1

Math 4130
Spring 2021

Anil Nerode

Elements of Series c

Given an in…nite sequence

a0; a1; :::

of real numbers, the in…nite se-

quence
2

Pn= a0+::: + an

of partial sums of the …rst n

elements

ofthe sequence de…nes the in-

…nite series
a0+::: + an+::: .

The in…nite series

a0+::: + an+:::

is called Cauchy convergent

3

if the sequence of partial sums

fP ng = f n
k=0 ang

is a Cauchy sequence.

Cauchy Criterion

Theorem: An in…nite series

a0+::: + an+:::

is Cauchy convergent i¤ for all

> 0,
4

there exists an integer N such

that

for all m; n > N with m < n; we have

jam+1+ : : : + anj < :

De…nition: The in…nite series

a0+::: + an+:::

is said to have limit L if

the sequence of partial sums

fP ng has limit L,
5

written

L= an= limn !1 n
k=0 ak

De…ne addition and scalar mul-

tiplication

of series term by term.

(a0+::: + an+:::) + (b0+::: + bn+:::) =

(a0+b0) + ::: + (an+bn) + :::

(a0+::: + an+:::) =

a0+::: + an+:::
6

With these operations

in…nite series form an

in…nite dimensional vector space

over the real numbers.

We can use the vector space

axioms

for addition and multiplication

freely.

Theorem: Convergent series

form a
7

vector subspace of the space of

series.

That is, a linear combination

an + bn of
a0+::: + an+::: and b0+::: + bn+:::

is also a convergent series and

( an + bn) = ( ( an+ bn)

Proof
n (
k=0 ak + bk ) = n
k=0 ak +
n b a
k=1 k k .
8

Take limits of each side.

Example: Let e0 be the in…-

nite series 1 + 0 + 0 + 0:::;

let e1 be the in…nite series 0 + 1 +

0 + 0 + 0 : : :, etc.

This sequence en is an in…nite

independent set,

so the vector space of conver-

gent series

is not …nite dimensional.

9

Theorem: If all an 0 , then

a0 + ::: + an + :::

converges if and only if

the sequence of partial sums

a0 ; a 0 + a1 , a0 + a1 + a2 ; : : :

is bounded from above.

Then the sum an is the

least upper bound of the par-

tial sums
10

Pn = a0 + : : : + an

Proof: A monotone sequence

of

non-negative reals converges

i¤

its terms have a common up-

per bound.

Apply this to partial sums.

Example: A telescoping se-

ries
11

is of the form
(b1 b0) + (b2 b1) + (b3 b2) : : :

Note that n (b
k=0 n+1 bn) = bn+1 b0

and the sequence fbn+1 b0g converges

i¤
fbngconverges i¤ limn !1 bn =b ex-
ists.

Exercise: 1a) Suppose a0; a1; :::

is a sequence of positive num-

bers
12

monotonely converging to zero.

Then a0 a1 + a2 : : : converges.

1b) Conclude that the harmonic

series
1 1+1 ::::
1 2 3

converges.

Example: the series of absolute

values

of the terms of the harmonic

series,
13

namely 1
1 + 12 + 13 + : : : :diverges,

even though the n th term has

limit zero.

Proof: For = 12 and for any

integer N ,

choose 2k > N: Then

k
1 +
2 +1 k
1
2 +2
+ ::: + 1
2 +2k
k
2k
2k +2k
1
2 ,

The Cauchy criterion is violated

so
1
1 + 12 + 31 + : : : series diverges.
14

Simple Tests for

Convergence

Comparison test: Suppose

that

for all n, an > 0 and bn > 0: Sup-

pose c > 0:

Suppose that there exists

an integer N such that

for all n > N; an < cbn .

15

If b0 + : : : + bn + ::: converges,

then a0 + : : : + an + :::also converges.

Proof:The partial sums a0 + : : : +

an are bounded by
c times the corresponding par-
tial sums b0 + : : : + bn

which is a monotone increas-

ing sequence

with its limit as an upper bound.

16

So the partial sums a0+: : :+an constitute

a monotone increasing sequence

bounded above and therefore

has a limit.

An in…nite series a0 + ::: + an + :::is

absolutely convergent if
ja0j + ::: + janj + :::

converges.

An in…nite series that is con-

vergent
17

but not absolutely convergent

is called conditionally conver-

gent.

The harmonic series is condi-

tionally convergent.

Tests for Convergence

Example: Suppose jxj < 11. Then

1 + x + x2 + : : : converges to 1 x .
18

This is a consequence of the

algebraic identity
(1 x) n k
k=0 x =
n
k=0 xk xk+1 = 1 xn+1 .

If jxj < 1, then certainly limn !1 x

n+1 =
0:

But if jxj > 1, the terms do not

tend to zero,

and the series diverges.

De…nition: Let f : A !R be a
function mapping
19

a set of reals A to the reals R:

Then is called uniformly

f
continuous on A
if for any > 0; there exists a >0
such that

for all x1 ; x 2 in A,
jx1 x2 j < implies jf (x1) f ( x2 ) j < .

Theorem: (Weierstrass M-
test)
20

Suppose that ffn(x)g is a sequence

of functions

de…ned on a set S of real num-

bers and that
fMng is a sequence of non-negative
reals

such that Mn < 1:

If for all integers n and all x 2 S ,

0 ,then ffng con-

jfn(x)j Mn
verges uniformly on S:
21

Proof: Using Mn < 1; and the

Cauchy criterion

for that sum, apply the Cauchy

criterion to
n+p
k=n+1 jfk (x)j
n+p
k=n+1 Mk , which holds
for all x2S .

De…nition: If a0; a1; : : : is a se-

quence of reals,

then U =lim sup an i¤ 1) and 2) be-

low:
22

1) for all > 0; there exists an

integer N

such that for all n > N; an < U + .

(That is, all but a …nite num-

ber of an

are less than U+ )

2) For all >0 , all integers M > 0;

there exists an integer n>M

23

such that an > U :

(That is, in…nitely many an are

larger than U )

De…nition: If a0; a1; : : : is a se-

quence of reals, then
L =lim inf an i¤ 1) and 2) below:

1) for all > 0; there exists an

integer N

such that for all n > N; an > L + .

24

(That is, all but a …nite num-

ber of an

are less than UL + )

2) For all >0 , all integers M > 0;

there exists an integer n>M

such that an > L :

(That is, in…nitely many an

are smaller than L :)

25

Ratio test: Suppose that a0 +

a1 + : : : is

an in…nite series of non-zero

reals.

1) If a
R = lim sup j n+1
a n j < 1;

the series converges absolutely.

2) If a
r = lim inf j n+1
an j > 1 ,

the series diverges.

26

Proof of 1 Suppose R<1 .

Choose an x; R < x < 1:

The de…nition of lim sup implies

there exists an integer N such

that

for all n>N , a

j n+1
an j < x:

Therefore for all n N;

jan+1 j janj jaN j

xn+1
< xn ;
xN
27

Let c = jaN jx N . Then for n N;

janj cxn.

Apply the comparison test.

Proof of 2 Since r > 1;

the de…nition of lim inf implies

there exists an integer N such

that

for all n N , jan+1j > janj ,

28

so we cannot have limn !1 an = 0:

Root Test: Given series a0 +

a1 + : : : ,

p
de…ne = lim supn !1 n an :

1) The series converges if <1

2) The series diverges if >1

Proof of 1: With < 1; choose

x with <x<1 .
29

The de…nition of

implies there exists an integer

N

such that for n N , janj < xn:

Apply the comparison test.

Proof of 2: >1 implies that

for in…nitely many n, janj > 1:

So limn !1 an is not zero.

30

Reorderings of Terms in
a Series

A series is non-negative if
every term is 0:

Theorem:
1. Any reordering b0 + ::: + bm + : : : ::
of all the terms

of a non-negative series a0 + ::: +

an + :::

is convergent with the same sum.

31

2. If a0 + ::: + an + ::: , b0 + ::: + bm + : : :

are both convergent non-negative

series

with sums A; B , respectively,

then
(a0+b0) + ::: + (an+bn) : : :

is a convergent non-negative se-

ries

with sum A+B .

3. If
32

a0 + ::: + an + ::: , b0 + ::: + bm + : : :

are convergent non-negative se-

ries

with sums A; B , respectively,

and c0 ; c 1 ; c 2 : : : is a sequence

consisting of all products aibj ;

where i; j range over 0; 1; 2; : : :then

c0 + c1 + c2 + : : :
33

is a convergent

non-negative series with sum

AB .

This is called the Cauchy Prod-

uct

of the two series.

4) If a0 + ::: + an + ::: , b0 + ::: + bm + : : :

are both convergent non-negative

series
34

with sums A; B , respectively,

then
(a0 b0) + ::: + (an bn) : : :

is a convergent series with sum

A B

Proof:
1)The partial sums
Pn= a0+::: + an

form a monotone increasing se-

quence
35

with sum an :

Let Qm = b0 + ::: + bm

be the m th partial sum of a

rearrangement
b0+::: + bm+ : : : ::

of the terms of a0 + ::: + an + :::

Since the terms of the two se-

ries

are the same in di¤erent or-

ders:
36

for all m there exists an n such

that Qm Pn;

for all n there exists an m such

that Pn Qm

Therefore the least upper bound

of the partial sums Pn

is the least upper bound

of the partial sums Qm:

37

This implies the series have the

same sum.

2), 3) have similar proofs.

In the case of 3), the sum is

written i;j aibj .

These identities extend stan-

dard rules

for …nite sums and products

of numbers
38

to …nite sums and products

of convergent non-negative se-

ries.

4) follows from the triangle in-

equality

since jai bij jaij + jbij implies that

n jai bij n ja j + jb j
i=1 i=1 i i

= n ja j + n jb j A + B:
i=1 i i=1 i

so the series
39

(a 0 b0) + ::: + (an bn) : : :

is absolutely convergent.

Let a0 + ::: + an : : : be an

absolutely convergent series.

De…ne non-negative series

a+
0 + :::a + + :::;
n

a0 +:::an +:::

by
40

a+
i = max(0; ai) , ai = min(ai; 0) .

So for all i, ai = a+
i ai .

By assumption 1 ja j
i=0 i < 1; so
1
i=0 a+
i < 1;
1
i=0 ai < 1:

Combining these results,

Theorem: a0 + :::an + ::: is ab-

solutely convergent

i¤ both a+ +
0 + :::an + :::; and a0 +:::an +:::
41

are absolutely convergent se-

ries.

Moreover, ai = a+
i ai ,

and any rearrangement

of terms of a0 + :::an + :::

has the sum ai :

Proof: We justify only the last

remark.
42

Rearranging all terms in

a0+:::an+:::

rearranges all terms in

a+ +
0 + :::an + :::

and all terms in

a0 + :::an + ::::

Those are non-negative conver-

gent

series whose sums are the same

43

no matter what the rearrange-

ment is.

But ai = a+
i ai :

Riemann

Reordering Theorem

Suppose that a0 + :::an + ::: is con-

vergent

but not absolutely convergent.

44

Then for any real r,

we can reorder the terms in the

series

to obtain a series converging

to r:

PROOF: We can assume

all terms are non-zero.

If there are only a …nite num-

ber
45

of negative terms

or only a …nite number

of positive terms,

the sequence is absolutely

convergent (exercise).

So there are in…nitely many

negative terms and in…nitely

46

many positive terms. Let pn be

the n th positive term, let qn

be

the n th negative term.

If either p0 + p1 + :::or q0 + q1 + ::: con-

verge,

then the convergence of a0+:::an+

:::

implies the other converges,

47

in which case a0 + :::an + ::: is the

di¤erence

of two absolutely convergent se-

ries

and is absolutely convergent.

Stage 1. Let m0 be least such

that
p0+ : : : + pm0 > r

Stage 2. Let n0 be least such

that
48

p0+ : : : + pm0 q 0 ::: q n0 < r

Stage 3 Let m1 > m0 be least

such that
p0+ : : : + pm0 q 0 ::: q n0 +pm0+1+::: + pm1 > r

Stage 4 Let n1 > n0 be least such

that
(p0+ : : : + pm0 ) (q 0 + : : : + q n0 )

+(pm0+1+::: + pm1 ) (q n0+1 + : : : + q n1 )< r

..
49

The successive intervals around

r

shrink to zero because

a0+:::an+::: converges implies limn !1 an =
0 ,

so limn !1 pn = 0 and limn !1 qn = 0:

We have de…ned a rearrange-

ment of
a0+:::an+:::

which converges to r.
50

Power Series
A power series in (z z0 ) is a
series

of the form
a0 + 1 a (z
n=1 n z0 )n

also written 1 a (z
n=0 n z0 )n :

De…nition: lim supfbng is by de…ni-

tion
51

the largest limit point of the (possi-

bly

non-convergent) sequence fbng:

q
Let = lim supn !1 n
janj .

Suppose 6= 0 and r = 1:

Then r is called the

radius of convergence

of the series around center z0.

52

Exercise 2: If x is a real num-

ber

and T is a non-empty compact

subset

of the reals R , then there ex-

ists

a p in T closest to x among
elements of T .
:

Theorem:
53

1. If jz z0j < r; the power series

converges absolutely,

that is, 1 ja (z
n=0 n z0 )n j < 1 .

2. The power series converges

uniformly

on every compact subset T of

[z 2 R j jz z0 j < r ]

3. If jz z0j > r, the power series

diverges.
54

Proof of 1,2: Apply the root

test.
q
jz z0 j
lim supn !1 n
jan(z z0 )n j = r

Proof of 3: :By the exercise

let p

be an element of T closest to z0

So for all z2T , jz z0 j jp z0j < r:

This implies that for all z 2 T;

jan(z z 0 )n j jan(p z 0)nj:

55

Apply the Weierstrass M -test.

Taylor’s Theorem

(Lagrange’s form)

Suppose the (k + 1) th derivative

f (k+1)

of f :R !R exists at each point

of

the closed interval I with end-

points z0 and z:
56

Then there exists a point p

between z0 and z such that

f (z ) = f (z0) + f 0(z0)(z z0)+

f (k) (z0 )
::: + k! (z z0 )k + Rk (z z0 );

f (k+1)(p)
where Rk ( z z0 ) = (k+1)!
(z z0)k+1

Corollary: If in addition jf (k+1)(z)j

M for every z 2 I; then
jRk (z z 0 )j M jz z 0jk+1
(k+1)!
57

and therefore limz Rk (z z0 )

!z0 (z z )k
0
=0

Example: z0 = 0; f (z) = ez;

f (z ) = 1 + z + z2 + z3 z k
+ : : : k! + Rk (z )
2 6

where p xk+1
Rk (z ) = e(k+1)!

for some p between 0 and z: