You are on page 1of 76

Lebanese University

Lebanese University Faculty of Sciences

Faculty of Sciences

M1104
Analysis II
Department of Mathematics

Spring Semester
2019 - 2020

©opyright Reserved
i
Integral Calculus and Series
LEBANESE UNIVERSITY
February 27, 2020

ii
CONTENTS

Introduction v

1 Indefinite Integral 2
1.1 Primitive, Antiderivative and Indefinite Integral. . . . . . . . . . . . . . . . . . . . . 2
1.2 Table of Elementary Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Integration by Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Integration by Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Integration of Rational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Integrating of rational function F in some usual
Z functions . . . . . . . . . . . . . . . 11
1.6.1 Rational function in sine and cosine F (sin x, cos x) dx. . . . . . . . . . . . . 11
Z
1.6.2 Rational function in sine hyperbolic and cosine hyperbolic F (sinh x, cosh x) dx. 12

Z p
1.6.3 2
Rational function in x and ax + bx + c i.e F (x, ax2 + bx + c) dx. . . . 12
1.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 The Riemann Integral 15


2.1 The Riemann Integral. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Properties Of The Riemann Integral. . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Integrable Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Mean Value And Fundamental Theorems Of Calculus. . . . . . . . . . . . . . . . . . 22
2.5 Integration By Parts And Substitution Methods. . . . . . . . . . . . . . . . . . . . . 25
2.6 Riemann’s Original Definition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7 Schwarz Inequality For Riemann Integrals. . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3 Improper Riemann Integrals 31


3.1 Definitions And Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Basic Properties Of Improper Integrals. . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Convergence Tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Schwarz Inequality For Improper Integrals. . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Integration Of Comparison Relations. . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 Series 49
4.1 Numerical Series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Alternating Series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

iii
4.3 Ratio And Root Test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Abel’s Summation By Parts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.5 Rearrangements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.6 Cauchy Product. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

iv
INTRODUCTION

v
vi
1
CHAPTER 1
INDEFINITE INTEGRAL

Integration is the inverse process of differentiation. In differentiation we are given a function


and we are required to find its derivative, while in integration the derivative of some function is
given and we are required to find that function. For example F 0 (x) = 2x, can we discover what the
function F (x) is? One function with this property is F (x) = x2 , or more generally F (x) = x2 + c
where c is any constant.
In this chapter, I and J unless otherwise stated will denote non empty intervals of R.
1.1 Primitive, Antiderivative and Indefinite Integral.

1.1 Definition. Let F, f : I →7 R be two functions defined on an interval I of R. We say that F


is a primitive of f on I if F is differentiable on I and F 0 (x) = f (x) for all x ∈ I.

1.2 Remark. If I is any subset of R, not necessarily interval, and F 0 (x) = f (x) for all x ∈ I,
then F is said to be an antiderivative of f on I.

1.3 Example. Let F (x) = ln |x| and f (x) = x1 . Then; (a)F is a primitive of f on ]0, ∞[, (b)F is
a primitive of f on ] − ∞, 0[, (c)F is an antiderivative of f on ] − ∞, 0[∪]0, ∞[.

1.4 Remark. If F (x) is a primitive of f on I, then for any constant c ∈ R, we have F (x) + c is
also a primitive of f on I.

Proof. Since (F (x) + c)0 = (F (x))0 = f (x) for all x ∈ I, then F (x) + c is a primitive of f on I.

1.5 Property. If F and G are two primitives of f on I, then there exists c ∈ R such that F (x) −
G(x) = c for all x ∈ I.

Proof. We have F 0 (x) = G0 (x) = f (x) for all x ∈ I. Then (F (x) − G(x))0 = 0 for all x ∈ I. Since
I is an interval, then there exists c ∈ R such that F (x) − G(x) = c.

From this property one can deduce that; If F is a primitive of f on I, then the set of all
primitives of f on I is {F (x) + c : c ∈ R}.

1.6 Remark. The difference between two antiderivatives is not necessarily a constant. F1 (x) =
arctan x and F2 (x) = − arctan x1 are two antiderivatives of f (x) = 1+x
1 ∗
2 on R , but F1 (x) =
π π
F2 (x) + 2 if x > 0 and F1 (x) = F2 (x) − 2 if x < 0.

2
1.7 Definition. The set {F (x) + c : c ∈ R} of all primitivesZ of f on I is said to be an indefinite
integral of the function f w.r.t x on I and it is denoted by f (x) dx and we weite
Z
f (x) dx = F (x) + c.

Z
The symbol is the integral sign, f (x) is called the integrand, x stands for variable of inte-
Z
gration, c is called constant of integration and f (x) dx is called integral (indefinite integral) of f
w.r.t x. The presence of indefinite constant c justifies the name indefinite integral.

1.8 Property. Let λ, µ ∈ R. If f and g admit primitives F and G on I respectively, then λf + µg


admits a primitive λF + µG on I and we write
Z Z Z
[λf (x) + µg(x)] dx = λ f (x) dx + µ g(x) dx.

Z Z
Proof. Since [λ f (x) dx + µ g(x) dx]0 = [λF (x) + µG(x)]0 = λf (x) + µg(x) for all x ∈ I, then
Z Z Z
[λf (x) + µg(x)] dx = λ f (x) dx + µ g(x) dx.

This
Z property can be generalized to the form: Z Z Z
[λ1 f1 (x) ± λ2 f2 (x) ± · · · ± λn fn (x)] dx = λ1 f1 (x) dx ± λ2 f2 (x) dx ± · · · ± λn fn (x) dx.

1.9 Theorem. If f : I 7→ R is a continuous function, then f admits a primitive on I.

Proof. We will see the proof in the next chapter.

Note that,
 there are non continuous functions which admit primitives. For example
x2 sin x1 , if x =6 0;
F (x) = is differentiable on R but it is not of class C 1 .
0, If x = 0.
Also note that, even if a primitive for some function f exists, it may not be possible to find a
2
primitive for f explicitly. For example f (x) = ex .

1.2 Table of Elementary Integrals


Based on differentiation and definition of primitives, the table of derivatives of the main elementary
functions can be read backwards as a list of primitives. The student is strongly advised to commit
these results to memory, because no further progress is otherwise possible.

3
Table of primitives of: usual functions(1 − 9), fractional rational functions (10 − 11) and
fractional irrational functions (12 − 14). Here a ∈ R∗ and the case a = 1 is considerable.

Z
f (x) dx = F (x) conditions on x I
Z Z
1
1) x−1 dx = dx = ln |x| x 6= 0 ] − ∞, 0[ or ]0, ∞[
x
Z α+1
x
2) xα dx =
α+1
i)α ∈ R \ {−1} x>0 ]0, ∞[
ii)α ∈ N ··· R
Z iii)α ∈ {−2, −3, · · · , } x 6= 0 ] − ∞, 0[ or ]0, ∞[
3) cos x dx = sin x ··· R
Z
4) sin x dx = − cos x ··· R
Z Z
1
5) 2
dx = (1 + tan2 x) dx = tan x x ∈ R \ { π2 + kπ; k ∈ Z} ]− π
2
+ kπ, π
2
+ kπ[, k ∈ Z
Z cos x
6) ex dx = ex ··· R
Z
7) cosh x dx = sinh x ··· R
Z
8) sinh x dx = cosh x ··· R
Z Z
1
9) 2 dx = (1 − tanh2 x) dx = tanh x ··· R
Z cosh x
dx 1 x
10) 2 2
= arctan ··· R
Z x +a a a
dx 1 x − a
11) 2 − a2
= ln x 6= ±a ]∞ , −|a|[ or ] − |a|, |a|[ or ]|a|, ∞[
x 2a x+a

Z
dx x
12) √ = arcsin |x| < |a| ] − |a|, |a|[
Z a2 − x2 |a|
dx x p
13) √ = arg sinh = ln(x + a2 + x2 ) ··· R
Z 2
a +x 2 a
dx p
14) √ = ln |x + x2 − a2 | |x| > |a| ]∞ , −|a|[ or ]|a|, ∞[
x2 − a2

1.3 Integration by Substitution

1.10 Theorem. Assume f : I 7→ R admits a primitive ( continuous ) on I and u : J 7→ R is


differentiable on J such that u(J)Z ⊆ I.
If the variable t in the integral f (u(t))u0 (t) dt be changed to x by putting u(t) = x, then
Z Z
0
f (u(t))u (t) dt = f (x) dx .

x=u(t)

If in addition u is strictly monotone, then


Z Z
f (x) dx = f (u(t))u0 (t) dt .

t=u−1 (x)

Z Z
dh i
Proof. Let F be a primitive of f on I. Then f (x) dx = F (u(t)). Hence f (x) dx|x=u(t) =

Z x=u(t) Z dt
d 0 0 0 0

F (u(t)) = F (u(t))u (t) = f (u(t))u (t). Thus f (u(t))u (t) dt = f (x) dx . If in addition

dt x=u(t)

4
u is strictly monotone, then
Z hZ i Z
f (u(t))u0 (t) dt −1 = f (x) dx = F (u(t))|t=u−1 (x) = F (x) = f (x) dx.

t=u (x) x=u(t) t=u−1 (x)

1.11 Examples. To calculate


Z
sin(ln t)
1. dt, we put ln t = x and 1t dt = dx. Hence
t
Z Z
sin(ln t)
dt = sin x dx = − cos x + c = − cos(ln t) + c.
t
Z
2t + 5
2. dt, we put t2 + 5t + 3 = x and (2t + 5)dt = dx. Hence
t2 + 5t + 3
Z Z
2t + 5 1
2
dt = dx = ln |x| + c = ln |t2 + 5t + 3| + c.
t + 5t + 3 x
Z
3. (6t + 15)(t2 + 5t + 3) dt, we put t2 + 5t + 3 = x and (2t + 5)dt = dx. Hence

3x2 3(t2 + 5t + 3)2


Z Z Z
2 2
(6t+15)(t +5t+3) dt = 3 (2t+5)(t +5t+3) dt = 3 x dx = +c = +c.
2 2
Z
2
4. te1−t dt, we put 1 − t2 = x and −2tdt = dx. Hence

−1 −1 −1 x −1 1−t2
Z Z Z
1−t2 1−t2
te dt = −2te dt = ex dx = e +c= e + c.
2 2 2 2
Z
5. sin3 t cos t dt, we put sin t = x and cos t dt = dx. Hence

x4 sin4 t
Z Z
3
sin t cos t dt = x3 dx = +c= + c.
4 4


Z p
6. 1 − x2 dx, we put x = sin t = u(t) and dx = cos tdt. We have f (x) = 1 − x2 is contin-
uous on I = [−1, 1] and u is differentiable and strictly monotone on J = [− π2 , π2 ] such that
Z p = [−1, 1] ⊆ I.
u(J) Z pHence Z Z Z
2 2 2 1
1 − x dx = 1 − sin t cos t dt = | cos t| cos t dt = cos t dt = (1+cos(2t)) dt =
2
1 1 1 1 p
[t + sin(2t)] + c = [t + sin t cos t] + c = [arcsin x + x 1 − x2 ] + c.
2 2 2 2
Z
1
7. √ dx, we put x = tan t = u(t) and dx = (1 + tan2 t)dt. We have f (x) =
(1 + x ) 1 + x2
2
1√
(1+x2 ) 1+x2
is continuous on I = R and u is differentiable and strictly monotone on J =
] − π2 , π2 [ such that u(J) = R ⊆ I. Hence
Z √
1 + tan2 t
Z Z Z
1 1
√ dx = √ dt = √ dt = cos2 t dt =
(1 + x 2 ) 1 + x2 (1 + tan 2 t) 1 + tan2 t 1 + tan 2t
Z Z
x
| cos t| dt = cos t dt = sin t + c = sin(arctan x) + c = · · · = √ + c.
1 + x2

5
1.4 Integration by Parts

1.12 Theorem. Let u and v be two differentiable functions over I. If u0 v admits a primitive on
I, then so is uv 0 and Z Z
u(x)v (x) dx = u(x)v(x) − u0 (x)v(x) dx.
0

Proof. Let H be any primitive of u0 v on I. Then

[u(x)v(x) − H(x)]0 = u0 (x)v(x) + u(x)v 0 (x) − u0 (x)v(x) = u(x)v 0 (x).


Z Z
Thus u(x)v 0 (x) dx = u(x)v(x) − u0 (x)v(x) dx.

We can apply the integration by parts formula if one of the functions is differentiable and the other
one is of class C 1 on I.
How to chose u and v 0 :

• Whatever you let v 0 be, you need to be able to find v.

• It helps if u0 v is simpler than uv 0 .

1.13 Examples. To calculate


Z
1. ln x dx, we put u(x) = ln x and v 0 (x) = 1. Then u0 (x) = 1
x and v(x) = x. Hence

Z Z
1
ln x dx = x ln x − x dx = x ln x − x + c.
x

Z
x3
2. x2 ln x dx, we put u(x) = ln x and v 0 (x) = x2 . Then u0 (x) = 1
x and v(x) = 3 . Hence

x3 1 x3 x3 x3
Z Z
x2 ln x dx = ln x − = ln x − + c.
3 x 3 3 9

Z
3. xex dx, we put u(x) = x and v 0 (x) = ex . Then u0 (x) = 1 and v(x) = ex . Hence

Z Z
x x
xe dx = xe − ex dx = xex − ex + c.

Tabular Method: If f (x) is n + 1 times differentiable (polynomial of degree n) and g(x) is


one of the following functions eax , sin(ax), cos(ax), sinh(ax), cosh(ax) where a ∈ R∗ or any function
has known n + 1 successive primitives g1 (x), · · · , gn+1 (x), then
Z Z
0 n n n+1
f (x)g(x) dx = f (x)g1 (x)−f (x)g2 (x)+· · ·+(−1) f (x)gn+1 (x)+(−1) f n+1 (x)gn+1 (x) dx.

6
1.5 Integration of Rational Functions
N (x)
If the integrand f (x) = is a rational function whose numerator N (x) and denominator D(x)
D(x)
are polynomials and degree of numerator is greater than or equal degree of denominator, then
divide the numerator by the denominator and then use the rule:
Numerator Remainder N (x) R(x)
= Quotient + ⇔ = Q(x) +
Denominator Denominator D(x) D(x)
with R(x) a polynomial of degree less than degree of D(x). Therefore
Z Z Z
N (x) R(x)
dx = Q(x) dx + dx.
D(x) D(x)
R(x)
The problem boils down to integrating a rational map in which the numerator’s degree is
D(x)
less than the denominator’s.
4x2 + 4x + 1 25
1.14 Example. = (4x + 12) + .
x−2 x−2
R(x)
From Algebra, can be decomposed uniquely as linear combinations of simple elements of
D(x)
1 ax + b
the form n
and/or 2 with ∆ = p2 − 4q < 0. We discuss the possible situations
(x − a) (x + px + q)n
which will turn out to be fundamental for treating the generic integrand.
1
(a) Primitive of simple elements of first kind: .
Z (x − a)n
1
If n = 1, then dx = ln |x − a| + c.
Z (x − a) Z
1 1
If n ≥ 2, then n
dx = (x − a)−n d(x − a) = + c.
(x − a) (1 − n)(x − a)n−1
ax + b
(b) Primitive of simple elements of second kind: with ∆ = p2 − 4q < 0.
(x2 + px + q)n
ax + b a 2x + p b − ap
2
Note that 2 = + . For the first fraction
(x + px + q)n 2 (x2 + px + q)n( (x2 + px + q)n
d(x2 + px + q) ln(x2 + px + q) + c Ifn = 1,
Z Z
2x + p
dx = = (x2 +px+q)−n+1 For the second
(x2 + px + q)n (x2 + px + q)n −n+1 +c ifn ≥ 2.
Z
1
one, let In = dx. Note that
(x + px + q)n
2

p p2 p2 p 4q − p2
x2 + px + q = x2 + 2 x + − + q = (x + )2 + = t 2 + h2 ,
2 4 4 2 4
r
4q − p2
Z
p 1
where t = x + and h = . Hence In = dt. We can evaluate In by two
2 4 (t2 + h2 )n
ways:
First way: we put t = h tan θ and dt = h(1 + tan2 θ)dθ. Then
Z Z
1 1
In = 2n−1 dθ = cos2n−2 θ dθ.
h (1 + tan2 θ)n−1
x+ p
Z
t
We compute cos2n−2 θ dθ and we replace θ by arctan = arctan q 2 .
h 4q−p2
4
Second way: We compute I1 , then we find a recurrent relation between In and In+1 for
n ≥ 2.

7
x + p2
Z
dt 1 t 1
– I1 = = arctan + c = arctan + c.
t2 + h2
q q
h h 4q−p2 4q−p2
4 4
– To find a relation between In and In+1 , we integrate In by parts.

Z
1 1
In = dt ( put u = 2 and v 0 = 1)
(t2
+h )2 n (t + h2 )n
t2
Z
t
= 2 + 2n dt
(t + h2 )n (t2 + h2 )n+1
Z 2
t t + h2 − h2
= 2 + 2n dt
(t + h2 )n (t2 + h2 )n+1
t 2

= 2 + 2n In − h In+1 .
(t + h2 )n
1 h t i p
Hence, In+1 = 2
(2n − 1)I n + 2 2 n
. Finally, replace t by x + and h by
r 2nh (t + h ) 2
4q − p2
.
4
Here is a version of the Fundamental Theorem of Algebra and the partial fraction decomposition.

1.15 Theorem.

1. A polynomial D(x) of degree d with real coefficients decomposes uniquely as a product

D(x) = α(x − α1 )r1 · · · (x − αm )rm (x2 + p1 x + q1 )s1 · · · (x2 + pn x + qn )sn

where α, αi , pj , qj are real and ri , sj are integers such that r1 + · · · + rm + 2s1 + · · · + 2sn = d.
The real roots αi of D(x) are distinct and counted with multiplicity ri . The factors x2 +pj x+qj
are pairwise distinct and irreducible over R i.e. ∆ = p2j − 4qj < 0.

R(x) R(x)
2. = can be decomposed
D(x) α(x − α1 ) · · · (x − αm ) (x + p1 x + q1 )s1 · · · (x2 + pn x + qn )sn
r1 r m 2
uniquely as sum of partial fractions

R(x) 1 
= L1 (x) + · · · + Lm (x) + Q1 (x) + · · · + Qn (x) (1.1)
D(x) α

where each Li takes the form


Ai1 Ai2 Airi
Li (x) = + + ··· +
(x − αi ) (x − αi )2 (x − αi )ri

and each Qj takes the form

Bj1 x + Cj1 Bj2 x + Cj2 Bjs x + Cjsj


Qj (x) = + 2 + ··· + 2 j
2
(x + pj x + qj ) (x + pj x + qj ) 2 (x + pj x + qj )sj

Proof. admitted.

To recover the undetermined coefficients Ai` , Bjµ and Cjµ we multiply equation (1.1) by D(x)
both sides and we take x = αi to obtain Airi , for the other coefficients we use the following rules:
Two polynomials of degree d coincide if and only if one of the following conditions holds

8
• The coefficients of corresponding monomials coincide.

• The polynomials assume the same values at d + 1 distinct points.

Once these coefficients have been determined, we can start integrating the right-hand side of (1.1)
and rely on the fundamental cases above. We illustrate this technique by a few examples.
1.16 Examples.
2x3 + x2 − 4x + 7
Z
1. (Non repeated linear factors in D(x)) Let us compute dx.
x2 + x − 2
The numerator has greater degree than the denominator, so by Euclidean division we obtain

2x3 + x2 − 4x + 7 x+5
= (2x − 1) + 2 .
x2 + x − 2 x +x−2

The denominator factorizes as x2 + x − 2 = (x − 1)(x + 2) non repeated linear factors, hence

x+5 x+5 a b
= = + .
x2 +x−2 (x − 1)(x + 2) x−1 x+2

That is,
x + 5 = a(x + 2) + b(x − 1). (1.2)
To find the coefficients a and b, we compute (1.2) at the zero’s x = 1, x = −2 of D(x). For
x = 1 we obtain a = 2 and for x = −2 we obtain b = −1.
A general way to find the coefficients, we compare the coefficients of corresponding monomials.
Equation (1.2) can be written

x + 5 = (a + b)x + (2a − b).

Comparing coefficients yields the linear system



a + b = 1,
2a − b = 5,

solved by a = 2 and b = −1. Therefore

2x3 + x2 − 4x + 7
Z Z h
2 1 i
dx = (2x−1)+ − dx = x2 −x+2 ln |x−1|−ln |x+2|+c.
x2 + x − 2 x−1 x+2

x2 + 2x
2. (Repeated linear factors in D(x)) Let us determine a primitive of f (x) = .
(x − 1)3 (x − 2)2
The numerator has less degree than the denominator. We must search for the reals a, b, c, d
and e such that
x2 + 2x a b c d e
3 2
= + 2
+ 3
+ + ,
(x − 1) (x − 2) x − 1 (x − 1) (x − 1) x − 2 (x − 2)2
or,

x2 + 2x = a(x − 1)2 (x − 2)2 + b(x − 1)(x − 2)2 + c(x − 2)2 + d(x − 1)3 (x − 2) + e(x − 1)3 . (1.3)

Now, x = 1 implies c = 3 and x = 2 implies e = 8. To find a, b and d we take three values


for x 6= 1, 2. For instance x = 0, 3, −1 gives the linear system

 4a − 4b + 2d = −4,
4a + 2b + 8d = −52,
36a − 18b + 24d = 36,

9
solved by a = 18, b = 10 and d = −18. Therefore

−18
Z Z h
18 10 3 8 i
f (x) dx = + + + + dx
x − 1 (x − 1)2 (x − 1)3 x − 2 (x − 2)2
10 3 8
= 18 ln |x − 1| − − 2
− 18 ln |x − 2| − + c.
(x − 1) 2(x − 1) (x − 2)

Another possibility to find a, b and d, we compare the coefficients of corresponding monomials


in (1.3) to obtain a linear system in three equations with three unknowns.

3. (Irreducible quadratic factors in D(x))Let us determine the partial fraction decomposi-


x4 + 2x2
tion of 2 .
(x + x + 1)3
The numerator has less degree than the denominator. We must search for the reals a, b, c, d, e
and f such that

x4 + 2x2 ax + b cx + d ex + f
= 2 + + .
(x2 + x + 1)3 (x + x + 1) (x2 + x + 1)2 (x2 + x + 1)3

To find the reals a, b, c, d, e and f , we take six different values for x to obtain a linear system in
six equations with six unknowns, or by comparing the coefficients of corresponding monomials.

4. (Mixed factors in D(x))Let us determine the partial fraction decomposition of

5x2
.
(x + 1)(x − 2)(x + 3)2 (x2 + 4)2

The numerator has less degree than the denominator. We must search for the reals a, b, c, d, e, f, g
and h such that

5x2 a b c d ex + f gx + h
2 2 2 = x + 1 + x − 2 + x + 3 + (x + 3)2 + x2 + 4 + (x2 + 4)2 ,
(x + 1) (x − 2) (x + 3) (x + 4)
| {z } | {z } | {z } | {z }
non rep non rep repeated quadratic

or,

5x2 = a(x − 2)(x + 3)2 (x2 + 4)2 + b(x + 1)(x + 3)2 (x2 + 4)2
+ c(x + 1)(x − 2)(x + 3)(x2 + 4)2 + d(x + 1)(x − 2)(x2 + 4)2
+ (ex + f )(x + 1)(x − 2)(x + 3)2 (x2 + 4) + (gx + h)(x + 1)(x − 2)(x + 3)2 . (1.4)

If we compute (1.4) at x = −1, 2, −3, we obtain a, b and d. The other coefficients c, e, f, g and
h can be determined by comparing the coefficients of corresponding monomials or by taking
five different values of x other than −1, 2, −3.

Note that many functions that are not rational can be transformed into a rational function in
a new variable. The table below included some special cases: Let F be a rational function in many
variables,

10
Z
F dx new variable t
Z
m r
F (x, x n , · · · , x s ) dx x = tk , k = LCM {n, · · · , s}
Z  ax + b  m  ax + b  r  
n s ax+b k
F (x, ,··· , ) dx cx+d = t , k = LCM {n, · · · , s} i.e
cx + d cx + d
m
Z
4
(a, b, c, d) ∈ R with ad − bc 6= 0 k is a common denominator of n ,··· , rs
F (eax ) dx, a ∈ R∗ (even F is not rational) t = eax , x = 1
a ln t and dx 1
= at dt
Z
F (sin x, cos x) dx t = tan x2 , x = 2 arctan t and dx = 2
1+t2
dt
2t 1−t2
Z sin x = 1+t2
and cos x = 1+t2

F (sinh x, cosh x) dx t = tanh x2 , x = 2 arg tanh t and dx = 2


1−t2
dt
2t 1+t2
Z sinh x = 1−t2
and cosh x = 1−t2

F (sin2 x, cos2 x, tan x) dx t = tan x, x = arctan t and dx 1


= 1+t 2 dt

t2 1
Z sin2 x = 1+t2
and cos x = 1+t2
p
F (x, ax2 + bx + c) dx, a 6= 0 and b2 − 4ac 6= 0
√ √
First case a > 0 ax2 + bx + c = x a + t
√ p
Second case a < 0 then b2 − 4ac > 0, roots: α, β ax2 + bx + c = a(x − α)(x − β) = t(x − α)

1.6 Integrating of rational function F in some usual functions


Z
1.6.1 Rational function in sine and cosine F (sin x, cos x) dx.

(1) Particular case. If F is a polynomial, then by linearity of integral it is sufficient to know


how to integrate Z
cosm x sinn x dx.

If m is odd, use the substitution t = sin x.


If n is odd, use the substitution t = cos x.
If m, n ∈ Q and m + n is a negative even integer, use the substitution t = tan x.
If m and n are even, or one is even and the other is zero, we transform cosm x sinn x into
linear combinations of sin(αx) and cos(βx) by using Euler’s formulae
eix + e−ix eix − e−ix
cos x = and sin x = ,
2 2i
or by using repeatedly the following formulae

(i) cos(2a) = 2 cos2 a − 1. (ii) cos(2a) = 1 − 2 sin2 a.


(iii) cos(3a) = 4 cos3 a − 3 cos a. (iv) sin(3a) = 3 sin a − 4 sin3 a.
1
(vi) sin a cos a = 21 sin(2a).

(v) sin a cos b = 2 sin(a + b) + sin(a − b) .
1
cos(a + b) + cos(a − b) . (viii) sin a sin b = 12 cos(a − b) − cos(a + b) .
 
(vii) cos a cos b = 2

(2) Bioche rule. Let W (x) = F (sin x, cos x) dx.


If W (−x) = W (x), use the substitution t = cos x.
If W (π − x) = W (x), use the substitution t = sin x.
If W (π + x) = W (x), use the substitution t = tan x.

11
Z
x
(3) General case. Always the change of variable t = tan 2 transform F (sin x, cos x) dx into
integral of rational function in t (see previous table).
Z
1.6.2 Rational function in sine hyperbolic and cosine hyperbolic F (sinh x, cosh x) dx.

(1) Bioche rule. Let W (x) = F (sin x, cos x) dx.


If W (−x) = W (x), use the substitution t = cosh x.
If W (π − x) = W (x), use the substitution t = sinh x.
If W (π + x) = W (x), use the substitution t = tanh x.

Z
x
(2) General case. Always the change of variable t = tanh 2 or t = ex transform F (sinh x, cosh x) dx
into integral of rational function in t (see previous table).


Z √
1.6.3 Rational function in x and ax2 + bx + c i.e F (x, ax2 + bx + c) dx.

If a 6= 0 and ∆ = b2 − 4ac 6= 0, by completing square√ method√ the the term 2
√ ax + bx + c can be
transformed
√ to one of the following general forms α − t , α + t or t − α2 .
2 2 2 2 2

For √α2 − t2 , use the substitution t = α sin θ or t = α cos θ or t = α tanh θ.


For t2 − α2 , use the substitution t = α cosh θ where  = sign (αt) or t = α cos1 θ .

For α2 + t2 , use the substitution t = α sinh θ or t = α tan θ.

12
1.7 Exercises
Exercise 1. Compute the indefinite integrals:
ex
Z Z Z Z
1 1
1) dx, 2) dx, 3) dx, 4) tan x dx,
ex + 1 ex + 1 (1 + x2 ) arctan x

Z Z Z Z
1
5) tan2 x dx, 6) tan4 x dx, 7) dx, 8) tan3 x dx,
tan4 x

sin4 x
Z Z Z Z
x 1 1
9) dx, 10) √ dx, 11) dx, 12) dx.
cos6 x 1 + x2 x ln |x| x ln x[ln(| ln x|)]
Exercise 2. Compute the indefinite integrals:
Z Z Z Z
1) arctan x dx, 2) x arctan x dx, 3) (1 + x2 )e−2x dx, 4) (1 − x3 ) cos(−2x) dx,

Z Z Z Z
5) (x − 1) 1 − x dx, 6) (x + x ) ln x dx, 7) ex sin x dx,
2 10 4
8) ex cos x dx,

Z Z Z Z
x cos x
9) xe sin x dx, 10) x 3
dx, 11) sin x dx, 12) sin(ln x) dx,
Z Z sin x Z Z
3
13) sin 2x cos 3x dx, 14) x sin x dx, 15) arccos3 x dx, 16) sin(2x)(ln(1 + cos2 x))2 dx.

Exercise 3. Assume that n ≥ 2 is an integer.


1. Prove the following reduction formulae:
n−1
Z Z
n 1 n−1
(a) sin x dx = − sin x cos x + sinn−2 x dx.
n n
n−1
Z Z
n 1 n−1
(b) cos x dx = cos x sin x + cosn−2 x dx.
n n
Z Z
n 1 n−1
(c) tan x dx = tan x − tann−2 x dx.
n−1
−1
Z Z
n n−1
(d) cot x dx = cot x − cotn−2 x dx.
n−1
n−2
Z Z
1
(e) secn x dx = secn−2 x tan x + secn−2 x dx.
n−1 n−1
−1 n−2
Z Z
(f) cscn x dx = cscn−2 x cot x + cscn−2 x dx.
n−1 n−1
You may deduce (b), (d), and (f) from (a), (c) and (e) respectively.
Z Z Z Z Z Z
4 4 4 4 4
2. Evaluate: sin x dx, cos x dx, tan x dx, cot x dx, sec x dx and csc4 x dx.
Z
dx
Exercise 4. Let n ∈ N and In = .
(x2
+ 1)n
Find a relation between In+1 and In , then evaluate I2 . Z
cos x
Exercise 5. Let a, b, c and d be non zero real numbers. Assume that I = dx
Z c cos x + d sin x
sin x
and J = dx
c cos x + d sin x
1. Calculate cI + dJ and dI − cJ, then deduce the values of I and J.
Z
a cos x + b sin x
2. Deduce dx.
c cos x + d sin x

13
Z Z
Exercise 6. Consider the integrals I = sin x sinh x dx and J = cos x cosh x dx

1. Find two relation between I and J , then deduce the values of I and J.
Z
2. Deduce sin2 x sinh2 x dx.

Exercise 7. Give the general form of the partial fraction decomposition for the following rational
functions.

x5 + 3x 1+x 1+x
1. , 2. , 3. .
(x2 − 2x)(x + 1)2 (x3 + 1)3 (x4 + 1)2
x+2
Exercise 8. Let f (x) = .
(|x| + 3)(x − 3)
1. Justify the existence of a primitive of f on ] − ∞, 3[.
2. Give a primitive of f on ] − ∞, 3[.
3. What is the primitive that vanishes at x = 0.
Exercise 9. Let n ∈ N∗ . Compute the indefinite integrals:
x2 x−1
Z Z Z Z
1 1
1) dx, 2) dx, 3) dx, 4) dx,
(x + 2)(x2 − 1) (x − 1)2 (x2 + 1) x4 − 1 x3 + 1

x4 + 1
Z Z Z Z
1 1 x+1
5) 3
dx, 6) dx, 7) dx, 8) dx,
x −1 x4 − 1 4
x +1 x2 + 4x + 8

e2x + ex
Z Z Z Z
x+1 1 1
9) dx, 10) dx, 11) dx, 12) dx.
e2x + 4ex + 8 1 − 6x − 9x2 x4 + x2 + 1 x(xn + 1)
Exercise 10. Compute the indefinite integrals:
Z Z Z
1) sin5 x cos4 x dx, 2) sin4 x cos5 x dx, 3) sin3 x cos3 x dx,4) sin2 x cos2 x dx,
Z Z Z Z
1 1
5) sin2 x cos4 x dx, 6) sin2 x cos6 x dx, 7) 3 5 dx, 8) dx,
Z Z Z sin 2 x3cos 2 x Z sin x
1 1 sin x 1
9) dx, 10) 3
dx, 11) dx, 12) dx,
Z sin 2x Z cos x Z 2 + cos x Z cos x(sin x − cos x)
1 1 x sin x sinh3 x
13) dx, 14) dx, 15) 2
dx, 16) dx,
Z 2 + sin x Z 1 + cos x Z (1 + cos x) Z 2 + cosh 2x
tanh x sinh x 1 √ x
17) 3 3 dx, 18) x + 1)2
dx, 19) ex + 1 dx, 20) dx.
sinh x + cosh x (e (x sin x + cos x)2
Exercise 11. Compute the indefinite integrals:
Z p Z Z Z
3x + 1 x+1 1
1) x − x2 dx, 2) √ dx, 3) √ dx, 4) √ dx,
2
x +x+1 2
x − 3x + 2 x 1 − x2
Z Z
2x + 1 1
5) √ dx, 6) √ dx.
2
−2x + 4x + 8 (2x + 1) x2 + x + 1
Exercise 12. Compute the indefinite integrals:
Z √ √ Z r
x+1− 3x+1 x−1
Z Z
1 x
1) √ dx, 2) √ dx, 3) √ √ dx, 4) dx,
1+x 4 3
x +1
3
x+1+ x+1 x−2
Z r Z
1 1+x 1
5) dx, 6) √ √ dx.
x 1−x x+1+ x

14
CHAPTER 2
THE RIEMANN INTEGRAL

The Riemann idea of integration for a bounded function f over a closed bounded intervals [a, b]
is to assign a real number A, called the area to the region R bounded by the curve x = a, x = b, y = 0
and y = f (x), where we assume that f is non-negative. The number A, the area under the curve,
Z b
is called the integral of f over [a, b] and denoted by the symbol f (x) dx. In this chapter we will
a
answer the following questions:
1. How do Bernhard Riemann and Jean-Gaston Darboux define the area under the curve?

2. What conditions imposed on f to obtain it Riemann integrable over [a, b]?

3. What are the basic properties of Riemann integral?

4. How to calculate Riemann integral?

2.1 The Riemann Integral.

2.1 Definition. (Partitions, Upper And Lower Sums)


Let f be a bounded function on [a, b].

• A partition P of [a, b] is a set of points {xk }nk=0 such that a = x0 < x1 < x2 < · · · < xn = b.

• For each subinterval [xk−1 , xk ] of [a, b], set


mk = inf f (x), and Mk = sup f (x).
x∈[xk−1 ,xk ] x∈[xk−1 ,xk ]

n
P
– The lower Darboux sum of f with respect to P is the sum L(f ; P ) = mk (xk −xk−1 ).
k=1
n
P
– The upper Darboux sum of f with respect to P is U (f ; P ) = Mk (xk − xk−1 ).
k=1

Let m = inf f (x) and M = supx∈[a,b] f (x). It is clear that, for any partition P of [a, b] we have
x∈[a,b]

m(b − a) ≤ L(f ; P ) ≤ U (f ; P ) ≤ M (b − a).

Also, if f ≥ 0, then the lower and the upper sums approximate the area under the curve from below
and from above.
2.2 Examples.
1. P = {a = x0 , x1 = b} is the trivial partition of [a, b]

15
2. Pn = {xi = a + ni (b − a); 0 ≤ i ≤ n} is the partition of [a, b] that divides it into n subintervals
with equal lengths.

2.3 Definition. (Refinement)


A partition Q is a refinement of a partition p if Q contains all of the points of P . In this case
we write P ⊆ Q.

2.4 Property. 1. If P ⊆ Q, then L(f ; P ) ≤ L(f ; Q) and U (f ; P ) ≥ U (f ; Q).

2. L(f ; P ) ≤ U (f ; Q) for P, Q any partitions of [a, b].

Proof. 1. Let P = {x0 , x1 , · · · , xn }. Assume Q = {x0 , · · · , xk−1 , t, xk · · · , xn } is a refinement of


P by adding a single point t to some subinterval [xk−1 , xk ] of P . For m0k = inf f (x) and
x∈[xk−1 ,t]
m00k = inf f (x) we have
x∈[t,xk ]

mk (xk − xk−1 ) = mk (xk − t) + mk (t − xk−1 )


≤ m0k (xk − t) + m00k (t − xk−1 ).

Then by induction, we obtain L(f ; P ) ≤ L(f ; Q), and analogous argument holds for the other
inequality.

2. Let P, Q be any partitions of [a, b]. Then L(f ; P ) ≤ L(f ; P ∪ Q) ≤ U (f ; P ∪ Q) ≤ U (f ; Q) .

Note that L(f ; Q) and U (f ; Q) produces a better approximation to the area under the curve.
This suggests that to get the real area we look at
the upper integral U (f ) = inf{U (f ; P ); P is a partition of [a, b]} and
the lower integral L(f ) = sup{L(f ; P ); P is a partition of [a, b]}.

2.5 Corollary. If f is a bonded function on [a, b], then L(f ) ≤ U (f ).

Proof. Let Q be a partition of [a, b], then L(f ) ≤ U (f ; Q). It follows that L(f ) ≤ U (f ).

Property 2.4 suggests the following definition for integrability

2.6 Definition. (Riemann Integrability In Sense Of Darboux)


Let f : [a, b] 7→ R be a bounded function. We say that f is Riemann integrable on [a, b] and we
Z b Z b
write f ∈ R([a, b]) if L(f ) = U (f ). In this case we define f or f (x) dx to be this common
Z b a a

value, that is f (x) dx = L(f ) = U (f ).


a

16
2.7 Examples.
1. Let f : [a, b] →
7 R be a constant function with f (x) = c for all x. We claim that f ∈
Z b
R([a, b]) and f (x) dx = c(b − a). Let P = {x0 , x1 , · · · , xn } be a partition of [a, b], then
a
n
P n
P n
P
L(f ; P ) = mk (xk −xk−1 ) = c(xk −xk−1 ) = c(b−a) and U (f ; P ) = Mk (xk −xk−1 ) =
k=1 k=1 k=1
n
P
c(xk − xk−1 ) = c(b − a). Hence L(f ) = U (f ) = c(b − a). Thus f ∈ R([a, b]) and we have
k=1
Z b
f (x) dx = L(f ) = U (f ) = c(b − a).
a

2. Let f : [0, 1] 7→ R be defined by f (x) = 1 if 0 ≤ x < 1 and f (1) = 2. We claim that


Z 1
f ∈ R([0, 1]) and f (x) dx = 1. Let P = {x0 , x1 , · · · , xn } be a partition of [a, b], then
0
mk = 1 for all 1 ≤ k ≤ n and Mk = 1 for 1 ≤ k < n and Mn = 2. Hence L(f ; P ) =
Pn n
P n
P n−1
P
mk (xk − xk−1 ) = (xk − xk−1 ) = 1 and U (f ; P ) = Mk (xk − xk−1 ) = Mk (xk −
k=1 k=1 k=1 k=1
n−1
P
xk−1 ) + Mn (xn − xn−1 ) = (xk − xk−1 ) + 2(1 − xn−1 ) = xn−1 + 2(1 − xn−1 ). It follows that
k=1
Z 1
L(f ) = U (f ) = 1, and f ∈ R([0, 1]) and we have f (x) dx = 1.
0

3. Let f : [0, 1] 7→ R be defined by f (x) = 1 if x ∈ Q and zero otherwise. We claim that


f∈/ R([0, 1]). By the density of rationales and irrationals in R, we obtain mk = 0 and Mk = 1
for all 1 ≤ k ≤ n. It follows that L(f ; P ) = 0, U (f ; P ) = 1 and L(f ) = 0 6= U (f ) = 1. Thus
f is not integrable on [0, 1].

2.8 Theorem. (Criteria For Integrability)


A bonded function f is integrable on [a, b] if and only if for every ε > 0, there exists a partition Pε
such that U (f ; Pε ) − L(f ; Pε ) < ε.

Proof. Let ε > 0 and Pε be a partition of [a, b] such that U (f ; Pε )−L(f ; Pε ) < ε. Then U (f )−L(f ) ≤
U (f ; Pε ) − L(f ; Pε ) < ε. It follows U (f ) = L(f ) and f ∈ R([a, b]).
For the proof of the converse statement: Assume that f ∈ R([a, b]) and let ε > 0, then by definitions
of inf and sup there exist two partitions P1 and P2 of [a, b] such that
ε ε
U (f ; P1 ) < U (f ) + , and L(f ; P2 ) > L(f ) − .
2 2
Let Pε = P1 ∪ P2 be the common refinement and since L(f ) = U (f ), we obtain
ε ε
U (f ; Pε ) − L(f ; Pε ) ≤ U (f ; P1 ) − L(f ; P2 ) ≤ (U (f ; P1 ) − U (f )) + (L(f ) − L(f ; P2 )) < + = ε.
2 2

2.9 Remark. Refining a partition can bring the upper and lower sums closer together. That is for
Pε ⊆ Qε , we have U (f ; Qε ) − L(f ; Qε ) ≤ U (f ; Pε ) − L(f ; Pε ) < ε.

2.10 Corollary. (Sequential Criterion For Integrability)


A bonded function f is integrable on [a, b] if and only if there exists a sequence of partitions (Pn )
satisfying lim [U (f ; Pn ) − L(f ; Pn )] = 0.
n→∞

17
2.11 Property. (Sequential Criterion For Integrability)
A bonded function f is integrable on [a, b] if and only if there exists a sequence of partitions (Pn )
such that L(f ; Pn ) and U (f ; Pn ) are convergent and lim L(f ; Pn ) = lim U (f ; Pn ). In this case
n→∞ n→∞
Z b
f = lim L(f ; Pn ) = lim U (f ; Pn ) = lim L(f ; Qn ) = lim U (f ; Qn ) for any Qn ⊇ Pn .
a n→∞ n→∞ n→∞ n→∞

Z b
Proof. Assume f ∈ R([a, b]) and let I = f (x) dx. Then by sequential criterion, there exists a
a
sequence of partitions (Pn ) satisfying lim [U (f ; Pn )−L(f ; Pn )] = 0. Since L(f ; Pn ) ≤ I ≤ U (f ; Pn ),
n→∞
then 0 ≤ I − L(f ; Pn ) ≤ U (f ; Pn ) − L(f ; Pn ) and L(f ; Pn ) − U (f ; Pn ) ≤ I − U (f ; Pn ) ≤ 0. By
passing to the limit, we obtain lim L(f ; Pn ) = lim U (f ; Pn ) = I. Also, by passing to the limit
n→∞ n→∞
in L(f ; Pn ) ≤ L(f ; Qn ) ≤ I and I ≤ U (f ; Qn ) ≤ U (f ; Pn ) we obtain I = limn→∞ L(f ; Qn ) =
limn→∞ U (f ; Qn ). The proof of the converse statement is a consequence of sequential criterion.

2.2 Properties Of The Riemann Integral.

2.12 Theorem. Let f : [a, b] 7→ R be a bounded function and c ∈ (a, b). Then, f is integrable on
Z b Z c Z b
[a, b] if and only if f is integrable on [a, c] and [c, b]. In this case we have f= f+ f.
a a c

Proof. If f is integrable on [a, b], then for ε > 0 there exists a partition P satisfying U (f ; P ) −
L(f ; P ) < ε. Since refining a partition can bring the upper and lower sums closer together, we can
add c to P if it is not already there and take P1 = P ∩ [a, c] and P2 = P ∩ [c, b] as partitions of
[a, c] and [c, b]. We have

U (f ; P1 ) − L(f ; P1 ) ≤ U (f ; P ) − L(f ; P ) < ε and U (f ; P2 ) − L(f ; P2 ) ≤ U (f ; P ) − L(f ; P ) < ε.

Thus f is integrable on [a, c] and [c, b].


Conversely, if f is integrable on [a, c] and [c, b], then for ε > 0 there exists a partition P1 of [a, c]
satisfying U (f ; P1 ) − L(f ; P1 ) < 2ε and a partition P2 of [c, b] satisfying U (f ; P2 ) − L(f ; P2 ) < 2ε .
Then P = P1 ∪ P2 produces a partition of [a, b] satisfying U (f ; P ) − L(f ; P ) < ε. Thus f is
integrable on [a, b] an we have
Z b Z c Z b
f ≤ U (f ; P ) < L(f ; P ) + ε = L(f ; P1 ) + L(f ; P2 ) + ε ≤ f+ f + ε,
a a c

Z c Z b Z b
f+ f ≤ U (f ; P1 ) + U (f ; P2 ) < L(f ; P1 ) + L(f ; P2 ) + ε = L(f ; P ) + ε ≤ f + ε.
a c a
Z b Z c Z b Z c Z b Z b
Since ε > 0 arbitrary, the above inequalities imply f≤ f+ f and f+ f≤ f.
Z b Z c Z b a a c a c a

consequently, we obtain f= f+ f.
a a c

18
2.13 Definition. Let f : [a, b] 7→ R be a bounded integrable function on [a, b] and c ∈ [a, b]. Define
Z a Z b Z c
f =− f and f = 0.
b a c

2.14 Remark. If f is an integrable function on some interval I, and a, b, c any three points chosen
Z b Z c Z b
in any order from I, then it is straightforward to verify that f= f+ f.
a a c

2.15 Theorem. If f is integrable on [a, b] and [c, d] ⊆ [a, b], then f is integrable on [c, d].

Proof. The proof is similar to Theorem 2.12.

2.16 Theorem. Let f, g ∈ R([a, b]) and λ, m, M ∈ R. Then


Z b Z b Z b
1. f + g ∈ R([a, b]) and (f + g) = f+ g.
a a a
Z b Z b
2. λf ∈ R([a, b]) and λf = λ f.
a a
Z b
3. If m ≤ f ≤ M , then m(b − a) ≤ f ≤ M (b − a).
a
Z b Z b
4. If f ≤ g, then f≤ g.
a a
Z b Z b
5. |f | ∈ R([a, b]) and f ≤ |f |.

a a

Proof. Since f, g ∈ R([a, b]), there exists a sequence of partitions (Pn ) satisfying lim [U (f ; Pn ) −
n→∞
Z b
L(f ; Pn )] = 0 with f = lim L(f ; Pn ) = lim U (f ; Pn ) and a sequence of partitions (Qn )
a n→∞ n→∞
Z b
satisfying lim [U (g; Qn ) − L(g; Qn )] = 0 with g = lim L(g; Qn ) = lim U (g; Qn ).
n→∞ a n→∞ n→∞

1. Let Sn = Pn ∪ Qn be the common refinement of Pn and Qn . Recall that

Mk (f + g) ≤ Mk (f ) + Mk (g) and mk (f + g) ≥ mk (f ) + mk (g).

It follows that

U (f + g; Sn ) ≤ U (f ; Sn ) + U (g; Sn ) and L(f + g; Sn ) ≥ L(f ; Sn ) + L(g; Sn ). (2.1)

Consequently,

U (f + g; Sn ) − L(f + g; Sn ) ≤ [U (f ; Sn ) − L(f ; Sn )] + [U (g; Sn ) − L(g; Sn )].

19
Then by passing to the limit both sides we obtain lim [U (f + g; Sn ) − L(f + g; Sn )] = 0.
n→∞
Thus f + g ∈ R([a, b]), and by passing to the limit in inequalities (2.1) we obtain
Z b Z b Z b
(f + g) = f+ g.
a a a

2. For λ ≥ 0, we have Mk (λf ) = λMk (f ) and mk (λf ) = λmk (f ). It follows


U (λf ; Pn ) = λU (f ; Pn ) and L(λf ; Pn ) = λL(f ; Pn ).
Z b Z b
By passing to the limit we obtain λf ∈ R([a, b]) and λf = λ f.
a a
For λ < 0, we have Mk (λf ) = λmk (f ) and mk (λf ) = λMk (f ). It follows
U (λf ; Pn ) = λL(f ; Pn ) and L(λf ; Pn ) = λU (f ; Pn ).
Z b Z b
By passing to the limit we obtain λf ∈ R([a, b]) and λf = λ f.
a a

3. Evident.
Z b Z b Z b
4. We have 0 ≤ (g−f ), then by 3. we obtain 0 ≤ (g−f ). It follows from 1. that f≤ g.
a a a

5. Knowing that Mk (|f |) − mk (|f |) ≤ Mk (f ) − mk (f ), it follows that U (|f |; Pn ) − L(|f |; Pn ) ≤


U (f ; Pn ) − L(f ; Pn ). Hence lim [U (|f |; Pn ) − L(|f |; Pn )] = 0 and so |f | ∈ R([a, b]).
n→∞
Z b Z b
Since −|f | ≤ f ≤ |f |, then by 4. we obtain f ≤ |f |.

a a

2.17 Theorem. Let f ∈ R([a, b]). Assume that m ≤ f (x) ≤ M ∈ for all x ∈ [a, b] and g :
[m, M ] 7→ R is continuous. Then g ◦ f is integrable.

Proof. Left to the reader.

2.18 Corollary. Let f, g ∈ R([a, b]). Then |f |, f 2 , f g, max{f, g} and min{f, g} are integrable.

Proof. Left as exercise. To show that f g is integrable, we may use f g = 14 [(f + g)2 − (f − g)2 ].

2.3 Integrable Functions.

2.19 Theorem. If f : [a, b] 7→ R is bounded monotone, then it is integrable.

Proof. We prove the result for an increasing function. Then mk = f (xk−1 ) and Mk = f (xk ).
Choose the regular subdivision Pn = {xi = a+ ni (b−a); 0 ≤ i ≤ n}. (Pn ) is the sequence of partitions
of [a, b] that divides it into n subintervals with equal lengths, that is 4xk = xk − xk−1 = b−a n . Then
b−a P n
lim [U (f ; Pn ) − L(f ; Pn )] = lim (f (xk ) − f (xk−1 )) = limn→∞ b−a
n (f (b) − f (a)) = 0. Thus
n→∞ n→∞ n k=1
f ∈ R([a, b]).

20
2.20 Theorem. If f : [a, b] 7→ R is continuous, then it is integrable.

Proof. We have f is continuous on a closed and bounded interval, it is uniformly continuous.This


ε
means that, given ε > 0, there exists a δ > 0 such that |x − y| < δ guarantees |f (x) − f (y)| < (b−a) .
Let P be a partition of [a, b] such that 4xk = xk − xk−1 < δ. From the extreme value theorem
Mk = f (Yk ) and mk = f (yk ) for some Yk , yk ∈ [xk−1 , xk ]. Since |Yk − yk | ≤ xk − xk−1 < δ, then
ε
Mk − mk = f (Yk ) − f (yk ) < (b−a) . It follows that

n n
P ε P
U (f ; P ) − L(f ; P ) = (Mk − mk ) 4 xk < (b−a) 4xk = ε
k=1 k=1

. Thus f is integrable on [a, b].

2.21 Theorem. Let f : [a, b] 7→ R be a bounded function.

1. If f ∈ R([c, b]) for all c ∈ (a, b) then f ∈ R([a, b]).

2. If f ∈ R([a, c]) for all c ∈ (a, b) then f ∈ R([a, b]).

Proof. We prove the first item. The second one can done analogously. Let M > 0 so that |f (x)| ≤
M for all x ∈ [a, b] and let P = {a = x0 , x1 , · · · , xn = b} be a partition of [a, b]. Then
n
P n
P
U (f ; P ) − L(f ; P ) = (Mk − mk ) 4 xk = (M1 − m1 )(x1 − a) + (Mk − mk ) 4 xk .
k=1 k=2

The idea is that for ε > 0, the first step is to choose x1 close enough to a so that (M1 −m1 )(x1 −a) <
ε
ε. This is possible because (M1 − m1 ) ≤ 2M , so we can pick x1 so that x1 − a ≤ 4M . The
second step, since f is integrable on [x1 , b], then there exists a partition P1 of [x1 , b] so that
U (f ; P1 ) − L(f ; P1 ) ≤ 2ε . Finally we produce P2 = P1 ∪ {a} a partition of [a, b] so that
ε ε
U (f ; P2 ) − L(f ; P2 ) = (M1 − m1 )(x1 − a) + U (f ; P1 ) − L(f ; P1 ) ≤ + = ε.
2 2

2.22 Examples.

1. Let f : [0, 2] 7→ R defined by f (x) = sinx x . Note that f ( more precisely its extension by
continuity) is bounded on [0, 2]. Since f is continuous on [c, 2] for all c ∈ (0, 2), then f ∈
R([c, 2]) for all c ∈ (0, 2). It follows from above theorem that f ∈ R([0, 2]).

2. Let f : (0, 1] 7→ R defined by f (x) = sin x1 . Note that f can be regarded as bounded function
on [0, 1]. Since f is continuous on [c, 1] for all c ∈ (0, 1), then f ∈ R([c, 1]) for all c ∈ (0, 1).
It follows from above theorem that f ∈ R([0, 1]).

21
2.23 Corollary. Let f : [a, b] 7→ R be a bounded function.

1. If f is continuous on [a, b] except one end point, then it is integrable on [a, b].

2. If f is continuous on [a, b] except one point, then it is integrable on [a, b].

3. If f is continuous on [a, b] except a finite number of points , then it is integrable on [a, b].

2.4 Mean Value And Fundamental Theorems Of Calculus.

2.24 Theorem. (First Mean-Value Theorem For Integrals)


Let f, g ∈ R([a, b]), m := inf f (x) and M := sup f (x). If g is of constant sign on [a, b], then
x∈[a,b] x∈[a,b]
there exists k ∈ [m, M ] such that
Z b Z b
f (x)g(x) dx = k g(x) dx.
a a

If in addition f ∈ C[a, b], then there exists a point ξ ∈ [a, b] such that
Z b Z b
f (x)g(x) dx = f (ξ) g(x) dx.
a a

Proof. Assume without loss of generality that g(x) ≥ 0 for all x ∈ [a, b]. Since m ≤ f (x) ≤ M and
g(x) ≥ 0 on [a, b], then
mg(x) ≤ f (x)g(x) ≤ M g(x).
Since mg(x), f.g, and M g(x) ∈ R([a, b]), then
Z b Z b Z b
m g(x) dx ≤ f (x)g(x) dx ≤ M g(x) dx.
a a a

WeZ discuss two cases: Z


b b
If g(x) dx = 0, then f (x)g(x) dx = 0. In this case k can be chosen arbitrary in [m, M ] so
a Z b a Z b
that the equality f (x)g(x) dx = k g(x) dx holds true.
a a
Z b R b Rb
a f (x)g(x) dx f (x)g(x) dx
If g(x) dx > 0, then m ≤ R b ≤ M . Let k = a R b , then k ∈ [m, M ] and
a g(x) dx g(x) dx
Z b Za b a

the equality f (x)g(x) dx = k g(x) dx holds true.


a a
If in addition f ∈ C[a, b], then by the intermediate-value theorem there exists a point ξ ∈ [a, b]
such that k = f (ξ). Then it follows that
Z b Z b
f (x)g(x) dx = f (ξ) g(x) dx.
a a

22
The case g = 1 is interested. In this case the first mean-value theorem becomes,
Z b
If f ∈ R([a, b]), then f (x) dx = k(b − a) for some k ∈ [m, M ].
Z b a

If f ∈ C[a, b], then f (x) dx = f (ξ)(b − a) for some ξ ∈ [a, b].


a

2.25 Theorem. (Second Mean-Value Theorem For Integrals)


If f ∈ R([a, b]) and g is a monotonic function on [a, b], then there exists a point ξ ∈ [a, b] such that
Z b Z ξ Z b
f (x)g(x) dx = g(a) f (x) dx + g(b) f (x) dx.
a a ξ

Proof. Admitted.

For the second mean-value theorem we have three interested particular cases:

1. If f = 1 and g is a monotonic function on [a, b], then there exists a point ξ ∈ [a, b] such that
Z b
g(x) dx = g(a)(ξ − a) + g(b)(b − ξ).
a

2. If f ∈ R([a, b]) and g is a decreasing and positive or increasing and negative function on [a, b],
then there exists a point ξ ∈ [a, b] such that
Z b Z ξ
f (x)g(x) dx = g(a) f (x) dx.
a a

This is because we can modify the value of g at the point b so that g(b) = 0.

3. If f ∈ R([a, b]) and g is a increasing and positive or decreasing and negative function on [a, b],
then there exists a point ξ ∈ [a, b] such that
Z b Z b
f (x)g(x) dx = g(b) f (x) dx.
a ξ

This is because we can modify the value of g at the point a so that g(a) = 0.

Z x
2.26 Property. If f : [a, b] 7→ R is integrable, then the function F (x) = f (t) dt is continuous
a
on [a, b].

Proof. Fix x0 ∈ [a, b] and let x ∈ [a, b] . Then by the first mean-value theorem, there exists
k ∈ [m, M ] such that
Z x Z x0 Z x
F (x) − F (x0 ) = f (t) dt − f (t) dt = f (t) dt = k(x − x0 ).
a a x0

It follows that lim F (x) = F (x0 ) and F is continuous at x0 . Since x0 is an arbitrary point of [a, b],
x→x0
then F is continuous on [a, b].

23
2.27 Theorem. (Fundamental Theorem OfZ Calculus I )
x
If f ∈ C([a, b]), then the function F (x) = f (t) dt satisfies F ∈ C 1 ([a, b]) and F 0 = f on[a, b].
a

Proof. Fix x0 ∈ [a, b] and let x ∈ [a, b] \ {x0 } . Then by the first mean-value theorem, there exists
ξ between x0 and x such that
Rx Rx Rx
F (x) − F (x0 ) f (t) dt − a 0 f (t) dt f (t) dt f (ξ)(x − x0 )
= a
= x0 = = f (ξ).
x − x0 x − x0 x − x0 x − x0

F (x) − F (x0 )
It follows that lim = lim f (ξ) = f (x0 ) and F is differentiable at x0 with F 0 (x0 ) =
x→x0 x − x0 x→x0
f (x0 ) . Since x0 is an arbitrary point of [a, b], then F is differentiable on [a, b] and F 0 = f . Since
f ∈ C([a, b]), then so F 0 ∈ C([a, b]). Thus F ∈ C 1 ([a, b]) and F 0 = f on[a, b].

Theorem 2.30 says that, if f ∈ C([a, b]), then F is a primitive of f on [a, b].

2.28 Corollary. If f : [a, b] 7→ R is continuous on [a, b] and u, v : I 7→ [a, b] are differentiable


Z u(x)
functions on an interval I ⊆ R with values in [a, b]. Then the function G(x) = f (t) dt is
v(x)
differentiable on I and we have

G0 (x) = u0 (x)f (u(x)) − v 0 (x)f (v(x)), for all x ∈ I.

2.29 Theorem. (Fundamental Theorem Of Calculus II )


Z b
If F is continuous on [a, b] and F is a primitive of f on (a, b), and if f ∈ R([a, b]), then f =
a
F (b) − F (a).

Proof. Let P be a partition of [a, b]. Then by applying the mean value theorem to F on the interval
[xk−1 , xk ], we obtain

F (xk ) − F (xk−1 ) = f (ck )(xk − xk−1 ), for some ck ∈ (xk−1 , xk ).

n
P
Since mk ≤ f (ck ) ≤ Mk , then L(f ; P ) ≤ [F (xk )−F (xk−1 )] ≤ U (f ; P ). It follows that L(f ; P ) ≤
k=1
Z b Z b
F (b) − F (a) ≤ U (f ; P ). but L(f ; P ) = U (f ; P ) = f , then we conclude that f = F (b) −
a a
F (a).

Z b
Note that, g 0 = g(b) − g(a) provided that g is continuous on [a, b], differentiable on (a, b)
a
and g 0 integrable on [a, b].

24
2.5 Integration By Parts And Substitution Methods.

2.30 Theorem. (Integration By Parts )


If u and v are continuous functions on [a, b] that are differentiable on (a, b) and if u0 and v 0 are
integrable on [a, b], then
Z b Z b
uv 0 = [u(b)v(b) − u(a)v(a)] − u0 v.
a a

Proof. We have uv 0 = (uv)0 −u0 v. Use the fact that the product of integrable functions is integrable,
Z b Z b
0
linearity of integral and Theorem 2.29 to conclude uv = [u(b)v(b) − u(a)v(a)] − u0 v.
a a

2.31 Theorem. (Integration By Substitution )


Let f, φ and φ0 be continuous where f : [a, b] 7→ R and φ : [α, β] 7→ [a, b]. Then
Z β Z φ(β)
f (φ(t))φ0 (t) dt = f (x) dx.
α φ(α)

If in addition φ possesses an inverse function, then


Z b Z φ−1 (b)
f (x) dx = f (φ(t))φ0 (t) dt.
a φ−1 (a)

Z φ(y) Z y
Proof. Let F (y) = f (x) dx and G(y) = f (φ(t))φ0 (t) dt. we have F 0 = G0 on [α, β].
φ(α) α
It follows F (y) − G(y) = F (α) − G(α) = 0 for all y ∈ [α, β]. In particular, we conclude that
F (β) = G(β).

2.6 Riemann’s Original Definition.


In addition to approach of Darboux for integration, there is another equivalent approach due to
Riemann, which starts with a direct approximation of the integral by a sum.

2.32 Definition. (Riemann Sum, Mesh)


Let P = {a = x0 , x1 , · · · , xn = b} be a partition of [a, b] and ck ∈ [xk−1 , xk ].
n
• The Riemann sum of f with respect toP and (ck )nk=1 is R(f ; P, (ck )nk=1 ) =
P
f (ck )(xk −
k=1
xk−1 ).
n
• The mesh of P is defined by ||P || := max{xk − xk−1 }, that is the length of the largest
k=1
subinterval [xk−1 , xk ].

25
2.33 Definition. (Riemann’s Original Definition)
Let f : [a, b] 7→ R be a bounded function. We say that f is Riemann integrable on [a, b] if and
only if there is a real number I with the following property: For every ε > 0, there exists a δ > 0
such that ||P || < δ guarantees |R(f ; P, (ck )nk=1 ) − I| < ε for every possible choice ck ∈ [xk−1 , xk ].
Pn
That is lim f (ck )(xk − xk−1 ) = I.
||P ||→0 k=1
Z b
The quantity I = f is called the Riemann integral of f .
a

2.34 Remark. A bounded function f is Darboux integrable on [a, b] if and only if it is Riemann
Z b
integrable on [a, b]. In this case f in both senses is the same.
a

Proof. Admitted.

If we choose the regular partition Pn = {xk = a + k b−a n ; 0 ≤ k ≤ n} and ck ∈]xk−1 , xk [ or


ck = xk or ck = xk−1 then we obtain the following sequential characterization.

2.35 Corollary. If f is integrable on [a, b], then


n b
b−a P
Z
lim f (ck ) = f (x) dx
n→∞ n k=1 a
n
Z b
b−a P b−a
lim f (a + k n ) = f (x) dx
n→∞ n k=1 a
Z b
b − a n−1
P b−a
lim f (a + k n ) = f (x) dx.
n→∞ n k=0 a

In particular, if f is integrable on [0, 1], then


n
Z 1
1 P
lim f (ck ) = f (x) dx
n→∞ n k=1 0
n
Z 1
1 P k
lim f(n) = f (x) dx
n→∞ n k=1 0
Z 1
1 n−1
P k
lim f(n) = f (x) dx.
n→∞ n k=0 0

2.36 Examples.
n
P n
1. Let us study the convergence of the sequence of general term un = n2 +k2
. We have
k=1

1 Pn n
1 1
f ( nk ),
P
un = 2 = n
n k=1 1+ nk 2 k=1

26
1
where f (x) = is continuous, hence integrable on [0, 1]. Thus
1 + x2
Z 1 1 π
lim un = f (x) dx = arctan x = .

n→∞ 0 0 4

2n−1
P n
2. Let us study the convergence of the sequence of general term un = k2
. We have
k=n

n−1 n−1 n−1


n 1 1 1
f ( nk ),
P P P
un = (k+n)2
= n k 2
(1+ n )
= n
k=0 k=0 k=0

1
where f (x) = is continuous, hence integrable on [0, 1]. Thus
(1 + x)2
Z 1
1
lim un = f (x) dx = .
n→∞ 0 2

2.7 Schwarz Inequality For Riemann Integrals.

2.37 Property. Let f, g ∈ R([a, b]), then


Z b 2 Z b Z b
2
1. f (t)g(t) dt ≤ f (t) dt g 2 (t) dt.
a a a
Z b 2 Z b Z b
2. f (t)g(t) dt = f 2 (t) dt g 2 (t) dt ⇔ f = λg, λ ∈ R.
a a a

Proof. admitted.

27
2.8 Exercises
Exercise 1. Let f (x) = x2 on [0, b].

1. Show that f is integrable on [0, b] by using ε− criteria for integrability.


Z b
2. Show, using the sequential criteria , that f is integrable on [0, b] and compute f (x) dx .
0
n
n(n+1)(2n+1)
k2 =
P
Hint: you may use 6
.
k=1
Exercise 2. Let f : R 7→ R be a continuous T −periodic function, a, b ∈ R and n ∈ N.
Z b+T Z b
1. Show that f (x) dx = f (x) dx.
a+T a
Z a+T Z T Z a+nT Z T
2. Prove that f (x) dx = f (x) dx and f (x) dx = n f (x) dx.
a 0 a 0

Exercise 3. Let f : [−a, a] 7→ R be a continuous function.


Z a
1. Show that if f is an odd function, then f (x) dx = 0.
−a
Z a Z a
2. Show that if f is an even function, then f (x) dx = 2 f (x) dx.
−a 0
Z 2 5 − x
2
3. Deduce the value of x ln dx.
−2 5+x

Exercise 4. Let f : [a, b] 7→ R be a continuous function and c ∈ R.

1. Show that
Z b Z c−b Z b Z b
f (x) dx = − f (c − x) dx and f (x) dx = f (a + b − x) dx.
a c−a a a

Z π
4 tan α + tan β
2. Deduce the value of I = ln(1 + tan x)dx.(Hint: Use tan(α + β) = .)
0 1 − tan α tan β

Exercise 5. Let f : [a, b] 7→ R be a continuous function such that f (a + b − x) = f (x) for all
x ∈ [a, b].
Z b Z b
a+b
1. Show that xf (x) dx = f (x) dx.
a 2 a
Z π
2 sin(2x)
2. Deduce the value of M = x dx.
0 1 + cos2 (2x)

Exercise 6. Let m ∈ R
Z π
2 sinm (x)
1. Calculate Im = m dx.
0 sin x + cosm x
Z π
2 dx
2. Deduce the value of J = .
0 1 + tan4 x

28
Exercise 7. Compute the following integrals
Z π Z π Z π
4 cos x[1 + tan2 x] 2 1 2 1
1) dx, 2) dx, 3) dx,
sin x + cos x 0 2 + cos x 0 √ 1 + sin x cos x
Z0 π 1 2p
cos11 x
Z Z
1
4) 17 dx, 5) √ dx, 6) 2 − x2 dx.
1 + sin x 1 + x2
0 0 0
Exercise 8. Let f : R 7→ R be a continuous function. Calculate g 0 (x) where g is defined by
Z x3 Z x Z x
3
1)g(x) = f (t) dt, 2)g(x) = x f (t) dt, 3)g(x) = [x2 + f (t)]2 dt.
3x 0 0
Exercise 9. Compute g 0 (x) and specify the domain of differentiability of the function g defined by
Z x Z x2 Z sin x
1)g(x) = ln(1 − t) dt, 2)g(x) = arctan t dt, 3)g(x) = arg tanh t dt.
0 −x 0
− cosh x
xex
Z
Exercise 10. Let G(x) = ln(1 − et ) dt and g(x) = − sinh x ln(1 − e− cosh x ).
ln(1−ex ) 1 − ex
1. Show that G0 = g on an interval J to be determined.
2. Are the domain of definition of G0 and g the same? Justify.
Exercise 11. Find the limit of the following sequences
n−1 n n+1 2
 1
n k (2n)! n
√k 1
P P P
1)un = n2 +k2
, 2)un = n2 +k2
, 3)un = , 4)un = ,
n2 3 n3 +k3 n n!
k=0 k=1 k=1 v
u n
n n n 1u
k − kπ
Y
k
cos nk , 6)un = 1
P P P n
5)un = n2 n2
e 2n , 7)un = √
n2 +2kn
, 8)un = t (n + k).
k=1 k=1 k=1 n
k=1
1
(1 − x)n ex
Z
Exercise 12. Consider the sequence In = dx.
0 n!
1. Show that lim In = 0.
n→∞

2. Find a relation between In and In+1 .


n
1 P
3. Deduce that lim = e.
n→∞ k=0 k!

1 x
Z
Exercise 13. Let f be a continuous function on ] − 2, 2[. Find lim f (t) dt.
x→0 x 0
Exercise 14. Let f be a continuous function on [0, 1]. Find the following limits
Z x Z 1
1 n nf (t)
1) lim 3 f (t) dt, 2) lim 2
dt.
x→0 x
+
sin x n→∞ 0 1 + (nt)
Exercise 15. Find the following limits
Z 2x Z 2x Z 2x
1 et sin t
1) lim dt, 2) lim dt, 3) lim dt,
x→∞ x ln t x→0+ x t x→∞ x t
1 Z x2 3
sin πt
Z 2x x
cos t
Z
1 1
4) lim dt, 5) lim dt, 6) lim dt.
x→∞ x t x→1+ x t ln t x→1 +
x ln t
Exercise 16. Let f [0, ∞[7→ R be a continuous function such that lim f (x) = ` ∈ R. Find the
x→∞
limit of the following sequences
Z n+2 Z n+√n Z n Z n2 +n+1
f (x) f (x) arctan x2
1) f (x) dx, 2) √ dx, 3) dx, 4) √ dx.
n n x 0 n n2 +1 x
Exercise 17. Let f : [a, b] 7→ R be a continuous function.

29
Z b
1. Show that if f ≥ 0 on [a, b] and f (x) dx = 0, then f (x) = 0 for all x ∈ [a, b].
a
Z b Z b Z b
2 3
2. Assume (f (x)) dx = (f (x)) dx = (f (x))4 dx. Show that f = 0 or f = 1 on [a, b].
a a a
Hint: Expand (f 2 (x) − f (x))2 .

Z b Z b
3. Show that f (x) dx = |f (x)| dx if and only if f ≥ 0 or f ≤ 0 on [a, b].

a a

Exercise 18. Let f : [0, 1] 7→ R be a differentiable function such that f 0 is continuous and f (1) = 0.
Z 1 Z 1
1. Show that f (x) dx = − xf 0 (x) dx
0 0
Z 1 Z 1
2 1
2. Prove that f (x) dx ≤ (f 0 (x))2 dx. Under which conditions the equality holds.
0 3 0

Exercise 19. Let f be a function of class C 2 on [0, 1] such that f (0) = f 0 (1) = 0 and I =
Z 1
f (x) dx.
0
Z 1
1
1. Show that I = (x2 − 2x)f 00 (x) dx.
2 0

1
2. Deduce that |I| ≤ sup |f 00 (x)|.
3 x∈[0,1]
Z 1
Exercise 20. Let a ∈]0, 1[ and In (a) = (ln x)n dx.
a

1. Find a relation between In−1 (a) and In (a) for n ≥ 1.

2. Show that lim In (a) = (−1)n n!.


a→0+

30
CHAPTER 3
IMPROPER RIEMANN INTEGRALS

Riemann integrals have been defined for bounded functions over closed bounded intervals. In
this chapter we extend the concept of Riemann integral to functions defined on unbounded in-
tervals or to unbounded functions. Such integrals are called improper and they are defined
in terms of limits of ordinary integrals.
In this chapter I will denote a nonempty semi-open or open possibly unbounded interval,[a, b), (a, b]
or (a, b), of the real line R.
3.1 Definitions And Examples.

3.1 Definition. Let f (t) be a function defined on I. We say that:

• f is locally integrable on [a, b) if f is integrable on every closed interval [a, c] ⊆ [a, b).

• f is locally integrable on (a, b] if f is integrable on every closed interval [c, b] ⊆ (a, b].

• f is locally integrable on (a, b) if f is integrable on every closed interval [c, d] ⊆ (a, b).

Let Rloc (I) denotes the set of locally integrable functions on I.

3.2 Remark. If f is a continuous, bounded continuous except on a finite number of points, or


monotone function on I then it is respectively continuous, bounded continuous except on a finite
number of points or monotone on every closed interval of I, hence locally integrable on I.

The three basic types of improper integrals are as follows:


Improper Integrals Of Type I; (Unbounded domains of integration.)

3.3 Definition. Let a, b ∈ R, f ∈ Rloc ([a, ∞)) and g ∈ Rloc ((−∞, b]). We formally set
Z +∞ Z c Z b Z b
f (t) dt = lim f (t) dt, g(t) dt = lim g(t) dt.
a c→+∞ a −∞ d→−∞ d

• If the limit exists and is finite, we say that the improper integral converges.

• If the limit is infinite or does not exist, we say that the improper integral diverges to ∞ or
diverges, respectively.

31
3.4 Examples. Let f denotes the integrands of the following improper integrals.
Z ∞
1. Consider e−t dt. Since f ∈ C([0, ∞)), then f ∈ Rloc ([0, ∞)) and we have,
Z ∞ 0 Z c h ic
−t
e dt = lim e−t dt = − lim e−t = − lim (e−c − e0 ) = 1. Hence the improper
0 c→+∞ 0 c→+∞ 0 c→+∞
integral converges.
Z 0
2. Consider e−t dt. Since f ∈ C((−∞, 0]), then f ∈ Rloc ((−∞, 0]) and we have,
Z 0 −∞ Z 0 h i0
−t
e dt = lim e−t dt = − lim e−t = − lim (e0 − e−d ) = +∞. Hence the
−∞ d→−∞ d d→−∞ d d→−∞
improper integral diverges to ∞.
Z ∞
3. Consider sin t dt. Since f ∈ C([0, ∞)), then f ∈ Rloc ([0, ∞)) and we have,
Z ∞ 0 Z c h ic
sin t dt = lim sin t dt = − lim cos t = − lim (cos c − cos 0), @. Hence the
0 c→+∞ 0 c→+∞ 0 c→+∞
improper integral diverges.

Improper Integrals Of Type II;(Unbounded integrands.)

3.5 Definition. Let a, b ∈ R, f ∈ Rloc ([a, b)) and g ∈ Rloc ((a, b]). We formally set
Z b Z c Z b Z b
f (t) dt = lim f (t) dt, g(t) dt = lim g(t) dt.
a c→b− a a d→a+ d

• If the limit exists and is finite, we say that the improper integral converges.

• If the limit is infinite or does not exist, we say that the improper integral diverges to ∞ or
diverges, respectively.

3.6 Examples.
Z 1
1
1. Consider dt. Since f ∈ C([0, 1)), then f ∈ Rloc ([0, 1)) and we have,
Z 1 0 t − 1 Z c
1 1 h ic
dt = lim dt = lim ln |t − 1| = lim ln |c − 1| = −∞. Hence the
0 t−1 c→1− 0 t − 1 c→1− 0 c→1−
improper integral diverges to −∞.
Z2
1
2. Consider dt. Since f ∈ C((1, 2]), then f ∈ Rloc ((1, 2]) and we have,
Z 2 1 t−1 Z
2
1 1 h i2
dt = lim dt = lim ln |t − 1| = lim − ln |d − 1| = ∞. Hence the
1 t−1 d→1+ d t − 1 d→1+ d d→1+
improper integral diverges to ∞.
Z 1
1
3. Consider dt. Since f ∈ C([0, 1)), then f ∈ Rloc ([0, 1)) and we have,
1
0 (1 − t) 2
Z 1 Z c
1 1 h 1 c
i  1

1 dt = lim 1 dt = −2 lim (1 − t) 2 = −2 lim (1 − c) 2 − 1 = 2.
0 (1 − t) 2 c→1− 0 (1 − t) 2 c→1− 0 c→1−
Hence the improper integral converges.

32
1 Z
1 1
4. Consider 2
sin( ) dt. Since f ∈ C([0, 1)), then f ∈ Rloc ([0, 1)) and we have,
Z 1 0 (t − 1) t−1
Z c
1 1 1 1 h 1 ic  1
2
sin( ) dt = lim 2
sin( ) dt = lim cos( ) = lim cos( )−
0 (t − 1) t−1 0 (t − 1) t−1 t − 1 0 c→1 c−1
c→1 − c→1 − −

cos 1 , @. Hence the improper integral diverges.

Improper Integrals Of Type III;(Build up as sum of Type I and Type II.)

Rb
• First, one can consider improper integrals a f (t) dt where a difficulty occurs at both end
points of the interval of integration, either because the endpoints is at infinity or because
the integrand is unbounded there. The trick here is to pick an intermediate point c ∈ (a, b)
Rb Rc Rb
and write a = a + c , thus reducing the integral to a sum of two integrals that are each
of type I or II; The original integral is said to be convergent if and only if each of the two
sub-integrals is convergent.
Rb
• Second, one can consider improper integrals a f (t) dt where a difficulty occurs at an interior
point c of the intervalR of integration, because f has no finite limit at c+ or c− . The trick here
b Rc Rb
is to write formally a = a + c , thus reducing the integral to a sum of two integrals that
are each of type II; The original integral is said to be convergent if and only if each of the
two sub-integrals is convergent.

3.7 Examples.
Z ∞
1. Consider e−t dt. Since f ∈ C((−∞, ∞)), then f ∈ Rloc ((−∞, ∞)) and we have,
Z ∞ −∞Z Z ∞
0
−t
e dt = e−t dt + e−t dt. Hence the improper integral diverges, because one of
−∞ −∞ 0
the improper integrals to the right diverges.
Z 3
1
2. Consider p dt. Since f ∈ C((1, 3)), then f ∈ Rloc ((1, 3)) and we have,
1 (t − 1)(3 − t)
Z 3 Z 2 Z 3
1 1 1
p dt = p dt + p dt := I1 + I2 . we have;
1 (t − 1)(3 − t) 1 (t − 1)(3 − t) 2 (t − 1)(3 − t)
Z 2
1 h i2 π
I1 = p dt = lim arcsin(t − 2) = is convergent and
−[(t − 2)2 − 1] d→1+ d 2
Z1 3
1 h ic π
I2 = p dt = lim arcsin(t − 2) = is convergent. Thus the improper
2 −[(t − 2)2 − 1] c→3− 2 2
integral converges and we have,
Z 3 Z 2 Z 3
1 1 1 π π
p dt = p dt + p dt = + = π.
1 (t − 1)(3 − t) 1 (t − 1)(3 − t) 2 (t − 1)(3 − t) 2 2

Z 2
1
3. Consider dt. Since f ∈ C([0, 1)) ∩ C((1, 2]), then f ∈ Rloc ([0, 1)) ∩ Rloc ((1, 2]) and
0 t−1
we
Z 2 have, Z 1 Z 2
1 1 1
dt = dt + dt. Hence the improper integral diverges, because one
0 t − 1 0 t − 1 1 t − 1
of the improper integrals to the right diverges.

33
3.8 Remark. If f locally integrable on (a, b), that is a difficulty occurs at both end
Z b
points of the interval of integration, then the improper integral f (t) dt is convergent if
a
 Z c 
limd→a+ limc→b− f (t) dt exists. The order in which limits are taken does not matter.
d

3.9 Remark. Let f ∈ Rloc ((−a, a)), g ∈ Rloc ([a, c)) ∩ Rloc ((c, b]) and k ∈ Rloc ((a, b)) then the
(Cauchy) principal value
Z a Z c Z a−h
PV f (t) dt := lim f (t) dt = lim f (t) dt,
−a c→a+ −c h→0+ −a+h
Z b Z c−h Z b 
PV g(t) dt := lim g(t) dt + g(t) dt ,
a h→0+ a c+h
Z b Z b−h
PV k(t) dt := lim k(t) dt.
a h→0+ a+h

It may happen that the principal Z ∞value Zexists although Z 1the improperZ 3 integral is divergent. For
1
t 1 1
example; the improper integrals t dt, 2
dt, dt and dt are divergent, has no
−∞ −1 1 − t −1 t −2 t
Z ∞ Z 1 Z 1
t 1
sense, while their principal values exist and P V t dt = 0,P V 2
dt = 0, P V dt =
−∞ −1 1 − t −1 t
Z 3
1 3
0 and P V dt = ln( ).
−2 t 2
One can easily see that; If the improper integral exists, then the principal value also exists and
equal to the improper integral. Thus the principal value of an improper integral constitutes a proper
generalization of the improper integral. When we know á-priori that the improper integral exists,
we can evaluate it by just computing its principal value.

3.10 Remark. Recall that, if f locally integrable on (a, b) and bounded on ]a, b[, then f is integrable
Z 1
1
on [a, b]. For example; The integral sin dt is not of Type I nor of Type II near zero. The
0 t
1
integrand f (t) = sin t is locally integrable on (0, 1] and can be regarded as bounded function on
[0, 1], Hence integrable on [0, 1].

3.2 Basic Properties Of Improper Integrals.


The basic properties of Riemann integral can be generalized to improper Riemann integral by a
limit process.

3.11 Property. (Chasles Relation)


Z b Z b
Let f ∈ Rloc ([a, b)), and c ∈ (a, b). Then f (t) dt and f (t) dt have same nature and if they
a c

34
are convergent, we have
Z b Z c Z b
f (t) dt = f (t) dt + f (t) dt.
a a c

Proof. Let x ∈ (a, b). Then by Chasles relation for Riemann integral we have
Z x Z c Z x
f (t) dt = f (t) dt + f (t) dt.
a a c

Pass to the limit as x → b− ,


we obtain
Z x Z c Z x Z c
lim f (t) dt = f (t) dt + lim f (t) dt with f (t) dt is constant.
x→b− a a x→b− c a
Z b Z b
Therefor f (t) dt and f (t) dt have same nature and if they are convergent, we have
a c
Z b Z c Z b
f (t) dt = f (t) dt + f (t) dt.
a a c

3.12 Property. (Change Of Variable)


Let u : [a, b) 7→ R be of class C 1 on [a, b), b ∈ R and f be continuous function on u([a, b)). Then
for β ∈ (a, b) we have
Z β Z u(β)
0
f (u(t))u (t) dt = f (x) dx.
a u(a)
Z β
Thus, if the integral to the right has finite limit as β → b− , then f (u(t))u0 (t) dt is convergent
a
and we have Z b Z u(β)
f (u(t))u0 (t) dt = lim f (x) dx.
a β→b− u(a)

1 Z 1 √ Z 1
cos( t)2
Z
cos t cos t
3.13 Example. √ dt = 2 √ dt formally. Let β ∈ (0, 1), then √ dt =
0 t 0 2 t t
Z 1 Z 1 Z 1 β
cos t
2 √ cos x2 dx −−−−→ 2 cos x2 dx proper integral. Thus the improper integral √ dt is con-
β β→0+ 0 0 t
Z 1 Z 1
cos t
vergent and we have √ dt = 2 cos x2 dx.
0 t 0

3.14 Property. (Change Of Variable Bijection)


Let a, b, c, d ∈ R, u : (a, b) 7→ (c, d) of class C 1 and strictly monotone on (a, b) and let f : (c, d) 7→ R
be a continuous function on (c, d). Then for α, β ∈ (c, d) with α < β we have
Z β Z u−1 (β)
f (x) dx = f (u(t))u0 (t) dt with u−1 (α), u−1 (β) ∈ (a, b).
α u−1 (α)
Z d
+ −
Thus, if the integral to the right has finite limit as (α, β) → (c , d ), then f (x) dx is convergent
c
and we have: Z d Z b
f (x) dx = f (u(t))u0 (t) dt if u is strictly increasing ,
c a

35
Z d Z a
f (x) dx = f (u(t))u0 (t) dt if u is strictly decreasing .
c b

Z ∞
1 1
3.15 Example. Consider the improper integral x
sin x dx. For x = u(t) = ln t, we have
0 e e
Z ∞ Z ∞
1 1 1 1 1
u is of class C and strictly increasing on [1, ∞). Then x
sin x dx = sin( ) 2 dt =
0 e e 1 t t
1 ∞
cos( ) = 1 − cos 1.
t 1

3.16 Property. (Integration By Parts)


Z b
Let u and v be two functions of class C1
on [a, b), b ∈ R. If lim u(t)v(t) is finite, then u(t)v 0 (t) dt
t→b− a
Z b
and u0 (t)v(t) dt have same nature; and in case of convergence, we have
a
Z b b− Z b
0
u(t)v (t) dt = u(t)v(t) − u0 (t)v(t) dt,

a a a
b−
where u(t)v(t) = lim u(t)v(t) − u(a)v(a).

− a t→b

Proof. Let β ∈ (a, b). Then


Z β Z β
0
u(t)v (t) dt = [u(β)v(β) − u(a)v(a)] − u0 (t)v(t) dt.
a a
Z b Z b
Since lim u(β)v(β) is finite, then u(t)v 0 (t) dt and u0 (t)v(t) dt have same nature; and in case
β→b− a a
of convergence,by passing to the limit as β → b− in the above identity we obtain
Z b b− Z b
0
u(t)v (t) dt = u(t)v(t) − u0 (t)v(t) dt.

a a a

The integration by parts formula works for differentiable functions u and v on [a, b), b ∈ R
such that u0 , v 0 ∈ Rloc ([a, b)).

3.17 Property. (Linearity Of Integral)


Z b Z b
Let f, g ∈ Rloc ((a, b)), a, b ∈ R, and λ ∈ R. If f (t) dt and g(t) dt are convergent, then
a a
Z b 
Z b 
Z b Z b
1. f (t) + g(t) dt is convergent and f (t) + g(t) dt = f (t) dt + g(t) dt.
a a a a
Z b Z b Z b
2. λf (t) dt is convergent and λf (t) dt = λ f (t) dt.
a a a

36
Proof. Let α, β ∈ (a, b) such that α < β. For both items we look for the proper integrals
Z β Z β Z β Z β Z β

f (t) + g(t) dt = f (t) dt + g(t) dt and λf (t) dt = λ f (t) dt,
α α α α α
then take the limit both sides as (α, β) → (a, b) to obtain the result.
Z β Z β

Note that, from passing the limit in the proper integrals f (t) + g(t) dt = f (t) dt +
Z β Z b Z b α Z b α

g(t) dt, we have the right to write f (t) + g(t) dt = f (t) dt + g(t) dt if both integrals
α a a a
converge, diverge to +∞, diverge to −∞ or if one integral converges and the other diverges to ±∞.
Z b

If both integrals are divergent , then f (t) + g(t) dt may converges, diverges or diverges to ∞.
Z ∞ aZ
∞ Z ∞
For example, the integrals x dx and −x dx are both divergent, but (x − x) dx = 0 and
Z ∞ 0 0 0

(x + x) dx = ∞.
0 Z b

If one integral is convergent and the other is divergent, then the integral f (t) + g(t) dt is
a
divergent.

3.3 Convergence Tests.

3.18 Property. (Examples Of Riemann)


Let a, b, α ∈ R with a < b, then

1. Z ∞
1
dx converges ⇔ α > 1.
1 xα
2. Z −1
1
dx converges ⇔ α > 1.
−∞ |x|α

3. Z 1 Z b Z b
1 1 1
dx, dx and dx converge ⇔ α < 1.
0 xα a (x − a)α a (b − x)α

( c (
Z c x1−α c1−α −1
1 1−α 1 if α 6= 1, if α 6= 1,
Proof. 1. Let c ∈ (1, ∞). Then, dx = = 1−α
1 xα ln x|c1 if α = 1 ln c if α = 1.
Z +∞  1
1 α−1 if α > 1,
Therefore, dx =
1 xα ∞ if α ≤ 1.
2. For x = u(t) = −t, we have u is of class C 1 and strictly decreasing on [1, ∞). Then
Z −1 Z 1
1 1
α
dt = α
dt. Hence the convergence occur if and only if α > 1.
−∞ |x| ∞ t

3. For x = u(t) = 1t , we have u is of class C 1 and strictly decreasing on [1, ∞). Then,
Z 1 Z 1
1 1
α
dx = − −α+2
dt. Hence the convergence occur if and only if α < 1. For the other
0 x +∞ t
integrals we take respectively x − a = u(t) = 1t and b − x = u(t) = 1t .

37
Z b Z b
3.19 Property. Let f ∈ Rloc ([a, b)), b ∈ R. If f (t) dt is convergent, then lim f (t) dt = 0.
a x→b− x

Proof. Since the integral is convergent, then by Chasles relation and for x ∈ (a, b) we have
Z b Z x Z b
f (t) dt = f (t) dt + f (t) dt.
a a x

Then by passing to the limit as x → b− we obtain the desired result


Z b Z b Z b Z b
f (t) dt = f (t) dt + lim f (t) dt ⇔ 0 = lim f (t) dt.
a a x→b− x x→b− x

3.20 Property. ZLet f ∈ Rloc ([a, b)), b ∈ R be such that f (x) ≥ 0 for all x ∈ [a, b). Then the
x
function F (x) = f (t) dt is an increasing function on [a, b) .
a

Proof. let x, y ∈ [a, b) so that x < y. Then


Z x Z y Z a Z x Z x Z y
F (x) − F (y) = f (t) dt − f (t) dt = f (t) dt + f (t) dt = f (t) dt = − f (t) dt ≤ 0.
a a y a y x

3.21 Corollary. Let f ∈ Rloc ([a, b)), b ∈ R be such that f (x) ≥ 0 for all x ∈ [a, b). Then the
Z b
integral f (t) dt = lim F (x) is convergent if F is bounded above and diverges to +∞ elsewhere.
a x→b−

3.22 Property. (Comparison Test)


Let f, g ∈ Rloc ([a, b)), b ∈ R be such that 0 ≤ f (t) ≤ g(t) for all t ∈ [a, b). Then
Z b Z b Z b Z b
1. If g(t) dt converges, then f (t) dt converges and we have 0 ≤ f (t) dt ≤ g(t) dt.
a a a a
Z b Z b
2. If f (t) dt diverges, then g(t) dt diverges to +∞.
a a

Proof. 1. We have 0 ≤ f (t) ≤ g(t) for all t ∈ [a, b), then


Z x Z x
0 ≤ F (x) = f (t) dt ≤ G(x) = g(t) dt.
a a

38
Z b
Since g(t) dt is convergent, then G(x) is bounded and so F (x) is bounded above. Therefore
Z b a

f (t) dt converges and by passing to the limit as x → b− of both sides in the above
a Z b Z b
inequality, we obtain 0 ≤ f (t) dt ≤ g(t) dt.
a a
Z b Z b
2. Suppose to the contrary that g(t) dt converges, then by first item we obtain f (t) dt
Z b a Z b a

converges. But we have f (t) dt diverges to ∞. Thus g(t) dt diverges to ∞.


a a

3.23 Examples.

(sin t)2
Z
1. Consider I := dt. Since the integrand f ∈ C([1, ∞)), then f ∈ Rloc ([1, ∞))
1 1 + cos t + t2 Z ∞
1 1
and we have 0 ≤ f (t) ≤ t2 for all t ∈ [1, ∞). We have dt is convergent (Riemann
1 t2
Example), then by comparison test I is convergent and we have
Z ∞ Z ∞
(sin t)2 1
0 ≤ I := 2
dt ≤ 2
dt = 1.
1 1 + cos t + t 1 t


(sin t)2
Z
2. Consider J := dt. Since the integrand f ∈ C([0, ∞)), then f ∈ Rloc ([0, ∞)).
0 1 + cos t + t2
Since I converges, then by Chasles Relation J also converges.
Z ∞
1
3. Consider K := dt. Since the integrand f ∈ C([2, ∞)), then f ∈ Rloc ([2, ∞)) and we
2 ln t Z ∞
1 1 1
have 0 ≤ t ≤ ln t for all t ∈ [2, ∞). We have dt is divergent (Riemann Example), then
1 t
by comparison test K is divergent to ∞.

3.24 Definition. (Absolute And Conditional Convergence)


Z b
Let f ∈ Rloc ([a, b)), b ∈ R. The improper integral f (t) dt converges absolutely if the integral
Z b Z b a Z b
|f (t)| dt converges. However, if f (t) dt converges and |f (t)| dt diverges, then the integral
a
Z b a a

f (t) dt is said to be Conditionally convergent.


a

The notion of absolute convergence is useful functions with arbitrary sign.

3.25 Property. (Absolute Convergence Test)


Z b Z b
Let f ∈ Rloc ([a, b)), b ∈ R. If f (t) dt converges absolutely then it also converges and f (t) dt ≤
Z b a a

|f (t)| dt.
a

39
Z b
Proof. Note that 0 ≤ f (t) + |f (t)| ≤ 2|f (t)| for any t ∈ [a, b). Since |f (t)| dt converges, then
Z b a

the comparison test tell us that [f (t) + |f (t)|] dt converges. Thus, due to linearity of integral,
Z b Z b a

f (t) dt = ([f (t) + |f (t)|] − |f (t)|) dt is convergent. Since 0 ≤ f (t) + |f (t)| ≤ 2|f (t)| for any
a a Z b Z b
t ∈ [a, b), then by comparison test and linearity of integral we obtain f (t) dt ≤ |f (t)| dt.
a a

3.26 Remark. (Modified Comparison Test)


Z b
Let f, g ∈ Rloc ([a, b)), b ∈ R such that |f (t)| ≤ g(t) for all t ∈ [a, b) or near b− . If g(t) dt
Z b a

converges, then f (t) dt is absolutely convergent and hence convergent.


a
∞ Z
sin t
= |tsin t|
sin t 1
3.27 Example. Consider I := α+1
dt, 0 < α. We have tα+1 α+1 ≤ tα+1 . Since
t

Z b 1
1
α+1
dt converges, Riemann Examples α + 1 > 1, then by Comparison Test I is absolutely
a t
convergent and hence convergent.

3.28 Property. (Cauchy Criterion)


Let f ∈ Rloc ([a, b)), b ∈ R. Then
Z b Z X2
f (t) dt converges ⇔ lim f (t) dt = 0. Equivalently;
a (X1 ,X2 )→(b− ,b− ) X1

Z b Z X2
f (t) dt converges ⇔ ∀ε > 0, ∃x0 ∈ [a, b) such that f (t) dt < ε holds ∀X1 , X2 ∈ [x0 , b).

a X1

Z x Z b
Proof. Let F (x) = f (t) dt for all x ∈ [a, b). By definition, f (t) dt converges if and only if
a a
lim F (x) exists. We have
x→b−
Z X2 Z X2 Z X1
f (t) dt = f (t) dt − f (t) dt = F (X2 ) − F (X1 ).
X1 a a

Therefore the equivalence is simply the Cauchy criterion for the existence of lim F (x).
x→b−
Z X2 Z X2
It is useful to note that the following inequality f (t) dt ≤ |f (t)| dt holds for proper

X1 X1
integrals. Hence, Cauchy Criterion gives that absolute convergence implies convergence.
Z ∞
cos t
3.29 Example. Consider I := dt, 0 < α < 1.
1 tα
• I is convergent by Cauchy Criterion;
By integration by parts for proper integrals and Cauchy Criterion we have
Z X2 Z X2
1 sin X2 sin X1 sin t
α
cos t dt = α − α +α α+1
dt −−−−−−−−−−−→ 0.
X1 t X2 X1 X1 t (X1 ,X2 )→(∞,∞)

Thus, again by Cauchy Criterion I is convergent.

40
• I is convergent by Integration By Parts For Improper Integrals;
∞ Z ∞ Z ∞
sin t ∞
Z
1 sin t sin t
α
cos t dt = α + α α+1
dt = − sin 1 + α dt.
1 t t 1 1 t 1 tα+1
Since the integral to the right side is convergent, then I is convergent.

• I is not absolutely convergent; Z ∞


cos t cos2 t 1 + cos(2t) 1 cos(2t) 1
We have, α ≥ α = = α+ . Since dt is divergent and

t t α
2t Z 2t 2tα 2tα
Z ∞ 1

cos(2t) cos2 t
dt is convergent, then dt is divergent. Consequently, and by compari-
1 2tα 1 tα
son test we obtain that I is not absolutely convergent.

3.30 Definition. (Landau Notations)


Let f and g be two functions defined in a neighborhood of a ∈ R . We say that

1. f is dominated by g in a neighborhood of a, and we write f = O (g), if there exist a function


x→a
b defined in a neighborhood of a such that

• f (x) = b(x)g(x) in a neighborhood of a.


• b is bounded in a neighborhood of a.

2. f is negligible with respect to g in a neighborhood of a, and we write f = o (g), if there exist


x→a
function ε defined in a neighborhood of a such that

• f (x) = ε(x)g(x) in a neighborhood of a.


• ε(x) −−−→ 0.
x→a

3. f is equivalent to g in a neighborhood of a , and we write f ∼ (g), if there exist function ε


x→a
defined in a neighborhood of a such that

• f (x) = (ε(x) + 1)g(x) in a neighborhood of a.


• ε(x) −−−→ 0.
x→a

Note that f ∼ g ⇔ f − g = o (g).


x→a x→a

3.31 Property. Let f and g be two functions defined in a neighborhood N of a ∈ R . If g never


vanishes on N \ {a}, then:
f
1. f = O (g) ⇔ the function is bounded in a neighborhood of a.
x→a g
f (x)
2. f = o (g) ⇔ −−−→ 0.
x→a g(x) x→a
f (x)
3. f ∼ (g) ⇔ −−−→ 1.
x→a g(x) x→a

41
3.32 Property. Let f, g ∈ Rloc ([a, b)), b ∈ R. Then
Z b Z b
1. If f = O (g) and g(t) dt converges absolutely, then f (t) dt converges absolutely.
x→b a a
Z b Z b
2. If f = o (g) and g(t) dt converges absolutely, then f (t) dt converges absolutely.
x→b a a

3. (Equivalent test)
Z b Z b
If f ∼ g, then |f (t)| dt and |g(t)| dt have same nature.
x→b a a

Proof. 1. Assume f = O (g), there exist a bounded function B defined in a neighborhood of


x→b
b such that f (x) = B(x)g(x) in a neighborhood of b. Hence there exists M > 0 such that
Z b
|f (x)| ≤ M |g(x)| in a neighborhood of b. Since |g(t)| dt converges, then by comparison
Z b a

test f (t) dt converges absolutely.


a

2. Evident. We use, f = o (g) ⇒ f = O (g).


x→b x→b

3. Evident. We use, f ∼ g ⇒ f = O (g) ∧ g = O (f ) .
x→b x→b x→b

3.33 Examples.
Z ∞
sin t
1. Consider I = ln(1 + ) dt. Since the integrand f ∈ C([1, ∞)), then f ∈ Rloc ([1, ∞))
1 t Z ∞
sin t sin t sin2 t sin2 t sin t
and we havef (t) = ln(1 + )= − 2
+ o( 2 ). Since dt is convergent,
Z ∞ t t 2t t 1 t
sin2 t
integration by parts, and dt is convergent, by comparison test and Riemmann ex-
1 t2
sin t
amples, then I is convergent. Note that we cannot use equivalent test here because does
t
not have constant sign near ∞.
Z ∞
sin2 t
2. Consider J = ln(1 + ) dt. Since the integrand f ∈ C([1, ∞)), then f ∈ Rloc ([1, ∞))
1 t Z ∞ Z ∞
sin2 t sin2 t sin2 t 1 − cos(2t)
and we have ln(1 + t ) ∼ t ≥ 0 near ∞. Since dt = dt is
x→∞ 1 t 1 2t
divergent, then J is divergent.

3.34 Theorem. (Limit Form Of Comparison Test)


Let f, g ∈ Rloc ([a, b)), b ∈ R be such that f (x) ≥ 0 and g(x) > 0 in neighborhood of b.
Z b Z b
f (x)
1. If lim = ` > 0, then g(t) dt and f (t) dt have same nature.
x→b− g(x) a a
Z b Z b
f (x)
2. If lim = 0 and g(t) dt converges, then f (t) dt converges.
x→b− g(x) a a

42
Z b Z b
f (x)
3. If lim = ∞ and g(t) dt diverges, then f (t) dt diverges to ∞.
x→b− g(x) a a

f (x)
Proof. 1. We use, lim = ` > 0 ⇒ f ∼ `g.
x→b− g(x) x→b

f (x)
2. We use, lim = 0 ⇒ f = o (g).
x→b− g(x) x→b

f (x) f (x)
3. We use, lim = ∞ ⇒ > M ⇒ f (x) > M g(x) for some positive M in in a
g(x)
x→b− g(x)
neighborhood of b.

3.35 Corollary. (xα Test)

• Let f ∈ Rloc ([1, ∞)) and α ∈ R.


Z ∞
1. If xα f (x) −−−→ ` 6= 0, then|f (t)| dt converges for α > 1 and diverges for α ≤ 1.
x→∞ 1
Z ∞
2. If xα f (x) −−−→ 0 and α > 1, then |f (t)| dt converges.
x→∞ 1
Z ∞
α
3. If x f (x) −−−→ ∞ and α ≤ 1, then f (t) dt diverges.
x→∞ 1

• Let f ∈ Rloc ([a, b)) and a, b, α ∈ R.


Z b
1. If (b − x)α f (x) −−−→ ` 6= 0, then |f (t)| dt converges for α < 1 and diverges for α ≥ 1.
x→b a
Z b
2. If (b − x)α f (x) −−−→ 0 and α < 1, then |f (t)| dt converges.
x→b a
Z b
3. If (b − x)α f (x) −−−→ ∞ and α ≥ 1, then f (t) dt diverges.
x→b a

• Let f ∈ Rloc ((a, b]) and a, b, α ∈ R.


Z b
1. If (x − a)α f (x) −−−→ ` 6= 0, then |f (t)| dt converges for α < 1 and diverges for α ≥ 1.
x→a a
Z b
2. If (x − a)α f (x) −−−→ 0 and α < 1, then |f (t)| dt converges.
x→a a
Z b
3. If (x − a)α f (x) −−−→ ∞ and α ≥ 1, then f (t) dt diverges.
x→a a

3.36 Property. (Bertrand Examples)


Let α, β ∈ R. Then
Z ∞
1
1. I(α, β) = dx converges ⇔ (α > 1) or (α = 1 and β > 1).
e x lnβ x
α

43
Z 1
e 1
2. J(α, β) = dx converges ⇔ (α < 1) or (α = 1 and β > 1).
0 xα | ln x|β

Proof. 1. We discuss three cases:

1+α xµ
• Case one α > 1: Choose µ ∈ (1, α) for example µ = 2 . Since lim = 0 and
x→∞ xα lnβ x
µ > 1, then I(α, β) converges.
1+α xµ
• Case two α < 1: Choose µ ∈ (α, 1) for example µ = 2 . Since lim = ∞ and
x→∞ xα lnβ x
µ < 1, then I(α, β) diverges.
• Case three α = 1: We perform the change Z ∞of variable x = Zu(t) = et which is of class C 1

1 1
and strictly increasing on [1, ∞). Then β
dx = dt converges if β > 1.
e x ln x 1 tβ
Thus we conclude that I(α, β) converges if and only if (α > 1) or (α = 1 and β > 1).

2. We perform the the change of variable x = u(t) = 1t which is of class C 1 and strictly decreasing
Z 1 Z ∞
e 1 1
on [e, ∞). Then J(α, β) = α | ln x|β
dx = dt. Hence by firs item;
0 x e t −α+2 lnβ t

J(α, β) converges ⇔ (−α+2 > 1) or (−α+2 = 1 and β > 1) ⇔ (α < 1) or (α = 1 and β > 1).

There are improper integrals which are convergent but not absolutely convergent. These are
often quite subtle to handle. We give Abel-Dirichlet test which is useful to deal with such integrals.
The basic tool for this test is the second mean value theorem and hence essentially the integration
by parts formula.

3.37 Property. (Abel-Dirichlet Test)


Z b
Let f, g ∈ Rloc ([a, b)), b ∈ R be such that g is monotonic. Then f (t)g(t) dt converges if one of
a
the following pairs of conditions hold:
Z b
C1 : g is bounded and f (t) dt converges. Or
a Z x
C2 : g(t) −−−→ 0 and F (x) = f (t) dt is bounded on [a, b).
t→b− a

Proof. Let X1 , X2 ∈ [a, b). Then by the second mean-value theorem, there exists ξ ∈ [X1 , X2 ] such
that Z X2 Z ξ Z X2
f (t)g(t) dt = g(X1 ) f (t) dt + g(X2 ) f (t) dt.
X1 X1 ξ

Hence by the Cauchy Criterion, we conclude that the integral converges if either of the two pairs
of conditions holds.

3.4 Schwarz Inequality For Improper Integrals.


We generalize Schwarz inequality for Riemann integrals to improper integrals by a limit process.

44
Z b Z b
3.38 Property. Let f, g ∈ Rloc ([a, b)), b ∈ R. If f 2 (t) dt and g 2 (t) dt are convergent, then
Z b a a

f (t)g(t) dt is convergent and we have


a
Z b 2 Z b Z b
2
1. f (t)g(t) dt ≤ f (t) dt g 2 (t) dt.
a a a
Z b 2 Z b Z b
2. f (t)g(t) dt = f 2 (t) dt g 2 (t) dt ⇔ f = λg, λ ∈ R.
a a a

b
(f 2 + g 2 )
Z
Proof. The inequality |f g| ≤ ensures the absolute convergence of f (t)g(t) dt.
2 a

1. Schwarz inequality for Riemann integral gives


Z x 2 Z x Z x
2
∀x ∈ [a, b), f (t)g(t) dt ≤ f (t) dt g 2 (t) dt.
a a a

By passing to the limit as x → b− we obtain


Z b 2 Z b Z b
2
f (t)g(t) dt ≤ f (t) dt g 2 (t) dt.
a a a

2. Schwarz equality for Riemann integral gives


Z x 2 Z x Z x
∀x ∈ [a, b), f (t)g(t) dt = f 2 (t) dt g 2 (t) dt ⇔ f = λg, λ ∈ R.
a a a

By passing to the limit as x → we obtain b−


Z b 2 Z b Z b
2
f (t)g(t) dt = f (t) dt g 2 (t) dt ⇔ f = λg, λ ∈ R.
a a a

3.5 Integration Of Comparison Relations.

Z b
3.39 Property. Let f, g ∈ Rloc ([a, b)) and g positive on [a, b), b ∈ R . Assume that g(t) dt is
a
convergent, then for x tends to b− we have
Z b Z b Z b 
1. If f = O (g) then, f (t) dt = O (g) dt = O g(t) dt .
t→b x x t→b x→b x
Z b Z b Z b 
2. If f = o (g) then, f (t) dt = o (g) dt = o g(t) dt .
t→b x x t→b x→b x
Z b Z b
3. If f ∼ g, then f (t) dt ∼ g(t) dt.
t→b x x→b x

45
Proof. Exercise.

3.40 Corollary. Let f ∈ Rloc ((a, b)), a, b ∈ R and x0 ∈ (a, b). Assume that f Zhas finite expansion
n x
ai xi + o (x − x0 )n . Then the function x 7→
P
up to order n at x0 , f (x) = f (t) dt has finite
i=0 x→x0 x0
Z x Z x n
P i
expansion up to order n + 1 at x0 and we have f (t) dt = ai t dt + o (x − x0 )n+1 .
x0 x0 i=0 x→x0

Z b
3.41 Property. Let f, g ∈ Rloc ([a, b)) and f, g be positive on [a, b), b ∈ R . Assume that g(t) dt
a
is divergent, then for x tends to b− we have
Z x Z x 
1. If f = O (g) then, f (t) dt = O g(t) dt .
t→b a x→b a
Z x Z x 
2. If f = o (g) then, f (t) dt = o g(t) dt .
t→b a x→b a
Z x Z x
3. If f ∼ g, then f (t) dt ∼ g(t) dt.
t→b a x→b a

3.42 Example. Z 1
1 1 1 1
∼ ⇒ dt ∼ ln .
ln(1 + t) t→0+ t x ln(1 + t) x→0+ x

46
3.6 Exercises
Exercise 1. Study the nature of the following improper integrals
Z ∞ Z ∞ Z ∞ Z ∞
x −x 1
1) e dx, 2) e dx, 3) x dx, 4) √ dx,
0 0 −∞ 1 x3
Z ∞ Z ∞ Z ∞ Z ∞
1 1 cos x
5) sin(5x) dx, 6) dx, 7) dx, 8) dx,
0 e x ln x e x(ln x)2 0 1 + x2
Z ∞ Z ∞ Z ∞ Z ∞
x+1 ln x
9) √ dx, 10) √ dx, 11) cos x2 dx, 12) sin(x2 + x + 1) dx.
0 x4 + 1 1 x 0 0

Exercise 2. Study the nature of the following improper integrals


Z 1 Z 1 Z 0 Z 2
1 1 1 1
1) dx, 2) √ dx, 3) dx, 4) √ dx,
−1 x2 0 x3 −2 (x + 1)2 0 2x
π π
Z 2
1
Z
2 1
Z
2 √ Z b
1
5) √ dx, 6) dx, 7) tan x dx, 8) dx,
1 + (sin x)2
p
0 2x − x 0 0 a (x − a)(b − x)
π

Z
2
Z ∞
e− x2 +x+1 Z 1
ln x
Z 1
ln(1 − x2 )
9) ln(1 + sin x) dx, 10) √ dx, 11) √ dx, 12) dx.
− π2 0 x 0 1−x 0 x2

Exercise 3. Prove that the following integrals are semi-convergent


Z ∞ Z ∞ Z ∞
cos x 2
1) √ dx, 2) cos x dx, 3) x2 sin x4 dx.
1 x −2 1

Exercise 4.

x−1
Z
1. Show that dx is convergent and determine its value.
1 (x + 1)(x2 − x + 1)
Z ∞
x+2
2. Study the nature of dx.
1 (x − 1)(x2 + x + 1)
Z x 1 + t
Exercise 5. Let x ∈ [0, 1[ and I(x) = t ln dt. Calculate I(x) and find lim I(x).
Z 1 Z 0∞ 1−t x→1−
ln x ln x
Exercise 6. Let I = 2
and J = .
0 1+x 1 1 + x2
1. Show that I and J are convergent. Calculate I + J.
Z ∞
∗ ln x π ln a
2. Let a ∈ R+ . Show that 2 2
=− .
0 1+a x 2a
Z 3
Exercise 7. Let p, q ∈ R. Study the nature of I(p, q) = (x + 2)p (3 − x)q−1 dx.
Z ∞ −2
p−1 −x
Exercise 8. Let p ∈ R and Γ(p) = x e dx.
0

1. Study the nature of Γ(p).

2. Find a relation between Γ(p + 1) and Γ(p).

3. Deduce that Γ(n + 1) = n! for all n ∈ N.

47
1
xα + x1−α
Z
Exercise 9. Study according to the values of α the nature of dx.
0 1+x
Exercise 10. Let n ∈ Z and a ∈]0, 1[.
Z 1
1. Calculate In (a) = xn ln x dx in terms of n and a.
a
Z 1
2. Establish for what values of n the integral xn ln x dx is convergent and give its value .
0
1
xn
Z
Exercise 11. Consider the sequence un = p dx.
0 x(1 − x)
1. Show that un is well defined.
2. Show that {un }n is decreasing and bounded below. Deduce the nature of {un }n .
3. Compute u0 , u1 and show that 2nun = (2n − 1)un−1 for n ≥ 1.
4. Determine the value of un in terms of n.
Exercise 12. Study according to the values of α ∈ R the nature of the following integrals
Z ∞ α+1 Z ∞ Z ∞
x + x3−α sin2 x 1
1) √ dx, 2) α
dx, 3) α [1 + (ln x)2 ]
dx,
0 4
x + x 3 0 x 0 x
1 cosα ( πx
2 )

x2α−2 + xα+1 ∞
Z Z Z
x ln x
4) dx, 5) √ 1 dx, 6) dx.
(1 + x2 )α
p
0 x(1 − x) 0 x3 x3 (3 + x)α+ 2 0

Exercise 13. Let x ∈ R∗+ .


Z 1 Z ∞
x sin t x sin t
1. Prove that dt and dt are convergent.
0 t(t + x) 1 t(t + x)
Z 1 Z ∞
x sin t x sin t
2. Let u(x) = dt and v(x) = dt.
0 t(t + x) 1 t(t + x)
x + 1
(a) Show that 0 ≤ u(x) ≤ x ln .
x
(b) Show that lim v(x) = 0.
x→0+
Z ∞
x sin t
3. Deduce that lim dt = 0.
x→0+ 0 t(t + x)
Z ∞
Exercise 14. Let f : [1, ∞[7→ R be a continuous function such that f (x) dx is convergent.
Z ∞ 1
f (x)
Show that dx is convergent for α > 0.
1 xα Z ∞
sin t
Exercise 15. Find lim dt.
x→∞ 0 t+x Z ∞
f (t)
Exercise 16. Let f : [0, ∞[7→ R be a continuous function such that dt is convergent.
1 t
Assume that 0 < ε < X and 0 < a < b.
Z X Z bε Z bX
f (at) − f (bt) f (t) f (t)
1. Show that dt = dt − dt.
ε t aε t aX t
Z ∞
f (at) − f (bt)  
2. Deduce that dt is convergent to f (0) ln ab .
0 t
Z ∞ Z ∞ −at
cos(at) − cos(bt) e − e−bt
3. Calculate dt and dt.
0 t 0 t

48
CHAPTER 4
SERIES

4.1 Numerical Series.


n
P
The sum of finitely many real numbers a0 , a1 , · · · , an is a given real number ak . Sometimes we
k=0

P
wish to form the sum of an infinite sequence of real numbers a0 + a1 + · · · + an + · · · = ak .
k=0
For example, if a man walks halfway across a road of length two units and repeats this process
indefinitely, then one can think the total distance he will travel is an infinite sum
1 1 1 1 ∞
P 1
2 = 1 + + + + ··· + n + ··· = 2k
.
2 4 8 2 k=0

In (n + 1) steps he will travel n-th partial sum units


1 1 1 1 n
P 1 1
1+ + + + ··· + n = 2k
=2− 2n−1
units.
2 4 8 2 k=0
Thus he will get closer and closer to the other side of the road as n → ∞ and we have
n ∞
P 1 P 1
lim 2k = 2 = 2k
,
n→∞ k=0 k=0
the limit of the n-th partial sum equals to the infinite sum. This infinite sum can be defined using
sequences and leads to the notion of series.

Given a sequence {ak }k≥0 we can form another sequence of partial sums {sn }n≥0 by
n
P
sn = ak = a0 + a1 + · · · + an−1 + an .
k=0

Let us (formally) define



P n
P
ak = lim ak = lim sn .
k=0 n→∞ k=0 n→∞

P P
The symbol ak (or ak ) is called series, ak = sk − sk−1 is the general term of the series
k=0 k≥0
and k is called the index. Note that the index is a dummy variable

P ∞
P ∞
P
a0 + a1 + · · · + an + · · · = ak = ai = an ,
k=0 i=0 n=0

and it may start from some m ∈ N. In this case, formally



P n
P n+m
P
ak = lim sn = lim ak = lim ak .
k=m n→∞ n→∞ k=m n→∞ k=m

49
n
P
4.1 Definition. Given a sequence {ak }k≥0 and sn = ak . Consider the limit lim sn .
k=0 n→∞

• The series is said to converge to a real number s if the sequence of partial sums converges
to s, and one writes
Pn ∞
P
s = lim sn = lim ak = ak .
n→∞ n→∞ k=0 k=0

• The series is said to diverge or diverge to ∞ if the sequence of partial sums diverges, or
diverges to ∞, respectively.

In the first case we say that the series is convergent and in the other case we say that the series
is divergent.

4.2 Property. (Necessary But Not Sufficient Condition For Convergence )



P
If the series an converges, then lim an = 0. Equivalently;
n=0 n→∞

P
If lim an 6= 0, then the series an is divergent.
n→∞ n=0

Proof. Suppose s = lim sn . Since an = sn − sn−1 , then lim an = lim (sn − sn−1 ) = s − s = 0.
n→∞ n→∞ n→∞

4.3 Examples.

P 1
1. The series 2k
converges to 2.
k=0
1 − xn+1
Using the identity 1 + x + x2 + · · · + xn = for x 6= 1, then we obtain
1−x
n 1−( 12 )n+1
P 1 1
sn = 2k
= 1− 12
=2− 2n .
k=0

∞ ∞
P 1 P 1
Since lim sn = 2, then the series 2k
converges to 2 and we have 2k
= 2.
n→∞ k=0 k=0


(−1)n diverges.
P
2. The series
k=0

(−1)n diverges.Or,
P
Since s2n = 1 and s2n+1 = 0, then lim sn does not exist. So the series
n→∞ k=0
since lim (−1)n 6= 0, then the series is divergent.
n→∞


(1)n diverges to ∞.
P
3. The series
k=0
n
1n = 1 + 1 + · · · + 1 = n + 1. Hence lim sn = ∞. So the series diverges
P
Note that sn =
k=0 n→∞
to ∞. Or, since lim 1n 6= 0, then the series is divergent.
n→∞

The above three examples are special cases of geometric series.

50
4.4 Example. (Geometric Series)

xk
P
The geometric series
k=0

1
xk =
P
• converges if |x| < 1 and 1−x .
k=0

• Diverges to ∞ if x ≥ 1 and diverges if x ≤ −1.


(1)n diverges to ∞.
P
Solution. If x = 1, then the series
k=0
1 − xn+1
Suppose x 6= 1, then sn = 1 + x + x2 + · · · + xn = . Hence
1−x
1

 = 1−x , if |x| < 1;
lim sn = ∞, if x > 1;
n→∞
does not exist, if x ≤ −1.

Another way to show the geometric series is divergent for |x| ≥ 1 is that lim an = lim xn 6= 0.
n→∞ n→∞

4.5 Remark. (Geometric Series)



xn converges if and only if |x| < 1 and
P
In general, for any m ∈ N, the series
k=m

∞ n+m n
1
xn = lim xn = lim (xm + xm+1 + · · · + xm+n ) = xm lim xn = xm 1−x
P P P
.
k=m n→∞ k=m n→∞ n→∞ k=0

4.6 Property. (Telescoping Series)



P
The series ak is called telescoping if ak = bk+1 − bk for a suitable sequence {bk }k≥m and
k=m


P ∞
P
ak = (bk+1 − bk ) = lim bn+1 − bm .
k=m k=m n→∞

Proof. we have
n
P
sn = ak = am + am+1 · · · + an = (bm+1 − bm ) + (bm+2 − bm+1 ) + · · · + (bn+1 − bn ) = bn+1 − bm .
k=m


P
Then ak = lim bn+1 − bm .
k=m n→∞

This property show how to construct series that converge to any given number or diverge to ∞.
4.7 Examples.
1. A typical example.
∞ ∞
1
( k1 − 1 1 1
P P
= − k−1 ) = −( lim − 2−1 ) = 1.
(k−1)k
k=2 k=2 n→∞ n

51
2. The condition lim ak = 0 is necessary but not sufficient condition for convergence.
k→∞
∞ ∞
1
P P 
ln(1 + k ) = ln(k + 1) − ln k = lim ln(n + 1) − ln 1 = ∞.
k=1 k=1 n→∞

4.8 Property. (Linearity Of Series)



P P∞
Let α ∈ R, an and bn be convergent series. Then
n=0 n=0

P ∞
P ∞
P
1. αan is a convergent series and αan = α an .
n=0 n=0 n=0

P ∞
P ∞
P ∞
P
2. (an + bn ) is a convergent series and (an + bn ) = an + bn .
n=0 n=0 n=0 n=0

Proof. For both items we look at the partial sums

n
P n
P n
P n
P n
P
αak = α ak and (ak + bk ) = ak + bk ,
k=0 k=0 k=0 k=0 k=0

then take the limit of both sides to obtain the result.

n
P n
P n
P
Note that, from passing the limit in (ak + bk ) = ak + bk , we have the right to write
k=0 k=0 k=0

P ∞
P ∞
P
(an + bn ) = an + bn if both series converge, diverge to +∞, diverge to −∞ or if one
n=0 n=0 n=0
series converges and the other diverges to ±∞.

P
If both series are divergent , then (an + bn ) may converges, diverges or diverges to ∞. For
n=0
∞ ∞ ∞
(−1)n −(−1)n are both divergent, but (−1)n − (−1)n = 0 and
P P P 
example, the series and
n=0 n=0 n=0

(−1)n + (−1)n diverges.
P 
n=0

P
If one series is convergent and the other is divergent, then the series (an + bn ) is divergent.
n=0


P ∞
P
4.9 Property. If ak is a convergent series and bk any series such that bk = ak except for
k=0 k=0

P
at most finitely-many k, then bk is also convergent.
k=0

n
P n
P n
P n
P n0
P
Proof. Let sn = ak and σn = bk . Note that σn = (bk − ak ) + ak = (bk − ak ) + sn
k=0 k=0 k=0 k=0 k=0
for some fixed n0 ∈ N. Take the limit of both sides to obtain the result.

From this property one can deduce that a series behavior does not change by adding, modifying or
eliminating a finite number of terms.

52

P
4.10 Property. If ak is a positive term series, ak ≥ 0 for all k ∈ N, the sequence of partial
k=0
sum {sn }n≥0 is an increasing sequence and so the series is either converges ( when the sequence of
partial sum {sn }n≥0 is bounded above ) or diverges to ∞.

Proof. We have sn+1 − sn = an+1 ≥ 0. Then the sequence {sn }n≥0 is an increasing sequence.

Note that, to show a positive-term series is convergent, it is sufficient to show the sequence of
partial sum is bounded above.


P
4.11 Definition. The infinite sum rn = ak is called the nth remainder or the tail of the
k=n+1

P
series . If the series ak converges to s, then rn = s − sn .
k=0

Here is another necessary condition for convergence.


P
4.12 Property. If ak is a convergent series, then lim rn = 0. Equivalently;
k=0 n→∞

P
If lim rn 6= 0, then the series ak is divergent.
n→∞ k=0


P
Proof. Suppose ak = s. Then lim rn = lim (s − sn ) = s − s = 0.
k=0 n→∞ n→∞


P 1
4.13 Example. The harmonic series n diverges to +∞.
n=1

1 1 1 1 1
≥ 12 , then lim rn 6= 0. So the positive
P
Since rn = k = n+1 + n+2 + · · · n+n + ··· ≥ n · n+n
k=n+1 n→∞

P 1
term series n diverges to +∞.
n=1

4.14 Theorem. (Comparison Test)



P ∞
P
Let ak and bk be positive-term series. suppose that there exists N ∈ N such that 0 ≤ ak ≤ bk
k=0 k=0
for all k ≥ N .

P ∞
P ∞
P ∞
P
1. If bk converges, then ak converges and ak ≤ bk .
k=0 k=0 k=N k=N

P ∞
P
2. If ak diverges to ∞ then bk diverges to ∞.
k=0 k=0

n
P n
P
Proof. 1. We have ak ≤ bk for all k ≥ N , then sn = ak ≤ σn = bk . Since σn is
k=N k=N

P
convergent, then it is bounded and so sn is bounded above. Therefore the series ak
k=N

53

P
converges and by taking the limit of both sides in the above inequality, we obtain ak ≤
k=N

P ∞
P
bk . Since a finite number of terms has no effect on the convergence of a series, then ak
k=N k=0
converges.

P ∞
P
2. Suppose to the contrary that bk converges, then by first item we obtain ak converges.
k=0 k=0

P ∞
P
But we have ak diverges to ∞. Thus bk diverges to ∞.
k=0 k=0

4.15 Examples.

P 1
1. The harmonic series k = ∞.
k=1
Since ln(1 + x) ≤ x for any x > −1, it follows 0 ≤ ln(1 + k1 ) ≤ k1 for any k ≥ 1. But from the

ln(1 + k1 ) we obtain that the harmonic series diverges
P
divergence of the telescoping series
k=1
to ∞.
∞ ∞
P 1 P 1
2. The series k2
converges and k2
≤ 2.
k=1 k=1
∞ ∞
1 1 P 1 P 1
Since 0 ≤ k2
≤ (k−1)(k) for any k ≥ 2, and (k−1)(k) = 1, then the series k2
converges
k=2 k=1
∞ ∞ ∞
P 1 P 1 P 1
and k2
=1+ k2
≤1+ (k−1)(k) ≤ 2.
k=1 k=2 k=2

3. The comparison test applies only to series with positive terms.


∞ ∞
We have ak = − k1 ≤ bk = k12 ,and the series
P P
bk converges while ak diverges to −∞.
k=1 k=1

P 1
4. The series kp converges if and only if p > 1.
k=1
1
• If p ≤ 0, then lim p 6= 0. Hence the series diverges to ∞.
k→∞ k
• If 0 < p < 1, then 0 ≤ k1 < k1p . Hence the series diverges to ∞.
• If p > 2, then 0 ≤ k1p < k12 . Hence the series is convergent.
• The case 1 < p < 2 will be examined later by Cauchy’s condensation test or integral test.

4.16 Definition. (Absolute And Conditional Convergence)



P ∞
P
The series ak is said to be absolutely convergent if the positive term series |ak | converges.
k=0 k=0
P∞ ∞
P ∞
P
However, if ak converges and |ak | diverges, then the series ak is said to be Conditionally
k=0 k=0 k=0
convergent.

The notion of absolute convergence is useful for series with arbitrary sign.

4.17 Property. (Absolute Convergence Test)



P ∞
P ∞
P
If ak converges absolutely then it also converges and ak ≤ |ak |.
k=0 k=0 k=0

54

P
Proof. Note that 0 ≤ ak + |ak | ≤ 2|ak | for any k ≥ 0. Since |ak | converges, then the comparison
k=0

P ∞
P ∞
P
test tell us that (ak + |ak |) converges. Thus, due to linearity of series, ak = ([ak + |ak |] −
k=0 k=0 k=0
|ak |) is convergent. Since 0 ≤ ak + |ak | ≤ 2|ak | for any k ≥ 0, then by comparison test we obtain

P ∞
P
ak ≤ |ak |.
k=0 k=0

4.18 Remark. (Modified Comparison Test)



P
Let bk be a convergent positive-term series. If there exists N ∈ N such that |ak | ≤ bk for all
k=0

P
k ≥ N , then the series (with arbitrary sign) ak is absolutely convergent and hence convergent.
k=0

4.19 Theorem. (Limit Form Of Comparison Test)



P ∞
P
Let ak and bk be positive-term series with bn > 0 for sufficiently large n.
k=0 k=0

1. If lim an = l > 0, then the series both converge or both diverge.


n→∞ bn
∞ ∞
an P P
2. If lim = 0 and bk converges, then ak converges.
n→∞ bn k=0 k=0
∞ ∞
an P P
3. If lim = ∞ and bk diverges, then ak diverges to ∞.
n→∞ bn k=0 k=0

l
Proof. 1. Choose 0 < ε = 2 < l. By definition of limit we have, ∃n0 ∈ N such that,
an
n ≥ n0 ⇒ | − l| < ε.
bn

Then 0 < 2l bn < an < 3l2 bn for n ≥ n0 . We conclude from comparison test the series both
converge or both diverge.

2. Choose ε = 1. By definition of limit we have, ∃n0 ∈ N such that,


an
n ≥ n0 ⇒ | | < ε.
bn

P
Then an < bn for n ≥ n0 . We conclude from comparison test that the convergence of bk
k=0

P
implies the convergence of ak .
k=0

3. Choose M = 1. By definition of limit we have, ∃n0 ∈ N such that,


an
n ≥ n0 ⇒ > M.
bn

P
Then an > bn for n ≥ n0 . We conclude from comparison test that the divergence of bk
k=0

P
implies the divergence of ak .
k=0

55
4.20 Definition. (Landau Notations)
Let {an } and {cn } be two sequences. We say that

1. {an } is dominated by {cn }, and we write an = O(cn ), if there exist a bounded sequence {bn }
and N ∈ N such that an = bn cn for any n ≥ N .

2. {an } is negligible with respect to {cn }, and we write an = o(cn ), if there exist a sequence {εn }
goes to zero as n goes to infinity and an N ∈ N such that an = εn cn for any n ≥ N .

3. {an } is equivalent to {cn }, and we write an ∼ cn , if there exist a sequence {εn } goes to zero
as n goes to infinity and an N ∈ N such that an = (1 + εn )cn for any n ≥ N . Equivalently;
an ∼ cn if and only if an − cn = o(cn ).

Note that an = O(cn ) respectively ( an = o(cn ), an ∼ cn ), if and only if there exist a se-
quence {µn } such that an = µn cn eventually(after a certain rank) and {µn } is bounded respectively
(µn −−−→ 0, µn −−−→ 1 ).
n→∞ n→∞

4.21 Property. Let {an } and {cn } be two sequences. If {cn } eventually(after a certain rank) never
vanishes, then:

1. an = O(cn ) ⇔ { acnn } is bounded.


an
2. an = o(cn ) ⇔ cn −−−→ 0.
n→∞
an
3. an ∼ cn ⇔ cn −−−→ 1.
n→∞

4.22 Property. Let {an } and {cn } be two sequences.



P ∞
P
1. If an = O(cn ) and ck converges absolutely , then ak converges absolutely.
k=0 k=0

P ∞
P
2. If an = o(cn ) and ck converges absolutely, then ak converges absolutely.
k=0 k=0

3. (Equivalent test)

P ∞
P
If an ∼ cn , then |ak | and |ck | have same nature.
k=0 k=0

Proof. 1. Assume an = O(cn ), there exist a bounded sequence {bn } and N ∈ N such that
an = bn cn for any n ≥ N . Hence there exists M > 0 such that |an | ≤ M |cn | eventually. Since

P ∞
P
|ck | converges, then by comparison test ak converges absolutely.
k=0 k=0

2. Evident. We use, an = o(cn ) ⇒ an = O(cn ).


 
3. Evident. We use an ∼ cn ⇒ an = O(cn ) ∧ cn = O(an ) .

56
We give some examples after alternating series test.

4.23 Property. (Cauchy’s Condensation Test)



P
Let {an }n≥1 be a positive decreasing sequence of real numbers, the series ak converges if and
k=1

2k a2k = a1 + 2a2 + 4a4 + 8a8 + · · · converges.
P
only if
k=0

n n
2k a2k . Since
P P
Proof. Let sn = ak and σn =
k=1 k=0

a2 ≤ a2 ≤ a1 ,
2a4 ≤ a3 + a4 ≤ 2a2 ,
4a8 ≤ a5 + a6 + a7 + a8 ≤ 4a4 ,
··················
2 a2n+1 ≤ a2n +1 + · · · + a2n+1 ≤ 2n a2n ,
n

then by adding the inequalities, we obtain


1
(σn+1 − a1 ) ≤ s2n+1 − a1 ≤ σn .
2
We can conclude that the increasing sequences sn and σn are either both bounded above or both
unbounded above. Then the two series converge or diverge together.

One can easily see that sn ≤ σn for any n ∈ N . Hence in the proof of the property we may obtain
a better inequalities
1
(σn+1 − a1 ) ≤ s2n+1 − a1 ≤ σ2n+1 − a1 .
2

By passing the limit in case of convergence, we obtain 21 σ + 12 a1 ≤ s ≤ σ, where s =
P
ak and
k=1

2k a
P
σ= 2k .
k=0

4.24 Corollary. (Riemann Series)



P 1
The series kp converges for p > 1 and diverges for p ≤ 1.
k=1

Proof. We discuss two cases according to the values of p.



1. If p > 0, then { n1p }n≥1 is a positive decreasing sequence. Hence the series 1
P
kp and the con-
k=1
∞ ∞ ∞
2k (2k1)p = 1 1
P P P
densed series (2p−1 )k
have same nature. But the geometric series (2p−1 )k
k=0 k=0 k=0
converges if and only if 2p−1 < 1, that is, p > 1.

2. If p ≤ 0, then lim k1p 6= 0. So the Riemann series diverges to ∞.


k→∞

P 1
Thus kp = ∞ for p ≤ 1 and converges for p > 1.
k=1

57

P 1
In case of convergence, that is p > 1, the sum is written ζ(p) = np , and known as Riemann’s
n=1
zeta function. Also the inequalities 12 σ + 21 a1 ≤ s ≤ σ ensures that 12 1 1
p−1 + 2 ≤ ζ(p) ≤
1
p−1 .
1−1/2 1−1/2
2
For example 1 ≤ ζ(2) ≤ 2 and 76 ≤ ζ(3) ≤ 34 . One can show that ζ(2) = π6 , but no closed expression
for ζ(3) is known.

4.25 Corollary. (nα − Test)



P
Consider the series ak with terms of arbitrary sign.
k=1

1. If nα an −−−→ ` 6= 0, then
P
ak converges absolutely if and only if α > 1.
n→∞ k=1

2. If nα an −−−→ 0 and α > 1, then
P
ak converges absolutely.
n→∞ k=1

3. If nα an −−−→ ∞ and α ≤ 1, then
P
ak diverges.
n→∞ k=1

Proof. Evident.
|`|
1. nα an −−−→ ` 6= 0 ⇒ |an | ∼ nα .
n→∞

2. nα an −−−→ 0 ⇒ an = o( n1α ).
n→∞

1
3. nα an −−−→ ∞ ⇒ an > nα eventually, by the definition of limit.
n→∞

4.26 Corollary. (Bertrand Series)


∞  
1
P 
The series q
kp ln k converges if and only if p > 1 or p = 1 and q > 1 .
k=2

Proof. We discuss three cases according to the values of p.

1+p 1 1
1. If p > 1, take α ∈]1, p[ for example α = 2 . Hence lim nα = lim = 0,
n→∞ np lnq n n→∞ n(p−α) lnq n
for any q ∈ R. Hence, by the nα − test, the series converges.

1+p 1 1
2. If p < 1, take α ∈]p, 1[ for example α = 2 . Hence lim nα = lim = ∞,
n→∞ np lnq n n→∞ n(p−α) lnq n
for any q ∈ R. Hence, by the nα − test, the series diverges.
q−1
3. If p = 1, we have ( x ln1q x )0 = − ln (x ln(x)(q+ln x)
q
x)2
< 0 for sufficiently large x. Hence, in this case
the general term of Bertrand series is a positive decreasing sequence for sufficiently large n
∞ ∞
P 1 1 P 1
and so the condensed series kq lnq 2 = lnq 2 kq have the same nature as Bertrand series
k=0 k=0

P 1
k lnq k . Thus, for p = 1, Bertrand series converges if and only if q > 1.
k=2

58
4.27 Theorem. (Integral Test) Z ∞
Let f be continuous, positive and decreasing on [k0 , ∞), for k0 ∈ N. Then the integral f (x) dx
k0

P
is convergent if and only if the series f (k) is convergent. Moreover, we have
k=k0


P R∞ ∞
P
f (k) ≤ k0 f (x) dx ≤ f (k).
k=k0 +1 k=k0

Proof. Since f is decreasing on [k0 , ∞), for k ≥ k0 we have

f (k) ≤ f (x) ≤ f (k − 1), ∀x ∈ [k − 1, k],

and as the function f is continuous on [k − 1, k], hence integrable on [k − 1, k],


Z k
f (k) ≤ f (x) dx ≤ f (k − 1).
k−1

Then for all n ∈ N with n > k0 we obtain


n
P Rn n
P n−1
P
f (k) ≤ bn := k0 f (x) dx ≤ f (k − 1) = f (k).
k=k0 +1 k=k0 +1 k=k0

If the integral converges, then the sequence {bn } is bounded. So the n-th partial sum {sn } is
increasing, bounded above and thus the series converges.
If the integral diverges, then from the second inequality we obtain that the series diverges.

Typical applications of the integral test are Riemann series and Bertrand series.

P
Note that, under the assumptions of the above theorem, if f (k) converges then for all n > k0 ,
k=k0
Z ∞ Z ∞ Z ∞ Z ∞
f (x) dx ≤ rn ≤ f (x) dx, and sn + f (x) dx ≤ s := sn + rn ≤ sn + f (x) dx,
n+1 n n+1 n


P 1
4.28 Example. Suppose we want to estimate the precision of the sum of k3
computed up to
R∞ 1 k=1

the first 10 terms.R Since s10 = 1.197532 · · · , and n x3 dx = 2n1 2 , then s10 + 11 f (x) dx ≤ s :=
R

s10 + R10 ≤ s10 + 10 f (x) dx, gives 1.201664 ≤ s ≤ 1.202532. The exact value for s is 1.202057 · · ·

4.2 Alternating Series.

4.29 Theorem. (Leibniz’s Alternating Series Test)



(−1)k bk with bk > 0 converges if the sequence {bk }k≥0 is decreasing and
P
An Alternating series
k=0
lim bk = 0. Moreover, for all n ≥ 0 on has
k→∞

|rn | = |s − sn | ≤ bn+1 and s2n+1 ≤ s ≤ s2n .

59
Proof. Since the sequence {bk }k≥0 is a decreasing sequence,
s2n = s2n−2 − b2n−1 + b2n ≤ s2n−2 and s2n+1 = s2n−1 + b2n − b2n+1 ≥ s2n−1 .
Moreover,
s2n = s2n−1 + b2n ≥ s2n−1 ≥ · · · ≥ s1 and s2n+1 = s2n − b2n+1 ≤ s2n ≤ · · · ≤ s0 .
Therefore {s2n }n≥0 and {s2n+1 }n≥0 are monotone bounded sequences. Thus
lim s2n = inf s2n := r and lim s2n+1 = sup s2n+1 := l.
n→∞ n≥0 n→∞ n≥0

As, r − l = limn→∞ (s2n − s2n+1 ) = limn→∞ b2n+1 = 0, then the series converges and
s2n+1 ≤ s = r = l ≤ s2n for all n ≥ 0.
In addition,
0 ≤ s − s2n+1 ≤ s2n+2 − s2n+1 = b2n+2 and 0 ≤ s2n − s ≤ s2n − s2n+1 = b2n+1 .
So |rn | = |s − sn | ≤ bn+1 for all n ≥ 0.

(−1)k bk with bk > 0 is called an alternating series because the signs of the terms
P
The series
k=0
alternate between + and −.
4.30 Examples.

P (−1)k
1. The series kp converges for p > 0 and diverges for p ≤ 0.
k=1

P (−1)k
2. The series k is conditionally convergent.
k=1

P (−1)k
3. Let us approximate the sum k! to the third digit, that means with a margin less than
k=0
10−3 . The series converges by Leibniz’s Test. As b7 = 50401
< 10−3 , then |s − s6 | ≤ b7 < 10−3 .
But s6 = 0.368056 · · · , the estimate s ≈ 0.368 is correct up to third place.
4. The condition the sequence {bk }k≥0 is decreasing in Leibniz’s test is only sufficient. In fact,
for k ≥ 2 let  1
 k, k even;
bk =
 k−1
k2
, kodd.
∞ ∞
(−1)k k1 +
P 1
(−1)k bk =
P P
We have bk > 0, lim bk = 0. The series k2
converges, for
k→∞ k=2 k=2 k≥3
k odd
the two series to the right converge. But the sequence is not decreasing since bk > bk+1 for k
even and bk < bk+1 for k odd.
∞ n n ∞
(−1)n (−1)n
sin( (−1) (−1)
+ o( n12 ). Since
P P
5. Consider the series n ). We have sin( n ) = n n
n=1 n=1

P 1
is convergent, by alternating series test, and n2
is absolutely convergent, by Riemann
n=1
∞ n
sin( (−1)
P
examples, then the series n ) is convergent. Note that we cannot use equivalent
n=1
(−1)n
test because n does not have constant sign eventually.
∞ ∞
sin( n1 ). We have sin( n1 ) ∼ 1 1
P P
6. Consider the series n ≥ 0 eventually. Since n is divergent,
n=1 n=1

sin( n1 ) is divergent
P
then
n=1

60
4.3 Ratio And Root Test.

4.31 Property. (Ratio Test)


Let an be a sequence such that an 6= 0 eventually.

1. If there exists a real r < 1 such that an+1
P
an ≤ r eventually, then ak converges absolutely.

k=k0

an+1 P
2. If an ≥ 1 eventually, then ak diverges.
k=k0

Proof. Note that, if un and vn are eventually strictly positive and uun+1
n
≤ vn+1
vn , then un = O(vn ).
un+1
Because the sequence vn+1 is decreasing and bounded below by zero, hence convergent and bounded.
 
rn+1
1. Evident. We use, an+1
an ≤ r = ⇒ |an | = O(rn ).

rn

1n+1
2. Evident. We use, an+1
an ≥ ⇒ 1 = O(|an |) ⇒ limn→∞ an 6= 0.

1n

4.32 Theorem. (d’Alembert’s Ratio Test) a


n+1
Let an be a sequence such that an 6= 0 eventually and lim = ` exists, finite or infinite.
n→∞ an

P
1. If ` < 1, the series ak converges absolutely.
k=k0


P
2. If ` > 1, the series ak diverges.
k=k0

3. If ` = 1, the test gives no information.

a
n+1
Proof. 1. If ` < 1, then choose ` < r < 1. Since lim = `, there exists an integer N such
n→∞ an

that an+1
P
< r for all n ≥ N .So we conclude from previous property that the series ak

an
k=k0
converges absolutely.

2. If ` > 1, there exists an integer N such that an+1
an > 1 for all n ≥ N .So we conclude from

P∞
previous property the the series ak diverges.
k=k0

3. The Riemann series for all values of p yields the limit ` = 1.

4.33 Theorem. (Cauchy’s Root Test)


1
Let an be a sequence such that an 6= 0 eventually and lim |an | n = ` exists, finite or infinite.
n→∞

61

P
1. If ` < 1, the series ak converges absolutely.
k=k0


P
2. If ` > 1, the series ak diverges.
k=k0

3. If ` = 1, the test gives no information.

1
Proof. 1. If ` < 1, then choose ` < r < 1. Since lim |an | n = ` , there exists an integer N
n→∞

n rk , the
P
such that |an | < r for all n ≥ N . Hence by comparing with the geometric series
k=k0

P
absolute convergence of ak follows.
k=k0

2. If ` > 1, there exists an integer N such that |an | > 1 for all n ≥ N and hence the n-th term
does not approach zero.

3. The Riemann series for p = 1 and p = 2 yields the limit ` = 1.

4.4 Abel’s Summation By Parts.


There are numerical series which are convergent but not absolutely convergent. These are often
quite subtle to handle. We give some tests which are useful to deal with such series. The basic tool
for these tests is the following Abel’s summation by parts formula.

4.34 Theorem. (Abel’s Summation By Parts Formula)


n
P
Let an and bn be two sequences and An := ak . We then have the identity
k=1

n
P n
P
ak bk = An bn+1 − Am bm+1 − Ak (bk+1 − bk ).
k=m+1 k=m+1

n
P n
P
It follows that ak bk converges if the sequences {An bn+1 } and Ak (bk+1 − bk ) converge.
k=m+1 k=m+1
Comments; Use the observation ∆Ak = Ak+1 − Ak = ak , the identity reads
n
P n
P
(∆Ak )bk = An bn+1 − Am bm+1 − Ak (∆bk ).
k=m+1 k=m+1

A particular case m = 0 of the summation by parts formula reads


n
P n
P
(∆Ak )bk = An bn+1 − Ak (∆bk ).
k=1 k=1

Proof. Let A0 = 0. We have


n
P n
P n
P n
P
ak bk = (Ak − Ak−1 )bk = Ak bk − Ak−1 bk .
k=m+1 k=m+1 k=m+1 k=m+1

62
By shifting index for the second summation on the right,
n
P n−1
P n
P
Ak−1 bk = Ak bk+1 = Am bm+1 − An bn+1 + Ak bk+1 ,
k=m+1 k=m k=m+1

the summation by parts follows.

4.35 Theorem. (Dirichlet’s Test)



P n
P
The series ak bk is convergent if the sequence {An } where An := ak is bounded and {bn } is
k=1 k=1
decreasing to zero.

n
P
Proof. In view of the summation by parts formula, it is sufficient to show {An bn+1 } and Ak (bk+1 −
k=1
bk ) converge. The first sequence {An bn+1 } converges to zero, since it is the product of bounded se-
n
P
quence by another which converges to zero. For the second one Ak (bk+1 −bk ); Assume |An | ≤ M
k=1

P
for all n and for some M > 0. Then |Ak (bk+1 −bk )| ≤ M (bk+1 −bk ). The series (bk+1 −bk ) = −b1
k=1
n
P
is telescoping, then the convergence of Ak (bk+1 − bk ) follows from the comparison test.
k=1

4.36 Theorem. (Dedekind’s Test)



P n
P
The series ak bk is convergent if the sequence {An } where An := ak is bounded, bn −−−→ 0
k=1 k=1 n→∞

P
and |bk+1 − bk | is convergent.
k=1

n
P
Proof. In view of the summation by parts formula, it is sufficient to show {An bn+1 } and Ak (bk+1 −
k=1
bk ) converge. The first sequence {An bn+1 } converges to zero, since it is the product of bounded se-
n
P
quence by another which converges to zero. For the second one Ak (bk+1 −bk ); Assume |An | ≤ M
k=1

P
for all n and for some M > 0. Then |Ak (bk+1 − bk )| ≤ M |bk+1 − bk |. Since the series |bk+1 − bk |
k=1
n
P
is convergent, then the convergence of Ak (bk+1 − bk ) follows from the comparison test.
k=1

4.37 Theorem. (Abel’s Test)



P ∞
P
The series ak bk is convergent if ak is convergent and {bn } is monotone and bounded.
k=1 k=1

n
P
Proof. In view of the summation by parts formula, it is sufficient to show {An bn+1 } and Ak (bk+1 −
k=1
bk ) converge. The first sequence {An bn+1 } converges, since it is the product of convergent sequences.
Pn
For the second one Ak (bk+1 − bk ); Assume |An | ≤ M for all n and for some M > 0 and {bn }
k=1

63

P
is monotone increasing. Then |Ak (bk+1 − bk )| ≤ M (bk+1 − bk ). The series (bk+1 − bk ) =
k=1
n
P
lim bn+1 − b1 is telescoping, then the convergence of Ak (bk+1 − bk ) follows from the comparison
n→∞ k=1
test.

4.5 Rearrangements
The purpose of this section is to expose one of the benefits of working with absolutely convergent
series.


P ∞
P ∞
P
4.38 Definition. If ak is a series. Then bk is a rearrangement of ak , if there is one-
k=1 k=1 k=1
to-one and onto function φ : N 7→ N such that aφ(n) . That is bn = aφ(n) , where φ(1), φ(2), · · ·
bn =
is any sequence of positive integers in which every positive integer appears once and only once.

We want to show that the sum of an absolutely convergent series can be rearranged in an
arbitrary way without changing the sum.


P ∞
P
4.39 Theorem. If ak is absolutely convergent, then every rearrangement bk is also abso-
k=1 k=1

P ∞
P ∞
P ∞
P
lutely convergent. Moreover, ak = bk and |ak | = |bk |.
k=1 k=1 k=1 k=1

∞ ∞
|ak | < 2ε . Let N = max{φ(1), · · · , φ(n0 )},
P P
Proof. Lets := ak . Let ε > 0 and n0 ∈ N such that
k=1 k=n0
then {a1 , a2 , · · · , an0 } ⊂ {b1 , b2 , · · · , bN }. Hence for n > N ,
n n ∞
|ak | < 2ε .
P P P
| ak − bk | ≤
k=1 k=1 k=n0

Consequently,

P n
P ∞
P n
P n
P n
P
| ak − bk | ≤ | ak − ak | + | ak − bk |
k=1 k=1 k=1 k=1 k=1 k=1
∞ ∞
P ε P ε ε ε
< |ak | + 2 ≤ |ak | + 2 ≤ 2 + 2 = ε.
k=n+1 k=n0

P ∞
P ∞
P ∞
P
Thus bk is convergent and ak = bk . Similarly, we can obtain |bk | is convergent and
k=1 k=1 k=1 k=1

P ∞
P
|ak | = |bk |.
k=1 k=1

This theorem is quite useful in finding the sum of a series which is absolutely convergent.
∞ ∞
(−1)n+1 1 2
= π6 . Then by rear-
P P
4.40 Example. The series n 2 is absolutely convergent and n2
n=1 n=1

P (−1)n+1
ranging the terms of the series n2
, we obtain
n=1
∞ ∞ ∞ ∞
P (−1)n+1 P 1 P 1 1 P 1 π2
n2
= n2
−2 (2n)2
= 2 n2
= 12 .
n=1 n=1 n=1 n=1

64

P ∞
P
It can be shown that, if every rearrangement of ak is convergent, then ak is absolutely
k=1 k=1
convergent.

4.41 Theorem. (Riemann Rearrangement Theorem )



P ∞
P
If ak is conditionally convergent, then ak can be rearranged to have any real number as its
k=1 k=1
sum, to be divergent to∞, and to be divergent to−∞.

4.6 Cauchy Product.


∞ ∞ ∞
1 1 3 1
= 56 . Then
P P P
Consider the geometric series, 2k
= 2, 3k
= 2 and 6k
k=1 k=1 k=1

∞ ∞ ∞
P 1 1 P 1 P 1
2k 3k
6= 2k 3k
.
k=1 k=1 k=1

One way to multiply series and preserve the above property is the so-called Cauchy product,
formally

P ∞
P
an bn = (a0 + a1 + a2 + · · · )(b0 + b1 + b2 + · · · )
n=0 n=0
= (a0 b0 ) + (a0 b1 + a1 b0 ) + (a0 b2 + a1 b1 + a2 b0 ) + · · ·

P
= cn ,
n=0

where,
n
P
cn = a0 bn + a1 bn−1 + · · · an b0 = ak bn−k .
k=0


P ∞
P ∞
P n
P
4.42 Definition. Let an and bn be two series. The series cn , where cn := ak bn−k ,
n=0 n=0 n=0 k=0
is their Cauchy product.

In general the cauchy product of two convergent series series may not be convergent. Take
n
(−1)n n 1

P
an = bn = √ n+1
. Then the n-th term of the Cauchy product is cn = (−1) .
k=0 (n−k+1)(k+1)

For k ≤ n, we have (n − k + 1)(k + 1) = − ( n2


− + 1)2 ( n2 k)2 ≤ ( n2 + 1)2 . Hence |cn | ≥ 2(n+1)
n+2 → 2.
It follows that the cauchy product is not convergent.

4.43 Theorem. (Mertens Theorem )



P ∞
P n
P ∞
P
Let an be absolutely convergent, bn be convergent and define cn := ak bn−k . Then cn
n=0 n=0 k=0 n=0
is convergent and we have

P ∞
P ∞
P
an bn = cn .
n=0 n=0 n=0

65

P ∞
P n
P n
P n
P
Proof. Let A = an , B = bn , An = ak , Bn = bk , Cn =: ck and Dn = B − Bn .
n=0 n=0 k=0 k=0 k=0
Note that,

Cn = a0 Bn + a1 Bn−1 + · · · + an B0
n
P 
= a0 (−B + Bn ) + a1 (−B + Bn−1 ) + · · · + an (−B + B0 ) + B ak
k=0
= An B − R n ,

P
where Rn = a0 Dn + a1 Dn−1 + · · · + an D0 . Let α = |an |. Since Dn → 0, then {Dn } is bounded,
n=0

P
say |Dn | < D for all n. Now, for ε > 0 given, there exists N ∈ N such that α = |an | < ε and
n=N
|Dn | < ε for all n > N . Then, for n ≥ 2N we have

|Rn | ≤ (|a0 | + · · · + |an−N |)ε + (|an−N +1 | + · · · + |an |)D ≤ αε + Dε.

Hence Rn → 0. Since An B → AB, then the result follows from passing to the limit in Cn =
An B − Rn .

66
4.7 Exercises
4
Exercise 1. Express uk = in partial fractions and hence find the sum
k(k + 2)
n
X
Sn = uk .
k=1


X
Deduce the nature of the series un .
n=0
Exercise 2. Find the value to which each of the following series converges:
∞ ∞ ∞
X 2 X (−1)n 3n−1 X 1
a) ; b) ; c) .
4n 4n n2 −n
n=0 n=1 n=2

X
Exercise 3. Apply the comparison test, find the nature of the series un , where
n=1

1


 for n even
un = n
1
 √
 for n odd
n
Exercise 4. Determine the nature of the following series:
∞ ∞  ∞ 
n+1 n
 n
X 1 X X n
a) ; b) ; c) .
nn n n+1
n=1 n=1 n=1
Exercise 5. Apply, if possible, the integral test on the following series. If the test cannot be
applied, state why.
∞ ∞ ∞ ∞
X n X en X  πn  X ln(n + 2)
a) ; b) ; c) sin ; d) .
n+2 (en + 1)2 2 n+2
n=1 n=1 n=1 n=0

Exercise 6. Let p ∈ N and a ∈ R satisfies 0 < a < 1. Use the ratio test, prove that the series

X
np an
n=0

converges. Deduce the value of the following limit

lim np an .
n→∞


X 1
Exercise 7. Consider the series un with positive terms un ≥ 0 and α > . Suppose that
2
n=1

X
un converges.
n=1


X
(i) Prove that the series (un )2 is convergent.
n=1
∞  2
X 1
(ii) Prove that the series un − α is convergent
n
n=1

67

X un
(iii) Deduce the nature of the series .

n=1
 
n+2
Exercise 8. Let un = ln . Find the n-th partial sum
n
n
X
Sn = uk .
k=1


X
Deduce the nature of the series un .
n=1
Exercise 9. Determine the nature of the following series:
∞ ∞ ∞ ∞
X arctan n X ln n X 3n − 4n X 3n+1 − n
a) ; b) √ ; c) ; d) .
n n 5n n3n
n=1 n=1 n=0 n=1

X
Exercise 10. Same question for the series un with terms
n=1
n
2n nn

n! 3n + 1
a) un = ; b) un = ; c) un = ; d) un = ;
n! n! (2n)n 4n − 3
1 1 1
e) un = n n ; f) un = √
√ ; g) un = √ ;
n+1− n n+n
(−1)n
 
π
h) un = 1 − ; i) un = sin 3 ; j) un = ln n e−n ;
n 2n 2
 2
n+1 n

1 1
k) un = ; l) un =   ; m) un = .
ln(ln(n + 1)) 1
1+ ln(n+1) 2n
n
Exercise 11. Determine whether each of the following series converges conditionally, converges
absolutely, or diverges.
∞ ∞ ∞
X
n 2n X n!
n
X (−1)n ln n
a) (−1) n ; b) (−1) ; c) ;
e −1 (2n)! n
n=1 n=1 n=1

∞ ∞ ∞ 1
X n X (−1)n X en
d) (−1)n ; e) ; f) (−1)n .
(ln n)2 (sin n)2 arctan n
n=2 n=1 n=1

Exercise 12. Euler proved that



X 1 π2
= .
n2 6
n=1

Deduce the values of the following series:


∞ ∞
X 1 X (−1)n
a) ; b) .
(2n − 1)2 n2
n=1 n=1

Exercise 13. Study the nature of the following series:


∞ √
X
n n
(−1) .
n+1
n=0

68
Exercise 14. Same question for the series
∞  
X 1 2πn
√ cos .
n 3
n=1

Exercise 15. Establish for what values of α ∈ R, the series



X αn+1 2−n
2n + 1
n=0

is convergent.

69

You might also like