Marian Mureşan

Introduction to
Mathematica®
with
Applications
Introduction to Mathematicar with Applications
Marian Mureşan

Introduction to
Mathematicar
with Applications

123
Marian Mureşan
Faculty of Mathematics and Computer Science
Babeş-Bolyai University
Cluj-Napoca, Romania

ISBN 978-3-319-52002-5 ISBN 978-3-319-52003-2 (eBook)

DOI 10.1007/978-3-319-52003-2

Library of Congress Control Number: 2016963590

© Springer International Publishing AG 2017

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.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To my wife Viorica, always with love
Foreword

The book (written by Professor Marian Mureşan) is intended to present the

Mathematica system in a manner in which the reader will find it easy to acquire
a part of the considerable number of mathematical instruments offered by the
producer, based on a large variety of examples taken from different scientific
branches.
The importance of the symbolic calculus is incontestable, since nowadays the
scientific calculus is not reduced to the numerical one. Modern mathematical
problems, or those appearing through modeling natural phenomena, lead to such
complicated symbolic expressions that we need the help of computers to process
them. The fast development of the computers’ technical capacity has enabled the
genesis of a new domain—the symbolic (or formal) calculus—through which a
genuine human–machine collaboration is achieved. The symbolic calculus systems
are aimed at automating difficult calculi. They provide simple access to many
sophisticated mathematical instruments, using a simple yet flexible language, that
allows avoidance of writing interminable subroutines (similarly to how it happens
with utilizing traditional languages).
It is mainly mathematical aspects that are covered in this book, so that the reader
can understand as precisely as possible how to handle the problem. At the same time,
the user is warned with regard to the type of problems which can be solved by means
of the computer and the necessity of the analysis and verification of the results.
Special emphasis is given to the fact that a session of Mathematica is typically
interactive. Often there are multiple ways to approach problems, the choice of one
of them being made on the spot, according to the answer of the system. The reader
is carefully guided to the most suitable choice.
The book contains 11 chapters on specific aspects of the Mathematica language
or branches of mathematics in general, such as ordinary differential equations, to
which Chap. 7 is dedicated. It is worth mentioning that part of the applications
presented in the book have been elaborated by the author in recent years, being
the subject of certain scientific articles mentioned in the bibliography. In this book,
standard linguistic constructions are used.

vii
viii Foreword

The book is at an incredibly high scientific level, and it is useful to many

categories of users; for researchers, professors, and students in scientific fields, the
knowledge of such a symbolic calculus system is nowadays indispensable. Math-
ematics researchers can use the programs for testing some conjectures or even for
proving some results that need a big amount of calculation (in differential geometry,
number theory, combinatorics, theory of functions of complex variable, numeric
calculus, differential equations). The book is useful for engineers, economists, and
IT specialists, who can therefore benefit from easy access to a volume of vast
mathematical knowledge. I would mention that parts of the content of this book have
been presented during the scientific seminars held at the Faculty of Mathematics
and Computer Science, having been highly appreciated. Thanks to all of the above
mentioned, I strongly recommend Prof. Marian Mureşan’s book.

Cluj-Napoca, Romania Prof. Valeriu Anisiu

August 2016
Preface

This book is the output of our work over several years. Being involved in
calculus of variations and optimal control problems, we have realized that an
exact calculation and a suggestive visualization are very useful, making the ideas
addressed many times only in an "-ı language clearer. Then we have chosen
Mathematica for computations and visualizations of the ideas. Why did our option
go to Mathematica? The answer is simple: because we had noticed the wonderful
results of Prof. J. Borwein and his colleagues regarding the decimals of number .
Their approach was based on an extensive use of Mathematica.
Wolfram Research, located at Champaign, IL, USA, is the company which has
been developing Mathematica.
Mathematica is continuously developing. We used Mathematica 10.3. It is very
likely there will be newer versions with extra facilities in the future.
We have introduced notions and results in Mathematica in our lectures to master
students at the Faculty of Mathematics and Computer Science of the Babeş-Bolyai
University in Cluj-Napoca, Romania. We did the same thing with our PhD students
at three summer schools organized in the framework of the grant “Center of
Excellence for Applications of Mathematics” supported by DAAD, Germany. The
summer schools have been organized in Struga (Macedonia, FYROM), Sarajevo
(Bosnia and Herzegovina), and Cluj-Napoca (Romania).
This book is not very large, but it collects many examples. In the first part of
the book, the examples are discussed in detail helping the reader to understand the
reasoning in and with Mathematica. Later on, the reader is led to use the benefit
of the Help and other sources freely offered by Wolfram Research. We take into
account mainly the Wolfram community forum as well as the video training and
conferences generously offered by Wolfram Research.
A well-motivated case for visualization in mathematics is contained in [58].
Here is the right place to express my gratitude to the following colleagues of
mine from the Faculty of Mathematics and Computer Science of the Babeş-Bolyai
University for their support: Anca Andreica, Valeriu Anisiu, Paul Blaga, Virginia

ix
x Preface

Niculescu, Adrian Petruşel, and Adrian Sterca. The existence and development of
the MOS (Modeling, Optimization, and Simulation) Research Center of our faculty
was a real help for us in the preparation of this book.

Cluj-Napoca, Romania Prof. Marian Mureşan

August 2016
Contents

1 About Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Warning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 First Steps to Mathematica. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 The Introductory Techniques for Using Mathematica . . . . . . . . . . . . . . 5
2.1.1 Numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Bracketing in Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3 Set or SetDelayed Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.4 Some Simple Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Basic Steps to Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Problems in Number Theory, Symbolic Manipulation,
and Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1 Problems in Number Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Symbolic Manipulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.3 Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Riemann  Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Some Numerical Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.1 The First Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.2 The Second Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.3 The Third Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5.1 Operations with Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5.2 Operations with Matrices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.5.3 Inner and Outer Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5.4 Again on the Third Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4 Sorting Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Sorting Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

xi
xii Contents

4.2.1 Selection Sort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.2.2 Insertion Sort. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2.3 Mergesort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.4 Heapsort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.5 Quicksort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.1.1 Functions with Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Differentiation and Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2.1 Differentiation of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2.2 Integration of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3 Functions and Their Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4 Plane and Space Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4.1 Plane Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4.2 Functions and Graphs in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6 Manipulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.1 Manipulate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.1.1 Circumcircle, Incircle, and the Main Points in
a Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.1.2 Euler’s Nine Points Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.1.3 Frenet–Serret Trihedron of a Helix . . . . . . . . . . . . . . . . . . . . . . . . 99
6.1.4 Hyperboloid of One Sheet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7 Ordinary Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.1 Simple Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.2 Systems of Ordinary Linear Homogeneous Differential
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
7.2.1 Singular Points of the Linear Homogeneous
Planar Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.2.2 Multiple Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.3 On Two Runge–Kutta Methods of the Fourth Order . . . . . . . . . . . . . . . 124
7.3.1 An Explicit Runge–Kutta Method . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.3.2 A Semi-explicit Runge–Kutta Method . . . . . . . . . . . . . . . . . . . . 128
8 Pi Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.1 Various Simple and Not So Simple Formulas. . . . . . . . . . . . . . . . . . . . . . . 133
8.1.1 Vandermonde Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.1.2 Sums Connected to Polygonal Numbers . . . . . . . . . . . . . . . . . . 134
8.1.3 Machbin’s and Machbin-Like Formulas . . . . . . . . . . . . . . . . . . 134
8.1.4 Gregory and Leibniz Formula. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.1.5 Vardi Formula. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
8.1.6 Abraham Sharp Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.1.7 Not So Simple Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.2 Newton’s Geometric Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.3 Euler Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Contents xiii

8.4  by Arcsin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

8.5  by the Golden Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.6  by Integrals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.6.1 Dantzell Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.6.2 Lucas Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.6.3 Backhouse-Lucas Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.7 BBP and Adamchik-Wagon Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.8 A Method for Finding AW Formulas and Proofs . . . . . . . . . . . . . . . . . . . 159
8.9 BBP-Type Formulas for  in Powers of 2k . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.9.1 A Case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.10  Formulas by Binomial Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.11 S. Ramanujan Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.12 R. W. Gosper Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.13 The Chudnovskys’ Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.14 B. Cloitre Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.15 F. Bellard Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
9 Optimization of Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
9.1 Necessary Conditions for a Mayer Optimal Control Problem . . . . . 171
9.2 Zermelo’s Navigation Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2.2 A Planar Form of the Navigation Problem . . . . . . . . . . . . . . . . 174
9.2.3 Numerical Approach to the Navigation Problem . . . . . . . . . 177
9.3 Optimal Guidance for Planar Lunar Ascent. . . . . . . . . . . . . . . . . . . . . . . . . 179
9.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.3.2 The Ascent from the Lunar Surface . . . . . . . . . . . . . . . . . . . . . . . 180
9.3.3 The Optimal Control Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.3.4 Computational Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
10 Miscellany in the Euclidean Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.1 Some Planar Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.1.1 Conchoid of Nicomedes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.1.2 Cycloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
10.1.3 Epicycloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
10.1.4 Hypocycloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
10.1.5 Maria Agnesi’s Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
10.1.6 Cassini Ovals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
10.1.7 Orthoptics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
10.1.8 Pedals of Some Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
10.2 Attractors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
10.2.1 Hénon Attractor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
10.2.2 Lorenz Attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
10.3 Limit Cycles and Hopf Bifurcation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
10.3.1 Van Der Pol Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
10.3.2 Hopf Bifurcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
xiv Contents

11 Miscellany in the Euclidean Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

11.1 Some Space Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
11.1.1 Viviani’s Window. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
11.1.2 The Frenet–Serret Trihedron Along Viviani’s Window . . 234
11.1.3 Plane Curves on a Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
11.2 Some Surfaces in R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
11.2.1 Torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
11.2.2 A Simple Tube Around Viviani’s Window . . . . . . . . . . . . . . . . 238
11.2.3 Another Tube Around Viviani’s Window . . . . . . . . . . . . . . . . . 238
11.2.4 Helicoid, Catenoid, and Costa’s Minimal Surface . . . . . . . . 239
11.3 Some Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.3.1 The Volume of Two Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.3.2 Reuleaux Tetrahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
11.4 Local Extrema of Real Functions of Several Real Variables . . . . . . . 248
11.4.1 Unconstrained Local Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
11.4.2 Constrained Local Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Chapter 1
About Mathematica

1.1 Introduction

Mathematica is a product of Wolfram Research on the address http://www.wolfram.

com/. The complete documentation for an effective use of Mathematica can be
found at that page. The documentation is clear in explanation, rich in examples,
and fundamental for everybody working with Mathematica. Therefore, we avoided
repeating explanations of the rules of arithmetic operations, syntaxes of the com-
mands, built-in functions, etc. We let this pleasure to the reader to discover them by
the references given herein.
The evolution of the logos of Mathematica is introduced at http://mathworld.
wolfram.com/Spikey.html and presented below by Fig. 1.1.
The versions of spikeys are shown in Fig. 1.2
GraphicsGrid[Partition[
MapIndexed[Show[#1,PlotLabel!”Version ”<>ToString[#2[[1]]]] &,
Take[spikeys,10]],4,4,f1,1g,fg],ImageSize!550,Alignment!Top]
and all are trademarks of Wolfram Research.
Along the present book, we used Mathematica 10:3; its logo is given in Fig. 1.3.
Mathematica is a very versatile and powerful package for numeric, symbolic, and
graphic calculus and even much more. Besides the documentation for Mathematica,
there are a lot of interesting books leading the reader into this miraculous world.
We mention here only a few titles [11, 20, 32, 33, 64–68, 72, 73], and [16].
The documentation issued by Wolfram Research is of main importance [75]
and [74]. A very important and useful activity is the series of video workshops
generously offered by Wolfram Research.

© Springer International Publishing AG 2017 1

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_1
2 1 About Mathematica

Fig. 1.1 Spikeys

Fig. 1.2 Versions of spikeys

1.1 Introduction 3

Fig. 1.3 The spikey of this

version

1.1.1 Warning

This book is intended to be only an introduction to Mathematica, mainly from the

point of view of a mathematician according to his experience. Mathematica is much
larger than we present it here and is continuously growing fast. Therefore, it is hard
to predict how it will look like over a decade.
Chapter 2
First Steps to Mathematica

2.1 The Introductory Techniques for Using Mathematica

When we intend to write a document in Mathematica, we firstly select from Format

menu the StyleSheet option and choose the type of document we are interested in.
Then we follow clicking either Text or Input in Style of Format, depending on
whether we want to introduce text or formulas to compute (evaluate). For various
formulas, the option Writing Assistant in Palettes is of real help.

2.1.1 Numbers

Suppose we want to compute the sum 2 C 5. We type

2+5
Then immediately appears
2+5 (* Here we press SHIFT+ENTER to get the result *)
7
Thus the result is 7. As one can already see, we have inserted a comment. The
comments are useful when we introduce initial data and/or mark different parts of
a longer code. They are written in gray. Below we introduce other examples with
arithmetic operations.
25-13 (* 25 minus 13 SHIFT+ENTER to get the result *)
12
2211 (* 22 times 11 SHIFT+ENTER to get the result *)
242
22/11 (* 22 divided by 11 SHIFT+ENTER to get the result *)
2

© Springer International Publishing AG 2017 5

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_2
6 2 First Steps to Mathematica

22^ 11 (* or *) 2211 (* 22 at the power 11 SHIFT+ENTER to get the

result *)
584318301411328

7:1=2:33 (* Division *)
3.04721

7/6 (* Division *)
7
6
(* The result is given as a fraction. If we wish a result under the form
of a decimal number, use the built-in function N[ ] and write *)

N[7/6] (* The standard output is with 5 decimals. Here we use the

built-in function N; the arguments of a function, built-in functions
included, are given by means of square brackets *)
1.16667
We can get the same result writing
7/6//N (* This means that the built-in function N[ ] is applied to the
fraction 7/6 *)
1.16667
Also,
7./6 (* We write either term as a decimal number and the result is
returned as decimal number *)
1.16667

N[7/6,10] (* If we need more than 5 decimals, we specify their number in

the second argument of the built-in function N[ ] *)
1.166666667
We now mention another format for the built-in function N
N@7/6 (* This means the built-in function N[ ] is applied to the fraction
7/6; equivalently, we write N[7/6] *)
1.16667
Suppose we want to know the first 15 decimals of the number pi, denoted Pi or
: If we write either of the inputs

Pi (* or *)  (* Input *)

we get that
 (* and *)

i.e., no numerical result is returned. To get the expected numerical result, we have
to write
N[Pi,15]
3.14159265358979
2.1 The Introductory Techniques for Using Mathematica 7

or
N[,15]
3.14159265358979
and Mathematica returns it with 15 decimals. Another example is given below:
 1 8
2
(* We wrote this expression with the Writing Assistent menu in
Palletes *)
1
256
If we want a numerical result with 10 decimals, we write
h 8 i
N 12 ; 10
0.003906250000
Two interesting cases follow:
2/0
2/1
Power::infy: Infinite expression 10 encountered. 
ComplexInfinity
0
In the first case, the answer contains a warning message and a result (Com-
plexInfinity), while in the second case the result is correct according to the usual
convention [53, p. 30].
We now present some notations and built-in functions.
N, Z, R, C (* Used for the sets of natural, integer, real, and complex
numbers *)
x+y
p I (* The complex number x+iy; I is reserved for the imaginary unit
1 *)
Re[z] (* The real part of a complex number z; Re is a built-in
function *)
Im[z] (* The imaginary part of a complex number z; Im is a built-in
function *)
Conjugate[z] (* The complex conjugate of a complex number z;
Conjugate is a built-in function *)
Abs[z] (* The absolute value of a complex number z; Abs is a built-in
function *)
Arg[z] (* The argument  in z D jzj.cos  C i sin /, the trigonometric
form of a complex number z; Arg is a built-in function *)
N, Z, R, C
x+y I
Re[z]
Im[z]
Conjugate[z]
Abs[z]
Arg[z]
8 2 First Steps to Mathematica

p p p p
Suppose we want to see whether 3 C 5 C p 7C p 11pis an p algebraic number.
Using the Writing Assistant Palette, we write 3 C 5 C 7 C 11 2 Algebraics
and understand it as a question, i.e., “is it true that the sum belongs to the set of
algebraic numbers?”
p p p p
3 C 5 C 7 C 11 2 Algebraics
True
Another way to check that the sum is an algebraic number consists in per-
forming
p p the p following
p steps. First we generate the list f3; 5; 7; 11g, transform it as
f 3; 5; 7; 11g, add its elements, and finally we verify whether it is true that
their sum is an algebraic number or not.
Plus@@Sqrt[f3,5,7,11g]2 Algebraicsnp (* First we mention that
p p p o
Sqrt[f3,5,7,11g] is equivalent to 3; 5; 7; 11 : Here @@ is a
substitution for the
h built-in
np function Apply and the left-hand side is
p p p oi
interpreted Apply Plus; 3; 5; 7; 11 , which gives
p p p p
3 C 5 C 7 C 11. Finally we ask if this sum is an algebraic
number or not *)
True

2.1.1.1 Divisibility

We want to find out how many zero digits the number 32Š has at the end.
 ı 
IntegerQ 32Š 107 (* Is 32Š=10^ 7 an integer? Yes. This means that
32Š ends withıat least
 7 zeros *) ^
IntegerQ 32Š 108 (* Is 32Š=10 8 an integer? No. This means that
32Š ends with exactly 7 zeros *)
True
False
Another example of divisibility follows below.
IntegerQŒ75^ 75  98^ 98=10^ 98 (* Is 75^ 75  98^ 98=10^ 98 an integer?
The answer is yes *)
ModŒ75^ 75  98^ 98; 10^ 98 (* We test the same thing with the
built-in function Mod *)
IntegerQŒ75^ 75  98^ 98=10^ 99 (* The answer is no and we conclude
the number 75^ 75  98^ 98 has exactly 98 zero digits at the end *)
True
0
False
2.1 The Introductory Techniques for Using Mathematica 9

2.1.2 Bracketing in Mathematica

There are four kinds of bracketing in Mathematica.

(term) parentheses for grouping
f[x] square brackets for arguments of functions
{a,b,c,d} curly braces for lists. A list is a sequence of elements in
which the order counts; it is not a set
a[[i]] the ith element of the list a.

Other examples of calculations by similar constructions are introduced below

together with the corresponding results.

2.1.3 Set or SetDelayed Operator

The = operator (called the Set operator) evaluates the expression on its right-hand
side only once, when the assignment is made.
x=3 (* or *) Set[x,3]
The value of the variable x is 3 and it is permanent. It is kept until a different
assignment regarding it occurs or is explicitly removed.
A certain value can be simultaneously set to several variables as
x=y=value (* The same value is set to both variables *)
value
To remove the value of a variable, we write
x=. (* or *) Clear[x]
The := operator (called the SetDelayed operator) used in defining functions
differs from the = operator. When we use := the expression appearing on its right
is evaluated anew each time the expression appearing on its left is called. There are
situations when = and := can be used interchangeably. There are cases when one is
appropriate and the other is not.
a:=3 (* or *) SetDelayed[a,3]
Mathematica uses both upper- and lowercase letters. There is a convention in
Mathematica that the built-in objects always have names starting with uppercase
(capital) letters. To avoid confusion, one should always choose names for one’s own
variables that start with lowercase letters.
mnop is a variable (independent or dependent) name introduced by
the user
Mnop is a built-in object whose name begins with a capital letter
10 2 First Steps to Mathematica

Here is a short list of some built-in functions.

DivideŒx; y . x/y  /
p
SqrtŒx . x  /
LogŒx . loge x  /
SinŒx . sin x  /
TanŒx . tan x  /
ArcSinŒx . arcsin x  /
ArcTanŒx . arctan x  /
TanhŒx . tanh x  /
ArcTanhŒx . arctanh x  /

We note the following facts regarding the variables:

xy means x times y
xy is the variable xy. We note that there is no space between
x and y
5x means 5 times x
x^ 2 y means x^ 2 times y or x2 times y

The following shortcuts are useful:

@ means that f @expr and is defined f [expr]
@@ means Apply. f @@fa; b; c; dg D f Œa; b; c; d
@@@ means applying at level 1. f @@@{{a,b,c},{d,e}} is
{f [a,b,c],f [d; e]}
=: means ReplaceAll. {x; y}/. x ! a is fa; yg
& specifies a pure function. #13 C #24 &[x,y] is x3 C y4
&& is the logical And

2.1.4 Some Simple Steps

Clear[a,b,c,x] (* Clear is a very useful built-in function. It removes any

previous definition regarding the arguments of it *)
{Plus@@fa,b,cg, (* Here @@ is again a substitution for the built-in
function Apply and the code is interpreted as Apply[Plus,fa,b,cg]
which gives a C b C c: a, b, and c are any objects that can be added:
complex numbers, real numbers, matrices that have the same number
of rows and the same number of columns, etc. To avoid long lines of
results, we put all these commands in one list *)
2.1 The Introductory Techniques for Using Mathematica 11

Plus@@.a  b  c/;
Plus@@.a=b=c/;
Plus@@fa  b  cg (* This is Apply[Plus, fa  b  cg], in the curly
braces we have precisely one number *)}
PowerŒx; yDDPower@@fx; ygDD Power@@.x C y/DD
Power@@.x  y/DDxy (* We conclude that we can write xy in either
way *)
 
1 1
a C b C c; a C b C c; a C C ; abc
b c
True

Range@10 (* The list of natural numbers from 1 to 10, i.e.,

1,2,3,4,5,6,7,8,9,10 *)
Table[Range@n,fn,10g] (* Table produces a list, in this case is a list of lists,
i.e., fRange[1], Range[2], Range[3], Range[4], Range[5], Range[6], Range[7],
Range[8], Range[9], Range[10]g *)
Flatten[Table[Range@n,fn,10g]] (* From the previous list of lists it
returns one list deleting all internal curly braces. Thus we get a single list *)
Plus@@Flatten[Table[Range@n,fn,10g]] (* The elements of the
previous list are added *)
{1,2,3,4,5,6,7,8,9,10}
{{1},{1,2},{1,2,3},{1,2,3,4},{1,2,3,4,5},{1,2,3,4,5,6},{1,2,3,4,5,6,7},
{1,2,3,4,5,6,7,8},{1,2,3,4,5,6,7,8,9},{1,2,3,4,5,6,7,8,9,10}}
{1,1,2,1,2,3,1,2,3,4,1,2,3,4,5,1,2,3,4,5,6,1,2,3,4,5,6,7,1,2,3,4,5,6,7,8,
1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,10}
220
The last three commands can be written under a shorter form.
Table[Range@n,fn,10g]
Flatten@% (* % takes the last result *)
Plus@@%
{{1},{1,2},{1,2,3},{1,2,3,4},{1,2,3,4,5},{1,2,3,4,5,6},{1,2,3,4,5,6,7},
{1,2,3,4,5,6,7,8},{1,2,3,4,5,6,7,8,9},{1,2,3,4,5,6,7,8,9,10}}
{1,1,2,1,2,3,1,2,3,4,1,2,3,4,5,1,2,3,4,5,6,1,2,3,4,5,6,7,1,2,3,4,5,6,7,8,
1,2,3,4,5,6,7,8,9,1,2,3,4,5,6,7,8,9,10}
220
We can show other examples.
Range@10; (* The semicolon ”;” stops printing the list
f1; 2; 3; 4; 5; 6; 7; 8; 9; 10g *)
Table[Range@n,fn,10,10g] (* This is a list containing only the list
f1,2,3,4,5,6,7,8,9,10g *)
Plus@@Table[Range@n,fn,10,10g] (* We get the list
f1; 2; 3; 4; 5; 6; 7; 8; 9; 10g *)
12 2 First Steps to Mathematica

{Plus@@Plus@@Table[Range@n,fn,10,10g] (* Now we add the

elements of the list *)
Plus@@Range@10 (* This addition can be performed easier in
this way *)}
Times@@Times@@Table[Range@n,fn,10,10g]==
Times@@Range[10]==10!
{1,2,3,4,5,6,7,8,9,10}
{1,2,3,4,5,6,7,8,9,10}
{55,55}
True
Some example with means follow.
Mean@@Table[Range@n,fn,10,10g]==Mean@@fRange@10g==
Plus@@Range@10/10==11/2 (* Here we suggest several ways to write
the arithmetic mean. We find the arithmetic mean of the first ten nonzero
natural numbers. Generally .a1 C a2 C . . . C an /= n: If all the terms are
integers, the result is given as a fraction of two integers *)
N[Plus@@Range@10/10] (* If at least one entry is given as a decimal
number or the result is required so, the result appears in decimal form *)
GeometricMean@@Table[Range@n,fn,10,10g] (* The geometric
p
mean of some nonnegative numbers, i.e., n a1 a2 : : : an ; ai  0 *)
N@%
HarmonicMean@@Table[Range@n,fn,10,10g] (* The harmonic
mean of some positive numbers, i.e., n=.1=a1 C 1=a2 C : : : C 1=an /;
ai > 0 *)
N[%]
True
5:5
24=5 32=5 51=5 71=10
4:52873
25200
7381
3:41417
Chapter 3
Basic Steps to Mathematica

3.1 Problems in Number Theory, Symbolic Manipulation,

and Calculus

3.1.1 Problems in Number Theory

3.1.1.1 Pythagorean Numbers

Find all the pairs .n; m/ for 1  n  m  20 such that n2 C m2 is a squared number
(Pythagorean pair or Pythagorean numbers). One approach consists in printing all
the satisfactory pairs step by step. If n and m are larger, the resulting list could be
rather large.
Table[ p
If[IntegerQ@ n2 C m2 ==True, (* IntegerQ tests whether its argument is
explicitly an integer *) p p
Print[”n = ”,n,”, m = ”,m, ”, n2 C m2 =”, n2 C m2 ]],fn,20g,fm,n,20g];

(*We produce pairs of integers with components from 1 to 20 and test

if each pair is a Pythagorean one. If we find a Pythagorean pair, display
its members and the square root of the sum of squares of members.
Obviously, weptake into account the commutativity of the operations *)
n = 3, m = 4, pn2 C m2 = 5
n = 5, m = 12,p n2 C m2 = 13
n = 6, m = 8, pn2 C m2 = 10
n = 8, m = 15, pn2 C m2 = 17
n = 9, m = 12, pn2 C m2 = 15
n = 12, m = 16, pn2 C m2 = 20
n = 15, m = 20, n2 C m2 = 25

© Springer International Publishing AG 2017 13

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_3
14 3 Basic Steps to Mathematica

Another approach to Pythagorean numbers is introduced below [33, p. 83]:

Table[fn,mg,fn,6g,fm,n,6g]
Flatten[Table[fn,mg,fn,6g,fm,n,6g],1] (* We get a list of pairs *)
Select[Flatten[Table[fn,mg,fn,20g,fm,n,20g],1] (* We flatten the list
of pairs and select the Pythagorean numbers. We may consider that
mn. The built-in function selects and picks out the pairs that satisfy
the logical condition *)
(SqrtŒ#ŒŒ1^ 2 C #ŒŒ2^ 2 2Integers)&] (* The result appears as a list
of Pythagorean pairs *)
fff1; 1g; f1; 2g; f1; 3g; f1; 4g; f1; 5g; f1; 6gg; ff2; 2g; f2; 3g; f2; 4g; f2; 5g;
f2; 6gg; ff3; 3g; f3; 4g; f3; 5g; f3; 6gg; ff4; 4g; f4; 5g; f4; 6gg; ff5; 5g; f5; 6gg;
ff6; 6ggg
ff1; 1g; f1; 2g; f1; 3g; f1; 4g; f1; 5g; f1; 6g; f2; 2g; f2; 3g; f2; 4g; f2; 5g; f2; 6g;
f3; 3g; f3; 4g; f3; 5g; f3; 6g; f4; 4g; f4; 5g; f4; 6g; f5; 5g; f5; 6g; f6; 6gg
ff3; 4g; f5; 12g; f6; 8g; f8; 15g; f9; 12g; f12; 16g; f15; 20gg
The above result on Pythagorean numbers
 can be given as a table using the built-
p
in function Grid. We display the triplet n; m; n2 C m2 ; where .n; m/ are some
Pythagorean numbers.
p
header=fn,m,” n2 C m2 ”g; (* This is the header of the grid. Thus, we
have three columnsp *)
f[fx ,y g]:=fx,y, x2 C y2 g; (* The rule by which we add the third
column *)
Grid[Join[fheaderg,f/@Select[Flatten[Table[fn,mg,fn,20g,fm,n,20g],1],
(SqrtŒ#ŒŒ1^ 2 C #ŒŒ2^ 2 2Integers)&]],
Frame!All, Alignment!Right,FrameStyle!Thin] (* f/@ expr is
equivalent to Map[f, expr] and applies f to each element on the first level
in expr. Then the header is added on the first position of the list, and
finally the list is displayed *)
p
n m n2 C m2
3 4 5
5 12 13
6 8 10
8 15 17
9 12 15
12 16 20
15 20 25

3.1.1.2 Euler’s Sum of Powers Conjecture

Euler’s conjecture was proposed in 1769. It states that for all integers n and k greater
than 1, if the sum of n kth powers of positive integers is itself a kth power, then n
3.1 Problems in Number Theory, Symbolic Manipulation, and Calculus 15

P
is greater than or equal to k, i.e., niD1 ai k D bk ; where n > 1 and a1 ; a2 ; : : : ; an ; b
are positive numbers, then n  k. He also provided a complete solution to the four
cubes problem 33 C 43 C 53 D 63 , i.e.,
33 C 43 C 53 DD 63
True
We will study some particular cases.
Do[
Do[
Do[
h i
p
If IntegerQ@ 3 i3 C j3 C k3 DD True; PrintŒ”i D ”; i; ”; j D ”; j; ”; k D ”; k ,
fk,j,20g],
fj,i,15g],
fi,10g]
i = 1, j = 6, k = 8
i = 2, j = 12, k = 16
i = 3, j = 4, k = 5
i = 3, j = 10, k = 18
i = 6, j = 8, k = 10
i = 7, j = 14, k = 17
i = 9, j = 12, k = 15
We can approach this exercise from another point of view.
n p o
header3= i; j; k; ” 3 i3 C j3 C k3 ” ; (* This is the header of the grid *)
n p o
f3[fx ,y ,z g]:= x; y; z; 3 x3 C y3 C z3 ; (* It transforms a triplet into a
four-component list *)
triple=fg; (* The list of triples is empty at the beginning *)

Do[Do[Do[
h i
p
If IntegerQ@ 3 i3 C j3 C k3 DD True; AppendToŒtriple; fi; j; kg ;
fk,j,30g],fj,i,25g],fi,10g] (* The commutativity allows us to consider
only the cases 1  i  j  k *)
Grid[Join[fheader3g,f3/@triple],Frame!All,Alignment!Right,
FrameStyle!Thin]
16 3 Basic Steps to Mathematica

p
3
i j k i3 C j3 C k3
1 6 8 9
2 12 16 18
3 4 5 6
3 10 18 19
3 18 24 27
4 17 22 25
6 8 10 12
7 14 17 20
9 12 15 18
11 15 27 29
12 16 20 24
15 20 25 30
18 19 21 28
18 24 30 36

On the same line, we have 304 C 1204 C 2724 C 3154 D 3534 , i.e.,
304 C 1204 C 2724 C 3154 DD3534
True
p
Table[If[IntegerQ@ 4 i4 C j4 C k4 C m4 ==True,
Print[“i = ”,i,“, j = ”,j,“, k = ”,k,“, m = ”,m]],
fm,300,315g,fk,250,272g,fj,100,120g,fi,25,30g];
i = 30, j = 120, k = 272, m = 315
For k D 5, the conjecture was disproved by L. J. Lander and T. R. Parkin in 1966
[39] when they found the following counterexample 275 C 845 C 1105 C 1335 D
1445 , i.e.,
275 C 845 C 1105 C 1335 DD1445
True
For k D 4, Noam Elkies in [22] found a counterexample which states that
26824404 C 153656394 C 187967604 D 206156734 , i.e.,
26824404 C 153656394 C 187967604 DD 206156734
True
In 1988, Roger Frye in [27] subsequently found the smallest possible counterex-
ample for k = 4: 958004 C 2175194 C 4145604 D 4224814 , i.e.,
958004 C 2175194 C 4145604 DD 4224814
True

3.1.1.3 A Conjecture of Fermat

n
In 1640, Fermat stated that every number of the form 2.2 / C 1 is a prime number.
Almost 200 years later, an example contradicted Fermat’s statement.
3.1 Problems in Number Theory, Symbolic Manipulation, and Calculus 17

 n
Table[PrimeQ 22 ,fn,10g] (* PrimeQ offers an answer whether its
argument is prime or not; then the corresponding Boolean value is
assigned to a list *)
{True,True,True,True,False,False,False,False,False,False}
   n 
Table If PrimeQ 22 C 1 DD False; Print Œ”for
 n D ”; n; ”; number D
n n
22 C1 D"; 22 C 1; “is not a prime” ; fn; 10g I (* Display the
n
nonprime numbers
h of ithe form 22 C 1 for all n2 f1,2,: : :,10g *)
5
FactorInteger 22 C 1 (* For a nonprime number, we display
its decomposition
h in powers
i of primes *)
6
FactorInteger 22 C 1
n
for n = 5, number = 22 +1 = 4294967297 is not a prime
n
for n = 6, number = 22 +1 = 18446744073709551617 is not a prime
2n
for n = 7, number = 2 +1 = 340282366920938463463374607431768211457
is not a prime
n
for n = 8, number = 22 +1 = 115792089237316195423570985008687907853
269984665640564039457584007913129639937 is not a prime
n
for n = 9, number = 22 +1 = 134078079299425970995740249982058461274
7936582059239337772356144372176403007354697680187429816690342
7690031858186486050853753882811946569946433649006084097 is not
a prime
n
for n = 10, number = 22 +1 = 1797693134862315907729305190789024733
61797697894230657273430081157732675805500963132708477322407536
02112011387987139335765878976881441662249284743063947412437776
78934248654852763022196012460941194530829520850057688381506823
42462881473913110540827237163350510684586298239947245938479716
304835356329624224137217 is not a prime
ff641; 1g; f6700417; 1gg
ff274177; 1g; f67280421310721; 1gg
Now we recompose the number from FactorInteger and compare it with the
initial number.
power[fx ,y g]:=Power[x,y]; h 5 i 5
Times@@power/@FactorInteger 22 C 1 ==22 C 1
True

3.1.2 Symbolic Manipulations

The binomial coefficient BinomialŒn; k is the number of ways of picking k

unordered elements from a set of n elements. By definition Binomial[n,0]
=Binomial[n,n]=1.
18 3 Basic Steps to Mathematica

Binomial[m+1,n+1]==Binomial[m,n+1]+Binomial[m,n] (* This is the

recurrence of binomials; no hint about its validity *)
FullSimplify[Binomial[m,n]+Binomial[m,1+n]] (* We can overcome
the previous situation by using the FullSimplify built-in function *)
Binomial[1+m,1+n]==Binomial[m,n]+Binomial[m,1+n]
Binomial[1+m,1+n]
P
Now we verify for some values of n the binomial identity nkD0 BinomialŒn; k
D 2n :
sumbinomials[n
P ]:=1+Plus@@Binomial[n,Range@n] (* Define the
sum nkD0 Binomial[n,k] *)
Table[Boole[sumbinomials[n]==2n ],fn,10g] (* Test if the sum is equal
to the corresponding power of 2. If yes, it inserts ”1” in the list,
otherwise inserts ”0” *)
With[fm=10g,
If[Times@@Table[Boole[sumbinomials[n]==2n ],fn,mg]==1,
Print[”True”],Print[”False”]]]
P P
{FullSimplify[ kD0 Binomial[,k]],FullSimplify[ kD0 Binomial[,2k]],
P
FullSimplify[ kD0 Binomial[,2k+1]] (* Here we calculate some sums
and simplify the P results *) P
FullSimplify[f kD0 Binomial[,k], kD0 Binomial[,2k],
P
kD0 Binomial[,2k+1]g] (* Here there are the previous computations
written in a shorter form *)}
f1; 1; 1; 1; 1; 1; 1; 1; 1; 1g
True
˚  1C 1C

˚2 ; 21C ; 21C

2 ;2 ;2
Some binomial identities are exhibited below:
Clear[n,m]
Pk Pk
FullSimplify[f
Pn mD0 Binomial[n+m,m],
Pn mD0 Binomial[n+m,n],
k
PmD1 m Binomial[n,m], kD0 Binomial[n+k,k]Divide[1,2
Pm ],
n i
Binomial[n,i]
˚ iDm -1+n Binomial[i,m], iD0 .1/ Binomial[n,i]g]

1; 1; 2 n; 2n ; 2-m+n Binomial[n,m]; .1/m Binomial[-1+n,m]

Now we take a look at the Bell numbers. Set n 2 N: The nth Bell number is the
number of partitions of an n-set. Its corresponding Mathematica built-in function is
BellB[n].
P
sumBell[n ]:=1+ n1 kD1 Binomial[n-1,k]sumBell[k] (* We define our
Bell numbers based on the corresponding recurrence *)
Table[sumBell[n],{n,1,10}] (* The first ten Bell numbers are displayed *)
Table[Boole[sumBell[n]==BellB[n]],{n,10}] (* We test if our Bell
numbers coincide with the built-in Bell numbers only for the first ten terms *)
3.1 Problems in Number Theory, Symbolic Manipulation, and Calculus 19

With[{m=10},
If[Times@@Table[Boole[sumBell[n]==BellB[n]],{n,m}]==1,
Print[”True”],
Print[”False”]]]
f1; 2; 5; 15; 52; 203; 877; 4140; 21147; 115975g
f1; 1; 1; 1; 1; 1; 1; 1; 1; 1g
True
Another example of symbolic p manipulation requires checking whether
tan.3=11/ C 4 sin.2=11/ D 11.
Simplify[Tan[3/11]+4Sin[2/11]] (* The built-in function Simplify is of
no help here *)
FullSimplify[Tan[3/11]+4Sin[2/11]] (* The built-in function
FullSimplify
  is of real
 help *)
Cot 5 C 4Sin 2
p 22 11
11
qp
By the next example, we show that 3 64 .22 C .1=2/2 /  1 D 4.
q 
p3
Simplify 2 2
64.2 C .1=2/ /  1 (* The result immediately follows *)
4
We show that eI D 1:
ExpŒI (* The result follows at once *)
1

3.1.3 Texts

We introduce an example of a text. Suppose the word “Mathematica” is lying on the

graph of the sine function.
Show[Graphics[MapThread[Text[StyleForm[#1,FontFamily!”Courier”,
FontWeight!”Bold”,FontSize!14,FontColor!Hue@#2],
f#2,Sin@#2g]&,fCharacters[”Mathematica”],Table[2i/11,fi,11g]g]]]
See Fig. 3.1.

Fig. 3.1 Mathematica on the

sine function
20 3 Basic Steps to Mathematica

An interesting example with texts similar to the one mentioned earlier may be
found in [65, pp. 6–7].

3.2 Riemann  Function

P1
Riemann’s zeta function .p/ is defined by the relation .p/ D kD1 kp (for
p > 1).
Riemann  function at point 2 is
Zeta@2
2
6
We get the same result as a limit
P 
Limit nkD1 k12 ; n ! Infinity
2
6

Here there are the values of the Riemann  function corresponding to some values
of the argument p.
Table[Zeta@p,fp,9g]
 P  
Table Limit nkD1 k1p ; n ! 1 ; fp,9g (* We look for the values of
  function
Riemann P as limits *)  
Table N Limit nkD1 k1p ; n ! 1 ; fp; 3; 9; 2g
n 2 4
o
6 8
ComplexInfinity; 6 ; ZetaŒ3; 90 ; ZetaŒ5; 945 ; ZetaŒ7; 9450 ; ZetaŒ9
n 2 4
o
6 8
1; 6 ; ZetaŒ3; 90 ; ZetaŒ5; 945 ; ZetaŒ7; 9450 ; ZetaŒ9
f1:20206; 1:03693; 1:00835; 1:00201g

3.3 Some Numerical Sequences

3.3.1 The First Sequence

p
Let us analyze the sequence an D n n, n 2 N; n  2:
p
Table[N@ n n; fn; 20g] (* The resulting list suggests that the sequence
is eventually decreasing *)
f1:; 1:41421; 1:44225; 1:41421; 1:37973; 1:34801; 1:32047; 1:29684; 1:27652;
1:25893; 1:24358; 1:23008; 1:21811; 1:20744; 1:19786; 1:18921; 1:18135;
1:17419; 1:16762; 1:16159g
We check it by the next code.
˚ p p

Limit N@ n n; n n ; n ! 1 (* The limit is 1 *)
f1; 1g
3.3 Some Numerical Sequences 21

p
n
Fig. 3.2 The sequence n

We want to visualize its convergence toward 1.

 ˚ p
 
ListPlot
˚ Table n; N@ n n ; # ˚; Filling ! Axis;

ImageSize ! 200 &=@

fn; 1; 20g; fn; 100; 1000; 100g; n; 103 ; 104 ; 103
The graphs suggest the behavior of this sequence; see Fig. 3.2.

3.3.2 The Second Sequence

We consider the sequence defined

p p
n 2 N :
nC1 n
an D .n C 1/Š  nŠ;

A classical solution to it may be found at [53, pp. 106–107]. With Mathematica

the solution is clear immediately, considering the next code.
h h p p ii
n
Assuming n 2 Integers&&n > 0; Limit nC1 .n C 1/Š  nŠ; n ! 1
1
e
22 3 Basic Steps to Mathematica

3.3.3 The Third Sequence

We now look at the convergence of a sequence defined by a recurrence [53, p. 141].

The sequence is defined by

xnC1 D 1  xn 2 ; x1 D a 2 0; 1Π; n 2 N :

We take a look at the behaviorof the terms of the sequence considering three
p .
different initial values :25; :9; and 1 C 5 2: The following code is helpful:

With[fm=7g, (* Number of terms *)

x[1]=.25;x[n ]:=1-x[n-1]2 ; (* The first term and the first recurrence *)
y[1]=.9;y[n ]:=1-y[n-1]2 ; (* The first term and the second recurrence *)
z[1]=(-1+Sqrt[5])/2;z[n ]:=1-z[n-1]2 ; (* The first term and the
third recurrence *)
header=f”index”,”sequence”,”sequence”,”sequence”g;
fGrid[Join[fheaderg,Table[fk,x[k],y[k],Simplify[z[k]]g,fk,mg]],
Frame!All,Alignment!ffRight,Left,Left,Leftgg],
data1=Table[fn,x[n]g,fn,mg];data2=Table[fn,y[n]g,fn,mg];
data3=Table[fn,z[n]g,fn,mg];
ListLinePlot[fTooltip[data1],Tooltip[data2],Tooltip[data3]g,
PlotStyle!fBlue,Red,Darker[Green]g,
Mesh!Full,MeshStyle!fBlack,PointSize[Medium]g,
ImageSize!f280,200g,Ticks!fTable[,f,mg],ff.25,”.25”g,
f(-1+Sqrt[5])/2, ” 1CSqrtŒ5
2
”g,f.9,”.9”g,1gg]g]
The results are exhibited
p in Fig. 3.3.
Denote " D .1 C 5/=2 2 0; 1Π. We have the following cases:
(i) 0 < xn < 1 H) 0 < xnC1 < 1. Thus, the sequence is lying in the open interval
0; 1Œ.

Fig. 3.3 The third sequence

3.4 Variables 23

(ii) Because of sgn.xnC1  xn / D sgn.xn  "/; it implies that if xn D ", then

xnC1 D ", and thus, the sequence is constant.
(iii) Because of sgn.xnC1 "/ D sgn.xn "/; it implies that the sequence oscillates
if x1 ¤ ". We also have jxnC2  xnC1 j > jxnC1  xn j, which implies that
jxnC2  xnC1 j > jx2  x1 j, so the convergence is excluded.
Thus, the sequence converges in a single case and diverges in all the other cases.

3.4 Variables

Just like in mathematics, we need and use variables of different sorts and for
different purposes.
The next command sets the value of the variable x to 7:
x=7
7
If we do not want to display this number, we write
x D 7I . A semicolon “I00 stops printing the value of the variable /
N@PowerŒx;
p .7/1  (* Instead, we want to know the numerical value
7
of 7 *)
1:32047
Now we can use this variable in computations such as
x+Ex +Log@x+Sin@x+Tan@x+Floor[x+.25]+Ceiling[x-.25]
21+e7 +Log[7]+Sin[7]+Tan[7]
In certain cases, if we want a numerical result, then we write
N[%,12] (* % means that the previous result is taken into account
and we ask for a result of 12 *)
1121.10750316
If the input variable is a non-integer real number, we immediately get the desired
result with 5 decimals without using the built-in function N.
x=7.00001
x+Ex +Log@x+Sin@x+Tan@x+Floor[x+.25]+Ceiling[x-.25]
7.00001
1121.12
As we have already seen, if we add a semicolon (;) at the end of an insertion line,
the result is not returned, but it is calculated and kept.
x=2.00001;
x+Ex +Log@x+Sin@x+Tan@x+Floor[x+.25]+Ceiling[x-.25]
N[%,8]
12.8066
24 3 Basic Steps to Mathematica

3.5 Lists

A list is a sequence of objects. It is not a set (in a mathematical sense) because the
rank of an element in a list counts, which is not the case of a set. A proper synonym
of list (in the sense of Mathematica) is sequence (in mathematical sense).

3.5.1 Operations with Lists

Here are two lists of numbers having the same number of elements. We add and
subtract them, rise to power, and perform other operations. The operations are
performed componentwise.
aa=f1,3,4,2,-3g; (* The first list *)
bb=f2,-3,8,-7,1g; (* The second list *)
faa+bb, (* In order to save space, the results of the operations on the
two lists appear as members of a list *)
bb-aag
faa2 +1, (* The operations are performed componentwise *)
Exp[%]//N,
aa/(bb-1.5), (* Division is operated componentwise *)
xxf1;4;11g ;
f1,2,5.1gf1;4;11g (* The two lists must have the same number of
elements *)g
ff3; 0; 12; 5; 2g; f1; 6; 4; 9; 4gg
ff2; 10; 17; 5; 10g; ff20:0855; 1:; 162755:; 0:00673795; 0:135335g;
{2.71828,0.00247875,54.5982,0.00012341,54.5982}},
{2.,-0.666667,0.615385,-0.235294,6.},fxx; xx4 ; xx111 g,
f1; 16; 1:647130704063058J 108 gg
We often need to know the values of certain elements in a list. There are two
ways to extract some elements in a list.
a=f1,4,5,2,-7g
fa[[3]], (* We extract the third element indicating the rank of the chosen
element *)
Part[a,4], (* We extract the fourth element *)
a[[f5,3g]], (* We extract the fifth and third element, in this order.
The result is a list of two elements *)
Part[a,f5,1g], (* We extract the fifth and first element, in this order.
Again the result is a list *)
a[[-2]] (* It counts from the end, i.e., the second element from the
end *)g
f1; 4; 5; 2; 7g
f5; 2; f7; 5g; f7; 1g; 2g
3.5 Lists 25

A list can be rearranged as the following example shows:

Clear[a]
a=f1,4,9,16g;
a[[f4,3,2,1g]]
{16,9,4,1}
The previous rearrangement can be performed by the next command.
a[[Table[Length@a-k,fk,0,Length@a-1g]]]
{16,9,4,1}
We now introduce a list of lists.
Clear[a]
a=f-1,1,f2,3g,f˛,ˇ, g,f,f,,birdggg
{-1,1,{2,3},{˛,ˇ, },{,{,,bird}}}
Then
fa[[3]], (* We pick the third element which is a list *)
a[[3,1]], (* From the third element, we pick the first *)
a[[4]],a[[4,f1,2g]], (* From the fourth element, we pick the first and
the second *)
a[[5]],a[[5,2]],a[[5,2]][[2]]g (* From the fifth element we pick the
second and from it again the second *)
{{2,3},2,{˛,ˇ, },{˛,ˇ},{,{,,bird}},{,,bird},}

Clear[a]
a=f1,4,9,13g
Length@a (* It returns the length of the list *)
fFirst@a, (* It extracts the first element *)
Last@a, (* It extracts the last element *)
Drop[a,3], (* It gives the list with its first three elements dropped *)
Drop[a,f2,4g], (* It gives the list with the second until the fourth
elements dropped *)
Rest@a, (* The first element is removed *)
Most@a, (* The last element is removed *)
Take[a,f2,4g], (* It gives the specified elements in a list *)
Most@Rest@a==Rest@Most@ag (* We ask if the left-hand side
coincides with the right-hand side *)
{1,4,9,13}
4
{1,13,{13},{1},{4,9,13},{1,4,9},{4,9,13},True}

fRiffle[a,picture], (* It inserts the variable picture between successive

elements of the list a *)
Riffle[a,fx,yg], (* It inserts x and y cyclically *)
Riffle[faa,bb,ccg,fx,y,zg]g
{{1,picture,4,picture,9,picture,13},{1,x,4,y,9,x,13},{aa,x,bb,y,cc,z}}
26 3 Basic Steps to Mathematica

Clear[b,c,d]
 
fFirst c2 C c C b2 C b C d2 ; (* The elements are sorted ascending;
thenthe first element is selected *)
Last c2 C b2 C d2 C d , (* The elements are sorted ascending; then
the last
 element is selected
 *)
Drop c2 C b2 C d2 ; 2 , (* The elements are sorted ascending; then
the first
 two elements  are dropped *)
Rest c2 C b2 C d2 ; (* The elements are sorted ascending; then the
first element
 is removed
 *)
Most b2 C d2 C c ; (* The elements are sorted ascending; then the last
element
˚ 2 2is removed *)g

b; d ; d ; c2 C d2 ; b2 C c

a=ff1,4g,9,f13,f14,15ggg
Flatten@a (* It flattens out the nested list *)
Flatten[fa[[2]],a[[3]],a[[1]]g]
{{1,4},9,{13,{14,15}}}
{1,4,9,13,14,15}
{9,13,14,15,1,4}

Clear[b,c]
FlattenAt[fa,fb,cg,fd,eg,ffgg,ff2g,f4gg] (* FlattenAt flattens the
positions 2 and 4 *)
{{1,4,9,13},b,c,{d,e},f}
A list can be rearranged as the following examples show. We use the built-in
function Sort and change the places of certain elements according to some criteria.
Clear[a]
a=f1,9,4,4,16g;
Sort@a (* Implicitly it sorts ascending *)
Sort[a,Greater] (* This sorts in descending order *)
Sort[a,#1>#2&] (* We ask for a descending order *)
Sort[ffa,2g,fc,1g,fd,3gg,#1ŒŒ2 < #2ŒŒ2&] (* It sorts ascending with
respect to the second component *)
Sort[ffa,2g,fc,1g,fd,3gg,#1ŒŒ1 < #2ŒŒ1&] (* It sorts ascending with
respect to the first component *)
{1,4,4,9,16}
{16,9,4,4,1}
{16,9,4,4,1}
{{c,1},{{1,9,4,4,16},2},{d,3}}
{{{1,9,4,4,16},2},{c,1},{d,3}}
A deeper discussion on the sorting problem will be given in the next chapter.
Some elements of a list can be reset. Let us reset the third element of the list.
3.5 Lists 27

u=f1,3,9g;
u[[3]]=~;
u
{1,3,~}
Manipulation of lists.
Clear[a,b,c,d,e,x,y,t,u]
Join[fa,c,bg,fx,yg,ft,ag] (* Join concatenates lists. The result is an
unsorted list *)
{a,c,b,x,y,t,a}

Union[fa,c,bg,fb,c,dg,fc,d,eg] (* Union combines lists, keeping only

distinct elements. It supplies a list sorted ascending *)
{a,b,c,d,e}

Intersection[f1,7,2,3g,f3,1,2g,f2,1,3,3,g] (* Intersection gives a

sorted list of the elements common to all the lists. It supplies a
list sorted ascending *)
{1,2,3}

Complement[f11,7,2,3g,f3,1,2g,f2,1,g] (* Complement gives the list

of elements in the first set not belonging to any other set. The list is
sorted ascending *)
{7,11}
Suppose we want to see if all elements of a sequence are the same.
a=f1,1,1g;
Length@Union@a==1
a=f1,1,2g;
Length@Union@a==1
True
False
Another approach is the following: http://community.wolfram.com/groups/-/m/
t/767736?_19.
a=f1,1,1g;
Equal@@a
a=f1,1,2g;
Equal@@a
True
False
A list of the form ff1; 2g; f3; 4gg can be considered a 2  2 matrix.
ma=ff1,2g,f3,4gg//MatrixForm

12
34
From a simple matrix, we can compose more complicated matrices.
28 3 Basic Steps to Mathematica

m=ff1,2g,f3,4gg
ArrayFlatten[ff0,0,mg,fm,m,0gg]//MatrixForm (* ArrayFlatten
creates a single flattened matrix from a matrix of matrices *)
{{1,2},{3,4}}
0 1
000012
B0 0 0 0 3 4C
B C
@1 2 1 2 0 0A
343400

m=ff1gg
ArrayFlatten[ffm,0,0,0g,f0,m,0,0g,f0,0,m,0g,f0,0,0,mgg]//MatrixForm
ff1gg
0 1
1000
B0 1 0 0C
B C
@0 0 1 0A
0001

fTable[a[i,j]=0,fi,4g,fj,4g]//MatrixForm, (* The null matrix *)

Table[a[i,i]=1,fi,4g]//MatrixForm, (* The column matrix *)
Table[a[i,j],fi,4g,fj,4g]//MatrixForm, (* The identity or unit matrix *)
DiagonalMatrix[fa,b,c,dg]//MatrixForm (* A matrix with prescribed
diagonal elements, all the other elements are null *)g
80 1 0 1 0 1 0 19
ˆ
ˆ 0000 1 1000 a000 > >
<B C B C B C B C=
B0 0 0 0C;B1C;B0 1 0 0C;B0 b 0 0C
ˆ@ 0 0 0 0 A @ 1 A @ 0 0 1 0 A @ 0 0 c 0 A> >
:̂ ;
0000 1 0001 000d

3.5.2 Operations with Matrices

In this subsection, we will take a look at some elementary operations on matrices

[1, 11]. For that, we introduce three sample matrices. As mentioned above, a matrix
is a list of lists, the lists having the same number of elements.
ma=ff0,0,-2,1g,f3,-1,7,2g,f-6,0,5,-1g,f-6,0,1,-2gg;
mb=ff-7,-6,-3,-7g,f2,-3,0,4g,f3,4,1,2g,f5,6,3,6gg;
mc=ff-7,-6,-3,-7,-2g,f2,-3,0,4,2g,f3,4,1,2,0g,f5,6,3,6,-2gg;
fMatrixForm@ma, (* Here we have the matrix ma under the
classical form, that is, a rectangular array of numbers, symbols, or
expressions *)
MatrixForm@mb,
MatrixForm@mcg
3.5 Lists 29

80 1 0 1 0 19
ˆ
ˆ 0 0 2 1 7 6 3 7 7 6 3 7 2 >>
<B C B C B =
B 3 1 7 2 C B 2 3 0 4 C B 2 3 0 4 2 C
C
; ;
ˆ@ 6 0 5 1 A @ 3 4 1 2 A @ 3 4 1 2 0 A>
>
:̂ ;
6 0 1 2 5 6 3 6 5 6 3 6 2
We calculate 3ma  2mb, mbT , ma  mb; jmaj, jmbj, jma  mbj, and ma1 .
matrix1=3ma-2mb//MatrixForm;
matrix2=Transpose@mb//MatrixForm;
matrix3=ma.mb; (* Product of matrices *)
deta=Det@ma; (* Determinant of a matrix *)
detb=Det@mb;
detab=Det@matrix3;
matrix7=Inverse@ma//MatrixForm; (* The inverse of a matrix *)
fmatrix1,matrix2g
fmatrix3//MatrixForm,deta,detb,detab,detab==deta  detbg
matrix7
80 1 0 19
ˆ
ˆ 14 12 0 17 7 2 3 5 > >
<B C=
B 5 3 21 2 C B
C ; B 6 3 4 6 C
ˆ@ 24 8 13 7 A @ 3 0 1 3 A> >
:̂ ;
28 12 3 18 7 4 2 6
80 1 9
ˆ
ˆ 1 2 1 2 >
>
<B C =
B 8 25 4 1 C ; 36; 24; 864; True
ˆ@ 52 50 20 46 A >
>
:̂ ;
35 28 13 32
0 1 1 1 1
 4 0  12  12
B  1
7 19
 12 C
25
B 12 12 C
@ 1 0 1 1 A
6 6 6
2 1
3
0 3
 13
We want to find the eigenvalues of a matrix ma.
ma=ff0,0,-2,1g,f3,-1,7,2g,f-6,0,5,-1g,f-6,0,1,-2gg;
Eigenvalues@ma (* Eigenvalues computed by the built-in function
Eigenvalues *)
soleingen=Solve[Det[ma- IdentityMatrix[4]]==0] (* Eigenvalues
computed based on definition *)
N@soleingen
˚  
Root 36  15#1  3#12 C #13 &; 1 ,
 
Root 36  15#13#12 C#13 &; 3 ,
 2 3


Root 36 15#1  3#1 C #1 &; 2p; 1   1=3 

1 1431 27 1945
1=3
1
p
f ! 1g; ! 1 C 3 2  2
C 2 53 C 1945 ,
30 3 Basic Steps to Mathematica

  p  p
27 1945
1=3
! 1  16 1 C i 3 1431
2
 2

   
1
p p 1=3
2
1  i 3 12 53 C 1945 ,
   p 1=3
p 27 1945
! 1  16 1  i 3 1431
2
 2

   
1
p p 1=3
2
1 C i 3 12 53 C 1945
ff ! 1:g; f ! 6:29279g; f ! 1:64639 C 1:735ig,
f ! 1:64639  1:735igg
We note that the matrix ma has two real and two complex conjugate eigenvalues.
The real eigenvalues are
soleingen[[f1,2g]]
Part[soleingen,f1,2g] (* Equivalent to the previous command *)
%//N
fsoleingen[[1]],soleingen[[1,1]],soleingen[[1,1,1]],soleingen[[1,1,2]]g
(*
 Steps to pick
 the first real eigenvalue *)
 p 1=3   p 1=3  
f ! 1g; ! 1 C 3 2  27 21945
1 1431
C 12 53 C 1945
   p 1=3   1=3  
27 1945
p
f ! 1g; ! 1 C 13 1431 2
 2
C 1
2
53 C 1945
ff ! 1:g; f ! 6:29279gg
ff ! 1g; ! 1; ; 1g
We can perform the previous computations directly, using some built-in
functions.
ma=ff0,0,-2,1g,f3,-1,7,2g,f-6,0,5,-1g,f-6,0,1,-2gg;
CharacteristicPolynomial[ma, ]
feigv=Eigenvalues[N@ma],eigv[[1]]g
eigvec=Eigenvectors@N@ma
MatrixForm@%[[3]] (* The third eigenvector *)
-36-51 -18 2 -2 3 + 4
ff6:29279 C 0:i; 1:64639 C 1:735i; 1:64639  1:735i; 1: C 0:ig;
6:29279 C 0:ig
ff0:187245 C 0:i; 0:654212 C 0:i; 0:699033 C 0:i; 0:21977 C 0:ig;
f0:0733515 C 0:107752i; 0:9019 C 0:i; 0:0170554 C 0:121983i;
 0:341825 C 0:193829ig; f0:0733515  0:107752i; 0:9019 C 0:i;
 0:0170554  0:121983i; 0:341825  0:193829ig;
f0: C 0:i; 1: C 0:i; 0: C 0:i; 0: C 0:igg
0 1
0:0733515 0:107752i
B 0:9019 C0:i C
B C
@ 0:0170554 0:121983iA
0:341825 0:193829i
3.5 Lists 31

3.5.3 Inner and Outer Commands

Given fx1 ; x2 ; . . . ; xn g and fy1 ; y2 ; . . . ; yn g, how can one produce the lists

fx1 ; y1 ; x2 ; y2; x3 ; y3 ; . . . ; xn; yn g ;

ffx1 ; y1 g ; fx1 ; y2 g ; . . . ; fx1 ; yn g ; fx2 ; y1 g ; fx2 ; y2 g ; . . . ; fx2 ; yn g ; . . . ;
fxn ; y1 g ; fxn ; y2 g ; . . . ; fxn ; yn gg‹

The first list can be obtained by the built-in function Riffle or by using the built-in
function Inner[f,fa,bg,fx,yg,g].
Some examples introduced below show the outputs for different uses of the built-
in function Inner.
Clear[a,b,c,d,u,v,w,x,y,z]
Inner[f,fa,b,cg,fx,y,zg,g] (* This is a quite good definition of the
built-in function Inner *)
g[f[a,x],f[b,y],f[c,z]]

fInner[f,ffa,bg,fc,dgg,ffu,vg,fw,xgg,g],
Inner[List,ffa,bg,fc,dgg,ffu,vg,fw,xgg,List]g
{{{g[f[a,u],f[b,w]],g[f[a,v],f[b,x]]},{g[f[c,u],f[d,w]],g[f[c,v],f[d,x]]}},
{{{{a,u},{b,w}},{{a,v},{b,x}}},{{{c,u},{d,w}},{{c,v},{d,x}}}}}

fInner[List,fa,b,cg,fx,y,zg,List],
Flatten[Inner[List,fa,b,cg,fx,y,zg,List]],
Inner[Sequence,fa,b,cg,fx,y,zg,List] (* It coincides with the previous
one *)g
{{{a,x},{b,y},{c,z}},{a,x,b,y,c,z},{a,x,b,y,c,z}}

Inner[Sequence,fa,b,cg,fx,y,zg,List]==Riffle[fa,b,cg,fx,y,zg] (* It
works *)
fInner[Times,fa,bg,fx,yg,Plus], (* Some concrete applications of the
built-in function Inner follow *)
Inner[Power,fa,b,cg,fx,y,zg,Times]g
True
fa x+b y; ax by cz g

mm=ff1,2,3g,f4,5,6gg; (* The addition of two lists can be performed by

the built-in function Inner *)
Inner[Plus,mm[[1]],mm[[2]],List]
{5,7,9}
We study the case of input sequences of arbitrary length (not necessarily equal),
i.e., the case Inner[List,list1,list2,List].
32 3 Basic Steps to Mathematica

Clear[a,b]
myInner[a List,b List]:=(
n=Length@a;m=Length@b;
list=fg;min =Min[n,m];max =Max[n,m];
If[n+m>0,
Do[AppendTo[list,a[[k]]];
AppendTo[list,b[[k]]],fk,ming];
If[n<max,Do[AppendTo[list,b[[k]]],fk,min +1,mg],
Do[AppendTo[list,a[[k]]],fk,min +1,ng],
If[m==0,list=a,list=b]]];
Print[list]
)
We check our code.
x=fg;y=f1,2g; (* The first list is empty *)
myInner[x,y]
x=f1g;y=fg; (* The second list is empty *)
myInner[x,y]
x=f3,4g;y=f1,2g; (* Both lists are nonempty and of equal lengths *)
myInner[x,y]
x=f3,4,5g;y=f1,2g; (* Both lists are nonempty and of different
lengths *)
myInner[x,y]
x=f1g;y=f1,2,7,5g;
myInner[x,y]
x=fg;y=fg; (* Both lists are empty *)
myInner[x,y] f1; 2g
f1g
f3; 1; 4; 2g
f3; 1; 4; 2; 5g
f1; 1; 2; 7; 5g
fg
Now we focus on the second command, i.e., Outer[f,list1 ,list2 ,. . . ].
Clear[a,b,c,d,u,v,w,x,y,z,f]
Outer[f,fa,bg,fx,y,zg,f˛; ˇg] (* It is a simple definition of this
command *)
Outer[List,fa,bg,fx,y,zg,f˛; ˇg]
Flatten[Outer[List,fa,bg,fx,y,zg,f˛; ˇg],1] (* It treats only sublists at
level 1 in the lists as separate elements *)
Flatten[Outer[List,fa,bg,fx,y,zg,f˛; ˇg],2]
Flatten[Outer[List,fa,bg,fx,y,zg,f˛; ˇg],3]
Flatten[Outer[List,fa,bg,fx,y,zg,f˛; ˇg],3]==
Flatten[Outer[List,fa,bg,fx,y,zg,f˛; ˇg]]
3.5 Lists 33

ffff[a,x,˛],f[a,x,ˇ]g,ff[a,y,˛],f[a,y,ˇ]g,ff[a,z,˛],f[a,z,ˇ]gg,
fff[b,x,˛],f[b,x,ˇ]g,ff[b,y,˛],f[b,y,ˇ]g,ff[b,z,˛],f[b,z,ˇ]ggg
ffffa,x,˛g,fa,x,ˇgg,ffa,y,˛g,fa,y,ˇgg,ffa,z,˛g,fa,z,ˇggg,
fffb,x,˛g,fb,x,ˇgg,ffb,y,˛g,fb,y,ˇgg,ffb,z,˛g,fb,z,ˇgggg
fffa,x,˛g,fa,x,ˇgg,ffa,y,˛g,fa,y,ˇgg,ffa,z,˛g,fa,z,ˇgg,
ffb,x,˛g,fb,x,ˇgg,ffb,y,˛g,fb,y,ˇgg,ffb,z,˛g,fb,z,ˇggg
ffa,x,˛g,fa,x,ˇg,fa,y,˛g,fa,y,ˇg,fa,z,˛g,fa,z,ˇg,fb,x,˛g,
fb,x,ˇg,fb,y,˛g,fb,y,ˇg,fb,z,˛g,fb,z,ˇgg
fa,x,˛,a,x,ˇ,a,y,˛,a,y,ˇ,a,z,˛,a,z,ˇ,b,x,˛,b,x,ˇ,b,y,˛;b,y,ˇ;b,z,˛,b,z,ˇg
True

Outer[List,fa,b,c,dg,fu,v,wg,List] (* If the lists are of different lengths,

the system returns a message *)
Outer::ipnfm W Positive machine-sized integer or infinity expected at position 4 in
Outer[List,fa,b,c,dg,fu,v,wg,List] 
Outer[List,{a,b,c,d},{u,v,w},List]
Note that we can overcome this situation in the following way. Given the lists
fx1 ; x2 ; . . . ; xn g and fy1 ; y2 ; . . . ; ym g we want to obtain the next list

ffx1 ; y1 g ; fx1 ; y2 g ; : : : ; fx1 ; ym g ; fx2 ; y1 g ; fx2 ; y2 g ; : : : ; fx2 ; ym g ; : : : ;

fxn ; y1 g ; fxn ; y2 g ; : : : ; fxn ; ym gg

Here is the code:

Clear[a,b,n,m]
myOuter[a List,b List]:=(
n=Length@a;m=Length@b;
If[nm>0,
list=fg;k=1;
While[kn,
i=1;
While[i m,AppendTo[list,fa[[k]],b[[i]]g];i++];k++];
Print@list;
];
)

x=f1,2,6,8g;y=f4,5,7g;
myOuter[x,y]
{{1,4},{1,5},{1,7},{2,4},{2,5},{2,7},{6,4},{6,5},{6,7},{8,4},{8,5},{8,7}}
We note that
Clear[a,b,c,x,y,z]
Flatten[Outer[List,fa,b,cg,fx,y,zg],1]
{{a,x},{a,y},{a,z},{b,x},{b,y},{b,z},{c,x},{c,y},{c,z}}
coincides with
34 3 Basic Steps to Mathematica

CartesianProduct[fa,b,cg,fx,y,zg]
{{a,x},{a,y},{a,z},{b,x},{b,y},{b,z},{c,x},{c,y},{c,z}}
CartesianProduct[fa,b,cg,fx,y,z,tg] (* Clearly the lists are of
arbitrary lengths *)
{{a,x},{a,y},{a,z},{a,t},{b,x},{b,y},{b,z},{b,t},{c,x},{c,y},{c,z},{c,t}}

3.5.4 Again on the Third Sequence

In order to provide a shorter code, we reconsider the third sequence, (Sect. 3.3.3).
The sequence under discussion is

xnC1 D 1  xn2 ; x1 D a 2 0; 1Π; n 2 N :

This code uses the idea of a list more clearly.

With[fm=7g, (* Number of terms *)
u[1]=f.25,0.9,(-1+Sqrt[5])/2g;
u[n ]:=1-u[n-1]2 ;
Table[data@i=Table[fn,u[n][[i]]g,fn,mg],fi,3g];
fGrid[Table[fk,u[k][[1]],u[k][[2]],Simplify[u[k][[3]]]g,fk,mg],
Alignment!Left,Frame!All,
Background!fffNone,RGBColor[0,0,1,.4],
RGBColor[1,0,0,.5],RGBColor[0,1,0,.5]gg,Noneg],
ListLinePlot[Table[data@i,fi,3g],
PlotStyle!fBlue,Red,Darker[Green]g,
Mesh!Full,MeshStyle!fBlack,PointSize[Medium]g,
Ticks!
n n n p p o oo
Table[; f; mg]; f.25,”.25”g; 1C2 5 ; ” 1C2 5 ” f:9; ":9"g; 1 ,
ImageSize ! f280,200gg]g
]
See Fig. 3.4 below.

Fig. 3.4 Again the third sequence

Chapter 4
Sorting Algorithms

4.1 Introduction

An important problem of theoretical computer science is the sorting problem. The

statement of the problem (in a very simple but satisfactory form) is: given a set of n
real numbers, find an algorithm that sorts this set of entries in a minimum number
of steps (comparisons and interchanges). We mention a fundamental work on this
topic by D. Knuth [38].
We have already seen that Sort is a built-in Wolfram language function. Some
examples are introduced below:
Clear[a,b,c,d,x,y]
Sort[fd,b,c,ag], (* Sort implicitly provides a list in ascending order *)
Sort[f4,1,3,2,2g,Greater], (* Here we ask to sort in descending
order *)
Sort[ffa,2g,fc,1g,fd,3gg,#1[[2]]< #2[[2]]&], (* Sort also supports
pairs of numbers. A criterion is given after the list of objects to be
sorted *)
Sort[y,x2 ,x+y,y3 ],
Sort[f“cat”,“fish”,“catfish”,“Cat”g], (* Sort supports text *)
Sort[f“aa”,“abb”,“ba”,b,“aaa”g],
Sort[fPi,E,2,3,1,1-I,Sqrt@2g], (* Complex numbers are present *)
Sort[f-11,10,2,1,-4, 1-3Ig,Abs@#1<Abs@#2&]
2
{{a,b,c,d},{4,3,2,2,1},{{c,1},{a,2},{d,3}},{x
n ,y,y3 ,x+y}, o
p
{cat,Cat,catfish,fish},{aa,aaa,abb,b,ba}, 1; 1  i; 2; 3; 2; e;  ,
{1,2,1-3 i,-4,10,-11}}

© Springer International Publishing AG 2017 35

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_4
36 4 Sorting Algorithms

Sort@Join[Log@Range@5,Sqrt@Range@5] (* The sorting is

performed alphanumerically; Log[2] is considered a symbol but a
number *)
Sort@N@Join[Log@Range@5,Sqrt@Range@5] (* Numerical it
worked properly *)
Sort[Join[Log@Range@5,Sqrt@Range@5],Less] (* Adding an
ordering
n relation, Sort works properly *) o
p p p
0; 1; 2; 2; 3; 5; LogŒ2; LogŒ3; LogŒ4; LogŒ5
nf0:; 0:693147; 1:; 1:09861; 1:38629;
p
1:41421; 1:60944;
p p o
1:73205; 2:; 2:23607g
0; LogŒ2; 1; LogŒ3; LogŒ4; 2; LogŒ5; 3; 2; 5

RandomInteger[f-5,5g,f10,3g]; (* A list of ten triples whose

components are random integers in [-5,5] *)
Sort[%,Norm@#1<Norm@#2&] (* It sorts the previous list
ascending with respect to their norms *)
ff3; 0; 3g; f1; 4; 2g; f4; 2; 1g; f1; 4; 3g; f3; 1; 4g; f2; 3; 5g; f1; 5; 4g;
f5; 4; 4g; f4; 5; 5g; f5; 5; 5gg
Sort[fI,1+I,1-I,2+3 Ig,Re@#1<Re@#2&] (* It sorts ascending with
respect to the real parts *)
{i,1-i,1+i,2+3 i}
Sort[f4,1,3,2g,(Print@f#1,#2g;#1>#2)&] (* The comparisons
are printed step by step *)
f4; 1g
f3; 2g
f1; 3g
f4; 3g
f1; 2g
f4; 3; 2; 1g

4.2 Sorting Methods

We have mentioned that Mathematica has a built-it Wolfram language symbol for
the sorting problem, that is, Sort. It was used earlier. At the same time each one can
imagine his/her own sorting algorithm. Here we review five sorting methods.

4.2.1 Selection Sort

First, we find the smallest element in the input list and interchange it with the first
element. Then we find the next smallest element and interchange it with the second
element. Continue in this way until the entire sequence is sorted.
4.2 Sorting Methods 37

In other words we say that the selection sort divides the input list into two parts:
the sublist of items already sorted, which is built up from left to right located at the
left-hand side of the list (input), and the sublist of items remaining to be sorted that
occupy the rest of the list. Initially, the sorted sublist is null, and the unsorted sublist
is the entire input list.
Mathematica has its own selection sort. It requires the “Combinatorica‘”
package.
Needs[”CombinatoricaJ”]
RandomInteger[f1,20g,15]
SelectionSort[%,Less]
f3; 12; 16; 1; 3; 18; 3; 1; 7; 16; 5; 4; 19; 3; 17g
f1; 1; 3; 3; 3; 3; 4; 5; 7; 12; 16; 16; 17; 18; 19g
A code for selection sort follows. The elements which will be interchanged are
given in blue.
list=RandomSample[Range@10,7];
Print[”initial list ”,list] (* Initial input list *)
Do[If[list[[i]]>list[[j]],plist=list;
plist[[i]]=Style[plist[[i]],Blue,Bold,16];
plist[[j]]=Style[plist[[j]],Blue,Bold,16];
Print@plist;
list[[fi,jg]]=list[[fj,ig]]],fi,Length@listg,fj,i+1,Length@listg];
Print[”sorted list ”,list] (* Sorted output list *)
initial list f7; 6; 4; 3; 9; 1; 5g
{7,6,4,3,9,1,5}
{6,7,4,3,9,1,5}
{4,7,6,3,9,1,5}
{3,7,6,4,9,1,5}
{1,7,6,4,9,3,5}
{1,6,7,4,9,3,5}
{1,4,7,6,9,3,5}
{1,3,7,6,9,4,5}
{1,3,6,7,9,4,5}
{1,3,4,7,9,6,5}
{1,3,4,6,9,7,5}
{1,3,4,5,9,7,6}
{1,3,4,5,7,9,6}
{1,3,4,5,6,9,7}
sorted list {1,3,4,5,6,7,9}
Another short code for selection sort algorithm was introduced in [53, Chap. 10].
selectionSort[a List]:=Flatten[fMin@a,
If[Length@a>1,selectionSort@Drop[a,First@Position[a,Min@a]],]g];
We test it.
38 4 Sorting Algorithms

RandomInteger[f1,9g,15]
selectionSort@%
f6; 2; 2; 7; 2; 3; 8; 5; 7; 8; 2; 2; 4; 6; 5g
f2; 2; 2; 2; 2; 3; 4; 5; 5; 6; 6; 7; 7; 8; 8g
We introduce below a short code for the selection sort algorithm. The feature of
this code is that the keys to be interchanged are written in augmented bold type,
whereas the keys compared are given in augmented italic type.
appendto[char ,ii ,jj ]:=Block[fcg,
c=a;
c[[ii]]=Style[a[[ii]],char,Blue,16];
c[[jj]]=Style[a[[jj]],char,Blue,16];
AppendTo[aa,c]];
With[fm=10,n=10g, (* For a random selection of n terms in
Range[m] *)
a=RandomSample[Range@m,n]];
Print[”Input (unsorted) list : ”,%]
aa=f%%g;
Do[min=a[[i]];minj=i;
Do[If[a[[j]]<min,appendto[Italic,minj,j];
min=a[[j]];minj=j],fj,i+1,Length@ag];
If[i¤minj,appendto[Bold,i,minj]];
a[[fi,minjg]]=a[[fminj,ig]],fi,Length@a-1g
];
AppendTo[aa,a];
aa
Print[”Output (sorted) list : ”,a]
The sorting steps are presented in Fig. 4.1.

4.2.2 Insertion Sort

Let a D fa1 ; : : : ; an g be the list to be sorted. At the beginning and after each iteration
of the algorithm, the list consists of two parts: the first part a1 ; : : : ; ai1 is already
sorted, and the second part ai ; : : : ; an is still unsorted i 2 f1; : : : ; ng: In order to
insert element ai into the sorted part, it is compared with ai1 ; ai2 ; etc. When an
element aj with aj  ai is found, ai is inserted behind it. If no such element is found,
then ai is inserted at the beginning of the sequence. After inserting the element ai , the
length of the sorted part has increased by one. In the next iteration, aiC1 is inserted
into the sorted part. While at the beginning the sorted part consists of element a1
only, at the end it consists of all the elements a1 ; : : : ; an . This is the insertion sort
algorithm.
4.2 Sorting Methods 39

Fig. 4.1 Selection sort algorithm

We introduce below a short code in Mathematica for the insertion sort algorithm
in [53, Chap. 10]. The keys to be moved are written in italic type, whereas the key
to be inserted is given in bold blue type:
insertionSort[list List]:=Module[fterm,a=listg, (* Declaring it as
module, the code acts as a function that can be called several times *)
For[i=2,iLength@a,i++,term=a[[i]];j=i-1;
While[j1&&a[[j]]>term,
b=a;b[[i]]=Style[a[[i]],Italic,Blue,16];
b[[j]]=Style[a[[j]],Italic,Blue,16];
a[[j+1]]=a[[j]];
AppendTo[aa,b];j- -];
a[[j+1]]=term;b=a;
b[[j+1]]=Style[a[[j+1]],Bold,Blue,16];
AppendTo[aa,b]];
AppendTo[aa,a]];
aa=ff5,2,4,6,1gg; (* Here is the input (unsorted) list; the list aa
will contain all the steps of the algorithm *)
insertionSort@Flatten[aa,1] (* The last element of the output
consists in the corresponding sorted list *)
The sorting steps are presented in Fig. 4.2.
We can visualize the insertion sort algorithm by converting the successive lists
into permutation plots with animation.
40 4 Sorting Algorithms

Fig. 4.2 Insertion sort algorithm

Fig. 4.3 Insertion sort with permutations and animation

permutationPlot[list List]:=ListPlot[list,
PlotRange!ff.5,Length@list+.5g,f.6,Length@list+.6gg,
PlotStyle!PointSize[.4/Length@list],Axes!None,
FrameTicks!None,Frame!True,AspectRatio!1,
ImageSize!3.2Length@list(Length@list-1)]
With[fm=5,n=5g,
a=RandomInteger[f1,mg,n];
Print[”Input (unsorted) sequence : ”,a];
aa=fag;
If[n>1,
Do[min=a[[i]];minj=i;appendto=0;
Do[If[a[[j]]<min,min=a[[j]];minj=j;appendto=1],fj,i+1,ng];
If[appendto==1,a[[fi,minjg]]=a[[fminj,ig]]; AppendTo[aa,a]],fi,n-1g
];
]]
Print[aa]
permutationPlot[#]&/@aa
ListAnimate[%] f2; 1; 5; 5; 1g
Input (unsorted) sequence : f2; 1; 5; 5; 1g
ff2; 1; 5; 5; 1gg
ff2; 1; 5; 5; 1g; f1; 2; 5; 5; 1g; f1; 1; 5; 5; 2g; f1; 1; 2; 5; 5gg
The sorting steps are presented in Fig. 4.3.
4.2 Sorting Methods 41

4.2.3 Mergesort

Mergesort algorithm is based on two steps, [53, Chap. 10]. The first one splits
the input list into smaller groups by halving it until the groups had only one or two
elements. Then it merges the groups back together so that their elements are in order.
This is an algorithm of the “divide and conquer” type. Below we introduce a code
for mergesort.
merge[left List,right List]:=
Module[flindex=1,rindex=1g,
Table[Which[lindex>Length@left,right[[rindex++]],
rindex>Length@right,left[[lindex++]],left[[lindex]]right[[rindex]],
left[[lindex++]],True,right[[rindex++]]],fLength@left+Length@rightg]]
mergeSort[m List]:=Module[fmiddleg,If[Length@m2,
middle=Ceiling[Length@m/2];
Apply[merge,Map[mergeSort,Partition[m,middle,middle,f1,1g,fg]]],m]]
a=RandomInteger[f1,13g,15]
mergeSort@a
f8; 8; 7; 1; 5; 4; 4; 5; 1; 1; 7; 8; 6; 2; 12g
f1; 1; 1; 2; 4; 4; 5; 5; 6; 7; 7; 8; 8; 8; 12g

4.2.4 Heapsort

We mention that there exists a built-in function heapsort which performs sorting
using this algorithm.
Needs[“CombinatoricaJ”]
HeapSort[f9,5,1,8,6,4,3,7,2g]
f1; 2; 3; 4; 5; 6; 7; 8; 9g
The heapsort code introduced below is based on the version of Floyd’s algorithm
published in [26] and the corresponding Mathematica code given by [43, p. 99].
A list a D fa1 ; : : : ; an g is said to be a heap if abi=2c  ai , for all i 2 f2; : : : ; ng.
We present the heapsort algorithm in full detail.
ı Heapsort is a sorting algorithm of class O.n ln.n//.
ı It is also known as sorting by the “method of heaps.”
ı It is not recursive but is comparable to the quicksort.
ı Heapsort is a sorting algorithm “in situ,” that is, it does not require additional
structures; the sorting is performed in the space of the input list. We mention that
there are many heapsort implementations, but it is not our aim to review them.
42 4 Sorting Algorithms

ı Heapsort is similar to some extent to the selection sort.

ı At each step, the smallest element is selected, moved behind the input list, and
ignored along the next steps. Thus, the amount of work is gradually reduced.
ı In order to find the smallest element of the input, it is transformed into a heap.
Each input list a in the next grid
Clear[a];
headerheap=Join[f”index”g,Table[k,fk,9g]];
list=Join[fag,Table[ak ,fk,9g]];
Grid[Join[fheaderheapg,flistg],Frame!All,Alignment!Center,
FrameStyle!Thin,Background!fNone,fGreen,Yellowgg]
can be transformed into a binary tree as the figure below shows.
See Fig. 4.4.
TreePlot[fa1 ! a2 ; a1 ! a3 ; a2 ! a4 ; a2 ! a5 ; a3 ! a6 ; a3 ! a7 ;
a4 ! a8 ;a4 ! a9 g,Automatic,VertexLabeling!True,
PlotStyle!Black,ImageSize!50]
See Fig. 4.5.
A binary tree is a tree-like structure that is rooted, and each vertex has at most
two children, and each child of a vertex is designed as its left or right child. There
is a vast literature on binary trees; we only mention [38, pp. 145–146].
ı We note that the binary tree just obtained has the property that the index of each
element is greater or equal to the index of its parent.
ı a1 is the root of the tree, so it has to be the minimal element of the input.
ı In a heap the minimal element is located on the top of it.
ı If one of the conditions ai  a2i or ai  a2iC1 is not satisfied, then we interchange
ai with min fa2i ; a2iC1 g.
ı In order to increase the speed of the algorithm, we perform the previous step from
right to left.

Fig. 4.4 The list a

Fig. 4.5 The binary tree of a

4.2 Sorting Methods 43

ı When the input is transformed step by step into a heap, the smallest element is
at the top. This is interchanged with the last one, that is, an , which is at the top
now, and we put it behind the smallest one, reducing by one the length of the new
input.
We consider an example and perform the heapsort algorithm. Let a D .9; 5; 1; 8;
6; 4; 3; 7; 2/ be the input sequence (list) to be sorted. The binary tree representation
of a is given by the Fig. 4.6.
TreePlot[f9 ! 5; 9 ! 1; 5 ! 8; 5 ! 6; 1 ! 4; 1 ! 3; 8 ! 7; 8 ! 2g;
Automatic,VertexLabeling!True,PlotStyle!Black,ImageSize!150]
The sorting steps start as follow.
F We interchange 8 ! 2.
F We interchange 5 ! 2.
F We interchange 9 ! minf2; 1g.
The steps introduced above are visualized below.
TreePlot[#,Automatic,9,VertexLabeling!True,PlotStyle!Black,
ImageSize!150]&/@
ff9 ! 5; 9 ! 1; 5 ! 8; 5 ! 6; 1 ! 4; 1 ! 3; 8 ! 7; 8 ! 2; 2 ! 8g,
f9 ! 5; 9 ! 1; 5 ! 2; 5 ! 6; 1 ! 4; 1 ! 3; 2 ! 7; 2 ! 8; 2 ! 5g,
f9 ! 2; 9 ! 1; 2 ! 5; 2 ! 6; 1 ! 4; 1 ! 3; 5 ! 7; 5 ! 8; 1 ! 9gg
The first steps are exhibited in Fig. 4.7.
Next we go on as follows.
F We interchange 9 ! minf4; 3g. Such interchanges will go on until 9 will arrive
on the last row, and thus we get a heap. The smallest element lies on top of the
binary tree.
F We interchange 1 with the last element of the last row, that is, 1 ! 8.
F We remove 1 and the link between 1 and 5.
Now we are in the situation introduced in the figure below.

Fig. 4.6 The binary tree of

new a
44 4 Sorting Algorithms

Fig. 4.7 The first steps of the heapsort of a

Fig. 4.8 The second steps of the heapsort of a

fTreePlot[f1 ! 2; 1 ! 9; 2 ! 5; 2 ! 6; 5 ! 7; 5 ! 8; 9 ! 4; 9 ! 3; 3 !
9g;
Automatic,1,VertexLabeling!True,PlotStyle!Black,ImageSize!150],
Show[Graphics[ffPointSize[Medium],Point[f0,0g]g,
Line[fff1,0g,f1.6,0g,f1.6,2.6gg,ff1,-.08g,f1.68,-.08g,f1.68,2.6ggg]g],
TreePlot[f1 ! 2; 1 ! 3; 2 ! 5; 2 ! 6; 5 ! 7; 5 ! 8; 3 ! 4; 3 ! 9g;
Automatic,1,VertexLabeling!True,PlotStyle!Black],PlotRange!All,
ImageSize!145],
Show[Graphics[ffPointSize[Medium],Point[f0,0g]g,
Line[fff1,0.1g,f1.6,0.1g,f1.6,-.2gg,ff1,.02g,f1.52,.02g,f1.52,-.2ggg]g],
TreePlotŒf8 ! 2; 8 ! 3; 2 ! 5; 2 ! 6; 5 ! 7; 5 ! 1; 3 ! 4; 3 ! 9g;
Automatic,8,VertexLabeling!True,PlotStyle!Black],PlotRange!All,
ImageSize!145]g
The second steps are exhibited in Fig. 4.8.
We further work with the remained binary tree whose length is decreased by 1.
F Due to the previous steps, the vertices 5; 3 and 2 satisfy the condition abi=2c  ai .
Furthermore, this condition is also satisfied by the vertices 8 and 2. Therefore,
we interchange 8 ! 2.
F We interchange 8 ! 5.
F We interchange 8 ! 7.
4.2 Sorting Methods 45

Fig. 4.9 The third steps of the heapsort of a

The steps that we made can be visualized below.

fTreePlotŒf8 ! 2; 8 ! 3; 2 ! 5; 2 ! 6; 5 ! 7; 3 ! 4; 3 ! 9; 2 ! 8g;
Automatic,8, VertexLabeling ! True; PlotStyle ! Black;
ImageSize ! 150;
TreePlotŒf2 ! 8; 2 ! 3; 8 ! 5; 8 ! 6; 5 ! 7; 3 ! 4; 3 ! 9; 5 ! 8g;
Automatic; 2; VertexLabeling ! True; PlotStyle ! Black;
ImageSize ! 150;
TreePlotŒf2 ! 5; 2 ! 3; 5 ! 8; 8 ! 7; 5 ! 6; 3 ! 4; 3 ! 9; 7 ! 8g;
Automatic; 2; VertexLabeling ! True; PlotStyle ! Black;
ImageSize ! 150g
The third steps are exhibited in Fig. 4.9.
The last tree is a heap; thus, we proceed in the following way:
F We interchange 2 ! 8.
F We remove 2 from the tree and the link between 2 and 7.
F The tree just obtained is ready for a new step of the algorithm.
These steps are shown in the figure below.
fShow[Graphics[fLine[fff0,.058g,f1.18,.058g,f1.18,2.6gg,
ff0,-.05g,f1.27,-.05g,f1.27,2.6ggg]g],
TreePlotŒf2 ! 5; 2 ! 3; 5 ! 7; 7 ! 8; 5 ! 6; 3 ! 4; 3 ! 9g;
Automatic,2,VertexLabeling!True,PlotStyle!Black],
ImageSize!145]g
Show[Graphics[{Line[{{{0,.058},{.65,.058},{.65,-.15}},
{{0,-.03},{.55,-.03},{.55,-.15}}}]}],
TreePlotŒf8 ! 5; 8 ! 3; 5 ! 7; 5 ! 6; 3 ! 4; 3 ! 9; 7 ! 2g,Automatic,8,
VertexLabeling!True,PlotStyle!Black],ImageSize!145],
TreePlotŒf8 ! 5; 8 ! 3; 5 ! 7; 5 ! 6; 3 ! 4; 3 ! 9g, Automatic,8,
VertexLabeling!True, PlotStyle!Black,ImageSize!145]}
46 4 Sorting Algorithms

Fig. 4.10 The fourth steps of the heapsort of a

Fig. 4.11 The fifth steps of the heapsort of a

The fourth steps are exhibited in Fig. 4.10.

We continue the algorithm.
F We interchange 8 ! minf5; 3g.
F We interchange 8 ! 4.
F We have a heap and interchange 9 ! 3.
These steps are visualized in the next figure.
{TreePlot[f8 ! 5; 8 ! 3; 5 ! 7; 5 ! 6; 3 ! 4; 3 ! 9; 3 ! 8g,
Automatic,8,VertexLabeling!True,PlotStyle!Black,ImageSize!145],
TreePlot[f3 ! 5; 5 ! 7; 5 ! 6; 3 ! 8; 8 ! 4; 8 ! 9; 4 ! 8g,
Automatic,3,VertexLabeling!True,PlotStyle!Black, ImageSize!145],
Show[Graphics[{Line[fff2.4,.058g,f2.4,2-.4g,f1+.2,2-.4gg,
ff2.48,.058g,f2.48,2-.32g,f1+.2,2-.32ggg]g],
TreePlot[f3 ! 5; 5 ! 7; 5 ! 6; 3 ! 4; 4 ! 8; 4 ! 9g,Automatic,3,
VertexLabeling!True,PlotStyle!Black],ImageSize!145]}
The fifth steps are exhibited in Fig. 4.11.
We go on to the next steps.
F We remove 3 from the tree and the link between 3 and 4.
F We interchange 9 ! 4.
F We interchange 9 ! 8 and get a heap.
4.2 Sorting Methods 47

Fig. 4.12 The sixth steps of the heapsort of a

These steps are visualized below.

{Show[Graphics[fLine[fff2.5,.058g,f2.8,.058g,f2.8,-.1gg,
ff2.5,-.01g,f2.72,-.01g,f2.72,-.1gg]g]],
TreePlot[f9 ! 5; 9 ! 4; 5 ! 7; 5 ! 6; 4 ! 8; 4 ! 3g;Automatic,9,
VertexLabeling!True,PlotStyle!Black,ImageSize!145]],
TreePlot[f9 ! 5; 9 ! 4; 5 ! 7; 5 ! 6; 4 ! 8; 4 ! 9g;Automatic,9,
VertexLabeling!True,PlotStyle!Black, ImageSize!145],
TreePlot[f4 ! 5; 4 ! 9; 5 ! 7; 5 ! 6; 9 ! 8; 8 ! 9g;Automatic,4,
VertexLabeling!True,PlotStyle!Black,ImageSize!145]}
The sixth steps are exhibited in Fig. 4.12.
We go on to the next steps.
F We interchange 9 ! 4.
F We remove 4.
F We interchange 9 ! 5 and get a heap.
These steps are visualized in the figure below.
{Show[Graphics[Line[fff1.75,.058g,f1.09,0.058g,f1.09,1.62gg,
ff1.75,.01g,f1.05,.01g,f1.05,1.62ggg]g],
TreePlot[f4 ! 5; 4 ! 8; 5 ! 7; 5 ! 6; 8 ! 9g;Automatic,4,
VertexLabeling!True, PlotStyle!Black],ImageSize!130],
Show[Graphics[fLine[fff1.75,.058g,f2.15,0.058g,f2.15,-.15gg,
ff1.75,-.03g,f2.08,-.03g,f2.08,-.15ggg]g],
TreePlot[f9 ! 5; 9 ! 8; 5 ! 7; 5 ! 6; 8 ! 4g;Automatic,9,
VertexLabeling!True, PlotStyle!Black],ImageSize!150],
TreePlot[f9 ! 5; 9 ! 8; 5 ! 7; 5 ! 6; 5 ! 9g;Automatic,9,
VertexLabeling!True,PlotStyle!Black,ImageSize!130]}
The seventh steps are exhibited in Fig. 4.13.
We continue to the next steps.
F We interchange 9 ! 5.
F We interchange 9 ! 6 and get a heap.
F We interchange 5 ! 9 and remove 5 and the link between 5 and 9.
F We interchange 9 ! 6.
48 4 Sorting Algorithms

Fig. 4.13 The seventh steps of the heapsort of a

Fig. 4.14 The eighth steps of the heapsort of a

The visualizations of these steps follow.

{TreePlot[f5 ! 9; 5 ! 8; 9 ! 7; 9 ! 6g;Automatic,5,
VertexLabeling!True,PlotStyle!Black,ImageSize!120],
TreePlot[f5 ! 6; 5 ! 8; 6 ! 7; 6 ! 9g;Automatic,5,
VertexLabeling!True,PlotStyle!Black,ImageSize!120],
Show[Graphics[fLine[fff1.52,.830g,f1.85,.830g,f1.85,.72gg,
ff1.52,.9g,f1.908,.9g,f1.908,.72ggg]g],
TreePlot[f9 ! 6; 9 ! 8; 6 ! 7; 6 ! 5g;Automatic,9,
VertexLabeling!True, PlotStyle!Black,ImageSize!120],
TreePlot[f9 ! 6; 9 ! 5; 6 ! 7; 6 ! 8g;Automatic,9,
VertexLabeling!True, PlotStyle!Black],ImageSize!120],
TreePlot[f9 ! 6; 6 ! 7; 6 ! 8; 6 ! 9g;Automatic,9,
VertexLabeling!True, PlotStyle!Black,ImageSize!110]}
The eighth steps are exhibited in Fig. 4.14.
We continue to the next steps.
F We interchange 9 ! 7 and get a heap.
F We interchange 9 ! 6.
F We remove 6 from the tree and the link between 6 and 7.
4.2 Sorting Methods 49

Fig. 4.15 The ninth steps of

the heapsort of a

The visualizations of these steps follow.

{TreePlot[f6 ! 9; 9 ! 7; 9 ! 8; 7 ! 9g;Automatic,6,
VertexLabeling!True,PlotStyle!Black,ImageSize!120],
Show[Graphics[fLine[fff.04,0g,f.04,2-.18g,f.38,2-.18gg,
ff.0,0g,f.0,2-.12g,f.38,2-.12ggg]g],
TreePlot[f6 ! 7; 7 ! 9; 7 ! 8g;Automatic,6,VertexLabeling!True,
PlotStyle!Black],ImageSize!90],
Show[Graphics[fLine[fff.04,0.04g,f.44,0.04g,f.44,-.1gg,
ff.04,-0.03g,f.4-.01,-0.03g,f.4-.01,-.1ggg]g],
TreePlot[f9 ! 7; 7 ! 6; 7 ! 8g;Automatic,9,VertexLabeling!True,
PlotStyle!Black],ImageSize!90]}
The ninth steps are exhibited in Fig. 4.15.
We continue to the next steps.
F We interchange 9 ! 7.
F We interchange 9 ! 8 and get a heap.
F We interchange 9 ! 7.
F We remove 7 from the tree and the link between 7 and 8.
These steps are visualized in the next figure.
{TreePlot[f9 ! 7; 7 ! 8; 7 ! 9g;Automatic,9,
VertexLabeling!True,PlotStyle!Black,ImageSize!f60,150g],
TreePlot[f7 ! 9; 9 ! 8; 8 ! 9g;Automatic,7,
VertexLabeling!True,PlotStyle!Black,ImageSize!f60,150g],
Show[Graphics[fLine[fff.04,0.03g,f.2,.03g,f.2,2-.051g,f.04,2-.051gg,
ff.04,-0.03g,f.3-.05,-.03g,f.3-.05,2g, f.04,2ggg]g]],
TreePlot[f7 ! 8; 8 ! 9g;Automatic,7,VertexLabeling!True,
PlotStyle!Black],ImageSize!f60,150g]
Show[Graphics[fLine[fff.04,0.03g,f.27,.03g,f.27,-.08gg,
ff.04,-0.03g, f.3-.08,-.03g,f.3-.08,-0.08ggg]g],
TreePlot[f9 ! 8; 8 ! 7g,Automatic,9,VertexLabeling!True,
PlotStyle!Black],ImageSize!f50,150g]}
50 4 Sorting Algorithms

Fig. 4.16 The tenth steps of

the heapsort of a

Fig. 4.17 The eleventh steps

of the heapsort of a

The tenth steps are exhibited in Fig. 4.16.

We continue with the final steps.

F We interchange 9 ! 8 and get a heap.
F We interchange 9 ! 8 and remove 8 and the link between 8 and 9.
F The last and largest element remained, i.e., 9.
These final steps are visualized in the next figure.
{TreePlot[f9 ! 8; 8 ! 9g,Automatic,9,
VertexLabeling!True,PlotStyle!Black,ImageSize!f40,80g],
Show[Graphics[fPoint[f0,0g],Point[f0,1g]g],TreePlot[f8 ! 9g,
Automatic,8,VertexLabeling!True,PlotStyle!Black],
ImageSize!f50,80g],
Show[Graphics[fLine[fff.04,0.03g,f.27,.03g,f.27,-.08gg,
ff.04,-0.03g,f.3-.08,-.03g,f.3-.08,-0.08ggg]g],
TreePlot[f9 ! 8g,Automatic,9,VertexLabeling!True,
PlotStyle!Black],ImageSize!f40,80g],
Show[Graphics[
fLine[ff-.03,-0.04g,f.03,-.04g,f.03,.04g,f-.03,0.04g,f-.03,-0.04gg],
Text[9,f0,0g]g],ImageSize!f18,33g]}
The eleventh steps are exhibited in Fig. 4.17
The heapsort algorithm sorts the list a D fa1 ; : : : ; an g, requiring no more than
2 .2p  2/ .p  1/, or approximately 2n .Log2 Œn -1) comparisons and half as many
interchanges in the worst case to sort n D 2p  1 items. The algorithm is easier to
follow if a is thought of as a tree, with abi=2c the father of ai for 1 < i < n:
4.2 Sorting Methods 51

Now a Mathematica code of the above introduced algorithm follows.

heaps=fg; (* We collect the partial heaps *)
heapSort[p List]:=Module[fheap=heapify@p,ming,
Append[Table[min =First@heap;
heap[[1]]=heap[[n]];

heap=heapify[Drop[heap,-1],1];AppendTo[heaps,heap];
min,fn,Length@p,2,-1g],Max@heap]]/;Length@p>0
heapify[p List]:=Module[fj,heap=pg,Do[heap=heapify[heap,j],
fj,Quotient[Length@p,2],1,-1g];heap]
heapify[p List,k Integer]:=Module[fhp=p,i=k,m,n=Length@pg,
While[(m=2 i)n,
If[m<n&&Less[hp[[m+1]],hp[[m]]],m++];
If[Less[hp[[m]],hp[[i]]],hp[[fi,mg]]=hp[[fm,ig]]];
i=m];
hp
]

a=f9,5,1,8,6,4,3,7,2g;
heapSort@a
Table[heaps[[i]],fi,1,Length@heapsg] (* Here is the list of partial
heaps *)
{1,2,3,4,5,6,7,8,9}
{{2,5,3,7,6,4,9,8},{3,5,4,7,6,8,9},{4,5,8,7,6,9},{5,6,8,7,9},{6,7,8,9},
{7,9,8},{8,9},{9}}
trees=fg;
Do[a=heaps[[m]];
n=Length@a
If[n==0,tree=fg;Goto[”print”],
If[n==1,tree=a;Goto[”print”],
tree=fg;
If[n==2,AppendTo[tree,a[[1]]!a[[2]]];Goto[”print”],
tree=fg;j=1;
Do[
j++;AppendTo[tree,a[[i]]!a[[j]]];
j++;If[j>n,Goto[“print”]];
AppendTo[tree,a[[i]]!a[[j]]],fi,bn=2cg]
] ] ];
Label[”print”];
AppendTo[trees,tree],
fm,Length@heaps-1g];
Table[TreePlot[trees[[k]],Automatic,trees[[k,1,1]],VertexLabeling!True,
ImageSize! 10.15  bk=2c/],fk,Length@treesg]
The steps of the above algorithm are exhibited in Fig. 4.18.
52 4 Sorting Algorithms

Fig. 4.18 The steps of the above heapsort algorithm for a

4.2.5 Quicksort

The quicksort algorithm was developed by C. A. R. Hoare and published in [34].

A series of interesting writings on this topic were published by R. Sedgewick. We
mention here [61] and the reference therein. Some details may be found in [53,
Chap. 10].
The quicksort algorithm starts by selecting an element, called the pivot point,
from a list and then divides the list into two, with all elements less than the pivot
element moved to or remained on its left and all elements greater than the pivot
element moved to or remained on its right. The process is then continuously repeated
in each sublist until the initial list is sorted, ascending in our case. The initial list is
sorted when the length of each sublist is 1.
A short Mathematica code of the quicksort algorithm follows.
Clear[x]
quickSortcases[x List]:=
Module[fpivotg,
If[Length@x1,Return@x]; (* If the list is empty or it contains a
single element, the algorithm stops *)
pivot=RandomChoice@x; (* It selects a random pivot from the list *)
Flatten@f (* It flattens the three sorted lists *)
quickSortcases@Select[x,#<pivot&], (* It selects all the
elements smaller than the pivot *)
4.2 Sorting Methods 53

Cases[x,y /;y==pivot],
quickSortcases@Cases[x,y /;y>pivot]g
]

With[fm=10,n=12g,
list=RandomChoice[Range@m,n]]
quickSortcases@list
f5; 3; 3; 4; 2; 1; 4; 7; 8; 8; 4; 9g
f1; 2; 3; 3; 4; 4; 4; 5; 7; 8; 8; 9g
Another quicksort code is introduced below.
quickSortselect[fg]=fg;
quickSortselect[fx ,xs g]:=Join[ (* It joins all the elements less
than, equal to, or greater than x *)
quickSortselect@Select[fxsg,#x&], (* It applies quickSortselect
to all picked elements in the list xs less than or equal to x *)
fxg,
quickSortselect@Select[fxsg,#>x&] (* It applies quickSortselect to
all picked elements in the list xs greater than x *)
];

With[fm=10,n=12g,
list=RandomChoice[Range@m,n]]
quickSortselect@list
f1; 5; 8; 7; 3; 6; 10; 4; 2; 7; 4; 9g
f1; 2; 3; 4; 4; 5; 6; 7; 7; 8; 9; 10g
Suppose we are interested in the time needed to sort a list.
Timing[expr] evaluates expr and returns a list of the time in seconds used,
together with the result obtained. Then, one can use the next procedure:
With[fm=10,000,n=15,000g,
list=RandomChoice[Range@m,n]]; (* The input list *)
Timing[quickSortcases@list;]
Timing[quickSortselect@list;]
f1:15625; Nullg
f0:65625; Nullg
Based on this simple test, we conclude that the quickSortselect code is clearly
faster than the quickSortcases code.
Chapter 5
Functions

5.1 Definitions

We have already seen some built-in functions along the previous chapters. We can
define our own functions depending on our needs or interests. Thus, we have the
freedom to use plenty of mathematics. First, we consider some simple examples.
f[x ]:=Sin[x]+x2 (* A function is defined f(x) = sin x + x2 *)
g[x ]:=f[f[x]+1]+2 (* A composition of a function g(x) = f(f(x) + 1) +
2 *)
fg[2], (* The value of g at an integer; it is returned noncomputed *)
FullSimplify@g[2], (* It is still noncomputed *)
TraditionalForm@g[2],
g[2.], (* The value is given with five decimal digits *)
g[2]//N,
N[g[2],15], (* The value is given with 15 decimals *)
N[#,15]&/@g[2]g (* The same result as before *)
{2+(5+Sin[2])2 +Sin[5+Sin[2]],2+(5+Sin[2])2 +Sin[5+Sin[2]],
2+(5+sin(2))2 +sin(5+sin(2)),36.5546,36.5546,36.5545586117012,
36.5545586117012}
It is not difficult to construct a function of several variables. An easy example
is the code of finding the solutions of the second-degree polynomial with real
coefficients.

© Springer International Publishing AG 2017 55

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_5
56 5 Functions

Clear[a,b,c] (* We want to avoid any interference of previous values of a,

b, or c *) n p p o
b2 4ac bC b2 4ac
rootsseconddegree[a ,b ,c ]:= b 2a ; 2a
;
frootsseconddegree[1,2,1], (* Now we test our formula for different sets
of data *)
u=rootsseconddegree[1,2,-4],
rootsseconddegree[1,2,4]g
n n  p  p o
f1; 1g; 12 2  2 5 ; 12 2 C 2 5 ;
n  p 1 p oo
1
2
2  2i 3 ; 2
2 C 2i 3

Suppose we need the second solution of the second equation, first under a simple
form and then under a numerical form. We write
Simplify@u[[2]]
N@u[[2]]
p
1 C 5
1:23607
There exists a built-in function which can solve the second-degree equation
easily. This is the built-in function Reduce.
Reduce[a x2+b x+c==0,x]
 p p
b2 4ac bC b2 4ac
a ¤ 0&& x== b 2a k 2a
 
 a==0&&b ¤ 0&&x==  c  .c==0&&b==0&&a==0/
b
We can use this function for constrained equations.
Clear[x,y]
 
Reduce x2  y2 DD
  8; fx; p
yg; Reals
p p
x  2 2&& y DD  8 C x2 ky DD 8 C x2 k
 p  p p
x  2 2&& y DD  8 C x2 ky DD 8 C x2

Clear[x,y,xy]
 
Reduce x2 y2 DD 8; fx; yg; Integers (* We look for integer solutions *)
xy=fx,yg//.fToRules[%]g (* The list of solutions *)
fx,yg=xy[[1]] (* We extract the first pair of solutions *)
.x DD 3&&y DD 1/k.x DD 3&&y DD 1/k
.x DD 3&&y DD 1/k.x DD 3&&y DD 1/
ff3; 1g; f3; 1g; f3; 1g; f3; 1gg
f3; 1g
5.1 Definitions 57

5.1.1 Functions with Conditions

We define the discontinuous function

8
ˆ
ˆ
<x; if jxj < 1;
f .x/ D sin x; if 1  x < 2;
ˆ
:̂cos x; otherwise:

Using the built-in function Which (making the graph continuous), we write
f[x ]:=Which[Abs@x<1, -x,
1 Abs@x<2, Sin@x,
True, Cos@x
];
Plot[f@x,fx,-3,3g,ImageSize!200] (* A graph is useful to see how Mathe-
matica interprets the built-in function Which. The graph is continuous; the
function is transformed into a set-valued function *)
f/@f0.5,Pi/2,Pig (* The values of the function f at some values of the
argument *)
Limit[f@x,x! 2] (* 2 is a discontinuity point. The limit is considered side
limit from the right-hand side *)
The graph of the above discontinuous function is exhibited by the built-in
function Which in Fig. 5.1.
We note that if the function is discontinuous at certain points, its graph is
continuous, and thus the function which represents that graph is a set-valued
function or a multifunction.

Fig. 5.1 The graph and some

values by the built-in function
Which
58 5 Functions

Fig. 5.2 The graph and some

values by the built-in function
Piecewise

Using the built-in function Piecewise, the previous function looks like
g[x ]:=Piecewise[ff-x,Abs@x<1g,fSin@x,1Abs@x<2gg,Cos@x]
Plot[g@x,fx,-3,3g,ImageSize!200]
g/@f0.5,Pi/2,Pig
Limit[g@x,x!2]
The graph of the above discontinuous function is exhibited by the built-in
function Piecewise in Fig. 5.2.
We note that the built-in function Piecewise supplies the true graph of the
function which in the present case has points of discontinuity.
The built-in function If supplies a continuous graph even if the function is
discontinuous. Its output coincides with the one given by Which.
h[x ]:=If[Abs@x<1,-x,If[1Abs@x<2,Sin@x,Cos@x]];
Plot[h@x,fx,-3,3g,ImageSize!200]
h/@f0.5,Pi/2,Pig
Limit[h@x,x!2] (* The limit is again considered side limit from the right-
hand side *)
The graph of the above discontinuous function is exhibited by the built-in
function If in Fig. 5.3.
p[x ]:=Piecewise[ff0,x<0g,f1,x0gg]
Plot[p@x,fx,-1,1g,PlotStyle!Thick,ImageSize!150,
Ticks!ff-1,1g,f1gg]
p/@f-1,0,1g (* At the discontinuity point x=0, it is returned the
value 1 *)
Limit[p@x,x!0],Limit[p@x,x!0,Direction!-1],
Limit[p@x,x!0,Direction!1] (* Because at x = 0 the function is
discontinuous, the limit is considered side limit from the right-hand
side *)
Another discontinuous function is exhibited by the built-in function Piecewise in
Fig. 5.4.
5.1 Definitions 59

Fig. 5.3 The graph and some

values by the built-in function
If

Fig. 5.4 Another graph and

some values by the built-in
function Piecewise

Fig. 5.5 The signum

function

A similar discontinuous function is the signum (sign) function.

Plot[Sign@x,fx,-2,2g,PlotStyle!fDarker[Green],Thickg,
Ticks!ff-2,2g,f-1,1gg,TicksStyle!Directive[Red,Bold,12],
ImageSize!200]
The signum function is presented in Fig. 5.5 above.
Let us examine another example. In the following code appears /; and it is a
shorthand of If.
Clear[f,g,h]
f[x ]:=Sqrt@x/;x>0
g[x ]:=Sqrt@x/;x<0
60 5 Functions

Fig. 5.6 Interpolation on

some finite and distinct data

h[x ]:=Sqrt@x2 /;x<0

ff[4],f[-4],g[4],g[-4],h[2],h[-2]g
{2,f[-4],g[4],2 i,h[2],2}
Consider the same function defined by If. The two functions work identically for
x > 0: Otherwise, If returns Null.
g1[x ?Positive]:=Sqrt@x
g2[x ]:=If[x>0,Sqrt@x]
ffg1[4],g2[4]g,fg1[-1],g2[-1]gg
ff2; 2g; fg1Œ1; Nullgg
Now an example of interpolation of some data follows.
data=Table[fi,RandomReal[i]g,fi,9g]
ListLinePlot[data,Epilog!fBlack,Line[data],PointSize[Medium],Red,
Point[data]g,InterpolationOrder!3,ImageMargins!10,
ImageSize!250]
ff1; 0:652455g; f2; 0:794146g; f3; 0:207637g; f4; 0:40818g; f5; 1:3127g;
f6; 4:69005g; f7; 4:38125g; f8; 0:94469g; f9; 7:26584gg
Below is presented an interpolation of some data in Fig. 5.6.

5.1.1.1 Collatz Conjecture

Collatz conjecture states that the algorithm consisting in substituting a given positive
integer n by Floor[n/2] provided n is even and by 3n C 1 provided n is odd reaches
eventually 1, [47, Chap. 1]. Formally we write
(
bn=2c; n is even;
f .n/ D
3n C 1; n is odd;
5.1 Definitions 61

and state that for any natural n, there exists a natural m such that f iterated m times
at n reaches 1, i.e., f .f .: : : f .n/// D 1.
„ ƒ‚ …
m times
For example, starting with number 7, we write the Collatz successive iterations
obtaining the following sequence:
7 ! 22 ! 11 ! 34 ! 17 ! 52 ! 26 ! 13 ! 40 ! 20 !
10 ! 5 ! 16 ! 8 ! 4 ! 2 ! 1
A code for a step of this algorithm can be written as follows:
f[n Integer]:=If[EvenQ[n],n/2,3n+1]
f[16] (* Because 16 is even, we divide it by 2 *)
f[17] (* Because 17 is odd, we multiply it by 3 and add 1 *)
8
52
We can write the above-defined function under the form:
g[n Integer?EvenQ]:=n/2
g[n ]:=3n+1
g[8]
g[7]
4
22
Suppose we want to find out how many times one needs to apply g to numbers
from 1 to 30 getting 1:
k=Length/@(NestWhileList[g,#,!#==1 &]&/@Range[30];)
m=Max@k (* Find the largest number of steps needed to reach 1 *)
ListPlot[k,Filling!Axis,PlotStyle!Black,ImageSize!250,
PlotRange!All]
array=Table[fi,k[[i]]g,fi,30g];
pointmaxat=Last[Sort[array,#1[[2]]< #2[[2]]&]][[1]] (* We find the point
which requires the largest number of iteration to reach 1 *)
112
Some simple tests for Range[30] of the Collatz conjecture are presented in
Fig. 5.7 below.

5.1.1.2 Square and Cubic Roots

p
Now we will introduce a very short code to compute a, a  0, using an
approximate method.
Clear[f,a,"]
f[a ; " ]:= Module[fx = 1.,xp=ag, While[Abs[xp - x] 10" ,xp = x;
x = (x+a/x)/2];x]
62 5 Functions

Fig. 5.7 Collatz conjecture

for Range[30]

Suppose we want to find the square roots of some numbers with certain
approximations. For a D 3 and " D 1011 , we write
N[f[3,11.],10]
1:73205
p
Another approach for 3, a  0, follows
Nest[(# + a/# )/2&,1.0,5]/.a!3
1:73205
p For the square root of a nonnegative number, we can use the built-in function
. For example,
hp i
N 3; 11
1:7320508076
p
3
Now we introduce a very short code to compute, a; a  0, using an
approximate method.
Clear[f,a,"]
"
1
 ı 2  = 1.,xp=ag, While[Abs[xp - x] 10 ,xp = x;
f[a ," ]:=Module[fx
x D 3 2x C a x ];N[x]]
Suppose we want to find the cube root of a number with certain approximations,
then we write
f[10,7.]
2:15443
p p
We can use the built-in function 3  to find 3 10. Indeed
hp i
N 3 10; 10
2:154434690
5.2 Differentiation and Integration 63

5.2 Differentiation and Integration

5.2.1 Differentiation of Functions

Suppose the function f .x/ D x3 C 2x C 5, x 2 R is defined. Then, its first-order

derivative is
Clear[f,x]
f[x ]:=x3 +2x+5
f’@x
2 C 3x2
or
D[f@x,x]
2 C 3x2
To find the derivative at a given point, say 2:28, we write
%/.x!2.28
17:5952
or
f ’@2:28
17:5952
For the second-order derivative, we write
f ”@x
6x
or any of the following two
fDerivative[2][f][x],D[f[x],x,x]g
{6x,6x}
Immediately we have
ff”[],Derivative[2][f][x]/.x! ,D[f[x],x,x]/.x! g
f6; 6; 6g
Suppose we have a function of three variables f .x; y; z/ D 2xy2 z3 C
6 f .x;y;z/
log.y2 C 1/ C y cos.z2 /, and we want to calculate @@x@y 2 @z3 . Then, we write

Clear[f,x,y,z]
f[x ,y ,z ]:=2xy2 z3 +Log[y2 +1]+y Cos[z2 ];
{Derivative[1,2,3][f][x,y,z],
2
Derivative[1,1,0][f][x,y,z], (* @ f.x;y;z/
@x@y
*)
Derivative[1,1,0][f][x,y,z]/.y!1.25,
Derivative[1,1,0][f][x,y,z]/.fy!1.25,z!
˚ 108 g}

24; 4yz3 ; 5:z3 ; 5.0000000000000005J ^  24

Suppose we want to calculate Derivative[1,1,0][f][x,y,z]/.y!1.25 at z=1. Then,
we write
%%/.z!1 (* It uses the last but one result *)
5:
64 5 Functions

5.2.2 Integration of Functions

We consider the following example:

Clear[f,h,x]
f[x ]:=x4 +1 (* f(x) = x4 + 1 *)
1
h[x ]:=1/f[x] (* h(x) = f .x/ *)
R
Integrate[h@x,x] (* Indefinite integral, i.e., h.x/ dx *)
R1
Integrate[h@x,fx,-1,1g] (* Definite integral, i.e., 1 h.x/ dx *)
TraditionalForm[%]
Simplify[%]
N[%] p p p p
2ArcTanŒ1 2xC2ArcTanŒ1C 2xLogŒ1 2xCx2 CLogŒ1C 2xCx2 
p
p 4 2
2LogŒ1712 2
p
4 2 p
2log.1712 2/
p
4 2 p
2LogŒ1712 2
p
4 2
1:73395
Sometimes apparently simple integrals lead to special functions.
Integrate[Sqrt[x]Sqrt[1+x],x]
FullSimplify[D[%,x]]
Integrate[Sqrt[x]Sqrt[1+x]Sqrt[2+x],x]
FullSimplify[D[%,x]]
 p p 
1 p
x 1 C x.1 C 2x/  ArcSinh x
4
p p
x 1Cx

1 p p
5
2 x.1 C x/3=2 2 C x  p 1 1 p 
1C x x
q q q h h i i
1Cx 1 p 2Cx
4 2Cx 1 C x .2 C x/ C i x x
EllipticE iArcSinh p1x ; 2
p p p
x 1Cx 2Cx
Here EllipticE[
,m] denotes the elliptic integral of the second kind. The incom-
plete elliptic integral of the second kind is defined
R
p
E.
; k/ D 0 1  k2 SinΠd;
with m D k2 . For more details, visit the page mathworld.wolfram.com/
EllipticIntegraloftheSecondKind.html.
We introduce some multiple integrals with an example.
5.3 Functions and Their Graphs 65

Integrate[1/(x2 +5),x,x,x] (* The output contains the imaginary unit i *)

FullSimplify[%] (* After changing to a simpler form, it looks more
familiar *)
FullSimplify[[D[%,x,x,x]]
 (* and we
h arrived
i to the starting
hp function
i *)
1
p  2
 x
p
20
10x C 2 5 10 C x ArcTan p C 5i 5Log 5  ix
hp i 5
p  
5i 5Log 5 C ix  10xLog 5 C x2
p   h i   
1
10
5 5 C x2 ArcTan px  5x 1 C Log 5 C x2
5
1
5Cx2

A multiple integral coincides with the iterative integral after simplifications if

certain assumptions are satisfied [53, Chap. 8].
  
FullSimplify Integrate1=.x2 C 5/;
 x; x; x DD 2    
FullSimplify Integrate Integrate Integrate 1= x C 5 ; x ; x ; x
True

5.3 Functions and Their Graphs

We can integrate and/or differentiate an interpolating function as follows:

Needs[“PlotLegendsJ”] (* We need the ”PlotLegendsJ” package *)
Clear[f,g,h,x]
f=Interpolation[f1,-2,5,-5g];
g[x ]=Integrate[f@x,x,x];
h[x ]=D[g@x,x,x];
Plot[ff@x,g@x,h@xg,fx,1,4g,Ticks!ff1,2,3,4g,f-5,-2,1,5gg,
AxesOrigin!f0,0g,PlotLegend!ff,g,hg,
LegendPosition!f1.1,-0.4g,ImageSize!400]
The built-in function PlotLegends is presented in Fig. 5.8 below.

Fig. 5.8 The built-in function PlotLegends

66 5 Functions

Fig. 5.9 The built-in function PlotLegends with color

We can change the color and style of the previous graph.

Plot[ff@x,g@x,h@xg,fx,1,4g,PlotStyle!fDirective[Red,Thick],
Directive[Thick,Blue,Dashing[f.025g]],
Directive[Thick,Black,Dashing[f.015g]]g,
Ticks!ff1,2,3,4g,f-5,-2,1,5gg,PlotLegend!ff,g,hg,
LegendPosition!f1.1,-0.4g,ImageSize!450]
Above we present the built-in function PlotLegends with color Fig. 5.9.
Define the function
(p
x; if x  0;
f .x/ D p
x; if x  0:

We plot its graph for 1  x  1:

Clear[f]
f[x ?Positive]:=Sqrt@x
f[x ?Negative]:=Sqrt@-x
fPlot[f@x,fx,-1,1g,ImageSize!200,Ticks!ff-1,1g,f.5,1gg],
Plot[Sqrt@Abs@x,fx,-1,1g,ImageSize!200,
Ticks!ff-1,1g,f.5,1gg]g
The built-in function Sqrt is presented in Fig. 5.10 below.
We can color such a picture. One way of coloring is as follows:
Clear[f]
f[x ?Positive]:=Sqrt@x
f[x ?Negative]:=Sqrt@-x
Plot[f@x,fx,-1,1g,ImageSize!200,PlotStyle!Thick,
ColorFunction!fHue@#&g,Ticks!ff-1,1g,f.5,1gg]
Presented below is the built-in function Sqrt with color Fig. 5.11.
5.3 Functions and Their Graphs 67

Fig. 5.10 The built-in function Sqrt

Fig. 5.11 The built-in

function Sqrt with color

Fig. 5.12 The built-in

function Plot acting on two
functions

We have seen the usage of the built-in function Plot for several functions. The
graphs of different functions are colored with different colors. Here is an example,
f .x/ defined above and g.x/ D 2 sin x on the same interval Œ5; 6.
Plot[ff@x,2Sin@xg,fx,-5,6g,ImageSize!250]
The built-in function Plot acting on two functions is presented in Fig. 5.12 above.
Two other possible ways of coloring are presented below:
Plot[ff@x,2Sin@xg,fx,-5,6g,PlotRange!All,ImageSize!250,#1]&/@
fPlotStyle!fBlack,Redg,fColorFunction!Function[fx,yg,Hue[y]]gg
Two functions colored in different ways are presented in Fig. 5.13.
68 5 Functions

Fig. 5.13 Different colored functions

5.4 Plane and Space Figures

5.4.1 Plane Figures

Mathematica is an excellent tool for drawing figures, simple or very complicated

ones.
We note that Mathematica considers two rectangular axes such that the desired
graph lies in a visible area. The scale of the two axes need not coincide. Maybe we
want to have a smaller graph colored in red and dashed. Therefore, we can choose
from some options.
{Plot[2Sin@t,ft,0,4g,PlotStyle!fThickness!0.008,Red,Dashedg,
AxesOrigin!f0,0g,Ticks! ff0; ; 2; 3; 4g,Automaticg,
ImageSize!200],
Plot[2Sin@t,ft,0,4 g,PlotStyle!fThickness!0.008,Blueg,
AxesOrigin! f,0g,AxesLabel!ft,xg,
AxesStyle!
fDirective[Thickness!0.01,Green,12],Directive[Purple,12]g,
Ticks! ff0; ; 2; 3; 4g;f-2,-1,1,2gg,ImageSize!200],
Plot[2Sin@t,ft,0,4g,PlotStyle!fThickness!0.008,Blueg,
AxesOrigin! f2,0g,AxesLabel!ft,Sin@tg,
AxesStyle!fDirective[Thickness!0.01,Green,12],
Directive[Purple,12]g,Ticks! ff0; ; 2; 3; 4g,f-2,-1,1,2gg,
AspectRatio!Automatic,ImageSize!200]}
One way of presentation of functions is introduced in Fig. 5.14 below.
The length of a dash and empty space can be modified. The graph can be
exhibited in dotted form.
Plot[2Sin@x,fx,-5,7.28g,ImageSize!200,
Ticks!ff-5,-2,2,4,7g,f-2,-1,1,2gg,
PlotStyle!fThickness!0.008,Black,#1g]&/@
fDashing[Table[n/100,fn,1,5,0.4g]],Dottedg
One way of presentation of functions is introduced in Fig. 5.15 below.
5.4 Plane and Space Figures 69

Fig. 5.14 Different presentations of functions 1

Fig. 5.15 Different presentations of functions 2

We consecutively display and label the graphs of the three functions. At the third
plot, the label uses the same font and style that is used in graphs and is not included
in the plot.
{Plot[3 Cos@x,fx,-5,7.28g,
PlotStyle!fThickness!0.008,Darker[Green],Dashing[0.03]g,
PlotLabel!”3cos(x)”,ImageSize!180,PlotRange!All],
Plot[2Sin@x,fx,-5,7.28g,PlotStyle!fThickness!0.008,Blue,Dottedg,
ImageSize!180,PlotLabel!Style[”sin(x)”,FontSize!18,
Background!Green],Background!Yellow,PlotRange!All]
Labeled[Plot[Tan@x,fx,-5,7.28g,ImageSize!180],
Style[”Tan(x)”,”Graphics”],Background!LightGreen]}
Different colors for the pictures are introduced in Fig. 5.16.
70 5 Functions

Fig. 5.16 Different colors for the pictures

Fig. 5.17 Different functions

on the same Cartesian system

Suppose we want to put several graphs of functions in the same Cartesian system
of coordinates. We can do it in the following way.
Plot[fSin[2x],Cos[x/2]g,fx,0,4g,
PlotStyle!fDirective[Thickness!0.007,Red],
Directive[Thickness!0.007,Blue]g,PlotLegend!f”sine”,”cosine”g,
ImageSize!300]
A way of presentation of functions is introduced in Fig. 5.17 above.
We can also insert the legend exactly where we want.
5.4 Plane and Space Figures 71

Fig. 5.18 A way of legend insertion

Fig. 5.19 A complete graph

with 13 vertices

ListPlot[Table[fx,f@xg,ff,Sin,Cos,Logg,fx,0,10,0.5g],
PlotLegend!f”Sine”,”Cosine”,”Log”g,LegendPosition!f1.1,-0.4g,
Joined!fTrue,True,Falseg,Ticks!ff5,10g,f-1,1,2gg,ImageSize!300,
PlotMarkers! #1]&/@fAutomatic,f”ı”,””,”*”gg
Next, a way of legend insertion is introduced in Fig. 5.18.
A complete graph with 13 nodes is presented in Fig. 5.19. Its vertices are labeled.
72 5 Functions

Fig. 5.20 Two complete graphs with 13 vertices

Tuples[Table[fCos@t,Sin@tg,ft,0,2,2/13g],2]; (* It generates all

vertices of the graph *)
Graphics[ffBlue,Line@%g,fText[Style[Row[fStyle[”A(”,Italic,12],#1,
Style[”)”,Italic,12]g]],fCos[#1 2/13],Sin[#1 2/13]g]&/@Range@13g,
ImageSize!200g]
The above complete graph can be introduced with a single command Fig. 5.20.
Clear[a]
p=CompleteGraph@13;
Show[p,Graphics[fText[Style[ Row[fStyle[”A(”,Italic,12],#1,
Style[”)”,Italic,12]g],Red],1.2fCos[#1 2 /13+/20],
Sin[#1 2 /13+/20]g]&/@Range@13g],ImageSize!200],
Show[p,Graphics[fText[Style[a#11 ,Italic,12,Red],
1.15fCos[(#1-1)2/13+/20],Sin[(#1-1) 2/13+/20]g]&/@
Range@13g],ImageSize!200]
See Fig. 5.20.
An old-fashioned hour plate can be designed as follows. Moreover, we can show
the exact time at the instant of running the code Fig. 5.21.
fhour,minute,secondg=
ToExpression[StringSplit[TextString[TimeObject[]],”:”]];
minute=minute+second/60;hour=hour+minute/60;hour=Mod[hour,12];
Show[Graphics[fTable[Text[Style[”-”,Plain,FontWeight!”Bold”,Blue,8],
fCos@,Sin@g1.01,f0,0g,fCos@ ,Sin@g],
f,0,2-2/60,2/60g],fBlack,PointSize[0.03],Point[f0,0g],
5.4 Plane and Space Figures 73

Fig. 5.21 An hour plate

Fig. 5.22 Discs covering a circle and the sine function

PointSize[0.02],Point[fCos[#1/6],Sin[#1/6]g],
fArrowheads[0.05],Thin,Black,Arrow[ff0,0g,
.97fCos[=2   second/30],Sin[=2   second/30]gg]g,
fArrowheads[0.07],Thick,Green,Arrow[ff0,0g,
.95fCos[=2   minute/30],Sin[=2   minute/30]gg]g,
fArrowheads[0.078],Thick,Blue,Arrow[ff0,0g,
.8fCos[=2   hour/6],Sin[=2   hour/6]gg]g,
Text[Style[RomanNumeral[#1],Italic,12,Blue],1.16fCos[(-#1+3)/6],
Sin[(-#1+3)/6]g]gg]&/@Range@12,ImageSize!200]
The color of a figure can vary, thanks to many options. One of them is the
graphics directive Hue. Two simple examples follow in Fig. 5.22.
Graphics[Table[fHue[k/7],Disk[fCos[k 2/7],Sin[k 2/7]g,1/2]g,fk,0,6g],
PlotRange!ff-2,2g,f-2,2gg,ImageSize!200]
We change the previous code so that the discs are laid on the sine function.
Graphics[Table[fHue[k/7],Disk[fk 2/7,Sin[k 2/7]g,1/2]g,fk,0,14g],
PlotRange!ff-1/2,14g,f-2,2gg,ImageSize!250]
See Fig. 5.22.
74 5 Functions

Fig. 5.23 Different spirals

We now present some spirals. In the first example, the polar radius and the angle
are equal. In the second example, the polar radius equals sin.3t/, whereas in the
third is sin.5t/: The next two examples show other polar closed curves Fig. 5.23.
{PolarPlot[t,ft,0,20g,PlotLabel!Style[PolarPlot,16],
Ticks!ff-15,-10,-5,5,10,15g,f-15,-10,-5,5,10,15gg],
p3=PolarPlot[Sin[3t],ft,0,g,PlotStyle!fThickness!0.01,
Darker@Greeng],
PolarPlot[Sin[5t],ft,0,g,PlotStyle!fThickness!0.01,Redg,
Ticks!ff-.5,.5g,f-.5,.5,1gg]}

p1=PolarPlot[f1,1/2,1+1/10 Sin[10t]g,ft,0,2g,
PlotStyle!fDirective[Thick,Darker@Green],Blue,
Directive[Dashed,Thick,Red]g];
p2=PolarPlot[f1,1-1/5,1+1/5,1+1/5 Sin[10t]g,ft,0,2g,
PlotStyle!fDirective[Thick,Darker@Blue],
Directive[Thick,Darker[Black]],Directive[Thick,Darker@Green],
Directive[Dashed,Thick,Red]g,Axes!False];
fp1,p2g
Show[p3,p2,PlotRange!All,Axes! fTrue,Falseg,ImageSize!200]
Some graphs with PolarGraph are shown in Fig. 5.24.
Now we show the simple cross, [19, pp. 130–131].
With[fa=6,b=4g,x[t ]:=a Sec@t; y[t ]:=b Csc@t;
ParametricPlot[ffx@t,y@tg,fx[t+/4],y[t+/4]gg,ft,-,g,
PlotStyle!Blue,Axes!False,ImageSize!180]
We suggest the next code for the Maltese cross.
p p
PolarPlot[f2/ SinŒ4t  =2,8/ SinŒ4t C =2g,ft,0,2g,Axes!False,
PlotStyle!fThick,Blueg,ImageSize!180]
The two crosses are presented in Fig. 5.25.
5.4 Plane and Space Figures 75

Fig. 5.24 Graphs with polar plot

Fig. 5.25 The cross and the Maltese cross

We present the next code for the Star of David Fig. 5.26.
p=Tuples[Table[fCos@t,Sin@tg,ft,/6,13/6,/3g],2]; (* This is
the complete graph with 6 vertices *) p
p=Select[p,EuclideanDistance[#[[1]],#[[2]]]== 3&]; (* From
the complete graph, we select the edges of a prescribed length *)
Show[Graphics[fThick,Blue,Line[p]g],ImageSize!180]
76 5 Functions

Fig. 5.26 Star of David

Fig. 5.27 Roses

The polar equation of a rhodonea (rose), [19, p. 175], is r D a cos.mt/. We admit

that m 2 Z. In such a case, there are m petals if m is odd and 2m petals if m is even
Fig. 5.27.
5.4 Plane and Space Figures 77

Fig. 5.28 Planar curves by parametric equations, 1

With[fa=4g,PolarPlot[a Cos[# t],ft,0,2g,Ticks!ff-a,ag,f-a,agg,

TicksStyle!Directive[Red,Bold]]&/@f2,3,4,5g]
We now present some planar curves defined by parametric equations. Giving
different domains to the independent variable, we present how the final graphs
appear Fig. 5.28.
xx[t ]:=11Cos@t-6Cos[11t/6];
yy[t ]:=11Sin@t-6Sin[11t/6];
ParametricPlot[fxx@t,yy@tg,ft,0,#g,ColorFunction!fHue@#&g,
Ticks!ff-15,15g,f-15,15gg,TicksStyle!Directive[Black,Bold]]&/@
f2,4,6,8,10,12g
Other graphs are introduced in Fig. 5.29.
p
xx[t ]:=11Cos@t-6Cos[p 2t/6];
yy[t ]:=11Sin@t-6Sin[ 2t/6];
ParametricPlot[fxx@t,yy@tg,ft,0,#g,ColorFunction!fHue@#&g,
Ticks!ff-15,15g,f-15,15gg,TicksStyle!Directive[Black,Bold]]&/@
f2,12,60g
The bean is a quartic curve given by the implicit equation
 
x4 C x2 y2 C y4 D x x2 C y2 :
78 5 Functions

Fig. 5.29 Planar curves by parametric equations, 2

Fig. 5.30 A bean and a simple butterfly

It has horizontal tangents at (2/3, ˙2/3) and vertical tangents at (0,0) and (1,0). Its
graph looks as the left-hand graph in Fig. 5.30.
ContourPlot[x4 +x2 y2 +y4 ==x(x2 +y2 ),fx,0,1g,fy,-2/3,2/3g,
ContourStyle!fDarker@Greeng,FrameTicks!fff-0.6,0,0.6g,Noneg,
ff0,1g,Nonegg,ImageSize!180]
A simple butterfly follows. Its equation is given implicitly, the right-hand graph
in Fig. 5.30.
ContourPlot[y6 DD x2  x6 ,fx,-1,1g,fy,-.9,.9g,ContourStyle!fBlueg,
FrameTicks!fff-0.8,0,0.8g,Noneg,ff-1,0,1g,Nonegg,
ImageSize!180]
A more realistic butterfly curve is a transcendental plane curve discovered by
Temple H. Fay in [23]. This curve is given by the following parametric equations
Fig. 5.31.
xx[t ]:=Sin@t(eCos@t -2Cos[4t]-Sin[t/12]5 );
yy[t ]:=Cos@t(eCos@t -2Cos[4t]-Sin[t/12]5 );
ParametricPlot[fxx@t,yy@tg,ft,0,#1g,ColorFunction!fHue@#&g,
Ticks!ff-3,0,3g,f-2,3gg]&/@f2Pi,4Pi,6Pi,8Pi,12Pi,24Pig
ParametricPlot with two parameters gives a region Fig. 5.32.
5.4 Plane and Space Figures 79

Fig. 5.31 Butterflies

Fig. 5.32 ParametricPlot with two parameters giving a region

Apply[fParametricPlot[fft,Sin@tg,ft,Sin@t+1gg,ft,0,2g,
Ticks! #1,#2],
ParametricPlot[ft,r Sin@t+(1-r)(Sin@t+1)g,ft,0,2g,fr,0,1g,
Frame!ffFalse,Trueg,fFalse,Truegg,
FrameStyle!ffThick,Directive[Thick,Dashed]g,fRed,Bluegg,
FrameTicks! #1,#2]g&,fff0,Pi/2,Pi,3 Pi/2,2 Pig,f-1,0,1,2gg,
ImageSize!200g]
By default the functions in the built-in function Plot are not evaluated until they
have numeric values for the variables Fig. 5.33.
{Plot[Table[2Sin[x+t],ft,0,4g],fx,0,2g,Ticks!ff,2g,f-2,2gg],
Plot[Evaluate[Table[2Sin[x+t],ft,0,4g]],fx,0,2g,Ticks!ff,2g,
f-2,2gg]}
Evaluating the argument symbolically can sometimes be faster, as the following
example shows Fig. 5.34.
80 5 Functions

Fig. 5.33 The use of built-in function Evaluate, 1

Fig. 5.34 The use of built-in function Evaluate, 2

Fig. 5.35 The use of built-in function Evaluate, 3

Clear[a,b]
fTiming[Plot[Integrate[Sin@x Cos@x,fx,0,ag],fa,0,2g,
Ticks!ff,2g,f.5gg]],
Timing[Plot[Evaluate[Integrate[Sin@x Cos@x,fx,0,bg]],fb,0,2g,
Ticks!ff,2g,f.5gg]]g
It is clear that the first graph needs more than 8 s whereas the second one needs
about 129 times less.
Sometimes numeric evaluation is faster, as the following example shows
Fig. 5.35.
{Timing[Plot[Nest[Sin[#]#&,a,20],fa,0,g,PlotStyle!Blue,
Ticks!ff1.5,2.5,g,f1.5gg]],
Timing[Plot[Evaluate[Nest[Sin[#]#&,a,20]],fa,0,g,PlotStyle!Blue,
Ticks!ff1.5,2.5,g,f1.5gg]]}
We show how the built-in function Nest is used to study the uniform convergence
of a sequence of functions Fig. 5.36.
5.4 Plane and Space Figures 81

Fig. 5.36 The use of built-in function Nest

Plot[Nest[Sin@#&,a,#(10)],fa,0,2g,
PlotLabel!ToExpression[”iterations”]==TraditionalForm[#(10)],
LabelStyle!Hue[#/10],
PlotStyle!fThickness!0.01,Hue[#/10]g,
Ticks!ff,2g,Automaticg,
TicksStyle!Directive[Bold]]&/@f1/10,3,6,9g
We know the behavior of the function f .x; n/ D sin.sin.: : : .sin x///, n times,
x 2 Œ0;  Fig. 5.37.
Table[
Plot[Nest[Sin@#&,a,n],{a,0,},
PlotRange!All,
Ticks!{{{/2,“/2”},},Automatic},
PlotLabel!ToExpression[“iterations”]==TraditionalForm[n],
PlotStyle!{Thickness!0.01,Hue[n/1000]}],
{n,1,1000,200}]
Taking into account that f .x/ D f .x/;, we note that limn!1 f .x; n/ D 0,
uniformly in x 2 R. A rigorous proof can be given, too. p
At the same time, we have the stronger result that limn!1 n=3f .x; n/ D
sgn.x/, for x 2   ; Π; Fig. 5.38. A proof is given in the Proposition 5.1.
82 5 Functions

Fig. 5.37 Iteration of the function f .x; n/

p
Fig. 5.38 The n=3f .x; n// function iteration

Table[Plot[Sqrt[n/3]Nest[Sin,x,n],fx,-,g,
PlotLabel!ToExpression[”iterations ”]==TraditionalForm[n],
PlotStyle!fHue[n/1000],Thickness!0.008g,
Ticks!ff-,g,Automaticg,PlotRange!All],
fn,1,1000,300g]
We can handle the last command dynamically by the next code. Changing by a
click the value of n, the color of the graph changes Fig. 5.39.
5.4 Plane and Space Figures 83

Fig. 5.39 Dynamical

changing of Fig. 5.38

Clear[n]
Dynamic[Plot[Sqrt[n/3]Nest[Sin,x,n],fx,-,g,ImageSize!200,
Ticks!ff-,g,f-1,1gg,PlotStyle!fHue[n/1000],Thickness!0.008g,
PlotLabel!Row[
fStyle[n,Italic],” = ”,PopupMenu[Dynamic[n],f1,301,601,901,1201g]g]]]

Proposition 5.1 (de Bruijn formula) Denote as above f .x; n/ D

sin.sin.: : : .sin x///, n times, n 2 N ; x 2   ; Π. Then,
p
lim n=3f .x; n/ D sgn x:
n!1

Proof Obviously f .x; n C 1/ D sin f .x; n/; n 2 N and x 2 R. Choose x 2 0; Π:

Then f .x; 1/ 2 0; 1  0; =2, 0  f .n C 1; x/ < f .n; x/, and f .n; x/ ! 0
uniformly in x. We have

1 1 t  sin t t  t
2
 2 D   C 1 ! 1=3 as t # 0:
sin t t t3 sin t sin t

Substitute t:=f .n; x/ to get

1 1
 ! 1=3 as n ! 1:
f 2 .n C 1; x/ f 2 .n; x/

According to the Corollary 1.5 at page 87 in [53], we further have that

n1 
1X 1 1

n iD1 f 2 .i C 1; x/ f 2 .i; x/

1 1 1
D  ! 1=3 as n ! 1
n f 2 .n; x/ f 2 .1; x/
84 5 Functions

and

nf 2 .n; x/ ! 3 as n ! 1: t
u

Example The built-in function NestList is a powerful tool in studying sequences

defined by recurrences. Below, we introduce two examples.
Suppose we want to study the following sequences defined by recurrences
q q
p p p
x1 D 2; x2 D 2 C 2; : : : ; xnC1 D 2 C xn ; and
p p pxn
x1 D 2; : : : ; xnC1 D 2 ; n 2 N :

Clear[a,n]
Assumptions!n2Integers&&n>0;
With[fn=10,a=2.0g,
p p 
fNestList a C # &; a; n , (* Exhibit the first ten terms of the first
sequenceh *) i
p # p
NestList a &; a; 3n (* Exhibit the first thirty terms of the
second sequence *)g]
{{1.41421,1.84776,1.96157,1.99037,1.99759,1.9994,1.99985,1.99996,
1.99999,2.,2.},
{1.41421,1.63253,1.76084,1.84091,1.89271,1.927,1.95003,1.96566,1.97634,
1.98367,1.98871,1.99219,1.99459,1.99626,1.99741,1.9982,1.99876,
1.99914,1.9994,1.99959,1.99971,1.9998,1.99986,1.9999,1.99993,1.99995,
1.99997,1.99998,1.99998,1.99999,1.99999}}
We note that the above evaluations suggest that both sequences tend toward 2.
The limits are found rigorously in [53, p. 140] and [54, p. 28]. M
Let us color some planar curves with different colors Figs. 5.40 and 5.41.
Plot[#1,ft,0,15g,PlotStyle!AbsoluteThickness@4,
ColorFunction!”Rainbow”,
ImageSize!200]&/@fSin@t,Sin@t/t,t Sin@tg

{Plot[Sin@t,ft,0,10g,PlotStyle!AbsoluteThickness[5],
ColorFunction!”BlueGreenYellow”,#1],
Plot[Sin@t/t,ft,0,15g,PlotStyle!AbsoluteThickness[4],
ColorFunction!Function[fx,yg,Hue[x]],#1],
Plot[t Sin@t,ft,0,10g,PlotStyle!AbsoluteThickness[2],
ColorFunction!Function[fx,yg,ColorData[”NeonColors”][y]],#1]g
&/@fImageSize!200g}
5.4 Plane and Space Figures 85

Fig. 5.40 Planar curves with rainbow colors, 1

Fig. 5.41 Planar curves with several colors, 2

5.4.2 Functions and Graphs in 3D

A surface of revolution can be obtained as the following examples show. Then, we

get a surface looking like a UFO Fig. 5.42.
86 5 Functions

Fig. 5.42 Getting a UFO

meshstyle=fMeshStyle!GrayLevel[1],MeshFunctions!f#3&g,
Mesh!None,fMesh!None,Boxed!False,Axes!Falsegg;
fPlot[Cos@t,ft,0,g,PlotStyle!fThickness!0.01g,
Ticks!ffg,f-1,1gg],
RevolutionPlot3D[Cos@t,ft,0,g,PlotStyle!fGreen,Opacity!0.4,
5.4 Plane and Space Figures 87

Fig. 5.43 A broken line in

the 3D space

Thickness[.02]g,ImageSize!200,
Ticks!ff-,0,g,f-,0,g,f-1,0,1gg]g
RevolutionPlot3D[Cos@t,ft,0,g,PlotPoints!40,
PlotStyle!fGreen,Opacity!0.4,Thickness[.02]g,ImageSize!200,
Ticks!ff-,0,g,f-,0,g,f-1,0,1gg#]&/@meshstyle
A 3D curve very often has several representations. The first example shows a
broken line connecting 3 points in the 3D space Fig. 5.43.
Graphics3D[ffThickness!0.008,Darker[Green],
Line[ff0,0,0g,f1,2,1g,f-1,3,3gg]g,
Text[Style[”A=(0,0,0)”,Bold,12],f0,0.2,-0.4g],
Text[Style[”B=(1,2,1)”,Italic,12,Darker[Blue]],f0.5,1.3,1.2g],
Text[Style[”C=(-1,3,3)”,Bold,16,Red],f0,2.7,3.1g],
fPointSize[0.03],Orange,Point[ff0,0,0g,f1,2,1g,f-1,3,3gg]gg,
Axes!True, PlotRange!All,ImageSize!200,
Ticks!ff-1,0,0.5g,f0,1,2,3g,f0,1,2,3gg]
We now introduce a cube, its vertices, and some facts about these elements.
88 5 Functions

Fig. 5.44 A broken line in the unit cube

p=Tuples[f0,1g,3] (* The vertices of the unit cube *)

MatrixForm/@%
fp[[2]], (* The second element of the list p *)
p[[2,3]]g (* The third element of the second element in the list p *)
q=Tuples[p,2]; (* This list is large, so we do not display it *)
q[[3]]
q[[3,2]]
ff0;
80 0;10g;0f0; 1 0; 1g; f0; 1; 0g; f0; 1; 1g; f1; 0; 0g; f1; 0; 1g; f1; 1; 0g; f1; 1; 1gg
0 1 0 1 0 1 0 1 0 1 0 19
< 0 0 0 0 1 1 1 1 =
@0A;@0A;@1A;@1A;@0A;@0A;@1A;@1A
: ;
0 1 0 1 0 1 0 1
ff0; 0; 1g; 1g
ff0; 0; 0g; f0; 1; 0gg
f0; 1; 0g
We now show some facts on the unit cube Fig. 5.44.
fGraphics3D[fRed,Thickness!0.008,Line@pg],
Graphics3D[fBlue,Line@qg,Boxed!False]g (* In the first cube, the vertices
are connected following the order of the list p; in the second cube, each
vertex is connected to the other ones *)
We plot some functions using different coordinate systems and different label
options.
fShow[ParametricPlot3D[f(2+Cos@v)Cos@u,(2+Cos@v)Sin@u,
Sin@vg,fu,0,2g,fv,; 2g,Mesh!None,
PlotLabel!Style[ParametricPlot3D,Italic,Green,16],
Ticks!ff-3,0,3g,f-3,0,3g,f-1,0,1gg],
ParametricPlot3D[fCos@u Sin@v,Sin@u Sin@v,
u v((/2)2 -v2 )Cos@vg,fu,0,2g,fv,0,/2g,Mesh!None],
ImageSize!200,ViewPoint!f2.5,2.25,2.5g,PlotRange!All],
5.4 Plane and Space Figures 89

Fig. 5.45 Parametric and spherical plot

SphericalPlot3D[Sin@+Sin[5
]/5,f;0,/2g,f
,3/2,11/6g,
Ticks!ff-1,0,1g,f-1,0,1g,f-.5,0,.5gg,Mesh!None,
PlotLabel!Style[SphericalPlot3D,14,Bold,Blue],ImageSize!200]g
See Fig. 5.45.
We now introduce two helices. The first is conic and the second is cylindrical.
Then, we put them together into the same frame Fig. 5.46.
Clear[r,t]
{p1=ParametricPlot3D[ft Cos@t,t Sin@t,tg,ft,0,2g,
PlotStyle!fDirective[Thick,Darker[Green]]g,ImageSize!200],
p2=ParametricPlot3D[ffCos@r,Sin@r,0g,fCos@r,Sin@r,rgg,fr,0,2g,
PlotStyle!fDirective[Thick,Red]g,ImageSize!f150,275g,
Axes!True,AxesOrigin!f0,0,0g,Ticks!ff-1,1g,f-1,0,1g,f,2gg,
TicksStyle!16,AxesStyle!Directive[Thickness!.0175,
Darker[Green]],ViewPoint!f3,0,/3g,Boxed!False],
Show[p1,p2,ImageSize!200]}
The Möbius surface is given below Fig. 5.47.
{pmoebius=ParametricPlot3D[
fCos[t](3+r Cos[t/2]),Sin[t](3+r Cos[t/2]),r Sin[t/2]g,fr,-1,1g,ft,0,2g,
Mesh!None,PlotPoints!f75,100g,MaxRecursion! 0,
PlotStyle!fGreen,Opacity!0.4,Thickness[.02]g,
PlotPoints!50,Ticks!ff-3,0,2,4g,f-3,0,3g,f-1,0,1gg,
ImageSize!250],
Show[pmoebius,Boxed!False,Axes!False,ImageSize!250]}
90 5 Functions

Fig. 5.46 Conic and cylindrical helices

Fig. 5.47 Möbius surface

Now the edge of the Möbius surface follows Fig. 5.48.

eqs[r ,t ]:=fCos[t](3+r Cos[t/2]),Sin[t](3+r Cos[t/2]),r Sin[t/2]g;
ParametricPlot3D[feqs[1,t],eqs[-1,t]g,ft,0,2g,
Ticks!ff-2,4g,f-2,2g,f-1,1gg,
PlotStyle!fBlue,Dashed,Thickness[.2]g,ImageSize!250]
We now plot a function in different coordinate systems Fig. 5.49.
{Plot3D[Sin[x y],fx,0,3g,fy,0,3g,Mesh!#,
Ticks!fAutomatic,Automatic,f-1,0,1gg],
DensityPlot[Sin[x y],fx,0,3g,fy,0,3g,Mesh!#]}&/@fAll,Noneg
An example of a saddle point on a 3D surface (hyperbolic paraboloid) is
presented below Fig. 5.50.
Show[ParametricPlot3D[fx,y,(x2 -y2 )/2g,fx,-2,2g,fy,-2,2g,
Boxed!False,Axes!False,MeshStyle!GrayLevel[1],
PlotStyle!fLighter[Green],Opacity!0.6g,PlotPoints!f100,60g],
5.4 Plane and Space Figures 91

Fig. 5.48 The edge of the

Möbius surface

Fig. 5.49 Different coordinate systems

ParametricPlot3D[fx,0,x2 /2g,fx,-2,2g,
PlotStyle!fDirective[Thick,Darker[Red]]g],
ParametricPlot3D[f0,y,-y2 /2g,fy,-2,2g,
PlotStyle!fDirective[Thickness!0.006,Darker[Red]]g],
Graphics3D[fPointSize[0.04],Black,Point[ff0,0,0gg]g],
2 2
PlotLabel!Style[”Saddle point of z D x2  y2 ”,FontSize!14,
Background!Yellow],ImageSize!250]
92 5 Functions

Fig. 5.50 Hyperbolic

paraboloid

Fig. 5.51 Examples with ContourPlot, RegionPlot3D, and RegionPlot

A few specialized visualization functions are given in the next sequence Fig. 5.51.
{ContourPlot3D[x2 +y2 -z2 ,fx,-2,2g,fy,-2,2g,fz,-2,2g,
PlotLabel!Style[ContourPlot3D,Red,16]],
RegionPlot3D[x2 +y2  z2 ,fx,-2,2g,fy,-2,2g,fz,-2,2g,
PlotLabel!Style[RegionPlot3D,Orange,16]],
RegionPlot[x2  y2 ,fx,-2,2g,fy,-2,2g,
PlotLabel!Style[RegionPlot,Blue,16]]}
Chapter 6
Manipulate

6.1 Manipulate

The Manipulate command is very useful, allowing a dynamical approach to results

and corresponding figures.
Along this book we will extensively use the Manipulate command concurrent by
the Table command.
Table[n,fn,15g]
Table[Prime@n,fn,15g] (* the nth prime number *)
{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15}
{2,3,5,7,11,13,17,19,23,29,31,37,41,43,47}
We can now use the Manipulate command as the next example shows Fig. 6.1.
Manipulate[Prime@n,fn,1,15,1g]
See Fig. 6.1.
In our example in the Table command, the variable n takes discrete values. In
this case in the Manipulate command, we require the same variable to be discrete,
too.
Let us compare the next two cases. In the first case n is a positive integer, whereas
in the second case n is a real positive number within the interval Œ1; 15.
{Manipulate[n!,fn,1,11,1g],Manipulate[n!,fn,1,11g]}
See Fig. 6.2.
Another simple example regarding the Manipulate command is the following:
{Manipulate[#,fn,1,15,1g],Manipulate[#,fn,1,15g]}&/@
{Plot[Sin[n x],fx,0,2g,ImageSize!150,Ticks!ff,2g,f-1,1gg]}
See Fig. 6.3.
The previous example is improved by adding the option
Appearance!”Labeled”. This option allows us to see the instant value of the
control variable.

© Springer International Publishing AG 2017 93

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_6
94 6 Manipulate

Fig. 6.1 Manipulate 1

Fig. 6.2 Manipulate 2

Fig. 6.3 Manipulate 3

Manipulate[Plot[Sin[n x],fx,0,2g,ImageSize!150,
Ticks!ff,2g,f-1,1gg],fn,1,15,Appearance!”Labeled”g]
See Fig. 6.4.
We can also use several control variables as the following example shows:
Manipulate[Plot[Sin[n x]+Sin[m x],fx,0,2g,PlotRange!2,
ImageSize!200,Ticks!ff,2g,Automaticg],
fn,1,10,Appearance!”Labeled”g,fm,1,15,Appearance!”Labeled”g]
See Fig. 6.5.
For a discrete variable, as we did above, let us consider for example Newton’s
binomial formula.
Manipulate[Expand[.˛ C ˇ/n ],fn,1,15,1,Appearance!”Labeled”g]
See Fig. 6.6.
6.1 Manipulate 95

Fig. 6.4 Manipulate 4

Fig. 6.5 Manipulate 5

Fig. 6.6 Manipulate 6

6.1.1 Circumcircle, Incircle, and the Main Points in a Triangle

A perpendicular bisector of a side of a triangle is a straight line passing through the

midpoint of the side and being perpendicular to it. The three perpendicular bisectors
meet in a single point, the triangle’s circumcenter. This point is the center of the
96 6 Manipulate

circumcircle, the circle passing through all the three vertices. The circumcircle’s
radius is called the circumradius. An altitude of a triangle is a straight line through a
vertex and perpendicular to the opposite side. The point where the altitude intersects
the opposite side is called the foot of the altitude. The three altitudes intersect in a
single point called the orthocenter of the triangle. An angle bisector of a triangle is
a straight line through a vertex which cuts the corresponding angle into halves. The
three angle bisectors intersect in a single point called the incenter. This is the center
of the triangle’s incircle. The incircle is the circle which lies inside the triangle and
touches all three sides. A median of a triangle is a straight line through a vertex and
the midpoint of the opposite side. The three medians intersect in a single point, the
triangle’s centroid.
We may assume that vertex A of a triangle is free, whereas the vertices B and C
are fixed.
Clear[a1,a2]

mainPointsinTriangle[a1 ,a2 ]:=

Module[fb1=b2=c2=0,c1=6,a,b,c,x,y,ahax,ahay,aha,bhbx,bhby,bhb,
chcx,chcy,chc,op,ox,oy,hx,hy,hp,centroidx,centroidy,centroid,aleft,
aright,bix,biy,bip,bbprime,rRadiusg,
vertexA=fa1,a2g;vertexB=fb1,b2g;vertexC=fc1,c2g;
a=EuclideanDistance[vertexB,vertexC];
b=EuclideanDistance[vertexC,vertexA];
c=EuclideanDistance[vertexA,vertexB];
If[ab+ckbc+akca+b,Print[”No triangle”],
centroid=fcentroidx,centroidyg=(vertexA+vertexB+vertexC)/3;
midAB=(vertexA+vertexB)/2;midBC=(vertexB+vertexC)/2;
midCA=(vertexC+vertexA)/2;
aha=fahax,ahayg=fa1,0g;
sol2=Solve[
y-b2==-(a1-c1)(x-b1)/(a2-c2)&&y-c2==(a2-c2)(x-c1)/(a1-c1),fx,yg];
bhb=fbhbx,bhbyg=fx,yg/.sol2[[1]];
sol3=Solve[y-c2==-a1(x-c1)/a2&&y==a2 x/a1,fx,yg];
chc=fchcx,chcyg=fx,yg/.sol3[[1]];
sol4=Solve[x==a1&&fx,yg2InfiniteLine[fvertexB,bhbg],fx,yg];
hp=fhx,hyg=fx,yg/.sol4[[1]];
sol5=Solve[
x==(b1+c1)/2&&(y-(c2+a2)/2)==-(a1-c1)(x-(c1+a1)/2)/(a2-
c2),fx,yg];
op=fox,oyg=fx,yg/.sol5[[1]]; faleft,arightg=afc,bg/(b+c);
bip=vertexC+fa1-c1,a2ga/(a+c);
sol7=Solve[y==a2(x-aleft)/(a1-aleft)&&y==
(c2+a a2/(a+c))x/(c1+a(a1-c1)/(a+c)),fx,yg];
6.1 Manipulate 97

bisec=fbisx,bisyg=fx,yg/.sol7[[1]];
sol8=Solve[y==a2 x/a1&&y-c2==(bisy-c2)(x-c1)/(bisx-c1),fx,yg];
cisec=fx,yg/.sol8[[1]];rRadius=Norm[op]; rinradius=Divide[a a2,a+b+c];
ab =vertexA+((vertexA-vertexB).(vertexA-bisec))(vertexB-vertexA)/c2 ;
bc =vertexB+((vertexB-vertexC).(vertexB-bisec))(vertexC-vertexB)/a2 ;
ca =vertexC+((vertexC-vertexA).(vertexC-bisec))vertexA-vertexC)/b2 ;
Show[Graphics[fText[Style[”A”,Italic,12],vertexA+f.01,0.3g],
Text[Style[”B”,Italic,12],vertexB-f.2,.15g],
Text[Style[”C”,Italic,12],vertexC-f-.2,.15g],
Text[Style[”HA ”,Italic,12],fahax,b2-.2g],
Text[Style[”IA ”,Italic,12],faleft,b2-.2g],
Text[Style[”MA ”,Italic,12],f(b1+c1)/2,b2-.2g],
Blue,Line[ffvertexA,vertexBg,fvertexB,vertexCg,
fvertexC,vertexAg,fvertexA,ahag,fvertexC,chcg,fvertexB,bhbg,
fvertexC,midABg,fvertexA,midBCg,fvertexB,midCAg,
fvertexA,opg,fvertexB,opg,fvertexC,opgg],
Darker[Green],Line[ffvertexA,faleft,0gg,fvertexB,bipg,
fvertexC,cisecg,fbisec,ab g,fbisec,bc g,fbisec,ca gg],
Black,Line[ffop,midABg,fop,midBCg,fop,midCAgg],
Black,PointSize[.015],Point[fvertexA,vertexB,vertexC,bip,aha,bhb,
chc,midAB,midBC,midCA,faleft,0g,cisec,ab ,bc ,ca g],
Red,Point[fhp,op,centroid,bisecg],Circle[bisec,rinradius],
Circle[op,rRadius]g],ImageSize!300,PlotRange!All]]]
We can now see the important points of a triangle as well as the circumcircle and
the incircle change. HA , IA , and MA are the feet of, respectively, the height, bisector,
and median of vertex A.
Manipulate[
Quiet@mainPointsinTriangle[a1,a2],
ffa1,1,”abscissa”g,1,6,0.0001,Appearance!”Labeled”g,
ffa2,4,”ordinate”g,4,10,0.001,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False
]
See Fig. 6.7.

6.1.2 Euler’s Nine Points Circle

Euler’s nine points circle is an interesting circle defined by nine points related to
a triangle. The midpoints of the three sides and the feet of the three altitudes all
lie on a single circle, the triangle’s nine-point circle. The remaining three points
for which it is named so are the midpoints of the segments of altitudes between
the vertices and the orthocenter. The radius of the nine-point circle is half of the
circumradius.
98 6 Manipulate

Fig. 6.7 Main points in a triangle

Clear[a1,a2]

eulersNinePointsCyrcle[a1 ,a2 ]:=

Module[fb1=b2=c2=0,c1,a,b,c,x,y,ap,ahax,ahay,bp,bhbx,bhby,cp,
chcx,chcy,op,ox,oy,hx,hy,hp,radiusg,
vertexA=fa1,a2g;vertexB=fb1,b2g;vertexC=fc1,c2g;
a=EuclideanDistance[vertexB,vertexC];
b=EuclideanDistance[vertexC,vertexA];
c=EuclideanDistance[vertexA,vertexB];
If[ab+ckbc+akca+b,Print[”No triangle”],
centroid=fcentroidx,centroidyg=(vertexA+vertexB+vertexC)/3;
midAB=(vertexA+vertexB)/2;midBC=(vertexB+vertexC)/2;
midCA=(vertexC+vertexA)/2;ap=fahax,ahayg=fa1,0g;
6.1 Manipulate 99

sol2=Solve[y-b2==-(a1-c1)(x-b1)/(a2-c2)&&y-c2==
(a2-c2)(x-c1)/(a1-c1),fx,yg];bp=fbhbx,bhbyg=fx,yg/.sol2[[1]];
sol3=Solve[y-c2==-a1(x-c1)/a2&&y==a2 x/a1,fx,yg];
cp=fchcx,chcyg=fx,yg/.sol3[[1]];
sol4=Solve[x==a1&&fx,yg2InfiniteLine[fvertexB,bpg],fx,yg];
hp=fhx,hyg=fx,yg/.sol4[[1]];
sol5=Solve[x==(b1+c1)/2&&(y-(c2+a2)/2)==
-(a1-c1)(x-(c1+a1)/2)/(a2-c2),fx,yg];
op=fox,oyg=fx,yg/.sol5[[1]];
radius=EuclideanDistance[midBC,(op+hp)/2];
Show[Graphics[fText[Style[A,Italic,12],vertexA+f.01,.2g],
Text[Style[B,Italic,12],vertexB-f.2,.15g],
Text[Style[C,Italic,12],vertexC-f-.2,.15g],
Text[Style[”HA ”,Italic,12],fahax,b2-.2g],
Text[Style[”MA ”,Italic,12],f(b1+c1)/2,b2-.2g],
Black,PointSize[.015],Point[fvertexA,vertexB,vertexC,hp,op,centroidg],
Red,Point[fap,bp,cp,midAB,midBC,midCA,(hp+vertexA)/2,
(hp+vertexB)/2,(hp+vertexC)/2g],PointSize[.02],Point[(hp+op)/2],
Blue,Line[ffvertexB,vertexAg,fvertexB,vertexCg,fvertexC,vertexAg,
fvertexA,apg,fvertexC,cpg,fvertexB,bpg,fvertexC,midABg,fvertexA,
midBCg,fvertexB,midCAg,fvertexA,opg,fvertexB,opg,fvertexC,opg,
fop,hpg,fvertexA,(hp+op)/2g,fvertexB,(hp+op)/2g,fvertexC,
(hp+op)/2gg],Red,Circle[(op+hp)/2,radius]g],PlotRange!All]]]
In the next example, we consider that the vertices B and C are fixed, whereas the
vertex A is free, i.e., its coordinates are free. Thus, we can see how the Euler’s circle
varies according to A.
Manipulate[
Quiet@eulersNinePointsCyrcle[a1,a2],
ffa1,1,”abscisa”g,1,6,0.0001,Appearance!”Labeled”g,
ffa2,4,”ordinate”g,3,7,0.001,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 6.8.

6.1.3 Frenet–Serret Trihedron of a Helix

The Frenet–Serret trihedron describes the kinematic properties of a particle mov-

ing along a continuously differentiable curve in the three-dimensional Euclidean
space R3 or the geometric properties of the curve itself irrespective of any
motion. More specifically, the formulas describe the derivatives of the so-called
tangent, normal, and binormal unit vectors in terms of each other. We assume that
100 6 Manipulate

Fig. 6.8 Euler’s nine points circle

x0 .t/2 Cy0 .t/2 Cz0 .t/2 ¤ 0 for all t, where r.t/ = (x.t/; y.t/; z.t// are the Cartesian
coordinates of a point r.t/ on the curve. Its arc length is
Z tp Z t
 0 
s.t/ D Œx0 .u/2 C Œy0 .u/2 C Œz0 .u/2 du D r .u/ du:
t0 t0

If x0 .t/2 C y0 .t/2 C z0 .t/2 ¤ 0 for all t in an interval, the curve is said to be smooth
on that interval. If the curve is smooth, the arc length function s.t/ is increasing, and
thus it has an inverse of the form t D t.s/. Then the curve r can be reparametrized
r.s/. In that case s is said to be the natural parameter of the curve.
The unit tangent vector is defined by
 
tangent.t/ D r0 .t/= r0 .t/

and its curvature by

 
 d tangent.s/  0 0 00
D  D ktangent .t/k D kr .t/  r .t/k :
 ds  0
kr .t/k kr0 .t/k3

We check that the circular helix has a constant curvature.

6.1 Manipulate 101

Clear[a,k,r,t]
$Assumptions=a>0&&k>0&&r2Reals&&t2Reals;
r[t ]:=fa Cos@t,a Sin@t,k tg; (* Circular helix with the usual
representation *)
Norm[Cross[r0 @t,r00 @t]]/Norm[r0 @t]3 //Simplify (* Curvature *)
a
a2 Ck2

We get the same result by the built-in function ArcCurvature.

ArcCurvature[r@t,t]//Simplify
a
a2 Ck2

The Frenet–Serret trihedron is discussed in the last chapter where the Viviani’s
window is exhibited (Sect. 11.1.2). The code presented here is rather elementary.
Clear[a,k,s,circle,helix,vt,vs,tangent,tprime,normal,binormal,sfinal,
vt,vs,t]
$Assumptions=a>0&&k>0&&s2Reals&&t2Reals;
vt[t ]:=fa Cos@t,a Sin@t,k tg; (* Circular helix with the usual
representation
n *)
h i h i o
vs[s ]:= a Cos p 2s 2 ; a Sin p 2s 2 ; p k2 s 2 (* Circular
a Ck a Ck a Ck
helix by the arc parameter; this form is useful when we discuss certain
elements of it: tangent, normal, binormal, etc. *)
tangent[s ]:=vs’@s (* Colored in green *)
tprime[s ]:=vs”@s
normal[s ]:=Simplify[tprime@s/Norm[tprime@s]] (* Colored in red *)
binormal[s ]:=Cross[tangent[s],normal[s]] (* Colored in blue *)
a=1;k=a/4; p (* Values of the constants *)
sfinal=4 a2 C k2 ; (* Final value of the arc parameter *)

helix=ParametricPlot3D[Evaluate[vs@s],fs,0,sfinalg,Axes!None,
Boxed! False];
circle= h hn h i h i oi
ParametricPlot3D Evaluate a Cos p 2s 2 ; a Sin p 2s 2 ; 0 ;
n oi a Ck a Ck
p
s; 0; 2 a2 C k2 ;

Animate[
Show[helix,circle,Graphics3D[ffPointSize[.02],Black,Point[vs@s]g,
fArrowheads[0.03],Thick,Darker[Green],
Arrow[fvs@s,vs@s+tangent@sg],Red,
Arrow[fvs@s,vs@s+normal@sg],Blue,
Arrow[fvs@s,vs@s+binormal@sg]gg],
ImageSize!180],fs,0,sfinalg,AnimationRate!.01,
AnimationRunning!False,SaveDefinitions!True]
See Fig. 6.9.
102 6 Manipulate

Fig. 6.9 Frenet-Serret

trihedron for a helix

Here we have approached the Frenet–Serret trihedron for the helix by an

elementary way. There exists also the built-in function FrenetSerretSystem. This
built-in function will be used later for the study of Viviani’s window.

6.1.4 Hyperboloid of One Sheet

Now we will introduce the hyperboloid of one sheet. This surface models a cooling
tower. A hyperboloid cooling tower was patented by the Dutch engineers Frederik
van Iterson and Gerard Kuypers in 1918. The general equation of the hyperboloid
of one sheet is

x2 y2 z2
C  D 1; a > 0; b > 0; c > 0:
a2 b2 c2
6.1 Manipulate 103

Each hyperboloid of one sheet admits two families of rectilinear generators.

These generators are of the form

x=a  z=c D .1  y=b/; x=a C z=c D .1= /.1 C y=b/;

and

x=a  z=c D .1 C y=b/; x=a C z=c D .1=/.1  y=b/;

where and  are real parameters. Obviously, each generator (straight line) of
each family lies entirely on the surface. We find the generators depending on the
variable z.
Clear[a,b,c,x,y,z, ,]
Solve[x/a-z/c== (1-y/b)&& (x/a+z/c)==1+y/b,fx,yg]
Solve[x/a-z/c==(1+y/b)&&(x/a+z/c)==1-y/b,fx,yg]
 
a.zC2c z 2 / b.cC2z Cc 2 /
x ! c 1C 2 ; y !
. / c.1C 2 /
 
a.zC2cz2 / b.cC2zCc2 /
x ! c 1C2 ; y !  c 1C2
. / . /
Thus, its rectilinear generators have the equations
   
a z C 2c  z 2 b c C 2z C c 2
xD ; yD ; z D z;
c .1 C 2 / c .1 C 2 /
this is a straight line depending on the parameter
   
a z C 2c  z2 b c C 2z C c2
xD ; yD ; z D z;
c .1 C 2 / c .1 C 2 /
this is a straight line depending on the parameter :

For an application, we consider below the elliptic hyperboloid of one sheet of

equation z2 D 4x2 C 4y2  4, and its two rectilinear generators, for z 2 Œ3; 2.
Suppose we introduce some particular values for and . Then their graphs given
below are obvious.
Clear[ ; ]

rectilinearGeneratorsHyperboloid[ ;  ]:=
Module[fa=b=1,c=2,denominator1=Divide[1,c(1+ 2 )],
denominator2=Divide[1,c(1+2 )],t,u,zg,
Show[ParametricPlot3D[fa Sqrt[1+u2 ]Cos[t],b Sqrt[1+u2 ]Sin[t],c ug,
ft,0,2g,fu,-3/c,2/cg,
104 6 Manipulate

Fig. 6.10 Elliptic

hyperboloid

PlotStyle!fDirective[Opacity[0.3],Lighter[Green],PlotPoints!200]g,
MaxRecursion!12,Mesh!None,Boxed!False,
ViewPoint!f15,0,1g,Axes!False],
ParametricPlot3D[fa(z+2c -z 2 )denominator1,
b(-c+2z +c 2 )denominator1,zg,fz,-3,2g,PlotStyle!Black],
ParametricPlot3D[fa(z+2c-z2 )denominator2,
-b(-c+2z+c2 )denominator2,zg,fz,-3.1,1.95g,PlotStyle!Black],
PlotRange!All,ImageSize!150]]

Manipulate[
Quiet@rectilinearGeneratorsHyperboloid[ ; ],
ff ,2,”parameter”g,-130,130,.1,Appearance!”Labeled”g,
ff,2,”parameter”g,-140,140,.1,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 6.10.
Chapter 7
Ordinary Differential Equations

Two tutorials issued by Wolfram Research are very helpful along this chapter,
namely, [74] and [75].

7.1 Simple Differential Equations

One of the simplest ordinary differential equations is x0 .t/ D x.t/: It is known that
if x W Œa; b ! R and is smooth enough, its general solution has the form x.t/ D cet ,
where c is a constant. We admit that the constant is a real number. A good source
for ordinary differential equations is [31]
Clear[t,x]
sol=DSolve[fx’[t]==x[t]g,x[t],t]
ffxŒt ! et CŒ1gg
We pick out a specific solution using /. (ReplaceAll).
y[t ]=x[t]/.sol
Plot[y[t]/.fC[1]!1g,ft,-1,1g,ImageSize!180,Ticks!ff-1,0,1g,f1,Egg]
fet CŒ1g
See Fig. 7.1.
In the next example, we study an initial value problem for a differential equation
of the first order.

© Springer International Publishing AG 2017 105

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_7
106 7 Ordinary Differential Equations

Fig. 7.1 A simple ordinary

differential equation

Fig. 7.2 Other simple ordinary differential equations

Clear[x,y]
Flatten[
fsol=DSolve[fx’[t]==x[t],x[0]-2==0g,x[t],t],eqn=fy’[t]==y[t],y[0]-2==0g;
sol1=DSolve[eqn,y,t] (* Here function y is given without argument and
the output form is changed *)g,1]
Flatten[fPlot[x[t]/.sol,ft,0,1g,#],Plot[y[t]/.sol1,ft,0,1g,#,
PlotStyle!Red]g&/@ffAxesOrigin!f0,0g,ImageSize!150,
Ticks!fff0.5,”1/2”g,1.g,f1,2,3,5gggg]
{{x[t]! 2 et },{y!Function[{t},2 et ]}}
See Fig. 7.2.

7.2 Systems of Ordinary Linear Homogeneous Differential

Equations

We consider the following matrix with real entries ma D ff1; 2; 2g; f1; 4; 2g;
f1; 5; 3gg and present some solutions.
Clear[ma,t,x,y,z, ]
ma=ff1,-2,2g,f1,4,-2g,f1,5,-3gg;
CharacteristicPolynomial[ma, ];
Eigenvalues[N[ma]] (* Matrix ma has real and distinct eigenvalues *)
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 107

Fig. 7.3 A system of simple ordinary differential equations

sol=Flatten[DSolve[fx’[t]==x[t]-2y[t]+2z[t],y’[t]==x[t]+4y[t]-2z[t],
z’[t]==x[t]+5y[t]-3z[t]g,fx[t],y[t],z[t]g,t]];
x[t ]=sol[[1,2]]; y[t ]=sol[[2,2]]; z[t ]=sol[[3,2]];
functionsarray=Table[fx[t],y[t],z[t]g/.fC[1]!i,C[2]!j,C[3]!kg,
fi,-5,5,2g,fj,-5,5,2g,fk,-2,2g];
tograph=Flatten[functionsarray,1];
ParametricPlot3D[Evaluate[tograph],ft,-2,2g,
PlotStyle!fBlue,Red,Darker[Green]g,Ticks!ff-2,0,2g,f-2,0,2g,
f-2,0,2gg,PlotRange!ff-2,2g,f-2,2g,f-2,2gg],
ParametricPlot3D[Evaluate[tograph],ft,-2,2g,
PlotStyle!fBlue,Red,Darker[Green]g,Boxed!False,
Axes!False,PlotRange!ff-3,3g,f-3,3g,f-3,3gg]
{2.,-1.,1.}
See Fig. 7.3.
The next code shows how the same system as an initial value problem looks like.
The solution is a curve in the solid space.
Clear[x,y,z,t]
A=ff1,-2,2g,f1,4,-2g,f1,5,-3gg;
X[t ]=fx@t,y@t,z@tg;
system=MapThread[#1==#2&,fX’[t],A.X[t]g];
sol=DSolve[Join[system,fx@0==1,y@0==2,z@0==-1g],fx,y,zg,t];
ParametricPlot3D[fx[t],y[t],z[t]g/.sol,ft,-2,1g,
ColorFunction!Function[fx,y,zg,Hue[z]],PlotStyle!fThickg,
PlotRange!All,Ticks!ff0,20g,f-20,0,20g,f-40,-20,0,20gg,
ImageSize!140]
See Fig. 7.4.
108 7 Ordinary Differential Equations

Fig. 7.4 An initial value

problem

An example of an initial value problem of a system of linear differential with

constant coefficients whose characteristic equation has one real root and two
complex roots is introduced below.
Clear[x,y,z]
A=ff-1,-2,2g,f-2,-1,2g,f-3,-2,3gg;
Eigenvalues[A]; (* One real and two complex eigenvalues, f1,i,-ig *)
X[t ]=fx@t,y@t,z@tg;
system=MapThread[#1==#2&,fX’[t],A.X[t]g];
sol=DSolve[system,fx,y,zg,t]/.fC[1]!-1,C[2]!1,C[3]!1g;
plot=ParametricPlot3D[fx@t,y@t,z@tg/.sol,ft,-9,1g,
ColorFunction!Function[fx,y,zg,Hue[z]],
PlotStyle!fThickg,PlotRange!All];
Show[plot,Graphics3D[ffPointSize[0.03],
ColorFunction!Function[fx,y,zg,Hue[z]],
Point[ffx[-9],y[-9],z[-9]g/.solg]g,fPointSize[0.02],
ColorFunction!Function[fx,y,zg,Hue[z]],
Point[ffx@1,y@1,z@1g/.solg]g,
fText[Style[”Initial point”,Bold,12],fx@-9,y@-9,z@-9+0.5g/.sol],
Text[Style[”Final point”,Bold,12],fx@1,y@1,z@1+0.5g/.sol]gg],
ImageSize!150]
See Fig. 7.5.
Two examples of planar systems of linear homogeneous differential equations
follow. In the first example, the eigenvalues are real and distinct in Fig. 7.6, whereas
in the second example, they are complex in Fig. 7.7, [74, pp. 43–44].
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 109

Fig. 7.5 An initial value

problem with a real and two
complex eigenvalues

solplanarsystemode[A ,x ,y ]:=
Module[fX,systemg,
X[t ]=fx@t,y@tg;
system=MapThread[#1==#2&,fX’[t],A.X[t]g];
sol=DSolve[system,fx,yg,t]]

Clear[x,y]
A=ff4,-6g,f1,-1gg;
Eigenvalues[A] (* Real and distinct eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-10,10,6g,fj,-10,10,6g]],2];
ParametricPlot[Evaluate[particularsols],ft,-3,3g,
PlotRange!ff-2,2g,f-1,1gg,ImageSize!300]
f2; 1g
See Fig. 7.6.
Clear[x,y]
A=ff7,-8g,f5,-5gg;
Eigenvalues[A] (* Complex eigenvalues *)
110 7 Ordinary Differential Equations

Fig. 7.6 Planar system with real and distinct eigenvalues

Fig. 7.7 Planar system with complex eigenvalues

solplanarsystemode[A,x,y];
particularsols=
Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-10,10,8g,fj,-10,10,8g]],2];
ParametricPlot[Evaluate[particularsols],ft,-35,35g,
PlotRange!ff-20,20g,f-10,10gg,ImageSize!300]
f1 C 2i; 1  2ig
See Fig. 7.7.

7.2.1 Singular Points of the Linear Homogeneous Planar

Differential Equations

In some examples we show the behavior of the solutions around their singular points
in the case of several linear homogeneous planar differential equations.
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 111

Fig. 7.8 Planar system with

null eigenvalues

Clear[x,y]
A=ff0,0g,f-3,0gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Both eigenvalues are null *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[
Table[fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-10,3g,
PlotRange!ff-2,2g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!200]
eigenvalues = f0; 0g
See Fig. 7.8.
Clear[x,y]
A=ff2,0g,f0,2gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Real positive equal
eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-3,3g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!200]
eigenvalues = f2; 2g
See Fig. 7.9.
Clear[x,y]
A=ff1/2,1g,f0,1/2gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Real equal positive
eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-15,4g,
PlotRange!ff-1.3,1.3g,f-1.1,1.1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180] ˚

eigenvalues = 12 ; 12
See Fig. 7.10.
112 7 Ordinary Differential Equations

Fig. 7.9 Planar system with

positive equal eigenvalues, 1

Fig. 7.10 Planar system with

positive equal eigenvalues, 2

Clear[x,y]
A=ff1/2,0g,f0,2gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Distinct positive eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-15,3g,
PlotRange!ff-1.3,1.3g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180] ˚

eigenvalues = 2; 12
See Fig. 7.11.
Clear[x,y]
A=ff0,1g,f1,0gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Distinct eigenvalues,
one positive, one negative *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,1.2g,fj,-3,2,1.2g]],2];
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 113

Fig. 7.11 Planar system with

distinct positive eigenvalues

Fig. 7.12 Planar system with

one positive and one negative
eigenvalues

ParametricPlot[Evaluate[particularsols],ft,-4,4g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180]
eigenvalues = f1; 1g
See Fig. 7.12.
Clear[x,y]
A=ff-2,0g,f0,-2gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Real equal negative
eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,1.2g,fj,-3,2,1.2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-3,3g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180]
eigenvalues = f2; 2g
See Fig. 7.13.
114 7 Ordinary Differential Equations

Fig. 7.13 Planar system with

equal negative eigenvalues

Fig. 7.14 Planar system with

complex distinct eigenvalues
with negative real parts

Clear[x,y]
A=ff-1,-1g,f1,-1gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Complex distinct eigenvalues
with negative real parts *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-10,4g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180]
eigenvalues = f1 C i; 1  ig
See Fig. 7.14.
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 115

Fig. 7.15 Planar system with

complex distinct eigenvalues
with positive real parts

Clear[x,y]
A=ff1/2,-2g,f2,0gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Complex distinct
eigenvalues, the real part is positive *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,2g,fj,-3,2,2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-10,3g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!180] n  p  p o
eigenvalues = 14 1 C 3i 7 ; 14 1  3i 7

See Fig. 7.15.

Clear[x,y]
A=ff0,-1g,f1,0gg;
Print[”eigenvalues = ”,Eigenvalues[A]] (* Pure complex eigenvalues *)
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[
Table[fx@t,y@tg/.sol/.fC[1]!1/i,C[2]!1/jg,fi,-3,2,1.2g,fj,-3,2,1.2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-4,4g,
PlotRange!ff-1,1g,f-1,1gg,Ticks!ff-1,1g,f-1,1gg,
ImageSize!200]
eigenvalues = fi; ig
See Fig. 7.16.
We collect all the above results in a single code.
116 7 Ordinary Differential Equations

Fig. 7.16 Planar system with

pure complex eigenvalues

Clear[x,y]
mat=fff0,0g,f-3,0gg,ff2,0g,f0,2gg,ff1/2.,1g,f0,1/2gg,
ff1./2,0g,f0,2gg,ff0,1g,f1,0gg,ff-2,0g,f0,-2gg,ff-1,-1g,f1,-1gg,
ff1,-3g,f3,1gg,ff0,-1g,f1,0ggg;
colour=fRed,Blue,Magenta,Green,Darker[Red],Brown,Blue,
Darker[Green],Redg;
font=f”Helvetica”,”Times”,”Courier”,”Old English Text MT”g;
Table[A=mat[[i]];
eigen=Eigenvalues[A];
solplanarsystemode[A,x,y];
particularsols=Partition[Flatten[Table[
fx@t,y@tg/.sol/.fC[1]!1/m,C[2]!1/jg,fm,-3,2,1.2g,fj,-3,2,1.2g]],2];
ParametricPlot[Evaluate[particularsols],ft,-8,3g,
PlotLabel!ToExpression[”eigenvalues”]==TraditionalForm[eigen],
LabelStyle!fcolour[[i]],FontFamily!font[[Mod[i,4]+1]]g,
PlotRange!ff-1.001,1g,f-1,1.1gg,Ticks!ff-1,1g,f-1,1gg,
PlotStyle!fDirective[Thickness!0.008,color[[i]]]g],fi,1,9g]
See Fig. 7.17.

7.2.2 Multiple Equilibria

7.2.2.1 Tunnel-Diode Circuit

Consider the following system of planar nonlinear ordinary differential equations

(
x0 D 0:5.h.x/ C y/;
y0 D 0:2.x  1:5y C 1:2/;
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 117

Fig. 7.17 All graphs of the earlier planar systems

where h.x/ D 17:76x  103:79x2 C 229:62x3  226:31x4 C 83:72x5 . The equilibrium

points of this system are the real solutions of the system
(
0:5.h.x/ C y/ D 0;
x  1:5y C 1:2 D 0:

Let P1 , P2 , and P3 be the equilibrium points. The first picture shows the three
equilibrium points. The second picture shows the trajectories of the system in the
phase-plane near the equilibrium points.
118 7 Ordinary Differential Equations

Clear[x,y,h,p,A]
h[x ]=17.76x-103.79x2 +229.62x3 -226.31x4 +83.72x5 ;
A=f0.5f-h[x],yg,0.2f-x,-1.5 y+1.2gg;
diffA=ffD[A[[1,1]],x],D[A[[1,2]],y]g,fD[A[[2,1]],x],D[A[[2,2]],y]gg;
soll=Solve[h[x]-y==0&&-x-1.5y+1.2==0,fx,yg,Reals];
Table[p[i]=fx,yg/.soll[[i]],fi,1,3g];
Table[Eigenvalues[diffA/.fx!p[i][[1]],y!p[i][[2]]g],fi,1,3g];
fShow[Plot[(1.2-x)/1.5,fx,0,1.2g,
PlotStyle!fDirective[Thickness!0.005,Blue]g],
Plot[h[x],fx,0,1.2g,PlotRange!ff0,1.25g,f0,1gg,
PlotStyle!fDirective[Thickness!0.005,Blue]g],
Graphics[ffPointSize[0.017],Red,Point[p[1]],Point[p[2]],Point[p[3]]g,
fText[Style[”P1 ”,Italic,12],p[1]+f0.068,0.04g],
Text[Style[”y=h(x)”,Italic,12],p[3]+f-.028,.67g],
Text[Style[”x+1.5y=1.2”,Italic,12],f.67,.56g],
Text[Style[”P2 ”,Italic,12],p[2]+f0.06,0.04g],
Text[Style[”P3 ”,Italic,12],p[3]+f0.07,0.04g]gg],
PlotRange!All,ImageSize!f250,180g],
eqs=Sequence[u’[t]==0.5(-h[u[t]]+v[t]),v’[t]==0.2(-u[t]-1.5 v[t]+1.2)];
Show[Table[sol2=NDSolve[feqs,u[0]==p[1][[1]]+i/4,
v[0]==p[1][[2]]+j/4g,fu,vg,ft,0,100g];
sol3=NDSolve[feqs,u[0]==p[2][[1]]+i/4,v[0]==p[2][[2]]+j/4g,fu,vg,
ft,0,100g];
sol4=NDSolve[feqs,u[0]==p[3][[1]]+i/4,v[0]==p[3][[2]]+j/4g,fu,vg,
ft,0,100g];
ParametricPlot[fEvaluate[fu[t],v[t]g/.sol2],Evaluate[fu[t],v[t]g/.sol3],
Evaluate[fu[t],v[t]g/.sol4]g,ft,0,100g],fi,-2,2g,fj,-2,2g],
Graphics[fPointSize[0.015],Red,Point[p[1]],Point[p[2]],Point[p[3]],
Black,
Text[Style[”P1 ”,Italic,12],p[1]+f0.068,0g],
Text[Style[”P2 ”,Italic,12],p[2]+f-0.065,0g],
Text[Style[”P3 ”,Italic,12],p[3]+f0.06,-0.065g]g],
PlotRange!ff-.4,1.4g,f-.3,1.3gg,AxesOrigin!f0,0g,
ImageSize!f320,300g]g
See Fig. 7.18.

7.2.2.2 Five Equilibrium Points

Here we introduce a planar system of ordinary differential equations with 5

equilibrium points. We introduce two codes.
In the first code we study certain trajectories in order to clearly see the phase
portrait. The second code is much shorter.
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 119

Fig. 7.18 Tunnel-diode circuit

The system under discussion is given below:

(
x0 D y  x3 =2 C x4 =2 C 3y4 =2  3y5 =4;
y0 D 4x  6y3 C x4 C 3y4  x5 :

Here is the first code:

Clear[x,y]
equipointreal=N[Simplify[Solve[-y-x3 /2+x4 /2+3y4 /2-3y5 /4==0&&
4x-6y3 +x4 +3y4 -x5 ==0,fx,yg,Reals]]];
equipointreal=fx,yg/.equipointreal;

listofpoints[m ,h ]:=
Module[fk,u,vg,
For[k=0,k<m,k++,
u D x C h.y  x3 =2 C x4 =2 C 3y4 =2  3y5 =4/I
v D y C h.4x  6y3 C x4 C 3y4  x5 /I
AppendTo[points,fu,vg];x=u;y=v]]

points=fequipointreal[[5]]g;
fx,yg=equipointreal[[5]]+f-0.01,0.01g;
h=0.02;nsteps=8;listofpoints[nsteps,h];
arrow5nw=fpoints[[Floor[nsteps/2]+3]],points[[Floor[nsteps/2]+4]]g;
plot5nw=ListLinePlot@N@points;

points=fequipointreal[[5]]g;
fx,yg=equipointreal[[5]]+f0.01,-0.01g;
h=0.02;nsteps=14;listofpoints[nsteps,h];
arrow5se=fpoints[[Floor[nsteps/2]+2]],points[[Floor[nsteps/2]+3]]g;
plot5se=ListLinePlot@N@points;
120 7 Ordinary Differential Equations

points=fequipointreal[[5]]g;
fx,yg=equipointreal[[5]]+f-0.01,-0.01g;
h=0.02;nsteps=150;listofpoints[nsteps,-h];
arrow5sw=fpoints[[Floor[nsteps/3]+4]],points[[Floor[nsteps/3]+3]]g;
plot5sw=ListLinePlot@N@points;

points=fequipointreal[[5]]g;
fx,yg=equipointreal[[5]]+f0.01,0.01g;
h=0.02;nsteps=48;listofpoints[nsteps,-h];
arrow5ne=fpoints[[Floor[nsteps/2]+5]],points[[Floor[nsteps/2]+4]]g;
plot5ne=ListLinePlot[N[points]];

points=fequipointreal[[4]]g;
fx,yg=equipointreal[[4]]+f-0.01,0.01g;
h=0.02;nsteps=9;listofpoints[nsteps,h];
arrow4nw=fpoints[[Floor[nsteps/2]+4]],points[[Floor[nsteps/2]+5]]g;
plot4nw=ListLinePlot@N@points;

points=fequipointreal[[4]]g;
fx,yg=equipointreal[[4]]+f0.00001,-0.00001g;
h=0.02;nsteps=265;listofpoints[nsteps,h];
arrow4se=fpoints[[Floor[nsteps/11]+2]],points[[Floor[nsteps/11]+5]]g;
plot4se=ListLinePlot[N[points]];

points=fequipointreal[[4]]g;
fx,yg=equipointreal[[4]]+f-0.01,-0.01g;
h=0.02;nsteps=29;listofpoints[nsteps,-h];
arrow4sw=fpoints[[Floor[nsteps/2]+3]],points[[Floor[nsteps/2]+2]]g;
plot4sw=ListLinePlot@N@points;

points=fequipointreal[[4]]g;
fx,yg=equipointreal[[4]]+f0.01,0.0001g;
h=0.02;nsteps=100;listofpoints[nsteps,-h];
arrow4e=fpoints[[Floor[nsteps/5]+3]],points[[Floor[nsteps/5]+2]]g;
plot4e=ListLinePlot@N@points;

points=fpoints[[Floor[nsteps/2]]]g;
fx,yg=Flatten[points]-f0,0.001g;
h=0.02;nsteps=100;listofpoints[nsteps,h];
arrow3ws=fpoints[[Floor[nsteps/4]+1]],points[[Floor[nsteps/4]+2]]g;
plot3ws=ListLinePlot@N@points;
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 121

points=fequipointreal[[3]]g;
fx,yg=equipointreal[[3]]+f-0.01,0.01g;
h=0.02;nsteps=30;listofpoints[nsteps,h];
arrow3nw=fpoints[[Floor[nsteps/2]+9]],points[[Floor[nsteps/2]+10]]g;
plot3nw=ListLinePlot@N@points;

points=fequipointreal[[3]]g;
fx,yg=equipointreal[[3]]+f-0.1,0.001g;
h=0.02;nsteps=50;listofpoints[nsteps,h];
arrow3w=fpoints[[Floor[nsteps/2]+2]],points[[Floor[nsteps/2]+3]]g;
plot3w=ListLinePlot@N@points;

points=fequipointreal[[2]]g;
fx,yg=equipointreal[[2]]+f0.01,0.005g;
h=0.02;nsteps=173;listofpoints[nsteps,h];
arrow2e=fpoints[[Floor[nsteps/2]+60]],points[[Floor[nsteps/2]+61]]g;
plot2e=ListLinePlot[N[points],InterpolationOrder!2];

points=fequipointreal[[2]]g;
fx,yg=equipointreal[[2]]+f-0.01,-0.001g;
h=0.02;nsteps=1000;listofpoints[nsteps,h];
arrow2sw=fpoints[[Floor[nsteps/6]+24]],points[[Floor[nsteps/6]+25]]g;
plot2sw=ListLinePlot@N@points;

points=fequipointreal[[2]]g;
fx,yg=equipointreal[[2]]+f-0.001,0.001g;
h=0.02;nsteps=150;listofpoints[nsteps,-h];
arrow2n=fpoints[[Floor[nsteps/4]+9]],points[[Floor[nsteps/4]+8]]g;
plot2n=ListLinePlot@N@points;

points=fequipointreal[[2]]g;
fx,yg=equipointreal[[2]]+f0.01,-0.01g;
h=0.02;nsteps=74;listofpoints[nsteps,-h];
arrow2s=fpoints[[Floor[nsteps/3]+9]],points[[Floor[nsteps/3]+8]]g;
plot2s=ListLinePlot@N@points;

fx,yg=f2.7,2.65g;
points=ffx,ygg;
h=0.002;nsteps=18;listofpoints[nsteps,h];
arrow5fs=fpoints[[Floor[nsteps/2]]],points[[Floor[nsteps/2]+1]]g;
plot5fs=ListLinePlot[N@points,InterpolationOrder!2];

fx,yg=f2.7,2.9g;
points=ffx,ygg;
h=0.002;nsteps=9;listofpoints[nsteps,h];
122 7 Ordinary Differential Equations

arrow5fn=fpoints[[Floor[nsteps/2]+3]],points[[Floor[nsteps/2]+4]]g;
plot5fn=ListLinePlot[N@points,InterpolationOrder!2];
h=0.02;

fx,yg=f-2,-1.5g;
points=ffx,ygg;
h=0.02;nsteps=200;listofpoints[nsteps,h];
arrow1fn=fpoints[[Floor[nsteps/5]+10]],points[[Floor[nsteps/5]+11]]g;
plot1fn=ListLinePlot[N[points],InterpolationOrder!4];

fx,yg=f-2.7,-1.1g;
points=ffx,ygg;
h=0.002;nsteps=32;listofpoints[nsteps,h];
arrowfsn=fpoints[[Floor[nsteps/2]+1]],points[[Floor[nsteps/2]+2]]g;
plotfsn=ListLinePlot[N[points]];

Show[plot5nw,plot5se,plot5ne,plot5sw,plot4nw,plot4se,plot4sw,
plot4e,plot3nw,plot2e,plot2sw,plot2n,plot2s,plot5fs,plot5fn,plot1fn,
plotfsn,plot3ws,
Graphics[ffPointSize[0.013],Red,Point[fequipointreal[[5]],
equipointreal[[4]],equipointreal[[3]],equipointreal[[2]],
equipointreal[[1]]g]g,
Arrowheads[0.028],Green,
Arrow[farrow5nw,arrow5sw,arrow5ne,arrow5se,arrow4nw,arrow4se,
arrow4sw,arrow4e,arrow3nw,arrow2e,arrow2sw,arrow2n,arrow2s,
arrow5fs,arrow5fn,arrow1fn,arrowfsn,arrow3wsg]g],
Ticks!ff-3,-2,-1,1,2,3g,f-3,-2,-1,1,2,3,4gg,AxesOrigin!f0,0g,
PlotPoints!40,AspectRatio!1,ImageSize!300,PlotRange!All]
See Fig. 7.19.
The second code follows:
Clear[x,y]
equipoints=Solve[-y-x3 /2+x4 /2+3y4 /2-3y5 /4==0&&
4x-6y3 +x4 +3y4 -x5 ==0,fx,yg,Reals];
eqreal=fx,yg/.equipoints;

vector=ffeqreal[[5]]+f-0.01,0.01g,eqreal[[5]]+f0.01,-0.01g,
eqreal[[5]]+f-0.01,-0.01g,eqreal[[5]]+f0.01,0.01g,
eqreal[[5]]+f0.01,-0.01g,eqreal[[5]]+f-0.01,-0.01g,
eqreal[[5]]+f0.01,0.01g,eqreal[[4]]+f-0.01,0.01g,g
eqreal[[4]]+f0.00001,-0.00001g,eqreal[[4]]+f-0.01,-0.01g,
eqreal[[4]]+f0.01,0.0001g,eqreal[[3]]+f-0.01,0.01g,
eqreal[[2]]+f0.01,0.005g,eqreal[[2]]+f-0.01,-0.001g,
eqreal[[2]]+f-0.001,0.001g,eqreal[[2]]+f0.01,-0.01g,
f2.7,2.65g,f2.7,2.9g,f-2,-1.5g,f-2.7,-1.1gg,
f8,14,150,48,14,150,48,9,265,29,100,30,173,1000,150,74,18,9,200,32g,
7.2 Systems of Ordinary Linear Homogeneous Differential Equations 123

Fig. 7.19 First code for the

five equilibrium case

f0.02,0.02,-0.02,-0.02,0.02,-0.02,-0.02,0.02,0.02,-0.02,-0.02,0.02,
0.02,0.02,-0.02,-0.02,0.002,0.002,0.02,0.002g,
f2,2,3,2, 2,3,2,2,11,2,5,2, 2,6,4,3,2,2,5,2g,
ff3,4g,f2,3g,f4,3g,f5,4g,f2,3g,f4,3g,f5,4g,f4,5g,f2,5g,f3,2g,f3,2g,
f9,10g,f60,61g,f24,25g,f9,8g,f9,8g,f0,1g,f3,4g,f10,11g,f1,2ggg;
aArrows=aPlots=fg;
Do[fx,yg=Flatten[vector[[1,m]]];points=ffx,ygg;
nsteps=vector[[2,m]];
For[k=0,k<nsteps,k++,
u=x+vector[[3,m]](-y-x3 /2+x4 /2+3y4 /2-3y5 /4);
v=y+vector[[3,m]](4x-6y3 +x4 +3y4 -x5 );
AppendTo[points,fu,vg];x=u;y=v];
aArrows=AppendTo[aArrows,
fpoints[[Floor[nsteps/vector[[4,m]]+vector[[5,m,1]]]]],
points[[Floor[nsteps/vector[[4,m]]+vector[[5,m,2]]]]]g];
aPlots=AppendTo[aPlots,ListLinePlot[N@points]],
fm,Length[vector[[1]]]g]
Show[Flatten[aPlots],
Graphics[ffPointSize[0.013],Red,
Point[feqreal[[5]],eqreal[[4]],eqreal[[3]],eqreal[[2]],eqreal[[1]]g]g,
Arrowheads[0.028],Green,Arrow[aArrows]g],AxesOrigin!f0,0g,
AspectRatio!1,Ticks!ff-3,-2,-1,1,2,3g,f-3,-2,-1,1,2,3,4gg,
PlotRange!All]
See Fig. 7.20.
124 7 Ordinary Differential Equations

Fig. 7.20 Second code for the five equilibrium case

7.3 On Two Runge–Kutta Methods of the Fourth Order

In this section we will introduce an explicit fourth-order Runge–Kutta method,

which requires only three evaluations at each step, and a semi-explicit Runge–
Kutta method of fourth order with two evaluations at each step. These methods
were worked out in [49, 50], and [51].

7.3.1 An Explicit Runge–Kutta Method

For the explicit fourth-order Runge–Kutta method which requires only three
evaluations at each step, we consider a system of autonomous ordinary differential
equations of the first order in the vectorial form

dx
D f .x/ (7.1)
dt

where x D .x1 ; : : : ; xs /, f D f .x1 ; : : : ; xs /, t 2 R.

7.3 On Two Runge–Kutta Methods of the Fourth Order 125

We suppose that the vector x D x0 is given at t D a. The interval Œa; b, where
the independent variable t belongs, is partitioned by a finite mesh hi > 0 and
m
X
aC hi D b: (7.2)
iD1

We also suppose that the function f is sufficiently smooth such that all that
follow are correct. This implies that the initial value problem stated at x.a/ D x0
has a unique solution which exists on the whole interval Œa; b. The sequence of
approximations .xn / of the unknown function x solution of the initial value problem
consisting in the system of differential equations (7.1) and the initial condition is
defined by

xnC1 D xn C hn .˛0 k0;n C ˛1 k1;n C ˛2 k2;n / ;

 
ki;n D f argi;n ;
0 1
X2
argi;n D xn C hn @ i;j kj;n1 C i;0 k0;n C i;1 k1;n A ;
jD0

i D 0; 1; 2 and n D 1; 2; : : : ; m:

For the first step, we take

ki;1 D x0 ; i D 0; 1; 2:

The method is explicit, so we have

0;0 D 0;1 D 1;1 D 0:

To get an accuracy up to the terms in h4 , the following relations concerning the

parameters involved have to hold

0;0 C 0;1 C 0;2 D 0 ;

1;0 C 1;1 C 1;2 C 1;0 D 1 ;
2;0 C 2;1 C 2;2 C 2;0 C 2;1 D 2 ;
˛0 C ˛1 C ˛2 D 1;
2 .˛0 0 C ˛1 1 C ˛2 2 / D 1;
3 ˛0 20 C ˛1 21 C ˛2 22  D 1;
4 ˛0 30 C ˛1 31 C ˛2 32 D 1;
6 .˛1 1;0 0 C ˛2 .2;0 0 C 2;1 1 // D 1;
8 .˛11;0 0 1 C ˛2 2 .2;0 0 C 2;1 1 // D 1;
12 ˛1 1;0 20 C ˛2 2;0 20 C 2;1 21 D 1;
126 7 Ordinary Differential Equations

24˛2 1;0 2;1 0 D 1;

0;0 0 C 0;1 1 C 0;2 2 D 0 ;
1;0 0 C 1;1 1 C 1;2 2 C 1;0 D 1 ;
2;0 0C 2;1 1 C 2;2 2 C 2;0 C 2;1 D 2 ;
P2 P2
2 iD0 ˛i kD0 i;k k C sS D 1;
P P
2 2 2
2 iD0 ˛i kD0 i;k k C sS D 1;
P
2
2 iD0 ˛i . i;1 1;0 0 C i;2 .2;0 0 C 2;1 1 // C sS D 1;

where

sS D ˛1 1;0 C ˛2 .2;0 C 2;1 / :

In [49] we have expressed these parameters in terms of 0 , 1 , 2 , 1;1 , and

2;1 . From the compatibility condition, we get 2 D 1 and the equality

60 1  2 .0 C 1 / C 1 D 0:

Successively we have the results

31  1 1  0
˛0 D ; 1;0 D ;
6 .1  0 / .1  0 / 40 .1  30 /
30  1 .1  0 / .41  1/ .30  1 /
˛1 D ; 2;0 D ;
6 .1  1/ .1  0 / 20 .1  0 / .1  60 1 /
1  60 1 .1  30 / .1  40 1 /
˛2 D ; 2;1 D ;
12 .1  1 / .1  0 / .1  0 / .1  60 1 /
sS D 1=2:

Further, if we denote

1  1
1;1 D ˇ; 2;1 D ; and ı0 D ;
0  1

it results

˛1 ˇ C ˛2 ı
1;0 D ıˇ; 2;0 D ı; 0;1 D  ;
˛0
ı .˛1 ˇ C ˛2 ı/ .ı  1/ .˛1 ˇ C ˛2 ı/
0;0 D ; 0;2 D 0  ;
˛0 ˛0
1;2 D 1  1;0 C .ı  1/ˇ; 2;2 D 2  .2;0 C 2;1 / C .ı  1/;
7.3 On Two Runge–Kutta Methods of the Fourth Order 127

Then we immediately have

31 1
˛0 D 6.10 /.1 0 /
I
30 1
˛1 D 6.1 1/.1 0 / I
160 1
˛2 D 12.1 1 /.10 /
I
(*0;0 D 0;1 D 1;1 D 0I *)
1 0
1;0 D 40 .13 0/
I
.10 /.41 1/.30 1 /
2;0 D 20 .1 0 /.160 1 / I
.130 /.140 1 /
2;1 D . 1 0 /.160 1 /
I
1;1 D 1  .1;0 C 0 1;0 C 2;1 /
1;1 D ˇ D 0:0I
2;1 D  D 0:0I
ı D 10 1
1
I
˛1 ˇC˛2 
0;0 D ı  ˛0
I
˛1 ˇC˛2 
0;1 D  ˛0 I
0;2 D 0  .ı  1/ ˛1 ˇC˛
˛0
2
I

1;0 D ı ˇI
1;2 D 1  1;0 C .ı  1/ˇI
2;0 D ı  I
2;2 D 2  .2;0 C 2;1 / C .ı  1/ I
0 C1
1  4.13 0 /0
 0 1;0  2;1
By some numerical experiments it seems reasonable to consider that
0 D 0:5; 1 D 0:0; 2 D 1:0; 1;1 D 1  .1;0 C 0 1;0 C 2;1 / :
We consider the next example. The approach by our Runge–Kutta method
follows.
f[fx ,y ,z g]:=f1,Cos[x]2 y-(1-Sin[x]Cos[x])z,(1+Sin[x]Cos[x])y+Sin[x]2 zg
(* The right-hand side of the system of differential equations *)
x0=f0,1,1g; (* Initial value *)
a=0; (* Leftmost boundary value *)
b=; (* Rightmost boundary value *)
m=1000; (* Number of steps *)
h=(b-a)/m; (* Step size uniformly distributed *)
kold0=kold1=kold2=x0;
Do[
knew0=f[x0+h( 0;0 kold0+ 0;1 kold1+ 0;2 kold2)];
knew1=f[x0+h( 1;0 kold0+ 1;1 kold1+ 1;2 kold2+1;0 knew0)];
knew2=f[x0+h( 2;0 kold0+ 2;1 kold1+ 2;2 kold2+2;0 knew0+2;1 knew1)];
x1=x0+h(˛0 knew0+˛1 knew1+˛2 knew2);
x0=x1;
kold0=knew0;kold1=knew1;kold2=knew2,
128 7 Ordinary Differential Equations

Fig. 7.21 The graphs of the exact and approximative solutions

fi,mg]
N[x1,15]
{3.14159,-23.1405,-0.989255}
The solution by the built-in function NDSolve is presented below.
Clear[x,y,z,t]
sol=NDSolve[fx’[t]==1,y’[t]==Cos[x[t]]2 y[t]-(1-Sin[x[t]]Cos[x[t]])z[t],
z’[t]==(1+Sin[x[t]]Cos[x[t]])y[t]+Sin[x[t]]2 z[t],x[0]==0,y[0]==1,z[0]==1g,
fx,y,zg,ft,g,AccuracyGoal!20,PrecisionGoal!20,
WorkingPrecision!35];
N[fx[],y[],z[]g/.sol,15] (* The solution by integration *)
N[ft,Exp[t]Cos[t]-Sin[t],Exp[t]Sin[t]+Cos[t]g/.t! ,15] (* The exact
solution at  *)
{{3.14159265358979,-23.1406926327793,-1.00000000000000}}
{3.14159265358979,-23.1406926327793,-1.00000000000000}

Plot[ft,Exp[t]Cos[t]-Sin[t],Exp[t]Sin[t]+Cos[t]g,ft,0,g,PlotRange!All]
(* The graph of the exact solution *)
See Fig. 7.21.

7.3.2 A Semi-explicit Runge–Kutta Method

We consider a system of first-order autonomous ordinary differential equations in

the vectorial form (7.1) where x D .x1 ; : : : ; xs /; f D f .x1 ; : : : ; xs /, t 2 Œa; b: We
admit that a vector is given x0 D x.a/ 2 Rs at t D a. The interval Œa; b, where
the independent variable t belongs, is partitioned by a finite mesh hi > 0 and (7.2)
holds.
We also suppose that the function f is sufficiently smooth such that all that
follow are correct. This implies that the initial value problem stated at x.a/ D x0
7.3 On Two Runge–Kutta Methods of the Fourth Order 129

has a unique solution, which exists on the whole interval Œa; b. The sequence of
approximations .xn / of the unknown function x solution of the initial value problem
consisting in the system of differential equations (7.1) and the initial condition is
well defined.
It is known that to each initial value problem, we assign a table of coefficients of
the form
Clear[a,b, ]
: ::
Grid[ffa ,. . . ,a , g,f::, ,::,::g,fa ,. . . ,a , g,fb ,. . . ,b ,gg,
1;1 1;q 1 q;1 q;q q 1 q
Dividers!ffFalse,False,False,Trueg,fFalse,False,False,Truegg]
a1;1 : : : a1;q 1
:: :: ::
: : :
aq;1 : : : aq;q q
b1 : : : bq
where
q
X
aj;i D j ;
iD1
q
X
xnC1 D xn C hn bi ki;n ; (7.3)
iD1
q
  X
ki;n D f argi;n ; argi;n D xn C hn ai;j kj;n :
jD1

A Runge–Kutta method is said to be explicit if ai;j D 0 for i  j; is said to be

semi-explicit if ai;j D 0 for i < j; is diagonally implicit if is semi-explicit and all the
ai;i are equal, [4].
We denote by p the order of the method. In [4] Alexander shows that there exists
a unique diagonally implicit formula with .q; p/ D .1; 2/, as the implicit midpoint
rule:
Grid[ff1/2,1/2g,f1,Nullgg,Dividers!ffFalse,Trueg,fFalse,Truegg]
1 1
2 2
1
and for .q; p/ D .2; 3/ and .q; p/ D .3; 4/, respectively, there exists exactly one
Astable diagonally implicit Runge–Kutta formula. These are given by
Print[”(q,p)=(2,3)”]
hnn o n o ˚
1 1 1 1 1 1 1 1 1 1 1

o
Grid 2
C p
2 3
; 0; 2
C p
2 3
; ;
3 2
C p ;
2 3 2
 p
2 3
; ;
2 2
; Null ,
Dividers ! ffFalse; False; Trueg; fFalse; False; Truegg
(q,p)=(2,3)
130 7 Ordinary Differential Equations

1 1 1 1
2
C p 0 2
C p
2 3 2 3
 13 1
2
C 1
p 1
 1
p
2 3 2 2 3
1 1
2 2
and
Clear[˛]
h i
p
Print “.q; p/ D .3; 4/; ”; “˛ D .2= 3/cos.=18/”
GridŒff.1 C ˛/=2; 0; 0; .1 C ˛/=2g; f˛=2; .1 C ˛/=2; 0; 1=2g;
f1
˚ C ˛; .1
 C 2˛/; .1 C  ˛/=2;
 .1 ˛/=2g;

1= 6˛ 2 ; 1  1= 3˛ 2 ; 1= 6˛ 2 ; Null ;
Dividers!ffFalse,False,False,Trueg,fFalse,False,False,Truegg
p
(q,p)=(3,4), ˛ D .2= 3/cos.=18/
1C˛ 1C˛
2
0 0 2
 ˛2 1C˛
2
0 1
2
1C˛ 1˛
1 C ˛ 1  2˛ 2 2
1
6˛ 2
1  3˛1 2 1
6˛ 2

The aim of the present subsection is to introduce a semi-explicit Runge–Kutta

method .q; p/ D .2; 4/.
We use the following relations to pass from a step to the next:
q
X  
xnC1 D xn C hn bi ki;n ; ki;n D f argi;n ;
iD1
2 3
q q
X X (7.4)
argi;n D xn C hn 4 i;j kj;n1 C ai;j kj;n 5 ;
jD1 jD1

ki;1 D x0 :

If we take
q
X  
i D i;j C ai;j ;
jD1

then instead of the first table, we get the next one

a=.;b=.;
: : : : :
Grid[ff 1;1 ; : : : ; 1;q ; a1;1 ; : : : ; a1;q ; 1 g; f::; ; ::; ::; ; ::; ::g,
f q;1 ; : : : ; q;q ; aq;1 ; : : : ; aq;q ; q g; f; ; ; b1 ; : : : ; bq ; gg;
Dividers!ffFalse,False,False,False,False,False,Trueg,
fFalse,False,False,Truegg]
7.3 On Two Runge–Kutta Methods of the Fourth Order 131

1;1 : : : 1;q a1;1 : : : a1;q 1

:: :: :: :: ::
: : : : :
q;1 : : : q;q aq;1 : : : aq;q q
b1 : : : bq
It is clear that if i;j D 0, then (7.4) reduces (7.3).
To achieve the order 4 in 2 stages, we expand xnC1 x .tn C hn / in Taylor
polynomials similar to [49]. To get the fourth order of accuracy, the next equalities
have to be true.

1;1 C 1;2 C a1;1 C a1;2 D 1 ;

2;1 C 2;2 C a2;1 C a2;2 D 2 ;
b1 C b2 D 1; (7.5)
b1 1 C b2 2 D 1=2; (7.6)
b1 21 C b2 22 D 1=3; (7.7)
b1 31 C b2 32 D 1=4; (7.8)
b1 .a1;1 1 C a1;2 2 / C b2 .a2;1 1 C a2;2 2 / D 1=6; (7.9)
b1 1 .a1;1 1 C a1;2 2 / C b2 2 .a2;1 1 C a2;2 2 / D 1=8; (7.10)
   
b1 a1;1 21 C a1;2 22 C b2 a2;1 21 C a2;2 22 D 1=12;
.b1 a1;1 C b2 a2;1 / .a1;1 1 C a1;2 2 / C
.b1 a1;2 C b2 a2;2 / .a2;1 1 C a2;2 2 / D 1=24:

From (7.5), (7.6), (7.7), and (7.8) it results

 p .  p .
b1 D b2 D 1=2; 1 D 3  3 3; 2 D 3 C 3 3;

and from (7.9)–(7.10) it follows

 p .
a1;1 1 C a1;2 2 D 2  3 12; (7.11)
 p .
a2;1 1 C a2;2 2 D 2 C 3 12; (7.12)

a1;1 C a2;1 D 2 ; (7.13)

a1;2 C a2;2 D 1 : (7.14)
132 7 Ordinary Differential Equations

Then from (7.11), (7.12), (7.13), and (7.14) we get

 p .
a2;2 D 3  3 6  ;
 p .  p
a1;1 D 2  3 12  2 C 3 ;
 p .  p
a2;1 D 1 C 3 4 C 2 C 3 ;

a1;2 D ;  real parameter:

If we choose  D a1;2 , then we get a semi-explicit two-step Runge–Kutta method

p .
of the fourth order, i.e., .q; p/ D .2; 4/. In that case we take 0;1 D 3 12 and
p .
1;1 D  3 12.
Chapter 8
Pi Formulas

Perhaps number  is the most famous and fascinating number in mathematics.

A long list of papers and books is dedicated to this number. We mention only
some of them: [2, 3, 6–9, 17, 18, 30, 63, 69, 71, 76], and [77]. More specific
references will be given as the context requires.
Along this chapter, we revise some results related to  and implicitly show how
powerful is Mathematica when handling very complicated expressions.
 is intimately related to the properties of circles and spheres in elementary
mathematics. At the advanced level, number  is present in complicated results
not obviously tied to mathematics. For a planar circle of radius r, its circumference
and area are given by

C D 2r and A D r2 :

8.1 Various Simple and Not So Simple Formulas

8.1.1 Vandermonde Identity

P1  .2k3/!! 2
vandermonde= kD0 .2k/!!
4

Regarding the speed of convergence we note some remarks.

© Springer International Publishing AG 2017 133

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_8
134 8 Pi Formulas

Fig. 8.1 Vandermonde

formula

P  2
vandermonde[n ]:= nkD0 .2k3/!!
.2k/!!
Table[N[vandermonde[k]],fk,6g] (* The first six terms *)
DiscretePlot[vandermonde[n],fn,50g,
Ticks!fAutomatic,f1.2728,1.27295,1.2731,f4/,“4/”ggg,
PlotStyle!Black,ImageSize!250]
Simplify[vandermonde[1]]
{1.25,1.26563,1.26953,1.27106,1.2718,1.27223}
See Fig. 8.1.

8.1.2 Sums Connected to Polygonal Numbers

A polygonal number is a number represented as dots or pebbles arranged in the

shape of a regular polygon. Some series of this kind follow.
P
{spol1= 1 3
kD1 k.2k1/.4k3/ ;
P1 p
3 3
spol2= kD1 .3k1/.3k2/ ;
p P1 12k5
spol3=4 3 kD1 8k.2k1/.3k1/.6k5/ ;
P1 864k.k1/C226
spol4=16 kD1 .12k1/.12k5/.12k7/.12k11/.4k1/.4k3/ }
f; ; ; g

8.1.3 Machbin’s and Machbin-Like Formulas

Relevant information on this topic may be found in many papers. We only mention
[70] and [76].
8.1 Various Simple and Not So Simple Formulas 135

˚    1 
FullSimplify 4ArcTan 15  ArcTan 239 ; (* Machbin’s original
formula 1706
1 *)  1   1 
8ArcTan 10  ArcTan 239  4ArcTan 515 ; (* Klingenstierna
1730 *)        
2ArcTan 12  ArcTan 17 ; ArcTan 12 C ArcTan 13 ; (* Euler 1738 *)
1 3
5ArcTan 7 C 2ArcTan 79 ; (* Euler 1755 *)
  1 1
4ArcTan 15  ArcTan 70 C ArcTan 99 ; (* Euler 1764 *)
1 1
ArcTan 2 C ArcTan 3 ; (* Hutton 1776 *)
   
2ArcTan 13 C ArcTan 17 ; (* Hutton 1776 *)
     
ArcTan 12 C ArcTan 15 C ArcTan 18 ; (* Strassnitzky 1844 *)
1 1  1 
12ArcTan 18 C 8ArcTan 57  5ArcTan 239 ; (* Gauss 1863 *)
1 1  1 
6ArcTan 8 C 2ArcTan 57 C ArcTan 239 ; (* Störmer 1896 *)
   

5ArcTan 1 C 2ArcTan 3
˚    7     79    

; ; ; ; ; ; ; ; ; ; ;
4 4 4 4 4 4 4 4 4 4 4 4

8.1.3.1 Kanada

In December 2002, Kanada computed  to over 1.24 trillion decimal digits. His
team first computed  in hexadecimal (base 16) to 1,030,700,000,000 places, using
the following two arctangent relations:
˚ 1 1  1 
FullSimplify 12ArcTan 49 C 32ArcTan 57  5ArcTan 239 C
 1 
12ArcTan 110443 ; (* Takano, [63] *)
1  1   1   1 

44ArcTan 57 C 7ArcTan 239  12ArcTan 682 C 24ArcTan 12943
˚(* Störmer

1896, [12] *)
;
4 4

8.1.4 Gregory and Leibniz Formula

This formula may be found in [71, (32)]. The convergence of this series is slow.
The graph below also shows the low speed of convergence.
Clear[s,n]
P k
s[n ]:=4 nkD0 .1/
2kC1
I
DiscretePlot[s[n],fn,50g,
Ticks!fAutomatic,f3,3.1,,3.2,3.3gg,
PlotStyle!Black,ImageSize!250]
s[1] (* The sum of the series *)
See Fig. 8.2.
136 8 Pi Formulas

Fig. 8.2 Gregory and

Leibniz formula

Fig. 8.3 Vardi formula

8.1.5 Vardi Formula

This formula may be found in [71, (8)]. In this formula number  is connected to
the  function of Riemann and is computed by it.
P k
vardi[n ]:= nkD1 3 41
k Zeta[k+1];

Table[N[vardi[k]],fk,1,26,5g] (* Some terms of the sequence *)

DiscretePlot[vardi[n],fn,50g,
Ticks!fAutomatic,f2.64,2.8,3,3.1,gg,
PlotStyle!Black,ImageSize!250]
vardi[1] (* The sum of the series *)
{0.822467,2.60687,3.01488,3.11152,3.13446,3.1399}
See Fig. 8.3.
8.1 Various Simple and Not So Simple Formulas 137

Fig. 8.4 Abraham Sharp

formula

8.1.6 Abraham Sharp Formula

This formula may be found in [71, (9)].

p P k
s[n ]:= 12 nkD0 3k.1/
.2kC1/
;
N[Table[s[n]-,fn,10g]] (* We note the speed of convergence
toward zero *)
DiscretePlot[s[n],fn,50g,
Ticks!fAutomatic,ff,””ggg,
PlotStyle!Black,ImageSize!250]
s[1] (* The sum of the series *)
{-0.0623912,0.0145888,-0.00373976,0.00101209,-0.000283868,
0.0000816591,-0.0000239376,7.1202217131194345J*^ -6,
-2.1426517125000544J*^ -6,6.509132886023394J*^ -7}
See Fig. 8.4.

8.1.7 Not So Simple Series

A good source of series connected to  is [71]. For some of them, we offer at

least a method of computation. In some cases, we use a second method based on an
appropriately chosen definite integral.
138 8 Pi Formulas

Fig. 8.5 Case 1

8.1.7.1 Case 1
p P  1 
case1[n ]:= 3 2 3 nkD0 .1/k 3kC1 1
C 3kC2 ;
Table[N[case1[k]],fk,6g] (* The first six terms of the sequence *)
DiscretePlot[case1[n],fn,50g,
Ticks!fAutomatic,f3,3.1,,3.2,3.3gg,
PlotStyle!Black,ImageSize!250]
case1[1] (* The sum of the series *)
{2.72798,3.42389,2.9279,3.31333,2.99812,3.26476}
See Fig. 8.5.
We can prove the previous result in the following way.
Clear[a,b]
P1  a 
k b
kD0 .1/ 3kC1
C 3kC2
FullSimplify[%]
p
1
p
9
3a C 3b C aLogŒ8  bLogŒ8
p
1
9
3.a C b/ C .a  b/LogŒ8

We obviously choose a D b and have

P1 k
 a a

kD0 .1/ 3kC1
C 3kC2
2a
p
3 3

The result is clear.

8.1.7.2 Case 2

We treat the next series in a similar way.

8.1 Various Simple and Not So Simple Formulas 139

Clear[a,b,c]
P1  a 
k b c
kD0 .1/ 4kC1
C 4kC2
C 4kC3
FullSimplify[%]

1
   2  2
16
a C c C 2bCot 8 C aCot 8 C cCot 8 C
p      p     
4 2aCot 8 Log Cos 8  4 2cCot 8 Log Cos 8 
p      p       
4 2aCot 8 Log Sin 8 C4 2cCot 8 Log Sin 8 Tan 8
p p p hp i
1
8
2a C b C 2c  C 2 2.a  c/ArcCoth 2

We consider that c ! a and solve the equation

h p p i
FullSimplify[Solve 18 2a C b C 2a ==1,b]
nn p oo
b ! 8  2 2a

We conclude that the series is now

hP  p i
1 k a 82 2a a
FullSimplify kD0 .1/ 4kC1
C 4kC2
C 4kC3

pThe last result shows that for any complex a one gets . Particularly, for a D
2 2, we have
h p P1  1 i
k 1
Simplify 2 2Simplify kD0 .1/ 4kC1
C 4kC3

We want to see the speed of convergence of this series.
p P  1 
case2[n ]:=2 2 nkD0 .1/k 4kC1 1
C 4kC3 ;
Table[N[case2[n]],fn,6g] (* The first six terms of the sequence *)
DiscretePlot[case2[n],fn,50g,
Ticks!fAutomatic,f3,3.1,,3.2gg,
PlotStyle!Black,ImageSize!250]
Simplify[case2[1]]
f2:80149; 3:37289; 2:96676; 3:282; 3:02434; 3:24223g
See Fig. 8.6.
A different approach is by means of a definite integral.
We introduce a different approach to this series based on a definite integral.
R 1 xi1
f[i ]:= 0 1Cx 4 dx;

as=Array[a,3];
product= p42 as.Array[f,3];
asArray=FullSimplify[product]
p p p
1
4
. 2aŒ2 C 2aŒ3 C 2aŒ1. C 2ArcCothŒ 2/ C 2aŒ3LogŒ3  2 2/
Because this intermediate result is complicated, we try to find relations between
the values of transcendental numbers so that the result is simplified. We have found
two such relations and we substitute them. Then
140 8 Pi Formulas

Fig. 8.6 Case 2

p p
simpler=Simplify[Together[asArray/.fArcCoth[
p p 2]!(1/2)Log[3+2 2],
Log[3-2 2]!-Log[3+2
p 2]g]] p
1
4
..2aŒ1 C 2aŒ2 C 2aŒ3/ C 2.aŒ1  aŒ3/LogŒ3 C 2 2/

GetTranscendentals[exp ]:=Union[Cases[exp,Pij Log,Infinity]]

transsimpler=GetTranscendentals[simpler]
n h p io
; Log 3 C 2 2

collected=Map[Factor,Collect[simpler,transsimpler]]
 h
1
p p i
4
 2aŒ1 C 2aŒ2 C 2aŒ3 C 12 .aŒ1  aŒ3/Log 3 C 2 2

We write the system of linear equations which follows.

syst=DeleteCases[Flatten[CoefficientList[collected,transsimpler]],0]
n o
aŒ1
2
 aŒ3
2
; aŒ1
2
C aŒ2
p C
aŒ3
2
2 2

The previous system is written under matrix form and we find its rank.
TableForm[syst,2]
(coeffMatrixx=Normal[CoefficientArrays[syst,as]][[2]])//MatrixForm
MatrixRank[coeffMatrixx]
aŒ1
2
 aŒ3
2
aŒ1
2
C aŒ2
p C
aŒ3
2 2 ! 2
1
2
0  12
1 p1 1
2 2 2 2
2
8.1 Various Simple and Not So Simple Formulas 141

Since the rank is 2, one coefficient becomes parameter and the system is solved
with respect to this parameter. Thus, we find out the next result.
sol=Solve[syst==f0,1g]
p p
ffaŒ2 ! 2 2  2 2aŒ1; aŒ3 ! aŒ1gg
We substitute and check the result.
P1  p p
p4 k a 2 22 2a a
2 kD0 .1/ 4kC1
C 4kC2
C 4kC3
Simplify[%]
p p
2.1C 2/
 p
2C 2

The last series converges to  for any complex number a.

8.1.7.3 Case 3

Now we study the series

P1 k
 a b c d

kD0 .1/ 5kC1
C 5kC2
C 5kC3
C 5kC4
;
FullSimplify[%]
1p
p p p p p
.2 2. 5  5(b+c) C 5 C 5.a C d//C
20 5   p
10.a C b  c  d/ArcCosh 32 C 5.a  b C c  d/LogŒ16/
We solve the system of equations
Solve[a+b-c-d==0&&a-b+c-d==0,fc,dg]
ffc ! b,d ! agg
and substitute
FullSimplify[%%%/.fc!b,
q p d!ag]
1 2
p p p
5 5
5 C 5 a C 5  5b 

If we consider fa; b; c; dg D f1; 0; 0; 1g, then

fa,b,c,dg=f1,0,0,1g;
P1 k
 a b c d

kD0 .1/ 5kC1
C 5kC2
C 5kC3
C 5kC4
;
FullSimplify[%]
q
1
5
2 C p2 
5

We note that for each complex a and b it is true that

Clear[a,b] P  a 
FullSimplify 1kD0 .1/
k
5kC1
b
C 5kC2 C b
5kC3
C a
5kC4
q p p p p
1 2
5 5
5 C 5a C 5  5b 
142 8 Pi Formulas

By a definite integral

We can try to approach this series by means of the corresponding definite integral.
We have
Clear[a]
R 1 xi1
f[i ]:= 0 1Cx 5 dx;

as=Array[a,4];
as.Array[f,4];
asArray=FullSimplify[%]
 p p p
1
p p
100
2 10 5  5.aŒ2 C aŒ3/ C 5 C 5.aŒ1 C aŒ4/ C
p h p i  p h i
5aŒ2ArcTanh 6765 15127
5
C 5aŒ1 2 5ArcCoth p3 C LogŒ16 
 p h i  p h 5i
5 2 5aŒ4ArcCoth p3 C aŒ3 2 5ArcCoth p3  LogŒ16 C
5 5
.aŒ2 C aŒ4/LogŒ16//

LogExpand[exp ]:=PowerExpand[exp]/.Log[n Integer]:!

Apply[#2.Log[#1]&,
Transpose[FactorInteger[n]]]
GetTranscendentals[exp ]:=Union[Cases[exp,Pij ArcCothj ArcTanhj
Log,Infinity]]
simpler=Together[ h p i h i
LogExpand[asArray]/.ArcTanh 676515127
5
! 10ArcCoth p35 ]
r  r r 
1
p  p p
50
10 5 C 5 aŒ1 C 10 5  5 aŒ2 C 10 5  5 aŒ3C
r  h i h i
p p p
10 5 C 5 aŒ4 C 5 5aŒ1ArcCoth p3 C 5 5aŒ2ArcCoth p3 
5 5
p h i p h i
3 3
5 5aŒ3ArcCoth p  5 5aŒ4ArcCoth p C
5 5
10aŒ1LogŒ2  10aŒ2LogŒ2 C 10aŒ3LogŒ2  10aŒ4LogŒ2/

transsimpler=GetTranscendentals[simpler]
n h i o
; ArcCoth p3 ; LogŒ2
5

collected=Map[Factor,Collect[simpler,transsimpler]]
p p p p
p p p p
p1  5 C 5aŒ1 C 5  5aŒ2 C 5  5aŒ3 C 5 C 5aŒ4 C
5 10 h i
p3
.aŒ1CaŒ2aŒ3aŒ4/ArcCoth
p 5
C 15 .aŒ1  aŒ2 C aŒ3  aŒ4/LogŒ2
2 5
8.1 Various Simple and Not So Simple Formulas 143

system=DeleteCases[Flatten[CoefficientList[collected,
Append[transsimpler,Sqrt[2]]]],0]
TableForm[system,4]
(coeffMatrix=Normal[CoefficientArrays[system,as]][[2]])//MatrixForm
MatrixRank[coeffMatrix]
 r 
aŒ1 aŒ2 aŒ3 aŒ4 aŒ1 aŒ2 aŒ3 aŒ4 1 1
p
5
 5
C 5
 ;
5 2 5
p C p  p  p ; 5 C 5 aŒ1 C
2 5 2 5 2 5 10 5
r  r  r  
1 1
p 1 1
p 1 1
p
10 5
5  5 aŒ2 C 10 5
5  5 aŒ3 C 10 5
5 C 5 aŒ4
aŒ1
5
 aŒ2
5
C aŒ3
5
 aŒ4
5
aŒ1 aŒ2 aŒ3 aŒ4
p C p  p  p
2 r 5 2 5 2 5 2 5 r  r 
 p p p
1 1 1 1 1 1
10 5
5 C 5 aŒ1 C 10 5
5  5 aŒ2 C 10 5
5  5 aŒ3C
r 
1 1
p
10 5
5 C 5 aŒ4
0 1 1
5
 15  15 1
5
B 1
p 1
p  p 1 C  p 1
B r 2 5 r 2 5 r  r 2 5 C 2 5
@  p p p p A

1 1 1 1 1 1 1 1
10 5
5 C 5 10 5 5  5 10 5 5  5 10 5 5 C 5
3
sol=FullSimplify[Solve[system==f0,0,1g]]

p p p
ffaŒ2 ! 12 .5 5 C 5  1 C 5 aŒ1/;
 p p  p
aŒ3 ! 12 5 5 C 5  1 C 5 aŒ1 ; aŒ4 ! aŒ1gg
Taking fa; b; c; dg D f1; 0; 0; 1g, we get the same result as before.

8.1.7.4 Case 4

Let us consider the next case.

Clear[a,b,c,d,e]
P1 k
 a b c d e

kD0 .1/ 6kC1
C 6kC2 C 6kC3
C 6kC4
C 6kC5
;
FullSimplify[%]
1
p p p p
36
..6a C 2 3b C 3c C 2 3d C 6e/ C 2.6 3.a  e/ArcCothΠ3 C
.b  d/LogŒ8//
Clearly there are two obvious options e D a and d D b. Then
FullSimplify[%/.fe!a,d!bg]

1
p
36
12a C 4 3b C 3c 
An almost easy option goes to b D c D 0 and a D 1, [71, (33)]. Then
P1 k
 1 1

kD0 .1/ 6kC1
C 6kC5

3
144 8 Pi Formulas

Fig. 8.7 Case 4

Another particular case is the next one, which is true for every complex c.
Clear[c]
P1  1 
k 0 c 0 1
kD0 .1/ 6kC1
C 6kC2
C 6kC3
C 6kC4
C 6kC5
1
12
.4 C c/
Another easy option is b D 0 and 4a C c D 12. Particularly, a D 1 and c D 8,
P1 k
 1 8 1

kD0 .1/ 6kC1
C 6kC3 C 6kC5

Regarding the convergence of this series, we remark the following:
P  1 
case4[n ]:= nkD0 .1/k 6kC1 8
C 6kC3 1
C 6kC5 ;
Table[N[case4[k]],fk,6g] (* The first six terms of the sequence *)
DiscretePlot[case4[n],fn,50g,
Ticks!fAutomatic,f3,3.1,,3.2,3.3gg,
PlotStyle!Black,ImageSize!250]
Simplify[case4[1]]
f2:74401; 3:41309; 2:93603; 3:30681; 3:00355; 3:2601g
See Fig. 8.7.

8.1.7.5 Case 5

Let us take a look at the series

Clear[a,b,c,d,e,f]
P1 k
 a b c d e f

kD0 .1/ 7kC1
C 7kC2 C 7kC3
C 7kC4
C 7kC5
C 7kC6
;
FullSimplify[%]
8.1 Various Simple and Not So Simple Formulas 145

1
  
2fCsc 7  bLogŒ4 C cLogŒ4  dLogŒ4 C eLogŒ4  fLogŒ4 C
28         
4dCos 7 Log Cos 3 Cot 7 C 4.b C e/Cos 7 
  14   
Log Root 8 C 20#1  12#12 C #13 &; 2  4fCos 7 
    
Log Root 8 C 20#1  12#12 C #13 &; 3 C 4cCos 7 
  3        
Log Sec 14 Tan 7 C 2cSec 14 C 2dSec 14 C 2bSec 3 C
 3    2 3
 14 
2eSec 14  4f Log Root 8 C 20#1  12#1 C #1 &; 2 Sin 14 C
   
4cLog Root 8 C 20#1  12#12 C #13 &; 3 Sin 14 
    

dLog Root 8 C 20#1  12#12 C #13 &; 3 Sin 14 C
   
4bLog Root 1 C 12#1  20#12 C 8#13 &; 3 Sin 14 
  2 3
 
4eLog Root 1 C 12#1  20#1 C 8#1 &; 3 Sin 14 C
 h   2 i h  2 i h   2 i
4 bLog 12 Csc 14 C dLog 12 Csc 3 14
C eLog 2Sin 14
C
h  3 2 i   3      3 
c Log 2Sin 14 C fLog Sec 14 Tan 7 Sin 14 C
     h   8 i
a 2Csc 7 C LogŒ4  Cos 7 Log 16Sin 14 C
  2 3
 
4Log Root 8 C 20#1  12#1 C #1 &; 2 Sin 14 C
      
4Log Cos 3 14
Cot 7 Sin 3 14
We note that the result is rather complicated and it is hard to believe that it can
be easily manipulated. Therefore, we pass to a particular form of it.
P  a 
7 1 kD0 .1/
k
7kC1
b
C 7kC2 c
C 7kC3 c
C 7kC4 b
C 7kC5 a
C 7kC6 ;
FullSimplify[%]
     
 a Csc 7 C c Sec 14 C b Sec 314
       
FullSimplify %=:Csc 7 ! Sec 14 C Sec 314
  
.a C c/Sec 14 C .a C b/Sec 3 14
So the result
   
 .a C c/Sec 14 C .a C b/Sec 3
14
depends on the (complex) coefficients a, b, and c.

8.1.7.6 Case 6

Let us consider the series

Clear[a,b,c,d]
P1  a 
k b c d c b a
kD0 .1/ 8kC1
C 8kC2 C 8kC3 C 8kC4 C 8kC5 C 8kC6
C 8kC7
;
sum8=FullSimplify[%]
 p
1
p  
16
 2 2b C d C 2.a C .1 C 2/c/Csc 8
h   p p p i
FullSimplify sum8=:Csc 8 ! 2 2 C 2
r  
1
p p   p
16
2 2b C 2 2 2 C 2 a C 1 C 2 c C d 
146 8 Pi Formulas

Then for a D b D c D 0, and d D 1, we get

sum8/.fa!0,b!0,c!0,d!1g

16
This is precisely the Gregory and Leibniz formula.
If we consider the above series under its general form, we have
P1 
k a b c d e f g
kD0 .1/ 8kC1
C 8kC2 C 8kC3 C 8kC4 C 8kC5 C 8kC6 C 8kC7 ;
   p p
 
FullSimplify Together %=: Cos 8 ! 2C 2
2
;
  p p   p p   p p
Sin 8 ! 2 2
2
; Tan 8
! p 2 2
p , Cot 8
! p2Cp 2 ,
2C 2
r 2
2
  p p p  p p
Csc 8 ! 2 2 C 2, Tan 16 ! 1  2 C 2 2 C 2 ,
r  r 
 p p   p p
Cot 16 ! 1 C 2 C 2 2 C 2 , Tan 3 16
! 1  2 C 2 2 2 ,
r   
 3  p p
Cot 16 ! 1 C 2 C 2 2  2 ;

The result is rather complicated and is out of our goal.

By a definite integral

Because the last result is rather long, we approach this series by means of a definite
integral.
Clear[a]
R 1 xi1
f[i ]:= 0 1Cx 8 dx;

as=Array[a,7];
asArray=FullSimplify[as.Array[f,7]]
 p p
1
p p
16
 2aŒ2 C aŒ4 C 4  2 2.aŒ3 C aŒ5/ C 2aŒ6 C
q p
2.2 C 2/.aŒ1 C aŒ7/ 
p    
2aŒ6ArcCoshŒ3 C 4 .aŒ1  aŒ7/ArcTanh Cos 8 C
     
.aŒ3 C aŒ5/ArcTanh Sin 8 Cos 8 
p h p i    
2aŒ2Log 3  2 2 C 4 .aŒ3  aŒ5/ArcTanh Cos 8 C
    
.aŒ1  aŒ7/ArcTanh Sin 8 Sin 8
8.1 Various Simple and Not So Simple Formulas 147

LogExpand[exp ]:=PowerExpand[exp]/.Log[n Integer]:!

Apply[#2.Log[#1]&,Transpose[FactorInteger[n]]]
GetTranscendentals[exp ]:=
Union[Cases[exp,Pi j ArcTanh j ArcCosh j Log j Cos j Sin, Infinity]]
simpler=h n h p i
Together LogExpandŒasArray=: ArcCoshŒ3 ! Log 3 C 2 2 ;
h p i h p ioi
Log 3  2 2 ! Log 3 C 2 2 I

transsimpler=GetTranscendentals[simpler]
n h
        p i
; ArcTanh Cos 8 ; ArcTanh Sin 8 ; Cos 8 ; Log 3 C 2 2 ;
 

Sin 8

collected=Map[Factor,Collect[simpler,transsimpler]];
system=DeleteCases[Flatten[CoefficientList[collected,
Append[transsimpler,Sqrt[2]]]],0]
TableForm[system,7];
(coeffMatrix=Normal[CoefficientArrays[system,as]][[2]])//MatrixForm
MatrixRank[coeffMatrix]
0 1
1 1
0 16
0 0 0  16 0
B 1
0 0 0 0 0  14 C
B 4 C
B 1 1 C
B 0 0 4 0 0 0 C
B 1
4 C
B 0 0 0  14 0 0 C
B 4 C
B 1
0 0 0 0 0  14 C
B 4 C
B 0 1 C
@ p p 0 p 0 p 16 p 0 p 0 p 0 p A
2C 2 1 2 2 2 2 1 2C 2
16 16 16
0 16 16 16
5
We consider a particular case.
sol=Solve[system==f0,0,0,0,0,1,0g]
p p
p p
ffaŒ2 !  2 C 2 p aŒ1  2  2 aŒ3;
p aŒ4 ! 16;
p p
aŒ5 ! aŒ3; aŒ6 !  2 C 2 aŒ1  2  2 aŒ3; aŒ7 ! aŒ1gg
If aŒ1 D aŒ3 D 0, then
sol/.fa[1]!0,a[3]!0g
ffaŒ2 ! 0; aŒ4 ! 16; aŒ5 ! 0; aŒ6 ! 0; aŒ7 ! 0gg
i.e.,
P .1/k
16 1 kD0 8kC4

The last series is the Gregory and Leibniz series.
148 8 Pi Formulas

8.1.7.7 Case 7

Now we consider the following series:

Clear[a,b,c,d,e,f,g,h]
P1  a 
k b c d e f g h
kD0 .1/ 9kC1
C 9kC2 C 9kC3 C 9kC4 C 9kC5 C 9kC6 C 9kC7
C 9kC8
;
FullSimplify[%];
P1  a 
k b c d d c b a
kD0 .1/ 9kC1
C 9kC2 C 9kC3 C 9kC4 C 9kC5 C 9kC6 C 9kC7
C 9kC8
;
FullSimplify[%]
 p
1
    2    
27
 2 3c C 3 a Csc 9
C b Csc 9
C d Sec 18

An easy choice for the second series is a D b D d D 0. Then one has

P1 k
 1 1

kD0 .1/ 9kC3
C 9kC6
2
p
9 3

The last series coincides up to a factor with the series in Case 1.

8.1.7.8 Case 8

By [71, (34)] one has

P1 k
 1 1 1 1 1

kD0 .1/ 10kC1
C 10kC3
C 10kC5
C 10kC7
C 10kC9

4
We want to take a closer look at the series
Clear[a,b,c,d,e]
P1  a
k b c d e d
kD0 .1/ 10kC1
C 10kC2

C 10kC3 C 10kC4
C 10kC5
C 10kC6
C
c b a
10kC7
C 10kC8
C 10kC9
I

FullSimplify[%]
 r 
1
p p  p
100
10 1 C 5 a C 2 10 5 C 5 b C 10 1 C 5 cC
p p
2 50  10 5d C 5e 

We may choose the following particular case

FullSimplify[%/.fa!1,b!0,c!-1,d!0,e!0g]

5
Regarding the convergence of the initial series, we note that
P  1 
case10[n ]:= 1 kD0 .1/
k
10kC1
1
C 10kC3 1
C 10kC5 1
C 10kC7 C 1
10kC9
Table[N[case10[k]],fk,6g] (* The first six terms *)
DiscretePlot[case10[n],fn,50g,
Ticks!fAutomatic,f.78,f/4,”/4”g,.79gg,
PlotStyle!Black,ImageSize!250]
8.1 Various Simple and Not So Simple Formulas 149

Fig. 8.8 Case 8

Simplify[case10[1]] (* The sum of the series *)

f0:76046; 0:802046; 0:772906; 0:795394; 0:777067; 0:79254g
See Fig. 8.8.

8.1.7.9 Case 9

Let us consider the series

Clear[a,b,c,d,e]
P1  a
k b c d e e
kD0 .1/ 11kC1
C 11kC2
C 11kC3
C 11kC4
C 11kC5
C 11kC6
C
d c b a

11kC7
C 11kC8 C 11kC9
C 11kC10
I
sum11=FullSimplify[%]
1
        
11
 a Csc 11 C b Csc 2
11
C e Sec 22 C d Sec 3
22
C c Sec 5
22
Now we choose a particular case
FullSimplifyŒ%=:fb
 ! 0; c ! 0; d ! 0; e ! 0g
1
11
aCsc 11
It means that
P1 k
 1 1

kD0 .1/
  11kC1
C 11kC10
(* Equals *)
1 
11
Csc 11
Also, one has
FullSimplify[sum11/.fa!1,b!0,c!1,d!0,e!-1,f!0g]

p
2 2
150 8 Pi Formulas

8.1.7.10 Case 10

Because the amount of computation is huge for the next series, we consider a
simplified version of it, namely,
Clear[a,b,c,d,e,f]

P1 k a b c d e f e
kD0 .1/ 12kC1
C 12kC2
C 12kC3
C 12kC4
C 12kC5
C 12kC6
C 12kC7
C
d c b a

12kC8
C 12kC9
C 12kC10
C 12kC11 ;
FullSimplify[%]
1
p p p p p p
72
.6. 2 C 6/a C 12b C 6 2c C 4 3d C 6 2.1 C 3/e C 3f/
We have a simple series
p
FullSimplify[%/.fa!1,e!1,b!-1,c!-1,d! 3=2,f!4g]

p
6

By definite integral

The general case is discussed using a definite integral.

R 1 xi1
f[i ]:= 0 1Cx 12 dx;

as=Array[a,11];
as.Array[f,11];
asArray=FullSimplify[TrigToExp[%]];
LogExpand[exp ]:=PowerExpand[exp]/.Log[n Integer]:!
Apply[#2.Log[#1]&,Transpose[FactorInteger[n]]]
GetTranscendentals[exp ]:=
Union[Cases[exp,Pi j ArcCosh j Log j ArcTanh,Infinity]]
simpler=Together[LogExpand[asArray]/.fArcCosh[19601]!
p p
ArcCosh[17]-Log[577-408
p p 2],ArcCosh[577]!-Log[577-408 2],p
Log[17-12 2]!Log[3-2 2]+Log[2], h pArcCosh[49]!-Log[49-20
i 6],
p 2 6
ArcCosh[3]!-Log[3-2 2],ArcTanh 5 !ArcCosh[5]g];

transsimpler=GetTranscendentals[simpler];
collected=Map[Factor,Collect[simpler,transsimpler]];
system=DeleteCases[Flatten[CoefficientList[collected,
Append[transsimpler,Sqrt[2]]]],0]
TableForm[system,11]
(coeffMatrix=Normal[CoefficientArrays[system,as]][[2]])//MatrixForm
MatrixRank[coeffMatrix]
system/.fa[1]!1,a[2]!0,a[3]!1,a[4]!0,a[5]!-1,a[6]!0,
a[7]!-1,a[8]!0,a[9]!1,a[10]!0,a[11]!1g
8.1 Various Simple and Not So Simple Formulas 151

Solve
  p p 
system DD 0;  16 3 ; 0;  72 ; 0; 0;  36 ; 0;  8 3 ; 0; 2C
1p 1 1 1
p
6
3 1
;6 1
p
4 3

ffaŒ1 ! 1; aŒ2 ! 0; aŒ3 ! 1; aŒ5 ! 1; aŒ6 !  4aŒ4

p ; aŒ7
3
! 1;
aŒ8 ! aŒ4; aŒ9 ! 1; aŒ10 ! 0; aŒ11 ! 1gg
We check the result
asa=f1,0,1,0,-1,0,-1,0,1,0,1g;
asa.Array[f,11];
2Sqrt[2]FullSimplify[TrigToExp[%]]


8.1.7.11 Case 11

We will study the next series

Clear[a,b,c,d,e,f]
P1  a
k b c d e f f
kD0 .1/ 13kC1
C 13kC2 C 13kC3 C 13kC4

C 13kC5
C 13kC6
C 13kC7
C
e d c b a
13kC8
C 13kC9
C 13kC10
C 13kC11
C 13kC12
;
FullSimplify[%]
1
        
 a Csc 13 C b Csc 2 C c Csc 3 C f Sec 26 C e Sec 3 C
13  5  13 13 26
d Sec 26
One option now is to consider one coefficient nonzero and all the others zero. For
example
P  1 
FullSimplify 1 kD0 .1/
k
13kC1
1
C 13kC12
2.1/15=26 
13.1C.1/2=13 /

The last number is equal to

1

13
Csc 13

8.1.7.12 Case 12

There exists a nice series in this case, namely

P  3
FullSimplify 1 kD0 .1/k 14kC1 3
C 14kC3 C 3
C 4
C 3
C
3 3
 14kC5 14kC7 14kC9
14kC11
C 14kC13

152 8 Pi Formulas

We consider now the series

P1 
k a b c d e f
kD0 .1/ 14kC1
C 14kC2 C 14kC3
C 14kC4
C 14kC5
C 14kC6
C

g f e d c b a
14kC7
C 14kC8
C 14kC9
C 14kC10
C 14kC11
C 14kC12
C 14kC13
;

FullSimplify[%]
1
         
 g C 2 a Csc 14 C b Csc 7 C c Csc 3 C f Sec 14 C e Sec 7
28  3  14
Cd Sec 14
We simplify the last sum
FullSimplify[%/.fb!0,d!0,f!0g]
      
1
28
 g C 2 a Csc 14 C c Csc 3
14
C e Sec 7
Because of
1       
FullSimplify 28 4 C 2 3Csc 14 3Csc 3
14
C 3Sec 7
1
we get the first series.
Remark In [71, (35)] the following series is exhibited
P1 k
 3 3 3 4 4 4 4

kD0 .1/ 14kC1
C 14kC3 C 14kC5 C 14kC7 C 14kC9
C 14kC11
C 14kC13
;
and it is stated that it converges toward . This is false since
N[%]
3:74057

8.1.7.13 Case 13

Clear[a,b,c,d,e,f,g,h]
P1  a
k b c d e f g
kD0 .1/ 16kC1
C 16kC2 C 16kC3 C 16kC4 C 16kC5 C 16kC6 C 16kC7 C

h g f e d c b a
16kC8
C 16kC9
C 16kC10
C 16kC11
C 16kC12
C 16kC13
C 16kC14
C 16kC15
;
FullSimplify[%]
 p      
1
32
 h C 2 2d C a Csc 16 C b Csc 8 C c Csc 3
16
C g Sec 16 C
 3  p   
e Sec 16 C 2 2f Sin 8

The simplest case appears whenever h D 1 and a D b D c D d D e D f D

g D 0. Then we have
P .1/k
32 1 kD0 16kC8

Thus, we get the Gregory and Leibniz formula.
Another particular case consists in d D 1 and a D b D c D e D f D g D h D 0.
In this case we get a series already discussed at Case 2.
8.1 Various Simple and Not So Simple Formulas 153

Suppose that g D 1 and all other coefficients are null. Then we have the following
series
P  1 
FullSimplify 1 kD0.1/
k 1
C 16kC9
1
   16kC7
8
Csc 8 Sin 16


The last result coincides with

p pq p p
1 2C 2 2 2C 2
8
p
2


8.1.7.14 Case 14

We have the result

P1 k
 a b c d d c
kD0 .1/ 
18kC1
C 18kC3
C 18kC5
C 18kC7
C 18kC11
C 18kC13
C
b a
18kC15
C 18kC17 ;
FullSimplify[%]
1
     
18
 2b C a Csc 18 C d Sec 9 C c Sec 2
9
Since
     
FullSimplify Csc 18  Sec 9 C Sec 2 9
6
we conclude that
P1 k
 2 3 2 2 2 2 3
kD0 .1/
 18kC1
C 18kC3 C 18kC5 C 18kC7 C 18kC11
C 18kC13
C 18kC15
C
2
18kC17
;
FullSimplify[%]

The convergence of the previous series is suggested by the following.
P  2
case14[n ]:= nkD0 .1/k 18kC1 3
C 18kC3 2
C 18kC5 2
C 18kC7 2
C 18kC11 C
2 3 2

18kC13
C 18kC15 C 18kC17 I
Table[N[case14[n]],fn,6g (* The first six terms of the sequence *)
DiscretePlot[case14[n],fn,50g,
Ticks!fAutomatic,f3.12,,3.163gg,
PlotStyle!Black,ImageSize!250]
FullSimplify[case14[1]]
f3:00528; 3:23335; 3:07251; 3:19696; 3:09541; 3:1812g
See Fig. 8.9.

8.1.7.15 Case 15

We have the known result

154 8 Pi Formulas

Fig. 8.9 Case 14

P1  3
kD0 .1/
k 3
C 22kC3 3
C 22kC5 C 3 C 3
C 8
C 3
C
3 3
22kC1
3 3
 22kC7 22kC9 22kC11 22kC13
22kC15
C 22kC19 C 22kC19 C 22kC21 I
FullSimplify[%]

Remark In [71, (37)] is exhibited the following series
P1  3 3
kD0 .1/
k
C 22kC3 3
C 22kC5 C 3 C 3
C 8
C 3
C
3 3
22kC1
3 1
 22kC7 22kC9 22kC11 22kC13
22kC15
C 22kC19 C 22kC19 C 22kC21 I
and it is stated that it converges toward . This is false since
N[%]
3:075

8.1.7.16 A Case

P .1/kC1
3+4 1kD1 2k.2kC1/.2kC2/


8.1.7.17 A Case

P
6 1 1
kD1 k2
2
This is exactly the  function of Riemann evaluated at 2, i.e., .2/.
8.5  by the Golden Ratio 155

8.1.7.18 A Case
P
8 1kD1
1
.2k1/2
2

8.2 Newton’s Geometric Construction

In [71, (18)] the next definite integral is presented

R 1=4 p
3 Sqrt[3]/4+24 0 x.1  x/ dx
Simplify[%]
p  p
3 3
4
C 14 3 3 C 4


8.3 Euler Series

From [71, (23] we see that

nP o
1 2k k.kŠ/2 P1 kŠ
2 kD0 .2kC1/Š ; kD0 .2kC1/ŠŠ
f; g

8.4  by Arcsin

From [71, (26] we see that

h i
Series ArcSinŒx
p ; fx; 0; 10g
1x2
P1 .2x/2kC1 kŠ2
kD0 2.2kC1/Š
%/.x!1/2
3 5 16x7 128x9
x C 2x3 C 8x
15
C 35
C 315
C OŒx11
ArcSinŒx
p
1x2

p
3 3

8.5  by the Golden Ratio

f[x ]:= ArcSinŒx

p ;
1x2
f[1/2]
156 8 Pi Formulas

FullSimplify[f[Sin[Pi/10]]]
Pi
%== 5pGoldenRatioC2 (* This is [71, (28)] *)
p
3q 3
1 2p
5

5C 5
True
p
GoldenRatio== 1C2 5
FullSimplify[TrigToExp[FunctionExpand[
h i
1 .kŠ/2
2
Sum 2kC1 ; fk; 0; Infinityg ]]]
h GoldenRatio i .2kC1/Š
Pi
N 5pGoldenRatioC2 ==N[%]
True  
1
5
Root 1  5#12 C 5#14 &; 3
True
 r  
p
FullSimplify 15 10
1
5 5  p 1
5 GoldenRatioC2
0

=GoldenRatio;  
p P1 .kŠ/2
==FullSimplify 5
C2 kD0
2
2kC1 .2kC1/Š
True
hP i
1 .kŠ/2
FullSimplify kD0
2kC1 .2kC1/Š
h P i
==FullSimplify
1 1 1
kD0
.2kC1/BinomialŒ2k;k
2k
h P R1 k i
==FullSimplify
1
1 2kC1 k
kD0
2k .2kC1/ 0 x .1  x/ dx
  
R 1 P1  x 1x k
==FullSimplify
1 0 kD0


dx
h R i
1 1
==FullSimplify
0
2x.1x/ dx
q q q
1
5
2  p2  DD 15 2  p2  DD 15 2  p2 
5 5 5
q
1 2 2GoldenRatio 
DD 5 2  p  DD p p
5 5 5C2 5

The last equality is true because of

 p

FullSimplify p 2 p 1C2 5
5 5C2 5
q
1 2
5
2 p 
5

We also have
8.6  by Integrals 157

Fig. 8.10  by the golden

ratio

h p R1 i
FullSimplify 5
2
C2 0
2 x.1x/
1
dx

The convergence of the first series is suggested by the following code.

=GoldenRatio; p Pn .kŠ/2
pibygoldenratio[n ]:= 5
C2
2 kD0
2kC1 .2kC1/Š
Table[N[pibygoldenratio[k],10],fk,6g] (* The first six terms *)
DiscretePlot[pibygoldenratio[n],fn,50g,
Ticks!fAutomatic,fgg,
PlotStyle!Black,ImageSize!250]
FullSimplify[pibygoldenratio[1]]
f3:126021252; 3:140314037; 3:141483900; 3:141583199; 3:141591819;
3:141592579g
See Fig. 8.10.

8.6  by Integrals

We recall some integrals connected to the  number.

8.6.1 Dantzell Formula

22
R1 x4 .1x/4
7
 0 dx (* [71, (50)] *)
1Cx2

Hence it immediately follows that  < 22=7, [41].
158 8 Pi Formulas

8.6.2 Lucas Formula

355 1
R 1 x8 .1x/8 .25C816x2 /
 3164
113 0 1Cx2
dx; (* [71, (51)] *)
Simplify[%]

Hence it immediately follows that  < 355=113, [41].

8.6.3 Backhouse-Lucas Formula

In [5, 42], and [71] the following integral is discussed.

R 1 m .1x/n
i[m Integers,n Integers]:= 0 x 1Cx 2 dx
i[m,n]
FullSimplify[Assuming[m>0,i[m,2m]]]
i[m,n]
i[m,2 m]
On the same line we mention four more results from [42].
nR  10
1 x .1x/8 5
0 2
4.1Cx /
C 138450312
dx== 355
113
 ;

R 1 8192 9 8

0 114291
 xp.1x/2 C 219772564011
15409
dx== 355
113
 ;
1x
R 1 x14 .1x/12 .124360C77159x2 /
755216.1Cx2 /
dx==  103993 ;
0 
33102
R 1 x12 .1x/12 .13491060x2 /
0 38544.1Cx2 /
dx== 104348
33215

fTrue; True; True; Trueg

8.7 BBP and Adamchik-Wagon Formulas

This section is mainly based on [2, 3, 6], and [69]. A BBP formula allows to
compute the n-th decimal of  without knowing its previous decimals. The acronym
BBP comes from the names of three mathematicians called (David) Bailey, (Peter)
Borwein, and (Simon) Plouffe.
An interesting result of this kind is proven in [2].
We underline that the next result, called an Adamchik-Wagon formula (AW
formula), is true for any complex number r.
FullSimplifyŒPowerExpand
P1 1  4C8r Π
8r 4r 2C8r 1C2r 1C2r r
kD0 16k 8kC1  8kC2  8kC3  8kC4
 8kC5
 8kC6
C 8kC7

8.8 A Method for Finding AW Formulas and Proofs 159

We can approach the previous identity by means of the next two commands
f[i ]:=2i=2 IntegrateŒSumŒz^ .8k C i  1/; fk; 0; Infinityg; fz; 0; 1=SqrtŒ2g
FullSimplify[PowerExpand[
(4+8r)f[1]-8r f[2]-4r f[3]-(2+8r)f[4]-(1+2r)f[5]-(1+2r)f[6]+r f[7]]]

The last command can be written under the form
FullSimplify[PowerExpand[
f4+8r,-8r,-4r,-(2+8r),-(1+2r),-(1+2r),rg.Array[f,7]]]

Setting r D 0 yields the original BBP formula, [6]
FullSimplify[PowerExpand[4f[1]-2f[4]-f[5]-f[6]]]

Setting r D 1=2 yields the BBP formula [71, (39)]
FullSimplify[PowerExpand[f0,4,2,2,0,0,-1/2g.Array[f,7]]]


8.8 A Method for Finding AW Formulas and Proofs

In this section we refer to [2] and [3].

as=Array[a,7];
asArray=FullSimplify[TrigToExp[as.Array[f,7]]];
LogExpand[exp ]:=PowerExpand[exp]/.Log[n Integer]:!
Apply[#2.Log[#1]&,Transpose[FactorInteger[n]]]
GetTranscendentals[exp ]:=Union[Cases[exp,Pi j ArcTan j Log j
ArcCoth,Infinity]]
simpler=Together[LogExpand[asArray]/.f
ArcCot[Sqrt[2]]!Pi/2-ArcTan[Sqrt[2]],ArcCoth[Sqrt[2]]!
Log[Sqrt[2]+1],ArcCoth[4]!Log[5]-Log[3]g];
transsimpler=GetTranscendentals[simpler]
n hp i h p io
; ArcTanŒ2; ArcTan 2 ; LogŒ3; LogŒ5; Log 1 C 2

collected=Map[Factor,Collect[simpler,transsimpler]];
system=DeleteCases[Flatten[CoefficientList[collected,
Append[transsimpler,Sqrt[2]]]],0];
160 8 Pi Formulas

TableForm[system,7];
(coeffMatrix=Normal[CoefficientArrays[system,as]][[2]])//MatrixForm
MatrixRank[coeffMatrix]
0 1 1
4
0 12 0 1 0 2
B 1 0 1 1 1 0 1 C
B 8 C
B 0 1 04 1 02 1 0 C
B C
B 1 4 1 C
B  4 0 2 0 1 0 2 C
B 1 C
B 4  12 21 0 1 2 2 C
B C
@ 0 1 0 0 0 1 0 A
4
1
8
0  14 0 12 0 1
6

sol=Solve[system==f0,0,0,0,0,1,0g]
nn
aŒ2 ! 4  aŒ1; aŒ3 ! 2  aŒ1 2
; aŒ4 ! 1  aŒ1
2
; aŒ5 !  aŒ1
4
;
oo
aŒ6 !  aŒ1
4
; aŒ7 !  12 C aŒ1
8

Let aŒ1 be a free parameter denoted 8r C 4. Then

a[1]=8r+4;
solution=Simplify[as/.First[sol]]
f4 C 8r; 8r; 4r; 1  4r; 1  2r; 1  2r; rg
This yields the AW formula for  at the beginning of the Sect. 8.7.
If we set r to 0, pop out the BBP coefficients
solution/.r!0
f4; 0; 0; 1; 1; 1; 0g
and thus the BBP formula is proven.

8.9 BBP-Type Formulas for  in Powers of 2k

We list the formulas (38)–(43) in [71].

˚P1 1  4 2 1 1

FullSimplify
P  kD0 16k 8kC1  8kC4  8kC5 
 8kC6 ;
1 1 1 8 4 4 1
C C  ;
2 kD0 16k 8kC2 1 P 8kC3
1 1
 64 8kC7 32
8kC4
16 16 4
FunctionExpand 16 kD0 
256k 16kC1
 16kC4  16kC5  16kC6 C 16kC9 
2 1 1
 16kC13
 ;
16kC12
1 P1
16kC14
1
 128 64 64 16 8
FunctionExpand 32 kD0 
256k 16kC2
C 16kC3 C 16kC4  16kC7 C 16kC10 C
4 4 1
C 16kC12  16kC15 ;
1 P1
16kC11
1
 256 192 256 96 96 16
32 kD0 4096k 24kC2 C 24kC3  24kC4  24kC6 
 24kC8 C 24kC10 
4 3 6 2 1
24kC12
 24kC15  24kC16  24kC18  24kC20 ;
8.9 BBP-Type Formulas for  in Powers of 2k 161

Fig. 8.11 BBP-type

formulas for  in powers of
2k

  1 P1 1
 256 256 384
TrigToExp FunctionExpand 64 kD0 4096k 24kC1 C 24kC2  24kC3 
256 64 96 64 16 8 4 6
 24kC5 C 24kC8 C 24kC9 C 24kC10 C 24kC12  24kC13 C 24kC15 C
24kC4
6 1 1 1 1

24kC16
C 24kC17
C 24kC18
 24kC20
 24kC21
g]
f; ; ; ; ; g
The last series converges rapidly. This fact is suggested by the following results.
1 Pn 1
 256 256 384 256 64
bbptypeŒn_WD 64 kD0 4096k 24kC1 C 24kC2  24kC3  24kC4  24kC5 C
96 64 16 8 4 6 6 1
C 24kC9 C 24kC10 C 24kC12  24kC13 C 24kC15 C 24kC16 C 24kC17 C
24kC8
1 1 1

24kC18
 24kC20
 24kC21

Table[N[bbptype[n],25],fn,6g] (* The first six terms of the sequence *)

DiscretePlot[bbptype[n],fn,1,50g,Ticks!fAutomatic,f3.1,gg,
PlotStyle!Black,ImageSize!250,AxesOrigin!f0,3g]
f3:141592653538106427571660; 3:141592653589789141720046;
3:141592653589793238020210; 3:141592653589793238462586;
3:141592653589793238462643; 3:141592653589793238462643g
See Fig. 8.11.

8.9.1 A Case

Another series of the same kind is formula (48) in [71] and we introduce it bellow.
1 P1 1
 256 64 128 352 64 288
sum24 D 96 kD0 4096k 24kC2 C 24kC3 C 24kC5 C 24kC6 C 24kC7 C 24kC8 C
128 80 20 16 1 6 2 1
C 24kC10 C 24kC12  24kC14  24kC15 C 24kC16  24kC17  24kC19 C
24kC9
1 2

24kC20
 24kC21
;
FullSimplify[FunctionExpand[%]];
162 8 Pi Formulas

Because of
FullSimplify[20ArcCot[4-Sqrt[3]]+20ArcCot[4+Sqrt[3]]]
5 7
FullSimplify[-16 Sqrt[3]ArcCot[ SqrtŒ3 ]-4 Sqrt[3]ArcCot[ SqrtŒ3 ]
-4 Sqrt[3]ArcCot[3
  Sqrt[3]]+12 Sqrt[3]ArcCot[11 Sqrt[3]]];
20ArcTan 23
p  h p i
 2 3  C ArcTan 153373

one has
  13651680  h p i
1
18
21  4SqrtŒ3  ArcTan 815616479  4SqrtŒ3ArcTan 153373 C
p  h 
1
p i
16.1/1=12 2 .1/5=6 ArcTan 11 43 3 C
h  p i   h  p i
1 1
ArcTan 11 4 C 3 3 C i C .1/1=3 ArcTan 11 4C3 3 ;

sum24a=ComplexExpand[%] h p i
7 2 1
 13651680  2ArcTan 15337 3
6
 p  18 ArcTan 815616479  p C
3 3 3 3 p
h  p i 1
4 1 4ArcTan Π11p .43 3 /
9
ArcTan 11 4  3 3  
3 3p
h  p i 1
4 1 4ArcTan Π11 . 43 3 /
9
ArcTan 11 43 3  p C
3 3
h  p i 1
p
4 1 4ArcTan Π11p . 4C3 3/
9
ArcTan 11 4 C 3 3 C 
3 3p
h  p i 1
4 1 4ArcTan Π11 . 4C3 3 /
9
ArcTan 11 4C3 3 C p C
3 3
   h p 2i
2 1
p 2 1
2Log 1C 121 .43 3/
i  9 Log 1 C 121 4  3 3 C p C
3 3
   h p 2i
2 1
p 2 1
2Log 1C 121 .43 3/
9
Log 1 C 121 4  3 3 C p 
3 3
   h p i
2 1
p 2 1
2Log 1C 121 .4C3 3/
2

9
Log 1 C 121 4 C 3 3  p C
3 3
  h p i !
 p 2 1
2Log 1C 121 .4C3 3/
2
2 1
9
Log 1 C 121
4 C 3 3  p
3 3

We now show that the above number is real and equals : Firstly we show that
its imaginary part is zero.
FullSimplify[Im[sum24a]]
0
8.10  Formulas by Binomial Sums 163

We pick up the real part of sum24a.

FullSimplify[TrigExpand[Re[sum24a]]]

1
p  13651680  p h p i
15 3
18
21  4 3   ArcTan  4 3ArcTan C
 p h  815616479p i 337
1
16 1 C 3 ArcTan 11 4 C 3 3 C
 p h  p i
1
1 C 3 ArcTan 11 4C3 3

And now the last step of calculations shows that

N[%-Pi]
0:
We conclude that sum24 D .

8.10  Formulas by Binomial Sums

We introduce three formulas by binomial sums, called in some references BBP like
binomial sums. For this kind of sums, a great help is offered by the identity
Z 1
1
mn D .m  n C 1/ xpn .1  x/.mp/n dx; (8.1)
pn 0

[53, p. 277].
Below we introduce three series of this sort. The first series belongs to W. Gosper,
whereas the second to S. Plouffe. The third one belongs to Gourevitch and Guillera,
[28]. The first two series follow.
P1 50k  6
D kD0
BinomialŒ3k; k2k
P k2k .kŠ/2 P k2k
3C D 1kD1 D 1kD1
.2k/Š BinomialŒ2k; k
We transform the right-hand sides substituting the binomial coefficients with the
definite integrals given above by (8.1).
h hnR P
1 1 .50k6/.3kC1/ k
FullSimplify FunctionExpand 0 kD0 2k
x .1  x/2k dx;
R 1 P1  oii
k k k
0 kD1 k2 .2k C 1/x .1  x/ dx
f; 3 C g
164 8 Pi Formulas

The Gourevitch-Guillera series follows.

1 P1 1  59296 
16807 kD0 7kC1
 10326
7kC2
 3200
7kC3
 1352
7kC4
792
 7kC5 C 552
7kC6
;
2k BinomialŒ7k; 2k
Chop@N[%-]
0

8.11 S. Ramanujan Series

We are referring to an identity of the form given below, [9, 10], and [7].
p
8 P1 .4k/Š 1103 C 26390k

9801 kD0 .kŠ/4 3964k
1

Regarding the convergence of this series we note the following.
p P
8 n .4k/Š 1103C26390k
ramanujan1[n ]:= 9801 kD0 .kŠ/4  3964k
fTable[N[1/ramanujan1[n],30],fn,4g], (* The first four terms *)
Table[N[ -1/ramanujan1[n],30],fn,4g]g (* It clearly appears that at
each step twelve exact decimals are added *)
DiscretePlot[1/ramanujan1[n],fn,50g,Ticks!fAutomatic,f3.1,gg,
PlotStyle!Black,ImageSize!250,AxesOrigin!f0,3g]
ramanujan1[1]
ff3:14159265358979387799890582631;
3:1415926535897932384626490657;
3:14159265358979323846264338328;
3:14159265358979323846264338328g;
f6.3953626244302651021001947563J30.*^ -16;
 5.68242325601395950808108722896J30.*^ -24;
 5.23889628048110452740902417593J30.*^ -32;
 4.94418757924803001220998690737J30.*^ -40gg
See Fig. 8.12.
Another example on the same line is the following, [30, (1)],
1
1 1 X 6k C 1  .1=2 C k/. .1=2//3
D :
 4 kD0 4k kŠ3

h h P ii
.GammaŒ1=2Ck=GammaŒ1=2/3
FullSimplify FunctionExpand 14 1
kD0
6kC1
4k
 kŠ3
p 2 2 2
CEllipticKΠ14 .2 3/ GammaΠ16  GammaΠ13 
2
 p
16 3 3
8.11 S. Ramanujan Series 165

The last result is written as

p 2 2 2

1 EllipticKΠ14 .2 3/ GammaΠ16  GammaΠ13 
 1 C   p
16 3 2

We substitute x D 1=6 in (6.42) of [53] and have

  p  
Gamma 16 D 21=3 p3  Gamma 13 ^ 2
We show that the difference of the terms in the complete elliptic integral and  ’s
is
 1 6 p h  p i2
1 3 1
22=3
 Pi
Gamma 3
 16 3EllipticK 4
2  3 I
N[%]
0:
The third example is presented below, [30, (2)],
h h P ii
1 1 42kC5 .GammaŒ1=2Ck=GammaŒ1=2/3
FullSimplify FunctionExpand 16 kD0 26k kŠ3
 
Chop@N %  1
1
p 2 2 2 2
1 5EllipticKΠ16 .83 7/ 5GammaΠ17  GammaΠ27  GammaΠ47 

C 4 2
 p
64 7 4
0
The fourth series of Ramanujan type is the next one

Fig. 8.12 An S. Ramanujan

series
166 8 Pi Formulas

1 P1  p
1
p
 D 32 kD0 42 5 C 30 k C 5 5  1 
3
 p 8k
.GammaŒ1=2Ck=GammaŒ1=2/ 51
kŠ3
 216k 2
FullSimplifyŒFunctionExpandŒ
 p
1 P1
p
32 kD0 42 5 C 30 k C 5 5  1 
p ii
.GammaŒ1=2Ck=GammaŒ1=2/3 1 51 ^
3  26k
.8k/
 kŠ  2
Chop@N %  1
h  p p i
1p 1
2EllipticK 16  7 3  15
.267119
r  5/
2 32
p  p h  p p i
1
2 6 27 C 7 5 125 C 56 5 EllipticE 32 16  7 3  15 C
r  
p p
328  267 15 C 5 234319  35224 15
h  p p i
1
EllipticK 32 16  7 3  15
0
A short series of Ramanujan follows, [59],
P1 3

kD0 .1/
k
.4k C 1/  .GammaŒ1=2Ck=GammaŒ1=2/ I
  kŠ3
Chop@N %  2
0

8.12 R. W. Gosper Series

Gosper proved the following identity, [71, (57)],

h 2
i
 C 4ArcTanŒz C 2Log 12zz
z2 C1
D
P1 1  4.zC1/8kC1 2.zC1/8kC4 .zC1/8kC5 .zC1/8kC6

kD0 16k 8kC1
 8kC4  8kC5  8kC6

Let us see how can one prove it.

P1 1  4.zC1/8kC1 2.zC1/8kC4 .zC1/8kC5 .zC1/8kC6

kD0 16k 8kC1
 8kC4  8kC5
 8kC6
I

FullSimplifyŒFunctionExpandŒ%
 h i  1Ci 
2
2 ArcCot .1Cz/ 2 C .1 C i/ArcCot 1Cz C
   
 1Ci  1C 12 .1Cz/2 1C 14 .1Cz/4
.1 C i/ArcCoth 1Cz  Log p 1 4
 Log p 1 8
1 4 .1Cz/ 1 16 .1Cz/

Define
8.12 R. W. Gosper Series 167

Fig. 8.13 A domain of

convergence for a Gosper
series

h 2
i
f1Œz WDPi C 4ArcTanŒz C 2Log 12zz 2
z C1
I
 h i  1Ci 
2
f2Œz WD2 ArcCot .1Cz/ 2 C .1 C i/ArcCot 1Cz C
   
 1Ci  1
1C 2 .1Cz/2
1C 14 .1Cz/4
.1 C i/ArcCoth 1Cz  Log p 1 4
 Log p 1 8
I
1 4 .1Cz/ 1 16 .1Cz/

We show that the derivatives of the two functions coincide and the functions
coincide at a value of the argument.
{FullSimplify[f1’[z]],FullSimplify[f2’[z]]}
{f1[0],FullSimplify[ComplexExpand[f2[0]]]}
n o
16z 16z
;
1C2zC2z3 Cz4 1C2zC2z3 Cz4
f; g
We find the domain of convergence of the series.
Clear[k]
SumConvergence
h  i
1 4.zC1/8kC1 2.zC1/8kC4 .zC1/8kC5 .zC1/8kC6
16k 8kC1
 8kC4
 8kC5
 8kC6
;k
AbsŒ1 C z8 < 16
Let us note that the point z D 0 belongs to the domain of convergence.
RegionPlot[Abs[1+x+I y]2 <2,fx,-2.6,.6g,fy,-1.6,1.6g,
FrameTicks!fff-1.5,-.5,0,.5,1.5g,Noneg,ff-2.5,-1.5,-.5,0,.5g,Nonegg,
ImageSize!200]
See Fig. 8.13.
168 8 Pi Formulas

8.13 The Chudnovskys’ Series

The series of the Chudnovsky brothers is as follows [17]:

1 P1 .1/k .6k/Š 13591409C545140134k
 D 12 kD0 .kŠ/^ 3.3k/Š  .640320/^ .3kC3=2/
Now we prove it.
P .1/k .6k/Š
seriesChudnovsky[n ]:= 12 nkD0 .kŠ/^ 3.3k/Š 
13591409C545140134k
.640320/^ .3kC3=2/
N[1/fseriesChudnovsky[0],seriesChudnovsky[1],
seriesChudnovsky[2]g-Pi,30]
f5.90307941897917012045434349361J30.*^ -14;
3.07847804278696946621268494680J30.*^ -28;
 1.72052025953306221680193717343J30.*^ -42g
Its remarkable property consists in the fact that, with each additional term, about
14 exact digits are added to the sum.
FullSimplify@FunctionExpand@seriesChudnovsky@1
1

A series given at [9] has a similar remarkable property that each additional term
in the series adds about 24 exact digits.
Clear[a,b,chudnov,c]
p
a=212175710912 61 p C 1657145277365;
b=13773980892672
 61 C 107578229802750;
p
c=5280 236674 C 30303 61 ;

P .1/k .6k/Š.aCk b/
chudnov[n ]:=12 nkD0 kŠ3 .3k/Šc3kC3=2
I
Block[fMaxExtraPrecision=1000g,N[chudnov[#]- 1 ,30]]&/@f0,1,2,3g
Chop@N[FullSimplify[FunctionExpand[chudnov[1]]]- 1 ]
f1.61283781816166378669791721927J30.*^ -25;
 1.42103169042706048703554408064J30.*^ -50;
1.33768392030277996437842472463J30.*^ -75;
 1.30889434942233465826255332062J30.*^ -100g
0

8.14 B. Cloitre Series

An interesting sum was found by Cloitre, sum in which the golden ratio is present.
The result appeared in [18] and follows below.
g:=GoldenRatio;
 2
P g1 g2 g5 2g5
50 1 1 g
kD0 g5k .5kC1/2  .5kC2/2
 .5kC3/2
C .5kC4/2
C .5kC5/2
2
8.15 F. Bellard Series 169

Fig. 8.14 A Cloitre series

Some details regarding the convergence of the series are given below.
P  2
g1 g2 g5 2g5
cloitre[n ]:=50 nkD0 g15k .5kC1/
g
2  .5kC2/2  .5kC3/2 C .5kC4/2 C .5kC5/2

Table[N[cloitre[n],10],fn,5g] (* The first five terms *)

Table[N[ 2 -cloitre[n]],fn,10g]
f9:870515488; 9:869648052; 9:869606822; 9:869604548; 9:869604411g
f0:000911086; 0:0000436505; 2.4206409001692464J*^ -6;
 1.473621864533925J*^ -7; 9.56164747378807J*^ -9;
 6.49704290367481J*^ -10; 4.5712766905126045J*^ -11;
 3.304023721284466J*^ -12; 2.4158453015843406J*^ -13;
 1.7763568394002505J*^ -14g
It appears that each additional term in the series adds about one exact digit.
DiscretePlot[cloitre[n],fn,50g,Ticks!fAutomatic,f9.8, 2 gg,
PlotStyle!Black,ImageSize!250,AxesOrigin!f0,9.7g]
cloitre[1] (* The sum of the series *)
See Fig. 8.14.

8.15 F. Bellard Series

Fabrice Bellard in [8] has introduced the following rapidly convergent series.
FullSimplify[FunctionExpand[
1 P1 .1/k
 32 1 256 64 4 4 1

64 kD0 210k 4kC1
 4kC3 C 10kC1  10kC3  10kC5  10kC7 C 10kC9
]]

Let us see the behaviour of the first partial sums of this series.
170 8 Pi Formulas

Fig. 8.15 A Bellard series

bellard[n ]:=
1 Pn .1/k  32 1 256 64 4 4 1

64 kD0 210k 4kC1
 4kC3
C 10kC1
 10kC3
 10kC5
 10kC7
C 10kC9

Table[N[bellard[n],15],fn,5g] (* The first five terms of the sequence *)

Table[N[-bellard[n],10],fn,10g] (* At each step about three exact
decimals are added *)
DiscretePlot[bellard[n],fn,50g,
Ticks!fAutomatic,f3.1,gg,
PlotStyle!Black,
ImageSize!250,
AxesOrigin!f0,3g]
f3:14159257186839; 3:14159265364205; 3:14159265358976;
3:14159265358979; 3:14159265358979g
f8.172140293208859087919603904841J10.*^ -8;
 5.225753148164040239440670995045J10.*^ -11;
3.787038213038850139206135235540J10.*^ -14;
 2.938073224786711749125149429567J10.*^ -17;
2.379281510127834004780657979821J10.*^ -20;
 1.984326124743791286407198329027J10.*^ -23;
1.690819411332111257720829708060J10.*^ -26;
 1.464440904932783554716997630963J10.*^ -29;
1.284761748110936503778867721027J10.*^ -32;
 1.138867079619021633370576416979J10.*^ -35g
See Fig. 8.15.
Chapter 9
Optimization of Trajectories

9.1 Necessary Conditions for a Mayer Optimal Control

Problem

For brevity, below we introduce an abbreviated version of Theorem 4.2.i in [15, 48].
Let the Mayer problem of optimal control be expressed as

Œx; u D g.a; x.a/; b; x.b// .cost functional/; (9.1)

8
ˆ dx
ˆ
ˆ dt
D f .t; x.t/; u.t//; t 2 Œa; b; a.e., (differential constraint)
ˆ
<eŒx D .a; x.a/; b; x.b// 2 B  R1CnC1Cn .boundary conditions/;
(9.2)
ˆ
ˆ.t; x.t// 2 A; t 2 Œa; b .time and state constraints/;
ˆ
:̂
u.t/ 2 U.t/; t 2 Œa; b .control constraint/:

A pair .x.t/; u.t//, a  t  b, is said to be admissible (or feasible) provided that

x W Œa; b ! Rn is absolutely continuous, [47, 53], u W Œa; b ! Rp is measurable,
and x and u satisfy (9.2) a.e. Let ˝ be the class of admissible pairs .x; u/. The
goal is to find the minimum of the cost functional (9.1) over ˝, that is, to find
an element .x ; u / 2 ˝ so that 1 <  Œx ; u   Œx; u for all .x; u/ 2 ˝.
We introduce the variables D . 1 ; : : : ; n /, called multipliers, and an auxiliary
function H.t; x; u; /, called the Hamiltonian, defined on T  U  Rn by
n
X
H.t; x; u; / D i fi .t; x; u/: (9.3)
iD1

Let us define

M.t; x; / D inf H.t; x; u; /:

u2U.t/

© Springer International Publishing AG 2017 171

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_9
172 9 Optimization of Trajectories

Further necessary assumptions:

1. There exists an element .x ; u / 2 ˝ such that 1 <  Œx ; u   Œx; u for
all .x; u/ 2 ˝.
2. A is closed in R1Cn .
3. The set S D f.t; x; u/j.t; x/ 2 A; u 2 U.t/g is closed in R1CnCp .
4. f 2 C1 .S; R/.
5. Notation:

X n
@fi @fi @H
fit D ; fixj D ; Hxj D D i fixj ;
@t @xj @xj iD1

X n
@H @H
Ht D D i fit ; H j D D fj :
@t iD1
@ j

6. The graph f .t; x .t//j a  t  bg of the optimal trajectory x belongs to the

interior of A.
7. U does not depend on time and is a closed set.
8. The endpoint e .x / D .a; x .a/; b; x .b// of the optimal trajectory x is a point
of B, where B has a tangent variety B0 (of some dimension ı, 0  ı  2n C 2)
whose vectors are denoted by

   
h D . 1 ; 1 ; 2 ; 2 / ; 1 D 11 ; : : : ; 1n ; 2 D 21 ; : : : ; 2n ;

or by
   
h D . da; dx1 ; db; dx2 / ; dx1 D d11 ; : : : ; d1n ; dx2 D d21 ; : : : ; d2n :

Theorem 9.1 Assume the above eight hypotheses and let .x ; u / be an optimal
pair for the Mayer problem (9.1) and (9.2). Then the optimal pair .x ; u /
necessarily has the following properties:
(a) There exists an absolutely continuous function .t/ D . 1 .t/; : : : ; n .t// such
that

d i  
D Hxi t; x .t/; u .t/; .t/ ; i D 1; : : : ; n; t 2 Œa; b .a:e:/:
dt

If dg is not identically zero at eŒx , then .t/ is never zero in Œa; b.
(b) For almost any fixed t 2 Œa; b (a.e.), the Hamiltonian, as a function depending
only on u, takes its minimum value in U at the optimal strategy u D u .t/. This
implies M .t; x .t/; .t// D H .t; x .t/; u .t/; .t//, t 2 Œa; b (a.e).
(c) The function M.t/ D M .t; x .t/; .t// coincides a.e. in Œa; b with an absolutely
continuous function, and
9.2 Zermelo’s Navigation Problem 173

dM d     
D M t; x .t/; .t/ D Ht t; x .t/; u .t/; .t/ ; t 2 Œa; b .a:e:/:
dt dt

(d) (transversality relation) There exists a constant 0  0 such that

n 
X
. 0 ga  M.a// da C 0 gxi C i .a/ dx1i C
1
iD1
(9.4)
n 
X
. 0 gb C M.b// db C 0 gxi  i .b/ dx2i D 0;
2
iD1

for every vector h D . da; dx1 ; db; dx2 / 2 B0 .

(e) (Legendre–Clebsch) If H.t; x.t/; u.t/; .t// has second-order partial derivatives
with respect to u1 ; : : : ; um at least in a neighborhood of u .t/, then we
necessarily have that
m
X
Huj uk j k > 0
j;kD1

for all  D .1 ; : : : ; m / 2 Rm , all a < t < b; where the derivatives Huj uk are
computed at .t; x .t/; u .t/; .t//.

9.2 Zermelo’s Navigation Problem

Zermelo’s navigation problem requires determination of the optimal trajectory and

the associated guidance of a boat (ship, aircraft) traveling between two given points
so that the transit time is minimized. In this section, we study this minimum time
optimal control problem mainly, numerically, by the power of Mathematica. By
the Manipulate command, we show the families of trajectories of the navigation
problem for certain values of parameters.

9.2.1 Introduction

According to [21, p. 150], Zermelo was the first to formulate and solve in [78]
and [79] a problem that now is called the navigation problem of Zermelo. The
problem came to Zermelo’s mind when the airship Graf Zeppelin circumnavigated
the Earth in August 1929. He considered a vector field given in the Euclidean plane
that describes the distribution of winds as depending on place and time and treated
the question of how an airship or ship, moving at a constant speed against the
174 9 Optimization of Trajectories

surrounding air, has to fly in order to reach a given point B from a given point A
in the shortest possible time. With:
• x D x.t/ and y D y.t/ the Cartesian coordinates of the airship at time t,
• u D u.t; x; y/ and v D v.t; x; y/ the corresponding components of the vector field
representing the velocity of the wind (water) with respect to the Cartesian system,
• ˇ D ˇ.t; x; y/ the angle between the momentary speed .u0 ; v0 / of the airship
against the surrounding air and the x-axis and normalizing to k .u0 ; v0 / k D 1;
one has the system of differential equations that describes the problem

dx dy
D u C cos ˇ and D v C sin ˇ:
dt dt
Using the calculus of variations, Zermelo obtained the following differential
equation for the heading angle ˇ:

dˇ @v @u @v @u
D sin2 ˇ  C sin ˇ cos ˇ   cos2 ˇ  :
dt @x @x @y @y

The previous differential equation, called the Zermelo’s differential equation, is

a necessary condition for ˇ to be the optimal guidance function.
An extensive list of references on Zermelo’s navigation problem and on methods
of its study may be found in [56].
Since the navigation problem is obviously nonlinear, up to the author’s knowl-
edge, there is no solution in closed form, and its approaches make intensive use
of numerical methods. Hereafter, we use the power of Mathematica to study the
navigation problem of Zermelo numerically.

9.2.2 A Planar Form of the Navigation Problem

Zermelo’s navigation problem is a minimum time paths through a region of

position-dependent time vector velocity. Under a general form, the problem sup-
poses that a boat travels through a zone of currents. The magnitude and direction of
the currents are given by the functions of time and position.

u D u.t; x; y/ and v D v.t; x; y/

where .x; y/ are Cartesian coordinates giving the position of the boat at time t, and
.u; v/ are the velocity components of the boat at the current point .x; y/ at time t
in the x and y directions, respectively. The magnitude of the velocity of the boat
relative to the water is supposed to be a constant V > 0.
9.2 Zermelo’s Navigation Problem 175

The problem requires to steer the boat in such a way as to minimize the time
necessary to travel from a given point A D .x1 ; y1 / at instant a to another given
point B D .x2 ; y2 / at instant b. The equations of motion are
(
x0 .t/ D V cos ˇ.t/ C u.t; x.t/; y.t//
y0 .t/ D V sin ˇ.t/ C v.t; x.t/; y.t//; t 2 Œa; b;

where ˇ is the heading angle of the boat’s axis relative to a fixed coordinate axis.
Let the fixed coordinate axis be the horizontal axis and ˇ the control function.
In a more compact form, the navigation problem can be stated as follows:
8
ˆ 0
ˆx .t/ D V cos ˇ.t/ C u.t; x.t/; y.t//;
ˆ
ˆ
ˆ
ˆ
ˆy0 .t/ D V sin ˇ.t/ C v.t; x.t/; y.t//;
ˆ
ˆ
ˆ
ˆ
ˆ
ˆx.a/ D x1 ; y.a/ D y1 ; (initial conditions);
ˆ
ˆ
ˆ
ˆ
<x.b/ D x2 ; y.b/ D y2 ; (final conditions);
Œa; b; (finite horizon); (9.5)
ˆ
ˆ
ˆ
ˆ ˇ 2 C.Œa; b; R/; (control function);
ˆ
ˆ
ˆ
ˆ
ˆ
ˆu; v W Œa; b  R  R ! R; (components of the velocity of water);
ˆ
ˆ
ˆ
ˆV; (relative speed of the water);
ˆ
ˆ
:̂
g.b/ D b ! min; (cost functional):

We suppose that the functions u and v are continuous in the first variable and of
class C1 in the second and third variables.
By the maximum principle of Pontryagin under the form in (a) and (b) of
Theorem 9.1, we have that

0 @H @u @v 0 @H @u @v
1 D  D  1  2 ; 2 D  D  1  2 ; (9.6)
@x @x @x @y @y @y
@H
0D D V . 1 sin ˇ C 2 cos ˇ/ H) 1 sin ˇ D 2 cos ˇ; (9.7)
@ˇ

where the Hamiltonian of the system is

H .t; x; y; 1 ; 2 / D 1 .V cos ˇ C u.t; x; y// C 2 .V sin ˇ C v.t; x; y// C 1: (9.8)

If we solve the system of differential Eqs. (9.5) and (9.6), we find x; y; ˇ; 1 ; and
2 . By (9.7) and the initial and final conditions, we find the solutions that solve
the navigation problem. Based on [44] or [45], we have that there exists such a
solution.
176 9 Optimization of Trajectories

Example If the functions u and v do not depend explicitly upon t, that is,
(
x0 .t/ D V cos ˇ.t/ C u.x.t/; y.t//;
(9.9)
y0 .t/ D V sin ˇ.t/ C v.x.t/; y.t//; t 2 Œa; b;

then the problem is autonomous, and therefore we take a D 0 as the initial instant
and .0; 0/ as the final point.
From now on, we suppose that the functions u and v do not depend on t, i.e.,
Eqs. (9.9) hold. Then the Hamiltonian does not explicitly depend on t, and then
H D constant is a prime integral. Because we minimize time, this constant has to
be 0. Then from (9.7) we have that H D 0. We invoke (9.6) and get

 cos ˇ  sin ˇ
1 D and 2 D : (9.10)
V C u cos ˇ C v sin ˇ V C u cos ˇ C v sin ˇ

Substituting
ı (9.10) in (9.6) (or asking for consistency between (9.8) and
d @ˇ H dt D 0), what follows is the Zermelo’s navigation formula:

dˇ @v @u @v @u
D sin2 ˇ C sin ˇ cos ˇ   cos2 ˇ : (9.11)
dt @x @x @y @y

Now the nonlinear Eqs. (9.9) and (9.11) give the general solution for our
navigation problem. If we take into account the initial and final conditions, we get
the concrete solution if the data are consistent.
We study a special case considering for the current of water the following
functions:

u.x; y/ D .V=h/y; v.x; y/ D 0; (9.12)

where h is a nonzero real number. Now we express the data of the problem as
functions depending on the angle ˇ. From (9.11) to (9.12), we write

dˇ V V 
D cos2 ˇ H) tan ˇ D tan ˇf C t  tf ; (9.13)
dt h h

where tf is the final time and ˇf is the final angle, both still unknown.
From the second equation of the system we have

dy  
D V sin ˇ H) y D y.ˇ/ D h sec ˇ  sec ˇf :
dt
9.2 Zermelo’s Navigation Problem 177

Now we take into account the first equation of the system, that is,

dx V
D V cos ˇ  y H) x D x.ˇ/ D
dt h
  
h sec ˇf C tan ˇf    
 ln C tan ˇf  tan ˇ sec ˇf  sec ˇf  sec ˇ tan ˇf :
2 sec ˇ C tan ˇ

The angular limits of the navigation problem, the initial angle ˇ0 , and the final
one ˇf can be obtained requiring the following system of nonlinear equation to be
valid:

x .ˇ0 / D x1 and y .ˇ0 / D y1 : (9.14)

Now all the elements of the trajectory are determined, and we pass to the
numerical approach.

9.2.3 Numerical Approach to the Navigation Problem

Clearly the initial time is a D 0. We choose the initial position at .x.0/; y.0// D
.7:32; 3:727/. The final position is at the origin .0; 0/. Then by (9.14), we have
that the initial angle is 105ı , whereas the final angle is 240:004ı . The minimum
time to steer the boat from the initial point to the origin by formula (31) is 5:46439:
The trajectory of this problem is given in blue, whereas the heading direction vectors
appear in red.
boatZermelo[vbig ,h ,x0 ,y0 ,m ]:=Module[
ftorig,thetaf,thetaff,theta0,degree0,degreef,t1,t2,t,timef,x,y,n,ig,
x[t ]:=
-(h/2)(Sec[thetaf](Tan[thetaf]-Tan[t])-Tan[t](Sec[thetaf]-Sec[t])+
Log[Divide[Tan[thetaf]+Sec[thetaf],Tan[t]+Sec[t]]]);
y[t ]:=h(Sec[t]-Sec[thetaf]);
sol=FindRoot[
f-(h/2)(Sec[thetaff](Tan[thetaff]-Tan[torig])-Tan[torig](Sec[thetaff]-
Sec[torig])+
Log[Divide[Tan[thetaff]+Sec[thetaff],Tan[torig]+Sec[torig]]])==x0,
h(Sec[torig]-Sec[thetaff])==y0g,fftorig,1.85g,fthetaff,4.17gg];
theta0=torig/.sol; (* Initial angle *)
thetaf=thetaff/.sol; (* Final angle *)
timef=(h/vbig)(Tan[thetaf]-Tan[theta0]); (* Final time *)
degree0=(theta0360)/N[2]; (* Initial angle *)
degreef=(thetaf360)/N[2]; (* Final angle *)
thetaa[i ]:=theta0+i(thetaf-theta0)/m; (* Arrow at each thetaa[i]
angle *)
178 9 Optimization of Trajectories

Show[ParametricPlot[fx[t],y[t]g,ft,theta0,thetafg,
PlotStyle!fBlue,Thickg],
Graphics[ffRed,Arrowheads[0.02],
Arrow[Table[ffx[thetaa[n]],y[thetaa[n]]g,
fx[thetaa[n]]+vbigCos[thetaa[n]],
y[thetaa[n]]+vbigSin[thetaa[n]]gg,fn,0,mg]]g,
fBlack,Arrowheads[0.02],Arrow[ffx[thetaa[0]],y[thetaa[0]]g,
fx[thetaa[0]]-(vbig/h)y[thetaa[0]],y[thetaa[0]]gg]g,
fBlue,Arrowheads[0.025],
Arrow[ffx[thetaa[0]],y[thetaa[0]]g,fx[thetaa[0]]-(vbig/h)y[thetaa[0]]+
vbigCos[thetaa[0]],y[thetaa[0]]+vbigSin[thetaa[0]]gg]g,
fPointSize[0.015],Blue,Point[fx[thetaa[0]],y[thetaa[0]]g],
Point[fx[thetaa[m]],y[thetaa[m]]g]g,
Text[Style[Row[fStyle[”minimum time of traveling”,Italic],” = ”,timefg],13],
fx0/2,y0/3g],
Text[Style[A,Italic,12],fx0-0.3,y0g],Text[Style[O,Italic,12],f-0.25,0.25g],
Text[Style[Row[fStyle[”from A”,Italic],” = (”, x0,”,”,y0,”) ”,
Style[”to O=(0,0)”,Italic]g],13],fx0/2,y0/2g],
Text[Style[Row[fStyle[”initial angle in A”,Italic],” = ”,
Superscript[degree0,” ı”]g],13],fx0/2,2y0/3g],
Text[Style[Row[fStyle[”final angle in O”,Italic],” = ”,
Superscript[degreef,” ı”]g],13],fx0/2,4y0/4.8g]g],
PlotRange!All,ImageSize!500]]
The graph below exhibits the dynamics of the navigation problem when the initial
data belong to some intervals. The black (horizontal) vector at A represents the
direction and magnitude of the current vector at the initial point. The blue (oblique)
vector at A represents the tangent to the trajectory.
Manipulate[
Quiet@boatZermelo[vbig,h,x0,y0,m],
ffvbig,2,Vg,.5,5,0.5,Appearance!”Labeled”g,
ffh,2,hg,0.5,5,0.5,Appearance!”Labeled”g,
ffx0,7,”x0 ”g,5,20,1,Appearance!”Labeled”g,
ffy0,-2,”y0 ”g,-7,3,1,Appearance!”Labeled”g,
ffm,15,”number of arrows”g,10,30,1,Appearance!”Labeled”g,
SaveDefinitions!True
]
See Fig. 9.1.
Remark A forthcoming interesting work on the Zermelo’s navigation problem is
[25].
9.3 Optimal Guidance for Planar Lunar Ascent 179

Fig. 9.1 Zermelo’s navigation problem

9.3 Optimal Guidance for Planar Lunar Ascent

The present section is focused on the equations of the optimal guidance for
planar Lunar ascent from rest to insertion in a circular parking orbit. The goal
is to minimize the flight time under certain assumptions. We point out some
computational and graphical aspects for flights reaching two altitudes.
The problem of soft landing a spacecraft on the Moon is a known topic, and we
mention [55] and the references therein.

9.3.1 Introduction

The studies on the problem of the ascent from the lunar surface are not new; they
started before 1965 as it clearly follows from [14, 46] to [40]. At that time, the
methods of calculus of variations were still in force, and more and more strongly
the optimal control approach pointed out to be the way to get newer results.
In the lunar ascent problem, the spacecraft is launched from the surface of the
moon and ascends to the orbit insertion point, which is defined by a given final
velocity vector and a given altitude (sometimes also called free downrange). In spite
180 9 Optimization of Trajectories

of the gravity perturbations, it is expected that the ascent trajectory will stay on or
close to the plane containing the great circle defined by the launch point and the
insertion point.
We point out some computational and graphical aspects of this problem in
connection with certain recent results given by D. G. Hull in [35, 36] and [37].
A pertinent introduction on this topic together with a large list of references may
be found in [36] and [37].
For constant thrust, minimum time and minimum fuel consumption are the same
problem; see [13] or [15].
This section is organized as follows: Sect. 9.3.2 introduces the problem of the
ascent from the lunar surface giving some references to make clear the framework
of the paper and introduces equations, approximations, and initial values of the
ascent problem; Sect. 9.3.3 contains an optimal control approach of determining
the optimal trajectory under certain assumptions; and Sect. 9.3.4 discusses some
computational and graphical issues of earlier results connected to the problem of
determining the optimal trajectory. The Sect. 9.3.4.1 exhibits some figures on the
case of 15240 m altitude, whereas the Sect. 9.3.4.2 introduces some figures for the
case of 100 km altitude.

9.3.2 The Ascent from the Lunar Surface

Following [36], we consider the ascent problem of a spacecraft from the surface
of Moon. The equations of motion in three-dimensional flight over a nonrotating,
spherical Moon are written in the local horizontal/local vertical frame as shown in
Fig. 9.2. For applications, we considered the data in [36].
The parameters involved in the flight are as follows:
1. Œ0; tf  is the horizon of time. The final instant tf of the insertion on the coast or
parking trajectory is to be found in many cases.
2. x and y are the in-plane curvilinear distance (downrange) and altitude. The
altitude is specified. More specifically, in Sect. 9.3.4, we discuss in detail the
problem for two altitudes of 15240 m and 100 km.
3. u and v are the in-plane horizontal and vertical velocity components. We assume
that v.tf / D 0:
4. z and w are the out-of-plane curvilinear distance and velocity components.
5.  is the thrust pitch angle.
6. is the thrust yaw angle.
7. rm D 1738:14  103 m is the radius of the Moon.
8. gm D m =rm2 D 1:619 m=s2 is the acceleration of gravity on the surface of the
Moon, m being the gravitational constant of the Moon.
9. is the thrust to mass ratio T=m; where the thrust is T D 60030 N, and the mass
equation is m.t/ D m0  ˇt; m0 D 17512:6836 kg: For constant thrust, T D ˇVe ;
9.3 Optimal Guidance for Planar Lunar Ascent 181

Fig. 9.2 Flight over a

spherical Moon T
V
y u v
x

rm

where ˇ D 19:118 kg=s is the constant propellant mass flow rate, and Ve is the
constant exhaust velocity.
From 9 it follows that D .t/ D T=m D T=.m0  ˇt/ D Ve =.t  ˛/; where
˛ D m0 =ˇ:
If z
w

0; then the ascent is said to be planar.
The equations of motion for a spacecraft in three-dimensional flight over a non-
rotating spherical moon are written in the local horizontal/local vertical coordinate
system. Then x; y; u; and v are the in-plane downrange, altitude, horizontal velocity,
and vertical velocity component, suggested by Fig. 9.2.
According to [36] and [37], the equations of motion are given by

x0 D rm u=r; u0 D cos  cos C .uw=r/ tan.z=rm /  uv=r;

y0 D v; v0 D sin   g C u2 =r C w2 =r; (9.15)
0 0 2
z D rm w=r; w D cos  sin  .u =r/ tan.z=rm /  vw=r:

The ascent trajectory from the lunar surface to a coast orbit consists of a
(sometimes vertical) rise, constant rate pitchover, and optimal guidance phases. It
is designed that the trajectory lies in the plane of the great circle plane containing
the launch point and the orbit insertion point. Small deviations from the plane occur
because of perturbations on gravity and spacecraft performance. Because the out-
of-plane motion is small, the trajectory is said to be quasiplanar.
Because the insertion point is relatively close to the lunar surface, we consider
the following approximations:
y=rm  1; so that r rm and g m =rm2 D gm :
The out-of-plane components z and w are assumed to be sufficiently small that
the terms z=rm and w=rm can also be neglected.
182 9 Optimization of Trajectories

Fig. 9.3 Flight over a flat

Moon
T

y
q
O
y

For typical launch trajectories, it can be shown that uv=rm  T=m; so the last
term in the equation of u0 is neglected.
Under these circumstances, the Eq. (9.15) can be written under the simplified
form as

x0 D u; u0 D cos  cos ;
y0 D v; v0 D sin   gm ; (9.16)
z0 D w; w0 D cos  sin :

By flight over a flat moon, as in [36], we mean flight over a spherical moon at
low altitude and low velocities. While the low altitude condition .y=rm  1/ is met
so that r rm ; g gm ; the low velocity requirement .u2 =.gm rm /  1/ is not; see
[36]. Flight over a flat moon is suggested by Fig. 9.3.
We note that for flight over a flat moon, the gravity vector is along the radius
vector.

9.3.3 The Optimal Control Approach

As we mentioned earlier, for constant thrust, the problem of minimizing fuel

consumption is the same as minimizing flight time. Hence, the optimal control
problem requires to minimize the performance index (cost functional)

 D Œx; y; z; ;  D tf (9.17)
9.3 Optimal Guidance for Planar Lunar Ascent 183

subject to the differential constraints (9.16), the following initial constraints

x.0/ D 0; y.0/ D 0; z.0/ D 0; u.0/ D 0;

v.0/ D 0; and w.0/ D 0;
(9.18)
and the final boundary constraints, as suggested in [35] and [36],

y.tf / D 15240; z.tf / D 0; u.tf / D 1625;

v.tf / D 0; and w.tf / D 0:
(9.19)
The control variables are the angles  D .t/ and D .t/:
The Hamiltonian of the optimal control problem (9.16)–(9.19) is of the form

H D 1 u C 2 v C 3 w C 4 cos  cos C 5 . sin   gm / C 6 cos  sin :

By (a) in Theorem 9.1, we have the first necessary condition under the form of
the system of differential equations to the multipliers

01 D 0; 02 D 0; 03 D 0; 04 D  1 ; 05 D  2 ; and 06 D  3 :

Immediately, the equations of the multipliers follow:

1 D c1 ; 2 D c2 ; 3 D c3 ;
4 D c1 tCc4 ; 5 D c2 t C c5 ; 6 D c3 t C c6 ;

where the cs are all constants of integration.

By (b) in Theorem 9.1, the equations of the optimal controls are given by

H D  4 sin  cos C 5 cos   6 sin  sin D 0;

(9.20)
H D  4 cos  sin C 6 cos  cos D 0:

Under the necessary condition of ascent, we suppose that cos  > 0, and
therefore from the latter equation in (9.20), it follows that

4 sin  6 cos D 0; (9.21)

and for j j small also 4 tan  6 D 0: Thus, the first equation in (9.20) is written
as

. 4 cos C 6 sin / tan   5 D 0: (9.22)

From (9.21) to (9.22), we have the multipliers 5 and 6 depending on 4 ; that is,
6 D 4 tan and 5 D 4 tan  sec :
The second-order partial derivatives of H with respect to the control variables are
H D  4 sec  sec ; H D  4 cos  sec ; and H D 0:
184 9 Optimization of Trajectories

By the Legendre–Clebsch necessary condition in Theorem 9.1 for a minimum,

we write H  0; H  0; and H H  .H /2  0: Then it follows that

sgn. 4 cos  cos /  0: (9.23)

Thus, necessarily 4  0:
From (9.21) for some real " D ˙1, we write

" 6 " 4
q cos q sin D 0:
24 C 26 24 C 26

Therefore, we can deduce the equations of the optimal yaw angle

q q
sin D " 6 = 24 C 26 and cos D " 4 = 24 C 26 ;

if 24 C 26 ¤ 0: Suppose for a while that 4

6
0: Then by (9.22), it follows
that 5
0: Thus, all the Lagrange multipliers are null, and therefore we get a
contradiction.
Naturally we admit that 24 C 26 ¤ 0: From the first equation in (9.20), we
successively have for some real ı D ˙1 that

H D 0 H)
" 4 " 6
 4 sin  q C 5 cos   6 sin  q D 0 H)
24 C 26 24 C 26

2 C 26
 "q4 sin  C 5 cos  D 0 H)
24 C 26
q
ı 24 C 26 ı 5
 "q sin  C q cos  D 0:
24 C 25 C 26 24 C 25 C 26

Now we impose condition (9.23) and get that ı D 1: For " we choose 1 because
4 is negative, and for small values of ; cos is positive. From the last equation,
we can now deduce the equations of the optimal pitch angle
q
5 24 C 26
sin  D  q and cos  D q : (9.24)
24 C 25 C 26 24 C 25 C 26
9.3 Optimal Guidance for Planar Lunar Ascent 185

With these results on the expressions of angles, by (9.16), the velocity and
acceleration equations of the optimal trajectory, if any, are of the form

4
x0 D u; u0 D  q ;
24 C 25 C 26
5
y0 D v; v0 D  q  gm ; (9.25)
24 C 25 C 26
6
z0 D w; w0 D  q :
4 C 25 C 26
2

Lemma 9.1 The tangential velocity function u is of class C1 ; positive and increas-
ing on the interval Œ0; tf :.
Proof It is enough to note that u0 in (9.25) is continuous and positive. t
u
1
Lemma 9.2 The radial velocity function v is of class C and positive on the interval
Œ0; tf :
Following [36] and [37], we introduce the functions A; B; and C by
Z t
ds
A.t/ D A.t; c1 ; : : : ; c6 / D p ;
0 .s  ˛/ ps2 C qs C r
Z t
s ds
B.t/ D B.t; c1 ; : : : ; c6 / D p ; (9.26)
0 .s  ˛/ ps2 C qs C r
Z t
s2 ds
C.t/ D C.t; c1 ; : : : ; c6 / D p ;
0 .s  ˛/ ps2 C qs C r

where the constants p; q; and r are defined as follows:

p D c21 C c22 C c23 ; q D 2.c1 c4 C c2 c5 C c3 c6 /; and r D c24 C c25 C c26 :

Lemma 9.3 Assume that t  0: The following two integral equalities hold
Z t Z t
A.s/ ds D tA.t/  B.t/ and B.s/ ds D tB.t/  C.t/: (9.27)
0 0

Proof The two integral equalities are obtained by integration by parts; see [53,
Chap. 6]. t
u
Theorem 9.2 From (9.25) with (9.26) and (9.27) by integration, we get the
equations of velocities and states.
186 9 Optimization of Trajectories

u.t/ D u.0/  Ve c1 B C Ve c4 A;
v.t/ D v.0/  Ve c2 B C Ve c5 A  gm t;
w.t/ D w.0/  Ve c3 B C Ve c6 A;
(9.28)
x.t/ D x.0/ C tu.t/ C Ve c1 C  Ve c4 B;
y.t/ D y.0/ C tv.t/ C Ve c2 C  Ve c5 B C gm t2 =2;
z.t/ D z.0/ C tw.t/ C Ve c3 C  Ve c6 B;

Remark By Eq. (9.28), we know the form of the optimal trajectories, if any. What
has remained is the step of determining the constants of integration cs such that the
seeking trajectory satisfies the boundary conditions.
In order to solve the minimum time optimal control problem, let us suppose that
all the initial and final values are given, that is, we know x.0/; y.0/; z.0/; u.0/; v.0/;
w.0/; y.tf /; z.tf /; u.tf /; v.tf /; and w.tf /: The variables to be found are tf and x.tf /:
Then we write

Vx D u.tf /  u.0/; Vy D v.tf /  v.0/ C gm tf ; Vz D w.tf /  w.0/;

Rx D x.tf /  x.0/  tf u.tf /; Ry D y.tf /  y.0/  tf v.tf /  gm tf 2 =2; (9.29)
Rz D z.tf /  x.0/  tf w.tf /;

and further

Vx D  Ve c1 B.tf / C Ve c4 A.tf /; Vy D Ve c2 B.tf / C Ve c5 A.tf /;

Vz D  Ve c3 B.tf / C Ve c6 A.tf /;
(9.30)
Rx D Ve c1 C.tf /  Ve c4 B.tf /; Ry D Ve c2 C.tf /  Ve c5 B.tf /;
Rz D Ve c3 C.tf /  Ve c6 B.tf /:

Both sides of the preceding equalities depend on the final instant tf :

Because tf is free, by the transversality condition in Theorem 9.1, we have that
H.tf / D 1; which gives

c1 u.tf / C c2 v.tf / C c3 w.tf /  .c2 tf C c5 /gm

q
 .T=.m0  ˇtf // .c1 tf C c4 /2 C .c2 tf C c5 /2 C .c3 tf C c6 /2 D 1:
(9.31)
Equations (9.30) and (9.31) form a system of seven nonlinear equations in the
seven unknowns tf ; c1 ; c2 ; c3 ; c4 ; c5 ; and c6 : Since x.tf / is free, by the transversality
condition, we have that c1 D 0; so we only have six unknowns to determine.
9.3 Optimal Guidance for Planar Lunar Ascent 187

Theorem 9.3 If the nonlinear system formed by the Eqs. (9.30) and (9.31) has a
unique system of real solutions, then the minimum time optimum control problem
has a unique solution. If the nonlinear system has several systems of real solutions,
then we have to check on which one the minimum time is achieved.
Remark Based on Filippov’s existence Theorem for Mayer problem of optimal
control as it is given in [15, p. 310], the problem discussed by Theorem (9.3) has a
solution.

9.3.4 Computational Issues

This subsection uses the results of the previous one and analyzes their effectiveness.
Remark Denote X D ax2 CbxCc; where a D p; b D 2p˛Cq; and c D p˛ 2 Cq˛Cr:
We have that
ˇ ˇ
Z
dx 1 ˇ 2c C bx C 2pcX ˇ
ˇ ˇ
P.x/ D p D  p ln ˇ ˇ ; [29, p. 97]
x X c ˇ x ˇ
ˇ ˇ
Z
dx 1 ˇ b C 2ax C 2paX ˇ
ˇ ˇ
Q.t/ D p D p ln ˇ p ˇ ; [29, p. 94]
X a ˇ b  4ac ˇ
2

Z p
x dx X b
R.x/ D p D  Q.x/; [29, p. 96]:
X a 2a

Then the functions A; B; and C are written as

A.t/ D P.t  ˛/  P.˛/;

B.t/ D Q.t  ˛/  Q.˛/ C ˛.P.t  ˛/  P.˛//;
C.t/ D R.t  ˛/  R.˛/ C 2˛.Q.t  ˛/  Q.˛// C ˛ 2 .P.t  ˛/  P.˛//:

These formulas are useful for an easier approach to numerical computation in

Theorem 9.3.
Remark In case of the planar flight, we have to consider that z D w D D 0:
These equalities imply that

3 D 6 D 0:

Remark Hereafter, we discuss the planar case using the initial and final values as
they are given in [35]. In this case, the multipliers are of the form

1 D c1 ; 2 D c2 ; 3 D c1 t C c3 ; and 4 D c2 t C c4 ;

188 9 Optimization of Trajectories

and the Eqs. (9.28) are written as

u.t/ D u.0/  Ve c1 B C Ve c3 A; v.t/ D v.0/  Ve c2 B C Ve c4 A  gm t;

x.t/ D x.0/ C tu.t/ C Ve c1 C  Ve c3 B; (9.32)
y.t/ D y.0/ C tv.t/ C Ve c2 C  Ve c4 B C gm t2 =2:

If we integrate the velocity equations in (9.32), then the equations of states can be
written also as

x.t/ D x.0/ C u.0/t C Ve c3 tA.t/  Ve .c1 t C c3 /B.t/ C Ve c1 C.t/;

y.t/ D y.0/ C v.0/t C Ve c4 tA.t/  Ve .c2 t C c4 /B.t/ C Ve c2 C.t/  gm t2 =2:

If we denote by .xc ; yc / the pair of Cartesian components of the .x; y/ point in

local horizontal/local vertical coordinates, then

xc .t/ D .rm C y.t// sin.x.t/=rm / and yc .t/ D .rm C y.t// cos.x.t/=rm /:

Immediately it follows that yc 0 .0/ D xc 0 .0/ D 0:

Lemma 9.4 We have that
p
00 Ve c4  gm ˛ r Ve c3
yc .0/ D p and xc 00 .0/ D p :
˛ r ˛ r

Then by the L’Hospital Theorem, [53, p. 207], the angle of the spacecraft at the
take-off moment is
p
Ve c4 C gm ˛ r
arctan : (9.33)
Ve c3

9.3.4.1 Ascent to 15,240 m Altitude

In this subsection, we present the numerical results and graphics for the ascent to
the 15;240 m altitude.
Remark Consider the data given in Sect. 9.3.2 and the guess (approximate) values
as they are listed in [35], i.e.,

tf D 404; c1 D 0; c2 D 5:87  105 ; c3 D 0:0399; and c4 D 0:029:

With these initial values, we have obtained the following final moment and
coefficients of the Lagrange multipliers:
9.3 Optimal Guidance for Planar Lunar Ascent 189

tf D 393:784; c1 D 0; c2 D 2:60297  104 ;

c3 D 0:170877; and c4 D 0:124794:

According to our calculation, the downrange in this setting is x.tf / D 279; 348:
If we introduce the following constants to measure the degree of accuracy of our
calculations

ı1 D Vx  Ve .c1 B.tf /  c3 A.tf //; ı2 D Vy  Ve .c2 B.tf /  c4 A.tf //;
ı3 D Rx C Ve .c1 C.tf /  c3 B.tf //; ı4 D Ry C Ve .c2 C.tf /  c4 B.tf //;
ı5 D c1 u.tf / C c2 v.tf /  .c2 tf C c4 /gm
q
 .T=.m0  ˇtf // .c1 tf C c3 /2 C .c2 tf C c4 /2 ;

then the results obtained are given in Table 9.1.

Lemma 9.5 By (9.24) the angle of the thrust vector at the take-off moment is given
by

lim arctan. 4 = 3 / D 36:1412ı :

t#0

Applying (9.33), the angle of the spacecraft in respect to the local horizon at the
take-off moment is 8:27585ı :
In the last part of the present subsection, we introduce some graphics visualizing
the numerical results.
Figures 9.4 and 9.5 introduce the variations of the tangential and the radial
velocities.

Table 9.1 Accuracy of computation

ı1 D 6:13909  1012 ı2 D 6:13909  1012
ı3 D 2:32831  109 ı4 D 6:69388  109 ı5 D 1:0

Fig. 9.4 Tangential velocity u

1625.

O t
393.784
190 9 Optimization of Trajectories

Fig. 9.5 Radial velocity v 59.2125

30

O t
231.997 393.784

Coast trajectory
A B
Surface of the Moon

Fig. 9.6 Lunar ascent—complete time

Fig. 9.7 Lunar

ascent—during the first A 8.27585 °
second
Surface of the Moon

Figure 9.6 represents the whole trajectory of the spacecraft with a part of the
lunar surface. A is the initial point, whereas B is the final point.
The picture that we introduce in Fig. 9.7 represents the trajectory of the spacecraft
constrained on the time interval t 2 Œ0; 1: The parameter of the lunar arc belongs to
the interval Œ=2  0:0000008; =2 C 0:0000002:
Figures 9.4, 9.5, 9.6, and 9.7 are obtained by means of the code which follows:
(* Optimal guidance for a planar Lunar ascent *)
Clear[”c*”,”t*”,”u*”,”v*”,”x*”,”y*”,m0,beta]
t0=.0; (* initial time s *)
x0=.0;y0=.0; (* initial site in local horizontal system of
coordinates *)
u0=.0;v0=.0; (* initial velocities m/s *)
rm=1738.14103 ; (* radius of the Moon m *)
gm=1.619; (* gravitational acceleration on the moon m/s2 *)
thrust= 60030.0; (* thrust N *)
m0=17512.6836; (* initial mass kg *)
beta=19.118;
9.3 Optimal Guidance for Planar Lunar Ascent 191

vexit=thrust/beta;
(* xb=279280.0; downrange m *)
yb=15240.0; (* final altitude m *)
ub=1625.0; vb=0.0; (* final velocities m/s *)
(* tb=394.0; final time s *)
velx=ub-u0; vely=vb-v0+gmtb; (* residues of velocities *)
radx=x[tb]-x0-ubtb; rady=yb-y0-vbtb-0.5gmtb2 ;
(* residues of states *)
alpha=m0/beta;
p=a=c12 +c22 ;q=-2(c1c3+c2c4);r=c42 +c32 ;
b=2 palpha+q;c=palpha2 +qalpha+r;
pbig[x ,c1 ,c2 ,c3 ,c4 ]:=-(1/Sqrt[c])Log[Abs[
Divide[2 c+bx+2Sqrt[c(ax2 +bx+c)],x]]];
qbig[x ,c1 ,c2 ,c3 ,c4 ]:=(1/Sqrt[a])Log[Abs[Divide[
2Sqrt[a(a x2 +b x+c)]+2ax+b,Sqrt[b2 -4ac]]]];
rbig[x ,c1 ,c2 ,c3 ,c4 ]:= Divide[Sqrt[ax2 +bx+c],a]-
Divide[b,2a]qbig[x,c1,c2,c3,c4];
aa[c1 ,c2 ,c3 ,c4 ]:=pbig[tb-alpha,c1,c2,c3,c4]-
pbig[-alpha,c1,c2,c3,c4];
b[c1 ,c2 ,c3 ,c4 ]:=qbig[tb-alpha,c1,c2,c3,c4]-
qbig[-alpha,c1,c2,c3,c4]+alpha(pbig[tb-alpha,c1,c2,c3,c4]-
pbig[-alpha,c1,c2,c3,c4]);
cc[c1 ,c2 ,c3 ,c4 ]:=rbig[tb-alpha,c1,c2,c3,c4]-
rbig[-alpha,c1,c2,c3,c4]+2alpha(qbig[tb-alpha,c1,c2,c3,c4]-
qbig[-alpha,c1,c2,c3,c4])+alpha2 (pbig[tb-alpha,c1,c2,c3,c4]-
pbig[-alpha,c1,c2,c3,c4]);
c1=0.0; (* because of the downrange is free *)
root=FindRoot[fc1ub+c2vb+(c2tb-c4)gm-
(thrust/(m0-betatb))Sqrt[(c1tb-c3)2 +(c2tb-c4)2 ]==-1,
velx==-vexitc1bb[c1,c2,c3,c4]+vexitc3aa[c1,c2,c3,c4],
vely==-vexitc2bb[c1,c2,c3,c4]+vexitc4aa[c1,c2,c3,c4],
rady==vexitc2cc[c1,c2,c3,c4]-vexitc4bb[c1,c2,c3,c4]g,
fftb,404.0g,fc2,-5.87105 g,fc3,-.0399g,fc4,-.029gg,
AccuracyGoal!10,PrecisionGoal!10];
tb=root[[1,2]];c2=root[[2,2]];c3=root[[3,2]];c4=root[[4,2]];
(* Build up the trajectory *)
aaa[t ]:=pbig[t-alpha,c1,c2,c3,c4]-pbig[-alpha,c1,c2,c3,c4];
bbb[t ]:=qbig[t-alpha,c1,c2,c3,c4]-qbig[-alpha,c1,c2,c3,c4]+
alpha(pbig[t-alpha,c1,c2,c3,c4]-pbig[-alpha,c1,c2,c3,c4]);
ccc[t ]:=rbig[t-alpha,c1,c2,c3,c4]-rbig[-alpha,c1,c2,c3,c4]+
2alpha(qbig[t-alpha,c1,c2,c3,c4]-qbig[-alpha,c1,c2,c3,c4])+
alpha2 (pbig[t-alpha,c1,c2,c3,c4]- pbig[-alpha,c1,c2,c3,c4]);
u[t ]:=u0-vexit(c1bbb[t]-c3aaa[t]);
v[t ]:=v0-vexit(c2bbb[t]-c4aaa[t])-gmt;
192 9 Optimization of Trajectories

x[t ]:=x0+u[t]t+vexit(c1ccc[t]-c3 bbb[t]);

y[t ]:=y0+v[t]t+vexit(c2ccc[t]-c4 bbb[t])+0.5gmt2 ;
radx==vexitc1cc[c1,c2,c3,c4]-vexit c3 bb[c1,c2,c3,c4];
fvalue1=-velx-vexitc1bb[c1,c2,c3,c4]+vexitc3aa[c1,c2,c3,c4],
value2=-vely-vexitc2bb[c1,c2,c3,c4]+vexitc4aa[c1,c2,c3,c4],
value3=-radx+vexitc1cc[c1,c2,c3,c4]-vexitc3bb[c1,c2,c3,c4],
value4=-rady+vexitc2cc[c1,c2,c3,c4]-vexitc4bb[c1,c2,c3,c4],
value5=c1ub+c2vb+(c2tb-c4)gm-
(thrust/(m0-betatb))Sqrt[(c1tb-c3)2 +(c2tb-c4)2 ],x[tb]g
xcart[t ]:=(rm+y[t])Sin[Divide[x[t],rm]];
ycart[t ]:=(rm+y[t])Cos[Divide[x[t],rm]];
xinit=xcart[t0];yinit=ycart[t0];
xfinal=xcart[tb];yfinal=ycart[tb];
theta[t ]:=ArcTan[Divide[c2t-c4,c1t-c3]]180/N[Pi];
(* Tangential and Radial velocities *)
fPlot[u[t],ft,t0,tbg],Plot[v[t],ft,t0,tbg],Plot[theta[t],ft,t0,1g]g;
roott=t/.Solve[vexit(c2t-c4)==-gm(t-alpha)Sqrt[pt2 +qt+r],t][[2]];
vroott=v[roott];
fShow[Plot[u[t],ft,t0,tbg],
Graphics[ffArrowheads[.02],Arrow[ff-40,0g,ftb+50,0gg],
Arrow[ff0,0g,f0,u[tb]+400gg]g,
fPointSize[0.015],Point[f0,0g],Point[ftb,0g],Point[f0,u[tb]g],
Point[ftb,u[tb]g]g,fDashed,Line[fftb,0g,ftb,u[tb]gg],
Line[ff0,u[tb]g,ftb,u[tb]gg]g,Text[Style[O,Italic,12],f-18,100g],
Text[Style[t,Italic,12],ftb+43,110g],
Text[Style[u,Italic,12],f-18,u[tb]+340g],
Text[Style[tb,Italic,12],ftb,-140g],
Text[Style[u[tb],Italic,12],f40,u[tb]+120g]g],
PlotRange!All,Axes!False,ImageSize!250],
Show[Plot[v[t],ft,t0,tbg],Graphics[ffArrowheads[.02],
Arrow[ff-40,0g,ftb+50,0gg],Arrow[ff0,0g,f0,69gg]g,
fPointSize[0.015],Point[f0,0g],Point[ftb,0g],Point[froott,0g],
Point[froott,vroottg],Point[f0,vroottg],Point[f0,30g],
Point[f0,vroottg],Point[ftb,v[tb]g]g,
fDashed,Line[ffroott,0g,froott,vroottgg],
Line[ff0,vroottg,froott,vroottgg]g,Text[Style[O,Italic,12],f-18,5g],
Text[Style[t,Italic,12],ftb+43,5g],Text[Style[v,Italic,12],f-16,66g],
Text[Style[tb,Italic,12],ftb,-6g],Text[Style[”30”,Italic,12],f-24,30g],
Text[Style[vroott,Italic,12],f50,63g],
Text[Style[roott,Italic,12],froott,-6g]g],
PlotRange!All,Axes!False,ImageSize!250]g
tau[t ]:=Divide[thrust,m0-betat];
vprime[t ]:=-tau[t]Divide[-c2t+c4,Sqrt[(c1t-c3)2 +(c2t-c4)2 ]]-gm;
uprime[t ]:=-tau[t]Divide[-c1t+c3,Sqrt[(c1t-c3)2 +(c2t-c4)2 ]];
9.3 Optimal Guidance for Planar Lunar Ascent 193

fPlot[tau[t],ft,t0,tbg],Plot[vprime[t],ft,t0,tbg]g
ArcTan[(ycart[1]-ycart[0])/(xcart[1]-xcart[0])]180/N[Pi];
theta[t0];Limit[theta[t],t!0]
Limit[Divide[ycart[t]-ycart[0],t],t!0];
Limit[Divide[xcart[t]-xcart[0],t],t!0];
anglestarting=Divide[vexitc4+gmalphaSqrt[r],vexitc3];
anglestartingdegree=ArcTan[Divide[vexitc4+gmalphaSqrt[r],
vexitc3]]180/N[Pi];
anglestartingarctan=ArcTan[anglestarting];
circle[radius , ]:=radiusfCos[],Sin[]g;
fShow[ParametricPlot[fxcart[t],ycart[t]g,ft,t0,1g],
ParametricPlot[circle[rm,a],fa,Pi/2-0.0000008,Pi/2+0.0000002g],
Graphics[fPointSize[0.015],fRed,Point[fxinit,yinitg]g,
Text[Style[” ”,Italic,12],fxinit,yinit+0000g],
Text[Style[A,Italic,12],fxinit-.07,yinit+.05g],
Text[Style[”Surface of the Moon”,Italic,12],fxinit+.2,yinit-.11g],
Text[Style[” ”,Italic,12],fxinit,yinit-0.2g],
Text[Style[anglestartingdegree”ı ”,Italic,12],fxinit+1.,yinit+.06g],
circle[fxinit,yinitg,0.5,f0,anglestartingarctang],
fArrowheads[.02],Arrow[ffxinit+0.5Cos[anglestartingarctan-0.0001],
yinit+0.5Sin[anglestartingarctan-0.0001]g,
fxinit+0.5Cos[anglestartingarctan],yinit+0.5Sin[anglestartingarctan]g
g]gg],
PlotRange!All,Axes!False,ImageSize!250],
radius=Sqrt[xfinal2 +yfinal2 ];
Show[ParametricPlot[fxcart[t],ycart[t]g,ft,t0,tbg],
ParametricPlot[circle[rm,a],fa,13.4Pi/30,24.2Pi/48g],
ParametricPlot[circle[radius,a],fa,13.3Pi/30,24.0Pi/48g],
Graphics[fPointSize[0.010],fRed,Point[fxinit,yinitg]g,
fGreen,Point[fxfinal,yfinalg]g,
Text[Style[” ”,Italic,12],fxinit,yinit+90000g],
Text[Style[A,Italic,12],fxinit-7000,yinit+7000g],
Text[Style[B,Italic,12],fxfinal,yfinal+11000g],
Text[Style[”Coast trajectory”,Italic,12],(radius+15000)fCos[Pi/2-0.1],
Sin[Pi/2-0.1]g],
Text[Style[”Surface of the Moon”,Italic,12],fxinit+50000,yinit-15000g],
Text[Style[” ”,Italic,12],fxinit,yinit-70000g],
Text[Style[” ”,Italic,12],fxfinal-10000,yfinal+45000g]g],
PlotRange!All,Axes!False,ImageSize!400]g
Show[ParametricPlot[fxcart[t],ycart[t]g,ft,t0,tbg],
ParametricPlot[circle[rm,a],fa,0,2Pig],
Graphics[fPointSize[0.010],fBlack,Point[f0,0g]g,
fRed,Point[fxinit,yinitg]g,fGreen,Point[fxfinal,yfinalg]g,
Text[Style[” ”,Italic,12],fxinit,yinit+130000g],
194 9 Optimization of Trajectories

Table 9.2 Accuracy of computation for 100 km altitude

ı1 D 2:27374  1012 ı2 D 4:09273  1012
ı3 D 1:04774  109 ı4 D 4:42378  109 ı5 D 1:0

Text[Style[A,Italic,18],fxinit,yinit+130000g],
Text[Style[B,Italic,18],fxfinal,yfinal+85000g],
Text[Style[”Surface of the Moon”,Italic,18],fxinit,yinit-210000g],
Text[Style[” ”,Italic,12],fxinit,yinit-70000g],
Text[Style[” ”,Italic,12],fxfinal-10000,yfinal+45000g]g],
PlotRange!All,Axes!False,ImageSize!400]

9.3.4.2 Ascent to 100,000 m Altitude

In this subsubsection, we present the numerical results and graphics for the ascent
to 100;000 m altitude, that is, y.tf / D 100;000: The results and figures of the present
case follow the previous code by changing the initial data.
Remark We use the same guess (approximate) values as they are given in the pre-
vious subsubsection. We have obtained the following final moment and coefficients
of the Lagrange multipliers

tf D 440:185; c1 D 0; c2 D 0:000715588;
c3 D 0:114418; and c4 D 0:244682:

According to our calculation, the downrange in this setting is x.tf / D 273; 612:
Similar to Table 9.1, we introduce the values of the degree of accuracy in the
present setting. The results are given by Table 9.2.
Lemma 9.6 The angle of the thrust vector at the take-off moment is given by

 D 64:9383ı ;

whereas the angle of the spacecraft at the take-off moment is

45:6646ı :

In the sequel similar to the previous subsubsection, we exhibit some figures

related to the case of ascent to 100 km altitude.
Figures 9.8 and 9.9 introduce the variations of the tangential and the radial
velocities.
Figure 9.10 represents the trajectory of the rocket and the entire lunar surface.
The lunar surface is assimilated to a circle.
9.3 Optimal Guidance for Planar Lunar Ascent 195

Fig. 9.8 Tangential velocity; u

100 km altitude 1625.

O t
440.185

Fig. 9.9 Radial velocity; v

100 km altitude 372.284

O t
287.121 440.185

B
A

Surface of the Moon

Fig. 9.10 Lunar ascent—complete figure; 100 km altitude

196 9 Optimization of Trajectories

Coast trajectory
B

Surface of the Moon

Fig. 9.11 Lunar ascent—complete time; 100 km altitude

Fig. 9.12 Lunar

ascent—during the first
second; 100 km altitude

45.6646 °
A
Surface of the Moon

Figure 9.11 represents the whole trajectory of the spacecraft with a part of the
lunar surface.
The picture that we introduce in Fig. 9.12 represents the trajectory of the
spacecraft constrained on the time interval t 2 Œ0; 1: The parameter of the lunar
arc belongs to the interval Œ=2  0:0000006; =2 C 0:00000035:
Remark The maximal orbit transfer problem is connected to the problem of lunar
ascent. An approach to the maximal orbit transfer problem with similar method is
realized in [57]. Other flying problems are discussed in [48].
Chapter 10
Miscellany in the Euclidean Plane

10.1 Some Planar Curves

In this section we present some applications of Mathematica inspired by various

sources in planar geometry. Many of the applications are based on the use of the
built-in functions Manipulate or Animate.

10.1.1 Conchoid of Nicomedes

The conchoid of Nicomedes is the locus of points a fixed distance away from a line
as measured along a line from the focus point. The polar equation of the conchoid
of Nicomedes is r.t/ D b C a sec.t/, whereas in Cartesian coordinates, we write the
parametric equations for it t ! .b C a cos.t//.cos.t/; sin.t//. Clearly the function
is 2-periodic. In the graph bellow we set b D 1.
Clear[a, ]
GraphicsGrid[With[feps=105 g,
Function[a,Show[ParametricPlot[(1+a Sec@ )fCos@ ,Sin@ g,
Evaluate[f,#[[1]]+eps,#[[2]]-epsg],PlotStyle!Red]&/@
Partition[Range[0,2,1/2],2,1],PlotRange!ff-2,2g,f-2,2gg,
Ticks!None,PlotLabel!ToExpression[“a”]==a]]/@f0,.03,.3,.5,1,2g]g,
ImageSize!600]
See Fig. 10.1.
ParametricPlot[(1+a Sec@)fCos@ ,Sin@g/.a!0.003,
f ,0,2g,ColorFunction!“Rainbow”,ImageSize!100,
Ticks!ff-1,-.5,.5,1g,f-3,-2,2,3gg]
See Fig. 10.2.

© Springer International Publishing AG 2017 197

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_10
198 10 Miscellany in the Euclidean Plane

Fig. 10.1 Conchoid of Nicomedes, 1

Fig. 10.2 Conchoid of

Nicomedes, 2
10.1 Some Planar Curves 199

Fig. 10.3 Conchoid of Nicomedes, 3

Below we want to see how a point describes this curve.

With[fa=0.03g,
point[ ]:=(1+a Sec[])fCos[],Sin[]g;
conchoid=ParametricPlot[point[],f ,0,2g,
ColorFunction!”Rainbow”,ImageSize!100,
Ticks!ff-1,1g,f-3,-2,2,3gg]];

Animate[
Show[conchoid,Graphics[ffPointSize[.07],Red,Point[point[ ]]gg],
ImageSize!75],f,0,2g,AnimationRate!.025,
AnimationRunning!False,SaveDefinitions!True]
See Fig. 10.3.
200 10 Miscellany in the Euclidean Plane

Fig. 10.4 Conchoid of

Nicomedes, 4

Now we show the evolution of the conchoid under the assumption that the
parameter a varies. We set a 2 Œ2; 2.
Clear[a]
conchoidofNicomedes[a ]:=Module[fg,
Show[ParametricPlot[(1+a Sec[])fCos[ ],Sin[]g,f,0,2 Pig,
PlotStyle!Red,PlotRange!ff-3,3g,f-3,3gg,ImageSize!200,
Ticks!None]]]

Manipulate[Quiet@conchoidofNicomedes[a],
ffa,-.46,”a”g,-2,2,.00001,Appearance!”Labelled”g,
SaveDefinitions!True]
See Fig. 10.4.

10.1.2 Cycloid

A cycloid is the curve described by a point P on the circumference of the circle of

radius a.> 0/ which rolls on a straight line. If the straight line is the horizontal real
axis, the parametric equations of P are given by t ! a.t  sin.t/; 1  cos.t//.
10.1 Some Planar Curves 201

Fig. 10.5 Cycloid, 1

Fig. 10.6 Cycloid, 2

Clear[a,x]
x[t ]:=aft-Sin[t],1-Cos[t]g
ParametricPlot[x[t]/.a!3,ft,0,#g,ImageSize!f140,140g,
PlotStyle!Red,Ticks!fNoneg]&/@(f1/2,1,3/2,2g)
See Fig. 10.5.
The visualization of the geometric definition of the cycloid follows.
cycloid[tend ]:=
Module[fa=3,x,y,t,xx,yy,gridlinex,gridlinesyg,
x[t ]:=a(t-Sin[t]); y[t ]:=a(1-Cos[t]);
gridlinesx=Table[fxx,GrayLevel[.5]g,fxx,0,4a,a/4g];
gridlinesy=Table[fyy,GrayLevel[.5]g,fyy,0,2a,a/2g];
ParametricPlot[fx[t],y[t]g,ft,0,tendg,
PlotStyle!fRed,Thickg,
Epilog!fThick,Blue,Circle[aftend,1g,a],
Red,PointSize[Large],Point[fx[tend],y[tend]g],
Black,PointSize[Medium],Point[aftend,1g]g,
GridLines!fgridlinesx,gridlinesyg,
Ticks!faRange[4],af1,2gg,PlotRange!2aff0,2g,f0,1gg]]

Manipulate[Quiet@cycloid[tend],ftend,.0001,4,
Appearance!”Labeled”g,SaveDefinitions!True]
See Fig. 10.6.
202 10 Miscellany in the Euclidean Plane

10.1.3 Epicycloid

An epicycloid is the curve described by a point P fixed on the circumference of the

circle of radius b.> 0/ which rolls on the outside of the circle of radius a.> 0/.
Its parametric equations are
(
x.t/ D .a C b/ cos t  b cos..a=b C 1/t/
y.t/ D .a C b/ sin t  b sin..a=b C 1/t/;

where a and b are positive parameters.

For a D 8 and b D 5, one has the next picture.
Clear[x,y,a,b]
x[t ,p :1]:=(a+b)Cos[t]-p b Cos[(a/b+1)t]
y[t ,p :1]:=(a+b)Sin[t]-p b Sin[(a/b+1)t]
ParametricPlot[fx[t],y[t]g/.fa!8,b!5g,ft,0,#g,
Ticks!ff-15,15g,f-15,15gg]&/@(2f1,2,5,10g)
See Fig. 10.7.

Fig. 10.7 Epicycloid, 1

10.1 Some Planar Curves 203

The visualization of the geometric definition of the epicycloid follows.

epicycloid[tend ]:=
Module[fa=3,b=1,x,y,p,xx,gridlinesg,
x[t ,p :1]:=(a+b)Cos[t]-p b Cos[(a+b)t/b];
y[t ,p :1]:=(a+b)Sin[t]-p b Sin[(a+b)t/b];
gridlines=Table[fxx,GrayLevel[0.5]g,fxx,-(a+2b),a+2b,(a+2b)/8g];
ParametricPlot[fx[t],y[t]g,ft,0,tendg,PlotStyle!fRed,Thickg,
ImageSize!300,Epilog!fThick,Blue,Circle[f0,0g,a],Black,
Circle[fx[tend,0],y[tend,0]g,b],Red,PointSize[Large],
Point[fx[tend],y[tend]g],Black,Line[ffx[tend,0],y[tend,0]g,
fx[tend],y[tend]gg]g,GridLines!fgridlines,gridlinesg,
PlotRange!(a+b)(1+1/5)f-1,1.1g] ]

Manipulate[Quiet@epicycloid[tend],ftend,0.001,2,
Appearance!”Labeled”g,SaveDefinitions!True]
See Fig. 10.8.

Fig. 10.8 Epicycloid, 2

204 10 Miscellany in the Euclidean Plane

The proof of the parametric equations of the epicycloid is given here.

Clear[a,b,x,y,p]
x[t ,p :1]:=(a+b)Cos[t]-p b Cos[(a+b)t/b]
y[t ,p :1]:=(a+b)Sin[t]-p b Sin[(a+b)t/b]
With[fa=3,b=1,t=/6g,
Show[ParametricPlot[fx[tt],y[tt]g,ftt,0,tg,Axes!None,
PlotStyle!fRed,Thickg],
Graphics[fCircle[f0,0g,a],fBlue,Circle[fx[t,0],y[t,0]g,b]g,
Line[fff0,0g,fx[t,0],y[t,0]gg,fafCos[t],Sin[t]g,afCos[t],0gg,
ffx[t,0],y[t,0]g,fx[t,0],0gg,ffx[t,0],y[t,0]g,fx[t],y[t]gg,
ffx[t],y[t]g,fx[t],0gg,ff0,0g,fa+b,0gg,ff0,0g,f0,a+b/2gg,
ffx[t,0],y[t]g,fx[t],y[t]ggg],
fText[Style[O,Italic,12],f0,-.3g],Text[Style[P,Italic,12],fx[t]+.35,y[t]g],
Text[Style[C,Italic,12],fx[t,0]+.3,y[t,0]g],
Text[Style[D,Italic,12],fa Cos[t],-.3g],Text[Style[T,Italic,12],fx[t],-.3g],
Text[Style[B,Italic,12],fx[t,0],-.3g],Text[Style[S,Italic,12],fa+.2,.2g],
Text[Style[R,Italic,12],fx[t,0]-.22,y[t]+.1g],
Text[Style[Q,Italic,12],fa Cos[t]-.26,a Sin[t]g]g,
PointSize[Medium],Point[ff0,0g,afCos[t],Sin[t]g,fx[t,0],y[t,0]g,
afCos[t],0g,fx[t,0],0g,fx[t],0g,fx[t,0],y[t]g,fa,0gg],
Red,Point[fx[t],y[t]g]g],
PlotRange!All,ImageSize!300]]
See Fig. 10.9.
Data: OQ D OS D a; CQ D CP D b, angle(DOQ) = t, arc(SQ) = at D arc(QP),
OB D .a C b/ cos t; CD D .a C b/ sin t; DQ D a sin t; OD D a cos t; angle(QCP) D
at=b; angle(OCB) = =2  t; angle(BCP) = at=b C t  =2:

Fig. 10.9 Epicycloid, 3

10.1 Some Planar Curves 205

Particular cases of the epicycloid are a D b when a cardioid is obtained and

a D 2b when a nephroid is obtained.

10.1.3.1 Cardioid

Geometrically, a cardioid is an epicycloid with cycles of equal radius. Therefore,

the parametric equations of a cardioid are
Clear[x,y,a];
x[t ,p :1]:=2a Cos[t]-p a Cos[2t];
y[t ,p :1]:=2a Sin[t]-p a Sin[2t];
ParametricPlot[fx[t],y[t]g/.fa!3g,ft,0,#g,Ticks!ff-8,4.5g,f-5,8gg,
ImageSize!f150,150g]&/@(f1/2,1,3/2,2g)
See Fig. 10.10.

Fig. 10.10 Cardioid

206 10 Miscellany in the Euclidean Plane

Fig. 10.11 Nephroid

10.1.3.2 Nephroid

As we mentioned earlier, geometrically a nephroid is an epicycloid with the radius

of the rolling circle equal to half of the radius of the fixed cycles. Therefore, the
parametric equations of a nephroid are
Clear[x,y,b]
x[t ,p :1]:=2b Cos[t]-p b Cos[3t]
y[t ,p :1]:=2b Sin[t]-p b Sin[3t]
ParametricPlot[fx[t],y[t]g/.fb!3g,ft,0,#g,Ticks!ff-8,4.5g,f-8,8gg,
ImageSize!f150,150g]&/@(f1/2,1,3/2,2g)
See Fig. 10.11.

10.1.3.3 Nephroid of Freeth

The polar equation of the Freeth’s nephroid is r D a.1 C 2 sin.=2//, [19, 6.8,
pp.175–178] or according to http://www-groups.dcs.st-and.ac.uk/~history/Curves/
Freeths.html. Below we use its parametric equations.
10.1 Some Planar Curves 207

Fig. 10.12 Nephroid of Freeth

ParametricPlot[(1+2 Sin[t/2])fCos[t],Sin[t]g,#,Ticks!None,
ImageSize!150]&/@fft,-2,0g,ft,0,2g,ft,-2,2gg
See Fig. 10.12.

10.1.4 Hypocycloid

A hypocycloid is the curve described by a point P fixed on the circumference of the

circle of radius b which rolls on the inside of the circle of a radius.
Its parametric equations are of the form
Clear[x,y,a,b]
x[t ]:=(a-b)Cos[t]+b Cos[(a/b-1)t]
y[t ]:=(a-b)Sin[t]-b Sin[(a/b-1)t]
where a and b are parameters. For a D 8 and b D 5; we have the next graphs.
ParametricPlot[fx[t],y[t]g/.fa!8,b!5g,ft,0,#g,ImageSize!150,
Ticks!ff-15,15g,f-15,15gg]&/@(f2,4,5,10g)
See Fig. 10.13.
The above given parametric equations of the hypocycloid are sometimes given
under the form
xx[t ]:=(a-b)Cos[t]+b Cos[t(k-1)]
yy[t ]:=(a-b)Sin[t]-b Sin[t(k-1)]
b:=a/k; (* k is a parameter *)
Changing the parameters a and k, the resulting figures change drastically.
With[fa=1.g,
ParametricPlot[f(a-a/#)Cos[t]+(a/#)Cos[t(#-1)],
(a-a/#)Sin[t]-(a/#)Sin[t(#-1)]g,ft,0,10g,ImageSize!150]&/@f.25,.33,
.5,1.3,1.4,1.6,1.8g]
See Fig. 10.14.
208 10 Miscellany in the Euclidean Plane

Fig. 10.13 Hypocycloid, 1

The visualization of the geometric definition of the hypocycloid follows.

hypocycloid[tend ]:=
Module[fa=4,b=1,x,y,p,xx,gridlinesg,
x[t ,p ]:=b(a/b-1)Cos[t]+b p Cos[(a/b-1)t];
y[t ,p ]:=b(a/b-1)Sin[t]-b p Sin[(a/b-1)t];
gridlines=Table[fxx,GrayLevel[0.5]g,fxx,-a,a,a/5g];
ParametricPlot[fx[t,1],y[t,1]g,ft,0,tendg,PlotStyle!fRed,Thickg,
Ticks!None,ImageSize!225,Epilog!fThick,Blue,Circle[f0,0g,a],
Black,Circle[fx[tend,0],y[tend,0]g,b],Red,PointSize[Large],
Point[fx[tend,1],y[tend,1]g],Black,Line[ffx[tend,0],y[tend,0]g,
fx[tend,1],y[tend,1]gg]g,GridLines!fgridlines,gridlinesg,
PlotRange!(1+1/20)f-a,ag]]

Manipulate[Quiet@hypocycloid[tend],ftend,.0001,2,
Appearance!”Labeled”g,SaveDefinitions!True]
See Fig. 10.15.
10.1 Some Planar Curves 209

Fig. 10.14 Hypocycloid, 2

The proof of the parametric equations of the hypocycloid is given here.

Clear[a,b,x,y]
With[fa=5,b=1,tend=/6g,
x[t ,p :1]:=b(a/b-1)Cos[t]+p b Cos[b(a/b-1)t];
y[t ,p :1]:=b(a/b-1)Sin[t]-p b Sin[b(a/b-1)t];
Show[ParametricPlot[fx[t],y[t]g,ft,0,tendg,Axes!None,
PlotStyle!fRed,Thickg],
Graphics[fCircle[f0,0g,a],fBlue,Circle[fx[tend,0],y[tend,0]g,b]g,
210 10 Miscellany in the Euclidean Plane

Fig. 10.15 Hypocycloid, 3

Line[fff0,0g,afCos[tend],Sin[tend]gg,
ffx[tend,0],y[tend,0]g,fx[tend,0],0gg,
ffx[tend,0],y[tend,0]g,fx[tend],y[tend]gg,
ffx[tend],y[tend]g,fx[tend],0gg,ff0,0g,fa,0gg,
ffx[tend,0],y[tend]g,fx[tend],y[tend]ggg],
fText[Style[O,Italic,12],f0,-.4g],
Text[Style[P,Italic,12],fx[tend]-.33,y[tend]-.25g],
Text[Style[C,Italic,12],fx[tend,0],y[tend,0]+.3g],
Text[Style[T,Italic,12],fx[tend],-.4g],
Text[Style[B,Italic,12],fx[tend,0],-.4g],
Text[Style[S,Italic,12],fa+.35,.15g],
Text[Style[R,Italic,12],fx[tend,0]+.28,y[tend]+.2g],
Text[Style[Q,Italic,12],fa Cos[tend]+.35,a Sin[tend]+.1g]g,
fPointSize[Medium],Point[ff0,0g,afCos[tend],Sin[tend]g,
fx[tend,0],y[tend,0]g,fx[tend,0],0g,fx[tend],0g,
fx[tend,0],y[tend]g,fa,0gg],Red,Point[fx[tend],y[tend]g]gg],
PlotRange!All,ImageSize!300]]
See Fig. 10.16.
Data: OQ D OS D a; CQ D CP D b, angle(SOQ) = t, arc(SQ) = at; arc(QP),
OB D .a  b/ cos t; CB = .a  b/ sin t; Q = a.sin t; cos t/; angle(QCP) D at=b;
angle(OCB) = =2  t; angle(BCP) = at=b  t  =2: 
10.1 Some Planar Curves 211

Fig. 10.16 Hypocycloid, 4

Fig. 10.17 Deltoid

10.1.4.1 Deltoid

A deltoid is a particular hypocycloid with a=b D 3. We show how it looks like.

Clear[x,y,a,b]
x[t ,p :1]:=(a-b)Cos[t]+p b Cos[(a/b-1)t]
y[t ,p :1]:=(a-b)Sin[t]-p b Sin[(a/b-1)t]
ParametricPlot[fx[t],y[t]g/.fa!6,b!2g,ft,0,2g,
Ticks!ff-2,6g,f-5,5gg,PlotStyle!Red,ImageSize!150]
See Fig. 10.17.
212 10 Miscellany in the Euclidean Plane

10.1.5 Maria Agnesi’s Curve

Maria Agnesi’s curve is obtained starting with a fixed circle, and a point O on the
circle is chosen. For any other point A on the circle, the secant line OA is drawn.
The point M is diametrically opposite to O. The line OA intersects the tangent of
M at the point N: The line parallel to OM through N and the line perpendicular
to OM through A intersect at P: As the point A varies, the path of P is the Maria
Agnesi’s curve. We consider O D .0; 0/, M D .0; 2a/, and a circle of diameter OM
of equation x2 C .y  a/2 D a2 . Let A be an arbitrary point on this circle. Then
A D a.sin.2t/, 1  cos.2a//, N D a.2ctan.t/; 2/, and P D a.2ctan.t/; 1  cos.2t//.
Thus the parametric equations of the curve are x.t/ D 2a cot t; y.t/ D a.1cos.2t//,
a > 0:
With[fa=6,eps=.3g,
ParametricPlot[f2a Cot[t], a(1-Cos[2t])g,ft,eps,-epsg,
PlotStyle!Red,Ticks!ff-30,30g,fa,2agg,Prolog!ffDashing[f.01g],
Blue,Circle[f0,ag,a]gg,ImageSize!500]]
See Fig. 10.18.
We can generate Maria Agnesi’s curve with the next code. The circle is
x2 C .y  a/2 D a2 ; where a is a real positive constant. We consider the parameter
m varying on the interval Œmend; Cmend; (mend > 0).
Animate[
With[fa=4,mend=15g,
Show[ParametricPlot[fmm,8 a3 /(4 a2 +mm2 )g,
fmm,-mend-0.00001,mg,Ticks!ff-mend,mendg,fagg,
PlotStyle!Red,AxesOrigin!f0,0g], Graphics[fBlue,Circle[f0,ag,a],
Line[fff-mend,2ag,fmend,2agg,ff0,0g,fm,2agg,
ff4a2 m/(4a2 +m2 /,8a3 /(4a2 +m2 )g,fm,8a3 /(4a2 +m2 )gg,
ffm,0g,fm,2aggg],
PointSize[.015],Black,
Point[ff0,0g,fm,2ag,fm,0g,f4a2 m/(4a2 +m2 ),8a3 /(4a2 +m2 )gg],
Red,Point[ffm,8a3 /(4a2 +m2 )gg],
Text[Style[M,Italic,12],f0,2.2ag],Text[Style[”0”,Italic,12],f0,-1.2g],
Text[Style[N,Italic,12],fm,2.2ag],
Text[Style[P,Italic,12],fm+a/4,8a3 /(4a2 +m2 )-a/5g],g],
PlotRange!All]],fm,-15,15g,
AnimationRunning!False,AnimationRate!.3]
See Fig. 10.19.

Fig. 10.18 Maria Agnesi’s curve

10.1 Some Planar Curves 213

Fig. 10.19 Maria Agnesi’s curve, animation

10.1.6 Cassini Ovals

The Cassini ovals are described by a point such that the product of its distances
from two fixed points (called foci) a distance 2a apart is a constant b2 . Immediately
the Cartesian equation follows, that is,
  
.x  a/2 C y2 .x C a/2 C y2 D b2 ; a; b > 0:

The polar equation follows immediately

r q
r D a cos.2t/ ˙ .b=a/4  sin2 .2t/:

We note that generally there are two arcs that form the ovals.
bmax=2;
cassini[b ]:=Module[fa=1g,
Show[PolarPlot[Sqrt[a2 (Cos[2t]+Sqrt[(b/a)4 -Sin[2t]2 ])],ft,0,2g,
PlotStyle!Red],
PolarPlot[Sqrt[a2 (Cos[2t]-Sqrt[(b/a)4 -Sin[2t]2 ])],ft,0,2g,
PlotStyle!Blue],Ticks!fAutomatic,Noneg,
ImageSize!225,PlotRange!All]];

Manipulate[Quiet@cassini[b],
ffb,1g,0,bmax,0.001,Appearance!”Labeled”g,SaveDefinitions!True]
See Fig. 10.20.
A Cassini oval with a D b is said to be a Bernoulli lemniscata.
214 10 Miscellany in the Euclidean Plane

Fig. 10.20 Cassini ovals

10.1.7 Orthoptics

In plane geometry, an orthoptic is the set of points for which two tangents of a given
curve meet at a right angle.
Example It is well-known that the orthoptics of
2 2
(i) an ellipse ax2 C by2 D 1 is the director (Monge) circle x2 C y2 D a2 C b2 ,
a > 0, b > 0;
2 2
(ii) a hyperbola ax2  by2 D 1 is the (Monge) circle x2 C y2 D a2  b2 ,
0 < b < a (in caseıof a  b there are no orthogonal tangents);
(iii) a parabola y D x2 .2p/, p > 0, is its directrix;
2=3 2=3
(iv) an astroidp x C y D 1 is a quadrifolium with the polar equation
r D .1= 2/ cos.2t/, 0  t < 2.
Below we introduce some codes for these orthoptics.

10.1.7.1 Orthoptic of an Ellipse

2 2
We consider the ellipse 3x2 C 2y2 D 1 and exhibit its Monge circle and the two
orthogonal tangents from each point in the orthoptic circle.
Clear[˛,m,x]

mongeEllipse[˛ ]:=
Module[fa=3,b=2,c,m,m1,m2,t,eps=0.0001,x,x0,y0,x1,y1g,
c=Sqrt[a2 +b2 ];pointA=cfCos[˛],Sin[˛]g;
10.1 Some Planar Curves 215

ellipse=ParametricPlot[fa Cos[t],b Sin[t]g,ft,0,2g,PlotStyle!Blue];

circleMonge=ParametricPlot[cfCos[t],Sin[t]g,ft,0,2g,PlotStyle!Red];
If[Abs[ArcTan[b/a]-˛]<eps,pointB=fa,0g;pointC=f0,bg;,
If[Abs[-ArcTan[b/a]-˛]<eps,pointB=f0,bg;pointC=f-a,0g;,
If[Abs[+ArcTan[b/a]-˛]<eps,pointB=f-a,0g;pointC=f0,-bg;,
If[Abs[2-ArcTan[b/a]-˛]<eps,pointB=f0,-bg;pointC=fa,0g;,
msol=Solve[(c2 Cos[˛]2 -a2 )m2 -(c2 Sin[2˛])m+c2 Sin[˛]2 -b2 ==0,m];
fm1,m2g=fRe[msol[[1,1,2]]],Re[msol[[2,1,2]]]g;
sol1=Solve[x2 /a2+(c Sin[˛]+m1(x-c Cos[˛]))2 /b2 -1==0,x];
x0=Re[sol1[[1,1,2]]];y0=c Sin[˛]+m1(x0-c Cos[˛]);
pointB=fx0,y0g;
sol2=Solve[x2 /a2 +(c Sin[˛]+m2(x-c Cos[˛]))2 /b2 -1==0,x];
x1=Re[sol2[[1,1,2]]];y1=c Sin[˛]+m2(x1-c Cos[˛]);
pointC=fx1,y1g;];];];];
Show[ellipse,circleMonge,
Graphics[ffRed,Thickness!0.008,Line[fpointC,pointA,pointBg]g,
Line[ffa,bg,f-a,bg,f-a,-bg,fa,-bg,fa,bgg],
fPointSize[.018],Point[fpointA,pointB,pointCg]gg],
PlotRange!All,Ticks!ff-4,-3,3,4g,f-4,-2,2,4gg,ImageSize!300]]

Manipulate[
Quiet@mongeEllipse[˛],
ff˛,0,”˛”g,0,2,.0001,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 10.21.

10.1.7.2 Orthoptic of a Hyperbola

2 2
We consider the hyperbola 3x2  2y2 D 1 and exhibit its Monge circle and the two
orthogonal tangents from each point in the orthoptic circle. We emphasize that here
it is necessary that a > b > 0:
Clear[˛]

mongeHyperbola[˛ ]:=
Module[fa=3,b=2,c,m,m1,m2,t,eps=0.01,x,x0,y0,x1,y1,pointA,pointB,
pointCg,c=Sqrt[a2 -b2 ];pointA=cfCos[˛],Sin[˛]g;
hyperbola=ParametricPlot[ffa Cosh[t],b Sinh[t]g,
f-a Cosh[t],b Sinh[t]gg,ft,-/2,/2g,PlotStyle!fBlueg];
circleMonge=ParametricPlot[cfCos[t],Sin[t]g,ft,0,2g,
PlotStyle!fRedg];
216 10 Miscellany in the Euclidean Plane

Fig. 10.21 Orthoptic curve of a ellipse

If[Abs[ArcTan[b/a]-˛]<eps,pointB=fa,0g;pointC=f0,bg;,
If[Abs[-ArcTan[b/a]-˛]<eps,pointB=f0,bg;pointC=f-a,0g;,
If[Abs[+ArcTan[b/a]-˛]<eps,pointB=f-a,0g;pointC=f0,-bg;,
If[Abs[2-ArcTan[b/a]-˛]<eps,pointB=f0,-bg;pointC=fa,0g;,
msol=Solve[(c2 Cos[˛]2 -a2 )m2 -(c2 Sin[2˛])m+c2 Sin[˛]2 +b2 ==0,m];
fm1,m2g=fRe[msol[[1,1,2]]],Re[msol[[2,1,2]]]g;
sol1=Solve[x2 /a2 -(c Sin[˛]+m1(x-c Cos[˛]))2 /b2 -1==0,x];
x0=Re[sol1[[1,1,2]]];y0=c Sin[˛]+m1(x0-c Cos[˛]);
pointB=fx0,y0g;
sol2=Solve[x2 /a2 -(c Sin[˛]+m2(x-c Cos[˛]))2 /b2 -1==0,x];
x1=Re[sol2[[1,1,2]]];y1=c Sin[˛]+m2(x1-c Cos[˛]);
pointC=fx1,y1g;];];];];
Show[hyperbola,circleMonge,
Graphics[ffRed,Thickness!0.008,Line[fpointC,pointA,pointBg]g,
fPointSize[.018],Point[fpointA,pointB,pointCg]gg],
PlotRange!All,Ticks!ff-3,3g,f-2,2gg,ImageSize!300]]

Manipulate[
Quiet@mongeHyperbola[˛],
ff˛,0,”˛”g,-,,.0001,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 10.22.
10.1 Some Planar Curves 217

Fig. 10.22 Orthoptic curve

of a hyperbola

10.1.7.3 Orthoptic of a Parabola

Proposition 10.1 If two tangents to a parabola are perpendicular to each other,

then they intersect on the directrix. Conversely, two tangents which intersect on the
directrix are perpendicular.
Proof Let x2 D 2py be a parabola where p is a positive constant. Suppose that two
tangents contact this parabola at the points .a; a2 =2p/ and .b; b2 =2p/. The equation
of the first tangent is y  a2 =2p D .a=p/.x  a/, whereas the equation of the second
tangent is y  b2 =2p D .b=p/.x  b/: From these two equations and the condition of
orthogonality, we get the equations x D .a C b/=2 and y D p=2. The last equation
is the equation of the directrix of the given parabola. t
u
The converse statement is similarly proven.
Clear[u]

mongeParabola[u ]:=
Module[fp=2,xg,
pointA=fu,-p/2g;
parabola=ParametricPlot[fx,x2 /(2p)g,fx,-6.65,6.65g,
PlotStyle!fBlueg];
circleMonge=ParametricPlot[fx,-p/2g,fx,-6,6g,PlotStyle!fRedg];
pointB=fu,-p/2g;pointA=fu-Sqrt[u2 +p2 ],(u-Sqrt[u2 +p2 ])2 /(2p)g;
pointC=fu+Sqrt[u2 +p2 ],(u+Sqrt[u2 +p2 ])2 /(2p)g;
Show[parabola,circleMonge,
Graphics[ffRed,Thickness!0.008,Line[fpointA,pointB,pointCg]g,
fPointSize[.018],Point[fpointA,pointB,pointCg]gg],PlotRange!All,
Ticks!fpf-3/2,3/2g,pf-1/2,1/2,1,2,4gg,ImageSize!200]]
218 10 Miscellany in the Euclidean Plane

Fig. 10.23 Orthoptic curve

of a parabola

Manipulate[
Quiet@mongeParabola[u],
ffu,0,”parameter”g,-3,3,.0001,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 10.23.

10.1.7.4 Orthoptic of an Astroid

An astroid can be described by the next parametric representation

 
a.t/ D cos3 t; sin3 t ; 0  t  2:

The tangent vector is a0 .t/. Then the condition of orthogonality is written

a .t/a0 .t C  /= 0, i.e., cos.t/ D 0: Immediately one has  D .2k C 1/=2, k 2 Z;
0

we set  D =2. Thus the equations of the orthoptic of an astroid result by solving
the system of equations
(  
y  sin3 t D tan.t/ x  cos3 t
 
y  cos3 t D  cot.t/ x C sin3 t :

Then we get the Cartesian equations of the orthoptic of an astroid is

(
x.t/ D sin t cos t.sin t  cos t/
y.t/ D sin t cos t.sin t C cos t/:
10.1 Some Planar Curves 219

This curve is said to be a quadrifolium.

Clear[a]
mongeAstroid[˛ ]:=
Module[fa,tg,
a[t ]:=fCos[t]3 ,Sin[t]3 g;
qfolium[t ]:=fSin[t]Cos[t](Sin[t]-Cos[t]),Sin[t]Cos[t](Sin[t]+Cos[t])g;
astroid=ParametricPlot[a[t],ft,0,2g,PlotStyle!Blue];
circleMonge=ParametricPlot[qfolium[t],ft,0,2g,PlotStyle!Red];
Show[astroid,circleMonge,
Graphics[ffRed,Thickness!0.008,Line[fqfolium[˛],a[˛ C /2]g],
Green,Line[fa[˛],qfolium[˛]g]g,
fPointSize[.018],Point[fa[˛],qfolium[˛],a[˛ C /2]g]gg],
PlotRange!All,Ticks!ff-1,1g,f-1,1gg,ImageSize!250]]

Manipulate[
Quiet@mongeAstroid[˛],
ff˛,0,”˛”g,0,2,.0001,Appearance!”Labeled”g,
SynchronousUpdating!False,SaveDefinitions!True]
See Fig. 10.24.

Fig. 10.24 Orthoptic curve

of an astroid
220 10 Miscellany in the Euclidean Plane

10.1.8 Pedals of Some Curves

The pedal curve of a curve and a fixed point is the locus of the orthogonal projection
of the fixed point on the tangent lines of the curve.
Our first example is the pedal curve of a circle. This pedal curve is a cardioid. Its
shape depends on the distance of the point to the center of the circle.
Clear[a,b,˛]
pedalofCircle[a ,b ,˛ ]:= (* a is the radius of the circle; we suppose that
it is centered in the origin of the axes *)
(* b is the distance of the fixed point from the center of the circle *)
Module[fx,y,t,solg,
sol=Solve[a Cos[t]x+a Sin[t]y==a2 &&Cos[t] y-Sin[t] x==-b Sin[t],fx,yg];
sol=Flatten[FullSimplify[sol]];
Show[ParametricPlot[fx,yg/.sol,ft,0,2g,PlotStyle!Green],
Graphics[ffBlack,Circle[f0,0g,a],
PointSize[.02],Point[ff0,0g,fb,0g,afCos[˛],Sin[˛]gg]g,
fBlue,Thickness[.007],
Line[ffa Cos[˛],a Sin[˛]g,fa Cos[˛]+b Sin[˛]2 ,(a-b Cos[˛])Sin[˛]gg]g,
fMagenta,PointSize[.03],
Point[fa Cos[˛]+b Sin[˛]2 ,(a-b Cos[˛])Sin[˛]g]g,
fRed,Thickness[.007],
Line[ffb,0g,fa Cos[˛]+b Sin[˛]2 ,(a-b Cos[˛])Sin[˛]gg]gg],
Axes!False,ImageSize!220,PlotRange!All]]

Manipulate[
Quiet@pedalofCircle[a,b,˛],
ffa,1,ag,1,4,0.01,Appearance!”Labeled”g,
ffb,1.5,bg,1,4,0.01,Appearance!”Labeled”g,
ff˛,0,”˛”g,0,2,0.01,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!True]
See Fig. 10.25.
The second example is the pedal curve of an astroid. The parametric equations
of the astroid are .a cos3 t; b sin3 t/ and the fixed point is .xp; yp/.
Clear[a,b,xp,yp,˛]
pedalofAstroid[a ,b ,xp ,yp ,˛ ]:=
Module[fx,y,t,solg, (* a is the horizontal semiaxis of the astroid;
we suppose that is centered in the origin of the axes *)
(* b is the vertical semiaxis of the astroid *)
(* xp is the abscisa of the fixed point and yp is the ordinate of it *)
sol=Flatten[Solve[b Sin[t] x+a Cos[t] y-a b Sin[t] Cos[t]==0&&
a Cos[t] x-b Sin[t] y-a Cos[t] xp +b Sin[t] yp==0,fx,yg]];
10.1 Some Planar Curves 221

Fig. 10.25 Pedal curve of a

circle

Show[ParametricPlot[ffa Cos[t]3 ,b Sin[t]3 g,fx,yg/.solg,ft,0,2g,

PlotStyle!fBlack,Greeng],
Graphics[ffBlack,PointSize[.02],
Point[ff0,0g,fxp,ypg,fa Cos[˛]3 ,b Sin[˛]3 gg]g,
fBlue,Thickness[.007],
Line[ffa Cos[˛]3 ,b Sin[˛]3 g,f-((-a2 xp Cos[˛]2 +a b yp Cos[˛]Sin[˛]-
a b2 Cos[˛] Sin[˛]2 )/(a2 Cos[˛]2 +b2 Sin[˛]2 )),
-((a b xp Cos[˛] Sin[˛]-a2 b Cos[˛]2 Sin[˛]-b2 yp Sin[˛]2 )/
(a2 Cos[˛]2 +b2 Sin[˛]2 ))gg]g,
fMagenta,PointSize[.03],
Point[f-((-a2 xp Cos[˛]2 +a b yp Cos[˛] Sin[˛]-a b2 Cos[˛]Sin[˛]2 )/
(a2 Cos[˛]2 +b2 Sin[˛]2 )),
-((a b xp Cos[˛] Sin[˛]-a2 b Cos[˛]2 Sin[˛]-b2 yp Sin[˛]2 )/
(a2 Cos[˛]2 +b2 Sin[˛]2 ))g]g,
fRed,Thickness[.007],
222 10 Miscellany in the Euclidean Plane

Line[ffxp,ypg,f-((-a2 xp Cos[˛]2 +a b yp Cos[˛]Sin[˛] -a b2 Cos[˛]

Sin[˛]2 )/(a2 Cos[˛]2 +b2 Sin[˛]2 )),
-((a b xpCos[˛]Sin[˛]-a2 b Cos[˛]2 Sin[˛]-b2 yp Sin[˛]2 ) /
(a2 Cos[˛]2 +b2 Sin[˛]2 ))gg]gg],
Axes!False,ImageSize!220,PlotRange!All]]

Manipulate[
Quiet@pedalofAstroid[a,b,xp,yp,˛],
ffa,1,”a”g,1,3,0.01,Appearance!”Labeled”g,
ffb,1,”b”g,1,3,0.01,Appearance!”Labeled”g,
ffxp,0,”xp”g,0,6,0.01,Appearance!”Labeled”g,
ffyp,0,”yp”g,0,6,0.01,Appearance!”Labeled”g,
ff˛,0,”˛”g,0,2,0.01,Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!True]
See Fig. 10.26.

Fig. 10.26 Pedal curve of an

astroid
10.2 Attractors 223

10.2 Attractors

10.2.1 Hénon Attractor

The Hénon map is a discrete-time dynamical system, [62, 12.2]. It is one of the
most studied examples of dynamical systems that exhibit a chaotic behavior. The
Hénon map takes a point .xn , yn / in the plane and maps it to a new point according
to the recurrences
(
xnC1 D 1  axn 2 C yn ;
ynC1 D bxn :

The map depends on two real parameters, a and b, which for the classical Hénon
map have values of a D 1:4 and b D 0:3. For the classical values, the Hénon map
is chaotic. For other values of a and b, the map may be chaotic, intermittent, or
converge to a periodic orbit. Below we show the behavior of the Hénon map for
a D 1:4 and b D 0:3.
For[n=2000;x=1;y=1;a=1.4;b=0.3;points=ffx,ygg;
k=0,kn,k++,u=1+y-a x2 ;v=b x;
AppendTo[points,fu,vg];x=u;y=v]
Show[ListPlot[points,PlotStyle!fRed,PointSize[0.005]g,
Axes!False,ImageSize!250],Graphics[ffPointSize[0.012],
Green,Point[First[points]],
PointSize[0.02],Blue,Point[Last[points]]g,
Text[Style[”initial point”,Italic,Darker@Green,12],First[points]+f0,-.1g],
Text[Style[”final point”,Italic,Blue,12],Last[points]+f-.10,.15g]g]]

See Fig. 10.27.

We study if the Hénon system has an equilibrium point.
Clear[x,y]
With[fa=1.4,b=0.3g,
Solve[x==1+y-a x2 &&y==b x,fx,yg,Reals]]

Fig. 10.27 The Hénon

attractor
224 10 Miscellany in the Euclidean Plane

Solve::ratnz: Solve was unable to solve the system with inexact

coefficients. The answer obtained by solving a corresponding exact system
and numericizing the result. 
ffx ! 1:13135; y ! 0:339406g; fx ! 0:631354; y ! 0:189406gg
Thus, there are two pairs of real equilibrium points. We test that a point of these
is indeed an equilibrium point.
Clear[x,y]
With[fa=1.4,b=0.3g,
s=Solve[x==1+y-a x2 &&y==b x,fx,yg,Reals];
x=x/.s[[1]];y=y/.s[[1]];points=ffx,ygg; (* Initial point *)
For[n=2000;k=0,kn,k++,u=1+y-a x2 ;v=b x;
AppendTo[points,fu,vg];x=u;y=v]
Show[ListPlot[points,PlotStyle!fRed,PointSize[0.005]g,
Axes!False,ImageSize!200],
Graphics[ffPointSize[0.012],Green,Point[First[points]],
PointSize[0.02],Blue,Point[Last[points]]g,
Text[Style[”initial point”,Italic,Darker@Green,12],First[points]+f0,-.1g],
Text[Style[”final point”,Italic,Blue,12],Last[points]+f-.10,.15g]g]]]
Solve::ratnz: Solve was unable to solve the system with inexact
coefficients. The answer obtained by solving a corresponding exact system
and numericizing the result. 
See Fig. 10.28.
It is interesting to consider the Hénon attractors in R3 . Then we get that
For[n=2000;x=1;y=1;a=1.4;b=0.3;points=ff0,x,ygg;
k=0,kn,k++,u=1+y-a x2 ;v=b x;
AppendTo[points,fk,u,vg];x=u;y=v]
Show[ListPointPlot3D[points,ColorFunction!Function[fx,y,zg,
Hue[x]]],
Graphics3D[ffPointSize[0.015],Green,Point[First[points]],
PointSize[0.025],Hue[n],Point[Last[points]]g,
Text[Style[”initial point”,Italic,Darker@Green,12],
First[points]+f0,0,-.2g],
Text[Style[”final point”,Italic,Hue[n],12],Last[points]+f0,.5,-.2g]g]]]
See Fig. 10.29.

Fig. 10.28 Initial and final

points of a Hénon attractor
10.2 Attractors 225

Fig. 10.29 3D Hénon attractor

10.2.2 Lorenz Attractor

The Lorenz attractor was found by Ed. N. Lorenz around 1963, [62, 9.4]. It was
derived from a simplified model of convection in the Earth’s atmosphere. The
system is most commonly expressed as three coupled nonlinear ordinary differential
equations.
8
ˆ
ˆ dx
D a.y  x/;
< dt
dy
D x.b  z/  y;
ˆ dt
:̂ dz
D xy  cz:
dt

One commonly used set of constants is a D 10; b D 28; c D 8=3. A code using
the Euler’s method of integration of this case follows.
For[n=300;x=0.0001;y=.0001;z=.0001;a=10;b=28;c=8/3;h=0.015;
point1=fx,y,zg;points=fpoint1g;
k=0,k<n,k++,
u=x+h a(y-x);v=y+h(x(b-z)-y);w=z+h(x y-c z);
AppendTo[points,fu,v,wg];
x=u;y=v;z=w]
point2=fx,y,zg;
226 10 Miscellany in the Euclidean Plane

Fig. 10.30 Discret Lorenz attractor

Show[
ListPointPlot3D[points,ColorFunction!”Rainbow”,
PlotStyle!PointSize[0.008],Axes!False],
Graphics3D[ffPointSize[0.018],Green,Point[point1],Red,Point[point2]g,
Text[Style[”initial point”,Italic,Darker@Green,12],point1+f0,0,-7g],
Text[Style[”final point”,Italic,Red,12],point2+f0,0,-9g]g],
ViewPoint!f25,2,6g,ImageSize!400]
See Fig. 10.30.
Here we use the built-in function NDSolve and consider the smooth case of the
Lorenz equations. In this case a good source is
http://reference.wolfram.com/language/example/VisualizeTheLorenzAttractor.
html. Then we color the trajectory in two ways.
Clear[x,y,z]
Module[fa=10,b=28,c=8/3,tend=100,s,tg,
s=NDSolve[fx’[t]==a(y[t]-x[t]),y’[t]==-x[t] z[t]+b x[t]-y[t],
z’[t]==x[t] y[t]-c z[t],x[0]==y[0]==z[0]==0.0001g,fx,y,zg,ft,0,tendg,
MaxSteps! 1];
fParametricPlot3D[Evaluate[fx[t],y[t],z[t]g/.s],ft,0,tendg,
PlotPoints!1000,PlotStyle!Directive[Thick,RGBColor[.8,0,0]],
Boxed!False,Axes!False,ImageSize!275,
ColorFunction!(ColorData[”SolarColors”,#4]&)],
ParametricPlot3D[Evaluate[fx[t],y[t],z[t]g/.s],ft,0,tendg,
PlotPoints!1000,Boxed!False,Axes!False,ImageSize!275,
ColorFunction!(ColorData[”Rainbow”][#4]&)]g
]
10.3 Limit Cycles and Hopf Bifurcation 227

The previous two pictures are all static. Below we consider the constants a; b; and
c parameters. It is also interesting to us the case when the final instant, here denoted
p, is a parameter. This case is interesting because one can see the appearance and
evolution of the chaotic movement.
Clear[a,b,c,p,x,y,z,xsol,ysol,zsol]
lorenzTrajectory[a ,b ,c ,p ]:=
Module[fx,y,z,xsol,ysol,zsolg,
fxsol,ysol,zsolg=NDSolveValue[fx’[t]==a(y[t]-x[t]),y’[t]==x[t](b-z[t])-y[t],
z’[t]==x[t] y[t]-c z[t],x[0]==0.0001,y[0]==0.0001,z[0]==0.0001g,fx,y,zg,
ft,0,pg];
Show[ParametricPlot3D[fxsol[t],ysol[t],zsol[t]g,ft,0,pg,PlotRange!All,
PlotPoints!250,ImageSize!250,Boxed!False,Axes!False,
ColorFunction!(ColorData[”Rainbow”][#4]&)],
Graphics3D[ffPointSize[0.018],Green,Point[fxsol[0],ysol[0],zsol[0]g]g,
Text[Style[”initial point”,Italic,Darker@Green,12],fxsol[0],ysol[0],
zsol[0]g+f0,0,-6g]g]]]

Manipulate[
Quiet@lorenzTrajectory[a,b,c,p],
ffa,10g,9,11,.01,Appearance!”Labeled”g,ffb,28g,6,40,.01,
Appearance!”Labeled”g,
ffc,8/3g,1.5,23,.01,Appearance!”Labeled”g,ffp,18g,0.001,80,.01,
Appearance!”Labeled”g,
SaveDefinitions!True,SynchronousUpdating!False]
See Fig. 10.31.

10.3 Limit Cycles and Hopf Bifurcation

A limit cycle is a planar singular periodic solution of a nonlinear and autonomous

system of ordinary differential equations of the form
(
x0 D f .x; y/;
y0 D g.x; y/;

[52]. We assume that each initial value problem has a solution which is continuable
(
x D x.t/;
y D y.t/:
228 10 Miscellany in the Euclidean Plane

Fig. 10.31 Continuous

Lorenz attractor

10.3.1 Van Der Pol Equation

The Van der Pol differential equation is a second-order nonlinear ordinary differen-
tial equation with nonlinear damping:
 
x00 .t/   1  x2 .t/ x0 .t/ C x.t/ D 0;

where x is the position coordinate depending on the time t and  is a scalar

parameter indicating the nonlinearity and the strength of the damping, [62, pp. 198–
199]. This equation can be written under the form of a planar system of differential
equations of the first order
(
x0 D y;
 
y0 D  1  x2 x0  x:
10.3 Limit Cycles and Hopf Bifurcation 229

Below we show the existence of a limit cycle for  2 Œ0; 5: The green closed
trajectory is the singular periodic orbit. The blue trajectory always remains outside
the periodic orbit, whereas the red orbit always remains inside the periodic orbit.
These two nonperiodic orbits converge to the periodic one.
Clear[]
vanderPol[ ]:= Module[fa1=a3=2,a2=1,b1=3,b2=1,b3=0,x,y,t,sol1,
sol2,sol3,p1,p2,p3,system,point1,point2,point3g,
(* Outer trajectory *)
point1=fa1,b1g;
system=fx’[t]==y[t],y’[t]==(1-x[t]2 )y[t]-x[t],fx[0],y[0]g==point1g;
sol1=NDSolve[system,fx,yg,ft,100g];
p1=ParametricPlot[Evaluate[fx[t],y[t]g/.sol1],ft,0,100g,
PlotStyle!Blue,PlotPoints!200];
freqouter=Plot[x[t]/.sol1,ft,0,100g,PlotStyle!Blue,PlotPoints!100,
Ticks!ff50,100g,f-3,-2,-1,1,2,3gg];
(* Inner trajectory *)
point2=fa2,b2g;
system=fx’[t]==y[t],y’[t]==(1-x[t]2 )y[t]-x[t],fx[0],y[0]g==point2g;
sol2=NDSolve[system,fx,yg,ft,100g];
p2=ParametricPlot[Evaluate[fx[t],y[t]g/.sol2],ft,0,100g,PlotStyle!Red,
PlotPoints!300];
freqinner=Plot[x[t]/.sol2,ft,0,100g,PlotStyle!Red,PlotPoints!100,
Ticks!ff50,100g,f-2,-1,1,2gg];
(* Periodic trajectory; limit cycle *)
point3=fa3,b3g;
system=fx’[t]==y[t],y’[t]==(1-x[t]2 )y[t]-x[t],fx[0],y[0]g==point3g;
sol3=NDSolve[system,fx,yg,ft,100g];
p3=ParametricPlot[Evaluate[fx[t],y[t]g/.sol3],ft,0,100g,
PlotStyle!Green,PlotPoints!200];
freq=Plot[x[t]/.sol3,ft,0,100g,PlotStyle!Green,PlotPoints!100,
Ticks!ff50,100g,f-2,-1,1,2gg];
fShow[p1,p2,p3,Graphics[f
Text[Style[”initial outer point”,Italic,Blue,12],point1+f-.3,0.2g],
Text[Style[”initial inner point”,Italic,Red,12],point2+f-.5,0.2g],
fPointSize[0.03],Blue,Point[point1],Red,Point[point2],PointSize[0.035],
Green,Point[point3]gg],PlotRange!All,Ticks!ff-2,1,2g,f-3,-2,2,3gg,
ImageMargins!5,ImageSize!300],
Show[freq,PlotRange!ff0,50g,Automaticg],
Show[freqouter,PlotRange!ff0,50g,f-2,3gg],
Show[freqinner,PlotRange!ff0,50g,f-2,2gg]g]
Manipulate[Quiet@vanderPol[],
ff,.4g,0,3.1,0.01,Appearance!”Labeled”g,SaveDefinitions!True,
SynchronousUpdating!False]
See Fig. 10.32.
230 10 Miscellany in the Euclidean Plane

Fig. 10.32 Van der Pol equation

10.3.2 Hopf Bifurcation

As an example of Hopf bifurcation, we discuss a planar system of nonlinear

differential equations depending on a real parameter , [62, Chap. 8].
(
x0 D   x2
y0 D y:
10.3 Limit Cycles and Hopf Bifurcation 231

Fig. 10.33 Hopf bifurcation

Clear[]

Manipulate[Block[f$PerformanceGoal=”Quality”g,
Show[fContourPlot[ -x2 ==0,fx,-3,3g,fy,-3,3g,Mesh!None,
ContourStyle!fDashed,Blueg],ContourPlot[y==0,fx,-3,3g,fy,-3,3g,
ContourStyle!fDashed,Blueg],
StreamPlot[f -x2 ,yg,fx,-3,3g,fy,-3,3g,
StreamColorFunction!Hue]g,ImageSize!300]],ff,0.08g,0,1g]
See Fig. 10.33.
232 10 Miscellany in the Euclidean Plane

Fig. 10.34 Supercritical Hopf bifurcation

Another example follows.

Clear[]
supercriticalHopfbifurcation[ ]:=
Module[fx,y,u,v,k,h=0.02,nsteps=1000,a=0.001,length,
initpoints,pointss,pointsg,
initpoints=Union[Flatten[Outer[List,f-2,0,2g,f-2,0,2g],1],
ffa,0g,f0,ag,f0,-ag,f-a,0gg];
length=Length@initpoints;
pointss=fg;
Do[points=finitpoints[[p]]g;fx,yg=Flatten[points];
k=0;
While[k<nsteps,
u=x+h(y-x(x2 +y2 -)); v=y+h(-x-y(x2 +y2 -));
AppendTo[points,fu,vg];x=u;y=v;k++];
AppendTo[pointss,points],fp,lengthg];
ListLinePlot[pointss,Ticks!ff-2,-1,1,2g,f-2,-1,1,2gg,
PlotRange!All]]

Manipulate[
Quiet@supercriticalHopfbifurcation[],
ff,.6,””g,-1,2,0.1,Appearance!”Labeled”g,
SynchronousUpdating!False,SaveDefinitions!True]
See Fig. 10.34.
Chapter 11
Miscellany in the Euclidean Space

11.1 Some Space Curves

11.1.1 Viviani’s Window

Viviani’s window is defined as the intersection of the cylinder of radius a and center
.a; 0/

.x  a/2 C y2 D a2 ; a>0

and the sphere with center .0; 0; 0/ and radius 2a

x2 C y2 C z2 D 4a2 ; a > 0:

In order to see how this curve looks like, we consider the next code:
Clear[s,c,x,y]
s=x2 +y2 +z2 -4a2 ;c=(x-a)2 +y2 -a2 ;
Block[fa=1g,
ContourPlot3D[fs==0,c==0g,fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag,
MeshFunctions!fFunction[fx,y,z,fg,s-c]g,MeshStyle!ffThick,Bluegg,
Mesh!ff0gg,ContourStyle!Directive[LightGreen,Opacity[0.1],
Specularity[White,30]],Boxed!False,Axes!False,ImageSize!250]]
See Fig. 11.1.
It is easy to find its parametric equations

x.t/ D a.1 C cos t/;

y.t/ D a sin t;
z.t/ D 2a sin.t=2/; t 2 Œ0; 4:

© Springer International Publishing AG 2017 233

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2_11
234 11 Miscellany in the Euclidean Space

Fig. 11.1 Generation of the

Viviani’s window

Its picture was given above by the red line. We show the Viviani’s window
dynamically. The red dot moves on the curve describing it.
Clear[a,t]
With[fa=1g,
v1=ParametricPlot3D[fa(1+Cos[t]),a Sin[t],2a Sin[t/2]g,ft,0,4g,
Boxed!False,Axes!False];
v2=ParametricPlot3D[ff2afCos[t],Sin[t],0gg,
f2afCos[t],0,Sin[t]gg,f2af0,Cos[t],Sin[t]ggg,ft,0,2g];
pointviviani[t ]:=af1+Cos[t],Sin[t],2Sin[t/2]g];

Animate[Show[v1,v2,
Graphics3D[fPointSize[.04],Red,Point[pointviviani[t]]g],
PlotRange!All,ImageSize!200,ViewPoint!f6a,a/4,a/3g],ft,0,4g,
AnimationRate!.02,AnimationRunning!False,
SaveDefinitions!True]
See Fig. 11.2.

11.1.2 The Frenet–Serret Trihedron Along Viviani’s Window

Suppose a smooth curve in Euclidean space is given, r.t/, with r0 .t/ ¤ 0, for all t.
Then we use the following vectors:

r0 .t/
tangent.t/ D ;
kr0 .t/k
tangent0 .t/
normal.t/ D ;
ktangent0 .t/k
binormal.t/ D tangent.t/  normal.t/: (cross product)
11.1 Some Space Curves 235

Fig. 11.2 Dynamic

generation of the Viviani’s
window

It is known that these vectors are of unitary norm, and each one is orthogonal
on the plane determined by the other two. Thus, they define a trihedron called the
Frenet–Serret trihedron. We do not need to find these three vectors because there
is a built-in function FrenetSerretSystem which supplies them. Then we take the
Viviani’s window and write its corresponding trihedron. The tangent unit vector is
colored in green, and the normal unit vector is colored in red, whereas the binormal
unit vector is colored in blue.
Clear[r,t,u,tangent,normal,binormal]
With[fa=1g,
r=af1+Cos[#],Sin[#],2Sin[#/2]g&;
vectors=Last[FrenetSerretSystem[r[t],t]]//Simplify;
ftangent,normal,binormalg=Map[Arrow[fr[t],r[t]+#g]&,vectors]];

Manipulate[
Show[ParametricPlot3D[r[s],fs,0,4Pig,Boxed!False,Axes!False,
ImageSize!225,PlotStyle!fThick,Blueg,ViewPoint!f8a,a/2,ag],
Graphics3D[ffPointSize[.03],Black,Point[r[t]]g,
Thick,Darker[Green],tangent,Red,normal,Blue,binormalg],
PlotRange!2]//Evaluate,ft,0,4Pi,Appearance!f”Labelled”gg]
See Fig. 11.3.
236 11 Miscellany in the Euclidean Space

Fig. 11.3 Frenet–Serret

trihedron of the Viviani’s
window

11.1.3 Plane Curves on a Surface

We consider the surface x D u2 C u C v; y D v2 C u  v; and z D uv, where u; v are

real parameters. We show that the curves u D a and v D a (a is a constant) on the
surface are planar.
(* A necessary and sufficient condition for a curve to lie in a
plane is that its binormal is constant. *)
Clear[a,u,v]
f[u ,v ]:=fu2 +u+v,v2 +u-v,u vg (* The surface *)
Derivative[2,0][f][u,v]/.v!a (* A binormal *)
Derivative[0,2][f][u,v]/.u!a (* The other binormal *)
s=ParametricPlot3D[f[u,v],fu,-5,5g,fv,-5,5g,Mesh!None,
PlotStyle!Directive[Green,Opacity[0.3],Specularity[White,30]]];
c1=ParametricPlot3D[f[u,2],fu,-5,5g,PlotStyle!Blue];
c2=ParametricPlot3D[f[2,v],fv,-5,5g,PlotStyle!Red];
Show[s,c1,c2,ImageSize!250,Axes!False,Boxed!False]
f2; 0; 0g
f0; 2; 0g
See Fig. 11.4.
11.2 Some Surfaces in R3 237

Fig. 11.4 Coplanar curves in

space

11.2 Some Surfaces in R3

Along this section we use the tangent, normal, and binormal unit vectors at a point
of a space curve to construct interesting surfaces.

11.2.1 Torus

We present a torus around the Oz vertical axis.

Clear[t, ,r,u,tube,normal,binormal]
With[fa=1,radius=0.2g,
r[t ]:=afCos[t],Sin[t],0g;
fss[t ]:=FrenetSerretSystem[afCos[t],Sin[t],0g,t];
normal[u ]:=fss[t][[2,2]]/.t!u;
binormal[u ]:=fss[t][[2,3]]/.t!u;
(* The circle is centered at r[t] in the plane of normal and binormal
unit vectors *)
tube[v , ]:=r[v]+radius(Cos[]normal[v]+Sin[ ]binormal[v]);
ParametricPlot3D[Evaluate[tube[u,]],fu,0,2g,f ,0,2g,
PlotPoints!f100,12g,Ticks!ff-a,0,ag,f-a,0,ag,f-radius,0,radiusgg,
MeshStyle!ffThick,Redgg,Mesh!3,
PlotStyle!Directive[Darker[Green],Opacity[0.3],
Specularity[White,40],ImageSize!200]]]
See Fig. 11.5.
238 11 Miscellany in the Euclidean Space

Fig. 11.5 Torus around the vertical axis

11.2.2 A Simple Tube Around Viviani’s Window

Clear[t,,r,u,tube,normal,binormal]
With[fa=1,radius=0.2g,
r[t ]:=af1+Cos[t],Sin[t],2 Sin[t/2]g;
fss[t ]:=FrenetSerretSystem[af1+Cos[t],Sin[t],2Sin[t/2]g,t];
normal[u ]:=fss[t][[2,2]]/.t!u;
binormal[u ]:=fss[t][[2,3]]/.t!u;
(* The circle is centered at r[t] in the plane of normal and binormal
unit vectors *)
tube[v , ]:=r[v]+radius(Cos[]normal[v]+Sin[ ]binormal[v]);
ParametricPlot3D[Evaluate[tube[u,]],fu,0,4g,f ,0,2g,
PlotPoints!f100,12g,Ticks!ff0,1,2ag,f-a,0,ag,f-2a,0,2agg,
Mesh!0,PlotStyle!Directive[Darker[Green],Opacity[0.5],
Specularity[White,40]],ImageSize!200]]
See Fig. 11.6.

11.2.3 Another Tube Around Viviani’s Window

Clear[t,,r,radius,u,v,tube,normal,binormal]
With[fa=1g,
r[v ]:=af1+Cos[v],Sin[v],2Sin[v/2]g;
fss[t ]:=FrenetSerretSystem[af1+Cos[t],Sin[t],2Sin[t/2]g,t];
11.2 Some Surfaces in R3 239

Fig. 11.6 Tube around the

Viviani’s window

normal[u ]:=fss[t][[2,2]]/.t!u;
binormal[u ]:=fss[t][[2,3]]/.t!u;
radius[v ]:=0.2+.04Sin[20v]; (* The radius of the tube *)
(* The circle is centered at r[t] in the plane of normal and binormal
unit vectors *)
Tube[v , ]:=r[v]+radius[v](Cos[]normal[v]+Sin[ ]binormal[v]);
ParametricPlot3D[Evaluate[tube[u,]],fu,0,4 Pig,f ,0,2 Pig,
PlotPoints!f100,12g,Ticks!ff0,1,2g,f-1,0,1g,f-2,0,2gg,
Mesh!None,PlotStyle!Directive[Darker[Green],Opacity[0.5],
Specularity[White,30]],ImageSize!200]]
See Fig. 11.7.

11.2.4 Helicoid, Catenoid, and Costa’s Minimal Surface

An elliptic helicoid is presented below:

With[fa=1,b=3,c=0.5g,
ParametricPlot3D[fa u Cos[t],b u Sin[t ],c tg,ft,0,2g,fu,0,2g,
PlotPoints!80,
240 11 Miscellany in the Euclidean Space

Fig. 11.7 Another tube

around the Viviani’s window

Fig. 11.8 An elliptic helicoid

PlotStyle!fGreen,Opacity!0.3,Specularity[White,40]g,
PlotRange!All,Mesh!None,
Ticks!ff-2a,0,2ag,Automatic,f0,2cgg,ImageSize!225]]
See Fig. 11.8.
11.2 Some Surfaces in R3 241

Fig. 11.9 Animated helicoid

We animate the above figure in the following way:

Animate[
With[fa=1,b=1,c=0.25,v=RotationTransform[-,f0,0,1g][f10,0,4g]g,
Show[ParametricPlot3D[fa u Cos[t],b u Sin[t],c tg,ft,0,3g,fu,0,2g,
PlotPoints!50,PlotRange!All,Mesh!None,Boxed!False,
PlotStyle!fGreen,Opacity!0.3,Specularity[White,40]g,Axes!None,
ImageSize!200],ViewPoint!v]],f,0,2g,AnimationRunning!False]
See Fig. 11.9.
We now present a catenoid.
With[fc=1g,
ParametricPlot3D[fc Cosh[u/c]Cos[t],c Cosh[u/c]Sin[t],ug,ft,0,2g,
fu,-2,2g,
PlotPoints!50,PlotStyle!fGreen,Opacity!0.3,
Specularity[White,40]g,PlotRange!All,Mesh!None,
ImageSize!250]]
See Fig. 11.10.
The circular helicoid can be continuously deformed into a catenoid through a
convex combination.
Animate[
ParametricPlot3D[f(1-˛)u Cos[t]+˛Cosh[u]Cos[t],
(1-˛)u Sin[t]+˛Cosh[u]Sin[t],t(1-˛)+u˛g,fu,-2,2g,ft,0,2 Pig,
PlotRange!All,PlotStyle!fGreen,Opacity!0.3,
Specularity[White,40]g,Mesh!None,Axes!None,Boxed!False,
242 11 Miscellany in the Euclidean Space

Fig. 11.10 Catenoid

Fig. 11.11 Convex

combination
helicoid-catenoid

PlotPoints!50,ImageSize!200],f˛,0,1g,AnimationRate!0.1,
AnimationRunning!False]
See Fig. 11.11.
11.3 Some Bodies 243

Fig. 11.12 Costa’s minimal

surface

Now we introduce Costa’s minimal surface. For reference, see [24]. The code
follows.
cw=N[WeierstrassInvariants[f1/2,I/2g],10][[1]];
ew=WeierstrassP[0.5,fcw,0g];
costa[u ,v ]:=
f(1/2)Re[-WeierstrassZeta[u+I v,fcw,0g]+Pi u+Pi2 /(2ew)+(Pi/(2ew))
(WeierstrassZeta[u+I v-1/2,fcw,0g]
-WeierstrassZeta[u+I v-I/2,fcw,0g])],
(1/2)Re[-I WeierstrassZeta[u+I v,fcw,0g]+Pi v+Pi2 /(2ew)-(Pi I)/(2ew)
(WeierstrassZeta[u+I v-1/2,fcw,0g]
-WeierstrassZeta[u+I v-I/2,fcw,0g])],
(Sqrt[2Pi]/4)Log[Abs[(WeierstrassP[u+I v,fcw,0g]-ew)/
(WeierstrassP[u+I v,fcw,0g]+ew)]]g
costaplot=
With[feps=106 g,
ParametricPlot3D[costa[u,v],fu,eps,1-epsg,fv,eps,1-epsg,
Boxed!False,Axes!False,ViewPoint!f2.9,-1.4,1.5g,
PlotStyle!Directive[Green,Opacity!0.3,Specularity[White,40]],
Mesh!3,PlotPoints!80,ImageSize!325]]
See Fig. 11.12.

11.3 Some Bodies

11.3.1 The Volume of Two Bodies

In this subsection we find the volumes of two bodies in R3 .

The first body is given by the intersection of a 3D ball with a full cylinder. The
shape of the body is given below:
244 11 Miscellany in the Euclidean Space

Fig. 11.13 Intersection of a

ball with a full cylinder

Clear[x,y,z]
With[fa=1g,
s=x2 +y2 +z2 -4a2 ;
c=(x-a)2 +y2 -a2 ;
reg=RegionPlot3D[s0&&c0,fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag,
PlotStyle!Directive[Green,Opacity[0.3],Specularity[0.5]],
PlotPoints!80,Mesh!None];
Show[reg,PlotRange!All,Boxed!False,Axes!False,
ImageSize!250]]
See Fig. 11.13.
The volume of the body is calculated in two ways as follows:
Integrate[Boole[s0&&c0],fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag],
Integrate[If[s0&&c0,1,0],fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag]
˚ 16

9
.4 C 3/; 16
9
.4 C 3/
The next body is the intersection of a 3D ball with a full double circular
paraboloid.
Clear[x,y,z]
With[fa=1g,
s=x2 +y2 +z2 -4a2 ; p=x2 +y2 -Abs[z];
RegionPlot3D[s0&&p0,fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag,
PlotStyle!Directive[Green,Opacity!0.15,Specularity[White,5]],
PlotPoints!50,Mesh!2,Boxed!False,Axes!False,
ImageSize!250]]
See Fig. 11.14.
11.3 Some Bodies 245

Fig. 11.14 Double circular

paraboloid

Its volume is found using one of the next codes:

FullSimplify[
fIntegrate[Boole[s0&&p0],fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag],
Integrate[If[s0&&p0,1,0],fx,-2a,2ag,fy,-2a,2ag,fz,-2a,2ag]g]
n 
1
p 1 p o
6
89  17 17 ; 6
89  17 17 

11.3.2 Reuleaux Tetrahedron

The Reuleaux tetrahedron is the intersection of four balls of radius s centered

at the vertices of a regular tetrahedron with the side length s; several interesting
figures connected to Reuleaux and Meissner tetrahedrons are available at [60, pp.
1194–1195]. The frontier of a ball through each vertex passes through the other
three vertices, which also form vertices of the Reuleaux tetrahedron. The Reuleaux
tetrahedron has the same face structure as a regular tetrahedron but with curved
faces: four vertices and four curved faces, connected by six circular-arc edges. The
vertices of the tetrahedron are given below as elements of a list:
nn p . o np . p . o
vertices= 0; 0;  6 4 ; 3 3; 0; 6 12 ;
n p . p . o n p . p . oo
 3 6; 1=2; 6 12 ;  3 6; 1=2; 6 12 ;

We verify that the four points are noncoplanar

MatrixRank[vertices[[4]]-#&/@Table[vertices[[i]],fi,1,3g]]
3
and check that the tetrahedron is regular having the side length 1.
Simplify[Norm[f#[[1]]-#[[2]]g]&/@Subsets[vertices,f2g]]
f1; 1; 1; 1; 1; 1g
246 11 Miscellany in the Euclidean Space

Fig. 11.15 Four spheres

Fig. 11.16 Intersection of

two spheres

Now we can see the four balls.

Graphics3D[fOpacity[.5],Green,Specularity[White,1],PlotPoints!50,
Sphere/@verticesg,Boxed!False,ImageSize!Small]
See Fig. 11.15.
Clear[x,y,z]
ineqs=Function[v,v.v][fx,y,zg-#]1&/@vertices;
solid=With[fr=.6g,
RegionPlot3D[And@@ineqs,fx,-r,rg,fy,-r,rg,fz,-r,rg,
PlotStyle!Directive[Green,Opacity[0.25],Specularity[White,40]],
Mesh!1,PlotPoints!50,Boxed!False,Axes!None,
ImageSize!250]]
See Fig. 11.16.
tetrahedron=Tetrahedron[vertices];
centroid=RegionCentroid[tetrahedron];
solid=With[fr=.6g,
RegionPlot3D[And@@ineqs,fx,-r,rg,fy,-r,rg,fz,-r,rg,
PlotStyle!Directive[Green,Opacity[0.3],Specularity[White,1]],
Mesh!0,PlotPoints!50,Boxed!False,Axes!None,
ImageSize!240] ];
fsolidobject=Show[solid,Graphics3D[fPink,tetrahedrong],
PlotRange!All,ImageSize!240],Rotate[solidobject,/3,f0,0g]g
11.3 Some Bodies 247

Fig. 11.17 Tetrahedrons

Fig. 11.18 Animated

tetrahedron

See Fig. 11.17.

Animate[
Rotate[solidobject,,f0,0g],f,0,2g,AnimationRate!0.1,
SaveDefinitions!True,AnimationRunning!False]
See Fig. 11.18.
248 11 Miscellany in the Euclidean Space

11.4 Local Extrema of Real Functions of Several Real

Variables

11.4.1 Unconstrained Local Extrema

11.4.1.1 The First Example

We introduce a code to find the local extrema which is suitable for functions of two
variables.
Clear[x,y,f]
f[x ,y ]:=y4 -y3 -3x2 y+x4
fdfx[x ,y ]=D[f[x,y],x],dfy[x ,y ]=D[f[x,y],y],dfxx[x ,y ]=D[f[x,y],x,x],
dfyy[x ,y ]=D[f[x,y],y,y],dfxy[x ,y ]=D[f[x,y],x,y],
dff[x ,y ]=dfxx[x,y]*dfyy[x,y]-dfxy[x,y]2 g;
solns=DeleteDuplicates[fx,yg/.Solve[fdfx[x,y]==0,dfy[x,y]==0g,fx,yg]];
If[Length@solns==0,Print[”No stationary point”],
solnsheader=f”no.”,”stationary point”g;
Print[Grid[Join[fsolnsheaderg,Table[Prepend[fsolns[[k]]g,k],
fk,Length@solnsg]],Frame!All,FrameStyle!Thin,
Alignment!Right]];
realsolns=Select[solns,FreeQ[#,Complex]&];
If[Length@realsolns==0,Print[”No real stationary point”],
realsolnsheader=f”no.”,”real stationary point”g;
Print[Grid[Join[frealsolnsheaderg,Table[Prepend[frealsolns[[k]]g,k],
fk,Length@realsolnsg]],Frame!All,FrameStyle!Thin,
Alignment!Right]];
extremum=
Select[realsolns,dff[#[[1]],#[[2]]]>0&&dfxx[#[[1]],#[[2]]]¤0&];
If[Length[extremum]==0,Print[”No point of extremum”],
line=fg;
Do[If[dfxx[extremum[[k,1]],extremum[[k,2]]]<0,text=”maximum”,
text=”minimum”];
line=AppendTo[line,fk,extremum[[k]],text,f@@extremum[[k]]g],
fk,Length@extremumg];
extremumheader=f”no.”,”point of extremum”,”nature”,
”value of the function”g;
Print[Grid[Join[fextremumheaderg,Table[line[[k]],
fk,Length@extremumg]],
Frame!All,FrameStyle!Thin,Alignment!Right]]]]]
11.4 Local Extrema of Real Functions of Several Real Variables 249

Fig. 11.19 The surface under

discussion

No. Stationary
n point
o
3i 3
1  2p ;  4 No. Real stationary point
n 2 o
2 3i
p ;  34 1 f0;
˚ 0g

2 2
2 0; 3
3 f0; ˚ 34 3

˚ 0g

4 3 3 ˚ ;
2
2
˚0; 34 3
4 3 3
;
5 ˚ ; 2 2
2
2
3 3
6 ;
2 2

No. Point
˚ 3 of
extremum Nature Value of the function
1  2 ; 32 minimum  27
˚3 3
8
2 ;
2 2
minimum  27
8

Now we visualize this function.

Plot3D[f[x,y],fx,-2,2g,fy,0.5,2g,ImageSize!250]
See Fig. 11.19.
Example The previous code works properly only if the nature of the points of
extremum are sensitive to the first and second order differentials. The situation is
similar to the one-dimensional case f .x/ D x3 ; x 2 R; where higher order derivatives
are necessary to settle down the nature of the stationary point.

11.4.1.2 The Second Example

The example we discuss below appears in [53, Chap. 7]. The statement of the
exercise reads as follows: find the local extreme values of the following function

f .x; y; z/ D 2x2  xy C 2xz  y C y3 C z2 ; x; y; z 2 R:

250 11 Miscellany in the Euclidean Space

First we approach it by the built-in functions NMinimize and NMaximize.

Clear[x,y,z,f]
variables=fx,y,zg;
f[x ,y ,z ]:=2x2 -x y+2x z-y+y3 +z2
NMinimize[f[x,y,z],variables]
NMaximize[f[x,y,z],variables]
f0:481481; fx ! 0:333333; y ! 0:666667; z ! 0:333333gg
NMaximize::cvdiv W Failed to converge to a solution. The function may be
unbounded. 
f9:67408528878118  10339 ; fx ! 1:4507  109 ; y ! 2:13077  10113 ;
z ! 9:0809  108 gg
We extract the coordinates of the point of minimum.
fx,y,zg/.Last[NMinimize[f[x,y,z],variables]]
f0:333333; 0:666667; 0:333333g
We also extract
x/.Last[NMinimize[f[x,y,z],variables]]
y/.Last[NMinimize[f[x,y,z],variables]]
x/.Last[NMinimize[f[x,y,z],variables]]
0:333333
0:666667
0:333333
The next approach is based on the built-in functions Maximize and Minimize.
Clear[x,y,z]
Minimize[f[x,y,z],fx,y,zg]
Maximize[f[x,y,z],fx,y,zg]
Minimize::natt: The minimum is not attained at any point satisfying the given
˚constraints
˚ 

1; x ! 33 10
; y ! 1; z ! 85
Maximize::natt: The maximum is not attained at any point satisfying the given
˚constraints
˚ 

1; x ! 1; y ! 33 10
; z ! 85
We conclude that this function has neither global maximum point nor global
minimum point.
Now we make use of the tools of differential calculus and find the stationary
points.
11.4 Local Extrema of Real Functions of Several Real Variables 251

Clear[x,y,z]
variables=fx,y,zg;
f[x ,y ,z ]:=2x2 -x y+2x z-y+y3 +z2
solns=DeleteDuplicates[
variables/.Solve[Thread[Grad[f[x,y,z],variables]==0],variables]];
If[Length[solns]==0,Print[”No stationary point”],
solnsheader=f”no.”,”stationary point”g;
Print[Grid[Join[fsolnsheaderg,Table[Prepend[fsolns[[k]]g,k],
fk,Length[solns]g]],Frame!All,FrameStyle!Thin,Alignment!Right]];
realsolns=Select[solns,FreeQ[#,Complex]&];
If[Length[realsolns]==0,Print[”No real stationary point”],
realsolnsheader=f”no.”,”real stationary point”g;
Print[Grid[Join[frealsolnsheaderg,Table[Prepend[frealsolns[[k]]g,k],
fk,Length[realsolns]g]],Frame!All,FrameStyle!Thin,
Alignment!Right]];
seconddiff[x ,y ,z ]:=D[f[x,y,z],fvariables,2g];
extremum=fg;
Do[matrix=seconddiff[x,y,z]/.Thread[variables!realsolns[[j]]];
If[PositiveDefiniteMatrixQ[matrix]kNegativeDefiniteMatrixQ[matrix],
AppendTo[extremum,realsolns[[j]]]],fj,Length@realsolnsg];
If[Length[extremum]==0,Print[”No point of extremum”],
line=fg;
Do[matrix=seconddiff[x,y,z]/.Thread[variables!extremum[[k]]];
If[NegativeDefiniteMatrixQ[matrix],text=”maximum”,text=”minimum”];
line=AppendTo[line,fk,extremum[[k]],text,f@@extremum[[k]]g],
fk,Length@extremumg];
extremumheader=f”no.”,”point of extremum”,”nature”,
”value of the function”g;
Grid[Join[fextremumheaderg,Table[line[[k]],fk,Length@extremumg]],
Frame!All,FrameStyle!Thin,Alignment!Right]
]]]
No. Stationary
˚ 1 point

No. Real
˚ 1 stationary
point
1  4 ;  12 ; 14 1  4 ;  12 ; 14
˚1 2
˚1 2

2 ; ;  13
3 3
2 ; ;  13
3 3

No. Point
˚ 1 2 of extremum

Nature Value of the function
1 ; ;  13
3 3
minimum  13
27
252 11 Miscellany in the Euclidean Space

11.4.2 Constrained Local Extrema

11.4.2.1 The First Example

Along this subsection we consider the problem of finding the extreme values of the
function
(
x2 C y2 C z2 D 1;
f Œx; y; z/ D xyz; subject to
x C y C z D 1; x; y; z 2 R:

In order to see where these results come from, we use the method of Lagrange
multipliers [53, Chap. 7].
We have three real variables and two constraints. Consequently, we have two
multipliers.
Clear[x,y,z, ,]
function[x ,y ,z ]:=x y z
variables=fx,y,zg;
condition1[x ,y ,z ]:=x2 +y2 +z2 -1
condition2[x ,y ,z ]:=x+y+z
multipliers=f ,g;
allvariables=Flatten[fvariables,multipliersg];
The function we need is the Lagrangian
lagrangian[x ,y ,z , , ]:=ffunction[x,y,z],condition1[x,y,z],
condition2[x,y,z]g.f1, ,g
We look for the stationary points.
solns=DeleteDuplicates[allvariables/.
Solve[Thread[Grad[lagrangian[x,y,z, ,],allvariables]==0],
allvariables]];
If[Length@solns==0,Print[”No stationary point”],
Print[Grid[Join[ff”no.”,”stationary point”gg,
Table[Prepend[fSimplify[solns[[k]]]g,k],fk,Length@solnsg]],
Frame!All,FrameStyle!Thin,Alignment!Right]];
realsolns=Select[solns,FreeQ[#,Complex]&];
If[Length@realsolns==0,Print[”No real stationary point”],
Print[Grid[Join[ff”no.”,”real stationary point”gg,
Table[Prepend[fSimplify[realsolns[[k]]]g,k],fk,Length@realsolnsg]],
Frame!All,FrameStyle!Thin,Alignment!Right]];
seconddiff[x ,y ,z , , ]=D[lagrangian[x,y,z, ,],fallvariables,2g];
dxyz=fdx,dy,dzg;
xyzsolns=Table[Take[realsolns[[k]],3],fk,Length@realsolnsg];
cond1:=Table[D[condition1[x,y,z],fvariables,1g].dxyz/.
Thread[variables!xyzsolns[[j]]],fj,Length@realsolnsg];
11.4 Local Extrema of Real Functions of Several Real Variables 253

cond2:=Table[D[condition2[x,y,z],fvariables,1g].dxyz/.
Thread[variables!xyzsolns[[j]]],fj,Length@realsolnsg];
soldydz=Table[Flatten[Solve[cond1[[k]]==0&&cond2[[k]]==0,
fdx,dy,dzg,Reals]],fk,Length@realsolnsg];
soldydz1=dxyz/.soldydz;
dxyzall=Table[Join[soldydz1[[k]],fd ,dg],fk,Length@realsolnsg];
lines=fg;
Do[
matrix=Simplify[
dxyzall[[k]].(seconddiff[x,y,z, ,]/.
Thread[allvariables!realsolns[[k]]]).dxyzall[[k]]];
value=FullSimplify[function[x,y,z]/.Thread[variables!xyzsolns[[k]]]];
point=Simplify[xyzsolns[[k]]];
If[PositiveDefiniteMatrixQ[matrix],
lines=Append[lines,fk,point,matrix,valueg],
If[NegativeDefiniteMatrixQ[matrix],
lines=Append[lines,fk,point,matrix,valueg],
lines=Append[lines,fk,point,matrix,valueg]]],
fk,Length@realsolnsg]]]
header=f”no.”,”point of extremum”,”second order diff.”,”value of
the function”g;
Grid[Join[fheaderg,Table[lines[[k]],fk,Length@realsolnsg]],
Frame!All,FrameStyle!Thin,Alignment!Right]

No. Stationary point No. Real stationary point

n q o n q o
1  23 ; p1 6 ; p1 6 ; 2p 1
;1
6 6
1  23 ; p1 6 ; p1 6 ; 2p 1
; 1
6
6
nq o nq o
2
2 3
;  p1 6 ;  p1 6 ;  2p 1
;1
6 6
2 2
3
;  p 1
;  p 1
;  1
p ;1
6 6 2 6 6
n q o n q o
3  p1 6 ;  p1 6 ; 23 ;  2p 1
; 1
6 3  p1 6 ;  p1 6 ; 23 ;  2p 1
; 1
6 6 6
n q o n q o
4  p1 6 ; 32 ;  p1 6 ;  2p 1
;1
6 6
4  p1 6 ; 23 ;  p1 6 ;  2p 1
;1
6 6
n q o n q o
1 2 p 1 1 1 1 2 p 1 1 1
5 p ;  3
; ; p ; 6 5 p ;  3
; ; p ;
6 6 2 6 6 6 2 6 6
n q o n q o
1 1 2 1
6 p
6
; p
6
;  ; p
3 2 6 6
;1 6 p1
; p1
;  2
; p
3 2 6 6
1
;1
6 6

Solve::svars: Equations may not give solutions for all “solve” variables. 
Solve::svars: Equations may not give solutions for all “solve” variables. 
Solve::svars: Equations may not give solutions for all “solve” variables. 
General::stop: Further output of Solve::svars will be suppressed during this
calculation. 
254 11 Miscellany in the Euclidean Space

No. Point of extremum Second order diff. Value of the function

n q o p
1  23 ; p1 6 ; p1 6 6dy2 1
 3p 6
nq o p
2
2 3
;  p1 6 ;  p1 6  6dy2 1
p
n q o p
3 6

3  p1 6 ;  p1 6 ; 23  6dx2 1
p
n q o p
3 6

4  p1 6 ; 32 ;  p1 6  6dx2 1
p
n q o p
3 6
1 2 p 1
5 p ;  3
; 6dx2 1
 3p
n 6 q6o p
6
1 1 2
6 p
6
; p
6
;  3
6dx2 1
 3p 6

We conclude that the first, fifth, and sixth points are points of minimum, whereas
the other ones are points of maximum. The values of the function on all these points
have been already established in the last column.

11.4.2.2 The Second Example

We want to find the points of local extreme of the function

f .x; y; z/ D 2x2 C 2y2  xy C z4  2z2 ; on the compact set

˚ ˇ

K D .x; y; z/ ˇx2 C y2 C 2z2  8 ; x; y; z 2 R:

We have two cases. The cases are exclusive and exhaustive.

Firstly, we find the points of local extreme lying in the interior of the compact set
K. Therefore, we repeat the algorithm in the unconstraint local extreme case.
Clear[x,y,z,function]
variables=fx,y,zg;
function[x ,y ,z ]:=2x2 +2y2 -x y+z4 -2z2 ;
Then we find the stationary points.
solns=Select[DeleteDuplicates[
variables/.Solve[Thread[Grad[function[x,y,z],variables]==0],variables]],
#ŒŒ12 C #ŒŒ22 C 2#ŒŒ32 < 8&;
If[Length[solns]==0,Print[”No stationary point interior”],
solnsheader=f”no.”,”stationary point interior”g;
Print[Grid[Join[fsolnsheaderg,Table[Prepend[fsolns[[k]]g,k],
fk,Length[solns]g]],Frame!All,FrameStyle!Thin,Alignment!Right]];
realsolns=Select[solns,FreeQ[#,Complex]&];
If[Length[realsolns]==0,Print[”No real stationary point interior”],
realsolnsheader=f”no.”,”real stationary point interior”g;
11.4 Local Extrema of Real Functions of Several Real Variables 255

Print[Grid[Join[frealsolnsheaderg,Table[Prepend[frealsolns[[k]]g,k],
fk,Length[realsolns]g]],Frame!All,FrameStyle!Thin,
Alignment!Right]];
seconddiff[x ,y ,z ]:=D[function[x,y,z],fvariables,2g];
extremum=fg;
Do[matrix=seconddiff[x,y,z]/.Thread[variables!realsolns[[j]]];
If[PositiveDefiniteMatrixQ[matrix]kNegativeDefiniteMatrixQ[matrix],
AppendTo[extremum,realsolns[[j]]]],fj,Length@realsolnsg];
If[Length[extremum]==0,Print[”No point of extremum interior”],
line=fg;
Do[matrix=seconddiff[x,y,z]/.Thread[variables!extremum[[k]]];
If[NegativeDefiniteMatrixQ[matrix],text=”maximum”,text=”minimum”];
line=AppendTo[line,fk,extremum[[k]],text,function@@extremum[[k]]g],
fk,Length@extremumg];
extremumheader=f”no.”,”point of extremum”,”nature”,”value of the
function”g;
Grid[Join[fextremumheaderg,Table[line[[k]],fk,Length@extremumg]],
Frame!All,FrameStyle!Thin,Alignment!Right]
]]]
No. Stationary point interior K No. Real stationary point interior K
1 f0; 0; 1g 1 f0; 0; 1g
2 f0; 0; 0g 2 f0; 0; 0g
3 f0; 0; 1g 3 f0; 0; 1g

No. Point of extremum Nature Value of the function

1 f0; 0; 1g minimum 1
2 f0; 0; 1g minimum 1

Let us look for the points of extreme located on the boundary of K. Then we
apply the method of Lagrange multipliers.
Clear[x,y,z,condition,lagrangian, ,seconddiff]
condition[x ,y ,z ]:=x2 +y2 +2z2 -8;
multiplier=f g;
allvariables=Flatten[fvariables,multiplierg];
The function we need is
lagrangian[x ,y ,z , ]:=ffunction[x,y,z],condition[x,y,z]g.f1, g
We look for the stationary points
solns=DeleteDuplicates[allvariables/.
Solve[Thread[Grad[lagrangian[x,y,z, ],allvariables]==0],allvariables]];
If[Length@solns==0,Print[”No critical point on the boundary of K”],
Print[Grid[Join[ff”no.”,”critical point on the boundary of K”gg,
Table[Prepend[fSimplify[solns[[k]]]g,k],fk,Length@solnsg]],
256 11 Miscellany in the Euclidean Space

Frame!All,FrameStyle!Thin,Alignment!Right]];
realsolns=Select[solns,FreeQ[#,Complex]&];
If[Length@realsolns==0,Print[”No real critical point on the boundary of
K”],
Print[Grid[Join[ff”no.”,”real critical point on the boundary
of K”gg,Table[Prepend[fSimplify[realsolns[[k]]]g,k],
fk,Length@realsolnsg]],Frame!All,FrameStyle!Thin,
Alignment!Right]];
seconddiff[x ,y ,z , ]=D[lagrangian[x,y,z, ],fallvariables,2g];
dxyz=fdx,dy,dzg;
xyzsolns=Table[Take[realsolns[[k]],3],fk,Length@realsolnsg];
cond:=Table[D[condition[x,y,z],fvariables,1g].dxyz/.
Thread[variables!xyzsolns[[j]]],fj,Length@realsolnsg];
soldydz=Table[Flatten[Solve[cond[[k]]==0,fdx,dy,dzg,Reals]],
fk,Length@realsolnsg];
soldydz1=dxyz/.soldydz;
dxyzall=Table[Join[soldydz1[[k]],fd g],fk,Length@realsolnsg];
lines=fg;
Do[
matrix=Simplify[
dxyzall[[k]].(seconddiff[x,y,z, ]/.Thread[allvariables!realsolns[[k]]]).
dxyzall[[k]]];
value=FullSimplify[function[x,y,z]/.Thread[variables!xyzsolns[[k]]]];
point=Simplify[xyzsolns[[k]]];
If[PositiveDefiniteMatrixQ[matrix],
lines=Append[lines,fk,point,matrix,valueg],
If[NegativeDefiniteMatrixQ[matrix],
lines=Append[lines,fk,point,matrix,valueg],
lines=Append[lines,fk,point,matrix,valueg]]],fk,Length@realsolnsg]
]]
header=f”no.”,”point of extremum”,”second order diff.”,”value of f ”,
”nature”g;
Grid[Join[fheaderg,Table[lines[[k]],fk,Length@realsolnsg]],
Frame!All,FrameStyle!Thin,Alignment!Right]
11.4 Local Extrema of Real Functions of Several Real Variables 257

No. Critical point on the boundary of K No. Real critical point on the boundary of K
˚
˚

1 2; 2; 0;  32 1 2; 2; 0;  32

˚
˚

2 2; 2; 0;  52 2 2; 2; 0;  52
3 f0; 0; 2; 3g 3 f0; 0; 2; 3g
4 f0; 0; 2; 3g 4 f0; 0; 2; 3g
˚
˚

5 2; 2; 0;  52 5 2; 2; 0;  52
˚ 3

˚ 3

6 n2;q 2; 0; q
2 q o 6 n2;q 2; 0; q
2 q o
7  3; 3; 5;3 7  3; 3; 5;3
n q 2 q 2 q 2 o2 n q 2 q 2 q 2 o2
8  3 ;  32 ; 52 ;  32 8  3 ;  23 ; 52 ;  32
nq 2q q o nq 2q q o
3
9 ; 3 ;  5 ;  32
2 q2 q 2
9 3
; 3 ;  5 ;  32
2 q2 q 2
nq o nq o
3
10 2
; 32 ; 52 ;  32 10 3
2
; 23 ; 52 ;  32
n q o n q o
11  p1 2 ; p1 2 ;  72 ;  52 11  p1 2 ; p1 2 ;  72 ;  52
n q o n q o
12  p1 2 ; p1 2 ; 72 ;  52 12  p1 2 ; p1 2 ; 72 ;  52
n q o n q o
1
13 p ;  p1 2 ;  72 ;  52 13 1
p ;  p1 2 ;  72 ;  52
n 2 q o n 2 q o
1
14 p
2
;  p1 2 ; 72 ;  52 14 1
p
2
;  p1 2 ; 72 ;  52
Solve::svars: Equations may not give solutions for all “solve” variables.
Solve::svars: Equations may not give solutions for all “solve” variables.
Solve::svars: Equations may not give solutions for all “solve” variables.
General::stop: Further output of Solve::svars will be suppressed during this calculation.

No. Point of extremum Second order diff. Value of f Nature

1 f2; 2; 0g 4 dx2  10dz2 12 saddle
 
2 f2; 2; 0g 2 2dx2 C 7dz2 20 max
 2 
3 f0; 0; 2g 2 dx C dxdy C dy2 8 max
 2 
4 f0; 0; 2g 2 dx C dxdy C dy2 8 max
 
5 f2; 2; 0g 2 2dx2 C 7dz2 20 max
6 f2;
n q 2; 0g q q o 4dx2  10dz2 12 saddle
 
7  32 ;  32 ;  52 4 dx2 C dxdy C dy2 23
4
min
n q q q o  2 
8  32 ;  32 ; 52 4 dx C dxdy C dy2 23
4
min
nq q q o  
3
9 ; 3; 5
2 q2 q 2
4 dx2 C dxdy C dy2 23
4
min
nq o  
3
10 2
; 32 ; 52 4 dx2 C dxdy C dy2 23
4
min
n q o
11  p1 2 ; p1 2 ;  72 4dxdy 31
4
saddle
n q o
12  p1 2 ; p1 2 ; 72 4dxdy 31
4
saddle
n q o
1 1 7 31
13 p ;  p ; 
2 q o2
4dxdy 4
saddle
n 2
1
14 p
2
;  p1 2 ; 72 4dxdy 31
4
saddle
References

1. Abell, M.L., Braselton, J.P.: Differential Equations with Mathematica. AP Professional, Boston
(1993)
2. Adamchik, V., Wagon, S.:  A 2000-year search changes direction. Math. Educ. Res. 5(1),
11–19 (1996). www.cs.cmu.edu/~adamchik/articles/pi/pi.htm
3. Adamchik, V., Wagon, S.: A simple formula for . Am. Math. Mon. 104(9), 852–855 (1997)
4. Alexander, R.: Diagonally implicit Runge-Kutta methods for stiff O.D.E’s. SIAM J. Numer.
Anal. 14(6), 1006–1021 (1977)
5. Backhouse, N.: Pancake functions and approximations to . Math. Gaz. 79, 371–374 (1995).
Note 79.36
6. Bailey, D.H., Borwein, P.B., Plouffe, S.: On the rapid computation of various polylogarithmic
constants. Math. Comput. 66(218), 903–913 (1997)
7. Baruah, D.N., Berndt, B.C., Chan, H.H.: Ramanujan’s series for 1=: a survey. Am. Math.
Mon. 116(7), 567–587 (2009)
8. Bellard, F.: Computation of the n’th digit of  in any base in O.n2 / (1997). fabrice.bellard.
free.fr/pi/
9. Borwein, J.M., Borwein, P.B.: The class three Ramanujan type series for 1=. J. Comput.
Appl. Math. 45(1–2), 281–290 (1993)
10. Borwein, J.M., Borwein, P.B., Bailey, D.H.: Ramanujan, modular equations, and approxima-
tions to  or how to compute one billion digits of . Am. Math. Mon. 96(3), 201–219 (1989)
11. Borwein, J.M., Skerritt, M.P.: An Introduction to Modern Mathematical Computing with
Mathematica. Springer Undergraduate Text in Mathematics and Technology. Springer, New
York (2012)
12. Brun, V.: Carl Störmer in memoriam. Acta Math. 100(1–2), I–VII (1958)
13. Bryson, A.E., Ho, Y.C.: Applied Optimal Control: Optimization, Estimation, and Control.
Halsted Press, New York (1975)
14. Burns, R.E., Singleton, L.G.: Ascent from the lunar surface. Technical report TN D-1644,
NASA, George C. Marshall Space Flight Center, Huntsville (1965)
15. Cesari, L.: Optimization–Theory and Applications. Problems with Ordinary Differential
Equations. Applications of Mathematics, vol. 17. Springer, New York (1983)
16. Champion, B.: General Visualization Quick Start. Wolfram Research, Champaign (2013).
www.wolfram.com/training/courses/vis412.html
17. Chudnovsky, D.V., Chudnovsky, G.V.: The computation of classical constants. Proc. Natl.
Acad. Sci. USA 86(21), 8178–8182 (1989)
18. Cloitre, B.: A BBP formula for  2 in golden base (2003). abcloitre@wanadoo.fr
19. Dennis Lawrence, J.: A Catalog of Special Plane Curves. Dover, New York (1972)

© Springer International Publishing AG 2017 259

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2
260 References

20. Don, E.: Mathematica. Schaum’s Outlines Series. McGraw Hill, New York (2009)
21. Ebbinghaus, H.D., Peckhaus, V.: Ernst Zermelo. An Approach to His Life and Work. Springer,
Berlin/Heidelberg (2007)
22. Elkies, N.D.: On a4 C b4 C c4 D d4 . Math. Comput. 51(184), 825–835 (1988)
23. Fay, T.H.: The butterfly curve. Am. Math. Mon. 96(5), 442–443 (1989)
24. Ferguson, H., Gray, A., Markvorsen, S.: Costa’s minimal surface via Mathematica. Math.
Educ. Res. 5(1), 5–10 (1966)
25. Finch, S.R.: Zermelo’s navigation problem. In: Mathematical Constants II. Encyclopedia of
Mathematics and Its Applications. Cambridge University Press, Cambridge (Forthcoming)
26. Floyd, R.W.: Algorithm 245: treesort. Commun. ACM 7(12), 701 (1964)
27. Frye, R.E.: Finding 958004 C 2175194 C 4145604 D 4224814 on the connection machine. In:
Proceedings of Supercomputing’88. Science and Applications, vol. 2, pp. 106–116 (1988)
28. Gourevitch, B., Guillera Goyanes, J.: Construction of binomial sums for  and polylogarithmic
constants inspired by BBP formula. Appl. Math. E-Notes 7, 237–246 (2007). www.math.nthu.
edu.tw/~amen
29. Gradshteyn, S.G., Ryzhik, I.M.: Tables of Integrals, Series, and Products, 7th edn. Elsevier,
Amsterdam (2007)
30. Guillera, J., Zudilin, W.: Ramanujan-type for 1=: the art of translation. In: Bruce, D.P.,
Berndt, C. (eds.) The Legacy of Srinivasa Ramanujan. Lecture Notes Series, vol. 20, pp. 181–
195. Ramanujan Mathematical Society (2013). arXiv 1302.0548
31. Hartman, P.: Ordinary Differential Equations, 1st edn. Wiley, Hoboken (1964)
32. Hastings, C., Mischo, K., Morrison, M.: Hands-On Start to Wolfram Mathematicar and
Programming with the Wolfram LanguageTM . Wolfram Media, Champaign (2015)
33. Hazrat, R.: Mathematicar : A Problem-Centered Approach. Springer Undergraduate Mathe-
matics Series, vol. 53. Springer, London (2010)
34. Hoare, C.A.R.: Algorithm 64: quicksort. Commun. ACM 4(7), 321 (1961)
35. Hull, D.G.: Optimal guidance for Lunar ascent. Adv. Astronaut. Sci. 134, 275–285 (2009).
Proccedings of the AAS Space Flight Machanics Meeting, Savannach
36. Hull, D.G.: Optimal guidance for quasi-planar Lunar ascent. J. Optim. Theory Appl. 151(2),
353–372 (2011)
37. Hull, D.G., Harris, M.W.: Optimal solutions for quasiplanar ascent over a spherical Moon. J.
Guid. Control Dyn. 35(4), 1218–1224 (2012). doi: 10.2514/1.55443
38. Knuth, D.E.: The Art of Computer Programming. Sorting and Searching. Computer Science
and Information Processing, vol. 3. Addison-Wesley, Reading (1973)
39. Lander, L.J., Parkin, T.R.: Counterexample to Euler’s conjecture on sums of like powers. Bull.
Am. Math. Soc. 72(6), 1079 (1966)
40. Lawden, D.F.: Analytical Methods of Optimization. Dover Books on Mathematics. Dover,
Mineola (2006)
41. Lucas, S.K.: Integral proofs that 355=113 > . Gaz. Aust. Math. Soc. 32(4), 263–266 (2005)
42. Lucas, S.K.: Integral approximations to  with nonnegative integrands (2007). carma.
newcastle.edu.au/jon/Preprints/Papers/By%20Others/more-pi.pdf
43. Mangano, S.: Mathematica Cookbook. O’Reilly, Sebastopol (2010)
44. Manià, B.: Sopra un problema di navigatione di Zermelo. Math. Ann. 113(1), 584–589 (1937)
45. McShane, E.J.: A navigation problem in the calculus of variations. Am. J. Math. 59(2), 327–
334 (1937)
46. Miele, A.: Flight Mechanics. Theory of Flight Path. Addison-Wesley Series in the Engineering
Sciences Space Science and Technology, vol. 1. Addison-Wesley, Reading (1962)
47. Mureşan, M.: Classical Analysis by Mathematica (Forthcoming)
48. Mureşan, M.: A Primer on the Calculus of Variations and Optimal Control. Trajectories
Optimization (Forthcoming)
49. Mureşan, M.: On a Runge-Kutta type method. Rev. Anal. Numér. Théorie Approx. 16(2),
141–147 (1987)
50. Mureşan, M.: Some computing results of a Runge-Kutta type method. Seminar on
Mathematical Analysis Nr. 7, Univ. Babeş-Bolyai, Cluj-Napoca, pp. 101–114 (1987)
References 261

51. Mureşan, M.: A semi-explicit Runge-Kutta method. Seminar on Differential Equations Nr. 8,
Univ. Babeş-Bolyai, Cluj-Napoca, pp. 65–70 (1988)
52. Mureşan, M.: Qualitative Properties of Differential Equations and Inclusions. Ph.D. thesis,
Babeş-Bolyai University, Cluj-Napoca (1996)
53. Mureşan, M.: A Concrete Approach to Classical Analysis. CMS Books in Mathematics.
Springer, New York (2009)
54. Mureşan, M.: Instructor Solution Manual, for A Concrete Approach to Classical Analy-
sis. Springer, New York (2012). www.springer.com/mathematics/analysis/book/978-0-387-
78932-3?changeHeader
55. Mureşan, M.: Soft landing on Moon with Mathematica. Math.r J. 14 (2012). doi:
dx.doi.org/doi:10.3888/tmj.14–16
56. Mureşan, M.: On Zermelo’s navigation problem with Mathematica. J. Appl. Funct. Anal.
9(3–4), 349–355 (2014)
57. Mureşan, M.: On the maximal orbit transfer problem. Math.r J. 17 (2015). doi:
dx.doi.org/10.3888/tmj.17–4
58. Palais, R.S.: The visualization of mathematics: towards a mathematical exploratorium. Not.
Am. Math. Soc. 46(6), 647–658 (1999)
59. Rao, K.S.: Ramanujan and important formulas. In: Nagarajan, K.R., Soundararajan, T. (eds.)
Srinivasa Ramanujan: 1887–1920: A Tribute, pp. 32–41. MacMillan India, Madras (1988)
60. Robinson, E.A., Jr.: Man Ray’s human equations. Not. Am. Math. Soc. 62(10), 1192–1198
(2015)
61. Sedgewick, R.: Implementing Quicksort programs. Commun. ACM 21(10), 847–857 (1978)
62. Strogatz, S.H.: Nonlinear Dynamics and Chaos: with Applications to Physics, Biology,
Chemistry, and Engineering. Perseus, New York (1994)
63. Takano, K.: Pi no arctangent relation wo motomete [finding the arctangent relation of ]. Bit
15(4), 83–91 (1983)
64. Torrence, B., Torrence, E.: The Student’s Introduction to Mathematicar . A Handbook for
Precalculus, Calculus, and Linear Algebra, 2nd edn. Cambridge University Press, Cambridge,
UK (2009)
65. Trott, M.: The Mathematica GuideBook for Graphics. Springer, New York (2004)
66. Trott, M.: The Mathematica GuideBook for Programming. Springer, New York (2004)
67. Trott, M.: The Mathematica GuideBook for Numerics. Springer, New York (2006)
68. Trott, M.: The Mathematica GuideBook for Symbolics. Springer, New York (2006)
69. Weisstein, W.E.: BBP-Type formula. Technical report, MathWorld-A Wolfram Web Resources.
mathworld.wolfram.com/BBP-TypeFormula.html
70. Weisstein, W.E.: Machbin-Like formulas. Technical report, MathWorld-A Wolfram Web
Resources. mathworld.wolfram.com/Machin-LikeFormulas.html
71. Weisstein, W.E.: Pi formulas. Technical report, MathWorld-A Wolfram Web Resources.
mathworld.wolfram.com/PiFormulas.html
72. Wolfram, S.: The Mathematicar Book, 5th edn. Wolfram Media, Champaign (2003)
73. Wolfram, S.: An Elementary Introduction to the Wolfram Language. Wolfram Media,
Champaign (2015)
74. Wolfram, S.: Differential Equation Solving with DSOLVE. Wolfram Research, Champaign
(2008). Wolfram Mathematicar Tutorial Collection. htpps:/reference.wolfram.com/language/
tutorial/DSolveOverview.html
75. Wolfram, S.: Advanced numerical differential equation solving in Mathematica. In: Wolfram
Mathematicar Tutorial Collection. Wolfram Research, Champaign (2008). htpps:/www.scrib.
com/doc/122203558/Advanced-Numerical-Differential-Equation-Solving-in-Mathematica
76. Wolfram, S.: http://numbers.computation.free.fr/Constants/Pi/piclassic.html
77. Wolfram, S.: http://numbers.computation.free.fr/Constants/Pi/piramanujan.html
78. Zermelo, E.: Über die Navigation in der Luft als Problem der Variationsrechnung. Jahresbericht
der deutschen Mathematiker – Vereinigung, Angelegenheiten 39, 44–48 (1930)
79. Zermelo, E.: Über das Navigationproblem bei ruhender oder veränderlicher Windverteilung.
Z. Angew. Math. Mech. 11(2), 114–124 (1931)
Index

A Maltese, 74
altitude, 96 curvature, 100
angle bisector, 96 curve
ascent Maria Agnesi’s, 212
Lunar, 179 pedal, 220
planar, 181 smooth, 100
cycloid, 200

B
BBP like D
binomial sum, 163 deltoid, 211
bean, 77
bifurcation
hopf, 230 E
binormal, 99 epicycloid, 202
butterfly, 78 equation
differential
Van der Pol, 228
C
cardioid, 205
centroid, 96 F
circle foot, 96
nine points formula
Euler’s, 97 BBP, 158
circumcenter, 95 binomial
circumcircle, 96 Newton’s, 94
circumradius, 96 de Bruijn, 83
coefficient function
binomial, 17 zeta
conchoid of Nicomedes, 197 Riemann, 20
conjecture
Collatz, 60
Euler’s, 14 H
Fermat, 16 Hamiltonian, 171
cross, 74 heap, 41

© Springer International Publishing AG 2017 263

M. Mureşan, Introduction to Mathematicar with Applications,
DOI 10.1007/978-3-319-52003-2
264 Index

hyperboloid P
of one sheet, 102 pair
hypocycloid, 207 admissible, 171
feasible, 171
Pythagorean, 13
I paraboloid
incenter, 96 hyperbolic, 90
incircle, 96 parameter
natural, 100
perpendicular bisector, 95
L pivot, 52
lemniscata point
Bernoulli, 213 equilibrium, 117
limit cycle, 227 saddle, 90
list, 24 Preface, ix
problem
M Mayer, 171
map navigation
Hénon, 223 Zermelo’s, 173
mean sorting, 35
arithmetic, 12
geometric, 12
harmonic, 12 Q
median, 96 quadrifolium, 214
method
Runge-Kutta, 124
diagonally implicit, 129
explicit, 124, 129 R
semi-explicit, 124, 129 rhodonea, 76
multiplier, 171 rose, 76
multipliers
method of Lagrange, 252
S
sort
N heap, 41
nephroid, 205 insertion, 38
Freeth’s, 206 merge, 41
normal, 99 quicksort, 52
number selection, 37
Bell, 18 star
polygonal, 134 of David, 75
numbers surface
Pythagorean, 13 minimal
Costa’s, 243
Möbius, 89
O
orbit
parking, 179
orthocenter, 96 T
orthoptic, 214 tangent, 99
ovals vector
Cassini, 213 unit, 100
Index 265

tetrahedron trihedron
Reuleaux, 245 Frenet-Serret, 99, 235
timing, 53
torus, 237
trajectory W
quasiplanar, 181 window
tree Viviani’s, 101, 233
binary, 42