, 0 R + such that
(i) The series converges for all x with [x[ < R
(ii) The series diverges for all x with [x[ > R
R =
1
where = limsup
n
n
_
[c
n
[ or = lim
n
|c
n+1
|
|c
n
|
if this limit exists.
If = 0 then R = +. If = + then R = 0. (For R = 0, it means the
series
c
n
x
n
converges only for x = 0.)
n=0
n!x
n
= 1+x+2!x
2
+3!x
3
+ Then c
n
= n!
= lim
n
[c
n+1
[
[c
n
[
= lim
n
(n + 1) = +
So R = 0 and
n!x
n
converges only for x = 0.
n=0
x
n
n!
= 1 +x +
x
2
2!
+
x
3
3!
+ Then c
n
=
1
n!
= lim
n
[c
n+1
[
[c
n
[
= lim
n
1
n + 1
= 0
So R = + and
x
n
n!
converges for all x R with [x[ < +, i.e. for all
x R.
56
n=1
x
n
n
. Then c
n
=
1
n
= limsup
n
n
1
n
= limsup
n
1
n
n
= 1
So R = 1 and
The series
x
n
n
is convergent for all x with [x[ < 1, i.e. 1 < x < 1
The series
x
n
n
is divergent for all x with [x[ > 1, i.e. x < 1, 1 < x
No information for [x[ = 1, i.e. x = 1
If x = 1,
n=1
1
n
n
=
n=1
1
n
is divergent. (p-series with p = 1)
If x = 1,
n=1
(1)
n
n
= 1 +
1
2

1
3
+
1
4

Abels Partial Summation Formula: Let a
n
and b
n
be two se-
quences.
A
n
=
n
k=1
a
k
= a
1
+a
2
+ +a
n
Then
n+1
k=1
a
k
b
k
= A
n+1
b
n+1
k=1
A
k
(b
k+1
b
k
)
Proof. Let A
0
= 0. Then a
k
= A
k
A
k1
, k = 1, 2, . . .
n+1
k=1
a
k
b
k
=
n+1
k=1
(A
k
A
k1
) b
k
=
n+1
k=1
A
k
b
k

n+1
k=1
A
k1
b
k
. .
P
n
k=1
A
k
b
k+1
=
n
k=1
A
k
b
k
+A
n+1
b
n+1
k=1
A
k
b
k+1
= A
n+1
b
n+1
+
n
k=1
A
k
(b
k
b
k+1
)
4.64 Theorem (Dirichlets Test). Let a
n
and b
n
be two sequences of
real numbers. A
n
=
n
k=1
a
k
= a
1
+ +a
n
. Assume
(a) The sequence A
n
is bounded
57
(b) b
1
b
2
b
3

(c) lim
n
b
n
= 0
Then
n=1
a
n
b
n
is convergent.
Proof. Let s
n
= a
1
b
1
+ + a
n
b
n
. Show s
n
converges in R. By Abels
formula
s
n+1
= A
n+1
b
n+1
k=1
A
k
(b
k+1
b
k
)
A
n
is bounded, so there is a constant M > 0 such that [A
n
[ M for all n.
M A
n+1
M
If we multiply by b
n+1
0 we get
Mb
n+1
A
n+1
b
n+1
Mb
n+1
So lim
n
A
n+1
b
n+1
= 0. Next, show lim
n
(
n
k=1
A
k
(b
k+1
b
k
)) exists in
R. This is the n-th partial sum of the series
n=1
A
n
(b
n+1
b
n
). So show
the series
n=1
A
n
(b
n+1
b
n
) is convergent. Use Cauchy criterion. Let > 0
be given. Let
=

2M
> 0. Since lim
n
b
n
= 0, we have n
0
N such that
for all n n
0
, [b
n
[ = [b
n
0[ <
. Let n, m n
0
and n m.
k=m
A
k
(b
k+1
b
k
)
k=m
[A
k
[
..
M
[b
k+1
b
k
[ M
n
k = m[b
k+1
b
k
[
. .
b
k
b
k+1
= M ((b
m
b
m+1
) + (b
m+1
b
m+2
) + + (b
n
b
n+1
))
= M(b
m
b
n+1
) M[b
m
b
n+1
[ M([b
m
[
..
<
[b
n+1
[
. .
<
)
< M2
=
1 +
1
1
2
1
+
2
3
+
1 +
1
4
2
4
+
2
6
+
1 +
1
7
2
7
+
2
9
+
Then b
1
=
1
1
, b
2
=
1
2
, b
3
=
1
3
b
n
=
1
n
b
n
satises (b) and (c).
Also a
1
= 1 +
1
1
2
, a
2
= 2, a
3
= 3, a
4
= 1 +
1
4
2
, a
5
= 2, a
6
= 3, a
7
= 1 +
1
7
2
,
a
8
= 2, a
9
= 3, . . .
Then A
1
= 1 +
1
1
2
, A
2
= 3 +
1
1
2
, A
3
=
1
1
2
, A
4
=
1
1
2
+1 +
1
4
2
, A
5
=
1
1
2
+3 +
1
4
2
,
58
A
6
=
1
1
2
+ 6 +
1
4
2

We have A
n
3 +
1
1
2
+
1
4
2
+
1
7
2
+
. .
2
6
So A
n
is bounded. So the series is
convergent by Dirichlets test.
Alternating Series Test Of Leibniz: Assume b
1
b
2
b
3
and
lim
n
b
n
= 0. Then
n=1
(1)
n
b
n
= b
1
+b
2
b
3
+ and
n=1
(1)
n+1
b
n
= b
1
b
2
+b
3

are convergent.
4.66 Example. Let b
n
=
1
n
. Then
n=1
(1)
n+1
1
n
= 1
1
2
+
1
3

1
4
+
1
5
= ln 2
is convergent.
The series
n=1
c
n
=
n=1
(1)
n+1
n
i.e. c
n
=
(1)
n+1
n
is convergent but
n=1
[c
n
[ =
n=1
1
n
is divergent.
4.67 Denition. Let
c
n
be a series. If
c
n
is convergent but
[c
n
[ is di-
vergent, we say
c
n
is conditionally convergent. If
[c
n
[ is also convergent,
we say
c
n
is absolutely convergent.
4.68 Theorem. If
[c
n
[ is convergent then
c
n
is also convergent.
Proof. Use Cauchy criterion. Let > 0 be given. Since
[c
n
[ is convergent,
there is n
0
such that for all n m n
0
we have [
n
k=m
[c
k
[[ < . Let
n m n
0
. Then

k=m
c
k
k=m
[c
k
[ <
So
c
n
satises Cauchy criterion. Then
c
n
is convergent.
For absolute convergence we can use root test, ratio test or comparison test.
For conditional convergence we can use Dirichlets test or alternating series
test. (Only for alternating series.)
59
n=1
(1)
n+1
n
= 1
1
2
+
1
3

1
4
+ = S
The series is convergent by the alternating series test.
S = 1
1
2
+
1
3
__
1
4

1
5
_
+
_
1
6

1
7
_
+
_
=
5
6
(a positive number) <
5
6
Consider the rearrangement
1 +
1
3

1
2
. .
+
1
5
+
1
7

1
4
. .
+
1
9
+
1
11

1
6
. .
+
1
13
+
1
15

1
8
. .
+
Each group is in the form
1
4n 3
+
1
4n 1

1
2n
=
(4n 1)2n + (4n 3)2n (4n 3)(4n 1)
(4n 3)(4n 1)2n
=
8n
2
2n + 8n
2
6n 16n
2
+ 4n + 12n 3

=
8n 3

> 0
If t
n
is the n-th partial sum of the rearrangement. Then t
3
< t
6
< t
9
<
Then limsup
n
t
n
> t
3
= 1 +
1
3

1
2
=
5
6
. It follows that lim
n
t
n
cannot
be S since S <
5
6
. This is a property of the conditionally convergent series.
Given a conditionally convergent series
a
n
and r +, it is
possible to nd a rearrangement of the series such that rearrangement has
sum= r.
4.70 Denition. Let : N N be a 1-1, onto function. Let
n=1
a
n
be a series. Let b
n
= a
(n)
. The series
n=1
b
n
=
n=1
a
(n)
is called a
rearrangement of the series
n=1
a
n
.
If
n=1
a
n
= 1
1
2
+
1
3

1
4
+
1
5

n=1
b
n
= 1 +
1
3

1
2
+
1
5
+
1
7

1
4
+
Then (1) = 1, (2) = 3, (3) = 2, (4) = 5, (5) = 7, (6) = 4, . . .
60
4.71 Theorem. Assume
n=1
a
n
is absolutely convergent, i.e.
n=1
[a
n
[
is convergent. Then every rearrangement of
n=1
a
n
is convergent and it
converges to the same sum.
Proof. Let
n=1
a
(n)
be a rearrangement of
n=1
a
n
. Let
s
n
= a
1
+a
2
+ +a
n
t
n
= a
(1)
+a
(2)
+ +a
(n)
Given s
n
s where s R (s ,= ). Show s
n
t
n
0 as n . Let
> 0 be given. Since
a
n
converges absolutely, there is N such that for all
n m N we have
n
k=m
[a
k
[ < . Find p N such that
1, 2, . . . , N (1), (2), . . . , (p)
Take n > p. Then
[s
n
t
n
0[ =[s
n
t
n
[
=[a
1
+a
2
+ +a
N
+a
N+1
+ +a
n
a
(1)
a
(2)
a
(p)
. .
N of them will be cancelled
a
(n)
[
. .
The remaining terms will be of the form a
k
where k>N
Let q = maxk : a
k
remains in the above. Then
[s
n
t
n
0[
q
k=N+1
[a
k
[ <
4.6 Operations With Series
4.72 Theorem. Let
a
n
,
b
n
be convergent series with sums A and B.
Let c R be a constant. Then
(a
n
+b
n
) is convergent and has sum A +B
ca
n
is convergent and has sum cA
4.73 Remark. The sum of two divergent series may be convergent.
61
4.74 Example.
n=1
1
n
and
n=1
1
n+1
are both divergent.
n=1
_
1
n

1
n+1
_
=
n=1
1
n(n+1)
is convergent.
4.75 Example.
n=0
_
_
2
3
_
n
5
(1)
n
4
n
_
is convergent.
n=1
_
2
3
_
n
5
n=1
_
1
4
_
n
=
1
1
2
3
5
1
1
_
1
4
_ = 3 5
4
5
= 1
Cauchy Product Of Two Series: Consider the series
n=0
a
n
= a
0
+a
1
+a
2
+ and
n=1
b
n
= b
0
+b
1
+b
2
+
We dene a new series
n=0
c
n
as follows
c
0
= a
0
b
0
c
1
= a
0
b
1
+a
1
b
0
c
2
= a
0
b
2
+a
1
b
1
+a
2
b
0
.
.
.
c
n
= a
0
b
n
+a
1
b
n1
+ +a
n1
b
1
+a
n
b
0
=
n
k=0
a
k
b
nk
n=0
c
n
is called the Cauchy product of
n=0
a
n
and
n=0
b
n
.
4.76 Theorem. Assume
(a)
n=0
a
n
is absolutely convergent and
n=0
a
n
= A
(b)
n=0
b
n
is convergent and
n=0
b
n
= B
(c) c
n
=
n
k=0
a
k
b
nk
n = 0, 1, 2, . . .
Then
n=0
c
n
is convergent and has sum C = AB.
4.77 Remark. The above theorem is not true if both series are conditionally
convergent.
62
4.78 Example. Consider the Cauchy product of the conditionally conver-
gent series
n=0
(1)
n
n + 1
=
1
2
+
1
4
+
with itself. We have a
n
= b
n
=
(1)
n
n+1
. Then
c
n
=
n
k=0
(1)
k
k + 1
(1)
nk
n k + 1
= (1)
n
n
k=0
1
_
(k + 1)(n k + 1)
We have
(nk+1)(k+1) = nk+nk
2
k+k+1 = n+1+k(nk) n+1+
n
2
4
=
_
n
2
+ 1
_
2
When 0 x n, max. of x(n x) is
n
2
4
so we have
[c
n
[ =
n
k=0
1
_
(k + 1)(n k + 1)
k=0
1
_
_
n
2
+ 1
_
=
n
k=0
1
n
2
+ 1
=
n + 1
n
2
+ 1
=
2n + 2
n + 2
1
So lim
n
c
n
= 0 cannot be true. So the series
c
n
is divergent.
63
5 Continuity
5.1 General
Let (X, d
X
) and (Y, d
Y
) be two metric spaces. Let E ,= be a non-empty
subset of X, f : E Y , p E
, q Y . We say lim
xp
f(x) = q (or f(x) q
as x p) if for every > 0 there is > 0 with the following property: For
every x E with d
X
(x, p) < we have d
Y
(f(x), q) < . Equivalently, for
every > 0 there is > 0 such that f
_
B
X
(p) E
_
B
Y
(q). > 0 depends

in general on > 0 and the point p. For lim
xp
f(x), f(p) need not be
dened.
5.1 Example. Let X = R
2
with d
2
and Y = R with [ [ metric.
Let E = (x, y) : (x, y) R
2
and xy ,= 0. f : E R, f(x, y) =
x
y
sin
_
y
x
_
Let p = (a, 0) where a > 0. Then p E
. Show lim
(x,y)(a,0)
f(x, y) = 1. Let
> 0 be given. Since lim
t0
sin t
t
= 1, we have
> 0 such that for all t with

0 < [t[ <
we have
sin t
t
1
< . Choose =
a
1+
. Then 0 < < a. Show

that for all (x, y) E with d
2
((x, y), (a, 0)) < we have that [f(x, y)1[ < .
Let (x, y) E be such that d
2
((x, y), (a, 0)) < i.e.
(x a)
2
+ (y 0)
2
<
2
[x a[ < and [y 0[ <
a < x < a + so 0 < x. Let t =
y
x
. Then
0 < [t[ =
[y[
[x[
=
[y[
x
<

a
=
a
1+
a
a
1+
=
a
1+
a+a
1+
=
a
a
=
So t =
y
x
satises 0 < [t[ <
, thus
sin t
t
1
< i.e.
sin
_
y
x
_
y
x
1
<
5.2 Theorem. lim
xp
f(x) = q for every sequence p
n
in E with
lim
n
p
n
= p and p
n
,= p we have that lim
n
f(p
n
) = q.
Proof.
(): Suppose lim
xp
f(x) = q. Let p
n
be an arbitrary sequence in E such
that lim
n
p
n
= p and p
n
,= p. Show lim
n
f(p
n
) = q. Let > 0
be given. Since lim
xp
f(x) = q, there is > 0 such that for all x E
with d
X
(x, p) < we have d
Y
(f(x), q) < . Since lim
n
p
n
= p,
there is n
0
0
, d
X
(p
n
, p) < . Let n n
0
.
Then d
Y
(f(p
n
), q) < since x = p
n
(for n n
0
) satises x E and
d
X
(x, p) < .
64
(): Proof by contraposition. Suppose lim
xp
f(x) ,= q. So there is an
0
> 0 such that for every > 0 there is x E such that d
X
(x, p) <
and d
Y
(f(x), q)
0
.
Let = 1, nd x = p
1
E s.t. d
X
(p
1
, p) < 1 and d
Y
(f(p
1
), q)
0
Let =
1
2
, nd x = p
2
E s.t d
X
(p
2
, p) <
1
2
and d
Y
(f(p
2
), q)
0
.
.
.
Let =
1
n
, nd x = p
n
E s.t. d
X
(p
n
, p) <
1
n
and d
Y
(f(p
n
), q)
0
Then p
n
is a sequence in E such that lim
n
p
n
= p and
lim
n
f(p
n
) ,= q.
5.3 Corollary. Let E X, p E
, f, g : E R such that lim

xp
f(x) = A
and lim
xp
g(x) = B. Then
lim
xp
(f(x) +g(x)) = A +B
lim
xp
f(x)g(x) = AB
lim
xp
f(x)
g(x)
=
A
B
if B ,= 0
5.4 Denition. Let (X, d
X
) and (Y, d
Y
) be metric spaces. Let , = E X,
f : E Y , p E. We say f is continuous at the point p if lim
xp
f(x) =
f(p), i.e. for every > 0 there is > 0 such that for all x E with
d
X
(x, p) < we have d
Y
(f(x), f(p)) < . In general depends on and p.
If f is continuous at every point p of E, we say f is continuous on E.
5.5 Theorem. Let (X, d
X
) and (Y, d
Y
) be metric spaces and f : X Y
(i.e. E = X). f is continuous on X for every open set V Y , the inverse
image f
1
(V ) is an open set.
Proof.
(): Let f be continuous on X. Let V Y be an arbitrary open set. Show
f
1
(V ) is an open set in X. Let p f
1
(V ) be an arbitrary point.
Then f(p) V . V is open, so there is s > 0 such that B
Y
s
(f(p)) V . f
is continuous at p. Then for = s > 0, we nd > 0 such that for all x
with d
X
(x, p) < we have d
Y
(f(x), f(p)) < . Show B
X
(p) f
1
(V ).
Let x B
X
(p), i.e. d
X
(x, p) < d
Y
(f(x), f(p)) < = s f(x)
B
Y
s
(f(p)) V . f(x) V , so x f
1
(V ).
65
(): Suppose f
1
(V ) is open for every open set V Y . Show f is continuous
on X, i.e. show f is continuous at every point of X. Let p X be
an arbitrary point. Let > 0 be given. The set V = B
Y
(f(p)) is an
open set in Y . Then f
1
(V ) is an open set in X. Also p f
1
(V ).
Then there is > 0 such that B
X
(p) f
1
(V ). Let x be such that
d
X
(x, p) < , i.e. x B
X
(p). Then x f
1
(V ), i.e. f(x) V , i.e.
d
Y
(f(x), f(p)) < .
5.6 Corollary. Let f : X Y . f is cotinuous on X for every closed set
F Y we have that the inverse image f
1
(F) is closed in X.
Proof. F Y is closed F
C
is open. Using f
1
(F
C
) = (f
1
(F))
C
and
f is continuous the inverse image of every open set is open we get the
result.
5.7 Theorem. Let (X, d
X
), (Y, d
Y
) and (Z, d
Z
) be metric spaces. , = E
X, f : E Y , g : f(E) Z, p E. If f is continuous at p and g is
continuous at q = f(p) then g f is continuous at p.
Proof. Let > 0 be given. Since g is continuous at q, we have a
> 0
such that for all y f(E) with d
Y
(y, q) <
we have d
Z
(g(y), g(q)) < .
Since f is continuous at p, we have a > 0 such that for all x E with
d
X
(x, p) < we have d
Y
(f(x), f(p)) <
. Let x E and d
X
(x, p) < . Then
d
Y
(f(x), f(p)) <
. So d
Z
(g(y), g(q)) < i.e. d
Z
(g(f(x)), g(f(p))) < .
Let (X, d) be a metric space and f : X R
k
. f(x) R
k
, so we have
f(x) = (f
1
(x), f
2
(x), . . . , f
k
(x)) where f
1
, f
2
, . . . , f
k
: X R.
5.8 Example. f : R
3
R
2
and f(x, y, z) = (x
2
y + 1
. .
f
1
(x,y,z)
, z
3
x 3
. .
f
2
(x,y,z)
) R
2
.
5.9 Theorem. f : X R
k
is continuous on X f
1
, f
2
, . . . , f
k
are all
continuous on X.
In the above example, f
1
, f
2
: R
3
R are continuous (since they are poly-
nomials) we have that f : R
3
R
2
is also continuous.
5.10 Theorem. Let f, g : X R be continuous at the point p. Then f +g
and f g are continuous at p.
f
g
is continuous at p if g(p) ,= 0.
66
5.11 Example. X = R
k
. Fix a coordinate, say j-th coordinate. Dene
f : R
k
R. f is continuous on R
k
. x = (x
1
, x
2
, . . . , x
k
) x
j
. Fix
p = (p
1
, p
2
, . . . , p
k
) in R
k
. Show f is continuous at p. Given > 0, choose
= . Let x R
k
be any point such that d
2
(x, p) < . Then
[f(x) f(p)[ = [x
j
p
j
[ =
_
(x
j
p
j
)
2
_
(x
1
p
1
)
2
+ + (x
k
p
k
)
2
= d
2
(x, p) < =
If n
1
, n
2
, . . . , n
k
are non-negative integers then dene g : R
k
R by
g(x) = x
n
1
1
x
n
2
2
x
n
k
k
. Then by the theorem, g is continuous on R
k
. So
every polynomial P(x) =
c
n
1
n
k
x
n
1
1
x
n
k
k
is continuous on X.
5.12 Example. P : R
2
R and P(x, y) = 5x
2
7x
3
y
4
+ 8y
6
+ 5xy
2
3 is
continuous on R
2
.
5.2 Continuity And Compactness
5.13 Theorem. Let f : X Y be continuous on X. Let E be a compact
subset of X. Then the image f(E) is a compact subset of Y . (Continuous
image of a compact set is compact.)
Proof. Let C = G
: A be an open cover of f(E), i.e. every G
is an open set and f(E)

A
G
. Let V
= f
1
(G
) and A. V
is open for every A. Do we have E

A
V
? Let x E. Then
f(x) f(E). Then f(x) G
0
for some
0
A. So x f
1
(G
0
) = V
0
.
So E
A
V
. So the collection C
= f
1
(G
) : A is an open cover
of E. Since E is compact, there are
1
, . . . ,
n
A such that
E f
1
(G
1
) f
1
(G
n
)
f(E) f
_
f
1
(G
1
) f
1
(G
n
)
_
= f
_
f
1
(G
1
)
_
f
_
f
1
(G
n
)
_
G
1
G
n
So C has a nite subcover G
1
, . . . , G
n
of f(E).
In the proof of the following corollary, we will need the following proposition.
5.14 Proposition. Let S ,= be a bounded subset of R. Then sup S, inf S
S.
67
Proof. For sup S only. Let = sup S. Show every neighborhood B of
contains a point s from S. B = ( , + ). Since < sup S,
cannot be an upper bound for S. So there is an element s S such that
< S. Also if s S then s < + . So < s < + , i.e.
s B.
5.15 Corollary. Let (X, d) be a compact metric space and f : X R be
continuous on X. Then there are points p, q X such that for all x X we
have f(p) f(x) f(q). (A continuous real valued function on a compact
set attains its min.= f(p) and max.= f(q))
Proof. The set S = f(X) ,= is a compact subset of R. S is bounded. Then
sup S S. S is closed, i.e. S = S so sup S = S = f(X). That is, there is
q X such that sup S = f(q). For all x X we have f(x) sup S = f(q).
Similarly, inf S S so inf S = f(p) for some p X.
5.16 Corollary. Let f : [a, b] R be continuous on [a, b]. Then there are
two points p, q [a, b] such that for all x [a, b] we have f(p) f(x) f(q).
5.17 Theorem. Let X be a compact metric space, Y be an arbitrary metric
space, f : X Y be continuous, 1-1, onto. Then the inverse function
g = f
1
: Y X is also continuous.
Proof. Show that for every closed set F X, the inverse image g
1
(F) is a
closed set in Y . We have g
1
(F) = f(F). X is compact, F is closed so F is
compact. f is continuous, so f(F) is compact. So f(F) is closed.
5.18 Remark. If compactness of X is removed then the theorem is not true.
5.19 Example. X = [0, 2] in R with d(x
1
, x
2
) = [x
1
x
2
[. Y = (x, y) :
(x, y) R
2
, x
2
+y
2
= 1 with d
2
metric restricted to Y . Dene f : X Y as
f(t) = (cos t, sin t). f is continuous, 1-1 and onto. But f
1
is not continuous
at the point p = (1, 0).
68
5.3 Continuity And Connectedness
5.20 Theorem. Let X, Y be metric spaces and f : X Y be continuous.
Assume X is connected. Then f(X) is also connected. (Continuous image
of a connected set is connected.)
Proof. Assume f(X) is disconnected. Then there are sets E, F Y such
that f(X) = E F and E F = , E F = , E ,= , F ,= . Let
A = f
1
(E) and B = f
1
(F). A ,= . Let q E f(X), so q = f(x) for
some x X. Since f(x) = q E, x f
1
(E) = A. Similarly, B ,= .
X f
1
(f(X)) = f
1
(E F) = f
1
(E) f
1
(F) = A B
Also A B X. So X = A B. Show A B = and A B = . Assume
A B ,= . Let p A B. Then p A and p B = f
1
(F)
. .
f(p)F
. p A,
then there is a sequence p
n
in A such that p
n
p. f is continuous, so
lim
n
f(p
n
) = f(p). p
n
A = f
1
(E) f(p
n
) E. So f(p) is the limit of
a sequence in E. It means that f(p) E. So f(p) E F
. .
. Contradiction.
So A B = . Then X is the union of the separated non-empty sets A, B.
It means that X is disconnected.
5.21 Corollary (Intermediate Value Theorem). Let f : [a, b] R be con-
tinuous on [a, b]. Assume f(a) and f(b) have dierent signs. Then there is a
point p such that a < p < b and f(p) = 0.
Proof. [a, b] is connected f([a, b]) is connected. So f([a, b]) = [c, d] is
an interval. The interval [c, d] contains both negative and positive numbers
(namely f(a), f(b)). So [c, d] contains y = 0. So 0 f([a, b]), i.e. there is
p [a, b] such that f(p) = 0.
5.4 Uniform Continuity
Let (X, d
X
) and (Y, d
Y
) be two metric spaces. E X and f : E Y . We
say
(i) f is continuous on E if for every p E, for every > 0 there is
= (p, ) > 0 such that for all q E with d
X
(q, p) < we have
d
Y
(f(q), f(p)) < . (In general > 0 depends on > 0 and the point
p E.)
69
(ii) f is uniformly continuous on E if for every > 0 there is =
() > 0 such that for all points p, q E with d
X
(p, q) < we have
d
Y
(f(p), f(q)) < . ( depends only on . The same works for all
p E.)
Uniform Continuity

Continuity
5.22 Example. Let X = R, Y = R, E = (0, 1), d
X
= d
Y
= [ [ and let
f : E R, f(x) =
1
x
Claim 1: f is continuous on E.
Claim 2: f is not uniformly continuous on E.
1) Let p E and > 0 be given. Then 0 0. If
q E such that [q p[ < then
[f(p) f(q)[ =
1
p

1
q
=
[q p[
pq
<

pq
We have [q p[ < , so p < q < p +. We have < p, i.e.
p
2
1 +p
< p p
2
< p +p
2
So 0 < p and
[f(p) f(q)[ <

pq
<

p(p )
=
p
2
1+p
p
_
p
p
2
1+p
_ =
p
2
1+p
p
p+p
2
p
2
1+p
=
p
2
p
2
=
So f is continuous at p E. Since p E is arbitrary, f is continuous on E.
Note that =
p
2
1+p
depends on both and p. So we are inclined to say that
f is not uniformly continuous on E. But maybe by some other calculation,
we can nd depending only on .
2) Show that f is not uniformly continuous on E, i.e. cannot be found
depending only on . Assume for = 1, we have a > 0 such that for all
p, q E with [p q[ < we have [f(p) f(q)[ < 1.
Case 1: 0 <
1
3
. Let p = and q = +

2
. Then p, q E and
[p q[ =

2
< . So [f(p) f(q)[ < 1.
[f(p) f(q)[ =
1
p

1
q
=
[p q[
pq
=
/2
3
2
=
1
3
1
70
Contradiction.
Case 2:
1
3
< . Let
=
1
3
< . As in case 1, let p =
and q =
2
. Then
[p q[ <
< and [f(p) f(q)[ 1. Contradiction.

5.23 Example. Let X = Y = R, E = [2, 5], f : E R, f(x) = x
2
. Then
f is uniformly continuous on E. Let > 0 be given. Let =

10
> 0. Let
p, q E be such that [p q[ < . Then
[f(p) f(q)[ = [p
2
q
2
[ = [(p q)(p +q)[
= [p q[[p +q[ = [p q[
. .
<
(p +q)
. .
10
< 10 =
5.24 Theorem. Let f : X Y be continuous on X and let E X be
compact. Then f is uniformly continuous on E.
Proof. Let > 0 be given. Given p E, since f is continuous at p, we have
a = (p, ) > 0 such that for all q E with d
X
(q, p) < (p, ) we have
d
Y
(f(q) f(p)) <

3
C =
_
B(p,)
3
(p) : p E
_
Do this for every p E. Then C is an open cover of E. Since E is compact,
this open cover has a nite subcover
C
=
_
B(p
1
,)
3
(p
1
), . . . , B(p
n
,)
3
(p
n
)
_
for some nite set p
1
, . . . , p
n
E. So
E B(p
1
,)
3
(p
1
) B(p
n
,)
3
(p
n
)
Let = min
_
(p
1
,)
3
, ,
(p
n
,)
3
_
. Then > 0. Show this > 0 has the
property in the denition of uniform continuity. Let p, q E be two ar-
bitrary points such that d
X
(p, q) < . We have that p B(p
i
,)
3
(p
i
) and
q B(p
j
,)
3
(p
j
) for some p
i
, p
j
from p
1
, . . . , p
n
.
p B(p
i
,)
3
(p
i
) q B(p
j
,)
3
(p
j
)

d
X
(p, p
i
) <
(p
i
,)
3
< (p
i
, ) d
X
(q, p
j
) <
(p
j
,)
3
< (p
j
, )

d
Y
(f(p), f(p
i
)) <

3
d
Y
(f(q), f(p
j
)) <

3
71
Assume (p
i
, ) (p
j
, ). Also
d
X
(p
i
, p
j
) d
X
(p
i
, p)
. .
<
(p
i
,)
3
+d
X
(p, q)
. .
<
+d
X
(q, p
j
)
. .
<
(p
j
,)
3
< (p
j
, ) d
Y
(f(p
i
), f(p
j
)) <

3
So we have
d
Y
(f(p), f(q)) d
Y
(f(p), f(p
i
))
. .
<
3
+d
Y
(f(p
i
), f(p
j
))
. .
<
3
+d
Y
(f(p
j
), f(q))
. .
<
3
<
72
6 Sequences And Series Of Functions
6.1 General
Consider the following sequence of functions dened for 0 x 1.
f
1
(x) = x, f
2
(x) = x
2
, f
3
(x) = x
3
, . . . , f
n
(x) = x
n
, . . .
Fix any x, 0 x 1 and consider lim
n
f
n
(x) = lim
n
x
n
(x: xed). If
0 x 1 then lim
n
x
n
= 0. If x = 1 then lim
n
x
n
= 1. Dene
f(x) =
_
0 if 0 x 1
1 if x = 1
Then for every xed x, 0 x 1, we have lim
n
f
n
(x) = f(x).
6.1 Denition. Let E be any non-empty set and f
n
: E R, n = 1, 2, . . .
f : E R. We say f
n
f pointwise on E if for every xed x E,
lim
n
f
n
(x) = f(x), i.e. for every x E and for every > 0, there is a
natural number N = N(x, ) such that for all n N, [f
n
(x) f(x)[ < .
f is called the pointwise limit of f
n
. In the above example, observe that
every f
n
is continuous but their pointwise limit f is not continuous on the
set E = [0, 1]. Also every f
n
is dierentiable on the interval E = [0, 1] but
their pointwise limit f is not dierentiable on E = [0, 1].
6.2 Example. Consider f
n
(x) =
sin nx
n
on E = R, f(x) = 0. Then for every
xed x R, lim
n
f
n
(x) = 0 = f(x). We have f
n
(x) =

ncos nx and
f
(x) = 0. But lim

n
f
n
(x) ,= f
(x). Take x = 0, then f
n
(0) =

n f
(0).
6.3 Example. On E = [0, 1], consider the following sequence
f
n
(x) =
_
_
4n
2
x if 0 x
1
2n
4n 4n
2
x if
1
2n
x
1
n
0 if
1
n
x 1
lim
n
f
n
(x) = 0 for every xed x so f(x) = 0. We have
_
1
0
f(x)dx = 0 and
_
1
0
f
n
(x)dx =
1
2
1
n
2n = 1 so lim
n
_
1
0
f
n
(x)dx = 1
So we have
lim
n
_
1
0
f
n
(x)dx ,=
_
1
0
lim
n
f
n
(x)dx
73
6.4 Example. f
n
(x) =
x
2
(1+x
2
)
n
and E = R. Consider f(x) =
n=0
f
n
(x)
where n = 0, 1, 2, . . .
f(x) = x
2
+
x
2
1 +x
2
+
x
2
(1 +x
2
)
2
+ +
x
2
(1 +x
2
)
n
+
= x
2
_
1 +
1
1 +x
2
+
_
1
1 +x
2
_
2
+ +
_
1
1 +x
2
_
n
+
_
. .
geometric series with r=
1
1+x
2
= x
2
1
1
1
1+x
2
= 1 + x
2
if x ,= 0
If x = 0 then f(0) = 0 + 0 + = 0. So
f(x) =
_
1 +x
2
if x ,= 0
0 if x = 0
So the sum f(x) of continuous functions
f
n
(x) is not continuous on R.
Pointwise convergence is not strong enough for the calculus of limits of se-
quences of functions.
6.2 Uniform Convergence
6.5 Denition. Let E be any non-empty set and f
n
: E R, n = 1, 2, . . .,
f : E R be functions. We say f
n
f uniformly on E if for every > 0
we have N = N() such that for all n N() and for all x E we have
[f
n
(x) f(x)[ < . Here N = N() depends on only and it works for every
x E.
6.6 Example. Let 0 < c < 1 be a xed constant. Let E = [0, c], f
n
(x) = x
n
.
We have for every xed x E, lim
n
f
n
(x) = lim
n
x
n
= 0. So f(x) = 0,
i.e. x
n
0 pointwise on E. Does x
n
0 uniformly on E ? Let > 0 be
given. Since lim
n
c
n
= 0, we have N such that c
N
< . Let n N, x E
[f
n
(x) f(x)[ = [x
n
0[ = x
n
c
n
c
N
<
6.7 Example. E = [0, 1), f
n
(x) = x
n
. For every xed x with 0 x < 1, we
have lim
n
f
n
(x) = lim
n
x
n
= 0. So f(x) = 0 and f
n
f, i.e. x
n
0
pointwise on E = [0, 1). But this convergence is not uniform. Assume f
n
f
74
uniformly on E. Then for =
1
4
we can nd N
1
such that for all n N
1
and
for all x E = [0, 1) we have [f
n
(x) f(x)[ <
1
4
i.e. x
n
<
1
4
. Also
lim
n
_
n + 1
n
_
n
= lim
n
_
1 +
1
n
_
= e lim
n
_
n
n + 1
_
n
=
1
e
So for =
1
e

1
3
> 0 we have N
2
such that for all n N
2
we have
_
n
n + 1
_
n
1
e
<
1
e

1
3

1
e
+
1
3
<
_
n
n + 1
_
n
1
e
<
1
e

1
3
1
3
<
_
n
n + 1
_
n
for all n N
2
Let N = maxN
1
, N
2
, x =
N
N+1
and x E. Since N N
1
, we have x
N
<
1
4
and since N N
2
, we have
1
3
< x
N
. So
1
3
<
_
N
N+1
_
N
<
1
4
i.e.
1
3
<
1
4
which is
not true.
Cauchy Criterion For Uniform Convergence
Let E ,= , f
n
: E R, n = 1, 2, . . . Assume for every > 0 there is a
natural number N = N() such that for all n, m N() and for all x E
we have [f
n
(x) f
m
(x)[ < . Then there is a function f : E R such that
f
n
f uniformly on E.
If we have a series of functions
n=1
f
n
(x) dened on a set E, we dene
s
n
(x) = f
1
(x) + + f
n
(x). If there is a function f : E R such that
s
n
f uniformly on E then we say the series
n=1
f
n
(x) = f(x) uniformly
on E.
Cauchy Criterion: Assume for every > 0, there is a natural number
N = N() such that for all n, m N() with n m and for all x E
we have [
n
k=m
f
k
(x)[ < . Then there is a function f : E R such that
n=1
f
n
(x) = f(x) uniformly on E.
Weierstrass M-Test: Let f
n
: E R, n = 1, 2, . . . Assume for every n
there is a number M
n
> 0 such that
(i) [f
n
(x)[ M
n
for all x E
(ii)
n=1
M
n
is convergent
Then the series
n=1
f
n
(x) converges uniformly to some function f(x) on E.
6.8 Example. E = R. Consider
n=1
cos(2nx)
(2n1)(2n+1)
. Then f
n
(x) =
cos(2nx)
(2n1)(2n+1)
[f
n
(x)[ =
[ cos(2nx)[
(2n 1)(2n + 1)

1
(2n 1)(2n + 1)
= M
n
for all x E
75
n=1
M
n
is convergent since 0 < M
n

1
n
2
. So there is a function f : R R
such that
n=1
f
n
(x) = f(x) uniformly on R.
n=1
x
n+n
2
x
2
and E = [0, +). We have
f
n
(x) = [f
n
(x)[ =
x
n +n
2
x
2
To nd M
n
we use calculus. Find max. of f
n
(x) for x 0.
f
n
(x) =
n +n
2
x
2
xn
2
2x
(n +n
2
x
2
)
2
=
n n
2
x
2
(n +n
2
x
2
)
2
= 0 x
2
=
1
n
x =
1
n
0 x
1
n
x
2
1
n
n
2
x
2
n 0 n n
2
x
2
f
n
(x) 0
1
n
x
1
n
x
2
n n
2
x
2
n n
2
x
2
0 f
n
(x) 0
So f
n
(x) has its max. at the point x =
1
n
.
M
n
= f
n
_
1
n
_
=
1
n
n +n
2
1
n
=
1
2n
3/2
M
n
=
1
2
1
n
3/2
is convergent. So there is a function f : E R such that
n=1
x
n+n
2
x
2
= f(x) uniformly on the set E = [0, +).
6.10 Example. Consider a power series
n=0
c
n
x
n
= c
0
+c
1
x +c
2
x
2
+
Assume it has radius of convergence R > 0. If x = limsup
n
n
_
[c
n
[ then
R =
1
. Let 0 < r < R and E = [r, r]. f

n
(x) = c
n
x
n
. For all x E
[f
n
(x)[ = [c
n
[[x[
n
[c
n
[r
n
. .
M
n
Is
M
n
convergent ? Use root test.
limsup
n
n
_
[M
n
[ = limsup
n
n
_
[c
n
[r = r limsup
n
n
_
[c
n
[
. .
= r < R = 1
So by the root test,
M
n
is convergent. So the power series
c
n
x
n
con-
verges uniformly on E = [r, r] where 0 < r < R.
76
6.3 Uniform Convergence And Continuity
6.11 Theorem. Let (X, d) be a metric space and E ,= subset of X.
f
n
: E R, n = 1, 2, . . . and f : E R. Assume f
n
f uniformly on E.
Let x
0
be a limit point of E and assume for every n, lim
xx
0
f
n
(x) = A
n
.
Then A
n
is convergent and lim
xx
0
f(x) = lim
n
A
n
. That is
lim
xx
0
lim
n
f
n
(x)
. .
f(x)
= lim
n
lim
xx
0
f
n
(x)
. .
A
n
The two limits can be interchanged.
Proof. Show A
n
is a Cauchy sequence in R. Given > 0, nd N = N()
such that for all n N() and for all x E, [f
n
(x) f(x)[ <

2
. Let
n, m N(). Then for any x E
[f
n
(x) f
m
(x)[ [f
n
(x) f(x)[
. .
<
2
+[f(x) f
m
(x)[
. .
<
2
<
This proof shows uniformly convergent uniformly Cauchy
Take n, m N() and x them. For every x E we have [f
n
(x)f
m
(x)[ < .
Let x x
0
. [A
n
A
m
[ . True for all n, m N(). So A
n
is Cauchy.
Since R is complete, lim
n
A
n
= A exists in R. To show lim
xx
0
f(x) = A,
let > 0 be given. f
n
f uniformly on E, so there is N
1
= N
1
() such that
for all n N
1
() and for all x E
[f
n
(x) f(x)[ <

3
(1)
A
n
A, so there is N
2
= N
2
() such that for all n N
2
() we have
[A
n
A[ <

3
(2)
Let N = maxN
1
(), N
2
(). Since lim
xx
0
f
N
(x) = A
N
, we have > 0 such
that for all x E with d
X
(x, x
0
) < we have
[f
N
(x) A
N
[ <

3
(3)
Let x E and d
X
(x, x
0
) < . Then
[f(x) A[ [f(x) f
N
(x)[
. .
<
3
by (1)
+[f
N
(x) A
N
[
. .
<
3
by (3)
+[A
N
A[
. .
<
3
by (2)
<
77
6.12 Corollary. Let (X, d) be a metric space. Let f
n
: X R, n = 1, 2, . . .
f : X R. Assume f
n
f uniformly on X and each f
n
is continuous on
X. Then f is also continuous on X. (Uniform limit of continuous functions
is continuous.)
Proof. Fix x
0
X. Show lim
xx
0
f(x) = f(x
0
).
lim
xx
0
f(x) = lim
xx
0
lim
n
f
n
(x) = lim
n
lim
xx
0
f
n
(x) = lim
n
f
n
(x
0
) = f(x
0
)
6.13 Remark. If each f
n
is uniformly continuous on X and f
n
f uni-
formly on X then f is also uniformly continuous on X.
6.14 Example. E = [0, 1], f
n
(x) = x
n
, n = 1, 2, . . .
f(x) =
_
0 if 0 x < 1
1 if x = 1
Each f
n
is continuous on E but f is not continuous. So f
n
does not
converge to f uniformly.
lim
x1
lim
n
x
n
. .
0
. .
0
,= lim
n
lim
x1
x
n
. .
1
. .
1
(1) x xed. Take limit as n (1) n xed. Take limit as x 1
(2) Take limit as x 1
(2) Take limit as n

6.15 Corollary. Let (X, d) be a metric space. Assume f
n
: X R, n =
1, 2, . . . is continuous on X for every n and
n=1
f
n
(x) = f(x) uniformly on
X. Then f is also continuous on X.
lim
xx
0
n=1
f
n
(x) and
n=1
lim
xx
0
f
n
(x)

lim
xx
0
f(x)
n=1
f
n
(x
0
)

f(x
0
)
equal
= f(x
0
)
So lim
xx
0
n=1
f
n
(x) =
n=1
lim
xx
0
f
n
(x) if
f
n
(x) = f(x) uniformly
on X.
78
n=1
x(1 x)
n
= x(1 x) +x(1 x)
2
+x(1 x)
3
+
= x(1 x)
_
1 + (1 x) + (1 x)
2
+
. .
geometric series with r=1x
Also for x = 0 we have 0 + 0 + so let f(x) =
n=1
x(1 x)
n
. Then
f(x) =
_
1 x if 0 < x < 2
0 if x = 0
So E = [0, 2). Do we have uniform convergence on E = [0, 2) ?
lim
x0
+
n=1
x(1 x)
n
?
=
n=1
lim
x0
+
x(1 x)
n
LHS=lim
x0
+ f(x) = lim
x0
+(1 x) = 1
RHS=
n=1
lim
x0
+ x(1 x)
n
= 0 + 0 + = 0
1 ,= 0 so convergence is not uniform.
6.17 Example. lim
x0
+
n=1
nx
2
n
3
+x
3
= ? Take E = [0, 1].
f
n
(x) = [f
n
(x)[ =
nx
2
n
3
+x
3

n
n
3
=
1
n
2
= M
n
M
n
is convergent so by Weierstrass M-test,
n=1
nx
2
n
3
+x
3
converges to some
f(x) uniformly on E. Then
lim
x0
+
n=1
nx
2
n
3
+x
3
=
n=1
lim
x0
+
nx
2
n
3
+x
3
=
n=1
0 = 0 + 0 + = 0
6.18 Example. Let
n=0
c
n
x
n
= c
0
+ c
1
x + c
2
x
2
+ be a power series
with radius of convergence R > 0. Then for all x with R < x < R, the
power series converges. Let x
0
be such that R < x
0
< R. Do we have
lim
xx
0
n=0
c
n
x
n
?
=
n=0
lim
xx
0
c
n
x
n
79
Find r > 0 such that R < r < x
0
< r < R. If E = [r, r] then the power
series converges uniformly on E and x
0
E. So
lim
xx
0
n=0
c
n
x
n
=
n=0
lim
xx
0
c
n
x
n
So given a power series
n=0
c
n
x
n
with R > 0 and given any x
0
such that
R < x
0
< R, we have
lim
xx
0
n=0
c
n
x
n
=
n=0
lim
xx
0
c
n
x
n
=
n=0
c
n
x
n
0
6.4 Uniform Convergence And Integration
6.19 Theorem. Assume f
n
: [a, b] R, n = 1, 2, . . . are integrable on [a, b]
(continuous functions are integrable) and f
n
f uniformly on [a, b] for some
f : [a, b] R. Then f is also integrable on [a, b] and
_
b
a
f(x)dx = lim
n
_
b
a
f
n
(x)dx
Proof. We omit the integrability proof. Let > 0 be given. Since f
n
f
uniformly on [a, b], we have N such that for all n N, for all a x b,
[f
n
(x) f(x)[ <
. Let n N. Then
_
b
a
f
n
(x)dx
_
b
a
f(x)dx
_
b
a
(f
n
(x) f(x))dx
_
b
a
[f
n
(x) f(x)[dx
_
b
a
dx =
(b a)
< 2
(b a)
. .
So let
=

2(ba)
6.20 Example. E = [0, 1], f
n
(x) = n
2
x
n
(1 x). Let x E be xed.
If x = 0 then f
n
(0) = 0 0
If x = 1 then f
n
(1) = 0 0
80
If x = 0 < x < 1 then f
n
(x) = n
2
x
n
(1 x) 0
So f(x) = 0 for all 0 x 1. Do we have f
n
f uniformly on [0, 1] ?
_
1
0
f(x)dx
?
= lim
n
_
1
0
f
n
(x)dx
For LHS we have
_
1
0
f(x)dx =
_
1
0
0dx = 0
For RHS we have
lim
n
_
1
0
f
n
(x)dx = lim
n
__
1
0
n
2
x
n
dx
_
1
0
n
2
x
n+1
dx
_
= lim
n
_
n
2
x
n+1
n + 1
1
0
n
2
x
n+2
n + 2
1
0
_
= lim
n
n
2
_
1
n + 1

1
n + 2
_
= lim
n
n
2
(n + 1)(n + 2)
= 1
0 ,= 1 so convergence is not uniform.
6.21 Corollary. Assume f
n
: [a, b] R, n = 1, 2, . . . are integrable on [a, b]
and the series
n=1
f
n
(x) converges uniformly on [a, b]. Then
n=1
f
n
(x) is
also integrable on [a, b] and
_
b
a
_

n=1
f
n
(x)
_
dx =
n=1
_
b
a
f
n
(x)dx
6.22 Example. Consider F(x) =
n=1
x
n(x+n)
where 0 x 1. Show the
series converges uniformly on E = [0, 1].
f
n
(x) = [f
n
(x)[ =
x
n(x +n)

1
n
2
= M
n
M
n
is convergent. So by Weierstrass M-test, the series
n=1
x
n(x+n)
is
uniformly convergent on E = [0, 1]. Let us call
_
1
0
F(x)dx =
81
Then we have
=
n=1
_
1
0
x
n(x +n)
dx =
n=1
_
1
0
_
1
n

1
x +n
_
dx
=
n=1
_
1
n
x ln(x +n)
_
1
0
=
n=1
_
1
n
ln(1 +n) + ln n
_
= lim
n
_
n
k=1
_
1
k
ln(1 +k) + ln k
_
_
= lim
n
_
n
k=1
1
k
ln 2 + ln 1 ln 3 + ln 2 ln(1 +n) + ln n
_
= lim
n
_
n
k=1
1
k
ln(n + 1)
_
Let us dene
n
=
n
k=1
1
k
ln(n + 1)
Then lim
n
n
= .
n
k=1
1
k
=
n
+ ln(n + 1)
=
n
+ ln n + ln(n + 1) ln n
= + ln n + ln(n + 1) ln n +
n
. .
call
n
= + ln n +
n
So
n
k=1
= ln n + +
n
where
n
0 as n . is called Eulers
constant. 0.57721. It is not known whether is rational or irrational.
So for large n, 1 +
1
2
+
1
3
+ +
1
n
ln n +.
6.23 Example. Let
n=0
c
n
x
n
= c
0
+ c
1
x + c
2
x
2
+ be a power series
with radius of convergence R > 0. Take any x
0
such that R < x
0
< R.
Then
_
x
0
0
_

n=0
c
n
x
n
_
dx =
n=0
_
x
0
0
c
n
x
n
dx
82
6.24 Example.
n=1
1
2
n
n
= ? Consider
n=0
x
n
with R = 1. Take x
0
=
1
2
.
_
1/2
0
_

n=0
x
n
_
dx =
_
1/2
0
1
1 x
dx = ln(1 x)
1/2
0
= ln
1
2
= ln 2
n=0
_
1/2
0
x
n
dx =
n=0
x
n+1
n + 1
1/2
0
=
n=0
1
2
n+1
(n + 1)
=
n=1
1
2
n
n
We know that
_
1/2
0
_

n=0
x
n
_
dx =
n=0
_
1/2
0
x
n
dx
So we get
n=1
1
2
n
n
= ln 2
6.5 Uniform Convergence And Dierentiation
6.25 Theorem. Let f
n
: [a, b] R, n = 1, 2, . . . be dierentiable functions.
Assume
(i) f
n
converges uniformly to some function g on [a, b].
(ii) There is x
0
[a, b] such that f
n
(x
0
) is convergent.
Then there is a dierentiable function f : [a, b] R such that f
n
f
uniformly on [a, b] and f
(x) = g(x) for all x [a, b].

Proof. f
n
is uniformly convergent on [a, b]. Use Cauchy criterion. Let
> 0 be given. Let
= Find N
1
such that for all n, m N
1
, [f
n
(x
0
)
f
m
(x
0
)[ <
. Find N
2
such that for all n, m N
2
and for all x [a, b],
[f
n
(x) f
m
(x)[ <
. Let N = maxN
1
, N
2
and n, m N. Take x [a, b].
Apply Mean Value Theorem to f
n
f
m
on the interval [x
0
, x] (or [x, x
0
]).
Then there is a point t between x
0
and x
f
n
(x) f
m
(x) (f
n
(x
0
) f
m
(x
0
)) = (f
n
(t) f
m
(t))(x x
0
)
[f
n
(x) f
m
(x) (f
n
(x
0
) f
m
(x
0
))[ = [f
n
(t) f
m
(t)[
. .
<
[x x
0
[
. .
ba
<
(b a)
Then we have
[f
n
(x) f
m
(x)[ [f
n
(x) f
m
(x) (f
n
(x
0
) f
m
(x
0
))[ +[f
n
(x
0
) f
m
(x
0
)[
<
(b a) +
(b a + 1)
83
So let
=

ba+1
. Then there is a function f : [a, b] R such that f
n
f
uniformly on [a, b] and f
(x) = g(x) for all x [a, b].

6.26 Corollary. Let f
n
: [a, b] R, n = 1, 2, . . . be dierentiable functions.
Assume
(i)
n=1
f
n
(x) converges uniformly on [a, b].
(ii) There is x
0
[a, b] such that
n=1
f
n
(x
0
) is convergent.
Then the series
n=1
f
n
(x) is uniformly convergent on [a, b] and
_

n=1
f
n
(x)
_
n=1
f
n
(x)
6.27 Example. Let the power series
n=0
c
n
x
n
= c
0
+c
1
x+c
2
x
2
+ have
radius of convergence R > 0. Then R =
1
where = limsup
n
n
_
[c
n
[.
Here f
n
(x) = c
n
x
n
.
n=0
f
n
(x) =
n=1
c
n
nx
n1
=
n=0
c
n+1
(n + 1)x
n
For this series
limsup
n
n
_
[c
n+1
(n + 1)[ = limsup
n
_
n+1
_
[c
n+1
(n + 1)[
_n+1
n
= limsup
n
_
n+1
_
[c
n+1
[
_n+1
n
_
n+1
n + 1
_n+1
n
=
So the dierentiated series
n=1
c
n
nx
n1
and the original series
n=0
c
n
x
n
have the same R. So if 0 < r < R then
n=1
c
n
nx
n1
converges uniformly
on [r, r]. Then for all x with r x r we have
_

n=0
c
n
x
n
_
n=1
c
n
nx
n1
Given any x with R < x < R, we can nd a number r such that 0 < r < R
and r x r. So
_

n=0
c
n
x
n
_
n=1
c
n
nx
n1
is true for all x with R < x < R.
84
6.28 Remark. This result is not true for a general series of functions.
6.29 Example.
n=1
n
2
n
= ? Let x =
1
2
. Then we have
n=1
nx
n
= x
n=1
nx
n1
= x
_

n=1
x
n
_
=
_
1
1 x
_
=
1
(1 x)
2
for 1 < x < 1. So we have
n=1
n
2
n
=
1
2
_
1
1
2
_
2
=
1
2
1
4
= 2
THE END
85
7 Figures
Figure 1: Vertical distances between two graphs
B
r
(p) B
r
[p]
Figure 2: Open and closed balls with d
2
metric
86
Figure 3: B
r
(p) with d
1
metric
Figure 4: B
r
(0) with d
metric
Figure 5: B
r
(f)
87
Figure 6: p is an interior point of E but q is not
The set E int E
Figure 7: The set E and int E
Figure 8: E is an open set
Figure 9: N is a neighborhood of p but not a neighborhood of q
88
The set E Set of limit points
of E
Figure 10: The set E and its limit points
E is closed F is not closed and
not open
Figure 11: A set may be neither closed nor open
E is perfect F is not perfect
Figure 12: E is perfect but F has an isolated point
89
E is bounded F is unbounded
Figure 13: Bounded and unbounded sets
Figure 14: A k-cell in R
2
90

Math 215 Lecture Notes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Math 215 Lecture Notes

Uploaded by

Copyright:

Available Formats

MATH 215

M.A. student at the Department of Economics, Bilkent University, Bilkent, 06800,

(c) exists and it is a number.

2 are all upper bounds for E.

any upper bound for E, then b b

1.13 Remark. sup E need not be a member of E. If sup E is in E then it

are two least upper bounds for E.

is an upper bound for E b b

= sup E and b is an upper bound for E b

). This way we can

A such that f(x) = f(x

. We have int E E. (See Figure 7).

is the set of all limit points of E. We dene E = EE

E. Since E E is always true, we have

contained in E such that p

) are also complete.)

A for all n 2. For n 2

= inf E (Considered in the set of

is called the upper limit (or limit superior) of x

is called the lower limit (or limit inferior) of x

p = 1 for all n, so lim

(q). > 0 depends

> 0 such that for all t with

. Then 0 < < a. Show

, f, g : E R such that lim

: A be an open cover of f(E), i.e. every G

is an open set and f(E)

is open for every A. Do we have E

< and [f(p) f(q)[ 1. Contradiction.

(x) = 0. But lim

(x). Take x = 0, then f

. Let 0 < r < R and E = [r, r]. f

(2) Take limit as x 1

(2) Take limit as n

(x) = g(x) for all x [a, b].

(x) = g(x) for all x [a, b].

You might also like