Intelligent Systems Reference Library 58

George A. Anastassiou
Iuliana F. Iatan

Intelligent
Routines II
Solving Linear Algebra and Differential
Geometry with Sage
Intelligent Systems Reference Library

Volume 58

Series editors
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: [email protected]

Lakhmi C. Jain, University of Canberra, Canberra, Australia

e-mail: [email protected]

For further volumes:

http://www.springer.com/series/8578
About this Series

The aim of this series is to publish a Reference Library, including novel advances
and developments in all aspects of Intelligent Systems in an easily accessible and
well structured form. The series includes reference works, handbooks, compendia,
textbooks, well-structured monographs, dictionaries, and encyclopedias. It con-
tains well integrated knowledge and current information in the field of Intelligent
Systems. The series covers the theory, applications, and design methods of
Intelligent Systems. Virtually all disciplines such as engineering, computer sci-
ence, avionics, business, e-commerce, environment, healthcare, physics and life
science are included.
George A. Anastassiou Iuliana F. Iatan
•

Intelligent Routines II
Solving Linear Algebra and Differential
Geometry with Sage

123
George A. Anastassiou Iuliana F. Iatan
Department of Mathematical Sciences Department of Mathematics and Computer
University of Memphis Science
Memphis Technical University of Civil Engineering
USA Bucharest
Romania

ISSN 1868-4394 ISSN 1868-4408 (electronic)

ISBN 978-3-319-01966-6 ISBN 978-3-319-01967-3 (eBook)
DOI 10.1007/978-3-319-01967-3
Springer Cham Heidelberg New York Dordrecht London

Library of Congress Control Number: 2012932490

 Springer International Publishing Switzerland 2014

This work is subject to copyright. All rights are reserved 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. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publisher’s location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
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.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Mathematics: the art of redescribing to the
simplest of logical structures, that is an art of
rewriting finite steps of thinking.

G. A. Anastassiou

The measure of success for a person is the

magnitude of his/her ability to convert
negative conditions to positive ones and
achieve goals.

G. A. Anastassiou

Labor omnia vincit improbus.

Virgil, Georgics

The greatest thoughts come from the heart.

Vauvenargues, Reflections and Maxims

Friends show me what I can do, foes teach

me what I should do.

Schiller, Friend and foe

Experience is by industry achieved, and
perfected by the swift course of time.

Shakespeare, The Two Gentlemen of Verona

Good name in man and woman, dear my

lord, Is the immediate jewel of their souls.
Who steals my purse steals trash; ’t is
something, nothing; ’T was mine, ’t is his,
and has been slave to thousands; But he that
filches from me my good name Robs me of
that which not enriches him, And makes me
poor indeed.

Shakespeare, Othello Act III, scene 3

L’ami du genre humain n’est point du tout

mon fait.

Molière, Le Misanthrope, I, 1

Honor is like an island, rugged and without

shores; we can never re-enter it once we are
on the outside.

Boileau, Satires

Modesty is an adornment, but you come

further without it.

German proverb
Preface

Linear algebra can be regarded as a theory of the vector spaces, because a vector
space is a set of objects or elements that can be added together and multiplied by
numbers (the result remaining an element of the set), so that the ordinary rules of
calculation are valid. An example of a vector space is the geometric vector space
(the free vector space), presented in the first chapter of the book, which plays a
central role in physics and technology and illustrates the importance of the vector
spaces and linear algebra for all practical applications.
Besides the notions which operates mathematics, created by abstraction from
environmental observation (for example, the geometric concepts) or quantitative
and qualitative research of the natural phenomena (for example, the notion of
number) in mathematics there are elements from other sciences. The notion of
vector from physics has been studied and developed creating vector calculus,
which became a useful tool for both mathematics and physics. All physical
quantities are represented by vectors (for example, the force and velocity).
A vector indicates a translation in the three-dimensional space; therefore we
study the basics of the three-dimensional Euclidean geometry: the points, the
straight lines and the planes, were in the second chapter.
The linear transformations are studied in the third chapter, because they are
compatible with the operations defined in a vector space and allow us to transfer
algebraic situations and related problems in three-dimensional space.
Matrix operations clearly reflect their similarity to the operations with linear
transformations; so the matrices can be used for the numerical representation of
the linear transformations. The matrix representation of linear transformations is
analogous to the representation of the vectors through n coordinates relative to a
basis.
The eigenvalue problems (also treated in the third chapter) are of great
importance in many branches of physics. They make it possible to find some
coordinate systems in which changes take the simplest forms. For example, in
mechanics the main moments of a solid body are found with the eigenvalues of a
symmetric matrix representing the vector tensor. The situation is similar in con-
tinuous mechanics, where the body rotations and deformations in the main
directions are found using the eigenvalues of a symmetric matrix. Eigenvalues
have a central importance in quantum mechanics, where the measured values of
the observable physical quantities appear as eigenvalues of operators. Also, the

vii
viii Preface

eigenvalues are useful in the study of differential equations and continuous

dynamical systems that arise in areas such as physics and chemistry.
The study of the Euclidean vector space in the fourth chapter is required to
obtain the orthonormal bases, whereas relative to these bases the calculations are
considerably simplified. In a Euclidean vector space, scalar product can be used to
define the length of vectors and the angle between them. In the investigation of the
Euclidean vector spaces very useful are the linear transformations compatible with
the scalar product, i.e. the orthogonal transformations. The orthogonal transfor-
mations in the Euclidean plane are: the rotations, the reflections or the composi-
tions of rotations and reflections.
The theory of bilinear and quadratic form are described in the fifth chapter.
These are used with analytic geometry to get the classification of the conics and of
the quadrics, presented in the Chap. 8.
In Analytic Geometry we replace the definitions and the geometrical study of
the curves and the surfaces, by the algebraic correspondence: a curve and a surface
are defined by algebraic equations, and the study of the curve and the surface is
reduced to the study of the equation corresponding to each one (see the seventh
chapter).
The above are used in physics, in particular to describe physical systems subject
to small vibrations. The coefficients of a bilinear form behave for certain trans-
formations like the tensor coordinates. The tensors are useful in theory of elasticity
(deformation of an elastic medium is described through the deformation tensor).
In the differential geometry, in the study of the geometric figures, we use the
concepts and methods of the mathematical analysis, especially the differential
calculus and the theory of differential equations, presented in the sixth chapter.
The physical problems lead to inhomogeneous linear differential equations of
order n with constant coefficients.
In this book we apply extensively the software SAGE, which can be found free
online http://www.sagemath.org/.
We give plenty of SAGE applications at each step of our exposition.
This book is usefull to all researchers and students in mathematics, statistics,
physics, engineering and other applied sciences. To the best of our knowledge this
is the first one.
The authors would like to thank Prof. Razvan Mezei of Lenoir-Rhyne Uni-
versity, North Carolina, USA for checking the final manuscript of our book.

Memphis, USA, May 13, 2013 George A. Anastassiou

Bucharest, Romania Iuliana F. Iatan
Contents

1 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Geometric Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Free Vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Operations with Free Vectors . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Addition of the Free Vectors . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Scalar Multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.3 Vector Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.4 Defining of the Products in the Set
of the Free Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4 Vector Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

2 Plane and Straight Line in E3 . . . . . . . . . . . . . . . . . . . . . . . . ... 59

2.1 Equations of a Straight Line in E3 . . . . . . . . . . . . . . . . . . . ... 59
2.1.1 Straight Line Determined by a Point
and a Nonzero Vector . . . . . . . . . . . . . . . . . . . . . . ... 59
2.1.2 Straight Line Determined by Two Distinct Points . . . ... 61
2.2 Plane in E3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 62
2.2.1 A a Point and a Non Zero Vector Normal
to the Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 62
2.2.2 Plane Determined by a Point and Two
Noncollinear Vectors . . . . . . . . . . . . . . . . . . . . . . . ... 63
2.2.3 Plane Determined by Three Noncollinear Points . . . . ... 65
2.2.4 Plane Determined by a Straight Line and a Point
that Doesn’t Belong to the Straight Line . . . . . . . . . ... 66
2.2.5 Plane Determined by Two Concurrent Straight Lines ... 66
2.2.6 Plane Determined by Two Parallel Straight Lines . . . ... 68
2.2.7 The Straight Line Determined by the Intersection
of Two Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 69
2.3 Plane Fascicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 73

ix
x Contents

2.4 Distances in E3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.4.1 Distance from a Point to a Straight Line . . . . . . . . . . . . 74
2.4.2 Distance from a Point to a Plane . . . . . . . . . . . . . . . . . 75
2.4.3 Distance Between Two Straight Lines . . . . . . . . . . . . . . 79
2.5 Angles in E3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.5.1 Angle Between Two Straight Lines . . . . . . . . . . . . . . . . 83
2.5.2 Angle Between Two Planes . . . . . . . . . . . . . . . . . . . . . 83
2.5.3 Angle Between a Straight Line and a Plane . . . . . . . . . . 84
2.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

3 Linear Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.1 Linear Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.2 Matrix as a Linear Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.3 Changing the Associated Matrix to the Change of Basis . . . . . . 107
3.4 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.4.1 Characteristic Polynomial of an Endomorphism . . . . . . . 111
3.4.2 Determining the Eigenvalues and the Eigenvectors
for an Endomorphism . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.4.3 Diagonalization Algorithm of an Endomorphism . . . . . . 117
3.4.4 Jordan Canonical Form . . . . . . . . . . . . . . . . . . . . . . . . 122
3.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

4 Euclidean Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

4.1 Euclidean Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.2 Linear Operators in Euclidean Vector Spaces . . . . . . . . . . . . . . 150
4.2.1 Orthogonal Transformations in the Euclidean Plane . . . . 152
4.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

5 Bilinear and Quadratic Forms . . . . . . . . . . . . . . . . . . . . ....... 165

5.1 Bilinear and Quadratic Forms . . . . . . . . . . . . . . . . . . ....... 165
5.2 Gauss-Lagrange Method for Reducing a Quadratic
Form to a Canonical Expression . . . . . . . . . . . . . . . . ....... 169
5.3 Reducing a Quadratic Form to a Canonical Expression
by Jacobi Method . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 175
5.4 Eigenvalue Method for Reducing a Quadratic
Form into Canonical Expression . . . . . . . . . . . . . . . . ....... 180
5.5 Characterization Criteria for Positive (Negative)
Definite Matrices. . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 184
5.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 190
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 196
Contents xi

6 Differential Geometry of Curves and Surfaces . . . . . . . . . . . . . . . 197

6.1 Analytical Definition of Curves in the Space . . . . . . . . . . . . . . 197
6.2 Tangent and Normal Plane to a Curve in the Space . . . . . . . . . 198
6.3 Frenet Trihedron. Frenet Formulas. . . . . . . . . . . . . . . . . . . . . . 199
6.4 Curvature and Torsion of the Space Curves . . . . . . . . . . . . . . 206
6.5 Envelope of a Family of Curves in Plane . . . . . . . . . . . . . . . . 214
6.6 Analytic Definition of Surfaces . . . . . . . . . . . . . . . . . . . . . . . . 217
6.7 Tangent Plane and Normal to a Surface . . . . . . . . . . . . . . . . . . 219
6.8 First Fundamental Form of a Surface. Curves on a Surface . . . . 223
6.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

7 Conics and Quadrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

7.1 General Equation of a Conic. . . . . . . . . . . . . . . . . . . . . . . . . . 235
7.2 Conics on the Canonical Equations . . . . . . . . . . . . . . . . . . . . . 238
7.2.1 Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
7.2.2 Ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
7.2.3 Hyperbola . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
7.2.4 Parabola. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
7.3 Reducing to the Canonical Form of a Conic Equation . . . . . . . . 248
7.4 General Equation of a Quadric . . . . . . . . . . . . . . . . . . . . . . . . 262
7.5 Quadrics on Canonical Equations . . . . . . . . . . . . . . . . . . . . . . 270
7.5.1 Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
7.5.2 Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
7.5.3 Cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
7.5.4 Two-Sheeted Hyperboloid . . . . . . . . . . . . . . . . . . . . . . 280
7.5.5 Elliptic Paraboloid . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
7.5.6 Hyperbolic Paraboloid . . . . . . . . . . . . . . . . . . . . . . . . . 284
7.6 Ruled Surfaces. Surface Generation . . . . . . . . . . . . . . . . . . . . . 286
7.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Symbols

E3, 1
!
AB , 1
AB, 3
V3, 3
d (A, B), 13
a
 b, 14
a  b, 17
(Rn,?,•), 25
(Rn [X],?,•), 29
(Mm, n (R),?,•), 25, 26
xB, 36
MðB1 ;B2 Þ , 39
Msn (R), 46, 47
Man (R), 48, 49
, 52
n, 62, 63
L(U, V), 92
End(V), 92
Ker T, 92
Im T, 92
eðB1 ;B2 Þ , 97
T
P (k), 112
gk0 , 117
ak0 , 117
Jp (k), 123
Bp (k), 123
\x; y[, 135
kxk, 138
ru (x), 152
sO (x), 153
f (x), 166
b (x, y), 166
s, 199

xiii
xiv Symbols

b, 199
m, 199
pN, 200
pO, 200
pr, 200
U1, 223
dr, 229
Chapter 1
Vector Spaces

1.1 Geometric Vector Spaces

1.1.1 Free Vectors

Besides the notions used in Mathematics, created by abstraction from environmental

observation (for example the geometric concepts) or quantitative and qualitative
research of the natural phenomena (for example the notion of number) in mathematics
there are elements from other sciences. The notion of vector from physics has been
studied and developed creating vector calculus, which became a useful tool for both
mathematics and physics. All physical quantities are represented by vectors (for
example the force, the velocity).
In examining the phenomena of nature, we can meet two kinds of quantities:
1. scalar quantities (the temperature, the length, the time, the volume, the density,
the area) which can be characterized by a number (which is measured by a specific
unit);
2. vector quantities (the force, the velocity, the acceleration) which to measure their
characterization is not sufficient, it is necessary to know the direction and the
sense in which they operate.
To represent a vector-oriented segment is used a method from mechanics.
One denotes by E3 the three-dimensional space of the Euclidean geometry.
Definition 1.1 (see [1], p. 108). We call oriented segment (or bound vector) an
−
→
ordered pair of points (A, B) ∈ E3 × E3 and we denote it by AB (see Fig. 1.1).

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 1

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_1,
© Springer International Publishing Switzerland 2014
2 1 Vector Spaces

Fig. 1.1 The representation

of an oriented segment AB

AB
A

THE CHARACTERISTICS OF AN ORIENTED SEGMENT

−
→
We consider the oriented segment AB, for any two points A, B ∈ E3 .
• The points A and B are called the origin (the starting point) and respectively
−→
the endpoint (the peak) of the oriented segment. If A = B then AA is the null
oriented segment.
• If A = B then the straight line determined by them is called the support line of
−
→
AB and is denoted by AB.
−
→
• The direction of the oriented segment AB (A = B) is the direction of the straight
line AB.
• The sense on the support line, from A to B is called the sense of the oriented
−→
segment AB.
• The distance between the points A and B means the length − (the norm or
−
→  →
magnitude) of the oriented segment AB and is denoted by AB. If the origin
of an oriented segment coincides with the endpoint (the null oriented segment)
then the length of this segment is equal to 0.
−
→ −→
Definition 1.2 (see [1], p. 108). Two oriented segments AB and CD, A = B, C = D
have the same directions if their support lines and are parallel or coincide.
−→ −→
Definition 1.3 (see [2], p. 4 and [3], p. 4). Two oriented segments AB and CD, A =
B, C = D with the same direction have the same sense if B and D belong to the
same half-plane determined by a straight line AC (see Fig 1.2).

Fig. 1.2 Example of two

oriented segments, that have
the same sense

−
→ −→
Definition 1.4 (see [1], p. 108). Two oriented segments AB and CD, A = B C = D
one call equipollent if they have the same direction, the same sense and the same
−
→ −→ −→ −→
norm; if AB is equipollent with CD we shall write AB ∼ CD.
1.1 Geometric Vector Spaces 3

Theorem 1.5 (see [1], p. 108). The equipollent relation defined on the set of the
oriented segments is an equivalence relation.
Proof
The equipollent relation is an equivalence relation since it is:
−→ − →
1. reflexive: AB ∼ AB;
−
→ −→ −→ − →
2. symmetric: AB ∼ CD involves CD ∼ AB;
−
→ −→ −→ − → −
→ − →
3. transitive: AB ∼ CD and CD ∼ EF involves AB ∼ EF.
We can classify the vectors in the following way:
(a) free vectors, which have the arbitrary origin at any point in space, but whose
direction, sense and length on space are prescribed;
(b) bound vectors, whose origin is prescribed;
(c) sliding vectors moving along the same support line, and their origin can be
anywhere on the line.

Definition 1.6 (see [4], p. 86). We call a free vector (geometric vector) characterized
−
→ −
→
by an oriented segment AB, the set of the oriented segments, equipollent with AB:
−→ −→ − →
AB = CD|CD ∼ AB .

Any oriented segment of this set is called the representative of the free vector
−→
AB; therefore CD ∈ AB.
A free vector of the length:
• 1 is called versor (unit vector); generally one denotes by e;
• 0 is called null vector; one denotes by 0.

Definition 1.7 (see [4], p. 86). The length, direction and sense of a free nonzero
vector means the length, direction and sense corresponding to the oriented segment
that it represents.
The set of the geometric vectors from the space E3 will be denote by V3 :
 
V3 = AB| A, B ∈ E3 ,

−
→
namely V3 means the set of equivalence
  classes of the oriented segment AB.
We can use the notations: a AB or d (A, B) in order to designate the length
of a free vector a or AB.
Definition 1.8 (see [1], p. 109). We say that two free vectors are equal and we write
a = b if their representatives are equipollent.
Definition 1.9 (see [1], p. 109). Two free non-null vector a and b are collinear if
they have the same direction (see Fig. 1.3).
4 1 Vector Spaces

Fig. 1.3 Example of collinear

vectors

Definition 1.10 (see [1], p. 109). Two collinear vectors which have the same length
but they have opposite directions are called opposite vectors. The opposite of a free
vector a is −a (see Fig. 1.4).

Fig. 1.4 Example of an

opposite vectors

Definition 1.11 (see [1], p. 109). Three free vectors a , b , c are called coplanar if
their support lines lie in the same plane (see Fig. 1.5).

Fig. 1.5 Example of some

coplanar vectors

1.2 Operations with Free Vectors

We can define the following opperations in the set V3 :

1. the addition of the free vectors;
2. the scalar multiplication;
3. the vector decomposition.

1.2.1 Addition of the Free Vectors

An internal operation is defined on V3 (the addition of the free vectors)

 
+ : V3 × V3 → V3 , a, b → a + b.
1.2 Operations with Free Vectors 5

Therefore, the sum of two or more vectors is also a vector, which can be obtained
through the following methods:
(A) if vectors are parallel or collinear and
(a) they have the same sense, then the sum vector has the direction and the
sense of the component vectors, and their length is equal to the sum of the
lengths corresponding to the component vectors;
(b) they have opposite sense, then the sum vector has the common direction,
the sense of the larger vector, and its magnitude is given by the differences
of the two vector magnitudes.
(B) if the vectors have only a common origin, then their sum is determined using
the parallelogram rule.

Definition 1.12 (the parallelogram rule, see [1], p. 110). Let a, b ∈ V3 be two
free vectors, which have a common origin and A ∈ E3 be an arbitrary fixed point.
−
→ −
→ −→
If OA ∈ a (OA means the representative of the free vector a ) and OC ∈ b then the
−
→
free vector c represented by the oriented segment OB is called the sum of the free
−→ − → −→
vectors a and b; one writes c = a + b or OB = OA + OC (Fig. 1.6).

Fig. 1.6 Illustration of the

parallelogram rule

(C) if the vectors are arranged so as one extremity to be the origin of the other, to
achieve their sum one applies the triangle rule.

Definition 1.13 (the triangle rule, see [1], p. 110). Let a, b ∈ V3 be two free vectors
−→ −
→
and A ∈ E3 an arbitrary fixed point. If AB ∈ a (AB is the representative of the free
−
→ −
→
vector a) and BC ∈ b then the free vector c represented by the oriented segment AC
−→ − → − →
is called the sum of the free vectors a and b; one writes c = a + b or AC = AB + BC
(Fig. 1.7).

Fig. 1.7 Illustration of the

triangle rule
6 1 Vector Spaces

Remark 1.14 (see [11], p. 118 and [10], p. 8). For many vectors, sitting in the same
way, one applies the polygon rule, which is a generalization of the triangle rule; the
sum vector is that which closes the polygon, joining the origin of the first component,
with the last extremity.
The addition of the vectors is based on some experimental facts (composition of
forces, velocities).

Example 1.15 (see [5], p. 58). We suppose a segment AB and the points M1 and M2
that divide the segment into three equal parts. If M is an arbitrary point outside the
segment, express the vectors MM1 and MM2 depending on the vectors MA = a and
MB = b.

Solution
Using the triangle rule we have:

MM1 = MA + AM1 .

Hence
AB = AM + MB = −a + b.

We deduce
1 b−a
AM1 = AB =
3 3
and
b−a 2a + b
MM1 = a + = .
3 3
Similarly,

2 b−a a + 2b
MM2 = MA + AM2 = MA + AB = a + 2 = .
3 3 3
1.2 Operations with Free Vectors 7

A solution in Sage will be:

Theorem 1.16 (see [6], p. 7). The addition of the free vectors determines an abelian
group structure (V3 , +) on the set of the free vectors.
Proof
We note that the addition of the free vectors is an internal well-defined algebrical
operation, i.e. the free vector c = a + b doesn’t depend on the choice point A since
from AB = A B and BC = B C  it results AC = A C  .
We check the properties of (Fig. 1.8):
8 1 Vector Spaces

1. associativity:    
a + b + c = a + b + c, (∀) a, b, c ∈ V3 .

Let O be a fixed point in space and OA = a, AB = b, BC = c.

Fig. 1.8 Illustration of the

associative property

 
⎫
a + b + c = OA + AB + BC = OB + BC = OC ⎬    
 
=⇒ a + b + c = a + b + c.
a + b + c = OA + AB + BC = OA + AC = OC ⎭

2. 0 is the neutral element:

(∃) 0 ∈ V3 such that a + 0 = 0 + a = a, (∀) a ∈ V3 .

Let OA = a, OO = 0, AA = 0.
We have:

OO + OA = OA ⇐⇒ 0 + a = a,
OA + AA = OA ⇐⇒ a + 0 = a.

3. simetrizable element:

(∀) a ∈ V3 , (∃) − a ∈ V3 such that a + (−a) = (−a) + a = 0.

Let a = AB, −a = BA.

We obtain

a + (−a) = AB + BA = AA = 0,
(−a) + a = BA + AB = BB = 0.


Remark 1.17 (see [4], p. 88). The existence of an opposite for a free vector allows
the substraction definition of the free vectors a, b ∈ V3 (Fig. 1.9):
 
a − b = a + −b .
1.2 Operations with Free Vectors 9

4. commutativity:
a + b = b + a, (∀) a, b ∈ V3 .

Let a = OA, b = AB.

Fig. 1.9 Illustration of the

commutative property

We note that OABC is a parallelogram.

Then we have:

OA + AB = OB,
OC + CB = OB.

1.2.2 Scalar Multiplication

We now define an external operation (scalar multiplication)

• : R × V3 → V3 , (t, a) → ta,

• R being the set of the real numbers,

• ta = 0 if t = 0 or a = 0;
• the free vector ta has:
– the same direction with a,
– the same sense with a if t > 0 and the opposite sense of a if t < 0;
– the magnitude ta = |t| · a.

Theorem 1.18 (see [6], p. 7). The multiplication of the free vectors with scalar has
the following properties:
1. distributivity of scalar multiplication with respect to the vector addition:
 
t a + b = ta + tb, (∀) t ∈ R, (∀) a, b ∈ V3 ;

2. distributivity of scalar multiplication with respect to the scalar addition:

(s + t) a = sa + ta, (∀) s, t ∈ R, (∀) a ∈ V3 ;

10 1 Vector Spaces

3. s (ta) = (st) a, (∀) s, t ∈ R, (∀) a ∈ V3 ;

4. 1 · a = a, (∀) a ∈ V3 .

1.2.3 Vector Decomposition

Proposition 1.19
  (the decomposition of a vector in a direction, see [7], p. 8). Let
be a, b ∈ V3 \ 0 . The vectors a and b are collinear if and only if (∃) t ∈ R unique
such that b = ta.
 
Theorem 1.20 (see [7], p. 8). Let be a, b ∈ V3 \ 0 . The vectors a and b are collinear
if and only if (∃) α, β ∈ R nonsimultaneous equals to zero (i.e. α2 + β 2 = 0) such
that αa + βb = 0.
The decomposition of a vector after two directions is the reverse operation to the
addition of the two vectors.
Proposition 1.21 (the decompositionofa vector after two noncollinear directions,
see [7], p. 9). Let be a, b, c ∈ V3 \ 0 . If a, b, c are coplanar then (∃) α, β ∈ R
uniquely determined such that c = αa + βb.
 
Theorem 1.22 (see [7], p. 10). Let be a, b, c ∈ V3 \ 0 . The vectors a, b, c are
coplanar if and only if (∃) α, β, γ ∈ R nonsimultaneous equal to zero (namely
α2 + β 2 + γ 2 = 0) such that αa + βb + γc = 0.
Proposition 1.23 (the decomposition of a vector  after three noncoplanar direc-
tions, see [7], p. 10). Let be a, b, c, d ∈ V3 \ 0 . If a, b, c are noncoplanar then
(∃) α, β, γ ∈ R uniquely determined such that d = αa + βb + γc.
We suppose a point O in E3 called origin and three non-coplanar versors i j k whose
we attach the coordinate axes Ox, Oy, Oz that have the same sense as the sense of these
versors (see Fig. 1.10). The ensemble O, i, j, k is called the Cartesian reference
in E3 .

Fig. 1.10 Representation of a

cartesian reference in E3
1.2 Operations with Free Vectors 11

Whereas the versors i, j, k are non-coplanar, then under the Proposition 1.23, for
any vector v ∈ V3 (∃) r, s, t ∈ R uniquely determined such that v is expressed as
v = ri + sj + tk, called analytical expression of the vector v. The numbers (r, s, t)
are called the Euclidean coordinates (the components) of v relative to the reference

O, i, j, k .

Definition 1.24 (see [8], p. 449 and [10], p. 10) Let M ∈ E3 be a fixed point. The
vector OM is called the position vector ofthe pointM. The coordinates of the
position vector OM relative to the reference O, i, j, k are called the coordinates
of the point M. If OM = xi + yj + zk then one writes M (x, y, z).
Example 1.25 (see [9], p. 50 ). Find λ ∈ R such that the vectors
⎧
⎨ v1 = 2i + (λ + 2) j + 3k
v2 = i + λj − k
⎩
v3 = 4j + 2k

to be coplanar.
With this values of λ decompose the vector v1 after the directions of the vectors
v2 and v3 .
Solution
Using the Theorem 1.22, v1 v2 , v3 are coplanar if and only if (∃) α, β, γ ∈ R, α2 +
β + γ 2 = 0 such that αv1 + βv2 + γv3 = 0.
2

We obtain:

(2α + β) i + [(λ + 2) α + λβ + 4γ] j + (3α − β + 2γ) k = 0,

i.e. ⎧
⎨ 2α + β = 0
(λ + 2) α + λβ + 4γ = 0
⎩
3α − β + 2γ = 0;

the previous homogeneous system admits some non-trivial solutions ⇐⇒

 
 2 1 0 

 λ + 2 λ 4  = 0 ⇐⇒ λ = −8.
 
 3 −1 2 

If v1 , v2 , v3 are coplanar, from the Proposition 1.21 we have: (∃) α , β  ∈ R

uniquely determined such that

v 1 = α v 2 + β  v 3 ;

it results
12 1 Vector Spaces

2i − 6j + 3k = α i + −8α + 4β  j + −α + 2β  k;

therefore ⎧
⎨ α = 2
−8α + 4β  = −6
⎩
−α + 2β  = 3 ⇒ 2 + 2β  − 3 ⇒ β  = 25 .

We achieve:
5
v1 = 2v2 + v3 .
2
Using Sage we shall have:

We shall represent in Sage the three vectors v1 , v2 , v3:

Definition 1.26 (see [10], p. 12 ). If A (x1 , y1 , z1 ) , B (x2 , y2 , z2 ) are two given points
from E3 (see Fig. 1.11)
1.2 Operations with Free Vectors 13

Fig. 1.11 The distance

between two points

then we have

AB = OB − OA = (x2 − x1 ) i + (y2 − y1 ) j + (z2 − z1 ) k, (1.1)

and the distance from the points A and B denoted d (A, B) is calculated using the
formula:
  
d (A, B) = AB = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 . (1.2)

  −
→ −
→
Definition 1.27 (see [1], p. 115). Let be a, b ∈ V3 0 , O ∈ E3 and OA ∈ a, OB ∈ b.
−
→ −
→
The angle ϕ ∈ [0, π] determined by the oriented segments OA and OB is called the
angle between the free vectors a and b (see Fig. 1.12).

Fig. 1.12 Representation

of the angle between the
respective free vectors

The free vectors a and b are called orthogonal if the angle between them is π2 .

1.2.4 Defining of the Products in the Set of the Free Vectors

1.2.4.1 Scalar Product in V3

 
We consider a, b ∈ V3 \ 0 . For a = 0 , b = 0 one denotes by ϕ ∈ [0, π] the angle
between a and b.
14 1 Vector Spaces

Definition 1.28 (see [1], p. 116). The scalar (dot) product of the free vectors a
and b is the scalar a · b given by
  
 
a b cos ϕ, a = 0, b = 0
a·b= (1.3)
0, a = 0, b = 0.

One important physical application of the scalar product is the calculation of work.
−
→
The scalar product represents (see [11], p. 126) the work done by the force F
−
→
required to move a mobile to a straight line, with the director vector d, that makes
−
→
the angle ϕ with the direction of the force F .
Proposition 1.29 (see [7], p. 15). The scalar product of the free vectors has the
following properties:
1. commutativiy:
a · b = b · a, (∀) a, b ∈ V3 ;
 
2. t a · b = ta · b = a · tb, (∀) a, b ∈ V3 , (∀) t ∈ R;
3. distributivity of scalar multiplication with respect to the vector addition:
 
a · b + c = a · b + a · c, (∀) a, b, c ∈ V3 ;
 
a + b · c = a · b + b · c, (∀) a, b, c ∈ V3 ;

  
a · a > 0, (∀) a ∈ V3 \ 0 ;
4.
a · a = 0 ⇔ a = 0;
 
5. a · b = 0 ⇔ a and b are orthogonal, (∀) a, b ∈ V3 \ 0 ;
6. if a, b ∈ V3 ,

a = a1 i + a2 j + a3 k
b = b1 i + b2 j + b3 k

then we obtain the analytical expression of the scalar product:

a · b = a1 b1 + a2 b2 + a3 b3 . (1.4)

Particularly,
a · a = a12 + a22 + a32 = a2 . (1.5)
 
7. the angle between the vectors a, b ∈ V3 \ 0 is given by the formula:

a·b a1 b1 + a2 b2 + a3 b3
cos ϕ =  =  , ϕ ∈ [0, π] ; (1.6)
 
a b a12 + a22 + a32 · b12 + b22 + b32
1.2 Operations with Free Vectors 15

one notices that the vectors and are orthogonal if and only if

a1 b1 + a2 b2 + a3 b3 = 0.

Example 1.30 (see [12], p. 161). Prove that the heights of a triangle are concurrent.
Solution

Let H be the intersection of the heights from A and B. We must show that CH is
perpendicular on AB, namely CH · AB = 0.
Using the triangle rule we have:

AB = AC − BC.

We shall obtain

CH · AB = CH · AC − BC = CH · AC − CH · BC

= − HB + BC · AC + HA + AC · BC,

i.e.

CH · AB = −HB · AC − BC · AC + HA · BC + AC · BC = −HB · AC + HA · BC = 0.

We have used the commutative property of the scalar product and the fact that BH
is perpendicular on AC and AH is perpendicular on BC.
We need the following Sage code to solve this problem:

Example 1.31 (see [12], p. 161). Prove that th diagonals in a rhombus are
perpendicular.
16 1 Vector Spaces

Solution

We shall deduce

(u − v) · (u + v) = u · (u + v) − v · (u + v) = u · u + u · v − v · u − v · v = u2 − v2 .

As in the case of rhombus we have u = v we shall obtain

(u − v) · (u + v) = 0,

i.e. the diagonals in a rhombus are perpendicular.

Solving in Sage we shall have:

1.2.4.2 Cross Product in V3

 
We consider a, b ∈ V3 \ 0 . For a = 0 b = 0 one denotes by ϕ ∈ [0, π] the angle
between a and b.
Definition 1.32 (see [12], p. 175 and [1], p. 118) . The cross product of the free
vectors a and b is the free vector a × b constructed as follows:
• The direction of a × b is orthogonal to the plane determined by vectors and a and
b;  
 
• Its magnitude, equal to a × b is given by the formula
  
   
  a b sin ϕ, a, b noncollinear
a × b = (1.7)
0, a, b collinear.
1.2 Operations with Free Vectors 17

• Its sense is given by the right-hand rule: if the vector a × b is grasped in the right
hand and the fingers curl around from a to b through the angle ϕ, the thumb points
in the direction of a × b (Fig. 1.13).

Fig. 1.13 Graphical repre-

sentation of cross product

Proposition 1.33 (see [2], p. 17). The algebraic properties of the cross product of
the free vectors are:
1. Anticommutativiy:  
a × b = − b × a , (∀) a, b ∈ V3 ;

a×

2. t lineb = ta × b = a × tb, (∀) a, b ∈ V3 , (∀) t ∈ R;

3. Distributivity with respect to the vector addition:
 
a × b + c = a × b + a × c, (∀) a, b, c ∈ V3 ;
 
a + b × c = a × b + b × c, (∀) a, b, c ∈ V3 ;

4. a × a = 0, (∀) a ∈ V3 ;
5. a × 0 = 0 × a = 0, (∀) a ∈ V3 ;
6. if a, b ∈ V3 ,

a = a1 i + a2 j + a3 k
b = b1 i + b2 j + b3 k

then we obtain the analytical expression of the cross product:

 
 i j k 
 
a × b = (a2 b3 − a3 b2 ) i + (a3 b1 − a1 b3 ) j + (a1 b2 − a2 b1 ) k =  a1 a2 a3  (1.8)
 b1 b2 b3 
18 1 Vector Spaces

Proposition 1.34 (see [12], p. 170 and [2], p. 17).The geometric properties of the
cross product of the free vectors are:
  ⎫
a · a × b = 0⎬
1.   namely a × b is orthogonal both on a and b.
b · a × b = 0⎭
2. Lagrange identity:
 2  2  2
   
a × b = a2 b − a · b , (∀) a, b ∈ V3 ;
 
 
3. a × b is the positive area of the parallelogram determined by the vectors a and
b, having the same origin (Fig. 1.14).

Fig. 1.14 Geometric inter-

pretation of the cross product

−   → − →  →
 → −  → − −
→ − 
AOBCA = OA OB sin ≺ OA, OB = OA × OB .

But
AOBCA = 2AπOAB .

It results that − →
→ − 
OA × OB
AπOAB = .
2

Example 1.35 (see [9], p. 52). Knowing two sides AB = 3i − 4j BC = i + 5j of a

triangle, calculate the length of its height CD.
Solution
  √
AB = 9 + 16 = 5

 
i j k 

AB × BC =  3 −4 0  = 19k
1 5 0
      
AB × BC  AB CD   AB × BC  19
AπABC = =  
=⇒ CD =   = .
2 2 AB 5

We shall give a solution using Sage, too:

1.2 Operations with Free Vectors 19

Example 1.36 (see [5], p. 65). Let the vectors a = 3m − n b = m + 3n be such that
m = 3 and n = 2 and ≺ (m, n) = π2 . Determine the area of the triangle formed
by the vectors a and b.
Solution
According to Proposition 1.34, the area of the triangle formed by the vectors a
and b is:
1


Aπ = a × b .
2
We have

a × b = (3m − n) × (m + 3n) = 3m × (m + 3n) − n × (m + 3n)

× m +9m × n −n
= 3 m  × m −3 n 
× n = 10m × n.
=0 =m×n =0

We shall obtain
1 π
Aπ = · 10 · m × n = 5 m · n · sin = 30.
2 2
20 1 Vector Spaces

We can also achieve this result using Sage:

1.2.4.3 Mixed Product in V3

 
Definition 1.37 (see [2], p. 19). Let a, b, c ∈ V3 \ 0 be three free vectors. The
mixed product (also called
 the scalar triple product or box product) of these free
vectors is the scalar a · b × c .
 
Remark 1.38 (see [1], p. 121). If the free vectors a, b, c ∈ V3 \ 0 are noncoplanar,
then the volume
 of the parallelepiped
 determined by these three vectors (see Fig. 1.15)
 
is given by a · b × c .

Fig. 1.15 Geometric inter-

pretation of the mixed product

If we denote by:
• θ the angle between the vectors b and c,
• ϕ the angle between the vectors a and d = b × c, then
     
   
a· b × c = a·d = a d  ·cos ϕ = b × c a cos ϕ = ±Ah = ±Vparalelipiped ,
     
±2h
A

i.e.   
 
a · b × c  = Vparalelipiped .
1.2 Operations with Free Vectors 21

Proposition 1.39 (see [2], p. 19 and [1], p. 121). The mixed product of the free
vectors has the following properties:
   
1. a · b × c = c · a × b = b · (c × a)
   
2. a · b × c = −a · c × b
     
3. ta · b × c = a · tb × c = a · b × tc , (∀) t ∈ R;
       
4. a + b · c × d = a · c × d + b · c × d
5. Lagrange identity:

     a · c a · d 
a × b · c × d =  
b·c b·d
 
6. a · b × c = 0 if and only if:

(a) at least one of the vectors a, b, c is null;

(b) two of the three vectors are collinear;
(c) the vectors a, b, c are coplanar.
7. if

⎧
⎨ a = a1 i + a2 j + a3 k
b = b1 i + b2 j + b3 k
⎩
c = c1 i + c2 j + c3 k

then one obtains the analytical expression of the mixed product:

 
   a1 a2 a3 
a · b × c =  b1 b2 b3  . (1.9)
 c1 c2 c3 

Example 1.40 (see [9], p. 52). Find λ ∈ R such that the volume of the parallelepiped
determined by the vectors a = 2i − 3j + k, b = i + j − 2k, c = λi + 2j be equal to
5.
Solution
The volume of the parallelepiped determined by these vectors is
 
 2 −3 1 
 
V = ±  1 1 −2  = ± (10 + 5λ) .
λ 2 0 

From the condition V = 5 we deduce 10 + 5λ = ±5 i.e. λ1 = −1 λ2 = −3.

The same solution can be obtained in Sage:
22 1 Vector Spaces

Example 1.41 (see [13]). The following points A (3, 2, 1) , B (4, 4, 0) ,

C (5, 5, 0) , D (−1, 5, −1) are given in space.
(a) Determine if A, B, C, D are coplanar.
(b) If the points are not coplanar calculate the volume of the tetrahedron ABCD.
(c) In the same situation calculate the height of A on the plane (BCD).
(d) Prove that ≺ BAC ≡≺ BAD.

Solution
(a)+(b) The points A, B, C, D are coplanar if the volume of the tetrahedron ABCD
is equal to 0.
We have

AB = i + 2j − k, AC = 2i + 3j + 4k, AD = −4i + 3j − 2k.

According to the Proposition 1.39, we shall obtain


  
AB · AC × AD   1 2 −1 
1   
VABCD = = ·  2 3 4  = 10;
6 6  −4 3 −2 

therefore A, B, C, D are non-coplanar, that form a tetrahedron.

(c) We know that
AπBCD · d (A, (BCD))
VABCD = ,
6
where
1 
BC × BD ;
AπBCD =
2
it results that
3VABCD
d (A, (BCD)) = .
AπBCD
1.2 Operations with Free Vectors 23

As
BC = i + j + 5k, BD = −5i + j − k

we shall have  
 i j k 
 
BC × BD =  1 1 5  = −6i − 24j + 6k
 −5 1 −1 

and   √ √
BC × BD = 36 + 576 + 36 = 18 2;

therefore
1 √ √
AπBCD = · 18 2 = 9 2.
2
It will result
30 10
d (A, (BCD)) = √ = √ .
9 2 3 2

(d) We have:

⎫
cos ≺ BAC = cos ≺ AB, AC =  AB·AC
  = √ 4√ ⎪
⎬
AB·AC  6· 29

=⇒
cos ≺ BAD = cos ≺ AB, AD =  AB·AD
  = √ 4√ ⎪
⎭
AB·AD 6· 29

cos ≺ BAC = cos ≺ BAD =⇒≺ BAC ≡≺ BAD.

The solution of this problem in Sage is:

The following Sage code allows us to represent the tetrahedron ABCD:

24 1 Vector Spaces

1.3 Vector Spaces

Linear algebra can be regarded as the theory of the vector spaces, as a vector space
is a set of some objects or elements, that can be added together and multiplied by
the numbers (the result remaining an element of the set), so that the ordinary rules
of calculation to be valid.
An example of a vector space is the geometric vector space (the free vector
space), which plays a central role in physics and technology and illustrates the
importance of the vector spaces and linear algebra for all practical applications.
Let K be a commutative field and V be a non-empty set. The elements of K are
called scalars and we shall denote them by Greek letters, and the elements of V are
called vectors and we shall denote them by Latin letters, with bar above.
Definition 1.42 (see [14], p. 1). The set V is called a vector space over the field K
if the following are defined:
1. an internal algebraic operation, denoted additive “+”, + : V × V → V , called
addition, in respect to which V is a commutative group;
2. an external algebraic operation, denoted multiplicativ “ • ”, •: K × V → V ,
called multiplication by a scalar, that satisfies the axioms:
(a) (α+ β) a = αa + βa, (∀) α, β ∈ K and (∀) a ∈ V3
(b) α a + b = αa + αb, (∀) α ∈ K and (∀) a, b ∈ V3
(c) α (βa) = (αβ) a, (∀) α, β ∈ K and (∀) a ∈ V3
(d) 1 · a = a, (∀) a ∈ V3 .
1.3 Vector Spaces 25

If K is the field of the real numbers, V is called the real vector space. In the case
when K is the field of the complex numbers, V is called the complex vector space.
Examples of vector spaces
(1) The real arithmetic vector space with n - dimensions (Rn , +, ·)

Rn = R × R ×
. . . × R
n times

is the set of the ordered systems formed with n reale numbers, namely
   
Rn = x = x (1) , x (2) , . . . , x (n) |x (i) ∈ R, i = 1, n .

Let α ∈ R, x, y ∈ Rn , x = x (1) , x (2) , . . . , x (n) , y = y(1) , y(2) , . . . , y(n) ; then

def
 
+ : Rn × Rn → Rn , x + y = x (1) + y(1) , x (2) + y(2) , . . . , x (n) + y(n)

def
 
· : R × Rn → Rn , αx = αx (1) , αx (2) , . . . , αx (n) .

(2) The vector space of the polynomials in the indeterminate X with real coef-
ficients, of degree ≤ n, i.e. (Rn [X], +, ·)
Rn [X] means the set of the polynomials in the indeterminate X, with real coeffi-
cients, of degree ≤ n.
Let α ∈ R, P, Q ∈ Rn [X],

P (X) = a0 + a1 X + . . . + an X n , Q (X) = b0 + b1 X + . . . + bn X n ;

then
def
+ : Rn × Rn → Rn , P (X) + Q (X) = a0 + b0 + (a1 + b1 ) X + . . . + (an + bn ) X n
(1.10)
def
· : R × Rn → Rn , αP (X) = αa0 + αa1 X + . . . + αan X n . (1.11)

(3) The vector space (Mm,n (R), +, ·) of the m×n matrices, with real coefficients.

Example 1.43 (see [13]). Show that the set of the matrices of real numbers with m
lines and n columns forms a vector space on R, toward the addition of the matrices
and the scalar multiplication from R.
Solution
Stage I. One proves that (Mm,n (R), +, ·) is a commutative group (abelian group).
If A, B ∈Mm,n (R) then A + B ∈Mm,n (R), i.e. Mm,n (R) is a stable part in relation
to the addition of the matrices (the addition is well defined). Since are easily to check
the axioms on:
26 1 Vector Spaces

• Associativity:

(A + B) + C = A + (B + C) , (∀) A, B, C ∈ Mm,n (R);

• The existence of the neutral element:

⎛ ⎞
0 ··· 0
⎜ .. .. ⎟ ;
A + O = O + A, (∀) A ∈ Mm,n (R), O = ⎝ . .⎠
0 ··· 0

• The fact that any element is symmetrizable:

A + (−A) = (−A) + A = O, (∀) A ∈ Mm,n (R)

it results that (Mm,n (R), +, ·) is a group.

Since the matrix addition is commutative, namely

A + B = B + A, (∀) A, B ∈ Mm,n (R)

it results that (Mm,n (R), +, ·) is an abelian group.

Stage II. We check the axioms (a), (b), (c), which must satisfy by the scalar
multiplication:
(a) (α + β) A = αA + βA (∀) α, β ∈ R, (∀) A, B ∈ Mm,n (R)
(b) α (A + B) = αA + αB, (∀) α ∈ R, (∀) A, B ∈ Mm,n (R)
(c) α (βA) = (αβ) A (∀) α ∈ R, (∀) A, B ∈ Mm,n (R)
(d) 1 · A = A (∀) A ∈ Mm,n (R)
(4) The vector space (F, +, ·) of the functions defined on the set of real numbers
with real values.
If F = {f |f : R → R} , f , g ∈ F, α ∈ R then
def
+ : F × F → F, (f + g) (x) = f (x) + g (x)

def
· : R × F → F, (αf ) (x) = αf (x) .

(5) The free vector space, denoted with V3 .

Theorem 1.44 (see [6], p. 9). If V is a real vector space then the following statements
occur:
(i) 0 · a = 0 (∀) a ∈ V
(ii) α · 0 = 0 (∀) α ∈ R
(iii) (−1) · a = −a (∀) a ∈ V
(iv) if (∀) α ∈ K, (∀) a ∈ V such that α · a = 0 then α = 0 or a = 0.
1.3 Vector Spaces 27

Definition 1.45 (see [14], p. 6). A vector system {x 1 , . . . , x m } from the space vector V
over K is linearly dependent if there are the scalars α(i) ∈ K, α(i) = 0, (∀) i = 1, m,
such that
α(1) x 1 + . . . + α(m) x m = 0.

If the previous relationship occurs only if

α(1) = . . . = α(m) = 0

then the system is liniarly independent.

Definition 1.46 (see [1], p. 6). Let V be a vector space over the field K and S =
{x 1 , . . . , x m } be a system of vectors in V . We say that the vector v ∈ V is a linear
combination of elements from S if

⎦
m
v= α(i) x i , x i ∈ S, α(i) ∈ K, i = 1, m.
i=1

Definition 1.47 (see [14], p. 6). Let V be a vector space over the field K. The finite
system S of vectors from V is called system of generators for V if any vector from
V is a linear combination of vectors from S.
Definition 1.48 (see [6], p. 12). Let V be a space vector over the field K. The finite
system B of vectors from V is called basis of V if:
(a) B is linearly independent;
(b) B is a system of generators for V .
 
Example 1.49 (see [13]). B = Eij , i = 1, m, j = 1, n is a basis in Mm,n (R).
Solution
We consider ⎛ ⎞
a11 · · · a1n
⎜ .. .. ⎟ .
A ∈ Mm,n (R), A = ⎝ . . ⎠
am1 · · · amn

We can write
⎛ ⎞ ⎛ ⎞
a11 0 · · · 0 0 0 a12 · · · 0 0
⎜ 0 0 ··· 0 0 ⎟ ⎜ 0 0 ··· 0 0 ⎟
A=⎜ ⎟ ⎜ ⎟
⎝ ··· ··· ··· ··· ···⎠ + ⎝··· ··· ··· ··· ···⎠ + ··· +
0 0 ··· 0 0 0 0 ··· 0 0
⎛ ⎞ ⎛ ⎞
0 0 · · · 0 a1n 0 0 ··· 0 0
⎜ 0 0 ··· 0 0 ⎟ ⎜ ⎟
⎜ ⎟ + ··· + ⎜ 0 0 ··· 0 0 ⎟
⎝··· ··· ··· ··· ··· ⎠ ⎝··· ··· ··· ··· ··· ⎠
0 0 ··· 0 0 0 0 · · · 0 amn
28 1 Vector Spaces
⎛ ⎞ ⎛ ⎞
1 0 ··· 0 0 0 1 ··· 0 0
⎜ 0 0 ··· 0 0 ⎟ ⎜ 0 0 ··· 0 0 ⎟
= a11 ⎜
⎝···
⎟ + a12 ⎜ ⎟
··· ··· ··· ···⎠ ⎝··· ··· ··· ··· ···⎠
0 0 ··· 0 0 0 0 ··· 0 0
⎛ ⎞
0 0 ··· 0 0
⎜ 0 0 ··· 0 0 ⎟
+ · · · + amn ⎜ ⎟
⎝··· ··· ··· ··· ···⎠.
0 0 ··· 0 1

We denote
⎛ j ⎞
0 ··· 0
⎜ .. .. ⎟.
Eij = i⎝ . 1 .⎠
0 ··· 0

Using this notation, we have

A = a11 E11 + a12 E12 + · · · + amn Emn ,

namely  
B = Eij , i = 1, m, j = 1, n

is a system of generators.
We consider the linear null combination of the matrices from B:
⎛ ⎞
a11 · · · a1n
⎦
m ⎦
n
⎜ .. ⎟ = O ⇐⇒ a = 0,
αij Eij = O ⇐⇒ ⎝ ... . ⎠ ij (∀) i = 1, m, j = 1, n;
i=1 j=1 am1 · · · amn

therefore  
B = Eij , i = 1, m, j = 1, n

is linearly independent.
It results that  
B = Eij , i = 1, m, j = 1, n

is a basis in Mm,n (R).

1.3 Vector Spaces 29

A solution in Sage will be given to prove that, in the case of M2,3 (R):

Definition 1.50 (see [6], p. 13). Let V be a vector space over K.

(a) The number n ∈ N∗ of the vectors from a basis of a vector space V is called the
finite dimension of V (over K ) and one denotes by:
dimV or dimK V .
(b) In the case when V is not finite generated is said to be of infinite dimension and
one denotes dimK V = ∞.
Remark 1.51 (see [6], p. 13). dim Mm,n (R) = m · n.
Remark 1.52 (see [1], p. 10). Two arbitrary bases in V have the same number of
vectors.
Example 1.53 (see [13]). Let
 
Rn [X] = a0 + a1 X + . . . + an X n |αi ∈ R, i = 0, n, n ∈ N∗

be the set of polynomials in the indeterminate X, with real coefficients, of degree

≤ n.
(i) Show that the set Rn [X] is a real vector space, denoted by (Rn [X], +, ·) with
the usual operations of addition of polynomials and of multiplication of the
polynomials with scalars from R.
(ii) Find the dimension of Rn [X].
30 1 Vector Spaces

Solution
(i) Let be α ∈ R, P, Q ∈ Rn [X], P (X) = a0 + a1 X + . . . + an X n , Q (X) =
b0 + b1 X + . . . + bn X n .
Stage I. It proves that (Rn [X], +, ·) is an abelian group, where “+” and “·” are
defined in (1.10) and respectively (1.11).
As
deg ree (P + Q) ≤ max (deg ree (P) , deg ree (Q)) ≤ n

it results
P + Q ∈ Rn [X].

We showed that for P, Q ∈ Rn [X] we have P + Q ∈ Rn [X], namely Rn [X] is

a stable parts in relation to the addition of the polynomials (the addition is well
defined).
As the axioms on associativity, the existence of neutral element and the fact that
any element is simetrizable are easily checked it results that (Rn [X], +, ·) is a group.
As the addition of the polynomials is commutative it results that (Rn [X], +, ·) is
an abelian group.
Stage II. We check the axioms (a), (b), (c), (d) from the Definition 1.42, which
must satisfy the scalar multiplication.
(a) (α + β) P = αP + βP, (∀) α, β ∈ R and (∀) P ∈ Rn [X]
(b) α (P + Q) = αP + αQ, (∀) α ∈ R and (∀) P, Q ∈ Rn [X]
(c) α (βP) = (αβ) P, (∀) α, β ∈ R and (∀) P ∈ Rn [X]
(d) 1 · P = P, (∀) P ∈ Rn [X].

(ii) We denote ⎧
⎪
⎪ P0 (X) = 1
⎪
⎨ P1 (X) = X
..
⎪
⎪ .
⎪
⎩
Pn (X) = X n .

Let be the null linear combination:

α(0) P0 (X) + α(1) P1 (X) + . . . + α(n) Pn (X) = ORn [X] .

As
α(0) + α(1) X + . . . + α(n) X n = ORn [X]

it results that α(0) = . . . = α(n) = 0; so the family of polynomials B =

{P0 , P1 , . . . , Pn } is linearly independent.
1.3 Vector Spaces 31

Note that B = {P0 , P1 , . . . , Pn } generates Rn [X] since (∀) P ∈ Rn [X] :

P (X) = a0 + a1 X + . . . + an X n = α(0) P0 (X) + α(1) P1 (X) + . . . + α(n) Pn (X) ;

therefore B = {P0 , P1 , . . . , Pn } is a system of generators.

As B = {P0 , P1 , . . . , Pn } is linearly independent and a a system of generators for
Rn [X], from the Definition 1.48 it results that B is a base for Rn [X].
It results that dim Rn [X] = n + 1.
We can prove that in Sage, too:

Example 1.54 (see [13]). Prove that the set

V = {f : R → R|f (x) = α sin (x + β) , α, β ∈ R}

equipped with the normal operations of addition of functions and of multiplication

of functions with real numbers is a vector space. Determine the dimension of V .
Solution


Let be f , g ∈ V , f (x) = α sin (x + β) , g (x) = α sin x + β  . We shall check
if (V , +) is a commutative group.
We show only that f + g ∈ V , namely (∃) α0 , β0 ∈ R such that

(f + g) (x) = α0 sin (x + β0 ) , (∀) x ∈ R.

32 1 Vector Spaces

It must be the case that

α sin (x + β) + α sin x + β  = α0 sin (x + β0 ) , (∀) x ∈ R,

i.e.

α sin x cos β + α sin β cos x + α sin x cos β  + α sin β  cos x

= α0 (sin x cos β0 + sin β0 cos x) ⇔

α cos β + α cos β  sin x + α sin β + α sin β  cos x

= α0 cos β0 sin x + α0 sin β0 cos x, (∀) x ∈ R.

We denote: 
A = α cos β + α cos β 
B = α sin β + α sin β  .

We deduce 
A = α0 cos β0
B = α0 sin β0 .

It results that
sin β0 B B B
= ⇒ tg β0 = ⇒ β0 = arctg + kπ, k ∈ Z,
cos β0 A A A

where tg x means the tangent function.

Although an infinite number of values are obtained for β0 , the function
α0 sin (x + β0 ) has the same value, so the sum of two functions from V is well
defined. Immediately, we can determine

A
α0 = .
cos β0

Therefore (∃) α0 , β0 ∈ R such that

(f + g) (x) = α0 sin (x + β0 ) , (∀) x ∈ R.

(ii) Let be f ∈ V ⇒ (∃) α, β ∈ R such that

f (x) = α sin (x + β) , (∀) x ∈ R.

We have

f (x) = α sin x cos β + α sin β cos x = (α cos β) sin x + (α sin β) cos x.

1.3 Vector Spaces 33

It is known that
π   π
cos x = sin − x = − sin x − .
2 2
We denote  π
f1 (x) = sin x, f2 (x) = − sin x − .
2
Hence
f = αf1 cos β + αf2 sin β.

We prove that f1 , f2 ∈ V .
As

sin x
= 1 · sin (x + 0) ⇒
f1 ∈ V , for α = 1, β = 0;
f1 (x) =
f2 (x) = − sin x − π2 = (−1) sin x − π2 ⇒ f1 ∈ V , for α = −1, β = − π2 .

Thereby, B = {f1 , f2 } is a system of generators for V .

We consider the null linear combination

α(1) f1 + α(2) f2 = OV ⇔ α(1) f1 (x) + α(2) f2 (x) = 0, (∀) x ∈ R;

it results that
α(1) sin x + α(2) cos x = 0, (∀) x ∈ R.

If x = 0 from the previous relation we get α(2) = 0, while for x = π2 we obtain

α(1) = 0.
It results that B = {f1 , f2 } is linear independently. Therefore B = {f1 , f2 } is a basis
for V ⇒ dim V = 2.
We can check the previous results in Sage, too:
34 1 Vector Spaces

Example 1.55 (see [13]). Let V be a real vector space.

(i) Show that CV = V × V = {(u1 , u2 )} is complex vector space (the complexifi-
cation of the real vector space V ) in relation to the operations:
+ : CV × CV → CV defined by (u1 , u2 ) + (v1 , v2 ) = (u1 + v1 , u2 + v2 ) , · : C ×
CV → CV defined by (α + iβ) · (v1 , v2 ) = (αv1 − βv2 , αv2 + βv1 ) , (∀) α, β ∈
R.
(ii) If dimR V = n < ∞ determine dimC CV .

Solution

(i) Check the axioms from the definition of the vector space.
(ii) Let be B = {a1 , a2 , . . . , an } a basis for V . We intend to show that B =
{ia1 , ia2 , . . . , ian } a basis for CV .

Let be (u1 , u2 ) ∈ CV , u1 , u2 ∈ V .
As B is a basis for V it results that
• (∀) u1 ∈ V , (∃) x (i) ∈ K, i = 1, n unique, such that

u1 = x (1) a1 + . . . + x (n) an ;

• (∀) u2 ∈ V , (∃) y(i) ∈ K, i = 1, n unique, such that

u2 = y(1) a1 + . . . + y(n) an .
1.3 Vector Spaces 35

We compute
 
(u1 , u2 ) = x (1) a1 + . . . + x (n) an , y(1) a1 + . . . + y(n) an
   
= x (1) a1 , y(1) a1 + . . . + x (n) an , y(n) an
 
 

= y(1) − ix (1) · 0, a1 + . . . + y(n) − ix (n) · 0, an .




It results that B = 0, a1 , . . . , 0, an is a system of generators for C V .
Let be a null combination of elements from B :

 
 

α(1) + iβ (1) · 0, a1 + . . . + α(n) + iβ (n) · 0, an = 0, 0 . (1.12)

From (1.12) we have

   

−β (1) a1 , α(1) a1 + . . . + −β (n) an , α(n) an = 0, 0 ⇔

 

−β (1) a1 − . . . − β (n) an , α(1) a1 + . . . + α(n) an = 0, 0 ,

namely
− β (1) a1 − . . . − β (n) an = 0 (1.13)

and
α(1) a1 + . . . + α(n) an = 0. (1.14)

As a1 , . . . , an are linearly independent from (1.13) and (1.14) it results


β (1) = . . . = β (n) = 0
(1.15)
α(1) = . . . = α(n) = 0.



From (1.12) and (1.15) we deduce that B = 0, a1 , . . . , 0, an is linearly
independent.
Thus, B is a basis and dimC CV = n.
Checking this problem in Sage, for n = 3 we have:
36 1 Vector Spaces

Proposition 1.56 (see [1], p. 12). Let V be a space vector of n dimension and
B = {a1 , . . . , an } ⊂ V . Then:
(a) if B is liniarly independent it results that B is a basis,
(b) if B is system of generators it results that B is a basis.

Theorem 1.57 (see [14], p. 7). Let V be a space vector over K and B = {a1 , . . . , an } ⊂
V . Then B is a basis of V if and only if any vectors from V can be written in an unique
way as a linear combination of the vectors from B.
Definition 1.58 (see [14], p. 7). The unique scalars x (i) ∈ K, i = 1, m that appear as
coefficients in the writing of the vector x ∈ V as a linear combination of the vectors
from the basis B are called the coordinates of the vector x relative to the basis B.
We shall denote by x B the column matrix formed with the coordinates of the vector
x relative to the basis B; therefore
⎛ ⎞
x (1)
⎜ x (2) ⎟
⎜ ⎟
xB = ⎜ . ⎟ .
⎝ .. ⎠
x (m)

Remark 1.59 (see [6], p. 14). The writing of a vector in a basis is unique.
Example 1.60 (see [9], p. 31). In R4 the following vectors are given:

v1 = (1, 1, 2, 1) , v2 = (1, −1, 0, 1) , v3 = (0, 0, −1, 1) , v4 = (1, 2, 2, 0) .

1.3 Vector Spaces 37

(i) Show that these vectors form a basis.

(ii) Find the coordinates of the vector v = (1, 1, 1, 1) relative to this basis.

Solution
(i) dim R4 = 4 ⇒ it suffices to show that the vectors are linearly independent.
Let be the null linear combination

α(1) v1 + α(2) v2 + α(3) v3 + α(4) v4 = 0.

It results that
     
α(1) , α(1) , 2α(1) , α(1) + α(2) , −α(2) , 0, α(2) + 0, 0, −α(3) , α(3) +
 
α(4) , 2α(4) , 2α(4) , 0 = (0, 0, 0, 0) .

One obtains the system:

⎧ (1)
⎪
⎪ α + α(2) + α(4) = 0
⎨ (1)
α − α(2) + 2α(4) = 0
⎪
⎪ 2α(1) − α(3) + 2α(4) = 0
⎩ (1)
α + α(2) + α(3) = 0.

As  
1 1 0 1 

1 −1 0 2 
d =  = −4 = 0
2 0 −1 2 
1 1 1 0

it results that the system has a unique solution

α(1) = α(2) = α(3) = α(4) = 0.

It follows that the vectors are linearly independent.

(ii) Let be
v = α(1) v1 + α(2) v2 + α(3) v3 + α(4) v4 .

We have
   
(1, 1, 1, 1) = α(1) , α(1) , 2α(1) , α(1) + α(2) , −α(2) , 0, α(2)
   
+ 0, 0, −α(3) , α(3) + α(4) , 2α(4) , 2α(4) , 0 .
38 1 Vector Spaces

One obtains the system:

⎧ (1)
⎪
⎪ α + α(2) + α(4) = 1
⎨ (1)
α − α(2) + 2α(4) = 1
⎪
⎪ 2α(1) − α(3) + 2α(4) = 1
⎩ (1)
α + α(2) + α(3) = 1

whose solution is
1 (2) 1 1 1
α(1) = , α = , α(3) = , α(4) = .
4 4 2 2
Solving this problem with Sage, we achieve:

Hence ⎛ ⎞
1/4
⎜ 1/4 ⎟
vB = ⎜ ⎟
⎝ 1/2 ⎠ .
1/2
 
Let B1 = {e1 , . . . , en } , B1 = f 1 , . . . , f n be two bases of V x ∈ V and

x = x (1) a1 + . . . + x (n) an (1.16)

x = y(1) f 1 + . . . + y(n) f n . (1.17)

We suppose that the vectors from B2 can be written as a linear combination of the
vectors from the basis B1 :
(1) (n)
f 1 = α1 e1 + . . . + α1 en (1.18)
..
.
f n = αn(1) e1 + . . . + αn(n) en .
1.3 Vector Spaces 39

Writing the column coefficients of these linear combinations we get (see [14],
p. 8) the matrix ⎛ ⎞
(1) (1)
α1 . . . αn
⎜ (2) ⎟
⎜ α . . . αn(2) ⎟
M(B1 ,B2 ) = ⎜ 1 ⎟, (1.19)
⎝ ... ... ... ⎠
(n) (n)
α1 . . . αn

which represents the transition matrix from the basis B1 to the basis B2 .
Remark 1.61 (see [6], p. 18). The matrix M(B1 ,B2 ) is always a nonsingular matrix
due to the linear independence of the basis vectors.
Example 1.62 (see [13]). Find for the space of  polynomials byat most four degree,
the transition matrix from the basis B1 = 1, X, X 2 , X 3 , X 4 to the basis B2 =
 
1, (X + 1) , (X + 1)2 , (X + 1)3 , (X + 1)4 .
Solution
The vectors from B2 can be written as a linear combination of the vectors from
B1 thus:

1=1
X +1 = 1+X
(X + 1)2 = 1 + 2X + X 2
(X + 1)3 = 1 + 3X + 3X 2 + X 3
(X + 1)4 = 1 + 4X + 6X 2 + 4X 3 + X 4 .

It results that ⎛ ⎞
1 1 1 1 1
⎜0 1 2 3 4⎟
⎜ ⎟
M(B1 ,B2 ) =⎜
⎜0 0 1 3 6⎟⎟.
⎝0 0 0 1 4⎠
0 0 0 0 1

The same matrix can be achieved using Sage:

40 1 Vector Spaces

Substituting (1.18) into (1.17) we obtain

   
(1) (n)
x = y(1) α1 e1 + . . . + α1 en + . . . + y(n) αn(1) e1 + . . . + αn(n) en
 
(1)
= y(1) α1 + . . . + y(n) αn(1) e1 + . . . +
 
(n)
y(1) α1 + . . . + y(n) αn(n) en . (1.20)

Due to the uniqueness of writing a vector into a basis, from (1.16) and (1.20) it
results (see [6], p. 18)
⎧ (1) (1) (1) (n) (1)
⎨ x = y α1 + . . . + y αn
⎪
.. (1.21)
⎪ .
⎩ (n) (n) (n)
x = y α1 + . . . + y(n) αn .
(1)

The formulas (1.21) are (see [6], p. 18) the formulas of changing a vector
coordinates when the basis of a vector space one changes.
The relations (1.21) can be written in the matrix form as:

x B1 = M(B1 ,B2 ) · x B2 (1.22)

or
x B2 = M−1
(B1 ,B2 ) · x B1 . (1.23)

Example 1.63 (see [13]). In the arithmetic vector space R3 the following vectors are
considered:

a1 = (2, −1, 2) , a2 = (1, −1, 2) , a3 = (0, 3, 2) ,

b1 = (0, 1, −1) , b2 = (2, 1, 1) , b3 = (−1, 2, 1) ,
x = (−1, 2, 3) .

(a) Prove that B1 = {a1 , a2 , a3 } is a basis of R3 .

(b) Determine the coordinates
  x relative to the basis B1 .
of
(c) Prove that B2 = b1 , b2 , b3 is a new basis of R3 and write the transition matrix
from the basis B1 to the basis B2 .
(d) Write the formulas of changing a vector coordinates when one passes from the
basis B1 to the basis B2 .

Solution
(a) dim R3 = 3 ⇒ it is enough to prove that the vectors are linearly independent.
Consider the null linear combination
1.3 Vector Spaces 41

α(1) a1 + α(2) a2 + α(3) a3 = 0.

The, it results that

     
2α(1) , −α(1) , 2α(1) + α(2) , −α(2) , 2α(2) + 0, 3α(3) , 2α(3) = (0, 0, 0) .

One obtains the system:

⎧
⎨ 2α(1) + α(2) = 0
−α(1)
− α(2) + 3α(3) = 0
⎩ (1)
2α + 2α(2) + 2α(3) = 0.

As  
 2 1 0
 
d =  −1 −1 3  = −8 = 0
 2 2 2

it results that the system has a unique solution

α(1) = α(2) = α(3) = 0.

It follows that the vectors are linearly independent.

(b) Let
x = α(1) a1 + α(2) a2 + α(3) a3 .

We have
     
(−1, 2, 3) = 2α(1) , −α(1) , 2α(1) + α(2) , −α(2) , 2α(2) + 0, 3α(3) , 2α(3) .

We obtain the system:

⎧
⎨ 2α(1) + α(2) = −1
−α(1)
− α(2) + 3α(3) = 2
⎩ (1)
2α + 2α(2) + 2α(3) = 3

whose solution is
13 (2) 9 7
α(1) = − , α = , α(3) = .
8 4 8
Hence ⎛ ⎞
−13/8
x B = ⎝ 9/4 ⎠ .
7/8

(c) One proceeds similarly that B2 is also a basis.

42 1 Vector Spaces

We know that ⎛ ⎞
(1) (1) (1)
α1 α2 α3
⎜ ⎟
M(B1 ,B2 ) = ⎝ α1(2) α2(2) α3(2) ⎠ .
(3) (3) (3)
α1 α2 α3

We can write
(1) (2) (3)
b1 = α1 a1 + α1 a2 + α1 a3 (1.24)

(1) (2) (3)

b2 = α2 a1 + α2 a2 + α2 a3 (1.25)

(1) (2) (3)

b3 = α3 a1 + α3 a2 + α3 a3 . (1.26)

From (1.24) we deduce

(1) (2) (3)

(0, 1, −1) = α1 (2, −1, 2) + α1 (1, −1, 2) + α1 (0, 3, 2) . (1.27)

From (1.27) it results the system

⎧
⎪ (1) (2)
⎨ 2α1 + α1 = 0
(1) (2) (3)
−α1 − α1 + 3α1 = 1
⎪
⎩ (1) (2) (3)
2α1 + 2α1 + 2α1 = −1

which has the solution

(1) (2) (3)
α1 = 0.625, α1 = −1.25, α1 = 0.125.

From (1.25) we deduce

(1) (2) (3)

(2, 1, 1) = α2 (2, −1, 2) + α2 (1, −1, 2) + α2 (0, 3, 2) . (1.28)

From (1.28) it results the system

⎧
⎪ (1) (2)
⎨ 2α2 + α2 = 2
(1) (2) (3)
−α2 − α2 + 3α2 = 1
⎪
⎩ (1) (2) (3)
2α2 + 2α2 + 2α2 = 1

which has the solution

(1) (2) (3)
α2 = 1.875, α2 = −1.25, α2 = 0.125.

From (1.26) we deduce

(1) (2) (3)
(−1, 2, 1) = α3 (2, −1, 2) + α3 (1, −1, 2) + α3 (0, 3, 2) . (1.29)
1.3 Vector Spaces 43

From (1.29) it results the system

⎧
⎪ (1) (2)
⎨ 2α3 + α3 = −1
(1) (2) (3)
−α3 − α3 + 3α3 = 2
⎪
⎩ (1) (2) (3)
2α3 + 2α3 + 2α3 = 1

which has the solution

(1) (2) (3)
α3 = −0.875, α3 = 0.75, α3 = 0.625.

One obtains ⎛ ⎞
0.625 1.875 −0.875
M(B1 ,B2 ) = ⎝ −1.25 −1.75 0.75 ⎠ .
0.125 0.375 0.625

(d) Let
vB1 = x (1) a1 + x (2) a2 + x (3) a3

vB2 = y(1) b1 + y(2) b2 + y(3) b3 .

We shall have
⎛ (1) ⎞ ⎛ (1) ⎞ ⎧ (1)
x y ⎨ x = 0.625y(1) + 1.875y(2) − 0.875y(3)
⎝ x (2) ⎠ = M(B1 ,B2 ) · ⎝ y(2) ⎠ ⇒ x (2) = −1.25y(1) − 1.75y(2) + 0.75y(3)
(3) (3) ⎩ (3)
x y x = 0.125y(1) + 0.375y(2) + 0.625y(3) .

We shall give a solution in Sage, too:

44 1 Vector Spaces

Example 1.64. In the vector space M2,2 (R) the following matrices are considered:
!
" ! " ! " ! "
10 1 −1 1 −1 1 −1
C1 = , C2 = , C3 = , C4 = ,
00 0 0 1 0 1 −1
! " ! " ! " ! "
00 0 0 0 1 −1 −1
A1 = , A2 = , A3 = , A4 = .
01 −1 1 −1 1 −1 1

One requires:
(a) Prove that both B1 = {C1 , C2 , C3 , C4 } and B2 = {A1 , A2 , A3 , A4 } are bases for
M2,2 (R).
(b) Find the transition matrix from the basis B1 to the basis B2 .
(c) Write the formulas of changing a vector coordinates when one passes from the
basis B1 to the basis B2 .

Solution
(a) Let

α(1) A1 + α(2) A2 + α(3) A3 = O2 .

We have
! " ! " ! " ! " ! "
00 0 0 0 1 −1 −1 00
α(1) + α(2) + α(3) + α(4) = ⇔
01 −1 1 −1 1 −1 1 00
! " ! " ! " ! " ! "
0 0 0 0 0 α(3) −α(4) −α(4) 00
+ + + = .
0 α(1) −α(2) α(2) −α(3) α(3) −α(4) α(4) 00

It results that −α(4) = 0; therefore α(4) = 0.

We obtain the system:
⎧
⎨ α(3) − α(4) = 0 ⇒ α(3) = α(4)
−α(2) − α(3) − α(4) = 0 ⇒ α(2) = 0
⎩ (1)
α + α(2) + α(3) + α(4) = 0 ⇒ α(1) = 0.

Hence: α(1) =# α(2) = α(3) = α(4) = 0 ⇒ B2 linearly independent.

dim M2,2 (R) = 4 Proposition 1.56
⇒ B2 basis.
B2 has 4 elements
Similarly, one proves that B1 is a basis, too.
(b) Using (1.19) we have
⎛ (1) (1) (1) (1)
⎞
α1 α2 α3 α4
⎜ (2) (2) (2) (2) ⎟
⎜ α1 α2 α3 α4 ⎟
M(B1 ,B2 ) = ⎜ (3) (3) (3) (3) ⎟ .
⎝ α1 α2 α3 α4 ⎠
(4) (4) (4) (4)
α1 α2 α3 α4
1.3 Vector Spaces 45

On the basis of the formulas from (1.18) we obtain:

(1) (2) (3) (4)

A1 = α1 C1 + α1 C2 + α1 C3 + α1 C4 (1.30)

(1) (2) (3) (4)

A2 = α2 C1 + α2 C2 + α2 C3 + α2 C4 (1.31)

(1) (2) (3) (4)

A3 = α3 C1 + α3 C2 + α3 C3 + α3 C4 (1.32)

(1) (2) (3) (4)

A4 = α4 C1 + α4 C2 + α4 C3 + α4 C4 . (1.33)

From (1.30) we deduce

! " ! " ! " ! " ! "
00 (1) 10 (2) 1 −1 (3) 1 −1 (4) 1 −1
= α1 + α1 + α1 + α1 . (1.34)
01 00 0 0 1 0 1 −1

From (1.34) it results the system:

⎧ (1) (2) (3) (4) (1)
⎪
⎪ α1 + α1 + α1 + α1 = 0 ⇒ α1 = 0
⎪
⎨ (2) (3) (4) (2)
−α1 − α1 − α1 = 0 ⇒ α1 = 0
⎪ (3) (4) (3)
⎪
⎪ α1 + α1 = 0 ⇒ α1 = 1
⎩ (4) (4)
−α1 = 0 ⇒ α1 = −1.

Finally, one obtains

⎛ ⎞
0 0 1 −2
⎜ 0 1 0 2 ⎟
M(B1 ,B2 ) =⎜
⎝ 1
⎟.
0 0 0 ⎠
−1 −1 −1 −1

(c) Let

x B1 = x (1) C1 + x (2) C2 + x (3) C3 + x (4) C4

x B2 = y(1) A1 + y(2) A2 + y(3) A3 + y(4) A4 .

Substituting these relations into (1.22) we shall have

⎛ ⎞ ⎛ (1) ⎞ ⎧
x (1) y ⎪
⎪ x (1) = y(3) − 2y(4)
⎜ x (2) ⎟ ⎜ y(2) ⎟ ⎨
⎜ ⎟ ⎜ ⎟ x (2) = y(2) + 2y(4)
⎝ x (3) ⎠ = M(B1 ,B2 ) · ⎝ y(3) ⎠ ⇒ ⎪ x (3) = y(1)
⎪
⎩ (4)
x (4) y (4) x = −y − y(2) − y(3) − y(4) .
(1)
46 1 Vector Spaces

The solution using Sage is:

1.4 Vector Subspaces

Definition 1.65 (see [14], p. 8). Let V be a vector space over the n -dimensional
K field. The nonempty subset W of V it’s called a vector subspace of V if the
following conditions are satisfied:

(1) x + y ∈ W (∀) x, y ∈ W
(2) αx ∈ W (∀) α ∈ K, (∀) x ∈ W .

Proposition 1.66 (see [14], p. 8). Let V be a vector space over the n-dimensional K
field. The nonempty subset W of V it’s called a vector subspace of V if and only if

αx + βy ∈ W , (∀)α, β ∈ K, (∀)x, y ∈ W .

Examples of vector subspaces



(1) W1 = x (1) , x (2) , . . . , x (n) ∈ Rn |x (1) + x (2) + . . . + x (n) = 0 is a subspace
vector of Rn . 
(2) Mns (R) = A ∈ Mn (R)|At = A the vector subspace of the symmetric matrices
is a vector subspace of Mn (R).
We shall prove that the subset Mns (R) is a vector subspace of Mn (R) and we shall
determine its dimension, highlighting a basis of it.
1.4 Vector Subspaces 47

Let α, β ∈ R and A, B ∈ Mns (R). It results that At = A and Bt = B.

We shall have

(αA + βB)t = αAt + βBt = αA + βB ⇒ αA + βB ∈ Mns (R),

namely Mns (R) is a vector subspace of Mn (R).

⎛ ⎞
a11 a12 . . . a1n
⎜ a12 a22 . . . a2n ⎟
⎜ ⎟
A ∈ Mns (R) ⇒ A = ⎜ . .. .. .. ⎟ .
⎝ .. . . . ⎠
a1n a2n . . . ann

We can write
⎛ ⎞ ⎛ ⎞
1 0 ... 0 0 0 1 ... 0 0
⎜ 0 0 ... 0 0 ⎟ ⎜ 1 0 ... 0 0 ⎟
A = a11 ⎜
⎝... ...
⎟ + a12 ⎜ ⎟
... ... ...⎠ ⎝... ... ... ... ...⎠
0 0 ... 0 0 0 0 ... 0 0
     
=F1 =F2
⎛ ⎞
0 0 ... 0 1 ⎛ ⎞
0 0 ... 0 0
⎜ 0 0 ... 0 0 ⎟ ... 0 0 ⎠
+ . . . + a1n ⎜
⎝... ...
⎟ + a22 ⎝ 0 1
... ... ...⎠ ... ... ... ... ...
1 0 ... 0 0 0 0 ... 0 0
     
=Fn+1
=Fn
⎛ ⎞
0 0 ... 0 0
+ . . . + ann ⎝ . 0. . . 0. . . . . 0 0 ⎠.
... ... ...
0 0 ... 0 1
  
=Fn(n+1)/2

Using this notation, we have

A = a11 F1 + a12 F2 + . . . + a1n Fn + a22 Fn+1 + . . . + ann Fn(n+1)/2

 
namely B = Fi , i = 1, n (n + 1) /2 is a system of generators.
We consider the linear combination of the null matrices from B
⎛ ⎞
α11 . . . α1n
⎝ .. .
.. ⎠ = O ⇔ αij = αji = 0, (∀) i = 1, n, j = 1, n;
.
α1n . . . αnn
 
therefore B = Fi , i = 1, n (n + 1) /2 is linearly independent.
It results that B is a basis in Mns (R); hence
48 1 Vector Spaces

n (n + 1)
dim Mns (R) = .
2
We can check that in Sage, too:

 
(3) Mna (R) = A ∈ Mn (R)|At = −A the vector subspace of the antisymmetric
matrices is a vector subspace of Mn (R).
We shall prove that the subset Mna (R) is a vector subspace of Mn (R) and we shall
determine its dimension, highlighting a basis of its.
Let α, β ∈ R and A, B ∈ Mna (R). It results that At = −A and Bt = −B.
We shall have

(αA + βB)t = αAt + βBt = −αA − βB ⇒ αA + βB ∈ Mna (R),

i.e. Mna (R) is a vector subspace of Mn (R).

⎛ ⎞
a11 a12 . . . a1n
⎜ −a12 a22 . . . a2n ⎟
⎜ ⎟
A ∈ Mna (R) ⇒ A = ⎜ . .. .. .. ⎟ .
⎝ .. . . . ⎠
−a1n −a2n . . . ann
1.4 Vector Subspaces 49

We can write
⎛ ⎞ ⎛ ⎞
0 1 ... 0 0 0 0 ... 0 1
⎜ −1 0 ... 0 0 ⎟ ⎜ ... 0 ⎟
⎜
A = a12 ⎝ ⎟ + . . . + a1n ⎜ 0 0 0 ⎟
... ... ... ... ...⎠ ⎝... ... ... ... ...⎠
0 0 ... 0 0 1 0 ... 0 0
     
=G1 =Gn−1
⎛ ⎞
0 0 ... 0 0
⎜ 0 0 ... 0 0 ⎟
+ . . . + ann ⎜
⎝...
⎟.
... ... ... ...⎠
0 0 ... 0 1
  
=Gn(n−1)/2

Using this notation, we have

A = a12 G1 + . . . + a1n Gn−1 + . . . + ann Gn(n−1)/2

 
i.e. B= Gi , i = 1, n (n − 1) /2 is a system of generators.
We consider the linear combination of the null matrices from B:
⎛ ⎞
0 α12 . . . α1n
⎜ −α12 0 . . . α2n ⎟
⎜ ⎟
⎜ .. .. .. .. ⎟ = O ⇔ αij = 0, (∀) i = 1, n, j = 1, n, i < j;
⎝ . . . . ⎠
−α1n −α2n . . . 0
 
therefore B= Gi , i = 1, n (n − 1) /2 is linearly independent.
It results that B is a basis in Mna (R); hence

n (n − 1)
dim Mna (R) = .
2
Using Sage we shall have:
50 1 Vector Spaces

Proposition 1.67 (see [6], p. 20). The set of the solutions corresponding to a linear
and homogeneous system with m equations and n unknowns is a vector subspace of
Rn , having the dimension n − r , r being the rank of the associated matrix system A.
Example 1.68 (see [13]). In the vector space R5 we consider
 
W = x = (x1 , x2 , x3 , x4 , x5 ) ∈ R5

which checks the system

⎧
⎨ x1 + x2 + x3 + x4 + x5 = 0
x2 − x3 + x4 + 2x5 = 0
⎩
x1 + 2x2 = 0.

Determine dimW and one of its basis.

Solution
Within the Proposition 1.67 we have dimW = n − r = 5 − 3 = 2, the associated
matrix of the system being
⎛ ⎞
1 1 1 −1 1
A = ⎝ 0 1 −1 1 2⎠
12 0 0 0

and
⎛ ⎞
 111 
 
π = ⎝ 0 1 2 ⎠ = −3 = 0.
 120 
1.4 Vector Subspaces 51

It results the system

⎧
⎨ x1 + x2 + x5 = −x3 + x4
x2 + 2x5 = x3 − x4
⎩
x1 + 2x2 = 0 ⇒ x1 = −2x2 .

We denote
x3 = t, x4 = u, t, u ∈ R;

we obtain 
−x2 + x5 = −t + u
x2 + 2x5 = t − u,

so we deduce that x5 = 0.
Finally, we get the solution of the system
⎧
⎪ x1 = −2t + 2u
⎪
⎪
⎪
⎨ x2 = t − u
x3 = t , (∀) t, u ∈ R.
⎪
⎪
⎪
⎪ x4 = u
⎩
x5 = 0

For
• x3 = 0 and x4 = 1 we obtain a1 = (2, −1, 0, 1, 0) ;
• x3 = 1 and x4 = 0 we obtain a2 = (−2, 1, 1, 0, 0)
and B1 = {a1 , a2 } is a basis of W .
We shall give the solution in Sage, too:

Proposition 1.69 (see [14], p. 8 and [6], p. 22). Let V be a vector space over the field
K and W1 W2 be two of its vector subspaces. Then W1 ∩ W2 is a vector subspace
of V .
52 1 Vector Spaces

Definition 1.70 (see [14], p. 8 and [6], p. 22). Let V be a vector space over the field
K and W1 W2 be two of its vector subspaces. The sum of two vector subspaces W1
and W2 is defined as:

W1 + W2 = {x 1 + x 2 |x 1 ∈ W1 and x 2 ∈ W2 } .

Proposition 1.71 (see [14], p. 8 and [6], p. 22). Let V be a vector space over the field
K and W1 W2 be two of its vector subspaces. Then W1 + W2 is a vector subspace
of V .
Theorem 1.72 (Grassmann- the dimension formula, see [14], p. 8 and [6],
p. 22). Let V be a finite n dimensional vector space over the field K, and W1 W2
be two of its vector subspaces. Then there is the relation

dim (W1 + W2 ) = dimW1 + dimW2 − dim (W1 ∩ W2 ) .

Definition 1.73 (see [14], p. 8 and [6], p. 23). Let V be a n dimensional finite vector
space over the field K, and W1 W2 be two of its vector subspaces. The sum of the
vector subspaces W1 and W2 is called the direct sum, denoted by W1 ⊕ W2 if any
vector from W1 + W2 can be uniquely written in the form:

x = x 1 + x 2 , x 1 ∈ W1 , x 2 ∈ W2 .

Proposition 1.74 (see [14], p. 9 and [6], p. 23). Let V be a n dimensional finite vector
space over the field K and W1 W2 be two of its vector subspaces. Then V = W1 ⊕ W2
if and only if the following conditions are satisfied:
 
(1) W1 ∩ W2 = 0
(2) dimW1 + dimW2 = dimV .

Proposition 1.75 (see [6], p. 21). Let V be a n dimensional finite vector space over
the field K. If W is a vector space of V , then the dimension of W is finite and
dimW ≤ dimV .
Proposition 1.76 (see [6], p. 22). Let V be a n dimensional finite vector space over
the field K. If W is a vector space of V , then dimW = dimV if and only if W = V .
Example 1.77 (see [13]). In the vector space R3 [X] we consider

W1 = {P ∈ R3 [X]|P (1) = 0} , W2 = {P ∈ R3 [X]|degree P = 0} .

(a) Show that W1 and W2 are two vector subspaces of R3 [X].

(b) Determine the basis and dimension of each of these subspaces.
(c) Prove that R3 [X] = W1 ⊕ W2 .

Solution
(a) Let be α, β ∈ K and P, Q ∈ W1 . It results that P(1) = 0 and Q(1) = 0.
1.4 Vector Subspaces 53

We shall have

(αP + βQ) (1) = αP (1) + βQ (1) = 0 ⇒ αP + βQ ∈ W1 ;

     
=0 =0

hence W1 is a vector subspace of R3 [X].

Let be α, β ∈ K and P, Q ∈ W2 . It results that degree P = 0 and degree Q = 0.
We shall have:

0 ≤ degree (αP + βQ) ≤ max {degree (αP) , degree (βQ)}

= max {α · degree (P) , β · degree (Q)} = 0 ⇒ degree (αP + βQ) = 0,

i.e. αP + βQ ∈ W2 , therefore W2 is a vector subspace of R3 [X].

(b) Let be P ∈ W1 . It results that P(1) = 0; we deduce
 
X − 1|P ⇒ P = (X − 1) Q, degree (Q) ≤ 2 ⇒ P = (X − 1) a0 + a1 X + a2 X 2

⇒ P = a0 (X − 1) + a1 X (X − 1) + a2 X 2 (X − 1) = a0 P1 + a1 P2 + a2 P3 ;
        
=P1 =P2 =P3

hence B1 = {P1 , P2 , P3 } is a system of generators for W1 .

Let

α1 P1 + α2 P2 + α3 P3 = 0 ⇒ α1 (X − 1) + α2 X (X − 1) + α3 X 2 (X − 1) = 0.

We obtain the system:

⎧
⎪
⎪ −α1 = 0 ⇒ α1 = 0
⎨
α1 − α2 = 0 ⇒ α2 = 0
⇒
⎪ α2 − α3 = 0 ⇒ α3 = 0
⎪
⎩
α3 = 0

B1 is a linearly independent system.

Therefore B1 is a basis of W1 and dim W1 = 3.
A solution in Sage will be given, too:
54 1 Vector Spaces

Let be P ∈ W2 . It results that degree P = 0 ⇒ P =ct= t, t ∈ R; we deduce

1 = t · P1
P = t · 
=P1

therefore B2 = {P1 } is a system of generators for W2 .

Let
α0 P1 = 0 ⇒ α0 = 0;

it results that B2 is a linearly independent system.

Hence B2 is a basis of W2 and dim W2 = 1.
(c) Within the Proposition 1.74

W1 ∩ W2 = {P0 }
R3 [X] = W1 ⊕ W2 ⇔
dimW1 + dimW2 = dimR3 [X],

where Q0 ∈ R3 [X] is the null polynomial.

Let P ∈ W1 ∩ W2 , namely:
• P ∈ W1 ; it results P (1) = 0, hence

X − 1|P ⇒ P = (X − 1) Q, degree Q ≤ 2

• P ∈ W2 ; it results degree P = 0.
We deduce
W1 ∩ W2 = {P0 } .

It is noticed that

dim W1 + dim W2 = 3 + 1 = 4 = dim R3 [X].

1.5 Problems

1. Decompose the vector v1 = i − 3j + 2k after the directions of the vectors a =

i + j, b = j + k, c = i + j + 3k.
1.5 Problems 55

Solution
We shall present the solution in Sage:

2. Show that the points A(3, −1, 1) B(4, 1, 4) and C(6, 0, 4) are the vertices of a
right triangle.

Solution
Using Sage we shall have:

We need the following Sage code to represent the right triangle:

56 1 Vector Spaces

3. Calculate the height of the parallelepiped built on the vectors v1 , v2 , v3 taking as

a basis the parallelogram built on vectors v1 and v2 knowing that

v1 = 2i + j − k, v2 = 3i + 2j + k, v3 = −j + 2k.

4. In the n arithmetic vector space R3 consider the vectors:

a1 = (−1, −1, 2) , a2 = (0, 1, −1) , a3 = (2, 1, 1) , a4 = (−1, −1, 7) .

(a) Prove that S = {a1 , a2 , a3 , a4 } is a system of generators for R3 .

(b) Extract a subsystem S  from S, that constitutes a basis of R3 .

5. The following vectors are given: OA = 12i−4j +3k, OB = 3i+12j −4k, OC =

2i + 3j − 4k.
(a) Find the lengths corresponding to the edges of ABC.
(b) Prove that AOC is a right triangle and OAB is an isosceles triangle.
(c) Calculate the scalar product AB · BC.
(d) Calculate the cross product AB × BC.
(e) Compute the area of ABC.
(f) Find the length of the height AA in ABC.

6. Investigate the linearly dependency of the following system of vectors: v1 =

(2, 1, 3, 1) , v2 = (1, 2, 0, 1) , v3 = (−1, 1, −3, 0) in R4 . Does this system form
a basis in R4 ?

Solution
The solution in Sage will be given:

7. Let be the vectors v1 = (2, 4, 1, 3) , v2 = (7, 4, −9, 5) , v3 = (4, 8, −3, 7) ,

v4 = (5, 5, −5, 5) , v5 = (8, 4, −14, 6). Which is the dimension of that subspace
generated by them?

Solution
The answer of this question will be find in Sage:
1.5 Problems 57

8. Establish the transformation formulas of the coordinates when passing from the
basis B to the basis B , if

B = {u1 = (1, 2, −1, 0) , u2 = (1, −1, 1, 1) , u3 = (−1, 2, 1, 1) , u4 = (−1, −1, 0, 1)} ,

B = {v1 = (2, 1, 0, 1) , v2 = (0, 1, 2, 2) , v3 = (−2, 1, 1, 2) , v4 = (1, 3, 1, 2)}

are some bases in R4 .

Solution
Using Sage we shall have:

9. In the arithmetic vector space C3 the following vectors are considered:

a1 = (1, 0, 1) , a2 = (i, 1, 0) , a3 = (i, 2, i + 1) , x = (1, −1, i)

b1 = (1, 1 + i, 1 + 2i) , b2 = (0, −1, i) , b3 = (i + 1, i − 1, i) .
58 1 Vector Spaces

(a) Prove that B1 = {a1 , a2 , a3 } is a basis of C3 .

(b) Determine the coordinates
  x relative to the basis B1 .
of
(c) Prove that B2 = b1 , b2 , b3 is a new basis of C3 and write the transition matrix
from the basis B1 to the basis B2 .
(d) Write the formulas of changing a vector coordinates when one passes from the
basis B1 to the basis B2 .  
10. In the vector space R4 let W = x = (x1 , x2 , x3 , x4 ) ∈ R4 , verifying the system

x1 − x2 + 2x3 + 3x4 = 0
x3 − 4x4 = 0.

Determine dimW and one of its bases.

References

1. V. Balan, Algebră liniară, geometrie analitică, ed (Fair Partners, Bucureşti, 1999)

2. P. Matei, Algebră liniară. Gometrie analitică şi diferenţială, ed (Agir, Bucureşti, 2002)
3. I. Iatan, Advances Lectures on Linear Algebra with Applications (Lambert Academic Publish-
ing, Saarbrücken, 2011)
4. C. Udrişte, Algebră liniară, geometrie analitică (Geometry Balkan Press, Bucureşti, 2005)
5. Gh Atanasiu, Gh Munteanu, M. Postolache, Algebr ă liniară, geometrie analitică şi diferenţială,
ecuaţii diferenţiale, ed (ALL, Bucureşti, 1998)
6. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie n- dimensională, ed (Radical,
Craiova, 1996)
7. P. Matei, Algebră liniară şi geometrie analitică. Culegere de probleme, ed. (MatrixRom,
Bucureşti, 2007)
8. V. Postelnicu, S. Coatu, Mică enciclopedie matematică, ed (Tehnică, Bucureşti, 1980)
9. S. Chiriţă, Probleme de matematici superioare, ed (Didactică şi Pedagogică, Bucureşti, 1989)
10. T. Didenco, Geometrie analitică şi diferenţială (Academia Militară, Bucureşti, 1977)
11. L. Constantinescu, C. Petrişor, Geometrie şi trigonometrie, ed (Didactică şi pedagogică,
Bucureşti, 1975)
12. W.K. Nicholson, Linear Algebra and with Applications (PWS Publishing Company, Boston,
1995)
13. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică (Universitatea
din Craiova, Note de curs şi aplica ţii, 1993)
14. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie analitică, ed (Universitaria, Craiova,
1993)
Chapter 2
Plane and Straight Line in E3

2.1 Equations of a Straight Line in E3

Definition 2.1 (see [2], p. 448). A vector is a translation of the three-dimensional

space; therefore it must be studied the basics of the three-dimensional Euclidean
geometry: the
 points, the
 straight lines and the planes.
Let R = O; i, j, k be a Cartesian reference. For M ∈E3 , the coordinates of the
point M are the coordinates of the position vector OM. If

OM = xi + y j + zk

then M (x, y, z) .

2.1.1 Straight Line Determined by a Point and a Nonzero Vector

A straight line from E3 can be determined by (see [6]):

(1) a point and a nonzero free vector;
(2) two distinct points;
(3) the intersection of the two planes.
 
Let be M0 ∈ E3 , M0 (x0 , y0 , z 0 ) and v ∈ V3 \ 0 , v = ai + b j + ck. We
intend to find the equation of the straight line determined by the point M0 and by the
nonzero vector v, denoted by d = (M0 , v) and represented in Fig. 2.1.

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 59

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_2,
© Springer International Publishing Switzerland 2014
60 2 Plane and Straight Line in E3

Fig. 2.1 The straight line

determined by a point and a
nonzero vector

We denote
r 0 = OM 0 = x0 i + y0 j + z 0 k.

Let be M ∈ E3 , M (x, y, z) and we denote

r = OM = xi + y j + zk.

The point M ∈ d → M0 M and v are collinear→

M0 M × v = 0. (2.1)

However

M0 M = r − r 0 . (2.2)

From (2.1) and (2.2) it results (see [6]) the vector equation of the straight line
d = (M0 , v) :

(r − r 0 ) × v = 0;

v is called the direction vector of the straight line.

If M0 M and v are collinear then (∃) t ∈ R unique, such that

M0 M = tv.

Taking account the relation (2.2) it results that r − r 0 = tv; we obtain (see [6])
the vector parametric equation of the straight line d:

r − r 0 = tv, t ∈ R. (2.3)

The Eq. (2.3) can be written as

xi + y j + zk = x0 i + y0 j + z 0 k + tai + tb j + tck;

we deduce the parametric equations (see [4], p. 49) of the straight line d:
2.1 Equations of a Straight Line in E3 61

 x = x0 + ta
y = y0 + tb , t ∈ R. (2.4)

z = z 0 + tc

If in the relation (2.4) we eliminate the parameter t we obtain (see [4], p. 49) the
Cartesian equations of the straight line d:

x − x0 y − y0 z − z0
= = . (2.5)
a b c
In the relation (2.5), there is the following convention: if one of the denominators
is equal to 0, then we shall also cancel that numerator.

2.1.2 Straight Line Determined by Two Distinct Points

Let be:
• M1 ∈E3 , M1 (x1 , y1 , z 1 ) , r 1 = OM 1 and
• M2 ∈E3 , M2 (x2 , y2 , z 2 ) , r 2 = OM 2 .
We want to determine the equation of straight line determined by the points M1
and M2 , denoted by d = (M1 , M2 ) and represented in Fig. 2.2.

Fig. 2.2 Straight line deter-

mined by two distinct points

Let be M ∈ d, M (x, y, z) and we denote r = OM. We consider that the straight

line is determined by M1 and M1 M 2 .
As

M1 M × M1 M 2 = 0

it results that vector equation (see [6]) of the straight line d is:

(r − r 1 ) × (r 2 − r 1 ) = 0.

Since (within the Proposition 1.1 from the Chap. 1), (∃) t ∈ R unique, such that

M1 M = t M1 M 2
62 2 Plane and Straight Line in E3

it results that
r − r 1 = t (r 2 − r 1 ) , (∀) t ∈ R,

namely the vector parametric equation (see [5], p. 166) is of form:

r = r 1 + t (r 2 − r 1 ) , (∀) t ∈ R; (2.6)

the parametric equations (see [6]) are:


 x = x1 + t (x2 − x1 )
y = y1 + t (y2 − y1 ) , (∀) t ∈ R (2.7)

z = z 1 + t (z 2 − z 1 )

and the Cartesian equations (see [4], p. 49) will be:

x − x1 y − y1 z − z1
= = . (2.8)
x2 − x1 y2 − y1 z2 − z1

2.2 Plane in E3
A plane can be determined (see [6]) in E3 as follows:
(1) a point and a nonnull vector normal to plane,
(2) a point and two noncollinear vectors,
(3) three non collinear points,
(4) a straight line and a point that does not belong to the straight line,
(5) two concurrent straight lines,
(6) two parallel straight lines.

2.2.1 A a Point and a Non Zero Vector Normal to the Plane

Let be M0 ∈ π, M0 (x0 , y0 , z 0 ) and the nonnull free vector n = ai + b j + ck

normal to the plane (see Fig. 2.3).

Fig. 2.3 Plane determined by

a point and a non zero vector
normal to plane
2.2 Plane in E3 63

Definition 2.2 (see [6]). The straight line d which passes through M0 and has the
direction of the vector n is called the normal to the plane through M0 ; the vector n
is the normal vector of the plane.
We propose to obtain the plane equation determined by the point M0 and by the
vector n, denoted with π = (M0 , n).
A point M (x, y, z) ∈ π → M0 M and n are orthogonal.
We denote

r = O M = xi + y j + zk.

As n is perpendicular on M0 M it results

M0 M · n = 0,

namely
(r − r 0 ) · n = 0;

from here we deduce the normal equation (see [6]) of the plane π :

r · n − r 0 · n = 0. (2.9)

Writing the relation (2.9) as

ax + by + cz − ax0 − by0 − cz 0 = 0

we obtain the Carthesian equation (see [6]) of the plane π :

a (x − x0 ) + b (y − y0 ) + c (z − z 0 ) = 0. (2.10)

If we denote

ax0 − by0 − cz 0 = −d

then from the equation (2.10) one deduces the general Carthesian equation (see [6])
of the plane π :
ax + by + cz + d = 0. (2.11)

2.2.2 Plane Determined by a Point and Two Noncollinear Vectors

Let u, v be two noncollinear vectors, namely u × v = 0, of the form

u = l1 i + m 1 j + n 1 k, v = l2 i + m 2 j + n 2 k

and let be M0 ∈ π, M0 (x0 , y0 , z 0 ) (see Fig. 2.4).

64 2 Plane and Straight Line in E3

Fig. 2.4 Plane determined by

a point and two non collinear
vectors

We want to find the equation of the plane determined by the point M0 and by the
free vectors u and v, denoted by π = (M0 , u, v).
Let
−−−∀
• M0 M 1 be a representative for the free vector u,
−−−∀
• M0 M 2 be a representative for the free vector v.
A point M (x, y, z) ∈ π → M0 M, M0 M 1 , M0 M 2 are coplanar. The coplanarity
of these vectors can be expressed as:

(a) using the Proposition 1.21 from the Chap. 1, (∃) t1 , t2 ∈ R uniquely determined,
such that

M 0 M = t1 M 0 M 1 + t2 M 0 M 2 ; (2.12)

(b) using the Proposition 1.34 from the Chap. 1, M0 M is perpendicular on u × v,

namely

M0 M · (u × v) = 0. (2.13)

Writing the relation (2.12) on the form:

r − r 0 = t1 u + t2 v

we deduce the vector parametric equation (see [6]) of the plane π :

r = r 0 + t1 u + t2 v, (∀) t1 , t2 ∈ R (2.14)

and then the parametric equations of the plane π :


 x = x 0 + t1 l 1 + t2 l 2
y = y0 + t1 m 1 + t2 m 2 , (∀) t1 , t2 ∈ R. (2.15)

z = z 0 + t1 n 1 + t2 n 2

From the relation (2.13) we obtain the vector equation (see [6]) of the plane π :

(r − r 0 ) · (u × v) = 0. (2.16)
2.2 Plane in E3 65

As

M0 M = (x − x0 ) i + (y − y0 ) j + (z − z 0 ) k (2.17)

we have  
 x − x0 y − y0 z − z 0 
 
M0 M · (u × v) =  l1 m1 n 1  ;
 l2 m2 n2 

so from the relation (2.13) we deduce the Carthesian equations (see [6]) of the
plane π :
 
 x − x0 y − y0 z − z 0 
 
 l1 m1 n 1  = 0. (2.18)

 l2 m2 n2 

2.2.3 Plane Determined by Three Noncollinear Points

Let M0 , M1 , M2 ∈ E3 be three noncollinear points, M0 (x0 , y0 , z 0 ), M1 (x1 , y1 , z 1 ),

M2 (x2 , y2 , z 2 ) . It results that M0 M 1 , M0 M 2 are noncollinear. We propose to obtain
the equation of the plane determined by these points, which is represented in Fig. 2.5
and it is denoted by π = (M0 , M1 , M2 ).

Fig. 2.5 Plane determined by

three noncollinear points


We note that π = (M0 , M1 , M2 ) coincides with π1 = M0 , M0 M 1 , M0 M 2 ,
namely we are in the case presented in the previous paragraph. We have

M0 M 1 = (x1 − x0 ) i + (y1 − y0 ) j + (z 1 − z 0 ) k,
M0 M 2 = (x2 − x0 ) i + (y2 − y0 ) j + (z 2 − z 0 ) k.

A point M (x, y, z) ∈ π → M0 M, M0 M 1 , M0 M 2 are coplanar, namely


M0 M · M0 M 1 × M0 M 2 = 0.

Using M0 M from (2.17) we obtain the following Cartesian equation (see [6]) of
the plane π :
66 2 Plane and Straight Line in E3
 
 x − x0 y − y0 z − z 0 
 
 x1 − x0 y1 − y0 z 1 − z 0  = 0. (2.19)
 
 x2 − x0 y2 − y0 z 2 − z 0 

2.2.4 Plane Determined by a Straight Line and a Point

that Doesn’t Belong to the Straight Line

Let d ⇒ E3 and a point M0 ∈ / d (see Fig. 2.6).

We want to obtain the equation of the plane determined by the straight line d and
by the point M0 , denoted by π = (M0 , d).

Fig. 2.6 Plane determined by

a straight line and a point that
doesn’t belong to straight line

Let be A ∈ d, hence we have d = (A, a). We note that π = (M0 , d) coincides

with π1 = M0 , a, M0 A . If r 0 is the position vector of the point M0 (denoted with
M0 (r 0 )), A (r A ) and M (x, y, z) ∈ π then the vector equation (see [6]) of the plane
π is:
(r − r 0 ) · [a × (r A − r 0 )] = 0 (2.20)

and the Carthesian equation (see [6]) of the plane π :

 
 x − x0 y − y0 z − z 0 
 
 a1 a2 a3  = 0, (2.21)

 x A − x0 y A − y0 z A − z 0 

where a = a1 i + a2 j + a3 k and A (x A , y A , z A ) .

2.2.5 Plane Determined by Two Concurrent Straight Lines

Let d1 ∃ d2 = {P}, see Fig. 2.7; the straight line

• d1 is the straight line which passes through P and has the direction vector a 1 ,
d1 = (P, a 1 ),
• d2 is the straight line which passes through P and has the direction vector a 2 ,
d2 = (P, a 2 ) .
2.2 Plane in E3 67

We want to find the equation of the plane determined by the straight lines d1
and d2 .

Fig. 2.7 Plane determined by

two concurrent straight lines

Noting that π = (d1 , d2 ) coincides with π = (P, a 1 , a 2 ), i.e. with the plane
which passes through P and has the direction vectors a 1 and a 2 . If M (x, y, z) ∈ π
we deduce that the vector equation (see [6]) of the plane is:

(r − r P ) · (a 1 × a 2 ) = 0; (2.22)

the Cartesian equation of the plane π will be:

 
 x − x P y − yP z − z P 
 
 l1 m1 n 1  = 0, (2.23)

 l2 m2 n2 

where: a 1 = l1 i + m 1 j + n 1 k, a 2 = l2 i + m 2 j + n 2 k, P (x P , y P , z P ) .
Example 2.3 Check if the following straight lines are concurrent:

y−7
d1 : x−1
2 = 1 = z−5
4
y+1
d2 : x−6
3 = −2 = z
1

and then write the equation of the plane which they determine.
Solution
We note that the direction vectors of the two straight lines are: a 1 = 2i + j + 4k
and respectively a 2 = 3i − 2 j + k. As
 
i j k
 
a 1 × a 2 =  2 1 4  = 9i + 10 j − 7k = 0
 3 −2 1 

it results that the vectors a 1 and a 2 are noncollinear, namely d1 ∃ d2 = ⇐.

Let P = d1 ∃ d2 . Since:
• P ∈ d1 we obtain x P − 1 = 2y P − 14;
• P ∈ d2 we obtain −2x P + 12 = 3y P + 3.
68 2 Plane and Straight Line in E3

Solving the system ⎫

xP − 1 = 2y P − 14
−2x P + 12 = 3y P + 3

we obtain: x P = −3, y P = 5, z P = −3.

The plane determined by the straight lines d1 and d2 will have the equation

π : 9 (x + 3) + 10 (y − 5) − 7 (z + 3) = 0,

namely
π : 9x + 10y − 7z − 44 = 0.

We shall present the solution of this problem in Sage, too:

2.2.6 Plane Determined by Two Parallel Straight Lines

Let d1 , d2 ∈E3 , d1 ||d2 ,see Fig. 2.8; the straight line:

• d1 is the straight line which passes through A1 and has the direction vector a,
d1 = (A1 , a) ,
2.2 Plane in E3 69

• d2 is the straight line which passes through A2 and has the direction vector a,
d2 = (A2 , a) .

Fig. 2.8 Plane determined by

two parallel straight lines

The plane determined by d1 and d2 is the plane determined by A1 and the two
non collinear vectors a and A1 A2 .
If M (x, y, z) ∈ π then the vector equation (see [6]) of the plane π is:
 
r − r A1 · a × A1 A2 = 0. (2.24)

The Cartesian equation of the plane π :

 
 x − xA y − y A1 z − z A1 
 1
 a1 a2 a3  = 0, (2.25)
 
 x A − x A yA − yA z A − z A 
2 1 2 1 2 1

 
where: a = a1 i + a2 j + a3 k, A1 x A1 , y A1 , z A1 , A2 x A2 , y A2 , z A2 .

2.2.7 The Straight Line Determined by the Intersection

of Two Planes
We consider π1 , π2 ∈ E3 (see Fig. 2.9) having the equations:
⎫
π1 : a1 x + b1 y + c1 z + d1 = 0
π2 : a2 x + b2 y + c2 z + d2 = 0.

Fig. 2.9 Straight line deter-

mined by the intersection of
two planes
70 2 Plane and Straight Line in E3

The intersection of the planes π1 and π2 is the set of solutions of the system of
equations determined by the equations of π1 and π2 .
We denote
⎬ ⎭
a1 b1 c1
A= .
a2 b2 c2

If rank (A) = 2 it results a compatible system which is undeterminedsimple

and the intersection of the two planes is a straight line. If rank (A) = rank A , A
being the extend matrix of the system it results an incompatible system, therefore
π1 ∃ π2 = ⇐, namely π1 ||π2 .
Let
• n 1 be the normal to π1 , n 1 = a1 i + b1 j + c1 k,
• n 2 be the normal to π2 , n 2 = a2 i + b2 j + c2 k.
We have
⎧
d ⇒ π1 ⇔ n 1 ≺ d
⇔ The straight line d has the direction vector n 1 × n 2 .
d ⇒ π2 ⇔ n 2 ≺ d

We denote
u = n 1 × n 2 , u = li + m j + nk.

We have
 
 i j k
 
n 1 × n 2 =  a1 b1 c1  = (b1 c2 − b2 c1 ) i + (a2 c1 − a1 c2 ) j + (a1 b2 − a2 b1 ) k.
 a2 b2 c2 

We deduce   
⎨  b1 c1 
⎨
⎨ l =  
⎨
⎨  b2 c2 
⎨
⎨
⎨
⎨  
  c1 a1 
m=  
⎨
⎨ c2 a2 
⎨
⎨
⎨
⎨  
⎨
⎨  a1 b1 
⎨
 n=  .
a2 b2 

The equation of the straight line is (see [6]):

x − x0 y − y0 z − z0
= = , (2.26)
l m n

(x0 , y0 , z 0 ) being a solution of the system.

2.2 Plane in E3 71

Example 2.4 Write the equation of a plane which:

(a) passes through the point M (−2, 3, 4) and is parallel with the vectors v 1 =
i − 2 j + k and v 2 = 3i + 2 j + 4k;
(b) passes through the point M (1, −1, 1) and is perpendicular on the planes π1 :
x − y + z − 1 = 0 and π2 : 2x + y + z + 1 = 0.

Solution
(a) The vector equation of the plane is

π : (r − r M ) · (v 1 × v 2 ) = 0.

Where as
 
i j k
 
• v 1 × v 2 =  1 −2 1  = −10i − j + 8k,
3 2 4
• r − r M = (x + 2) i + (y − 3) j + (z − 4) k
we obtain
π : −10 (x + 2) − (y − 3) + 8 (z − 4) = 0

or
π : −10x − y + 8z − 49 = 0.

This equation can be determined with Sage, too:

and it can also be plotted:

72 2 Plane and Straight Line in E3

(b) The normal of the required plane is n = n 1 × n 2 , where n 1 = i − j + k and

n 2 = 2i + j + k; therefore n = −2i + j + 3k.
The Cartesian equation of the plane will be

π : (−2) (x − 1) + (y + 1) + 3 (z − 1) = 0 → π : −2x + y + 3z = 0.

We can also find this equation in Sage:

A graphical solution in Sage is:

2.3 Plane Fascicle 73

2.3 Plane Fascicle

Let be d = π1 ∃ π2 ⎫
π1 : a1 x + b1 y + c1 z + d1 = 0
π2 : a2 x + b2 y + c2 z + d2 = 0.

Definition 2.5 (see [1], p. 62 and [2], p. 681). The set of the planes which contain the
straight line d is called a plane fascicle of axis d (see Fig. 2.10). The straight line d
is called the fascicle axis and π1 , π2 are called the base planes of the fascicle.

Fig. 2.10 Plane fascicle

An arbitrary plane of the fascile has the equation of the form:

π : a1 x + b1 y + c1 z + d1 + λ (a2 x + b2 y + c2 z + d2 ) = 0, λ ∈ R√ .
74 2 Plane and Straight Line in E3

Example 2.6 Determine a plane which passes through the intersection of the planes
π1 : x + y + 5z = 0 and π2 : x − z + 4 = 0 and which forms with the plane
π : x − 4y − 8z + 12 = 0 an angle ϕ = π2 .
Solution
Let be d = d1 ∃ d2 . A plane of the plane fascicle of axis d has the equation

π ≡ : x + 5y + z + λ (x − z + 4) = 0,

namely
π ≡ : (1 + λ) x + 5y + (1 − λ) z + 4λ = 0.

It results that
n ≡ = (1 + λ) i + 5 j + (1 − λ) k.

As n = i − 4 j − 8k we shall deduce

 π n · n≡ −27 + 9λ
cos ≤ π, π ≡ = cos = ⎩ ⎩= ∞ → λ = 3.
2 ⎩ ≡⎩ 27 + 2λ2
∗n∗ · ⎩n ⎩ 9

The equation of the required plane is

π ≡ : (1 + 3) x + 5y + (1 + 3) z + 4 · 3 = 0

or
π ≡ : 4x + 5y − 2z + 12 = 0.

The solution in Sage of this problem is:

2.4 Distances in E3

2.4.1 Distance from a Point to a Straight Line

Let be d = (A, a) with A (x A , y A , z A ) , a = a1 i + a2 j + a3 k and M ∈ E3 .

2.4 Distances in E3 75

−−∀
Let A A≡ be a representative for a. The equation of the straight line is:
x − xA y − yA z − zA
= = .
a1 a2 a3

We build the parallelogram A A≡ P M (see Fig. 2.11).

Fig. 2.11 The distance from

a point to a straight line

We know that ⎩ ⎩
⎩ ⎩
A A A≡ P M = ⎩A A≡ × M A⎩ . (2.27)

However ⎩ ⎩
⎩ ⎩
A A A≡ P M = ⎩ A A≡ ⎩ · ρ (M, d) . (2.28)

From (2.27) and (2.28) it results that the distance formula from a point to a straight
line is [6] ⎩ ⎩
⎩ ≡ ⎩ ⎩ ⎩
⎩A A × M A⎩ ⎩a × M A⎩
ρ (M, d) = ⎩ ⎩ = . (2.29)
⎩ ≡⎩ ∗a∗
⎩A A ⎩

2.4.2 Distance from a Point to a Plane

We consider the plane π : ax + by + cz + d = 0 and the point M0 (x0 , y0 , z 0 ),

M0 ∈
/ π . Let M1 be the projection of M0 on the plane π , M1 (x1 , y1 , z 1 ), see Fig. 2.12.

Fig. 2.12 The distance from

a point to a plane

The distance from the point M0 to the plane π is

76 2 Plane and Straight Line in E3
⎩ ⎩
ρ (M0 , π ) = ⎩ M0 M1 ⎩ .

Let d = (M0 , n) be the normal line to the plane which passes through M0 ,
n = ai + b j + ck . The equation of this straight line is

x − x0 y − y0 z − z0
d: = = =t
a b c
or 
 x = x0 + ta
d : y = y0 + tb, t ∈ R.

z = z 0 + tc

As M1 ∈ d we deduce
x1 − x0 y1 − y0 z1 − z0
d: = = =t ⇔
a b c

 x1 = x0 + ta
y1 = y0 + tb, t ∈ R. (2.30)

z 1 = z 0 + tc

Since M1 ∈ π ⇔ ax1 + by1 + cz 1 + d = 0 ⇔

ax1 + by1 + cz 1 = −d (2.31)

Multiplying the first equation of (2.31) with a, the second with b and the third
with c we have: 
 ax1 = ax0 + ta 2
by = by0 + tb2 , t ∈ R. (2.32)
 1
cz 1 = cz 0 + tc2

Adding the three equations of (2.32) it results

 
ax1 + by1 + cz 1 = ax0 + by0 + cz 0 + t a 2 + b2 + c2 . (2.33)

Substituting (2.31) into (2.33) we deduce

 
ax0 + by0 + cz 0 + d = −t a 2 + b2 + c2 ,

namely
ax0 + by0 + cz 0 + d
t =− . (2.34)
a 2 + b2 + c2

We have
2.4 Distances in E3 77

M0 M1 = (x1 − x0 ) i + (y1 − y0 ) j + (z 1 − z 0 ) k;

therefore
⎩ ⎩   
⎩ M0 M1 ⎩ = (x1 − x0 )2 + (y1 − y0 )2 + (z 1 − z 0 )2 (2.30)
= t 2 a 2 + b2 + c2 ⇔

⎩ ⎩ 
⎩ M0 M1 ⎩ = |t| a 2 + b2 + c2 . (2.35)

Substituting (2.34) into (2.35) we can deduce [6] the distance formula from a
point to a plane:
|ax0 + by0 + cz 0 + d|
ρ (M0 , π ) = ∞ . (2.36)
a 2 + b2 + c2

Example 2.7 (see [6]). One gives:

• the plane π : x + y − z + 2 = 0,
• the straight line
⎫
x −y−1=0
d:
x + 2y + z − 4 = 0

• the point A = (1, 1, 2).

(a) Compute the distance from the point A to the plane π.
(b) Find the distance from the point A to the straight line d.

Solution
(a) Using the formul (2.36) we achieve:
∞
2 2 3
ρ (A, π ) = ∞ = .
3 3

(b) The distance from the point A to the straight line d (Fig. 2.13) is computed using
the formula
⎩ ⎩
⎩ ⎩ ⎩ ⎩
⎩ M A≡ × AM ⎩ ⎩a × AM ⎩
ρ (A, d) = ⎩ ⎩ = .
⎩ ⎩ ∗a∗
⎩ M A≡ ⎩
78 2 Plane and Straight Line in E3

Fig. 2.13 The distance from

a point to a straight line

The direction vector of the straight line d is

 
i j k 

a =  1 −1 1  = −i − j + 3k.
1 2 1
∞
We obtain: ∗a∗ = 11.
From M ∈ d we have
⎫ ⎫
x M − yM − 1 = 0 x M − yM − 1 = 0
→
x M + 2y M + z M − 4 = 0 x M + 2y M = 4 − z M .

Denoting z M = u ∈ R we deduce

1
3y M = 3 − u → y M = 1 − u
3
and
1
x M = y M + 1 = 2 − u.
3

 xM = 2
We can suppose that u = 0; we obtain: y M = 1 =⇔ M (2, 1, 0) . We have:

zM = 0
AM = i − 2k and
 
 i j k 
 
a × AM =  −1 −1 3  = 2i + j + k;
 1 0 −2 

⎩ ⎩ ∞
therefore ⎩a × AM ⎩ = 6. We shall obtain
∞
6
ρ (A, d) = ∞ = 0.739.
11
2.4 Distances in E3 79

Solving this problem with Sage, we shall have:

2.4.3 Distance Between Two Straight Lines

Let d1 , d2 be two noncoplanar straight lines (see Fig. 2.14).

80 2 Plane and Straight Line in E3

Fig. 2.14 the distance

between two straight lines

The distance between the straight lines d1 and d2 is

ρ (d1 , d2 ) = ρ (A1 , A2 ) = ρ (A2 , π ) ,

where:
• π is the plane which passes through d1 and it is parallel with d2 ,
• ρ (A2 , π ) is the height which corresponds to the vertex A2 of the oblique paral-
lelepiped built on the vectors a, b, A1 A2 .
Therefore, the distance formula between two straight lines is [6]:
  
V parallelepi ped a · b × A1 A2 
ρ (d1 , d2 ) = = ⎩ ⎩ . (2.37)
Abase ⎩a × b⎩

Definition 2.8 (see [6]). The support straight line corresponding to the segment which
represents the distance between two straight lines is called the common perpendic-
ular of the two straight lines.
Let δ be the straight line which represents the common perpendicular. To deter-
mine the equations of the straight line δ:
1. find the direction of the the common perpendicular n = a × b, a and b being the
direction vectors of the two straight lines;
2. write the equation of a plane π1 , which passes through d1 and contains n;
3. write the equation of a plane π2 , which passes through d2 and contains n;
4. δ = π1 ∃ π2 is the common perpendicular searched by us.
If
• d1 = (A1 , a) , A1 (x1 , y1 , z 1 ) ,
• d2 = (A2 , a) , A2 (x2 , y2 , z 2 ) ,
2.4 Distances in E3 81

• n = n 1 i + n 2 j + n 3 k,
• a = a1 i + a2 j + a3 k,
• b = b1 i + b2 j + b3 k,
then the equations of the common perpendicular δ are [6]:
 
⎨  x − x1 y − y1 z − z 1 
⎨
⎨  
⎨
⎨  a1 a2 a3  = 0
⎨
 
n1 n2 n 3 
δ :  (2.38)
⎨
⎨ x − x2 y − y2 z − z 2 
⎨ 
⎨
⎨
⎨ b1 b2 b3  = 0.
 
n1 n2 n3 

Example 2.9 Let be the straight lines

 
 x = 1 + 2r x = 1+s
d1 : y = 3 + r and d2 : y = −2 − 4s, r, s ∈ R.
 
z = −2 + r z = 9 + 2s
Find:
(a) the angle between these straight lines;
(b) the equation of the common perpendicular;
(c) the distance between the two straight lines.

Solution
(a) We have
 a·b
cos ≤ (d1 , d2 ) = cos ≤ a, b = ⎩ ⎩,
∗a∗ ⎩b⎩

where
• d1 = (A,
 a) , A (1, 3, −2) , a = 2i + j + k,
• d2 = B, b , B (1, −2, 9) , b = i − 4 j + 2k.
We obtain π
cos ≤ (d1 , d2 ) = 0 ⇔≤ (d1 , d2 ) = .
2
(b) The direction of the common perpendicular is

n = a × b,

namely  
i j k
 
n =  2 1 1  = 6i − 3 j − 9k.
 1 −4 2 

As
• the equation of the plane which passes through d1 and contains n is
82 2 Plane and Straight Line in E3

 
x − 1 y − 3 z + 2
 
π1 :  2 1 1  = 0 → π1 : x − 4y + 2z + 15 = 0;
 6 −3 −9 

• the equation of the plane which passes through d2 and contains n is

 
x − 1 y + 2 z − 9
 
π2 :  1 −4 2  = 0 → π2 : 2x + y + z − 9 = 0.
 6 −3 −9 

The equations of the common perpendicular will be:

⎫
x − 4y + 2z + 15 = 0
2x + y + z − 9 = 0.

(c) Using (2.37), we have

  
a · b × AB 
ρ (d1 , d2 ) = ⎩ ⎩ ,
⎩a × b⎩

where
• AB = (x B − x A ) i + (y B −
 y A ) j + (z B − z A ) k = −5 j + 11k,
2 1 1 
  
• a · b × AB =  1 −4 2  = −84,
 0 −5 11 
⎩ ⎩  ∞
• ⎩a × b⎩ = 62 + (−3)2 + (−9)2 = 126;
therefore
84 ∞
ρ (d1 , d2 ) = ∞ = 2 14.
126

We shall solve this problem using Sage:

2.5 Angles in E3 83

2.5 Angles in E3

2.5.1 Angle Between Two Straight Lines

Definition 2.11 (see [3], p. 112). Let d1 , d2 be two straight lines, which have the
direction vectors a = a1 i + a2 j + a3 k and respectively b = b1 i + b2 j + b3 k. The
angle between the straight lines d1 and d2 is the angle between the vectors a and
b (see Fig. 2.15).

Fig. 2.15 The angle between

two straight lines

Hence

a·b a1 b1 + a2 b2 + a3 b3
cos ϕ = ⎩ ⎩=  , ϕ ∈ [0, π ] . (2.39)
∗a∗ ⎩b⎩ a12 + a22 + a32 b12 + b22 + b32

Remark 2.12 (see [3], p. 112).

π
(1) ϕ = 2 ⇔the straight lines are perpendicular →

a · b = 0 → a1 b1 + a2 b2 + a3 b3 = 0.

(2) ϕ = 0 ⇔the straight lines are parallel→ a × b = 0 →

 
 i j k 
 
 a1 a2 a3  = 0 → (a2 b3 − a3 b2 ) i + (a3 b1 − a1 b2 ) j + (a1 b2 − a2 b1 ) k = 0 →
 
 b1 b2 b3 

 a2 b3 − a3 b2 = 0 a1 a2 a3
a3 b1 − a1 b2 = 0 → = = . (2.40)
 b b b3
a1 b2 − a2 b1 = 0 1 2

Therefore d1 ||d2 the relation (2.40) occurs.

2.5.2 Angle Between Two Planes

Let be
• π1 : a1 x + b1 y + c1 z + d1 = 0 and the normal vector n 1 = a1 i + b1 j + c1 k,
84 2 Plane and Straight Line in E3

• π2 : a2 x + b2 y + c2 z + d2 = 0 and the normal vector n 2 = a2 i + b2 j + c2 k.

Fig. 2.16 The angle between

two planes

Definition 2.13 (see [3], p. 113). The angle ϕ between the planes π1 and π2 is the
angle between the vectors n 1 and n 2 (see Fig. 2.16).
Hence

n1 · n2 a1 a2 + b1 b2 + c1 c2
cos ϕ = =  , ϕ ∈ [0, π ] . (2.41)
∗n 1 ∗ ∗n 2 ∗ a12 + b12 + c12 a22 + b22 + c22

Remark 2.14 (see [3], p. 113).

(1) π1 ||π2 ⊂⇔ n 1 and n 2 are collinear ⊂⇔ n 1 = tn 2 , t ∈ R; therefore

π1 ||π2 ⊂⇔ a 1 = ta 2 , b1 = tb2 , c1 = tc2 .

(2) π1 ≺π2 ⊂⇔ n 1 · n 2 = 0 ⊂⇔ a1 a2 + b1 b2 + c1 c2 = 0.

2.5.3 Angle Between a Straight Line and a Plane

Let d be the straight line with the direction vector a = a1 i + a2 j + a3 k and the plane
π having the normal vector n = n 1 i + n 2 j + n 3 k.
Definition 2.15 (see [3], p. 113). The angle ϕ between the straight line d and the
plane π is the angle between the straight line d and the projection of this straight
line on the plane π (see Fig. 2.17).
2.5 Angles in E3 85

Fig. 2.17 The angle between

a straight line and a plane

The angle between the straight line d and the plane π is related to the angle θ , the
angle of the vectors a and n, through the relations: θ = π2 ± ϕ as the vectors are on
the same side of the or in different parts. Hence:
π   π
cos θ = cos ± ϕ = ± sin ϕ, θ ∈ [0, π ] ⇔ ϕ ∈ 0, .
2 2
As
n·a
cos θ = , θ ∈ [0, π ]
∗n∗ ∗a∗

it results that
|a1 n 1 + a2 n 2 + a3 n 3 |  π
sin ϕ =   , ϕ ∈ 0, . (2.42)
n 21 + n 22 + n 23 a12 + a22 + a32 2

Remark 2.16 (see [3], p. 113).

(1) d||π → n · a = 0 → a1 n 1 + a2 n 2 + a3 n 3 = 0.
π (2.40) n 1 n2 n3
(2) d≺π ⇔ ϕ = 2 ⇔ θ = 0 ⇔ n||a =⇔ a1 = a2 = a3 .

Example 2.17 Are given

• the planes π1 : 2x − y + 7 = 0 and π2 : x − 5y + 3z = 0,
y+3
2 = −1 = 5 .
• the straight line d : x−1 z

Compute:
(a) the angle of these planes;
(b) the angle between the straight line and the plane π1 .

Solution
(a) We have n 1 = (2, −1, 0) , n 2 = (1, −5, 3) ; hence
∞
(2.41) n1 · n2 7
cos ≤ (π1 , π2 ) = cos ϕ = = = 0.529.
∗n 1 ∗ ∗n 2 ∗ 5
86 2 Plane and Straight Line in E3

(b) As the direction vector of the straight line d is a = (2, −1, 5) we obtain

(2.42) |a1 n 1 + a2 n 2 + a3 n 3 | 1
sin ≤ (d, π1 ) = sin ϕ =   =∞ .
n 21 + n 2 + n 3 a1 + a2 + a3
2 2 2 2 2 6

We need the following code in Sage to solve this problem:

2.6 Problems

1. Check if the points M1 (3, 0, 1), M2 (0, 2, 4) , M3 1, 43 , 3 are collineare.

Solution
Using Sage we shall have:
2.6 Problems 87

2. Write the equation of the plane determined by the points: M1 (3, 1, 0),
M2 (0, 7, 2), M3 (4, 1, 5) .

Solution
We shall give a solution in Sage:

3. Write the equation of a plane perpendicular on the planes π1 : x − y + z − 1 = 0

and π2 : 2x + y + z + 1 = 0 and which passes through the point M0 (1, −1, 1).
4. Write the equation of a plane which passes through the points M1 (3, 1, 2),
M2 (4, 6, 5) and is parallel with the vector v = i + 2 j + 3k.
Solution
Solving this problem in Sage we have:
88 2 Plane and Straight Line in E3

5. Let d be the straight line determined by the point P0 (2, 0, −1) and the direction
vector v = i − j. Compute the distance from the point P (1, 3, −2) to the straight
line d.
6. Write the ecuation of the perpendicular from the point M (−2, 0, 3) on the plane
π : 7x − 5y + z − 11 = 0.
Hint. The perpendicular from a point to a plane is the straight line which passes
through that point and has the normal vector of the plane as a direction vector.
Solution
Solving this problem with Sage, we shall have:
2.6 Problems 89

7. It gives a tetrahedron ABC D defined by the points A (3, 0, 0), B (2, 4, 0),
C (−3, −1, 0), D (0, 0, 5). Write the equations of its faces, the edge equations
and the equations corresponding to the heights of the tetrahedron ABC D.
8. Write the equation of the plane which passes through Oz and is perpendicular
on the plane π : 8x + y + 2z − 1 = 0.
Hint. The equation of the plane which passes through Oz is: ax + by = 0.
Solution
We shall present the solution in Sage:

9. Determine the projection equation of the straight line having the equations

⎫
x − 3z + 1 = 0
d:
y − 2z − 3 = 0

on the plane π : x − y + 2z − 1 = 0.
10. Let be the straight lines:
⎫
x − 2y + z + 1 = 0
d1 :
y−z =0
90 2 Plane and Straight Line in E3

and
x −1 y+3 z
d2 : = = .
2 1 8
(a) Find the equation of the common perpendicular.
(b) Compute the distance between the two straight lines.

References

1. S. Chiriţă (ed.), Probleme de matematici superioare (Didactică şi Pedagogică, Bucureşti, 1989)
2. V. Postelnicu, S. Coatu (eds.), Mică enciclopedie matematică (Tehnică, Bucureşti, 1980)
3. C. Udrişte, Algebră liniară, geometrie analitică (Geometry Balkan Press, Bucureşti, 2005)
4. I. Vladimirescu, M. Popescu (eds.), Algebră liniară şi geometrie n- dimensională (Radical,
Craiova, 1996)
5. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie analitică, ed. Universitaria,
Craiova,1993
6. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică, Note de curs şi
aplica ţii, Universitatea din Craiova, 1993
Chapter 3
Linear Transformations

3.1 Linear Transformations

The linear transformations should be studied [1] because they are compatible with
the operations defined in a vector space and allow the transfer of some algebraic
situations or problems from a space to another. Matrix operations clearly reflect
their similarity with the operations with linear transformations; hence the matrices
can be used for numerical representation of the linear transformations. The matrix
representation of linear transformations is [1] analogous to the representation of the
vectors through n coordinates, relative to a basis.
Definition 3.1 (see [2], p. 41). Let U and V two vector spaces over the field K . The
mapping T : U ∈ V is called a Linear transformation (or a linear mapping) if
the following conditions are satisfied:
(1) T (x + y) = T (x) + T (y) , (→) x, y ∈ U, namely T is additive;
(2) T (πx) = πT (x) , (→) x ∈ U, namely T is homogeneous.
The two properties of the linear maps can be formulated in a single.
Proposition 3.2 (see [2], p. 41). The mapping T : U ∈ V is linear if and only if

T (πx + λ y) = πT (x) + λT (y) , (→) π, λ ∈ K , (→) x, y ∈ U. (3.1)

Examples of linear transformations

1. T : R ∈ R, T (x) = ϕx, ϕ ∈ R;
2.  : V ∈ V,  (x) = x, (→) x ∈ V identity mapping;
3. ∅ : U ∈ V, ∅ (x U ) = 0V zero mapping;
 t
4. T : Rn ∈ Rm , T (x) = Ax, (→) x = x (1) , . . . , x (n) ∈ Rn , A ∈Mm,n (R) is
given; n
5. T : Mn (K ) ∈ K , T (A) = trace(A) = aii ;
i=1
6. T : Mm,n (K ) ∈Mn,m (K  ) , T (A) = A

t;

7. T : V3 ∈ R3 , T (v) = v (1) , v (2) , v (3) , v = v (1) i + v (2) j + v (3) k.

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 91

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_3,
© Springer International Publishing Switzerland 2014
92 3 Linear Transformations

We shall denote by L(U, V ) the set of all linear transformations defined on U

with their values in V .
Definition 3.3 (see [2], p. 42). Any linear mapping T : V ∈ V is called an endo-
morphism of the vector space V . We shall denote by End (V ) the set of the all
endomorphisms corresponding to the vector space V .
Definition 3.4 (see [2], p. 42). A linear mapping T : V ∈ K is called a linear form.
Definition 3.5 (see [2], p. 42). Let be T ∈ L(U, V ) .
(a) The set  
Ker T = x ∈ U |T (x) = 0V (3.2)

is called the kernel of the linear transformation T .

(b) The set
Im T = T (U ) = {v ∈ V | (∀) u ∈ U, T (u) = v} (3.3)

is called the image of the linear transformation T .

Definition 3.6 (see [2], p. 42). A linear transformation T : U ∈ V is called:
 
(1) injective if and only if Ker T = 0V ;
(2) surjective if and only if Im T = V .
Definition 3.7 (see [3], p. 57). Any bijective endomorphism of the vector space V is
called the automorphism of the vector space.
We shall denote by Aut(V ) the set of all automorphisms of the vector space V .
Proposition 3.8 (see [4], p. 31). Let T : U ∈ V be a bijective linear transformation.
If B1 = {e1 , . . . , en } is a basis in U , then B2 = {T (e1 ) , . . . , T (en )} is a basis in V .

Theorem 3.9 (see [5], p. 45). Let T : Un ∈ Vn be a linear mapping between two
vector spaces with the same dimension. Then the following statements are equivalent:
(i) T is injective;
(ii) T is surjective;
(iii) T is bijective.
Definition 3.10 (see [2], p. 42). A linear mapping T : U ∈ V which is injective
and surjective is called isomorphism between the vector spaces U and V .
Theorem 3.11 (see [3], p. 29). Let U and V be two finite dimensional vector spaces
over the field K . Then they are isomorphic if and only if dim U = dim V .
Definition 3.12 (see [4], p. 26). Let S, T ∈ L(U, V ). The sum of the two linear
transformations is the linear transformation R ∈ L (U, V ) ,

R (x) = (S + T ) (x) = S (x) + T (x) , (→) x ∈ U. (3.4)

Definition 3.13 (see [4], p. 26). Let T ∈ L(U, V ). The scalar multiples of the linear
transformation T, denoted πT ∈ L (U, V ) is defined as:
3.1 Linear Transformations 93

(πT ) (x) = πT (x) , (→) π ∈ U, (→) x ∈ U. (3.5)

Definition 3.14 (see [5], p. 41). The composition of the two linear transformations
is called the product (or the multiplication) and is defined as in the case of the
functions.
Remark 3.15 (see [5], p. 41). The composition isn’t commutative but it is associative.
Proposition 3.16 (see [4], p. 27). If U, V, W are some vector spaces over K and
T : U ∈ V , S : V ∈ W are two linear transformations, then the mapping
S ⇒ T : U ∈ V,
(S ⇒ T ) (x) = S (T (x)) , (→) x ∈ U (3.6)

is linear.
Example 3.17 (see [6], p. 36). Let be the linear transformations

f : R2 ∈ R2 , f (x) = f (x1 , x2 ) = (2x1 − x2 , 3x2 )

g : R2 ∈ R2 , g (x) = g (x1 , x2 ) = (x2 − x1 , 4x1 ) .

(a) Show that they are linear maps.

(b) Determine f ⇒ g, g ⇒ f and check if they are linear maps.

Solution
(a) f is a linear mapping∃

f (πx + λ y) = π f (x) + λ f (y) , (→) π, λ ∈ R, (→) x, y ∈ R2 .

Let π, λ ∈ R and (→) x, y ∈ R2 , x = (x1 , x2 ), y = (y1 , y2 ). We have:

f (πx + λ y) = f (πx1 + λy1 , πx2 + λy2 )

= (2 (πx1 + λy1 ) − (πx2 + λy2 ) , 3 (πx2 + λy2 ))
= (π (2x1 − x2 ) + λ (2y1 − y2 ) , π · 3x2 + λ · 3y2 )
= (π (2x1 − x2 ) , π · 3x2 ) + (λ (2y1 − y2 ) , λ · 3y2 )
= π (2x1 − x2 , 3x2 ) + λ (2y1 − y2 , 3y2 ) = π f (x) + λ f (y) .

Therefore, f is a linear mapping. Similarly, one shows that g is also a linear

mapping.

(b) We obtain

( f ⇒ g) (x) = f (g (x)) = f (x2 − x1 , 4x1 )

= (2 (x2 − x1 ) − 4x1 , 12x1 ) = (2x2 − 6x1 , 12x1 ) ;
94 3 Linear Transformations

(g ⇒ f ) (x) = g ( f (x)) = g (2x1 − x2 , 3x2 ) = (3x2 − (2x1 − x2 ) , 4 (2x1 − x2 ))

= (4x2 − 2x1 , 8x1 − 4x2 ) .

Using the Proposition 3.16 it results that the transformations f ⇒ g and g ⇒ f are
linear.
We can check this result in Sage, too:

Definition 3.18 (see [5], p. 42). The natural powers of an endomorphism T : V ∈

V are defined inductive as:

T0 = 
T = T · T n−1 , n = 1, 2, . . .
n

Definition 3.19 (see [5], p. 42). Let T ∈ L(U, V ) be a bijective linear mapping
(hence it is invertible). Its inverse, T −1 ∈ L(U, V ) is a linear mapping.
Example 3.20 (see [6], p. 36). Let be the linear mapping

f : R3 ∈ R3 , f (x) = f (x1 , x2 , x3 ) = (x1 − 2x2 + x3 , 2x1 + x2 − x3 , x2 − 3x3 ) .

(a) Show that f is invertible and determine its inverse f −1 : R3 ∈ R3 .

(b) Check if the mapping f −1 is linear too.
3.1 Linear Transformations 95

Solution
 
(a) We know (see the Definition 3.6) is injective if and only if Ker T = 0R3 . Let
be x ∈ R3 .
f (x) = 0R3 ∃ (x1 − 2x2 + x3 , 2x1 + x2 − x3 , x2 − 3x3 ) = (0, 0, 0)

 x1 − 2x2 + x3 = 0
∃ 2x1 + x2 − x3 = 0

x2 − 3x3 = 0.

Since the determinant of the matrix A, associated to the previous system is:
⎫ ⎫
⎫ 1 −2 1 ⎫
⎫ ⎫
ρ = ⎫⎫ 2 1 −1 ⎫⎫ ⇐= 0 ⇔ rank (A) = 3 ⇔
⎫ 0 1 −3 ⎫

the homogeneous system is compatible determined, having the unique solution x =

(0, 0, 0).  
Therefore, it results that Ker T = 0R3 , i.e. that f is injective. If f is injective
then according to the Theorem 3.9 it results that f is bijective, namely (see the
Definition 3.19) f is invertible.
Let be y ∈ R3 , y = (y1 , y2 , y3 ). Then

f (x) = y ∃ (x1 − 2x2 + x3 , 2x1 + x2 − x3 , x2 − 3x3 ) = (y1 , y2 , y3 )


 x1 − 2x2 + x3 = y1
∃ 2x1 + x2 − x3 = y2

x2 − 3x3 = y3 .

The system solution is:


⎬ x1 = (2y1 + 5y2 − y3 )
1
⎬
 12
x2 = 1
(−6y1 + 3y2 − 3y3 )
⎬
⎬
12

x3 = 1
12 (−2y1 + y2 − 5y3 ) .

We obtain f −1 : R3 ∈ R3 ,
⎭ ⎧
1 1 1
f −1 (x) = (2x1 + 5x2 − x3 ) , (−6x1 + 3x2 − 3x3 ) , (−2x1 + x2 − 5x3 ) .
12 12 12

(b) Using the Definition 3.19 it results that the mapping f −1 is linear.

Solving this problem in Sage we shall have:

96 3 Linear Transformations

Theorem 3.21 (see [5], p. 41). The set L(U, V ) is a vector space over the field K
relative to the addition of the linear transformations and the scalar multiplication of
a linear transformation.
Theorem 3.22 (see [5], p. 43). Let T ∈ L(U, V ). We have the following properties:
(i) Ker T is a vector subspace of U ;
(ii) Im T is a vector subspace of V .
Proposition 3.23 (see [5], p. 40 and [3], p. 27). If T ∈ L (U, V ). Then:
 
(1) T 0U = 0V , i.e. a linear transformation maps the null vector, in the null vector;
(2) T (−x) = −T (x) , (→) x ∈ U ;
(3) If W is a vector subspace of U , then T (W ) is a vector subspace of V ;
(4) If the vectors x 1 , . . . , x n ∈ U are linearly dependent then the vectors T (x 1 ) ,
. . . , T (x n ) ∈ V are linearly dependent, too;
(5) Being given the vectors x 1 , . . . , x n ∈ U , if the vectors T (x 1 ) , . . . , T (x n ) ∈ V
are linearly independent then the vectors x 1 , . . . , x n are linearly independent,
too.
Proposition 3.24 (see [3], p. 28). Let T ∈ L (U, V ) be an injective linear mapping.
If the system of vectors {x 1 , . . . , x n } ≺ U is linearly independent, the system of
vectors T (x 1 ) , . . . , T (x n ) is linearly independent.
Proposition 3.25 (see [3], p. 28). Let U, V be two vector spaces over the field K and
T ∈ L (U, V ). The following statements hold:
(a) If {x 1 , . . . , x n } ≺ U is a system of generators for U , then T (x 1 ) , . . . , T (x n )
is a system of generators for Im T .
(b) If T is surjective and {x 1 , . . . , x n } ≺ U is a system of generators for U , then
T (x 1 ) , . . . , T (x n ) is a system of generators for V .
Theorem 3.26 (see [3], p. 29 and [2], p. 44). Let U, V be two vector spaces over the
field K . If dim U < √ and T ∈ L(U, V ), then

dim KerT + dim ImT = dim U.

Definition 3.27 (see [2], p. 44). Let T ∈ L(U, V ).

3.1 Linear Transformations 97

(a) The T ’s kernel dimension is called defect of T .

(b) The T ’s image dimension is called the rank of T .
Let U, V be two finite dimensional vector spaces over the field K , dim U =  n,
dim V = m and T ∈ L(U, V ). Let B1 = {e1 , . . . , en } and B2 = f 1 , . . . , f m be
two bases in U and respectively in V . We consider the expressions of the vectors
T (e1 ) , . . . , T (en ) ∈ V in the basis B2 :

(1) (2) (m)
⎬
⎬ T (e1 ) = π1 f 1 + π1 f 2 + · · · + π1 f m
⎬
⎬
 (1) (2) (m)
T (e2 ) = π2 f 1 + π1 f 2 + · · · + π2 f m
⎬ ..
⎬
⎬ .
⎬
 (1) (2) (m)
T (en ) = πn f 1 + πn f 2 + · · · + πn f m .

We shall denote by
⎩ (1) (1) 
π1 . . . πn
⎨(B1 ,B2 )  .. .. 
T = . .  (3.7)
(m) (m)
π1 ... πn

the m × m matrix, where the column by index i contains the coordinates of the vector
T (ei ).
⎨(B1 ,B2 ) is called the associated matrix
Definition 3.28 (see [2], p. 43). The matrix T
of the linear transformation T relative to the bases B1 and B2 , fixed in the vector
spaces U and V .
Example 3.29 (see [9]). Let be

f : R3 ∈ R2 , f (x) = f (x1 , x2 , x3 ) = (2x1 + x2 − x3 , x2 + 7x3 ) .

(a) Show that f is a linear transformation.

(b) Write the associated matrix of f relative to the canonical bases of the two spaces
R3 and R2 .
(c) Determine Ker f and Im f .
(d) Is f surjective?
Solution
(a) Let π, λ ∈ R and (→) x, y ∈ R3 , x = (x1 , x2 , x3 ), y = (y1 , y2 , y3 ). We shall
have:
f (πx + λ y) = f (πx1 + λy1 , πx2 + λy2 , πx3 + λy3 )
= (2 (πx1 + λy1 ) + πx2 + λy2 − (πx3 + λy3 ) , πx2 + λy2 + 7 (πx3 + λy3 ))
= (π (2x1 + x2 − x3 ) + λ (2y1 − y2 − y3 ) , π (x2 + 7x3 ) + λ (y2 + 7y3 ))
= (π (2x1 + x2 − x3 ) , π (x2 + 7x3 )) + (λ (2y1 − y2 − y3 ) , λ (y2 + 7y3 ))
= π (2x1 + x2 − x3 , x2 + 7x3 ) + λ (2y1 − y2 − y3 , y2 + 7y3 )
= π f (x) + λ f (y) ;
98 3 Linear Transformations

therefore using Proposition 3.2 it results that f is a linear mapping.

(b) As B1 = {e1 = (1, 0, 0) , e2 = (0, 1, 0) , e3 = (0, 0, 1)} is a basis in R3 and
B2 = f 1 = (1, 0) , f 2 = (0, 1) is a basis from R2 it results that

f (e1 ) = f (1, 0, 0) = (2, 0) = 2 · f 1 + 0 · f 2

f (e2 ) = f (0, 1, 0) = (1, 1) = 1 · f 1 + 1 · f 2
f (e3 ) = f (0, 0, 1) = (−1, 7) = (−1) · f 1 + 7 · f 2 .

We shall obtain ⎭ ⎧
⎨ 2 1 −1
f (B1 ,B2 ) = .
01 7

(c) We have (based on the Definition 3.5)

 
Ker f = x ∈ R3 | f (x) = 0R2 .

Let be x ∈ R3 .

f (x) = 0R2 ∃ (2x1 + x2 − x3 , x2 + 7x3 ) = (0, 0)


2x1 + x2 − x3 = 0
∃
x2 + 7x3 = 0.

The kernel of f is the set of solutions corresponding to the system:


2x1 + x2 = x3
⇔ 2x1 = 8x3 ⇔ x1 = 4x3 .
x2 = −7x3

Hence, x ∈Ker f ∃ x = (4x3 , −7x3 , x3 ) , x3 ∈ R3 .

As, within the Proposition 1.67

dim Ker f = n − r = 3 − 2 = 1,

taking into account that (see the Theorem 3.26)

dim R3 = dim Ker f + dim Im f

we obtain
dim Im f = 2.

(d) We know that Im f ≡ R2 i.e. Im f is a vector space of R2 (see the Theorem 3.22).
3.1 Linear Transformations 99

As 
Im f ≡ R2 ⎪
Definition 3.6
dim Im f = 2 ⇔ Im f = R2 =⇔ f is surjective.
⎛
dim R2 = 2

We can solve this problem in Sage, too:

Example 3.30 (see [9]). Let V be a vector space, dim V = 3 and f : V ∈ V ,

⎩ 
111
⎨
f (B) = 1 1 1, (3.8)
111

where B = {e1 , e2 , e3 }. Determine dim Ker f and dim Im f .

Solution
Let be
x ∈ V ⇔ x = π1 e1 + π2 e2 + π3 e3 ;

hence
f (x) = π1 f (e1 ) + π2 f (e2 ) + π3 f (e3 ) .

From (3.8) it results 

 f (e1 ) = e1 + e2 + e3
f (e2 ) = e1 + e2 + e3

f (e3 ) = e1 + e2 + e3 .

We obtain
 
Ker f = x ∈ V | f (x) = 0V = {x ∈ V |π1 + π2 + π3 = 0} ;

therefore
dim Ker f = n − r = 3 − 1 = 2.

Whereas
100 3 Linear Transformations

π1 = −π2 − π3

we deduce

x = (−π2 − π3 ) e1 + π2 e2 + π3 e3 = π2 (e2 − e1 ) + π3 (e3 − e1 ) ⇔

⎜ ⎝⎞ ⎟ ⎜ ⎝⎞ ⎟
c1 c2

{c1 , c2 } is a system of generators for Ker f .

However, as dimKer f = 2 it results that {c1 , c2 } is a basis in Ker f .
We have

dim V = dim Ker f + dim Im f ⇔ 3 = 2 + dim Im f ⇔ dim Im f = 1.

We know (see the Definition 3.5) that

Im f = { f (x) |x ∈ V } .

We shall obtain

f (x) = (π1 + π2 + π3 ) (e1 + e2 + e3 ) ⇔

⎜ ⎝⎞ ⎟
c3

{c3 } is a system of generators for Im f .

However, since dim Im f = 1 it results that {c3 } is a basis in Im f . We shall
present the solution of this problem in Sage:

Example 3.31 (see [9]). Let T : M2 (R) ∈ M2 (R) be a linear mapping, defined by
⎭ ⎧ ⎭ ⎧
1 −2 10
T (A) = A , (→) A ∈ M2 (R) .
0 1 11

Build the matrix of the linear mapping T in the canonical basis of that space.
3.1 Linear Transformations 101

Solution
We know that
⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧⎠
10 01 00 00
B = {E 11 , E 12 , E 21 , E 22 } = , , ,
00 00 10 01

is a canonical basis in M2 (R). We compute:

⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧
1 −2 1 0 1 0 10
T (E 11 ) = · · = = E 11
0 1 0 0 1 1 00
⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧
1 −2 0 1 1 0 11
T (E 12 ) = · · = = E 11 + E 12
0 1 0 0 1 1 00
⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧
1 −2 0 0 1 0 −2 0
T (E 21 ) = · · = = −2E 11 + E 21
0 1 1 0 1 1 1 0
⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧
1 −2 0 0 1 0 −2 −2
T (E 22 ) = · · =
0 1 0 1 1 1 1 1
= − 2E 11 − 2E 12 + E 21 + E 22 .

It results ⎩ 
1 1 −2 −2
0 1 0 −2 
⎨(B)
T =
0
.
0 1 1 
0 0 0 1

Using Sage, we achieve:

Example 3.32 (see [9]). Let be f : R4 [X ] ∈ R3 [X ] defined by

f (P) = P ≤ , (→) P ∈ R4 [X ],
102 3 Linear Transformations

where P ≤ means the formal derivative of the polynomial P.

(a) Show that the mapping f is linear.
(b) Write ⎨ f (B1 ,B2 ) namely
 the matrix of f relative to the canonical
 bases B1 =
1, X, X 2 , X 3 , X 4 from R4 [X ] and respectively to B2 = 1, X, X 2 , X 3 from
R3 [X ].
(c) Determine Ker f and Im f .
Solution
(a) Let π, λ ∈ R and P, Q ∈ R 4 [X ]. We have
f (π P + λ Q) = (π P + λ Q)≤ = π P ≤ + λ Q ≤
= π f (P) + λ f (Q) , (→) π, λ ∈ R, (→) P, Q ∈ R4 [X ],

namely the mapping f is linear.

We can check that in Sage, too:

(b) We compute


⎬ f (1) = 1≤ = 0 = 0 · 1 + 0 · X + 0 · X 2 + 0 · X 3
⎬
⎬
⎬
 f (X ) = X≤ = 1 = 1 · 1 + 0 · X + 0 · X2 + 0 · X3
 2   2 ≤
f X = X = 2X = 0 · 1 + 2 · X + 0 · X 2 + 0 · X 3
⎬
⎬    ≤
⎬
⎬ f X 3 = X 3 = 3X 2 = 0 · 1 + 0 · X + 3 · X 2 + 0 · X 3
  4   4 ≤
f X = X = 4X 3 = 0 · 1 + 0 · X + 0 · X 2 + 4 · X 3 .

We deduce that ⎩ 
0 1 0 0 0
0 0 2 0 0
⎨
f (B1 ,B2 ) =
0
.
0 0 3 0
0 0 0 0 4

We can obtain the same matrix with Sage, too:

3.1 Linear Transformations 103

(c) We have

 
Ker f = P ∈ R4 [X ]| f (P) = O R3 [X ] .

Let P ∈R4 [X ], hence

P = a0 + a1 X + a2 X 2 + a3 X 3 + a4 X 4 ;

then

f (P) = OR3 [X ] ∃ P ≤ = OR3 [X ]

∃ a1 + 2a2 X + 3a3 X 2 + 4a4 X 3 = 0 + 0 · X + 0 · X 2 + 0 · X 3 .

It results that: a1 = . . . = a4 = 0, i.e. P = a0 . Therefore

Ker f = {P ∈ R4 [X ]|P = a0 = a0 · 1} ,

i.e. B = {1} is a basis for Ker f ⇔ dim Ker f = 1.

As

dim R4 [X ] = dim Ker f + dim Im f ⇔ 5 = 1 + dim Im f ⇔ dim Im f = 4.

However 
Im f ≡ R3 [X ] ⎪
dim Im f = 4 ⇔ Im f = R3 [X ].
⎛
dim R3 [X ] = 4

3.2 Matrix as a Linear Mapping

We propose to define the matrix operations, starting from the corresponding opera-
tions of linear maps.
104 3 Linear Transformations
 
Let be S,T ∈ L(U, V ) and B1 = {e1 , . . . , en }, B2 = f 1 , . . . , f m be two bases
in U and respectively in V. Let A, B be the associated matrices of S and T relative
to the two bases: A = ⎨
S(B1 ,B2 ) , A = ai j 1∗i∗m and B = T ⎨(B1 ,B2 ) , B = bi j 1∗i∗m.
1∗ j∗n 1∗ j∗n
We have: 
⎬
⎬ S (e1 ) = a11 f 1 + · · · + am1 f m
⎬
⎬ ..
⎬
⎬
   .
S e j = a1 j f 1 + · · · + am j f m
⎬
⎬ ..
⎬
⎬
⎬
⎬ .

S (en ) = a1n f 1 + · · · + amn f m ,

⎬
⎬ T (e1 ) = b11 f 1 + · · · + bm1 f m
⎬
⎬ ..
⎬
⎬
   .
T e j = b1 j f 1 + · · · + bm j f m
⎬
⎬ ..
⎬
⎬
⎬
⎬ .

T (en ) = b1n f 1 + · · · + bmn f m .

(1) Matrix equality

   
S and T are equal∃ S e j = T e j , (→) j = 1, n ∃

a1 j f 1 + · · · ai j f i + · · · + am j f m = b1 j f 1 + · · · + bi j f i + · · · + bm j f m (3.9)
 
As f 1 , . . . , f m is a basis and we know that the writing of a vector into a basis
is unique, from (3.9) it results that ai j = bi j , (→) i = 1, m, (→) j = 1, n.
   
Definition 3.33 (see [8], p. 60). The matrices A = ai j 1∗i∗m and B = bi j 1∗i∗m
1∗ j∗n 1∗ j∗n
are equal if and only if

ai j = bi j , (→) i = 1, m, (→) j = 1, n.

(2) Matrix addition

We denote by C the associated matrix of the linear mapping S + T . We have

  de f    
(S + T ) e j = S e j + T e j
= a1 j f 1 + · · · ai j f i + · · · + am j f m + b1 j f 1 + · · · + bi j f i + · · · + bm j f m ,

namely
       
(S + T ) e j = a1 j + b1 j f 1 + · · · ai j + bi j f i + · · · + am j + bm j f m .
(3.10)
3.2 Matrix as a Linear Mapping 105

However
 
(S + T ) e j = c1 j f 1 + · · · + ci j f i + · · · + cm j f m . (3.11)

From (3.10) and (3.11) it results

ci j = ai j + bi j , (→) i = 1, m, (→) j = 1, n.
 
Definition 3.34 (see [8], p. 60). The sum of matrices A = ai j 1∗i∗m and B =
    1∗ j∗n
bi j 1∗i∗m is the matrix C = ci j 1∗i∗m , ci j = ai j +bi j , (→) i = 1, m, (→) j = 1, n.
1∗ j∗n 1∗ j∗n
We denote C = A + B.

(3) Scalar multiplication of the matrices

We shall denote by C the associated matrix of the linear mapping πS, π ∈ K .

 
(πS) e j = πa1 j f 1 + · · · + πai j f i + · · · + πam j f m . (3.12)

However  
(πS) e j = c1 j f 1 + · · · + ci j f i + · · · + cm j f m . (3.13)

From (3.12) and (3.13) it results

ci j = πai j , (→) i = 1, m, (→) j = 1, n.

 
Definition 3.35 (see [8], p. 60). By multiplying a matrix A = ai j 1∗i∗m with a
1∗ j∗n
scalar π ∈ K it results the matrix π A, whose elements are obtained by multiplying
the all elements of A with π.

(4) Matrix multiplication

 Let S ∈ L(U,  B1 = {e1 , . . . , en } be a basis in U , B2 =

 V ) , T ∈ L(V, W ) and
f 1 , . . . , f m be a basis in V , B3 = g 1 , . . . , g p be a basis in W. We denote by:
 
• A = ai j 1∗i∗m the associated matrix of S relative to the bases B1 and B2 ;
 1∗ j∗n
• B = bi j 1∗i∗ p the associated matrix of T relative to the bases B2 and B3 .
1∗ j∗m

We have:
      
(T ⇒ S) e j = T S e j = T a1 j f 1 + · · · ak j f k + · · · + am j f m
     
= a1 j T f 1 + · · · + ak j T f k + · · · + am j T f m ,

i.e.
106 3 Linear Transformations
   
(T ⇒ S) e j = a1 j b11 g 1 + · · · + bi1 g i + · · · + b p1 g p + · · · + (3.14)
 
ak j b1k g 1 + · · · + bik g i + · · · + b pk g p + · · · +
 
am j b1m g 1 + · · · + bim g i + · · · + b pm g p .
 
We denote by C = ci j 1∗i∗ p the associated matrix of the linear mapping T ⇒ S ∈
1∗ j∗n
L (U, W ). It results
 
(T ⇒ S) e j = c1 j g 1 + · · · + ci j g i + · · · + c pj g p . (3.15)

We write (3.14) as
   
(T ⇒ S) e j = b11 a1 j + · · · + b1k ak j + · · · + b1m am j g 1 (3.16)
 
+ · · · + bi1 a1 j + · · · + bik ak j + · · · + bim am j g i
 
+ · · · + b p1 a1 j + · · · + b pk ak j + · · · + b pm am j g p .

From (3.15) and (3.16) it results

ci j = bi1 a1 j + · · · + bik ak j + · · · + bim am j , (→) i = 1, p, (→) j = 1, n. (3.17)

 
Definition 3.36 (see [8], p. 60). The product of the matrices B = bi j 1∗i∗ p and
    1∗ j∗m
A = ai j 1∗i∗m is the matrix C = ci j 1∗i∗ p , ci j , (→) i = 1, p, (→) j = 1, n being
1∗ j∗n 1∗ j∗n
defined in (3.17). We denote C = B · A.
Remark 3.37 (see [8], p. 61). The product B · A is defined if and only if the number
of columns of B is equal to the number of lines of A.
Proposition 3.38 (see [8], p. 60). If A, B, C are matrices with proper dimensions,
such that the following products are defined and ϕ ∈ K , then:
(a) A (BC) = (AB) C
(b) A (B + C) = AB + AC
(c) (B + C) A = B A + C A
(d) A (ϕB) = (ϕA) B = ϕ (AB).
Remark 3.39 (see [8], p. 61). In general, the matrix multiplication is not commutative.

(5) Matrix inversion

Definition 3.40 (see [8], p. 61). The matrix A ∈ Mn (K ) is invertible if there is a

unique matrix B ∈ Mn (K ) such that

AB = B A = In .

The inverse of A is denoted with A−1 .

3.2 Matrix as a Linear Mapping 107

(6) Rank of a matrix

Theorem 3.41 (see [8], p. 61). Let U, V be two vector spaces over  the field K and
T ∈ L(U, V ). If B1 = {e1 , · · · , en } is a basis in U and B2 = f 1 , · · · , f n is a
basis in V and A is the associated matrix of the linear mapping T relative to the bases
B1 and B2 , then
rank T = rank A.

Proposition 3.42 (see [8], p. 62). A square matrix is invertible if and only if it is
nonsingular.

3.3 Changing the Associated Matrix to the Change of Basis

Let x ∈ U. We can write

x = x (1) e1 + · · · + x (n) en .

We shall have

T (x) = x (1) T (e1 ) + · · · + x (n) T (en )

⎦ !
(1) (2) (m)
= x (1) π1 f 1 + π1 f 2 + · · · + π1 f m
⎦ !
+ · · · + x (n) πn(1) f 1 + πn(2) f 2 + · · · + πn(m) f m ,

namely
⎦ ! ⎦ !
(1) (m)
T (x) = x (1) π1 + · · · + x (n) πn(1) f 1 + · · · + x (1) π1 + · · · + x (n) πn(m) f m .
(3.18)
As T (x) ∈ V it results

T (x) = y (1) f 1 + · · · + y (m) f m . (3.19)

From (3.18) and (3.19) we deduce

 (1) (1) (1) (n) (1)
 y = x π1 + · · · + x πn
⎬
..
⎬ .
 (m) (m) (m)
y = x (1) π1 + · · · + x (n) πn ;

therefore
⎨(B1 ,B2 ) · x B1 .
(T (x)) B2 = T (3.20)
108 3 Linear Transformations

Let B1≤ be another basis of U and B2≤ be another basis of V . Let C be the transition
matrix from the basis B1 to the basis B1≤ and D the transition matrix from the basis
B2 to the basis B2≤ .
Within the relation (3.20) we have

⎨ B ≤ ,B ≤ · x B ≤ .
(T (x)) B2≤ = T (3.21)
( 1 2) 1

We know that
x B1 = C( B1 ,B ≤ ) · x B1≤ (3.22)
1

and
(T (x)) B2 = D( B2 ,B ≤ ) · (T (x)) B2≤ . (3.23)
2

Equating (3.20) and (3.23) it results

⎨(B1 ,B2 ) · x B1 .
D( B2 ,B ≤ ) · (T (x)) B2≤ = T (3.24)
2

Substituting (3.24) into (3.22) it results

⎨(B1 ,B2 ) · C B ,B ≤ · x B ≤ .
D( B2 ,B ≤ ) · (T (x)) B2≤ = T (3.25)
2 ( 1 ) 1 1

If in the relation (3.25) we multiply at the left, the both members with the inverse
of D, we obtain

(T (x)) B2≤ = D(−1 ⎨(B1 ,B2 ) · C B ,B ≤ · x B ≤ .

·T (3.26)
B2 ,B2≤ ) ( 1 1) 1

From (3.21) and (3.26) it results

⎨ B ≤ ,B ≤ · x B ≤ = D −1 ≤ · T
T ⎨ ·C ≤ · x ≤,
( 1 2) 1 ( B2 ,B2 ) (B1 ,B2 ) ( B1 ,B1 ) B1

i.e.
⎨ B ≤ ,B ≤ = D −1 ≤ · T
T ⎨ ·C ≤ . (3.27)
( 1 2) ( B2 ,B2 ) (B1 ,B2 ) ( B1 ,B1 )

The formula (3.27) constitutes [9] the changing formula of the associated matrix
of a linear mapping when one changes the bases in the two vector spaces U and V .
 
Example 3.43 (see [5], p. 48). Let T1 , T2 ∈ End R3 , be defined as
⎦ !
T1 (x) = 5x (1) − x (2) − 5x (3) , 20x (1) − 15x (2) + 8x (3) , 3x (1) − 2x (2) + x (3)
⎦ !
T2 (x) = 10x (1) − 10x (2) + 10x (3) , 0, 5x (1) − 5x (2) + 5x (3) ,
⎦ !
(→) x = x (1) , x (2) , x (3) ∈ R3 .
3.3 Changing the Associated Matrix to the Change of Basis 109

Find the sum of the two endomorfisme matrix T = T1 + T2 relative to the basis
B ≤ = {v 1 = (2, 3, 1) , v 2 = (3, 4, 1) , v 3 = (1, 2, 2)} ≺ R3 .
Solution
We have:

T (x) = (T1 + T2 ) (x) = T1 (x) + T2 (x)

⎦ !
= 15x (1) − 11x (2) + 5x (3) , 20x (1) − 15x (2) + 8x (3) , 8x (1) − 7x (2) + 6x (3) .

Let B1 = {e1 = (1, 0, 0) , e2 = (0, 1, 0) , e3 = (0, 0, 1)} be the canonical base of

the space R3 .
Computing 
 T (e1 ) = (15, 20, 8)
T (e2 ) = (−11, −15, −7)

T (e3 ) = (5, 8, 6)

we shall obtain ⎩ 
15 −11 5
⎨(B)
T =  20 −15 8  .
8 −7 6

We denote by C the transition matrix from the basis B to the basis B ≤ .

As 
 v 1 = (2, 3, 1) = 2e1 + 3e2 + e3
v 2 = (3, 4, 1) = 3e1 + 4e2 + e3

v 3 = (1, 2, 2) = e1 + 2e2 + 2e3

we have ⎩ 
231
C = 3 4 2.
112

Hence ⎩ 
100
⎨(B ≤ )
T ⎨(B) · C =  0 2 0  .
= C −1 · T
003

We can check this result using Sage:

110 3 Linear Transformations

3.4 Eigenvalues and Eigenvectors

The eigenvalue problems are of great importance in many branches of physics. They
make it possible to find some coordinate systems in which the linear transformations
take the simplest forms.
For example, in mechanics the main moments of a solid body one finds with the
eigenvalues of a symmetric matrix representing the vector tensor. The situation is
similar in continuous mechanics, where a body rotations and deformations in the
main directions are found using the eigenvalues of a symmetric matrix.
Eigenvalues have [1] a central importance in quantum mechanics, where the mea-
sured values of the observable physical quantities appear as eigenvalues of operators.
Also, the eigenvalues are useful in the study of differential equations and contin-
uous dynamical systems that arise in areas such as physics and chemistry.
Definition 3.44 (see [2], p. 45). Let V be a vector space over the field K and
T ∈End(V ). The subspace vector W of V is called an invariant subspace rela-
tive to T if from x ∈ W it results T (x) ∈ W or T (W ) ≡ W .
Definition 3.45 (see [2], p. 45). Let V be a vector space over the field K and
T ∈End(V ). We say that the scalar ϕ ∈ K is an eigenval for T if there is x ∈ V \ 0
such that

T (x) = ϕx. (3.28)

 
Definition 3.46 (see [2], p. 45). The vector x ∈ V \ 0 for which there is ϕ ∈ K
such that T (x) = ϕx is called the eigenvector corresponding to the eigenvalue ϕ.
Remark 3.47 (see [9]). If ϕ is an eigenval for T then there are an infinite number of
eigenvectors corresponding to ϕ.
Proposition 3.48 (see [2], p. 45). Let V be a vector space over the field K and ϕ ∈ K
an eigenvalue for T . Then the set

Vϕ = {x ∈ V |T (x) = ϕx} (3.29)

is a vector subspace of V, invariant relative to T .

Definition 3.49 (see [2], p. 45). Let V be a vector space over the field K . If T ∈End(V )
and ϕ ∈ K an eigenvalue for T , then the vector subspace Vϕ is called the eigensub-
space associated to the eigenvalue ϕ.
3.4 Eigenvalues and Eigenvectors 111

Proposition 3.50 (see [3], p. 64). Let T : V ∈ V be an endomorphism of the

vector space V and a 1 , . . . a p the eigenvectors of the endomorphism T respectively
corresponding to the distinct eigenvalues ϕ1 , . . . ϕ p , i.e.

T (a i ) = ϕi a i , i = 1, p,

then a 1 , . . . a p are linearly independent.

3.4.1 Characteristic Polynomial of an Endomorphism

Let be the endomorphism T : V ∈ V and B = {e1 , . . . , en } be a basis of V . We

consider the expressions of the vectors T (e1 ) , . . . , T (en ) ∈ V in the basis B :

(1) (2) (n)
⎬
⎬ T (e1 ) = π1 e1 + π1 e2 + · · · + π1 en
⎬
⎬
 T (e2 ) = π (1) e1 + π (2) e2 + · · · + π (n) en
2 1 2
⎬ .. (3.30)
⎬
⎬ .
⎬
 (1) (2) (n)
T (en ) = πn e1 + πn e2 + · · · + πn en .

⎨(B) be the associated matrix of the linear mapping T relative to the basis B,
Let T
⎩ (1) (1) 
π1 . . . πn
⎨(B)  .. ..  .
T = . .  (3.31)
(n) (n)
π1 ... πn

Let be x ∈ V . We can write

x = x (1) e1 + · · · + x (n) en .

The relation (3.28) becomes

⎦ ! ⎦ !
T x (1) e1 + · · · + x (n) en = ϕ x (1) e1 + · · · + x (n) en . (3.32)

As T is a linear mapping, from (3.32) it results

⎦ !
x (1) T (e1 ) + · · · + x (n) T (en ) = ϕ x (1) e1 + · · · + x (n) en . (3.33)

If in (3.33) we take into account the relations (3.30) we obtain

⎦ !
(1) (2) (n)
x (1) π1 e1 + π1 e2 + · · · + π1 en + · · · + (3.34)
112 3 Linear Transformations
⎦ !
x (n) πn(1) e1 + πn(2) e2 + · · · + πn(n) en
⎦ !
= ϕ x (1) e1 + · · · + x (n) en .

From (3.34) we deduce

⎦ !
⎬ (1) (1) (1)
⎬
⎬ π − ϕ x (1) + π2 x (2) + · · · + πn x (n) = 0
⎬
⎬
1 ⎦ !
⎬
 π (2) x (1) + π (2) − ϕ x (2) + · · · + πn(2) x (n) = 0
1 2
.. (3.35)
⎬
⎬
⎬
⎬ . ⎦ !
⎬
⎬
 π (n) x (1) + π (n) x (2) + · · · + πn(n) − ϕ x (n) = 0
1 2

i.e. a linear and homogeneous system in the unknowns x (1) , x (2) , . . . , x (n) .
The scalar ϕ is an eigenvalue of T if and only if the system (3.35) admits the
nonbanal solutions, i.e. there is a scalar system x (1) , x (2) , . . . , x (n) , not all null, that
verifies the system (3.35).
This is achieved if and only if
⎫ ⎫
⎫ π (1) − ϕ π (1) · · · π (1) ⎫
⎫ 1 2 n ⎫
⎫ ⎫
⎫ π1(2) π2(2) − ϕ · · · πn(2) ⎫
ρ=⎫ ⎫ = 0,
⎫ ··· ··· ··· ··· ⎫
⎫ ⎫
⎫ π (n) π2
(n) (n)
· · · πn − ϕ ⎫
1

i.e.  
det T⎨(B) − ϕIn = 0, (3.36)

In being the unit matrix of order n.

In conclusion, ϕ is an eigenvalue of T if and only if there is a root in K of the
Eq. (3.36).
We denote  
P (ϕ) = det T⎨(B) − ϕIn . (3.37)

Remark 3.51 (see [3], p. 66). The polynomial P (ϕ) is a polynomial in ϕ, of degree n:

P (ϕ) = (−1)n ϕn + (−1)n−1 δ1 ϕn−1 + (−1)n−2 δ2 ϕn−2 + · · · + (−1) δn−1 ϕ + δn ,

where
3.4 Eigenvalues and Eigenvectors 113

⎬
⎬    n
(i)
⎬ δ = trace ⎨(B) =
T πi
⎬
⎬ 1
⎬
⎬ ⎫ i=1 ⎫
⎬
⎬
  ⎫⎫ π (i) π (i) ⎫⎫
δ2 = ⎫ (i j) (j j) ⎫
⎬
⎬ ⎫ πi π j ⎫
⎬
⎬ 1∗i< j∗n
⎬
⎬ ..
⎬
⎬ .
⎬
 
δn = det T ⎨(B) .

Definition 3.52 (see [2], p. 45). The polynomial P (ϕ) defined in (3.37), in the
indeterminate ϕ, of degree n is called the characteristic polynomial associated of
the endomorphism T .
Definition 3.53 (see [3], p. 66). The equation P (ϕ) = 0 is called the characteristic
equation associated of the endomorphism T .
Remark 3.54 (see [3], p. 66). The scalar ϕ is an eigenvalue of T if and only if ϕ is
the root of the characteristic equation.
Example 3.55 (see [9]). Let T be be an endomorphism of R3 such that T has
the eigenvalues: ϕ1 = 1, ϕ2 = −1, ϕ3 = 2 with the eigenvectors x 1 = (1, 0, 1),
x 2 = (−1, 2, 1), x 3 = (2, 1, −1). Write the associated matrix of T in the canonical
basis from R3 .
Solution
We note that 
 x 1 = e1 + e3
x 2 = −e1 + 2e2 + e3

x 3 = 2e1 + e2 − e3 .

As
T (x 1 ) = ϕ1 x 1

we deduce
T (e1 + e3 ) = ϕ1 (e1 + e3 ) . (3.38)

Taking into account that T is a linear mapping from (3.38) we obtain

T (e1 ) + T (e3 ) = ϕ1 (e1 + e3 ) . (3.39)

Similarly, because
T (x 2 ) = ϕ2 x 2

we deduce
T (−e1 + 2e2 + e3 ) = ϕ2 (−e1 + 2e2 + e3 ) . (3.40)

Taking into account that T is a linear mapping from (3.40) we achieve:

114 3 Linear Transformations

− T (e1 ) + 2T (e2 ) + T (e3 ) = ϕ2 (−e1 + 2e2 + e3 ) . (3.41)

Similarly, as
T (x 3 ) = ϕ3 x 3

we shall have
T (2e1 + e2 − e3 ) = ϕ3 (2e1 + e2 − e3 ) . (3.42)

As T is a linear mapping from (3.42), it results:

2T (e1 ) + T (e2 ) − T (e3 ) = ϕ3 (2e1 + e2 − e3 ) . (3.43)

We shall solve the system of equations resulting from the relations (3.39), (3.41)
and (3.43) to determine the expression of T (ei ) , i = 1, 3 as a linear combination of
the elements of the basis, i.e.:

 T (e1 ) + T (e3 ) = e1 + e3
−T (e1 ) + 2T (e2 ) + T (e3 ) = e1 − 2e2 − e3 (3.44)

2T (e1 ) + T (e2 ) − T (e3 ) = 4e1 + 2e2 − 2e3 .

Adding the first two equations we deduce:

2T (e2 ) + 2T (e3 ) = 2 (e1 − e2 ) ,

i.e.
T (e2 ) + T (e3 ) = e1 − e2 . (3.45)

Adding the last two equations we deduce:

T (e1 ) + 3T (e2 ) = 5e1 − 3e3 . (3.46)

From (3.45) it results

T (e3 ) = e1 − e2 − T (e2 ) . (3.47)

From (3.46) it results

T (e1 ) = 5e1 − 3e3 − 3T (e2 ) . (3.48)

Substituting (3.47) and (3.48) into the first equation of the system (3.44) we obtain

5e1 − 3e3 − 3T (e2 ) + e1 − e2 − T (e2 ) = e1 + e3 ,

i.e.
5e1 − e2 − 4e3 = 4T (e2 ) ;
3.4 Eigenvalues and Eigenvectors 115

hence
5 1
T (e2 ) = e1 − e2 − e3 .
4 4
We shall have
⎭ ⎧
5 1 5 3
T (e1 ) = 5e1 − 3e3 − 3 e1 − e2 − e3 = e1 + e2
4 4 4 4

and ⎭ ⎧
5 1 1 3
T (e3 ) = e1 − e2 − e1 − e2 − e3 = − e1 − e2 + e3 .
4 4 4 4

We obtain ⎩ 
5/4 5/4 −1/4
⎨(B)
T =  3/4 −1/4 −3/4  .
0 −1 1

The same matrix can be achieved using Sage:

Definition 3.56 (see [5], p. 64). The set of the roots of the characteristic equation, that
is associated to the endomorphism T is called the spectrum of the endomorphism
T . If all the roots are simple on says that T is an endomorphism with a simple
spectrum. We denote by θ (T ) the spectrum of T .
Theorem 3.56 (Hamilton-Cayley, see [2], p. 49). Let V be a n dimensional vector
space over K , n ∞ 1 and be an endomorphism T ∈End(V ). If P (ϕ) is the character-
istic polynomial of A = T⎨(B) (the matrix of the endomorphism T relative to a basis
B of V , then P (A) = 0End(V ) .
Example 3.57 (see [7], p. 29). Compute

P (A) = A4 − 8A3 + 24 A2 − 32 A + 16I4 ,

where A is the matrix

116 3 Linear Transformations
⎩ 
2 0 0 0
 1 3 1 1 
A=
 0
.
0 1 −1 
−1 −1 0 2

Solution
The characteristic polynomial, that is associated to the matrix A is
⎫ ⎫
⎫2 − ϕ 0 0 0 ⎫⎫
⎫
⎫ 1 3−ϕ 1 1 ⎫⎫
P (ϕ) = ⎫⎫ ⎫ = ϕ − 8ϕ + 24ϕ − 32ϕ + 16.
4 3 2
⎫ 0 0 1 − ϕ −1 ⎫
⎫ −1 −1 0 2 − ϕ⎫

Using the Hamilton- Cayley theorem we obtain P (A) = OM4 (R4 ) .

We shall use Sage to check the previous result:

3.4.2 Determining the Eigenvalues and the Eigenvectors

for an Endomorphism

To determine the eigenvalues associated of an endomorphism we proceed [9] as:

• write the characteristic equation;
• solve the characteristic equation;
• achieve the eigenvalues of the endomorphism as roots of the characteristic equa-
tion, that are in K .
To determine the eigenvectors corresponding to an eigenvalue ϕ0 of T we proceed
as:
• rewrite the system (3.35), by replacing ϕ with ϕ0 ;
• determine the vector subspace Wϕ0 of the solutions of the obtained linear and
homogeneous system (we find its dimension and a basis), called the associated
eigensubspace of the eigenvalue ϕ0 ;
• all non-zero vectors from the vector subspace are eigenvectors for the associated
eigenvalue ϕ0 .
3.4 Eigenvalues and Eigenvectors 117

Definition 3.58 (see [3], p. 66). The dimension of the eigensubspace Wϕ0 associated
of the eigenvalue ϕ0 is called the geometric multiplicity of ϕ0 and it is denoted
by gϕ0 .
Definition 3.59 (see [3], p. 66). The algebraic multiplicity of the eigenvalue ϕ0 ,
denoted by aϕ0 means the multiplicity of ϕ0 as a root of the characteristic polynomial
P (ϕ), associated to the endomorphism T .
Proposition 3.60 (see [3], p. 66). The characteristic polynomial P (ϕ) is invariant
relative to the basis changing in the vector space V .
Theorem 3.61 (see [3], p. 66). Let V be a vector space over the field K , dimV =
n < √, T ∈End(V ) and ϕ0 an eigenvalue of T . Then the geometric multiplicity of
ϕ0 is not greater than the algebraic multiplicity of ϕ0 , i.e. gϕ0 ∗ aϕ0 .
Proposition 3.62 (see [5], p. 64). Let V be a vector space over the field K and
T ∈End(V ) . Then each eigenvector of T corresponds to a single eigenvalue ϕ ∈
θ (T ).

3.4.3 Diagonalization Algorithm of an Endomorphism

We suppose that T ∈ End (V ) has n distinct eigenvalues ϕ1 , . . . , ϕn and dim V = n.

We denote by a 1 , . . . , a n the eigenvectors of the endomorphism T corresponding
respectively to the eigenvalues ϕ1 , . . . , ϕn .
Let B = {a 1 , . . . , a n } be a basis of eigenvectors corresponding to T ; therefore

 T (a 1 ) = ϕ1 a 1
..
 .
T (a n ) = ϕn a n .

The matrix ⎩ 
ϕ1 0 ··· 0
0 ϕ2 0 
⎨(B)
T =
0 . 
0 .. 0 
0 0 ϕn

is a diagonal matrix.
Definition 3.63 (see [5], p. 69). Let V be a vector space over the finite dimensional
field K , n ∞ 1. We say that the endomorphism T ∈End(V ) is diagonalizable if
there is a basis of V relative to which its matrix is a diagonal matrix.
Theorem 3.64 (see [3], p. 68). Let V be a vector space over the n finite dimen-
sional field K , n ∞ 1. The necessary and sufficient condition that the endomorphism
T ∈End(V ) to be diagonalizable is that the characteristic polynomial P (ϕ) to have
all the roots in K and the geometric multiplicity of each eigenvalue to be equal to its
algebraic multiplicity.
118 3 Linear Transformations

The diagonalization algorithm of an endomorphism T ∈End(V ) consists [9] in

the following steps:
1. choose a certain basis B ≡ V and determine the matrix T ⎨(B) associated to the
endomorphism T in this basis;
2. find the eigenvalues ϕ1 , . . . , ϕ p and of their corresponding algebraic multiplicites:
aϕ1 , . . . , aϕ p ;
3. achieve the eigensubspaces Wϕ1 , . . . , Wϕ p corresponding to the eigenvalues
ϕ1 , . . . , ϕ p ;
4. determine the basis Bi of the eigenspace Wϕi associated to the eigenvalue ϕi ,
i = 1, p and of the geometric multiplicity gϕi , corresponding to the eigenvalue
ϕi , i = 1, p;
5. check the condition (from the Theorem 3.64) that the endomorphism T to be
diagonalizable;
6. obtain the basis B ≤ of the vector space V relative to which the associated matrix
of T has the canonical diagonal form:

B ≤ = B1 ⊂ B2 ⊂ . . . ⊂ B p .

The associated matrix of T in the basis B ≤ is a diagonal matrix, having on its

diagonal the eigenvalues ϕ1 , . . . , ϕ p , each of them appearing of a number of times
equal to its order of multiplicity:
⎩ 
ϕ1 0
 .. 
 . 
 
 ϕ1 
 
⎨(B ≤ )  .. 
T = . 
 
 ϕp 
 
 .. 
 . 
0 ϕp

7. build the transition matrix from the basis B to the basis B ≤ , i.e. M(B,B ≤ ) ;
8. test the correctness of the calculations using the relation

⎨(B ≤ ) = M−1 ≤ · T
T ⎨(B) · M(B,B ≤ ) .
(B,B )

Example 3.65 (see [9]). On the vector space of matrices of second order one considers
the mapping:
T : M2 (R) ∈ M2 (R) , T (A) = At .

(a) Write the associated matrix of T relative to the canonical basis of the space
M2 (R).
(b) Determine the eigenvalues and the corresponding eigenspaces.
3.4 Eigenvalues and Eigenvectors 119

(c) Determine a basis B ≤ of the vector space M2 (R) relative to which the associated
matrix of T has a diagonal form.
Solution

(a) We know that (see the Example 3.31)

⎭ ⎧ ⎭ ⎧ ⎭ ⎧ ⎭ ⎧⎠
10 01 00 00
B = {E 11 , E 12 , E 21 , E 22 } = , , ,
00 00 10 01

is a canonical basis in M2 (R).

We compute
⎭ ⎧
10
T (E 11 ) =E 11
t
= = E 11
00
⎭ ⎧
00
T (E 12 ) =E 12
t
= = E 21
10
⎭ ⎧
01
T (E 21 ) =E 21
t
= = E 12
00
⎭ ⎧
00
T (E 22 ) =E 22
t
= = E 22 .
01

The associated matrix of T relative to the canonical basis of the space M2 (R)
will be ⎩ 
1000
 
⎨(B) =  0 0 1 0  .
T 0 1 0 0
0001

(b) We determine

⎫ ⎫
⎫1 − ϕ 0 0 0 ⎫⎫
⎫
  ⎫ 0 ⎫⎫
P (ϕ) = det T⎨(B) − ϕI4 = ⎫ 0 −ϕ 1
⎫ 0 1 −ϕ 0 ⎫⎫
⎫
⎫ 0 0 0 1 − ϕ⎫
⎦ !
= (1 − ϕ)2 ϕ2 − 1 = (ϕ − 1)3 (ϕ + 1) .

The roots of the characteristic equation are:

• ϕ1 = 1, having aϕ1 = 3;
• ϕ2 = −1, having aϕ2 = 1.
The associated eigenspace of the eigenvalue ϕ1 is
120 3 Linear Transformations

Wϕ1 = {A ∈ M2 (R) |T (A) = ϕ1 A} .

As T (A) = At and ϕ1 = 1 we obtain

 
Wϕ1 = A ∈ M2 (R) |At = A .
⎭ ⎧
a11 a12
Let be A ∈M2 (R) , A = . From the condition At = A we deduce
a21 a22
⎭ ⎧ ⎭ ⎧
a11 a21 a11 a12
= ∃ a12 = a21 .
a12 a22 a21 a22

Hence  ⎭ ⎧⎠
a11 a12
Wϕ1 = A ∈ M2 (R) |A = .
a12 a22

We can write
⎭ ⎧ ⎭ ⎧ ⎭ ⎧
10 01 00
A = a11 + a12 + a22
00 10 01
⎭ ⎧
01
= a11 E 11 + a12 + a22 E 22 ;
10

it results that
 ⎭ ⎧⎠
01
B1 = E 11 , E 22 , (3.49)
10

is a system of generators for Wϕ1 .

We note that B1 is also linearly independent as if
⎭ ⎧
01
a11 E 11 + a12 + a22 E 22 = OM2 (R)
10

it results ⎭ ⎧ ⎭ ⎧
a11 a12 00
= ,
a12 a22 00

i.e. a11 = a12 = a22 = 0.

Therefore B1 from (3.49) is a basis of Wϕ1 and gϕ1 = 3.
The associated eigenspace of the eigenvalue ϕ2 is

Wϕ2 = {A ∈ M2 (R) |T (A) = ϕ2 A} ,

i.e.
3.4 Eigenvalues and Eigenvectors 121
 
Wϕ2 = A ∈ M2 (R) |At = −A .

From the condition At = −A we deduce


⎭ ⎧ ⎭ ⎧  a11 = 0
a11 a21 −a11 −a12
= ∃ a12 = −a21
a12 a22 −a21 −a22 
a22 = 0.

Therefore  ⎭ ⎧⎠
0 a12
Wϕ2 = A ∈ M2 (R) |A = .
−a12 0

We can write ⎭ ⎧
0 1
A = a12 .
−1 0

Similarly, we obtain ⎭ ⎧⎠
0 1
B2 = (3.50)
−1 0

is a basis of Wϕ2 and gϕ2 = 1.

(c) As


aϕ1 = gϕ1 ⎪
aϕ2 = gϕ2 ⇔ T is diagonalizable.
⎛
the characteristic equation has some real roots

The basis of the vector space M2 (R) relative to which the associated matrix of T
has the canonical diagonal form is
 ⎭ ⎧ ⎭ ⎧⎠
≤ (3.49)+(3.50) 01 0 1
B = B1 ⊂ B2 = E 11 , E 22 , , = {F1 , F2 , F3 , F4 }
10 −1 0

and
⎩ 
1 0 0 0
0 1 0 0 
⎨(B ≤ )
T =
0
.
0 1 0 
0 0 0 −1

We shall have:

F1 = E 11 = 1 · E 11 + 0 · E 12 + 0 · E 21 + 0 · E 22
F2 = E 22 = 0 · E 11 + 0 · E 12 + 0 · E 21 + 1 · E 22
122 3 Linear Transformations

F3 = E 12 + E 21 = 0 · E 11 + 1 · E 12 + 1 · E 21 + 0 · E 22
F4 = E 12 − E 21 = 0 · E 11 + 1 · E 12 − 1 · E 21 + 0 · E 22 .

Therefore
⎩ 
1 0 0 0
0 0 1 1 
M(B,B ≤ ) =
0
.
0 1 −1 
0 1 0 0

It results that
⎩ 
1 0 0 0
0 1 0 0 
M−1 ⎨
(B,B ≤ ) · T(B) · M(B,B )
≤ =
0
=T
⎨(B ≤ ) .
0 1 0 
0 0 0 −1

The same matrix can be achieved in Sage:

3.4.4 Jordan Canonical Form

Let V be a n finite dimensional vector space over the field K .

If in the previous paragraph we showed the necessary and sufficient conditions
that an endomorphism T ∈End(V ) to be diagonalizable, in this section we propose
to determine a basis of V relative to which the associated matrix of T to have a
simpler form called the Jordan canonical form.
3.4 Eigenvalues and Eigenvectors 123

Definition 3.66 (see [2], p. 47). A square matrix of the form

⎩ 
ϕ 1 0 0 ··· 0 0 0
 0 ϕ 1 0 ··· 0 0 0 
 
 · · · · · · · · · · · · · · · · · · · · · · · ·  ∈ M p (K ) (3.51)
 
 0 0 0 0 ··· 0 ϕ 1 
0 0 0 0 0 ··· 0 ϕ

is called a Jordan cell of the order p, denoted with J p (ϕ).

Example 3.67 The matrix ⎩ 
ϕ 1 00
0 ϕ 1 0
J4 (ϕ) = 
0

0 ϕ 1
0 0 0ϕ

is a Jordan cell of the fourth order.

This matrix can be generated with Sage:

Definition 3.68 (see [2], p. 47). A square matrix of the form

⎩ 
J p1 (ϕ) O
 J p2 (ϕ) 
 
B p (ϕ) =  ..  ∈ M p (K ) , (3.52)
 . 
O J pr (ϕ)

where p1 + p2 + · · · + pr = p is called a Jordan block of the order p.

Example 3.69. The matrices

⎩ 
ϕ1 0 00
⎭ ⎧ 0 ϕ 1 0 0
J2 (ϕ) O  
B5 (ϕ) = =
0 0 ϕ 1 0
,
O J3 (ϕ) 0 0 0 ϕ 1
00 0 0ϕ
124 3 Linear Transformations
⎩ 
ϕ 10 00 0 00
0 ϕ1 00 0 0 0
 
 0ϕ 0 0
⎭ ⎧ 0 10 0 
J4 (ϕ) O 0 00 ϕ1 0 0 0
B8 (ϕ) = =
0

O J4 (ϕ)  00 0ϕ 1 0 0

0 00 00 ϕ 1 0
 
0 00 00 0 ϕ 1
0 00 00 0 0ϕ

are some Jordan blocks of five and respectively eight order.

We shall use Sage to generate B5 (ϕ) :

We can also obtain B8 (ϕ) in Sage, too:

Definition 3.70 (see [2], p. 47). A square matrix of order n, which has Jordan blocks
on the main diagonal, i.e. of the form
⎩ 
Bn 1 (ϕ1 ) O
 Bn 2 (ϕ2 ) 
 
J= .  ∈ Mn (K ) , (3.53)
 .. 
O Bnr (ϕr )

where n 1 + n 2 + · · · + nr = n is called a matrix in the Jordan canonical form.

3.4 Eigenvalues and Eigenvectors 125

Example 3.71. The matrix

⎩ 
ϕ1 1 0 0 0 0
0 ϕ1 1 0 0 0
⎭ ⎧  
B3 (ϕ1 ) O 0 0 ϕ1 1 0 0
J= =
0

O B3 (ϕ2 )  0 0 ϕ2 1 0 
0 0 0 0 ϕ2 1
0 0 0 0 0 ϕ2

is a matrix in the Jordan canonical form.

The matrix J can be generated in Sage:

Definition 3.72 (see [2], p. 47). The endomorphism T ∈End(V ) is jordanizable

if there is a basis of V relative to which the associated matrix of T has a Jordan
canonical form.
Definition 3.73 (see [2], p. 47). A matrix A ∈Mn (K ) is called jordanizable if there
is a nonsingular matrix C ∈Mn (K ) such that C −1 AC be a matrix in the Jordan
canonical form.
Theorem 3.74 (see [2], p. 47). Let V be a n dimensional vector space over K , n ∞ 1
and T ∈End(V ) be an andomorphism whose characteristic polynomial is

P (ϕ) = (−1)n (ϕ − ϕ1 )m 1 · . . . · (ϕ − ϕr )m r , (3.54)

where the eigenvalues ϕ1 , ϕ2 , . . . , ϕr are distinct and m 1 + m 2 + · · · + m r = n.

Then there is a basis of V relative to which the associated matrix of T has the
Jordan canonical form.

3.4.4.1 Jordanization Algorithm of an Endomorphism

Let V be a finite n dimensional vector space over the field K , n ∞ 1 and let
T ∈End(V ) be an endomorphism.
The jordanization algorithm of an endomorphism T consists [2] of the following
steps:

1. Choose a basis B of V and write the associated matrix of T relative to this basis,
⎨(B) .
i.e. the matrix T
2. Determine the characteristic polynomial P (ϕ) using (3.37); there are two cases:
126 3 Linear Transformations

(a) the characteristic equation P (ϕ) = 0 hasn’t n roots in K , therefore T is not

jordanizable;
(b) the characteristic equation has the roots ϕ1 , ϕ2 , . . . , ϕr with the multiplicities
m 1 , m 2 , . . . , m r , with m 1 + m 2 + · · · + m r = n.
⎨(B) − ϕi In .
3. For the eigenvalue ϕi one computes T
4. Determine the number of the Jordan cells for the eigenvalue ϕi ,

n i = dim Ker (T − ϕi V ) , (3.55)

V being the identity mapping.

Therefore, the number of Jordan cells for the eigenvalue ϕi is equal to the maxi-
mum number of the corresponding linearly independent eigenvectors. There are
two situations:
(a) if n i = m i then a basis of the eigenspace Wϕi associated of the eigenvalue
ϕi will consist of n i linearly independent eigenvectors;
(b) if n i < m i then go to the next step.
5. Find the smallest natural number si ∈ N∩ , si ∗ m i such that

dim Ker (T − ϕi V )si = m i . (3.56)

6. Determine the number of the Jordan cells of order h ∈ {1, 2, . . . , si } within the
formula
dh = rank (T − ϕi V )h+1 + rank (T − ϕi V )h−1 − 2 · rank (T − ϕi V )h , (3.57)

where

⎬
⎬ rank (T − ϕi V )0 = n
⎬
 rank (T − ϕi V )si +1 = rank (T − ϕi V )si
⎬ si (3.58)
⎬
⎬ h · dh = m i .

h=1

7. Repeat the steps 3–6 for each eigenvalue of T .

8. Write the matrix J (see (3.53)) of T in the Jordan canonical form.
9. Achieve the basis B ≤ of V relative to which T has the matrix J.

Example 3.75 (see [2], p. 88). Let T : R3 ∈ R3 be an endomorphism whose matrix

relative to the canonical basis of R3 is
⎩ 
3 −2 1
A =  2 −2 2  .
3 −6 5
3.4 Eigenvalues and Eigenvectors 127

(a) Show that T is jordanizable, write its matrix to the Jordan canonical form J
and find the basis of R3 relative to which T has the matrix J.
(b) Compute An , n ∈ N∩ .

Solution

(a) Using (3.54), the characteristic polynomial will be:

P (ϕ) = (2 − ϕ)3

and the characteristic equation has the root ϕ1 = 2, with m 1 = 3 = n; hence T is

jordanizable.
The associated matrix of the endomorphism T − ϕi V is
⎩ 
1 −2 1
A − 2I3 =  2 −4 2  .
3 −6 3

Since according to the Theorem 3.22, Ker(T − ϕi V ) is a vector subspace of R3 ,

from the Proposition 1.67 it results that

dim Ker (T − ϕi V ) = 3 − r,

where
r = rank (A − 2I3 ) = 1;

therefore
n 1 = dim Ker (T − ϕi V ) = 3 − 1 = 2.

As n 1 < m 1 we shall determine (with (3.56)) the smallest natural number s1 ∈ N∩ ,

s1 ∗ m 1 such that
dim Ker (T − ϕ1 V )s1 = m 1 .

We note that

(A − 2I3 )2 = O3 (null matrix of the third order);

hence

rank (A − 2I3 )2 = 0 ⇔ dim Ker (T − ϕ1 V )2 = 3 − 0 = 3 = m 1 ,

i.e. s1 = 2.
We have

d1 = 0 + 3 − 2 = 1,
128 3 Linear Transformations

d2 = 0 + 1 − 0 = 1.

Therefore, there will be two Jordan cells in the achieved matrix in the Jordan
canonical form, namely a first order cell J1 (2) and a second-order cell J2 (2). The
matrix J of T will be the following in the Jordan canonical form:
⎩ 
⎭ ⎧ 200
J1 (2) O
J= = 0 2 1
O J2 (2)
002
 
and the basis of R3 relative to which T has this form is B ≤ = f 1 , f 2 , f 3 .
Using the definition of the associated matrix to an endomorphism we shall have:
  
 T  f 1 = 2 f 1
T f = 2f2
   2
T f 3 = f 2 + 2 f 3.

As the number of Jordan cells for the eigenvalue ϕ1 is equal to the maximum
number of the corresponding linearly independent eigenvectors, it results that the
vectors f 1 , f 2 are some eigenvectors for T , i.e. their coordinates are the solution of
system: ⎩ ⎩  ⎩ 
3 −2 1 x x
 2 −2 2   y  = 2  y  ∃
3 −6 5 z z
 
 3x − 2y + z = 2x  x − 2y + z = 0
2x − 2y + 2z = 2y ∃ 2x − 4y + 2z = 0
 
3x − 6y + 5z = 2z 3x − 6y + 3z = 0.

The rank of the associated matrix of the system is equal to 1, so we denote

y = t, z = u, (→) t, u ∈ R;

it results x = 2t − u.
We shall have
 
Vϕ1 = f ∈ R3 | f = (2t − u, t, u) , t, u ∈ R .

For t = 1, u = 0 we consider f 1 = (2, 1, 0). The eigenvector f 2 belongs (as f 1 )

to the family of vectors

f = (2t − u, t, u) , t, u ∈ R, |t| + |u| ⇐= 0.

The vector f 3 one determines from the relation

3.4 Eigenvalues and Eigenvectors 129

 
T f 3 = f 2 + 2 f 3,

i.e. from the system

⎩ ⎩  ⎩  ⎩ 
3 −2 1 x 2t − u x
 2 −2 2   y  =  t  + 2  y  ∃
3 −6 5 z u z
 
 3x − 2y + z = 2t − u + 2x  x − 2y + z = 2t − u
2x − 2y + 2z = t + 2y ∃ 2x − 4y + 2z = t
 
3x − 6y + 5z = u + 2z 3x − 6y + 3z = u.

The previous system is compatible if and only if: 3t − 2u = 0.

The rank of the associated matrix of the system is equal to 1, therefore we denote

y = v, z = w, (→) v, w ∈ R;

it results x = 2t − u + 2v − w.
For t = 2, u = 3 we consider f 2 = (1, 2, 3). In the case when t = 2, u = 3 for
v = w = 0 we obtain f 3 = (1, 0, 0) .

(b) The transition matrix from the canonical basis B to the basis B ≤ will be

⎩ 
211
C = M(B,B ≤ ) = 1 2 0.
030

We shall obtain
⎩  ⎩  ⎩ 
211 200 0 1 −2/3
A = CJC −1 =  1 2 0  ·  0 2 1  ·  0 0 1/3 
030 002 −1 2 1

and

⎩  ⎩ n  ⎩ 
211 2 0 0 0 1 −2/3
An = CJn C −1 =  1 2 0  ·  0 2n n · 2n−1  ·  0 0 1/3 
030 0 0 2n −1 2 1
⎩ n 
2 +n·2 n−1 −n · 2 n n·2 n−1

⇔ An =  n · 2n (1 − 2n) 2n n · 2n .
3n · 2 n−1 −3n · 2 (2 + 3n) 2n−1
n

We shall present the solution in Sage, too:

130 3 Linear Transformations

3.5 Problems

1. Let be the linear mapping

T : R2 ∈ R2 , T (x) = T (x1 , x2 ) = (x1 , 3x1 − x2 ) .

Show that T 2 =  (the identity mapping).

Solution
Using Sage, we have:
3.5 Problems 131

2. Find the matrix A in each of the following cases:

⎭ ⎭ ⎧⎧t ⎭ ⎧
10 80
(a) 3At + 2 = ;
02 31
  t  t
(b) 2 A − 3 · 1 2 0 = 3At + 2 1 −1 .

Solution
We shall use Sage to find the matrix A :

3. One considers the mapping

T : R2 [X ] ∈ R2 [X ] , T (P) = (4X + 1) P ≤ .

(a) Show that T is a linear transformation.

(b) Write theassociated matrix of T relative to the canonical basis B =
1, X, X 2 of the vector space R2 [X ].
(c) Determine KerT and ImT .

4. In the case of the previous problem:

(a) Find the eigenvalues and the eigenspace of the corresponding eigenvectors.
(b) Decide if T is diagonalizable or not and if so write the diagonal form of the
matrix and specify the form relative of which the matrix is diagonal.

5. Using the Hamilton-Cayley theorem compute the inverse of the matrix

⎩ 
1 3 −1
1 0 2 .
22 1

Solution
The solution of the problem in Sage is:
132 3 Linear Transformations

6. Let be the endomorphism T : R3 ∈ R3 , given through the matrix

⎩ 
1 −1 2
A = 1 0 1 
1 0 −1

relative to the canonical basis of R3 . Determine the matrix of T relative to the basis
 
B1 = f 1 = (1, 2, 3) , f 2 = (3, 1, 2) , f 3 = (2, 3, 1) .

Solution
This matrix can be determined in Sage:

7. Let V be a finite dimensional vector space, dim V = 3, T be an endomorphism

of V and B = {e1 , e2 , e3 } be a basis of V such that
⎩ 
111
⎨(B) =  1 1 1  .
T
111

(a) Determine the eigenvalues and the eigenvectors associated to the endomor-
phism T .
(b) Check if T is diagonalizable and then determine a space basis relative to
which the associated matrix of T has the canonical diagonal form.
 
8. Let the mapping T :R3 [X ] ∈ R4 , T (P) = P (−3) P (−1) P (1) P (3) .
3.5 Problems 133

(a) Show that T is a linear transformation.

(b) Compute T (Q), for Q (t) = 5 + 2t − t 2 .
(c) Write the2 associated
 matrix of T relative to the canonical basis B1 =
1, X, X , X 3 from R3 [X ] and respectively to the canonical basis from R4 .

9. Let be the mapping


2 "1
T : R3 [X ] ∈ R3 [X ] , T (P) = X k
t k P (t) dt.
k=1 −1

(a) Write the associated matrix of T relative to the canonical basis of the vector
space R3 [X ].
(b) Determine KerT and ImT .

Solution
Using Sage to solve the problem we achieve:

Hence:
Ker f = {P ∈ R3 [X ] |P = a2 Q 1 + a3 Q 2 }

and
⎭ ⎧ ⎭ ⎧ ⎠
2 2 2 2
Im f = { f (P) |P ∈ R3 [X ]} = a1 + a3 X + a0 + a2 X 2 .
3 3 3 5
134 3 Linear Transformations

10. Let T : R3 ∈ R3 be an endomorphism whose matrix relative to the canonical

basis of R3 is ⎩ 
2 1 0
A =  −1 1 −1  .
1 0 3

(a) Show that T is jordanizable, write its matrix in the Jordan canonical form J
and determine the basis of R3 relative to which J is the matrix of T .
(b) Compute An , n ∈ N∩ .

References

1. V. Postelnicu, S. Coatu, Mică enciclopedie matematică, ed (Tehnică, Bucureşti, 1980)

2. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie analitică, ed (Universitaria, Craiova,
1993)
3. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie n- dimensională, ed (Radical, Craiova,
1996)
4. C. Udrişte, Algebră liniară, geometrie analitică (Geometry Balkan Press, Bucureşti, 2005)
5. V. Balan, Algebră liniară, geometrie analitică, ed (Fair Partners, Bucureşti, 1999)
6. S. Chiriţă, Probleme de matematici superioare, ed (Didactică şi Pedagogică, Bucureşti, 1989)
7. Gh. Atanasiu, Gh. Munteanu, M. Postolache, Algebr ă liniară, geometrie analitică şi diferenţială,
ecua ţii diferenţiale, ed (ALL, Bucureşti, 1998)
8. P. Matei, Algebră liniară. Gometrie analitică şi diferenţială, ed (Agir, Bucureşti, 2002)
9. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică (Universitatea
din Craiova, Note de curs şi aplicaţii, 1993)
Chapter 4
Euclidean Vector Spaces

4.1 Euclidean Vector Spaces

The study of the Euclidean vector space is required to obtain the orthonormal bases,
whereas relative to these bases, the calculations are considerably simplified. In a
Euclidean vector space, scalar product can be used to define the length of vectors
and the angle between them.
Let E be a real vector space.
Definition 4.1 (see [6], p. 101). The mapping <, >: E∈E is called a scalar product
(or an Euclidean structure) on E if the following conditions are satisfied:
(a) < x, y >=< y, x >, (→) x, y ∈E
(b) < x + y, z >=< x, z > + < y, z >, (→) x, y, z ∈E
(c) α < x, y >=< αx, y >, (→) x, y ∈E, (→) α ∈ R
(d) < x, x > ≥ 0, (→) x ∈E; < x, x > = 0 ⇔ x = 0.
The scalar < x, y >∈ R is called the scalar product of vectors x, y ∈E.
Definition 4.2 (see [6], p. 101). A real vector space on which a scalar product
is defined, is called an Euclidean real vector space and it should be denoted by
(E, <, >).
Proposition 4.3 (see [6], p. 101). A scalar product on E has the following properties:
(i) < 0, x >=< x, 0 >= 0, (→) x ∈E
(ii) < x, y + z >=< x, y > + < x, z >, (→) x, y, z ∈E
(iii) < x, α y >= α < x, y >, (→) x, y ∈E, (→) α ∈ R

n 
m 
n 
m
(iv) < α (i) x i , β ( j) y j >= α (i) β ( j) < x i , y j >, (→) x i , y j ∈ E,
i=1 j=1 i=1 j=1
(i) (
(→) α , β ∈ R, i
j) = 1, n, j = 1, m.

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 135

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_4,
© Springer International Publishing Switzerland 2014
136 4 Euclidean Vector Spaces

Examples of Euclidean vector spaces

1. (Rn , <, >) the canonical Euclidean real space. The mapping <, >: Rn ∈ Rn
defined by


n    
< x, y >= x (i) y (i) , (→) x, y ∈ Rn , x = x (1) , x (2) , . . . , x (n) , y = y (1) , y (2) , . . . , y (n)
i=1

is a scalar product on Rn .
 
2. Mn,n (R) , <, > the real Euclidean space of the square matrices, with the scalar
product

 
< A, B >= trace At B , (→) A, B ∈ Mn,n (R) .

3. (V3 , <, >) where the mapping <, >: V3 × V3 ∈ R defined by


∀x∀ ∀y∀ cos ⇒ (x, y), x ∃= 0, y ∃= 0
< x, y >=
0, x = 0 or y = 0

is a scalar product on V3 .
This concret scalar product was the model from which, by abstraction has reached
the concept of the scalar product.
Example 4.4 (see [8]). Let V3 be the vector space of geometric vectors of the usual
physical space. Indicate if the next transformation is a scalar product on V3 :

<, >: V3 × V3 ∈ R, < x, y >= ∀x∀ ∀y∀ , (→) x, y ∈ V3 .

Solution
If the transformation would be a scalar product, then according to the Definition
4.1 it should be satisfied the condition:

< αx, y >= α < x, y >, (→) x, y ∈ V3 , (→) α ∈ R,

i.e.
∀αx∀ ∀y∀ = α ∀x∀ ∀y∀ , (→) x, y ∈ V3 , (→) α ∈ R.

Since the previous relationship can not be held for α < 0 and x, y ∃= 0 it results
that the given transformation is not a scalar product.
This fact can also be checked using Sage:
4.1 Euclidean Vector Spaces 137

Example 4.5 Let B = 1, 1 + X, 1 + X + X 2 be a basis of the space R2 [X ]. We

define the scalar product

< x, y > B = x1 y1 + x2 y2 + x3 y3 ,

where xi , yi , i = 1, 3 are the coordinates of the vectors x and y in the basis B.

Compute the scalar products:
(a) < 1 + X, 1 + X > B
(b) < 3 − X 2 , 2 + 4X + 6X 2 > B .
Solution
(a) Noting that
 
1 + X = 0 · 1 + 1 · (1 + X ) + 0 · 1 + X + X 2

it results ⎫ ⎭
0
(1 + X ) B = ⎬ 1 ⎧ ,
0

i.e. the coordinates of 1 + X in the basis B are: 0, 1, 0; therefore

< 1 + X, 1 + X > B = 0 · 1 + 1 · 1 + 0 · 0 = 1.

(b) Noting that

 
3 − X 2 = 3 + 1 + X − 1 − X − X 2 = 3 · 1 + 1 · (1 + X ) + (−1) · 1 + X + X 2

and
138 4 Euclidean Vector Spaces

2 + 4X + 6X 2 = −2 − 2 − 2X + 6 + 6X + 6X 2
 
= (−2) · 1 + (−2) · (1 + X ) + 6 · 1 + X + X 2

it results ⎫ ⎭
  3
3 − X2 = ⎬ 1 ⎧ ,
B
−1
⎫ ⎭
  −2
2 + 4X + 6X 2 = ⎬ −2 ⎧ ,
B
6

i.e. the coordinates of

• 3 − X 2 in the basis B are: 3, 1, −1;
• 2 + 4X + 6X 2 in the basis B are: −2, −2, 6;
hence

< 3 − X 2 , 2 + 4X + 6X 2 > B = 3 · (−2) + 1 · (−2) + (−1) · 6 = −14.

We shall give a solution in Sage, too:

Definition 4.6 (see [7], p. 155). Let (E, <, >) be a real Euclidean vector space. It’s
called a norm on E, a mapping ∀∀ :E∈ R+ which satisfies the properties:
(i) ∀x∀ > 0, (→) x ∈E and ∀x∀ = 0 ⇔ x = 0 (positivity)
(ii) ∀αx∀ = |α| ∀x∀ , (→) x ∈E, (→) α ∈ R (homogeneity)
(iii) ∀x + y∀ ⇐ ∀x∀ + ∀y∀ , (→) x, y ∈E (Minkowski’s inequality or the triangle
inequality).
4.1 Euclidean Vector Spaces 139

Theorem 4.7 (see [7], p. 154). Let (E, <, >) be a real Euclidean vector space. The
function ∀∀ :E∈ R+ , defined by
⎨
∀x∀ = < x, x >, (→) x ∈ E (4.1)

is a norm on E. The norm defined in the Theorem 4.7 is called the Euclidean norm.
Remark 4.8 (see [7], p. 154). If x ∈ V3 , then its norm (length), in the sense of
Theorem 4.7 coincides with the geometric meaning of its length.
Proposition 4.9 (see [7], p. 154). Let (E, <, >) be a real Euclidean vector space.
(a) For all x, y ∈E, the Cauchy- Schwartz- Buniakowski inequality occurs

< x, y >⇐ ∀x∀ ∀y∀ ; (4.2)

(b) For all x, y ∈E, the parallelogram identity occurs

 
∀x + y∀2 + ∀x − y∀2 = 2 ∀x∀2 + ∀y∀2 . (4.3)

Definition 4.10 (see [7], p. 155). Let (E, <, >) be a real Euclidean vector space.
It’s called the angle of the non-zero vectors x, y ∈E, the unique number ϕ ∈ [0, π ]
for which
< x, y >
cos ϕ = . (4.4)
∀x∀ ∀y∀

Definition 4.11 (see [7], p. 156). Let (E, <, >) be a real Euclidean vector space.
We shall say that the vectors x, y ∈E are orthogonal and we shall denote x⇔y if
< x, y >= 0.
Remark 4.12 (see [7], p. 156). The null vector is orthogonal on any vector x ∈E.
Example 4.13 (see [8]). In the vector space R2 one considers B = {e1 , e2 } , ∀e1 ∀ = 2,
∀e2 ∀ = 4, ⇒ (e1 , e2 ) = π3 . The vectors a = 2e1 − 3e2 , b = −e1 + e2 are given.
 
Compute ⇒ a, b .
Solution
We have:

< e1 , e1 >= ∀e1 ∀2 = 4,

< e2 , e2 >= ∀e2 ∀2 = 16.

From the relation (4.4) we deduce

< x, y >= ∀x∀ ∀y∀ cos ϕ;

hence, we obtain:
1
< e1 , e2 >= ∀e1 ∀ ∀e2 ∀ cos ⇒ (e1 , e2 ) = 2 · 4 · = 4.
2
140 4 Euclidean Vector Spaces

It results that
< a, a >=< 2e1 − 3e2 , 2e1 − 3e2 >= 4 < e1 , e1 > −12 < e1 , e2 > +9 < e2 , e2 >= 112
< a, b >=< 2e1 − 3e2 , −e1 + e2 >= −2 < e1 , e1 > +5 < e1 , e2 > −3 < e2 , e2 >= −36
< b, b >=< −e1 + e2 , −e1 + e2 >=< e1 , e1 > −2 < e1 , e2 > + < e2 , e2 >= 12.

Using (4.1), we have:

⎨ ≺
∀a∀ = < a, a > = 4 7
⎩ ⎩  ≺
⎩b⎩ = < b, b > = 2 3.

Therefore,
  (4.4) < a, b > 36
cos ⇒ a, b = ⎩ ⎩ =− ≺ ≺ .
⎩
∀a∀ b ⎩ 4 7·2 3

We shall check this result in Sage, too:

Proposition 4.14 (see [7], p. 156). Let (E, <, >) be a real Euclidean vector space and
a 1 , a 2 , . . . , a p ∈E be some nonnull vectors which are pairwise orthogonal. Then,
the vectors a 1 , a 2 , . . . , a p are linearly independent.
Definition 4.15 (see [7], p. 156). If x is a nonnull vector of the real Euclidean vector
space (E, <, >) then the vector
x
x0 =
∀x∀

is called the versor of x.

Remark 4.16 (see [7], p. 156). The length of the vector x 0 is equal to 1.
Definition 4.17 (see [7], p. 156). The dimesion of the real Euclidean vector space
(E, <, >) constitutes the dimension of the associated vector space E.
4.1 Euclidean Vector Spaces 141

Definition 4.18 (see [7], p. 156). Let (E, <, >) be a real Euclidean vector space and
S √ E.
(a) The system S is orthogonal if its vectors are nonnull and pairwise orthogonal.
(b) The system S is orthonormal (or orthonormat) if it is orthogonal and each of
its vectors has the length equal to 1.
(c) If dim E = n < ≡ then the basis B = {e1 , e2 , . . . , en } of E is called orthonor-
mal if < ei , e j >= δi j , (→) i, j = 1, n,
where

1, i = j
δi j = , i, j = 1, n
0, i ∃= j

is called the symbol of Kronecher.

The canonical basis Rn is an orthonormal basis for canonical Euclidean real vector
space (Rn , <, >).
Definition 4.19 (see [7], p. 157). Let (E, <, >) be a n finite dimensional real Euclid-
ean vector space and B = {e1 , e2 , . . . , en } is a basis of E. Let be x, y ∈ E; it results

  n

 x= x (i) ei
i=1
 n

y = y ( j) e j .
j=1

We have:


n 
n 
n 
n
(i) ( j)
< x, y >=< x ei , y e j >= x (i) y ( j) < ei , e j > . (4.5)
i=1 j=1 i=1 j=1

We denote

< ei , e j >= gi j , (→) i, j = 1, n;

we note that gi j = g ji , (→) i, j = 1, n, i.e.

⎫ ⎭
g11 g12 · · · g1n
 g12 g22 · · · g2n 
G=⎬  ;
··· ··· ··· ··· ⎧
g1n g2n · · · gnn

the matrix G signifies the matrix of the scalar product <, > relative to the
basis B.
Remark 4.20 (see [7], p. 157). The
matrix

G is symmetric (G t = G ) and positive
definite ( x Gx > 0, (→) x ∈ R \ 0 ).
t n

Definition 4.21 (see [7], p. 157). From (4.5) we obtain

142 4 Euclidean Vector Spaces


n 
n
< x, y >= x (i) y ( j) gi j , (4.6)
i=1 j=1

relation that constitutes the analytical expression of the scalar product <, >
relative to the basis B.
Definition 4.22 (see [7], p. 157). The equality (4.6) is equivalent to the equality

< x, y >= x tB G y B , (4.7)

called the matrix representation of the scalar product <, > relative to the
basis B.
Example 4.23. One considers the mapping <, >: R3 × R3 ∈ R which, relative to
the canonical basis B = {e1 , e2 , e3 } of R3 has the analytical expression:

< x, y >= (x1 − 2x2 ) (y1 − 2y2 ) + x2 y2 + (x2 + x3 ) (y2 + y3 ) ,

(→) x = (x1 , x2 , x3 ) , y = (y1 , y2 , y3 ) ∈ R3 .

 
(a) Show that R3 , <, > is a real Euclidean vector space.
(b) Prove that the vectors a = e1 − e2 + 2e3 , b = −e1 + e2 + 9e3 are orthogonal.
(c) Compute ∀x∀, where x = e1 − e2 + 2e3 .
(d) Compute the angle between the vectors x = e1 − e2 + 2e3 and y = e2 + 2e3 .
(e) Write the matrix of the scalar product <, > relative to the canonical basis of R3 .
Solution
 
(a) R3 , <, > is a real Euclidean vector space if the mapping <, > is a scalar
product. We will check that the conditions of the Definition 4.1 of a scalar product
are satisfied.
We note that

< x, y >= (x1 − 2x2 ) (y1 − 2y2 ) + x2 y2 + (x2 + x3 ) (y2 + y3 ) =< y, x >, (→) x, y ∈ R3
< x + z, y >= (x1 − 2x2 ) (y1 − 2y2 ) + (z 1 − 2z 2 ) (y1 − 2y2 ) + x2 y2 + z 2 y2 +
(x2 + x3 ) (y2 + y3 ) + (z 2 + z 3 ) (y2 + y3 ) =< x, y > + < z, y >, (→) x, y, z ∈ R3
< αx, y >= α [(x1 − 2x2 ) (y1 − 2y2 ) + x2 y2 + (x2 + x3 ) (y2 + y3 )]
= α < x, y >, (→) x ∈ R3 , (→) α ∈ R

< x, x >= (x1 − 2x2 )2 + x22 + (x2 + x3 )2 > 0, (→) x ∈ R3 \ 0


 x1 − 2x2 = 0

< x, x >= 0 ⇔ x2 = 0 ⇔ x1 = x2 = x3 = 0,

 x2 + x3

i.e. the mapping is a scalar product.

We shall prove that in Sage, too:
4.1 Euclidean Vector Spaces 143

(b) As
• a = e1 − e2 + 2e3 ≤ a = (1, −1, 2) ,
• b = −e1 + e2 + 9e3 ≤ b = (−1, 1, 9)
we shall obtain < a, b >= 0; hence a and b are orthogonal.
(c) As

x = e1 − e2 + 2e3 = (1, 0, 0) − (0, 1, 0) + 2 · (0, 0, 1) = (1, −1, 2)

≺
it results that < x, x >= 6 and using (4.1): ∀x∀ = 6.
≺ ≺
(d) Taking into account that: < x, y >= −4, ∀y∀ = 5, ∀x∀ = 6, we can
deduce:
(4.4) 4
cos ⇒ (x, y) = − ≺ ≺ .
6· 5

(e) We have
⎫ ⎭
1 −2 0
 
G = ⎬ −2 6 1 ⎧ ,
0 1 1

where gi j =< ei , e j >, (→) i, j = 1, 3.

Solving in Sage the points (b)–(e) of this problem, we achieve:
144 4 Euclidean Vector Spaces

If B ∗ = {e1 , e2 , . . . , en } is an orthonormal basis of E, then the matrix of the scalar

product relative to this basis is the unit matrix In . In this case, from (4.6) we achieve:


n 
n
< x, y >= x (i) y ( j) (4.8)
i=1 j=1

and from (4.7) we deduce

< x, y >= x tB ∗ y B ∗ . (4.9)

The equalities (4.8) and (4.9) justifies the importance of considering the ortho-
normal basis which consists in the fact that relative to these bases, the computations
are more simplified.
Definition 4.24 (see [7], p. 159). The matrix A ∈ Mn (R) is orthogonal if At A =
A At = In , In being unit matrix of order n.
Theorem 4.25 (theorem of change the orthonormal bases, see [7], p. 160).
Let (E, <, >) be a real Euclidean vector space of finite dimension n and B1 =
{e1 , e2 , . . . , en }, B2 = {u 1 , u 2 , . . . , u n } are two orthonormal bases of E. Then the
transition matrix from the basis B1 to the basis B2 is orthogonal.
Example 4.26 (see [6], p. 112). In the real Euclidean vector space E one considers
the bases B1 = {e1 , e2 , e3 }, B2 = {u 1 , u 2 , u 3 } . If x is an arbitrary vector from E,
x (i) , y (i) , i = 1, 3 are the coordinates of the vector relative to the bases B1 , B2 , B1
is an orthonormal basis and
 (1)

 x = 27 y (1) + 37 y (2) + 67 y (3)

x (2) = 67 y (1) + 27 y (2) + αy (3)


 (3)
x = −αy (1) − 67 y (2) + 27 y (3)

determine α ∈ R such that B2 be an orthonormal basis, too.

4.1 Euclidean Vector Spaces 145

Solution
In order that B2 be an orthonormal basis it is necessary (within the Theorem 4.25)
that the transition matrix from the basis B1 to the basis B2 , i.e.
⎫ 2 3 6⎭
7 7 7
 
M(B1 ,B2 ) = 
⎬
6
7
2
7 α
⎧
−α − 67 2
7

to be an orthogonal one.
From the condition

M(B1 ,B2 ) · Mt(B1 ,B2 ) = I3

it results that α = − 37 .
The solution is Sage is:

Theorem 4.27 (Gram-Schmidt orthogonalization, see [1], p. 26). If (E, <, >) is
a real Euclidean vector space of finite dimension n and B = {a 1 , a 2 , . . . , a n } is a
basis of E then there is a basis B ∗ = {e1 , e2 , . . . , en } of E which has the following
properties:
(i) the basis B ∗ is orthonormal;
(ii) the sets {a 1 , . . . , a k } and {e1 , . . . , ek } generate the same vector subspace Wk √
E, for each k = 1, n.
The Gram- Schmidt orthogonalization procedure (described in detail in [2],
p. 150) can be summarized as:

(1) build an orthogonal set B1 = b1 , b2 , . . . , bn which satisfies the property ii) of

the Theorem 4.27, where:



 b1 = a 1

 (1)

 b 2 = α2 b 1 + a 2



 (1) (2)
b 3 = α3 b 1 + α3 b 2 + a 3



 .
.
. (4.10)

 (1) (i−1)

 bi = αi b1 + . . . + αi bi−1 + a i



 .

 .

 .

 (1) (n−1)
b n = αn b 1 + . . . + αn bn−1 + a n ,

( j)
where the scalars αi ∈ R, i = 2, n, j = 1, i − 1 are determined from the condition
that bi ⇔ b j , i, j = 1, n, i ∃= j.
146 4 Euclidean Vector Spaces

(2) determine the orthonormal basis B ∗ = {e1 , e2 , . . . , en } of E, such that:

bi
ei = ⎩ ⎩ , i = 1, n. (4.11)
⎩b i ⎩

Example 4.28 (see [8]). In the space R2 [X ] we define

1
< P, Q >= P (t) Q (t) dt.
−1

Orthonormate the canonical basis of the space R2 [X ] (i.e. B = 1, X, X 2 ), with

respect to this scalar product.
Solution

Stage I. We build the orthogonal basis B ∗ = f 1 , f 2 , f 3 , with


 f1 = 1
f2 = X +αf1

f 3 = X 2 + α1 f 1 + α2 f 2 .

From the orthogonality condition of f 1 and f 2 we deduce:

< f 1 , f 2 >= 0 ⇔< f 1 , X + α f 1 >= 0 ⇔

< f 1, X >
< f 1 , X > +α < f 1 , f 1 >= 0 ⇔ α = − .
< f 1, f 1 >

As
1
< f 1 , X >=< 1, X >= 1 · tdt = 0
−1

it results that α = 0 and f 2 = X .

From the orthogonality condition f 1 and f 3 we deduce:

< f 1 , f 3 >= 0 ⇔ < f 1 , X 2 + α1 f 1 + α2 f 2 >= 0 ⇔

< f 1, X 2 >
< f 1 , X 2 > +α1 < f 1 , f 1 > +α2 < f 1 , f 2 >= 0 ⇔ α1 = − .
⎪ ⎛⎜ ⎝ < f 1, f 1 >
=0

We obtain
1
2
< f 1 , X >=< 1, X >=
2 2
1 · t 2 dt =
3
−1
4.1 Euclidean Vector Spaces 147

and

1
< f 1 , f 1 >=< 1, 1 >= 1 · 1 dt = 2;
−1

hence α1 = −1/3.
From the orthogonality condition f 2 and f 3 we deduce:

< f 2 , f 3 >= 0 ⇔< f 2 , X 2 + α1 f 1 + α2 f 2 >= 0 ⇔

< f 2, X 2 >
< f 2 , X 2 > +α1 f 2 , f 1 > + α2 < f 2 , f 2 >= 0 ⇔ α2 = − .
⎪ ⎛⎜ ⎝ < f 2, f 2 >
=0

As
1
< f 2 , X 2 >=< X, X 2 >= t · t 2 dt = 0
−1

it results α2 = 0. Therefore
1
f 3 = X2 − .
3
∗∗

Stage II. We build the orthonormal basis B = g 1 , g 2 , g 3 :

(4.11) f
g i = ⎩ i ⎩ , (→) i = 1, 3.
⎩ f i⎩

Whereas
⎩ ⎩  ≺
⎩ f 1⎩ = < f 1, f 1 > = 2

it results that

f 1
g1 = ⎩ 1 ⎩ = ≺ .
⎩ f 1⎩ 2

We compute
1
2
< f 2 , f 2 >=< X, X >= t · t dt = ;
3
−1
148 4 Euclidean Vector Spaces

we shall have ≺
f2 3
g 2 = ⎩ ⎩ = ≺ X.
⎩ f 2⎩ 2

We achieve that:
1 ⎞ ⎟
1 1 1 2 8
< f 3 , f 3 >=< X − , X 2 − >=
2
t −
2
dt =
3 3 3 45
−1

and
≺ ⎞ ⎟
f 3 5 1
g3 = ⎩ 3 ⎩ = ≺ X2 − .
⎩ f 3⎩ 2 2 3

We can also solve this problem in Sage:

Example 4.29 We consider the real vector space of the symmetric matrices, of the
order n, with real elements, Mns (R) and
 
<, >: Mns (R) × Mns (R) ∈ R, < A, B >= trace At B .
4.1 Euclidean Vector Spaces 149

Orthonormate the system of matrices:

 ⎞ ⎟ ⎞ ⎟ ⎞ ⎟⎠
11 01 −1 0
A1 = , A2 = , A3 = .
10 12 0 1

Solution  
We consider the orthogonal system {B1 , B2 , B3 }, trace Bit B j = 0, (→) i ∃= j, as
follows:

 B1 = A1
B2 = A2 + α B1

B3 = A3 + α1 B1 + α2 B2 .

From the condition < B2 , B1 >= 0, we have

0 =< A2 + α B1 , B1 >=< A2 , B1 > +α < B1 , B1 >,

i.e.

0 = trace (A2 B1 ) + αtrace (B1 B1 ) ;

therefore
trace (A2 B1 ) 2
α=− =− .
trace (B1 B1 ) 3

We obtain
⎞ ⎟ ⎞ ⎟ ⎞ ⎟
01 2 11 −2/3 1/3
B2 = − = .
12 3 10 1/3 2

The condition < B3 , B1 >= 0 involves

0 = < A3 + α1 B1 + α2 B2 , B1 >=< A3 , B1 > +α1 < B1 , B1 > +α2 < B2 , B1 >

trace (A3 B1 ) 1
= trace (A3 B1 ) + α1 trace (B1 B1 ) ≤ α1 = − = .
trace (B1 B1 ) 3

The condition < B3 , B2 >= 0 involves

0 = < A3 + α1 B1 + α2 B2 , B2 >=< A3 , B2 > +α1 < B1 , B2 > +α2 < B2 , B2 >

trace (A3 B2 ) 4
= trace (A3 B2 ) + α2 trace (B2 B2 ) ≤ α2 = − =− .
trace (B2 B2 ) 7
150 4 Euclidean Vector Spaces

We shall obtain
1 4
B3 = A3 + B1 − B2 .
3 7
The orthonotormal system {C1 , C2 , C3 } will be:
Bi
Ci = , i = 1, 3.
∀Bi ∀

We shall also determine these matrices in Sage:

4.2 Linear Operators in Euclidean Vector Spaces

In investigating the Euclidean vector spaces are very useful the linear transformations
compatible with the scalar product, i.e. the orthogonal transformations.
Definition 4.30 (see [7], p. 199). Let (E, <, >) be a finite dimensional real Euclid-
ean space. The endomorphism T ∈ End(V ) is called orthogonal operator or
orthogonal transformation if T transforms the orthonormal basis into some ortho-
normal basis, i.e. if B = {e1 , e2 , . . . , en } is an orthonormal basis of E then
B ∗ = {T (e1 ) , T (e2 ) , . . . , T (en )} is an orthonormal basis of E, too.
Theorem 4.31 (see [7], p. 199). For an operator T ∈ End(V ) the following statements
are equivalent:
4.2 Linear Operators in Euclidean Vector Spaces 151

1. T is orthogonal,
2. T is bijective and T −1 is orthogonal,
3. T preserves the scalar product, i.e., < T (x) , T (y) >=< x, y >, (→) x, y ∈ E,
4. T stores the length of vectors, i.e., ∀T (x)∀ = ∀x∀, (→) x ∈ E,
5. the operator matrix T relative to an orthonormal basis of E is orthogonal.
Corollary 4.32 (see [7], p. 200). If T ∈ End(V ) is orthogonal then T preserves the
vector angles, i.e.

< x, y > < T (x) , T (y) >

cos ⇒ (x, y) = = = cos ⇒ (T (x) , T (y)) . (4.12)
∀x∀ ∀y∀ ∀T (x)∀ ∀T (y)∀

Proposition 4.33 (see [7], p. 200). Let T, S ∈ End(V ) be two orthogonal operators
and be α ∈ R. Then:
1. T ∞ S is an orthogonal operator,
2. αT is orthogonal⇔ α = ±1.
We denote by
O (E) = {T ∈ End (V ) | T orthogonal} .

Proposition 4.34 (see [7], p. 200). If T ∈ O (E) and A is the associated matrix of T
relative to an orthonormal basis B of E then det A = ±1.
Definition 4.35 (see [7], p. 200). It’s called an orthogonal operator of the first
kind or the rotation operator, an orthogonal operator for which the determinant of
the associated matrix in an orthonormal basis of E is equal to −1.
Definition 4.36 (see [7], p. 200). It’s called an orthogonal operator of the second
kind, an orthogonal operator for which the determinant of the associated matrix in
an orthonormal basis of E is equal to −1.
We denote by:
• O+ (E) = the set of the orthogonal operators of the first kind,
• O− (E) = the set of the orthogonal operators of the second kind.
Proposition 4.37 (see [7], p. 200). The roots of the characteristic equation of an
orthogonal operator have their absolute values equal to 1. In particular, the eigenval-
ues of an orthogonal operator are equal to ±1.
Proposition 4.38 (see [7], p. 202). For an orthogonal operator, the eigenvectors that
correspond to different eigenvalues are orthogonal.
Theorem 4.39 (see [3], p. 95). The orthogonal matrices of M2 (R) are of the form:
⎞ ⎟ ⎞ ⎟
cos ϕ sin ϕ − cos ϕ sin ϕ
, , (4.13)
sin ϕ − cos ϕ − sin ϕ − cos ϕ
⎞ ⎟ ⎞ ⎟ ⎞ ⎟
− cos ϕ sin ϕ − cos ϕ − sin ϕ cos ϕ − sin ϕ
, , ,
sin ϕ cos ϕ − sin ϕ cos ϕ sin ϕ cos ϕ
⎞ ⎟ ⎞ ⎟
cos ϕ sin ϕ − cos ϕ − sin ϕ
, ,
− sin ϕ cos ϕ sin ϕ − cos ϕ
152 4 Euclidean Vector Spaces

(→) ϕ ∈ [0, 2π ].
Definition 4.40 (see [7], p. 203 and [3], p. 95). An orthogonal matrix with det A = 1
is called a rotation matrix in Rn .

4.2.1 Orthogonal Transformations in the Euclidean Plane

Theorem 4.41 (see [8]). The orthogonal transformations in the Euclidean plane are:
the rotations, the reflections or the compositions of rotations with reflections.
Proposition 4.42 (see [8]). The rotation of the plane vectors around the origin, in
the counterclockwise, with the angle ϕ, rϕ : R2 ∈ R2 ,
   
rϕ (x) = rϕ x (1) , x (2) = x (1) cos ϕ − x (2) sin ϕ, x (1) sin ϕ + x (2) cos ϕ
(4.14)

is an orthogonal transformation.
Remark 4.43 (see [8]). If O is the center of rotation then each point M has associated
the point M ∗ , such that (see Fig. 4.1):
⎦ ⎩ ⎩ ⎩ ⎩
⎩O M ⎩ = ⎩ ⎩
⎩O M ∗ ⎩ = a
⇒ M O M ∗ = the rotation angle in the counterclockwise.

Fig. 4.1 Rotation through

angle ϕ

We have:
T (x) = rϕ (x) = Ax,

where
⎞ ⎟
cos ϕ − sin ϕ
A=
sin ϕ cos ϕ

determines a rotation through the angle ϕ in the plane.

4.2 Linear Operators in Euclidean Vector Spaces 153

Proposition 4.44 (see [8]). The rotation through the angle π around the origin,
 
T (x) = sO (x) = −x (1) , −x (2) (4.15)

coincides with the reflection with respect to the origin (see Fig. 4.2).
Fig. 4.2 Rotation through
angle π around the origin

Proposition 4.45 (see [8]). The reflection across the Ox axis (see Fig. 4.3),
 
T (x) = sd (x) = x (1) , −x (2) (4.16)

is an orthogonal transformation.
Fig. 4.3 Reflection across the
axis Ox

Proposition 4.46 (see [8]). The reflection across the Oy axis (see Fig. 4.4),
 
T (x) = sd∗ (x) = −x (1) , x (2) (4.17)

is an orthogonal transformation.

Fig. 4.4 Reflection across the

axis Oy
154 4 Euclidean Vector Spaces

Proposition 4.47 (see [8]). The composition of the rotation rϕ with the reflection sd
is an orthogonal transformation.
Proof
We shall have
  (4.15)
 
T (x) = rϕ ∞ sd (x) = rϕ (sd (x)) = rϕ x (1) , −x (2)
(4.13)
 
= x (1) cos ϕ + x (2) sin ϕ, x (1) sin ϕ − x (2) cos ϕ .

We can note that

⎞ ⎟ ⎞ ⎟⎞ ⎟
cos ϕ sin ϕ cos ϕ − sin ϕ 1 0
A= =
sin ϕ − cos ϕ sin ϕ cos ϕ 0 −1

is an orthogonal transformation.
Proposition 4.48 (see [8]). The composition of the rotation rϕ with the reflection sO
is an orthogonal transformation.
Proof
We shall achieve
  (4.14)
 
T (x) = rϕ ∞ sO (x) = rϕ (sO (x)) = rϕ −x (1) , −x (2)
(4.13)
 
= −x (1) cos ϕ + x (2) sin ϕ, − x (1) sin ϕ − x (2) cos ϕ .

We can note ⎞ ⎟
− cos ϕ sin ϕ
A=
− sin ϕ − cos ϕ

is an orthogonal transformation, A being a matrix from (4.12).

Example 4.49 (see [5], p. 51). The coordinate axes Ox and Oy one rotates with the
angle ϕ = π3 and one considers the new system is oriented opposite to the original
≺ ≺ 
system. Knowing that a point A has the coordinates 3, −2 3 in the new system,
find its coordinates in the old coordinate system.
Solution
Case 1. We have a rotation, followed by a reflection across the Oy ∗ axis (Fig. 4.5).

Fig. 4.5 Rotation, followed

by a reflection across the Oy ∗
axis
4.2 Linear Operators in Euclidean Vector Spaces 155

We deduce
      (4.16)
T (x) = x ∗ , y ∗ = rϕ ∞ sd∗ (x) = rϕ sd∗ (x) = rϕ (−x, y)
(4.13)
= (−x cos ϕ − y sin ϕ, − x sin ϕ + y cos ϕ) .

The transformation T has the equations:


x ∗ = −x cos ϕ − y sin ϕ
y ∗ = −x sin ϕ + y cos ϕ.
≺ ≺
As x ∗ = 3, y ∗ = −2 3, then solving the above system it results
⎦ ≺
x = 3 − ≺23 = 2.134
y = − 23 − 3 = −3.232.

Using Sage, we achieve:

Case 2. We have a rotation, followed by a reflection across the Ox ∗ axis (Fig. 4.6).
We obtain
156 4 Euclidean Vector Spaces

Fig. 4.6 Rotation, followed

by a reflection across the Ox ∗
axis

    (4.15)
T (x) = x ∗ , y ∗ = rϕ ∞ sd (x) = rϕ (sd (x)) = rϕ (x, −y)
(4.13)
= (x cos ϕ + y sin ϕ, x sin ϕ − y cos ϕ) .

The transformation T has the equations:


x ∗ = x cos ϕ + y sin ϕ
y ∗ = x sin ϕ − y cos ϕ.
≺ ≺
As x ∗ = 3, y ∗ = −2 3, then solving the above system it results
⎦ ≺
x= ≺ 2 − 3 = −2.134
3

y = 3 + 23 = 3.232.

We can obtain these coordinates with Sage:

4.2 Linear Operators in Euclidean Vector Spaces 157

Proposition 4.50 (see [4], p. 463). The rotation of a rectangular coordinate system
around the origin, in the counterclockwise, through the angle ϕ, rϕ : R2 ∈ R2 is an
orthogonal transformation.
Proof
By rotating the rectangular coordinate system xOy around the origin, in the coun-
terclockwise, through the angle ϕ one gets the system x ∗ Oy ∗ . A point
 M which has
the coordinates (x, y) in the old system will have the coordinates x ∗ , y ∗ in the new
system.
We choose in the plane an orthonormal reference with the origin in the center of
rotation (Fig. 4.7).

Fig. 4.7 Rotation of a coor-

donate system through the
angle ϕ

We note that: 

 OC1 = x ∗ cos ϕ

AC1 = y ∗ sin ϕ

 OC2 = x ∗ sin ϕ

C2 B = y ∗ cos ϕ

and
158 4 Euclidean Vector Spaces

O A = OC1 − AC1 = x ∗ cos ϕ − y ∗ sin ϕ
O B = OC2 + C2 B = x ∗ sin ϕ + y ∗ cos ϕ.

It turns out that the equations corresponding to the transformation of the coordinate
system xOy by rotating it in the counterclockwise, through the angle ϕ will be:

x = x ∗ cos ϕ − y ∗ sin ϕ
y = x ∗ sin ϕ + y ∗ cos ϕ

i.e. 
x ∗ = x cos ϕ + y sin ϕ
(4.18)
y ∗ = −x sin ϕ + y cos ϕ.

We obtain:
 
Rϕ (x) = Rϕ (x, y) = x ∗ cos ϕ − y ∗ sin ϕ, x ∗ sin ϕ + y ∗ cos ϕ ,

T (x) = Rϕ (x) = Ax,

where ⎞ ⎟
cos ϕ − sin ϕ
A= ;
sin ϕ cos ϕ

Rϕ is an orthogonal transformation as A is a matrix from (4.12) and At · A =I2 .

Example 4.51. One gives the point M (1, 1) in the plane reported to the rectangular
axes Ox, Oy. Determine that angle with which the axes should be rotated so that the
point M belongs to the Ox ∗ axis. Find the new coordinates of M in these conditions.
Solution
Using (4.17) we achieve:
 ∗
x = x cos θ + y sin θ
y ∗ = −x sin θ + y cos θ,

where:
• x ∗ , y ∗ are the coordinates of the point M in the plane reported to the rectangular
axes Ox ∗ , Oy ∗ ;
• x, y, are the coordinates of the point M in the plane reported to the rectangular
axes Ox, Oy;
• θ is the angle to be rotated the axes.
Multiplying the first equation with cos ϕ and the second with − sin ϕ and adding
the obtained equations, we deduce:

x ∗ cos θ − y ∗ sin θ = x,

while multiplying the first equation with sin θ and the second with cos θ and adding
the obtained equations, we deduce:
4.2 Linear Operators in Euclidean Vector Spaces 159

x ∗ sin θ + y ∗ cos θ = y.

By emphasizing the condition that the point M belongs to the Ox ∗ axis (i.e. y ∗ = 0)
we have
 ∗  ∗
x cos θ = x x cos θ = 1 ≺
⇔ ⇔ x ∗2 = 2 ⊂≤ x ∗ = ± 2.
x ∗ sin θ = y x ∗ sin θ = 1
≺
In the case when x ∗ = 2 it results
 ≺
 cos θ = ≺1 = 2
2 ≺2
 sin θ = ≺1 = 2
2 2

i.e. θ = π/4. ≺
In the case when x ∗ = − 2 it results
 ≺
 cos θ = − ≺1 = − 2
2 2 ≺
 sin θ = ≺1 = 2
2 2

i.e.
π 5π
θ =π+ = .
4 4
So, the new coordinates of M if:
≺ 
π
• the axes one rotate with the angle θ = 4 are M 2, 0 ,
 ≺ 
• axes one rotate with the angle θ = 5π
4 are M − 2, 0 .

We shall given a solution in Sage, too:

160 4 Euclidean Vector Spaces

4.3 Problems

1. Show that the function <, >: R2 × R2 ∈ R2 , defined by

< x, y >= 5x1 y1 − 2x1 y2 − 2x2 y1 + 3x2 y2 , (→) x = (x1 , x2 ) ,

y = (y1 , y2 ) ∈ R2

is a scalar product.
2. In the real Euclidean vector space (E, <, >) having a basis B = {e1 , e2 } such
that: ∀e1 ∀ = 1, ∀e2 ∀ = 4, ⇒ (e1 , e2 ) = π/4 one considers the vector x =
e1 + 5e2 . Compute ∀x∀.
3. In the real Euclidean vector space R3 one assumes the vectors: a 1 = (1, 0, 3) ,
a 2 = (1, 1, 0) , a 3 = (1, 1, 1).
(a) Determine if B1 = {a 1 , a 2 , a 3 } constitutes a basis of R3 .
(b) Map the system of vectors B1 into one orthonormal.
Solution
Using Sage we shall have:

,
4.3 Problems 161

4. Let R3 be the arithmetic vector space and the scalar products <, >1 , <, >2 :
R3 × R3 ∈ R3 ,defined by:

<x, y >1 = x1 y1 + 2x1 y2 + 2x2 y1 + 5x2 y2 + x3 y3 ,

<x, y >2 = 2x1 y1 + 3x1 y2 + 3x2 y1 + 7x2 y2 + 2x3 y3 ,

(→) x = (x1 , x2 , x3 ) , y = (y1 , y2 , y3 ) ∈ R3 .

Compute
 3 the
 length of the vector
 x = (1, 2, 3) in the Euclidean vector spaces
R , <, >1 and R3 , <, >2 .

Solution
Using Sage, we shall have:

5. In the real vector space M2 (R) one assumes the matrices:

⎞ ⎟ ⎞ ⎟ ⎞ ⎟ ⎞ ⎟
12 23 3 1 4 2
A1 = , A2 = , A3 = , A4 =
11 10 1 −2 −1 −6

(a) Check if B1 = {A1 , A2 , A3 , A4 } determine a basis of M2 (R).

(b) Transform the basis B1 into an orthonormal one. 
 
6. Let be the vectors v 1 = 21 , 21 , 21 , 21 and v 2 = 16 , 16 , 21 , − 56 in the Euclidean
space R4 . Check that these vectors have their norm equal to 1 and they are
orthogonal. Then, bulid an ortonormate basis of this space, that contains the
vectors v 1 and v 2 .

Solution
With Sage, it will result:
162 4 Euclidean Vector Spaces

7. Prove that the transformation T : R3 ∈ R3 , where

⎞ ⎟
2 2 1 2 1 2 1 2 2
T (x) = x1 + x2 − x3 , x1 − x2 + x3 , − x1 + x2 + x3
3 3 3 3 3 3 3 3 3

(→) x = (x1 , x2 , x3 ) ∈ R3 is orthogonal in R3 with the usually scalar product.

Solution
We shall use Sage:

8. One gives the point M (1, 1) in the plane reported to the rectangular axes
Ox, Oy. Determine the angle that the axes should be rotated so that the
point M belongs to the Oy ∗ axis. Find the new coordinates of M in these
conditions.
9. Let be a triangle, having the vertices A (3, 1) , B (7, 1) , C (7, 4) . Find its image
through the rotation with the center O and the angle π3 .

Solution
Solving this problem in Sage, we achieve:
4.3 Problems 163

10. One considers the rotation through the angle ϕ in the counterclockwise, rϕ :
R2 ∈ R2 .
(a) Justify the linearity of the transformation rϕ .
(b) Build the associated matrix in the canonical basis from R2 and in the basis
B1 = {e1 + e2 , e1 − e2 }.
164 4 Euclidean Vector Spaces

(c) What is the relationship between the two matrices?

(d) Compute the kernel and the image of this linear mapping.
(e) Is rϕ injective? But is it surjective?

References

1. V. Balan, Algebră liniară, geometrie analitică, ed (Fair Partners, Bucureşti, 1999)

2. I. Iatan, Advances Lectures on Linear Algebra with Applications (Lambert Academic Publishing,
2011)
3. P. Matei, Algebră liniară. Gometrie analitică şi diferenţială, ed. (Agir, Bucureşti, 2002)
4. V. Postelnicu, S. Coatu, Mică enciclopedie matematică, ed (Tehnică, Bucureşti, 1980)
5. C. Udrişte, Aplicaţii de algebră, geometrie şi ecuaţii diferenţiale, ed (Didactică şi Pedagogică
R.A, Bucureşti, 1993)
6. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie analitică, ed. (Universitaria, Craiova,
1993)
7. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie n- dimensională, ed.
(Radical, Craiova, 1996)
8. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică, Note de curs şi
aplicaţii, (Universitatea din Craiova, 1993)
Chapter 5
Bilinear and Quadratic Forms

5.1 Bilinear and Quadratic Forms

The theory of bilinear form and quadratic form is used [5] in the analytic geometry
for getting the classification of the conics and of the quadrics.
It is also used in physics, in particular to describe physical systems subject to small
vibrations. The coefficients of a bilinear form one behave to certain transformations
like the tensors coordinates. Tensors are useful in theory of elasticity (the deformation
of an elastic medium is described through the deformation tensor).
Definition 5.1 (see [1], p. 150). A mapping b : V × V ∈ K is called a bilinear
form on V if it satisfies the conditions:
1. b (πx + λ y, z) = πb (x, z) + λb (y, z) , (→) π, λ ∈ K , (→) x, y, z ∈ V,
2. b (x, π y + λz) = πb (x, y) + λb (x, z) , (→) π, λ ∈ K , (→) x, y, z ∈ V.
Definition 5.2 (see [1], p. 150). We say that the bilinear form b : V × V ∈ K is sym-
metric (antisymmetric) if b (x, y) = b (y, x) (respectively, b (x, y) = −b (y, x).
Consequences 5.3 (see [2], p. 116). If the mapping b : V × V ∈ K is a bilinear
form then:
   
(1) b 0, x = b x, 0 =  0, (→) x∈V
 n n
(2) (a) b π (i) x i , y = π (i) b (x i , y) ,
i=1 i=1

(1) , . . . , π (n) ∈ K , (→) x , . . . , x , y ∈ V

(→) π  1 n
 n  n
 
(b) b x, (i)
λ yi = b x, y i ,
i=1 i=1
(→) λ (1) , . . . , λ (n) ∈ K , (→) x, y 1 , . . . , y n ∈ V.

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 165

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_5,
© Springer International Publishing Switzerland 2014
166 5 Bilinear and Quadratic Forms

Definition 5.4 (see [1], p. 150). If b : V × V ∈ K is a symmetric bilinear form,

the mapping f : V ∈ K , defined by f (x) = b (x, x), for all x ∈ V is called the
quadratic form associated to b.
Remark 5.5 (see [1], p. 150). Knowing the quadratic form f, allows to get the
symmetric bilinear form, associated to f as:

1
b (x, y) = [ f (x + y) − f (x) − f (y)] , (→) x, y ∈ V (5.1)
2
Definition 5.6 (see [1], p. 150). The symmetric bilinear form b associated to the
quadratic form f is called the polar form of the quadratic form f .
Example 5.7 (see [3], p. 93). The quadratic form corresponding to the real scalar
product (which is a symmetric bilinear form) is the square of the Euclidean norm:

f (x) =< x, x >= x2 , (→) x ∈ V.

Let V be an n finite dimensional vector space over K , n ≥ 1 and B = {a 1 , . . . , a n }

one of its basis. If b : V × V ∈ K is a bilinear form then (→) x, y ∈ V it results:


n 
n
x= x (i) a i , y = y ( j) a j ;
i=1 j=1

therefore

n 
n
b (x, y) = ai j x (i) y ( j) , (5.2)
i=1 j=1

where  
ai j = b a i , a j , (→) i, j = 1, n.

The expression (5.2) constitutes [1] the analytic   expression of the bilinear form b
relative to the basis B, and A ∈Mn (K ), A = ai j 1∀i, j∀n represents the associated
matrix of the bilinear form b relative to the basis B.
From (5.2) one obtains [1] the analytic expression of the bilinear form f : V ∈ K
relative to the basis B of V :
n  n 
n
f (x) = ai j x (i) y ( j) , (→) x = x (i) a i ∈ V. (5.3)
i=1 j=1 i=1

Definition 5.8 (see [1], p. 151) We call the associated matrix of a quadratic
form f : V ∈ K relative to a basis of V , the matrix of the bilinear mapping
b : V × V ∈ K from which derives f relative to the considered basis.
Example 5.9 (see [4]) Let be b : R4 × R4 ∈ R,

b (x, y) = x1 y1 + 2x2 y1 + 2x2 y2 − 4x2 y3 + x3 y3 + x4 y1 − x4 y4 .

5.1 Bilinear and Quadratic Forms 167

(a) Prove that b is a bilinear functional.

(b) Find the associated matrix of b relative to the basis
 
B ⇒ = f 1 = (1, 1, 0, 0) , f 2 = (0, 1, 0, 0) , f 3 = (0, 1, 0, 1) , f 4 = (1, 0, 0, 1)

and relative to the canonical basis and highlight the link between them.
(c) Determine the expression of the quadratic form f associated to b.
Solution
(a) According to the Definition 5.1, b is a bilinear functional if those two conditions
are accomplished. We shall check the first condition as for the others one proceeds
similarly. Let be π, λ ∈ K and x, y, z ∈ V ; we have:

b (πx + λ y, z) = (πx1 + λy1 ) z 1 + 2 (πx2 + λy2 ) z 1 + 2 (πx2 + λy2 ) z 2

− 4 (πx2 + λy2 ) z 3 + (πx3 + λy3 ) z 3
+ (πx4 + λy4 ) z 1 − (πx4 + λy4 ) z 4
= πx1 z 1 + λy1 z 1 + 2πx2 z 1 + 2λy2 z 1 + 2πx2 z 2 + 2λy2 z 2
− 4πx2 z 3 − 4λy2 z 3 + πx3 z 3 + λy3 z 3 + πx4 z 1 + λy4 z 1
− πx4 z 4 − λy4 z 4
= π (x1 z 1 + 2x2 z 1 + 2x2 z 2 − 4x2 z 3 + x3 z 3 + x4 z 1 − x4 z 4 )
+ λ (y1 z 1 + 2y2 z 1 + 2y2 z 2 − 4y2 z 3 + y3 z 3 + y4 z 1 − y4 z 4 )
= πb (x, z) + λb (y, z) .

(b) To determine the associated matrix of b relative to the basis B ⇒ we have to

compute:   

 b  f 1, f 1 = 5

⎫ b f 1, f 2 = 2
 
 b f 1, f 3 = 2

 ..
⎬
.

We achieve: ⎭ ⎩
5 2 2 3
⎧ 4 2 2 2
A⇒ = ⎧
⎨5
.
2 1 2
2 0 −1 2

The associated matrix of b relative to the canonical basis is

⎭ ⎩
10 0 0
⎧ 2 2 −4 0 
A=⎧ ⎨0 0 1 0 .


1 0 0 −1
168 5 Bilinear and Quadratic Forms

We can note that: ⎭ ⎩

1 0 0 1
⎧1 1 1 0
M(B,B ⇒ ) ⎧
=⎨ .
0 0 0 0
0 0 1 1

We have:
b (x, y) = x tB Ay B

and
A⇒ = Mt(B,B ⇒ ) AM(B,B ⇒ ) .

(c) The expression of the quadratic form f associated to b is

f (x) = x12 + 2x2 x1 + 2x22 − 4x2 x3 + x32 + x4 x1 − x42 .

The Solution in Sage will be presented, too:

5.1 Bilinear and Quadratic Forms 169

Definition 5.10 (see [1], p. 152) The rank of the quadratic form f is the rank of its
matrix relative to a basis of V and one denotes with rank f .
Remark 5.11 (see [2], p. 122) Because of the symmetry of the associated matrix of
a quadratic form, relative to a basis B of V , the relation (5.3) is written


n  2 
n
f (x) = aii x (i) + 2 ai j x (i) x ( j) . (5.4)
i=1 i, j=1
i< j

Definition 5.12 (see [2], p. 122) If the associated matrix of the quadratic form f :
V ∈ K relative to the basis B = {e1 , e2 , . . . , en } of V is diagonal, i.e. A =
diag (π1 , . . . , πn ) ; we shall say that:
• the basis B is a canonical basis for f,
• the analytical expression of f relative to the basis B, i.e.


n  2 
n
f (x) = πi x (i) , (→) x = x (i) ei ∈ V (5.5)
i=1 i=1

is a canonical expression for f .

We shall present three methods for obtaining a canonical expression for a quadratic
form.

5.2 Gauss-Lagrange Method for Reducing a Quadratic

Form to a Canonical Expression

Theorem 5.13 (Gauss-Lagrange, see [1], p. 152) Let V be an n finite dimensional

vector space over K and f : V ∈ K a quadratic form. Then there is a basis
B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n of V , relative to which f has a canonical expression.
Proof
We present the proof this theorem because it provides a method for getting a
canonical expression for a quadratic form.
Let B = {e1 , e2 , . . . , en } be a basis of V relative to which f has the analytical
expression:
 n  n 
n
f (x) = ai j x (i) y ( j) , (→) x = x (i) ei .
i=1 j=1 i=1

If f is the quadratic null form, then f has the canonical expression in any basis
of V . Hence, we can assume that isn’t null.
We can also assume that (∃) i = 1, n such that aii ⇐= 0. Otherwise, if ar p ⇐= 0,
for r ⇐= p then we make the change of coordinates:
170 5 Bilinear and Quadratic Forms

⎫ x (r ) = t (r ) + t ( p)
x ( p) = t (r ) − t ( p) (5.6)
⎬
x (i) (i)
= t , i ∈ {1, . . . , n} \ {r, p}

and we get an analytical expression, having non null its coefficients.

We suppose that a11 ⇐= 0. By grouping the terms that contain the variable x (1) ,
from (5.4) we obtain:

 2 
n 
n
f (x) = a11 x (1) + 2 a1k x (1) x (k) + ai j x (i) x ( j) . (5.7)
k=2 i, j⇐=1

We shall add and subtract the necessary terms in (5.7) to write it in the form:

1  2 
n
f (x) = (1) (2)
a11 x + a12 x + . . . + a1n x (n)
+ ai⇒ j x (i) x ( j) , (5.8)
a11
i, j=2


n
where ai⇒ j x (i) x ( j) doesn’t contain x (1) .
i, j=2
We make the change of coordinates:
 (1)

 z = a11 x (1) + a12 x (2) + . . . + a1n x (n)

⎫ z (2) = x (2)
..

 .

⎬
z (n) = x (n)

hence  (1) 1 (1)

x =
 a11 z − aa12 z (2) − . . . − a1n (n)
a11 z

⎫
11
x = z (2)
(2)

 ..

 .
⎬
x (n) = z (n) .
 
The transition to the new coordinates z (1) , z (2) , . . . , z (n) is achieved through the
relation:
x B = M(B,B1 ) · x B1 , (5.9)

with the transition matrix

⎭ ⎩
1
a11 − aa12
11
· · · − aa1n
11
⎧ 0 1 ··· 0 
⎧ 
M(B,B1 ) = ⎧ .. .. . . ..  .
⎨ . . . . 
0 0 0 1
5.2 Gauss-Lagrange Method for Reducing a Quadratic Form to a Canonical Expression 171
 
The new coordinates correspond to the new basis B1 = f 1 , f 2 , . . . , f n , where:


 f 1 = a111 e1


⎫ f 2 = − a12 e1 + e2
a11
 ..

 .

⎬ f = − a1n e + e .
n a11 1 n

The form Q has the following analytical expression relative to the basis B1 :

1  (1) 2 
n
f (x) = z + ai⇒ j x (i) x ( j) . (5.10)
a11
i, j=2

The sum

n
Q1 = ai⇒ j x (i) x ( j)
i, j=2

from the right member of the relation (5.10) is a quadratic form in n − 1 variables,
therefore can be treated by the process described above, as well
 as the form Q. 
Finally, after at most n − 1 steps we obtain a basis B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n of V ,
relative to which the quadratic form Q is reduced to the canonical expression.
Example 5.14 (see [4]). Let be the quadratic form
 2  2
Q : R4 ∈ R, Q (x) = x (1) + 2x (1) x (2) + 2 x (2) − 4x (2) x (3)
 2  2
+ x (3) + x (1) x (4) − x (4) .

Using the Gauss-Lagrange method, we can bring Q to the canonical expression

and highlight the transition matrix from the initial basis to that basis, relative to which
Q has the canonical expression.
Solution
The associated matrix of Q relative to the canonical basis of the space R4 , i.e.

B = {e1 = (1, 0, 0, 0) , e2 = (0, 1, 0, 0) , e3 = (0, 0, 1, 0) , e4 = (0, 0, 0, 1)}

is ⎭ ⎩
1 1 0 1/2
⎧ 1 2 −2 0 
A=⎧
⎨ 0
.
−2 1 0 
1/2 0 0 −1

We note that a11 ⇐= 0. We can write Q in the form

172 5 Bilinear and Quadratic Forms

 2  2
x (4)
Q (x) = x (1) + x (2) + + x (2) − 4x (2) x (3)
2
5  (4) 2  2
− x − x (2) x (4) + x (3) .
4
By making the change of coordinates:

x (4)

 y (1) = x (1) + x (2) +
⎫ (2) 2
y = x (2)

 y (3) = x (3)
⎬ (4)
y = x (4)

it results  y (4)

 x (1) = y (1) − y (2) −
⎫ (2) 2
x = y (2)
 (3) = y (3)

⎬ x (4)
x = y (4) .

The transition matrix associated with this change of coordinates will be:
⎭ ⎩
1 −1 0 −1/2
⎧0 1 0 0 
M(B,B1 ) =⎧
⎨0
,
0 1 0 
0 0 0 1

the new basis being

 
1
B1 = f 1 = e1 , f 2 = − e1 + e2 , f 3 = e3 , f 4 = − e1 + e4 .
2

The expression of the quadratic form Q relative to the basis B1 is

 2  2 5  (4) 2  2
Q = y (1) + y (2) − 4y (2) y (3) − y − y (2) y (4) + y (3) .
4
The associated matrix of Q relative to the basis B1 is
⎭ ⎩
1 0 0 0
⎧ 0 1 −2 −1/2 
A⇒ = ⎧
⎨ 0 −2
.
1 0 
0 −1/2 0 −5/4

⇒ ⇐ = 0, we can write Q in the form:

Noting that a22
5.2 Gauss-Lagrange Method for Reducing a Quadratic Form to a Canonical Expression 173

 2  2  2 3  2
(1) (2) (3) 1
Q= y + y − 2y − y (4) − 3 y (3) − y (4) − 2y (3) y (4) .
2 2

We make the change of coordinates:

 (1)

 z = y (1)
⎫ (2)
z = y (2) − 2y (3) − 21 y (4)

 z (3) = y (3)
⎬ (4)
z = y (4) ;

hence  (1)

 y = z (1)
⎫ (2)
y = z (2) + 2z (3) + 21 z (4)

 y (3) = z (3)
⎬ (4)
y = z (4) .
 
The transition to the new coordinates z (1) , z (2) , . . . , z (n) is achieved through the
relation
x B1 = M(B1 ,B2 ) · x B2 ,

with the transition matrix

⎭ ⎩
1 0 0 0
⎧0 1 2 1/2 
M(B1 ,B2 ) =⎧
⎨0
.
0 1 0 
0 0 0 1

The new coordinates correspond to the new basis

 
1
B2 = g 1 = f 1 , g 2 = f 2 , g 3 = 2 f 2 + f 3 , g 4 = f 2 + f 4 .
2

The expression of the quadratic form Q relative to the basis B2 is

 2  2  2 3  2
Q = z (1) + z (2) − 3 z (3) − z (4) − 2z (2) z (4) .
2
The associated matrix of Q relative to the basis B2 is
⎭ ⎩
1 0 0 0
⎧0 1 0 0 
A =⎧
⇒⇒
⎨0
.
0 −3 −1 
0 0 −1 −3/2

⇒⇒ ⇐ = 0. We will form a perfect square in Q for those terms that

We note that a33
(3)
contain z ; is follows
174 5 Bilinear and Quadratic Forms

 2  2 1   7  2
Q = z (1) + z (2) − 3z (3) + z (4) − z (4) .
3 6
We shall make the change of coordinates:
 (1)

 t = z (1)


⎫ t (2) = z (2)
 t (3) = 3z (3) + z (4)



⎬ (4)
t = z (4) ;

we have ⎭ ⎩
1 0 0 0
⎧0 1 0 0 
M(B2 ,B3 ) =⎧
⎨0
.
0 1/3 −1 
0 0 0 1

The expression of the quadratic form Q relative to the basis

 
1
B3 = h 1 = g 1 , h 2 = g 2 , h 3 = g 3 , h 4 = −g 3 + g 4
3

is  2  2 1  2 7  2
Q = t (1) + t (2) − t (3) − t (4) ,
3 6
so we have obtained the canonical expression of Q.
We shall get:

x B = M(B,B1 ) x B1 = M(B,B1 ) M(B1 ,B2 ) x B2 = M(B,B1 ) M(B1 ,B2 ) M(B2 ,B3 ) x B3 .

It follows that transition matrix from the initial basis B of the space R4 to the
basis B3 , relative to which Q has the canonical expression is:

M(B,B3 ) = M(B,B1 ) M(B1 ,B2 ) M(B2 ,B3 ) .

A solution in Sage will be given, too:

5.2 Gauss-Lagrange Method for Reducing a Quadratic Form to a Canonical Expression 175

5.3 Reducing a Quadratic Form to a Canonical Expression

by Jacobi Method

Theorem 5.15 (Jacobi, see [3], p. 100). Let V be ann finite dimensional vector space
over K , f : V ∈ K , a quadratic form and A = ai j 1∀i, j∀n its relative matrix to
the basis B = {e1 , e2 , . . . , en } of V .
If all the principal minors


 ϕ1 = a⎪ 11 ⎪

 ⎪ ⎪

⎫ ϕ2 = ⎪ a11 a12 ⎪
⎪ a21 a22 ⎪
(5.11)

 ..

 .

⎬
ϕn = det A
 
are all non- null, then there is a basis B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n of V , relative to which
the quadratic form Q has the canonical expression
 ϕi−1  2
n
f (x) = y (i) , (5.12)
ϕi
i=1

where
• y (i) , i = 1, n are the coordinates of x in the basis B ⇒ ,
• ϕ0 = 1.
Proof
We are looking for the vectors e⇒1 , e⇒2 , . . . , e⇒n by the form
 ⇒

 e1 = c11 e1

 e⇒2 = c21 e1 + c22 e2



 ..
⎫
.
⇒ (5.13)

 e = ci1 e 1 + ci2 e2 + . . . + cii ei


i
..



 .
⎬ ⇒
en = cn1 e1 + cn2 e2 + . . . + cnn en ,

where ci j , i, j = 1, n will be determined by imposing the conditions:

176 5 Bilinear and Quadratic Forms

  0, 1 ∀ j < i ∀ n
b ei⇒ , e j = (5.14)
1, i = j

and b : V × V ∈ K is the bilinear form from which f derives.

We compute
   
b ei⇒ , e j = b ci1 e1 + ci2 e2 + . . . + cii ei , e j (5.15)
     
= ci1 b e1 , e j + ci2 b e2 , e j + . . . + cii b ei , e j
= ci1 a1 j + ci2 a2 j + . . . + cii ai j .

We obtain:

  

 j = 1 : b ei⇒ , e1  = ci1 a11 + ci2 a12 + . . . + cii a1i = 0



⎫ j = 2 : b ei⇒ , e2 = ci1 a21 + ci2 a22 + . . . + cii a2i = 0
.. (5.16)
   .

 ⇒

 j = i − 1 : b e , e
i  i−1  = ci1 i−1,1 + ci2 ai−1,2 + . . . + cii ai−1,i = 0
a
⎬
j = i : b ei⇒ , ei = ci1 ai1 + ci2 ai2 + . . . + cii aii = 1

i.e. a compatible system that is determined, since its determinant is ϕi ⇐= 0 (hence

the vector ei⇒ is uniquely determined).
Using the formulas of Crammer we get solutions of the system (5.16):
⎪ ⎪
⎪ a11 · · · a1,i−1 0 ⎪⎪
⎪
⎪ .. .. .. .. ⎪
⎪ . . . . ⎪⎪
⎪
⎪ ai−1,1 · · · ai−1,i−1 0 ⎪⎪
⎪
⎪ ai1 · · · ai,i−1 1⎪ ϕi−1
cii = = , (→) i = 1, n. (5.17)
ϕi ϕi

To  determine the  expression of the quadratic form in the basis

B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n we shall calculate the elements of the matrix A⇒ , associated
of f relative to the basis B ⇒ .
We have
   
ai⇒ j = b ei⇒ , e⇒j = b ei⇒ , c j1 e1 + . . . + c j j e j
     
= c j1 b ei⇒ , e1 + c j2 b ei⇒ , e2 + . . . + c j j b ei⇒ , e j , (→) i, j = 1, n.
 
But, from (5.14) we know that b ei⇒ , e⇒j = 0 for j < i; hence ai⇒ j = 0 for j < i.
Because of the symmetry of the bilinear form b it results ai⇒ j = 0 for j > i. Therefore
ai⇒ j = 0 for j ⇐= i. For j = i we have
5.3 Reducing a Quadratic Form to a Canonical Expression by Jacobi Method 177
   
aii⇒ = b ei⇒ , ei⇒ = b ei⇒ , ci1 e1 + . . . + cii ei
       
= ci1 b ei⇒ , e1 + ci2 b ei⇒ , e2 + . . . + ci,i−1 b ei⇒ , ei−1 + cii b ei⇒ , ei
ϕi−1
= cii = , (→) i, j = 1, n.
ϕi

We deduce that the quadratic form has the following canonical expression in the
basis B ⇒ :
 ϕi−1  (i) 2
n
f (x) = ai⇒ j y (i) y ( j) = y
ϕi
i, j=1

and its associated matrix is

⎭ϕ ⎩
0
ϕ1 O
⎧ .. 
A⇒ = ⎧
⎨ . .

ϕn−1
O ϕn

Example 5.16 (see [3], p. 101). Using the Jacobi method find the canonical expression
and the basis in which to do this for the quadratic form

Q : R3 ∈ R, Q (x) = x12 + 7x22 + x32 − 8x1 x2 − 8x2 x3 − 16x1 x3 ,

× (→) x = (x1 , x2 , x3 ) ∈ R3 . (5.18)

Solution
The matrix of the quadratic form relative to the canonical basis of the space R3 is
⎭ ⎩
1 −4 −8
A = ⎨ −4 7 −4  .
−8 −4 1

Its principal minors ϕi , i = 0, 3 are:



 ϕ0 = 1


⎫ 1 = a⎪ 11 = 1 ⎪
ϕ
⎪ 1 −4 ⎪

 ϕ2 = ⎪⎪ ⎪ = −9

 −4 7 ⎪
⎬
ϕ3 = det A = −729.

The quadratic form Q will have the following canonical expression:

178 5 Bilinear and Quadratic Forms


n
ϕi−1 ϕ 0 2 ϕ 1 2 ϕ2 2
Q (x) = yi2 = y + y + y
ϕi ϕ 1 1 ϕ2 2 ϕ 3 3
i=1
1 2 1 2
= y12 −
y2 + y .
9 81 3
 
We shall determine the new basis B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n , relative to which Q has
the canonical expression:
 ⇒
⎫ e1 = c11 e1
e⇒ = c21 e1 + c22 e2
⎬ 2⇒
e3 = c31 e1 + c32 e2 + c33 e3 ,

where ci j , i, j = 1, 3 will be determined by imposing the conditions (5.14), b being

the associated bilinear form Q of the quadratic form in the basis B ⇒ , i.e.

1 1
b (x, y) = x1 y1 − x2 y2 + x3 y3 .
9 81
We have:

  
b ei⇒ , e1 = b (c11 e1 , e1) = c11
 b (e1 , e1 ) = c11 a11 = c11
⇔ c11 = 1;
b e⇒1 , e1 = 1

therefore: e⇒1 = e1 .
We shall compute:
 
b e⇒2 , e1 = b (c21 e1 + c22 e2 , e1 ) = c21 b (e1 , e1 ) + c22 b (e2 , e1 )
= c21 a11 + c22 a21 = c21 − 4c22
 ⇒ 
b e2 , e2 = b (c21 e1 + c22 e2 , e2 ) = c21 b (e1 , e2 ) + c22 b (e2 , e2 )
= c21 a12 + c22 a22 = −4c21 + 7c22 .

Taking into account (5.14) we obtain the system:

⎛
c21 − 4c22 = 0 4 1
⇔ c21 = − , c22 = − ,
−4c21 + 7c22 =1 9 9

i.e.
4 1
e⇒2 = − e1 − e2 .
9 9
5.3 Reducing a Quadratic Form to a Canonical Expression by Jacobi Method 179

We have also to calculate:

 
b e⇒3 , e1 = b (c31 e1 + c32 e2 + c33 e3 , e1 ) = c31 b (e1 , e1 )
+ c32 b (e2 , e1 ) + c33 b (e3 , e1 )
= c31 a11 + c32 a12 + c33 a13 = c31 − 4c32 − 8c33
 ⇒ 
b e3 , e2 = b (c31 e1 + c32 e2 + c33 e3 , e2 ) = c31 b (e1 , e2 )
+ c32 b (e2 , e2 ) + c33 b (e3 , e2 )
= c31 a21 + c32 a22 + c33 a23 = −4c31 + 7c32 − 4c33
 ⇒ 
b e3 , e3 = b (c31 e1 + c32 e2 + c33 e3 , e3 ) = c31 b (e1 , e3 )
+ c32 b (e2 , e3 ) + c33 b (e3 , e3 )
= c31 a31 + c32 a32 + c33 a33 = −8c31 − 4c32 + c33 .

Taking into account (5.14) we obtain the system:


⎫ c31 − 4c32 − 8c33 = 0

8 4 1
−4c31 + 7c32 − 4c33 = 0 ⇔ c31 = − , c32 = − , c33 = ;

⎬ 81 81 81
−8c31 − 4c32 + c33 = 1

it results:
8 4 1
e⇒3 = − e1 − e2 + e3 .
81 81 81
The solution in Sage will be given, too:
180 5 Bilinear and Quadratic Forms

5.4 Eigenvalue Method for Reducing a Quadratic Form

into Canonical Expression

Theorem 5.17 (Eigenvalue method, see [1], p. 153). Let V be an Euclidean real
vector space and let f : V ∈  R be a real quadratic form. Then there is an orthonormal
basis B ⇒ = e⇒1 , e⇒2 , . . . , e⇒n of the vector space V relative to which the canonical
expression of the form is
 n  2
f (x) = ρi y (i) , (5.19)
i=1

where:
• ρ1 , . . . , ρn are the eigenvalues of the associated matrix of the quadratic form,
relative to an orthonormal basis B (each eigenvalue being included in sum such
many times as its multiplicity),
• y (1) , . . . , y (n) are the coordinates of the vector x relative to the basis B ⇒ .
To apply the eigenvalue method for reducing a quadratic form to canonical expres-
sion one determines as follows:
1. choose an orthonormal basis B = {e1 , e2 , . . . , en } of V and write the matrix A,
associated to f relative to the basis B;
2. determine the eigenvalues: ρ1 , . . . , ρr ∈ R of the matrix A, with the correspond-
ing algebraic multiplicities aρ1 , . . . , aρr , with aρ1 + . . . + aρr = n ;
3. for the eigensubspaces Wρ1 , . . . , Wρr associated to the eigenvalues ρ1 , . . . , ρr
determine the orthonormal bases B1 , . . . , Br , using the Gram-Schmidt orthogo-
nalization procedure;
4. one considers the orthonormal basis: B ⇒ = B1 ≺ . . . ≺ Br of V and one writes
the canonical expression of f relative to the basis B ⇒ with (5.19), where x B ⇒ =
 (1) t
y , . . . , y (n) .
Example 5.18 (see [4]). Use the eigenvalue method to determine the canonical
expression and the basis relative to which can be made this, for the quadratic form:
5.4 Eigenvalue Method for Reducing a Quadratic Form into Canonical Expression 181

f : R3 ∈ R, f (x) = x12 +x22 +x32 +x1 x2 +x2 x3 +x1 x3 , (→) x = (x1 , x2 , x3 ) ∈ R3 .

Solution
The associated matrix of f relative to the canonical basis of the space R3 is
⎭ ⎩
1 1/2 1/2
A = ⎨ 1/2 1 1/2  .
1/2 1/2 1

We have:
⎪ ⎪
⎪ 1 − ρ 1/2 1/2 ⎪  2
⎪ ⎪ 1
P (ρ) = ⎪⎪ 1/2 1 − ρ 1/2 ⎪⎪ = (2 − ρ) −ρ ,
⎪ 1/2 1/2 1 − ρ ⎪ 2

which has the roots 

ρ1 = 2, aρ1 = 1
ρ2 = 1/2, aρ2 = 2.

The associated eigenspace of the eigenvalue ρ1 is

⎜ ⎝
Wρ1 = x ∈ R3 | Ax = ρ1 x .

We deduce:
 
⎫ x1 + 21 x2 + 21 x3 = 2x1 ⎫ −x1 + 21 x2 + 21 x3 = 0
1
x + x + x = 2x2 √
1 1
x − x + 1x = 0
⎬ 21 1 12 2 3 ⎬ 12 1 1 2 2 3
2 x 1 + 2 x 2 + x 3 = 2x 3 2 x 1 + 2 x 2 − x 3 = 0.

Denoting x3 = t, t ∈ R we achieve:

−2x1 + x2 = −t
⇔ x1 = t, x2 = t.
x1 − 2x2 = −t

Therefore
 

⎫ 

Wρ1 = x ∈ R3 | x = (t, t, t) = t · (1, 1, 1) = tc1 .

⎬ ⎞ ⎟⎠ ⎦ 

c1

 
The orthonormal basis B1 will be B1 = f 1 , where
 
c1 1 1 1 1
f1 = = ≡ c1 = ≡ ,≡ ,≡ .
c1  3 3 3 3
182 5 Bilinear and Quadratic Forms

The associated eigenspace of the eigenvalue ρ2 is

⎜ ⎝
Wρ2 = x ∈ R3 | Ax = ρ2 x .

We achieve:
 1 1 1

 x1 + x2 + x3 = x1



⎫
2 2 2
1 1 1
x1 + x2 + x3 = x2 √
2
 2 2



⎬ 1x + 1x + x = 1x
1 2 3 3
2 2 2
1 1 1
x1 + x2 + x3 = 0 √ x1 + x2 + x3 = 0.
2 2 2
We denote x1 = t1 , x2 = t2 , t1 , t2 ∈ R; hence x3 = −t1 − t2 .
Therefore:


⎫
Wρ2 = x ∈ R3 | x = (t1 , t2 , −t1 − t2 ) = t1 (1, 0, −1)

⎬ ⎞ ⎟⎠ ⎦
c2



+ t2 (0, 1, −1) = t1 c2 + t2 c3 . (5.20)
⎞ ⎟⎠ ⎦ 

c3

⎜ ⇒ ⇒
⎝
We consider the orthogonal system f 2 , f 3 , where
⎛ ⇒
f 2 = c2
⇒ ⇒
f 3 = c3 + π f 2 ,

⇒ ⇒
where π is obtained from the condition that f 3 and f 2 to be orthogonal, i.e.
⇒ ⇒
< f 3 , f 2 >= 0.

From
⇒
⇒ ⇒ < c3 , f 2 > 1
< c3 + π f 2 , f 2 >= 0 ⇔ π = − ⇒ ⇒ =− .
< f 2, f2 > 2

It results
 
⇒ 1 ⇒ 1 1 1
f 3 = c3 − f 2 = (0, 1, −1) − (1, 0, −1) = − , 1, − .
2 2 2 2
5.4 Eigenvalue Method for Reducing a Quadratic Form into Canonical Expression 183
 
The basis B = f 2 , f 3 is orthonormal, where

⇒  
f2 1 ⇒ 1 1
$ $
f 2 = ⇒ = ≡ f 2 = ≡ , 0, − ≡ ,
$ $ 2 2 2
$ f 2$
⇒ ≡  ≡ 
f3 2 ⇒ 1 2 1
f 3 =$ ⇒ $ = ≡ f 3 = −≡ , ≡ , −≡ .
$ $ 3 6 3 6
$ f 3$

We achieve:
B ⇒ = B1 ≺ B2 = { f 1 , f 2 , f 3 }.

The associated matrix of f in the basis B ⇒ will be

⎭ ⎩
2 0 0
A⇒ = ⎨ 0 1/2 0 
0 0 1/2

and the canonical expression of f relative to the basis B ⇒ :

1 2 1 2
f (x) = 2y12 + y + y .
2 2 2 3
We need the following Sage code to implement this method:
184 5 Bilinear and Quadratic Forms

5.5 Characterization Criteria for Positive (Negative)

Definite Matrices

Definition 5.19 (see [2], p. 125). Let V be a real vector space.

(a) The quadratic form f : V ∈ R is called positive definite (negative definite)
if f (x) > 0 ( respectively, f (x) < 0 ), (→) x ∈ V, x ⇐= 0;
(b) The quadratic form f : V ∈ R is called positive semidefinite (negative
semidefinite) if f (x) ≥ 0 ( respectively, f (x) ∀ 0), (→) x ∈ V and (∃) a ∈ V ,
a ⇐= 0, for which f (a) = 0;
(c) The quadratic form  f : V ∈ R is called nondefinite if (∃) a, b ∈ V such that
f (a) > 0 and f b < 0.
Definition 5.20 (see [3], p. 104). A symmetric matrix is positive (negative) definite
if its associated quadratic form is positive (negative) defined.
Proposition 5.21 (see [2], p. 125). Let V be an n finite dimensional real vector
space and A = ai j 1∀i, j∀n , A ∈Mn (R) the associated symmetric matrix of the
positive definite quadratic form f : V ∈ R, relative to the basis B = {e1 , e2 , . . . , en }
of V . Then, the following statements take place:
(a) aii > 0, (→) i = 1, n
(b) det A > 0
(c) rank f = n.
Theorem 5.22 (Sylvester’s criterion, the law of inertia, see [1], p. 154). Let V be
an n finite dimensional real vector space and f : V ∈ R a quadratic form. Then the
number of positive and respective negative coefficients from a canonical expression
of f doesn’t depend on the choice of the canonical basis.
Definition 5.23 (see [3], p. 104).
(i) The number p of the positive coefficients from a canonical expression of the
quadratic form f is called the positive index of f .
5.5 Characterization Criteria for Positive (Negative) Definite Matrices 185

(ii) The number q of the negative coefficients from a canonical expression of the
quadratic form f is called the negative index of f .
(iii) The pair ( p, q, d) is called the signature of the quadratic form, where d =
n − ( p + q) is the number of the null coefficients.
The following theorem allows us to decide if a quadratic form is positive or negative
definite, without being obliged to determine one of its canonical expression.
Theorem 5.24 (Sylvester’s criterion, inertia theorem,  see [3], p. 104). Let V be
an n finite dimensional real vector space and A = ai j 1∀i, j∀n , A ∈Mn (R) be a
symmetric matrix associated of the quadratic form f : V ∈ R relative to the basis
B = {e1 , e2 , . . . , en } of V . Then
1. f is positive definite if and only if all the principal minors ϕ1 , ϕ2 , . . . , ϕn of
the matrix A are strictly positive,
2. f is negative definite if and only if (−1)k ϕk > 0, (→) k = 1, n.
Remark 5.25 (see [3], p. 104).
(i) The quadratic form f is positive (negative) definite if and only if rank f =
n = p (respectively rank f = n = q).
(ii) The law of inertia states that following any of the three methods to obtain the
canonical expression of a quadratic form, the signature of the quadratic form
(inferred from obtained canonical the expression) is always the same.
(iii) Given a quadratic form f : V ∈ R, its associated matrix relative to a basis of
the space V , f is positive definite if and only if any of the following conditions
are satisfied:

• the quadratic form f has the signature (n, 0, 0)

• the determinants ϕi > 0, (→) i = 1, n
• the eigenvalues of the matrix A are strictly positive.

Example 5.26 (see [1], p. 156). Let f : R4 ∈ R be a quadratic form whose analytical
expression form relative to the canonical basis of R4 is

f (x) = x1 x2 − x2 x3 + x3 x4 + x4 x1 , (→) x = (x1 , x2 , x3 , x4 ) ∈ R4 .

(a) Write the matrix of f relative to the canonical basis of R4 and the analytical
expression of the polar of f , relative to the same basis.
(b) Use the Gauss method to determine a canonical expression for f and a basis of
R4 , relative to which f has this canonical expression.
(c) Indicate the signature of f .

Solution
(a) The matrix of the quadratic form relative to the canonical basis of R4 is
186 5 Bilinear and Quadratic Forms

y1 y2 y3 y4
⎭ ⎩
x1 0 1/2 0 1/2
x ⎧ 1/2 0 −1/2 0 
A = 2⎧ .
x3 ⎨ 0 −1/2 0 1/2 
x4 1/2 0 1/2 0

Remark 5.27 (see [4]). In the writing of the matrix A occurs both x1 , x2 , x3 , x4 and
y1 , y2 , y3 , y4 to obtain the analytical expression of the polar of f : multiply each
element of the matrix A with the index cooresponding to the line denoted by xi
respectively of the column, denoted by y j to intersection which is this element.
The analytical expression of the polar of f relative to the canonical basis of R4
will be
1 1 1 1 1 1
b (x, y) = x1 y2 + x1 y4 + x2 y1 − x2 y3 − x3 y2 + x3 y4
2 2 2 2 2 2
1 1
+ x4 y1 + x4 y3 , (→) x, y ∈ R4 . (5.21)
2 2
(b) As a12 ⇐= 0 we make the change of coordinates:
 

 x1 = y1 + y2 
 y1 = 1
2 x1 + 21 x2
⎫ ⎫
x2 = y1 − y2 y2 = 1
2 x1 − 21 x2
⇔
 x3
 = y3  y3
 = x3
⎬ ⎬
x4 = y4 y4 = x4 .

The transition associated matrix to this change of coordinates will be:

⎭ ⎩
1 1 0 0
⎧1 −1 0 0
M(B,B1 ) =⎧
⎨0
,
0 1 0
0 0 0 1

the new basis being

 
B1 = f 1 = e1 + e2 , f 2 = e1 − e2 , f 3 = e3 , f 4 = e4 .

The expression of the quadratic form f relative to the basis B1 is

f (x) = (y1 + y2 ) (y1 − y2 ) − (y1 − y2 ) y3 + y3 y4 + y4 (y1 + y2 ) ,

i.e.
f (x) = y12 − y22 − y1 y3 + y2 y3 + y3 y4 + y1 y4 + y2 y4 .
5.5 Characterization Criteria for Positive (Negative) Definite Matrices 187

The associated matrix of f relative to the basis B1 is

⎭ ⎩
1 0 −1/2 1/2
⎧ 0 −1 1/2 1/2 
A⇒ = ⎧
⎨ −1/2
.
1/2 0 1/2 
1/2 1/2 1/2 0

⇒ ⇐ = 0. We can write f in the form

We note that a11
 2
1 1 1 2 1 2 3
f (x) = y1 − y3 + y4 − y3 − y4 − y22 + y3 y4 + y2 y3 + y2 y4 .
2 2 4 4 2

By making the change of coordinates:



 z = y1 − 21 y3 + 21 y4
⎫ 1
z2 = y2

 z = y3
⎬ 3
z4 = y4 ;

it results 

 y = z 1 + 21 z 3 − 21 z 4
⎫ 1
y2 = z2

 y = z3
⎬ 3
y4 = z4.

The transition matrix associated with this change of coordinates, through the
relation
x B1 = M(B1 ,B2 ) x B2

will be: ⎭ ⎩
1 0 1/2 −1/2
⎧0 1 0 0 
M(B1 ,B2 ) =⎧
⎨0
,
0 1 0 
0 0 0 1

the new basis being

 
1 1 1
B2 = g 1 = f 1 , g 2 = f 2 , g 3 = e1 + f 3 , f 4 = − f 1 + f 4 .
2 2 2

The expression of the quadratic form f relative to the basis B2 is

1 1 3
f (x) = z 12 − z 22 − z 32 − z 42 + z 3 z 4 + z 2 z 3 + z 2 z 4 .
4 4 2
188 5 Bilinear and Quadratic Forms

The associated matrix of f relative to the basis B2 is

⎭ ⎩
1 0 0 0
⎧ 0 −1 1/2 1/2 
A⇒⇒ = ⎧
⎨0
.
1/2 −1/4 3/4 
0 1/2 3/4 −1/4

⇒⇒ ⇐ = 0. We can write f in the form

We note that a22
 2
1 1
f (x) = z 12 − −z 2 + z 3 + z 4 + 2z 3 z 4 .
2 2

By making the change of coordinates:



 t1 = z1
⎫
t2 = −z 2 + 21 z 3 + 21 z 4

 t = z3
⎬ 3
t4 = z4;

it results 

 z1 = t1
⎫
z2 = −t2 + 21 t3 + 21 t4

 z = t3
⎬ 3
z4 = t4 .

The transition matrix associated with this change of coordinates, through the
relation
x B2 = M(B2 ,B3 ) x B3

will be: ⎭ ⎩
1 0 0 0
⎧0 −1 1/2 1/2 
M(B2 ,B3 ) =⎧
⎨0
,
0 1 0 
0 0 0 1

the new basis being

 
1 1
B3 = h 1 = g 1 , h 2 = − g 2 , h 3 = g 2 + g 3 , g 4 = f 2 + f 4 .
2 2

The expression of the quadratic form f relative to the basis B3 is

f (x) = t12 − t22 + 2t3 t4 .

5.5 Characterization Criteria for Positive (Negative) Definite Matrices 189

The associated matrix of f relative to the basis B3 is

⎭ ⎩
1 0 0 0
⎧ 0 −1 0 0
A⇒⇒⇒ = ⎧
⎨0
.
0 0 1
0 0 1 0

⇒⇒⇒ = 0, a ⇒⇒⇒ ⇐ = 0.
We note that a33 34
Making the change of coordinates


 t1 = u1
⎫
t2 = u2

 t3 = u3 + u4
⎬
t4 = u3 − u4;

it results 

 u 1 = t1
⎫u = t
2 2

 u 3 = 2 (t3 + t4 )
1
⎬
u 4 = 21 (t3 − t4 ) .

The associated transition matrix of this change of coordinates, through the relation

x B3 = M(B3 ,B4 ) x B4

will be ⎭ ⎩
1 0 0 0
⎧0 1 0 0 
M(B3 ,B4 ) =⎧
⎨0
,
0 1 1 
0 0 1 −1

the new basis being

 
B4 = v 1 = h 1 , v 2 = h 2 , v 3 = h 3 + h 4 , v 4 = h 3 − h 4 .

The expression of the quadratic form f relative to the basis B4 is

f (x) = u 21 − u 22 + 2u 23 − 2u 24 .

The associated matrix of f relative to the basis B4 is

⎭ ⎩
1 0 0 0
⎧0 −1 0 0 
B=⎧
⎨0
.
0 2 0 
0 0 0 −2
190 5 Bilinear and Quadratic Forms

(c) We have
rank f = rank A = 4, p = 2, q = 2.

We obtain that the quadratic form f has the signature (2, 2, 0).
We can also solve this problem in Sage, too:

5.6 Problems

1. Let f : R3 ∈ R be a quadratic form whose analytical expression form relative

to the canonical basis of R3 is

f (x) = x1 x2 + x2 x3 .

(a) Write the matrix of f relative to the canonical basis of R3 and the analytical
expression corresponding to the polar of f , relative to the same basis.
5.6 Problems 191

(b) Use the Gauss method to determine a canonical expression for f and a basis
of R3 , relative to which f has this canonical expression.
(c) Indicate the signature of f .

Solution
Solving this problem in Sage, we obtain:

2. Determine an orthonormal basis of vector space R3 relative to which the quadratic

form

f (x) = −x12 + x22 − 5x32 + 6x1 x3 + 4x2 x3

has a canonical expression.

3. Let f : R4 ∈ R be a quadratic form whose analytical expression form relative
to the canonical basis of R4 is

f (x) = x12 + 5x22 + 4x32 − x42 + 6x1 x2 − 4x1 x3 − 12x2 x3 − 4x2 x4 − 8x3 x4 .

Use the Gauss method to determine a canonical expression for f and a basis of
R4 , relative to which f has this canonical expression.
192 5 Bilinear and Quadratic Forms

4. Let be the bilinear functional

b ( x, y) = x1 y2 − x2 y1 + x1 y3 − x3 y1 + x1 y4 − x4 y1 + x2 y3 − x3 y2
+ x2 y4 − x4 y2 + x3 y4 − x4 y3 , (→) x, y ∈ R4 (5.22)

(a) Prove that b is a antisymmetic bilinear functional.

(b) Find the matrix corresponding to the bilinear functional b : R4 × R4 ∈ R
relative to the basis
 
B ⇒ = f 1 = (1, 1, 1, 0) , f 2 = (0, 1, 1, 1) , f 3 = (1, 1, 0, 1) , f 4 = (1, 0, 1, 1) .

Solution
Using Sage, we shall have:

5. Let B = {e1 , e2 , e3 } be the canonical basis of the arithmetic vector space R3 and
let b : R3 × R3 ∈ R be the bilinear form for which:

⎫ b (e1 , e1 ) = −1, b (e2 , e2 ) = 3, b (e3 , e3 ) = −6
b (e1 − e2 , e2 ) = 2, b (e2 , e1 + 2e2 ) = 5, b (e3 − e1 , e1 ) = 4
⎬
b (2e1 + e2 , e3 ) = −7, b (e1 + e3 , e2 ) = 4, b (e1 − 2e2 , e3 ) = −1.

(a) Write the matrix corresponding to the bilinear functional b, relative to the
basis B.
(b) Is b a symmetic bilinear functional?
Solution
With Sage, we achieve:
5.6 Problems 193

6. (a) Write the analytical expression of the quadratic form f : R3 ∈ R, defined

by f (x) = b (x, x) , (→) x ∈ R3 , b being the bilinear functional from the
previous problem.
(b) Use the Jacobi method to determine a canonical expression for the quadratic
form f from the previous problem and that basis of R3 , relative towhich f
has this canonical expression.
(c) Prove that f is negative definite.

Solution
We shall solve in Sage this problem:
194 5 Bilinear and Quadratic Forms

7. Let b :M2 (R) ×M2 (R) ∈ R be a bilinear functional, defined by

b (A, B) = 2 · trace (AB) − trace (A) trace (B) , (→) A, B ∈ M2 (R) .

(a) Prove that b is symmetrically.

(b) Write the analytical expression of b, relative to the canonical basis of
M2 (R) .
(c) Build the matrix associated to b, relative to the canonical basis of M2 (R) .
(d) Indicate the signature of the quadratic form f :M2 (R) ∈ R, defined by
f (A) = b (A, B) , (→) A ∈ M2 (R).

Solution
The solution in Sage of this problem is:
5.6 Problems 195

8. Use the eigenvalue method to determine the canonical expression and the basis
in which makes this for the quadratic form:

f : R3 ∈ R, f (x) = 5x12 + 6x22 + 4x32 − 4x1 x2 − 4x1 x3 , (→) x ∈ R3 .

Indicate the signature of f .

9. Use the Jacobi method to find the canonical expression and the basis in which
makes this for the quadratic form:

f : R3 ∈ R, f (x) = 4x12 + 2x22 − 2x1 x2 + 2x1 x3 + 2x2 x3 , (→) x ∈ R3 .

10. Reduce to the canonical expression through the three method, the following
quadratic form:
 4
f : R4 ∈ R, f (x) = xi x j .
i< j
i=1
196 5 Bilinear and Quadratic Forms

References

1. I. Vladimirescu, M. Popescu (eds.), Algebră liniară şi geometrie analitică (Universitaria,

Craiova, 1993)
2. I. Vladimirescu, M. Popescu (eds.), Algebră liniară şi geometrie n- dimensională (Radical,
Craiova, 1996)
3. V. Balan (ed.), Algebră liniară, geometrie analitică (Fair Partners, Bucureşti, 1999)
4. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică. Note de curs şi
aplicaţii (Universitatea din Craiova, 1993)
5. V. Postelnicu, S. Coatu (ed.), Mică enciclopedie matematică (Tehnică, Bucureşti, 1980)
Chapter 6
Differential Geometry of Curves and Surfaces

6.1 Analytical Definition of Curves in the Space

Definition 6.1 (see [5], p. 544). The regular arc of a curve is defined as the set π
of the points M (x, y, z) from the real three-dimensional Euclidean space R3 , whose
coordinates x, y, z check one of the following systems of equations:

F (x, y, z) = 0
, (∈) (x, y, z) → D ⊆ R3 (the implicit representation) (6.1)
G (x, y, z) = 0

z = f (x, y)
, (∈) (x, y) → D ⊆ R2 (the explicit representation) (6.2)
z = g (x, y)

 x = f 1 (t)
y = f 2 (t) , (∈) t → (a, b) (the parametric representation) , (6.3)

z = f 3 (t)

where the functions F, G, f, g, f 1 , f 2 , f 3 satisfy the following conditions of regu-

larity:
(a) are some real and continuous functions,
(b) the functions f 1 , f 2 , f 3 establish a biunivocal correspondence between the values
of the parameter t → (a, b) and the points M → π,
(c) allow first order derivatives, not all null,
(d) at least one of the functional determinants

D (F, G) D (F, G) D (F, G)

, ,
D (y, z) D (z, x) D (x, y)

is not equal to 0.

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 197

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_6,
© Springer International Publishing Switzerland 2014
198 6 Differential Geometry of Curves and Surfaces

Definition 6.2 (see [5], p. 545). The regular curve is the reunion of the regular curve
arcs.

Definition 6.3 (see [5], p. 545). Let π be a curve given by its parametric equations

 x = x (t)
y = y (t) , (∈) t → (a, b) (6.4)

z = z (t)

and let M (x, y, z) → π be a current point.

If we consider a system of rectangular axes of versors i, j, k and if r is the position
vector of the point M, then the relation

r = x (t) i + y (t) j + z (t) k, (∈) t → (a, b) (6.5)

is called the vector equation of the curve π.

6.2 Tangent and Normal Plane to a Curve in the Space

Definition 6.4 (see [5], p. 552). One considers the regular curve π and let be the
points M, M1 → π. The limit position of the chord M, M1 when M1 tends to M is
called the tangent to the curve at the point M (see Fig. 6.1).

Fig. 6.1 Tangent to a curve

The equations of the tangent to the curve at a point M0 (x0 , y0 , z 0 ) → π in the

case when:
(a) the curve is given parametrically (as in the relation (6.4)) are :

x − x0 y − y0 z − z0
T : 
=  =  , (6.6)
x (t0 ) y (t0 ) z (t0 )
6.2 Tangent and Normal Plane to a Curve in the Space 199

where 
 x = x (t0 )
y = y (t0 ) , (∈) t0 → (a, b) ;

z = z (t0 )

(b) the curve is given implicitly (see the relation (6.1)) are

x − x0 y − y0 z − x0
T : D(F,G)
= D(F,G)
= D(F,G)
; (6.7)
D(y0 ,z 0 ) D(z 0 ,x0 ) D(x0 ,y0 )

(c) the curve is given explicitly (see the relation (6.2)) are

x − x0 y − y0 z − x0
T : =  =   . (6.8)
g y0 − f y0 f x0 − gx 0 f x0 g y0 − f y0 gx 0

Definition 6.5 (see [5], p. 553). One considers the regular curve π and let be M → π.
The normal plane to the curve π in the point M is the plane π N perpendicular to
the tangent T to the curve π in the point M.
The equation of the normal plane π N to the curve π in the point M, in the case
when:
(a) the curve is given parametrically (as in the relation (6.4)) is:

π N : x  (t0 ) ( x − x (t0 )) + y  (t0 ) (y − y (t0 )) + z  (t0 ) (z − z (t0 )) = 0; (6.9)

(b) the curve is given implicitly (see the relation (6.1)) is

D (F, G) D (F, G) D (F, G)

πN : ( x − x (t0 )) + (y − y (t0 )) + (z − z (t0 )) = 0;
D (y0 , z 0 ) D (z 0 , x0 ) D (x0 , y0 )
(6.10)

(c) the curve is given explicitly (see the relation (6.2)) is

  
π N : g y0 − f y0 ( x − x (t0 )) + f x0 − gx 0 (y − y (t0 ))
 
+ f x0 g y0 − f y0 gx 0 (z − z (t0 )) = 0. (6.11)

6.3 Frenet Trihedron. Frenet Formulas

Definition 6.6 (see [5], p. 576). Let π be a regular curve and the point M0 → π.
The Frenet Trihedron attached to the curve π in the point M0 is a right trihedron
determined by the versors τ , β, ν (Fig. 6.2).
200 6 Differential Geometry of Curves and Surfaces

Fig. 6.2 Frenet trihedral

The pair of versors τ , β, ν determine the following planes:

• the normal plane π N , determined by ν and β;

• the osculator plane π0 , determined by τ and ν;
• the rectified plane πr , determined by τ and β.

The equation of the normal plane to the curve π in a point M0 (x0 , y0 , z 0 ) → π

can be expressed with the relations (6.9), (6.10) or (6.11).
The equation of the osculator plane to the curve π in a point M0 (x0 , y0 , z 0 ) → π
in the case when:

(a) the curve is given parametrically (as in the relation (6.4)) is

x − x (t0 ) y − y (t0 ) z − z (t0 )

π0 :

x (t0 ) y  (t0 ) z  (t0 )

= 0; (6.12)

x  (t0 ) y  (t0 ) z  (t0 )

(b) the curve is given implicitly (see the relation (6.1)) is

x − x0 y − y0 z − z0

D(F,G) D(F,G) D(F,G)

π0 :

D(y0 ,z 0 ) D(z 0 ,x0 ) D(x0 ,y0 )

= 0; (6.13)

D2 (F,G) D2 (F,G) D2 (F,G)

2
D (y0 ,z 0 ) D (z 0 ,x0 ) D (x0 ,y0 )
2 2

(c) the curve is given explicitly (see the relation (6.2)) is

x − x0 y − y0 z − z0

g
π0 :
y0 − f  f  − g f  g − f  g

y0 x0 x0 x0 y0 y0 x0
= 0. (6.14)

g  − f  f  − g  f  g  − f  g 

y0 y0 x0 x0 x0 y0 y0 x0
6.3 Frenet Trihedron. Frenet Formulas 201

If
v = a1 i + a2 j + a3 k, (6.15)

then the equation of the rectified plane to the curve π in the point M0 (x0 , y0 , z 0 ) → π
is:
πr : a1 ( x − x (t0 )) + a2 (y − y (t0 )) + a3 (z − z (t0 )) = 0. (6.16)

The Frenet’s trihedron axes in the point M0 (x0 , y0 , z 0 ) → π are:

1. the tangent in M0 to the curve (the tangent versor is denoted by τ ); the equations
of the tangent to the curve π in a point M0 (x0 , y0 , z 0 ) → π can be expressed with
the relations (6.6), (6.7) or (6.8);
2. the binormal of a curve in M0 (the normal which is perpendicular to the osculator
plane, that passes through the point M0 ; the binormal versor is denoted by β) has
the equations:

  x − x0 y − y0 z − z0
β :

=

=

.

y (t0 ) z  (t0 )

z (t0 ) x  (t0 )

x (t0 ) y  (t0 )



y (t0 ) z  (t0 )

z  (t0 ) x  (t0 )

x  (t0 ) y  (t0 )

3. the principal normal of a curve in M0 (the straight line contained in the normal
plane and in the osculator plane passing through M0 ; the principal normal versor
is denoted by ν) has the equations:

x − x (t0 ) y − y (t0 ) z − z (t0 )

(ν) : = = , (6.17)
a1 a2 a3

ν being expressed in (6.15).

Since the rectified plane is that plane which passes through M0 (x0 , y0 , z 0 ) and
contains the directions of the tangent and of the binormal it results that the equation
of the rectified plane is:

x − x0 y − y0 z − z 0

πr :

t1 t2 t3

= 0, (6.18)

b1 b2 b3

• τ = t1 i + t2 j + t3 k is the director vector of the tangent in the point M0 (x0 , y0 , z 0 ) ;

• β = b1 i + b2 j + b3 k is the director vector of the binormale in the point
M0 (x0 , y0 , z 0 ).

Definition 6.7 (see [5], p. 550). The arc element of a curve π is the differential ds
of the function s = s (t), which signifies the length of the respective arc, from the
curve π.
If the curve π is given by the parametric equations (6.4), then
202 6 Differential Geometry of Curves and Surfaces
⎫
ds = dx 2 + dy 2 + dz 2 . (6.19)

If we dispose by the vector equation from the relation (6.5) of the curve π, then

ds = dr . (6.20)

The versors of the Frenet trihedron are:

1. the tangent versor:

dr r x  (t) i + y  (t) j + z  (t) k

τ= = ⎬ ⎬ = ⎫ (6.21)
ds ⎬r ⎬
x  (t)2 + y  (t)2 + z  (t)2

2. the binormal versor:

r  × r 
β=⎬  ⎬ (6.22)
⎬r × r  ⎬

3. the principal normal versor:

ν = β × τ, (6.23)

where r is expressed in the relation (6.5).

 
Theorem 6.8 (see [5], p. 576). Let π be a regular curve and τ , β, ν be the Frenet
trihedron attached to the curve π in the point M → π. There are the following relations
among the versors τ , β, ν:

⎭
⎭ τ =ν×β
⎭
⎭
⎭
 ν =β×τ
β⎬=⎬τ × ν (6.24)
⎭
⎭ ⎬ ⎬
⎭
⎭ τ  = ⎬β ⎬ = ν = 1
⎭

τ · ν = τ · β = ν · β = 0.

Example 6.9 (see [1], p. 137). Let be the curve:

r (t) = 2ti + t 2 j + ln tk, t > 0.

Find the equations of the edges (the tangent, the principal normal and the binormal)
and of the planes (the normal plane, the rectified plane and the osculator plane)
corresponding to the Frenet trihedron in the point t = 1.
Solution
The tangent equations are:
6.3 Frenet Trihedron. Frenet Formulas 203

x − 2t y − t2 z − ln t
(τ ) : = = 1
.
2 2t t

The tangent equations in the point t = 1:

x −2 y−1 z
(τ ) : = = .
2 2 1
The binormale equations in the point t = 1:
  x − x (1) y − y (1) z − z (1)
β :

=

=

,

y (1) z  (1)

z (1) x  (1)

x (1) y  (1)



y (1) z  (1)

z  (1) x  (1)

x  (1) y  (1)

i.e.
  x −2 y−1 z
β :

=

=

2 1

1 2

2 2

2 −1

−1 0

0 2

or
  x −2 y−1 z
β : = = .
−4 2 4

The equation of the normal plane is:

  1
π N : 2 (x − 2t) + 2t y − t 2 + (z − ln t) = 0;
t
in the point t = 1, the equation of the normal plane becomes:

π N : 2 (x − 2) + 2 (y − 1) + z = 0

or

π N : 2x + 2y + z − 6 = 0.

Using (6.12), the equation of the normal plane in the point t = 1:

x − x (1) y − y (1) z − z (1)

π0 :

x (1) y  (1) z  (1)

= 0;

x  (1) y  (1) z  (1)

i.e
204 6 Differential Geometry of Curves and Surfaces

x − 2 y − 1 z − 1

π0 :

2 2 1

= 0

0 2 −1

or

πo : 2x − y − 2z = 3.

The principal normal is the intersection between the osculator plane and the normal
plane:

2x − y − 2z = 3
(ν) :
2x + 2y + z = 6.

As the rectified plane is the plane that passes through the point t = 1 and contains
the director vectors of the tangent and of the binormal in this point, based on (6.18),
the equation of the rectified plane will be:

x − x (1) y − y (1) z − z (1)

πr :

a1 a2 a3
= 0,

b1 b2 b3

where:

• a = (a1 , a2 , a3 ) is the director vector of the tangent in the point t = 1, i.e.

a = (2, 2, 1),
• b = (b1 , b2 , b3 ) is the director vector of the binormal in the point t = 1, i.e.
b = (−4, 2, 4);

it will result:

x −2 y −1 z

πr :

2 2 1

= 0.

−4 2 4

A solution in Sage will be given, too:

6.3 Frenet Trihedron. Frenet Formulas 205

The Frenet’s formulas establish some relations between the edge versors that bears
its name and their derivatives.
Theorem 6.10 (see [5], p. 582). Let π be a regular curve and the point M → π be a
current point, having the position vector r . Let τ , β, ν be the versors of the tangent,
binormal and the principale normal in M. If ds is the arc element on the curve π,
then the following relations are satisfied, called Frenet’s formulas:
1. the first Frenet formula:

dτ ν
= (6.25)
ds R
2. the second Frenet formula:

dβ ν
=− (6.26)
ds T
3. the third Frenet formula:
⎧ ⎨
dν τ β
=− − . (6.27)
ds R T
206 6 Differential Geometry of Curves and Surfaces

6.4 Curvature and Torsion of the Space Curves

The scalars 1/R and 1/T, that are introduced through the Frenet formulas are called
the curvature and the torsion of the given curve in the point M.
We shall examine the geometric interpretation of these scalars.
Definition 6.11 (see [5], p. 579). The mean curvature is the variation of the tangent
direction, per arc unit; the curvature at a point of a curve is the limit of the mean
curvature, when the considered arc element tends to 0 (see Fig. 6.3), i.e.

λτ

K = lim

. (6.28)
λs∀0 λs

Fig. 6.3 Mean curvature

Definition 6.12 (see [5], p. 579). The curvature radius in a point from a curve is
equal to the inverse of curvature at that point and is denoted with R.
Remark 6.13 (see [2], p. 30). The curvature of the curve indicates the speed with
which the curve moves away from the tangent.
Theorem 6.14 (see [5], p. 586). Let π be a regular curve. The necessary and sufficient
condition that this curve to be a straight line is that K = 0.
Definition 6.15 (see [5], p. 581). The torsion in a point of a curve is the variation of
the binormal direction, per unit of arc, when the considered arc element tends to 0.
Definition 6.16 (see [5], p. 581). The torsion radius in a point of a curve is equal to
the inverse of the torsion at that point and is denoted by T .
Remark 6.17 (see [2], p. 30). The torsion of the curve indicates the speed with which
the curve moves away from the osculator plane.
6.4 Curvature and Torsion of the Space Curves 207

Theorem 6.18 (see [5], p. 587). Let π be a regular curve. The necessary and sufficient
condition that this curve to be a plane curve is that T1 = 0.
Theorem 6.19 (computing the curvature, see [5], p. 589). Let π be a regular curve
and M → π be the current point, with position vector r , ds be the arc element on the
curve π and R be the radius of curvature of the the curve π in the point M. Then
⎬ ⎬
⎬  ⎬
1 ⎬r × r "⎬
K = = ⎬ ⎬3 . (6.29)
R ⎬ ⎬
⎬r ⎬

Proof
Using the first Frenet formula we have:

dτ ν d2 r ν
= ⇒ 2 = ∃
ds R ds R
d2 r
ν = R 2. (6.30)
ds
We shall achieve:
⎬ 2 ⎬ ⎬ 2 ⎬
⎬d r ⎬ ⎬ ⎬
⎬ ⎬ = 1 · ν ⇒ ⎬ d r ⎬ = 1 ;
⎬ ds 2 ⎬ R ⎩ ⎬ ds ⎬
2 R
=1

therefore

 2  2  2
1 d2 x d2 y d2 z
= + + .
R ds 2 ds 2 ds 2

Using the third formula from (6.24) and the relation (6.30) we deduce:
⎬ ⎬ ⎬ ⎬
⎬ ⎬ ⎬ dr d2 r ⎬
⎬
⎬β ⎬ = ⎬ × R 2 ⎬ =1∃
ds ds ⎬
⎬ ⎬
1 ⎬ dr d2 r ⎬
=⎬
⎬ ds × ⎬. (6.31)
R ds 2 ⎬

As
dr dr dt
= · (6.32)
ds dt ds
208 6 Differential Geometry of Curves and Surfaces

we have:
     
d2 r d dr d dr dt d dr dt dr d2 t
= = · = · + ·
ds 2 ds ds ds dt ds ds dt ds dt ds 2
   
d dr dt dt dr d2 t
= · · + · ,
dt dt ds ds dt ds 2

i.e.
 2
d2 r d2 r dt dr d2 t
= · + · . (6.33)
ds 2 dt 2 ds dt ds 2

We deduce
  ⎪ 2  2 ⎛
dr d2 r dr dt d r dt dr d2 t
× 2 = · × · + ·
ds ds dt ds dt 2 ds dt ds 2
 3    
dt dr d2 r dt d2 t dr dr
= · × 2 + · · × .
ds dt dt ds ds 2 dt dt
⎩  
=0

Therefore  3   d2 r
dr d2 r dt dr d2 r
dr
dt × dt 2
× 2 = · × 2 =  ds 3 ,
ds ds ds dt dt
dt

i.e. d2 r
dr d2 r
dr
(6.20) dt × dt 2
× 2 = ⎬ ⎬3 . (6.34)
ds ds ⎬ dr ⎬
⎬ dt ⎬

From (6.31) and (6.34) we obtain

⎬  ⎬
d2 r ⎬r × r  ⎬
dt × dt 2
dr
1
= ⎬ ⎬3 = ⎬ ⎬3 .
R ⎬ dr ⎬ ⎬r  ⎬
⎬ dt ⎬

Theorem 6.20 (computing the torsion, see [5], p. 593). Let π be a regular curve
and M → π be the current point, with position vector r , ds be the arc element on the
curve π and R be the radius of curvature of the the curve π in the point M. Then

1 r  · r  × r 
= ⎬ ⎬ . (6.35)
T ⎬r  × r  ⎬2
6.4 Curvature and Torsion of the Space Curves 209

Proof
Taking into account the relations (6.32) and (6.33), we have
⎧  2 ⎨ ⎧  2 ⎨  
d3r d d2 r dt dr d2 t d d2 r dt d dr d2 t
= · + · = · + ·
ds 3 ds dt 2 ds dt ds 2 ds dt 2 ds ds dt ds 2
   2 ⎧  ⎨   2
d d2 r dt d2 r d dt 2 d dr d t dr d3 t
= · + · + · + ·
ds dt 2 ds dt 2 ds ds ds dt ds 2 dt ds 3
  2    2  
d d r dt dt d2 r d dt dt
= · · + · ·
dt dt 2 ds ds dt 2 ds ds ds
    2
d dr dt d t dr d3 t
+ · · 2+ ·
dt dt ds ds dt ds 3
 3
d3r dt d2 r d2 t dt d2 r dt d2 t d2 r dt d2 t
= 3 · + 2 · 2· + 2 · · 2+ 2 · ·
dt ds dt ds ds dt ds ds dt ds ds 2
dr d3 t
+ · ,
dt ds 3
i.e.  3
d3r d3 r dt d2 r dt d2 t dr d3 t
= · +3 · · + · . (6.36)
ds 3 dt 3 ds dt 2 ds ds 2 dt ds 3

Using the first Frenet formula we achieve:

 
d3 r d 1 1 dν
= ν+ · . (6.37)
ds 3 ds R R ds

Using the third Frenet formula, from (6.37) we shall have:

  ⎧ ⎨  
d3 r d 1 1 τ β d 1 1 1
= ν − − = ν − 2τ + β. (6.38)
ds 3 ds R R R T ds R R R·T

By multiplying with β in (6.38), it will result:

 
d3r d 1   1   1  
Theorem 6.8 1
· β = ν · β − τ · β + β · β =
ds 3 ds R ⎩   R ⎩   R · T ⎩  
2 R·T
=0 =0 =1
1 d3 r
∃ = R 3 · (τ × ν) ;
T ds
i.e.
     
1 d3 r dr d2 r d3 r dr d2 r dr d2 r d3r
=R 3 · ×R 2 = R2 · × 2 = R2 · × .
T ds ds ds ds 3 ds ds ds ds 2 ds 3
210 6 Differential Geometry of Curves and Surfaces

We obtain:

   2   
d2 r 3 3 d3 r d2 r
1
dr
ds · ds 2
× dds r3 dr
(6.31) ds
· dds r2 × dds r3 ds 3
· dr
ds × ds 2
= = ⎬ ⎬ = ⎬ ⎬2 . (6.39)
T 1 ⎬ dr 2 ⎬2 ⎬ dr d2 r ⎬
R2 ⎬ ds × dds r2 ⎬ ⎬ ds × ds 2
⎬

We have:
d2 r
dr d2 r
dr
(6.34) ds × ds 2
× 2 = ⎬ ⎬3 ;
ds ds ⎬ dr ⎬
⎬ dt ⎬

i.e
  3  
d3 r dr d2 r 1 d r dr
(6.34) d2 r
· × 2 = ⎬ ⎬3 · ×
ds 3 ds ds ⎬ dr ⎬ ds 3 ds ds 2
⎬ dt ⎬
⎪⎧  3 ⎨  ⎛
1 d3 r dt d2 r d2 t dt dr d3 t dr d2 r
= ⎬ ⎬3 · +3 2 · 2 · + · · × 2
⎬ dr ⎬ dt 3 ds dt ds ds dt ds 3 dt dt
⎬ dt ⎬
⎜
 3 3    
1 ⎝ ⎝ dt d r dr d2 r d2 t dt d2 r dr d2 r
= ⎬ ⎬3 ⎝ · 3 · × 2 +3 2 · · · × 2
⎬ dr ⎬ ⎞ ds dt dt dt ds ds dt 2 dt dt
⎬ dt ⎬ ⎩  
=0
⎟
 ⎠
d3 t dr dr d2 r ⎠
+ · · × 2 ⎠.
ds 3 dt dt dt ⎦
⎩  
=0

Finally, we get
 dt 3    
  · d3 r
· dr
× d2 r d3 r
· dr
× d2 r
d3 r dr d2 r ds dt 3 dt dt 2 dt 3 dt dt 2
· × 2 = ⎬ ⎬3 = ⎬  ds 3 ,
⎬3
ds 3 ds ds ⎬ dr ⎬ ⎬ dr ⎬
⎬ dt ⎬ ⎬ dt ⎬ · dt

i.e.
   
  d3 r
·
2
× ddt 2rdr d3 r
· ×
dr d2 r
d3r dr d2 r dt 3 dt dt 3 dt dt 2
· × 2 = ⎬ ⎬ ⎬3 =⎬3 ⎬ ⎬6 . (6.40)
ds 3 ds ds ⎬ dr ⎬ ⎬ ⎬ ⎬ dr ⎬
⎬ dt ⎬ · ⎬ dr
dt ⎬ ⎬ dt ⎬
6.4 Curvature and Torsion of the Space Curves 211

Substituting (6.40) into (6.39) we achieve:

 
dr d2 r 3
· × d 3r
dt
⎬ ⎬6 dt dt 2  2 
⎬ dr ⎬
3 
1 ⎬ dt ⎬
dr
dt · ddt 2r × ddt 3r r  · r  × r 
= ⎬ ⎬ = ⎬ ⎬ = ⎬ ⎬ .
T ⎬ dr d2 r ⎬2
⎬ ds × 2 ⎬ ⎬ dr 2 ⎬2 ⎬r  × r  ⎬2
⎬ ds⎬6
⎬ ds × dds r2 ⎬
⎬ dr ⎬
⎬ dt ⎬

Example 6.21 (see [4], p. 37). Let be the curve:


 x = 3 cos t
y = 3 sin t

z = 4t.

Determine:

(a) the tangent versor in any point M;

(b) the curvature of the curve in the point M;
(c) the principal normal versor in M;
(d) the equations of the principal normal;
(e) the binormal versor in M;
(f) the equations of the binormal;
(g) the torsion of the curve in M;
(h) the equation of the osculator plane in M.

Solution
(a) Using (6.21), the tangent versor is:

−3 sin t i + 3 cos t j + 4k 1
τ= ⇐ = −3 sin t i + 3 cos t j + 4k .
9 + 16 5

We can get the tangent in Sage, too:

We need the following Sage code to plot the curve:

212 6 Differential Geometry of Curves and Surfaces

(b) + (c) The given curve has the vector equation

r = 3 cos t i + 3 sin t j + 4t k.

We compute:

dτ dτ dt 1 dτ 1 dτ 1 dτ 1 
= · = dt · =⎬ ⎬· = ⎬ ⎬ · = −3 cos t i − 3 sin t j .
ds dt ds dt ⎬ ⎬ dt ⎬r ⎬ dt 25
ds ⎬ dr
dt ⎬

Using the first Frenet formula, it results:

1 3 
ν= − cos t i − sin t j ;
R 25
therefore:

• the curvature of the curve is

1 3
= ;
R 25
• the principal normal versor is:

ν = − cos t i − sin t j.
6.4 Curvature and Torsion of the Space Curves 213

The curvature of the curve can be also determined using the formula (6.29).
Solving (b) + (c) + (e) with Sage, we have:

(d) The equations of the principal normal are:

 x−3 cos t y−3 sin t
(ν) : − cos t = − sin t .
z = 4t

(e) the binormal versor can be determined using one of the two formulas: (6.22)
or the third relation from (6.24). We achieve:
1
β= 4 sin t i − 4 cos t j + 3k .
5
(f) The binormal equations are determined using the formulas (6.16); it results:
  x − 3 cos t y − 3 sin t z − 4t
β : = = .
4 sin t −4 cos t 3

(g) As
dβ dβ dt 1 dβ 1 dβ
= · = ds · =⎬ ⎬·
ds dt ds dt ⎬ dr ⎬ dt
dt ⎬ dt ⎬
1 dβ 1 
= ⎬ ⎬ · = 4 cos t i + 4 sin t j ,
⎬r ⎬ dt 25

using the second Frenet formula we shall have:

1  4 
− − cos t i − sin t j = cos t i + sin t j ;
T 25
214 6 Differential Geometry of Curves and Surfaces

hence, the torsion of the curve is:

1 4
= .
T 25
The same value can be found using Sage:

(h) The equation of the osculator plane in M can be determined with (6.12):

x − 3 cos t y − 3 sin t z − 4t

π0 :

−3 sin t 3 cos t 4

= 0.

−3 cos t −3 sin t 0

Using Sage, we achieve:

6.5 Envelope of a Family of Curves in Plane

Definition 6.22 (see [3], p. 121). A relation of the form

F (x, y, λ) = 0, (6.41)

where λ is a real parameter represents a family of curves in the plane xOy, each
curve from the family being determined by the value of the respective parameter λ.

Definition 6.23 (see [3], p. 121). The envelope of a family of curves is the tangent
curve in every of its points, to a curve from that family (see Fig. 6.4).
As the considered family of curves depends on the parameter λ it results that as
well as the envelope points will depend on the values of λ. Therefore, the envelope
of a family of curves has the parametric representation:

x = x (λ)
I : (6.42)
y = y (λ) .
6.5 Envelope of a Family of Curves in Plane 215

The common points of the envelope and of the curves that belong to the respective
family check the equation:

F (x (λ) , y (λ) , λ) = 0. (6.43)

By differentiating with respect to λ the relation (6.43) we achieve:

∂ F ∂x ∂F ∂y ∂F
· + · + = 0, (6.44)
∂x ∂λ ∂ y ∂λ ∂λ

i.e.

Fx xλ + Fy yλ + Fλ = 0. (6.45)

The equation of the tangent in a point M (x, y) to a curve from the family of
curves that has the Eq. (6.41) is

Fx
Y −y=− (X − x) , (6.46)
Fy

where X, Y are some current coordinates on the straight line.

The equation of the tangent in a point M (x, y) to the envelope of a family of
curves, characterized by the Eq. (6.42) is

yλ
Y−y= (X − x) . (6.47)
xλ

As the envelope and the family of curves are tangent in the common points it
results that the slopes of tangents coincide, i.e. from the relations (6.46) and (6.47)
we deduce

yλ Fx
= − ⇒ Fx xλ + Fy yλ = 0. (6.48)
xλ Fy

Substituting (6.48) into (6.45) we achieve

Fλ (x, y, λ) = 0. (6.49)

Fig. 6.4 The envelope of a

family of curves
216 6 Differential Geometry of Curves and Surfaces

Therefore, in conclusion, to get the equation of the considered family envelope it

must that the parameter λ to be eliminated between the Eqs. (6.41) and (6.49).
If the parameter λ can’t be eliminated, then from the Eqs. (6.41) and (6.49) one
determines the coordinates of a current point of the envelope, depending on λ and
these will constitute the parametric equations of the envelope.

Example 6.24 (see [4], p. 331). Determine the envelope of the family of straight
lines :
Fλ (x, y, λ) = x cos λ + y sin λ − a = 0,

λ being a real parameter.

Solution
We have
Fλ (x, y, λ) = −x sin λ + y cos λ.

Solving the system, which contains the Eqs. (6.41) and (6.49), i.e.:

Fλ (x, y, λ) = 0
Fλ (x, y, λ) = 0

it will result:

x cos λ + y sin λ − a = 0
(6.50)
−x sin λ + y cos λ = 0.

By multiplying the first equation of the system (6.50) with sin λ and the second
equation with cos λ we achieve

y = a sin λ.

By multiplying the first equation of the system (6.50) with cos λ and the second
equation with − sin λ we achieve

x = a cos λ.

We shall deduce

x 2 + y 2 = a 2 sin2 λ + a 2 cos2 λ = a 2 .

Hence, the elimination of the parameter λ from the system (6.50) gives us the
equation

x 2 + y2 = a2,
6.5 Envelope of a Family of Curves in Plane 217

i.e. the envelope is a circle centered in the origin and having the radius a.
Solving this problem in Sage, we obtain:

6.6 Analytic Definition of Surfaces

Definition 6.25 (see [5], p. 602). A regular portion of a surface is the set ϕ of
the points M (x, y, z) from the three-dimensional Euclidean real space R3 , whose
coordinates x, y, z check one of the following systems of equations:

F (x, y, z) = 0, (x, y, z) → D ⊆ R3 (the implicit representation) (6.51)

z = f (x, y) , (x, y) → D ⊆ R2 (the explicit representation) (6.52)


 x = f 1 (u, v)
y = f 2 (u, v) = 0, (u, v) → D ⊆ R2 (the parametric representation) , (6.53)

z = f 3 (u, v)

where the functions F, f 1 , f 2 , f 3 satisfy the following regularity conditions:

(a) they are real and continuous functions;

(b) the functions f 1 , f 2 , f 3 establish a biunivocal correspondence between the points
M → ϕ and the ordered pairs (u, v), where u and v are some real parameters;
(c) they admit first order derivatives, continuous, that aren’t all null;
f 1 , f 2 ) D( f 2 , f 3 ) f3 , f1 )
(d) at least one of the functional determinants D(D(u,v) , D(u,v) and D(D(u,v) isn’t
equal to 0.
Definition 6.26 (see [5], p. 604). Let ϕ be a regular portion of surface, given by its
parametric equations
218 6 Differential Geometry of Curves and Surfaces

 x = x (u, v)
y = y (u, v) , (u, v) → D ⊆ R2 (6.54)

z = z (u, v)

and M (x, y, z) → ϕ be a current point.

If we consider a system of rectangular axes of versors i, j, k and if r is the
position vector of the point M, then the relation

r = x (u, v) i + y (u, v) j + z (u, v) k (6.55)

constitutes the vector representation of ϕ.

Definition 6.27 (see [5], p. 603). A regular surface is the union of the regular surface
portions.
Definition 6.28 (see [6], p. 708). Let ϕ be a regular surface, given by its parametric
equations (6.54). The ordered pairs (u, v), that determines the position of a point
from the surface, are called the curvilinear coordinates on the surface ϕ.
Definition 6.29 (see [5], p. 647). Let ϕ be a regular surface, given by its parametric
equations (6.54). The set of the poins M (x, y, z) → ϕ, whose coordinates verify the
equations 
 x = x (u (t) , v (t))
y = y (u (t) , v (t)) , t → (a, b) (6.56)

z = z (u (t) , v (t))

forms a curve π, called a curve traced on the surface ϕ, π ⇔ ϕ.

The Eq. (6.56) are called the parametric equations of the curve π, traced on the
surface ϕ.

Theorem 6.30 (see [5], p. 609). Let ϕ be a regular surface. If (u, v) is a curvilinear
coordinate system on the surface ϕ, then any curve π ⇔ ϕ can be analytically
represented by one of the following equation:

u = u (t)
(6.57)
v = v (t)

f (u, v) = 0 (6.58)

u = g (v) . (6.59)
6.7 Tangent Plane and Normal to a Surface 219

6.7 Tangent Plane and Normal to a Surface

Definition 6.31 (see [5], p. 615). The tangent plane in a point of the surface ϕ is
set of the tangents pursued to all of the curves from the surface, passing through that
point (see Fig. 6.5).

Fig. 6.5 Tangent plane

The equation of the tangent plane to the surface ϕ in a point M0 (x0 , y0 , z 0 ) → ϕ,

in the case when:

(a) the surface is given parametrically (as in the relation (6.54)) is

x − x0 y − y0 z − z 0

πT :

xu 0 yu 0 z u 0

= 0, (6.60)

xv yv0 z v0

where

⎭
⎭ x0 = x (u 0 , v0 ) , y0 = y (u 0 , v0 ) , z 0 = z (u 0 , v0 )
⎭
⎭
 ∂x ∂y ∂z
xu 0 = |u=u 0 ,v=v0 , yu 0 = |u=u 0 ,v=v0 , z u 0 = |u=u 0 ,v=v0
⎭ ∂u ∂u ∂u
⎭
⎭
 xv = ∂x |u=u ,v=v , yu = ∂ y |u=u ,v=v , z v = ∂z |u=u ,v=v .
⎭
0
∂v 0 0 0
∂v 0 0 0
∂v 0 0

(b) the surface is given implicitly (see the relation (6.51)) is

∂F ∂F
πT : (x − x0 ) |x=x0 ,y=y0 ,z=z 0 + (y − y0 ) |x=x0 ,y=y0 ,z=z 0
∂x ∂y
∂F
+ (z − z 0 ) |x=x0 ,y=y0 ,z=z 0 = 0 (6.61)
∂z

(c) the surface is given explicitly (see the relation (6.52)) is

∂z ∂z
πT : (X − x0 ) |x=x0 ,y=y0 +(Y − y0 ) |x=x0 ,y=y0 −(Z − z 0 ) = 0. (6.62)
∂x ∂y
220 6 Differential Geometry of Curves and Surfaces

Example 6.32 (see [5], p. 618). Let be the surface


x = u +v
ϕ : y =u−v

z = uv.

Find the equation of the tangent plane to the surface ϕ in the point M, for u 0 =
2, v0 = 1.

Solution
We have 
xu = 1, yu = 1, z u = v
xv = 1, yv = −1, z v = u.

Considering u 0 = 2, v0 = 1 it results x0 = 2, y0 = 1, z 0 = 2; by replacing them

in the equation of the tangent plane (6.60) we have

x − 3 y − 1 z − 2

πT :

1 1 1

= 0,

1 −1 2

i.e πT : 3 (x − 3) − (y − 1) − 2 (z − 2) = 0 ⇒
πT : 3x − y − 2z − 4 = 0.

A solution in Sage will be given, too:

Example 6.33 (see [5], p. 619). Let be the surface

ϕ : x 2 + 2x y + y 2 + 4x z + z 2 + 2x + 4y − 6z + 8 = 0.

Find the equation of the tangent plane to the surface in the point M (0, 0, 2).

Solution.
We have

F (x, y, z) = x 2 + 2x y + y 2 + 4x z + z 2 + 2x + 4y − 6z + 8.
6.7 Tangent Plane and Normal to a Surface 221

Whereas  ∂F
⎭
⎭ = 2x + 2y + 4z + 2
 ∂x
∂F
⎭ ∂ y = 2x + 2y + 4
⎭
 ∂F
∂z = 4x + 2z − 6,

with (6.61), the equation of the tangent plane to the surface in the point M (0, 0, 2)
will be
πT : 10x + 4y − 2 (z − 2) = 0,

i.e
πT : 5x + 2y − z + 2 = 0.

The problem can be also solved in Sage:

Definition 6.34 (see [5], p. 621). A normal in a point of a surface is the straight line
perpendicular to the tangent plane to surface in that point (Fig. 6.6).

Fig. 6.6 Normal in a point of

surface
222 6 Differential Geometry of Curves and Surfaces

The equations of the normal to a surface ϕ in a point M0 (x0 , y0 , z 0 ) → ϕ, in the

case when:
(a) the surface is given parametrically (as in the relation (6.54)) is

x − x0 y − y0 z − z0
λN :

=

=

; (6.63)

yu z u

z u xu

xu yu

0 0

0 0

0 0

yv z v

z v xv

xv yv

0 0 0 0 0 0

(b) the surface is given implicitly (see the relation (6.51)) is

x − x0 y − y0 z − z0
λN : ∂F
= ∂F
= ∂F
; (6.64)
∂x |x=x0 ,y=y0 ,z=z 0 ∂ y |x=x0 ,y=y0 ,z=z 0 ∂z |x=x0 ,y=y0 ,z=z 0

(c) the surface is given explicitly (see the relation (6.52)) is

X − x0 Y − y0 Z − z0
λN : ∂z
= ∂z
= . (6.65)
∂x |x=x0 ,y=y0 ∂ y |x=x0 ,y=y0
−1

Example 6.35 (see [5], p. 623). Find the equation of the normal to the surface from
the Example 6.32, in the point u 0 = 2, v0 = 1.

Solution
For u 0 = 2, v0 = 1 it results that x0 = 2, y0 = 1, z 0 = 2.
We shall have:
!
∂x ∂y ∂z
∂u = 1, ∂u = 1, ∂u = v
∂x ∂y ∂z
∂v = 1, ∂v = −1, ∂v = u;

therefore, based on (6.63), the equations of the normal to the surface will be:
x −3 y − y0 z − z0
λN :

=

=

1 1

1 1

1 1

−1 2

2 1

1 −1

or
x −3 y−1 z−2
λN : = = .
3 −1 −2

We shall present a solution in Sage, too:

6.7 Tangent Plane and Normal to a Surface 223

Example 6.36 (see [5], p. 623). Find the equation of the normal λ N to the surface
from the Example 6.33, in the point M0 (0, 0, 2) → ϕ.

Solution
As in the Example 6.33 we computed: ∂∂xF , ∂∂Fy , ∂∂zF , using the relation (6.64) we
achieve
x −3 y z−2 x y z−2
λN : = = ⇒ λN : = = .
10 4 −2 5 2 −1

The solution with Sage is:

6.8 First Fundamental Form of a Surface. Curves on a Surface

Definition 6.37(see [5], p. 637). The first fundamental form of a surface ϕ, denoted
by ρ1 is the square of the arc element corresponding to an arbitrary arc from the
surface:
ρ1 = ds 2 . (6.66)

Remark 6.38 (see [5], p. 637). The first fundamental form ρ1 is also called the
metric of the surface ϕ.
Theorem 6.39 (expression of the first fundamental form, see [5], p. 637). Let ϕ
be a regular surface, given through its parametric equations (6.54).
224 6 Differential Geometry of Curves and Surfaces

Then

ρ1 = Edu 2 + 2Fdudv + Gdv 2 , (6.67)

where

 E = xu2 + yu2 + z u2
F = xu xv + yu yv + z u z v (6.68)

G = xv2 + yv2 + z v2

and !
∂y
xu = ∂x ∂z
∂u , yu = ∂u , z u = ∂u
∂y (6.69)
xv = ∂x ∂z
∂u , yu = ∂u , z u = ∂u .

Definition 6.40 (see [5], p. 643). Let π ⇔ ϕ be the curve given by (6.54). If M1 , M2
are two points corresponding to the values of t = t1 and t = t2 then the length of
the arc M1 M2 is

"

" t   2  2

t2

2 du du dv dv

s =

ds

E + 2F · +G dt

, (6.70)
t1
t1 dt dt dt dt

ds being the arc element on the curve π.

Example 6.41 (see [5], p. 643). One considers the surface

 x = u cos v
y = u sin v

z = av, a > 0.

Let π1 , π2 , π3 be three curves traced on the surface ϕ, as shown below

given by the parametric equations in the curvilinear coordinates


 π1 : v = 1
π2 : u = 21 av 2

π3 : u = − 21 av 2 .
6.8 First Fundamental Form of a Surface. Curves on a Surface 225

If M1 → π1 ≺π2 , M2 → π2 ≺π3 , M3 → π1 ≺π3 and s1 , s2 , s3 are the lengths of the

arcs M1 M2 , M2 M3 , M3 M1 find the perimeter of the curvilinear triangle M1 M2 M3 .

Solution
The curvilinear coordinates of the points are M1 , M2 , M3 are
   
1 1
M1 u = a, v = 1 , M2 (u = 0, v = 0) , M3 u = − a, v = 1 .
2 2

The parametric equations of the curves π1 , π2 , π3 traced on the surface ϕ are:


 x = u cos 1
π1 : y = u sin 1 ∃

z=a

 dx = du cos 1
⎭
(6.19)
dy = du sin 1 ∃ ds = du;
⎭

dz = da

 x = 2 av cos v
1 2
⎭
π2 : y = 21 av 2 sin v ∃
⎭

z = av
 1
⎭
 dx = 2 a · 2v cos v − 1
2 av 2 sin v dv

dy = 21 a · 2v sin v + 21 av 2 cos v dv ∃
⎭

dz = adv
 
(6.19) 1 2
ds = a v + 1 dv;
2

 x = − 2 av cos v
1 2
⎭
π3 : y = − 21 av 2 sin v ∃
⎭

z = av
  1
 dx = − 2 a · 2v cos v + 2 av sin v dv
1 2
⎭
dy = − 21 a · 2v sin v − 21 av 2 cos v dv ∃
⎭

dz = adv
 
(6.19) 1 2
ds = a v + 1 dv.
2
226 6 Differential Geometry of Curves and Surfaces

We shall have

# 

7
⎭
0 1 2
⎭
⎭ s 1 =
a v + 1 dv
= 6 a,
⎭
⎭

1 2

# 1 

s2 =
0 a 21 v 2 + 1 dv
= 76 a
⎭
⎭

⎭
⎭
# 21 a

⎭
 s3 =
1 du

= a.

−2a

The perimeter of the curvilinear triangle M1 M2 M3 will be

7 7 20 10
P = s1 + s2 + s3 = a+ a+a = a= a.
6 6 6 3
This problem will be also solve in Sage:
6.8 First Fundamental Form of a Surface. Curves on a Surface 227

Definition 6.42 (see [5], p. 644). Let ϕ be a regular surface and π1 ⇔ ϕ, π2 ⇔ ϕ

be two curves traced on the surface ϕ. The angle between the curves π1 , π2 is the
angle of the tangents to the two curves in their point of intersection.

Fig. 6.7 The angle between

the two curves

Theorem 6.43 (see [5], p. 645). Let the surface ϕ given by (6.54 ) and π1 ⇔ ϕ, π2 ⇔
ϕ be two curves traced on the surface ϕ. The angle between the curves π1 , π2 in
the point M → π1 ≺ π2 is given by the relation

Edu∂u + F (du∂u + dv∂v) + Gdv∂v

cos ϕ = ⇐ ⇐ , (6.71)
Edu 2 + 2Fdudv + Gdv 2 · E∂u 2 + 2F∂u∂v + G∂v 2

where
• E, F, G are the coefficients from the metric of the surface ϕ computed in M,
• du, dv are two differentials along the curve π1 , and ∂u, ∂v are two differentials
along the curve π2 .
Remark 6.44 (see [3], p. 146). If π1 : u = C1 , π2 : v = C2 , C1 , C2 → R (the curves
have constant coordinates); as du = 0, ∂v = 0 it results that cos ϕ = ⇐ F .
EG
Example 6.45 (see [3], p. 147). Compute the angle between the curves π1 : u = v
and π2 : u + v = 2, traced on the surface
   
ϕ : r = u 2 + v 2 i + u 2 − v 2 j + 2uvk.

Solution
The parametric equations of the surface ϕ will be

 x = u 2 + v2
y = u 2 − v2

z = 2uv.
228 6 Differential Geometry of Curves and Surfaces

Using (6.68), we shall compute the coefficients of the first fundamental form:

 E = 4u 2 + 4u 2 + 4v 2 = 8u 2 + 4v 2
F = 2u · 2v − 2u · 2v + 2v · 2u = 4uv

G = 4v 2 + 4v 2 + 4u 2 = 4u 2 + 8v 2 .

As π1 : u = v and π2 : u + v = 2 it results that du =dv and ∂u = −∂v.

We shall achieve:
(6.71)
$     %
cos ϕ = −4 2u 2 + v 2 dv∂v + 4uv (dv∂v − dv∂v) + 4 2v 2 + u 2 dv∂v ·
1
= ⎫  
4 2u 2 + v 2 dv 2 + 8uvdv 2 + 4 2v 2 + u 2 dv 2 ·
⎫  
4 2u 2 + v 2 ∂v 2 − 8uv∂v 2 + 4 2v 2 + u 2 ∂v 2

4 −u 2 + v 2 dv∂v
=⎫  ⎫ 
4 3u 2 + 2uv + 3v 2 dv 2 · 4 3u 2 − 2uv + 3v 2 ∂v 2
−u 2 + v 2 −u 2 + v 2
= ⎫  = ⎫  2 .
3u 2 + 2uv + 3v 2 3u 2 − 2uv + 3v 2 9 u 2 + v 2 − 4u 2 v 2

Taking into account that the point of intersection of two curves is u = 1, v = 1;

therefore 0
cos ϕ = ⇐ ,
36 − 4

i.e. the two curves are orthogonal.

The same result can be achieved if we shall solve the problem in Sage:
6.8 First Fundamental Form of a Surface. Curves on a Surface 229

Theorem 6.46 (see [5], p. 655). Let ϕ be a regular surface. The area of the surface
ϕ is calculated using the following surface integral

" " &

σ= E G − F 2 dudv, (6.72)
ϕ

where E, F, G are the coefficients of the first fundamental form.

Definition 6.47 (see [5], p. 655). Let ϕ be a regular surface and σ from (6.72) be its
area. The surface area element corresponding to ϕ, denoted by dσ is the expression

&
dσ = E G − F 2 dudv. (6.73)

Theorem 6.48 (see [5], p. 655). Let ϕ be a regular surface. The area and the surface
area element have respectively the expressions from bellow; if

(a) the surface is given parametrically (as in the relation (6.54)), then

 ## ⇐
σ= ⇐ ϕ A√2 + B √2 + C √2 dudv
(6.74)
dσ = A + B √2 + C √2 dudv,
√2

where


√ =
yu z u

⎭
⎭ A
⎭
⎭
yv z v

⎭
⎭


z x

B √ =

u u

(6.75)
⎭
⎭ z v xv
⎭
⎭

⎭
⎭
x y

⎭
 C √ =

u u

;
xv yv

(b) the surface is given implicitly (see the relation (6.54)), then

 ' 2  2  2
⎭
⎭ ## ∂F
+ ∂∂Fy + ∂∂zF
⎭
 ∂x
σ= ϕ ∂F dxdy
' 
∂z
2  2
(6.76)
⎭
⎭
⎭
 dσ = 1+ ∂z
+ ∂z
dxdy.
∂x ∂y
230 6 Differential Geometry of Curves and Surfaces

6.9 Problems

1. Let be the curve ⇐

π : r = e−t i + et j + 2tk.

(a) Determine the analytical expressions of the versors of the Frenet trihedron
in an arbitrary point of the curve.
(b) Write the equations of the edges and of the planes of the Frenet trihedron in
an arbitrary point of the curve.
(c) Find the curve and the torsion of the curve in an arbitrary point.

2. Find the versors of the Frenet trihedron in the origin for the curve:

π : r = t i + t 2 j + t 3 k.

Solution
Solving in Sage, we get:
6.9 Problems 231

3. Write the tangent equations and the equation of the normal plane to the curve:

 x (t) = et cos 3t
π : y (t) = et sin 3t

z (t) = e−2t

in the point t = 0.
4. Find the relation between the curvature and the torsion of the curve:

t2 t3
π:r =ti+ j + k.
2 6
Solution
We shall use Sage to determine the asked relation:
232 6 Differential Geometry of Curves and Surfaces

5. Determine the position vectors of those points belonging to the curve

 
π : r = t −1 i + t j + 2t 2 − 1 k

for which the binormale is perpendicularly on the line


x+y=0
d:
4x − z = 0.

Solution
We need the following Sage code to solve this problem:

6. Let be the surface


 x (u, v) = u cos v
ϕ : y (u, v) = u sin v

z (u, v) = av.
6.9 Problems 233

Write the first fundamental form of the surface ϕ.

Solution
The solution in sage of this problem is:

7. Let be the surface


 x (u, v) = uev
ϕ : y (u, v) = ue−v

z (u, v) = 4uv.

(a) Find the equation of the tangent plane to the surface in the point u = 0, v = 2;
(b) Find the equation of the normal to the surface in that point.
8. Let be the surface

ϕ : z = 5x 2 + 4y − 3.

(a) Find the equation of the tangent plane in the point M (1, 0, 2).
(b)Write the equation of the of normal to the surface ϕ in the point M (1, 0, 2).
Solution
The solution in Sage of this problem is:

9. Calculate the angle between the curves π1 : u−ev = 0 and π2 : u 2 +u+1−e−v =

0, traced on the surface
234 6 Differential Geometry of Curves and Surfaces

ϕ : r = u cos v i + u sin v j + (u + v) k.

10. Find the area of the quadrilateral, bounded by the curves: u = 0, u = a, v = 0,

v = π, traced on the surface

ϕ : r = u cos v i + u sin v j + av k.

References

1. G.h. Atanasiu, G.h. Munteanu, M. Postolache, Algebră liniară, geometrie analitică şi
diferenţială, ecuaţii diferenţiale, ed (ALL, Bucureşti) (1998)
2. I. Bârză, Elemente de geometrie diferenţială, ed (MatrixRom, Bucureşti) (2007)
3. T. Didenco, Geometrie analitică şi diferenţială (Academia Militară, Bucureşti) (1977)
4. C. Ionescu- Bujor, O. Sacter, Exerciţii şi probleme de geometrie analitică şi diferenţială, vol II,
ed (Didactică şi pedagogică, Bucureşti) (1963)
5. E. Murgulescu, S. Flexi, O. Kreindler, O. Sacter, M. Tîrnoveanu, Geometrie analitică şi
diferenţială, ed (Didactică şi pedagogică, Bucureşti) (1965)
6. V. Postelnicu, S. Coatu, Mică enciclopedie matematică, ed (Tehnică, Bucureşti) (1980)
Chapter 7
Conics and Quadrics

7.1 General Equation of a Conic

The analytical geometry replaces the definition and the geometrical study of the
curves and the surfaces with that algebraic: a curve respectively a surface is defined
by an algebraic equation and the study of the curves and of the surfaces is reduced
to the study of the equation corresponding to each of them.
Definition 7.1 (see [1], p. 158). We consider the function

f : R2 ∈ R, f (x, y) = a11 x 2 + 2a12 x y + a22 y 2 + 2b1 x + 2b2 y + c. (7.1)

A second order algebraic curve or a conic is the set π of the points M (x, y)
from the plane, whose coordinates relative to an orthonormal Cartesian reference
check thegeneral equation
f (x, y) = 0 (7.2)

where the coefficients a11 , a12 , a22 , b1 , b2 , c are some real constants, with a11
2 +

a12 + a22 →= 0; therefore

2 2

 
π = M (x, y) | (x, y) ∈ R2 , f (x, y) = 0 . (7.3)

Definition 7.2 (see [2]). The invariants of a conic are those expressions made with
the coefficients of the conic equation, that keep the same value to the changes of an
orthonormal reference.
Proposition 7.3 (see [2] and [3], p. 56). We can assign three invariants of the conics
from (7.2):

1) I = a11 + a22 , (7.4)

 
a a 
2) δ =  11 12  ,
a12 a22

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 235

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3_7,
© Springer International Publishing Switzerland 2014
236 7 Conics and Quadrics
 
 a11 a12 b1 
 
3) λ =  a12 a22 b2  ;
 b1 b2 c 

the first invariant is linear, the second is quadratic, while the third is a cubical one.
The invariant λ determines the nature of a conic. Thus, if:
• λ →= 0, we say that π is a non-degenerate conic (the circle, the ellipse, the hyper-
bola and the parabola)
• λ = 0, we say that π is a degenerate conic.
With the help of δ one establishes the type of a conic. Thus, if:
• δ > 0, we say that π has an elliptic type
• δ < 0, we say that π is of a hyperbolic type
• δ = 0, we say that π has a parabolic type.
Definition 7.4 (see [1], p. 168). The center of symmetry of a conic π (in the case
when this exists, the conic is called a conic with center) is a point C from the plane,
which has the property that for any point M ∈ π, the reflection of M with respect to
C satisfies the equation of the conic π, too.
Theorem 7.5 (see [1], p. 168). The conic π from (7.2) admits a unique center of sym-
metry C (x0 , y0 ) if and only if its invariant δ is non-null; in this case, its coordinates
are the solutions of the linear system
 
∂f
∂x =0 a11 x + a12 y + b1 = 0
∂f ⇔ (7.5)
∂y =0 a12 x + a22 y + b2 = 0.

The non-degenerate conics

• with center are: the circle, the ellipse, the hyperbola
• without center is: the parabola.
The General Table of the Conic Discussion
(I) If λ →= 0 then for:
(1) δ > 0, the conic is
(a) a real ellipse when I · λ < 0
(b) an imaginary ellipse when I · λ > 0
(2) δ = 0, the conic is a parabola
(3) δ < 0, the conic is a hyperbola
(II) If λ = 0 then for:
(1) δ > 0, we achieve two imaginary concurrent lines with a real intersection
(2) δ = 0, we can obtain:
(a) two parallel lines if δ1 < 0
(b) two confounded lines if δ1 = 0
7.1 General Equation of a Conic 237

(c) two imaginary parallel lines if δ1 > 0

(3) δ < 0, we achieve two real concurrent lines,
where
δ1 = a11 c − b12 . (7.6)

Theorem 7.6 (see [4], p. 174). Each conic has one of the following canonical forms:
(1) imaginary ellipse:

x2 y2
2
+ 2 + 1 = 0; (7.7)
a b
(2) real ellipse:

x2 y2
2
+ 2 − 1 = 0; (7.8)
a b
(3) hyperbola:

x2 y2
2
− 2 − 1 = 0; (7.9)
a b
(4) two imaginary concurrent lines having a real intersection

x2 y2
2
+ 2 = 0; (7.10)
a b
(5) two real concurrent lines

x2 y2
2
− 2 = 0; (7.11)
a b
(6) parabola

y 2 = 2 px; (7.12)

(7) two imaginary parallel lines

x2
+ 1 = 0; (7.13)
a2
(8) two real parallel lines
x2
− 1 = 0; (7.14)
a2
238 7 Conics and Quadrics

(9) a pair of confounded lines

x 2 = 0. (7.15)

7.2 Conics on the Canonical Equations

7.2.1 Circle

Definition 7.7 (see [5], p. 117). The circle is the set of the equally spaced points
from the plane by a fixed point called the center, the distance from the center to the
points of the circle, being the radius.  
We report the circle plane to an orthogonal Cartesian reference O, i, j . Let
C (a, b) be the center of the circle and M(x, y) be an arbitrary point of it (see
Fig. 7.1).

Fig. 7.1 Circle

From the Definiton 7.7, it results that the distance from C and M is constant and
equal to the radius r of the circle, i.e.:

C M = r ;

hence

(x − a)2 + (y − b)2 = r.

By raising to the power of 2, from the previous relation we get:

(x − a)2 + (y − b)2 = r 2 . (7.16)

If we open the squares from (7.16) we achieve the circle equation in the form:

x 2 + y 2 − 2ax − 2by + a 2 + b2 − r 2 = 0. (7.17)

7.2 Conics on the Canonical Equations 239

Denoting by ⎫
⎬ m = −a
n = −b (7.18)
⎭
p = a 2 + b2 − r 2 ,

the Eq. (7.17) becomes

x 2 + y 2 + 2mx + 2ny + p = 0. (7.19)

In the case of the circle, the set π will be:

 
π = M (x, y) | (x, y) ∈ R2 , x 2 + y 2 + 2mx + 2ny + p = 0 . (7.20)

As the Eq. (7.20) can be written in the form

(x + m)2 + (y + n)2 = m 2 + n 2 − p

it results that:
1. if m 2 + n 2 ⎧
− p > 0 then the circle π will have the center C(−m, −n) and the
radius r = m 2 + n 2 − p;
2. if m 2 + n 2 − p = 0 then the circle π one reduces to the point C(−m, −n);
3. if m 2 + n 2 − p < 0 then π = ϕ (empty set).
For m 2 + n 2 − p > 0, the Eq. (7.19) is called the general Cartesian equation of
the circle π.
If θ is that angle made by the radius with the positive direction of the axis Ox,
then the parametric equations of the circle will be:

x = a + r cos θ
, θ ∈ [0, 2π] .
y = b + r sin θ

Example 7.8 (see [5], p. 121). Find the equation of the circle determined by the
points: M (−1, 1) , N (2, −1) , P (1, 3).
Solution
Using the Eq. (7.19) we deduce
⎫ ⎫
⎬ 2m − 2n − p = 2 ⎬ m = − 10
11

4m − 2n + p = −5 ⇒ n = − 10
9
⎭ ⎭
2m + 6n + p = −10; p=−5.12

The equation of the circle will be:

22 18 12
x 2 + y2 − x− y− =0
10 10 5
240 7 Conics and Quadrics

or ⎨ ⎩
5 x 2 + y 2 − 11x − 9y − 12 = 0.

The solution in Sage is:

7.2.2 Ellipse

Let c > 0 be a positive real number and F, F ∀ be two fixed points of the plane such
that F F ∀ = 2c.
Definition 7.9 (see [5], p. 150). The ellipse is the set of the points M of the plane
that satisfy the relation
M F + M F ∀ = 2a = ct, (7.21)

which has a constant sum of the distances to two fixed points.

7.2 Conics on the Canonical Equations 241

We choose F F ∀ as Ox axis and the mediator of the segment F F ∀ as Oy axis (see

Fig. 7.2). We denote F F ∀ ⇒ B B ∀ = {O}.

Fig. 7.2 Ellipse

Therefore F(c, 0) and F ∀ (−c, 0); the points F and F ∀ are called the foci of the
ellipse and the distance F F ∀ constitutes the focus distance of the ellipse. M F, M F ∀
are the focus radiuses (focus radii) of the point M. The ellipse admits a unique center
of symmetry O and two axes of symmetry Ox,Oy. The ellipse is a bounded curve
(there is a rectangle that contains all its points).
To find the ellipse equation we shall transform analytically the Eq. (7.21). If
M (x, y), then the relation (7.21) becomes

(x − c)2 + y 2 + (x + c)2 + y 2 = 2a. (7.22)

As we want to simplify the relation (7.22), we write:

(x − c)2 + y 2 = 4a 2 − 4a (x + c)2 + y 2 + (x + c)2 + y 2

or

a (x + c)2 + y 2 = a 2 + cx;

therefore
⎨ ⎩ ⎨ ⎩
a 2 − c2 x 2 + a 2 y 2 − a 2 a 2 − c2 = 0. (7.23)

In the triangle M F F ∀ it is known that M F + M F ∀ > F F ∀ or 2a > 2c, hence a

> c; so that a 2 − c2 > 0.
Therefore, we can denote
a 2 − c2 = b2 . (7.24)

Dividing by a 2 b2 in (7.23) it results the implicit Cartesian equation (7.8) of the

ellipse. If a = b = r, the Eq. (7.8) becomes

x 2 + y2 = r 2 (7.25)
242 7 Conics and Quadrics

and it represents a circle centered in origin and having the radius r . Therefore, the
circle is a special case of an ellipse.
Thereby:  
2 y2
2 x
π = M (x, y) | (x, y) ∈ R , 2 + 2 = 1 . (7.26)
a b

To find the points of intersection of the curve with the coordinate axes we shall
make by turn y = 0 and x = 0. It results:
• A (a, 0), A∀ (−a, 0) are situated on Ox,
• B (0, b), B ∀ (0, −b) are situated on Oy.
The segment
• A A∀ = 2a is called the major axis of the ellipse;
• B B ∀ = 2b is called the minor axis of the ellipse.
Their halves, i.e. OA = a and OB = b are the semi-axes of the ellipse. The points
A, A∀ , B, B ∀ are called the vertices of the ellipse.
From the Eq. (7.8) one deduces the explicit Cartesian equations of the ellipse:

b⎧ 2
y=± a − x 2 , x ∈ [−a, a] . (7.27)
a
To obtain the parametric equations of the ellipse, one proceeds (see [6], p. 378)
as follows (see Fig. 7.3):
(1) build two concentric circles with the radii a and respectively b, a > b;
(2) trace a semi- line through origin, which intersects the two circles in the points
A and respectively B;
(3) bulid through the points A and B some lines, that are parallel with the axes; the
intersection of these points will be a point M of the ellipse;
(4) denote by θ the angle formed by the radius OA with the axis Ox.

Fig. 7.3 Concentric circles

7.2 Conics on the Canonical Equations 243

We shall deduce that the parametric equations of the ellipse are:


x = a cos θ
, θ ∈ [0, 2π] . (7.28)
y = b sin θ

7.2.3 Hyperbola

As in the case of the ellipse, we shall also consider c > 0 a real positive number and
F, F ∀ two fixed points from the plane such that F F ∀ = 2c.
Definition 7.10 (see [5], p. 156). The hyperbola is the set of the points M of the
plane that satisfy the relation
 
 M F − M F ∀  = 2a = ct, (7.29)

i.e. that have constantly the difference of the distances to two fixed points.
−−∈
We choose F F ∀ as Ox axis and the mediator of the segment F F ∀ as Oy axis (see
Fig. 7.4).

Fig. 7.4 Hyperbola

M F, M F ∀ are the focus radii of the point M. The hyperbola admits a unique
center of symmetry O and two axes of symmetry Ox, Oy.
From Fig. 7.4 we can note that the hyperbola is an unbounded curve.
The points F (c, 0) and F ∀ (c, 0) are called the foci of the hyperbola and the
distance F F ∀ constitutes the focus distance of the hyperbola.
To find the hyperbola equation we shall transform analytically the Eq. (7.29). If
M (x, y), then the relation (7.29) becomes:

(x − c)2 + y 2 − (x + c)2 + y 2 = ±2a. (7.30)

244 7 Conics and Quadrics

We shall obtain:

(x + c)2 + y 2 = (x − c)2 + y 2 ± 4a (x − c)2 + y 2 + 4a 2 .

By reducing the similar terms and passing in the first part of all those terms that
don’t contain radical signs, we have:

cx − a 2 = ±a (x − c)2 + y 2 ;

hence ⎨ ⎩ ⎨ ⎩
c2 − a 2 x 2 − a 2 y 2 − a 2 c2 − a 2 = 0. (7.31)

By multiplying (7.30) with −1 and by changing the signs in the parentheses, we

note that we have obtained the same Eq. (7.23) from the case of ellipse. The difference
of the hyperbola towards the ellipse (where
 we have a > c) comes from the fact that
in the triangle M F F ∀ : from F F ∀ >  M F − M F ∀  = 2a we have c > a; so that
c2 − a 2 = b2 . Therefore, we can denote

c2 − a 2 = b2 . (7.32)

By dividing the relation (7.31) with a 2 b2 it results the implicit Cartesian equation
(7.9) of the hiperbola.
To find the points of intersection of the curve with the coordinate axes we shall
make by turn y = 0 and x = 0. It results that

• A (a, 0), A∀ (−a, 0) are situated on Ox,

• we don’t have any real point on the Oy axis; therefore the Oy axis doesn’t cross
the hyperbola.

That is why, the Ox axis is called the transverse axis and the Oy axis is called the
untransverse axis.
The points A, A∀ represent the vertices of the hyperbola.
From the Eq. (7.9) one deduces the explicit Cartesian equations of the hyperbola:

b⎧ 2
y=± x − a 2 , x ∈ (−∃, −a] ⇐ [a, ∃) . (7.33)
a
The hyperbola admits two oblique asymptotes:

b
y = ± x. (7.34)
a
From (7.9) we shall have:
⎨x y⎩⎨x y⎩
+ − = 1.
a b a b
7.2 Conics on the Canonical Equations 245

If x ∈ [a, ∃), by denoting


y
a + yb = e
x t et + e−t
−t ⇒x =a ;
a − b =e
x
2

it results that the parametric equations of the hyperbola are:


x = a · cosh t
, t ∈ R.
y = b · sinh t

If x ∈ (−∃, −a], by denoting


y
a + yb = −e
x t et + e−t
−t ⇒ x = −a ;
a − b = −e
x
2

it results that the parametric equations of the hyperbola are:


x = −a · cosh t
, t ∈ R.
y = −b · sinh t

Example 7.11 (see [5], p. 158). Determine the vertices, the foci and the asymptotes
of the hyperbola
2x 2 − 5y 2 − 8 = 0.

Solution
Writing the equation of the hyperbola in the form

x2 y2
− 8 − 1 = 0,
4 5

we deduce: ⎫
⎬ a2 = 4
b2 = 85
⎭ 2
c = a 2 + b2 = 5 .
28



The hyperbola vertices are: A (2, 0), A∀ (−2, 0) and its foci: F 2 5 , 0 ,
7


∀
F −2 5 , 0 . The equations of the hyperbola asymptotes are:
7


2
y=± x.
5
246 7 Conics and Quadrics

Using Sage, we achieve:

7.2.4 Parabola

Definition 7.12 (see [5], p. 182). The parabola is the set of the points from plane,
that are equidistant from a fixed line and a fixed point.
The fixed line is called the directrix of the parabola and the fixed point is called
the focus of the parabola.
To find the equation of the parabola we choose a Cartesian reference whose axes
are:
• the perpendicular from the focus F on the directrix d as Ox axis,
• the parallel to d going to the midway between the focus and the directrix d as Oy
axis.
We denote A = d⇒Ox. Let M be a point of the parabola and N be its projection
on the directrix (see Fig. 7.5).
7.2 Conics on the Canonical Equations 247

Fig. 7.5 Parabola

The parabola has no center of symmetry and it has a single axis of symmetry, Ox.
It is an unbounded curve.    
We denote AF = p; it results F 2p , 0 and A − 2p , 0 . If M is an arbitrary
point of the parabola, then within the Definition 7.12, the point M has to satisfy the
relation:
M F = M N. (7.35)

As ⎨ p⎩ p
MN = x − − =x+ ,
2 2
the relation (7.35) becomes:
⎨
p ⎩2 p
x− + y2 = x + ;
2 2

hence one achieves the implicit Cartesian equation of the parabola:

y 2 = 2 px. (7.36)

Remark 7.13 (see [1], p. 165). In the case when x ⇔ 0, the implicit Cartesian
equation of the parabola will become:

y 2 = −2 px.

The Ox axis cuts the parabola in the point O(0, 0) called the parabola vertex.
From the Eq. (7.36) one deduces the explicit Cartesian equations of the parabola:
⎧
y = ± 2 px, x ≺ 0, (7.37)

p being a positive number called the parabola parameter, which indicates its form.
As p is smaller, both the focus and the directrix one approach [6] by the Oy axis,
and the parabola one approaches by the Ox axis (when p ∈ 0 then the parabola
degenerates into the Ox axis).
248 7 Conics and Quadrics

As p is greater, both the focus and the directrix one depart [6] by the Oy axis,
and the parabola one approaches by the Oy axis (when p ∈ ∃ then the parabola
degenerates into the Oy axis).
The parametric equations of the parabola are:
 2
x = 2t p
, t ∈ R. (7.38)
y=t

Definition 7.14 (see [5], p. 187). The eccentricity denoted by e is an element which
characterizes the circle, the ellipse, the hiperbola and the parabola, representing the
ratio of the distances from an arbitrary point of the respective conic to the focus and
the directrix, namely:
 
• e = 0 characterizes the circle e = ac = a0 ,
 
• 0 < e < 1 characterizes the ellipse e = ac ,
 
• e > 1 characterizes the hyperbola e = ac ,
 
• e = 1 characterizes the parabola e = M MF
N .

7.3 Reducing to the Canonical Form of a Conic Equation

We propose to determine an orthonormal Cartesian reference, relative to which the

general equation from (7.2) of has one of the canonical forms from (7.8), (7.9)
or (7.12).
We can distinguish [1], [2] the following situations:
Case 1. δ →= 0, i.e. the conic admits an unique center of symmetry C (x0 , y0 ).
The stages that are necessary in this case to obtain a canonical equation of a conic are:
(1) attach the square form

Q (v) = a11 x 2 + 2a12 x y + a22 y 2 , (√) v = (x, y) ∈ R2 (7.39)

to the Eq. (7.2)

(2) write the matrix
 
a11 a12
A=
a12 a22

associated to the form Q relative to the canonical basis B of R2 and then build the
characteristic polynomial:
 
 a − λ a12 
P (λ) =  11  = λ2 − I λ + δ. (7.40)
a12 a22 − λ 
7.3 Reducing to the Canonical Form of a Conic Equation 249

(3) make a change of an orthonormal reference such that the center of symmetry to
be the origin of the new reference.
 
The transition from the coordinates (x, y) to the coordinates x ∀ , y ∀ in the
new reference is achieved by a translation of the vector OC, characterized by the
equations: 
x = x0 + x ∀
(7.41)
y = y0 + y ∀ .

By this transformation, the general Eq. (7.2) of the conic becomes:

 2     2  
a11 x0 + x ∀ + 2a12 x0 + x ∀ y0 + y ∀ + a22
2
y0 + y ∀ + 2b1 x0 + x ∀
 
+ 2b2 y0 + y ∀ + c = 0,

i.e.
a11 x ∀2 + 2a12 x ∀ y ∀ + a22
2 ∀2
y + c∀ = 0, c∀ = f (x0 , y0 ) . (7.42)

(4) determine an orthonormal basis B ∀ , composed of the eigenvectors corresponding

to the eigenvalues λ1 and λ2 of the matrix A.
(5) rewrite the Eq. (7.42) relative to the basis B ∀ ; it will become

λ1 x ∀∀2 + λ2 y ∀∀2 + c∀ = 0. (7.43)

 
The transition to the new coordinates x ∀∀ , y ∀∀ one achieves through the relation:
   
x∀ x ∀∀
= M(B,B ∀ ) · . (7.44)
y∀ y ∀∀

We can note that for the Eq. (7.43) we have

⎫  
  λ1 0 

 δ=   = λ1 λ2
 
⎬  0 λ2 
 λ1 0 0 
  

 λ =  0 λ2 0  = λ1 λ2 c∀ .

⎭  0 0 c∀ 

As δ →= 0 it results that
λ
= c∀ .
δ
(6) obtain the canonical equation of the conic:

λ
λ1 x ∀∀2 + λ2 y ∀∀2 + = 0. (7.45)
δ
250 7 Conics and Quadrics

Remark 7.15 (see [2]). If λ1 and λ2 have the same sign and λδ has an opposite one,
then from (7.45) one obtains an ellipse. If λ1 and λ2 have different signs then from
(7.45) one obtains an hyperbola.
Example 7.16 (see [4], p. 209). Let be the conic

π : 9x 2 − 4x y + 6y 2 + 16x − 8y − 2 = 0.

Bring it to the canonical form, pointing the necessary reference changes, recognize
the obtained conic and plot its graph.
Solution
Using (7.4), we have:

I = 9 + 6 = 15,
 
 9 −2 
δ=   = 50 > 0
−2 6 
 
 9 −2 8 
 
λ =  −2 6 −4  = −500 →= 0.
 8 −4 −2 

As
• δ > 0, the conic π has an elliptic type
• λ →= 0, the conic π is non-degenerate
• δ →= 0, the conic π admits an unique center of symmetry.
The center of the conic is given by the system:

18x − 4y + 16 = 0
−4x + 12y − 8 = 0,
 
i.e. it is the point C − 45 , 25 .
We shall make a change of an orthonormal reference so that the center of symmetry
to be the origin of the new reference:
 
x∀ = x + 4
x = − 45 + x ∀
5 ⇔
y∀ = y − 2
5 y = 25 + y ∀ .

By this transformation, the equation of the conic becomes:

 2     2  
4 4 2 2 4
9 − + x∀ − 4 − + x∀ + y∀ + 6 + y ∀ + 16 − + x ∀
5 5 5 5 5
 
2
−8 + y∀ − 2 = 0
5
7.3 Reducing to the Canonical Form of a Conic Equation 251

or
π : 9x ∀2 − 4x ∀ y ∀ + 6y ∀2 − 10 = 0. (7.46)

We note that
 
(7.43) 4 2
c∀ = f − , ,
5 5

where

f (x, y) = 9x 2 − 4x y + 6y 2 + 16x − 8y − 2.

The matrix associated to the quadratic form (7.46) is

 
9 −2
A= .
−2 6

As the characteristic polynomial associated to the matrix A is

(7.40)
P (λ) = λ2 − 15λ + 50 = (λ − 10) (λ − 5)

it results the eigenvalues: λ1 = 5, λ2 = 10.

The eigensubspace associated to the eigenvalue λ1 will be
 
Vλ1 = v ∈ R2 |Av = λ1 v .

We achieve:
    
9 −2 v1 v1
=5 ;
−2 6 v2 v2
therefore

9v1 − 2v2 = 5v1
⇒ 2v1 = v2 ;
−2v1 + 6v2 = 5v2

it results
⎫ ⎝
⎬ ⎞
Vλ1 = v ∈ R2 | v = v1 (1, 2), (√) v1 ∈ R .
⎭  ⎪⎛ ⎜ ⎟
w
 
We shall obtain the orthonormal basis B1 = f 1 , where:
 
w 1 2
f1 = = ≤ ,≤ .
≡w≡ 5 5
252 7 Conics and Quadrics

The eigensubspace associated to the eigenvalue λ2 will be

 
Vλ2 = u ∈ R2 |Au = λ2 u .

We achieve:
    
9 −2 u1 u1
= 10 ;
−2 6 u2 u2

therefore

9u 1 − 2u 2 = 10u 1
⇒ u 1 = −2u 2 ;
−2u 1 + 6u 2 = 10u 2

it results
⎫ ⎝

⎬ 
⎞
Vλ2 = u ∈ R |u = u 2 (−2, 1), (√) u 2 ∈ R .
2

⎭  ⎪⎛ ⎜ 
⎟
z

 
We shall obtain the orthonormal basis B2 = f 2 , where:
 
z 2 1
f2 = = −≤ , ≤ .
≡z≡ 5 5

It will result  
B = B1 ⇐ B2 = f 1 , f 2 .

The equation of the conic becomes:

π : 5x ∀∀2 + 10y ∀∀2 − 10 = 0.

 
The transition to the new coordinates x ∀∀ , y ∀∀ one achieves through the relation
(7.44), where ⎠ ⎦
≤1 ≤2
M(B,B ∀ ) = 5 5 ;
− ≤ ≤1
2
5 5

therefore 
x∀ = ≤1 x ∀∀ + ≤2 y ∀∀
5 5 (7.47)
y∀ = − ≤2 x ∀∀ + ≤1 y ∀∀ .
5 5

The conic equation will have the canonical form

7.3 Reducing to the Canonical Form of a Conic Equation 253

x ∀∀2
π: + y ∀∀2 − 1 = 0
2
and it represents an ellipse.
The ellipse axes have the equations x ∀∀ = 0 and respectively y ∀∀ = 0.
Solving the system from (7.47) we deduce
⎫
⎬ x ∀∀ = x ∀≤
−2y ∀
5 (7.48)
∀ +y ∀
⎭ y ∀∀ = 2x≤
.
5

Therefore, the axes of the ellipse have the equations:


x ∀ − 2y ∀ = 0
(7.49)
2x ∀ + y ∀ = 0.

Taking into account that


x∀ = x + 4
5
y∀ = y − 2
5

we deduce that the ellipse axes will have the equations:

  
x + 45 −2 y − 25 = 0
2 x + 45 + y − 25 = 0,

i.e.

x − 2y + 85 = 0
(7.50)
2x + y + 65 = 0.
 
The ellipse center will be O∀∀ x ∀∀ = 0, y ∀∀ = 0 . We have:
⎫ ∀ ∀
 ⎬x≤−2y
x ∀∀
= 0 (7.48) = 0 (7.49)
∀∀ ⇔ 5
∀ +y ∀ ⇔
y =0 ⎭ ≤ =0
2x
5
 ∀ 
x − 2y ∀ = 0 (7.50) x − 2y + 85 = 0
⇔
2x ∀ + y ∀ = 0 2x + y + 65 = 0

x = − 45
⇔
y = 25 ;
 
hence the ellipse center will be O∀∀ x = − 45 , y = 25 .
The vertices of the ellipse will be:
254 7 Conics and Quadrics
⎨ ≤ ⎩ ⎨ ≤ ⎩
A x ∀∀ = 2, y ∀∀ = 0 , A∀ x ∀∀ = − 2, y ∀∀ = 0 ,
   
B x ∀∀ = 0, y ∀∀ = 1 , B ∀ x ∀∀ = 0, y ∀∀ = −1 .

We deduce
⎫ ∀ ∀ ≤
 ≤ ⎬x≤−2y
x ∀∀ = 2 (7.48) = 2 (7.49)
⇔ ∀
5
∀ ⇔
y ∀∀ = 0 ⎭ 2x≤+y = 0
5
 ∀ ≤  ≤
x − 2y ∀ = 10 (7.50) x − 2y + 85 = 10
⇔
2x ∀ + y ∀ = 0 2x + y + 65 = 0
⎫ ≤
⎬ x = 10−4
⎨ 5 ⎩ ≤
⇔
⎭ 2 1− 10
y= 5 ;
⎫ ≤
 ≤ ⎬ x = − 10−4
x ∀∀ =− 2 ⎨ ≤ 5 ⎩
⇔
y ∀∀ = 0 ⎭ 2 1+ 10
y= 5 ;

⎫ ∀ ∀
 ⎬x≤−2y
x ∀∀
= 0 (7.48) = 0 (7.49)
∀∀ ⇔ 5
∀ +y ∀ ⇔
y =1 ⎭ ≤ =1
2x
5
 ∀ 
x − 2y ∀ =≤0 (7.50) x − 2y + 85 =≤0
⇔
2x ∀ + y ∀ = 5 2x + y + 65 = 5
 ≤
x=2 ≤ 5−4
⇔ 5
y= 5 ;2+ 5

  ≤
x ∀∀ = 0 x = −2 5≤5−4
⇔
y ∀∀ = −1 y = 2−5 5 .

We obtain the following vertices of the ellipse:

 ⎨ ≤ ⎩  ⎨ ≤ ⎩
≤ 2 1 − 10 ≤ 2 1 + 10
10 − 4  , A∀  x = − 10 − 4
A x = ,y = ,y = 
5 5 5 5
⎠ ≤ ≤ ⎦ ⎠ ≤ ≤ ⎦
2 5−4 2+ 5 ∀ −2 5 − 4 2− 5
B x= ,y = ,B x = ,y = .
5 5 5 5

The obtained ellipse has the following graphic representation:

7.3 Reducing to the Canonical Form of a Conic Equation 255

The solution in Sage is:

256 7 Conics and Quadrics
7.3 Reducing to the Canonical Form of a Conic Equation 257

Case 2. δ = 0, i.e. namely the conic hasn’t an unique center of symmetry.

The stages that are necessary in this case to obtain a canonical equation of a conic
are (1), (2), (4) from the Case 1 followed by:
(1’) relative to the basis B ∀ , the Eq. (7.1) will become

λ1 x ∀2 + λ2 y ∀2 + 2b1∀ x ∀ + 2b2∀ y ∀ + c∀ = 0. (7.51)

 
The transition to the new coordinates x ∀ , y ∀ one achieves through the relation:

   ∀
x x
= M(B,B ∀ ) · . (7.52)
y y∀

We can note that for the Eq. (7.51) we have

 
λ 0 
δ =  1  = λ1 λ2 .
0 λ2 

We suposse λ2 = 0 for δ = 0. We shall obtain the equation

λ1 x ∀2 + 2b1∀ x ∀ + 2b2∀ y ∀ + c∀ = 0. (7.53)

(2’) Form a perfect square

 
b∀ 2 b∀2
λ1 x ∀ + 1 + 2b2∀ y ∀ + c∀ − 1 = 0.
λ1 λ1

(3’) Make the change of coordinates


b∀
x ∀∀ = x ∀ + λ11
y ∀∀ = y ∀ ;

it results the equation:

λ1 x ∀∀2 + 2b2∀ y ∀∀ + c∀∀ = 0.

We note that:
 
 λ1 0 0 
 
λ =  0 0 b2∀  = −λ1 b2∀2 .
 0 b∀ c∀∀ 
2

(4’) λ →= 0 ⇔ b2∀ →= 0; therefore, we can write

 
c∀∀
λ1 x ∀∀2 + 2b2∀ y ∀∀ + ∀ = 0.
2b2
258 7 Conics and Quadrics

(5’) Making the change of coordinates


X = x ∀∀
c∀∀
Y = y ∀∀ + 2b ∀
2

it results the canonical equation

λ1 X 2 + 2b2∀ Y = 0, (7.54)

which corresponds to a parabola.

Example 7.17 (see [2]). Let be the conic

π : 4x 2 − 4x y + y 2 − 3x + 4y − 7 = 0.

Bring it to the canonical form, pointing the necessary reference changes, recognize
the obtained conic and plot its graph.
Solution
Using (7.4), we have:

I = 4 + 1 = 5,
 
 4 −2 
δ=  =0
−2 1 
 
 4 −2 −3/2 
  25
λ =  −2 1 2  = − →= 0.
 −3/2 2 −7  4

As
• λ →= 0, the conic π is non-degenerate
• δ = 0, the conic π doesn’t admit an unique center of symmetry.
The matrix associated to the quadratic form is
 
4 −2
A= .
−2 1

As the characteristic polynomial associated to the matrix A is

(7.40)
P (λ) = λ2 − 5 = λ (λ − 5)

it results the eigenvalues: λ1 = 0, λ2 = 5.

The eigensubspace associated to the eigenvalue λ1 will be
 
Vλ1 = u ∈ R2 | Au = λ1 u .
7.3 Reducing to the Canonical Form of a Conic Equation 259

We achieve:     
4 −2 u1 u1
=0· ;
−2 1 u2 u2

therefore 
4u 1 − 2u 2 = 0
⇒ u 2 = 2u 1 ;
−2u 1 + u 2 = 0

it results ⎫ ⎝
⎬ ⎞
Vλ1 = v ∈ R2 |u = u 1 (1, 2), (√) u 1 ∈ R .
⎭  ⎪⎛ ⎜ ⎟
w
 
We shall obtain the orthonormal basis B1 = f 1 , where:
 
w 1 2
f1 = = ≤ ,≤ .
≡w≡ 5 5

The eigensubspace associated to the eigenvalue λ2 will be

 
Vλ2 = v ∈ R2 |Av = λ2 v .

We achieve:
    
4 −2 v1 v
=5 1 ;
−2 1 v2 v2

therefore

4v1 − 2v2 = 5v1
⇒ v1 = −2v2 ;
−2v1 + v2 = 5v2

it results
⎫ ⎝

⎬ 
⎞
Vλ2 = v ∈ R |v = v2 (−2, 1), (√) u 2 ∈ R .
2

⎭  ⎪⎛ ⎜ 
⎟
z

 
We shall obtain the orthonormal basis B2 = f 2 , where:
 
z 2 1
f2 = = −≤ , ≤ .
≡z≡ 5 5

It will result
 
B = B1 ⇐ B2 = f 1 , f 2 .
260 7 Conics and Quadrics

Relative to the basis B ∀ , the conic equation will become:

   
∀2 ∀2 1 ∀ 2 ∀ 2 ∀ 1 ∀
λ1 x + λ2 y − 3 ≤ x − ≤ y + 4 ≤ x + ≤ y − 7 = 0
5 5 5 5
or
11 2
5y ∀2 − ≤ x ∀ − ≤ y ∀ − 7 = 0. (7.55)
5 5

The transition to the new coordinates (x ∀ , y ∀ ) one achieves through the relation
(7.52), where ⎠ 1 ⎦
2 ≤ ≤
M(B,B ∀ ) = 5 5 ;
− ≤ ≤1
2
5 5

therefore

x= ≤1 x ∀ + ≤2 y ∀
5 5
y= − ≤2 x ∀ + ≤1 y ∀ .
5 5

We will form a perfect square, writing the Eq. (7.55) in the form
 2
1∀ 11 176
5 y − ≤ − ≤ x∀ − = 0.
5 5 5 25

Making the change of coordinates


x ∀∀ = x ∀
y ∀∀ = y∀ − 1
≤
5 5

it results the equation

11 176
5y ∀∀2 − ≤ x ∀∀ − = 0.
5 25

We can write:
⎠ ≤ ⎦
∀∀2 11 ∀∀ 176 5
5y −≤ x − · = 0.
5 25 11

Making the change of coordinates

 ≤
X = x ∀∀ − 176
25 ·
5
11 ,
Y = y ∀∀

the conic equation will have the canonical form

7.3 Reducing to the Canonical Form of a Conic Equation 261

11
π : Y2 = ≤ X
5

and it represents a parabola.

Solving this problem in Sage we achieve:
262 7 Conics and Quadrics

 The obtained
 parabola can be represented not only in the (Ox y) plane, but in the
Ox ∀∀ y ∀∀ plane, too:

7.4 General Equation of a Quadric

Definition 7.18 (see [7], p. 378). We consider the function f : R3 ∈ R,

f (x, y, z) = a11 x 2 + a22 y 2 + a33 z 2 + 2a12 x y (7.56)

+ 2a13 x z + 2a23 yz + 2b1 x + 2b2 y + 2b3 z + c.

An algebraic surface of the second order or a quadric is the set ρ of the points
M (x, y, z) from the space, whose coordinates relative to an orthonormal Cartesian
reference verify general the equation

f (x, y, z) = 0, (7.57)
7.4 General Equation of a Quadric 263

where the coefficients a11 , a22 , a33 , a12 , a13 , a23 , b1 , b2 , b3 , c are some real con-
%
3 %3
stants, with ai2j →= 0; therefore
i=1 j=1

 
ρ = M (x, y, z) | (x, y, z) ∈ R3 , f (x, y, z) = 0 . (7.58)

Definition 7.19 (see [2]). The invariants of a quadric are those expressions formed
with the coefficients of the conic equation, that keep the same value to the changes
of an orthonormal reference.
Proposition 7.20 (see [3], p. 69). We can assign four invariants of the quadric from
(7.58):

1) I = a11 + a22 + a33 (7.59)

     
a a  a a  a a 
2) J =  11 12  +  11 13  +  22 23 
a12 a22 a13 a33 a23 a33
 
 a11 a12 a13 
 
3) δ =  a12 a22 a23 
 a13 a23 a33 
 
 a11 a12 a13 b1 
 
a a a b 
4) λ =  12 22 23 2  .
 a13 a23 a33 b3 
 b1 b2 b3 c 

The invariant λ determines the nature of the quadric. Thus, if:

• λ →= 0, the quadric is called non- degenerate (the sphere, the ellipsoid, the hyper-
boloids and the paraboloids)
• λ = 0, the quadric is called degenerate (the cone, the cylinders).
As in the case of the conics, the center of symmetry for a quadric ρ, having the
Eq. (7.57) is the solution of the linear system:

⎫∂f ⎫

⎬ ∂x = 0 ⎬ a11 x + a12 y + a13 z + b1 = 0
∂f
∂ = 0 ⇔ a12 x + a22 y + a23 z + b2 = 0 (7.60)

⎭∂f
y ⎭
a13 x + a23 y + a33 z + b3 = 0.
∂z = 0

Theorem 7.21 (see [4], p. 176). Any quadric has one of the following canonical
forms:
(1) imaginary ellipsoid:

x2 y2 z2
+ + + 1 = 0, a ≺ b ≺ c > 0 (7.61)
a2 b2 c2
264 7 Conics and Quadrics

(2) real ellipsoid:

x2 y2 z2
+ + − 1 = 0, a ≺ b ≺ c > 0 (7.62)
a2 b2 c2
(3) hyperboloid of one sheet:

x2 y2 z2
2
+ 2 − 2 − 1 = 0, a ≺ b > 0, c > 0 (7.63)
a b c
(4) hyperboloid of two sheets

x2 y2 z2
+ − + 1 = 0, a ≺ b > 0, c > 0 (7.64)
a2 b2 c2
(5) imaginary second-order cone

x2 y2 z2
+ + = 0, a ≺ b > 0, c > 0 (7.65)
a2 b2 c2
(6) real second-order cone

x2 y2 z2
2
+ 2 − 2 = 0, a ≺ b > 0, c > 0 (7.66)
a b c
(7) elliptic paraboloid
x2 y2
+ − 2z = 0, a ≺ b > 0 (7.67)
a2 b2
(8) hyperbolic paraboloid

x2 y2
− − 2z = 0, a > 0, b > 0 (7.68)
a2 b2
(9) imaginary elliptic cylinder

x2 y2
+ + 1 = 0, a ≺ b > 0 (7.69)
a2 b2

(10) real elliptic cylinder

x2 y2
+ − 1 = 0, a ≺ b > 0 (7.70)
a2 b2
(11) hyperbolic cylinder

x2 y2
2
− 2 − 1 = 0, a > 0, b > 0 (7.71)
a b
7.4 General Equation of a Quadric 265

(12) pair of imaginary planes, having the intersection a real line

x2 y2
+ = 0, a, b > 0 (7.72)
a2 b2
(13) pair of secant planes
x2 y2
2
− 2 = 0, a, b > 0 (7.73)
a b
(14) parabolic cylinder
x2
− 2z = 0, a > 0 (7.74)
a2
(15) pair of parallel imaginary planes

x2
+ 1 = 0, a > 0 (7.75)
a2
(16) pair of parallel real planes

x2
− 1 = 0, a > 0 (7.76)
a2
(17) pair of confounded planes
x 2 = 0. (7.77)

The General Table of the Quadric Discussion

(I) If λ →= 0 then [8] for:

(1) δ →= 0, the quadric has the canonical equation

λ
S1 x 2 + S2 y 2 + S3 z 2 + = 0, (7.78)
δ
where S1 , S2 , S3 are the roots of the secular equation

S 3 − I S 2 + J S − δ = 0. (7.79)

(2) δ = 0, the quadric has the canonical equation


λ
S1 x + S2 y = 2 −
2 2
z. (7.80)
J

If the coefficients of (7.78) have the signs:

• + + ++ the quadric is an imaginary ellipsoid
266 7 Conics and Quadrics

• + + +− the quadric is a real ellipsoid

• + + −− the quadric is a hyperboloid of one sheet
• + − −− the quadric is a hyperboloid of two sheets.
If the coefficients of (7.80) have the signs:
• ++ the quadric is an elliptic paraboloid
• +− the quadric is a hyperbolic paraboloid
(II) If λ = 0 then for:
(1) δ →= 0, the quadric has the canonical equation

S1 x 2 + S2 y 2 + S3 z 2 = 0. (7.81)

(2) δ = 0, the quadric is a cylinder or a pair of planes.

If the coefficients of (7.81) have the signs:
• + + + the quadric is an imaginary cone
• + + − the quadric is a real cone.

Example 7.22 (see [8]). Determine the nature of the following quadrics:
(a) x 2 + y 2 + z 2 + 7x y + yz − 6z = 0
(b) 36x 2 + y 2 + 4z 2 + 72x + 6y − 40z + 109 = 0.
Solution
(a) We have

f (x, y, z) = x 2 + y 2 + 7x y + yz − 6z.

We identify the coefficients of the quadric:

⎫
⎬ a11 = 1, a22 = 1, a33 = 1, a12 = 27 , a13 = 0, a23 = 1
2
b1 = 0, b2 = 0, b3 = −3
⎭
c = 0.

We shall obtain:

I = 1 + 1 + 1= 2   
 1 7/2   1 0   1 1/2 
J =  + +  = − 19
7/2 1   0 1   1/2 1  2
 
 1 7/2 0 
  23
δ =  7/2 1 1/2  = − →= 0
 0 1/2 1  2
7.4 General Equation of a Quadric 267
 
 1 7/2 0 0 

 7/2 1 1/2 0  405
λ =  = →= 0.
 0 1/2 1 −3  4
 0 0 −3 0 

As δ →= 0, the quadratic admits an unique center of symmetry; the center coordi-

nates result solving the system (7.60), which becomes:
⎫
⎬ 2x + 7y = 0  
21 3 135
7x + 2y + z = 0 ⇒ C − , , .
⎭ 46 23 46
y + 2z − 6 = 0

We shall solve the secular Eq. (7.79), being:

⎫
19 23 ⎬ S1 = −2.53548
S 3 − 3S 2 − S+ =0⇒ S2 = 4.53516
2 2 ⎭
S3 = 1.

The quadric equation will have the canonical form:

x 2 + 4.53516y 2 − 2.53548z 2 − 8.80469 = 0,

i.e. it is an one-sheeted hyperboloid.

The solution in Sage is:
268 7 Conics and Quadrics

(b) We have

f (x, y, z) = 36x 2 + y 2 + 4z 2 + 72x + 6y − 40z + 109.

The quadric invariants will be:

I = 36 + 1 + 4 = 41  
 36 0   36 0   1 0 
J =   +   +   = 184
0 1  0 4 0 4
 
 36 0 0 
 
δ =  0 1 1/2  = 144 →= 0
0 0 4 
 
 36 0 0 36 
 
 0 1 1/2 3 
λ=   = −5184 →= 0.

 0 0 4 −20 
 36 3 −20 109 

The secular equation will be:

7.4 General Equation of a Quadric 269
⎫
⎬ S1 = 1
S 3 − 41S 2 + 184S − 144 = 0 ⇒ S2 = 36
⎭
S3 = 4.

The quadric equation will have the canonical form:

x 2 + 4y 2 + 36z 2 − 36 = 0

or
x2 y2
+ + z2 − 1 = 0
36 9
i.e. it is a real ellipsoid.
Using Sage, we shall have:
270 7 Conics and Quadrics

7.5 Quadrics on Canonical Equations

If
• δ →= 0 the quadric has an unique centre of simmetry; ex: sphere, ellipsoid, cone,
hyperboloids
• δ = 0 the quadric is called without center; ex: elliptic and hyperbolic paraboloid.

7.5.1 Sphere
Definition 7.23 (see [1], p. 185). The sphere is the set of the points from space
equally distant from a fixed point called the center of the sphere, the distance from
the center to the points of the sphere is called the radius of the sphere.
We shall report the sphere plane at an orthonormal Cartesian reference.
Let C (a, b, c) be center of the sphere and M (x, y, z) an arbitrary point of the
sphere (see Fig. 7.6).

Fig. 7.6 Sphere

7.5 Quadrics on Canonical Equations 271

Remark 7.24 (see [2]). The sphere is a quadric of rotation which is obtained by
rotating a circle (semicircle) around one of its diameter.
From the Definition 7.23 it results that the distance between C and M is constant
and equal to the radius R of the sphere:

C M = R,

i.e.

(x − a)2 + (y − b)2 + (z − c)2 = R ⇔

(x − a)2 + (y − b)2 + (z − c)2 = R 2 . (7.82)

If we open the squares in (7.82) we get the general Cartesian equation of the
sphere

x 2 + y 2 + z 2 − 2ax − 2by − 2cz + a 2 + b2 + c2 − R 2 = 0. (7.83)

To obtain other forms of the equation of a sphere we introduce the spherical

coordinates: (ρ, θ, ϕ) (see Fig. 7.7).

Fig. 7.7 The spherical coor-

dinates of a point from space

The relations between the Cartesian coordinates (x, y, z) of a point M from space
and its spherical coordinates (ρ, θ, ϕ) are:
272 7 Conics and Quadrics
⎫
⎬ x = ρ sin θ cos ϕ
y = ρ sin θ sin ϕ (7.84)
⎭
z = ρ cos θ,

where:
• ρ ≺ 0 is the distance from the point M to the origin of axes,
• θ, θ ∈ [0, π] is angle made by the position vector of the point M with the Oz axis,
• ϕ, ϕ ∈ [0, 2π] means the angle made by the projection of the position vector of
the point M on the plane (xOy) with the Ox axis.
Remark 7.25 (see [6], p. 667). Each triplet of spherical coordinate corresponds to a
point, but not any point corresponds to a triplet, as is the case when M is on Oz or
in origin.
If we make a change of an orthonormal reference, such that C (a, b, c) constitutes
the origin of thenew reference,
 then the transition from the coordinates (x, y, z) to
the coordinates x ∀ , y ∀ , z ∀ in the new reference is achieved by a translation of the
vector OC, characterized by the equations:

⎫
⎬ x = a + x∀
y = b + y∀
⎭
z = c + z∀.

The parametric equations of the sphere with the center C (a, b, c) and the radius
ρ ≺ 0 will be:

⎫
⎬ x = a + ρ sin θ cos ϕ
y = b + ρ sin θ sin ϕ , θ ∈ [0, π] , ϕ ∈ [0, 2π] . (7.85)
⎭
z = c + ρ cos θ.

Example 7.26 (see [9], p. 104). Write the equation of the sphere with the center on
the line:
x y−1 z+2
d: = = ,
1 −1 1
≤
having the radius R = 2 and passing through the point A (0, 2, −1).
Solution
We have
x y−1 z+2
= = = t, t ∈ R.
1 −1 1

The parametric equations of the straight line d will be:

7.5 Quadrics on Canonical Equations 273
⎫ ⎫
⎬ x =t ⎬ x =t
y − 1 = −t ⇒ y = 1 − t
⎭ ⎭
z+2=t z = t − 2.

As the center of the sphere is situated on the straight line d, it results that the point
C (t, 1 − t, t − 2) is the center of the sphere.
From the condition
2
C A = R 2

we deduce

3t 2 = 0 ⇒ t = 0.

It follows that C (0, −1, 2) is the the center of the sphere and the equation of the
sphere will be

S : x 2 + (y − 1)2 + (z + 2)2 = 2.

We shall give a solution in Sage, too:

274 7 Conics and Quadrics

7.5.2 Ellipsoid

Definition 7.27 (see [7], p. 351 and [10], p. 131). The ellipsoid is the set of the
points in space, whose coordinates relative to an orthonormal reference check the
Eq. (7.62), where a, b, c are some strictly positive real numbers called the semi-axes
of the ellipsoid.
We consider the orthonormal reference, relative to which is given the ellipsoid
equation.
To plot the ellipsoid we shall determine its intersections with: the coordinate axes,
the coordinate planes, the planes parallel to the coordinate planes.


y=0 x2
Ox : ⇒ 2 − 1 = 0 ⇒ x = ±a;
z=0 a

hence the ellipsoid crosses Ox in two points: A (a, 0, 0) and A∀ (−a, 0, 0).


x =0 y2
Oy : ⇒ 2 − 1 = 0 ⇒ y = ±b;
z=0 b

hence the ellipsoid crosses Oy in two points: B (0, b, 0) and B ∀ (0, −b, 0).


x =0 z2
Oz : ⇒ 2 − 1 = 0 ⇒ z = ±c;
y=0 c

hence the ellipsoid crosses Oz in two points: C (0, 0, c) and C ∀ (0, 0, −c).
The points A, A∀ , B, B ∀ , C, C ∀ are the vertices of the ellipsoid, while the symmetry
axes of the ellipsoid are Ox,Oy,Oz.
2 2
(Ox y) : z = 0 ⇒ ax 2 + by2 − 1 = 0 ⇒ an ellipse of semi-axes a and b.
x2 z2
(Ox z) : y = 0 ⇒ a2
+ c2
− 1 = 0 ⇒ an ellipse of semi-axes a and c.
y2 z2
(Oyz) : x = 0 ⇒ b2
+ c2
− 1 = 0 ⇒ an ellipse of semi-axes b and c.
The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:

x2 y2 k2
+ + − 1 = 0.
a2 b2 c2

If k ∈ (−c, c) then the intersections with the planes parallel to the plane (Ox y)
are some ellipses having the equations (Fig. 7.8):
7.5 Quadrics on Canonical Equations 275

x2 y2
⎨ ≤ ⎩2 + ⎨ ≤ ⎩2 − 1 = 0.
a
c 2 − k2 b
c 2 − k2
c c

Fig. 7.8 Ellipsoid

The ellipsoid has: an unique center of symmetry (the origin), symmetry axes (the
coordinate axes), planes of symmetry (the coordinate planes).
The parametric equations corresponding to an ellipsoid are:

⎫
⎬ x = a sin u cos v
y = b sin u sin v , u ∈ [0, π] , v ∈ [0, 2π] . (7.86)
⎭
z = c cos u

Remark 7.28 (see [10], p. 132). The sphere is a special case of ellipsoid, obtained if
all the semi-axes of the ellipsoid are equal between themselves. If two semi-axes are
equal, then one achieves a rotating ellipsoid, which can be generated by the rotation
of an ellipse around of an axis. For example, if a = b, then the ellipsoid is of rotation
around of Oz.
The next figure shows an ellipsoid built in Sage, using (7.86):
276 7 Conics and Quadrics

7.5.3 Cone

Definition 7.29 (see [10], p. 153). The cone of the second order is the set of the
points in space, whose coordinates relative to an orthonormal reference check the
Eq. (7.66).
We consider the orthonormal reference, relative to which is given the cone equa-
tion. To plot the cone we shall determine its intersections with: the coordinate axes,
the coordinate planes, the planes parallel to the coordinate planes.


y=0 x2
Ox : ⇒ 2 =0⇒x =0
z=0 a

hence the cone crosses Ox in origin. Similarly, the cone crosses Oy and O z in origin,
too. 2
2
(Ox y) : z = 0 ⇒ ax 2 + by2 = 0 ⇒ x = y = 0.
x2 z2
(Ox z) : y = 0 ⇒ a2
− c2
=0⇔

x
− z
=0
a c ⇒ two lines concurrent in the origin.
x
a + z
c =0

y2 z2
(Oyz) : x = 0 ⇒ b2
− c2
=0⇔

 y
− z
=0
by c ⇒ two lines concurrent in the origin.
b + z
c =0
7.5 Quadrics on Canonical Equations 277

The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:
x2 y2 k2
+ − = 0;
a2 b2 c2
therefore the intersections with the planes parallel to the plane (Ox y) are some
ellipses having the equations (Fig. 7.9):

x2 y2
 a 2 +  b 2 − 1 = 0.
ck ck

Fig. 7.9 Cone

The cone has an unique center of symmetry.

Remark 7.30 (see [10], p. 129). If a = b then one obtains the rotating cone, which
can be generated by the rotation of a conic (which represents two concurrent straight
lines) by equation

y2 z2
− =0 (7.87)
a2 c2
around the Oz axis.
278 7 Conics and Quadrics

The parametric equations corresponding to the cone of the second order are:
⎫
⎬ x = av cos u
y = bv sin u , u ∈ [0, 2π] , v ∈ R. (7.88)
⎭
z = ±c

We shall plot in Sage a cone of the second order, defined by the Eq. (7.66):

Definition 7.32 (see [10], p. 141). The one-sheeted hyperboloid is the set of the
points from space, whose coordinates relative to an orthonormal reference check the
Eq. (7.63).
We consider the orthonormal reference, relative to which is given the one-sheeted
hyperboloid equation.

y=0 x2
Ox : ⇒ 2 − 1 = 0 ⇒ x = ±a;
z=0 a

hence the one-sheeted hyperboloid crosses Ox in two points: A (a, 0, 0) and

A∀ (−a, 0, 0).

x =0 y2
Oy : ⇒ 2 − 1 = 0 ⇒ y = ±b;
z=0 b

hence the one-sheeted hyperboloid crosses Oy in two points: B (0, b, 0) and

B ∀ (0, −b, 0).

x =0 z2
Oz : ⇒ − 2 − 1 = 0 ⇒ z 2 = −c2 ;
y=0 c
7.5 Quadrics on Canonical Equations 279

hence the one-sheeted hyperboloid dosn’t cross Oz axis.

The points A, A∀ , B, B ∀ are called the vertices of the one-sheeted hyperboloid.
2 2
(Ox y) : z = 0 ⇒ π1 : ax 2 + by2 − 1 = 0 ⇒ an ellipse of semi-axes a and b.
y2 z2
(Oyz) : x = 0 ⇒ π2 : b2
− c2
− 1 = 0 ⇒ a hyperbola.
x2 z2
(Ox z) : y = 0 ⇒ π3 : a2
− − 1 = 0 ⇒ a hyperbola.
c2
The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:
x2 y2 k2
+ − − 1 = 0.
a2 b2 c2
The intersections with the planes parallel to the plane (Ox y) are some ellipses
having the equations:

x2 y2
⎨ ≤ ⎩2 + ⎨ ≤ ⎩2 − 1 = 0
a
c 2 + k2 b
c 2 + k2
c c

called clamp ellipse.

The one-sheeted hyperboloid is an unbounded quadric with an unique center of
symmetry (see Fig. 7.10) and that has the following parametric equations:
⎫ ≤
⎬ x = a ≤1 + u 2 cos v
y = b 1 + u 2 sin v , v ∈ [0, 2π] , u ∈ R. (7.89)
⎭
z = cu

Fig. 7.10 One-sheeted

hyperboloid

Remark 7.32 (see [10], p. 127) If a = b then the one-sheeted hyperboloid is of

rotation around of Oz, i.e. it can be generated by the rotation of the hyperbola
280 7 Conics and Quadrics

y2 z2
− −1=0 (7.90)
b2 c2
around of the Oz axis.
We shall use parametric equation (7.89) to represent the one-sheeted hyperboloid
in Sage:

7.5.4 Two-Sheeted Hyperboloid

Definition 7.33 (see [10], p. 137). The two-sheeted hyperboloid is the set of the
points from space, whose coordinates relative to an orthonormal reference check the
Eq. (7.64).
The number of the sheets is given by the number of the squares that have the same
sign with the free term.
We consider the orthonormal reference, relative to which is given the two-sheeted
hyperboloid equation. 
y=0 x2
Ox : ⇒ 2 = −1;
z=0 a

hence the two-sheeted hyperboloid doesn’t cross the Ox axis.


x =0 y2
Oy : ⇒ 2 = −1 ⇒
z=0 b

the two-sheeted hyperboloid doesn’t cross the Oy axis.

7.5 Quadrics on Canonical Equations 281

x =0 z2
Oz : ⇒ 2 = 1 ⇒ z = ±c;
y=0 c

hence the one-sheeted hyperboloid crosses in two points: C (0, 0, c) and C ∀ (0, 0, −c).
2 2 2 2
(Oyz) : x = 0 ⇒ by2 − cz 2 + 1 = 0 ⇒ π1 : cz 2 − by2 = 1 a hyperbola.
2 2 2 2
(Ox z) : y = 0 ⇒ ax 2 − cz 2 + 1 = 0 ⇒ π2 : cz 2 − ax 2 = 1 a hyperbola.
The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:
x2 y2 k2
+ = − 1.
a2 b2 c2
If k ∈ (−∃, −c) ⇐ (c, ∃) then the intersections with the planes parallel to the
plane (Ox y) are some ellipses having the equations:

x2 y2
⎨ ≤ ⎩2 + ⎨ ≤ ⎩2 − 1 = 0.
c c −k c c −k
a 2 2 b 2 2

The two-sheeted hyperboloid is an unbounded quadric with an unique center of

symmetry (see Fig. 7.11) and that has the following parameter equations:
⎫
⎬ x = a sinh u cos v
y = b sinh u sin v , v ∈ [0, 2π] , u ∈ R. (7.91)
⎭
z = ±c cosh u

Fig. 7.11 Two-sheeted hyper-

boloid
282 7 Conics and Quadrics

Remark 7.34 (see [10], p. 128) If a = b then the two-sheeted hyperboloid is of

rotation around of Oz, i.e. it can be generated by the rotation of the hyperbola

y2 z2
− +1=0 (7.92)
b2 c2
around of the Oz axis.
The following Sage code allows us to plot a two-sheeted hyperboloid:

7.5.5 Elliptic Paraboloid

Definition 7.35 (see [10], p. 150). The elliptic paraboloid is the set of the points in
space, whose coordinates relative to an orthonormal reference check the Eq. (7.67),
where a, b, z are some strictly positive real numbers.
We consider the orthonormal reference, relative to which is given the elliptic
paraboloid.

y=0 x2
Ox : ⇒ 2 = 0;
z=0 a

hence the elliptic paraboloid crosses the Ox axis in the origin.


x =0 y2
Oy : ⇒ 2 =0⇒
z=0 b

the elliptic paraboloid crosses the Oy axis in the origin.

7.5 Quadrics on Canonical Equations 283

x =0 z2
Oz : ⇒ 2 =0⇒
y=0 c

the elliptic paraboloid crosses the Oz axis in the origin.

2 2
(Ox y) : z = 0 ⇒ ax 2 + by2 = 0 ⇒ x = y = 0 ⇒ the intersection is the origin.
y2
(Oyz) : x = 0 ⇒ b2
− 2z = 0 ⇒ π1 : y 2 = 2b2 z a parabola.
x2
(Ox z) : y = 0 ⇒ a2
− 2z = 0 ⇒ π2 : x 2 = 2a 2 z a parabola.
The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:
x2 y2
+ = 2k.
a2 b2
If k > 0 then the intersections with the planes parallel to the plane (Ox y) are
some ellipses having the equations:

x2 y2
πλ : ⎨ ≤ ⎩2 + ⎨ ≤ ⎩2 = 1.
a 2k b 2k

The the elliptic paraboloid is an unbounded quadric without a center of symmetry

(see Fig. 7.12) and that has the following parameter equations:

⎫ ≤
⎬ x = a ≤2v cos u
y = b 2v sin u , u ∈ [0, 2π] , v > 0. (7.93)
⎭
z=v

Fig. 7.12 Elliptic paraboloid

284 7 Conics and Quadrics

Remark 7.36 (see [10], p. 128) If a = b then elliptic paraboloid is of rotation around
of Oz, i.e. it can be generated by the rotation of the parabola

y 2 = 2a 2 z (7.94)

around of the Oz axis.

We need the following Sage code to represent an elliptic paraboloid:

7.5.6 Hyperbolic Paraboloid

Definition 7.37 (see [10], p. 146). The hyperbolic paraboloid is the set of the
points in space, whose coordinates relative to an orthonormal reference satisfy the
Eq. (7.68).
We consider the orthonormal reference, relative to which is given the hyperbolic
paraboloid. 
y=0 x2
Ox : ⇒ 2 = 0;
z=0 a

hence the hyperbolic paraboloid crosses the Ox axis in the origin.


x =0 y2
Oy : ⇒ 2 =0⇒
z=0 b

the hyperbolic paraboloid crosses the Oy axis in the origin.

7.5 Quadrics on Canonical Equations 285

x =0 z2
Oz : ⇒ 2 =0⇒
y=0 c

the hyperbolic paraboloid crosses the Oz axis in the origin.

2 2 x
− by = 0
(Ox y) : z = 0 ⇒ ax 2 − by2 = 0 ⇔ ax y ⇒ two line concurrent in the
a + b =0
origin.
2
(Oyz) : x = 0 ⇒ − by2 − 2z = 0 ⇒ π1 : y 2 = −2b2 z a parabola with Oz as axis
of symmetry, pointing in the negative direction of the straight line Oz.
2
(Ox z) : y = 0 ⇒ ax 2 − 2z = 0 ⇒ π2 : x 2 = 2a 2 z a parabola with Oz as axis of
symmetry, pointing in the positive direction of the straight line Oz.
The intersections with the planes parallel to the plane (Ox y), by equation z = k,
can be determined from:
x2 y2
− = 2k.
a2 b2
If k > 0 then the intersections with the planes parallel to the plane (Ox y) are
some ellipses having the equations:

x2 y2
πλ : ⎨ ≤ ⎩2 − ⎨ ≤ ⎩2 = 1.
a 2k b 2k

The the elliptic paraboloid is an unbounded quadric without a center of symmetry

(see Fig. 7.12) and that has the following parameter equations (Fig. 7.13):
⎫ ≤
⎬ x = a ≤2v cos u
y = b 2v sin u , u ∈ [0, 2π] , v > 0. (7.95)
⎭
z = v cos 2u

Fig. 7.13 Hyperbolic

paraboloid
286 7 Conics and Quadrics

Remark 7.38 (see [6], p. 686). There is not a hyperbolic paraboloid of rotation; the
hyperbolic paraboloid is the only surface of second degree, which is not a surface of
rotation (because any section through a hyperbolic parabolid is not an ellipse).
The hyperbolic paraboloid is a surface of translation, this being obtained by the
translation of a parabola (which has the opening in the bottom)

y 2 = −2b2 z (7.96)

on a parabola (which has the opening upward)

x 2 = 2a 2 z. (7.97)

We shall illustrate a hyperbolic paraboloid, made in Sage:

7.6 Ruled Surfaces. Surface Generation

Definition 7.39 (see [1], p. 196). A surface that can be generated by moving a straight
line G, which one relieses on a given curve π from space is called a ruled surface.
The straight line G is called the generatrix (or generator) of the surface.
Definition 7.40 (see [2]). A family of straight lines, which has the property that
every straight line of the family can generate a surface represents the rectilinear
generators of the respective surface.
7.6 Ruled Surfaces. Surface Generation 287

Proposition 7.41 (see [4], p. 174). The straight line d passing through the point
A (x0 , y0 , z 0 ) and which has the director vector v = (l, m, n) constitutes the recti-
linear generators of the quadric from (7.58) if and only if the following conditions
are satisfied:

⎫
⎬ f (x0 , y0 , z 0 ) = 0
a11 l 2 + a22 m 2 + a33 n 2 + 2a12 lm + 2a13ln + 2a23 mn = 0 (7.98)
⎭ ∂f
l ∂x (x0 , y0 , z 0 ) + m ∂∂ yf (x0 , y0 , z 0 ) + n ∂∂zf (x0 , y0 , z 0 ) = 0.

The only ones ruler surfaces are the two non-degenerate quadrics: the one-sheeted
hyperboloid and the hyperbolic paraboloid.
Example 7.42 (see [4], p. 239). Determine the rectilinear generators of the quadric
by equation:

ρ : x 2 + y 2 + z 2 + 2x y − 2x z − yz + 4x + 3y − 5z + 4 = 0

passing through the point A (−1, −1, 1).

Solution
We check firstly the first condition from (7.99), i.e. if f (−1, −1, 1) = 0, where:

f (x, y, z) = x 2 + y 2 + z 2 + 2x y − 2x z − yz + 4x + 3y − 5z + 4.

We want to determine the vector v = (l, m, n) so that, it should be the director

vector of the rectilinear generatrix.
We shall check the the second condition of (7.98):

l 2 + m 2 + n 2 + 2lm − 2ln − mn = 0. (7.80)

As
⎫∂f ⎫∂f

⎬ ∂x (x, y, z) = 2x + 2y − 2z + 4 
⎬ ∂x (−1, −1, 1) = −2
∂f ∂f
∂y (x, y, z) = 2y + 2x − z + 3 ⇒ ∂ y (−1, −1, 1) = −2

⎭ ∂f 
⎭ ∂f
∂z (x, y, z) = 2z − 2x − y − 5 ∂z (−1, −1, 1) = 0,

the third condition of (7.98) becomes: l = −m. Substituting it in (7.80) we shall

achieve:
 
n=0 m = −n
n (m + n) = 0 ⇔ or
l = −m l = n.

We determined the two director vectors:

288 7 Conics and Quadrics
 
v 1 = (−m, m, 0) = m (−1, 1, 0) m=1,n=1 v 1 = (−1, 1, 0)
⇒
v 2 = (n, −n, n) = n (1, −1, 1) v 2 = (1, −1, 1) ;

therefore, the two rectilinear generators pass through the point A (−1, −1, 1). Their
equations are:
 x+1
= y+1
d1 : −1 1
z=1

and, respectively:
x +1 y+1 z−1
d2 : = = .
1 −1 1

The solution in Sage will be given, too:

7.6 Ruled Surfaces. Surface Generation 289

Proposition 7.43 (see [4], p. 176). The one-sheeted hyperboloid has two families of
rectilinear generators: 
x y
a − c = λ 1 − b
z
(7.81)
λ ax + cz = 1 + by

and
x  
−x
z
= λ 1 + by
a c , (√) λ ∈ R. (7.81)
λ a + cz = 1 − by

Proposition 7.44 (see [4], p. 176). The hyperbolic paraboloid has two families of
rectilinear generators:
 x y
a− b = 2λz (7.82)
λ ax + by = 1

and  y
a − b = λ , (√) λ ∈ R.
x
(7.83)
λ ax + by = 2z

All of the degenerate quadrics have generators. Therefore, the cylindrical surfaces
and the conical surfaces (including the quadrics by the cone type and the cylinder
type) are ruled surfaces.
Let V (a, b, c) be a fixed point and

f (x, y, z) = 0
π: (7.84)
g (x, y, z) = 0

be a given curve.
290 7 Conics and Quadrics

Definition 7.45 (see [11], p. 87). The surface generated by moving a straight line D
called the generatrix, passing through the fixed point V called vertix and one relieses
on the given curve π, called director curve is called conical surface (Fig. 7.14).

Fig. 7.14 A conical surface

We want to find the equation of the conical surface. An arbitrary straight line,
which passes through the point V has the equations:

x −a y−b z−c 2
= = , l + m 2 + n 2 →= 0.
l m n
We denote  l
=α
n , n →= 0. (7.85)
m
n =β

The straight line, which generates the conical surface (the generatrix) will be

x −a y−b z−c
= = . (7.86)
α β 1

The condition of supporting the generatrix on the director curve returns to the
algebraic condition of compatibility of the system:
⎫
⎬ f (x, y, z) = 0
g (x, y, z) = 0 (7.87)
⎭ x−a y−b
α = β = 1 .
z−c

Eliminating x, y, z from the equations of the system, the compatibility condition

become
ϕ (α, β) = 0. (7.88)

Eliminating now α and β from (7.86) and (7.88) it results that the requested
conical surface has the equation:
7.6 Ruled Surfaces. Surface Generation 291
 
x −a y−b
ϕ , = 0. (7.89)
z−c z−c

Example 7.46. Find the Cartesian equation of the conical surface, which has the
vertix V (2, 1, 3) and the director curve


z=0
π:
x 2 + y 2 = 9.

Solution
The generatrix of the conical surface has the equations:

x −2 y−1 z−3 2
G: = = , l + m 2 + n 2 →= 0.
l m n
Imposing the condition that the straight line G to support on the curve π we
deduce that the system
⎫ x−2
⎬ l = y−1 m = n
z−3

z=0 (7.90)
⎭
x 2 + y2 = 9

is compatible.
Using the notation (7.85) we achieve:

x −2 y−1 z−3
G: = = . (7.91)
α β 1

The system (7.90) becomes

⎫

 βx − αy = 2β − α
⎬
y = 1 − 3β
⇔ (2 − 3α)2 + (1 − 3β)2 = 9;

 x = 2 − 3α
⎭
x 2 + y2 = 9

therefore, the condition of compatibility is reduced to the relation

9α2 + 9β 2 − 12α − 6β − 4 = 0. (7.92)

From (7.91) we deduce 

α = x−2
z−3 (7.93)
β = y−1
z−3 .
292 7 Conics and Quadrics

Substituting (7.93) in (7.92) it results that

 2  2
x −2 y−1 x −2 y−1
9 +9 − 12 −6 − 4 = 0.
z−3 z−3 z−3 z−3

The Cartesian equation of the conical surface will be

9 (x − 2)2 + 9 (y − 1)2 − 12 (x − 2) (z − 3) − 6 (y − 1) (z − 3) − 4 (z − 3)2 = 0.

The solution in Sage is:

Definition 7.47 (see [11], p. 89). We call a cylindrical surface, the surface generated
by a straight line which remains parallel to a given direction and which supports on
a given curve, called the director curve (Fig. 7.15).
7.6 Ruled Surfaces. Surface Generation 293

Fig. 7.15 A cylindrical

surface

Let v = (l, m, n) be the given direction and π from (7.84) be the director curve.
An arbitrary straight line, by the director vector v has the equations:


nx − lz = λ
(7.94)
ny − mz = μ.

The condition of supporting the generator on the director curve returns to the
algebraic condition of compatibility of the system:

⎫

 f (x, y, z) = 0
⎬
g (x, y, z) = 0

 nx − lz = λ
⎭
ny − mz = μ.

Eliminating x, y, z from the equations of the system, the compatibility condition

becomes:

ϕ (λ, μ) = 0. (7.95)

Eliminating now λ and μ from (7.94) and (7.95) it results that the requested
cylindrical surface has the equation:

ϕ (nx − lz, ny − mz) = 0. (7.96)

Example 7.48 (see [4], p. 215). Write the equation of the cylindrical surface, which
has the director curve

z=0
π:
x 2 + y 2 + 2x − y = 0

and the generators parallel to the straight line by equation

294 7 Conics and Quadrics

x −1 y−2 z
d: = = .
1 −1 2

Solution
We note that v = (1, −1, 2) is the director vector of the straight line d. An arbitrary
straight line, by the director vector v has the equations:


x +y−3=λ
, (√) λ, μ ∈ R. (7.97)
2x − z − 2 = μ

The condition of supporting the generatrix on the director curve returns to the
algebraic condition of compatibility of the system:

⎫ 2
 x + y 2 + 2x − y = 0 ⎫ 2

⎬ ⎬ x + y 2 + 2x − y = 0
z=0
⇔ x +y−3=λ

 x +y−3=λ ⎭
⎭ 2x − 2 = μ.
2x − z − 2 = μ

From the third equation of the system we deduce

μ+2
x= ; (7.98)
2
substituting this expression of x in the second equation of the system we get:

μ+2 2λ − μ + 4
y =λ+3− = . (7.99)
2 2
If in the first equation of the system, we take into account of (7.98) and (7.99) we
achieve:

 2  2
μ+2 2λ − μ + 4 μ + 2 2λ − μ + 4
+ +2· − = 0,
2 2 2 2

i.e

(μ + 2)2 + (2λ − μ + 4)2 + 6μ − 4λ = 0. (7.100)

Eliminating μ and λ between (7.100) and (7.97) it follows that the cylindrical
surface equation is
7.6 Ruled Surfaces. Surface Generation 295

(2x − z)2 + (2y + z)2 + 8x − 4y − 6z = 0.

Solving with Sage this problem, we have:

Definition 7.49 (see [11], p. 91). We call a surface of rotation, a surface generated
by rotating a curve around of a straight line called the axis of rotation (Fig. 7.16).
296 7 Conics and Quadrics

Fig. 7.16 A surface of

rotation

The axis of rotation has the equations:

x − x0 y − y0 z − z0 2
= = , l + m 2 + n 2 →= 0
l m n
and the curve π, which one rotates being defined in (7.84).
The surface from the above figure can also result by the displacement of a circle,
parallel with itself, with variable radius, perpendicular to the axis of rotation and
which one relies on the given curve.
In order to satisfy the conditions (7.95), the circle equations will be achieved by
crossing a sphere with a variable radius and having the center on the rotation axis
with a family of planes, perpendicular on the rotation axis; therefore


(x − x0 )2 + (y − y0 )2 + (z − z 0 )2 = α2
G: (7.101)
lx + my + nz = β.

The condition of supporting the generatrix on the director curve returns to the
algebraic condition of compatibility of the system:

⎫

 f (x, y, z) = 0
⎬
g (x, y, z) = 0

 (x − x0 )2 + (y − y0 )2 + (z − z 0 )2 = α2
⎭
lx + my + nz = β.

Eliminating x, y, z from the equations of the system, the compatibility condition

becomes (7.88).
7.6 Ruled Surfaces. Surface Generation 297

Eliminating now α and β from (7.101) and (7.88) it results that the requested
cylindrical surface has the equation:
⎨ ⎩
ϕ (x − x0 )2 + (y − y0 )2 + (z − z 0 )2 , lx + my + nz = 0. (7.102)

7.7 Problems

1. Let be the points: A (−1, 4) , B (3, −2). Write the equation of the circle, which
has AB as a diameter.
Solution
Solving this problem in Sage, we have:
298 7 Conics and Quadrics

2. Determine the vertices and the semi-axes of the ellipse

2x 2 + 4y 2 − 5 = 0.

3. Let be the conic

π : 4x 2 − 12x y + 9y 2 − 2x + 3y − 2 = 0.

Bring it to the canonical form, indicating the required reference changes and
recognize the achieved conic.
4. Write the equation of the parabola which passes through the points:

O (0, 0) , A (1, 0) , B (0, 1) , C (2, 3) .

Solution
The solution in Sage is:
7.7 Problems 299

We shall also use Sage to represent the achieved parabola:

5. Let be the points A (−1, 1, 2) , B (1, 3, 3). Write the equation of the sphere with
center in the point and which passes through the point B.
6. Write the equations of the rectilinear generators

x2 y2
− = z,
16 4
that are parallel to the plane π : 3x + 2y − 4z = 0.
7. Find the rectilinear generators of the quadrics :

ρ : x 2 + 3y 2 + 4yz − 6x + 8y + 8 = 0.

Solution
we need the following Sage code to solve this problem:
300 7 Conics and Quadrics
7.7 Problems 301
302 7 Conics and Quadrics

8. Find the equation of the conical surface having as vertex the point V (−3, 0, 0)
and the director curve:
 2
3x + 6y 2 − z = 0
π:
x + y + z − 1 = 0.

9. Determine the equation of the cylindrical surface that has the director curve:

x 2 + y2 − 1 = 0
π:
z=0

and the generators parallel to the straight line

d : x = y = z.

10. Write the equation of the surface generated by rotating the parabola

y 2 = 2 px
p:
z=0

around the Ox axis.

Solution
A solution in Sage of this problem is:
References 303

References

1. V. Balan, Algebră liniară, geometrie analitică, ed (Fair Partners, Bucureşti, 1999)

2. I. Vladimirescu, M. Popescu, M. Sterpu, Algebră liniară şi geometrie analitică (Universitatea
din Craiova, Note de curs şi aplicaţii, 1993)
3. C. Udrişte, Aplicaţii de algebră, geometrie şi ecuaţii diferenţiale, ed (Didactică şi Pedagogică
R.A, Bucureşti, 1993)
4. I. Vladimirescu, M. Popescu, Algebră liniară şi geometrie analitică, ed (Universitaria, Craiova,
1993)
5. Gh. D. Simionescu, Geometrie analitică, ed (Didactică şi pedagogică, Bucureşti, 1968)
6. V. Postelnicu, S. Coatu, Mică enciclopedie matematică, ed (Tehnică, Bucureşti, 1980)
7. E. Murgulescu, S. Flexi, O. Kreindler, O. Sacter, M. Tîrnoveanu, Geometrie analitică şi
diferenţială, ed (Didactică şi pedagogică, Bucureşti, 1965)
8. V.T. Postelnicu, I.M. Stoka, Gh Vrânceanu, Culegere de probleme de geometrie analitică şi
proiectivă, ed (Tehnică, Bucureşti, 1962)
9. Gh Atanasiu, Gh Munteanu, M. Postolache, Algebr ă liniară, geometrie analitică şi diferenţială,
ecua ţii diferenţiale, ed (ALL, Bucureşti, 1998)
10. G. Mărgulescu, P. Papadapol, Curs de geometrie analitică, diferenţială şi algebră liniară
(Catedra de Matematici, 1976)
11. T. Didenco, Geometrie analitică şi diferenţială (Academia Militară, Bucureşti, 1977)
Index

A Dimension formula, 52
Algebraic multiplicity, 117 Direct sum, 52
Angle, 13, 14, 83–85, 139, 227 Distance, 13, 74, 75, 79, 80
Arc element of a curve, 201
Associated matrix, 97
Automorphism, 92 E
Eigensubspace, 110
Eigenval, 110
B Eigenvalue method, 180
Basis, 27 Eigenvector, 110, 111
Bilinear form, 165 Ellipse, 240, 241
Binormal, 201 Ellipsoid, 274
Binormal versor, 202 Elliptic paraboloid, 282, 283
Endomorphism, 92
Envelope, 214
C Euclidean coordinates, 11
Canonical basis, 169 Euclidean real vector space, 135
Cartesian reference, 10
Center of symmetry, 236
F
Change the orthonormal bases, 144
Family of curves, 214
Characteristic equation, 113
Fascicle axis, 73
Characteristic polynomial, 113
First Frenet formula, 205
Circle, 238
First fundamental form, 223
Common perpendicular, 80
Formulas of changing a vector coordinates,
Cone, 276
40
Conic, 235
Free vector, 3
Conical surface, 289, 290
Frenet Trihedron, 199
Coordinates, 36, 37
Cross product, 16, 17
Curvature, 206 G
Curvilinear coordinates, 218 Gauss-Lagrange, 169
Cylindrical surface, 292, 293 Geometric multiplicity, 117
Gram-Schmidt orthogonalization, 145

D
Defect, 97 H
Diagonalizable, 117 Hamilton-Cayley, 115

G. A. Anastassiou and I. F. Iatan, Intelligent Routines II, 305

Intelligent Systems Reference Library 58, DOI: 10.1007/978-3-319-01967-3,
© Springer International Publishing Switzerland 2014
306 Index

Hyperbola, 243 Parallelogram rule, 5

Hyperbolic paraboloid, 284, 285 Plane fascicle, 73
Polar form, 166
Position vector, 11
I Positive index, 184
Image, 92 Principal normal, 201
Injective, 92 Principal normal versor, 202
Invariants, 235, 263
Isomorphism, 92
Q
Quadratic form, 166
J
Quadric, 262
Jacobi, 175
Jordan block, 123
Jordan canonical form, 122, 124, 125
Jordan cell, 123 R
Jordanizable, 125 Rank, 97
Rectified plane, 200
Rectilinear generators, 286, 287
K Regular arc of a curve, 197
Kernel, 92 Regular curve, 198
Regular portion of a surface, 217
Regular surface, 218
L Rotation matrix, 152
Linear combination, 27
Linear transformation, 91
Linearly dependent, 27 S
Linearly independent, 27 Scalar (dot) product, 14, 135
Second Frenet formula, 205
Secular equation, 265
M Signature, 185
Matrix of the scalar product, 141, 142 Sphere, 270, 271
Mixed product, 20 Surface area element, 229
Surface of rotation, 295
Surjective, 92
N System of generators, 27
Negative index, 185
Norm, 138
Normal, 62, 63, 221, 222
T
Normal plane, 199, 200
Tangent, 198, 201
Null vector, 3
Tangent plane, 219
Tangent versor:, 202
O Third Frenet formula, 205
One-sheeted hyperboloid, 278, 279 Torsion, 206
Oriented segment, 1, 2 Transition matrix, 39
Orthogonal, 13, 139–141 Triangle rule, 5, 6
Orthogonal transformation, 150 Two-sheeted hyperboloid, 280
Orthonormal, 141
Osculator plane, 200
V
Vector space, 24, 25
P Vector subspace, 46, 47
Parabola, 246 Versor, 3