Gavriil Paltineanu

Ileana Bucur
Mariana Zamfir

Integral
Calculus
for Engineers
Integral Calculus for Engineers
Gavriil Paltineanu · Ileana Bucur · Mariana Zamfir

Integral Calculus
for Engineers
Gavriil Paltineanu Ileana Bucur
Department of Mathematics and Computer Department of Mathematics and Computer
Science Science
Technical University of Civil Engineering Technical University of Civil Engineering
Bucharest Bucharest
Bucharest, Romania Bucharest, Romania

Mariana Zamfir
Department of Mathematics and Computer
Science
Technical University of Civil Engineering
Bucharest
Bucharest, Romania

ISBN 978-981-19-4792-6 ISBN 978-981-19-4793-3 (eBook)

https://doi.org/10.1007/978-981-19-4793-3

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

The basic ideas of integral calculus have been outlined since antiquity, consenting
in development of methods for computing different areas and volumes.
To understand the basic idea of the notion of integral, let us start from the following
example. To compute the area of the circle portion lying the upper half-plane, we
can proceed as follows: divide this plane set into rectangles of equal widths, with
one of the bases on the axis O x and at least one end of the other base on the curve,
and then add their areas. It is clear that if the number of rectangles is big, then the
sum of their areas approximates well the area of the semicircle.
In other words, the area of the semicircle is the limit toward which the sum of the
areas of the rectangles tends, when their number tends to infinity. This limit is called
integral.
Archimedes (287–212 BC) used various variants of this method to calculate the
areas of the circle, sphere, cone, and so on, as well as the volumes of the sphere,
cone, revolution ellipsoid, and so on.
The methods developed by Archimedes were, two millennia later, the basis of
integral calculus.
At the end of the seventeenth century, Isaac Newton (1643–1727) and Gottfried
Wilhelm von Leibniz (1646–1716) formulated the basic notions and theorems of inte-
gration theory, including the fundamental theorem known as the “Leibniz–Newton
formula” which allows the computation of the integrals using the antiderivatives.
The full clarification of the notion of integral was reached, a century later, by the
contributions of the French mathematician Augustin Cauchy (1789–1857) and the
German mathematician Georg Bernhard Riemann (1826–1866).
The paper is based on the lectures delivered by the authors at the Department of
Mathematics and Computer Science of Technical University of Civil Engineering
Bucharest, Romania.
This book, which is an improved version, in English, of the book in Romanian
[7], includes the following chapters: Indefinite integrals, definite integrals, improper
integrals, integrals depending on parameter, line integrals, double and triple integrals,
and surface integrals.

v
vi Preface

The basic elements of integral calculus are presented, indispensable for students
in higher technical education to successfully approach other theoretical or tech-
nical disciplines, such as: Mechanics, physics, strength of materials, statistics and
dynamics of constructions, theory of elasticity, finite element method, mathematical
geodesy, compensation of measurements and statistics, and so on.
The main concern of the authors was to maintain a balance between accessibility
and scientific rigor.
The book contains many examples, completely solved exercises, drawings
and comments designed to facilitate the understanding of notions, theorems and
computation algorithms.
The book is useful primarily for students in higher technical education, but also
for engineers, for Ph.D. students, as well as for other specialists who use different
types or methods of integration.

Bucharest, Romania Gavriil Paltineanu

March 2022 Ileana Bucur
Mariana Zamfir
Contents

1 Indefinite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The Notion of Primitive Function (Antiderivative)
of a Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Basic Properties of Indefinite Integrals . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Primitives of Rational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Primitives of Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5 Primitives of Irrational Functions. Binomial Integrals . . . . . . . . . . . 21
2 Definite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1 Area of a Curvilinear Trapezoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Darboux Sums. Definition of Definite Integral . . . . . . . . . . . . . . . . . 28
2.3 Integrability of Continuous and Monotonic Functions . . . . . . . . . . . 33
2.4 Riemann Sums. Riemann Criterion for Integrability . . . . . . . . . . . . 34
2.5 Lebesgue Criterion for Integrability . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.6 Properties of Integrable Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7 Area of a Plane Figure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.8 Approximating Definite Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.1 Convergence and Divergence of Improper Integrals . . . . . . . . . . . . . 71
3.2 Convergence Criteria for Improper Integrals . . . . . . . . . . . . . . . . . . . 76
4 Integrals Depending on Parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.1 Proper Integrals Depending on a Parameter . . . . . . . . . . . . . . . . . . . . 87
4.2 Improper Integrals Depending on a Parameter . . . . . . . . . . . . . . . . . 93
4.3 Euler Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.1 Parameterized Paths. Definition of a Curve . . . . . . . . . . . . . . . . . . . . 113
5.2 Rectifiable Paths and Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3 Natural Parameterization of a Curve . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.4 Line Integrals of the First Kind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.5 Line Integrals of the Second Kind . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.6 Independence on the Path of Line Integral of the Second Kind . . . 149
vii
viii Contents

6 Double and Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

6.1 Double Integral. Definition and Properties . . . . . . . . . . . . . . . . . . . . 159
6.2 Basic Properties of the Double Integral . . . . . . . . . . . . . . . . . . . . . . . 165
6.3 Reducing a Double Integral to an Iterated Single Integral . . . . . . . . 166
6.4 Change of Variables in Double Integral . . . . . . . . . . . . . . . . . . . . . . . 173
6.5 Applications of the Double Integral in Geometry
and Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.5.1 Mass of a Lamina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.5.2 Coordinates of the Center of Mass of a Lamina . . . . . . . . . 181
6.5.3 Moments of Inertia of a Lamina . . . . . . . . . . . . . . . . . . . . . . 183
6.6 Riemann–Green Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
6.7 Improper Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.8 Volume of a Space Figure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.9 Triple Integrals. Definition and Basic Properties . . . . . . . . . . . . . . . 199
6.10 Computing Triple Integral. Change of Variables in Triple
Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.11 Applications of the Triple Integral in Geometry
and Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.11.1 Volume of a Space Domain . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.11.2 Mass of a Solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.11.3 Coordinates of the Center of Mass of a Solid . . . . . . . . . . . 211
6.11.4 Moment of Inertia of a Solid . . . . . . . . . . . . . . . . . . . . . . . . . 212
7 Surface Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
7.1 Parameterized Surface Canvases. Definition of a Surface . . . . . . . . 215
7.2 The Area of a Smooth Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.3 Surface Integral of the First Kind . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7.4 Surface Integral of the Second Kind . . . . . . . . . . . . . . . . . . . . . . . . . . 234
7.5 Integral Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Chapter 1
Indefinite Integrals

1.1 The Notion of Primitive Function (Antiderivative)

of a Function

To understand the need to introduce the notion of antiderivative of a function we will

start from a simple problem of mechanics. It is known that the law of the rectilinear
motion of a mobile is given by an equation of the form

s = f (t)

where t represents the time, and s is the distance traveled by the mobile until the
moment t. The derivative f ' (t) represents the instantaneous speed of movement of
the mobile at the moment t.
In practice, we encounter more often the inverse problem: knowing the speed of
movement of the mobile v = v(t), find the law of motion of the mobile, i.e. the
relationship between the distance traveled s to time t, and the time t. Because v(t) =
f ' (t), we are led to the following problem: knowing the derivative f ' (t) = v(t), find
the function f (t). It is obvious that this problem is the inverse of the fundamental
problem of differential calculus (which consists in finding the derivative of a given
function). Here, on the contrary, we are asked to find a function knowing its derivative.

Definition 1.1.1 Let I ⊂ R be an interval and f : I → R. A function F : I → R

is said to be a primitive function (antiderivative) of the function f on the interval
I if F is differentiable function on I and

F ' (x) = f (x), ∀ x ∈ I.

Theorem 1.1.1 If F : I → R is an antiderivative of the function f : I → R, then

for any real constant C, the function G : I → R given by

G(x) = F(x) + C, ∀ x ∈ I

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 1
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_1
2 1 Indefinite Integrals

is also an antiderivative of f on the interval I . Moreover, any another antiderivative

of f on I has this form.

Proof If G = F + C on I , then G ' = F ' + 0 = F ' = f on I , so G is a primitive

function of f on I .
Conversely, let G be another primitive function of f on I and let H = G − F on
I . Obviously, we have:

H ' (x) = G ' (x) − F ' (x) = f (x) − f (x) = 0, ∀ x ∈ I.

Let a ∈ I be an arbitrary fixed interior point. From Lagrange Theorem it follows

that for any x ∈ I , there is a point ξ in the open interval of ends a and x such that:

H (x) − H (a) = H ' (ξ )(x − a) = 0

whence it results that H (x) = H (a).

Definition 1.1.2 The set of all antiderivatives of the function f : I → R is called

the indefinite integral of the function f on the interval I and it is denoted by
∫
f (x) dx.

If F : I → R is an antiderivative of f , then from Theorem 1.1.1, it results that

∫
f (x) dx = F(x) + C , ∀x ∈ I

where C is the set of all real constant functions on I .

Remark 1.1.1 If the function f : I → R admits antiderivatives on I and F : I → R

is such a function, then F ' (x) = f (x) or dF(x) = f (x) dx, whence it results
∫ ∫
f (x) dx = dF(x) = F(x) + C , ∀x ∈ I.

Remark 1.1.2
(1) Chapter 2 will prove that any continuous function on a real interval admits
antiderivatives on that interval (Theorem 2.6.3).
(2) If f ∈ C 1 (I ), then
∫
f ' (x) dx = f (x) + C , ∀x ∈ I.
1.1 The Notion of Primitive Function (Antiderivative) of a Function 3

Next, we recall the table of antiderivatives of the basic elementary functions:

∫
0 dx = C

∫
x a+1
x a dx = + C , x ∈ R, a /= −1
a+1
∫ 
1 ln x + C , x ∈ (0, ∞)
dx = ln|x| + C =
x ln(−x) + C , x ∈ (−∞, 0)
∫
ax
a x dx = + C , x ∈ R, a > 0, a /= 1
ln a
∫
ex dx = ex + C , x ∈ R
∫
sin x dx = − cos x + C , x ∈ R
∫
cos x dx = sin x + C , x ∈ R
∫
sinhx dx = coshx + C , x ∈ R
∫
coshx dx = sinhx + C , x ∈ R
∫ {π }
1
2
dx = tanx + C , x ∈ R\ + kπ ; k ∈ Z
cos x 2
∫
1
dx = −cotx + C , x ∈ R\{kπ ; k ∈ Z}
sin2 x
∫ {π }
tanx dx = − ln|cosx| + C , x ∈ R\ + kπ ; k ∈ Z
2
∫
cotx dx = ln|sinx| + C , x ∈ R\{kπ ; k ∈ Z}
∫ | |
1 1 || x − a ||
dx = ln +C
x 2 − a2 2a | x + a |
  x−a
1
ln x+a + C , x ∈ (−∞, −a) ∪ (a, ∞)
= 1  a−x
2a , a /= 0
2a
ln x+a + C , x ∈ (−a, a)
4 1 Indefinite Integrals

∫
1 1 x
dx = tan−1 + C , x ∈ R, a /= 0
x 2 + a2 a a
∫
1 x
√ dx = sin−1 + C , x ∈ (−a, a), a > 0
a −x
2 2 a
∫ √
1
√ dx = ln x + x 2 + a 2 + C , x ∈ R, a /= 0
x 2 + a2
∫ | √ |
1 | |
√ dx = ln|x + x 2 − a 2 | + C
x −a
2 2
⎧ √
⎨ ln x + x 2 − a 2 + C , x ∈ (a, ∞)
= √ , a > 0.
⎩ ln −x − x 2 − a 2 + C , x ∈ (−∞, −a)

To demonstrate these formulas it is sufficient to derive the functions from the right
side to obtain the functions below the integral sign; e.g.:
( )'
x a+1 1
+C = x a , a /= −1, (tanx + C )' = and so on.
a+1 cos2 x

So for now we know how to calculate the antiderivatives of a small number of

elementary functions that actually appear in the table of derivatives of elementary
functions. The above table of antiderivatives does not exhaust even the class of the
simplest elementary functions, such as, for example, the functions ln x or tan−1 x.
Indeed, so far we have not encountered any function whose derivative is ln x or
tan−1 x.
The problem of integration (antiderivation) is considerably more difficult than
the problem of derivation. The derivation of a function is much facilitated by the
definition of the derivative, which has a constructive (algorithmic) character. We
recall that:
f (x + h) − f (x)
f ' (x) = lim .
h→0 h

E.g.:

 ' (x + h)2 − x 2
x 2 = lim
h→0 h
2xh + h 2
= lim = lim (2x + h) = 2x.
h→0 h h→0

Therefore, starting from the definition, we can find the derivatives of the basic
elementary functions. In the example above the derivative of the function f (x) = x 2
is f ' (x) = 2x.
1.2 Basic Properties of Indefinite Integrals 5

Things are completely different in the case of integration.∫ The definition of the
antiderivative gives us no clue how to calculate, for example ln x dx. To this diffi-
culty is added the fact that, in the case of integration, we do not have the chain
of derivation rules that allow us to start from the derivatives of several functions, to
find the derivatives of their different combinations: sums, products, ratios, composed
functions, inverse functions, and so on.
Finally, we also mention the fact that there are functions that cannot be inte-
grated; more precisely, there are functions that have antiderivatives, but they cannot
be expressed by elementary functions. Here are some examples of such primitive
functions:
∫ x ∫ ∫ √
e sin x
dx, dx, 1 − ε2 · sin2 x dx, |ε| < 1,
x x
∫ ∫ ∫
 1
sin x 2 dx, e−x dx,
2
dx.
ln x

Emphasizing the difficulties of integral computation, we should not be discour-

aged, because, even in the case of integration there are rules and methods, truly less
than in the case of derivation and with a smaller area of applicability, which allow
the computation of antiderivatives of some important classes of functions.
In the following we will present the general computation methods for primitive
functions.

1.2 Basic Properties of Indefinite Integrals

The first property, very useful in applications, is the linearity property. In particular,
it allows us to integrate polynomial functions.

Proposition 1.2.1 (Linearity Property) Let f, g : I → R be two functions that admit

antiderivatives on I and let α, β ∈ R such that α 2 + β 2 > 0. Then α · f + β · g
admits antiderivative on I and
∫ ∫ ∫
(α · f (x) + β · g(x)) dx = α · f (x) dx + β · g(x) dx. (1.1)

Proof The statement results from the linearity property of the derivation operation:

(α · F(x) + β · G(x))' = α · F ' (x) + β · G ' (x)

= α · f (x) + β · g(x), x ∈ I,

where F and G are primitive functions of f and g, respectively.

6 1 Indefinite Integrals
∫
Example 1.2.1 Compute 2x 3 − 4x 2 + 5x − 7 dx.
Taking into account Proposition 1.2.1, Formula (1.1), we have:
∫ ∫ ∫ ∫ ∫
 3
2x − 4x 2 + 5x − 7 dx = 2x 3 dx + −4x 2 dx + 5x dx + −7 dx
∫ ∫ ∫ ∫
= 2 x dx − 4 x dx + 5 x dx − 7 1 dx
3 2

x4 x3 x2
=2· −4· +5· − 7x + C
4 3 2
x4 4x 3 5x 2
= − + − 7x + C .
2 3 2
∫ 2 √
Example 1.2.2 Compute x
− 3 x + x14 − √4x dx.
Applying Proposition 1.2.1 again and taking into account the table of antideriva-
tives of the usual elementary functions we have:
∫ ( ) ∫ ∫ ∫ ∫
2 √ 1 4 1
x −4 dx − 4 x − 2 dx
1 1
− 3x+ 4−√ dx = 2 dx − x 3 dx +
x x x x
x 3 +1 x −4+1 x − 2 +1
1 1

= 2 ln x − 1 + −4· 1 +C
3
+1 −4 + 1 −2 + 1
3√ 1 √
= 2 ln x − x4 − − 8 x + C , x > 0.
3

4 3x 3

Proposition 1.2.2 (The change of variable formula. First method) Let I, J ⊂ R be

two intervals, let F : J → R be an antiderivative of the function f : J → R and
let u : I → J be a differentiable function on I .
Then the function ( f ◦ u) · u ' : I → R admits antiderivatives on I and
∫
f (u(x)) · u ' (x) dx = F(u(x)) + C , ∀x ∈ I. (1.2)

Proof The statement results immediately from the rule of derivation of composed
functions:

(F(u(x)))' = F ' (u(x)) · u ' (x) = f (u(x)) · u ' (x), ∀x ∈ I.

Remark 1.2.1 From Proposition 1.2.2, it results that in order to calculate the
antiderivative of the function we can proceed as follows:
If we make the substitution t = u(x), x ∈ I , then we have dt = du(x) = u ' (x) dx
and further
∫ ∫
f (u(x)) · u ' (x) dx = f (t) dt = F(t) + C = F(u(x)) + C , x ∈ I.
1.2 Basic Properties of Indefinite Integrals 7
∫ ∫
We specify that equality f (u(x)) · u ' (x) dx = f (t) dt is a formal equality,
because the function from the left side is defined on the interval I and the function
from the right side is defined on the interval J .

Example 1.2.3 Compute the following indefinite integrals:

∫ ∫ ∫ ∫
u ' (x) 2x − 1 ' 2
dx, dx, e u(x)
· u (x) dx, 2xex dx,
u(x) x2 − x + 4
∫ ∫
'
cos u(x) · u (x) dx, 3x 2 cos x 3 dx.

If we denote by t = u(x), then dt = u ' (x) dx, and we have:

∫ ∫
u ' (x) dt
dx = = ln|t| + C = ln|u(x)| + C .
u(x) t

In particular, we obtain:
∫
2x − 1 | 2 | 
dx = ln |x − x + 4| + C = ln x 2 − x + 4 + C .
x2 − x + 4

In the same way, we get:

∫ ∫
' 2 2
eu(x)
· u (x) dx = e u(x)
+ C and 2xex dx = ex + C
∫ ∫
'
cos u(x) · u (x) dx = sin u(x) + C and 3x 2 cos x 3 dx = sin x 3 + C .

∫ 1 ∫
Example 1.2.4 Compute x 2 +a 2 dx and
√ 1 dx.
a 2 −x 2
If we denote by t = a , then dx = a dt, and thus
x

∫ ∫ ∫ ∫
1 1 1 1 a 1 1
dx = 2 x dx = dt = dt
x 2 + a2 a +1
2 a2 t2 + 1 a t2 + 1
a
1 −1 1 x
= tan t + C = tan−1 + C , x ∈ R, a /= 0.
a a a
Similarly, we have
∫ ∫ ∫
11 1 1 a
√ dx = /  dx = √ dt
a −x
2 2 a
1 − ax
2 a 1 − t2
∫
1
= √ dt = sin−1 t + C
1 − t2
x
= sin−1 + C , x ∈ (−a, a), a > 0.
a
8 1 Indefinite Integrals
∫ 1
∫ 1
Example 1.2.5 Compute √x 2 −2x+5 dx and √5−2x−x 2
dx.
Using the canonical form of the second degree function, we have:

x 2 − 2 x + 5 = (x − 1)2 + 4 and 5 − 2x − x 2 = 6 − (x + 1)2 .

For the first integral, using the change of variable t = x − 1, we get dt = dx and
∫ ∫ ∫
1 1 1
√ dx = √ dx = √ dt
x2 − 2x + 5 (x − 1) + 4
2 t +4
2
√ √
= ln t + t 2 + 4 + C = ln x − 1 + x 2 − 2x + 5 + C .

Similarly, with the change of variable t = x + 1, we have dt = dx and

∫ ∫ ( )
1 1 t x +1
√ dx = √ dt = sin−1 √ + C = sin−1 √ + C.
5 − 2x − x 2 6 − t2 6 6

Proposition 1.2.3 (The change of variable formula. Second method) Let I, J ⊂ R

be two intervals and let u : I → J be a bijective function with the properties:
u ∈ C 1 (I ), u ' (x) /= 0, ∀x ∈ I , and its inverse u −1 : J → I is a function of C 1 —
class on J . Furthermore, if f : J →  R is' a continuous function and H : J → R is
an antiderivative of the function f · u −1 : J → R, then the function f ◦u : I → R
admits antiderivative on the interval I and
∫
f (u(x)) dx = H (u(x)) + C , ∀x ∈ I. (1.3)

Proof If we compute the derivative of the identity u −1 (u(x)) = x, ∀x ∈ I , it results:

 −1 '
u (u(x)) · u ' (x) = 1, ∀x ∈ I.

Therefore we have:
∫ ∫
 '
f (u(x)) dx = f (u(x)) · u −1 (u(x)) · u ' (x) dx
∫
 '
= f · u −1 (u(x)) · u ' (x) dx.

 '
Since H is an antiderivative of the function f · u −1 , from Proposition 1.2.2 it
results that
∫
 '
f · u −1 (u(x)) · u ' (x) dx = H (u(x)) + C .
1.2 Basic Properties of Indefinite Integrals 9

Remark∫1.2.2 According to Proposition 1.2.3, for computation of the indefinite

integral f (u(x)) dx we proceed as follows:
We make the change of variable t = u(x), x ∈ I , and we accept the following
formal computation:
 '
x = u −1 (t), dx = u −1 (t) dt,

hence
∫ ∫
 '
f (u(x)) dx = f (t) · u −1 (t) dt = H (t) + C = H (u(x)) + C .

∫ 
Example 1.2.6 Compute tan6 x dx, x ∈ − π2 , π2 .
If we denote by t = tan x, then x = tan−1 t and dx = 1+t 1
2 dt.

Further we have:
∫ ∫ ∫ ( )
t6 1
tan x dx =
6
dt = t −t +1−
4 2
dt
1 + t2 1 + t2
t5 t3 tan5 x tan3 x
= − + t − tan−1 t + C = − + tanx − x + C .
5 3 5 3
Proposition 1.2.4 (The integral by parts formula) If f, g : I → R are two functions
of C 1 —class on I , then the functions f · g, f · g ' and f ' · g admit antiderivatives
on I and we have the formula:
∫ ∫
'
f (x) · g (x) dx = f (x) · g(x) − f ' (x) · g(x) dx. (1.4)

Proof According to the differentiation rule for the product of two functions, we
have:
( f · g)' = f ' · g + f · g ' , hence f · g ' = ( f · g)' − f ' · g.
From Proposition 1.2.1 and Remark 1.1.2, it results that:
∫ ∫ ∫
' '
f (x) · g (x) dx = ( f (x) · g(x)) dx − f ' (x) · g(x) dx
∫
= f (x) · g(x) − f ' (x) · g(x) dx.

∫
Example 1.2.7 Compute ln x dx, x ∈ (0, ∞).
If we denote by f (x) = ln x and by g ' (x) = 1, then we have f ' (x) = 1
x
and
g(x) = x. From Proposition 1.2.4, it results that:
∫ ∫ ∫
1
ln x dx = x ln x − x· dx = x ln x − 1 dx = x ln x − x + C .
x
10 1 Indefinite Integrals
∫
Example 1.2.8 Compute tan−1 x dx.
If we denote by f (x) = tan−1 x and by g ' (x) = 1, then it follows f ' (x) = 1
1+x 2
and g(x) = x.
Further we have:
∫ ∫
−1 −1 x
tan dx = x tan x − dx.
1 + x2

For the computation of the last integral we make the change of variable t = 1+ x 2
and we obtain dt = 2x dx, and thus:
∫ ∫
x 1 1 1 1 
dx = dt = ln|t| + C = ln 1 + x 2 + C .
1+x 2 2 t 2 2

Therefore, we have:
∫
1 
tan−1 x dx = xtan−1 x − ln 1 + x 2 + C .
2
∫√ ∫√
Example 1.2.9 Compute the integrals a 2 − x 2 dx and x 2 + a 2 dx.
First, we notice that:
∫ √ ∫ ∫
a2 − x 2 x x
a 2 − x 2 dx = √ dx = a 2 sin−1 − x·√ dx.
a2 − x 2 a a2 − x 2

To compute the last integral we denote by f (x) = x, g ' (x) = √a 2x−x 2 , and we
√
obtain f ' (x) = 1, g(x) = − a 2 − x 2 .
Further, we have:
∫ √ ∫ √
x
x·√ dx = −x a − x +
2 2 a 2 − x 2 dx.
a2 − x 2

Thus, we have:
∫ √ √ ∫ √
x
a 2 − x 2 dx = a 2 sin−1 + x a2 − x 2 − a 2 − x 2 dx,
a

whence it results that:

∫ √ √
x
2 a 2 − x 2 dx = a 2 sin−1 + x a2 − x 2 + C
a

and further:
∫ √
a2 x x √ 2
a 2 − x 2 dx = sin−1 + a − x2 + C .
2 a 2
1.3 Primitives of Rational Functions 11

Similarly, it is shown that:

∫ √ √
a2 x √ 2
x 2 + a 2 dx = ln x + x 2 + a 2 + x + a2 + C .
2 2
∫√ ∫√
Example 1.2.10 Compute x 2 + 4x + 8 dx and −x 2 + 4x − 3 dx.
Using the canonical form of the second degree function, we obtain:
∫ √ ∫ /
x2 + 4x + 8 dx = (x + 2)2 + 4 dx.

If we consider the change of variable t = x + 2, and according to Example 1.2.9,

it results that:
∫ √ ∫ / ∫ √
x 2 + 4x + 8 dx = (x + 2)2 + 4 dx = t 2 + 4 dt
√ t √ 2
= 2 ln t + t2 + 4 + t +4+C
2
√
= 2 ln x + 2 + x 2 + 4x + 8
x + 2 √ 2
+ x + 4x + 8 + C .
2
In the same manner, we have:
∫ √ ∫ /
−x 2 + 4x − 3 dx = 1 − (x − 2)2 dx
1 −1 x − 2 √ 2
= sin (x − 2) + −x + 4x − 3 + C .
2 2

1.3 Primitives of Rational Functions

A rational function is a ratio of two polynomials (polynomial functions), meaning

a function of the form R(x) = Q(x)
P(x)
, x ∈ I , where I ⊂ R is an interval, P, Q are
two real polynomial functions on I and Q(x) /= 0, ∀x ∈ I .
If the degree of the numerator P is greater than or equal to the degree of the
denominator Q, then we can apply Euclid’s division algorithm and we obtain:

P(x) P1 (x)
= C(x) +
Q(x) Q(x)

where C and P1 are polynomial functions and the degree of P1 is less than the degree
of Q. Since C(x) is a polynomial function, it is easy to compute a primitive. Now,
12 1 Indefinite Integrals

one can reduce everything to the computation of a primitive of the proper rational
1 (x)
function PQ(x) , i.e. a rational function which has the property that the degree of P1 is
less than the degree of Q.
If Q(x) has the following decomposition into irreducible polynomials:
 l1  ln
Q(x) = (x − a1 )k1 · . . . · (x − am )km · x 2 + b1 x + c1 · . . . · x 2 + bn x + cn

where ai ∈ R, i = 1, m and b j , c j ∈ R, b2j − 4c j < 0, j = 1, n, then, according

1 (x)
to a basic theorem of Algebra, the ratio PQ(x) can be uniquely written as a sum of
partial fractions of the form:
m ( )
P1 (x) Σ Ai1 Ai2 Aiki
= + + ··· +
Q(x) i=1
x − ai (x − ai )2 (x − ai )ki
( )
Σn
B j1 · x + C j1 B j2 · x + C j2 B jl j · x + C jl j
+ + + ··· + 
j=1
x + bj · x + cj
2
x2 + bj · x + cj
2
x2 + bj · x + cj
lj

where Ai p , B js , C js are real constants that will be determined, p = 1, ki , s = 1, l j ,

i = 1, m, j = 1, n.
1 (x)
Therefore, to compute the primitive of the rational function PQ(x) it is enough to
know how to compute primitives of the following types:
∫ A ∗
∫ Bx+C
1. (x−a) k dx, x / = a, k ∈ N ; 2. dx, b2 − 4c < 0, k ∈ N∗ .
(x 2 +bx+c)k
∫ A ∗
Type 1. (x−a) k dx, x / = a, k ∈ N

∫ ∫
A A
dx, if k = 1
dx = ∫ x−a
(x − a)k
A
(x−a)k
dx, if k /= 1

= ∫A ln|x − a| −k
+ C , if k = 1
A(x − a) dx, if k /= 1

A ln|x − a| + C , if k = 1
=
A (x−a)
1−k

1−k
+ C , if k /= 1.
∫ 1
∫ 4
Example 1.3.1 Compute x−3
dx and (x−2)5
dx.
∫
1
dx = ln|x − 3| + C .
x −3
∫ ∫
4
dx = 4 (x − 2)−5 dx
(x − 2)5
(x − 2)−5+1 1
=4 +C =− + C.
−5 + 1 (x − 2)4
1.3 Primitives of Rational Functions 13
∫ 1
Example 1.3.2 Compute x 2 −a 2 dx, x ∈ R\{±a}, a / = 0.

Because we have the following decomposition into partial fractions

( )
1 1 1 1
= −
x 2 − a2 2a x −a x +a

it follows that:
∫ (∫ ∫ )
1 1 1 1
dx = dx − dx
x 2 − a2 2a x −a x +a
1
= (ln|x − a| − ln|x + a|) + C
2a | |
1 || x − a ||
= ln + C.
2a | x + a |

∫
Type 2. Bx+C
dx, b2 − 4c < 0, k ∈ N∗
(x 2 +bx+c)k
∫
Type 2.1. 1
x 2 +bx+c
dx, b2 − 4c < 0
∫ ∫
1 1
dx =  dx.
x 2 + bx + c x+ b 2
+ 4c−b2
2 4

4c−b2
If we consider the change of variable t = x + b2 and we denote by a 2 = 4
,
then we obtain:
∫ ∫ ( )
1 1 1 −1 t
dx = dt = tan +C
x 2 + bx + c t 2 + a2 a a
( )
2 2x + b
=√ tan−1 √ + C.
4c − b 2 4c − b2
∫ 1
Example 1.3.3 Compute x 2 +2x+5 dx.
Proceeding as above, we have:
∫ ∫ ( )
1 1 1 −1 x + 1
dx = dx = tan + C.
x 2 + 2x + 5 (x + 1)2 + 4 2 2
14 1 Indefinite Integrals
∫
Type 2.2. x 2Bx+C
+bx+c
dx, b2 − 4c < 0
We have successive:
∫ ∫ ∫
Bx + C x + CB B 2x + b + 2C −b
dx = B dx = B
dx
x 2 + bx + c x 2 + bx + c 2 x 2 + bx + c
∫ ∫
B 2x + b 2C − Bb 1
= dx + dx.
2 x + bx + c
2 2 x + bx + c
2

For the first integral, we make the change of variable t = x 2 + bx + c and we

obtain dt = (2x + b) dx, and further:
∫ ∫
2x + b 1
dx = dt = ln|t| + C
x 2 + bx + c t
 2
= ln x + bx + c + C .

The second integral is obviously of Type 2.1.

∫
Example 1.3.4 Compute x 23x+1+x+1
dx.
Proceeding as above, we have:
∫ ∫ ∫
3x + 1 x + 13 3 2x + 1 + 23 − 1
dx = 3 dx = dx
x2 + x + 1 x2 + x + 1 2 x2 + x + 1
∫ ∫
3 2x + 1 1 1
= dx −  dx
2 x2 + x + 1 2 1 2
x + 2 + 43
( )
3  1 2x + 1
= ln x 2 + x + 1 − √ tan−1 √ + C.
2 3 3

∫
Type 2.3. Bx+C
dx, b2 − 4c < 0, k ∈ N∗
(x 2 +bx+c)k
We have:
∫ ∫
Bx + C Bx + C
 k
dx =  k
dx.
x2 + bx + c x+ b 2
+ 4c−b2
2 4

2
If we make the change of variable t = x + b2 and we denote by a 2 = 4c−b
4
, then
it results that:
∫ ∫ ∫
Bx + C B 2t 2C − Bb 1
 k
dx =  k
dt +  k
dt.
x + bx + c
2 2 t +a
2 2 2 t + a2
2

 '
The first indefinite integral can be compute very easy because t 2 + a 2 = 2t. It
is obvious that:
1.3 Primitives of Rational Functions 15

∫  
2t ln t 2 + a 2 + C , if k = 1
 k
dt = (t 2 +a 2 )1−k
t 2 + a2 1−k
+ C , if k /= 1.

For the other integral, using the integral by parts formula, we establish the
following recurrence relation:
∫ ∫ ( ∫ )
1 1 t 2 + a2 − t 2 1 t2
Ik =  k
dt = 2  k
dt = 2
I k−1 −  k
dt .
t 2 + a2 a t 2 + a2 a t 2 + a2

Indeed, if we denote by f (t) = t and g ' (t) = 2 t 2 k , then we get f ' (t) = 1 and
∫ (t +a )
g(t) = 21 2t
dt = − 1
, whence it results that:
(t 2 +a 2 )k 2(k−1)(t 2 +a 2 )
k−1

∫
t2 t 1
 dt = −  + Ik−1 .
t 2 + a2
k
2(k − 1) t 2 + a 2
k−1 2(k − 1)

Furthermore, we have:
( )
1 t 1
Ik = 2 Ik−1 +  − Ik−1
a 2(k − 1) t 2 + a 2
k−1 2(k − 1)

or
∫ ( )
1 1 t 2k − 3
Ik =  dt = 2  + Ik−1 . (1.5)
t 2 + a2
k a 2(k − 1) t 2 + a 2 k−1 2(k − 1)
∫ x+1
Example 1.3.5 Compute dx.
(x 2 +x+1)2
Proceeding as above, we have:
∫ ∫ ∫
x +1 1 2x + 2 1 2x + 1 1
 2
dx =  2
dx =  2
dx + I2
x2 +x +1 2 x2 +x +1 2 x2 +x +1 2

where it was denoted by:

∫ ∫
1 1
I2 =  2
dx =  2
dx.
x2 +x +1 x+ 1 2
+ 3
2 4

For the first integral, we make the change of variable t = x 2 + x + 1 and we find:
∫ ∫
2x + 1 1 1 1
 dx = dt = − + C = − 2 + C.
x2 + x + 1
2 t 2 t x +x +1
16 1 Indefinite Integrals

For the second integral, we make the change of variable t = x + 1

2
and we use
the recurrence relation (1.5), for k = 2. Thus, we have:
∫ ( ∫ )
1 4 t 1 1
I2 =  dt =  + dt
t 2 + 43
2 3 2 t 2 + 43 2 t 2 + 43
( )
2t 4 2t
=  2 3 + √ tan−1 √ +C
3 t +4 3 3 3
( )
2x + 1 4 −1 2x + 1
=  + √ tan √ + C.
3 x2 + x + 1 3 3 3

Finally, we have:
∫
x +1 1 2x + 1
 dx = −  2 +  2
x2 +x +1
2
2 x +x +1 6 x +x +1
( )
2 2x + 1
+ √ tan−1 √
3 3 3
( )
x −1 2 2x + 1
=  2 + √ tan−1 √ + C.
3 x +x +1 3 3 3

The integration of rational functions is well highlighted by the following example:

∫ 7 6 +4x 5 −5x 4 +4x 3 −5x 2 −x
Example 1.3.6 Compute xx 6−2x −2x 5 +3x 4 −4x 3 +3x 2 −2x+1
dx.
It is easy to see that the polynomial at the denominator has the double root x = 1
and then it admits the following decomposition:
 2
x 6 − 2x 5 + 3x 4 − 4x 3 + 3x 2 − 2x + 1 = (x − 1)2 x 2 + 1 .

By polynomial remainder theorem, it follows that:

x 7 − 2x 6 + 4x 5 − 5x 4 + 4x 3 − 5x 2 − x x 5 − x 4 + x 3 − 3x 2 − 2x
= x + 
x 6 − 2x 5 + 3x 4 − 4x 3 + 3x 2 − 2x + 1 (x − 1)2 x 2 + 1
2

thus:
∫
x 7 − 2x 6 + 4x 5 − 5x 4 + 4x 3 − 5x 2 − x
dx
x 6 − 2x 5 + 3x 4 − 4x 3 + 3x 2 − 2x + 1
∫
x2 x 5 − x 4 + x 3 − 3x 2 − 2x
= +  2
dx.
2 (x − 1)2 x 2 + 1

x 5 −x 4 +x 3 −3x 2 −2x
The function admits the following decomposition into partial
(x−1)2 (x 2 +1)
2

rational functions:
1.3 Primitives of Rational Functions 17

x 5 − x 4 + x 3 − 3x 2 − 2x A B Cx + D Ex + F
 = + + 2 + .
(x − 1) x + 1
2 2 2 x − 1 (x − 1) 2 x + 1 x2 + 1
2

To determine the coefficients A, B, C, D, E, F can be done as follows: we

multiply both members of the above equality by (x − 1)2 , we put x = 1 and it
follows B = −1. Then, we pass the term − (x−1) 1
2 to the left side and by using

elementary computations, it results the following equality:

x 4 + x 3 + 2x 2 + x − 1 A Cx + D Ex + F
 = + 2 + .
(x − 1) x + 1
2 2 x − 1 x + 1 x2 + 1
2

Similarly, if we multiply both members of the last equality by x − 1 and making

x = 1, we find A = 1. Next, we pass the term x−1
1
to the left side and we deduce the
identity:

x2 + x + 2 Cx + D Ex + F
 = 2 +
x2 + 1
2 x +1 x2 + 1
2

whence

x 2 + x + 2 = C x 3 + Dx 2 + (C + E)x + D + F.
⎧
⎪
⎪ C = 0
⎨
D = 1
From the last identity we can write the system , which has the
⎪
⎪ C +E = 1
⎩
D+F = 2
solution C = 0, D = 1, E = 1, F = 1.
Therefore, we have:
∫ ∫ ∫
x 5 − x 4 + x 3 − 3x 2 − 2x 1 1
 dx = dx − dx
(x − 1)2 x 2 + 1
2 x −1 (x − 1)2
∫ ∫
1 x +1
+ dx +  dx
x2 + 1 x2 + 1
2

1
= ln|x − 1| + + tan−1 x
∫ x − 1 ∫
1 2x 1
+  2
dx +  2
dx
2 x +1
2 x +1
2

1 1
= ln|x − 1| + + tan−1 x −  2 + I2 .
x −1 2 x +1

From relation (1.5), it results that:

18 1 Indefinite Integrals
∫ ∫
1 x 1 1
I2 =  dx =  2 + dx
x2 + 1
2
2 x +1 2 x2 +1
x 1
=  2 + tan−1 x + C .
2 x +1 2

Finally, we have:
∫
x 7 − 2x 6 + 4x 5 − 5x 4 + 4x 3 − 5x 2 − x
dx
x 6 − 2x 5 + 3x 4 − 4x 3 + 3x 2 − 2x + 1
x2 1
= + ln|x − 1| +
2 x −1
x −1 3
+  2 + tan−1 x + C .
2 x +1 2

1.4 Primitives of Trigonometric Functions

∫
In this section we consider the integrals of the form R(cos x, sin x) dx, where R is
a rational function of two variables, i.e. a function of the form R(u, v) = Q(u,v)
P(u,v)
,P
and Q being polynomials of two variables written as follows:

Σ
m Σ
n Σ
p
Σ
r
P(u, v) = ai j u i v j , Q(u, v) = bks u k v s .
i=0 j=0 k=0 s=0

We suppose that I ⊂ (−π, π ) is an ∫interval and Q(cos x, sin x) /= 0, ∀x ∈ I .

In general, to compute the integral R(cos x, sin x) dx, we consider the substi-
tution t = tan x2 , x ∈ I . Inversing the function, we obtain x = 2 tan−1 t and
dx = 1+t 2
2 dt.

On the other hand, we have:

1 − tan2 x
1 − t2 2 tan x2 2t
cos x = 2
= and sin x = = .
1 + tan2 x
2
1 + t2 1 + tan2 x
2
1 + t2

On account of the above change of variable, we deduce that:

∫ ∫ ( ) ∫
1 − t2 2t 2
R(cos x, sin x) dx = R , · dt = R1 (t) dt
1 + t2 1 + t2 1 + t2

where R1 is a rational function with respect to the variable t.

1.4 Primitives of Trigonometric Functions 19

Remark 1.4.1 The interval I can be replaced by any other interval J on which the
function x → tan x2 is strictly monotonous and Q(cos x, sin x) /= 0, ∀x ∈ J .
∫ 1
Example 1.4.1 Compute 2−sin x
dx, x ∈ (−π, π ).
Putting the change of variable t = tan x2 , it results that:
∫ ∫ ∫
1 1 2 2
dx = · dt = dt
2 − sin x 2 − 1+t 2
2t 1 + t2 2t 2 − 2t + 2
∫ ( )
dt 2 −1 2t − 1
=  2
= √ tan √ +C
t − 21 + 43 3 3
( )
2 2 tan x2 − 1
= √ tan−1 √ + C.
3 3

In the following, we present three particular cases of other changes of variable,

which lead to simple computations of integrals of trigonometric functions.
∫ ∫  ∫
Case 1 R(cos x, sin x) dx = R1 cos2 x, sin2 x dx = R2 (tan x) dx
where R1 and R2 are rational functions.

Moreover, we assume that I ⊂ − π2 , π2 and Q(cos x, sin x) /= 0, ∀x ∈ I .
In this particular case, the substitution is t = tan x, x ∈ I . Inverting the function,
we get x = tan−1 t and dx = 1+t 1
2 dt.

Using the trigonometric formulas

1 1 tan2 x t2
cos2 x = = and sin 2
x = =
1 + tan2 x 1 + t2 1 + tan2 x 1 + t2

it results that:
∫ ∫ ( )
 2 1 t2 1
R1 cos x, sin x dx =
2
R1 , · dt.
1 + t2 1 + t2 1 + t2
∫ 
Example 1.4.2 Compute 2−sin 1
dx, x ∈ − π2 , π2 .
2
x 
If we consider the change of variable t = tan x, x ∈ − π2 , π2 , we obtain:
∫ ∫ ∫
1 1 1 1
dx = · dt = dt
2 − sin2 x t2
2 − 1+t 2 1 + t 2 t 2+2
( ) ( )
1 t 1 tanx
= √ tan−1 √ + C = √ tan−1 √ + C.
2 2 2 2
∫ 
Example 1.4.3 Compute 2−sin1x cos x dx, x ∈ − π2 , π2 .
20 1 Indefinite Integrals

First, we notice that:

∫ ∫ ∫
1 1 1 + tan2 x
dx = dx = dx.
2 − sin x cos x 2 − tan x cos2 x 2 tan2 x − tan x + 2

If we make the change of variable t = tan x, x ∈ − π2 , π2 , it follows that:
∫ ∫ ∫
1 1 + t2 1 1
dx = · dt = dt
2 − sin x cos x 2t − t + 2 1 + t
2 2 2t − t + 2
2
∫
1 1
=  dt
2 1 2
t − 4 + 15
( 16
)
1 4 4t − 1
= · √ tan−1 √ +C
2 15 15
( )
2 4 tan x − 1
= √ tan−1 √ + C.
15 15

∫ ∫
Case 2 R(cos x, sin x) dx = R1 (sin x) cos x dx, x ∈ (−π, π )
where R1 is a rational function.
Here we use the substitution sin x = t. Then cos x dx = dt and the integral
becomes:
∫ ∫ ∫
R(cos x, sin x) dx = R1 (sin x) cos x dx = R1 (t) dt.

∫
Example 1.4.4 Compute cos3 x dx.
If we make the change of variable t = sin x, then dt = cos x dx, and further:
∫ ∫ ∫

cos3 x dx = cos2 x cos x dx = 1 − sin2 x cos x dx
∫
 t3 sin3 x
= 1 − t 2 dt = t − + C = sin x − + C.
3 3

∫ ∫
Case 3 R(cos x, sin x) dx = R1 (cos x) sin x dx, x ∈ (−π, π )
where R1 is a rational function.
In this case, it is recommended the following change of variable: cos x = t. As
sin x dx = −dt, we have:
∫ ∫ ∫
R(cos x, sin x) dx = R1 (cos x) sin x dx = − R1 (t) dt.
1.5 Primitives of Irrational Functions. Binomial Integrals 21
∫
Example 1.4.5 Compute sin5 x dx.
If we make the change of variable cos x = t, then sin x dx = −dt and we have:
∫ ∫ ∫
 2
sin5 x dx = sin4 x sin x dx = 1 − cos2 x sin x dx
∫ ∫
 2  2t 3 t5
=− 1 − t 2 dt = − 1 − 2t 2 + t 4 dt = −t + − +C
3 5
3 5
2 cos x cos x
= − cos x + − + C.
3 5

1.5 Primitives of Irrational Functions. Binomial Integrals

∫
The binomial integrals have the form x m (ax n + b) p dx, where a, b ∈ R, a /= 0
and m, n, p ∈ Q.
The Russian mathematician Pafnuty Lvovich Chebyshev (1821–1894) showed
that these integrals can be reduced to the computation of the integrals of rational
functions, only in the following three cases:
Case 1 p ∈ Z.
If we denote by r the common denominator of the numbers m and n, and make
the natural substitution x = t r , we obtain:
∫ ∫
 
x m ax n + b dx = t mr at nr + b r t r −1 dt.
p p

Since m r ∈ Z and n r ∈ Z, it follows that the function under the integral sign is
a rational function.
∫
Example 1.5.1 Compute √ √41 6 dx, x ∈ (0, ∞).
x ( x+2)
The integral can be written as follows:
∫ ∫ −6
1
x− 2 x 4 + 2
1 1
√ √ 6
dx = dx.
x x +2
4

Thus, we have m = − 21 , n = 1
4
and p = −6 ∈ Z.
22 1 Indefinite Integrals

Therefore r = 4 and we make the substitution x = t 4 . It results dx = 4t 3 dt and

∫ ∫ ∫
1 4t 3 t
√ √ dx = 2 (t + 2)6
dt = 4 dt
x x +2
4 6
t (t + 2)6
∫ ∫ ∫
t +2−2 1 1
=4 dt = 4 dt − 8 dt
(t + 2) 6
(t + 2) 5
(t + 2)6
1 8 1
=− + · +C
(t + 2) 4 5 (t + 2)5
1 8
= − √ 4
+ √ 5
+ C.
4
x +2 5 x +2
4

Case 2 m+1n
∈ Z and p ∈
/Z
In this case, we make the substitution ax n + b = t r , where r is the denominator
of the number p. We get:
( ) n1 ( )1
tr − b r t r − b n −1 r −1
x= , dx = t dt
a na a

and further
∫ ∫ ( ) mn ( ) n1 −1
 r tr − b tr − b
t r −1 dt
p
x ax n + b
m
dx = t rp
na a a
∫ ( ) m+1
n −1
r tr − b
= t r ( p+1)−1 dt.
na a

Since m+1
n
− 1 ∈ Z and r p ∈ Z, it results that the function under the integral sign
is a rational function with respect to t.
∫
Example 1.5.2 Compute / 1 dx, x ∈ (−∞, −1) ∪ (1, ∞).
x (x 2 −1)
3

The integral can be written as follows:

∫ ∫
1  − 23
/ dx = x −1 x 2 − 1 dx.
3
x x2 − 1

Therefore, we have m = −1, n = 2, p = − 23 .

1.5 Primitives of Irrational Functions. Binomial Integrals 23

Since m+1
n√
= 0 ∈ Z and p = − 23 ∈ / Z, let us make the substitution x 2 − 1 = t 2 .
We get x = t 2 + 1, dx = √t 2t +1 dt, thus the last integral becomes
∫ ∫
1 2 t
− 21 −3
/ dx = t +1 · t 1 dt

x x2 − 1
3
t2 + 1 2
∫ ∫ ∫
1 1 1
=  dt = dt − dt
t2 t2 + 1 t2 t2 + 1
1
= − − tan−1 t + C
t
1 √
= −√ − tan−1 x 2 − 1 + C .
x2 − 1

Case 3 m+1n
+ p ∈ Z and m+1 n
∈
/ Z, p ∈ /Z
It can be shown, that if we make the substitution axx n+b = t r , x /= 0, where r is
n

the denominator of p, the problem is reduced to the computation of the primitive

function of a rational function.
∫
Example 1.5.3 Compute / 1 dx, x ∈ (1, ∞).
x 2 (x 2 −1)
3

The integral can be written as follows:

∫ ∫
1  − 23
/ dx = x −2 x 2 − 1 dx.
3
x2 x2 −1

Therefore, we have m = −2, n = 2, p = − 23 . Since m+1 n

+ p = −2 ∈ Z, then
the obvious substitution is x = 1−t 2 . Thus it follows dx =
√ 1 / t
dt, and further
(1−t 2 )3
∫ ∫
1   3 t
/ dx = 1 − t 2 1 − t 2 2 t −3 ·  3 dt

x2 x2 − 1
3
1 − t2 2
∫ ∫ ∫
1 − t2 1 1
= 2
dt = 2
dt − dt = − − t + C
t t t
√
x x −1
2
= −√ − + C.
x2 − 1 x
Chapter 2
Definite Integrals

2.1 Area of a Curvilinear Trapezoid

The notion of definite integral appeared out of necessity to solve some important
problems in geometry and physics, related mainly to the computation of areas and
volumes, respectively to the computation of the mass of an inhomogeneous body
and of mechanical work done by a variable force, and so on.
In the following, we will study the problem of computation the area of a plane
figure bounded by a closed curve. Elementary geometry teaches us to calculate only
the areas of plane figures bounded by segments of straight lines and by arcs of circles.
The problem of finding the area of a plane region bounded by a closed curve remains
open.
It is easily seen that a plane figure bounded by a closed curve can be divided up
into a finite number of “curvilinear trapezoids” (Fig. 2.1), using several straight lines
parallel to the coordinate axes. A curvilinear trapezoid is a plane figure bounded
from three parts by straight lines, the two of them are parallel to one of axis and the
third is perpendicular to them, and the fourth side of the trapezoid is the graph of a
function (Fig. 2.2).
More precisely, if f : [a, b] → R+ is a continuous positive function, then the set:
{ }
Γf = (x, y) ∈ R2 ; a ≤ x ≤ b, 0 ≤ y ≤ f (x)

is a curvilinear trapezoid.
Let Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b be an arbitrary partition
of the interval [a, b]. Let us denote by m i the greatest lower ⎡bound, respectively
⎤
by Mi the least upper bound of the function f on the interval xi−1 , xi , i = 1, n.
Furthermore, we consider the sums:

v
n v
n
sΔ = m i (xi − xi−1 ) and SΔ = Mi (xi − xi−1 ).
i=1 i=1

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 25
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_2
26 2 Definite Integrals

Fig. 2.1 The division of a

plane figure bounded by a
closed curve in “curvilinear
trapezoids”

Fig. 2.2 A curvilinear

trapozid

We notice that sΔ is the sum of the areas of “inscribed rectangles”:

⎡ ⎤
di = xi−1 , xi × [0, m i ] (Fig. 2.3)

while SΔ represents the sum of the areas of “circumscribed rectangles”:

⎡ ⎤
Di = xi−1 , xi × [0, Mi ] (Fig. 2.4).
∪n ∪n
If we denote by E Δ = i=1 di and by FΔ = i=1 Di , then E Δ ⊂ Γf ⊂ FΔ and

sΔ = area(E Δ ), SΔ = area(FΔ ).

From Figs. 2.3 and 2.4, we notice that:

sΔ ≤ area(Γf ) ≤ SΔ .

If a partition Δ' is a refinement of the partition Δ, i.e. Δ' contains the points of
the partition Δ and maybe some addition other points, then:

sΔ ≤ sΔ' ≤ area(Γf ) ≤ SΔ' ≤ SΔ .

We know intuitively that:

area(Γf ) = sup sΔ = inf SΔ .

Δ Δ

Firstly, we specify that for the time being we do not have a rigorous definition of
the notion of area of the set Γf , but only an intuitive concept. On the one hand, the
2.1 Area of a Curvilinear Trapezoid 27

Fig. 2.3 The sum of the y

areas of “inscribed
rectangles”

mi

x
O a x1 x i –1 x i b

Fig. 2.4 The sum of the y

areas of “circumscribed
rectangles”

Mi

x
O a x1 x i –1 x i b

above considerations suggest to introduce a new mathematical concept, namely the

definite integral:

∫b
def
f (x) dx = sup sΔ = inf SΔ ,
Δ Δ
a

and, on the other hand, to define:

∫b
area(Γf ) = f (x) dx.
a

Next, we will rigorously present these fundamental notions in mathematical

analysis.
28 2 Definite Integrals

2.2 Darboux Sums. Definition of Definite Integral

Definition 2.2.1 A partition of the compact interval [a, b] is any finite subset of
distinct points having the form:

Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b.

⎤ Δ is denoted by ∥Δ∥, and it is the maximum length of

The norm of ⎡the partition
the subintervals xi−1 , xi , i = 1, n:

∥Δ∥ = max (xi − xi−1 ).

1≤i≤n

The partition Δ' is said to be a refinement of the partition Δ, and denoted by

Δ ≺ Δ' , if Δ' contains the points of the partition Δ and maybe some additional other
points.
∥ ∥
Obviously, if Δ ≺ Δ' , then ∥Δ' ∥ ≤ ∥Δ∥.
Example 2.2.1
(1) Δ : a = x0 < x1 = b, ∥Δ∥ = b − a;
(2) Δ : a = x0 < x1 = a+b
2
< x2 = b, ∥Δ∥ = b−a 2
;
(3) Δ : a = x0 < x1 < . . . < xi < . . . < xn = b, where xi = a + i b−a
n
, i = 0, n,
∥Δ∥ = b−a
n
.
Furthermore, for any bounded function f : [a, b] → R, we shall denote by:

m = inf{ f (x); x ∈ [a, b]}, M = sup{ f (x); x ∈ [a, b]}

{ ⎡ ⎤} { ⎡ ⎤}
m i = inf f (x); x ∈ xi−1 , xi , Mi = sup f (x); x ∈ xi−1 , xi , 1 ≤ i ≤ n.

It is easily seen that the following inequalities hold:

m ≤ m i ≤ Mi ≤ M, ∀1 ≤ i ≤ n. (2.1)

Definition 2.2.2 We define the lower Darboux sum of the function f with respect
to the partition Δ by

v
n
sΔ = m i (xi − xi−1 ),
i=1

and the upper Darboux sum of the function f with respect to the partition Δ by

v
n
SΔ = Mi (xi − xi−1 ).
i=1
2.2 Darboux Sums. Definition of Definite Integral 29

Geometrically, the Darboux sums represent the shaded areas in Figs. 2.5 and 2.6.
From (2.1), it results that for any partition Δ of the interval [a, b], we have:

m · (b − a) ≤ sΔ ≤ SΔ ≤ M · (b − a). (2.2)

Lemma 2.2.1 For any two partitions Δ and Δ' of the interval [a, b] such that
Δ ≺ Δ' , we have the inequalities sΔ ≤ sΔ' ≤ SΔ' ≤ SΔ .
Proof Let us consider an arbitrary partition of the interval [a, b]:

Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b.

We assume that the partition Δ' of [a, b] contains the points of the partition Δ and
only one addition point, namely the point c, situated between xi−1 and xi (i fixed).
Let us denote by:
{ ⎡ ⎤}
m i' = inf f (x); x ∈ xi−1 , c and m i'' = inf{ f (x); x ∈ [c, xi ]}.

Fig. 2.5 Lower Darboux

sum

Fig. 2.6 Upper Darboux

sum
30 2 Definite Integrals

Since m i ≤ m i' and m i ≤ m i'' , it results that:

sΔ' − sΔ = m i' (c − xi−1 ) + m i'' (xi − c) − m i (xi − xi−1 )

≥ m i (c − xi−1 + xi − c) − m i (xi − xi−1 ) = 0.

Therefore, we proved that sΔ ≤ sΔ' .

Obviously, if we assume that the division Δ' has all the points of the partition Δ
and some addition distinct points c1 , . . . , c p , the proof is similar.
The proof of inequality SΔ' ≤ SΔ is very similar and the details are left as an
exercise.

Lemma 2.2.2 For any two arbitrary partitions Δ' and Δ'' of the interval [a, b] we
have sΔ' ≤ SΔ'' .

Proof Let Δ = Δ' ∪ Δ'' be the partition of [a, b] that includes every point that is
either in Δ' , or in Δ'' (or in both). Obviously, we have Δ' ≺ Δ and Δ'' ≺ Δ. From
Lemma 2.2.1, it follows that:

sΔ' ≤ sΔ ≤ SΔ ≤ SΔ'' .

From the inequalities (2.2), we deduce that the set {sΔ }Δ of lower Darboux sums
is upper bounded by the real number M · (b − a) and the set {SΔ }Δ of upper Darboux
sums is lower bounded by the real number m · (b − a).
Let us denote by

I∗ = sup sΔ
Δ

the lower Darboux integral of the function f on [a, b], and by

I ∗ = inf SΔ
Δ

the upper Darboux integral of the function f on [a, b].

Lemma 2.2.3 I∗ ≤ I ∗ .

Proof From Lemma 2.2.2, it results that sΔ' ≤ SΔ'' , whatever the partitions Δ' and
Δ'' . Firstly, if the partition Δ'' is fixed, we get I∗ = supΔ' sΔ' ≤ SΔ'' .
Since Δ'' is arbitrary, therefore we have I∗ ≤ inf Δ'' SΔ'' = I ∗ .

Definition 2.2.3 A bounded function f : [a, b] → R is said to be (D)—integrable

(integrable in the Darboux sense) on [a, b] if I∗ = I ∗ = I .
∫b
The common value I is denoted a f (x) dx, and this number is called the definite
integral of the function f on the interval [a, b]; the real number I is uniquely
determined by the function f and by the interval [a, b].
2.2 Darboux Sums. Definition of Definite Integral 31

Example 2.2.2 Any constant function f : [a, b] → R, f (x) = c, ∀x ∈ [a, b], is

(D)—integrable on [a, b].
Indeed, if f (x) = c, ∀x ∈ [a, b], then sΔ = SΔ = c · (b − a), for any partition
Δ of the interval [a, b], thus:

∫b
∗
I∗ = I = c · (b − a) = c dx.
a

Lemma 2.2.4 For any ε > 0, there is δε > 0 such that for any partition Δ of the
interval [a, b], with ∥Δ∥ < δε , we have:

I∗ − ε ≤ sΔ ≤ SΔ ≤ I ∗ + ε. (2.3)

Proof We will prove the inequality I∗ − ε ≤ sΔ . The proof of the other inequality
from (2.3) is left as an exercise to the readers.
Since I∗ = supΔ sΔ , it results that for any ε > 0, there is a partition Δ0 of [a, b]
such that
ε
I∗ − ≤ s Δ0 .
2
Let us suppose that Δ0 : a = c0 < c1 < . . . < ck−1 < ck < . . . < c p = b.
Let μ = min1≤k≤ p (ck − ck−1 ) and let Δ : a = x0 < x1 < . . . < xi < . . . <
xn = b be a ⎡partition⎤of [a, b], with ∥Δ∥ < μ. If we denote by Δ = Δ ∪ Δ0 , then in
the interval xi−1 , xi there is at most one point ck of the partition Δ0 . If we assume
that xi−1 < ck < xi , then we denote by:
{ ⎡ ⎤}
m i' = inf f (x); x ∈ xi−1 , ck and m i'' = inf{ f (x); x ∈ [ck , xi ]}.

Furthermore, we have:

m i' (ck − xi−1 ) + m i'' (xi − ck ) − m i (xi − xi−1 )

( ) ( )
= m i' − m i (ck − xi−1 ) + m i'' − m i (xi − ck )
≤ (M − m)(xi − xi−1 ) ≤ (M − m)∥Δ∥.

Since the partition Δ has at most p − 1 interior points ck , it follows that:

sΔ − sΔ ≤ ( p − 1)(M − m)∥Δ∥. (2.4)

32 2 Definite Integrals
{ }
ε
Let δε = min μ, 2( p−1)(M−m) , let Δ be a partition of the interval [a, b], with
∥Δ∥ < δε and Δ = Δ ∪ Δ0 . Since δε ≤ μ, it results that ∥Δ∥ < μ, and according
to (2.4), we have:
ε ε
sΔ − sΔ ≤ ( p − 1)(M − m)∥Δ∥ ≤ ( p − 1)(M − m) = .
2( p − 1)(M − m) 2

Therefore, we have:
ε ε
I∗ − ≤ s Δ0 ≤ s Δ ≤ s Δ + ,
2 2
hence

I ∗ − ε ≤ sΔ .

Theorem 2.2.1 (Darboux criterion for integrability) Let f : [a, b] → R be a

bounded function. The necessary and sufficient condition that the function f to
be (D)—integrable on the interval [a, b] is that for every ε > 0 there is δε > 0
such that for any partition Δ of [a, b], with the property ∥Δ∥ < δε the following
inequality is verified

SΔ − sΔ < ε.

Proof First, suppose that the function f is (D)—integrable on [a, b]. According to
Definition 2.2.3, we have I∗ = I ∗ = I . From Lemma 2.2.4, it results that for every
ε > 0 there is δε > 0 such that for any partition Δ, with ∥Δ∥ < δε , we have
ε ε
I− ≤ sΔ ≤ SΔ ≤ I + .
2 2
Obviously, from the last inequalities we deduce that:
( ε) ( ε)
SΔ − sΔ < I + − I− = ε.
2 2
Conversely, suppose that for every ε > 0 there is δε > 0 with the property that
for any partition Δ, with ∥Δ∥ < δε , we have SΔ − sΔ < ε.
Then, since sΔ ≤ I∗ ≤ I ∗ ≤ SΔ , it results that

0 ≤ I ∗ − I∗ ≤ SΔ − sΔ < ε.

As ε > 0 is arbitrary, it follows that I ∗ − I∗ = 0, hence that f is (D)—integrable

on [a, b].
2.3 Integrability of Continuous and Monotonic Functions 33

2.3 Integrability of Continuous and Monotonic Functions

Theorem 2.3.1 A continuous function f : [a, b] → R on the compact interval [a, b]

is (D)—integrable on [a, b].

Proof Let Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b be an arbitrary
partition of the interval [a, b]. Taking into account to Theorem 5.6.2 (see Differential
Calculus, Part I), since the function f is continuous on the compact interval [a, b],
then f is bounded and it ⎡ attains⎤its minimum and ⎡ maximum ⎤ values, m i and Mi , on
every compact interval xi−1 , xi . Let μi , ηi ∈ xi−1 , xi such that m i = f (μi ) and
Mi = f (ηi ).
On the other hand, from Theorem 5.6.3 (see Differential Calculus, Part I), it results
that f is uniformly continuous on [a, b], thus, | for every | ε > 0 there is δε > 0 such
that for all x ' , x '' ∈ [a, b], with the property |x ' − x '' | < δε , it follows that:
| ( ') ( )| ε
| f x − f x '' | < .
b−a

If we suppose ∥Δ∥ < δε , it results that |ηi − μi | ≤ xi − xi−1 ≤ ∥Δ∥ < δε ,

whence

v
n
ε v
n
SΔ − sΔ = ( f (ηi ) − f (μi )) · (xi − xi−1 ) < (xi − xi−1 )
i=1
b − a i=1
ε
= · (b − a) = ε,
b−a

and therefore Theorem 2.2.1 implies that f is (D)—integrable on the interval [a, b].

Theorem 2.3.2 A monotonic function f : [a, b] → R on the compact interval [a, b]

is (D)—integrable on [a, b].

Proof We present the proof for the case when the function f is monotonic
increasing and non-constant. If f is a constant function, then f is (D)—integrable
(Example 2.2.2).
Let us suppose that f is monotonic increasing and non-constant on [a, b]. Then
we have f (a) < f (b).
Let Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b be an arbitrary partition
of [a, b]. Since f is increasing, it results that m i = f (xi−1 ) and Mi = f (xi ), hence
we get:

v
n
SΔ − sΔ = ( f (xi ) − f (xi−1 ))(xi − xi−1 ).
i=1

For any ε > 0, we denoted by δε = f (b)−ε f (a) . If we suppose that ∥Δ∥ < δε , then
the upper and lower Darboux sums of f satisfy:
34 2 Definite Integrals

v
n
SΔ − sΔ < δε · ( f (xi ) − f (xi−1 ))
i=1
ε
= · ( f (b) − f (a)) = ε.
f (b) − f (a)

Therefore, we proved that for any ε > 0 there is δε > 0 such that for any partition
Δ of [a, b], with ∥Δ∥ < δε , it follows that SΔ − sΔ < ε.
From Theorem 2.2.1, we deduce that f is (D)—integrable on the interval [a, b].

2.4 Riemann Sums. Riemann Criterion for Integrability

Definition 2.4.1 Let f : [a, b] → R be a given real function on the interval [a, b],
let Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b be an arbitrary
partition
⎡ ⎤of [a, b], and let ξi be an arbitrary intermediate point in the subinterval
xi−1 , xi , i = 1, n.
The Riemann sum for the function f , associated to the partition Δ and the chosen
intermediate points ξi is denoted by σΔ ( f ; ξ ), and it is defined as:

v
n
σΔ ( f ; ξ ) = f (ξi )(xi − xi−1 )
i=1

where ξ is the vector ξ = (ξ1 , ξ2 , . . . , ξn ).

Furthermore, suppose that f is a bounded and positive function on [a, b]. If we

denote by m i the greatest lower bound,
⎡ respectively,
⎤ by Mi the least upper bound of
the function f on the subinterval xi−1 , xi , i = 1, n, then:
⎡ ⎤
m ≤ m i ≤ f (ξi ) ≤ Mi ≤ M, ∀ξi ∈ xi−1 , xi and

m · (b − a) ≤ sΔ ≤ σΔ ( f ; ξ ) ≤ SΔ ≤ M · (b − a),

where m = inf{ f (x); x ∈ [a, b]} and M = sup{ f (x); x ∈ [a, b]}.
On the other hand, geometrically, σΔ ( f ; ξ ) is the sum of the areas of the hatched
rectangles in Fig. 2.7 (rectangles with height f (ξi ), 1 ≤ i ≤ n and width xi − xi−1 ),
which approximates the area of the curvilinear trapezoid
{ }
Γf = (x, y) ∈ R2 ; a ≤ x ≤ b, 0 ≤ y ≤ f (x) .

We notice that the approximation is better if we refine Δ.

Intuitively, we are led to believe that if ∥Δ∥ → 0, then σΔ ( f ; ξ ) “converges” to
a finite number I , whatever the intermediate points ξi are.
2.4 Riemann Sums. Riemann Criterion for Integrability 35

Fig. 2.7 Riemann sum

In other words, “there exists” lim∥Δ∥→0 σΔ ( f ; ξ ) = I . We notice that this limit is

not an ordinary limit, because the sum σΔ ( f ; ξ ) is not a function of ∥Δ∥. Indeed, for
a value of the “independent variable” ∥Δ∥, there are an infinite number of partitions
Δ' which have the same norm as Δ, and implicitly there are an infinity of values for
σΔ ( f ; ξ ). To this we must add the fact that we can choose the points ξi in an infinite
number of ways.
We are led to consider a notion of limit in a large sense. The exact sense of the
limit lim∥Δ∥→0 σΔ ( f ; ξ ) = I is given by Definition 2.4.2.

Definition 2.4.2 It is said that the function f is (R)—integrable (integrable in the

Riemann sense) on [a, b] if there is a finite real number I such that for any ε > 0
there is δε > 0 with⎡the property
⎤ that for any partition Δ of [a, b], with ∥Δ∥ < δε
and any points ξi ∈ xi−1 , xi , i = 1, n, we have:

|σΔ ( f ; ξ ) − I | < ε.

In the following, the real number I is called the definite integral (in the sense of
Riemann) of the function f on [a, b] and is denoted by:

∫b
I = f (x) dx = lim σΔ ( f ; ξ ).
∥Δ∥→0
a

Remark 2.4.1 In a less precise writing, in which we do not go into details regarding
∑npartition Δ and the choice of intermediate∑
the points ξi , the Riemann sum σΔ ( f ; ξ ) =
b
∑i=1 f (ξi )(x i − x i−1 ) can be noted as well a f (x) Δx. Replacing the Greek letter
by the Latin letter S, the Riemann sum can be∫written as Sab f (x) Δx.
(The limit of this expression ∫b
is denoted
) by , i.e. a kind of elongated form
S lim∥Δ∥→0 Sa f (x) Δx = a f (x) dx . This is the historical origin of notation
b
∫b
a f (x) dx.
36 2 Definite Integrals

Theorem 2.4.1 An (R)—integrable function f : [a, b] → R on the compact interval

[a, b] is bounded on [a, b].

Proof By hypothesis, if f is (R)—integrable on [a, b], then there is I ∈ R such that

for ε = 1, there is δ1 > 0 with the property that for any partition Δ, with ∥Δ∥ < δ1
and for any points ξi we have:

I − 1 < σΔ ( f ; ξ ) < I + 1. (2.5)

Let us consider the partition⎡Δ : a =⎤ x0 < x1 < . . . < xi−1 < xi < . . . < xn =
b, with ∥Δ∥ < δ1 and let ξi ∈ xi−1 , xi , i = 1, n, be some arbitrary points. Let us
⎡ ⎤
fix a number j = 1, n and let us consider an arbitrary point ξ 'j ∈ x j−1 , x j and the
( )
vector ξ ' = ξ1 , ξ2 , . . . , ξ j−1 , ξ 'j , ξ j+1 , . . . , ξn . We have:
|( ( ) ( ))( )| | ( ) ( )|
| f ξ j − f ξ ' x j − x j−1 | = |σΔ f ; ξ j − σΔ f ; ξ ' |
j j
≤ (I + 1) − (I − 1) = 2

whence, it results that:

| ( ) ( )| 2 ⎡ ⎤
| f ξj − f ξ' | ≤ , for any ξ 'j ∈ x j−1 , x j ,
j
x j − x j−1
⎡ ⎤
thus f is bounded on x j−1 , x j , for all j = 1, n. Hence f is bounded on [a, b].

Remark 2.4.2 From Theorem 2.4.1 we deduce that the study of (R)—integrability
can be reduced to the study of the integrability of bounded functions.

Remark 2.4.3 The reciprocal statement of Theorem 2.4.1 is not generally true. There
are bounded functions which are not (D)—integrable. A such example is Dirichlet
function:
⎧
1 if x ∈ Q
f : [0, 1] → R, f (x) = .
0 if x ∈ [0, 1]\Q

Indeed, any subinterval of an arbitrary partition Δ of [0, 1] contains both rational

points and irrational points, thus:

m i = 0, Mi = 1, sΔ = 0, SΔ = 1, I∗ = 0, I ∗ = 1

whence, it results that f is not (D)—integrable on [0, 1].

Remark 2.4.4 If f : [a, b] → R is continuous on [a, b], then f is bounded on

[a, b] and it attains its supremum
⎡ and infimum
⎤ on any compact subinterval of [a, b].
Therefore, there are μi , ηi ∈ xi−1 , xi such that m i = f (μi ) and Mi = f (ηi ).
It results that Darboux sums sΔ and SΔ are Riemann sums, namely:
2.4 Riemann Sums. Riemann Criterion for Integrability 37

v
n v
n
sΔ = f (μi )(xi − xi−1 ) and SΔ = f (ηi )(xi − xi−1 ).
i=1 i=1

If f is bounded, the link between Riemann sums and Darboux sums is given by
the following lemma:

Lemma 2.4.1 Let f : [a, b] → R be a bounded function, ⎡ let ⎤Δ be an arbitrary

partition of [a, b] and ε > 0. Then there are αi , βi ∈ xi−1 , x i , 1 ≤ i ≤ n, such
that:

v
n
SΔ − ε < f (αi )(xi − xi−1 ) = σΔ ( f ; α) and
i=1

v
n
sΔ + ε > f (βi )(xi − xi−1 ) = σΔ ( f ; β)
i=1

where α = (α1 , . . . , αn ) and β = (β1 , . . . , βn ).

{ ⎡ ⎤}
Proof Let Mi = sup f (x); x ∈ xi−1 , x⎡i . According ⎤ to the definition of the least
upper bound, it results that there is αi ∈ xi−1 , x i such that:

ε
Mi − < f (αi ). (2.6)
b−a

Multiplying the inequality (2.6) by xi − xi−1 and making the sum, it results that:

v
n
SΔ − ε < f (αi )(xi − xi−1 ) = σΔ ( f ; α).
i=1

⎡ ⎤
Similarly, there is βi ∈ xi−1 , x i such that:

ε
mi + > f (βi ). (2.7)
b−a

Multiplying the inequality (2.7) by xi − xi−1 and making the sum, it follows that:

v
n
sΔ + ε > f (βi )(xi − xi−1 ) = σΔ ( f ; β).
i=1

Remark 2.4.5 Between Darboux sums and Riemann sums associated to a bounded
function f and a partition Δ we have the following relations:
{ ⎡ ⎤ }
sΔ = inf σΔ ( f ; ξ ); ξi ∈ xi−1 , x i , 1 ≤ i ≤ n
{ ⎡ ⎤ }
SΔ = sup σΔ ( f ; ξ ); ξi ∈ xi−1 , x i , 1 ≤ i ≤ n .
38 2 Definite Integrals

The following theorem prove that the both definitions of integrability are
equivalent.

Theorem 2.4.2 Let f : [a, b] → R be a bounded function. Then the function f is

(D)—integrable on [a, b] if and only if f is (R)—integrable on [a, b].

Proof If we assume that f is (D)—integrable on [a, b], then I∗ = I ∗ = I . On

the other hand, from Theorem 2.2.1, it results that ∀ε > 0, ∃ δε > 0 such that
SΔ − sΔ < ε, for any partition Δ, with ∥Δ∥ < δε .
Since sΔ ≤ I ≤ SΔ and sΔ ≤ σΔ ( f ; ξ ) ≤ SΔ , ∀ξ , it follows that:

|σΔ ( f ; ξ ) − I | < SΔ − sΔ < ε

for any Δ, with ∥Δ∥ < δε and for any vector ξ , whence, it results that f is (R)—
integrable on [a, b].
Conversely, let us assume that f is (R)—integrable. Then there is a finite number
I ∈ R with the property: ∀ε > 0, ∃ δε > 0 such that ∀Δ, with ∥Δ∥ < δε and ∀ξ we
have:
ε ε
I− < σΔ ( f ; ξ ) < I + . (2.8)
4 4
Let Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b be an arbitrary partition
with ∥Δ∥ < δε . ⎡ ⎤
From Lemma 2.4.1, it results that there is αi ∈ xi−1 , x i , 1 ≤ i ≤ n, such that:

ε
SΔ − < σΔ ( f ; α),
4
where α = (α1 , . . . , αn ).
According to (2.8), we get:
ε
SΔ < I + . (2.9)
2
Similarly, using again Lemma 2.4.1 and inequalities (2.8), it is shown that:
ε
sΔ > I − . (2.10)
2
From (2.9) and (2.10), it follows that SΔ − sΔ < ε, for any Δ, with ∥Δ∥ < δε ,
therefore f is (D)—integrable (according to Theorem 2.2.1).

Theorem 2.4.3 (Riemann criterion for integrability) The necessary and sufficient
condition that the function f : [a, b] → R to be integrable on the interval [a, b] is
there is a real number I such that for any sequence {Δn } of partitions of [a, b], with
limn→∞ ∥Δn ∥ = 0 and any choice of intermediate points ξ (n) , we have
2.4 Riemann Sums. Riemann Criterion for Integrability 39
( )
lim σΔn f ; ξ (n) = I.
n→∞

Proof First, from Definition 2.4.2, it results that there is I ∈ R such that ∀ε > 0,
∃ δε > 0 with the property that for any Δ, with ∥Δ∥ < δε and for any ξ , we have:

|σΔ ( f ; ξ ) − I | < ε. (2.11)

Let {Δn } be a sequence of partitions of [a, b] with the property limn→∞ ∥Δn ∥ = 0.
Then there is n 0 ∈ N∗ such that ∥Δn ∥ < δε , for any| n ≥( n 0 . ) |
Taking into account to (2.11), we deduce that |σΔn f ; ξ (n) − I | < ε, for any
n ≥ n 0 and any intermediate point system ξ (n) corresponding to the partition Δn ,
whence, it results that:
( )
lim σΔn f ; ξ (n) = I.
n→∞

( let us)suppose that there is a real number I ∈ R with the property that
Conversely,
limn→∞ σΔn f ; ξ (n) = I , for any sequence of partitions {Δn }, such that ∥Δn ∥ → 0
and for any intermediate point system ξ (n) .
Furthermore, if we suppose that the function f is not integrable on [a, b], then
∀I ∈ R, ∃ ε0 > 0 such that ∀δ > 0, there is a partition Δδ , with ∥Δδ ∥ < δ and an
intermediate point system ξ δ , with the property
| ( ) |
|σΔ f ; ξ δ − I | ≥ ε0 .
δ

In particular, for δ = n1 , it results that there

| is( a partition Δ| n , with ∥Δn ∥ < n1 and
)
(n) | (n)
− I | ≥ ε0 . This means that
an intermediate
( point
(n)
) system ξ such that σΔn f ; ξ
limn→∞ σΔn f ; ξ /= I , although ∥Δn ∥ → 0, which contradicts the hypothesis.

Remark 2.4.6 Let f : [a, b] → R be an integrable function and let Δn be an

equidistant partition of [a, b], i.e.:

Δn : a = x0(n) < x1(n) < . . . < xi−1

(n)
< xi(n) < . . . < xn(n) = b

where xi(n) = a + i b−an

, i = 0, n.
It is easily seen that limn→∞ ∥Δn ∥ = limn→∞ b−a n
= 0. Since f is an inte-
⎡ grable function,
⎤ we can arbitrarily choose the intermediate points ξi(n) in the interval
(n)
xi−1 , xi(n) , i = 1, n. Thus, if we choose ξi(n) = xi(n) = a +i b−a
n
, then the associated
Riemann sum becomes:
( ) ( )
b−a v
n
b−a
σΔn f ; xi(n) = f a+i .
n i=1 n

From Theorem 2.4.3, it results that:

40 2 Definite Integrals

∫b ( )
b−a v
n
b−a
f (x)dx = lim f a+i . (2.12)
n→∞ n i=1 n
a

∫1
Example 2.4.1 Compute 0 x dx.
Replacing in (2.12) a = 0, b = 1, f (x) = x, it follows that:

∫1
1vi
n
1 + 2 + ··· + n n(n + 1) 1
x dx = lim = lim 2
= lim 2
= .
n→∞ n n n→∞ n n→∞ 2n 2
0 i=1

∫π
Example 2.4.2 Compute 02 cos x dx.
To compute this integral we will use the identity:

sin nx
cos (n+1)x
cos x + cos 2x + · · · + cos nx = 2 2
, x /= 2kπ, k ∈ Z.
sin 2x
π
Replacing in (2.12) a = 0, b = 2
, f (x) = cos x, we obtain:
π
∫2
π v
n
πi
cos x dx = lim cos
n→∞ 2n 2n
0 i=1
( )
π π 2π nπ
= lim cos + cos + · · · + cos
n→∞ 2n 2n 2n 2n
nπ (n+1)π
π sin 2n
· cos 2n

= lim · 2
π
2
n→∞ 2n
sin 2n
2
π sin π
· cos (n+1)π
= lim · 4
π
4n
n→∞ 2n sin 4n
π
π π
= lim 2 · π · sin
4n
· cos
n→∞ sin 4n 4 4
√ √
2 2
=2· · = 1.
2 2
Remark 2.4.7 If f : [a, b] → R is an integrable function and {Δn } is a sequence
of partitions such that limn→∞ ∥Δn ∥ = 0, then limn→∞ sΔn = limn→∞ SΔn =
∫b
a f (x) dx.
Indeed, according to Lemma 2.4.1, it results that ∀n ∈ N, ∃ α (n) , β (n) such that:

( ) 1 ( ) 1
0 ≤ SΔn − σΔn f ; α (n) < and 0 ≤ σΔn f ; β (n) − sΔn < .
n n
Passing to the limit in these inequalities, we obtain:
2.5 Lebesgue Criterion for Integrability 41

∫b
( (n)
)
lim SΔn = lim σΔn f ; α = f (x) dx, respectively
n→∞ n→∞
a

∫b
( (n)
)
lim sΔn = lim σΔn f ; β = f (x) dx.
n→∞ n→∞
a

2.5 Lebesgue Criterion for Integrability

Definition 2.5.1 A set A ⊂ R is said to be null set if for any ε > 0, there is a
sequence {In } of open intervals with the properties:
∪
∞
(a) A ⊂ In ;
n=1
∑
∞
(b) l(In ) < ε.
n=1

where l(In ) is the length of the interval In .

(We note that some of these intervals In may be the empty set).
Proposition 2.5.1 Any singleton subset of R is a null set.
( )
Proof Let A = {x0 }. We can choose I1 = x0 − 3ε , x0 + 3ε and In = ∅, for any
∪∞ ∑∞
n ≥ 2. Obviously, we have A ⊂ n=1 In and n=1 l(In ) = l(I1 ) = 2ε3 < ε.
The following statement is obvious.
Proposition 2.5.2 If A ⊂ B and B is a null set, then A is also a null set.
Proposition 2.5.3 Any countable union of null sets is also a null set.
Proof Let An ⊂ R be a null set, for any n ∈ N∗ . According to Definition 2.5.1, it
results that for any ε > 0, there is a sequence {In m } of open intervals such that:
∞
⊔ ∞
v ε
An ⊂ In m and l(In m ) < .
m=1 m=1
2n

Furthermore, we have:
∞ ∞
(∞ ) ∞ v ∞ ∞
⊔ ⊔ ⊔ v v ε
An ⊂ In m and l(In m ) < =ε
n=1 n=1 m=1 n=1 m=1 n=1
2n
∪∞
hence n=1 An is a null set.
42 2 Definite Integrals

Corollary 2.5.1 Any finite or countable subset of R is a null set.

Proof The statement follows from Propositions 2.5.1 and 2.5.3.

In the following, we present, without proof, the result due to Henri Lebesgue
(1875–1941).

Theorem 2.5.1 (Lebesgue criterion for integrability) A bounded function

f : [a, b] → R is integrable on the interval [a, b] if and only if f is continuous
almost everywhere (i.e. the set of its points of discontinuity is a null set).

Remark 2.5.1
(1) Using Theorem 2.5.1, the proof of Theorem 2.3.1 becomes immediate, because
the empty set is a null set.
(2) According to Theorem 2.5.1, the proof of Theorem 2.3.2 is done quickly, because
the set of discontinuity points for a monotonic function is finite or countable,
thus a null set.

2.6 Properties of Integrable Functions

∫b
Proposition 2.6.1 a 1 dx = b − a.

Proof The statement follows from the fact that, in this case, any Riemann sum is:

σΔ (1; ξ ) = b − a.

Proposition 2.6.2 (Linearity Property) If f, g : [a, b] → R are two integrable

functions on [a, b], then for any real numbers α, β ∈ R, the function α · f + β · g is
also an integrable function on [a, b] and

∫b ∫b ∫b
(α · f (x) + β · g(x)) dx = α · f (x) dx + β · g(x) dx.
a a a

Proof Let {Δn } be a sequence of partitions of [a, b] such that limn→∞ ∥Δn ∥ = 0
and let ξ (n) be an arbitrary system of intermediate points of partition Δn . Obviously
we have:
( ) ( ) ( )
σΔn α f + β g; ξ (n) = α σΔn f ; ξ (n) + β σΔn g; ξ (n) .

Because the equality from the right side has a finite limit when n → ∞, namely
∫b ∫b
α · a f (x) dx + β · a g(x) dx, it results that the left member also has a finite limit,
hence the function α · f +β · g is integrable on [a, b] and the equality in the statement
holds.
2.6 Properties of Integrable Functions 43

Proposition 2.6.3 (Monotony Property) If f, g : [a, b] → R are integrable such

that f (x) ≤ g(x), ∀x ∈ [a, b], then

∫b ∫b
f (x) dx ≤ g(x) dx.
a a

Proof The inequality from the proposition statement follows immediately if we pass
to limit (on ∥Δ∥ → 0) in the obvious inequality

σΔ ( f ; ξ ) ≤ σΔ (g; ξ ).

Corollary 2.6.1 If f : [a, b] → R is integrable and f (x) ≥ 0, ∀x ∈ [a, b], then

∫b
f (x) dx ≥ 0.
a

∫b
The assertion follows from Proposition 2.6.3 and the obvious remark a 0 dx = 0.

Corollary 2.6.2 If f : [a, b] → R is an integrable function on [a, b], then

∫b
m · (b − a) ≤ f (x) dx ≤ M · (b − a),
a

where m = inf{ f (x); x ∈ [a, b]} and M = sup{ f (x); x ∈ [a, b]}.

Proof The assertion follows from the inequalities m ≤ f (x) ≤ M and from
Propositions 2.6.1, 2.6.2 and 2.6.3. Indeed:

∫b ∫b ∫b
m · (b − a) = m dx ≤ f (x) dx ≤ M dx = M · (b − a).
a a a

Proposition 2.6.4 If f : [a, b] → R is an integrable function on [a, b], then | f | is

integrable on [a, b] and we have
| b |
|∫ | ∫b
| |
| f (x) dx | ≤ | f (x)| dx.
| |
| |
a a

Proof Let A be the set of all points of discontinuity of the function | f | in [a, b] and
let B be the set of all points of discontinuity of the function f in [a, b]. Because it
is known that if the function f is continuous at a point, then the function | f | is also
44 2 Definite Integrals

continuous at that point, it follows that A ⊂ B. Since B is a null set, we deduce that
A is also a null set. According to Theorem 2.5.1, it results that | f | is integrable.
On the other hand we have:

−| f (x)| ≤ f (x) ≤ | f (x)|, ∀x ∈ [a, b].

From Proposition 2.6.3, it results that:

∫b ∫b ∫b
− | f (x)| dx ≤ f (x) dx ≤ | f (x)| dx
a a a

hence
| b |
|∫ | ∫b
| |
| f (x) dx | ≤ | f (x)| dx.
| |
| |
a a

Proposition 2.6.5 If f, g : [a, b] → R are two integrable functions on [a, b], then
their product f · g is also an integrable function on [a, b].

Proof Indeed, let us denote by A, B, respectively C, the sets of all discontinuity

points of the functions f, g, respectively f · g in [a, b]. Because it is known that
if the functions f and g are continuous at a point, then their product f · g is also
continuous at that point, it results that C ⊂ A ∪ B. Because A and B are null sets
(Theorem 2.5.1), it follows that A ∪ B is a null set (Proposition 2.5.3), hence C is a
null set (Proposition 2.5.2). Now the assertion follows from Theorem 2.5.1.

Proposition 2.6.6 If f : [a, b] → R is an integrable function on [a, b], then for

any c ∈ [a, b], f is integrable on [a, c] and [c, b] and the following relation holds:

∫b ∫c ∫b
f (x) dx = f (x) dx + f (x) dx.
a a c

Proof From Theorem 2.5.1, it results

{ ' } that f is integrable on [a, c] and [c, b].
Let us consider a ∥sequence
∥ Δn of the partitions
{ } of the interval [a, c], with
the property lim ∥Δ' ∥ = 0 and a sequence Δ'' of partitions of [c, b], with
∥ ∥ n→∞ n n
limn→∞ ∥Δ''n ∥ = 0. If we denote by Δn = Δ'n ∪ Δ''n , i.e. the partition obtained from
partition Δ'n by adding points of the partition Δ''n , then Δn is a partition of the interval
[a, b] and limn→∞ ∥Δn ∥ = 0.
Let α (n) (respectively β (n) ) be an arbitrary intermediate point system for the
partition Δ'n (respectively Δ''n ). If we denote by ξ (n) = α (n) ∪ β (n) , then ξ (n) is
an intermediate point system for the partition Δn .
2.6 Properties of Integrable Functions 45

If we pass to limit in the equality:

( ) ( ) ( )
σΔn f ; ξ (n) = σΔ'n f ; α (n) + σΔ''n f ; β (n)

we get:

∫b ∫c ∫b
f (x) dx = f (x) dx + f (x) dx.
a a c

Remark 2.6.1 If f : [a, b] → R is integrable on [a, b], then we have:

∫b ∫a
(1) f (x) dx = − f (x) dx
a b
∫a
(2) f (x) dx = 0.
a

Theorem 2.6.1 (The mean value formula) Let f, g : [a, b] → R be two integrable
functions on [a, b]. We assume that the function g has a constant sign on [a, b]. If
we denote by m = inf{ f (x); x ∈ [a, b]} and by M = sup{ f (x); x ∈ [a, b]}, then
there is m ≤ μ ≤ M such that:

∫b ∫b
f (x) · g(x) dx = μ · g(x) dx. (2.13)
a a

Proof Let us suppose that g(x) ≥ 0, ∀x ∈ [a, b]. Then, because m ≤ f (x) ≤ M,
∀x ∈ [a, b], it results that m · g(x) ≤ f (x) · g(x) ≤ M · g(x), ∀x ∈ [a, b].
From Propositions 2.6.2 and 2.6.3, we have:

∫b ∫b ∫b
m· g(x) dx ≤ f (x) · g(x) dx ≤ M · g(x) dx. (2.14)
a a a

∫b ∫b
If a g(x) dx = 0, then a f (x) · g(x) dx = 0, and the equality (2.13) holds for
any μ ∈ R.
∫b
Let us suppose that a g(x) dx /= 0. Since g ≥ 0, from Proposition 2.6.4, we
∫b
deduce that a g(x) dx > 0.
∫b ∫b
f (x)·g(x) dx
Dividing (2.14) by a g(x) dx, we obtain m ≤ a ∫ b ≤ M.
g(x) dx
∫b a
f (x)·g(x) dx
If we denote by μ = a
∫b , then m ≤ μ ≤ M and the relation (2.13) is
a g(x) dx
proved.
Corollary 2.6.3 Let f, g : [a, b] → R be two functions as in Theorem 2.6.1. If we
additionally assume that f is continuous on [a, b], then there is ξ ∈ [a, b] such that:
46 2 Definite Integrals

∫b ∫b
f (x) · g(x) dx = f (ξ ) · g(x) dx. (2.15)
a a

Proof Because f is continuous on the compact interval [a, b], it results that there
are α, β ∈ [a, b] such that m = f (α) and M = f (β). According to Theorem 2.6.1,
∫b ∫b
there is m = f (α) ≤ μ ≤ f (β) = M such that a f (x) · g(x) dx = μ · a g(x) dx.
On the other hand, f has the Darboux property on [a, b], whence it results that
there is a point ξ between α and β, so in [a, b], such that μ = f (ξ ). Therefore, the
relation (2.15) is verified.
Corollary 2.6.4 If f : [a, b] → R is integrable, then there is m ≤ μ ≤ M such
that:

∫b
f (x) dx = μ · (b − a).
a

Proof The statement holds immediately from Theorem 2.6.1, for the particular case
g = 1.
Corollary 2.6.5 If f : [a, b] → R is continuous, then there is ξ ∈ [a, b] such that:

∫b
f (x) dx = f (ξ ) · (b − a).
a

Proof The statement follows immediately from Corollary 2.6.2, for the particular
case g = 1.
Theorem 2.6.2 (The fundamental theorem of integral calculus) Any function f :
[a, b] → R continuous on the interval (a, b) has an antiderivative on (a, b). One of
the antiderivatives of f is the function

∫x
F(x) = f (t) dt, ∀x ∈ (a, b)
c

where c is an arbitrary fixed point from (a, b).

Proof Taking into account to Proposition 2.6.6, Remark 2.6.1 and Corollary 2.6.4,
for any x ∈ (a, b) and for any h ∈ R such that x + h ∈ (a, b), we have:
∫ x+h ∫x
F(x + h) − F(x) f (t) dt − c f (t) dt
= c
h h
∫x ∫ x+h ∫x
f (t) dt + x f (t) dt − c f (t) dt
= c
h
2.6 Properties of Integrable Functions 47
∫ x+h
f (t) dt f (ξx )h
= x
= = f (ξx )
h h
where ξx is a point between x and x + h.
Because f is continuous at the point x, further we have:

F(x + h) − F(x)
lim = lim f (ξx ) = f (x).
h→0 h h→0

Therefore F ' (x) = f (x), hence F is an antiderivative of f .

Theorem 2.6.3 (Newton–Leibniz formula) Let f : [a, b] → R be an integrable

function that admits antiderivatives on [a, b]. If F : [a, b] → R is an arbitrary
antiderivative of f on the interval [a, b], then:

∫b |
|
f (x) dx = F(x)|b = F(b) − F(a). (2.16)
a
a

Proof Let us take Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b an arbitrary
partition of the interval [a, b]. We notice that:

v
n
F(b) − F(a) = (F(xi ) − F(xi−1 )).
i=1

On the other hand, from Lagrange Theorem, it results that there is ξi ∈ (xi−1 , xi )
such that:

F(xi ) − F(xi−1 ) = F ' (ξi )(xi − xi−1 ) = f (ξi )(xi − xi−1 ).

If we denote by ξ = (ξ1 , ξ2 , . . . , ξn ), then:

v
n
F(b) − F(a) = f (ξi )(xi − xi−1 ) = σΔ ( f ; ξ ).
i=1

Let {Δn } be a sequence of partitions of [a, b], such that limn→∞ ∥Δn ∥ = 0, and
let ξ (n) be the intermediate point system of Δn , that follows from Lagrange Theorem.
Then we have:
( )
F(b) − F(a) = σΔn f ; ξ (n)

and therefore:
48 2 Definite Integrals

∫b
( )
f (x) dx = lim σΔn f ; ξ (n) = F(b) − F(a).
n→∞
a

∫1( )
Example 2.6.1 Compute 0 2x 3 − 4x 2 + 5x − 7 dx.
From Example 1.2.1, an antiderivative function of f (x) = 2x 3 − 4x 2 + 5x − 7
4 3 2
on the interval [0, 1] is F(x) = x2 − 4x3 + 5x2 − 7x.
According to Newton–Leibniz formula, we have:

∫1 | ( 4 )|
( 3 ) | x 4x 3 5x 2 |
2x − 4x + 5x − 7 dx = F(x)||1 =
2
− + − 7x ||1
0 2 3 2 0
0
1 4 5 16
= − + −7−0=− .
2 3 2 3
Theorem 2.6.4 (The integral by parts formula) If f, g : [a, b] → R are two
functions of C 1 —class on [a, b], then:

∫b | ∫b
' |b
f (x) · g (x) dx = f (x) · g(x)| − f ' (x) · g(x) dx. (2.17)
a
a a

Proof Formula (2.17) is obtained immediately if we apply Proposition 2.6.2 and

Theorem 2.6.3 to the relation:

( f (x) · g(x))' = f ' (x) · g(x) + f (x) · g ' (x), x ∈ [a, b].
∫1
Example 2.6.2 Compute 0 2x ex dx.
If we denote by f (x) = 2x and by g ' (x) = ex , it results that f ' (x) = 2 and
g(x) = ex .
From Theorems 2.6.3, 2.6.4 and Proposition 2.6.2, it follows that:

∫1 | ∫1 | |
| x |1
|
x |1
|
x |1
2x e dx = 2x e | − 2e dx = 2x e | − 2e |
x x
0 0 0
0 0
= 2e − 0 − 2e + 2 = 2.

Theorem 2.6.5 (The change of variable formula) If J ⊂ R is an interval, f : J → R

is a continuous function and u : [a, b] → J is a function of C 1 —class, then:

∫b ∫u(b)
'
f (u(x)) · u (x) dx = f (t) dt. (2.18)
a u(a)
2.7 Area of a Plane Figure 49

Proof From Theorem 2.6.2, it results that f admits antiderivative functions. If F

is an antiderivative function of f , then F ' (t) = f (t), ∀t ∈ J , and according to
Theorem 2.6.3, we have:

∫u(b) ∫u(b)
f (t) dt = F ' (t) dt = F(u(b)) − F(u(a)).
u(a) u(a)

On the other hand, it results that:

(F(u(x)))' = F ' (u(x)) · u ' (x) = f (u(x)) · u ' (x), ∀x ∈ [a, b].

Applying Theorem 2.6.3, we obtain:

∫b ∫b |
|
'
f (u(x)) · u (x) dx = (F(u(x)))' dx = F(u(x))|b
a
a a
∫u(b)
= F(u(b)) − F(u(a)) = f (t) dt.
u(a)

∫2
Example 2.6.3 Compute 0 x 22x−1
−x+4
dx.
If we denote by t = x 2 − x + 4, then dt = (2x − 1) dx and t ∈ [4, 6]. From
(2.18), we obtain:

∫2 ∫6 |
2x − 1 1 | 3
dx = dt = ln|t|||6 = ln 6 − ln 4 = ln .
x2 − x + 4 t 4 2
0 4

2.7 Area of a Plane Figure

We are familiar with the concept of area of a polygonal plane set (plane figure
bounded by a polygon) from the course of elementary mathematics. In what follows,
we shall introduce the concept
{ of squareable plane figure. }
If D = [a, b]×[c, d] = (x, y) ∈ R2 ; a ≤ x ≤ b, c ≤ y ≤ d is a rectangle with
sides parallel to the coordinate axes, then:

def
area(D) =(b − a)(d − c).
50 2 Definite Integrals

The same formula of area remains true even if the rectangle does not contain one
or more sides.

Definition 2.7.1 An elementary plane set E ⊂ R2 is a finite union of rectangular

plane sets with sides parallel to the coordinate axes, which may or may not contain
one or more sides and which have in common two by two only at most one side, i.e.:

⊔
p
◦ ◦
E= Di , Di ∩ D j = ∅, for i /= j.
i=1

By definition, the area of the elementary set E ⊂ R2 is:

def
v
p
v
p
area(E) = area(Di ) = (bi − ai )(di − ci ). (2.19)
i=1 i=1

Furthermore, we shall denote by E the family of all elementary plane sets.

Remark 2.7.1 The family of elementary plane sets has the following properties:
∪p
(1) The representation of an elementary set by the form E = i=1 Di is not unique.
(2) If E 1 , E 2 ∈ E , then E 1 ∪ E 2 , E 1 ∩ E 2 and E 1 \E 2 ∈ E .
∪ pelementary set E∪⊂q R does not depend by ∑
2
(3) The area of an its representation,
p
i.e. if E = i=1 Di and E = G , then area(E) = i=1 area(Di ) =
∑q ( ) j=1 j

j=1 area G j .
◦ ◦
(4) If E 1 , E 2 ∈ E and E 1 ∩ E 2 = ∅, then

area(E 1 ∪ E 2 ) = area(E 1 ) + area(E 2 ).

(5) If E 1 , E 2 ∈ E and E 1 ⊂ E 2 , then

area(E 1 ) ≤ area(E 2 ) and area(E 2 \E 1 ) = area(E 2 ) − area(E 1 ).

We recall that a subset A ⊂ R2 is bounded if there is r > 0 such that A ⊆ B(O, r )

(the ball with center at origin and radius r ).
If A ⊂ R2 is a bounded set, then we denote by
S∗ ( A) = sup{area(E); E ∈ E , E ⊂ A} the interior area of the set A
respectively by
S ∗ (A) = inf{area(F); F ∈ E , F ⊃ A} the exterior area of the set A.
If the set A does not contain any elementary set, we define S∗ (A) = 0.
Clearly, if the set A is bounded, then the both numbers S∗ ( A) and S ∗ (A) exist and

S∗ (A) ≤ S ∗ (A).
2.7 Area of a Plane Figure 51

Definition 2.7.2 We say that a bounded set A ⊂ R2 is squareable (has area) if

S∗ ( A) = S ∗ (A) = S(A).

In this case, the number S( A) is denoted by area(A) and it is called the area of
A.

Remark 2.7.2 Any elementary plane set is squareable and its area is equal to the
area defined in (2.19).

Remark 2.7.3 Any polygonal set has area defined by Definition 2.7.2, and this area
is the same as the known area of elementary geometry.
Indeed, since any polygonal set is a finite union of triangular sets and any triangle
is the union or the difference of two right triangles, it is enough to show that any
plane set whose boundary is a right triangle has area.
Let us take the right triangle ABC with: A = 90◦ , AB = a, AC = b (Fig. 2.8).
Let us divide the cathetus AB into n equal parts and let us consider rectangles
of type M N Q P, where M N = an and M P is parallel to AC. Let us suppose that
B M = i · an . Since the triangles B M P and B AC are similar, it follows that:

BM MP b
= , hence M P = i · .
a b n

Therefore, the area of the rectangle M N Q P is i · ab

n2
.
If we denote by E the union of these rectangles, then E ∈ E , the set E is included
in the triangle ABC and we have the equality:

ab ab(n − 1)
area(E) = (1 + 2 + · · · + n − 1) = .
n2 2n
Similarly, if we denote by F the union of rectangles of type M N S R, then F is an
elementary set that includes the triangle ABC and:

Fig. 2.8 Any right triangle B

is squarable

M P R

N S
Q

A C
52 2 Definite Integrals

ab ab(n + 1)
area(F) = 2
(1 + 2 + · · · + n) = .
n 2n
Further, we have:

ab ab(n − 1) ab(n + 1) ab
= sup ≤ S∗ ( ABC) ≤ S ∗ (ABC) ≤ inf =
2 n 2n n 2n 2

hence
ab
S∗ ( ABC) = S ∗ (ABC) = .
2
Therefore, the triangular set ABC has the area defined by Definition 2.7.2, and
this area is the known area of a right triangle in elementary geometry.

Remark 2.7.4 There are plane sets⎧that are non-squareable.

1 if x ∈ Q
Indeed, if f : R → R, f (x) = is the Dirichlet function and
0 if x ∈ R\Q
{ }
A = (x, y) ∈ R2 ; 0 ≤ x ≤ 1, 0 ≤ y ≤ f (x) ,

then S∗ (A) = 0 and S ∗ (A) = 1, whence, it results that A is non-squareable set.

The following result gives us examples of squareable sets. More precisely, we
will show that the curvilinear trapezoids corresponding to the integrable functions
are squareable sets.

Proposition
{ 2.7.1 If f : [a, b] → R+ is integrable,
} then the curvilinear trapezoid
Γf = (x, y) ∈ R2 ; a ≤ x ≤ b, 0 ≤ y ≤ f (x) is squareable and we have:

∫b
area(Γf ) = f (x) d x. (2.20)
a

Proof Let us take Δ : a = x0 < x1 < . . . < xi−1 < xi < . . . < xn = b an arbitrary
partition of [a, b] and let
{ ⎡ ⎤} { ⎡ ⎤}
m i = inf f (x); x ∈ xi−1 , xi , Mi = sup f (x); x ∈ xi−1 , xi , 1 ≤ i ≤ n.
∪n ⎡ ⎤
If we denote by E Δ = i=1 xi−1 , xi × [0, m i ], then E Δ ∈ E , E Δ ⊂ Γf and

v
n
area(E Δ ) = m i (xi − xi−1 ) = sΔ .
i=1

It results that sΔ ≤ S∗ (Γf ) (Fig. 2.9).

2.7 Area of a Plane Figure 53

Fig. 2.9 The subgraph of an

integrable function is
squarable

∪n ⎡ ⎤
Analogously, if we denote by FΔ = i=1 xi−1 , xi × [0, Mi ], then FΔ ∈ E ,
FΔ ⊃ Γf and

area(FΔ ) = SΔ ≥ S ∗ (Γf ).

Therefore, we have the inequalities:

sΔ ≤ S∗ (Γf ) ≤ S ∗ (Γf ) ≤ SΔ . (2.21)

The fact that f is integrable on [a, b] involves:

∫b
∗
I∗ = sup sΔ = inf SΔ = I = f (x) dx.
Δ Δ
a

Finally, from (2.21), it results that:

∫b
∗
area(Γf ) = S∗ (Γf ) = S (Γf ) = f (x) dx.
a

Corollary 2.7.1 Let f, g : [a, b] → R be two integrable functions on [a, b] such

that f (x) ≤ g(x), ∀x ∈ [a, b].
If we denote by Γ f g the intergraph of these functions (Fig. 2.10), i.e. the set
{ }
Γ f g = (x, y) ∈ R2 ; a ≤ x ≤ b, f (x) ≤ y ≤ g(x) ,

then Γ f g is a squareable set and we have the formula:

∫b
( )
area Γ f g = (g(x) − f (x))dx.
a
54 2 Definite Integrals

Fig. 2.10 The area of an

intergraph

Example 2.7.1 Compute the area of the ellipse.

2 2
The ellipse equation is ax 2 + by2 − 1 = 0.
For symmetry reasons, it is sufficient to compute a quarter of the area of the
ellipse, for example the hatched area from Fig. 2.11.
∩
The arc B A is the graph of the function:

b √ 2
f (x) = a − x 2 , x ∈ [0, a].
a
Applying Proposition 2.7.1 and Example 1.2.9, we obtain:

∫a
1 b √ 2
· Area (ellipse) = a − x 2 dx
4 a
0
( )|
b x √ 2 a 2 −1 ( x ) ||a
= a − x + sin
2
a 2 2 a |0
b a π 2
πab
= · · = .
a 2 2 4
Therefore, the area of the ellipse with the semiaxes a and b is equal to πab.

Fig. 2.11 The area of an

ellipse
2.7 Area of a Plane Figure 55

Fig. 2.12 The area of a

circulars ector

In particular, for a = b = R, we obtain the area of the circle: π R 2 .

Example 2.7.2 Compute the area of a circular sector.
We intend to calculate the area of the circular sector O AB in Fig. 2.12.
From elementary geometry, it is known that the area of O AB is R 2·α .
2

We will show that the circular sector O AB is squareable, in the sense of

Definition 2.7.2, and its area is equal to the above area.
Since the equation of the circle of radius R centered at the origin is x 2 + y 2 = R 2 ,
∩
we deduce that the arc AB is the graph of the function:
√
f (x) = R 2 − x 2 , x ∈ [0, R sin α].

On the other hand, the straight line segment O B is the graph of the function:
(π )
g(x) = x · tan − α = x · ctan α, x ∈ [0, R sin α].
2
According to Corollary 2.7.1, we have:

R∫sin α
(√ )
Area (O AB) = R 2 − x 2 − x · ctan α dx
0
( √ )|
x R 2 −1 ( x ) x 2 |
= R2 − x 2 + sin − · ctan α || R sin α
2 2 R 2 0

R sin α cos α
2
R ·α
2
R sin α cos α
2 2
R2 · α
= + − · = .
2 2 2 sin α 2
Therefore, any circular sector is squareable and its area is equal to the known area
of elementary geometry.
Theorem 2.7.1 The necessary and sufficient condition that a bounded subset A ⊂
R2 to be squareable is that for any ε > 0, there are two elementary sets E ε and Fε
such that E ε ⊂ A ⊂ Fε and

area(Fε ) − area(E ε ) < ε.

56 2 Definite Integrals

Proof First, if S∗ (A) = S ∗ (A) = S( A), on account of the definition of the least
upper bound and the greatest lower bound, it follows that there are E ε , Fε ∈ E ,
E ε ⊂ A ⊂ Fε such that:
ε ε
S( A) − < area(E ε ) and area(Fε ) < S( A) + .
2 2
Therefore we have:

area(Fε ) − area(E ε ) < ε.

Conversely, if for any ε > 0, there are E ε , Fε ∈ E such that E ε ⊂ A ⊂ Fε and

area(Fε ) − area(E ε ) < ε, then we have:

0 ≤ S ∗ ( A) − S∗ (A) < area(Fε ) − area(E ε ) < ε.

Since ε > 0 is arbitrary, it results that S∗ (A) = S ∗ (A), thus A is squareable.

Definition 2.7.3 A set Γ ⊂ R2 is said to be of area zero if for ∀ε > 0, ∃ Fε ∈ E
such that Γ ⊂ Fε and area(Fε ) < ε.
In particular, we have S ∗ (Γ) = 0, and taking into account that 0 ≤ S∗ (Γ) ≤
∗
S (Γ), it results that the set Γ is squareable and area(Γ) = 0.
Using this definition, Theorem 2.7.1 can be reformulated as follows:
Theorem 2.7.2 The necessary and sufficient condition that a bounded subset A ⊂
R2 to be squareable is that its boundary to be of area zero.
Proof First, if A is squareable, then ∀ε > 0, ∃ E ε , Fε ∈ E such that:

E ε ⊂ A ⊂ Fε and area(Fε \E ε ) = area(Fε ) − area(E ε ) < ε.

◦ ◦
Since Γ = ∂ A ⊂ Fε \ E ε and Fε \ E ε is also an elementary set, it results that Γ is
of area zero.
Conversely, we assume that for any ∀ε > 0, ∃ Fε ∈ E such that Γ = ∂ A ⊂ Fε
and area(Fε ) < ε.
Let D = [a, b] × [c, d] be a rectangle such that A ⊂ D. Obviously

⊔
p
◦ ◦
D\Fε ∈ E , hence D\Fε = Di , Di ∩ D j = ∅, for i /= j.
i=1

◦ ◦ ◦
For any i ≤ p, we have either Di ⊂ A, either Di ⊂ D\A, because otherwise, Di
◦
being a convex set, Di ∩ ∂ A /= ∅. But ∂ A ⊂ Fε and Fε ∩ Di = ∅. We have:
2.7 Area of a Plane Figure 57

⎛ ⎞
◦ ⎜ ⊔ ◦⎟
A ⊂ A∪∂A ⊂ ⎜
⎝ Di ⎟
⎠∪ ∂A
◦
Di ⊂A
⎛ ⎞ ⎛ ⎞
⊔ ◦ ⎜ ⊔ ◦⎟ ⎜ ⊔ ◦⎟
Di ⊂ A ⊂ ∂ A ∪ ⎜
⎝ Di ⎟ ⎜
⎠ ⊂ Fε ∪ ⎝ Di ⎟
⎠
◦ ◦ ◦
Di ⊂A Di ⊂A Di ⊂A
⎛ ⎞ ⎛ ⎞
⎜ ⊔ ◦⎟ ⎜ ⊔ ◦⎟
Fε ∪ ⎜
⎝ Di ⎟ ⎜
⎠\⎝ Di ⎟
⎠ ⊂ Fε
◦ ◦
Di ⊂A Di ⊂A
⎛ ⎛ ⎞⎞ ⎛ ⎞
⎜ ⎜ ⊔ ◦ ⎟⎟ ⎜ ⊔ ◦⎟
Area⎜ ⎜
⎝ Fε ∪ ⎝ Di ⎟⎟ ⎜
⎠⎠ − Area ⎝ Di ⎟
⎠ < ε.
◦ ◦
Di ⊂A Di ⊂A

The assertion follows now from Theorem 2.7.1.

◦ ◦
Proposition 2.7.2 Let A1 and A2 be two squareable sets such that A1 ∩ A2 = ∅.
Then their union A = A1 ∪ A2 is also a squareable set and:

area(A) = area(A1 ) + area( A2 ).

Proof The set A = A1 ∪ A2 is squareable because we use Theorem 2.7.2 and the
fact that the boundary of A is embedded in the set Fr A1 ∪ Fr A2 , and this set is of
area zero.
According to Theorem 2.7.1, for any ε > 0, there are E i , Fi ∈ E , i = 1, 2, with
the properties (Fig. 2.13):

E 1 ⊂ A1 ⊂ F1 , E 2 ⊂ A2 ⊂ F2

area(F1 ) − area(E 1 ) < ε

area(F2 ) − area(E 2 ) < ε.

Further, we have:

area(E 1 ) + area(E 2 ) ≤ area(A) ≤ area(F1 ∪ F2 ) ≤ area(F1 ) + area(F2 )

Fig. 2.13 The union of two

squarable sets is a squarable
set A1 A2

A
58 2 Definite Integrals

Fig. 2.14 A plane curve

defined by a polar equation is
squarable

area(E 1 ) + area(E 2 ) ≤ area(A1 ) + area(A2 ) ≤ area(F1 ) + area(F2 ).

From these inequalities, we deduce:

| area(A) − (area(A1 ) + area(A2 ))|

≤ area(F1 ) − area(E 1 ) + area(F2 ) − area(E 2 ) < 2ε

hence area(A) = area(A1 ) + area(A2 ).

Remark 2.7.5 Every finite union of squareable sets with no common interior points
is a squareable set and its area is equal to the sum of the areas of the sets of the
family; in the case of an infinite union, the property is no longer preserved.
Suppose now that a plane curve is defined by the polar equation:

ρ = ρ(θ ), θ ∈ [α, β].

The plane figure bounded by this curve and two rays which start at the pole and
make the angles α and β, respectively, with the polar axis is called a curvilinear
sector (Fig. 2.14).
Proposition 2.7.3 If ρ = ρ(θ ), θ ∈ [α, β] is a continuous positive function, then
the corresponding curvilinear sector

A = {(θ, ρ); α ≤ θ ≤ β, 0 ≤ ρ ≤ ρ(θ )}

is squareable and has the area given by the formula:

∫β
1
area(A) = ρ 2 (θ )dθ . (2.22)
2
α

Proof Let us consider Δn : α = θ0 < θ1 < . . . < θi−1 < θi < . . . < θn = β be an
arbitrary equidistant partition of the interval [α, β].
2.7 Area of a Plane Figure 59

We denote by m i (respectively, Mi ) the greatest ⎡ lower⎤ bound (respectively, the

least upper bound) of the function ρ = ρ(θ ), θ ∈ θi−1 , θi .
The area of the circular sector O Ri Pi = {(θ, ρ); θi−1 ≤ θ ≤ θi , 0 ≤ ρ ≤ ρ(θ )}
is 21 m i2 (θi − θi−1 ), and the area of the circular sector O Q i Ri−1 is 21 Mi2 (θi − θi−1 )
(Fig. 2.14).
If we denote by Pn (respectively, Q n ) the union of these n circular sectors O Ri Pi
(respectively, O Q i Ri−1 ), then Pn ⊂ A ⊂ Q n and we have:

v
n
1 v
n
1
area(Pn ) = m i2 (θi − θi−1 ), area(Q n ) = Mi2 (θi − θi−1 ).
i=1
2 i=1
2

We note that the both sums are in fact the Darboux sums of the function 21 ρ 2 (θ ),
θ ∈ [α, β] with respect to the partition Δn . Taking into account that limn→∞ ∥Δn ∥ =
limn→∞ β−αn
= 0 and the function 21 ρ 2 (θ ) is integrable on [α, β], it results that:

∫β
1
lim area(Pn ) = lim area(Q n ) = ρ 2 (θ ) dθ . (2.23)
n→∞ n→∞ 2
α

On the other hand, we know that Pn and Q n are squareable sets (Example 2.7.2),
whence it results that for any ε > 0, there are two elementary sets E n and Fn with
the properties:

E n ⊂ Pn ⊂ A ⊂ Q n ⊂ Fn

and
ε ε
area(Pn ) − area(E n ) < , area(Fn ) − area(Q n ) < .
3 3

In addition, according to (2.23), we can assume that area(Q n ) − area(Pn ) < 3ε .

Therefore, we have area(Fn ) − area(E n ) < ε, hence, according to Theorem 2.7.1,
the set A is squareable and:

∫β
1
area(A) = ρ 2 (θ )dθ .
2
α

Example 2.7.3 Compute the area of the plane figure bounded by the cardioid.
The polar equation of the cardioid is ρ = a(1 + cos θ ), θ ∈ [0, 2π ].
Since the curve is symmetric with respect to the axis O x, it is sufficient to compute
a half of its area (Fig. 2.15). According to (2.22), we have:
60 2 Definite Integrals

Fig. 2.15 Cardioid

∫π
1 1
· Area(cardioid) = a 2 (1 + cos θ )2 dθ
2 2
0
∫π
a 2 ( )
= 1 + 2 cos θ + cos2 θ dθ
2
0
∫π ( )
a 2
1 + cos 2θ
= 1 + 2 cos θ + dθ
2 2
0
( )|
a2 θ sin 2θ ||π 3a 2 π
= θ + 2 sin θ + + |0 = .
2 2 2 4

3a 2 π
Therefore, we obtain the area of the cardioid: area(cardioid) = 2
.

2.8 Approximating Definite Integral

The computation of the definite integral is usually done using Leibniz–Newton

formula:

∫b |
|
f (x) dx = F(x)|b = F(b) − F(a),
a
a

where F is an antiderivative of f .
Unfortunately, for very few functions we can find the antiderivatives, even if we
know that these antiderivatives exist. Here are some examples of antiderivatives that
cannot be calculated (in the sense that these antiderivatives are not expressed by
elementary functions):
∫ ∫ ∫ x ∫ ∫
sin x cos x e x 1
dx, dx, dx, dx, dx,
x x x ln x ln x
∫ ∫ ∫ ∫
( ) ( )
sin x 2 dx, cos x 2 dx, ex dx, e−x dx,
2 2
2.8 Approximating Definite Integral 61
∫ √ ∫
1
1 − k 2 · sin2 x dx, √ dx, |k| < 1 and so on.
1 − k 2 · sin2 x

If we cannot find an antiderivative of the function f , then, for the computation of

∫b
the integral a f (x) dx we use approximating methods. We will do the same if the
function is not defined by an analytical formula but is given by a table of values.
The basic idea is to approximate the function f by a polynomial function Pn and
∫b ∫b
to approximate a f (x) dx by a Pn (x) dx, because it is easy to integrate polynomial
functions.

Theorem 2.8.1 For any function f : [a, b] → R and for any (n+1)—distinct points
x0 , x1 , . . . , xn in the interval [a, b], called nodes, there is an unique polynomial L n
of degree n that interpolates the function f with respect to the nodes xi , i = 0, n,
i.e.:

L n (xi ) = f (xi ), i = 0, n.

This polynomial is called the Lagrange interpolation polynomial.

Proof It is easily verified that the following polynomial of degree n:

vn ∏ n
x − xj (x − x1 ) · . . . · (x − xn )
L n (x) = f (xi ) = f (x0 )
x
i=0 j=0 i
− x j (x 0 − x 1 ) · . . . · (x 0 − x n )
j/=i
(x − x0 ) · . . . · (x − xi−1 )(x − xi+1 ) · . . . · (x − xn )
+ ··· + f (xi )
(xi − x0 ) · . . . · (xi − xi−1 )(xi − xi+1 ) · . . . · (xi − xn )
(x − x0 ) · . . . · (x − xn−1 )
+ ··· + f (xn ) (2.24)
(xn − x0 ) · . . . · (xn − xn−1 )

has the property L n (xi ) = f (xi ), i = 0, n.

To prove the uniqueness, suppose that there is another polynomial Q n of degree
n, that interpolates the function f with respect to the nodes xi , i = 0, n, thus with
the property:

Q n (xi ) = f (xi ), i = 0, n.

If we denote by R(x) = Q n (x) − L n (x), then R is a polynomial of degree almost

n and R is equal to zero at n + 1 distinct points, namely:

R(xi ) = f (xi ) − f (xi ) = 0, i = 0, n.

Such a polynomial can be only the identical null polynomial, therefore Q n = L n .

Example 2.8.1 Find the expression of the Lagrange polynomial that takes the values
y0 = 3, y1 = 2, y2 = −3 with respect to the nodes x0 = −1, x1 = 0, x2 = 2.
62 2 Definite Integrals

According to Formula (2.24), we have:

(x − 0)(x − 2) (x + 1)(x − 2) (x + 1)(x − 0)

L 2 (x) = ·3+ ·2+ · (−3)
(−1 − 0)(−1 − 2) (0 + 1)(0 − 2) (2 + 1)(2 − 0)
−x 2 − 3x + 4
= .
2
Next, we assume that the nodes are equidistant, i.e.:

xn − x 0
xi = x0 + i h, i = 0, n, where h = .
n
If in the expression of the Lagrange interpolation polynomial (2.24) we make the
change of variable x = x0 + th, then we obtain:
⎛ ⎞
v
n
⎜ ∏
n
t − j⎟
~
L n (t) = L n (x0 + th) = ⎜ ⎟ · f (xi )
⎝ i − j ⎠
i=0 j=0
j/=i
vn
t(t − 1) · . . . · (t − i + 1) · (t − i − 1) · . . . · (t − n)
= · f (xi )
i=0
i (i − 1) · . . . · 2 · 1 · (−1) · (−2) · . . . · (−(n − i ))
1 v
n
n! t(t − 1) · . . . · (t − i ) · . . . · (t − n)
= (−1)n−i · · f (xi )
n! i=0 i! · (n − i )! t −i
v
n
Cni · πn+1 (t)
= (−1)n−i · f (xi )
i=0
(t − i ) · n!

where we denoted by

∏
n
πn+1 (t) = (t − j ). (2.25)
j=0

The expression of the Lagrange interpolation polynomial for equidistant nodes

is:

v
n
Cni πn+1 (t)
~
L n (t) = (−1)n−i · · f (xi ). (2.26)
i=0
n! (t − i )

If we denote by E( f ; x) = f (x) − L n (x) the approximation error at the point

x, then it results that:

E( f ; xi ) = 0 and E( f ; x) /= 0, ∀x /= xi , i = 0, n.
2.8 Approximating Definite Integral 63

We will also denote by:

∏
n
( ) ( )
Un+1 (x) = x − x j = (x − x0 ) · . . . · x − x j · . . . · (x − xn ). (2.27)
j=0

Obviously, Un+1 is a polynomial of degree n + 1. The following theorem can be

proved (for proof see [13], Theorem 4.1.2):

Theorem 2.8.2 If f : [a, b] → R is a function of C n+1 —class on [a, b], then for
any x ∈ [a, b], there is ξx ∈ (a, b) such that:

f (n+1) (ξx )
E( f ; x) = · Un+1 (x). (2.28)
(n + 1)!
| |
Corollary 2.8.1 If there is M > 0 such that | f (n+1) (x)| < M, ∀x ∈ [a, b], then:

M
|E( f ; x)| ≤ · |Un+1 (x)|, ∀x ∈ [a, b].
(n + 1)!

Example 2.8.2 Evaluate the approximation error of the function

f (x) = ln x, x ∈ (0, ∞)

at the point x = 0.6, using the Lagrange interpolation polynomial with respect to
the nodes x0 = 0.4; x1 = 0.5; x2 = 0.7; x3 = 0.8.

In this case, we have:

6 | | 6
f (4) (x) = − and | f (4) (x)| ≤ ≈ 234.4, x ∈ [0.4; 0.8].
x 4
(0.4)4

From Corollary 2.8.1, we deduce:

234.4
|E(ln x; x)| ≤ · |(x − 0.4)(x − 0.5)(x − 0.7)(x − 0.8)|.
4!
In particular, for x = 0.6, we have:

234.4
|E(ln x; 0.6)| ≤ · 0.0004 ≈ 0.0039.
24
According to Formulas (2.25), (2.27) and (2.28), we obtain the expression of the
approximation error for equidistant nodes, i.e.:
(n+1)
~ f ; t) = f (ξt )
E( · πn+1 (t)h n+1 . (2.29)
(n + 1)!
64 2 Definite Integrals

Let L n be the Lagrange polynomial interpolating the function f : [a, b] → R

with respect to the equidistant nodes xi = x0 + i h, i = 0, n, where h = b−a n
.
By integrating the relation f (x) = L n (x) + E( f ; x), x ∈ [a, b], we obtain:

∫b ∫b ∫b
f (x) dx = L n (x) dx + E( f ; x) dx. (2.30)
a a a

If in the right member of the equality (2.30) we make the change of variable
x = a + t · h and take into account the relations (2.26) and (2.29), it results that:

∫b ∫n ∫n
f (x) dx = ~
L n (t) · h dt + E( f ; t) · h dt
a 0 0
⎛ ⎞
v
n ∫n
⎝(−1)
i
n−i C n πn+1 (t) ⎠
=h · dt · f (xi )
i=0
n! (t − i )
0
∫n
h n+2
+ πn+1 (t) · f (n+1) (ξt ) dt. (2.31)
(n + 1)!
0

Next, we use the notations:

∫n
Cni πn+1 (t)
di = (−1) n−i
dt, i = 0, n (2.32)
n! (t − i )
0

Ai = h · di , i = 0, n (2.33)

∫n
h n+2
R( f ) = πn+1 (t) f (n+1) (ξt ) dt. (2.34)
(n + 1)!
0

Taking into account in (2.31) the above notations, we obtain:

∫b v
n
f (x) dx = Ai · f (xi ) + R( f ). (2.35)
a i=0

The numbers di are called the Newton–Côtes coefficients and Formula (2.35) is
called the Newton–Côtes quadrature formula.
For n ∈ {1, 2, 3}, the Newton–Côtes quadrature formula have special names:
trapezoidal formula (n = 1), Simpson formula (n = 2) and Newton 3/8 formula
(n = 3).
Next, we present the trapezoidal formula and Simpson formula.
2.8 Approximating Definite Integral 65

Trapezoidal formula (n = 1)
In this case we have only two nodes, x0 = a, x1 = b, and h = b − a. From (2.32)
and (2.33), we obtain the following Newton–Côtes coefficients:

∫1 ∫1
(−1)1 C10 t · (t − 1) 1 h b−a
d0 = · dt = − (t − 1) dt = , A0 = =
1! t 2 2 2
0 0
∫1 ∫1
(−1) 0
C11 t · (t − 1) 1 h b−a
d1 = · dt = t dt = , A1 = = .
1! t −1 2 2 2
0 0

Trapezoidal formula is:

∫b
b−a
f (x) dx = ( f (a) + f (b)) + R( f ). (2.36)
2
a

∫b
Therefore, a f (x) dx is approximated by b−a 2 (
f (a) + f (b)).
From a geometrical point of view, the trapezoidal formula approximates the area
of the plane figure bounded by the graph of the function f , the axis O x and the
straight lines x = a, x = b to the area of the hatched trapezoid in Fig. 2.16. From
(2.34), we deduce that the error is:

∫1
h3
R( f ) = t(t − 1) f '' (ξt ) dt.
2
0

{| | }
If we denote by M2 = sup | f '' (x)|; x ∈ [a, b] , then it results that:

Fig. 2.16 The y

approximation of the area of
f
a subraph of a function by a
trapezoid

O a b x
66 2 Definite Integrals

∫1 ∫1
(b − a)3 M2 (b − a)3 M2
|R( f )| ≤ |t(t − 1)| dt = t(1 − t) dt
2 2
0 0
(b − a)3 M2
= .
12
∫b
Therefore, if we approximate a f (x) dx by 2 (
b−a
f (a) + f (b)), then the absolute
(b−a) M2
3
error is bounded by 12
.
Obviously, such error is too big, so in practice is used the repeated trapezoidal
method.
The repeated trapezoidal method consists in dividing the interval [a, b] into
n equal subintervals and approximating the integral on each subinterval by the
trapezoidal formula.
Therefore, we have:

∫b vn
xi − xi−1
f (x) dx ≈ ( f (xi−1 ) + f (xi )).
i=1
2
a

Since x0 = a, xn = b and xi − xi−1 = b−a

n
, the repeated trapezoidal formula is:

∫b ( )
b−a v
n−1
f (x) dx ≈ f (a) + f (b) + 2 f (xi ) . (2.37)
2n i=1
a

The error that is obtained admits the following upper bound:

M2 (b − a)3 (b − a)3 M2
|R( f )| ≤ n · · = . (2.38)
12 n3 12n 2
∫2 x2
Example 2.8.3 Compute the definite integral −2 e− 2 dx using the repeated trape-
zoidal formula, with a smaller error than ε = 10−3 .
We notice that this integral cannot be computed by Leibniz–Newton formula,
x2
because the antiderivative of the function f (x) = e− 2 is not expressed by elementary
functions.
Using the above algorithm we have:

x2
a = −2, b = 2, f (x) = e− 2

x2 ( ) x2 ( ) x2
f ' (x) = −xe− 2 , f '' (x) = x 2 − 1 e− 2 , f ''' (x) = x 3 − x 2 e− 2 .
2.8 Approximating Definite Integral 67

√ √
x −∞ −2 − 3 0 3 2 +∞
f ''' (x) + + + + 0 − 0 + 0 − − − −
− 23 − 23
f '' (x) 0 ↗ 3e−2 ↗ 2e ↘ −1 ↗ 2e ↘ 3e−2 ↘ 0

Since 2 < e < 3, it results that e32 < 43 < 1 and e√2 e < 2√2 2 < 1.
From the table of variations of the function f '' , we deduce that:
{| | } | |
M2 = sup | f '' (x)|; x ∈ [−2, 2] = | f '' (0)| = 1.

If we put (b−a)
3

12 n 2
M2
= 43
12 n 2
< ε = 10−3 , then the absolute error becomes |R( f )| <
−3
10 .
On the other hand,

43 16
< 10−3 ⇒ n 2 > · 103 = 5333, (3) ⇒ n > 73.03.
12 n 2 3
Therefore, in Formula (2.37) we take n = 74 and we find:

∫2
x2 2 − (−2)
e− 2 dx ≈ ( f (−2) + f (2) + 2( f (x1 ) + · · · + f (x73 ))) = 2.3924
2 · 74
−2

where xi = −2 + i · 4
74
= −2 + i · 2
37
, i = 0.74.
Simpson Formula (n = 2)
Simpson formula corresponds to three nodes, x0 = a, x1 = a+b
2
, x2 = b, and
h = b−a
2
.
The Newton–Côtes coefficients are:

∫2 ∫2
(−1)2 C20 t(t − 1)(t − 2) 1 (2 ) 1 b−a
d0 = dt = t − 3t + 2 dt = , A0 =
2! t 2 3 6
0 0
∫2 ∫2
(−1)1 C21 t(t − 1)(t − 2) (2 ) 4 4(b − a)
d1 = dt = − t − 2t dt = , A1 =
2! t −1 3 6
0 0
∫2 ∫2
(−1) C22 2
t(t − 1)(t − 2) 1 (2 ) 1 b−a
d2 = dt = t − t dt = , A2 = .
2! t −2 2 3 6
0 0
68 2 Definite Integrals

Simpson formula is:

∫b ( ( ) )
b−a a+b
f (x) dx ≈ f (a) + 4 f + f (b) . (2.39)
6 2
a

If f : [a, b] → R is a function of C 4 —class on [a, b], then it can be shown that:

(b − a)5 M4
|R( f )| ≤ (2.40)
2880
where:
{| | }
M4 = sup | f (4) (x)|; x ∈ [a, b] .

From a geometrical point of view, the Simpson formula approximates the area of
the subgraph of the function f to the hatched area in Fig. 2.17, i.e. the area of the
plane figure bounded by the parabola P2 (the 2nd degree Lagrange polynomial that
interpolates the function f with respect to the nodes a, a+b
2
, b), the axis O x and the
straight lines x = a, x = b.
In the same manner as the trapezoidal method, the repeated Simpson method is
used for the best possible approximation. In this case, we divide the interval
⎡ [a, b]
⎤
into 2n equal subintervals, and we apply Formula (2.39) for each interval x2i−2 , x2i .
Thus, we have:

∫b vn ( ( ) )
x2i − x2i−2 x2i−2 + x2i
f (x) dx ≈ f (x2i−2 ) + 4 f + f (x2i ) .
i=1
6 2
a

Since x2i − x2i−2 = b−a

n
, we deduce that the repeated Simpson formula is:

Fig. 2.17 The y f

approximation of a subraph
of a function f by a subgraph
of a parabola P2

O a a+b b x
2
2.8 Approximating Definite Integral 69

∫b ( )
b−a v
n v
n−1
f (x) dx ≈ f (a) + f (b) + 4 f (x2i−1 ) + 2 f (x2i ) . (2.41)
6n i=1 i=1
a

From (2.40), it follows that:

(b − a)5 M4
|R( f )| ≤ . (2.42)
2880 n 4
∫2 x2
Example 2.8.4 Compute the definite integral −2 e− 2 dx using the repeated
Simpson formula, with a smaller error than ε = 10−3 .
x2
We have: a = −2, b = 2, f (x) = e− 2 ;
( ) x2 ( ) x2
f (4) (x) = x 4 − 6x 2 + 3 e− 2 , f (5) (x) = −x x 4 − 10x 2 + 15 e− 2 .

√ √ √ √
x −2 − 5 − 10 0 5 − 10 2
f (5) − − 0 + 0 − 0 + +
f (4) −5e−2 ↘ m ↗ 3 ↘ m ↗ −5e−2

We remark that:
( / ) (/ ) √
√ √ 8 − 4 10 not
f (4) − 5 − 10 = f (4) 5 − 10 = √
5− 10
= m < 0 and
e 2
| |
| 8 − 4√10 | 4√10 − 8
| |
|m| = | √ |= √ < 3.
| e 5−2 10 | 5− 10
e 2

Indeed, using the known inequality ex > x + 1, x /= 0, it results that:

( √ ) √
√
5− 10 5− 10 21 3 10 √
3e 2 >3 1+ = − > 4 10 − 8,
2 2 2

whence, we deduce that:

√
4 10 − 8
√
5− 10
< 3.
e 2

Therefore
{| | } | |
M4 = sup | f (4) (x)|; x ∈ [−2, 2] = | f (4) (0)| = 3
70 2 Definite Integrals

According to (2.42), to find the number of nodes we put the condition:

3 · 45
< 10−3
2880 n 4
whence, it results that:
/ /
4 3 · 45 · 1000 3200 √
n≥ = ≈ 1067 ≈ 5.715.
4 4

2880 3

It can be taken n = 6 and applying Formula (2.41), we obtain:

∫2 ( )
2 1 v
6 v
5
− x2
e dx ≈ f (−2) + f (2) + 4 f (x2i−1 ) + 2 f (x2i )
18 i=1 i=1
−2

≈ 2.3925

where xi = −2 + i · 12 4
= −2 + i · 13 , i = 0.12.
We notice that in this example the computations for the Simpson method are less
than for the trapezoidal method. Indeed, the Simpson method computes the values
of function in 12 nodes, while the trapezoidal method calculates the values of the
function in 74 nodes.
At the end of this paragraph we mention the fact that, in practice, the approximate
computation of integrals is done using specialized program packages, of which the
most used are MATLAB, Maple, Mathcad, and so on.
Chapter 3
Improper Integrals

3.1 Convergence and Divergence of Improper Integrals

The definite integral theory was made for bounded functions defined on closed and
bounded intervals. In the following, we will generalize the notion of integral for
situations in which only one or ∫both of these ∫conditions are∫no longer fulfilled, giving
∞ a ∞
a sense of integrals of the form a f (x)dx, −∞ f (x)dx, −∞ f (x)dx, respectively,
∫b
a f (x)dx, where a and b are finite and f is unbounded on [a, b) or (a, b] or (a, b).
The first integrals, for which the integration interval is infinite, are called improper
integrals of the first kind, while the integrals for which a and b are finite and the
function f is not bounded (i.e. lim x→b | f (x)| = ∞ or lim x→a | f (x)| = ∞) are called
x<b x>a
improper integrals of the second
∫∞ kind. We will study from the first type of improper
integrals, the integral of form a f (x)dx and from the second type, integrals of form
∫b
a f (x)dx, with lim x→b f (x) = ±∞. We will present both cases as unitary.
x<b

Definition 3.1.1 A function f : [a, b) → R, where b is finite or not, is called locally

integrable on [a, b) if it is integrable on any compact interval [a, u] ⊂ [a, b).
Let f : [a, b) → R be a locally integrable function on [a, b). It is natural to try to
define:

∫b ∫u
f (x)dx = lim f (x)dx.
u→b
a u<b a

It is possible that the limit of the right member exists and is finite, or it is infinite or it
∫b
does not exist. In the first case, the improper integral a f (x)dx is called convergent,
and in the other cases it is divergent.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 71
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_3
72 3 Improper Integrals

Example 3.1.1
∫ ∞ ∫ u
1 1 |u
dx = limu→∞ dx = limu→∞ tan−1 x |1
(1) 1 1+x 2
1 1+x
2
( −1 −1
) π π π
= limu→∞ tan u − tan 1 = − = .
∫1 1 ∫u 12 4 4 |u
|
(2) 0 1−x dx = limu→1 0 1−x dx = limu→1 (− ln|1 − x|)| =
u<1 u<1 0
limu→1 (− ln|1 − u|) = ∞.
∫ ∞ u<1 ∫ u
cos x dx = limu→∞ cos x dx = limu→∞ sin x|u0
(3) 0 0
= limu→∞ sin u (the limit does not exist).
Definition 3.1.2 Let f : [a, b) → R be a locally integrable function, where b is finite
or not. ∫u
If the limit limu→b a f (x)dx exists and it is finite, then we say that the integral
∫b u<b
f (x)dx is convergent. Otherwise, if the limit does not exist or it is infinite, we
a ∫b
say that the integral a f (x)dx is divergent.
Therefore, in the case of convergence, the value of the improper integral is:

∫b ∫u
f (x)dx = lim f (x)dx
u→b
a u<b a

∫∞ π
In Example 3.1.1, the first integral is convergent and its value is 1
1
1+x 2
dx = 4
,
and the other two integrals are divergent.

Remark 3.1.1 Let f : [a, b) → R be a locally integrable function, with b finite or

∫b ∫b
not. Then for any a < c < b, the integrals a f (x)dx and c f (x)dx have the same
nature (the nature of an improper integral means its convergence or divergence).
The statement results immediately from the decomposition:

∫b ∫c ∫b
f (x)dx = f (x)dx + f (x)dx.
a a c

Definition 3.1.3 Let f : (a, b] → R be a locally integrable function on (a, b] (i.e.

integrable on any compact interval [v, b] ⊂ (a, b]), where a is finite or not.
∫b ∫b
If the limit limv→a v f (x)dx exists and it is finite, then we say that a f (x)dx
v>a
is convergent. Otherwise, if the limit does not exist or it is infinite, we say that
∫b
a f (x)dx is divergent.
3.1 Convergence and Divergence of Improper Integrals 73

Therefore, in the case of convergence, we have:

∫b ∫b
f (x)dx = v→a
lim f (x)dx.
v>a
a v

Example 3.1.2 Compute the following improper integrals:

∫ 0 ∫ 0 ( √ )|0
1 1 |
√ dx = limv→−∞ √ dx = limv→−∞ −2 1 − x |
(1) −∞ 1−x v 1−x v
( √ )
= limv→−∞ −2 + 2 1 − v = ∞.
Therefore, this improper integral is divergent.
∫ 0
1
∫ 0
1 ( √ )||0
√ dx = limv→−1 √ dx = limv→−1 2 1 + x ||
(2) −1 1+x v>−1 v 1+x v>−1 v
( √ )
= limv→−1 2 − 2 1 + v = 2.
v>−1

In what follows, we present three improper test integrals.

∫∞
Example 3.1.3 Prove that 1 x1α dx is convergent if α > 1 and divergent if α ≤ 1.
For any u > 1, we have:

∫u ∫ ∫
1 ln x||u1 , if α = 1 ln u, if α = 1
dx = x 1−α |
u = u 1−α −1
xα 1−α |
, if α /= 1 1−α
, if α /= 1.
1 1

We notice that if α > 1, then:

∫u
1 u 1−α − 1 1
lim α
dx = lim =− ∈ R,
u→∞ x u→∞ 1−α 1−α
1

∫∞ 1
whence we deduce that the improper integral 1 xα
dx is convergent and has the
value:
∫∞
1 1
dx = .
xα α−1
1

∫u ∫∞
If α ≤ 1, then limu→∞ 1 x1α dx = ∞, thus 1 x1α dx is divergent.
∫∞ 1
In particular, the improper integral 1 x 2 dx is convergent and its the value is
∫∞ 1 ∫∞ 1
1 x 2 dx = 1, and the improper integral 1
√ dx is divergent.
x
∫b 1 ∫b 1
Example 3.1.4 Prove that the improper integrals a (x−a) α dx and
a (b−x)α
dx are
convergent if α < 1 and divergent if α ≥ 1, where a and b are finite.
74 3 Improper Integrals

For any a < v < b, we have:

∫b ∫
1 ln b−a
v−a
if α = 1
dx =
(x − a)α (b−a)1−α −(v−a)1−α
1−α
if α /= 1.
v

We notice that if α < 1, then limv→a (v − a)1−α = 0, which implies that the
∫b 1 v>a
integral a (x−a) α dx is convergent and has the value:

∫b
1 (b − a)1−α
α dx = ,
(x − a) 1−α
a

∫b 1
and if α ≥ 1, the improper integral a (x−a) α dx is divergent.
∫2 1 ∫2
In particular, 1 √x−1 dx is convergent and 1 (x−1)1√x−1 dx is divergent.
∫b 1
Similarly, it is shown that a (b−x) α dx is convergent if α < 1 and divergent if
∫1 1 ∫1 1
α ≥ 1. In particular, 0 √1−x dx is convergent and 0 1−x dx is divergent.

Definition 3.1.4 Let f : [a, c) ∪ (c, b] → R be a locally integrable function on

[a, c) ∪ (c, b] (i.e. integrable on any compact interval [u, v] ⊂ [a, c) ∪ (c, b]), where
a and b are finite. ∫b
We say that the improper integral a f (x)dx is convergent if both improper
∫c ∫b
integrals a f (x)dx and c f (x)dx are convergent, that is, if the limit:
⎛ c−ε ⎞
∫ ∫b
lim+ ⎝ f (x)dx + f (x)dx ⎠
ε→0
η→0+ a c+η

exists and it is finite. In this case, the value of the integral is given by:

∫b ∫u ∫b
f (x)dx = u→c
lim f (x)dx + v→c
lim f (x)dx.
u<c v>c
a a v

Definition 3.1.5 Let f : [a, c) ∪ (c, b] → R be a locally integrable function on

[a, c) ∪ (c, b].
We say that f is Cauchy integrable on [a, b] if there is the finite limit
⎛ c−ε ⎞
∫ ∫b ∫b
lim ⎝ f (x)dx + f (x)dx ⎠ = (v.p.) f (x)dx.
ε→0
ε>0 a c+ε a

The limit is called the principal value of the improper integral in Cauchy sense.
3.1 Convergence and Divergence of Improper Integrals 75
∫2 1
Example 3.1.5 The improper integral −1 x dx is divergent because:
⎛ −ε ⎞
∫ ∫2
1 1 2ε
lim+ ⎝ dx + dx ⎠ = lim+ (ln ε + ln 2 − ln η) = lim+ ln
ε→0
+
x x ε→0
+
ε→0
+
η
η→0 −1 η η→0 η→0

and this limit does not exists.

On the other hand, we have:
⎛ −ε ⎞
∫2 ∫ ∫2
1 1 1
(v.p.) dx = lim+ ⎝ dx + dx ⎠ = lim (ln ε + ln 2 − ln ε) = ln 2.
x ε→0 x x ε→0
−1 −1 ε

Definition 3.1.5 Let f : R → R be a locally integrable function on R (i.e. f is

integrable on any∫ ∞compact interval [u, v] ⊂ R). ∫u
We say that −∞ f (x)dx is convergent if the limu→∞ v f (x)dx exists and it
v→−∞
is finite.
Therefore, if the integral is convergent, we have:

∫∞ ∫u
f (x)dx = lim f (x)dx.
u→∞
v→−∞
−∞ v

The following limit is called the principal value of the integral in Cauchy sense.

∫∞ ∫u
(v.p.) f (x)dx = lim f (x)dx.
u→∞
−∞ −u

Convergence in Cauchy sense∫ ∞does not imply the convergence of improper inte-
gral. Sometimes the integral −∞ f (x)dx is divergent, but its principal value is
finite.
∫∞
Example 3.1.6 Study the nature of the improper integral −∞ sin x dx.
∫u
Since limu→∞ v sin x dx = limu→∞ (− cos u + cos v) does not exists, it
∫ v→−∞
∞
v→−∞
results that −∞ sin x dx is divergent. On the other hand, we have:

∫∞ ∫u
(v.p.) sin x dx = lim sin x dx = lim (− cos u + cos u) = 0.
u→∞ u→∞
−∞ −u
76 3 Improper Integrals

3.2 Convergence Criteria for Improper Integrals

Theorem 3.2.1 Let f : [a, b) → R be a locally integrable function on [a, b), where
b is finite or not. The necessary and sufficient condition for the improper integral
∫b
a f (x)dx to be convergent is for any ε > 0, there is a < δε < b such that

| u '' |
|∫ |
| |
| f (x)dx | < ε, for all u ' , u '' ∈ [δε , b).
| |
|' |
u

∫u
Proof For any a < u < b, we shall denote by F(u) = a f (x)dx.
∫b
According to Definition 3.1.2, the integral a f (x)dx is convergent if and only if
the limit limu→b F(u) exists and it is finite.
u<b
On the other hand, from Cauchy–Bolzano theorem (Theorem 5.4.3 from [6]), it
results that this limit exists and it is finite if and only if for any ε > 0, there is a
neighborhood Vε of b such that:
| ( ') ( )|
| F u − F u '' | < ε, for any u ' , u '' ∈ Vε ∩ [a, b).

If b is finite, we can assume that Vε has the form Vε = (b − ηε , b + ηε ), where

a < b − ηε < b, and we shall choose δε = b − ηε .
If b = ∞, we can assume that Vε has the form Vε = (δε , ∞), where δε > a.
In both cases, if u ' , u '' ∈ (δε , b), it results that u ' , u '' ∈ Vε ∩ [a, b), hence
| ( ') ( )|
| F u − F u '' | < ε.

On the other hand, it is easily seen that:

| u' | | u '' |
|∫ ∫u '' | |∫ |
| ( ') ( '' )| | | | |
| F u − F u | = | f (x)dx − f (x)dx | = | f (x)dx ||.
| |
|
| | |' |
a a u

∫b
Therefore, a f (x)dx is convergent if and only if ∀ε > 0, ∃a < δε < b such that
for any u ' , u '' ∈ (δε , b), we have:
| u '' |
|∫ |
| |
| f (x)dx | < ε.
| |
|' |
u

∫b
Definition 3.2.1 We say that the improper integral a f (x)dx is absolutely
∫b
convergent if the improper integral a | f (x)|dx is convergent.
3.2 Convergence Criteria for Improper Integrals 77

Theorem 3.2.2 Any absolutely convergent improper integral is convergent.

Proof The statement follows immediately from Theorem 3.2.1 and from the
inequality:
| u '' | | u '' |
|∫ | |∫ |
| | | |
| f (x)dx | ≤ | | f (x)|dx |.
| | | |
|' | |' |
u u

Remark 3.2.1 The reciprocal statement of Theorem 3.2.2 is not generally true.
There are improper convergent integrals without being absolutely convergent; a such
∫∞
example is Dirichlet integral 0 sinx x dx which is convergent (Example 3.2.9), but it
is not absolutely convergent.

Theorem 3.2.3 (The first comparison test) Let f, g: [a, b) → R+ be two locally
integrable functions on [a, b), where b is finite or not, such that 0 ≤ f (x) ≤
g(x), ∀x ∈ [a, b).
∫b ∫b
(1) If a g(x)dx is convergent, then a f (x)dx is convergent and:

∫b ∫b
0≤ f (x)dx ≤ g(x)dx.
a a

∫b ∫b
(2) If a f (x)dx is divergent, then a g(x)dx is divergent.
∫u ∫u
Proof Let F(u) = a f (x)dx and G(u) = a g(x)dx, where a < u < b.
Since 0 ≤ f (x) ≤ g(x), ∀x ∈ [a, b), from Proposition 2.6.3 and Proposition
2.6.4, it results that:

0 ≤ F(u) ≤ G(u), ∀a < u < b.

On the other hand, the functions F and G are increasing functions, because
F ' (u) = f (u) ≥ 0 and G ' (u) = g(u) ≥ 0, ∀a < u < b.
∫b
(1) If we suppose now that a g(x)dx is convergent, it follows that the limit L =
limu→b G(u) exists and it is finite.
u<b
Since F ≤ G, it results that F(u) ≤ G(u) ≤ L , ∀a < u < b. The fact that
F is increasing and upper bounded by L on [a, b), implies that there is the limit.
∫b
limu→b F(u) ≤ L, thus a f (x)dx is convergent.
u<b ∫b
(2) If we suppose that a f (x)dx is divergent, it results that limu→b F(u) = ∞ and
∫b u<b
even more limu→b G(u) = ∞, hence a g(x)dx is divergent.
u<b
78 3 Improper Integrals
∫∞
Example |3.2.1 | Study the nature of the improper integral 1 sinx 32x dx.
∫∞ 1
Since | sinx 32x | ≤ x13 , ∀x ∈ [1, ∞) and dx is convergent (Example 3.1.3),
∫ ∞ | sin 2x1 | x 3
from Theorem 3.2.3, it results that 1 | x 3 |dx is convergent.
∫∞
Therefore 1 sinx 32x dx is absolutely convergent, hence it is convergent according
to Theorem 3.2.2.

Theorem 3.2.4 (The second comparison test) Let f, g: [a, b) → R+ be two locally
integrable functions on [a, b), where b is finite or not, such that f (x) ≥ 0, g(x) >
0, ∀x ∈ [a, b) and there is the limit

f (x)
lim = l ∈ [0, ∞].
x→b g(x)
x<b

∫b ∫b
(1) If l ∈ (0, ∞), then the integrals a f (x)dx and a g(x)dx have the same
nature. ∫b ∫b
(2) If l = 0 and a g(x)dx is convergent, then a f (x)dx is also convergent.
∫b ∫b
(3) If l = ∞ and a g(x)dx is divergent, then a f (x)dx is also divergent.

Proof
f (x)
(1) We choose ε > 0 with the property l − ε > 0. Since lim x→b g(x)
= l, it results
x<b
that there is δε > 0 such that a < b − δε and we have:

f (x)
l −ε < < l + ε, ∀x ∈ [b − δε , b).
g(x)
∫b ∫b
If we suppose that a f (x)dx is convergent, then both integrals b−δε f (x)dx
∫ b f (x) f (x)
and b−δε l−ε dx are convergent. Since g(x) < l−ε , ∀x ∈ [b − δε , b), from
∫b
Theorem 3.2.3, it follows that the integral b−δε g(x)dx is convergent.
∫b ∫ b−δ ∫b ∫b
Since a g(x)dx = a ε g(x)dx + b−δε g(x)dx, it results that a g(x)dx is
convergent (see Remark 3.1.1).
∫b ∫b
If we assume that a g(x)dx is convergent, then the integrals a (l + ε)g(x)dx and
∫b
b−δε (l + ε)g(x)dx are convergent. Since f (x) < (l + ε)g(x), ∀x ∈ [b − δε , b),
∫b
from Theorem 3.2.3, it results that the integral b−δε f (x)dx is convergent.
∫b ∫ b−δ ∫b ∫b
Since a f (x)dx = a ε f (x)dx + b−δε f (x)dx, we deduce that a f (x)dx is
convergent.
∫b
(2) If l = 0, and the integral a g(x)dx is convergent, then the following two
∫b ∫b
integrals a ε g(x)dx and b−δε ε g(x)dx are convergent. From the inequality
< ε g(x), x ∈ [b − δε , b) and from Theorem 3.2.3, it results that
∫ b(x)
f
∫b
b−δε (x)dx is convergent, hence the integral a f (x)dx is also convergent.
f
3.2 Convergence Criteria for Improper Integrals 79

f (x)
(3) If l = ∞, then for any ε > 0, there is δε > a such that g(x) > ε, for any x ∈
∫b ∫b
[b − δε , b). Because the integrals a g(x)dx and b−δε ε g(x)dx are divergent,
then from inequality f (x) > ε g(x), x ∈ [b − δε , b) and from Theorem 3.2.3,
∫b ∫b
we deduce that b−δε f (x)dx is divergent, hence a f (x)dx is divergent.
Example 3.2.2 Study the nature of the improper integrals:
∫ ∞ −1 ∫1 x
(1) 1 tanx x dx and (2) 0 cos
√ dx.
x

tan−1 x
(1) Let us denote by f (x) = x
> 0 and by g(x) = 1
x
> 0, ∀x ∈ [1, ∞).
f (x) −1
Since lim x→∞ g(x) = lim x→∞ tanx x · x = lim x→∞ tan−1 x = π2 ∈ (0, ∞) and
∫∞ 1 ∫ ∞ tan−1 x
1 x dx is divergent (Example 3.1.3), from Theorem 3.2.4, it results that 1 x
dx
is divergent.
(2) If we denote by f (x) = cos
√x
x
> 0 and by g(x) = √1
x
> 0, ∀x ∈ (0, 1], then

f (x) cos x √
lim = lim √ · x = lim cos x = 1 ∈ (0, ∞).
x→0 g(x) x→0 x x→0
x >0 x>0 x>0

∫1 ∫1
Since the integral 0 √1x dx is convergent (Example 3.1.4), it results that 0 cos
√ x dx
x
is convergent, according to Theorem 3.2.4.
Theorem 3.2.5 Let f : [a, ∞) → R+ be a locally integrable function on [a, ∞).
∫∞
(1) If there is α > 1 such that lim x→∞ x α · f (x) /= ∞, then the integral a f (x)dx
is convergent. ∫∞
(2) If there is α ≤ 1 such that lim x→∞ x α · f (x) > 0, then the integral a f (x)dx
is divergent.
∫∞
In particular, if there is lim x→∞ f (x) and a f (x)dx is convergent, then
lim x→∞ f (x) = 0.
Proof
(1) Let α > 1 and l = lim x→∞ x α · f (x) < ∞. On account of the definition of the
limit of a function, it results that for ∀ ε > 0, ∃δε > 0 such that:

l − ε < x α · f (x) < l + ε, ∀x > δε

and further:
l −ε l +ε
< f (x) < α , ∀x > δε .
xα x
∫∞
As δε l+ε dx is convergent (Example 3.1.3), according to Theorem 3.2.3,
xα ∫∞
it results that the integral δε f (x)dx is convergent. From Remark 3.1.1, we
∫∞
deduce now that the integral a f (x)dx is convergent.
80 3 Improper Integrals

(2) Let α ≤ 1 and l = lim x→∞ x α · f (x) ∈ (0, ∞). Since l > 0, we can assume
that there is ε > 0 with the property 0 < ε < l. For such ε > 0, there is δε > a
such that:

l − ε < x α · f (x) < l + ε, ∀x > δε

and further:
l −ε l +ε
α
< f (x) < α , ∀x > δε .
x x
∫∞
Because the integral δε l−ε dx is divergent (Example 3.1.3), from Theorem 3.2.3,
∫ x∞α
we deduce that the integral δε f (x)dx is divergent, and from Remark 3.1.1, we have
∫∞
that the integral a f (x)dx is divergent.
If there is α ≤ 1 such that l = lim x→∞ x α · f (x) = ∞, then for ∀ε > 0, ∃ δε > a
with the property x α · f (x) > ε, ∀x ∈ (δε , ∞). Therefore
ε
f (x) > , ∀x > δε .
xα
∫∞ ∫∞
Since the integral δε xεα dx is divergent, it follows that δε f (x)dx is divergent,
∫∞
hence a f (x)dx is divergent (Remark 3.1.1).
∫ ∞ P(x)
Example 3.2.3 Study the convergence of the integral a Q(x) dx, where P and Q are
polynomial functions, |such that
| degree(Q) ≥ degree(P)+2 and Q(x) /= 0, ∀x > a.
2 | P(x) |
∫ ∞ P(x)
Since lim x→∞ x · | Q(x) | < ∞, from Theorem 3.2.5, it results that a Q(x) dx is
absolutely convergent, so it is convergent, according to Theorem 3.2.2.
∫∞
Example 3.2.4 Prove that the Euler–Poisson integral −∞ e−x dx is convergent.
2

∫ ∞ −x 2 ∫ ∞ −x 2
It is obvious that −∞ e dx = 2 0 e dx.
Since lim x→∞ x 2 · e−x = lim x→∞ exx 2 = 0, according to Theorem 3.2.5, it results
2 2

∫∞
that the integral 0 e−x dx is convergent.
2

Theorem 3.2.6 Let f : [a, b) → R+ , be a locally integrable function on [a, b), with
b finite or not.

(1) If there is α < 1 such that lim x→b (b − x)α · f (x) < ∞, then the integral
∫b x<b

a f (x)dx is convergent.
(2) If there is α ≥ 1 such that lim x→b (b − x)α · f (x) > 0, then the integral
∫b x<b

a f (x)dx is divergent.

Proof The proof is similar with the proof of Theorem 3.2.5, using the fact that
∫b 1
a (b−x)α dx is convergent for α < 1 and divergent for α ≥ 1 (Example 3.1.4).
3.2 Convergence Criteria for Improper Integrals 81

Example 3.2.5 Study the nature of the improper integrals:

∫3 ∫3
1 1
√ dx and √ dx.
(x + 1)(3 − x) x(3 − x)5
2 2

Using Theorem 3.2.6, it results that the first integral is convergent because:

1 1 1 1 1
lim (3 − x) 2 · √ = lim √ = < ∞ and α = < 1,
x→3
x<3
(x + 1)(3 − x) x→3
x<3
x + 1 2 2

while the second integral is divergent because:

5 1 1 1 5
lim (3 − x) 2 · √ = lim √ = √ > 0 and α = ≥ 1.
x→3
x< 3 x(3 − x) x→3
x<3
5 x 3 2

The following theorem is also true.

Theorem 3.2.7 Let f : (a, b] → R+ , be a locally integrable function on (a, b], with
a finite or not.

(1) If there is α < 1 such that lim x→a (x − a)α · f (x) < ∞, then the integral
∫b x>a

a f (x) dx is convergent.
(2) If there is α ≥ 1 such that lim x→a (x − a)α · f (x) > 0, then the integral
∫b x>a

a f (x)dx is divergent.

Proof We proceed in the same manner as in the proof of Theorem 3.2.5, using the
∫b 1
fact that a (x−a) α dx is convergent for α < 1 and divergent for α ≥ 1 (Example

3.1.4).

Example 3.2.6 Study the nature of the following improper integrals:

∫2 ∫2
1 1
√ dx and √ dx.
x(x + 1)3 x 3 (x + 1)
0 0

Using Theorem 3.2.7, it results that the first integral is convergent because:

1 1 1 1
lim x 2 · √ = lim √ = 1 < ∞ and α = < 1,
x→0
x>0 x(x + 1) 3 x→0
x>0 (x + 1) 3 2

while the second integral is divergent because:

3 1 1 3
lim x 2 · √ = lim √ = 1 > 0 and α = ≥ 1.
x→0
x>0 x 3 (x + 1) x→0
x>0
x +1 2
82 3 Improper Integrals

The following theorem establishes a strong connection between an improper inte-

gral with infinite limits and the numerical series that can be constructed using the
integrated function.

Theorem 3.2.8 (Cauchy integral criterion) If ∫f : [1, ∞) → R+ is a monotonically

∞
decreasing function, then the improper integral 1 f (x)dx and the numerical series
Σ ∞
n=1 f (n) have the same nature.

Proof Since f (n) ≤ f (x) ≤ f (n − 1), for any x ∈ [n − 1, n], we have:

∫n
f (n) ≤ f (x)dx ≤ f (n − 1), ∀n ≥ 2.
n−1

Adding these inequalities, we get:

Σ
m ∫m Σ
m−1
f (n) ≤ f (x)dx ≤ f (n), ∀m ≥ 2. (3.1)
n=2 1 n=1

Σ
If we assume that the series ∞n=1 f (n) is convergent, it results that there is M > 0
Σm−1
such that n=1 f (n) < M, ∀m ≥ 2. Using (3.1), we deduce that:

∫m Σ
m−1
f (x)dx ≤ f (n) < M, ∀m ≥ 2.
1 n=1

Let u > 1 be arbitrary and let m ∈ N∗ , m > u. Since f ≥ 0, it results that:

∫u ∫m
f (x)dx ≤ f (x)dx < M.
1 1

∫u ∫∞
Therefore, exists limu→∞ 1 f (x)dx ≤ M,Σ hence 1 f (x)dx is convergent.
∞
Σmnow that the numerical series n=1 f∫(n)
If we assume
m
is divergent, then it results
that limm→∞ n=1 f (n)∫ = ∞ and so that limm→∞ 1 f (x)dx = ∞, whence we
∞
deduce that the integral 1 f (x)dx is divergent.
∫∞ Σ
Example 3.2.7 The improper integral 1 x1α dx and the numerical series ∞ 1
n=1 n α
have the same nature, i.e. are convergent for α > 1 and divergent for α ≤ 1.

Theorem 3.2.9 (Abel–Dirichlet criterion) Let f, g: [a, b) → R, with b finite or not,

be two functions having the properties:
3.2 Convergence Criteria for Improper Integrals 83
|∫ u |
(1) f is continuous and there is M > 0 such that | a f (x)dx | ≤ M, ∀a < u < b.
(2) g is positive, monotonically decreasing, of C 1 —class on [a, b) and
lim x→b g(x) = 0.
x<b
∫b
Then the improper integral f (x)g(x)dx is convergent.
a
∫u
Proof If we denote by F(u) = a f (x)dx, then F ' (u) = f (u).
For any u ' , u '' ∈ (a, b), from Theorem 2.6.5, we have:
|u ''
∫u '' ∫u '' | ∫u ''
|
f (x)g(x)dx = F ' (x)g(x)dx = F(x)g(x)|| − F(x)g ' (x)dx.
|
u' u' '
u' u

On the other hand, g ' (x) ≤ 0, ∀x ∈ [a, b), because g is monotonically

decreasing. According to Corollary 2.6.2, there is ξ in the interval with limits u '
and u '' such that:

∫u '' ∫u ''
( ( ) ( ))
F(x)g ' (x)dx = F(ξ ) g ' (x)dx = F(ξ ) g u '' − g u ' .
u' u'

Therefore, we have:

∫u ''
( ) ( ) ( ) ( ) ( ( ) ( ))
f (x)g(x)dx = F u '' g u '' − F u ' g u ' − F(ξ ) g u '' − g u ' .
u'

|∫ u |
Taking into account to |F(u)| = | a f (x)dx | ≤ M, ∀a < u < b, it results that:
| u '' |
|∫ |
| | ( ( ) ( ))
| f (x)g(x)dx | ≤ 2M g u '' + g u ' .
| |
|' |
u

Since lim x→b g(x) = 0, it follows that for any ε > 0, there is a < δε < b such
x<b
ε
that |g(x)| < 4M
, ∀x ∈ (δε , b). Therefore, if u ' , u '' ∈ (δε , b), then we have:
| u '' |
|∫ | ( )
| |
| f (x)g(x)dx | ≤ 2M ε + ε = ε
| |
|' | 4M 4M
u

∫b
hence the integral a f (x)g(x)dx is convergent, according to Theorem 3.2.1.
84 3 Improper Integrals
∫∞
Example 3.2.8 Prove that the improper integral 1 sinx x dx is convergent.
Let us consider the functions f (x) = sin x and g(x) = x1 , x ∈ [1, ∞).
One can easily seen that the functions f and g verify the conditions of
Theorem 3.2.9: f is continuous on [1, ∞), g is positive, monotonically decreasing,
of C 1 —class on [1, ∞) and lim x→∞ g(x) = lim x→∞ x1 = 0. In addition, we have:
| u | | u |
|∫ | |∫ |
| | | | | |
| f (x)dx | = | sin x dx | = | − cos x|u | = |cos 1 − cos u| ≤ 2, ∀u > 1.
| | | | 1
| | | |
1 1

∫∞ sin x
Therefore, according to Theorem 3.2.9, the integral 1 x
dx is convergent.

Remark 3.2.2
(1) Let f : [a, b) → R be a locally integrable function on [a, b), with b finite.
∫b
If the limit lim x→b f (x) exists and is finite, then the integral a f (x)dx is
x<b
convergent.
(2) Let f : (a, b] → R be a locally integrable function on (a, b], with a finite.
∫b
If the limit lim x→a f (x) exists and is finite, then the integral a f (x)dx is
x>a
convergent.

Proof
1
(1) Since lim x→b (b − x) 2 | f (x)| = 0, from Theorem 3.2.6, it results that
∫b x<b ∫b
|
a f (x)|dx is convergent, hence a f (x)dx is convergent, according to
Theorem 3.2.2.
1
(2) Using the limit lim x→a (x − a) 2 | f (x)| = 0, the proof is similar.
x>a
∫∞
Example 3.2.9 Prove that the Dirichlet integral 0 sinx x dx is convergent.
The improper integral can be decomposed as follows:

∫∞ ∫1 ∫∞
sin x sin x sin x
dx = dx + dx.
x x x
0 0 1

∫1
Since lim x→0 sin x
= 1, it results that the integral sin x
dx is convergent,
x>0 x
∫0∞ x
sin x
according to Remark 3.2.2. On the other hand, the integral 1 x
dx is convergent,
as shown in Example 3.2.8.

Theorem 3.2.10 (The change of variable formula for improper integral) Let
f : [a, b) → R be a continuous function on [a, b), with b finite or not, and let
ϕ: [α, β) → [a, b) be a strictly monotonic function on [α, β), with β finite or
not. We also assume that ϕ is a function of C 1 —class on (α, β), ϕ(α) = a and
limt→β ϕ(t) = b.
t<β
3.2 Convergence Criteria for Improper Integrals 85
∫b ∫β
Then the improper integrals a f (x)dx and α f (ϕ(t)) · ϕ ' (t)dt have the same
nature, and, if they are convergent, we have the formula:

∫b ∫β
f (x)dx = f (ϕ(t)) · ϕ ' (t)dt.
a α

∫∞ ( ) ∫∞ ( )
Example 3.2.10 Prove that the Fresnel integrals 0 sin x 2 dx and 0 cos x 2 dx
are convergent.
∫∞ ( ) ∫1 ( ) ∫∞ ( ) ∫1 ( )
We have 0 sin x 2 dx = 0 sin x 2 dx + 1 sin x 2 dx. Since 0 sin x 2 dx
is
∫ ∞a proper
( 2 ) integral, it remains to study the convergence
√ of the second integral
1 sin x dx. We make the change of variable x = t, and we get x 2 = t,
dx = 2√t dt, t ∈ [1, ∞).
1
∫∞ ( ) ∫∞
From Theorem 3.2.10, it results that 1 sin x 2 dx = 1 2sin√tt dt.
On the other hand, from Theorem 3.2.9, for f (t) = sin t and g(t) = 2√ 1
, we
∫ ∞ sin t ∫ ∞ ( t2)
√
deduce that the integral 1 2 t dt is convergent. Therefore, the integral 1 sin x dx
∫∞ ( )
is convergent, and finally 0 sin x 2 dx is convergent.
∫∞ ( )
Similarly, it is shown that the integral 0 cos x 2 dx is convergent.
Chapter 4
Integrals Depending on Parameter

4.1 Proper Integrals Depending on a Parameter

In this chapter we present a new manner to define a function, using an integral. An

example of such a function was presented in Chap. 2, Theorem 2.6.3.
Indeed, let f : [a, b] → R ∫be a continuous function and let F: [a, b] → R be the
t
function defined by F(t) = a f (x) dx. As we have shown in Theorem 2.6.3, this
function is differentiable and F ' = f ; in other words, F is an antiderivative of f .
Therefore, the antiderivative of the function f is defined as a definite integral whose
upper limit is variable.
Such an integral is a particular case of an integral depending on a parameter for
which the parameter t appears at one of the limits of the integral.
Next, we consider the case when the parameter does not appear in the limits of
the integral but appears in the integrated function.
Let us consider f : D → R, where D = [a, b] × [c, d] ⊂ R2 is a rectangle.
If for any t ∈ [c, d], the function x → f (x, t): [a, b] → R is integrable (with
∫b
respect to x) on [a, b], then the integral a f (x, t)dx will depend on t. A function
can be defined as

∫b
F: [c, d] → R, F(t) = f (x, t) dx, ∀ t ∈ [c, d].
a

It can be considered a more general situation, for which the parameter also appears
within the limits of the integral. Specifically, we have:

Definition 4.1.1 Let us consider the functions f : D → R, where D = [a, b]×[c, d]

and α, β: [c, d] → [a, b].
If for any t ∈ [c, d], the function x → f (x, t): [a, b] → R is integrable (with
respect to x) on [a, b], then the function F: [c, d] → R, defined by:

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 87
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_4
88 4 Integrals Depending on Parameter

∫β(t)
F(t) = f (x, t) dx, ∀ t ∈ [c, d] (4.1)
α(t)

makes sense and it is called the integral depending on a parameter t.

The properties of the function F depend on the properties of the functions f, α, β,
and, further, we will analyze under what conditions the function F is continuous,
differentiable, integrable, and so on.

Theorem 4.1.1 If f : [a, b] × [c, d] → R is continuous on D = [a, b] × [c, d]

and α, β: [c, d] → [a, b] are continuous on [c, d], then the function F: [c, d] → R,
∫ β(t)
defined by F(t) = α(t) f (x, t)dx, ∀ t ∈ [c, d], is continuous on [c, d].

Proof Let t0 ∈ [c, d] an arbitrary fixed point. Then:

∫β(t) ∫ 0)
β(t

F(t) − F(t0 ) = f (x, t)dx − f (x, t0 )dx

α(t) α(t0 )

∫ 0)
α(t ∫ 0)
β(t ∫β(t)
= f (x, t)dx + f (x, t)dx + f (x, t)dx
α(t) α(t0 ) β(t0 )

∫ 0)
β(t ∫ 0)
β(t

− f (x, t0 )dx = ( f (x, t) − f (x, t0 ))dx

α(t0 ) α(t0 )

∫β(t) ∫α(t)
+ f (x, t)dx − f (x, t)dx. (4.2)
β(t0 ) α(t0 )

Because f is continuous on the compact set D, then, according to Theorem 5.6.2

(Part 1, Differential Calculus), it results that f is bounded on D, hence there is M > 0
such that | f (x, t)| < M, ∀(x, t) ∈ D. Furthermore, we have:
| |
| β(t |
| ∫ 0) |
| |
|F(t) − F(t0 )| ≤ | | f (x, t) − f (x, t0 )|dx |
| |
|α(t0 ) |
+ M|β(t) − β(t0 )| + M|α(t) − α(t0 )|.

On the other hand, the function f is uniformly continuous on D (see ( Theorem ) ( 5.6.3, )
Part 1, Differential Calculus),
| ' |hence ∀ε | > 0,| ∃δε' > 0 such that ∀ x ' , t ' , x '' , t '' ∈
D, with the property |x − x '' | < δε' , |t ' − t '' | < δε' , we have:
| ( ' ') ( )| ε
| f x , t − f x '' , t '' | < . (4.3)
3(b − a)
4.1 Proper Integrals Depending on a Parameter 89

Taking into account now to the continuity of the functions α and β at the point t0 ,
it results that there is δε'' > 0 such that ∀t ∈ [c, d], with |t − t0 | < δε'' , we have:

ε ε
|α(t) − α(t0 )| < and |β(t) − β(t0 )| < . (4.4)
3M 3M
{ }
Let δε = min δε' , δε'' and let t ∈ [c, d] be with the property |t − t0 | < δε . Taking
into account to (4.3) and (4.4), it follows that:
ε ε ε
|F(t) − F(t0 )| ≤ · |β(t0 ) − α(t0 )| + M · +M·
3(b − a) 3M 3M
ε ε ε
≤ · (b − a) + + = ε.
3(b − a) 3 3

Therefore, we prove that for any ε > 0, there is δε > 0 such that for any t ∈ [c, d],
with |t − t0 | < δε , we have |F(t) − F(t0 )| ≤ ε, hence F is continuous at the point
t0 . Since t0 was chosen arbitrary in [c, d], it follows that F is continuous on [c, d].

Remark 4.1.1 If f is continuous on D = [a, b] × [c, d] and α, β are continuous on

[c, d], then F is continuous on [c, d] and we have the equalities:

∫β(t) ∫ 0)
β(t

lim F(t) = lim f (x, t)dx = f (x, t0 ) dx = F(t0 ).

t→t0 t→t0
α(t) α(t0 )

( )
∫b ∫b
In particular, lim f (x, t)dx = lim f (x, t) dx.
t→t0 a a t→t0
This remark is very useful in applications.
∫4
Example 4.1.1 Compute limt→0 0 x cos5 (t x)dx.
From Remark 4.1.1, we have:

∫4 ∫4 ∫4 |
( ) x 2 ||4
lim x cos (t x)dx =
5
lim x cos (t x) dx =
5
x dx = = 8.
t→0 t→0 2 |0
0 0 0

The reader can easily realize that without Theorem 4.1.1 computation of the
limit from Example 4.1.1 would be very difficult. Indeed, we should first compute
∫4
0 x cos (t x)dx, and this calculation is extremely difficult and then must calculate
5

the limit of the function obtained, an operation that can also be very difficult.
Next, we present the differentiation theorem of the integral depending on a
parameter.
90 4 Integrals Depending on Parameter

Theorem 4.1.2 Let f : [a, b] × [c, d] → R be a continuous function. We further

assume that ∂∂tf exists, and it is continuous on D = [a, b]×[c, d] and that the functions
α, β: [c, d] → [a, b] are differentiable on [c, d]. Then the function F: [c, d] → R,
∫ β(t)
defined by F(t) = α(t) f (x, t)dx, ∀t ∈ [c, d], is differentiable on [c, d] and we
have the rule:

∫β(t)
' ∂f
F (t) = (x, t)dx + β ' (t) · f (β(t), t)
∂t
α(t)

− α ' (t) · f (α(t), t). (4.5)

The rule (4.5) is known as the Leibniz formula for differentiation of the integral
depending on a parameter.

Proof Let t0 ∈ [c, d] be a fixed point and let t ∈ [c, d], t /= t0 .

Using the decomposition (4.2), it results that:

∫ 0)
β(t
F(t) − F(t0 ) f (x, t) − f (x, t0 )
= dx
t − t0 t − t0
α(t0 )

∫β(t) ∫α(t)
1 1
+ f (x, t)dx − f (x, t)dx.
t − t0 t − t0
β(t0 ) α(t0 )

From Corollary 2.6.2, it results that there are ξ between β(t0 ) and β(t),
respectively, η between α(t0 ) and α(t) such that:

∫ 0)
β(t
F(t) − F(t0 ) f (x, t) − f (x, t0 )
= dx
t − t0 t − t0
α(t0 )
β(t) − β(t0 ) α(t) − α(t0 )
+ f (ξ, t) · − f (η, t) · .
t − t0 t − t0

On the other hand, from Lagrange theorem, it follows that there is θ in the open
interval of limits t0 and t such that:

∂f
f (x, t) − f (x, t0 ) = (x, θ ) · (t − t0 ).
∂t
4.1 Proper Integrals Depending on a Parameter 91

Therefore, we have:

∫ 0)
β(t
F(t) − F(t0 ) ∂f
= (x, θ )dx
t − t0 ∂t
α(t0 )
β(t) − β(t0 ) α(t) − α(t0 )
+ f (ξ, t) · − f (η, t) · . (4.6)
t − t0 t − t0

Furthermore, taking into account to Theorem 4.1.1 and the fact that f and ∂∂ ft are
continuous on D = [a, b] × [c, d], and α, β are differentiable on [c, d], it follows
that the right member of equality (4.6) has limit, hence there is:

∫ 0)
β(t
F(t) − F(t0 ) ∂f
lim = (x, t0 )dx
t→t0 t − t0 ∂t
α(t0 )

+ β ' (t0 ) · f (β(t0 ), t0 ) − α ' (t0 ) · f (α(t0 ), t0 ).

Therefore, F is differentiable at the point t 0 and

∫ 0)
β(t
' ∂f
F (t0 ) = (x, t0 )dx + β ' (t0 ) · f (β(t0 ), t0 )
∂t
α(t0 )

− α ' (t0 ) · f (α(t0 ), t0 ).

Since t 0 is arbitrary, it results that F is differentiable on [c, d] and (4.5) holds.

Remark 4.1.2 If α and β are constant, then the rule from Theorem 4.1.2 becomes:

∫β
' ∂f
F (t) = (x, t)dx.
∂t
α

Example 4.1.2 Let f : R2 → R be a function of C 1 − class on R2 and F: R → R,

∫ bt
F(t) = a t f (x + t, x − t)dx, t ∈ R, a, b ∈ R, a /= b. Compute F ' (t).
According to Leibniz Formula (4.5), we have:

∫b t ( )
' ∂f ∂f
F (t) = (x + t, x − t) − (x + t, x − t) dx
∂x ∂t
at
+ b · f (b t + t, b t − t) − a · f (a t + t, a t − t).
92 4 Integrals Depending on Parameter
∫ sin t 2
ex dx
Example 4.1.3 Compute the limit limt→π ∫0tan t .
0 ex 2 dx
0
We are in the case of indeterminate form and we can apply L’Hôspital rule. 0
,
Leibniz Formula (4.5) is used to compute two derivatives:
(∫ )
∫ sin t x2
d sin t
ex dx
2
e dx 0
dt 0
lim ∫0tan t = lim
0
(∫ )
t→π
0 ex 2 dx t→π d tan t
e x 2 dx
dt0
∫ sin t ∂ ( x 2 ) 2
0 ∂t
e dx + esin t · cos t − e0 · 0
= lim ∫ tan t ∂ ( 2 )
ex dx + etan t · cos12 t − e0 · 0
t→π 2
0 ∂t
∫ sin t 2
0 dx + esin t · cos t
2
esin t · cos t
= lim ∫ tan t
0
= lim tan2 t
t→π
0 0 dx + etan t · cos12 t
2 t→ π e · cos12 t
e0 · (−1)
= = −1.
e0 · 1
Example 4.1.4 Let us consider the elliptic integrals:
π π
∫2 √ ∫2
E(κ) = 1 − κ sin ϕ dϕ and K (κ) = √ 12
2 2
dϕ, 0 < κ < 1.
0 0 1−κ sin2 ϕ
Prove that:
(κ)
dE(κ)
= E(κ)−K and dKdκ(κ) = κ E(κ) − K κ(κ) .
dκ κ (1−κ 2 )
We prove the first equality, and the second one can be verified analogously. We
notice that in the case of elliptic integrals the parameter is κ. From Remark 4.1.2, it
results that:
π π
∫2 ∫2
dE(κ) −κ sin2 ϕ 1 1 − κ 2 sin2 ϕ − 1
= √ dϕ = √ dϕ
dκ 1 − κ sin ϕ
2 2 κ 1 − κ 2 sin2 ϕ
0 0
π π
∫ √ 2 ∫2
1 1 1 E(κ) − K (κ)
= 1 − κ sin ϕ dϕ −
2 2
√ dϕ = .
κ κ 1 − κ sin ϕ
2 2 κ
0 0

Theorem 4.1.3 (Fubini’s theorem) If f : [a, b] × [c, d] → R is continuous, then the

∫b
function F: [c, d] → R, defined by F(t) = a f (x, t) dx, t ∈ [c, d] is continuous
on [c, d] (thus integrable on [c, d]) and the following equality holds:
⎛ d ⎞
∫d ∫b ∫
F(t) dt = ⎝ f (x, t) dt ⎠ dx
c a c
4.2 Improper Integrals Depending on a Parameter 93

which is equivalent to:

⎛ b ⎞ ⎛ ⎞
∫d ∫ ∫b ∫d
⎝ f (x, t) dx ⎠ dt = ⎝ f (x, t) dt ⎠ dx.
c a a c

Proof For any u ∈ [a, b], we shall denote by:

∫u ∫d
g(u, t) = a f (x, t) dx and G(u) = c g(u, t) dt.
∫d ∫u
h(x) = c f (x, t) dt and H (u) = a h(x) dx.
∂g
Since g(u, t) and ∂u = f are continuous, from Theorems 4.1.1 and 4.1.2, it results
that: ∫ d ∂g ∫d ∫d
G ' (u) = c ∂u (u, t) dt = c f (u, t) dt and H ' (u) = h(u) = c f (u, t) dt.
Therefore, G ' (u) = H ' (u), ∀u ∈ [a, b], whence it results that the difference
between the two functions is a constant, hence there is C ∈ R such that:

G(u) = H (u) + C, ∀u ∈ [a, b].

Since G(a) = H (a) = 0, it follows that C = 0, hence:

G(u) = H (u), ∀u ∈ [a, b].

In particular, for u = b, we have G(b) = H (b), i.e.:

⎛ b ⎞ ⎛ d ⎞
∫d ∫ ∫b ∫
⎝ f (x , t) dx ⎠ dt = ⎝ f (x , t) dt ⎠ dx
c a a c

4.2 Improper Integrals Depending on a Parameter

Definition 4.2.1 Let the function f : [a, b) × [c, d] → R, with b finite or not. If
∫b
for any fixed point t0 ∈ [c, d] the improper integral a f (x, t0 ) dx is convergent, we
∫b
say that the improper integral depending on a parameter a f (x, t) dx is pointwise
(simple) convergent on the interval [c, d].
∫b
In other words, a f (x, t) dx is pointwise
∫ u convergent on the interval [c, d] if:
∀ t ∈ [c, d], the limit F(t) = limu→b a f (x, t) dx exists and it is finite.
∫b ∫u u<b ∫b
Since a f (x, t) dx = a f (x, t) dx + u f (x, t) dx, we deduce that the integral
∫b
a f (x, t) dx is pointwise convergent
∫b
on the interval [c, d] if:
∀ t ∈ [c, d], the limit limu→b u f (x, t) dx exists and it is equal to 0.
which is equivalent with:
94 4 Integrals Depending on Parameter
|∫ |
| b |
∀ t ∈ [c, d] and ∀ ε > 0, ∃ a ≤ δt,ε < b such that | u f (x, t) dx | < ε,

∀ δt,ε ≤ u < b.

Taking into account to Theorem 3.2.1, we deduce:

Remark 4.2.1 Let the function f : [a, b) × [c, d] → R, with b finite or not.
∫b
Then a f (x, t)dx is pointwise convergent
|∫ '' on [c,| d] if and only if ∀ t ∈ [c, d]
| u | ( )
and ∀ ε > 0, ∃ a < δt,ε < b such that | u ' f (x, t) dx | < ε, ∀ u ' , u '' ∈ δt, ε , b .

There is another type of convergence, with properties better than pointwise conver-
gence, in which δ depends only on ε and does not depend on t. This type of
convergence is called uniform convergence. Specifically, we have:

Definition 4.2.2 Let the function f : [a, b) × [c, d] → R, with b finite or not.
∫b
The improper integral depending on a parameter a f (x, t) dx is |called uniformly |
|∫ b |
convergent on the interval [c, d] if ∀ ε > 0, ∃ a < δε < b such that | u f (x, t) dx | <
ε, ∀ u ≥ δε and ∀ t ∈ [c, d].

Remark 4.2.2 Let the function f : [a, b) × [c, d] → R, with b finite or not.
∫b
Then a f (x, t) dx |is uniformly convergent
| on [c, d] if and only if ∀ ε > 0,
|∫ u '' |
∃ a < δε < b such that | u ' f (x, t) dx | < ε, ∀ u , u '' ∈ (δε , b), ∀ t ∈ [c, d].
'

Remark 4.2.3 Uniform convergence implies pointwise convergence, according to

Remarks 4.2.2 and 4.2.1; the reciprocal statement is not generally true.

Theorem 4.2.1 (Weierstrass criterion) Let the function f : [a, b) × [c, d] → R, with
b finite or not. If there is a function ϕ: [a, b) → R+ with the properties:

(1) | f (x, t)| ≤ ϕ(x), ∀(x, t) ∈ [a, b) × [c, d]

∫b
(2) a ϕ(x)dx is convergent then the improper integral depending on a parameter
∫b
a f (x, t) dx is uniformly convergent on the interval [c, d].
∫b
Proof Since the improper integral ϕ(x) dx
a |∫ is'' convergent,
| by Theorem 3.2.1, it
| u |
follows that ∀ ε > 0, ∃ a < δε < b such that | u ' ϕ(x) dx | < ε, ∀ u ' , u '' ∈ (δε , b).
Therefore, ∀ u ' , u '' ∈ (δε , b), ∀ t ∈ [c, d], we have:
| u '' | | u '' | | u '' |
|∫ | |∫ | |∫ |
| | | | | |
| f (x, t) dx | < | | f (x, t)| dx | < | ϕ(x) dx | < ε
| | | | | |
|' | |' | |' |
u u u

∫b
whence it results that the integral a f (x, t) dx is uniformly convergent on [c, d].
4.2 Improper Integrals Depending on a Parameter 95
∫∞
Example 4.2.1 Prove that the improper integral 0 e−x cos(t x) dx is uniformly
convergent on R.
Indeed, for any t ∈ R we have
| −x |
|e cos(t x)| ≤ e−x , ∀ x ∈ [0, ∞).
∫∞
On the other hand, the improper integral 0 e−x dx is convergent, because:

∫∞ ∫u |
−x −x
| −x |u
e dx = lim e dx = − lim e |
u→∞ u→∞ 0
0 0
( )
= lim −e−u + 1 = 1 ∈ R.
u→∞

∫∞
According to Theorem 4.2.1, the improper integral 0 e−x cos(t x) dx is
uniformly convergent on R..

Lemma 4.2.1 Let the function f : [a, b) × [c, d] → R, with b finite or not, let
{bn } be a real number sequence such that a < bn < b and limn→∞ bn = b and
∫b ∫b
let us consider Fn (t) = a n f (x, t) dx, ∀ t ∈ [c, d]. If a f (x, t) dx is uniformly
convergent on [c, d], then the sequence of functions {Fn } is uniformly convergent on
[c, d] to the function F, where:

∫b ∫u
F(t) = f (x, t) dx = lim f (x, t) dx, ∀ t ∈ [c, d].
u→b
a u<b a

∫b
Proof Since the integral a f (x, t) dx is uniformly convergent on [c, d], it follows
that ∀ ε > 0, ∃ a < δε < b such that ∀ u ' , u '' ∈ (δε , b) and ∀ t ∈ [c, d], we have:
| u '' |
|∫ |
| |
| f (x, t) dx | < ε. (4.7)
| |
|' |
u

Since limn→∞ bn = b, it results that there is n ε ∈ N∗ such that bn ∈ (δε , b), for
any n ≥ n ε .
If we assume now that n ≥ n ε and m ≥ n ε , from (4.7), we get:
|b |
|∫ m |
| |
|Fn (t) − Fm (t)| = || f (x, t) dx || < ε, ∀ t ∈ [c, d] (4.8)
| |
bn
96 4 Integrals Depending on Parameter

Therefore, the sequence of functions {Fn } is uniformly Cauchy (fundamental) on

[c, d], hence it is uniformly convergent on [c, d]. On the other hand, it is obvious
that for any t ∈ [c, d] we have:

∫bm
lim Fm (t) = lim f (x, t) dx = F(t).
m→∞ m→∞
a

Passing to the limit in (4.8) on m → ∞, we get:

|Fn (t) − F(t)| ≤ ε, ∀ t ∈ [c, d], hence Fn →u[c,d] F.

Theorem 4.2.2 Let f : [a, b) × [c, d] → R be a continuous function, where b is

∫b
finite or not. If we assume that the integral a f (x, t) dx is uniformly convergent on
∫b
[c, d], then the function F: [c, d] → R, defined by F(t) = a f (x, t) dx, ∀ t ∈ [c, d]
is also continuous on [c, d].
∫b
Proof Let a < bn < b, limn→∞ bn = b, and let Fn (t) = a n f (x, t) dx, t ∈ [c, d].
According to Theorem 4.1.1, we deduce that the functions Fn are continuous on
[c, d], n ∈ N∗ .
On the other hand, from Lemma 4.2.1, it follows that Fn →u[c,d] F.
Finally, from Theorem 3.2.1 from [6], we deduce that F is continuous on [c, d].

Remark 4.2.4 Under the conditions of Theorem 4.2.2, the following equality holds:

∫b ∫b
lim F(t) = lim f (x, t) dx = f (x, t0 ) dx = F(t0 ).
t→t0 t→t0
a a

Theorem 4.2.3 Let us consider the function f : [a, b) × [c, d] → R, with b finite or
not, which verified the conditions:

(i) f and ∂∂tf are continuous on [a, b) × [c, d]

∫b
(ii) the integral a f (x, t) dx is uniformly convergent on [c, d]
∫b
(iii) the integral a ∂∂tf (x, t) dx is uniformly convergent on [c, d].
∫b
Then the function F: [c, d] → R, F(t) = a f (x, t) dx, ∀ t ∈ [c, d] is
differentiable on [c, d] and the following formula holds:
⎛ b ⎞
∫ ∫b
' ∂ ⎝ ⎠ ∂f
F (t) = f (x, t)dx = (x, t) dx, ∀ t ∈ [c, d].
∂t ∂t
a a

Proof Let the real number sequence a < bn < b, with limn→∞ bn = b and let us
∫b
take Fn (t) = a n f (x, t) dx, t ∈ [c, d]. Obviously, the sequence of functions {Fn } is
uniformly convergent on [c, d] to the function F.
4.2 Improper Integrals Depending on a Parameter 97

On the other hand, from Theorem 4.1.2, we deduce that the functions Fn are
differentiable on [c, d] and:

∫bn
∂f
Fn' (t) = (x, t) dx, ∀t ∈ [c, d], ∀ n ∈ N∗ .
∂t
a

∫b
We also notice that, if we denote by G(t) = a ∂∂tf (x, t)dx, ∀t ∈ [c, d], according
to Lemma 4.2.1, it results that Fn' →u[c,d] G.
Finally, from Theorem 3.2.2 from [6], it follows that F is differentiable on [c, d]
and F ' (t) = G(t), ∀t ∈ [c, d].
∫π
Example 4.2.2 Compute F(t) = 0 ln(1+t cos x)
cos x
dx, t ∈ (−1, 1).
First, we remark that lim x→ 2
π
ln(1+t cos
cos x
x)
= lim x→ π2 t · ln(1+t cos x)
t cos x
= t, i.e. the point
x = π2 is not a real singularity for the function f (x, t) = ln(1+t cos x)
cos x
, t ∈ (−1, 1).
∂f
We have ∂t (x, t) = cos x · 1+t cos x = 1+t cos x .
1 cos x l
∫π ∫π
We notice that the integral 0 ∂∂tf (x, t)dx = 0 1+t lcos x dx is uniformly convergent
on any interval [a,[ b] ⊂] (−1, 1).
Indeed, if x ∈ 0, π2 , then cos x ≥ 0 and:

l l
0< < , ∀ t ∈ [a, b] ⊂ (−1, 1).
1 + t cos x 1 + a cos x
∫ π
2 l
Since 0 dx is a proper integral, so convergent, from Weierstrass criterion
1+a cos x
∫π
(Theorem 3.2.2), it results that 02 1+t lcos x dx is uniformly convergent on [a, b].
Similarly, we have:

l l [π ]
0< < ,x ∈ , π , t ∈ [a, b]
1 + t cos x 1 + b cos x 2
∫π
whence we deduce that the integral π 1+t lcos x dx is uniformly convergent on [a, b].
2
Furthermore, from Theorem 4.2.3, we have:

∫π ∫π
' ∂f l
F (t) = (x, t)dx = dx, t ∈ (−1, 1).
∂t 1 + t cos x
0 0

If we make the change of variable u = tan 2x , then dx = 2

1+u 2
du, and further:
98 4 Integrals Depending on Parameter

∫∞ ∫∞
' 1 2 1
F (t) = · du = 2 du
1+ t 1−u
2
1+u 2
1 + u2 (1 − t)u 2 + 1 + t
0 0
√ ⎛ ⎞|∞
∫∞ |
2 1 2 1 − t −1 ⎝ u ⎠|| π
= du = · tan √ | =√ .
1−t u2 + 1+t
1−t
1−t 1+t 1+t | 1 − t2
0 1−t 0

Therefore F ' (t) = √π , whence it follows that:

1−t 2
∫ ∫
π
F(t) = F ' (t)dt = √ dt = π · sin−1 t + C .
1− t2

Since F(0) = 0, it results that C = 0, hence F(t) = π · sin−1 t, t ∈ (−1, 1).

∫π ( )
Example 4.2.3 Compute F(t) = 02 ln t 2 − sin2 x dx, t ∈ (1, ∞).
(2 ) [ ]
We notice that the function f (x, t) = ln t − sin2 x , (x, t) ∈ 0, π2 × (1, ∞),
verifies the assertions from Theorem 4.2.3 on any compact subset
[ π] [ π]
D = 0, × [c, d] ⊂ 0, × (1, ∞).
2 2
Indeed,
π π π
∫2 ∫2 ∫2
∂f 2t 1
(x, t) dx = dx = 2 t dx
∂t t − sin2 x
2 t2 − sin2 x
0 0 0

is uniformly convergent on the interval (a, ∞), with a > 1, according to

∫ π2
1
Theorem 4.2.1, because t 2 −sin 2
x
< a 2 −sin
1
2 , if t > a > 1 and
x
1
0 a 2 −sin2 x dx is
a proper integral.
From Theorem 4.2.3, it follows that:
π π
∫2 ∫2
∂f 2t
F ' (t) = (x, t)dx = dx, t > 1.
∂t t2 − sin2 x
0 0
4.2 Improper Integrals Depending on a Parameter 99

If we make the change of variable u = tan x2 , then dx = 2

1+u 2
du, and we have:

∫∞ ∫∞
2t 1 1
F ' (t) = · du = 2 t ( ) du
t2 − u2
1+u 2
1 + u2 t 2 − 1 u2 + t 2
0 0
∫∞ (√ )|∞
√ t2 − 1 |
2t 2 1 |
= 2 du = · t 2 − 1 tan−1 u |
t −1 2
u + t 2 −1
2 t t 2−1 t |
0 0
(√ )|∞
2 −1 t 2 − 1 || π
=√ tan u | =√ .
t −1
2 t | t −1
2
0

Therefore, we obtain:
∫ ∫ ( √ )
π
F(t) = F ' (t)dt = √ dt = π ln t + t 2 − 1 + C , t > 1.
t2 − 1

On the other hand, we have:

⎛ π ⎞
∫2 ( √ )
⎜ ( ) ⎟
C = lim ⎝ ln t 2 − sin2 x dx − π ln t + t 2 − 1 ⎠
t→∞
0
⎛ π
⎞
∫2 ( ) ( √ )
⎜ sin2 x ⎟
= lim ⎝ ln t 2 1 − 2 dx − π ln t + t 2 − 1 ⎠
t→∞ t
0
π
⎛ ⎞
∫2 ( ( 2 )) ( √ )
⎜ sin x ⎟
= lim ⎝ 2 ln t + ln 1 − 2 dx − π ln t + t 2 − 1 ⎠
t→∞ t
0
π
∫ 2 ( ) ( ( √ ))
sin2 x
= lim ln 1 − 2 dx + lim π ln t − π ln t + t 2 − 1
t→∞ t t→∞
0
π
∫ 2 () ( )
sin2 x t 1
= lim ln 1 − 2 dx + lim π ln √ = π ln .
t→∞ t t→∞ t + t −1
2 2
0

Finally, we get:

∫2
π
( √ )
(2 ) t + t2 − 1
F(t) = ln t − sin x dx = π ln
2
, t > 1.
2
0
100 4 Integrals Depending on Parameter
∫∞
Example 4.2.4 Prove that the Dirichlet integral has the value 0 sinx x dx = π2 .
∫∞
The fact that the Dirichlet integral 0 sinx x dx is convergent was established in the
Example 3.2.9.
We consider now the following improper integral depending on parameter t ∈
[0, ∞):

∫∞
sin x
F(t) = e−t x dx, t ∈ [0, ∞)
x
0

and we notice that

∫∞
sin x
F(0) = dx.
x
0

First, integrating by parts, we obtain:

∫ ∫
e−t x sin x dx = −e−t x cos x − t e−t x cos x dx
( ∫ )
= −e−t x cos x − t e−t x sin x + t e−t x sin x dx

whence we deduce that:

∫
e−t x (t sin x + cos x)
e−t x sin x dx = − + C = ϕ(x, t) + C (4.9)
1 + t2

We notice that the function ϕ, which is an antiderivative of the function e−t x sin x,
is bounded. Indeed:
1+t
|ϕ(x, t)| ≤ ≤ 2, if t ≥ 0. (4.10)
1 + t2
∫∞
Next, we evaluate the integral u e−t x sinx x dx, for u > 0.
Because the integral is convergent, integrating by parts, we obtain:
|∞ | | |
|∫ | | | ∫∞ | ∫ ∞
| | | |
| e−t x sin x dx | = | ϕ(x, t) ||∞ + ϕ(x, t) dx | ≤ |ϕ(x, t)| + |ϕ(x, t)|
dx
| | | x 2 | x2
| x | | u x | u u
u u

Taking into account to (4.10), it follows that:

4.2 Improper Integrals Depending on a Parameter 101
|∞ |
|∫ | ∫∞
| sin x | 2 1 4
| e−t x |
dx | ≤ + 2 dx = . (4.11)
| 2
| x | u x u
u u

∫∞
From (4.11), we deduce that limu→∞ u e−t x sinx x dx = 0, hence the improper
∫ ∞ −t x sin x
integral 0 e x
dx is uniformly convergent on [0, ∞), according to Definition
4.2.2.
We remark that the function under the sign of integral is not defined in x = 0, but
has a finite limit in x = 0, namely lim x→0 e−t x sinx x = 1.
√ −t x sin x
e , if x > 0
Let us consider the continuous function f (x, t) = x .
1, if x = 0
∫∞
Obviously F(t) = 0 f (x, t)dx and because the improper integral
∫ ∞ −t x sin x
0 e x
dx is uniformly convergent on [0, ∞) and f is continuous, it results
that F is continuous on [0, ∞) (Theorem 4.2.2). In particular, we have F(0) =
limt→0 F(t).
t>0
On the other hand, we get:

∫∞ ( ) ∫∞
∂ −t x sin x
e dx = − e−t x sin x dx.
∂t x
0 0

| |
Let a > 0 be arbitrary. Since |e−t x sin x | ≤ e−a x , ∀ x ∈ [0, ∞), ∀ t ∈ [a, ∞)
∫ ∞ −a x ∫∞
and 0 e dx = a1 is convergent, it follows that the integral 0 e−t x sin x dx
is uniformly convergent on [a, ∞), ∀ a > 0 (Theorem 4.2.1), hence uniformly
convergent on (0, ∞).
From (4.9) and Theorem 4.2.3, it results that for any t > 0, we have:

∫∞ |
' −t x e−t x (t sin x + cos x) ||∞ 1
F (t) = − e sin x dx = | 0 = − 1 + t2
1 + t2
0

| |
(we notice that lim x→∞ t sin x+cos x
et x
= 0, since | t sin x+cos
et x
x|
≤ t+1
et x
→ 0 if x → ∞).
Therefore:

F(t) = −tan−1 t + C , ∀t > 0. (4.12)

On the other hand, we have:

∫∞
1
|F(t)| ≤ e−t x dx = , ∀t > 0,
t
0

whence it results that:

102 4 Integrals Depending on Parameter

lim F(t) = 0. (4.13)

t→∞

π
From (4.12) and (4.13), we deduce that C = 2
, and so:

π
F(t) = −tan−1 t + , ∀ t > 0. (4.14)
2
π
∫∞ π
Using (4.14), we obtain F(0) = limt→0 F(t) = 2
, hence 0
sin x
x
dx = 2
.
t>0
We can immediately prove that:
⎧π
∫∞ ⎨ 2 , if α > 0
sin α x
dx = 0, if α = 0 .
x ⎩ π
0 − 2 , if α < 0

Theorem 4.2.4 Let f : [a, b) × [c, d] → R be a continuous function. If the improper

∫b
integral depending on parameter a f (x, t)dx is uniformly convergent on [c, d],
∫b
then the function F: [c, d] → R, defined by F(t) = a f (x, t)dx, ∀ t ∈ [c, d] is
continuous (thus integrable) on [c, d] and:
⎛ d ⎞
∫d ∫b ∫
F(t)dt = ⎝ f (x, t) dt ⎠dx
c a c

which is equivalent to:

⎛ b ⎞ ⎛ ⎞
∫d ∫ ∫b ∫d
⎝ f (x, t)dx ⎠dt = ⎝ f (x, t)dt ⎠dx.
c a a c

∫b
Proof Let a < bn < b, limn→∞ bn = b and let Fn (t) = a n f (x, t)dx, t ∈ [c, d].
According to Theorem 4.1.3, the functions Fn are continuous on [c, d], ∀ n ∈ N∗
and:
⎛ ⎞
∫d ∫bn ∫d
Fn (t)dt = ⎝ f (x, t)dt ⎠dx.
c a c

On the other hand, from Lemma 4.2.1, it follows that Fn →u[c,d] F, whence we
deduce that:

∫d ∫d
lim Fn (t)dt = F(t)dt.
n→∞
c c
4.2 Improper Integrals Depending on a Parameter 103

Therefore, we have:
⎛ ⎞
∫d ∫d ∫bn ∫d
F(t)dt = lim Fn (t)dt = lim ⎝ f (x, t)dt ⎠dx,
n→∞ n→∞
c c a c

( )
∫u ∫d ∫d
whence, it results that there is limu→b f (x, t)dt dx = F(t)dt and it is finite.
u<b a c c
∫ b (∫ d )
Thus, the improper integral a c f (x, t)dt dx is convergent and:

⎛ d ⎞ ⎛ ⎞
∫b ∫ ∫d ∫d ∫b
⎝ f (x, t)dt ⎠dx = F(t)dt = ⎝ f (x, t)dx ⎠dt.
a c c c a

Remark 4.2.5 Similar results to the previous theorem are obtained in the case of
intervals of the form [a, b) or (a, b) (with corresponding changes).
∫∞
Example 4.2.5 Compute the Euler–Poisson integral −∞ e−x dx.
2

∫ ∞ −x 2
The improper integral −∞ e dx is convergent, according to Example 3.2.4.
∫∞ ∫∞ not
Since −∞ e−x dx = 2 0 e−x dx = 2 I , it is enough to compute the improper
2 2

integral denoted by

∫∞
e−x dx.
2
I =
0

∫∞
To compute I = 0 e−x dx, we make the change of variable x = t u, with t > 0,
2

and multiplying both members by e−t . Since dx = t du, we get:

2

∫∞
−t 2
t e−t (1+u ) du
2 2
I ·e = (4.15)
0

and further:
⎛ ⎞
∫∞ ∫∞ ∫∞
I · e−t dt = ⎝ t e−t (1+u ) du ⎠dt.
2 2 2

0 0 0

which is equivalent to:

104 4 Integrals Depending on Parameter
⎛ ⎞
∫∞ ∫∞ ∫∞
e−t dt = I 2 = ⎝ t e−t (1+u ) du ⎠dt.
2 2 2
I·
0 0 0

According to Theorem 4.2.4, changing the order of integration, we can write:

⎛ ⎞
∫∞ ∫∞ ∫∞ ( | )
1 2 |
I 2 = ⎝ t e−t (1+u ) dt ⎠du = ) · e−t (1+u ) ||∞ du
2 2 2
(
−2 1 + u 2 0
0 0 0
∫∞ ∫∞ |
1 1 1 1 −1 ||∞ π
= ( ) du = du = tan u | =
2 1+u 2 2 1 + u2 2 0 4
0 0

√
π
whence it results that I = 2
. Therefore, we find:

∫∞
√
e−x dx =
2
π.
−∞

Immediately, it follows that:

∫∞ √
1 π
e−α x dx =
2
, α > 0.
2 α
0

Remark 4.2.6 Analogously, using similar reasons as above (Examples 4.2.4 and
4.2.5), it is shown that the common value of the Fresnel integrals from Example
3.2.10 is:
∫∞ ∫∞ √
( ) ( ) 1 π
sin x 2 dx = cos x 2 dx = .
2 2
0 0

4.3 Euler Integrals

Definition 4.3.1 The Euler integral of the first kind or beta-function is the
following improper integral:

∫1
B( p, q) = x p−1 (1 − x)q−1 dx, p, q > 0. (4.16)
0

The Euler integral of the second kind or gamma-function is the improper

integral:
4.3 Euler Integrals 105

∫∞
┌( p) = x p−1 e−x dx, p > 0. (4.17)
0

Both integrals define important non-elementary functions, which prove to be

extremely useful in mathematical and physical applications. Their properties have
been thoroughly studied, and tables with their values have been drawn up.

Theorem 4.3.1 The beta-function defined in (4.16) is convergent and continuous for
any p > 0 and q > 0.

Proof First, we will show that the integral (4.16) is pointwise convergent for any
p > 0 and q > 0. To prove that we will decompose the integral into a sum of two
integrals:
1
∫1 ∫2 ∫1
x p−1
(1 − x) q−1
dx = x p−1
(1 − x)q−1
dx + x p−1 (1 − x)q−1 dx.
0 0 1
2

∫1
If p ≥ 1, then 02 x p−1 (1 − x)q−1 dx is a proper integral, hence it is convergent.
If 0 < p < 1, then 1 − p < 1 and lim x→0 x 1− p x p−1 (1 − x)q−1 = 1.
x>0
∫1
By Theorem 3.2.7, it follows that the improper integral 02 x p−1 (1 − x)q−1 dx is
convergent.
∫1
If q ≥ 1, then 1 x p−1 (1 − x)q−1 dx is a proper integral, hence it is convergent.
2
If 0 < q < 1, then 1 − q < 1 and lim x→1 (1 − x)1−q x p−1 (1 − x)q−1 = 1. From
∫1 x<1
Theorem 3.2.6, it follows that 1 x p−1 (1 − x)q−1 dx is convergent.
2
Therefore we proved that the beta-function B( p, q) is pointwise convergent for
any p > 0 and q > 0.
Moreover, we note that the beta-function B( p, q) is continuous for any p > 0
and q > 0. Indeed, let p0 > 0 and q0 > 0 be two fixed positive numbers. From the
∫1
above considerations, it results that the improper integral 0 x p0 −1 (1 − x)q0 −1 dx is
convergent.
On the other hand, since

x p−1 (1 − x)q−1 ≤ x p0 −1 (1 − x)q0 −1 , ∀ p ≥ p0 , q ≥ q0 , x ∈ (0, 1),

∫1
from Theorem 4.2.1, it follows that 0 x p−1 (1 − x)q−1 dx is uniformly convergent
for any p ≥ p0 , q ≥ q0 . According to Theorem 4.2.2, it results that the beta-function
B( p, q) is continuous on the domain [ p0 , ∞) × [q0 , ∞). As p0 > 0 and q0 > 0 are
arbitrary, we deduce that the beta-function B( p, q) is continuous for any p > 0 and
q > 0.

Remark 4.3.1 The beta-function B( p, q) is uniformly convergent on any domain

[ p0 , ∞) × [q0 , ∞), with p0 > 0, q0 > 0.
106 4 Integrals Depending on Parameter

Theorem 4.3.2 The beta-function B( p, q) has the following properties:

B( p, q) = B(q, p), p > 0, q > 0. (4.18)

If p > 1, then the following recurrence relation holds:

p−1
B( p, q) = B( p − 1, q). (4.19)
p+q −1

In particular, for any m, n ∈ N∗ , m ≥ 2, we have:

(m − 1)! · (n − 1)!
B(m, n) = (4.20)
(m + n − 1)!

∫∞ ∫∞
t p−1 t q−1
B( p, q) = dt = dt. (4.21)
(1 + t) p+q (1 + t) p+q
0 0

π
∫2
B( p, q) = 2 sin2 p−1 u · cos2 q−1 u du. (4.22)
0

Proof Assertion (i) follows immediately if we make the change of variable x = 1−t.

(ii) Integrating by parts the relation (4.16), for p > 1 and q > 0, it results that:

| ∫1
1 p−1 |
q |1 p−1
B( p, q) = − x (1 − x) | + x p−2 (1 − x)q dx
q 0 q
0
∫1
p−1 ( )
= x p−2 (1 − x)q−1 − (1 − x)q−1 x dx
q
0
p−1 p−1
= B( p − 1, q) − B( p, q).
q q

Further, we have:
( )
p−1 p−1
1+ B( p, q) = B( p − 1, q)
q q
or
4.3 Euler Integrals 107

p−1
B( p, q) = B( p − 1, q).
p+q −1

Similarly, it is shown that if q > 1, then:

q −1
B( p, q) = B( p, q − 1).
p+q −1

Since B( p, 1) = 1p , for any n ∈ N∗ , we get:

n−1 n−2 n − (n − 1)
B( p, n) = · · ... · B( p, 1)
p+n−1 p+n−2 p + n − (n − 1)
(n − 1)!
= .
p · ( p + 1) · . . . · ( p + n − 1)

In particular, for m, n ∈ N∗ , m ≥ 2 we have:

(m − 1)! · (n − 1)!
B(m, n) = .
(m + n − 1)!

(iii) In (4.16) we consider the change of variable x = t

1+t
and we obtain:

∫∞ ∫∞
t p−1 1 t p−1
B( p, q) = · dt = dt.
(1 + t) (1 + t)q−1 (1 + t)2
p−1
(1 + t) p+q
0 0

Moreover
∫∞
t q−1
B( p, q) = B(q, p) = dt.
(1 + t) p+q
0

(iv) In (4.16) we consider the change of variable x = sin2 u , dx =

2 sin u cos u du and we have

π π
∫2 ∫2
B( p, q) = 2 sin2 p−2 u cos2 q−2 u sin u cos u du = 2 sin2 p−1 u cos2 q−1 u du.
0 0

Theorem 4.3.3 The gamma-function (4.17) is convergent and continuous for any
p > 0.
Proof For the beginning we will show that the gamma-function ┌( p) is pointwise
convergent for any p > 0. To prove that we will decompose the integral into a sum
of two integrals:
108 4 Integrals Depending on Parameter

∫∞ ∫1 ∫∞
p−1 −x p−1 −x
x e dx = x e dx + x p−1 e−x dx.
0 0 1

∫1
If p ≥ 1, then 0 x p−1 e−x dx is a proper( integral, thus it) is convergent.
If 0 < p < 1, then 1− p < 1 and lim x→0 x 1− p x p−1 e−x = 1. By Theorem 3.2.7,
∫1 x>0
the improper integral 0 x p−1 e−x dx is convergent. ( )
On the other hand, we notice that lim x→∞ x 2 x p−1 e−x = 0. From
∫ ∞ p−1 −x
Theorem 3.2.5, it follows that the improper integral 1 x e dx is convergent.
Therefore we proved that the gamma-function ┌( p) is pointwise convergent for
any p > 0.
Let r > 0, s > 0 be two positive arbitrary numbers, with r < s. If p ∈ [r, s],
then. ( )
x p−1 e−x ≤ x r −1 + x s−1 e−x , for any x > 0.
∫∞( )
As the improper integral 0 x r −1 + x s−1 e−x dx is convergent, from
∫ ∞ p−1 −x
Theorem 4.2.1, we deduce that ┌( p) = 0 x e dx is uniformly convergent on
[r, s]. From Theorem 4.2.2 and from to the fact that the function f (x, p) = x p−1 e−x
is continuous on (0, ∞) × (0, ∞), it results that the gamma-function ┌( p) is
continuous on the interval [r, s], so on (0, ∞), because r > 0, s > 0 are arbitrary.

Remark 4.3.2 The gamma-function ┌( p) is uniformly continuous on any compact

interval [r, s] ⊂ (0, ∞), where 0 < r < s < ∞.

Theorem 4.3.4 The gamma-function ┌( p) has the following properties:

┌(1) = 1.

┌( p + 1) = p ┌( p), p > 0. (4.23)

In particular, ┌(n + 1) = n!, n ∈ N∗ .

┌( p) · ┌(q)
B( p, q) = , p > 0, q > 0. (4.24)
┌( p + q)
π
B( p, 1 − p) = ┌( p) · ┌(1 − p) = , p ∈ (0, 1). (4.25)
sin(π p)
(1) √
In particular, ┌ 2
= π.

Proof
|
∫∞ |
(i) ┌(1) = 0 e−x dx = −e−x ||∞ = 1.
0
4.3 Euler Integrals 109
∫ |
∞ |
┌( p + 1) = x p e−x dx = −x p e−x ||∞
0 0
(ii) ∫ ∞
+p x p−1 e−x dx = p ┌( p).
0
In particular, for n ∈ N∗ we have:

┌(n + 1) = n · ┌(n) = n · (n − 1) · ┌(n − 1)

= · · · = n · (n − 1) · . . . · 2 · ┌(1).

Since ┌(1) = 1, it results that ┌(n + 1) = n!.

This formula is interesting because it provides an analytical expression for n!,
namely:

∫∞
n! = x n e−x dx, n ∈ N∗ .
0

(iii) First, we notice that if we make the change of variable x = t y, t > 0, then
dx = t dy and we get:

∫∞
┌( p) = t p
y p−1 e−t y dy. (4.26)
0

Substituting in (4.26), p with p + q and t with 1 + t, we get:

∫∞
┌( p + q)t p−1
= t p−1 y p+q−1 e−(1+t)y dy
(1 + t) p+q
0

Taking into account now the formula (4.21), we deduce:

∫∞
┌( p + q)t p−1
┌( p + q)B( p, q) = dt
(1 + t) p+q
0
⎛ ⎞
∫∞ ∫∞
= ⎝t p−1 y p+q−1 e−(1+t)y dy ⎠dt
0 0
⎛ ⎞
∫∞ ∫∞
= ⎝t p−1 y p+q−1 e−(1+t)y dt ⎠dy
0 0
⎛∞ ⎞
∫∞ ∫
p+q−1 ⎝
= e −y
y t p−1 −t y
e dt ⎠dy.
0 0
110 4 Integrals Depending on Parameter

From (4.26), it follows that:

∫∞
y p+q−1 e−y
┌( p + q) · B( p, q) = · ┌( p)dy
yp
0
∫∞
= ┌( p) y q−1 e−y dy = ┌( p) · ┌(q).
0

┌( p)·┌(q)
Therefore, we proved that B( p, q) = ┌( p+q)
.

(iv) The proof of the formula (4.25) will not be done because it is too long and
requires superior knowledge of mathematics.
( ) ( ) √
From (4.25), for p = 21 , we get ┌ 2 21 = sinπ π = π , whence ┌ 21 = π .
( ) 2
The value for ┌ 21 can also be found by direct computation, using the change the
variable x = t 2 and Example 4.2.5. Indeed:

( ) ∫∞ ∫∞ √
1 π √
= x e dx = 2 e−t dt = 2 ·
− 21 −x 2
┌ = π.
2 2
0 0

Example 4.3.1 Compute the following integrals using the Euler integrals B and ┌:
∫∞ ∫∞ 2 n −x 2
(1) x 3 e−3 x dx; (2) x e dx, n ∈ N∗ ;
0 −∞
π
∫1 √ ∫2 7 5
(3) x2 − x3 dx; (4) sin 2 x cos 2 x dx;
0 0

(1) If we make the change of variable 3 x = t, then dx = 13 dt, t ∈ [0, ∞), and:

∫∞ ∫∞
3 −3 x 1 t 3 −t
x e dx = e dt
3 33
0 0
∫∞
1 ┌(4) (4.23) 3! 2
= 4 t 3 e−t dt = = = .
3 81 81 27
0

∫∞ ∫∞
x 2 n e−x dx = 2 x 2 n e−x dx.
2 2
(2) −∞ 0

If we make the change of variable x 2 = t ≥ 0, then dx = 1

√
2 t
dt, t ∈ [0, ∞) and:
4.3 Euler Integrals 111

∫∞ ∫∞ ∫∞ ∫∞
2 n −x 2 2 n −x 2 1 n −t
√ dt = t n− 2 e−t dt
1
x e dx = 2 x e dx = 2 t e
2 t
−∞ 0 0 0
( ) ( ) ( )
1 1 (4.23) 1
=┌ n+ =┌ n− +1 = n−
2 2 2
( )
1 (4.23) (4.23)
┌ n− = ... =
2
( ) ( ) ( ) ( )
1 3 5 3 1 1
= n− · n− · n− ··· · ┌
2 2 2 2 2 2
(2 n − 1) · (2 n − 3) · (2 n − 5) · · · 3 · 1
=
2n
√ not (2 n − 1)!! √
· π= · π.
2n
∫∞ √
For n = 0, we obtain the Euler–Poisson integral −∞ e−x dx = π.
2

A new property of the gamma-function can be retained:

( )
1 1 · 3 · · · (2 n − 3) · (2 n − 1) √ (2 n − 1)!! √
┌ n+ = n
π= π , n ∈ N∗ (4.27)
2 2 2n

∫1 √ ∫1 √ (
∫1 )
1 3 (4.24)
x 2 − x 3 dx = x(1 − x) 2 dx = B 2,
x 2 (1 − x) dx =
2
0 0 0
(3) (3) (3)
(3) ┌(2) · ┌ 2 1! · ┌ 2 ┌
= (7) = ( ) = 5 (25 )
┌ 2 ┌ 1+ 2 5
2
┌ 2
(3) (3)
┌ ┌ 4
= 5 ( 2 3) = 5 3 2 (3) = .
2
┌ 1 + 2
·
2 2
· ┌ 2
15
(4) Using Formula (4.22), we get:

π ( ) ( )
∫2 ( ) 9 7
7 5 1 9 7 (4.24) 1 ┌ 4 · ┌ 4
sin 2 x cos 2 x dx = · B , = ·
2 4 4 2 ┌(4)
0
( ) ( )
5 3 ( ) ( )
1 ┌ 1 + 4 · ┌ 1 + 4 (4.23) 1 5 3 5 3
= · = · · ·┌ ·┌
2 3! 12 4 4 4 4
( ) ( ) ( ) ( )
5 1 3 (4.23) 5 1 1 3
= ·┌ 1+ ·┌ = · ·┌ ·┌
64 4 4 64 4 4 4
( ) ( ) √
5 1 1 (4.25) 5 π 5 2π
= ·┌ ·┌ 1− = · = .
256 4 4 256 sin π4 256
Chapter 5
Line Integrals

5.1 Parameterized Paths. Definition of a Curve

Definition 5.1.1 A parameterized path in R3 of C k -class, k ∈ N∗ is any vector

function r : I ⊂ R → R3 of C k -class, defined on an interval I ⊂ R.

If we denote by x, y, z the scalar components of the vector function r , then:

r (t) = (x(t), y(t), z(t)), ∀t ∈ I.

The fact that the vector function r : I → R3 is of C k -class is equivalent to the

fact that the scalar functions x, y, z : I → R are of C k -class on the interval I ; i.e.
they are k-times differentiable
⎧ and their derivatives are continuous on I .
⎨ x = x(t)
The equations y = y(t) , t ∈ I , are called the parametric equations of the
⎩
z = z(t)
path or a parametric representation of the path and t is called a parameter.
We draw attention to the fact that according to Definition 5.1.1, a parameterized
path is a vector function (application) and not a set of points. To better mark this we
will note a path parameterized with (I, r ) or (I, r = r (t)).
The direct image r (I ) of the interval I through the vector function r , i.e. the subset

r (I ) = {r (t); t ∈ I } = {(x(t), y(t), z(t))t ∈ I } ⊂ R3

is called the support of parameterized path r .

If I = [a, b] is a compact interval, then its support r (I ) is a compact and connected
subset of R3 . In this case, the points r (a) and r (b) are called the ends (extremities)
of the path. If r (a) = r (b), then the path is called closed.
−
→ −→ −→
Let O x yz be an orthogonal coordinate system and let i , j , k be the unit
vectors of the axis O x, O y, respectively Oz. If we identify any point M(x, y, z) ∈ R3
−−→
with its position vector O M, then we obtain the vector equation of the path:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 113
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_5
114 5 Line Integrals

Fig. 5.1 The circle of radius

R centered at the origin O

−
→ −
→ −
→
r (t) = −
→
r (t) = x(t) i + y(t) j + z(t) k , ∀ t ∈ I.

Example 5.1.1 Let r : [0, 2π ] → R3 be the path defined by:

−
→ −
→ −
→
r (t) = R cos t i + R sin t j , t ∈ [0, 2π ], R > 0.
⎧
⎨ x = R cos t
The parametric equations are: y = R sin t , t ∈ [0, 2π ].
⎩
z =0
We noticethat for any t ∈ [0, 2π ], the point (x(t), y(t), 0) verifies the equation
x 2 + y2 = R2
of the circle .
z = 0
It results that the support of the path is the circle of radius R centered at the origin
O that lies in the plane O x y. The parameter t is the angle between the position vector
−−→
O M and the positive direction of the axis O x (Fig. 5.1).
We notice also that this path is closed, because r (0) = r (2π ) = (R, 0, 0).

Example 5.1.2 Let r : [0, 2π ] → R3 be the path defined by:

−
→ −
→ −
→ −
→
r (t) = R cos t i + R sin t j + ht k , t ∈ [0, 2π ], R, h > 0.
⎧
⎨ x = R cos t
The parametric equations are: y = R sin t , t ∈ [0, 2π ].
⎩
z = ht
We notice that for any t ∈ [0, 2π ], we have (x(t))2 + (y(t))2 = R 2 , whence, it
results that the support of this path is situated on the right circular cylinder of radius
R and axis of symmetry Oz given by x 2 + y 2 = R 2 .
5.1 Parameterized Paths. Definition of a Curve 115

Fig. 5.2 The cylindrical z

helix

y
A


The support of this path is the arc AB of the cylindrical helix of radius R and
slope h/R, where A(R, 0, 0) and B(R, 0, 2π h) (Fig. 5.2).
Example 5.1.3 Let r : R → R3 be the path defined by:

−
→ −
→ −
→ −
→
r (t) = (x0 + lt) i + (y0 + mt) j + (z 0 + nt) k , t ∈ R

where x0 , y0 , z 0 , l, m, n are real constants.

The parametric equations are:
⎧
⎨ x = x 0 + lt
y = y0 + mt , t ∈ R.
⎩
z = z 0 + nt

The support of this path is the straight line passing through the point M0 (x0 , y0 , z 0 )
and having the direction parameters l, m, n (Fig. 5.3). In other words, this straight
−
→ −
→ −
→
line has the same direction as − →
v , where − →v =l i +m j +n k .
Definition 5.1.2 A path (I, r ) is called simple if the vector function r is injective,
i.e. ∀t1 , t2 ∈ I, t1 /= t2 ∀t1 , t2 ∈ I, t1 /= t2 , it results that r (t1 ) /= r (t2 ). A closed
path is simple if the equality r (t1 ) = r (t2 ) implies t1 = t2 or t1 = a and t2 = b,
respectively t1 = b and t2 = a, where a and b are the ends of the interval I .
The paths presented in Example 5.1.1, Example 5.1.2 and Example 5.1.3 are
simple.
Definition 5.1.3 Let (I, r ) be a parameterized path of C 1 -class.
A point M0 (x(t0 ), y(t0 ), z(t0 )) is called singular point of (I, r ) if

x ' (t0 ) = y ' (t0 ) = z ' (t0 ) = 0.

116 5 Line Integrals

Fig. 5.3 The support of a

straight line

A smooth path is a path without singular points, i.e.:

x '2 (t) + y '2 (t) + z '2 (t) > 0, ∀t ∈ I.

A smooth path has a tangent at each point and the position of the tangent
continuously depends of the point of tangency.
Every path is oriented with respect to the increasing parameter. For example, in
Example 5.1.2 the path is oriented from A to B.

Definition 5.1.4 Two parameterized paths (I1 , r1 ) and (I2 , r2 ) of C k -class are called
equivalent if there is a function λ : I1 → I2 bijective, of C k -class, with the inverse
λ−1 : I2 → I1 of C k -class and with the property λ' (t1 ) /= 0, ∀t1 ∈ I1 such that:

r2 (λ(t1 )) = r1 (t1 ), ∀t1 ∈ I1 .

We use the notation (I1 , r1 ) ∼ (I2 , r2 ).

Such a function λ is called change of parameter. Since λ' is continuous on I1 and
λ' /= 0 on I1 , it follows that λ' > 0 or λ' < 0 on I1 .
If λ' > 0 on I1 , then λ is strictly increasing and we say that the paths are equivalent
to the same orientation. Otherwise, the paths are equivalent to opposite orientation.
It is obvious that two equivalent paths have the same support.

Example 5.1.4 Let r1 : I1 → R2 and r2 : I2 → R2 be two paths defined by:

−
→ −
→ ( π)
−
→
r 1 (t1 ) = R sin t1 i + R cos t1 j , t1 ∈ I1 = 0, ,R >0
/ 2
−
→ −
→ −
→
r 2 (t2 ) = t2 i + R 2 − t22 j , t2 ∈ I2 = (0, R).


These paths have the same support, namely the arc AB of the circle of radius R
centered at the origin (Fig. 5.4).
5.1 Parameterized Paths. Definition of a Curve 117

Fig. 5.4 Three parametrical

representations of the arc AB

We notice that the function λ : I1 → I2 defined by

( π)
λ(t1 ) = R sin t1 , ∀t1 ∈ I1 = 0,
2
( )
is bijective, of C ∞ -class on I1 and λ' (t1 ) = R cos t1 > 0, ∀t1 ∈ 0, π2 .
Moreover, for any t1 ∈ I1 we have:

−
→ → √
− −
→ −
→ −
→ →
r 2 (λ(t1 )) = λ(t1 ) i + R 2 − λ2 (t1 ) j = R sin t1 i + R cos t1 j = −
r 1 (t1 ).

It follows that λ is a change of parameter, and therefore the two paths are equivalent
to the same orientation.
Let us consider now the path r3 : I3 → R2 defined by:
−
→ −
→ ( π)
−
→
r 3 (t3 ) = R cos t3 i + R sin t3 j , t3 ∈ I3 = 0, .
2
We notice that the function μ : I3 → I2 , defined by

μ(t3 ) = R cos t3 , ∀t3 ∈ I3

is a change of parameter. Since

( π)
μ' (t3 ) = −R sin t3 < 0, ∀t3 ∈ 0,
2
we deduce that μ is strictly decreasing, hence the paths (I3 , r3 ) and (I2 , r2 ), respec-
tively, the paths (I3 , r3 ) and (I1 , r1 ) are equivalent to opposite orientation. Indeed,
the orientation of the paths (I1 , r1 ) and (I2 , r2 ) is from A to B, while the orientation
of the path (I3 , r3 ) is from B to A.

Definition 5.1.5 A parameterized curve is any class of equivalent parameterized

paths.
118 5 Line Integrals

) γ is a parameterized curve if there is a parameterized path r : I → R

3
( Therefore,
r : I → R such that:
2

{ ( ) }
γ = ρ : I → R3 R2 parameterized path; (I, ρ) ∼ (I, r ) .

Because (I, r ) ∼ (I, r ), it follows that r ∈ γ .

A parameterized curve is simple (closed, smooth) if the path that defines it is
simple (closed, smooth). A simple curve is considered to be positively oriented if
the path that defines it is oriented in the direction of increasing the parameter, and
negatively orientedotherwise. ( )
Let γ be a simple and smooth parameterized curve and let r : I → R3 R2
be the parameterized path that defines it, oriented in the direction of increasing the
parameter. We will denote by γ+ the family of all parameterized paths equivalent to
r and having the same orientation as r . Obviously, r ∈ γ+ . We will note with γ− the
family of all equivalent parameterized paths with r having the opposite orientation
to r .
The support of a parameterized curve γ is the support of the path that defines it
and, obviously, it coincides with the support of any representative path of γ .
Let γ be the parameterized curve defined by the path r1 in Example 5.1.4. Its
 
support is the arc AB from Fig. 5.4. The support of the curve γ+ is the arc AB

(oriented from A to B), while the support of the curve γ− is the arc B A. Obviously,
r2 ∈ γ+ and r3 ∈ γ− . Whenever there is no confusion, we will identify a curve γ by
any of its representative.

Definition 5.1.6 Let r1 : [a, b] → R3 and r2 : [b, c] → R3 be two parameterized

paths with the property r1 (b) = r2 (b). It is called the union of the paths and is
denoted by r1 ∪ r2 the following path that can be written as:

r1 (t) if t ∈ [a, b]
r1 ∪ r2 : [a, c] → R3 , (r1 ∪ r2 )(t) = .
r2 (t) if t ∈ [b, c]

If γi is the curve defined by ri , i = 1, 2, then γ1 ∪ γ2 is the curve defined by the

path r1 ∪ r2 .
A curve is called piecewise smooth if it is a union of finite number of smooth
curves.

5.2 Rectifiable Paths and Curves

The notion of curve (path) introduced in paragraph 5.1 is quite general, and therefore,
in some cases, especially in the case of curves that admit multiple points, the support
of a curve may differ significantly from the intuitive image that we have of a curve.
5.2 Rectifiable Paths and Curves 119

Fig. 5.5 The length of a M1 M i –1

polygonal line inscribed in a
curve

M0 Mi

The Italian mathematician Giuseppe Peano proved that two continuous functions
x = x(t), y = y(t) can be defined on the interval [0, 1] (thus a path), such that, when
the parameter t crosses the interval [0, 1], the corresponding point M(x(t), y(t))
starts to the point (0, 0) (corresponding to the value t = 0), passes through all points
of the square [0, 1] × [0, 1] and ends to the point (1, 1) (corresponding to the value
t = 1). In other words, the support of this path fills a square. It is clear that the notion
of length for such a path does not make sense.
In the following, we will introduce the notion of rectifiable path (which has length)
and we will show how to compute the length of a rectifiable path using the definite
integral.
Let r : [a, b] → R3 be a path and let x = x(t), y = y(t), z = z(t), t ∈ [a,] be
its parametric equations. Consider an arbitrary partition Δ of the interval [a, b], i.e.

Δ : a = t0 < t1 < ... < ti−1 < ti < ... < tn = b

and we will denote by Mi the coordinate points (x(ti ), y(ti ), z(ti )), i = 0, n, which
lie on the support of the path.
Σ
n
Let L Δ (r ) = Mi−1 Mi > 0 be the length of the polygonal line inscribed in
i=1
the support of the path r with vertices at the point Mi corresponding to the values
t = ti (Fig. 5.5).
The set {L Δ (r )}Δ , when Δ is any possible partition of the interval [a, b], is a set
of positive numbers upper bounded or not.

Definition 5.2.1 The path r is called rectifiable (has length) if the set {L Δ (r )}Δ is
upper bounded, i.e. ∃M > 0such that L Δ (r ) < M, for any partition Δ of [a, b].
If the path r is rectifiable, then its length is defined by:

L(r ) = sup{L Δ (r )} < ∞.

Δ

Lemma 5.2.1 For any four real numbers a1 , a2 , b1 , b2 the following inequality is
established:
|/ / |
| 2 |
| a + a 2 − b2 + b2 | ≤ |a1 − b1 | + |a2 − b2 |. (5.1)
| 1 2 1 2|
120 5 Line Integrals

Proof Amplifying with the conjugate and taking into account the inequality of the
triangle, we obtain:

|/ / | | 2 |
| 2 | |a + a 2 − b2 − b2 |
| a + a 2 − b2 + b2 | = / 1 2
/1 2
| 1 2 1 2|
a12 + a22 + b12 + b22
|a1 − b1 ||a1 + b1 | + |a2 − b2 ||a2 + b2 |
≤ / / . (5.2)
a12 + a22 + b12 + b22

On the other hand, we have:

/ /
|a1 + b1 | ≤ |a1 | + |b1 | ≤ a1 + a2 + b12 + b22
2 2

and similarly:
/ /
|a2 + b2 | ≤ |a2 | + |b2 | ≤ a1 + a2 + b12 + b22 .
2 2

Taking into account these inequalities in (5.2), it results that:

|/ / |
| 2 |
| a + a 2 − b2 + b2 | ≤ |a1 − b1 | + |a2 − b2 |.
| 1 2 1 2|

Remark 5.2.1 The inequality (5.1) remains established for any 2n real numbers
ai , bi ∈ R, i = 1, n. For example, for n = 3 we have:
|/ / |
| 2 |
| a + a 2 + a 2 − b2 + b2 + b2 | ≤ |a1 − b1 | + |a2 − b2 | + |a3 − b3 |. (5.3)
| 1 2 3 1 2 3|

The proof is the same as the proof of Lemma 5.2.1.

Theorem 5.2.1 Let r : [a, b] → R3 , r (t) = (x(t), y(t), z(t)), t ∈ [a, b] be a

smooth parameterized path. Then r is rectifiable and its length is given by:

∫b √
L = L(r ) = x '2 (t) + y '2 (t) + z '2 (t) dt.
a

Proof Let Δ : a = t0 < t1 < ... < ti−1 < ti < ... < tn = b be an arbitrary partition
of the interval [a, b] and let L Δ (r ) the length of the polygonal line inscribed in the
support of the path r (Fig. 5.5), i.e.:
5.2 Rectifiable Paths and Curves 121

n /
Σ
L Δ (r ) = (x(ti ) − x(ti−1 ))2 + (y(ti ) − y(ti−1 ))2 + (z(ti ) − z(ti−1 ))2 .
i=1

From Lagrange theorem, it follows that there are ξi , ηi , μi ∈ (ti−1 , ti ), i = 1, n,

such that:

x(ti ) − x(ti−1 ) = x ' (ξi )(ti − ti−1 )

y(ti ) − y(ti−1 ) = y ' (ηi )(ti − ti−1 )
z(ti ) − z(ti−1 ) = z ' (μi )(ti − ti−1 ).

Taking into account to this equalities in L Δ (r ), we get:

Σ
n √
L Δ (r ) = x '2 (ξi ) + y '2 (ηi ) + z '2 (μi ) · (ti − ti−1 ). (5.4)
i=1

The function g : [a, b] → R given by

√
g(t) = x '2 (t) + y '2 (t) + z '2 (t), t ∈ [a, b]

is continuous on [a, b], because the derivatives functions x ' , y ' , z ' are continuous,
hence g is integrable on [a, b].
Consider the Riemann sum:

Σ
n √
σΔ (g; ξ ) = x '2 (ξi ) + y '2 (ξi ) + z '2 (ξi ) · (ti − ti−1 ). (5.5)
i=1

Since g is integrable on [a, b], it follows that for any ε > 0 there is δε' > 0 such
that for any partition Δ of [a, b], with the property Δ < δε' and for any intermediate
points ξ = (ξ1 , ξ2 , ..., ξn ) we have:
| |
| ∫b |
| |
|σΔ (g; ξ ) − g(t)dt | < ε. (5.6)
| |
| |
a

On the other hand, from the inequality (5.3) and the generalized inequality of the
triangle, it results that:

|L Δ (r ) − σΔ (g; ξ )|
Σ
n
(| ' | | |)
<< | y (ηi ) − y ' (ξi )| + |z ' (μi ) − z ' (ξi )| · (ti − ti−1 ). (5.7)
i=1
122 5 Line Integrals

Since the functions y ' and z ' are uniformly| continuous | on [a, b], it results that
there is δε'' > 0 such that ∀ t', t'' ∈ [a, b], with |t ' − t ' | < δε'' , we have:
| '( ') ( )| ε | ( ) ( )| ε
| y t − y ' t '' | < and |z ' t ' − z ' t '' | < . (5.8)
b−a b−a

If we now choose the partition Δ such that Δ < δε'' , then:

|ηi − ξi | ≤ ti − ti−1 ≤ Δ < δε''

and similarly
|μi − ξi | < δε'' .
From (5.8), we deduce that:
| ' | ε | | ε
| y (ηi ) − y ' (ξi )| < and |z ' (μi ) − z ' (ξi )| < . (5.9)
b−a b−a

Taking into account to (5.9) in (5.7), it results that:

ε Σ
n
|L Δ (r ) − σΔ (g; ξ )| < (ti − ti−1 ) = ε.
b − a i=1

Therefore, we proved that for any Δ, with Δ < δε'' , we have:

|L Δ (r ) − σΔ (g; ξ )| < ε. (5.10)

Since g is integrable on [a, b], it follows that σΔ (g; ξ ) is bounded for any Δ and
any ξ and, taking into account to (5.10), we have that the set {L Δ (r )}Δ is upper
bounded. Therefore, we proved that the path is rectifiable.
We denote L = sup{L Δ (r )}. From the definition of the least upper bound, it
Δ
follows that for any n ∈ N∗ , there is a partition Δn of the interval [a, b] such that:

1
L− < L Δn (r ) ≤ L . (5.11)
n

Moreover, we can assume that Δn < n1 , because, otherwise, we refine this

partition until we obtain a partition Δ'n with this property. As L Δn (r ) ≤ L Δ'n (r ), it
follows that L Δ'n (r ) satisfies (5.11).
We now consider a partition Δn of the interval [a, b] with the property:
( )
1
Δn < min , δε' , δε''
n

for which the inequalities (5.11) are verified.

5.2 Rectifiable Paths and Curves 123

By (5.6), (5.10), and (5.11), it follows that:

| |
| ∫b |
| | | | | |
| L − g(t)dt | ≤ | L − L Δ (r )| + | L Δ (r ) − σΔ (g; ξ )|
| | n n n
| |
a
| |
| ∫b |
| | 1
|
+ |σΔn (g; ξ ) − g(t)dt || < + 2 ε.
| | n
a

Because the last inequality holds for any n ∈ N∗ and any ε > 0, it follows that:

∫b ∫b √
L= g(t)dt = x '2 (t) + y '2 (t) + z '2 (t)dt
a a

and the theorem is proved.

It can be shown that if a path is rectifiable, then any other path equivalent to it is
rectifiable and has the same length.

Definition 5.2.2 A curve is called rectifiable if it is an equivalence class of a recti-

fiable paths. The length of a rectifiable curve is the length of any representative of
this equivalence class.

Example 5.2.1 Compute the length of the cylindrical helix:

⎧
⎨ x = R cos t
y = R sin t , t ∈ [0, 2 π ], R > 0, h > 0.
⎩
z =ht

According to Theorem 5.2.1, we have:

∫2 π√ ∫2 π√ √
L= R 2 sin2 t + R 2 cos2 t + h 2 dt = R 2 + h 2 dt = 2π R2 + h2.
0 0

Remark 5.2.1 Let r : [a, b] → R2 be a smooth parameterized plane path defined

by:

r (t) = (x(t), y(t)), t ∈ [a, b].

Then r is rectifiable and its length is:

∫b √
L = L(r ) = x '2 (t) + y '2 (t)dt.
a
124 5 Line Integrals

Example 5.2.2 Compute the length of the circle x 2 + y 2 = R 2 .

A parametric representation of the circle is:


x = R cos t
, t ∈ [0, 2 π ], R > 0.
y = R sin t

Using Remark 5.2.1, we have:

∫2π √ ∫2π √ ∫2π

L= x '2 (t) + y '2 (t)dt = R2 sin t +
2
R2 cos2 tdt = R dt = 2π R.
0 0 0

x2 y2
Example 5.2.3 Compute the length of the ellipse a2
+ b2
− 1 = 0, a, b > 0.

A
 parametric representation of the ellipse is:
x = a cos t
, t ∈ [0, 2 π ], a, b > 0.
y = b sin t
It is enough to calculate a quarter of the length of the ellipse. By Remark 5.2.1,
we have:
π π
∫ 2 √ ∫2 /
L ( )
= a sin t + b cos t dt =
2 2 2 2 a 2 − a 2 − b2 cos2 tdt.
4
0 0

If we denote by c the focal length and by ε the eccentricity of the ellipse, then
a 2 − b2 = c2 and ε = ac ∈ (0, 1).a 2 − b2 = c2 and ε = ac ∈ (0, 1).
Next, we have:

∫2 /
π π
( ) ∫ 2 √
L c 2
=a 1− cos2 tdt = a 1 − ε2 cos2 t dt
4 a
0 0
π π
∫ / 2
(π ) ∫ 2 √
=a 1 − ε2 sin2 − t dt = a 1 − ε2 sin2 t dt.
2
0 0

Remains to calculate the integral:

π
∫ 2 √
1 − ε2 sin2 t dt, 0 < ε < 1.
0

The antiderivative of this function is not an elementary function, and therefore

the computation of this integral cannot be done with the Leibniz–Newton formula.
5.2 Rectifiable Paths and Curves 125

Attempting to calculate the length of the ellipse led to an integral that cannot be
computed exactly. Such an integral is called an elliptic integral. The following types
of elliptic integrals are known:
(1) Elliptic integral of the first kind:
π
∫2
1
K (k) = √ dϕ, k ∈ (0, 1).
1 − k 2 sin2 ϕ
0

(2) Elliptic integral of the second kind:

π
∫2 /
E(k) = 1 − k 2 sin2 ϕ dϕ, k ∈ (0, 1).
0

(3) Elliptic integral of the third kind:

π
∫2
1
F(k, h) = ( )√ dϕ, k ∈ (0, 1).
1 + h sin ϕ 1 − k 2 sin2 ϕ
2
0

The computation of these integrals is done with approximate methods, and tables
have been drawn up with their (approximate) values for different values of the
parameters k, respectively, k and h.

Remark 5.2.2 Let f : [a, b] → R be a function of C 1 -class. The length of the

graph of this function is:

∫b √
L= 1 + f '2 (x) dx.
a

Proof Indeed, the function f defines a smooth parameterized path, namely:


x=t
, t ∈ [a, b].
y = f (t)

The graph of f coincides with the support of this path. The statement results now
from Remark 5.2.1.

Example 5.2.4 Compute the length of the graph of the function:

(x )
f : [0, b] → R, f (x) = a ch , x ∈ [0, b], a > 0.
a
126 5 Line Integrals

Fig. 5.6 Catenary y

O b x

The graph of this function is called catenary, and it is shown in Fig. 5.6.
From Remark 5.2.2, we deduce that:

∫b / ( ) ∫b (x )
2 x
L= 1 + sinh dx = cosh dx
a a
0 0
( x )|b ( )
| b
= a sinh | = a sinh .
a 0 a

Remark 5.2.3 Let ρ : [α, β] → R, ρ = ρ(θ ), be a function of C 1 -class and let

r : [α, β] → R2 be the path given by r (θ ) = (ρ(θ ) cos θ, ρ(θ ) sin θ ), θ ∈ [α, β].
Then r is rectifiable and its length is:

∫β √
L = L(r ) = ρ 2 (θ ) + ρ '2 (θ ) dθ.
α

Proof Indeed, a parametric representation of the path is:


x = ρ(θ ) cos θ
, θ ∈ [α, β].
y = ρ(θ ) sin θ


The support of this path is the arc AB, represented in Fig. 5.7. According to
Theorem 5.2.1, we have:

∫β /
L= (ρ ' (θ ) cos θ − ρ(θ ) sin θ )2 + (ρ ' (θ ) sin θ + ρ(θ ) cos θ )2 dθ
α
∫β √
= ρ 2 (θ ) + ρ '2 (θ ) dθ.
α
5.2 Rectifiable Paths and Curves 127

Fig. 5.7 The length of a

curve in polar coordinates

Fig. 5.8 The length of the

cardioid

Example 5.2.5 Compute the length of the cardioid:

ρ(θ ) = a(1 + cos θ ), θ ∈ [0, 2 π ], a > 0.

The cardioid is represented in Fig. 5.8.

For reasons of symmetry, it is enough to compute only a half of the length of this
curve. From Remark 5.2.3 we have:
∫π / ∫π √
L
= a (1 + cos θ ) + a sin θ dθ =
2 2 2 2
2a 2 (1 + cos θ )dθ
2
0 0
∫π / ( )
θ
= 2a · 2 cos
2 2 dθ
2
0
|
∫π ( ) ( )|π
θ θ ||
= 2a cos dθ = 4a sin = 4a.
2 2 ||
0 0

Therefore, the length of the cardioid is L = 8a.

128 5 Line Integrals

Remark 5.2.2 According to Theorem 5.2.1, it follows that if r1 : [a, b] → R3 and

r2 : [b, c] → R3 are two smooth parameterized paths and if r = r1 ∪ r2 is the path
obtained by their union, then r is rectifiable and:

L(r ) = L(r1 ) + L(r2 ).

Moreover, any piecewise smooth curve is rectifiable and its length is the sum of
the lengths of its smooth parts.

5.3 Natural Parameterization of a Curve

A curve admits an infinite parametric representations. Next, we will present an impor-

tant parameterization of the rectifiable curves, namely the natural parameterization.
Let r : [a, b] → R3 be a smooth parameterized path defined by:

r (t) = (x(t), y(t), z(t)), t ∈ [a, b].

According to Theorem 5.2.1, the path r is rectifiable and its length is:

∫b √
L = L(r ) = x '2 (t) + y '2 (t) + z '2 (t) dt.
a

For any t ∈ [a, b], we consider the function:

∫ t √
s : [a, b] → [0, L], s = λ(t) = x '2 (u) + y '2 (u) + z '2 (u) du.
a

If A(x(a), y(a), z(a)) and B(x(b), y(b), z(b)) are the ends of the path support

and M(x(t), y(t), z(t)) belongs to the support AB, then s = λ(t) represents the

length of the arc √ AM (Fig. 5.9).
Since λ' (t) = x '2 (t) + y '2 (t) + z '2 (t) > 0, ∀ t ∈ [a, b], λ(a) = 0, λ(b) = L,
then, it results that the function λ : [a, b] → [0, L] is of C 1 -class, strictly increasing
and bijective.
Its inverse λ−1 : [0, L] → [a, b] is also a function of C 1 -class.
( )
Let the vector function r̃ : [0, L] → R3 , given by r̃ (s) = r λ−1 (s) , s ∈ [0, L].
Since

r̃ (λ(t)) = r (t), ∀ t ∈ [a, b]

5.3 Natural Parameterization of a Curve 129

Fig. 5.9 The length of the

arc AM

Fig. 5.10 The diagramme of

parametric representation of
the curve in function of the
arc element

it results that the paths ([a, b], r ) and ([0, L], r̃ ) are equivalent to the same orientation,
and hence λ is a change of parameter.
⎧ ( )
⎨ x̃(s) = x (λ−1 (s))
Definition 5.3.1 The parametric equations ỹ(s) = y λ−1 (s) , s ∈ [0, L],
⎩ ( )
z̃(s) = z λ−1 (s)
define a new parametric representation of the path ([a, b], r ) called the natural
parameterization.

Example 5.3.1 Find the natural parametric representation of the curve:

x 2 + y 2 = R 2 , x, y ≥ 0.

The support of this plane curve is the quarter of circle in the first quadrant, which
has the parametric representation:
 [ π]
x = R sin t
, t ∈ 0, , R > 0.
y = R cos t 2

∫2 √
π

πR
Its length is: L = R 2 cos2 t + R 2 sin2 t dt = 2
.
0
−−→ → 
If O M = −
r (t), then the length of the arc AM, which we denote by s (Fig. 5.11),
is:
130 5 Line Integrals

Fig. 5.11 The parametrical

representation of the arc AM
in function of S (the arc
element)

∫ t √
s = λ(t) = R 2 sin2 u + R 2 cos2 u du = Rt.
0

[ ] [ ]
Obviously, λ is a function strictly increasing. It follows that λ : 0, π2 → 0, π2R
[ ] [ ]
is bijective and its inverse λ−1 : 0, π2R → 0, π2 is defined by:
[ ]
−1 s πR
λ (s) = t = , s ∈ 0, .
R 2

The natural representation of this path is:

 ( ) [ ]
x̃(s) = R sin (Rs ) πR
, s ∈ 0, .
ỹ(s) = R cos Rs 2

Example 5.3.2 Find the natural parameterization of the cylindrical helix.

A parametric representation of the circular helix is:

⎧
⎨ x = R cos t
y = R sin t , t ∈ [0, 2π ], R > 0, h > 0
⎩
z = ht
√
and its length is L = 2 π R 2 + h 2 (Example 5.2.1). Then we have:

∫ t √ √
s = λ(t) = R 2 sin2 u + R 2 cos2 u + h 2 dt = t R 2 + h 2 , t ∈ [0, 2π ].
0

[ √ ]
The inverse function is λ−1 : 0, 2 π R 2 + h 2 → [0, 2π ],
5.3 Natural Parameterization of a Curve 131

s [ √ ]
λ−1 (s) = √ , s ∈ 0, 2π R 2 + h 2 .
R2 + h2

Therefore, the natural parametric representation of the cylindrical helix is:

⎧ ( )
⎪
⎪ x̃(s) = R cos √ s
⎨ ( R +h )2 2
[ √ ]
ỹ(s) = R sin √ s , s ∈ 0, 2 π R 2 + h2 .
⎪
⎪ R 2 +h 2
⎩ z̃(s) = h √ s
R 2 +h 2

Theorem 5.3.1 Let us consider r̃ (s) = (x̃(s), ỹ(s), z̃(s)), s ∈ [0, L] the natural
parametric representation of the path (I, r ). Then:
∥ ∥
∥ dr̃ ∥
∥ ∥ = 1 and x̃ '2 (s) + ỹ '2 (s) + z̃ '2 (s) = 1.
∥ ds ∥
( )
Proof Since r̃ (s) = r λ−1 (s) , s ∈ [0, L], it results that:
( ) ( )
dr̃ dr λ−1 (s) d λ−1 (s)
= ( ) · .
ds d λ−1 (s) ds

On the other hand, on account of inverse function theorem, we have:

( )
d λ−1 (s) 1
= '
ds λ (t)

where t = λ−1 (s). √

Taking into account to the fact that λ' (t) = x '2 (t) + y '2 (t) + z '2 (t), t ∈ I , we
deduce:

dr̃ dr 1 1 ( ' )
= · ' = √ x (t), y ' (t), z ' (t) (5.12)
ds dt λ (t) x '2 (t) + y '2 (t) + z '2 (t)

hence
∥ ∥
∥ dr̃ ∥ 1 √
∥ ∥ = √ · x '2 (t) + y '2 (t) + z '2 (t) = 1.
∥ ds ∥
x '2 (t) + y '2 (t) + z '2 (t)
( )
Since dr̃
ds
= x̃ ' (s), ỹ ' (s), z̃ ' (s) , it follows that:

x̃ '2 (s) + ỹ '2 (s) + z̃ '2 (s) = 1. (5.13)

132 5 Line Integrals

d−
→r
The vector dt
is known to be tangent to the curve. From (5.12), we deduce that
−
→
d r̃ d−
→r
ds
is collinear to dt
, and thus it is also tangent to the curve. On the other hand,
−
→
d r̃
from (5.13), it results that ds
is a unit vector.
−
→ −
→ −
→ −
→
Remark 5.3.1 The vector ddsr̃ = x̃ ' (s) i + ỹ ' (s) j + z̃ ' (s) k is the unit tangent
vector to the curve at the point M(x̃(s), ỹ(s), z̃(s)).

Remark 5.3.2 Although any smooth curve has a natural parametric representation,
it cannot always be found, because we cannot always calculate the definite integral
in the right-hand side of the formula:

∫ t √
s = λ(t) = x '2 (u) + y '2 (u) + z '2 (u) du.
a

However, the theoretical importance of the natural representation of a curve is

very high, as we will see in the next paragraph.

5.4 Line Integrals of the First Kind

The line integral is an extension of the definite integral, in the sense that the integrated
interval [a, b] is replaced by a space (or plane) curve.
Let γ be a smooth curve and let x = x(t), y = y(t), z = z(t), t ∈ [a, b] be a
parametric representation of it. Such a curve is rectifiable and its length is:

∫b √
L= x '2 (t) + y '2 (t) + z '2 (t) dt.
a

Let x = x̃(s), y = ỹ(s), z = z̃(s), s ∈ [0, L] be its natural parameterization,

and let f : Ω ⊂ R3 → R be a continuous function. We assume also that the curve
support is included in Ω.

Definition 5.4.1 The line integral of the first kind (line integral
∫ with respect to
arc length) of the function f along the curve γ is denoted by γ f (x, y, z) ds and
is defined by

∫ ∫L
f (x, y, z)ds = f (x̃(s), ỹ(s), z̃(s))ds. (5.14)
γ 0
5.4 Line Integrals of the First Kind 133
∫ ( )
Example 5.4.1 Compute γ x − y + z 2 ds, where γ is the cylindrical helix:
⎧
⎨ x = R cos t
y = R sin t , t ∈ [0, 2 π ], R > 0, h > 0.
⎩
z = ht

According to Example 5.3.2, the natural parametric representation of γ is:

( ) ) (
s s
hs
x̃ = R cos √ , ỹ = R sin √ , z̃ = √ ,
R2 + h2 R +h
2 2 R + h2
2

[ √ ]
where s ∈ 0, 2 π R 2 + h 2 . From Definition 5.4.1, we have:
∫
( )
x − y + z 2 ds
γ
√
2 π ∫R 2 +h 2( ( ) ( ) )
s s h2s2
= R cos √ − R sin √ + 2 ds
R2 + h2 R2 + h2 R + h2
0
√ ( ( ) ( ))|2π √ R 2 +h 2
s s|
=R R2 + h2 · sin √ + cos √ |
R2 + h2 R +h
2 2 |
0
|2 π √
| R 2 +h 2
h s |
2 3
8h π
2 3 √
+ ( 2 )| = · R2 + h2.
3 R + h2 | 3
0

Next, we present the physical interpretation of the line integral of the first kind.
Let us suppose that a non-homogeneous wire of negligible thickness is described
by the smooth curve γ .
 
We denote by AB the support of curve γ and by f : AB → R+ the continuous
positive function that expresses the linear density of the wire.
Let Δ be an arbitrary partition of the interval [0, L]:

Δ : 0 = s0 < s1 < · · · < si− 1 < si < · · · < sn = L


and let Mi (x̃(si ) , ỹ(si ), z̃(si )) ∈ AB, i = 0, n, where A = M0 and B = Mn

(Fig. 5.12). We specify that si represents the length of the arc AMi . If the partition
Δ is refining enough, we can assume that the density of the wire is constant on the

arc Mi−1 Mi ; i.e. the density is equal to the value of the function f at one of the ends
of this arc, e.g.

f (M) = f (Mi ), ∀M ∈ Mi−1 Mi .
134 5 Line Integrals

Fig. 5.12 The mass of a

material wire


It follows that the mass of the piece of the wire Mi−1 Mi is approximately equal
to f (Mi )(si − si−1 ) and the total mass of the wire is approximated by the sum:

Σ
n Σ
n
f (Mi )(si − si−1 ) = f (x̃(si ), ỹ(si ), z̃(si ))(si − si−1 ).
i=1 i=1

The exact value of the total mass of the wire γ will be:

Σ
n
Mass(γ ) = lim f (x̃(si ), ỹ(si ), z̃(si ))(si − si−1 ).
Δ →0
i=1

Taking into account to Definition 5.4.1, it results that the mass of the wire is given
by:

∫L ∫
Mass(γ ) = f (x̃(s), ỹ(s), z̃(s))ds = f (x, y, z)ds.
0 γ

The exact meaning of the above limit is as follows: ∀ε > 0, ∃δε > 0 such that
for any partition Δ of the interval [0, L], with Δ < δε , we have:
| |
| Σ
n |
| |
|Mass(γ ) − f (x̃(si ), ỹ(si ), z̃(si ))(si − si−1 )| < ε.
| |
i=1

∫
In conclusion, the line integral of the first kind γ f (x, y, z)ds is the mass of a
non-homogeneous wire of negligible thickness, which is described by the smooth
curve γ and has the continuous density function f .
If we denote by x G , yG , z G the coordinates of the center of mass G of the
non-homogeneous wire γ , then it is shown that:
5.4 Line Integrals of the First Kind 135
∫ ∫ ∫
γ x f (x, y, z)ds γ y f (x, y, z)ds γ z f (x, y, z)ds
xG = ∫ , yG = ∫ , zG = ∫ .
γ f (x, y, z)ds γ f (x, y, z)ds γ f (x, y, z)ds
( )

In the case of a homogeneous wire i.e., f (M) = k = ct. , ∀ M ∈ AB , it
results that:
∫ ∫ ∫ ∫
1 1 1
Mass(γ ) = k 1ds = k · L , x G = · x ds, yG = · y ds, z G = · z ds.
L L L
γ γ γ γ

Using the same argument as above, we deduce that the moments of inertia of
the wire γ with respect to the origin O(0, 0, 0) of the axes and with respect to the
coordinate axes and the coordinate planes are given by the formulas:
∫
( )
IO = x 2 + y 2 + z 2 f (x, y, z) ds
γ
∫
( )
IO x = y 2 + z 2 f (x, y, z) ds
γ
∫
( )
IO y = x 2 + z 2 f (x, y, z) ds
γ
∫
( )
I Oz = x 2 + y 2 f (x, y, z) ds
γ
∫
IO x y = z 2 f (x, y, z) ds
γ
∫
IO x z = y 2 f (x, y, z) ds
γ
∫
I O yz = x 2 f (x, y, z) ds.
γ

If we knew the natural parametric representation of a curve, then Formula (5.14)

would be sufficient to calculate the line integral of the first kind along this curve.
Usually, a curve is given by a parametric representation in which the param-
eter t is arbitrary, and its natural representation cannot be found. The following
theorem allows the computation of the line integral of the first kind if the parametric
representation is arbitrary.

Theorem 5.4.1 Let γ be a smooth curve and let x = x(t), y = y(t), z = z(t),
t ∈ [a, b] be a parametric representation of it. If Ω ⊂ R3 is a domain that contains
the support of the curve γ and f : Ω → R is a real continuous function on Ω, then
136 5 Line Integrals

we have:

∫ ∫b √
f (x, y, z)ds = f (x(t), y(t), z(t)) x '2 (t) + y '2 (t) + z '2 (t) dt. (5.15)
γ a

Proof Since f is continuous and the functions x, y, z are of C 1 -class on [a, b],
it follows that the integral from the right member of (5.15) exists. According to
Definition 5.4.1, we have:

∫ ∫L
f (x, y, z)ds = f (x̃(s), ỹ(s), z̃(s)) ds.
γ 0

If we make the change of variable s = λ(t), t ∈ [a, b], it results that:

x̃ ◦ λ = x, ỹ ◦ λ = y, z̃ ◦ λ = z.
√
ds = λ' (t)dt = x '2 (t) + y '2 (t) + z '2 (t) dt.

and further:

∫ ∫L
f (x, y, z)ds = f (x̃(s), ỹ(s), z̃(s))ds
γ 0
−1
λ∫ (L)

= f (x(t), y(t), z(t))λ' (t)dt

λ−1 (0)

∫b √
= f (x(t), y(t), z(t)) · x '2 (t) + y '2 (t) + z '2 (t) dt.
a

∫( )
Example 5.4.2 Compute x − y + z 2 ds, where γ is the cylindrical helix:
γ

x = R cos t, y = R sin t, z = ht, t ∈ [0, 2π ], R, h > 0.

According to Theorem 5.4.1, we get:

∫ ( ) ∫2 π( )√
x − y + z 2 ds = R cos t − R sin t + h 2 t 2 R 2 sin2 t + R 2 cos2 t + h 2 dt
γ 0
( )|2π
√ h2t 3 | 8h 2 π 3 √ 2
|
= R 2 + h 2 R sin t + R cos t + | = · R + h2.
3 | 3
0
5.4 Line Integrals of the First Kind 137

Example 5.4.3 Find the mass and the moment of inertia with respect to O z axis for
a wire γ lying along the cylindrical helix with the constant density f (x, y, z) = 1.
The parametric equations of the wire are:

x = R cos t, y = R sin t, z = ht, t ∈ [0, 2 π ], R, h > 0.

√
According to Example 5.2.1, the length of the wire is L = 2π R2 + h2.
Then the mass of the wire is:
∫ √
Mass(γ ) = 1ds = L = 2 π R 2 + h 2 .
γ

The moment of inertia with respect to O z axis is calculated as follows:

∫
( )
I Oz = x 2 + y 2 ds
γ

∫2π
( )√
= R 2 cos2 t + R 2 sin2 t R 2 sin2 t + R 2 cos2 t + h 2 dt
0
∫2π √ √
= R 2 R 2 + h 2 dt = 2π R 2 R 2 + h 2 .
0

Remark 5.4.1 The definition and formula for computing the line integral of the first
kind along a space smooth curve are directly transposed to the case when the function
is defined on the points of a smooth plane curve having the parametric representation:

x = x(t), y = y(t), t ∈ [a, b].

In this case, the computation Formula (5.15) becomes:

∫ ∫b √
f (x, y)ds = f (x(t), y(t)) x '2 (t) + y '2 (t) dt. (5.16)
γ a

∫
Example 5.4.4 Compute x y ds, where γ is the arc of the first quadrant of the
γ
x2 y2
ellipse a2
+ b2
− 1 = 0, a, b > 0.

x = a cos t [ ]
A parametric representation of curve γ is , t ∈ 0, π2 , a, b > 0.
y = b sin t
Using Remark 5.4.1, we have:
138 5 Line Integrals

π
∫ ∫2 √
x y ds = ab cos t sin t a 2 sin2 t + b2 cos2 t dt.
γ 0

[ ]
If we consider the change of variable u = a 2 sin2 t + b2 cos2 t, t ∈ 0, π2 , then
( 2 )
du = 2 a − b2 sin t cos t dt, and further we obtain:

∫ ∫a 2 |a 2
ab √ ab √ ||
x y ds = ( 2 ) u du = ( 2 ) · u u|
2 a − b2 3 a − b2 |2
γ b2 b
( 3 ) ( )
ab a − b3 ab a 2 + ab + b2
= ( 2 ) = .
3 a − b2 3(a + b)

Remark 5.4.2 If the smooth plane curve γ is given by an explicit representation,

y = y(x), x ∈ [a, b], then x can be taken as a parameter on the curve and Formula
(5.16) becomes:

∫ ∫b √
f (x, y)ds = f (x, y(x)) 1 + y '2 (x) dx. (5.17)
γ a

∫ 1
Example 5.4.5 Compute γ y−x ds, where γ is the segment on the straight line
y = 2x + 1 from point A(0, 1) to point B(2, 5).

According to Remark 5.4.2, we have:

∫ ∫2 √ √ ∫
2
1 1 1
ds = · 1 + 22 dx = 5 dx
y−x 2x + 1 − x x +1
γ 0 0
√ |2 √
|
= 5 ln|x + 1|| = 5 ln 3.
0

Remark 5.4.1 If γ is a piecewise smooth curve and the function f is continuous and
bounded on each smooth piece of the curve γ , then:
∫ p ∫
Σ
f (x, y, z) ds = f (x, y, z)ds,
γ i=1 γ
i

where γ = γ1 ∪ γ2 ∪ · · · ∪ γ p .

Remark 5.4.2 The line integral of the first kind does not depend on the orientation
of the curve γ .
5.4 Line Integrals of the First Kind 139

Indeed, a parametric representation of the curve γ− is the following:

x = x̃(L − s), y = ỹ(L − s), z = z̃(L − s), s ∈ [0, L].

If we make the change of variable u = L − s, we obtain:

∫ ∫L
f (x, y, z) ds = f (x̃(L − s), ỹ(L − s), z̃(L − s)) ds
γ− 0

∫0 ∫L
=− f (x̃(u), ỹ(u), z̃(u))du = f (x̃(u), ỹ(u), z̃(u))du
L 0
∫
= f (x, y, z)ds.
γ+

Remark 5.4.3 Since the line integral of the first kind was defined as a definite integral,
it preserves the basic properties of the definite integral. In what follows, we will
present the basic properties of the line integral of the first kind, frequently used in
computations.

Proposition 5.4.1 (Linearity Property)

∫
(α f (x, y, z) + βg(x, y, z)) ds
γ
∫ ∫
=α f (x, y, z) ds + β g(x, y, z) ds.
γ γ

Proposition 5.4.2 (Monotony Property) ∫

If f is positive and continuous function, then f (x, y, z) ds ≥ 0.
γ

Proposition 5.4.3 (Additivity Property)

  
If an arc ABis composed by two arcs AC and C B, then:
∫ ∫ ∫
f (x, y, z) ds = f (x, y, z) ds + f (x, y, z) ds.
  
AB AC CB

or
∫ ∫ ∫
f (x, y, z) ds + f (x, y, , z) ds + f (x, y, z) ds = 0.
  
AB BC CA
140 5 Line Integrals

Proposition 5.4.4 (Estimation of the modulus of the integral)

| |
|∫ | ∫
| |
| f (x, y, z) ds | ≤ | f (x, y, z)| ds.
| |
| |
γ γ

Theorem 5.4.2 (The mean value theorem)

 
If f is continuous on AB = γ , then there is a point (ξ, η, ζ ) ∈ AB such that:
∫
f (x, y, z) ds = f (ξ, η, ζ ) · L ,
γ

where L is the length of the curve γ .

5.5 Line Integrals of the Second Kind

Let γ be a smooth curve, let us consider x = x(t), y = y(t), z = z(t), t ∈ [a, b]

a parametric representation of γ and let x = x̃(s), y = ỹ(s), z = z̃(s), s ∈ [0, L]
be its natural parametric representation.

We denote also by AB the support of γ .
We will note by − →
τ = − →τ (M) the unit tangent vector at the point

M(x̃(s) , ỹ(s) , z̃(s)) ∈ AB, oriented in the sense of increasing the parameter
s (Fig. 5.13).
According to Remark 5.3.1, we have:

−
→ −
→ −
→ −
→
τ = x̃ ' (s) i + ỹ ' (s) j + z̃ ' (s) k .

Fig. 5.13 The unit vector of

the arc AB
5.5 Line Integrals of the Second Kind 141

−
→
Let F = (P, Q, R) : Ω ⊂ R3 → R3 be a continuous vector field. We further

assume that AB ⊂ Ω. We have:
−
→ −
→ −
→ −
→
F (x, y, z) = P(x, y, z) i + Q(x, y, z) j + R(x, y, z) k , ∀(x, y, z) ∈ Ω.

Definition 5.5.1 The line integral of the second kind (line integral with respect
−
→
to coordinates) of the vector field F = (P, Q, R) along the oriented curve γ+ is
denoted by
∫
P(x, y, z) dx + Q(x, y, z)dy + R(x, y, z) dz
γ+

and it is given by the following definite integral:

∫
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
γ+

∫ ∫L
−
→ − −
→
= F ·→
τ = F (x̃(s), ỹ(s), z̃(s)) · −
→
τ ds
γ+ 0

∫L
(
= P(x̃(s), ỹ(s), z̃(s)) · x̃ ' (s) + Q(x̃(s), ỹ(s), z̃(s)) · ỹ ' (s)
0
)
+ R(x̃(s), ỹ(s), z̃(s)) · z̃ ' (s) ds. (5.18)

Remark 5.5.1 The line integral of the second kind depends on the orientation of the
curve γ .

Indeed, the unit tangent vector to the curve γ− at the point M ∈ AB is −−
→
τ,
whence, it results that:

∫ ∫L
−
→ ( →)
P(x, y, z) dx + Q(x, y, z)dy + R(x, y, z) dz = F (x, y, z) · −−
τ ds
γ− 0

∫L ∫
−
→
=− F (x, y, z) · −
→
τ ds = − P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz.
0 γ+

∫
Example 5.5.1 Compute x dx + y dy + z dz, where γ is cylindrical helix:
γ+
142 5 Line Integrals
⎧
⎨ x = 3 cos t
y = 3 sin t , t ∈ [0, 2 π ].
⎩
z = 4t

According to Example 5.3.2, the natural representation of the curve γ is:

⎧ (s)
⎨ x̃(s) = 3 cos( 5)
ỹ(s) = 3 sin 5s , s ∈ [0, 10π ].
⎩
z̃(s) = 45s

From Formula (5.18), it results that:

∫
x dx + y dy + z dz
γ+

∫10 π( ( s ) ( 3 ( s )) (s ) 3 ( s ) 4s 4 )
= 3 cos · − sin + 3 sin · cos + · ds
5 5 5 5 5 5 5 5
0
|10π
∫10π |2|
16s 16 s |
= ds = · | = 32π 2 .
25 25 2 |
0 0

For the physical interpretation of the line integral of the second kind, we consider a

smooth curve γ such that the arc AB is its support.
Let x = x̃(s), y = ỹ(s), z = z̃(s), s ∈ [0, L] be the natural parameterization
−
→ 
of γ and let F = (P, Q, R) : AB ⊂ R3 → R3 be a continuous vector field. We
consider also an arbitrary partition Δ of the interval [0, L]:

Δ : 0 = s0 < s1 < · · · < si−1 < si < · · · < sn = L .


We denote by Mi (x̃(si ), ỹ(si ), z̃(si )) ∈ AB, i = 0, n (Fig. 5.14). The length of

the arc Mi−1 Mi is si − si−1 .

Fig. 5.14 The mechanical

work of a variable force
which acts along the curve
5.5 Line Integrals of the Second Kind 143
[ ]
Let ξi ∈ si−1 , si be an arbitrary point, let Ni (x̃(ξi ), ỹ(ξi ), z̃(ξi )) be the corre-

sponding point on the arc M M , and let −
i−1
→τ be the unit tangent vector to the curve
i i
γ+ at the point Ni . If the partition Δ is refining enough, we can assume that the vector
−
→ 
field F = (P, Q, R), which we interpret as a force, is constant on the arc Mi−1 Mi ;
i.e. it is equal to its value at the point Ni .
−
→
Under these conditions, the work done by the force F acting on an object, moving

it along the arc Mi−1 Mi can be approximated by:
−
→
F (Ni ) · −
→
τ i (si − si−1 )

−
→
where by F (Ni ) · −
→
τ i means the scalar product of the two vectors.
−→ 
The total work done by the force F to move an object along the entire arc AB is
approximated by the sum:

Σ
n
−
→ Σ
n
−
→
F (Ni ) · −
→
τ i (si − si−1 ) = F (x̃(ξi ), ỹ(ξi ), z̃(ξi )) · −
→
τ i (si − si−1 )
i=1 i=1
Σ
n
(
= P(x̃(ξi ), ỹ(ξi ), z̃(ξi )) · x̃ ' (ξi )
i =1
+ Q(x̃(ξi ), ỹ(ξi ), z̃(ξi )) · ỹ ' (ξi )
)
+ R(x̃(ξi ), ỹ(ξi ), z̃(ξi )) · z̃ ' (ξi ) (si − si−1 ).

The exact value of the total work will be equal to:

Σ
n
−
→
W = lim F (Ni ) · −
→
τ i (si − si−1 )
Δ →0
i=1
∫
= P(x, y, z) dx + Q(x, y, z)dy + R(x, y, z)dz.
γ+

Therefore, the line integral of the second kind

∫
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
γ+

−
→ −
→ −
→ −
→
is the mechanical work done by the variable force F = P i + Q j + R k for
moving an object along the oriented curve γ+ .
Dependence of line integral of the second kind on the orientation of the integrated
path is coherent with physical interpretation of the integral as being the work of a
144 5 Line Integrals

field of force along a path. Indeed, if the direction of tracing the trajectory is reversed
the work performed by the force field changes its sign in the opposite.
The following theorem allows the computation of the line integral of the second
kind when the parametric representation of the curve is arbitrary.

Theorem 5.5.1 Let γ be a smooth curve and let us take x = x(t), y = y(t), z = z(t),
t ∈ [a, b], one of its parametric representation. We denote by γ+ the curve oriented
in the direction of increasing the parameter. If Ω ⊂ R3 is a domain including the
−
→
support of γ and F = (P, Q, R) : Ω → R3 is a continuous vector field, then:
∫
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
γ+

∫b
(
= P(x(t), y(t), z(t)) · x ' (t) + Q(x(t), y(t), z(t)) · y ' (t)
a
)
+ R(x(t), y(t), z(t)) · z ' (t) dt. (5.19)

Proof Clearly, the integral of the right member of (5.19) exists, because x, y, z are
of C 1 -class on [a, b] and P, Q, R are continuous. According to Definition 5.5.1, we
have:
∫
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
γ+

∫L
(
= P(x̃(s), ỹ(s), z̃(s)) · x̃ ' (s) + Q(x̃(s), ỹ(s), z̃(s)) · ỹ ' (s)
0
)
+ R(x̃(s), ỹ(s), z̃(s)) · z̃ ' (s) ds.

If we make the change of variable change:

∫ t √
s = λ(t) = x '2 (u) + y '2 (u) + z '2 (u) du, t ∈ [a, b]
a

then we obtain:
( )
x̃(λ(t)) = x λ−1 (λ(t)) = x(t)
ỹ(λ(t)) = y(t), z̃(λ(t)) = z(t).

Also, taking into account the differential rules for composed and inverse functions,
we have:
5.5 Line Integrals of the Second Kind 145
( ) ( )
' d ( ( −1 )) dx λ−1 (s) d λ−1 (s) 1
x̃ (s) = x λ (s) = ( −1 ) · = x ' (t) · ' .
ds d λ (s) ds λ (t)

Similarly, we have:
ỹ ' (s) = y ' (t) · λ'1(t) ; z̃ ' (s) = z ' (t) · 1
λ' (t)
and
√
ds = λ' (t) dt = x '2 (t) + y '2 (t) + z '2 (t) dt.

Therefore, we obtain:

∫L
(
P(x̃(s), ỹ(s), z̃(s)) · x̃ ' (s) + Q(x̃(s), ỹ(s), z̃(s)) · ỹ ' (s)
0
)
+ R(x̃(s), ỹ(s), z̃(s)) · z̃ ' (s) ds
∫b
(
= P(x(t), y(t), z(t)) · x ' (t) + Q(x(t), y(t), z(t)) · y ' (t)
a
)
+ R(x(t), y(t), z(t)) · z ' (t) dt.
∫
Example 5.5.2 Compute y dx + z dy + x dz, where γ has the parametric repre-
γ+
⎧
⎪
⎨ x = R
2 (1 + cos t)
sentation: y = 2 (1 − cos t) , t ∈ [0, 2π ], R > 0.
R
⎪
⎩ z = √R sin t
2

According to Theorem 5.5.1 (Formula (5.19)), we have:

∫
y dx + z dy + x dz
γ+

Fig. 5.15 The circle lied in z

the plane x+y = R, parallel to
Oz axis and passes through
the points A(R, 0, 0) and B(0, P
R, 0)

B
O y
A
Q
x
146 5 Line Integrals

∫2π( ( ) )
R R R R R R
= (1 − cos t) · − sin t + √ sin t · sin t + (1 + cos t) · √ cos t dt
2 2 2 2 2 2
0
∫2π ∫2π ∫2π ∫2π
R2 R2 R2 R2 π R2
=− sin t dt + sin t cos t dt + √ dt + √ cos t dt = √ .
4 4 2 2 2 2 2
0 0 0 0

We notice that from a geometric point of view, the support of curve γ is the circle:

x 2 + y2 + z2 = R2
x+y = R.

This circle lies in the plane x + y = R that is parallel to Oz axis and passes
through the points A(R, 0, 0) and ( B(0, R, ) 0). The straight line segment [AB] is a
diameter of the circle. The point R2 , R2 , 0 is its center and √R2 is its radius.

Remark 5.5.2 On an arbitrary curve γ we have two orientations. If we take two para-
metric representations (I1 , r1 ), (I2 , r2 ) of the curve γ , then the change of parameter
λ : I1 → I2 is increasing or decreasing. In the first case, (I1 , r1 ) and (I2 , r2 ) have
the same orientation of γ . In the second case, they have different orientations of γ .
Which of them is considered as positive is a choice who must be specified.

Remark 5.5.3 If the curve γ is given by a parametric representation, then γ+ repre-

sents the curve γ oriented in the sense of increasing the parameter. If the curve γ
is a closed space curve and it is given as an intersection of two surfaces, then the
orientation of the curve cannot be deduced from the context and must be specified by
the statement. If the curve γ is a closed plane curve, usually the positive orientation
of such a curve is counterclockwise.

For the line integral of the second kind on a closed curve, a special symbol is
used, namely the integral symbol provided with a circle in the middle of it, a circle
on which appears an arrow representing the direction of orientation of the curve.
When the arrow does not appear∮ on the circle placed on the integral sign, so when
it is denoted by the symbol , we will understand that the integration is performed
on a closed curve oriented in a positive direction. For example, in the case of the
closed space curve in Example 5.5.2, it can be specified that the orientation of curve
is counterclockwise if we look from the point O− the origin of the axis system.
Example 5.5.2 can be reformulated as follows:
∮
Compute y dx + z dy + x dz, where γ+ is the space circle
γ+


x 2 + y2 + z2 = R2
x+y =R

traced out in the counterclockwise direction, if we look straight down from the center
of the sphere.
5.5 Line Integrals of the Second Kind 147

Remark 5.5.1 For a plane curve γ having the parametric representation:

x = x(t), y = y(t), t ∈ [a, b]

−
→
and a vector field F = (P, Q), Formula (5.19) becomes:
∫
P(x, y) dx + Q(x, y) dy
γ+

∫b
( )
= P(x(t), y(t)) · x ' (t) + Q(x(t), y(t)) · y ' (t) dt. (5.20)
a

∮
Example 5.5.3 Compute x dy − y dx, where γ is the circle x 2 + y 2 = R 2 , R > 0,
γ
traced out in a counterclockwise direction.
We use the parametric representation of the circle:

x = R cos t
, t ∈ [0, 2π ].
y = R sin t

and we obtain:

∮ ∫2π ∫2π
x dy − y dx = (R cos t · R cos t − R sin t · (−R sin t))dt = R 2 dt = 2π R 2 .
γ 0 0

Remark 5.5.2 If the plane curve γ is given by the explicit representation y = y(x),
x ∈ [a, b], then the formula for computing the line integral of the second kind
becomes:

∫ ∫b
( )
P(x, y)dx + Q(x, y)dy = P(x, y(x)) + Q(x, y(x)) · y ' (x) dx. (5.21)
γ+ a

Remark 5.5.3 Similarly, if the plane curve γ is given by the explicit representation
x = x(y), y ∈ [c, d], then the formula for computing the line integral of the second
kind becomes:

∫ ∫d
( )
P(x, y)dx + Q(x, y)dy = P(x(y), y) · x ' (y) + Q(x(y), y) dy. (5.22)
γ+ c
148 5 Line Integrals

∫ ( ) 
Example 5.5.4 Compute 2 x y dx − x 2 + y dy, where γ+ is the arc AB of the
γ+
parabola y = x 2 , with A(1, 1) and B(2, 4).

The explicit representation of the arc AB is: y(x) = x 2 , x ∈ [1, 2].
From Remark 5.5.2, it results that:

∫ ∫2
( ) ( 3 ( 2 ) )
2x y dx − x + y dy =
2
2x − x + x 2 · 2x dx
 1
AB

∫2 |2
x 4 || 15
=− 2x dx = − | = − .
3
2 1 2
1

Remark 5.5.4 If γ is a union of smooth curves, γ = γ1 ∪ γ2 ∪ · · · ∪ γ p , then:

∫ p ∫
Σ
P dx + Q dy + R dz = P dx + Q dy + R dz.
γ+ i=1 (γ )
i +

∮
Example 5.5.5 Compute (z − y) dx + (x − z)dy + (y − x) dz, where γ is the
γ
triangle ABC with vertices A(a, 0, 0), B(0, b, 0), C(0, 0, c), a, b, c > 0 and the
orientation A → B → C → A.

Obviously, the curve is the union of the sides of the triangle: γ = AB ∪ BC ∪C A.

According to Remark 5.5.4, we have:
∮ ∫ ∫ ∫
(z − y)dx + (x − z)dy + (y − x)dz = + +
γ AB BC CA

Fig. 5.16 The triangle ABC

5.6 Independence on the Path of Line Integral of the Second Kind 149

The canonical equations of the straight line (AB) are:

x −a y−0 z−0
(AB) : = = = t,
0−a b−0 0−0

hence the parametric representation of the segment AB is:

⎧
⎨ x = a − at
AB : y = bt , t ∈ [0, 1].
⎩
z =0

Further, we have:
∫
(z − y) dx + (x − z) dy + (y − x) d z
AB
∫1 ∫1
= ((−bt) · (−a) + (a − at) · b + (bt − a + at) · 0)dt = a b dt = ab.
0 0

Similarly, it is shown that:

∫
(z − y) dx + (x − z) dy + (y − x) dz = bc
BC
∫
(z − y) dx + (x − z) dy + (y − x) dz = ca.
CA

Therefore, we have:
∮
(z − y)dx + (x − z)dy + (y − x) dz = ab + bc + ca.
γ

5.6 Independence on the Path of Line Integral

of the Second Kind

In this paragraph we will analyze the case when the value of the line integral of the
second kind depends only on the extremities of the curve and does not depend on
the shape of the curve itself. This case is interesting both from a mathematical point
of view, because the computation of such an integral is simpler, and from a practical
point of view, because it has applications in thermodynamics.
150 5 Line Integrals

Definition 5.6.1 Let D ⊂ R3 be an open subset and let P, Q, R : D → R be three

arbitrary functions. The following expression:

ω = P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz, (x, y, z) ∈ D

is called differential 1-form on D with coefficients P, Q, R.

Moreover, if the functions P, Q, R are of C k -class on D, then ω is called a
differential 1-form of C k -class.
Example 5.6.1 If f : D ⊂ R3 → R is differentiable function on D, then the
first differential d f = ∂∂ xf dx + ∂∂ yf dy + ∂∂ zf dz is a differential 1-form on D with the
∂f ∂f ∂f
coefficients , , .
∂x ∂y ∂z

Definition 5.6.2 A differential 1-form ω = P dx + Q dy + R dz on D is called

exact if there is a function f : D → R of C 1 -class on D such that ω = d f , which
means that the following equalities on D are verified:

∂f ∂f ∂f
P= , Q= , R= .
∂x ∂y ∂z
−
→
Remark 5.6.1 If we consider the vector field V : D ⊂ R3 → R3 ,
−
→ −
→ −
→ −
→
V (x, y, z) = P(x, y, z) i + Q(x, y, z) j + R(x, y, z) k , ∀(x, y, z) ∈ D,

−
→
then the differential 1-form ω = P dx + Q dy + R dz is exact on D if V is
a conservative vector field (potential vector field); i.e. there is a scalar function
−
→
f : D → R of C 1 -class on D such that V = grad f .
Definition 5.6.3 Let D ⊂ R∫3 be an open and connected subset. We say that the line
integral of the second kind P dx + Q dy + R dz is independent on the path in
γ
D if for any two points A , B ∈ D and any piecewise smooth curves γ1 and γ2 ,
which have the supports included in D and with the same ends A and B, we have:
∫ ∫
P dx + Q dy + R dz = P dx + Q dy + R dz.
γ1 γ2

Theorem 5.6.1 Let D ⊂ R3 be an open and connected set and let P, Q, R : D →

R three continuous functions on D. Then the differential 1-form ω = P dx +
∫Q dy + R dz is exact on D if and only if the line integral of the second kind
P d x + Q d y + R d z is independent on the path in D.
γ

Proof First, by hypothesis, there is a function f of C 1 -class on D such that:

∂f ∂f ∂f
P= , Q= , R= . (5.23)
∂x ∂y ∂z
5.6 Independence on the Path of Line Integral of the Second Kind 151

Let A and B be two arbitrary points in D and let γ be a piecewise smooth curve

whose support AB is included in D. If x = x(t), y = y(t), z = z(t), t ∈ [a, b] is
a parametric representation of γ , then A has the coordinates (x(a), y(a), z(a)) and
B has the coordinates (x(b), y(b), z(b)).
Let F : [a, b] → R be the composed function:

F(t) = f (x(t), y(t)z(t)), t ∈ [a, b].

Taking into account to the differential rules of the composed functions and the
equalities (5.23), it results that:

∂f ∂f
F ' (t) = (x(t), y(t), z(t)) · x ' (t) + (x(t), y(t), z(t)) · y ' (t)
∂x ∂y
∂f
+ (x(t), y(t), z(t)) · z ' (t)
∂z
= P(x(t), y(t), z(t)) · x ' (t) + Q(x(t), y(t), z(t)) · y ' (t)
+ R(x(t), y(t), z(t)) · z ' (t).

The equality is valid for any point t ∈ [a, b] except a finite number of points,
namely those points that correspond to the union points set of the smooth curves that
compose the curve γ . Since the equality is true excepting a null set, it follows that:
∫
P(x, y, z) dx + Q(x, y, z) dy + R(x, y, z) dz
γ+

∫b
(
= P(x(t), y(t), z(t)) · x ' (t) + Q(x(t), y(t), z(t)) · y ' (t)
a
)
+ R(x(t), y(t), z(t)) · z ' (t) dt
∫b
= F ' (t) dt = F(b) − F(a) = f (B) − f ( A).
a

Therefore, the value of the integral does not depend on the shape of the curve but
depends only on its ends.
Conversely, let M0 (x0 , y0 , z 0 ) ∈ D be a fixed point, M(x, y, z) ∈ D an arbitrary

point and γ a piecewise curve whose support M0 M is included in D (Fig. 5.17).
Since, by hypothesis, the line integral is independent on the path in D, it results that
we can define a function f : D → R such that for any M(x, y, z) ∈ D, we have:
∫
f (x, y, z) = P dx + Q dy + R dz.

M0 M
152 5 Line Integrals

Fig. 5.17 The independence

of the line integral on the
path in D M

N
z
M0

y D
x

Let N (x + h, y, z) ∈ D be a point such that the straight line segment M N ⊂ D

(we use here the fact that D is open set). A parametric representation of M N is:

x = t, y = y, z = z, t ∈ [x, x + h].

Further, we have:
∫
f (x + h, y, z) = P dx + Q dy + R dz
 
M0 M ∪ MN
∫ ∫
= P dx + Q dy + R dz + P dx + Q dy + R dz.
 
M0 M MN

According to Corollary 2.6.4, it results that:

∫
f (x + h, y, z) − f (x, y, z)  P dx + Qdy + Rdz
= MN
h h
∫ x+h
P(t, y, z) dt
= x
h
P(ξ, y, z)h
= = P(ξ, y, z),
h
where ξ is a point between x and x + h. Using again the fact that the function P is
continuous, it results that there is

f (x + h, y, z) − f (x, y, z)
lim = P(x, y, z),
h→0 h

that is:
∂f
∂x
= P.
Similarly, by substituting the segment M N with a segment parallel to O y axis,
respectively, Oz axis, it is shown that ∂∂ yf = Q, respectively, ∂∂ zf = R; hence ω is
exact.
5.6 Independence on the Path of Line Integral of the Second Kind 153

Fig. 5.18 The

representation of the closed
curve γ as the union of two
curves γ 1 and γ 2

Theorem 5.6.2 Let D ⊂ R3 be an open and connected set ∫ and let P, Q, R : D →

R three continuous functions on D. Then the line integral P dx + Q dy + R dz is
∮ γ
independent on path in D if and only if the line integral P dx + Q dy + R dz = 0,
γ
for any closed piecewise smooth curve γ whose support is included in D.

Proof First, let γ1 , γ2 be two piecewise smooth curves that have the same ends and
that have their supports included in D (Fig. 5.18) and let γ = γ1 ∪ (γ2 )− . Obviously
γ is a closed piecewise smooth curve, whose support is included in D. From the
hypothesis, it follows that:
∮
P dx + Q dy + R dz = 0.
γ

On the other hand, we have:

∮ ∫ ∫
P dx + Q dy + R dz = P dx + Q dy + R dz + P dx + Q dy + R dz
γ γ1 (γ2 )−
∫ ∫
= P dx + Q dy + R dz − P dx + Q dy + R dz = 0.
γ1 γ2

It results that:
∫ ∫
P dx + Q dy + R dz = P dx + Q dy + R dz,
γ1 γ2

∫
that is the line integral P dx + Q dy + R dz is independent on path in D.
γ
Conversely, let γ be a closed, piecewise smooth curve, whose support is included
in D and let r (t) = (x(t), y(t), z(t)), t ∈ [a, b] be an arbitrary parametric repre-
sentation of γ . For any a < c < b, we denoted by γ1 the curve r = r (t), t ∈ [a, c],
and by γ2 the curve r = r (t), t ∈ [c, b]. Obviously, γ = γ1 ∪ γ2 . By hypothesis,
we have:
154 5 Line Integrals
∫ ∫
P dx + Q dy + R dz = P dx + Q dy + R dz,
(γ1 )+ (γ2 )+

whence, we deduce:
∫ ∫ ∫
P dx + Q dy + R dz = P dx + Q dy + R dz + P dx + Q dy + R dz = 0.
γ (γ1 )+ (γ2 )−

Definition 5.6.4 A differential 1-form ω = P dx + Q dy + R dz is called closed

in D ⊂ R3 if its coefficients P, Q, R : D → R are functions of C 1 -class on D and
for any point in D we have:

∂P ∂Q ∂P ∂R ∂Q ∂R
= , = , = .
∂y ∂x ∂z ∂ x ∂z ∂y
−
→
Remark 5.6.2 Let V = (P, Q, R) : D ⊂ R3 → R3 be a vector field of C 1 -class
on D. Then the differential 1-form ω = P dx + Q dy + R dz is closed in D if and
⇀
only if the field V is irrotational, i.e.:
( ) ( ) ( )
−
→ ∂R ∂Q − → ∂P ∂R − → ∂Q ∂P − → −→
Curl V = − i + − j + − k = 0 , ∀(x, y, z) ∈ D.
∂y ∂z ∂z ∂x ∂x ∂y

Theorem 5.6.3 If the differential 1-form ω = P dx + Q dy + R dz with coefficients

P , Q , R : D → R of C 1 -class is exact in D, then ω is closed in D.

Proof By hypothesis, there is a function f : D → R of C 2 -class on D such that

P = ∂∂ xf , Q = ∂∂ yf , R = ∂∂ zf . Since the second-order derivatives of the function f are
continuous, it follows that the mixed derivatives are equal.

Therefore, we have:

∂P ∂2 f ∂2 f ∂Q
= = =
∂y ∂ y∂ x ∂ x∂ y ∂x

∂P ∂2 f ∂2 f ∂R
= = =
∂z ∂z∂ x ∂ x∂ z ∂x
∂Q ∂2 f ∂2 f ∂R
= = = .
∂z ∂ z∂ y ∂ y∂z ∂y

Definition 5.6.5 A subset S ⊂ R3 is said to be a star domain (star convex set)

if there is a point A ∈ S with the property that the straight line segment [A , M] is
included in S, for any M ∈ S.

We recall that [A, M] = {(1 − t)A + t B; t ∈ [0, 1]}.

5.6 Independence on the Path of Line Integral of the Second Kind 155

Remark 5.6.3 Any convex set is a star domain, but the reciprocal statement is not
generally true. For example, the set R2 \{(x, 0); x > 0} is a star domain (with
respect to the point O(0 , 0)), but it is not convex.

Theorem 5.6.4 If D ⊂ R3 is an open star domain, then any closed differential

1-form in D is exact in D.

Proof By hypothesis, there is A ∈ D such that [A, M] ⊂ D, ∀M ∈ D.

For simplicity, we suppose that A coincides with the origin O(0 , 0 , 0) and M
has the coordinates (x , y , z).
Let t ∈ [0, 1] be arbitrary and let us take T = (1 − t)O + t M = (t x, t y, t z) the
corresponding point on the segment [O, M]. We define the function f : D → R as
follows:

∫1
f (x, y, z) = (P(T ) · x + Q(T ) · y + R(T ) · z)dt
0
∫1
= (P(t x, t y, t z) · x + Q(t x, t y, t z) · y + R(t x, t y, t z) · z)dt.
0

Taking into account the differentiation theorem of the integral depending on a

parameter (Theorem 4.1.2), it results that:

∫1 (
∂f ∂P
= (t x, t y, t z) · t · x + P(t x, t y, t z)
∂x ∂x
0
)
∂Q ∂R
+ (t x, t y, t z) · t · y + (t x, t y, t z) · t · z dt.
∂x ∂x

On the other hand, by hypothesis, we have:

∂Q ∂P ∂R ∂P
= and = ,
∂x ∂y ∂x ∂z

Hence:

∫1 (
∂f ∂P ∂P
= (t x, t y, t z) · t · x + P(t x, t y, t z) + (t x, t y, t z) · t · y
∂x ∂x ∂y
0
) ∫1
∂P d
+ (t x, t y, t z) · t · z dt = (P(t x, t y, t z) · t)dt
∂z dt
0
= P(t x, t y, t z) · t|10 = P(x, y, z) · 1 − P(0, 0, 0) · 0 = P(x, y, z).
156 5 Line Integrals

Therefore ∂∂ xf = P, and analogously ∂∂ yf = Q, ∂∂zf = R; that is ω is exact.

From Theorems 5.6.1, 5.6.2, 5.6.3, 5.6.4, we can summarize the following result:

Remark 5.6.1 Let ω = P dx + Q dy + R dz be a differential 1-form of C 1 -class

on the open star domain D ⊂ R3 . Then the following assertions are equivalent:

(1) ω is exact in D (i.e.

( ∃ f ∈ C (D) such that d f = P dx
1
) + Q dy + R dz).
∂P ∂Q ∂P ∂R ∂Q ∂R
(2) ω is closed in D i.e., ∂ y = ∂ x , ∂ z = ∂ x , ∂z = ∂ y .
∫
(3) P dx + Q dy + R dz is independent on path in D.
γ
∮
(4) P dx + Q dy + R dz = 0, for any closed piecewise smooth curve γ , whose
γ
support is included in D.

Remark 5.6.4 Under the hypotheses from Remark 5.6.1, we have:

∫
(1) If the line integral P dx + Q dy + R dz is independent on the path in D, then
γ
we get the Leibniz–Newton formula for line integral:
∫
P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz
γ
∫
= P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz

AB
B(x∫
2 ,y2 ,z 2 )

= P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz

A(x1 ,y1 ,z 1 )

= f (B) − f (A) = f (x2 , y2 , z 2 ) − f (x1 , y1 , z 1 )

where f is a differentiable function on D such that d f = P dx + Q dy + R dz, and

the points A(x1 , y1 , z 1 ), B(x2 , y2 , z 2 ) ∈ D are the ends of the support of the curve
γ.
(2) The differentiable function f from remark (1) is easily determined using the
formula:

∫x ∫y ∫z
f (x, y, z) = P(x, y0 , z 0 )dx + Q(x, y, z 0 )dy + R(x, y, z)dz
x0 y0 z0

for any (x, y, z) ∈ D, where (x0 , y0 , z 0 ) ∈ D is an arbitrary fixed point and under
the hypothesis that the corresponding path is included in D.
5.6 Independence on the Path of Line Integral of the Second Kind 157

∫
(2,3,4)
Example 5.6.2 Compute yz dx + zx dy + x y dz.
(1,1,1)

The differential 1-form ω = yz dx + zx dy + x y dz is closed in R3 , because:

∂Q ∂P ∂R ∂Q ∂P ∂R
P = y z, Q = zx, R = x y and = = z, = = x, = = y.
∂x ∂y ∂y ∂z ∂z ∂x

From Remark 5.6.1, we deduce that this line integral is independent on the path
in R3 . Then, for the computation of the line integral we can choose the curve γ
as the polygonal line determined by the points A(1, 1, 1),B(2, 1, 1), C(2, 3, 1), and
D(2, 3, 4) (the polygonal line that joints the points) (1, 1, 1) and (2, 3, 4) by straight
line segments parallel to the coordinate axes.

∫
(2,3,4) ∫
yz dx + zx dy + x y dz = yz dx + zx dy + x y dz
(1,1,1) AB
∫
+ yz dx + zx dy + x y dz
BC
∫
+ yz dx + zx dy + x y dz
CD
∫
(2,1,1)

= yz dx + zx dy + x y dz
(1,1,1)

∫
(2,3,1)

+ yz dx + zx dy + x y dz
(2,1,1)

∫
(2,3,4)

+ yz dx + zx dy + x y dz
(2,3,1)

∫2 ∫3 ∫4
= dx + 2dy + 6dz = 1 + 4 + 18 = 23.
1 1 1

Another solution can be given if we notice that ω = d f , where f (x, y, z) = x yz.

Then:
|(2,3,4)
∫
(2,3,4) |
|
|
yz dx + zx dy + x y dz = x yz | = 24 − 1 = 23.
|
(1,1,1) |
(1,1,1)
158 5 Line Integrals

Remark 5.6.2 A differential 1-form ω = P dx + Q dy is closed in D ⊂ R2 if:

P, Q : D → R are functions of C 1 -class on D and ∂∂Py = ∂∂Qx .
∮ ( )
Example 5.6.3 Compute 4y(x − 1)dx + 2x 2 − 4x + y dy, where γ is the ellipse
γ
x2 y2
4
+ = 1.
25
We denote by P(x, y) = 4y(x − 1) and Q(x, y) = 2x 2 − 4x + y. Obviously.
P, Q : R2 → R are functions of C 1 -class on R2 and ∂∂Py = ∂∂Qx = 4(x − 1).
It follows that the differential 1-form
( )
ω = 4y(x − 1)dx + 2x 2 − 4x + y dy.

is closed in R2 , and hence it is exact in R2 . Since γ is a closed curve, from

Theorem 5.6.1 and Theorem 5.6.2, we deduce that
∮
( )
4y(x − 1)dx + 2x 2 − 4x + y dy = 0
γ

thus no computation is necessary.

Chapter 6
Double and Triple Integrals

In this chapter we will study the integration of the functions of two and three real
variables. The obtained results extend without difficulty to functions of n-variables.
In the case of functions of one variable, the integrated domain is a closed and
bounded interval, and the definition of the integral is made with the help of Riemann
or Darboux sums, in which the lengths of the subintervals corresponding to the
partition of the integrated interval appear.
For functions of two variables, the integrated domain is naturally to be a bounded
plane set. In order to be able to define the Riemann or Darboux sums in this case as
well, we must assume that the integrated domain is squareable (has area). In Chap. 2,
Sect. 2.7, we showed that a bounded plane set is squareable if and only if its boundary
is a set of area zero (Theorem 2.7.2).

6.1 Double Integral. Definition and Properties

We begin this paragraph by presenting a particular case of a squareable bounded

plane set, a very common case in applications.
Theorem 6.1.1 The support of a piecewise smooth curve is a set of area zero.
Proof Since a finite union of area zero sets is also an area zero set, it is sufficient to
show that the support of a smooth curve γ is a set of area zero.
Let r : [a, b] → R2 be the representative parameterized path of smooth curve γ ,
r (t) = (x(t), y(t)), t ∈ [a, b]. We denote by L the length of this path (Theorem 5.2.1)
and let x = x̃(s), y = ỹ(s) be its natural parameterization.
Let us take Δ : 0 = s0 < s1 < · · · < si−1 < si < · · · < sn = L an equidistant
partition of the interval [0, L] and let Mi (x̃(si ), ỹ(si )), i = 0, n be points on the
͡
curve support (Fig. 6.1). The length of the arc Mi−1 Mi is Ln . Consider a square Di
centered at Mi and with the sides parallel to the coordinate axes, having the length
2 · Ln .
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 159
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_6
160 6 Double and Triple Integrals

Fig. 6.1 The support of a L

smooth curve
n

Mi

M i −1 Di

∪n
It is obvious that the curve support is included in i=0 Di and that:
( n )
∪ Σ
n
4 L2
area Di ≤ area(Di ) = (n + 1).
i=0 i=0
n2

2 ∪n
Since lim 4nL2 (n + 1) = 0, for n sufficiently large, the area of the set i=0 Di
n→∞
is however small, hence the support of the curve γ is a set of area zero.

Corollary 6.1.1 Any bounded plane set whose boundary is piecewise smooth is
squareable.

Proof The statement follows from Theorem 6.1.1 and Theorem 2.7.2.
Let A ⊂ R2 be a bounded subset. Then there is a circle (disk) that contains the set
A. It follows that the distance between any two points of the set A is less than the diam-
eter of this circle. Therefore, the set of positive real numbers {dist(M, N ); M, N ∈ A}
is upper bounded, hence it has a least upper bound (Fig. 6.2).

Definition 6.1.1 Let A ⊂ R2 be a bounded subset. The following number is called

the diameter of the set A:
not
d( A) = diam(A) = sup{dist(M, N ); M, N ∈ A}.

In the following we will denote by D a squareable bounded set of R2 .

Definition 6.1.2 It is called a partition of D any finite family of subsets {Di },

◦ ◦ ∪p
Di ⊂ D, i = 1, p, with the properties: Di ∩ D j = ∅, for i /= j and D = i=1 Di .
If we denote by ρ the partition D1 , D2 , . . . , D p of D, then the norm of this
partition is defined as follows:

Fig. 6.2 The diamenter of a

A
set
d ( A)
6.1 Double Integral. Definition and Properties 161

Fig. 6.3 The partition of the

domain D D1

D2
Dp
Di

∥ρ∥ = max{diam(Di ); 1 ≤ i ≤ p}.

From Proposition 2.7.3, we deduce (Fig. 6.3):

Σ
p
area(D) = area(Di ).
i=1

Definition 6.1.3 The partition ρ ' of the domain D is said to be a refinement of the
partition ρ of D, and denoted by ρ ' ≻ ρ, whether each subdomain of the partition
ρ is a finite union of subdomains of the partition ρ ' which have no common interior
points. ( )
Therefore, if ρ is the partition (Di )1≤i≤ p , then ρ ' has the form Di' j 1≤i≤ p , where
∪i ∥ ∥ 1≤ j≤ni
Di = nj=1 Di' j , ∀ 1 ≤ i ≤ p. Obviously, if ρ ' ≻ ρ, then ∥ρ∥ ≥ ∥ρ ' ∥.
Let ρ : D1 , D2 , . . . , D p be a partition of the domain D ⊂ R2 and let f : D → R
be a bounded function on D. We denote by:

m = inf{ f (x, y); (x, y) ∈ D}, M = sup{ f (x, y); (x, y) ∈ D}

m i = inf{ f (x, y); (x, y) ∈ Di },
Mi = sup{ f (x, y); (x, y) ∈ Di }, 1 ≤ i ≤ p.

Definition 6.1.4 The lower Darboux sum, respectively the upper Darboux sum
associated to the bounded function f and the partition ρ are defined as follows:

Σ
p
Σ
p
sρ = m i · area(Di ) and Sρ = Mi · area(Di ).
i=1 i=1

Σp
Since m ≤ m i ≤ Mi ≤ M, ∀1 ≤ i ≤ p and area(D) = i=1 area(Di ), it
follows:

m · area(D) ≤ sρ ≤ Sρ ≤ M · area(D). (6.1)

162 6 Double and Triple Integrals

Lemma 6.1.1 If ρ ≺ ρ ' , then sρ ≤ sρ ' ≤ Sρ ' ≤ Sρ .

Proof We assume that the partition ρ consists of domains (Di )1≤i≤ p and the partition
( )
ρ ' consists of domains Di' j 1≤i≤ p .
1≤ j≤n i ∪i
Since ρ ≺ ρ ' , it results that for any 1 ≤ i ≤ p, we have Di = nj=1 Di' j . If
{ }
we denote by m i' j = inf f (x, y); (x, y) ∈ Di' j , then m i ≤ m i' j , ∀ i = 1, p, ∀ j =
1, n i .
Further, we have:
⎛ ⎞
Σ
p
Σ
p
Σni
( ')
sρ = m i · area(Di ) = m i ·⎝ area Di j ⎠
i=1 i=1 j=1

Σ
p
Σ
ni
( )
≤ m i' j · area Di' j = sρ ' .
i=1 j=1

Therefore, we proved that sρ ≤ sρ ' . Similarly, it is shown that Sρ ' ≤ Sρ .

Lemma 6.1.2 For any two partitions ρ ' and ρ '' of domain D, we have sρ ' ≤ Sρ '' .
( )
Proof Let us suppose that the partition ρ ' consists of the domains Di' 1≤i≤ p and
( )
the partition ρ '' consists of the domains D ''j . If we denote by ρ the partition
( ) 1≤ j≤q
consisting of the domains Di' ∩ D ''j 1≤i≤ p , then ρ is finer as ρ ' and ρ '' . According
1≤ j≤q
to Lemma 6.1.1, we have:

sρ ' ≤ sρ ≤ Sρ ≤ Sρ '' .

Next, let us denote by P the family of all partitions of the domain D and we define
the lower (upper) Darboux double integral as follows:
{ } { }
I∗ = sup sρ ; ρ ∈ P and I ∗ = inf Sρ ; ρ ∈ P .

The existence of these real numbers, I∗ and I ∗ , results from the inequalities (6.1).
On the other hand, from Lemma 6.1.2, we deduce that:

I∗ ≤ I ∗ .

Definition 6.1.5 Let f : D → R be a bounded function on the squareable bounded

domain D ⊂ R2 . We say that the function f is integrable in the Darboux sense on
D if:

I∗ = I ∗ = I.
6.1 Double Integral. Definition and Properties 163
˜
The common value I is denoted by f (x, y)dx dy and it is called the double
D
integral of the function f on the domain D; the real number I is uniquely determined
by f and D.
Example 6.1.1 Let D ⊂ R2 be an arbitrary squareable bounded subset. Prove that
any constant function f : D → R, f (x, y) = k, ∀(x, y) ∈ D is integrable on D and:
¨
1 dx dy = area(D).
D

Indeed, for any partition ρ : D1Σ

, . . . , D p of D we have m i = Mi = k,∀ i = 1, p,
p
whence, it results that sρ = Sρ = i=1 k · area(Di ) = k · area(D), and further that:
¨ ¨
∗
I∗ = I = k · area(D) = k dx dy, hence 1 dx dy = area(D).
D D

Theorem 6.1.2 (Darboux criterion for integrability) Let D ⊂ R2 be a squareable

bounded subset and let f : D → R be a bounded function. The necessary and suffi-
cient condition that f to be integrable on D is that for every ε > 0, there is δε > 0
with the property that for any partition ρ of D, with ∥ρ∥ < δε the following inequality
is verified

Sρ − sρ < ε.

The proof is similar to the proof presented in the one-dimensional case

(Theorem 2.2.1).
Theorem 6.1.3 If D ⊂ R2 is a compact squareable subset, then any continuous
function f : D → R is integrable on D.
The proof is similar to the proof presented in the one-dimensional case
(Theorem 2.3.1).
The following theorem can also be proved:
Theorem 6.1.4 (Lebesgue criterion for integrability) Let D ⊂ R2 be a compact
squareable subset. The necessary and sufficient condition that the function f : D →
R to be integrable on D is that f to be bounded and the set of all its points of
discontinuity to be a null set.
Next, we define the Riemann sums.
Definition 6.1.6 Let f : D → R be a bounded function, let ρ : D1 , D2 , . . . , D p be
an arbitrary partition of D and let (ξ, η) = (ξi , ηi )1≤i≤ p be such that (ξi , ηi ) ∈ Di ,
∀ i = 1, p, an arbitrary system of intermediate points associated to the partition ρ.
The Riemann sum of the function f , associated to the partition ρ and the
intermediate points (ξi , ηi ) is defined as follows:
164 6 Double and Triple Integrals

Σ
p
σρ ( f ; ξ, η) = f (ξi , ηi ) · area(Di ).
i=1

Since m i ≤ f (ξi , ηi ) ≤ Mi , ∀ i = 1, p, it results that:

sρ ≤ σρ ( f ; ξ, η) ≤ Sρ

for any intermediate point system (ξ, η).

Definition 6.1.7 Let f : D → R be a bounded function. We say that the function

f is integrable in the Riemann sense on D if there is a finite real number I with
the property that for any ε > 0, there is δε > 0 such that for any partition ρ of D,
with ∥ρ∥ < δε and any intermediate points (ξi , ηi ) ∈ Di , the following inequality is
verified:
| |
|σρ ( f ; ξ, η) − I | < ε.

The real number I is called the double integral of f on D and it is denoted by:
¨
I = f (x, y)dx dy.
D

Remark 6.1.1 For any ε > 0, there are (αi , βi ) ∈ Di and (λi , μi ) ∈ Di such that:

Sρ − σρ ( f ; α, β) < ε and σρ ( f ; λ, μ) − sρ < ε

where (α , β) = (αi , βi )1 ≤ i ≤ p , (λ , μ) = (λi , μi )1 ≤ i ≤ p .

Indeed, from the definition of the least upper bound, it follows that for any ε > 0,
ε
there is (αi , βi ) ∈ Di such that Mi − f (αi , βi ) < area(D) . Further, we have:

Σ
p
Sρ − σρ ( f ; α, β) = (Mi − f (αi , βi ))area(Di )
i=1
ε
< · area(D) = ε
area(D)

The proof of the other inequality is similar.

Using this remark and proceeding in the same manner as for the proof of
Theorem 2.4.1, one can show that the two definitions of the double integral, with
Riemann sums and Darboux sums, coincide. It can also be shown, as in the case of
the definite integral (Theorem 2.4.2), that the following criterion for integrability is
valid.
6.2 Basic Properties of the Double Integral 165

Theorem 6.1.5 (Riemann criterion for integrability) The necessary and sufficiency
condition that the bounded function f : D → R to be integrable on D is the existence
of a finite real number I with the property that for any sequence of partitions
( {ρn)}
of D such that lim ∥ρn ∥ = 0 and for any choice of intermediate points ξ (n) , η(n) ,
n→∞
we have
( )
lim σρn f ; ξ (n) , η(n) = I.
n→∞

Remark 6.1.2 From Theorem 6.1.5 and Remark 6.1.1, it follows that if f is integrable
on D, then for any sequence of partitions {ρn } of D, with the property lim ∥ρn ∥ = 0,
n→∞
it results that:
¨
lim sρn = lim Sρn = f (x, y)dx dy.
n→∞ n→∞
D

6.2 Basic Properties of the Double Integral

The properties of the double integral are analogous to the properties of the definite
integral, and therefore we will only enumerate these properties, leaving the proofs
to the reader.
˜
Proposition 6.2.1 1 dx dy = area(D), where D is a bounded squareable subset.
D

Proposition 6.2.2 (Linearity property) If f, g: D → R are integrable on D and

α, β ∈ R, then the function α f + βg is integrable on D and:
¨ ¨
(α f (x, y) + βg(x, y))dx dy = α f (x, y)dx dy
D D
¨
+β g(x, y)dx dy.
D

Proposition 6.2.3 (Monotony property) If f, g: D → R are integrable on D and

f (x, y) ≤ g(x, y), ∀(x, y) ∈ D, then:
¨ ¨
f (x, y)dx dy ≤ g(x, y)dx dy.
D D

In particular, if f : D → R+ is integrable and nonnegative on D, then:

166 6 Double and Triple Integrals
¨
f (x, y)dx dy ≥ 0.
D

Proposition 6.2.5 If f : D → R is integrable on D, then | f | is integrable on D and:

| |
|¨ | ¨
| |
| f (x, y)dx dy || ≤ | f (x, y)|dx dy.
|
| |
D D

Proposition 6.2.6 (Additivity property) If D = D1 ∪ D2 and D1 , D2 are two

squareable bounded subsets, without common interior points and f is integrable on
D1 and D2 , then f is integrable on D and we have:
¨ ¨ ¨
f (x, y)dx dy = f (x, y)dx dy + f (x, y)dx dy.
D D1 D2

Theorem 6.2.1 (The mean value theorem) Let f : D → R be a bounded integrable

function, m = inf{ f (x, y); (x, y) ∈ D} and M = sup{ f (x, y); (x, y) ∈ D}. Then
there is m ≤ μ ≤ M such that:
¨
f (x, y)dx dy = μ · area(D).
D

Moreover, if we assume that f is continuous on the compact and connected subset

D, then there is a point (ξ, η) ∈ D such that:
¨
f (x, y)dx dy = f (ξ, η) · area(D).
D

6.3 Reducing a Double Integral to an Iterated Single

Integral

Definition 6.3.1 A domain D ⊂ R2 is called simple with respect to O y axis if there

are two continuous functions ϕ, ψ : [a, b] → R such that ϕ(x) < ψ(x), ∀a < x <
b and:
{ }
D = (x, y) ∈ R2 ; a ≤ x ≤ b, ϕ(x) ≤ y ≤ ψ(x) .

Similarly, a domain D ⊂ R2 is called simple with respect to O x axis; if there are

two continuous functions u, v : [c, d] → R with the property u(y) < v(y), ∀c <
6.3 Reducing a Double Integral to an Iterated Single Integral 167

Fig. 6.4 A simple domain y ψ

with respect to Oy

ϕ
O a b x

y < d such that:

{ }
D = (x, y) ∈ R2 ; c ≤ y ≤ d, u(y) ≤ x ≤ v(y) .

Remark 6.3.1 In the case of a simple domain with respect to O y axis, any straight
line parallel to the axis O y cuts the boundary of the domain at the most two points
(Fig. 6.4).

Remark 6.3.2 In the case of a simple domain with respect to O x axis, any straight
line parallel to the axis O x cuts the boundary of the domain at the most two points
(Fig. 6.5).
There are domains that are simple with respect to both axes, for example, rect-
angles, circles, and so on; there are also domains that are not simple with respect to
any axis, for example, circular crowns.

Lemma 6.3.1 Let D ⊂ R2 be a domain simple with respect to O y axis and let
f : D → R be a continuous function. If we denote by m = inf{ f (x, y); (x, y) ∈ D}
and M = sup{ f (x, y); (x, y) ∈ D}, then:
⎛ ⎞
∫b ∫
ψ(x)
⎜ ⎟
m · area(D) ≤ ⎝ f (x, y)dy ⎠dx ≤ M · area(D).
a ϕ(x)

Fig. 6.5 A simple domain y

with respect to Ox
d

u D v

O x
168 6 Double and Triple Integrals

Proof First, we notice that from the continuity theorem of the integral depending on
∫ ψ(x)
a parameter (Theorem 4.1.1), it results that the function F(x) = ϕ(x) f (x, y)dy,
x ∈ [a, b] is continuous on [a, b], hence integrable on [a, b].
By hypothesis, we have:

m ≤ f (x, y) ≤ M, ∀(x, y) ∈ D.

From the monotony property of the definite integral (Proposition 2.6.3), it results
that:

∫
ψ(x) ∫
ψ(x) ∫
ψ(x)

m dy ≤ f (x, y)dy ≤ M dy, ∀ x ∈ [a, b]

ϕ(x) ϕ(x) ϕ(x)

or:

∫
ψ(x)

m · (ψ(x) − ϕ(x)) ≤ f (x, y)dy ≤ M · (ψ(x) − ϕ(x)), ∀ x ∈ [a, b].

ϕ(x)

Using again the monotony property of the definite integral, we obtain:

⎛ ⎞
∫b ∫b ∫
ψ(x) ∫b
⎜ ⎟
m (ψ(x) − ϕ(x))dx ≤ ⎝ f (x, y)dy ⎠dx ≤ M (ψ(x) − ϕ(x))dx.
a a ϕ(x) a

∫b
It remains to be seen that a (ψ(x) − ϕ(x))dx = area(D) (Corollary 2.7.1) and
thus lemma is proved.

Theorem 6.3.1 Let D ⊂ R2 be a simple domain with respect to O y axis and let
f : D → R be a continuous function. Then:
⎛ ⎞
¨ ∫b ∫
ψ(x)
⎜ ⎟
f (x, y)dx dy = ⎝ f (x, y)dy ⎠dx. (6.2)
D a ϕ(x)

Proof Let us take Δ : a = x0 < x1 < · · · < xi−1 < xi < · · · < xn = b an
equidistant partition of the interval [a, b]. Then xi = a + i · b−a
n
, i = 0, n and
∥Δ∥ = n . b−a

We consider the functions

ϕ j : [a, b] → R, j = 0, n, defined as:

j
ϕ j (x) = ϕ(x) + (ψ(x) − ϕ(x)).
n
6.3 Reducing a Double Integral to an Iterated Single Integral 169

Fig. 6.6 A partition of the y ϕn =ψ

simple domain D
ϕj
Di j
ϕ j −1

ϕ0 = ϕ x
O a = x0 x i −1 xi xn = b

Obviously, we have ϕ0 = ϕ and ϕn = ψ. ( )

We denote by ρn the partition of D consisting by the sets Di j 1≤i≤n , where:
{ } 1≤ j≤n
Di j = (x, y) ∈ R2 ; xi−1 ≤[ x ≤ xi ,]ϕ j−1 (x) ≤ y ≤ ϕ j (x) (Fig. 6.6).
We notice that for any x ∈ xi−1 , xi , we have:

( ) b−a 1
diam Di j ≤ + ∥ψ − ϕ∥∞ ,
n n
whence, it results that lim ∥ρn ∥ = 0. Let us denote by m i j and by Mi j the greatest
n→∞
lower bound, respectively the least upper bound of the function f on Di j . From
Lemma 6.3.1, it follows that:
⎛ ⎞
∫xi ϕ∫j (x)
( ) ⎜ ⎟ ( )
m i j · area Di j ≤ ⎝ f (x, y)dy ⎠dx ≤ Mi j · area Di j , i, j = 1, n.
xi−1 ϕ j−1 (x)

Making the sum, successively following i and j we obtain:

⎛ ⎛ ⎞⎞
ϕ∫j (x)
Σ
n Σ n ∫xi
Σ Σ
n
( ) ⎜
n
⎜ ⎟⎟
m i j · area Di j ≤ ⎝ ⎝ f (x, y)dy ⎠⎠dx
i=1 j=1 i=1 x j=1 ϕ j−1 (x)
i−1

Σ
n Σ
n
( )
≤ Mi j · area Di j .
i=1 j=1

Since
ϕ∫j (x) ∫
ψ(x)
Σ
n
f (x, y)dy = f (x, y)dy
j=1ϕ ϕ(x)
j−1 (x)
170 6 Double and Triple Integrals

and
⎛ ⎛ ⎞⎞ ⎛ ⎞
ϕ∫j (x)
n ∫xi
Σ Σ
n ∫b ψ(x)
∫
⎜ ⎜ ⎟⎟ ⎜ ⎟
⎝ ⎝ f (x, y)dy ⎠⎠dx = ⎝ f (x, y)dy ⎠dx
i=1 x j=1 ϕ j−1 (x) a ϕ(x)
i−1

it results that:
⎛ ⎞
∫b ∫
ψ(x)
⎜ ⎟
sρn ≤ ⎝ f (x, y)dy ⎠dx ≤ Sρn . (6.3)
a ϕ(x)

Since f is integrable on D, from Remark 6.1.2, it follows that:

¨
lim sρn = lim Sρn = f (x, y)dx dy.
n→∞ n→∞
D

Passing to the limit on n in the inequalities (6.3), we obtain:

⎛ ⎞
¨ ∫b ∫
ψ(x)
⎜ ⎟
f (x, y)dx dy = ⎝ f (x, y)dy ⎠dx.
D a ϕ(x)

Remark 6.3.3 If the domain D ⊂ R2 is simple with respect to O x axis and the
function f : D → R is continuous, then we have the following formula for computing
the double integral:
⎛ ⎞
¨ ∫d ∫v(y)
⎜ ⎟
f (x, y)dx dy = ⎝ f (x, y)dx ⎠dy. (6.4)
D c u(y)

{ }
Remark 6.3.4 If the domain D = (x, y) ∈ R2 ; a ≤ x ≤ b, c ≤ y ≤ d ⊂ R2 is a
rectangle and f : D → R is continuous, then the double integral is calculated using
one of the iterated formulas:
⎛ ⎞ ⎛ ⎞
¨ ∫b ∫d ∫d ∫b
f (x, y)dx dy = ⎝ f (x, y)dy ⎠dx = ⎝ f (x, y)dx ⎠dy. (6.5)
D a c c a
6.3 Reducing a Double Integral to an Iterated Single Integral 171

Fig. 6.7 A bounded y

rectangle
1

−1 O 1 x

Remark 6.3.5 If f (x, y) = g(x) · h(y) and the functions g: [a, b] → R and h :
[c, d] → R are continuous (or integrable), then f is integrable on the rectangle
D = [a, b] × [c, d] and we have the equality:

¨ ∫b ∫d
f (x, y)dx dy = g(x)dx · h(y)dy.
D a c

˜
Example 6.3.1 Compute x 3 y dx dy, where D is the rectangle bounded by the
D
straight lines: x = −1, x = 1, y = 0, y = 1 (Fig. 6.7).
From Remark 6.3.5, it results that:

¨ ∫1 ∫1
x y dx dy =
3
x dx ·
3
y dy
D −1 0
( | ) ( 2| )
x 4 || 1 y ||1
= | · = 0.
4 −1 2 |0

We obtain the same result if Remark 6.3.4 is applied:

⎛ ⎞
¨ ∫1 ∫1 ∫1 ( | ) ∫1 (
⎝ x 4 y || 1 y y)
x 3 y dx dy = x 3 y dx ⎠dy = | dy = − dy = 0.
4 −1 4 4
D 0 −1 0 0

˜
Example 6.3.2 Compute D x 2 y 2 dx dy, where D is the domain bounded by the
curves: y = x 2 , y = 1.
We notice
{ that D is a domain simple with respect
} to O y axis:
D = (x, y) ∈ R2 ; −1 ≤ x ≤ 1, x 2 ≤ y ≤ 1 (Fig. 6.8).
From Theorem 6.3.1 (formula 6.2), it results that:
172 6 Double and Triple Integrals

Fig. 6.8 A simple domain y

bounded by a parabola and a
y = x2
straight line

1 y =1

−1 O 1 x

⎛ ⎞
¨ ∫1 ∫1
x 2 y 2 dx dy = ⎝ x 2 y 2 dy ⎠dx
D −1 x2
∫1 ( | ) ∫1 ( 2 )
x 2 y 3 || 1 x x8
= dx = − dx
3 |x 2 3 3
−1 −1
( 3 )|
x x 9 || 1 4
= − = .
9 27 |−1 27

On the other hand, it is easy to see that D is simple also with respect to O x axis.
Indeed:
{ √ √ }
D = (x, y) ∈ R2 ; 0 ≤ y ≤ 1, − y ≤ x ≤ y .

Therefore, according to formula (6.4) we have:

⎛ √ ⎞
¨ ∫1 ∫ y ∫1 ( | )
⎜ ⎟ x 3 y 2 || √ y
x y dx dy =
2 2
⎝ x y dx ⎠dy =
2 2
dy
√
3 |−√ y
D 0 − y 0

∫1 ( √ √ ) √ |
y3 y y3 y 4 y 4 y || 4
= + dy = 1 = .
3 3 27 |0 27
0

Remark 6.3.6 If the integrated domain is not simple with respect to any of the
coordinate axes, then we decompose the domain, by straight lines parallel to the
coordinate axes, into a finite number of subdomains
∪n D1 , D2 , . . . , Dn ⊂ D, simple
with respect to the same axis, such that D = i=1 Di . Using the additivity property
of the double integral (Proposition 6.2.6), we have:
6.4 Change of Variables in Double Integral 173

¨ n ¨
Σ
f (x, y)dx dy = f (x, y)dx dy (6.6)
D i=1 D
i

such that for all double integrals of the right member of equality (6.6) one of the
computation formulas (6.2) or (6.4) applies.

6.4 Change of Variables in Double Integral

We have seen what an efficient tool is the method of changing the variable for the
computation of single integrals. By choosing the right variable change, we can replace
the expression under the sign of the integral with another expression that integrates
easily.
We recall the most used formula for changing the variable in the single integral.
Let u : [c, d] → [a, b] be a surjective function of C 1 —class, such that u ' (t) /= 0,
∀ t ∈ (c, d). In these conditions, we have either u ' (t) > 0, ∀ t ∈ (c, d), either
u ' (t) < 0, ∀ t ∈ (c, d), hence u is injective on [c, d]. Since u is also surjective, we
deduce that u is bijective.
If f : [a, b] → R is a continuous function, then the following formula holds:
−1
∫b u∫ (b)

f (x)dx = f (u(t)) · u ' (t) dt.

a u −1 (a)

If u it is strictly increasing, then u −1 (a) = c, u −1 (b) = d, hence:

∫b ∫d
f (x)dx = f (u(t)) · u ' (t)dt
a c

If u it is strictly decreasing, then u −1 (a) = d, u −1 (b) = c, hence:

∫b ∫c
f (x)dx = f (u(t)) · u ' (t) dt.
a d

In both situations, we have:

∫b ∫d
| |
f (x)dx = f (u(t)) · |u ' (t)|dt. (6.7)
a c
174 6 Double and Triple Integrals

∫π
Example 6.4.1 Compute I = π2 sin1 x dx.
3
If we make the change of variable t = tan x2 , it results that:

2
x = u(t) = 2 tan−1 t; u ' (t) = > 0;
1 + t2
π π 1
a= , b = , c = √ , d = 1.
3 2 3
2 tan x2
Since sin x = 1+tan2 x2
= 2t
1+t 2
, according to formula (6.7), we have:

∫1 ∫1 |
1 + t2 | √
2 1 | 1
I = · dt = dt = ln|t| | 1 = ln 3 = ln 3.
2t 1 + t2 t | √13 2
√1 √1
3 3

Next, we present a change of variable formula for the double integral analogous
to formula (6.7).
Let Ω ⊂ R2 be an open, bounded, squareable subset. To keep the analogy with
the one-dimensional case (it is known that intervals are the only connected sets in
R), we assume in addition that set Ω is connected.
Let F: Ω → R2 , F(u, v) = (x(u, v), y(u, v)), ∀(u, v) ∈ Ω be a vector function
with the properties:

(i) F is of C 1 -class on Ω
(ii) F: Ω → D is bijective
(iii) F is a regular transformation on Ω, this means that:
| ∂x |
D(x, y) | ∂x |
det JF (u, v) = (u, v) = || ∂u
∂y
∂v
∂y
|(u, v) /= 0, ∀(u, v) ∈ Ω
|
D(u, v) ∂u ∂v

and let D = F(Ω) ⊂ R2 .

Such a vector function is also called a change of coordinates or a change of
variables. Since a regular transformation transforms open and connected sets into
open and connected sets, it results that D ⊂ R2 is also an open and connected set.

Proposition 6.4.1 By a change of variables, any point on the boundary of D

corresponds to a point on the boundary of Ω, and coverse. In other words:

F(∂Ω) = ∂ D.

Proof Indeed, since Ω and D are open sets and F: Ω → R2 is continuous we have:
( )
Ω = Ω ∪ ∂Ω, D = D ∪ ∂ D, D = F(Ω) ⊂ F Ω ⊂ F(Ω) = D.
6.4 Change of Variables in Double Integral 175
( )
Using the continuity of F and the fact that Ω is compact, we deduce that F Ω
( ) ( )
is compact, and therefore D ⊂ F Ω , i.e. D = F Ω or

D ∪ ∂ D = F(Ω ∪ ∂Ω) = F(Ω) ∪ F(∂Ω) = D ∪ F(∂ D).

Since D ∩ ∂ D = ∅, we deduce from above that ∂ D ⊂ F(∂Ω).

We prove now that F(∂Ω) ⊂ ∂ D. In( the ) contrary case, we choose z 0 ∈ ∂Ω such
/ ∂ D. Since F(∂Ω) ⊂ F Ω = D = D ∪ ∂ D, we get F(z 0 ) ∈ D.
that F(z 0 ) ∈
Since F is an homeomorphism between Ω and D, there is z 1 ∈ Ω such that F(z 1 ) =
= F(z 0 ) ∈ D.
We consider now r > 0 such that B(z 1 , r ) ⊂ Ω and B(z 1 , r ) ∩ B(z 0 , r ) = ∅.
From the above considerations F(B(z 1 , r ))( is(an open)) subset of D containing F(z 0 ).
If 0 < r ' < r is sufficiently small, then F B z 0 , r ' ⊂ F(B(z 1 , r )), i.e. we arrive
to the contradictory relations:
( ) ( )
B z 0 , r ' ∩ Ω /= ∅, B(z 1 , r ) ∩ B z 0 , r ' ∩ Ω = ∅,
( ( ) )
F(B(z 1 , r )) ∩ B z 0 , r ' ∩ Ω /= ∅.

Remark 6.4.1 The statement of Proposition 6.4.1 remains valid even when F is a
continuous map from Ω to R2 or F is an homeomorphism between Ω and D. These
conditions are verified in all further situations.
Taking into account that the Jacobian (u, v) → D(x,y)
D(u,v) (u,
v): Ω → R is contin-
uous on the connected subset Ω, it results that it keeps constant sign on Ω. Therefore,
either
D(x, y)
(u, v) > 0, ∀(u, v) ∈ Ω
D(u, v)

either
D(x, y)
(u, v) < 0, ∀(u, v) ∈ Ω.
D(u, v)

We notice that we find the situation from the one-dimensional case, with the
Jacobian D(x,y)
D(u,v)
instead of the derivative u ' .

The following formula can be proved, analogous to formula (6.7):

Theorem 6.4.1 (The change of variables formula in double integral) Let F: Ω →

D, F(u, v) = (x(u, v), y(u, v)), ∀(u, v) ∈ Ω be a change of variables and let
f : D → R be a continuous function. Then the domain D = F(Ω) is squareable and
the following formula holds:
¨ ¨ | |
| D(x, y) |
f (x, y)dx dy = |
f (x(u, v), y(u, v))| (u, v)||du dv. (6.8)
D(u, v)
D Ω
176 6 Double and Triple Integrals

The formula (6.8), as well as formula (6.7), is extremely used in practice and
transforming the initial integral into an easier to compute integral. In the case of
formula (6.8) a new element intervenes, which is missing in the case of the simple
integral: formula (6.8) aims at a simplification not only of the function f , but also
the integrated domain.

The choice of changing variables makes the domain Ω to have a simpler shape
than the domain D (to be, e.g., a rectangle). Then the computation of the initial
integral is considerably simplified.
Case 1. The most used change of variables is the transition to the polar coordinates:

{
x = ρ cos θ
, 0 < ρ < ∞, 0 < θ < 2π. (6.9)
y = ρ sin θ

If we denote by:

A = {(ρ, θ ); 0 < ρ < ∞, 0 < θ < 2π }, and B = R2 \{(x, 0); x ≥ 0}

F(ρ, θ ) = (ρ cos θ, ρ sin θ ),

then F: A → B is a regular transformation, because, its Jacobian is:

| |
D(x, y) || cos θ −ρ sin θ ||
det JF (ρ, θ ) = =| = ρ > 0, ∀(ρ, θ ) ∈ A.
D(ρ, θ ) sin θ ρ cos θ |

Let 0 < α < β < 2π and let ϕ : [α, β] → R be a continuous function. We

denote also by:

Ω = {(ρ, θ ); α < θ < β, 0 < ρ < ϕ(θ )} and D = F(Ω).

Then F: Ω → D is a change of variables. If f : D → R is continuous, then

according to formula (6.8) we have:
¨ ¨
f (x, y)dx dy = f (ρ cos θ, ρ sin θ ) · ρdρ dθ
D Ω
⎛ ϕ(θ ) ⎞
∫β ∫
= ⎝ f (ρ cos θ, ρ sin θ ) · ρ dρ ⎠dθ .
α 0
6.4 Change of Variables in Double Integral 177
˜ √
Example 6.4.2 Compute x 2 + y 2 dx dy if:
D

{ }
x √
D = (x, y); x + y < a , √ < y < x 3, x > 0 .
2 2 2
3
( )
In this case, Ω = F −1 (D) is the rectangle π6 , π3 × (0, a).
Indeed, taking into account the formulas (6.9) in the inequalities that define the
domain D, we obtain:
{ }
cos θ √
Ω = (ρ, θ ); ρ < a , √ < sin θ < 3 cos θ
2 2
3
{ } (
1 √ π π)
= (ρ, θ ); 0 < ρ < a, √ < tanθ < 3 = , × (0, a).
3 6 3

Therefore we have (Figs. 6.9 and 6.10):

¨ √ ¨ √
x 2 + y 2 dx dy = ρ 2 cos2 θ + ρ 2 sin2 θ · ρ dρ dθ
D Ω
π ⎛ a ⎞ π
∫3 ∫ ∫3 3
⎝ ρ 2 dρ ⎠dθ = a π · a3
= dθ = .
3 18
π 0 π
6 6

˜
Example 6.4.3 Compute x y dx dy, where:
D

{ }
D = (x, y); x 2 + y 2 < 2ax, y > 0, a > 0 .

We notice that the equation x 2 + y 2 − 2 a x = 0 is the equation of the circle of

radius a centered at the point (a, 0) (Fig. 6.11). Substituting (6.9) in the inequalities
of D, we obtain (Fig. 6.12):

Fig. 6.9 A circular sector of y

a circle centered in the origin π
and of radius a 3
D π
6
O a x
178 6 Double and Triple Integrals

Fig. 6.10 The variation ρ

domain of the polar
coordinates θ and ρ
a

O π π θ
6 3

{ }
Ω = (ρ, θ ); ρ 2 < 2 a ρ cos θ, ρ sin θ > 0
{ π}
= (ρ, θ ); 0 < ρ < 2 a cos θ, 0 < θ < .
2
Further, we have:
¨ ¨
x y dx dy = (ρ cos θ · ρ sin θ ) · ρ dρ dθ
D Ω
π ⎛ 2 a cos θ ⎞ π
∫2 ∫ ∫2
= ⎝ ρ 3 cos θ sin θ dρ ⎠dθ = 4 a 4 cos5 θ sin θ dθ
0 0 0
π
∫2 |
' cos6 θ || π 2 a4
= −4 a 4
cos θ · (cos θ ) dθ = −4 a ·
5 4 2 = .
6 |0 3
0

Fig. 6.11 The superior y

semicircle centered in (a, 0)
and of radius a
D

O a 2a x

Fig. 6.12 The variation ρ

domain of the polar 2a ρ ( θ ) = 2 a cos θ
coordinates θ and ρ
Ω
O π θ
2
6.4 Change of Variables in Double Integral 179

Fig. 6.13 The ellipse y

centered in the origin and of b
semi-axis a and b
D
O a x

Case 2. Another change of variables is the transition to generalized polar coordinates

and it is used if the integrated domain is bounded by an ellipse. If the equation of
2 2
the ellipse is ax 2 + by2 = 1, then the generalized polar coordinate is defined by the
relations:
{
x = a ρ cos θ
, 0 < ρ < 1, 0 < θ < 2π. (6.10)
y = b ρ sin θ

The Jacobian is:

| |
D(x, y) || a cos θ −a ρ sin θ ||
=| = a b ρ > 0.
D(ρ, θ ) b sin θ b ρ cos θ |
˜
Example 6.4.4 Compute (x + y − 2)dx dy, where (Fig. 6.13):
D

{ }
x2 y2
D = (x, y) ∈ R2 ; 2 + 2 ≤ 1
a b

In this case, the generalized polar coordinates (6.10) are used and it is obtained
(Fig. 6.14):
⎛ ⎞
¨ ∫2 π ∫1
(x + y − 2)dx dy = ⎝ (a ρ cos θ + b ρ sin θ − 2) · a b ρ dρ ⎠dθ
D 0 0
∫2π | ∫2π |
ρ 3 ||1 ρ 3 ||1
=a b 2
cos θ | dθ + a b 2
sin θ dθ
3 0 3 |0
0 0
∫2π |
2|
ρ |1
− 2a b dθ = −2π a b.
2 |0
0
180 6 Double and Triple Integrals

Fig. 6.14 The variation ρ

domain of the polar
coordinates θ and ρ
1
Ω

O 2π θ

6.5 Applications of the Double Integral in Geometry

and Mechanics

6.5.1 Mass of a Lamina

By thin plate or lamina we mean a material flat object of negligible thickness in

the shape of a bounded squareable plane domain D ⊂ R2 . The lamina is gener-
ally considered non-homogeneous, its density being given by a continuous positive
function f : D → R+ .
Let ρ : D1 , D2 , . . . , D p be an arbitrary partition of D and let (ξi , ηi ) ∈ Di , i =
1, p be an arbitrary point. The mass of each plate Di may be approximated with the
product f (ξi , ηi ) · area(Di ). Obviously, the mass of the entire plate is approximated
by:

Σ
p
Mass(D) ≈ f (ξi , ηi ) · area(Di ).
i=1

The exact value of the mass of lamina is given by the formula:

Σ
p ¨
Mass(D) = lim f (ξi , ηi ) · area(Di ) = f (x, y)dx dy. (6.11)
∥ρ∥→0
i=1 D

Remark 6.5.1 If the lamina D is homogeneous, i.e. it has a constant density at each
point ( f (x, y) = k = ct., ∀(x, y) ∈ D), then its mass is equal to:

Mass(D) = k · area(D).
6.5 Applications of the Double Integral in Geometry and Mechanics 181

6.5.2 Coordinates of the Center of Mass of a Lamina

Let D ⊂ R2 be a non-homogeneous lamina of density f : D → R+ and let

(x G , yG ) be the coordinates of the center of mass G. We consider a partition
ρ : D1 , D2 , . . . , D p of the domain D and some arbitrary points (ξi , ηi ) ∈ Di .
The mass of the plate Di is approximated by the product f (ξi , ηi ) · area(Di ). If
we consider the mass of the plate Di concentrated at a single point, namely at the
point (ξi , ηi ), then the coordinates of the center of mass will be (Fig. 6.15):

Σ
p Σ
p
ξi f (ξi , ηi ) · area(Di ) ηi f (ξi , ηi ) · area(Di )
i=1 i=1
xG ≈ , yG ≈ .
Σp Σ
p
f (ξi , ηi ) · area(Di ) f (ξi , ηi ) · area(Di )
i=1 i=1

Assuming that f is continuous on D, at the limit we get the exact coordinates:

Σ
p ˜
ξi f (ξi , ηi ) · area(Di ) x f (x, y)dx dy
= ˜
i=1 D
xG = lim
∥ρ∥→0 Σp
f (x, y) dx dy
f (ξi , ηi ) · area(Di ) D
i=1
Σ
p ˜
ηi f (ξi , ηi ) · area(Di ) y f (x, y)dx dy
= ˜
i=1 D
yG = lim (6.12)
∥ρ∥→0 Σp
f (x, y) dx dy
f (ξi , ηi ) · area(Di ) D
i=1

Remark 6.5.2 In the particular case of a homogeneous plate, it results that:

˜ ˜ ˜ ˜
x dx dy x dx dy y dx dy y dx dy
xG = ˜ , yG = ˜
D D D D
= = . (6.13)
1 dx dy area(D) 1 dx dy area(D)
D D

Example 6.5.1 Find the coordinates of the center of mass of a homogeneous lamina
D, with the constant density f (x, y) = 2, where:

Fig. 6.15 The center of y

mass of a non-homogeneous
lamina D G
D

O x
182 6 Double and Triple Integrals
{ π }
D = (x, y) ∈ R2 ; 0 ≤ x ≤ , 0 ≤ y ≤ cos x .
2
Using the formulas (6.13), we have successively (Fig. 6.16):
π ⎛ cos x ⎞ π
¨ ∫2 ∫ ∫2
1 dx dy = ⎝ dy ⎠ dx = cos x dx = 1.
D 0 0 0
⎛ cos x ⎞ π π
¨ ∫ ∫ 2 ∫2
x dx dy = ⎝ x dy ⎠dx = x cos x dx
D 0 0 0
π
| ∫ 2
|π π
|
= x sin x | − sin x dx = − 1.
2
0 2
0
⎛ cos x π ⎞ π
¨ ∫ ∫ 2 ∫2
cos2 x
y dx dy = ⎝ y dy ⎠dx = dx
2
D 0 0 0
π
∫ 2 ( )|
cos 2x + 1 1 sin 2x |π π
= dx = + x || 2 = .
4 4 2 0 8
0

(π )
Therefore the center of mass has the coordinates (x G , yG ) = 2
− 1, π8 .

Remark 6.5.3 If the lamina is symmetrical with respect to O x axis (respectively O y

axis), then its center of mass G ∈ O x, hence yG = 0 (respectively G ∈ O y, hence
x G = 0).

Fig. 6.16 The center of y

mass of a non-homogeneous
lamina D 1
cos x

O π x
2
6.5 Applications of the Double Integral in Geometry and Mechanics 183

6.5.3 Moments of Inertia of a Lamina

It is known that the moment of inertia of a material point with respect to a certain
axis is equal to the product of the mass of the point and the square of the distance
from the point to the axis. In the case of a system of material points, the moment of
inertia with respect to an axis is the sum of the moments of inertia of the material
points that constituting the system with respect to the same axis.
Let D be a lamina with a continuous density f : D → R+ , let ρ : D1 , D2 , . . . , D p
be any partition of D and (ξi , ηi ) ∈ Di some arbitrary points. We approximate
as before the mass of the plate Di with the product f (ξi , ηi ) · area(Di ) and
consider this mass concentrated at the point (ξi , ηi ). The moment of inertia of
Σ p system
this of material points with respect to O y axis will be equal to the sum
ξ
i=1 i
2
f (ξi , ηi ) · area(Di ). If the norm of the partition ρ is small, this sum can be
considered as an approximate value of the moment of inertia I y of the lamina D with
respect to O y axis.
The exact value of the moment of inertia I y with respect to O y is:

Σ
p ¨
I y = lim ξi2 f (ξi , ηi ) · area(Di ) = x 2 f (x, y)dx dy. (6.14)
∥ρ∥→0
i=1 D

Similarly, the moment of inertia Ix of the lamina D with respect to O x axis is:

Σ
p ¨
Ix = lim ηi2 f (ξi , ηi ) · area(Di ) = y 2 f (x, y) dx dy. (6.15)
∥ρ∥→0
i=1 D

Taking into account that the moment of inertia of a material ( point )of mass m
(placed at the point (x, y)) with respect to the origin O(0, 0) is m· x 2 + y 2 , applying
the same arguments we find that:
¨
( )
I O = Ix + I y = x 2 + y 2 f (x, y)dx dy. (6.16)
D

Remark 6.5.4 If the plate is homogeneous, namely it has a constant density at each
point ( f (x, y) = k = ct., ∀(x, y) ∈ D), then:
¨ ¨
Ix = k y dx dy; I y = k
2
x 2 dx dy; I O
D D
¨
( )
=k x + y dx dy.
2 2

D
184 6 Double and Triple Integrals

Fig. 6.17 The quarter disk y

of radius R centered at the
origin
D

O R x

Example 6.5.2 Find the moments of inertia with respect to the axis O x, respec-
tively with respect to the origin O of the non-homogeneous lamina having density
f (x , y){ = x, where: }
D = (x, y) ∈ R2 ; x 2 + y 2 ≤ R 2 , x ≥ 0, y ≥ 0 (Fig. 6.17).
The plane domain D is the quarter disk of radius R centered at the origin that lies
in the first quadrant of the axis system. By passing to polar coordinates (Fig. 6.18):
{
x = ρ cos θ π
, 0 ≤ ρ ≤ R, 0 ≤ θ ≤ ,
y = ρ sin θ 2

we get:
π ⎛ R ⎞ π
¨ ∫2 ∫ 5 ∫
2

⎝ ρ 4 sin2 θ cos θ dρ ⎠dθ = R

Ix = y 2 x dx dy = sin2 θ cos θ dθ
5
D 0 0 0
π
∫2 |
R5 ' R 5 sin3 θ || π R5
= sin θ (sin θ ) dθ =
2
· 2 =
5 5 3 |0 15
0
π ⎛ ⎞ π
¨ ∫2 ∫R ∫2
( ) R5 R5
IO = x 2 + y x dx dy =
2 ⎝ ρ 4 cos θ dρ ⎠dθ = cos θ dθ = .
5 5
D 0 0 0

Fig. 6.18 The variation ρ

domain of the polar
coordinates θ and ρ R
Ω

O π θ
2
6.6 Riemann–Green Formula 185

6.6 Riemann–Green Formula

The Riemann–Green formula makes the connection between the double integral and
the line integral of the second kind, being used mainly for the processing of line
integrals on closed curves.
Let D ⊂ R2 be a bounded plane domain whose boundary C = ∂ D is a piecewise
smooth curve formed by a finite union of simple closed curves. Let P, Q : D → R
be two real continuous functions with the property that there are the first order partial
derivatives ∂∂Py and ∂∂Qx continuous on D.
With these details, the Riemann–Green formula is expressed as follows:
¨ ( ) ∫
∂Q ∂P
(x, y) − (x, y) dx dy = ↺ P(x, y)dx + Q(x, y)dy (6.17)
∂x ∂y
D C

In this formula, the orientation of the curve C = ∂ D is chosen so that when we

traverse the curve, the domain D must always be on the left.
In Fig. 6.19 we have exemplified the orientation of the curve for a domain whose
boundary consists of a single closed curve and in Fig. 6.20 for a domain whose
boundary consists of a finite union of closed curves.
Definition 6.6.1 By elementary domain of Green type (G− elementary domain)
we mean any of the four domains represented in Fig. 6.21.
We will prove the Riemann–Green formula (6.17) in stages.
Stage 1.

Fig. 6.19 The orientation of

the curve for a domain
whose boundary consists of a
single closed curve

Fig. 6.20 The orientation of

the curve for a domain
whose boundary consists of a
finite union of closed curves
186 6 Double and Triple Integrals

Fig. 6.21 Elementary domains of Green-type

Fig. 6.22 Elementary y

domain of Green-type
d E
f
y

Δ
A
c B

O a x b x

Lemma 6.6.1 The Riemann–Green formula is established for any G—elementary

domain.

Proof First, we consider an elementary domain of Green type Δ as in Fig. 6.22.

More specifically, such a domain is defined as follows:
{ }
Δ = (x, y) ∈ R2 ; a < x < b, c < y < f (x)

where f : [a, b] → [c, d] is a continuous, strictly increasing, surjective function

(Fig. 6.22).
͡
Its boundary consists of the segments AB, B E and the arc E A. According to the
computation method of the double integral for simple domains with respect to O y
axis (formula 6.2), we have:
¨
∂P
− (x, y)dx dy
∂y
Δ
⎛ ⎞
∫b ∫f (x) ∫b ∫b
⎝ ∂P
=− (x, y)dy ⎠dx = − P(x, f (x))dx + P(x, c)dx. (6.18)
∂y
a c a a
6.6 Riemann–Green Formula 187

͡
Considering the following parametric representations of the arc AE and the
segments AB and B E:
͡
AE : x = t, y = f (t), t ∈ [a, b]
AB : x = t, y = c, t ∈ [a, b]
B E : x = b, y = t, t ∈ [c, d]

we deduce:

∫ ∫b
P(x, y)dx = P(t, f (t))dt
͡ a
AE
∫ ∫b
P(x, y)dx = P(t, c)dt
AB a
∫
P(x, y)dx = 0. (6.19)
BE

From (6.18) and (6.19), it follows that:

¨ ∫ ∫ ∫ ∫
∂P
− dx dy = P dx + P dx + P dx = ↺ P dx. (6.20)
∂y
Δ AB BE
͡ ∂Δ
AE

On the other hand, we have:

⎛ ⎞
¨ ∫d ∫b
∂Q ⎜ ∂Q ⎟
(x, y)dx dy = ⎝ (x, y)dx ⎠dy
∂x ∂x
Δ c f −1 (y)

∫d ∫d
( )
= Q(b, y)dy − Q f −1 (y), y dy. (6.21)
c c

͡
If we consider for arc AE the parametric representation:
͡
AE : x = f −1 (t), y = t, t ∈ [c, d]

we obtain:
188 6 Double and Triple Integrals

∫ ∫d
( )
Q(x, y)dy = Q f −1 (t), t dt. (6.22)
͡ c
AE

For the segments AB and B E, we have:

∫ ∫ ∫d
Q(x, y)dy = 0 and Q(x, y)dy = Q(b, t)dt. (6.23)
AB BE c

From (6.21), (6.22) and (6.23), it results that:

¨ ∫ ∫ ∫ ∫
∂Q
dx dy = Qdy + Q dy + Q dy = ↺ Q dy. (6.24)
∂x
Δ AB BE
͡ ∂Δ
EA

Adding formulas (6.20) and (6.24), we obtain the Riemann–Green formula for
the G-elementary domain Δ considered in Fig. 6.22.
It is obvious that the proofs of the Riemann–Green formula for the other G−
elementary domains in Fig. 6.21 are absolutely analogous.
Stage 2.

Lemma 6.6.2 The Riemann–Green formula is valid for any triangular domain.

Proof Let Δ be an arbitrary triangular domain having as boundary the triangle ABC
(Fig. 6.23).
Let E be the point of intersection between the straight line parallel to the axis
O y and passing through A and the straight line parallel to the axis O x and passing
through B. Let F be the intersection point between the straight lines AE and BC.

Fig. 6.23 Triangular domain A

y

Δ1 Δ3
B E
Δ2
F

G C

O x
6.6 Riemann–Green Formula 189

We will denote also by G the intersection point between the straight lines AE and
the straight line parallel to the axis O x and passing through C.
The domain Δ is the union of domains Δ1 , Δ2 , Δ3 , where Δ1 has the boundary
AB E, Δ2 has the boundary B E F and Δ3 has the boundary AFC.
We notice that Δ1 and Δ2 are G—elementary domains, while Δ3 does not have
this property. On the other hand, it is clear that Δ3 can be described as the difference
of two G—elementary domains. Indeed, if we denote by Δ4 the domain whose
boundary is AGC and by Δ5 the domain whose boundary is F GC, then Δ4 , Δ5 are
G—elementary domains and Δ3 = Δ4 \Δ5 .

Applying the Riemann–Green formula for each of the G—elementary domains

Δ1 , Δ2 , Δ4 , Δ5 , we obtain:
¨ ( ) ¨ ¨ ¨ ¨ ¨ ¨ ¨
∂Q ∂P
− dx dy = + + = + + −
∂x ∂y
Δ Δ1 Δ2 Δ3 Δ1 Δ2 Δ4 Δ5
⎛ ⎞ ⎛ ⎞⎛ ⎞
∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫
⎜ ⎟ ⎜ ⎟⎜ ⎟
= ⎝ + + ⎠ + ⎝ + + ⎠⎝ + + ⎠
AB BE EA BF FE EB AG GC CA
⎛ ⎞
∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫
⎜ ⎟
−⎝ + + ⎠= + + + + + + +
FG GC CF AB EA BF FE AE EF FG GC
⎛ ⎞
∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫ ∫
⎜ ⎟
+ − − + = +⎝ + ⎠+ = + +
CA FG GC FC AB BF FC CA AB BC CA
∫
= ↺ P dx + Q dy.
∂Δ

Theorem 6.6.1 The Riemann–Green formula is valid for any polygonal domain.

Proof Clearly, any polygonal domain is a finite union of triangular domains which
have no common interior points.
For simplicity, let us consider the domain Δ represented in Fig. 6.24, whose
boundary is the quadrilateral ABC D. Obviously, Δ = Δ1 ∪ Δ2 , where Δ1 is the
triangular domain AB D and Δ2 is the triangular domain BC D. From Lemma 6.6.2,
we have:
¨ ( ) ¨ ( ) ¨ ( )
∂Q ∂P ∂Q ∂P ∂Q ∂P
− dx dy = − dx dy + − dx dy
∂x ∂y ∂x ∂y ∂x ∂y
Δ Δ1 Δ2
∫ ∫
= ↺ P dx + Q dy + ↺ P dx + Q dy
AB D BC D
190 6 Double and Triple Integrals

Fig. 6.24 Polygonal domain B

A
Δ1
Δ2

D C

⎛ ⎞ ⎛ ⎞
∫ ∫ ∫ ∫ ∫ ∫
⎜ ⎟ ⎜ ⎟
=⎝ + + ⎠+⎝ + + ⎠
AD DB BA DC CB BD
⎛ ⎞
∫ ∫ ∫ ∫ ∫
⎜ ⎟
=⎝ + + + ⎠ = ↺ P dx + Q dy.
AD DC CB BA ∂Δ

Stage 3.

Theorem 6.6.2 The Riemann–Green formula is valid for any domain whose
boundary is a simple closed piecewise smooth curve.

Proof Indeed, it can be shown that there is a sequence of polygonal lines {Cn },
inscribed in C = ∂ D such that:
∫ ∫
lim P(x, y)dx + Q(x, y)dy = P(x, y)dx + Q(x, y)dy.
n→∞
Cn C

If we denote by Dn the bounded domain whose boundary is the curve Cn , then

from Lemma 6.6.2, we deduce:
¨ ( )
∂Q ∂P
(x, y) − (x, y) dx dy
∂x ∂y
D
¨ ( )
∂Q ∂P
= lim (x, y) − (x, y) dx dy
n→∞ ∂x ∂y
Dn
∫ ∫
= lim P(x, y)dx + Q(x, y)dy = P(x, y)dx + Q(x, y)dy.
n→∞
Cn C

∫
Example 6.6.1 Compute ↺ (y − x y)dx + (x y − x)dy, where:
∂D
6.6 Riemann–Green Formula 191

Fig. 6.25 Ellipse centered y

in the origin and of semi-axis b
a and b
D
O a x

{ 2 2
}
D = (x, y) ∈ R2 ; ax 2 + by2 ≤ 1 (Fig. 6.25).
If we denote by P(x, y) = y − x y and by Q(x, y) = x y − x, then

∂P ∂Q
= 1 − x and = y − 1.
∂y ∂x

From the formula Riemann–Green (6.17), we have:

∫
↺ (y − x y)dx + (x y − x)dy
∂D
¨
= (x + y − 2)dx dy.
D

According to Example 6.4.4, we obtain:

¨
(x + y − 2)dx dy = −2 π a b.
D

−
→ −
→ −
→ −
→
Remark 6.6.1 If V : D ⊂ R2 → R2 , V (x, y) = P(x, y) i + Q(x, y) j is a
vector field of C -class on the squareable bounded domain D and C = ∂ D, then.
1

∫ ∫
−
→ −
V ·→
τ = ↺ P(x, y)dx + Q(x, y)dy
C+ C
¨ ( )
∂Q ∂P
= (x, y) − (x, y) dx dy.
∂x ∂y
D

Remark 6.6.2 If D ⊂ R2 is a squareable domain for which the Riemann–Green

formula is valid, then:
∮
1
area(D) = x dy − y dx. (6.25)
2
∂D
192 6 Double and Triple Integrals

Proof Let us denote by P(x, y) = − 2y and by Q(x, y) = 2x .

Then, using the Riemann–Green formula and Proposition 6.2.1, it results that:
∮ ¨ ( )
1 ∂Q ∂P
x dy − y dx = (x, y) − (x, y) dx dy
2 ∂x ∂y
∂D D
¨ ( ) ¨
1 1
= + dx dy = 1 dx dy = area(D).
2 2
D D

x2 y2
Example 6.6.2 Compute the area of the elliptical domain D : a2
+ b2
≤ 1.
x2 y2
Because the boundary of the domain D is the ellipse a2
+ b2
= 1, then, using the
parametric equations of the ellipse:
{
x = a cos t
, t ∈ [0, 2 π ], a, b > 0
y = b sin t

and the formula (6.25), we obtain:

∮ ∫2 π
1 1
area(D) = x dy − y dx = (a cos t b cos t + b sin t a sin t) dt
2 2
∂D 0
∫2 π ∫2 π
1 ( ) 1
= ab cos2 t + ab sin2 t dt = a b dt = π a b.
2 2
0 0

6.7 Improper Double Integrals

In this paragraph we present the notion of improper double integral in case where
the integrated domain is unbounded. Let D ⊂ R2 be an unbounded domain and
let f : D → R. We will assume that f is integrable on any squareable bounded
subdomain of D.
˜
Definition 6.7.1 We say that the double integral f (x, y)dx dy is convergent if
D
for any sequence of squareable bounded domains {Dn } with the properties:
(i) D 1 ⊂ D2 ⊂ · · · ⊂ Dn ⊂ · · ·
(ii) D n ⊂∪Dn+1 , ∀ n ∈ N∗
(iii) D = ∞ n=1 Dn
˜
the limit lim f (x, y)dx dy exists, it is finite and does not depend on the choice
n→∞
Dn
of the sequence {Dn }. In case of convergence, we will write
6.7 Improper Double Integrals 193
¨ ¨
f (x, y)dx dy = lim f (x, y)dx dy.
n→∞
D Dn

˜
Otherwise, if the limit lim f (x, y)dx dy does not exist or it is infinite, we say
n→∞
˜ Dn
that the integral f (x, y)dx dy is divergent.
D
˜
Example 6.7.1 Establish the nature of the improper double integral x y dx dy.
{ } R2
Let us denote by Dn = (x, y) ∈ R2 ; x 2 + y 2 < n 2 , n ∈ N∗ . It is obvious that
the sequence {Dn } of bounded domains has the properties (i)–(iii) of Definition 6.7.1.
Since Dn is the disk of radius n centered at the origin, passing to polar coordinates
we have:
⎛ ⎞
¨ ∫2 π ∫n ∫2 π
⎝ ⎠ n4
x y dx dy = ρ cos θ sin θ dρ dθ =
3
cos θ sin θ dθ = 0.
4
Dn 0 0 0

On the other hand, we consider the domains:

{ }
Dn' = (x, y) ∈ R2 ; −n < x < 2 n, −n < y < 2 n , n ∈ N∗ .
{ }
Clearly, the sequence Dn' fulfills also the properties (i)–(iii) of Definition 6.7.1.
Furthermore, we have:
⎛ 2n ⎞ ⎛ 2n ⎞ ⎛ 2n ⎞
¨ ∫2 n ∫ ∫ ∫ 4
x y dx dy = ⎝ x y dx ⎠dy = ⎝ x dx ⎠ · ⎝ y dy ⎠ = 9n .
4
Dn' −n −n −n −n

˜ ˜ 4
Since x y dx dy = 0 and lim x y dx dy = lim 9n4 = ∞, we deduce that
n→∞ ' n→∞
Dn
˜ Dn
the improper double integral x y dx dy is divergent.
R2

The improper double integral as defined above, retains the main properties of
the double integral on a bounded domain. Thus, linearity, monotony, additivity with
respect to the integrated domain and the change of variables formula are properties
that remain true even in the case of double integral on an unbounded domain.
For positive functions defined on an unbounded domain, the following theorem
can be proved.

Theorem 6.7.1 Let D ⊂ R2 be an unbounded domain and let f : D → R+ be a

positive function,˜integrable on any squareable subdomain of D. Then the improper
double integral f (x, y)dx dy is convergent if and only if there is at least one
D
194 6 Double and Triple Integrals

sequence of bounded squareable domains,

˜ with the properties (i)–(iii) for which the
sequence {an } is bounded, where an = f (x, y)dx dy.
Dn
Moreover, we have:
¨ ¨
f (x, y)dx dy = lim an = lim f (x, y)dx dy.
n→∞ n→∞
D Dn

˜
e−x −y dx dy is
2 2
Example 6.7.2 Prove that the following improper double integral
R2
∫∞
convergent, and then find that the value of the Euler–Poisson integral is −∞ e−x dx =
2

√
π.
Since the function f (x, y) = e−x −y ≥ 0, ∀(x, y) ∈ R2 is positive, it is enough
2 2

to find a sequence {Dn } of bounded squareable domains for which the sequence of
positive numbers {an } is bounded, where:
¨
an = f (x, y)dx dy.
Dn

{ }
If we choose Dn = (x, y) ∈ R2 ; x 2 + y 2 < n 2 , n ∈ N∗ , then the sequence
{Dn } fulfills the properties (i)–(iii) from Definition 6.7.1.
Because Dn is the disk of radius n centered at the origin, passing to polar
coordinates we obtain:
⎛ ⎞
¨ ∫2 π ∫n
e−x −y dx dy = ⎝ e−ρ · ρ dρ ⎠dθ
2 2 2
an =
Dn 0 0

∫2 π( | ) ∫2 π
1 −ρ 2 ||n 1 ( ) ( )
1 − e−n dθ = 1 − e−n · π.
2 2
= − e | dθ =
2 0 2
0 0

˜
e−x −y 2
2
Because lim an = π , from Theorem 6.7.1, it follows that dx dy is
n→∞
R2
convergent and its value is:
¨
e−x −y 2
2
dx dy = lim an = π.
n→∞
R2

On the other hand, let us consider the domains:

{ }
Dn' = (x, y) ∈ R2 ; |x| < n, |y| < n , n ∈ N∗ .
6.8 Volume of a Space Figure 195
{ }
Obviously, the sequence of domains Dn' fulfills also the properties (i)–(iii) from
Definition 6.7.1. Further, we have:
⎛ ⎞
¨ ∫n ∫n
e−x −y dx dy = ⎝ e−x e−y dx ⎠dy
2 2 2 2

Dn' −n −n
⎛ ⎞ ⎛ ⎞ ⎛ ⎞2
∫n ∫n ∫n
=⎝ e−x dx ⎠ · ⎝ e−y dy ⎠ = ⎝ e−x dx ⎠ .
2 2 2

−n −n −n

From Theorem 6.7.1, it follows that:

¨ ¨
e−x −y dx dy = lim e−x −y dx dy
2 2 2 2
π=
n→∞
D Dn'
⎛ n ⎞2 ⎛ ∞ ⎞2
∫ ∫
= lim ⎝ e−x dx ⎠ = ⎝ e−x dx ⎠
2 2

n→∞
−n −∞

∫∞
(here we used the fact that −∞ e−x dx is convergent, see Example 3.2.4).
2

We thus calculated the Euler–Poisson integral, namely:

∫∞
√
e−x dx =
2
π.
∞

The most remarkable result is another property of improper double integral that
has no analogue in the one-dimensional case, that is, the convergence of a improper
double integral implies its absolute convergence. The following result can be proved:

Theorem ˜ 6.7.2 The necessary and sufficient condition

˜ for the improper double
integral f (x, y)dx dy to be convergent is that | f (x, y)|dx dy to be convergent.
D D

In other words, for improper double integral the concepts of convergence and
absolute convergence are equivalent.

6.8 Volume of a Space Figure

As we have seen in this chapter, the transition from the simple integral to the double
integral, in addition to many analogies, also implies some substantial changes, both
in terms of concepts and in terms of reasoning. These changes have their origin
mainly in the theory of measurable plane sets (which have area).
196 6 Double and Triple Integrals

In contrast to this situation, the transition from the double integral to the triple
integral does not involve any complication.
To begin with, it is necessary to introduce the notion of volume.
From elementary geometry, it is known that the volume of a rectangular paral-
lelepiped is equal to the product of the lengths of its edges. In particular, if T is a
rectangular parallelepiped with edges parallel to the coordinate axes, this means that:

T = [a1 , a2 ] × [b1 , b2 ] × [c1 , c2 ],

then:
def
Vol(T ) = (a2 − a1 )(b2 − b1 )(c2 − c1 ).

The same formula of volume remains true even if the rectangular parallelepiped
does not contain one or more faces.

Definition 6.8.1 By an elementary space subset E ⊂ R3 we mean any finite union

of rectangular parallelepipeds with edges parallel to the coordinate axes, which may
or may not contain one or more faces and which have in common two by two only
at most one face.

More precisely, E ⊂ R3 is an elementary set if there is a finite number of rectan-

gular parallelepipeds Ti , i = 1, p, with edges parallel to the coordinate axes, such
that:

∪
p
◦ ◦
E= Ti and Ti ∩ T j = ∅, for i /= j.
i=1

The volume of such elementary set is by definition the sum of the volumes of the
parallelepipeds that compose it:

def Σ
p
Vol(E) = Vol(Ti ).
i=1

Next, we note by T the family of all elementary space sets.

We recall that a set T ⊂ R3 is bounded if there is r > 0 such that T is included
in the ball with the center at origin and radius r (i.e. T ⊆ B(O, r )).
If T ⊂ R3 is a bounded set, then we will define:
V∗ (T ) = sup{Vol(E); E ∈ T , E ⊂ T } the interior volume of T .
V ∗ (T ) = inf{Vol(F); F ∈ F , F ⊃ T } the exterior volume of T .
If the set T does not contain any elementary set, we will define V∗ (T ) = 0. Clearly
we have:

V∗ (T ) ≤ V ∗ (T ).
6.8 Volume of a Space Figure 197

Definition 6.8.2 We say that the bounded set T ⊂ R3 is cubable (has volume) if:

V∗ (T ) = V ∗ (T ) = V (T ).

The common value V (T ) is called the volume of T and it is denoted by Vol(T ).

Remark 6.8.1 Any elementary space set is cubable and its volume coincides with
the volume defined in Definition 6.8.1.

Example 6.8.1 Let D ⊂ R2 be a squareable bounded domain and let h > 0 be

a positive number. By a generalized right cylinder with base D and height h we
understand the following solid:
{ }
T = (x, y, z) ∈ R3 ; (x, y) ∈ D, 0 ≤ z ≤ h .

We will show that the generalized right cylinder T is cubable and its volume is:

Vol(T ) = h · area(D).

Indeed, we recall that E denotes the family of all elementary plane sets. Since D
is a squareable bounded domain, from Theorem 2.7.1, it follows that for any ε > 0,
there are E ε , Fε ∈ E with the properties:

E ε ⊂ D ⊂ Fε and area(Fε ) − area(E ε ) < ε.

∪p
We remark that if E = i=1 [ai , bi ] × [ci , di ] ∈ T , then

∪
p
PE = [ai , bi ] × [ci , di ] × [0, h] ∈ E .
i=1

Further, we have:

h · area(E ε ) ≤ h · area(D) = h · sup{area(E); E ⊂ D, E ∈ E } ≤ V∗ (T )

≤ V ∗ (T ) ≤ h · sup{area(F); F ⊃ D, F ∈ E } = h · area(D) ≤ h · area(Fε ).

Taking into account that h · (area(Fε ) − area(E ε )) < h · ε and ε > 0 is arbitrary,
it follows that:

V∗ (T ) = V ∗ (T ) = h · area(D).

Next, we will show that any curvilinear cylinder (cylindroid) is cubable.

By a curvilinear cylinder with base D ⊂ R2 , we understand a solid T bounded
by the base D, a surface z = f (x, y), (x, y) ∈ D, and the lateral cylindrical surface
(Fig. 6.26).
198 6 Double and Triple Integrals

Fig. 6.26 Curvilinear z = f ( x, y)

cylinder z

O y
D
x ∂D

More precisely, if D ⊂ R2 is a squareable bounded domain and f : D → R+ is a

positive continuous function, then the corresponding curvilinear cylinder is:
{ }
T = (x, y, z) ∈ R3 ; (x, y) ∈ D, 0 ≤ z ≤ f (x, y) .

Theorem 6.8.1 Any curvilinear cylinder is cubable and its volume is given by the
formula:
¨
Vol(T ) = f (x, y)dx dy. (6.26)
D

Proof If we consider that ρ : D1 , D2 , . . . , D p is an arbitrary partition of D and

(ξi , ηi ) ∈ Di , i = 1, p, are arbitrary points, then we shall denote by:
{ }
Ti = (x, y, z) ∈ R3 ; (x, y) ∈ Di , 0 ≤ z ≤ f (ξi , ηi ) , 1 ≤ i ≤ p.

Taking into account to Example 6.8.1, we have:

Vol(Ti ) = f (ξi , ηi ) · area(Di ).

Therefore, the volume of the cylindrical solid T can be approximated by the sum:

Σ
p
Σ
p
Vol(T ) ≈ Vol(Ti ) = f (ξi , ηi ) · area(Di ).
i=1 i=1

But we notice that the last sum is the Riemann sum of the function f , associated
to the partition ρ and the intermediate points (ξi , ηi ).
The exact value of the volume of the curvilinear cylinder T is:

Σ
p
Σ
p
Vol(T ) = lim Vol(Ti ) = lim f (ξi , ηi ) · area(Di ).
∥ρ∥→0 ∥ρ∥→0
i=1 i=1
6.9 Triple Integrals. Definition and Basic Properties 199

Here it is taken into account that f is continuous on D, hence integrable on D

and we obtained
¨
Vol(T ) = f (x, y)dx dy.
D

Corollary 6.8.1 Let D ⊂ R2 be a squareable bounded domain and let f, g: D → R

be two continuous functions such that f (x, y) ≤ g(x, y), ∀(x, y) ∈ D. Then the
space domain:
{ }
T = (x, y, z) ∈ R3 ; (x, y) ∈ D, f (x, y) ≤ z ≤ g(x, y)

is cubable and its volume is determined by the formula:

¨
Vol(T ) = (g(x, y) − f (x, y))dx dy.
D

The following theorem results immediately from Definition 6.8.2; its proof is
completely analogous to the case of plane sets (Theorem 2.7.1).

Theorem 6.8.2 A bounded subset T ⊂ R3 is cubable if and only if for any ε > 0,
there are two elementary space subsets Pε and Q ε with the properties:

Pε ⊂ T ⊂ Q ε and Vol(Q ε ) − Vol(Pε ) < ε.

Definition 6.8.3 A subset S ⊂ R3 is said to be of volume zero if for any ε > 0,

there is an elementary space set Q ε ∈ T such that S ⊂ Q ε and Vol(Q ε ) < ε.

The following theorem can be proved.

Theorem 6.8.3 A subset T ⊂ R3 is cubable if and only if its boundary is of volume

zero.

6.9 Triple Integrals. Definition and Basic Properties

In the following we will denote by T ⊂ R3 a bounded cubable subset.

Definition 6.9.1 It is called a partition of T any finite family of cubable subdomains

∪p
ρ : T1 , T2 , . . . , T p which have no pairwise common interior points and T = i=1 Ti .

By the norm of the partition ρ we mean the largest of the diameters of the domains
Ti , i = 1, p, i.e.:
200 6 Double and Triple Integrals

∥ρ∥ = max{diam(Ti ); 1 ≤ i ≤ p}
{ ( ) }
where diam(Ti ) = sup dist M ' , M '' ; M ' , M '' ∈ Ti .
If ρ : T1 , T2 , . . . , T p is a partition of T and f : T → R is a bounded function, then
we will denote by:

m = inf{ f (x, y, z); (x, y, z) ∈ T }, M = sup{ f (x, y, z); (x, y, z) ∈ T }

m i = inf{ f (x, y, z); (x, y, z) ∈ Ti }, 1 ≤ i ≤ p

Mi = sup{ f (x, y, z); (x, y, z) ∈ Ti }, 1 ≤ i ≤ p.

Definition 6.9.2 The lower Darboux sum, respectively the upper Darboux sum
associated to the bounded function f and the partition ρ are defined as follows:

Σ
p
Σ
p
sρ = m i · Vol(Ti ), Sρ = Mi · Vol(Ti ).
i=1 i=1

Σ
p
Because m ≤ m i ≤ Mi ≤ M, ∀ 1 ≤ i ≤ p and Vol(T ) = Vol(Ti ), it results:
i=1

m · Vol(T ) ≤ sρ ≤ Sρ ≤ M · Vol(T ). (6.27)

The properties of Darboux sums for three-variable functions are similar to the
properties of Darboux sums for two-variable functions, so we will not recall them.
If we denote by P the family of all partition of T , then we will define the lower
(upper) Darboux triple integral:
{ } { }
I∗ = sup sρ ; ρ ∈ P and I ∗ = inf Sρ ; ρ ∈ P .

The existence of I∗ and I ∗ results from inequalities (6.27). From the properties
of the Darboux sums, it follows that:

I∗ ≤ I ∗ .

Definition 6.9.3 Let f : T → R be a real bounded function defined on the cubable

bounded domain T ⊂ R3 . We say that the function f is integrable in the Darboux
sense on T if.

I∗ = I ∗ = I.
6.9 Triple Integrals. Definition and Basic Properties 201
˝
The common value I is denoted by I = f (x, y, z)dx dy dz and it is called
T
the triple integral of the function f on the domain T ; the real number I is uniquely
determined by the function f and T .

Example 6.9.1 Prove that any constant function on a bounded cubable set T ⊂ R3
is integrable. If f (x, y, z) = k, ∀(x, y, z) ∈ T , then
˚
k dx dy dz = k · Vol(T ).
T

Indeed, if ρ : T1 , T2 , . . . , T p is an arbitrary partition of T , then:

Σ
p
m i = Mi = k, ∀ i = 1, p and sρ = Sρ = k · Vol(Ti ) = k · Vol(T ),
i=1

whence, it results that:

˚
∗
I∗ = I = k · Vol(T ) = k dx dy dz.
T

In particular, we have:
˚
1 dx dy dz = Vol(T ).
T

Next, we define the Riemann sums.

Definition 6.9.4 Let T ⊂ R3 be a bounded cubable subset, f : T → R a bounded

function, ρ : T1 , T2 , . . . , T p an arbitrary partition of T and Pi (ξi , ηi , ζi ) ∈ Ti an
arbitrary intermediate point, ∀ i = 1, p.
The Riemann sum of the function f , associated to the partition ρ and the
intermediate points Pi is given by:

Σ
p
Σ
p
σρ ( f ; Pi ) = f (Pi ) · Vol(Ti ) = f (ξi , ηi , ζi ) · Vol(Ti )
i=1 i=1

Since m i ≤ f (Pi ) ≤ Mi , ∀ i = 1, p, it results that:

sρ ≤ σρ ( f ; Pi ) ≤ Sρ

for any system of intermediate points Pi ∈ Ti .

202 6 Double and Triple Integrals

Definition 6.9.5 We say that the function f : T → R is integrable in the Riemann

sense on the bounded cubable domain T if there is a finite number I , with the property
that for any ε > 0, there is δε > 0 such that for any partition ρ of T , with ∥ρ∥ < δε
and any intermediate points Pi ∈ Ti , we have:
| |
|σρ ( f ; Pi ) − I | < ε.

The number I is called the triple integral of function f on the domain T and it
is denoted by:
˚
I = f (x, y, z)dx dy dz.
T

We will also write:

˚
f (x, y, z)dx dy dz = lim σρ ( f ; Pi ),
∥ρ∥→0
T

the exact meaning being that of Definition 6.9.5.

Remark 6.9.1 The properties of the triple integral are completely analogous to the
properties of the double integral. In particular, it can be shown that any continuous
function on a compact cubable domain is integrable. Moreover, it can be shown
that both definitions of the triple integral, with Riemann sums and Darboux sums,
coincide.

6.10 Computing Triple Integral. Change of Variables

in Triple Integral

Definition 6.10.1 A domain T ⊂ R3 is called simple with respect to Ozaxis if

there are a bounded squareable domain D ⊂ R2 and two functions ϕ, ψ : D → R
◦
with the property ϕ(x, y) < ψ(x, y), ∀(x, y) ∈ D , such that:
{ }
T = (x, y, z) ∈ R3 ; (x, y) ∈ D, ϕ(x, y) ≤ z ≤ ψ(x, y) .

From Corollary 6.8.1, it results that such a domain has volume and:
¨ ¨
Vol(T ) = ψ(x, y)dx dy − ϕ(x, y)dx dy.
D D
6.10 Computing Triple Integral. Change of Variables in Triple Integral 203

Remark 6.10.1 In the case of a simple domain with respect to Oz axis, any straight
◦
line parallel to the Oz axis through an inner point (x, y) ∈ D , cuts the boundary of
the domain in at most two points.

The following theorem shows us how to compute the triple integral on a simple
domain with respect to Oz axis.

Theorem 6.10.1 Let T ⊂ R3 be a simple domain with respect to Oz axis and let
f : T → R be a continuous function. Then:
⎛ ⎞
˚ ¨ ∫
ψ(x,y)
⎜ ⎟
f (x, y, z)dx dy dz = ⎝ f (x, y, z)dz ⎠dx dy. (6.28)
T D ϕ(x,y)

Similarly, simple domains with respect to O x, respectively O y axis can be

considered, and formula (6.28) can be adapted for such simple domains.

Remark 6.10.1 If T is a rectangular parallelepiped with edges parallel to the

coordinate axes, i.e.:
{ }
T = (x, y, z) ∈ R3 ; a1 ≤ x ≤ a2 , b1 ≤ y ≤ b2 , c1 ≤ z ≤ c2 ⊂ R3

then, for any continuous function f : T → R the triple integral is calculated by the
formula:
⎛ ⎛ ⎞ ⎞
˚ ∫a2 ∫b2 ∫c2
f (x, y, z)dx dy dz = ⎝ ⎝ f (x, y, z)dz ⎠dy ⎠dx. (6.29)
T a1 b1 c1

In the right member of the formula (6.29) the order of integration can be changed
and other 5 computation formulas analogous to (6.29) are obtained (Fig. 6.27).

Remark 6.10.2 If f : [a1 , a2 ] → R, g: [b1 , b2 ] → R, h : [c1 , c2 ] → R are three

continuous functions and T = [a1 , a2 ] × [b1 , b2 ] × [c1 , c2 ], then:

˚ ∫a2 ∫b2 ∫c2

f (x) · g(y) · h(z)dx dy dz = f (x) dx · g(y) dy · h(z) dz.
T a1 b1 c1

˝
Example 6.10.1 Compute x y z dx dy dz, where:
T

{ }
T = (x, y, z) ∈ R3 ; 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c

From Remark 6.10.2, it results that:

204 6 Double and Triple Integrals

Fig. 6.27 Rectangular z

parallelepiped c

O
b y
a
x

⎛ ⎛ c ⎞ ⎞
˚ ∫a ∫b ∫
x y z dx dy dz = ⎝ ⎝ x y z dz ⎠dy ⎠dx
T 0 0 0
∫a ∫b ∫c
a 2 b2 c2
= x dx · y dy · z dz = .
8
0 0 0

˝
Example 6.10.2 Compute T z dx dy dz, where T is the conical domain bounded
√
by the surfaces: z = 0, z = 1, z = x 2 + y 2 (Fig. 6.28).
Geometrically, the equation x 2 + y 2 = z 2 represents a cone with the vertex at the
origin and symmetrical around the Oz axis. We notice that if we denote by D the
disk x 2 + y 2 < 1, then:
{ √ }
T = (x, y, z) ∈ R3 ; (x, y) ∈ D, x 2 + y 2 < z < 1

According to formula (6.28), we obtain:

Fig. 6.28 Cone with the z

vertex at the origin and
symmetrical around the Oz
1
axis

O y
x
6.10 Computing Triple Integral. Change of Variables in Triple Integral 205
⎛ ⎞
˚ ¨ ∫1
⎜ ⎟
z dx dy dz = ⎜ z dz ⎟
⎝ ⎠dx dy
T D
√
x 2 +y 2
¨
1 ( ( ))
= 1 − x 2 + y 2 dx dy
2
D
⎛ ⎞
∫2 π ∫1 ∫2 π
1 ⎝ ( )
⎠ 1 1 π
= 1 − ρ ρ dρ dθ =
2
dθ = .
2 2 4 4
0 0 0

Remark 6.10.3 If the integrated domain T is not simple with respect to any of the
coordinate axes, then the domain T can be divided, by planes parallel to one of the
subdomains T1 , T2 , . . . , Tn , which
coordinate planes, into a finite number of simple∪
n
do not have common interior points. Since T = i=1 Ti , we have:
˚ n ˚
Σ
f (x, y, z)dx dy dz = f (x, y, z)dx dy dz.
T i=1 T
i

Next, we present the change of variables theorem in triple integral.

Let Ω ⊂ R3 be an open, cubable, connected, bounded subset and let F: Ω → R3 ,
let

F(u, v, w) = (x(u, v, w), y(u, v, w), z(u, v, w)), ∀(u, v, w) ∈ Ω

be a vector function with the properties:

(i) F is of C 1 -class on Ω
(ii) F: Ω → T = F(Ω) is bijective
(iii) F is a regular transformation on Ω, i.e.

D(x, y, z)
det JF (u, v, w) = (u, v, w) /= 0, ∀(u, v, w) ∈ Ω.
D(u, v, w)

Such a vector function is called a change of variables.

As in the two-dimensional
( ) case,
( )it can be shown that.
T \T = F Ω\Ω and T = F Ω .

Theorem 6.10.2 (The change of variables formula in triple integral) Let F: Ω ⊂

R3 → T ⊂ R3 be a change of variables defined by

F(u, v, w) = (x(u, v, w), y(u, v, w), z(u, v, w)), ∀(u, v, w) ∈ Ω

and let f : T → R be a continuous function. Then:

206 6 Double and Triple Integrals
˚
f (x, y, z) dx dy dz
T
˚ | |
| D(x, y, z) |
= |
f (x(u, v, w), y(u, v, w), z(u, v, w))| (u, v, w)||du dv dw.
D(u, v, w)
Ω
(6.30)

Formula (6.30) is used in practice in order to transform the initial integral into an
easier to compute integral, either by simplifying the integrated domain Ω, either by
simplifying the integrated function f .
Mapping F transforms the domain Ω into the domain T . Consequently, the spec-
ification of a point (u, v, w) belonging to Ω uniquely determines the corresponding
point (x, y, z) of T . In other words, the numbers u, v, w can be regarded as coordi-
nates (different from Cartesian coordinates) of the points of the domain T . They are
called curvilinear coordinates.

In the following we will present three space coordinate systems that are
most frequently used in applications, namely: spherical coordinates, cylindrical
coordinates and generalized spherical coordinates.
Case 1. Spherical coordinates.
The spherical coordinates are expressed by the formulas:
⎧
⎨ x = ρ sin θ cos ϕ
y = ρ sin θ sin ϕ , 0 < ρ < ∞, 0 < θ < π, 0 < ϕ < 2π. (6.31)
⎩
z = ρ cos θ

The meaning of the notations is shown in Fig. 6.29, namely: ρ is the distance from
the origin O to the point M(x, y, z), θ is the angle between the positive direction
of the Oz axis and the straight line segment O M and ϕ is the angle formed by the
positive direction of the O x axis and the projection O N of the segment O M to O x y
plane.
Obviously we have:

x = ∥O A∥ = ∥O N ∥ cos ϕ = ρ sin θ cos ϕ

y = ∥O B∥ = ∥O N ∥ sin ϕ = ρ sin θ sin ϕ
z = ∥OC∥ = ρ cos θ

The Jacobian of the transformation is given by:

6.10 Computing Triple Integral. Change of Variables in Triple Integral 207

Fig. 6.29 The spherical z

coordinates
C

M ( x, y , z )
ρ
θ

O B
y
ϕ
A
N ( x, y , 0 )
x

| |
| sin θ cos ϕ ρ cos θ cos ϕ −ρ sin θ sin ϕ |
D(x, y, z) | |
= | sin θ sin ϕ ρ cos θ sin ϕ ρ sin θ cos ϕ || = ρ 2 sin θ > 0.
D(ρ, θ, ϕ) || |
cos θ −ρ sin θ 0
˝
Example 6.10.3 Compute x y z dx dy dz, where T is the solid bounded by the
T
surfaces: x = 0, y = 0, z = 0, x 2 + y 2 + z 2 = 1.
From a geometric point of view, the domain T is the first octant of the ball (solid
sphere) of radius R = 1 centered at the origin, having the equation x 2 + y 2 + z 2 ≤ 1
(Fig. 6.30).
In spherical
{ coordinates, the integrated domain T becomes} Ω, where:
Ω = (ρ, θ, ϕ) ∈ R3 ; 0 < ρ < 1, 0 < θ < π2 , 0 < ϕ < π2 (Fig. 6.31).
From Theorem 6.10.2, we deduce:

Fig. 6.30 The first octant of z

the solid sphere of radius
R = 1 centered at the origin
T

O y

x
208 6 Double and Triple Integrals

Fig. 6.31 The variation ϕ

domain of the spherical π
coordinates ρ, θ, ϕ 2

O π θ
1
ρ 2

˚ ˚
x y z dx dy dz = ρ 3 sin2 θ cos θ sin ϕ cos ϕ · ρ 2 sin θ dρ dθ dϕ
T Ω
⎛ π ⎞ ⎛ π ⎞ ⎛ ⎞
∫2 ∫2 ∫1
⎜ ⎟ ⎜ ⎟
= ⎝ sin ϕ cos ϕ dϕ ⎠ · ⎝ sin3 θ cos θ dθ ⎠ · ⎝ ρ 5 dρ ⎠
0 0 0
( | ) ( 4 | ) ( 6| )
sin ϕ || π
2
sin θ || π ρ ||1 1
= · · = .
2 |0 4 |0 6 |0
2 2
48

Case 2. Cylindrical coordinates.

The cylindrical coordinates are expressed by the formulas:

⎧
⎨ x = ρ cos θ
y = ρ sin θ , 0 < ρ < ∞, 0 < θ < 2 π, z ∈ R. (6.32)
⎩
z =z

The meaning of the notations is shown in Fig. 6.32. The Jacobian is:
| |
| cos θ −ρ sin θ 0 |
D(x, y, z) || |
= | sin θ ρ cos θ 0 || = ρ > 0.
D(ρ, θ, z) |
0 0 1|

Let us return to Example 6.10.2 and compute the triple integral using the change
of variables in cylindrical coordinates.
˝
Example 6.10.4 Compute z dx dy dz, where T is the domain bounded by the
√
T
surfaces: z = 0, z = 1, z = x 2 + y 2 .
In cylindrical coordinates, the integral over the conical domain T is the integral
over Ω, where:
{ }
Ω = (ρ, θ, z) ∈ R3 ; 0 < ρ < z, 0 < θ < 2 π, 0 < z < 1 .
6.10 Computing Triple Integral. Change of Variables in Triple Integral 209

Fig. 6.32 Cylindrical z

coordinates

M ( x, y , z )

z
O
ρ y
θ
N ( x, y , 0 )
x

From Theorem 6.10.2, it results that:

⎛ ⎛ ⎞ ⎞
˚ ˚ ∫2 π ∫1 ∫ z
z dx dy dz = z · ρ dρ dθ dz = ⎝ ⎝ z · ρ dρ ⎠dz ⎠dθ
T Ω 0 0 0
⎛ 2 π⎛ 1 ⎞ ⎞ ⎛ 2π ⎞ ⎛ 1 ⎞
∫ ∫ 3 ∫ ∫ 3
z z 1 π
=⎝ ⎝ dz ⎠dθ ⎠ = ⎝ 1 dθ ⎠ · ⎝ dz ⎠ = 2π · = .
2 2 8 4
0 0 0 0

Case 3. Generalized spherical coordinates.

The generalized spherical coordinates are expressed by the formulas:

⎧
⎨ x = a ρ sin θ cos ϕ
y = b ρ sin θ sin ϕ , 0 < ρ < ∞, 0 < θ < π, 0 < ϕ < 2 π. (6.33)
⎩
z = c ρ cos θ

D(x,y,z)
The Jacobian of transformation is D(ρ,θ,ϕ)
= a b c ρ 2 sin θ > 0.
˝ ( x2 z2
)
Example 6.10.5 Compute 9
+ y2 + 25
dx dy dz, where:
T

{ }
x2 z2
T = (x, y, z) ∈ R3 ; 1 ≤ + y2 + ≤4 .
9 25

In this case, the generalized spherical coordinates are:

210 6 Double and Triple Integrals
⎧
⎨ x = 3 ρ sin θ cos ϕ
y = ρ sin θ sin ϕ
⎩
z = 5 ρ cos θ

and the domain T is described by:

{ }
Ω = (ρ, θ, ϕ) ∈ R3 ; 1 ≤ ρ ≤ 2, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2 π .

According to Theorem 6.10.2, we have:

˚ ( ) ˚
x2 z2
+y +
2
dx dy dz = ρ 2 · 15 ρ 2 sin θ dρ dθ dϕ
9 25
T Ω
⎛ ⎛ ⎞ ⎞
∫2 π ∫π ∫2
= 15 ⎝ ⎝ ρ 4 sin θ dρ ⎠dθ ⎠dϕ
0 0 1
⎛ 2π ⎞ ⎛ π ⎞ ⎛ 2 ⎞
∫ ∫ ∫
= 15⎝ 1 dϕ ⎠ · ⎝ sin θ dθ ⎠ · ⎝ ρ 4 dρ ⎠
0 0 1
( | ) ( | ) ( 5| )
| | ρ ||2
= 15 ϕ ||2 π · − cos θ ||π · = 372π.
0 0 5 |1

6.11 Applications of the Triple Integral in Geometry

and Mechanics

6.11.1 Volume of a Space Domain

If T ⊂ R3 is a cubable bounded domain, then the volume of T is given by the

formula:
˚
Vol(T ) = 1 dx dy dz. (6.34)
T
6.11 Applications of the Triple Integral in Geometry and Mechanics 211

6.11.2 Mass of a Solid

By solid we mean a material set in the shape of a cubable bounded domain T ⊂ R3 ,

together with a positive and continuous function f : T → R+ , called density.
The solid is generally considered non-homogeneous, its density being variable.
Proceeding similarly as in the case of the mass of a non-homogeneous lamina,
we find that the mass of the solid is expressed through the triple integral as:
˚
Mass(T ) = f (x, y, z) dx dy dz. (6.35)
T

6.11.3 Coordinates of the Center of Mass of a Solid

The coordinates of the center of mass of a non-homogeneous solid T with the density
function f are described by the expressions:
˝
x f (x, y, z) dx dy dz
xG = ˝
T
f (x, y, z) dx dy dz
T
˝
y f (x, y, z)dx dy dz
yG = ˝
T
f (x, y, z)dx dy dz
T
˝
z f (x, y, z)dx dy dz
zG = ˝
T
. (6.36)
f (x, y, z)dx dy dz
T

Remark 6.11.1 In particular, if the solid is homogeneous ( f (x, y, z) = k = ct.,

∀(x, y, z) ∈ T ), the calculation formulas for the mass and the coordinates of the
center of mass become:
˚
Mass(T ) = k dx dy dz = k · Vol(T )
T
˝ ˝ ˝
x dx dy dz y dx dy dz z dx dy dz
T T T
xG = , yG = , zG = .
Vol(T ) Vol(T ) Vol(T )
212 6 Double and Triple Integrals

6.11.4 Moment of Inertia of a Solid

The moments of inertia of a non-homogeneous solid T having the density f with

respect to Oz axis (respectively, to O x y plane or to the origin O) are expressed by
the formulas:
˚
( 2 )
I Oz = x + y 2 f (x, y, z)dx dy dz
T
˚
IO x y = z 2 f (x, y, z)dx dy dz
T
˚
( )
IO = x 2 + y 2 + z 2 f (x, y, z)dx dy dz. (6.37)
T

Formulas (6.37) take equivalent forms in the event of a change in the coordinate
axis or the coordinate plane for which the moment of inertia is computed.

Remark 6.11.2 For a homogeneous solid of constant density f (x, y, z) = k, the

formulas (6.37) are given by:
˚
( )
I Oz = k · x 2 + y 2 dx dy dz
T
˚
IO x y = k · z 2 dx dy dz
T
˚
( )
IO = k · x 2 + y 2 + z 2 dx dy dz.
T

Example 6.11.1 Determine the coordinates of the center of mass and the moment
of inertia with respect to O y axis of the non-homogeneous solid T with the density
f , where:
{ }
T = (x, y, z) ∈ R3 ; x 2 + z 2 ≤ 9, 1 ≤ y ≤ 3 and f (x, y, z) = 2 y.

From a geometric point of view, the space domain T is the region that lies inside
the cylinder given by x 2 + z 2 = 9 and between the planes y = 1 and y = 3.
Let us denote by D the projection of the domain T to the coordinate plane O x z.
Then D is the disk of radius 3 centered at the origin:

x 2 + z 2 ≤ 9.

Therefore:
6.11 Applications of the Triple Integral in Geometry and Mechanics 213

Fig. 6.33 Cylinder z

symmetrical around the Oy
axis

D T
O 1 3 y

{ }
T = (x, y, z) ∈ R3 ; (x, z) ∈ D, 1 ≤ y ≤ 3 .

Obviously, because the cylinder T is symmetrical around the O y axis, then its
center of mass is situated on the axis O y, hence x G = z G = 0 (Fig. 6.33).
It remains to compute the coordinate yG . According to formulas (6.36), we have:
˝ ˝ ˝
y f (x, y, z)dx dy dz 2 y 2 dx dy dz y 2 dx dy dz
yG = ˝ = ˝ = ˝
T T T
f (x, y, z) dx dy dz 2 y dx dy dz y dx dy dz
T T T

whence:
⎛ ⎞ ⎛ ⎞
˚ ∫3 ¨ ∫3 ¨
y dx dy dz = ⎝ y dx dz ⎠dy = y⎝ 1 dx dz ⎠dy
T 1 D 1 D
∫3
= Area(D) · y dy = 9 π · 4 = 36 π
1
⎛ ⎞
˚ ∫3 ¨
y 2 dx dy dz = ⎝ y 2 dx dz ⎠dy
T 1 D
⎛ ⎞
∫3 ¨
= y2⎝ 1 dx dz ⎠dy = 78 π.
1 D

Thus, the coordinate yG of the center of mass is:

78 π 13
yG = = .
36 π 6
The moment of inertia with respect to O y axis is:
214 6 Double and Triple Integrals
˚
( )
IO y = x 2 + z 2 f (x, y, z)dx dy dz
T
⎛ ⎞
˚ ∫3 ¨
( ) ( )
= x2 + z 2
· 2 y dx dy dz = 2 ⎝ x 2 + z y dx dz ⎠dy
2

T 1 D
⎛ ⎞
∫3 ¨
( )
=2 y⎝ x 2 + z 2 dx dz ⎠dy
1 D
⎛ ⎛ ⎞ ⎞
∫3 ∫2π ∫3
=2 y ⎝ ⎝ ρ 3 dρ ⎠dθ ⎠dy
1 0 0
⎛ ⎞
∫3 ∫2π ∫3
81 ⎠
=2 y⎝ dθ dy = 81 π y dy = 324 π.
4
1 0 1
Chapter 7
Surface Integrals

7.1 Parameterized Surface Canvases. Definition

of a Surface

The notion of parameterized surface canvas is a natural generalization of the notion

of parameterized path.

Definition 7.1.1 Let Δ ⊂ R2 be an open connected subset (domain). Any vector

function r : Δ → R3 of C 1 − class is called a parameterized surface canvas.

If we denote by x, y, z the scalar components of the vector function r , then:

r (u, v) = (x(u, v), y(u, v), z(u, v)), ∀(u, v) ∈ Δ.

⎧
⎨ x = x(u, v)
The equations y = y(u, v), (u, v) ∈ Δ , are called the parametric equations
⎩
z = z(u, v)
of the canvas or a parametric representation of the canvas, and u and v are called
parameters.
We will denote by (Δ, r ) a parameterized surface canvas.
The direct image r (Δ) of the domain Δ through the vector function r , i.e. the
subset:

r (Δ) = {r (u, v); (u, v) ∈ Δ} = {(x(u, v), y(u, v), z(u, v)); (u, v) ∈ Δ} ⊂ R3

is called the support of parameterized surface canvas (Δ, r ).

The fact that the vector function r : Δ → R3 is of C 1 − class is equivalent to the
fact that the scalar functions x, y, z : Δ → R are of C 1 − class on Δ, that is, they
are continuous and their first order partial derivatives are continuous on Δ.
Next we will use some notations specific to differential geometry.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 215
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3_7
216 7 Surface Integrals
{ −
→ − → − →{
With respect to the Cartesian coordinate system O, i , j , k , the vector
equation of a surface canvas (Δ, r ) is:
−
→ −
→ −
→
r (u, v) = −
→
r (u, v) = x(u, v) i + y(u, v) j + z(u, v) k , ∀(u, v) ∈ Δ.

Partial derivatives with respect to the parameters u and v are denoted by:

∂x ∂x ∂y ∂y ∂z ∂z
xu = ; xv = ; yu = ; yv = ; zu = ; zv =
∂u ∂v ∂u ∂v ∂u ∂v
∂ r→ ∂ →
r
r→u = = xu →i + yu →j + z u k;
→ r→v = = xv →i + yv →j + z v k.
→
∂u ∂v
We also note with:
| | | |
| yu z u | | xu z u |
|
A = A(u, v) = | | |
; B = B(u, v) = −| |;
yv z v | xv z v |
| |
|x y |
C = C(u, v) = || u u ||
xv yv
∥− ∥2
E = E(u, v) = ∥→ r u ∥ = xu2 + yu2 + z u2
F = F(u, v) = r→u · r→v = xu xv + yu yv + z u z v
∥→ ∥2
G = G(u, v) = ∥− r ∥ = x 2 + y2 + z2.
v v v v

We remark that: ∥→ ∥ √
−
→ −
→ −
→ −
→
r u ×−→r v = A i + B j + C k and ∥ − r u×− →
r v ∥ = A2 + B 2 + C 2 .
If we denote by θ the angle between the vectors −
→
r u and −
→
r v , then:

ru ∥2 ∥→
E · G − F 2 = ∥→ rv ∥2 − (→
ru · r→v )2 =
∥ 2∥
= ∥r→u ∥∥→
rv ∥2 − ∥→
ru ∥2 ∥→rv ∥2 cos2 θ =
ru ∥2 ∥→
= ∥→ rv ∥2 sin2 θ = ∥→
ru × r→v ∥2 = A2 + B 2 + C 2

Thus, we obtained the following identity, called Lagrange identity:

E · G − F 2 = A2 + B 2 + C 2 . (7.1)

Definition 7.1.2 A parameterized surface canvas r : Δ → R3 is called simple if

the vector function r is injective, i.e. ∀(u 1 , v1 ), (u 2 , v2 ) ∈ Δ, (u 1 , v1 ) /= (u 2 , v2 )
implies r (u 1 , v1 ) /= r (u 2 , v2 ).

Definition 7.1.3 A parameterized surface canvas r : Δ → R3 is called smooth

if:
∥− ∥2
∥→
r u ×−
→
r v ∥ = A2 + B 2 + C 2 > 0, ∀(u, v) ∈ Δ.
7.1 Parameterized Surface Canvases. Definition of a Surface 217

Example 7.1.1 Let r : Δ → R3 be the parameterized surface canvas defined by:

−
→ −
→ −
→ −
→
r (u, v) = R sin u cos v i + R sin u sin v j + R cos u k , ∀(u, v) ∈ Δ

where
{ π π{
Δ = (u, v) ∈ R2 ; 0 < u < , 0 < v <
2 2
The parametric equations are:
⎧
⎨ x = R sin u cos v
y = R sin u sin v, (u, v) ∈ Δ .
⎩
z = R cos u

We observe that for any (u , v) ∈ Δ, the point (x(u, v), y(u, v), z(u, v)) verifies
the equation x 2 + y 2 + z 2 = R 2 , whence we deduce that the support of this canvas is
the portion of the sphere of radius R centered at the origin that lies in the first octant
(x > 0 , y > 0, z > 0) (Fig. 7.1).
Using the above notations, we have:

xu = R cos u cos v, xv = −R sin u sin v

yu = R cos u sin v, yv = R sin u cos v
z u = −R sin u, z v = 0

Fig. 7.1 The portion of the z

sphere of radius R centered
at the origin that lies in the
first octant

O y

Fig. 7.2 The diagramme of Φ = (ϕ 1 , ϕ 2 )

the parameter change Δ1 ⊂ R 2 Δ2 ⊂ R2

Φ −1
r1 r2

S ⊂R3
218 7 Surface Integrals

and further:

A = R 2 sin2 u cos v, B = R 2 sin2 u sin v

C = R 2 sin u cos u
E = R 2 , F = 0, G = R 2 sin2 u.

Since E · G − F 2 = A2 + B 2 + C 2 = R 4 sin2 u > 0, ∀(u, v) ∈ Δ, it results that

this surface canvas is smooth. We also notice that the considered canvas is simple.

Definition 7.1.4 Two parameterized surface canvases (Δ1 , r1 ) and (Δ2 , r2 ) are
called equivalent with the same orientation if there is a diffeomorphism.

ϕ : Δ1 ⊂ R2 → Δ2 ⊂ R2

with the properties:

det Jϕ (u 1 , v1 ) > 0, ∀ (u 1 , v1 ) ∈ Δ1 and r1 = r2 ◦ ϕ.
Recall that ϕ : Δ1 → Δ2 is diffeomorphism if ϕ is bijective, of C 1 − class on
Δ1 and its inverse ϕ−1 : Δ2 → Δ1 is of C 1 − class on Δ2 . Moreover, we recall that
if ϕ1 and ϕ2 are the scalar components of ϕ, then:
| |
| ∂ϕ1 (u , v ) ∂ϕ1
,v |
D(ϕ1 , ϕ2 ) | ∂u 1 1 1 ∂v1 (u 1 1 ) |
det Jϕ (u 1 , v1 ) = (u 1 , v1 ) = | ∂ϕ ∂ϕ2 |.
D(u 1 , v1 ) | ∂u 12 (u 1 , v1 ) ,v |
∂v1 (u 1 1 )

The equality r1 = r2 ◦ ϕ means:

r1 (u 1 , v1 ) = r2 (ϕ1 (u 1 , v1 ), ϕ2 (u 1 , v1 )), ∀ (u 1 , v1 ) ∈ Δ1

If det Jϕ < 0 on Δ1 , we say that the two surface canvases are equivalent to
opposite orientation. The function ϕ is also called parameter change.
We will use the notation (Δ1 , r1 ) ∼ (Δ2 , r2 ).

Remark 7.1.1 Any two equivalent parameterized surface canvases have the same
support.

Remark 7.1.2 Any parameterized surface canvas equivalent to a simple (smooth)

parameterized surface canvas is also simple (smooth).

Example 7.1.2 Show that the following parameterized surface canvases (Δ1 , r1 )
and (Δ2 , r2 ) are equivalent, where:

−
→ −
→ −
→ −
→
r 1 (u 1 , v1 ) = R sin u 1 cos v1 i + R sin u 1 sin v1 j + R cos u 1 k
{ π π{
Δ1 = (u 1 , v1 ) ∈ R2 ; 0 < u 1 < , 0 < v1 < ;
√ 2 2
−
→ −
→ −
→ −
→
r 2 (u 2 , v2 ) = u 2 i + v2 j + R 2 − u 22 − v22 k
7.1 Parameterized Surface Canvases. Definition of a Surface 219
{ }
Δ2 = (u 2 , v2 ) ∈ R2 ; u 22 + v22 < R 2 , u 2 > 0, v2 > 0 .

Indeed, it is verified by direct computation that the application ϕ : Δ1 → Δ2 ,

defined by

ϕ(u 1 , v1 ) = (R sin u 1 cos v1 , R sin u 1 sin v1 ) , (u 1 , v1 ) ∈ Δ1

satisfies the conditions of Definition 7.1.4, thus (Δ1 , r1 ) ∼ (Δ2 , r2 ).

Definition 7.1.5 A subset S ⊂ R3 is called elementary surface if there is a simple

and smooth parameterized surface canvas (Δ , r ) such that S = r (Δ) and, in
addition, the function r −1 : S → Δ to be continuous.

The pair (Δ , r ) is by definition a parametric representation of the surface S.

Therefore, a subset S ⊂ R3 is an elementary surface if there is an open connected
domain Δ ⊂ R2 and a vector function r : Δ → S of C 1 − class, with the properties:
(i) r is bijective
(ii) r −1 : S → Δ is continuous
(iii) A2 + B 2 + C 2 > 0, ∀(u, v) ∈ Δ.

Remark 7.1.3 It can be shown that if (Δ, r ) is a parametric representation of the

elementary surface S and Δ' ⊂ R2 (is another )domain such that the mapping ϕ :
Δ' → Δ is a diffeomorphism, then Δ' , r ◦ ϕ is also a parametric representation
of S.

Remark 7.1.4 The most convenient particular case of elementary surface is the
graph of a function f : D ⊂ R2 → R, also known as an explicit surface.

Let D ⊂ R2 be a domain and let f : D → R be a function of C 1 − class (Fig. 7.2).

We associate this function the parameterized surface canvas (D , r ):

r (x, y) = r→(x, y) = x →i + y →j + f (x, y)k,

→ ∀(x, y) ∈ D.

Next, we will show that the support of this canvas, i.e.,

not
S = r (D) = {(x, y, f (x, y)); (x, y) ∈ D} = G f ⊂ R3

that represents the graph of the function f , is an elementary surface (Fig. 7.3).
If we denote by p = ∂∂ xf and q = ∂∂ yf , then:
| | | | | |
|0 p| |1 p| |1 0 ||
|
A=| | |
= − p; B = −| | = −q; C = || =1
1q | 0q | 0 1|

and
A2 + B 2 + C 2 = 1 + p 2 + q 2 > 0, ∀(x, y) ∈ D,
220 7 Surface Integrals

Fig. 7.3 The elementary z

surface
S

O
y
D
x

whence we deduce that the canvas (D, r ) is smooth. On the other hand, it is obvious
that r : D → S is bijective. Finally, it is clear that the function r −1 : S → D, defined
by

r −1 (x, y, z) = (x, y), ∀(x, y, z) ∈ S

is continuous.

Definition 7.1.6 A set S ⊂ R3 is called a surface if any point of it belongs to an

open subset V of R3 such that V ∩ S is an elementary surface.

It is obvious that any elementary surface is a surface within the meaning of

Definition 7.1.6. In particular, any explicit surface is a surface (Remark 7.1.4).

Remark 7.1.5 The support of a smooth parameterized surface canvas is a surface.

Indeed, let (Δ, r ) be a smooth parameterized surface canvas, defined by:

r (u, v) = r→(u, v) = x(u, v)→i + y(u, v) →j + z(u, v)k,

→ ∀(u, v) ∈ Δ.

Let S = r (Δ) be its support and let M0 (x(u 0 , v0 ), y(u 0 , v0 ), z(u 0 , v0 )) ∈ S be

an arbitrary point. Since at least one of the parameters A , B , C is nonzero, we can
assume that:
| |
| xu yu |
|
C(u 0 , v0 ) = | |(u 0 , v0 ) = D(x, y) (u 0 , v0 ) /= 0.
xv yv | D(u, v)

Consider the vector function ϕ : Δ → R2 defined by:

ϕ(u, v) = (x(u, v), y(u, v)), ∀(u, v) ∈ Δ.

Since D(x,y) , v /= 0, from the local inversion theorem, it results that there are
D(u,v) (u 0 0 )
an open neighborhood U0 ⊂ Δ of the point (u 0 , v0 ) and an open neighborhood D
of the point (x0 , y0 ) = ϕ(u 0 , v0 ) such that the map ϕ : U0 → D is diffeomorphism
and D(x,y)
D(u,v) (u,
v) /= 0, ∀(u, v) ∈ U0 .
7.1 Parameterized Surface Canvases. Definition of a Surface 221

If we denote by u = u(x, y), v = v(x, y) the scalar components of the inverse

function ϕ−1 : D → U0 and by f (x, y) = z(u(x, y), v(x, y)), ∀(x, y) ∈ D, then
f is function of C 1 − class on D. On the other hand, since:

x(u 0 , v0 ) = x0 ;
y(u 0 , v0 ) = y0 ;
z(u 0 , v0 ) = z(u(x0 , y0 ), v(x0 , y0 )) = f (x0 , y0 )

we deduce that M0 ∈ G f = {(x, y, f (x, y)); (x, y) ∈ D}.

As G f ⊂ R3 is an elementary surface (Remark 7.1.4), it follows that the support
S = r (Δ) is a surface within the meaning of Definition 7.1.6.
Another way of analytic representation of a surface is the implicit representation.

Remark 7.1.6 Let D ⊂ R3 be an open subset and let F : D → R be a function of

( )2 ( )2 ( )2
C 1 − class on D. Moreover, if we assume that ∂∂ Fx + ∂∂Fy + ∂∂zF > 0 on D,
then S = {(x, y, z) ∈ D; F(x, y, z) = 0} is a surface called implicit surface.

Indeed, let M0 (x0 , y0 , z 0 ) ∈ S. Since the first-order partial derivatives of F are

not simultaneously zero at M0 , we can assume that ∂∂zF (M0 ) /= 0. From the implicit
function theorem, it follows that there are an open neighborhood U of x0 , an open
neighborhood V of y0 , an open neighbourhood W of z 0 and an unique implicit
function (x, y) → f (x, y) : U × V → W of C 1 − class with the properties:
(a) f (x0 , y0 ) = z 0 .
(b) F(x, y, f (x, y)) = 0, ∀(x, y) ∈ U × V .
If the set G f = {(x, y, f (x, y)); (x, y) ∈ U × V } represents the graph of the
implicit function f : U × V → W , from the property (a), it results that M0 ∈ G f ,
and from (b), it follows that G f ⊂ S. Since G f is an elementary surface, we deduce
that the set S = {(x, y, z) ∈ D ; F(x, y, z) = 0} is a surface within the meaning of
Definition 7.1.6.
In conclusion, a surface can be represented analytically in three ways: parametric,
explicit, or implicit. These three analytical representations are locally equivalent, in
the sense that any point on a surface belongs to an elementary surface that can be
represented in all three possible ways. In practice, all three analytical representations
are used, but the parametric representation proves to be more useful in theoretical
problems. It will be used mainly in the following.

Definition 7.1.7 A parameterized surface is any equivalence class of parameterized

surface canvases.

Therefore, ~
S is a parameterized surface if there is a parameterized surface canvas
r : Δ → R3 such that:
{ }
~
S = r1 : Δ1 → R3 parametrized surface canvas; (Δ1 , r1 ) ∼ (Δ, r ) .
222 7 Surface Integrals

Since (Δ, r ) ∼ (Δ, r ), we deduce that (Δ , r ) ∈ ~S.

A parameterized surface S̃ is called simple (smooth) if the surface canvas that
determines it is simple (smooth). The support of ~ S is the support S of any canvas
from ~
S. Usually, we will identify the surface ~
S with its support S.
Let S = r (Δ) be a smooth parameterized surface and let (u 0 , v0 ) ∈ Δ. Since

∥→
ru (u 0 , v0 ) × r→v (u 0 , v0 )∥2 = A2 (u 0 , v0 ) + B 2 (u 0 , v0 ) + C 2 (u 0 , v0 ) > 0

it results that the vectors −

→
r u (u 0 , v0 ) and −
→
r v (u 0 , v0 ) are non-collinear, thus
determine a plane.

Definition 7.1.7 The plane determined by the vectors − →r u (u 0 , v0 ) and −

→
r v (u 0 , v0 )
passing through the point M0 (x(u 0 , v0 ), y(u 0 , v0 ), z(u 0 , v0 )) of the surface S is
called the tangent plane to the surface S at M0 . The straight line perpendicular to
the tangent plane at the point M0 is called the normal line to the surface S at the
point M0 .

This plane does not depend on the local representation of S.

ϕ
Indeed, if we consider (s, t) ∈ Δ1 ⊂ R2 −→(u, v) ∈ Δ ⊂ R2 a diffeomorphism,
then r1 = r ◦ ϕ is another local representation of S. At the point (x0 , y0 , z 0 ), we
have:
∂u ∂v
r1 )s (s0 , t0 ) = r→u (u 0 , v0 ) ·
(→ + r→v (u 0 , v0 ) ·
∂s ∂s
∂u ∂v
r1 )t (s0 , t0 ) = r→u (u 0 , v0 ) ·
(→ + r→v (u 0 , v0 ) ·
∂t ∂t
and
( )
∂u ∂v ∂v ∂u D(u, v)
r1 )s × (→
(→ r1 )t = (→
ru × r→v ) · · − · = (→
ru × r→v ) · ,
∂s ∂t ∂s ∂t D(s, t)

whence, we deduce that the normal line to the surface S at the point (x0 , y0 , z 0 ) is
uniquely determined.
−
→ −
→ −
→
Since the vector −→
r u ×−→r v = A i + B j +C k is perpendicular to the tangent
plane, means it is collinear with the normal line to the surface, it follows that A, B, C
are the direction parameters of the normal line. Further, we denote by:

−
→ −
→ −
→ −
→
−
→ r u ×−
→
r v A i + B j +C k
n = ∥− ∥ = √
∥→
r u ×−
→
r v∥ A2 + B 2 + C 2

the unit normal vector to surface S.

In the case of an explicit surface, the direction parameters of the normal line to
surface are: − p, −q, 1 (Remark 7.1.4).
7.2 The Area of a Smooth Surface 223

7.2 The Area of a Smooth Surface

To begin with, we approach the problem of the area of an explicitly defined smooth
surface.
Let D ⊂ R2 be a squareable bounded domain and let f : D → R be a function
of C 1 − class on D. If we denote by p = ∂∂ xf and q = ∂∂ yf , then p, q are continuous
on D.
Let S (respectively, S) be the graph of the function f : D → R (respectively,
the graph of the function f : D → R). Thus:
{ }
S = {(x, y, f (x, y)); (x, y) ∈ D} and S = (x, y, f (x, y)); (x, y) ∈ D .

If we denote by ┌ = S \ S the boundary of the surface S and by C the boundary

of the domain D, then:

┌ = {(x, y, f (x, y)); (x, y) ∈ C}.

The sets D and C are the orthogonal projections to the plane O x y of the sets S
and ┌, respectively.
Let ρ : D1 , D2 , .., D p be an arbitrary partition of D and let Mi (xi , yi ) ∈ Di ,
1 ≤ i ≤ n, be some arbitrary points. We will denote by Pi ((xi , yi , f (xi , yi ))) the
corresponding points on the surface S (Fig. 7.4).
Let us denote by πi the tangent plane to S at the point Pi and by − → n i the unit
normal vector to S at Pi , oriented upward.
If we denote by γi the angle formed by the unit vector − →
n i with Oz axis, then:

1
cos γi = √
1 + pi2 + qi2

where pi = ∂∂ xf (xi , yi ) and qi = ∂∂ yf (xi , yi ).

Let Si (respectively, Ti ) be the portion from the surface S (respectively, from the
tangent plane πi ) cuts by the cylinder whose generatrix is parallel to Oz and whose
directrix curve is the boundary ∂ Di .
We will approximate the area of the surface portion Si by the area of the plane
Σn
portion Ti and the area of S by area(Ti ). Therefore:
i=1

area(Si ) ≈ area(Ti ) and area(S)

Σ
n Σ
n
= area(Si ) ≈ area(Ti ).
i =1 i =1
224 7 Surface Integrals

Fig. 7.4 The area of an

elementary surface γi
ni
Pi
S

Γ
z

O y
x Mi
Di
D C

Because γi is the angle between the plane πi and the coordinate plane O x y, from
the elementary geometry, it follows that:

area(Di ) = area(Ti ) · cos γi

and further:
√
area(Ti ) = 1 + pi2 + qi2 · area(Di ). (7.2)

From (7.2), we have:

Σ
n n √
Σ
area(Ti ) = 1 + pi2 + qi2 · area(Di ). (7.3)
i =1 i =1

Definition 7.2.1 It is called the area of the smooth surface S the following real
number:
n √
Σ
not
area(S) = area (S) = A = lim 1 + pi2 + qi2 · area(Di ). (7.4)
∥ρ∥→0
i=1
7.2 The Area of a Smooth Surface 225

The exact meaning is this: There is a finite real number A ∈ R+ with the property
that for any ε > 0, there is δε > 0 such that for any partition ρ : D1 , D2 , ... , D p
of D, with ∥ρ∥ < δε , and for any points Mi (xi , yi ) ∈ Di , we have:
| n √
|
| Σ |
| |
|A − 1 + pi + qi · area(Di )| < ε.
2 2
| |
i =1

Notice that the sum from the

√ right member of the relation (7.3) is the Riemann
sum of the function g(x, y) = 1 + p 2 (x, y) + q 2 (x, y), associated to the partition
ρ and the intermediate points Mi (xi , yi ) ∈ Di . Since g is continuous on D, thus
integrable on D, according to (7.4), it follows that:
n √
Σ
area(S) = lim 1 + pi2 + qi2 · area(Di ) = lim σρ (g; Mi ) =
∥ρ∥→0 ∥ρ∥→0
i=1
¨ ¨ √
= g(x, y)dx dy = 1 + p 2 (x, y) + q 2 (x, y)dx dy.
D D

Therefore, we proved the following result.

Theorem 7.2.1 Any smooth explicit surface S : z = f (x, y), (x, y) ∈ D, where
D ⊂ R2 is a bounded squareable domain and f : D → R is a function of C 1 −
class on D, has area given by the formula:
¨ √
area(S) = 1 + p 2 (x, y) + q 2 (x, y)dxdy. (7.5)
D

R2 2
Example 7.2.1 Compute the area of the surface S : x 2 + y 2 = h2
z , 0 ≤ z ≤ h.

Geometrically, the surface S represents a cone with the vertex at the origin, the
radius R and the height h (Fig. 7.5). Notice also that the surface S is the graph of the
function:
h √ 2
f (x, y) =
x + y 2 , (x, y) ∈ D
R
{ }
where D = (x, y) ∈ R2 ; x 2 + y 2 < R 2 .
Next, we have:

∂f h x ∂f h y
p= = · √ ; q= = · √
∂x R x +y
2 2 ∂y R x + y2
2

and

R2 + h2
1 + p2 + q 2 = .
R2
226 7 Surface Integrals

Fig. 7.5 A cone with the z

vertex at the origin, the
radius R and the height h
h

O y
x

Taking into account to formula (7.5), the area of the surface S is expressed by:
¨ √ √ ¨
R2 + h2 R2 + h2
area(S) = dx dy = ·1 dx dy =
R2 R
√D √ D
R2 + h2 R2 + h2
= · area(D) = · π R2 = π · R · G
R R
√
where we denoted by G = R 2 + h 2 the generatrix of the cone.
Next, we study the area of a simple smooth parameterized surface.
If S is a simple smooth parameterized surface, then S is the support of a simple
smooth parameterized surface canvas. Therefore, there is a vector function r : Δ ⊂
R2 → R3 ,

r (u, v) = (x(u, v), y(u, v), z(u, v)), ∀ (u, v) ∈ Δ

such that:

S = r (Δ) = {(x(u, v), y(u, v), z(u, v)); (u, v) ∈ Δ}.

∥→ ∥2
Obviously, r is of C 1 − class on Δ, ∥−
r u ×−
→
r v ∥ = A2 +B 2 +C 2 > 0, ∀(u, v) ∈
Δ, and r : Δ → S is bijective. Also, if the vector function r is extendable by
continuity to Δ, we will denote by:

S = r (Δ) = {(x(u, v), y(u, v), z(u, v)); (u, v) ∈ Δ}.

If we denote by ┌ = S \ S the boundary of S and by C the boundary of Δ, then:

┌ = r (C) = {(x(u, v), y(u, v), z(u, v)), (u, v) ∈ C}.

7.2 The Area of a Smooth Surface 227

We mention that, in general, the correspondence between ┌ and C is not bijective.

The surface S is called closed if S = S. A closed parameterized surface has no
boundary.

Example 7.2.2. Let S be the parameterized surface having the representation:

r (u, v) = (R sin u cos v, R sin u sin v, R cos u), (u, v) ∈ Δ = (0, π ) × (0, 2π ).

The parametric equations are:

⎧
⎨ x = R sin u cos v
y = R sin u sin v, u ∈ (0, π ), v ∈ (0, 2π ) .
⎩
z = R cos u

We notice that r (0 , v) = (0 , 0 , R) , ∀ v ∈ [0 , 2 π]. Therefore, the image

of any point on the segment AE through the vector function r is the point P(0, 0, R)
(Figs. 7.6 and 7.7). Similarly, the image of any point on the segment B F is the point
P ' (0, 0, −R).
On the other hand, the image of any point M ∈ AB ∪ E F will be a point of
coordinates:
x = R sin u, y = 0, z = R cos u , u ∈ [0 , π].

Fig. 7.6 The variation v

domain of parameters u and v E ( 0, 2 π ) F (π , 2 π )

A B (π , 0 ) u

Fig. 7.7 The sphere of z

radius R centered at the P ( 0, 0, R )
origin, without the meridian
P Q P'

O y
Q ( R , 0, 0 )

x P ' ( 0, 0, − R )
228 7 Surface Integrals

Since x 2 + y 2 + z 2 = R 2 and x ≥ 0, it follows that the image of the boundary of

the domain Δ through the vector function r is the meridian P Q P ' on the sphere of
radius R centered at the origin. Therefore:
S = r (Δ) is the sphere of radius R centered at the origin, without the meridian
P Q P '. ( )
S = r Δ is the sphere of radius R centered at the origin.
The boundary of S is ┌ = S\S = P Q P ' .

Definition 7.2.2 Let S be a simple and smooth parameterized surface and let

r (u, v) = (x(u, v), y(u, v), z(u, v)), ∀(u, v) ∈ Δ s

be a parametric representation of S. Then the surface S is squareable and its area

is described by the expressions:
¨ √
area(S) = area(S) = E · G − F 2 du dv =
Δ
¨ √ ¨
= A2 + B 2 + C 2 du dv = ∥→
ru × r→v ∥du dv. (7.6)
Δ Δ

Remark 7.2.1 It can be shown that the area of a smooth parameterized surface does
not depend on the chosen parameterization.
ϕ
Indeed, if we consider (s , t) ∈ Δ1 ⊂ R2 −→ (u , v) ∈ Δ ⊂ R2 a diffeomor-
phism, then r1 = r ◦ ϕ is another local representation of S. From Definition
7.1.7, we have:
| |
| D(u, v) |
∥(→ r1 )t ∥ = ∥→
r1 )s × (→ |
ru × r→v ∥ · | |.
D(s, t) |

The assertion follows using the change of variable ϕ in the last term of (7.6).

Remark 7.2.2 Formula (7.6) generalizes formula (7.5) which expresses the area of
an explicit surface.

(Indeed,
) let the explicit surface S : z = f (x, y), (x, y) ∈ D ⊂ R2 with f ∈
C1 D .
From Remark 7.1.4, we have A = − p, B = −q, C = 1, whence, it results that:
¨ √
area(S) = 1 + p 2 + q 2 dxdy.
D

Example 7.2.3 Compute the area of the sphere S : x 2 + y 2 + z 2 = R 2 .

A parametric representation of the sphere is:

7.2 The Area of a Smooth Surface 229
⎧
⎨ x = R sin u cos v
y = R sin u sin v, (u, v) ∈ Δ = (0, π ) × (0, 2π ) .
⎩
z = R cos u

As shown in Example 7.1.1, in this case we have:

E · G − F 2 = A2 + B 2 + C 2 = R 4 sin2 u

hence, applying formula (7.6), we obtain:

¨ ∫ 2π ∫ π
area(S) = R 2 sin u dudv = R 2 dv sin u du
Ʌ 0 0
| π
= − 2π R 2 cos u |0 = 4π R 2 .

Example 7.2.4 Find the area of the torus.

Let us consider a circle of radius a centered at the point (b, 0), where 0 < a < b,
lying in the plane O x y. The torus is the surface T obtained by revolving, in space,
this circle around the axis O y.
If θ is the angle in Fig. 7.8 and ϕ is the angle of rotation of the circle around the
axis O y, then the parametric equations of the torus are (Fig. 7.9):
⎧
⎨ x = (b + a cos θ ) cos ϕ
y = a sin θ, (θ, ϕ) ∈ Δ = (0, 2π ) × (0, 2π ), a, b > 0 .
⎩
z = (b + a cos θ ) sin ϕ

Further, we have:

xθ = −a sin θ cos ϕ, xϕ = −(b + a cos θ ) sin ϕ

yθ = a cos θ, yϕ = 0
z θ = −a sin θ sin ϕ, z ϕ = (b + a cos θ ) cos ϕ
E = a 2 , F = 0, G = (b + a cos θ )2
E · G − F 2 = a 2 (b + a cos θ )2 .

Fig. 7.8 The circle centered y

in (b, 0) and of radius a
a
θ
O b x
230 7 Surface Integrals

¨ ∫2π ∫2π
area(T ) = a(b + a cos θ )dθ dϕ = a dϕ (b + a cos θ )dθ =
Δ 0 0
= 2π a(bθ + a sin θ )|2π
0 = 4π ab.
2

7.3 Surface Integral of the First Kind

Let S be a simple smooth parameterized surface that is given by the parameterization

r (u, v) = (x(u, v), y(u, v), z(u, v)), ∀(u, v) ∈ Δ

( )
and let F : S = r Δ → R be a continuous function.
Of course, we suppose that Δ ⊂ R2 is a squareable bounded domain.

Definition 7.3.1 The surface

˜ integral of the first kind of the function F over the
surface S is denoted by F(x, y, z)dσ and is defined by the following double
S
integral:
¨
F(x, y, z)dσ =
S
¨ √
= F(x(u, v), y(u, v), z(u, v)) E(u, v) · G(u, v) − F 2 (u, v)du dv. (7.7)
Δ

˜
Remark 7.3.1 The surface integral of the first kind F(x, y, z)dσ is independent
S
of its parameterization, in the sense that for two equivalent parameterizations, the
value of the integral is the same, as a result of Theorem 6.4.1 of changing variables
in the double integral.

Remark 7.3.1 Under the above conditions, if F(x, y, z) = 1, (x, y, z) ∈ S, then,

from (7.6), it results that:
¨ ¨ √
1 dσ = E(u, v) · G(u, v) − F 2 (u, v)dudv = area(S).
S Δ

˜ ( )
Example 7.3.1 Compute S x − y + z 2 dσ , where the surface S is given by:

S : x 2 + y 2 + z 2 = R 2 , z > 0.
7.3 Surface Integral of the First Kind 231

The surface S represents the upper half of the sphere of radius R centered at the
origin. A parametric representation of this surface is:
⎧
⎨ x = R sin u cos v ( )
y = R sin u sin v, (u , v) ∈ 0, π2 × (0, 2 π )
⎩
z = R cos u

(see Example 7.1.1).

Taking into account that in this case E · G − F 2 = R 4 sin2 u, from formula (7.7),
we get:
¨
( )
x − y + z 2 dσ =
S
¨
( )
= R sin u cos v − R sin u sin v + R 2 cos2 u R 2 sin u du dv =
Δ
∫ π (∫ 2π )
2 ( 2 )
=R 3
sin u cos v − sin u sin v + R sin u cos u dv du =
2 2
0 0
∫ π |π
2
4 cosu || 2
3
2π R 4
= 2π R 4
sin u cos u du = − 2π R
2
| = .
0 3 0 3

Remark 7.3.2 If S : z = f (x , y) , (x , y) ∈ D is an explicit smooth surface,

where D ⊂ R2 is a squareable bounded domain, f : D → R is of C 1 − class on
D, and F : S → R is continuous, then:
¨
F(x, y, z)dσ =
S
¨ √
= F(x, y, f (x, y)) 1 + p 2 (x, y) + q 2 (x, y)dx dy. (7.8)
D

˜
Example 7.3.2 Compute S (x + z)dσ , where the surface S is the portion of the
√
cone z = x 2 + y 2 that lies inside the circular cylinder given by x 2 + y 2 = 2 y.

We notice that the projection of the surface S to the plane O x y is the domain:
{ } { }
D = (x, y) ∈ R2 ; x 2 + y 2 − 2y ≤ 0 = (x, y) ∈ R2 ; x 2 + (y − 1)2 ≤ 1

√ 1 centered at the point (0 , 1); Fig. 7.10).

(i.e., the disk of radius
Therefore, S : z = x 2 + y 2 , (x, y) ∈ D. Further, we have:

∂z x ∂z y
p= = √ ; q= = √ ; 1 + p 2 + q 2 = 2.
∂x x +y
2 2 ∂ y x +y
2 2
232 7 Surface Integrals

Fig. 7.9 The portion of a z

cone that lies inside a
circular cylinder

O
y
D
x

Fig. 7.10 The disk of radius y

1 centered at the point (0,1)
2

D 1

O x

According to formula (7.8), we have:

¨ ¨ ( √ ) √
I = (x + z)dσ = x+ x 2 + y 2 · 2dx dy.
S D

Since we are integrating over a disk, it makes sense to use polar coordinates:
∫
x = ρ cos θ
, 0 ≤ ρ ≤ 2 sin θ, 0 ≤ θ ≤ π
y = ρ sin θ

and therefore, we get:

⎛ 2 sin θ ⎞
√ ∫ ∫
π

I = 2 ⎝ (ρ cos θ + ρ)ρ dρ ⎠dθ =

0 0
|
√ ∫
π
ρ 3 ||2 sin θ
= 2 (cos θ + 1) · | dθ
3 0
0
√ ∫π
8 2 ( 3 )
= sin θ cos θ + sin3 θ dθ =
3
0
7.3 Surface Integral of the First Kind 233

√ ∫π √ ∫π
8 2 8 2 ( )
= sin θ dθ =
3
1 − cos2 θ sin θ dθ =
3 3
0 0
√ ( 3 )| √
8 2 cos θ | 32 2
= · |
− cos θ | =
π .
3 3 0 9

Remark 7.3.2 If the surface S is a piecewise smooth surface, i.e., it is a finite union
of simple smooth surfaces Si , i = 1 , n, such that two by two have no common
interior points, then:

Σ
n ¨ n ¨
Σ
area(S) = area(Si ) and F(x, y, z)dσ = F(x, y, z)dσ .
i=1 S i=1 S
i

Remark 7.3.3 The surface integral of the first kind is frequently used in engineering
applications. For example, having a thin shell which is a material surface S, smooth
or piecewise smooth, whose variable density is given by the continuous positive
function F : S → R+ , then the mass, the center of mass and the moments of
inertia of the surface are expressed through the surface integrals of the first kind as
follows.

1. The Mass of the Shell S is Given by:

¨
Mass(S) = F(x, y, z)dσ .
S

2. The Coordinates of the Center of Mass of the Shell S Are Defined by:
˜ ˜
x F(x, y, z)dσ y F(x, y, z)dσ
xG = ˜ yG = ˜
S S
,
F(x, y, z)dσ F(x, y, z)dσ
S S
˜
z F(x, y, z)dσ
zG = ˜
S
.
F(x, y, z)dσ
S

3. The moments of inertia of the thin shell S with respect to the origin O(0, 0, 0),
with respect to the coordinate axes O x, O y, Oz and with respect to the coordinate
planes O x y, O x z, O yz are given by the formulas:
¨
( )
IO = x 2 + y 2 + z 2 F(x, y, z)dσ
S
234 7 Surface Integrals
¨
( )
IOx = y 2 + z 2 F(x, y, z)dσ ,
S
¨
( )
IOy = x 2 + z 2 F(x, y, z)dσ
S
¨
( )
IOz = x 2 + y 2 F(x, y, z)dσ
S
¨
IOx y = z 2 F(x, y, z)dσ ,
S
¨
IOx z = y 2 F(x, y, z)dσ
S
¨
IO yz = x 2 F(x, y, z)dσ .
S

7.4 Surface Integral of the Second Kind

Before we really define the surface integral of the second kind, we first need to
introduce the idea of an oriented surface, similar to an oriented curve.
Let S be a smooth surface and let
−
→ −
→ −
→ −
→
r (u, v) = x(u, v) i + y(u, v) j + z(u, v) k , ∀(u, v) ∈ Δ

be one of its parametric representation.

−
→
Because the surface is smooth, it follows that −
→
r u ×−
→r v /= 0 , for any (u, v) ∈ Δ.
We notice that at each interior point M(x(u, v), y(u, v), z(u, v)) ∈ S, there are
two unit normal vectors to S, namely − →n (M) and −− →
n (M), where:
−
→r u ×−
→
−
→
n (M) = ∥−
r v
∥.
→ −
→
∥ r u × r v∥

Definition 7.4.1 The smooth surface S is said to be orientable (two-sided) if the

map M → − →n (M) : S → R3 is continuous.

It is obvious that if the application M → −→

n (M) : S → R3 is continuous, then
−
→
also the map M → − n (M) : S → R is continuous. If a surface is orientable,
3

then choosing a side of this surface returns to choose one of the two continuous
applications M → ±− →n (M).
Therefore, we have two possible orientations (or two sides) of the surface, namely:
7.4 Surface Integral of the Second Kind 235
( →) ( )
S+ = S, −
n and S− = S, −−
→
n

Remark 7.4.1 The property of the mapping M → − →n (M) : S → R3 to be

continuous on the orientable surface S is a global property and it covers the entire
surface.

This implies the following property: let M0 ∈ S be an arbitrary interior point and
let C be an arbitrary closed curve lying on the surface S, which passes through the
point M0 and has no common points with the boundary of S; let us suppose that
we choose an orientation to the normal line to S at M0 , and that is the orientation
given by the unit normal vector −→n (M0 ). Imagine that we are moving the unit normal
−
→
vector n (M) on the curve C, starting from the point M0 , in such a way the vector
always remains perpendicular to the surface, we return to the point M0 with the same
orientation of the normal, that is:

lim −
→
n (M) = −
→
n (M0 ).
M → M0
M ∈C

If there is a closed curve lying on the surface S such that after it has been traversed,
the unit normal vector changes its orientation to the opposite, the surface is called
one-sided.

Example 7.4.1 The simplest orientable surface is the plane. Also, quadric surfaces
(sphere, ellipsoid, hyperboloids, paraboloids, cone, cylinders, pairs of planes) are
orientable surfaces. When the surface is closed, so when it is the boundary of a
bounded spatial domain, then it is orientable and its two faces are called: the outer
face and the inner face.

Example 7.4.2 Any elementary surface is an orientable surface.

Indeed, let

−
→ −
→ −
→ −
→
r (u, v) = x(u, v) i + y(u, v) j + z(u, v) k , ∀(u, v) ∈ Δ

be a parametric representation of an elementary surface S.

We notice that the application M → − →n (M) : S → R3 is continuous on S,
because it is the composition of continuous functions r − 1 : S → Δ and (u, v) →
r u ×−
−
→ →
∥−
r v
r u ×−
→ →
r v∥
: Δ → R3 .
In particular, any explicit smooth surface is orientable having the sides: the upper
side which corresponds to the normal oriented upward (which forms an acute angle
with the positive half-axis Oz) (Fig. 7.11) and the lower side which corresponds
to the normal oriented downward (which forms an obtuse angle with the positive
half-axis Oz) (Fig. 7.12).
236 7 Surface Integrals

Fig. 7.11 The orientation of γ

z
the upper side of an explicit
smooth surface

O y
x

Fig. 7.12 The orientation of z

the lower side of an explicit
smooth surface γ

O y
x

{ }
Example 7.4.3 The sphere S = (x, y, z) ∈ R3 ; x 2 + y 2 + z 2 = R 2 is orientable
closed surface; thus, it has two faces: the outer face corresponding to the unit normal
vector that point outward and the inner face corresponding to the unit normal vector
facing inward.

Indeed, for any point M(x , y , z) on the sphere, the unit vector of the outer
−−→
n (M) = R1 O M. Clearly, the map M → −
normal is −
→ →n (M) : S → R3 is continuous
on S.

Example 7.4.4 The classic example of a non-orientable (one-sided) surface is the

Möbius strip. A model of this surface is obtained by twisting a rectangular strip
of paper ABC D such that the point A coincides with the point C and B with D
(Fig. 7.13).

It is easily seen that after we traverse the middle line E F of the Mőbius strip, the
orientation of the normal to the surface is reversed.

Fig. 7.13 The Möbius strip

7.4 Surface Integral of the Second Kind 237

Fig. 7.14 The outer face of z

the sphere centered in the
γ n
origin and of radius R
u
M
u' y
O
γ'
x M'
n

Fig. 7.15 The lower side of z

the cone with vertex in the
origin h

O y
x

Fig. 7.16 A simple domain z

γ n
with respect to the three
coordinate axes
S2

S3 γ

S1 n
O y
D
∂D
x
238 7 Surface Integrals

Let S be a simple smooth orientable parameterized surface and let

−
→ −
→ −
→ −
→
r (u, v) = x(u, v) i + y(u, v) j + z(u, v) k , ∀(u, v) ∈ Δ

be one of its parametric representation, where Δ ⊂ R2 is a squareable bounded ( →)

domain and x, y, z are functions of C 1 − class on Δ. We will denote by S+ = S, − n
−
→ −
→
the face of the surface corresponding to the unit normal vector −→n = ∥− r u× r v
r u ×−
→ →r v∥
.
−
→
We also consider a continuous vector field F = (P, Q, R) : Ω ⊂ R → R3 , 3

where Ω ⊂ R3 is a space domain that contains the surface S.

Definition 7.4.2 The surface integral of the second kind of the vector field
−
→ ( →)
F = (P, Q, R) over the face S+ = S, −n of the surface S, means the following
expression:
¨
P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy =
S+
¨
= F→ · n→dσ =
S
¨
= (P(x, y, z) cos α + Q(x, y, z) cos β + R(x, y, z) cos γ )dσ (7.9)
S

where α, β, γ are the angles formed by the unit normal vector − →n to the surface S
with the positive directions of the coordinate axes O x, O y, Oz, respectively, i.e.:

n→(x, y, z) = cos α(x, y, z)→i + cos β(x, y, z) →j + cos γ (x, y, z)k,

→ (x, y, z) ∈ S.

If we use the usual notations presented in paragraph 7.1, then:

| | | | | |
| yu z u | | xu z u | | xu yu |
|
A=| | |
; B = −| | |
; C =| |
yv z v | xv z v | xv yv |
A
cos α = ± √
A2 + B2 + C 2
B
cos β = ± √
A2 + B 2 + C 2
C
cos γ = ± √ .
A2 + B2 + C 2

Note that we must have to choose the +" or "−" sign in front of the square root
to match the orientation of the normal to the surface.
7.4 Surface Integral of the Second Kind 239

Taking into account the computation of the surface integral of the first kind, the
formula (7.9) becomes:
¨
P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy =
S+
¨
=± [P(x(u, v), y(u, v), z(u, v)) · A(u, v)+
Δ
+ Q(x(u, v), y(u, v), z(u, v)) · B(u, v)+
+R(x(u, v), y(u, v), z(u, v)) · C(u, v)]du dv. (7.10)

Remark 7.4.2 The surface integral of the second kind depends on the surface
orientation.
( )
Indeed, if S− = S, −−
→
n is the other face of the surface S, then:
¨
P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy =
S−
¨ ¨
−
→ −
→
= F · (−→
n )dσ = − F · n→dσ =
S S
¨
=− P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy.
S+

Remark 7.4.3 In physical applications, the surface integral of the second kind
˜ − → −
→
Ss F ·n →dσ is often referred to as the total flux ϕ of the vector field F through the
−
→
face S+ (respectively, S− ) of the surface S. Specifically, let us suppose that F is the
particle velocities field of a fluid in stationary flow (the velocities of fluid particles
˜ − →
depend on space positions but not on the time). Then, ϕ = S F · n→dσ represents
the volume of fluid flowing through the surface S per time unit, in the direction of
the unit normal vector that defines the selected face.
˜
Example 7.4.5 Compute x 2 dy ∧ dz + y 2 dz ∧ dx + zdx ∧ dy,,
S+

where S+ is the outer face of the sphere x 2 + y 2 + z 2 = R 2 .

The
⎧ parametric equations of the sphere are:
⎨ x = R sin u cos v
y = R sin u sin v , u ∈ [0, π ], v ∈ [0, 2π ].
⎩
z = R cos u
From Example 7.1.1, we have:

A = R 2 sin2 u cos v, B = R 2 sin2 u sin v, C = R 2 sin u cos u

240 7 Surface Integrals

A2 + B 2 + C 2 = R 4 sin2 u, hence:

cos α = ± sin u cos v, cos β = ± sin u sin v, cos γ = ± cos u. (7.11)

Since S+ is the outer face of the sphere, we notice that for the normal to the
outward-facing sphere ( we ) must choose the sign “+” in the formulas (7.11).
Indeed, if u ∈ 0, π2 , then the corresponding point M on the sphere is on the
upper hemisphere and the outer normal will describe a acute angle γ with the axis
Oz (cos γ = ( cos)u > 0) (Fig. 7.14).
If u ' ∈ π2 , π , then the corresponding point M ' on the sphere is on the lower
hemisphere( and' the outward-facing
) normal will make an obtuse angle γ ' with the
'
axis Oz cos γ = cos u < 0 .
From the formula (7.10), it results that:
¨
x 2 dy ∧ dz + y 2 dz ∧ dx + zdx ∧ dy =
S+
⎛ ⎞
∫2π ∫π
( 4 4 )
= ⎝ R sin u cos3 v + R 4 sin4 u sin3 v + R 3 sin u cos2 u du ⎠dv =
0 0
⎛ 2π ⎞ ⎛ π ⎞
∫ ∫
= R 4 ⎝ cos3 vdv ⎠ · ⎝ sin4 udu ⎠
0 0
⎛ 2π ⎞ ⎛ π ⎞
∫ ∫
+ R 4 ⎝ sin3 v dv ⎠ · ⎝ sin4 u du ⎠+
0 0
⎛ ⎞
∫2π ∫π
+ R3 ⎝ sin u cos2 u du ⎠dv.
0 0

∫ 2π ∫ 2π ( ) ( )||
Since we have 0 cos vdv = 0 1 − sin2 v cos vdv = sin v −
3 sin3 v |2π =
3 |0
∫ 2π
0, and similarly, 0 sin3 vdv = 0, then it follows that:
¨
x 2 dy ∧ dz + y 2 dz ∧ dx + zdx ∧ dy =
S+
⎛ ⎞
∫2π ∫π |
cos3 u ||π 4π R 3
= R 3 ⎝ sin u cos2 u du ⎠dv = −2π R 3 · = .
3 |0 3
0 0
7.4 Surface Integral of the Second Kind 241

Remark 7.4.1 If the smooth surface is explicitly given by

S : z = f (x, y), (x, y) ∈ D

where D ⊂ R2 is a squareable bounded domain and f : D → R is a function of

C 1 − class on D, then we obtain

−p −q 1
cos α = √ ; cos β = √ ; cos γ = √ .
± 1+ p +q
2 2 ± 1+ p +q
2 2 ± 1 + p2 + q 2

If S+ is the upper side of the surface, corresponding to the normal oriented upward,
then cos γ > 0 and we choose the sign “+” in front of the square root from the
denominator.
For the lower side S− , the unit normal vector will point downward, hence cos γ <
0 and we will choose the sign “−” in front of the square root.
In this case, formula (7.9) becomes:
¨
P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy =
S+
¨
= (− p · P(x, y, f (x, y)) − q · Q(x, y, f (x, y))+
D
+ R(x, y, f (x, y)))dxdy. (7.12)

Example 7.4.6 Compute the total flux ϕ of the vector field:

−
→ −
→ −
→ −
→
V (x, y, z) = (y + z) i − (z + x) j − (x + y) k

through the lower side of the cone S : x 2 + y 2 = z 2 , 0 ≤ z ≤ h.

−
→
The total flux of the vector field V through the lower face S− of the cone is given
by (Fig. 7.15):
¨
ϕ= (y + z)dy ∧ dz − (z + x)dz ∧ dx − (x + y)dx ∧ dy.
S−

The surface S is explicitly defined as:

√
S:z= x 2 + y 2 , (x, y) ∈ D,

where
{ }
D = (x, y) ∈ R2 ; x 2 + y 2 ≤ h 2 .
242 7 Surface Integrals

Also, we have:

∂z x ∂z y √ √
p= = √ ; q= = √ ; 1 + p 2 + q 2 = 2.
∂x x 2 + y2 ∂y x 2 + y2

Since S− is the lower side of the cone, then cos γ < 0 and further, we have:
cos γ = −√1
; cos α = √ √x 2 2 ; cos β = √ √y 2 2 .
2 2· x +y 2· x +y
The total flux becomes:
¨ ( )
x
ϕ= (y + z) √ √
2 · x 2 + y2
S
)
y 1
−(z + x) √ √ − (x + y) √ dσ =
2 · x 2 + y2 − 2
⎛( √ ) x
¨ y + x 2 + y 2 √ √
⎜ 2 · x 2 + y2
= ⎜ (√ )
⎝ y
D
− x 2 + y 2 + x √ √ +
2 · x 2 + y2
⎛ ⎞
) ¨ ∫2π ∫h
x+y √
+ √ 2dxdy = 2 xdxdy = 2 ⎝ ρ 2 cos θ dρ ⎠dθ =
2
D 0 0
∫2π
2h 3
= cos θ dθ = 0.
3
0

7.5 Integral Formulas

Integral formulas are formulas for linking different types of integrals. Such a formula
has already been presented in Chap. 6, namely the Riemann–Green formula, which
establishes the connection between the double integral on a domain and the line
integral of the second kind on the boundary of the domain.
The analogous formula for three-dimensional space is the Gauss–Ostrogradsky
formula, which establishes an important link between the triple integral and the
surface integral of the second kind and is used, in particular, for the computation of
integrals on closed surfaces.

Theorem 7.5.1 (Gauss–Ostrogradsky formula).

Let T ⊂ R3 be a domain simple with respect to the three coordinate axes and
let P , Q, R : T → R be three continuous functions with the property that there
7.5 Integral Formulas 243

are partial derivatives ∂∂ Px , ∂∂Qy , ∂∂Rz continuous on T . We also assume that the closed
surface S = T \T that bounded the domain T is a piecewise smooth orientable
surface and S+ is its outer face. Then we have the formula:
¨
P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy =
S+
˚ ( )
∂P ∂Q ∂R
= (x, y, z) + (x, y, z) + (x, y, z) dxdydz. (7.13)
∂x ∂y ∂z
T

Proof Since the domain T ⊂ R3 is simple with respect to O z axis, it follows

that there are a squareable bounded domain D ⊂ R2 and two continuous functions
ϕ, ψ : D → R, with the property ϕ(x) < ψ(x), ∀(x, y) ∈ D such that:
{ }
T = (x, y, z) ∈ R3 ; ϕ(x, y) < z < ψ(x, y), ∀(x, y) ∈ D .

We denote by S1 the graph of the function z = ϕ(x, y), (x, y) ∈ D, respectively

by S2 the graph of the function z = ψ(x, y), (x, y) ∈ D, and by S3 the lateral
cylindrical surface with the generators parallel to O z axis and the directrix curve
∂ D. We notice that the surface S = S1 ∪ S2 ∪ S3 is the boundary of the domain T .
The assumption that S is piecewise smooth means that ϕ and ψ are of C 1 − class on
D.
The outer face of the surface S means the face corresponding to the normal facing
outward (Fig. 7.16).
This means for S1 the lower face surface, and for S2 the upper face surface.
Because, for the lower face S1 of the surface, the angle formed by O z axis with
the downward-facing normal is obtuse, it results cos γ < 0, hence:

1
cos γ = − √ ( )2 ( )2 .
∂ϕ ∂ϕ
1+ ∂x
+ ∂y

Furthermore, we have:
¨ ¨
−1
R(x, y, z)dx ∧ dy = R(x, y, z) · √ ( )2 ( )2 dσ =
∂ϕ ∂ϕ
(S1 )− S1 1+ ∂x
+ ∂y
√ ( )2 ( )2
¨
R(x, y, ϕ(x, y)) ∂ϕ ∂ϕ
=− √ ( )2 ( )2 · 1+ + dxdy =
∂x ∂y
D 1 + ∂ϕ
∂x
+ ∂ϕ∂y
¨
=− R(x, y, ϕ(x, y))dxdy. (7.14)
D
244 7 Surface Integrals

Similarly, for the upper face of the surface S2 , cos γ > 0, hence:
¨ ¨
1
R(x, y, z)dx ∧ dy = R(x, y, z) · √ ( )2 ( )2 dσ =
∂ψ ∂ψ
(S2 )+ S2 1+ ∂x
+ ∂y
√ ( )2 ( )2
¨
R(x, y, ψ(x, y)) ∂ψ ∂ψ
= √ ( )2 ( )2 · 1+ + dxdy =
∂x ∂y
D 1 + ∂ψ∂x
+ ∂ψ∂y
¨
= R(x, y, ψ(x, y))dxdy. (7.15)
D

For the outer face of the lateral cylindrical surface S3 , cos γ = 0, because γ = π2 .
It turns out that:
¨ ¨
R(x, y, z)dx ∧ dy = R(x, y, z) cos γ dσ = 0. (7.16)
(S3 )+ S3

Therefore, we have:
¨ ¨
R(x, y, z)dx ∧ dy = R(x, y, z)dx ∧ dy+
S+ (S1 )−
¨
+ R(x, y, z)dx ∧ dy
(S2 )+
¨
+ R(x, y, z)dx ∧ dy =
(S3 )+
¨
= R(x, y, ψ(x, y))dxdy
D
¨
− R(x, y, ϕ(x, y))dxdy. (7.17)
D

On the other hand, from Theorem 6.10.1, it results that:

⎛ ⎞
˚ ¨ ∫
ψ(x,y)
∂R ⎜ ∂R ⎟
(x, y, z)dxdydz = ⎝ (x, y, z)dz ⎠dxdy =
∂z ∂z
T D ϕ(x,y)
¨ |
|
= R(x, y, z)||ψ(x,y) dxdy =
ϕ(x,y)
D
7.5 Integral Formulas 245
¨
= R(x, y, ψ(x, y))dxdy
D
¨
− R(x, y, ϕ(x, y))dxdy. (7.18)
D

From Eqs. (7.17) and (7.18), we deduce:

˚ ¨
∂R
(x, y, z)dxdydz = R(x, y, z)dx ∧ dy. (7.19)
∂z
T S+

Similarly, using the fact that the domain T is simple also with respect to O y axes,
respectively O x, we get:
˚ ¨
∂Q
(x, y, z)dxdydz = Q(x, y, z)dz ∧ dx (7.20)
∂y
T S+

and, respectively:
˚ ¨
∂P
(xs, y, z)dxdydz = P(x, y, z)dy ∧ dz. (7.21)
∂x
T S+

Finally, adding Eq. (7.19), (7.20) and Eq. (7.21), we obtain the Gauss–Ostro-
gradsky formula:
˚ ( )
∂P ∂Q ∂R
(x, y, z) + (x, y, z) + (x, y, z) dxdydz =
∂x ∂y ∂z
T
¨
= P(x, y, z)dy ∧ dz + Q(x, y, z)dz ∧ dx + R(x, y, z)dx ∧ dy.
S+

Remark 7.5.1 Among the examples of simple domains with respect to the three
coordinate axes we mention: the sphere, the ellipsoid, the rectangular parallelepiped
with the edges parallel to the axes, and so on. Without going into details, we mention
that the Gauss–Ostrogradsky formula also remains valid for domains T that are
finite unions of simple domains with respect to the three coordinate axes, two by
two having in common at most piecewise smooth surfaces. By writing the Gauss–
Ostrogradsky formula for each of the simple domains Ti that make up the domain
T , adding these formulas and using the additivity property of the triple integral and
the surface integral, the Gauss–Ostrogradsky formula for the domain T is obtained.
This is explained by the fact that the surface integral oven an intersection surface of
two neighboring simple domains appears in the sum of the right side twice, once on
246 7 Surface Integrals

the upper face and once on the lower face, so its contribution to the right side is zero.
Thus, only the integral on the outer face of the domain T remains in the right side.

Remark 7.5.2 Taking into account the connection between the surface integral of
the second kind and the surface integral of the first kind, the Gauss–Ostrogradsky
formula is also written:
¨ ˚ ( )
∂P ∂Q ∂R
(P cos α + Q cos β + R cos γ )dσ = + + dxdydz
∂x ∂y ∂z
S T
(7.22)

where α, β, γ are the angles that the outer normal to the surface makes with the
positive directions of the coordinate axes O x, O y and Oz.
−
→
Remark 7.5.3 If we consider the vector field V : T ⊂ R3 → R3 of C 1 − class on
T having the components P, Q, R, then:

−
→ −
→ −
→ −
→ → ∂P
− ∂Q ∂R
V = P · i + Q · j + R · k and div V = + + .
∂x ∂y ∂z
−
→ −
→ −
→
Let also −
→n = cos α · i + cos β · j + cos γ · k be the unit normal vector of
the outer normal to the surface S.
With these notations, the Gauss–Ostrogradsky formula becomes:
¨ ˚
−
→ − −
→
V ·→
n dσ = div V dxdydz (7.23)
S T

and in this form it is also called the flux—divergence formula, because it shows that
−
→
the flux of the vector field V through the surface S, according to the outer normal,
is equal to the triple integral of its divergence on the domain T .

Example 7.5.1 Using the Gauss–Ostrogradsky formula compute the integral.

¨
x 3 dy ∧ dz + y 3 dz ∧ dx + z 3 dx ∧ dy,
S+

where S+ is the outer face of the cube:

{ }
T = (x, y, z) ∈ R3 ; 0 ≤ x ≤ a, 0 ≤ y ≤ a, 0 ≤ z ≤ a .

First, we denote by P(x, y, z) = x 3 , Q(x, y, z) = y 3 , R(x, y, z) = z 3 and from

the Gauss–Ostrogradsky formula (7.13), we deduce:
7.5 Integral Formulas 247
¨
x 3 dy ∧ dz + y 3 dz ∧ dx + z 3 dx ∧ dy =
S+
˚
( 2 )
= 3x + 3y 2 + 3z 2 dxdydz
T
∫a ∫a ∫a
( )
=3 dx dy x 2 + y 2 + z 2 dz =
0 0 0
∫a ∫a ( )|
z 3 ||a
=3 dx x z+y z+
2 2
dy
3 |0
0 0
∫a ∫a ( )
a3
=3 dx ax + ay +
2 2
dy =
3
0 0
∫a ( )|
y3 a 3 ||a
=3 ax y + a + y | dx
2
3 3 0
0
∫a ( )
a4 a4
=3 a x +
2 2
+ dx =
3 3
0
( )|
2x
3
2a 4 ||a
=3 a + x | = 3a 5 .
3 3 0

The Stokes formula expresses a relationship between surface integrals and line
integrals. It generalizes Riemann–Green formula, the latter being a special case of
the former when the surface in question is a part of the plane O x y.
We will present this formula for the particular case of explicit surfaces.

Theorem 7.5.2 (Stokes formula)

Let S : z = f (x, y), (x, y) ∈ D be an explicit smooth surface, where D ⊂ R2 is a
bounded domain whose boundary γ is a smooth curve and f : D → R is a function
of C 2 − class on D. Let Ω ⊂ R3 be a space domain that includes the surface S and
let P, Q, R : Ω → R be three functions of C 1 − class on Ω.
If we denote by ┌ = S\S = {(x, y, f (x, y)); (x, y) ∈ γ } the boundary of the
surface S, then we have the formula:
∫
 P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz =
┌
¨ ( ) ( ) ( )
∂R ∂Q ∂P ∂R ∂Q ∂P
= − dy ∧ dz + − dz ∧ dx + − dx ∧ dy .
∂y ∂z ∂z ∂x ∂x ∂y
S+
(7.24)
248 7 Surface Integrals

Fig. 7.17 An explicit z n

smooth surface S

O y
D
x
γ

Fig. 7.18 A finite union of

explicit surfaces
S1 Sp

S2

There is the following compatibility link (* ) between the direction of travel of

the curve ┌ and the face of the surface S on which the integral is made in the
right member of the formula (7.24): if the curve ┌ is traversed counter-clockwise
(respectively, clockwise), then the surface integral in the right member is made on
the upper (respectively, lower) face of the surface (Fig. 7.17) (*) Assuming that the
coordinate system is rectangular).

Proof If x = ϕ(t), y = ψ(t), t ∈ [a, b] is a parametric representation of the plane

curve γ , then the spatial curve ┌ = {(x, y, f (x, y)); (x, y) ∈ γ } is described by the
following parametric equations:
⎧
⎨ x = ϕ(t)
y = ψ(t) , t ∈ [a, b].
⎩
z = f (ϕ(t), ψ(t))

Taking into account the computation mode for the line integral of the second kind,
we have:

∫ ∫b
 P(x, y, z)dx = P(ϕ(t), ψ(t), f (ϕ(t), ψ(t))) · ϕ'(t)dt =
┌ a
∫
=  P(x, y, f (x, y))dx. (7.25)
γ

Further, from Riemann–Green formula, it results that:

7.5 Integral Formulas 249
∫ ¨ ( )
∂P ∂P ∂f
 P(x, y, f (x, y))dx = − + · dxdy. (7.26)
∂y ∂z ∂ y
γ D

∂f ∂f
If we note by p = ∂x
and q = ∂y
, further we have:
¨ ¨ √
∂P ∂P 1
− dxdy = − · √ · 1 + p 2 + q 2 dxdy =
∂y ∂y 1+ p +q
2 2
D D
¨ ¨
∂P ∂P
=− cos γ dσ = − dx ∧ dy. (7.27)
∂y ∂y
S S+

Similarly, we have:
¨ ¨ √
∂P ∂f ∂P −q
− · dxdy = · √ · 1 + p 2 + q 2 dxdy =
∂z ∂y ∂z 1+ p +q
2 2
D D
¨ ¨
∂P ∂P
= cos βdσ = dz ∧ dx. (7.28)
∂z ∂z
S S+

From Eqs. (7.25), (7.26), and Eq. (7.28), we deduce:

∫ ¨ ¨
∂P ∂P
 P(x, y, z)dx = dz ∧ dx − dx ∧ dy. (7.29)
∂z ∂y
┌ S+ S+

Similarly, it is shown that:

∫ ¨ ¨
∂Q ∂Q
 Q(x, y, z)dy = dx ∧ dy − dy ∧ dz (7.30)
∂x ∂z
┌ S+ S+
∫ ¨ ¨
∂R ∂R
 R(x, y, z)dz = dy ∧ dz − dz ∧ dx. (7.31)
∂y ∂x
┌ S+ S+

Adding the relations (7.29), (7.30), and (7.31) we obtain Stokes formula (7.24)
from the theorem statement.

Remark 7.5.4 Stokes formula also holds for surfaces that are finite unions of explicit
surfaces such as those in Theorem 7.5.2, two by two having in common at most
oriented curve arcs that are portions of the oriented boundaries of these surfaces.
Indeed, writing Stokes formula for each of the surfaces Si and adding up the obtained
∪p
formulas we obtain Stokes formula for the surface S = Si (Fig. 7.18).
i=1
250 7 Surface Integrals

The explanation consists in the fact that the line integral along an intersection
curve of two neighboring surfaces intervenes twice in the sum of the left member
of formula (7.24), with opposite orientations, hence its contribution in this sum is
zero. In this way, on the left side of the formula appears only the line integral on the
boundary of surface.
On the other hand, it is obvious that:
¨ p ¨
Σ
= .
S i=1 S
i

Remark 7.5.5 Taking into account the connection between the surface integral of
the second kind and the surface integral of the first kind, Stokes formula is also
written:
∫
 P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz =
┌
¨ (( ) ( ) ( ) )
∂R ∂Q ∂P ∂R ∂Q ∂P
= − cos α + − cos β + − cos γ dσ .
∂y ∂z ∂z ∂x ∂x ∂y
S
(7.32)

Remark 7.5.6. Under the conditions of Theorem 7.5.2, if we denote by:

−
→ −
→ −
→ −
→
V = P· i +Q· j +R· k
( ) ( ) ( )
−
→ ∂R ∂Q −
→ ∂P ∂R −
→ ∂Q ∂P −
→
Curl V = − · i + − · j + − · k
∂y ∂z ∂z ∂x ∂x ∂y
−
→ −
→ −
→ −
→
n = cos α · i + cos β · j + cos γ · k

−
→ −
→ −
→
d−
→
r = dx · i + dy · j + dz · k

then Stokes formula (7.24) is also written in vector form:

∫ ¨
−
→ − → −
→ →
 V ·d r = Curl V · − n dσ . (7.33)
┌ S

Of course, here −
→
n is the unit normal vector to the surface oriented upward.
The Stokes formula, in the form (7.33), shows that the circulation of the vector
−
→
field V along the oriented curve ┌ = ∂ S is equal to the flux of the curl of vector
−
→
field V through that surface in the direction of normal oriented upward.
7.5 Integral Formulas 251

−
→
Remark 7.5.7. If the vector field V = (P, Q, R) is irrotational on Ω, that is
−
→ −
→
curl V = 0 on Ω, from (7.33), we deduce that the line integral of the second
−
→
kind of the vector field V along any closed curve included in Ω is zero, hence the
differential form:

ω = P(x, y, z)dx + Q(x, y, z)dy + R(x, y, z)dz

is exact on Ω.
−
→
Moreover, in this case, the line integral of the second kind of the vector field V
is independent on the path in Ω (Remark 5.6.1).

The Stokes formula is used to compute the line integral of the second kind of the
−
→ −
→
vector field V along a closed curve ┌ when the flux of curl V through the surface
bounded by the curve ┌ is easier to compute.

Example 7.5.2 Using Stokes formula, compute the line integral of the second kind:
∫
 (z − y)dx + (x − z)dy + (y − x)dz
ABC

where A(a, 0, 0), B(0, b, 0), C(0, 0, c), a, b, c > 0, are the vertices of the triangle
ABC and the direction of travel is A → B → C → A.
The plan determined by the points A, B, C has the equation: ax + by + cz = 1.
Notice that the triangle ABC is the boundary of the explicit surface (Fig. 7.19):
( x y)
S : z =c 1− − , (x, y) ∈ D
a b
where D is the right triangle (full) O AB, namely (Fig. 7.20):
{ ( x ){
D = (x, y) ∈ R2 ; 0 ≤ x ≤ a, 0 ≤ y ≤ b 1 − .
a

Fig. 7.19 The triangle ABC z

c C
γ n

S
b
O B y
a
A
x
252 7 Surface Integrals

Fig. 7.20 projection of the y

triangle ABC in the plane
xOy b

O a x

If we denote by P(x, y, z) = z − y, Q(x, y, z) = x − z, R(x, y, z) = y − x,

then we have:
∂P ∂P ∂Q ∂Q ∂R ∂R
= −1; = 1; = 1; = −1; = −1; = 1.
∂y ∂z ∂x ∂z ∂x ∂y

From Stokes formula (7.32), it results that:

∫ ¨
 (z − y)dx + (x − z)dy + (y − x)dz = 2(cos α + cos β + cos γ )dσ
ABC S

where α, β, γ are the angles formed by the normal to the surface S, oriented upward,
with the axes O x, O y, Oz. On the other hand, we have:

∂z c ∂z c
p= = − ,q = =−
∂x a ∂y b

and:

a 2 b2 + b2 c2 + c2 a 2
1 + p2 + q 2 = .
a 2 b2
Since cos γ > 0, we deduce:

−p bc
cos α = √ =√
1+ p +q
2 2 a b + b2 c2 + c2 a 2
2 2

−q ca
cos β = √ =√
1+ p +q
2 2 a b + b2 c2 + c2 a 2
2 2

1 ab
cos γ = √ =√ .
1+ p +q
2 2 a b + b2 c2 + c2 a 2
2 2

Therefore, we have:
7.5 Integral Formulas 253
∫
 (z − y)dx + (x − z)dy + (y − x)dz
ABC
¨ ( )
c c
=2 + + 1 dxdy =
a b
D
(c c ) ¨ 2(ab + bc + ca)
=2 + +1 · 1dxdy = · area(D) =
a b ab
D
2(ab + bc + ca) ab
= · = ab + bc + ca.
ab 2
Remark 7.5.8 We recall that the computation of the integral in Example 7.5.2 as a
line integral of the second kind was performed in Example 5.5.5.
References

1. Bucur, I., Analyse mathématique. Calcul intégral, Conspress, Bucarest, 2014.

2. Ilyn, V.A., Poznyak, E.G., Fundamentals of mathematical analysis, Mir Publishers, Moscow,
Part I and Part II, 1982.
3. Krasnov, M., Kiselev, A., Makarenko, G., Shikin, E., Mathematical analysis for engineers, Mir
Publishers, Moscow, Part I, 1989 and Part II, 1990.
4. Lange, S., Calculus of several variables, Third Edition, Springer, New York Inc., 1987.
5. Nikolsky, S.M., A course of mathematical analysis, Mir Publishers, Moscow, Part I and Part
II, 1977.
6. Paltineanu, G., Bucur, I., Zamfir, M., Differential Calculus for Engineers, Springer, 2022,
ISBN: 978-981-19-2553-5
7. Păltineanu, G., Zamfir, M., Higher Mathematics 2 (Integral Calculus), Conspress, Bucharest,
2016 (In Romanian).
8. Păltineanu, G., Matei, P., Mateescu, G.D., Numerical analysis, Conspress, Bucharest, 2010 (In
Romanian).
9. Popescu, S.A., Mathematical analysis II. Integral calculus, Conspress, Bucharest, 2016.
10. Rudin, W., Principles of mathematical analysis, McGraw-Hill, Inc., 1964.
11. Rudin, W., Real and complex analysis, Third Edition, McGraw-Hill, Inc., 1987.
12. Stănăşilă, O., Mathematical analysis, E.D.P., Bucharest, 1981 (In Romanian).
13. Trench, W.F., Introduction to real analysis, Library of Congress Cataloging-in-Publication
Data, 2003.

© The Editor(s) (if applicable) and The Author(s), under exclusive license 255
to Springer Nature Singapore Pte Ltd. 2022
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3
Index

A Closed curve, 118

Abel-Dirichlet criterion for improper Closed parameterized path, 113
integral, 82 Computation of Dirichlet integral, 100
Absolutely convergent improper integral, Computation of Euler-Poisson integral, 103
77 Computing triple integral on simple
Antiderivative, 1, 2 domain, 202, 203
Approximation error at a point for Convergent and divergent for improper
Lagrange polynomial, 62, 63 double integral, 192, 193
Area of a circular sector, 55 Convergent and divergent for improper
Area of a curvilinear sector, 58 integral, 71
Area of a curvilinear trapezoid, 25, 52 Coordinates of center of mass of a lamina,
Area of a plane figure, 49 181
Area of a simple smooth parameterized Coordinates of center of mass of a shell,
surface, 226 233
Area of a smooth explicit surface, 225 Coordinates of center of mass of a solid,
Area of a torus, 229 211
Area of cardioid, 60 Coordinates of center of mass of wire, 134
Area of ellipse, 54 Cubable bounded space set, 197
Area of sphere, 228 Curvilinear cylinder (cylindroid), 197
Curvilinear sector, 58
Curvilinear trapezoid, 25
B Cylindrical coordinates, 208
Binomial integral, 21 Cylindrical helix, 115

C D
Cauchy integral criterion, 82 Darboux criterion for integrability for
Change of variable for definite integral, 48 function of one variable, 32
Change of variable for improper integral, 84 Darboux criterion for integrability for
Change of variable for indefinite integral. function of two variables, 163
First method, 6 Darboux integrals for function of one
Change of variable for indefinite integral. variable, 30
Second method, 8 Darboux integrals for function of two
Change of variables in double integral, 175 variables, 162
Change of variables in triple integral, 205 Darboux sums for function of one variable,
Circular sector, 55 28, 161
© The Editor(s) (if applicable) and The Author(s), under exclusive license 257
to Springer Nature Singapore Pte Ltd. 2022
G. Paltineanu et al., Integral Calculus for Engineers,
https://doi.org/10.1007/978-981-19-4793-3
258 Index

Darboux sums for function of three Improper integral depending on a

variables, 200 parameter uniform convergent, 94
Darboux sums for function of two variable, Improper integral of the first kind, 71
200 Improper integral of the second kind, 71
Definite integral, 30, 35 Indefinite integral, 2
Diameter of a plane set, 160 Integrable function of one variable in the
Diameter of a space domain, 199 Darboux sense, 30
Diffeomorphism, 218 Integrable function of one variable in the
Differential closed form, 154 Riemann sense, 35
Differential exact form, 150 Integrable function of three variables in the
Differential form, 150 Darboux sense, 200
Dirichlet integral, 84 Integrable function of three variables in the
Domain in, 202 Riemann sense, 202
Domain in ϒ 2 simple with respect to Integrable function of two variables in the
coordinate axis, 166 Darboux sense, 162
Double integral, 163, 164 Integrable function of two variables in the
Riemann sense, 164
Integral depending on a parameter, 88
E Integration by parts for definite integral, 48
Elementary domain of Green type, 185 Integration by parts for indefinite integral, 9
Elementary plane set, 50
Elementary space set, 196
Elementary surface, 219 L
Elliptic integrals, 125 Lagrange interpolation polynomial, 61
Equivalent paths, 116 Lagrange interpolation polynomial for
Equivalent surface canvases, 218 equidistant nodes, 62
Euler integral of the first kind Lamina, 180
(beta-function), 104 Lebesque criterion for integrability for
Euler integral of the second kind function of one variable, 42
(gamma-function), 104 Lebesque criterion for integrability for
Euler-Poisson integral, 80 function of two variables, 163
Explicit surface, 219 Leibniz formula for differentation of the
integral depending on a parameter,
91, 92
F Length of a rectifiable curve (path), 119,
First Comparison test for improper integral, 123
71 Length of the cardioid, 127
Flux-divergence formula, 246 Length of the circle, 124
Fresnel integrals, 85 Length of the cylindrical helix, 123
Fubini Theorem, 92 Line integral independent on the path, 150
Line integral of the first kind, 132
Line integral of the second kind, 140, 141
G Locally integrable function, 71
Gauss-Ostrogradsky formula, 242
Generalized polar coordinates, 179
Generalized right cylinder, 197 M
Generalized spherical coordinates, 209 Mass of a lamina, 181
Mass of a shell, 233
Mass of a solid, 211
I Mass of a wire, 134
Implicit surface, 221 Mean value formula for double integral,
Improper double integral, 192 166
Improper integral depending on a Mean value formula for line integral, 140
parameter pointwise convergent, 93 Mean value formula for simple integral, 45
Index 259

Mechanical work, 143 Reducing a double integral to an iterated

Mőbius strip, 236 simple integral, 168, 170
Moments of inertia of a lamina, 183, 184 Refinement of a partition of a bounded
Moments of inertia of a shell, 233 plane domain, 161
Moments of inertia of a solid, 212 Refinement of a partition of a compact
Moments of inertia of a wire, 135 interval, 28
Repeated Simpson formula, 68
Repeated trapezoidal formula, 66
N Riemann criterion for integrability for
Natural parameterization of a path, 129 function of one variable, 38
Newton-Côtes coefficients, 64 Riemann criterion for integrability for
Newton-Côtes quadrature formula, 64 function of two variables, 165
Newton-Leibniz formula, 47 Riemann-Green formula, 185
Normal line, 222 Riemann sum for function of one variable,
Norm of a partition of a compact interval, 34
28 Riemann sums for function of three
Norm of a partition of a plane bounded variables, 201
domain, 161 Riemann sums for function of two
Norm of a partition of a space bounded variables, 163
domain, 199, 200
Null subset, 41
S
Second Comparison test for improper
O integral, 78
Orientable (two-sider) surface, 234 Set of area zero, 56
Orientation of a curve, 118 Set of volume zero, 199
Shell, 233
Simple curve, 118
P Simple path, 115
Parameter change, 116, 218 Simpson formula (rule), 67
Parameterized curve, 117 Singular point of a path, 115
Parameterized path, 113 Smooth curve, 118
Parameterized simple surface canvas, 216 Smooth path, 116
Parameterized smooth surface canvas, 216 Solid, 211
Parameterized surface, 221 Spherical coordinates, 206
Parameterized surface canvas, 215 Squareable bounded plane set, 51
Parametric representation of a path, 113 Star domain, 154
Parametric representation of a surface Stokes formula, 247
canvas, 215 Support of a curve, 118
Partition of a bounded plane domain, 161 Support of a parametrized path, 113
Partition of a bounded space domain, 199 Support of a parametrized surface canvas,
Partition of a compact interval, 28 221
Piecewise smooth curve, 118 Surface, 220
Polar coordinates, 176 Surface integral of the first kind, 230
Polygonal plane set, 49 Surface integral of the second kind, 238
Primitive function, 1
Primitives of rational functions, 11
Primitives of trigonometric functions, 18 T
Principal value of improper integral in Table of antiderivatives of the basic
Cauchy sense, 74 elementary functions, 3
Tangent plane, 222
Total flux of a vector field through a
R surface, 239
Rectifiable curve (path), 118, 123 Trapezoidal formula (rule), 64
260 Index

Triple integral, 201, 202 Volume of a space domain, 210

Volume of a space figure, 195
U
Union of two paths, 118
Unit normal vector, 222
W
Weierstrass criterion for uniform
V convergence of an improper integral,
Volume of a curvilinear cylinder, 198 94