PROGRAMMING

MATHEMATICS
USING MATLAB®
PROGRAMMING
MATHEMATICS USING
MATLAB®

LISA A. OBERBROECKLING
Department of Mathematics and Statistics
Loyola University Maryland
Baltimore, MD, United States
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2021 Elsevier Inc. All rights reserved.

MATLAB® is a trademark of The MathWorks, Inc. and is used with permission.

The MathWorks does not warrant the accuracy of the text or exercises in this book.
This book’s use or discussion of MATLAB® software or related products does not constitute endorsement or
sponsorship by The MathWorks of a particular pedagogical approach or particular use of the MATLAB® software.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopying, recording, or any information storage and retrieval system, without
permission in writing from the publisher. Details on how to seek permission, further information about the
Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center
and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other
than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our
understanding, changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any
information, methods, compounds, or experiments described herein. In using such information or methods they
should be mindful of their own safety and the safety of others, including parties for whom they have a professional
responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability
for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or
from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-817799-0

For information on all Academic Press publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Katey Birtcher

Editorial Project Manager: Rafael G. Trombaco
Production Project Manager: Beula Christopher
Designer: Bridget Hoette
Typeset by VTeX
To Rob and James, for the support, patience (especially when teaching MATLAB), and
laughter.

To Christos Xenophontos for your encouragement and for introducing me to MATLAB.

Contents

Preface xiii
Introduction xv

Part 1. MATLAB®

1. Introduction to MATLAB® 3
1.1. Basic MATLAB® information 3
1.1.1. Starting MATLAB 3
1.1.2. Good commands to know 3
1.2. Basic mathematics 4
1.2.1. Built-in mathematical functions 5
1.2.2. Precedence rules 6
1.2.3. Formats 8
1.3. Variables 9
1.4. Diaries and script files 10
1.5. Exercises 12

2. Vectors and Matrices (Arrays) 15

2.1. One-dimensional arrays (vectors) 15
2.1.1. Constant spaced vectors 15
2.1.2. Equally spaced vectors 16
2.2. Two-dimensional arrays (matrices) 17
2.3. Addressing elements of vectors/arrays 18
2.4. Component-wise calculations 22
2.5. Random numbers 25
2.6. Exercises 28

3. Plotting in MATLAB® 33
3.1. Basic 2D plots 33
3.2. Bad domain examples 34
3.3. Axis settings 35
3.4. Multiple plots 40
3.5. Color, line, and marker modifications 43
3.5.1. Clf/close all 46
3.5.2. Subplots 46
3.6. Other 2D plots 49
3.6.1. Parametric curves 49
3.6.2. Polar curves 50
3.7. Exercises 52

4. Three-Dimensional Plots 59
4.1. Vector functions or space curves 59

vii
viii Contents

4.2. Plotting surfaces 62

4.2.1. The meshgrid command 63
4.2.2. Domain issues 64
4.2.3. Level curves 65
4.2.4. Multiple plots and modifying colors 66
4.3. View command 68
4.4. Axis settings, revisited 70
4.5. Other coordinate systems and 3D graphs 72
4.5.1. The sphere and cylinder commands 72
4.5.2. Cylindrical coordinates 75
4.5.3. Spherical coordinates 77
4.6. Exercises 78

5. Functions 83
5.1. The lookfor and help commands 83
5.2. File format 84
5.3. Function examples 86
5.3.1. Basic function examples 86
5.3.2. More function examples – multiple inputs 86
5.3.3. Multiple outputs 87
5.3.4. Bad examples 89
5.4. Exercises 90

6. Control Flow 93
6.1. Relational and logical operators 93
6.2. If statements 97
6.3. Switch/case 99
6.4. Use of characteristic functions 99
6.5. For loops 100
6.6. While loops 102
6.7. Useful commands break, continue, return, and error 103
6.8. Optional inputs and outputs of functions 104
6.9. Exercises 107

7. Miscellaneous Commands and Code Improvement 115

7.1. Miscellaneous commands 115
7.1.1. The fprintf command 115
7.1.2. The sprintf command 117
7.1.3. Formats revisited 118
7.1.4. The save/load commands 118
7.1.5. The tic/toc commands 119
7.1.6. The fill command 119
7.1.7. The command alpha 122
7.1.8. The syms, diff, int, and subs commands 125
7.1.9. Commands for polynomials 127
 Contents ix

7.2. Code improvement 129

7.2.1. Vectorization of code 130
7.2.2. Preallocation 131

Part 2. Mathematics and MATLAB®

8. Transformations and Fern Fractals 135

8.1. Linear transformations 135
8.2. Affine transformations 139
8.3. Fern fractals 140
8.4. Exercises 141

9. Complex Numbers and Fractals 147

9.1. Complex numbers 147
9.1.1. Adding complex numbers 147
9.1.2. Multiplication by a real numbers (scalars) 147
9.1.3. Multiplication and de Moivre’s theorem/formula 147
9.1.4. Plotting complex numbers in MATLAB® 150
9.1.5. Creating line segments with complex numbers 151
9.2. The Chaos Game 153
9.3. Line replacement fractals 154
9.3.1. Snowflake fractals 154
9.3.2. Gosper Island 155
9.4. Geometric series 156
9.5. Exercises 158

10. Series and Taylor Polynomials 167

10.1. Review of series 167
10.2. Power series 169
10.3. Taylor polynomials and Taylor series 173
10.4. Exercises 177

11. Numerical Integration 183

11.1. Approximating integrals/numerical integration 183
11.2. Riemann sums 183
11.3. Error bounds 185
11.4. Simpson’s rule 186
11.5. Exercises 189

12. The Gram–Schmidt Process 193

12.1. General vector spaces and subspaces 193
12.1.1. Vector spaces 193
12.1.2. Subspaces 194
12.2. Linear combinations of vectors 195
12.3. Linear independence and bases 196
x Contents

12.3.1. Linear independence 196

12.3.2. Bases 197
12.4. Rank 200
12.5. Orthonormal vectors and the Gram–Schmidt process 201
12.5.1. Orthogonal and orthonormal vectors 201
12.5.2. The Gram–Schmidt process 204
12.6. Answers to example problems 210
12.7. Exercises 211

A. Publishing and Live Scripts 215

A.1. Live scripts 215
A.2. Basic scripts or M-files 215
A.3. Publishing M-files 216
A.4. Using sections 216
A.4.1. Using sections for publishing 217
A.4.2. Using sections for running/debugging files 222
A.5. Formatting text 223
A.5.1. Basic text formatting 223
A.5.2. Lists 224
A.5.3. HTML links 225
A.5.4. Inserting images 225
A.5.5. Pre-formatted text 226
A.5.6. Inserting HTML code 227
A.5.7. Inserting LATEXequations 227

B. Final Projects 229

B.1. Ciphers 229
B.1.1. Substitution cipher 229
B.1.2. Columnar transposition cipher 230
B.2. Game of Pig 231
B.3. Linearization and Newton’s method 232
B.3.1. Linearization 232
B.3.2. Newton’s method 232
B.4. Disk and Shell method 234
B.5. Power ball data 235

C. Linear Algebra Projects 237

C.1. Matrix calculations and linear systems 237
C.1.1. First handout 237
C.1.2. Exercises 239
C.2. The Hill cipher 243
C.2.1. Useful commands 245
C.2.2. Exercises 251
C.3. Least-squares solutions 252
C.3.1. Brief overview 252
 Contents xi

C.3.2. Curve fitting 253

C.3.3. Exercises 254
C.4. Markov matrices 256
C.4.1. Brief overview 256
C.4.2. Exercises 256

D. Multivariable Calculus Projects 261

D.1. Lines and planes 261
D.2. Vector functions 262
D.2.1. 2D example plots 262
D.2.2. 3D example plot 262
D.2.3. Bad domain example 263
D.2.4. Adjusting the view 263
D.2.5. Sphere command 263
D.2.6. Multiple plots on one figure 265
D.2.7. Exercises 265
D.3. Applications of double integrals 266
D.3.1. Calculating integrals and viewing regions 266
D.3.2. Exercises 268

References 271
Index 273
Preface

This book started in 2004 when I started to use MATLAB® in my courses at Loyola
University Maryland. I started including a few MATLAB projects within the intro-
duction to linear algebra and multivariable calculus courses. I also taught a one-credit
MATLAB course that was required of every mathematics major. Later the course was
changed to a three-credit course. I expanded the previous assignments and added some
others. The class notes and assignments from over the years have expanded into this
book.
My philosophy has always been to use MATLAB to practice basic programming skills
with mathematics topics students had seen previously, such as numerical integration, and
topics they likely had not seen such as fractals. Visualizing mathematics has always been
important.

Supplements
Student companion site: Please visit https://www.elsevier.com/books-and-journals/
book-companion/9780128177990
Instructor-only site: Qualified instructors can register and access teaching materials
at https://textbooks.elsevier.com/web/Manuals.aspx?isbn=9780128177990

xiii
Introduction

The goal of the course and thus book is to introduce MATLAB® and to practice basic
programming techniques. There is a lot more to MATLAB than what is covered in this
book. Most students have already had some programming experience in another lan-
guage before taking the course this book has stemmed from, although it is not necessary.
On completion of the course/book, one should be familiar enough with MATLAB to
explore more complicated features and commands. Deepening your understanding of
mathematics and learning new topics are bonuses!

xv
CHAPTER 1

Introduction to MATLAB®
1.1. Basic MATLAB® information
1.1.1 Starting MATLAB
MATLAB has many different windows or panels, the first three of which are on the
main screen by default (see Table 1.1).
You can always modify the layout of the panels including “docking” or “undock-
ing” them, in the “Home” view, select “Layout” from the menu and the top item is
“Default.” In order to use MATLAB successfully, you should pay attention to the Cur-
rent Folder. Otherwise, MATLAB may not be able to save and run your programs/files
successfully.

1.1.2 Good commands to know

The first four commands are useful to “start fresh” without closing and reloading MAT-
LAB.
• clc Clears the command window of all previous commands and output. These
commands are still stored in the command history and can be accessed with the
up-arrow.
• clear Clears all defined variables in memory. BE CAREFUL WITH THIS COM-
MAND! You can also clear certain variable names by typing clear varname1
varname2.
• clf and close Clears the current figure (plot). If no figure window was open, it
will open a blank figure window. Another command is close. This will close the
current figure window. The command close all will close all figure windows.
Note that there is no clf all. Both of these commands have other variations to

Table 1.1 Main MATLAB Windows.

Command Window Enter commands and variables, run programs
Workspace Window Information about the current variables
Current Folder Window Shows the files in the current folder/directory
Command History Window History of commands entered in Command Window
Figure Window Output from graphic commands
Editor Window Creates and debugs script and function files
Help Window Help information
Launch Pad Window Access to tools, demos, and documentation

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00006-5 All rights reserved. 3
4 Programming Mathematics Using MATLAB®

clear/close named figures, etc. such as clf(2) and close(2) that will clear or close
Figure 2, respectively.
• format Sends the output display back to the default format. The command format
compact will not have as much white space (blank lines) between subsequent
commands and output.
• exit or quit Quit MATLAB. You can also quit/exit MATLAB by using the
MENU option under “File” or the usual close application icon within a Mac or
Windows environment.

Other good commands

• who Lists current variables
• help command Displays the help for command. For example, help atan2.
• disp('text') Displays text as output in the command window.
• exist text or exist('text') Checks if variables or functions are defined (see
help exist for more details).
• lookfor text Searches for “text” as a keyword in help entries of functions.

>> exist average

ans =

>> lookfor average

mean - Average or mean value.
HueSaturationValueExample - Compute Maximum Average HSV of Images with MapReduce
ewmaplot - Exponentially weighted moving average chart.

1.2. Basic mathematics

Mathematical operations
MATLAB uses the typical symbols for addition, subtraction, multiplication, division,
and exponentiation (+, -, ∗, /, and ^). These are considered matrix arithmetic op-
erations and follow the rules from linear or matrix algebra.

>> 10/3

ans =

3.3333

MATLAB has another division operator, the “divided into” operator.

 Introduction to MATLAB® 5

>> 10\3

ans =

0.3000

Thus the forward slash is our usual division, “3 divided by 2” while the backslash is “3
divided into 2.” The need for both of these become more apparent when working with
matrices.
“Dot” operations or component-wise operations are useful and/or necessary for
use with vectors and matrices (discussed in Section 2.4). These are considered “array
arithmetic operations” and are carried out element or component-wise.
MATLAB does NUMERICAL, rather than algebraic computations, as seen below.
Think about what is expected versus what is given as the answer to the subraction
calculation.

>> asin(1/2)

ans =

0.5236

>> pi/6 - asin(1/2)

ans =

-1.1102e-16

1.2.1 Built-in mathematical functions

Table 1.2 has common mathematical functions in MATLAB. This is not a complete list.
The modulo function mod calculates the modulus after division. In other words,
mod(x,y) returns the remainder when you divide x by y. In some languages like Python
or Perl this is equivalent to % is the modulus operator. Thus mod(x,y) in MATLAB is
the same as x % y in other languages. There is also a remainder function rem that has
the same functionality as mod EXCEPT when the divisor and quotient are opposite
signs or the quotient is 0 (see Exercise 9).

>> mod(10,3)

ans =

1
6 Programming Mathematics Using MATLAB®

>> rem(10,3)

ans =

>> mod(10,-3)

ans =

-2

>> rem(10,-3)

ans =

>> mod(10,0)

ans =

10

>> rem(10,0)

ans =

NaN

NOTE: NaN stands for “not a number.”

1.2.2 Precedence rules

While some languages go strictly from left to right when there are multiple operations
within one line, most now follow the mathematical rules for order of operations. Many
of you may know the pneumonic PEMDAS (Please excuse my dear Aunt Sally); paren-
theses, exponentiation, multiplication, division, addition, subtraction. There are some
issues with that pneumonic, however. There are also some discrepancies, to be explored
in the exercises. Table 1.3 shows the precedence rules for the arithmetic operations
discussed so far, going from highest precedence to lowest.
One thing to consider is: where do functions go on this list? For example, how does
MATLAB interpret the command cos(x)^2? Is this equivalent to cos(x2 ), cos2 (x), or
something else entirely? This will be explored in the exercises.
 Introduction to MATLAB® 7

Table 1.2 Mathematical functions.

sqrt(x) Square root
exp(x) Exponential (ex )
abs(x) Absolute value
mod(x,y) Modulus
rem(x,y) Remainder
log(x) Natural logarithm (ln(x))
log10(x) Common logarithm
sin(x) Sine of x (radians)
sind(x) Sine of x (degrees)
cos(x) Cosine of x
tan(x) Tangent of x
cot(x) Cotangent of x
asin(x) Inverse sine of x (radians)
asind(x) Inverse sine of x (degrees)
pi π

Table 1.3 Basic math precedence.

( ) parentheses
^ exponentiation
∗ / \ multiplication and division
+ - addition and subtraction

If you have a long calculation/expression, you can continue on the next line in
the command window or within a MATLAB file with the ellipses or continuation
operator. Note when the ellipses work with/without spaces:

>> 1+2*3-12^2/3 ...

*2

ans =

-89

>> 1+2*3-12^2/3 *...

2

ans =

-89

>> 1+2*3-12^2/3*...
2
8 Programming Mathematics Using MATLAB®

ans =

-89

>> 1+2*3-12^2/3...
1+2*3-12^2/3...
|
Error: Unexpected MATLAB operator.

1.2.3 Formats
As mentioned above, the format command returns the format back to the default
format, which is the same as format short. Generally speaking, this will display a
number up to four decimal places, while format long will display 15. In scientific
notation, this amounts to five and 16 significant digits, respectively. See Table 1.4.

>> pi

ans =

3.1416

>> format long

>> pi

ans =

3.141592653589793

There are other built-in formats, including how numbers in scientific notation are dis-
played. See help format for more examples.
Another useful command is format compact and format loose (default). This
will change how the output is displayed.

>> pi

ans =

3.1416

>> format compact

>> pi
ans =
3.1416
>> format loose
 Introduction to MATLAB® 9

>> pi

ans =

3.1416

>>

The subsequent commands shown in this text will use format compact (see Table 1.4).

Table 1.4 Basic formats displaying 10∗pi.

format short (default) 31.4159
format long 31.415926535897931
format rat 3550/113
format bank 31.42
format short e 3.1416e+01
format long e 3.141592653589793e+01
format short g 31.416
format long g 31.4159265358979
format hex 403f6a7a2955385e
format compact (no blank lines)
format loose (default) (some blank lines)

1.3. Variables
The format for a variable assignment is as follows:
Variable name = Numerical value or computable expression
Some conventions:
• The = is the assignment operator which assigns a value to a variable.
• Left-hand side can include only one variable name.
• Right-hand side can be a number or an expression made up of numbers, functions,
and/or variables previously assigned numerical values.
• Variables must begin with a letter.
• Press the Enter/Return key to make the assignment.
• The variable ans is the value of the last expression that is not assigned.
• Be careful with variable names. For example, do not name a variable help or sin.
• Variable names are case sensitive; thus A is not the same as a.
Remember:
• Use semicolon (;) to suppress screen output.
• Multiple commands can be typed on one line by typing a comma (,) between them
if they are not already ended with a semicolon (;).
10 Programming Mathematics Using MATLAB®

√
Example 1.3.1. Assign the number 3 to variable a and 4 to variable b. Print a2 + b2
and assign the solution to the variable c.

>> a=3; b=4; c = sqrt(a^2+b^2), a+b+c

c =
5
ans =
12

Notice in the above example, you do not need spaces around the “=” for variable
assignments but you may use them for aesthetic reasons.

Example 1.3.2. Experiment with the equation

x tan x + sin x
cos2 =
2 2 tan x
by calculating each side of the equation for x = π/5.

>> x = pi/5;
>> LHS = (cos(x/2))^2, RHS = (tan(x)+sin(x))/(2*tan(x))
LHS =
0.9045
RHS =
0.9045
>> format long
>> LHS, RHS
LHS =
0.904508497187474
RHS =
0.904508497187474
>> LHS-RHS
ans =
-1.110223024625157e-16

1.4. Diaries and script files

You can record your commands and output to the command window with the diary
command. The commands you enter in the command window and any output are
stored as an ASCII (plain text) file. The command diary toggles the recording on and
off. If you do not specify a filename, it will create a file in the current folder of the
name “diary.” The command diary filename will save the recording to a file of the
name “filename.” The commands diary off and diary on will pause and restart the
recording, respectively, to the active file. NOTE: when you use the diary filename
 Introduction to MATLAB® 11

more than once (within the same current folder), it will continue to APPEND to the
file.

>> diary filename

>> 1+1
ans =
2
>> diary off
>> 2+2
ans =
4
>> diary on
>> 3^2
ans =
9
>> diary

For example, the commands above will generate the following text in the file “file-
name”:
1+1
ans =
2
diary off
3^2
ans =
9
diary

Script files, or m-files, are extremely useful for running and rerunning code. You
may be required to turn in script files for your assignments.
• Script files are ASCII files (plain text files) with extension .m; thus they are also
called m-files. These are basically batch files.
• When you run a SCRIPT file, MATLAB executes each line as if you typed each
line into the command window.
• Script files are very useful; you can edit them, save them, execute them many times
and “tweak” them to experiment with commands.
• The MATLAB editor window is the best way to create and edit script files.
• To avoid extraneous output to the command window, put “;” after variable assign-
ments or intermediate calculations.
• Comments within MATLAB files begin with the percent symbol (%).
Running script files:
There are many ways to run an m-file of name filename.m. First, you must MAKE
SURE CURRENT FOLDER IS CORRECT!
12 Programming Mathematics Using MATLAB®

1. >> filename
2. >> run filename
3. >> run('filename')
4. Within the Editor tab, chose run...

1.5. Exercises
1. Basic calculations Use MATLAB to do the following calculations. Be careful! The
following are displayed using regular mathematical notation; you need to figure out
what MATLAB functions are needed.
7 .753 292 644/3
(a) (2.4)(64 ) + 8 , (b) + + 20 · 9−3 ,
16 2 − 225 5 11
(c) cos(360), (d) cos(360◦ ),

(e) cos(2π), (f) cos(2π ◦ ),

(g) e + 5, (h) |π − 5|,

(i) 3 ln 7, (j) 3 log 7,

6
(k) sin−1 (0.5) + 4, (l) 4 cos(5 arctan(13/4)).
π
2. Order of operations
(a) Calculate, without using any parentheses, −42 using THREE of the following
and write your answers on your own paper, specifying which you did and
what answers they gave:
(i) calculator (specify type/model) (ii) Google.com
(iii) Excel or Google Sheets (iv) Desmos.com or Wolfram Al-
pha
Now calculate −42 using MATLAB. Are there differences in the answers?
Based on your knowledge of Order of Operations, what should be the an-
swers?
(b) Do the same for the calculation of − cos(π/4)2 (you may use parentheses
around the π/4; i.e. you should calculate -cos(pi/4)^2). NOTE: In Excel,
to calculate with π use “PI()”, again noting the differences (if any) in the
answers.
(c) Do the same for the calculation of −212 and −81/3 (without using any paren-
theses). Should parentheses be used to get the proper calculations? Where?
(d) How should Table 1.3 for precedence rules within MATLAB be changed or
expanded to include functions and negation (unary minus)? Answer this by
rewriting the table on paper.
 Introduction to MATLAB® 13

3. Using variables. Define variables with the assignments x = 8, y = 3.5, and X =

√
1/9. Calculate the following within MATLAB. For the n z calculations, use the
nthroot function. √
4(y − x) 3 X
(a) , (b) ,
3X − 20 10
√3
(c) 3 cos x tan y, (d) e(X +y)/x + 3 x .
4. Suppose x = 3 and y = 5 (define the variables at the beginning of the problem).
 
1 −1 3y 4(y − 5)
(a) 3π x , (b) 1 − 5
2
, (c) , (d) , (e) 2 sin(x) cos(y).
x 4x − 8 3x − 6
5. Define the following variables: tablePrice = 1256.95, chairPrice = 89.99, and
gasPrice = 3.499. Using the variables and format bank for parts a–c, write your
answers to the following questions on paper, interpreting the MATLAB output as a
meaningful answer:
(a) Find the cost of one table and eight chairs.
(b) Find the same cost as above but with 6.5% sales tax.
(c) Find the cost of 14.25 gallons of gas that you would have to pay.
(d) Compute the actual cost of 14.25 gallons of gas using the default format.
6. More calculations Define the variables x = 256 and y = 125. Calculate the follow-
ing within MATLAB. When radical notation is used in the problem, use the sqrt
and nthroot functions and use exponential calculations when exponential notation
is used in the problem.
√ 1 √ 1 1
(a) x, (b) x 2 , (c) −x, (d) (−x) 2 , (e) x 4 ,
√ 1 1 1 √
(f) 4
x, (g) y 3 , (h) −y 3 , (i) (−y) 3 , (j) 3 −y.

(k) From the above calculations, do you see anything surprising with the answers?
1 √
(l) Calculate (−x) 4 and 4 −x. What are the differences?
(m) Calculate (−8)2/3 and 82/3 on paper using your exponent rules. Now do the
calculations within MATLAB, Excel, Google.com, WolframAlpha.com, and
a scientific or graphing calculator (specifying what model you have used).
What are the differences, if there are any?
7. Calculator precision √
(a) Within an Excel spreadsheet or Google Sheet, calculate 12 178212 + 184112
using exponential notation for the calculations. Write your answer clearly.
(b) Rewrite the above expression with your answer to part (a) into an equation
and algebraically simplify the equation so there are no radicals or rational
exponents. √
(c) Now calculate 12 178212 + 184112 using MATLAB.
(d) Compare the left-hand side and right-hand side of the equation you get in
part (b) by subtraction within MATLAB, using format long).
14 Programming Mathematics Using MATLAB®

(e) Calculate “3 quadrillion and 18 minus 3 quadrillion and 14” in your head
and write it down. Now translate this into mathematics so you can calculate
it within Google.com, Excel, and MATLAB. (You may need to look up how
many zeros you will need!) Compare your answers in a table.
(f) Do the same with 2.000000000000018 − 2.000000000000014.
8. Ambiguities with notation. Define variables with the assignments x = 10 and
y = π/4. Calculate the following within MATLAB. You may have to adjust from
mathematical notation to correct MATLAB notation. Make sure you are using the
default format!

(a) cos y, (b) cos y2 , (c) cos(y2 ), (d) cos2 y, (e) (cos y)2 ,

(f) x−1 , (g) cos−1 (x/20), (h) cos(x/20)−1 , (i) (cos(x/20))−1 .

(j) Redo part (g) and then use the MATLAB variable ans to calculate

cos−1 (x/20)
.
4y
(k) Are any of the above calculations ambiguous in how they are written (which
ones and why)? What could be done to make the calculations clearer to the
person performing/entering the calculations?
9. Exploring rem and div. It is common to use either of the functions mod or rem
to tell whether positive integers are even or odd, among other uses. We will explore
the differences and similarities of these functions.
(a) Here is another simple use for these functions. You are given a list of 10-digit
numbers. You would like to only use the last seven digits of these numbers (for
example, for display purposes). Use both the mod or rem functions to easily
get the last seven digits of the number 4108675309. Do you see a difference
in their use for this?
(b) Use the same commands on the number -4108675309. Do you see a differ-
ence? Explain in your own words what you think is the difference between
the mod and rem functions. Is there a difference when using these functions
to tell whether any integer is even or odd?
(c) Can you come up with a way, using MATLAB functions such as mod, rem,
round, ceil, fix, etc. to capture the “area code” (first three digits) of
4108675309? Experiment with at least two phone numbers with different
area codes.
(d) What about capturing the “central office” part of the number (867)? Do it
for 4108675309 and 4106172000.
CHAPTER 2

Vectors and Matrices (Arrays)

2.1. One-dimensional arrays (vectors)
Arrays are used to store and manipulate numbers. They are arranged in rows and/or
columns and are defined within MATLAB® using brackets [ ].
One-dimensional arrays, usually called vectors, can represent points or vectors in
space (of any dimension), or can be used to store data. For example: the point (1, 2)
in 2D or the point (1, 2, −5) in 3D, or the data (70, 75, 72, 77). These vectors can be
represented as rows or columns.
To enter a row vector, use spaces or commas between numbers.

>> x = [1 2 3], y = [4, 5, 6]

x =
1 2 3
y =
4 5 6

To enter a column vector, use semi-colons between numbers:

>> y = [4;5;6]
y =
4
5
6

There are ways to define vectors without having to enter every element. The two most
common ways are to define constant spaced and equally spaced (linearly spaced) vectors.

2.1.1 Constant spaced vectors

If the difference between the elements (increment) of the vector is important, use
“colon-notation”. For a vector from m to n, incremented by q, use the format

variablename = [m: q : n]

If no increment q is given (as in the variable x below), it is 1 by default.

>> x=[1:7], y=[1:2:6], z=[1:3:25], w=[1:3:24]

x =
1 2 3 4 5 6 7
y =

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00007-7 All rights reserved. 15
16 Programming Mathematics Using MATLAB®

1 3 5
z =
1 4 7 10 13 16 19 22 25
w =
1 4 7 10 13 16 19 22

Notice in the vectors y and w above that the last number (n) may not actually be a value
in the vector if the increment (q) is such that it will not occur based on the starting
value (m).
One can also increment backwards with a negative increment value.

>> a=[6:-1:1], b=[10:-2:-2]

a =
6 5 4 3 2 1
b =
10 8 6 4 2 0 -2

2.1.2 Equally spaced vectors

If the number of elements in the vector is important but not necessarily the actual
values, use linspace. The command linspace stands for linearly spaced. This is
especially useful for plotting. The command
variablename = linspace(m,n)
will give you a vector of 100 equally-spaced elements between m and n. For a vector of
q elements from m to n, use
variablename = linspace(m,n,q)

>> x = linspace(1,7,3), y = linspace(0,1,5)

x =
1 4 7
y =
0 0.2500 0.5000 0.7500 1.0000

Notice that both the values m and n are elements in the vector.
There is also a command logspace that is also useful for plotting (among other
things). The command logspace(a,b) will generate a row vector of 50 logarithmically
spaced points from 10a and 10b . If b is pi, then the points are between 10a and π .

variablename = logspace(a,b)
variablename = logspace(a,b,q)

Just as in linspace, you can specify the number of elements in the vector by stating
the value q.
 Vectors and Matrices (Arrays) 17

>> logspace(0,5,6)
ans =
1 10 100 1000 10000 100000
>> logspace(5,1,5)
ans =
100000 10000 1000 100 10
>> linspace(1,pi,4)
ans =
1.0000 1.7139 2.4277 3.1416
>> logspace(0,pi,4)
ans =
1.0000 1.4646 2.1450 3.1416
>> logspace(2,pi,3)
ans =
100.0000 17.7245 3.1416

2.2. Two-dimensional arrays (matrices)

Two-dimensional arrays or matrices can be used for many things. You can use matrices
to store data or information as a table or spreadsheet. Matrices can also be used to solve
systems of linear equations such as

2x + 3y + z = 4,
x − 5y + 3z = 3,
4x − 2y + 3z = 2.

The format for defining matrices expands on the format for defining row and column
vectors; spaces or commas between elements within a row with semicolons between
rows.
variable = [1st row ; 2nd row ; ... ; last row]
You can also use vectors of the same size to be rows (or columns) of a matrix, as we do
to define the matrix B below.

>> A = [2 3 1;1 -5 3;4 -2 3; 0 1/2 66], x=0:4;...

y=linspace(0,pi,5); B=[x;y]
A =
2.0000 3.0000 1.0000
1.0000 -5.0000 3.0000
4.0000 -2.0000 3.0000
0 0.5000 66.0000
B =
0 1.0000 2.0000 3.0000 4.0000
18 Programming Mathematics Using MATLAB®

0 0.7854 1.5708 2.3562 3.1416

2.3. Addressing elements of vectors/arrays

For many reasons, we may want to capture one element, or a portion of a vector or
matrix. This section explains several ways to accomplish this.

Elements of vectors
• v(k) picks the kth element of v.
• v(m:n) picks the mth through the nth elements of v.

>> v = linspace(0,1,5)
v =
0 0.2500 0.5000 0.7500 1.0000
>> v(3)
ans =
0.5000
>> v(2:4)
ans =
0.2500 0.5000 0.7500

Notice the difference if v is a matrix.

>> v=[1 2 3;4 5 6]

v =
1 2 3
4 5 6
>> v(2)
ans =
4
>> v(3)
ans =
2
>> v(3:6)
ans =
2 5 3 6

Elements of matrices
• A(m,n) picks the (m, n)th element (element in the mth row, nth column) of the
matrix A.
• A(m:n, p:q) gives the submatrix from the elements (m : n) × (p : q)
• A(m:n, :) gives rows m through n, and every column of A
 Vectors and Matrices (Arrays) 19

• A(:, p:q) gives every row, and columns p through q of A

>> A = [2 3 1;1 -5 3; 4 -2 3; 0 1 6]
A =
2 3 1
1 -5 3
4 -2 3
0 1 6
>> A(3,2) % 3rd row, 2nd column of A
ans =
-2
>> A(2:4,2:3) %2-4 rows, 2-3 columns of A
ans =
-5 3
-2 3
1 6
>> A(:,1) %every row, 1st column of A
ans =
2
1
4
0
>> A(2:3,:) %2-3 row, every column of A
ans =
1 -5 3
4 -2 3

Adding elements to arrays

You can easily add elements to arrays. You must be careful that what you are adding is
of appropriate size.

>> B = [1 4 2 3;3 6 9 2;1 4 9 7]

B =
1 4 2 3
3 6 9 2
1 4 9 7
>> C=[B; 2 5 1 8] % adding a row to B
C =
1 4 2 3
3 6 9 2
1 4 9 7
2 5 1 8
>> D = [C [1;2;3;4]] % adding a column to C
D =
1 4 2 3 1
3 6 9 2 2
20 Programming Mathematics Using MATLAB®

1 4 9 7 3
2 5 1 8 4

Deleting elements
One can delete elements by assigning nothing to these elements.

>> v=[2:2:10]
v =
2 4 6 8 10
>> v(2)=[]
v =
2 6 8 10
>> v(2:3)=[]
v =
2 10
>> B
B =
1 4 2 3
3 6 9 2
1 4 9 7
>> B(2:3,:) = [ ]
B =
1 4 2 3
>> C
C =
1 4 2 3
3 6 9 2
1 4 9 7
2 5 1 8

>> C(:,1:2) = [ ]
C =
2 3
9 2
9 7
1 8

Notice the colon: recall that this indicates “all columns” or “all rows.”

Transpose
There are actually several commands for transposing an array, with subtle but important
differences.

>> x, A=[1 2/3; -3 pi]

x =
 Vectors and Matrices (Arrays) 21

1 4 7
A =
1.0000 0.6667
-3.0000 3.1416
>> x', A'
ans =
1
4
7
ans =
1.0000 -3.0000
0.6667 3.1416
>> x.', A.'
ans =
1
4
7
ans =
1.0000 -3.0000
0.6667 3.1416
>> transpose(x), transpose(A)
ans =
1
4
7
ans =
1.0000 -3.0000
0.6667 3.1416

>> format compact, format rat

>> B = [-1+2i, 2+3i;4-5i, 6-7i]
B =
-1 + 2i 2 + 3i
4 - 5i 6 - 7i
>> B'
ans =
-1 - 2i 4 + 5i
2 - 3i 6 + 7i
>> B.'
ans =
-1 + 2i 4 - 5i
2 + 3i 6 - 7i
>> transpose(B)
ans =
-1 + 2i 4 - 5i
2 + 3i 6 - 7i
>> ctranspose(B)
ans =
22 Programming Mathematics Using MATLAB®

-1 - 2i 4 + 5i
2 - 3i 6 + 7i

Notice that if every number is real, there are no differences between the commands.
But if there are complex numbers, there is a conjugate transpose, and a transpose (see if
you can figure out which ones are which). Other good commands are flip, fliplr,
flipud, etc.

Strings
Strings are arrays of characters rather than numbers, but addressing them can be done
in a similar way.
• An array of characters
• Created by typing characters within single quotes
• Can include letters, digits, symbols and spaces

>> s = 'MATLAB is AWESOME'

s =
'MATLAB is AWESOME'
>> s(5)
ans =
'A'
>> s(1:6)
ans =
'MATLAB'
>> s(1:6) = 'Maths*'
s =
'Maths* is AWESOME'

2.4. Component-wise calculations

When doing calculations on arrays, one must be mindful as to whether the calculation is
done component-wise or not and the sizes of what you are trying to combine. Just as in
arithmetic on matrices, addition/subtraction and scalar multiplication is automatically
done component-wise.

>> x=[1 2 3]; y=[4 5 6]; z=[10 11];

>> x+y
ans =
5 7 9
>> x+z
Error using +
Matrix dimensions must agree.
>> x+2
 Vectors and Matrices (Arrays) 23

ans =
3 4 5
>> -3*x
ans =
-3 -6 -9

Multiplication of arrays is using the matrix multiplication definition.

>> x*y
Error using *
Inner matrix dimensions must agree.
>> A=[1 2;3 4], B = [0 1;-1 1], A*B
A =
1 2
3 4
B =
0 1
-1 1
ans =
-2 3
-4 7
>> x^2
Error using ^
Inputs must be a scalar and a square matrix.
To compute elementwise POWER, use POWER (.^) instead.
>> A^2
ans =
7 10
15 22

In order to do calculations component-wise on an array, use the “.” before the oper-
ator. Thus “.∗” is component-wise multiplication, “.^2” will square every component,
etc.

>>
x.*y
ans =
4 10 18
>> x.^2
ans =
1 4 9
>> A = [1 2 3;4 5 6;-7 -8 -9]
A =
1 2 3
4 5 6
-7 -8 -9
>> A^2
ans =
24 Programming Mathematics Using MATLAB®

-12 -12 -12

-18 -15 -12
24 18 12
>> A.^2
ans =
1 4 9
16 25 36
49 64 81

Note that, for scalar multiplication, the “.∗” is not necessary, but 3.∗z or z.∗3 will still
work. For addition and subtraction, it will only work if the scalar is first.

>> 3.*z
ans =
30 33
>> z.*3
ans =
30 33
>> 3.+z
ans =
13 14
>> z.+3
z.+3
|
Error: Unexpected MATLAB operator.
>> x.+y
x.+y
|
Error: Unexpected MATLAB operator.

Mathematical functions on vectors and matrices are automatically done component-

wise.

>> abs(A)
ans =
1 2 3
4 5 6
7 8 9
>> exp(A)
ans =
2.7183 7.3891 20.0855
54.5982 148.4132 403.4288
0.0009 0.0003 0.0001
>> sind([0 30 45 60 90])
ans =
0 0.5000 0.7071 0.8660 1.0000
 Vectors and Matrices (Arrays) 25

Other useful functions

1. length
2. size
3. numel
4. end
5. sort
6. max
7. min
8. sum
9. ones
10. zeros
11. eye
12. reshape
13. and many others!

2.5. Random numbers

There are many uses for generating random numbers, vectors, or matrices. The com-
mand rand will return a uniformly distributed pseudorandom number between 0 and
1. The command randn is similar, except that it is normally distributed. To generate a
random integer between 1 and k, use randi(k). To permute the numbers from 1 to n,
use randperm(n).

>> x=rand
x =
0.6348
>> y=randn
y =
0.4598
>> z=randi(10)
z =
6
>> p=randperm(5)
p =
5 3 2 4 1

All of the above commands can be modified to generate random vectors and matri-
ces, and other modifications can be made to randi and randperm. See Table 2.1 for
a summary of the basic modifications. See the MATLAB documentation about fancier
modifications, including information on the random number generator.

>> x=rand, y=randn, z=randi(10)

x =
26 Programming Mathematics Using MATLAB®

0.1492
y =
0.7923
z =
6
>> p=randperm(5)
p =
5 3 1 2 4
>> x2=rand(2), y2=randn(2,3)
x2 =
0.1903 0.2686
0.0302 0.9827
y2 =
-1.7616 -0.4471 0.1665
-0.9166 0.1737 -1.0049
>> z2=randi(10,2), z3=randi(10,2,3), z4=randi([-10,10],1,4)
z2 =
7 3
8 9
z3 =
9 1 2
9 5 7
z4 =
-9 1 6 -7
>> p2=randperm(10,3)
p2 =
10 2 7
>> p3=randperm(10,3), z6=randi(10,1,3)
p3 =
1 5 2
z6 =
5 3 3

Note that we can modify rand to generate a random number between any two
numbers and we can modify randn to have a different mean and standard deviation
than the standard normal distribution. See examples below.

Example 2.5.1. Create 1000 random numbers between 5 and 8.

>> x=rand(1000,1)*(3) + 5;
>> histogram(x)

In Fig. 2.1 the command histogram is used to quickly display the generated random
numbers.
 Vectors and Matrices (Arrays) 27

Table 2.1 Useful random number generators.

rand Uniformly distributed number between 0 and 1
rand(n) n × n matrix of random numbers
rand(m,n) m × n matrix of random numbers
randn Normally distributed number (N (0, 1))
randn(n) n × n matrix of normally distributed numbers
randn(m,n) m × n matrix of normally distributed numbers
randi(k) Uniformly distributed integer between 1 and k
randi([a,b]) Integer between a and b
randi(k,n) n × n matrix of integers between 1 and k
randi([a,b],n) n × n matrix of integers between a and b
randi(k,m,n) m × n matrix of integers between 1 and k
randi([a,b],m,n) m × n matrix of integers between a and b
randperm(n) Random permutation of integers from 1 to n
randperm(n,k) Random permutation of k unique integers from 1 to n

Figure 2.1 Visualizing rand modification. (A) histogram(x), (B) histogram(x,[5,6,7,8]).

Example 2.5.2. Create 10,000 random numbers normally distributed with mean
of 100 and standard deviation of 25. Note that our sampling of 10,000 data points
from N (100, 25) would not give a mean and standard deviation of exactly 100 and 25,
respectively.

>> y=randn(10000,1)*25 + 100;

>> [mean(y), std(y)]
ans =
99.7555 25.0802
>> histogram(y)
28 Programming Mathematics Using MATLAB®

Figure 2.2 Visualizing randn modification. (A) histogram(y), (B) histogram(y,100).

In Fig. 2.2 the command histogram is used to display these normally distributed ran-
dom numbers.

2.6. Exercises
NOTE: additional exercises that use vectors and/or matrices appear in Appendix C.
1. Do not use MATLAB help or documentation to answer these questions; only
experiment with the commands.
(a) Using the rand command, create a random matrix A1 with 4 rows and 6
columns.
(b) Create a matrix B1 that is the transpose of A1 (there are multiple ways to do
this with one command; choose one).
(c) Using the randi command, create a random ROW VECTOR named u1
with eight elements of integers from 1 to 100.
(d) Create a random COLUMN VECTOR named v1 with nine elements of
integers from −10 to 10.
(e) By using the MATLAB functions length, size, and numel on these ma-
trices and vectors, figure out the differences between these functions, and
come to your own conclusions as to when these functions would be useful.
Write your conclusions nicely on a separate sheet of paper. Write complete
sentences so each part that you are answering is clear. You will be graded
on following directions and not on the accuracy of the descriptions of the
functions, but whether you experimented sufficiently and your conclusions
reflect your experimentation.
 Vectors and Matrices (Arrays) 29

2. Do not use MATLAB help or documentation to answer these questions; only

experiment with the commands.
(a) Create a vector v2 with the numbers 2.1, 2.5, 2.6, −2.1, −2.5, and −2.6.
(b) By using the MATLAB functions ceil, floor, fix, and round on v2 ex-
plain in your own words on paper what the functions do, making clear what
the differences are between them. Just as above, you will be graded not on
the accuracy of the descriptions of the functions but whether your experi-
mented sufficiently and your conclusions reflect your experimentation.
3. Use the colon operator to generate the following vectors (do not suppress the
output):
(a) u3 = -5 -4 -3 -2 -1 0 1 2 3 4 5
(b) v3 = 0 0.2500 0.5000 0.7500 1.0000
(c) w3 = 12 9 6 3 0
4. Create these vectors and matrices as efficiently as possible.
(a) Create a vector x that has nine equally spaced values between −2 and 2.
(b) Create a matrix A4 that has the values of x as the first column, the values of
3x as the second column, and the values of 4x − 7 as the third column.
(c) Create a matrix B4 whose first row has the values π x and the second row
has values that are cosine of the corresponding values of π x.
5. Create these vectors and matrices as efficiently as possible.
(a) Create the 4 × 4 identity matrix called M.
(b) Create a 10 × 10 matrix T filled with 1s except a 0 in the third row, fourth
column position.
(c) Create a 8 × 9 matrix F filled with 0s except a 1 in the fifth row, seventh
column position.
(d) Create a vector called lengthu that stores the length of the vector u3 created
above.
(e) Create variables r and c that store the number of rows and columns of the
matrix F above, respectively.
6. Each row in the matrix M6 follows a different mathematical pattern. Using these
patterns, enter the following matrix into MATLAB. Thus you should be able to do
this without entering each value by hand. Using M6, do parts (a)–(d) as efficiently
as possible:
⎡ ⎤
1 2 3 4
⎢9 6⎥
⎢ 8 7 ⎥
M6 = ⎢ ⎥.
⎣1 4 9 16⎦
−1 1 −1 1
(a) Create a vector A6 consisting of the elements of the third column of M6
and a vector W6 consisting of the elements of the second row of M6.
30 Programming Mathematics Using MATLAB®

(b) Create a 4 × 3 matrix B6 consisting of all elements of the second through

fourth columns of M6.
(c) Create a 3 × 4 matrix C6 consisting of all elements in the first through third
rows of M6.
(d) Create a 2 × 3 matrix D6 consisting of all elements of the second and third
rows and last three columns of M6.
7. Use the matrix M6 above to find the following using MATLAB functions and
commands as efficiently as possible. Think what you would need to use if M6 had
1000 rows and columns.
(a) Find the minimum values in each column of M6 and store the answers in
the variable called mincols.
(b) Find the maximum values in each row of M6 and the location of these
values. State your answers on your sheet of paper nicely. For example, “Max-
imum of row 1 = ___ and is in the i,j entry of M6.”
(c) Sort each column of M6 and store the result in the matrix M6colsort and
sort each row of M6 and store the result in the matrix M6rowsort. NOTE:
there is a way to have MATLAB do this work for you without having to
look at the matrix M6 yourself!
(d) Total the values in each column of M6 and store the result in M6coladd and
total the values in each row of M6 and store the result in M6rowadd.
8. Use the matrices C and D below to find the following in MATLAB:
 
11 5 −7 − 8
C= , D= .
−9 4 −9 −4

(a) E = CD, F = DC, and G = each element of C multiplied with its corre-
sponding element of D.
(b) H = C + D, J = C 2 , and K = each element of C squared.
9. Try to do these as efficiently as possible.
(a) Using rand, generate a random real number between −2π and 2π and store
it in the variable prob9a (do not suppress output).
(b) Using randi, simulate a roll of six dice for the game of Farkle by creating a
vector of six random integers between 1 and 6 (so repeats are allowed) and
store it in the variable farkle (do not suppress output).
(c) Generate a random order for 20 presentations (thus repeats are not allowed,
unlike above) by using the randperm function if the presenters are numbered
1 through 20 (do not suppress output).
(d) Generate a matrix called PowerballTix that will simulate ten Powerball
tickets/drawings (each row will be one drawing). The first five entries of
each row are the “white balls” and the last entry of each row is the “power-
ball” for the drawing. For each row/drawing, the first five columns should
 Vectors and Matrices (Arrays) 31

be random integers from 1 to 69 with no repeats and not necessarily in

order and the last column should be a random integer from 1 to 26 [3] (do
not suppress output).
(e) A CDC study [7, p. 14] in the years 2011 to 2014 showed American females
aged 20–29 had a mean height of 64.1 inches with a standard error of 0.12
inches. Using these numbers as the mean and standard deviation of a normal
distribution, generate one random height and store it in a variable called
randomheight (do not suppress output).
10. We will use the command histogram along with the random generator functions
used above to visualize some of the data.
(a) Create a vector called tenrolls that will store ten rolls of a standard die
(suppress the output!) and then use the histogram command to display the
outcome of those ten rolls. Use six “nbins” for your histogram.
(b) Do the same as in part (a) but have the vector called manyrolls for 100,000
rolls of a standard die.
(c) The GRE guide [5, p. 18] reports that between July 1, 2015 and June 30,
2018 there were 1,695,463 test takers that took the Quantitative Reasoning
part of the GRE with a mean of 153.07 and standard deviation 9.24. Sim-
ulate this normal distribution by generating a vector called GRE that has the
same number of randomly generated scores as reported test takers (suppress
this output!). Then use the histogram command to display your vector,
using 100 “nbins” for your histogram.

11. Use the matrix M6 in #6 above to answer the following problems. Show the
commands and/or experimentation that helped you reach your conclusions.
(a) What is the difference between sum(M6), sum(M6,2), sum(M6,1), sum(
M6'), and sum(M6, 'all')? Can you come up with an equivalent but
different command for sum(M6,2)?
(b) What is the difference between the commands sort(M6), sort(M6,1), and
sort(M6,2)?
(c) Can you use the same idea for min and max? In other words, do you get
similar results for the commands min(M6), min(M6,2), min(M6,'all'),
etc. as you do with the sum or sort command? If not, what is going on with
these commands and how can you get the same results as in sum(M6,2)?
12. (a) Generate a vector named theta1 of values from 0 to 2π with increments of
π/6 using linspace. (Hint: how many values should you have?)
(b) Generate a vector named theta2 of values from 0 to 2π with increments of
π/6 WITHOUT using linspace.
(c) Calculate the sine of these angles using either theta1 or theta2.
(d) Generate a table of Cartesian coordinates corresponding to these angles on
the unit circle by building a matrix called Cart. The first column should be
32 Programming Mathematics Using MATLAB®

either theta1 or theta2, the second column should be the x-coordinate,

and the third column should be the y-coordinate of the points on the unit
circle.
(e) Add a fourth column that contains the angle measurements in degrees and
call this new matrix Cart2.
(f) Add rows (in the appropriate spots!) corresponding to π/4, 3π/4, 5π/4 and
7π/4 and call this new matrix Cart3 (this may take several steps).
CHAPTER 3

Plotting in MATLAB®
3.1. Basic 2D plots
There are many ways to plot data, curves, and/or functions. We will not cover all
plotting commands, but hopefully after this chapter you will be able to use other plotting
commands (such as pie, etc.) and be able to customize your plots easier.
We will start with plotting functions. If you are familiar with the command ezplot,
this command is outdated and should be replaced by fplot, so we will not discuss
ezplot. Both ezplot and fplot are useful but do not have as much versatility as the
plot command. With fplot you can control the domain, colors, markers, etc. as we
will discuss in this chapter but MATLAB® decides which data points to use within your
domain to create the plot. For example, notice what the commands fplot(@(x)x.^3),
fplot(@(x)sin(x)), and fplot(@(x)sin(x), [-2,2]) create.
With the command plot we can control which points, and thus how many points
are used to create the plot(s). We can also plot data points. Most of the examples in this
chapter will use plot. Much of the syntax for modifying colors, creating multiple plots,
etc. are the same with fplot and plot.
In order to plot a mathematical function using plot, follow these steps.
Step 1: define the domain as a vector. This is where the linspace command can come
in handy, and when you will want to SUPPRESS THE OUTPUT.
x = linspace(-5,5);
Step 2: calculate the y-values for each of the x-values in your domain USING
COMPONENT-WISE CALCULATIONS. Note that fplot needs component-wise
calculations as well.
y = x.^3;
Step 3: Plot the inputs and outputs using the plot command. What the plot(x,y)
command does is plot each point (xk , yk ) in the vectors x and y (or whatever you have
called these variables). By default, it connects these points in the order of the vectors
with a blue line. If there are enough points in the vectors, you get a smooth curve as
expected in this case.
We will have to add labels and titles if we want them (shown later). One can also
just as easily define the vectors within the plot command. For example, the following
code would produce the same figure as in Fig. 3.1:

x=linspace(-5,5);
plot(x,x.^3)

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00008-9 All rights reserved. 33
34 Programming Mathematics Using MATLAB®

Figure 3.1 plot(x,y).

3.2. Bad domain examples

Remember, MATLAB will plot points as they are defined in the vectors. Thus if the
vectors do not have enough points, or they are in a certain order, the pictures may not
come out as planned.
What we could do is make sure x has more elements, either by using linspace
(oftentimes preferable), or by making the increment used in colon operator be small. Be
careful: too many elements in the vector (from having the n in linspace(a,b,n) too
large, or having too small of an increment if using colon operator) can cause MATLAB
to slow down, especially when defining domains used in 3D plots.
The following have different domain definitions, followed by the commands y=sin
(x); plot(x,y). Notice the difference in the domain definitions, and thus the figures
shown in Fig. 3.2.
Another issue is having the points in the wrong order. If we use the commands
below to create a square, we see in Fig. 3.3(A) that the vertices are listed in the wrong
order.

x=[0 0 1 1]; y=[0 1 0 1];

plot(x,y)
axis([-0.2 1.2 -0.2 1.2])

Notice also the use of the axis command so that the plot can be seen better. We
must make sure the vertices are in the proper order so that when they are connected,
we get the desired effect.
 Plotting in MATLAB® 35

Figure 3.2 Domain examples. (A) Bad domain: x = -5:5, (B) Better domain: x = -5:0.01:5.

Figure 3.3 Reordering points. (A) x=[0 0 1 1]; y=[0 1 0 1], (B) x=[0 0 1 1];
y=[0 1 1 0].

We can make this even better by remembering to include the first point as the
last point so all vertices are connected. Also, for Fig. 3.4(B), we added the command
axis equal after the axis([-0.2 1.2 -0.2 1.2]) command so that what is shown
does indeed look like a square as desired.

3.3. Axis settings

There are many settings and properties of the axis figure. Not all will be covered here.
The reader is directed to the MATLAB documentation for a more thorough treatment.
36 Programming Mathematics Using MATLAB®

Figure 3.4 Better squares. (A) x=[0 0 1 1 0]; y=[0 1 0 1 0], (B) Using axis equal.

As seen in Figs. 3.3 and 3.4, you can control the length of the axes shown and/or
aspect ratios. You can even set axis off. You can change the limits of the x-values
or y-values shown in the figure by using xlim and/or ylim commands. Fig. 3.12 uses
the xlim command. You can also use the axis command. The axis command is in
the form [xmin xmax ymin ymax], with the addition of zmin zmax elements for a
three-dimensional figure (discussed in Chapter 4). Thus axis([-2 2 -5 5]) will set
the axes to be [−2, 2] × [−5, 5]. This command is equivalent to xlim([-2,2]), ylim
([-5,5]).
To change the aspect ratio, especially when plotting circles, squares, etc. you may
want to use axis equal or axis square. Another useful one is axis tight. To reset
to the default, use axis normal. In Fig. 3.5 are different examples showing the differ-
ences between some of the axis settings to graph y = sin(x) for x ∈ [−2π, 2π ]. A word
of advice: if you are expecting lines and/or planes to look orthogonal (perpendicular)
and they do not, it could be the aspect ratio of the figure. Learn from me, first try
axis equal before spending too much time double checking and triple checking your
math! This is demonstrated in Section 4.4.1.
To see the importance of setting the aspect ratio using axis, consider the following
MATLAB code that generates Fig. 3.6(A):

t=linspace(0,2*pi);
x=cos(t); y=sin(t);
plot(x,y)

If you look closely, this is a parameterization of the unit circle, and yet the figure
looks more like an ellipse. If we use the command axis([-1.5 1.5 -1.5 1.5]) to
 Plotting in MATLAB® 37

Figure 3.5 Different axis settings with sine. (A) Default axis, (B) axis tight, (C) axis equal
and (D) axis square.

establish that [xmin xmax ymin ymax] should have values [-1.5 1.5 -1.5 1.5], it
still looks like an ellipse (Fig. 3.6(B)).
Setting the axis limits in Fig. 3.6(B) still does not achieve the desired picture. Other
axis properties then need to be set, such as axis square and axis equal (Fig. 3.7).
A word of caution: when you are combining both setting the limits and the aspect
ratios, the order of these commands is important! You may need to experiment
with different orders to see the difference and which order gives you the desired effect.
For example, in the y = sin(x) plot below, if you have the axis equal before or after
38 Programming Mathematics Using MATLAB®

Figure 3.6 Unit circle. (A) Default axis and (B) axis([-1.5 1.5 -1.5 1.5]).

Figure 3.7 Axis Examples. (A) The default axis, (B) axis square and (C) axis equal.

the ylim([-2,2]) commands, you get different results (see Fig. 3.8). Also, if you add
plots after setting the axes, the axes are already “set” so the additional plot will not
change the limits or aspect ratio.

x=linspace(-10,10);
y=sin(x);
plot(x,y)
axis equal
axis([-5, 5, -2,2])
 Plotting in MATLAB® 39

Figure 3.8 Effects of order of commands. (A) axis equal first, (B) axis equal second, (C) Only
axis equal and (D) Only axis([-5, 5, -2,2]).

Experiment with plotting the circle, setting the axes from −1.5 to 1.5 and axis
equal. Have both commands, in different orders, and just one of those commands to
see the differences in the plots created. Experimenting you may notice the following:

axis([-1.5 1.5 -1.5 1.5]), axis square

%% is equivalent to
axis square, axis([-1.5 1.5 -1.5 1.5])
%% is equivalent to
axis equal, axis([-1.5 1.5 -1.5 1.5])
%% but is different from
axis([-1.5 1.5 -1.5 1.5]), axis equal
40 Programming Mathematics Using MATLAB®

Figure 3.9 Multiple plots with one plot command.

3.4. Multiple plots

There are several ways you can put multiple plots within one figure. Keep in mind that
whenever MATLAB sees a new plot command, whether it be plot, scatter, or mesh,
etc., it replaces any previous plot in the active figure window with the new command
unless we tell it otherwise.
The most basic way to have multiple plots is to put all of them within one command,
as in Fig. 3.9. The drawback is that this can be more cumbersome when modifying how
the curves/points look.

x = linspace(-2*pi,2*pi);
y = sin(x);
y2 = 3*sin(x);
y3 = sin(3*x);
plot(x,y,x,y2,x,y3)
xlabel('x'),ylabel('y')
title('Example 1 of Multiple Plots')

Notice that we also put in axes labels and a title in the above plot.
You can easily create a legend, making sure the text within the legends is in the same
order as the order within the plot command. By comparing Figs. 3.9 and 3.10, notice
the effect of the axis tight command. In Fig. 3.10(B), we have changed the location
of the legend by using the command
 Plotting in MATLAB® 41

Figure 3.10 Using legend and axis tight. (A) Default legend location, (B) Changing location
of legend.

legend('y=sin(x)', 'y=3sin(x)', 'y=sin(3x)', 'Location','SouthWest').

x = linspace(-2*pi,2*pi);
y = sin(x);
y2 = 3*sin(x);
y3 = sin(3*x);
plot(x,y,x,y2,x,y3)
xlabel('x'),ylabel('y')
title('Example of Multiple Plots')
legend('y=sin(x)', 'y=3sin(x)', 'y=sin(3x)')
axis tight

Another way to have multiple plots is to use the hold on and hold off commands.
The hold on command tells MATLAB to hold the active figure window open to add to
it with any subsequent plotting commands until the hold off command appears. Thus
you can mix various plot commands in one figure window (when appropriate). For
example, you can use the plot command and the scatter command on one figure, as
in Fig. 3.11(A).

x = randi(10,1,50); % 1x50 vector of random integers from 1-10

y = randi(10,1,50); % create another similar vector
scatter(x,y) % scatter plot of the points (x,y)
hold on
plot(x,x) % plot the line y=x
hold off
42 Programming Mathematics Using MATLAB®

Figure 3.11 Using hold on and hold off command. (A) Using scatter, (B) Using plot.

Note that you can just as easily use the command plot instead of scatter, although
for the plot command you must specify to just plot the points with a marker that you
designate, rather than connect the points (see Fig. 3.11(B)).

plot(x,y, 'o') % plot the points (x,y) using marker 'o'

hold on
plot(x,x) % plot the line y=x
hold off

When using hold on and hold off for multiple plots, the lines/markers will cycle
through the default colors (new from MATLAB version R2014b on). The grid on
command can also be useful.

x = linspace(-2*pi,2*pi);
y= sin(x);
y2=3*sin(x);
y3 = sin(3*x);
plot(x,y)
hold on
plot(x,y2)
plot(x,y3)
xlabel('x'),ylabel('y')
title('Example of Multiple Plots')
xlim([-4,4])
grid on
legend('y = sin x', 'y = 3sin x', 'y = sin(3x)', 'Location', 'NorthWestOutside')
hold off %% DON'T FORGET TO HOLD OFF!!
 Plotting in MATLAB® 43

Figure 3.12 Using hold on and hold off command.

3.5. Color, line, and marker modifications

You can modify the solid line to dotted, dashed, etc., create markers at certain point(s),
and/or adjust the size of the lines and/or markers as in Fig. 3.13.

plot(x,y,'g','LineWidth',2) % line green, thicker

hold on
plot(x,y2,'r--', 'LineWidth',2) % red, dashed, thick line
plot(x,y3,'b-') % blue solid line (default)
plot(x,cos(x), 'k:o', 'LineWidth', 2) % black, dotted, thick
x2 = -2*pi:pi/4:2*pi; % creating points
y4 = sin(3*x2);
plot(x2,y4,'b*') % blue * marker (no line)
plot(0,0,'kx', 'LineWidth',2,'MarkerSize',10) % big black x
xlabel('x'),ylabel('y')
title('Example of Multiple Plots')
legend('y = sin x','y = 3sin x','y = sin(3x)','y = cos(x)')
axis([-pi, pi, -5,5])
hold off %% DON'T FORGET TO HOLD OFF!!

Note that the default first color is a different blue ([0, 0.4470, 0.7410])
from 'b'= [0, 0, 1], which is new from MATLAB version R2014b on. The
default line style is solid, the default LineWidth is 0.5 and the default Mark-
erSize is 6. Table 3.1 lists the possible basic colors, line styles, and markers. One, two,
or three of these specifications can be defined in any order.
44 Programming Mathematics Using MATLAB®

Figure 3.13 Adding markers and modifying lines/markers.

Table 3.1 List of basic colors, line styles, and markers.

Colors 'b' blue Markers '+' plus symbol
'c' cyan 'o' open circle
'g' green '∗' asterisk
'k' black 'x' x
'm' magenta 's' square
'r' red 'd' diamond
'w' white 'p' pentagram
'y' yellow 'h' filled hexagram
Line Styles '-' solid (default) '<' left-pointing triangle
'--' dashed '>' right-pointing triangle
':' dotted '^' upward triangle
'-.' dash-dot 'v' downward triangle

Note that you can use the “short name” shown above, the long name, or the RGB
Value of these basic eight colors as shown in Table 3.2. Also the dashes, dash-dots, etc.
may not appear nicely if you have too many points in the vectors plotted.
Consider Fig. 3.11. Because the default colors are cycled through (which is new
starting in MATLAB version R2014b), the line and data points are different colors but
we may want them to be the same color. One way is to specify the colors.

x = randi(10,1,50);
y = randi(10,1,50);
 Plotting in MATLAB® 45

scatter(x,y, 'b')
hold on
plot(x,x, 'b')
hold off

Another way is to reset to the first default color as is done in the following code:

scatter(x,y)
hold on
ax = gca;
ax.ColorOrderIndex = 1; % resets back to first color for next plot
plot(x,x)
hold off

The lines command is very useful to capture the colors of the current color map.
Since there are 7 different colors, lines(7) will return a matrix where each row is the
RGB code for the colors. Thus you can store those and assign them. The code below
would return a similar plot as in the code above.

c=lines(7); % stores 7 default colors

c1=c(1,:); % assigns first color
scatter(x,y)
hold on
plot(x,x, 'Color', c1)
hold off

Another way is to specify different colors other than the basic eight colors dis-
cussed above is to specify colors by their RGB triple (for some examples, see http://
www.rapidtables.com/web/color/RGB_Color.html). While many RGB colors are
triples with numbers from 0 to 255 where (000) is black and (255, 255, 255) is white,
MATLAB expects each number in the triple to be a number between 0 and 1. Thus
you can take the RGB triple (255, 69, 0) for the color “orange red” and define it in
MATLAB as OrangeRed = 1/255∗[255,69,0];. Likewise, you can define the color

Table 3.2 Basic eight colors.

Short name Long name RGB value
'b' blue [0 0 1]
'c' cyan [0 1 1]
'g' green [0 1 0]
'k' black [0 0 0]
'm' magenta [1 0 1]
'r' red [1 0 0]
'w' white [1 1 1]
'y' yellow [1 1 0]
46 Programming Mathematics Using MATLAB®

Figure 3.14 Using other colors.

LoyolaGreen = 1/255∗[0, 104, 87]; then one can use the defined color triple in
your plot commands (see Fig. 3.14).

OrangeRed = 1/255*[255,69,0]; LoyolaGreen=1/255*[0, 104, 87];

x=linspace(0,10,50);
plot(x,cos(x),'-', 'color',OrangeRed,'LineWidth',2)
hold on
plot(x,cos(x),'s', 'MarkerFaceColor',LoyolaGreen,...
'color',LoyolaGreen, 'LineWidth',2,'MarkerSize',10)
hold off

3.5.1 Clf/close all

The clf command stands for “clear figure”. This will clear most figure settings such as
axis, etc. that may have been applied to previous figure. If more than one figure window
is open, it applies to the current or active figure window. It is also needed to clear
the subplots. Note that some figure settings are not reset with the clf command. The
command close all will close all figure windows. This will reset all figure commands,
but can also slow things down so sometimes clf is preferred.

3.5.2 Subplots
The command subplot(m,n,k) is used to create a matrix of subplots within the same
figure window. The number m is how many rows, n is how many columns, and k is which
 Plotting in MATLAB® 47

Figure 3.15 Subplot example.

subplot the following commands apply to. NOTE: you do not need to use commas to
separate the m, n, and k. It is easiest to explain by viewing an example (see Fig. 3.15).

subplot(2,2,1)
x = linspace(-10,10);
y = cos(x);
plot(x,y)
title('y = cos(x)')
ylim([-2,2])
subplot(2,2,2)
y = 1/2*cos(x);
plot(x,y)
title('y = 1/2cos(x)')
ylim([-2,2])
subplot(2,2,3)
y = 2*cos(x);
plot(x,y)
title('y = 2cos(x)')
axis([-5 5 -2, 2])
subplot(2,2,4)
y = cos(x + pi/4);
plot(x,y)
title('y = cos(x + pi/4)')
ylim([-2, 2])
48 Programming Mathematics Using MATLAB®

Figure 3.16 Subplot stretching across elements.

Notice that the numbering goes across the rows, first, then down. There is no way
to make a “global title” without using the text command, and there it is not as nice
because of having to specify the exact placement of the text.
One can have a subplot stretch across elements (see Fig. 3.16):

clf
subplot(141)
x=linspace(-5,5);
y=sin(x);
plot(x,y)
subplot(142)
y = sin(2*x);
plot(x,y)
subplot(1,4,3:4)
y=sin(1/2*x);
plot(x,y)

Issue with subplots: once you have used subplot commands, any plotting com-
mands will apply to the subplot currently set. Thus if I now have enter additional
plotting commands, it will place it in the last subplot, in this case it will place it in the
3-4 spot from the above. Here is an example. If this code immediately follows the above
code, the result is Fig. 3.17. Thus one should use the command clf or close all
between subplots and subsequent plots to “reset” things.

plot(x,-abs(x))
 Plotting in MATLAB® 49

Figure 3.17 Not clearing figure after subplot.

One can use loops (discussed in Chapter 6) with subplots.

3.6. Other 2D plots

One nice thing about understanding the plot command is you can easily plot curves
in which y is not necessarily a function of x. For example, we can easily plot the curve
x = y2 + 2y − 3, as shown in Fig. 3.18.

y = linspace(-5,5);
x=y.^2+2*y-3;
plot(x,y)
xlabel('x'),ylabel('y')

3.6.1 Parametric curves

You can also easily graph parametric equations such as

x = 3 cos t,
y = 2 sin t,

for t ∈ [0, 2π ].

t = linspace(0,2*pi);
x = 3*cos(t);
50 Programming Mathematics Using MATLAB®

y = 2*sin(t);
plot(x,y)
xlabel('x'),ylabel('y')
title('Example of Parametric Equations')
axis equal
axis([-5 5 -5 5])
grid on

3.6.2 Polar curves

We can plot polar equations as parametric equations using the conversion equations.
Notice in Fig. 3.19(A) that the curve appears somewhat jagged. This is because we are
connecting only 100 points for the curve. We can create a smoother curve, by increasing
the number of points in our linspace command. Fig. 3.19(B) was created by defining
t = linspace(0,2∗pi,300);.

t = linspace(0,2*pi);
r = 1 + sin(10*t);
x = r.*cos(t);
y = r.*sin(t);
plot(x,y)
xlabel('x'),ylabel('y')

Instead of using the conversion equations, one can have MATLAB do the conversion
using the command pol2cart. You can have a polar axis with the command polarplot

Figure 3.18 Other plots.

 Plotting in MATLAB® 51

Figure 3.19 Polar curves. (A) Default linspace, (B) 300 elements.

Figure 3.20 Polar curves. (A) With pol2cart (A) and (B) with polarplot.

(see Fig. 3.20). This command wants values for θ and r in polar coordinates. Note that
the command polar is no longer advised to use.

%% using pol2cart
t = linspace(0,2*pi,300);
r = 1 + sin(10*t);
[x,y]=pol2cart(t,r);
plot(x,y)
52 Programming Mathematics Using MATLAB®

xlabel('x'),ylabel('y')
axis equal
%% using polarplot
t = linspace(0,2*pi,300);
r = 1 + sin(10*t);
polarplot(t,r)

3.7. Exercises
1. The main span of the Golden Gate Bridge [9] can be roughly modeled with a
catenary equation
 x 
y = 4200 cosh − 3954
4346
for |x| ≤ 2100. You can also model it with a quadratic over the same domain:

y = 0.00011338x2 + 246.

(a) Plot both of them on the same figure, making the catenary in red and the
quadratic in black (default width). CAREFUL! How many elements should
be in your domain?
(b) Plot the catenary curve, making the line thicker than default and setting
'color' [8,21] to be the RGB vector for an approximation to Golden
Gate’s International Orange:
1
(155, 25, 0).
255
Also use the command “axis equal”.
2. The Gateway Arch in St. Louis was designed by Eero Saarinen and the central
curve is modeled by a catenary

y = 693.8597 − 68.7672 cosh(0.0100333x),

with |x| ≤ 299.2239 (x and y are both distances in feet) [20].

(a) Plot the curve, making the line black.
(b) Plot the curve again, making the line thicker than the default and setting
'color' to be the RGB color for silver:

1
(155, 25, 0).
255
(c) What is the height of the arch at its center? State the exact value and use
MATLAB to estimate the answer if needed (work on paper).
 Plotting in MATLAB® 53

(d) At which point(s) is the height 100 m? State the exact value and use MAT-
LAB to estimate the answer if needed (work on paper).
(e) What is the slope of the arch at the points in part (d)? Use MATLAB to
estimate the answer if needed and show any work on paper.
(f) Create a second plot where the curve of the arch is shown, and the tangent
lines at these points is shown in red. Mark the points with a black “x”.
2x2 − 5x + 11
3. Consider the function f (x) = 2 . We will work to make a decent graph
x − 7x − 8
(closer to one you may see in a textbook) in steps.
(a) Notice that the function has (a) vertical asymptote(s). What are they? State
your answer as a full sentence, showing all work on paper. (Your answers
should be equations for vertical lines!)
(b) This function also has horizontal asymptote(s). What are they? State your
answer as a full sentence, showing all work on paper.
(c) Plot the function y = f (x) using fplot without any other settings specified
other than axes labels and a title.
(d) Plot the function y = f (x) for −20 ≤ x ≤ 20 using plot without any other
settings other than axes labels and a title. Notice this may not look like the
way you may see the graph for this function in a textbook.
(e) Now use plot with y = f (x) for −20 ≤ x ≤ 20 with some modifications.
Set the range of the y-axis from −20 to 20. Notice again this may not look
“textbook quality”.
(f) Now use plot with y = f (x) (same domain and range for the y as above)
by dividing the domain into separate domains for each side of the vertical
asymptotes. The graphs of each piece should all be the same color so that it
looks like the same function.
(g) Do the same as in part (3f) but also graph the vertical asymptote(s) as red,
dashed lines. Graph the horizontal asymptotes as black, dotted lines.
4. Consider the function f (x) = 13 x4 − 8x2 + 4x + 1.
(a) Using calculus, find the first and second derivatives of f (x).
(b) Plot y = f (x), y = f  (x), y = f  (x) on the same graph/figure for x ∈ [−5, 5].
Make sure you use a descriptive legend for each of the functions plotted.
(c) EXTRA CREDIT TWO POINTS: use axis off and create horizontal
and vertical lines (in black) for the x and y axes. The axes should be of
appropriate length for the graph. (There is a way to do this without knowing
the height and width ahead of time!)
5. Consider the functions f (x) = cos(3x) − 1 and g(x) = 2x sin(x).
f (x)
(a) Using calculus, find the exact value of lim on paper, showing all work.
x→0 g(x)
(b) To demonstrate the calculus, plot y = f (x)/g(x), y = f  (x)/g (x), and y =
f  (x)/g (x) near x = 0 on the same graph. You may have to adjust your axes
54 Programming Mathematics Using MATLAB®

to make it a nice looking graph and make sure you use a descriptive legend
so one can tell which functions are which.
6. Consider the functions f (x) = sec(5x) − 1 and g(x) = 7x sin(x).
f (x)
(a) Using calculus, find the exact value of lim on paper, showing all work.
x→0 g(x)
(b) To demonstrate the calculus, plot y = f (x)/g(x), y = f  (x)/g (x), and y =
f  (x)/g (x) near x = 0 on the same graph. You may have to adjust your axes
to make it a nice looking graph and make sure you use a descriptive legend
so one can tell which functions are which.
7. Create a script that displays the following. The amount A(t) of an initial investment
P in an account paying an annual interest rate r at time t is given by
 r nt
A(t) = P 1 +
n
where n is the number of times the interest is compounded in a year and t is the
number of years. If the interest is compounded continuously, the amount is given
by
A(t) = Pert .
Consider an investment of $7500 put into a trust-fund account at an annual interest
rate of 2.5% for 21 years. Show the difference in the value of the account when
the interest is compounded annually, quarterly, and continuously by plotting A(t)
from t = 15 to t = 21 for these three situations on the same figure. Use a different
line type (color and/or type of line), label the axes, create a meaningful legend and
title for the plot.
8. For a calculus problem, you want to plot the region enclosed by the curves

x = y2 − 3, x = ey + 1, y = −1, y = 1.

Create a script file to plot these four curves, making the first two blue and red,
respectively, and the last two curves in black. Set the window to be [−5, 5] ×
[−5, 5].
9. If a projectile is thrown or fired with an initial velocity of v0 meters per second at
an angle α above the horizontal and air resistance is assumed to be negligible, then
its position after t seconds is given by the parametric equations
1
x = (v0 cos α)t, y = (v0 sin α)t − gt2 ,
2
where g is the acceleration due to gravity (9.8 m/s2 ).
(a) According to Guinness World Records, the fastest lacrosse shot was recorded
at 53.6 m/s in 2015 [12]. Use this speed as initial velocity v0 , and α = 8.5◦
to answer the following:
 Plotting in MATLAB® 55

i. When will the lacrosse ball hit the ground?

ii. A lacrosse field is 100 m long. If a player is standing at one end of the
field, does the ball make it to or past the other end of the field? How far
from the player will it hit the ground?
iii. What is the maximum height reached by the ball?
iv. Calculate these showing all work and graph the position of the ball to
check and demonstrate your answers. Label the axes appropriately.
(b) Using subplot, graph the path of the ball for the original value of α and
three other values (with appropriate titles!) demonstrating how your answers
to part (a) may change.
(c) Using subplot, graph the path of the ball for the original value of v0 and
three other values (with appropriate titles!) demonstrating how your answers
to part (a) may change.
10. We will use the command histogram(y,nbins) along with the random gener-
ator functions to visualize data. Some languages only have one random number
generator similar to rand in MATLAB. Common ways to simulate a roll of the
die is to multiply these numbers by 6 and then modify them in some way to get
integers between 1 and 6. Notice that you can then adjust this method to sim-
ulate any sided die that you would like. There is an easier way to simulate dice
rolls in MATLAB using randi. We will create vectors that will store 10,000 rolls
of a standard die (suppress the output!) using different methods and then visually
compare the methods with histograms.
(a) Create vectors called diceRolls1, diceRolls2, diceRolls3, and
diceRolls4 that will store 10,000 rolls of a standard die (suppress the out-
put!).
• The vector diceRolls1 will use rand and ceil appropriately.
• The vector diceRolls2 will use rand and floor appropriately.
• The vector diceRolls3 will use rand and fix appropriately.
• The vector diceRolls4 will use randi appropriately.
(b) Using subplot, create a 2 × 2 grid of plots comparing the four methods
using the histogram command to display the outcome of those rolls. Use
six “nbins” for your histogram. Have titles in each of the four subplots to
specify which method was used.
(c) Based on your histograms, is there a method that seems better, or a method
that creates an “unfair” die, or are they about the same? You may want
to rerun the section of code multiple times within MATLAB to be more
confident of your answer.
(d) Why did we not have a method that uses rand and round? Could you
create a fair die that would use rand and round? Support your answer by
simulating 10,000 rolls and creating a histogram.
56 Programming Mathematics Using MATLAB®

Table 3.3 Orbit data.

a (AU ) e
Mercury 0.3871 0.2056
Venus 0.7233 0.0068
Earth 1.0000 0.0167
Mars 1.5273 0.0930
Neptune 30.0699 0.0090
Pluto 39.4869 0.2489

11. Plot the following polar plots on one figure using subplot. Be sure that the titles
name the curves, or give the equation if it is unnamed. Also be sure to choose
the correct domain for each to make sure for each to make sure you produce the
entire curve. Plot these using the pol2cart command.
(a) Freeth’s nephroid  r = 1 + 2 sin(θ/2);
(b) Hippopede r = 1 − 0.8 sin2 θ ;
(c) Butterfly curve [6] r = esin θ − 2 cos(4θ ) + sin5 (θ/12), θ ∈ [0, 50π ];
(d) r = sin2 (4θ ) + cos(4θ );
(e) r = 2 + 6 cos(4θ );
(f) r = 2 + 3 cos(5θ ).
12. Using Kepler’s first law of planetary motion, one can obtain a basic model of the
orbits of planets and dwarf planets around the sun can be modeled using the polar
equation [24, p. 727]
a(1 − e2 )
r=
1 + e cos(θ )
where a is the semi-major axis and e is the eccentricity of the planet. The values
of several semi-major axes (a) and eccentricities (e) for our solar system [23] are in
Table 3.3.
(a) Plot the orbits of Mercury, Venus, Earth and Mars in one figure using
polarplot on the polar plane, making sure to have a legend, title, etc.
You may have to order the commands appropriately so all orbits appear on
the figure. Plot Mercury in blue, Venus in yellow, Earth in green, and Mars
in red.
(b) Plot the orbits of Earth, Neptune, and Pluto in one figure using polarplot
on the polar plane, making sure to have a legend, title, etc. As above, you
may have to order the commands appropriately so all orbits appear on the
figure. Plot Earth in green, Neptune in cyan and Pluto in magenta.
13. This problem will show the value of working with log-scales in plots. The efficacy
of two drugs are given in Table 3.4 by measuring the percentage of binding of the
 Plotting in MATLAB® 57

Table 3.4 Drug efficacy data.

Concentration Binding of Binding of
level (M) drug A (%) drug B (%)
2 × 10−9 2 0
5 × 10−9 5 1
2 × 10−8 19 4
7 × 10−8 42 13
2 × 10 −7 68 31
7 × 10−7 87 57
2 × 10−6 95 80
9 × 10−6 99 95
3 × 10−5 100 98
5 × 10−4 100 100

drug to the necessary receptor at different concentration levels in moles per liter,
or molars (M).
(a) Plot the data points with concentration level of the drug on the horizontal
axis and binding percentage on the vertical scale. First use a linear scale
(using plot), being sure to label the axes and create a legend. Use different
markers for drug A and drug B and connect the markers with a different
line (solid, dotted, etc.).
(b) Create a similar plot of the data but instead use a semi-log scale (using
semilogx).
CHAPTER 4

Three-Dimensional Plots
4.1. Vector functions or space curves
Vector functions (3D parametric equations) are defined and plotted similarly to 2D
plots, except one must use the command fplot3 instead of fplot and plot3 instead
of plot.

fplot3(@(t) cos(7*t), @(t) sin(3*t), @(t) cos(11*t))

One can fine-tune the domain used for t and color(s) used, but for more flexibility
and control use plot3.
Step 1: Define your inputs (the t). This is where linspace comes in handy; make sure
you SUPPRESS THE OUTPUT.

t = linspace(0,10*pi,200);

Step 2: Calculate the x, y, and z for each t value USING COMPONENT-WISE

CALCULATIONS.

x = t.*cos(t);
y = 4*t+1;
z = t.*sin(t);

Step 3: Plot the defined vectors using the plot3 command to get Fig. 4.1(A).

plot3(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')

As discussed in the previous chapter, you must be careful on how you define your
domain to make sure you get the expected plot, as seen in Fig. 4.1(B).

% bad domain example

t = 0:10*pi;
x = t.*cos(t);
y = 4*t+1;
z = t.*sin(t);
plot3(x,y,z)

You can add plots, titles, label axes, etc. to your figure just as in 2D plots.

% multiple plots

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00009-0 All rights reserved. 59
60 Programming Mathematics Using MATLAB®

t = linspace(0,3*pi);
x = t.*cos(t);
y = 4*t+1;
z = t.*sin(t);
s=linspace(-5,5);
x2 = 2*pi + s;
y2 = 2*pi*s;
z2 = 8*pi + 4*s;
plot3(x,y,z,x2,y2,z2)
xlabel('x'),ylabel('y'),zlabel('z')

You can add plots using hold on and hold off, and here we are also adding a point
to the graph (see Fig. 4.2).

% multiple plots using hold on/hold off

t = linspace(0,10*pi,150);
x = t.*cos(t);
y = t.*sin(t);
z = 4*t;
s=linspace(-5,5);
x2 = 2*pi + s;
y2 = 2*pi*s;
z2 = 8*pi + 4*s;
plot3(x,y,z)
hold on
plot3(x2,y2,z2,'k')
xlabel('x'),ylabel('y'),zlabel('z')
plot3(2*pi,0,8*pi,'*k') % plotting a point
hold off

As in 2D, you can change the window using xlim, ylim and/or zlim commands (try
in the command window) or the axis command (see Fig. 4.3). Notice these commands

Figure 4.1 Vector functions. (A) Vector function example, (B) Bad domain example.
 Three-Dimensional Plots 61

Figure 4.2 Multiple 3D plots.

Figure 4.3 Multiple 3D plots.

may not get the desired result and it may be better to adjust your domains instead. The
grid on command is also used in Fig. 4.3.

t = linspace(0,10*pi);
x = t.*cos(t);
y = t.*sin(t);
z = 4*t;
s=linspace(-5,5);
x2 = 2*pi + s;
y2 = 2*pi*s;
z2 = 8*pi + 4*s;
plot3(x,y,z)
hold on
plot3(x2,y2,z2,'k')
xlabel('x'),ylabel('y'),zlabel('z')
62 Programming Mathematics Using MATLAB®

plot3(2*pi,0,8*pi,'*k')
zlim([0,60])
title('Example of Multiple Plots and Other Commands')
legend('vector curve', 'tangent line')
grid on
hold off %% DON'T FORGET TO HOLD OFF!!

4.2. Plotting surfaces

As for previous plots, you can use fmesh and/or fsurf.

fmesh(@(x,y) 0.5*cos(0.5*x).*sin(y))
%%
fsurf(@(x,y) exp(y).*cos(pi*x), [-4,4,0,2])

The commands that plot the three-dimensional surfaces establish points (x, y, z)
of the surface to be corresponding elements in the matrices X, Y , and Z and then
“connect-the-dots” as in the two-dimensional plots. One can create these matrices sep-
arately, but if they are not formed from datasets then many times the surface(s) are the
result of functions. Thus the steps to establish the domain and then using component-
wise calculations are especially important.
4y
Example 4.2.1. Graph the surface f (x, y) = for −10 ≤ x ≤ 10 and −5 ≤
x2 + y2 + 1
y ≤ 5.
Step 1: Establish the domain by creating vectors. They do not need to be the same size.
Step 2: Create the matrices X and Y based on the domain. The easiest way to do this
is by using the command meshgrid.
Step 3: Calculate the corresponding Z using component-wise calculations on the ma-
trices X and Y .
Step 4: Plot the surface. The most common commands are mesh and surf as seen in
Fig. 4.4.

u = linspace(-10,10,50); v = linspace(-5,5,50); % Step 1

[x,y] = meshgrid(u,v); % Step 2
z = (4*y)./(x.^2 + y.^2 + 1); % Step 3
mesh(x,y,z) % Step 4

Note that Step 1 and Step 2 can be combined into one step, but for clarity they are
shown separately.
 Three-Dimensional Plots 63

Figure 4.4 Plotting surfaces. (A) Using mesh(x,y,z), (B) Using surf(x,y,z).

4.2.1 The meshgrid command

The meshgrid command takes as inputs vectors to define matrices corresponding ma-
trices. Once can create two or three matrices. For the purpose of this example, we will
create two matrices. Note that the vectors for the domains below are not the same size,
and are created within the meshgrid command.

>> [x,y]=meshgrid(1:5,6:8)
>> z=x.*y
x =
1 2 3 4 5
1 2 3 4 5
1 2 3 4 5
y =
6 6 6 6 6
7 7 7 7 7
8 8 8 8 8
z =
6 12 18 24 30
7 14 21 28 35
8 16 24 32 40

Once the matrices are created, three-dimensional plot commands will use as points the
corresponding elements of these matrices, and then connect those points. Thus the first
point plotted in the above example will be the point (x, y, z) = (1, 6, 6), and the last
point will be (5, 8, 40) (see Fig. 4.5).
64 Programming Mathematics Using MATLAB®

Figure 4.5 meshgrid examples. (A) plot3(x,y,z,'∗k'), (B) mesh(x,y,z).

Figure 4.6 Domain examples. (A) Default linspace, (B) Using 40 values.

4.2.2 Domain issues

There can be a fine line between not enough and too many points for creating accu-
rate, aesthetically pleasing figures. You may notice that in the above example to create
Fig. 4.4, the command linspace was used to create the vectors but the number of
values was decreased from the default 100 to 50. Using the default can create a surface
that may not be aesthetically pleasing and/or may slow MATLAB down by performing
more calculations than necessary to create the surface (see Fig. 4.6).

[x,y]=meshgrid(linspace(-5,5));
z=5-sqrt(x.^2 + y.^2);
surf(x,y,z)
 Three-Dimensional Plots 65

Figure 4.7 Bad domain examples. (A) linspace(-9,9,25), (B) [x,y]=meshgrid(-9:.05

:9).

Notice in the code above, the meshgrid command has one input vector. This com-
mand is equivalent to the command meshgrid(linspace(-5,5),linspace(-5,5))
but is more efficient.
If one does not use enough values, the surface can appear jagged, as in Fig. 4.7(A).
Too many values as in Fig. 4.7(B) can create a bad figure, also. Note that for some sur-
faces even having 40 instead of 60 in your vector can make it appear jagged. This second
example shown in Fig. 4.7(B) was created by using the command [x,y]=meshgrid
(-9:.05:9); for z=y.∗sin(x). This is common if one is in the habit of using the
colon operator for creating domain vectors in 2D plots. When using this notation to
create 3D plots, this can create vectors and thus matrices of a much larger size than
needed, and can greatly slow down MATLAB in performing both the calculations com-
mand(s) and plotting command(s). Think about the size of the matrices created if the
domain was from −100 to 100 and one created a vector -100:.01:100!

4.2.3 Level curves

In multivariable calculus and other disciplines, level curves are useful to study functions
of two variables or three-dimensional surfaces. In MATLAB® , level curves are contour
plots. One sets up these as one does with surfaces, but uses commands such as contour,
contourf or contour3.

Plot the level curves of f (x, y) = −xye−x

2 −y2
Example 4.2.2. .

[x,y]=meshgrid(linspace(-4,4));
z = -150*x.*y.*exp(-x.^2-y.^2);
contour(x,y,z)
66 Programming Mathematics Using MATLAB®

Figure 4.8 Level Curves. (A) contour(x,y,z), (B) contourf(x,y,z) and (C) contour(x,y,
z,20).

The above code creates Fig. 4.8(A). You can create a filled contour as in Fig. 4.8(B)
using contourf.
One can adjust the number of level curves shown by having a fourth argument as
in Fig. 4.8(C); otherwise MATLAB automatically chooses the number of level curves
drawn.

To create labels, or to adjust which values the level curves are shown; see the code
below. The output of the code below is in Fig. 4.9.

[c,h]=contour(x,y,z);
clabel(c,h)
%% Forcing values of levels (with labels)
cvalues=-20:4:20;
[c,h]=contour(x,y,z,cvalues);
clabel(c,h)

You can show level curves in 3D, and even add a mesh surface or filled surface as in
Fig. 4.10.

4.2.4 Multiple plots and modifying colors

Using hold on and hold off, one can add multiple plots to 3D graphs as in the 2D
case. One can also change the colors in the mesh and surf plots by setting EdgeColor
and/or FaceColor as in Fig. 4.11. One can also combine surfaces, vector functions
and/or points within one figure with the hold on and hold off commands. Common
mistakes are to forget to use plot3 rather than plot in these instances, or to use plot3
when mesh or surf is more appropriate.

[x,y]=meshgrid(linspace(-2,2,55));
z1=y.*exp(x.^2-5);
 Three-Dimensional Plots 67

Figure 4.9 Modifying level curves. (A) Level Curve with Labels, (B) Specifying Levels.

Figure 4.10 Level curves in 3D. (A) contour3(x,y,z,20), (B) meshc(x,y,z) and (C) surfc
(x,y,z).

Figure 4.11 Multiple plots and modifying colors. (A) No modifications, (B) Specifying Colors.
68 Programming Mathematics Using MATLAB®

mesh(x,y,z1)
xlabel('x'),ylabel('y'),zlabel('z')
hold on
z2=1/2*x.*cos(y);
mesh(x,y,z2)
hold off
%% Modifying colors
LoyGreen = 1/255*[0, 104, 87];
mesh(x,y,z1,'EdgeColor',LoyGreen)
xlabel('x'),ylabel('y'),zlabel('z')
hold on
surf(x,y,z2,'FaceColor','y', 'EdgeColor','r')
hold off

The above code above creates Figs. 4.11(A) and 4.11(B).

The default “colormap” of 3D plots changed in R2014b. Previous to this edition
of MATLAB, the default colormap was “jet.” It is now “parula” (see Fig. 4.12). There
are many other pre-set colormaps, and you can create your own. See MATLAB’s doc-
umentation on colormap for a full treatment.

4.3. View command

Within the MATLAB figure window, you may have found the “Rotate 3D” capability
quite useful. If instead, you want a static picture to have a certain perspective in 3D, you
can use the view command. When you rotate the 3D figure window, you may notice
some numbers appearing in the bottom left corner of the window (see Fig. 4.13).
The azimuth is the angle measured from the negative y-axis in the positive (coun-
terclockwise) direction on the xy-plane. The elevation is the angle from the xy-plane
towards the positive z-axis. These angles (measured in degrees) determine the viewpoint
looking at the plot, as shown in Fig. 4.14.

Figure 4.12 Colormaps. (A) colormap jet, (B) colormap default and (C) colormap bone.
 Three-Dimensional Plots 69

Figure 4.13 Rotate 3D screen shot.

Figure 4.14 Explanation of view.

The default view is view(-37.5, 30), as seen in Fig. 4.15(A). By changing the
viewpoint to be similar to what is shown in Fig. 4.14 (using view(30,30)), we get
Fig. 4.15(B).
Using the view command, you can look down at the xy-plane, xz-plane, and
yz-plane (see Fig. 4.16). This will be explored in the exercises.
As mentioned in the MATLAB documentation, there are some limitations of control
with this command. To gain more control of the view, you can use the camera properties
but these are not discussed here.
70 Programming Mathematics Using MATLAB®

Figure 4.15 View examples. (A) The default view, (B) Using view(30,30).

Figure 4.16 View examples. (A) xy-plane view, (B) xz-plane view.

4.4. Axis settings, revisited

As discussed in Section 3.3, the axis settings can really change the look of the plots.
These settings can be especially important and especially tricky in 3D graphs, as the
next example demonstrates.

Example 4.4.1. Graph the surface f (x, y) = 2x2 + y2 along with its tangent plane and
normal vector at the point (1, 1, 3).
From multivariable calculus, we get that the equation for the tangent plane is z =
4x + 2y − 3 and the parametric equations for the normal vector are

x = 1 + 4t,
y = 1 + 2t,
 Three-Dimensional Plots 71

z = 3 − t.

Graphing it and then adjusting the view gives us the figures in Fig. 4.17.

% surface
[x,y]=meshgrid(linspace(-4,4,50));
z=2*x.^2+y.^2;
% plane
z2 = 4*x+2*y-3;
% vector
t=linspace(-1,1.5);
x3=1+4*t;
y3=1+2*t;
z3=3-t;
% graphing
mesh(x,y,z,'EdgeColor','black')
hold on
surf(x,y,z2,'FaceColor','blue')
plot3(x3,y3,z3,'r', 'LineWidth',2)
hold off
xlabel('x'),ylabel('y'),zlabel('z')

When first seeing these figures, you may check and recheck their calculus and won-
der what is going on as the normal vector does not appear to be perpendicular to the
tangent plane. The problem is not the calculus; it is the aspect ratios of these figures.
These can be fixed by using axis equal with some adjustments. If using zlim still does

Figure 4.17 Different view settings for Example 4.4.1. (A) default view: view(-37.5,30), (B)
view(65,25).
72 Programming Mathematics Using MATLAB®

Figure 4.18 axis equal, view(37,40), zlim([-10,10]).

not achieve the desired picture, as mentioned previously you may also want to adjust
the original domains to get the desired picture (see Fig. 4.18).

4.5. Other coordinate systems and 3D graphs

4.5.1 The sphere and cylinder commands
The sphere and cylinder commands are quite useful commands.
To plot a unit sphere, use the commands sphere and axis equal. As discussed
above, without the axis equal command it will not look like a perfect sphere. One
can also force more or less faces to create the sphere by using the command sphere(n),
where n is the number of desired faces (see Fig. 4.19). The default is 20, thus sphere is
equivalent to sphere(20).
The most flexibility comes from using the sphere command in the form of [x,y,
z]=sphere(n). This generates three (n + 1) × (n + 1) matrices x, y, and z that when
combined with the mesh(x,y,z) or surf(x,y,z) commands, will generate a sphere
with n faces. This allows you to translate, resize, and/or recolor the sphere as you would
any 3D graph (see Fig. 4.20).

[x,y,z]=sphere(30);
mesh(x,y,z) % unit sphere
hold on
surf(x+3,y+2,z-1) % sphere with center (3,2,-1)
r=0.5;
mesh(r*x-1,r*y+2,r*z+1,'FaceColor',[0.5,0.5,0.5],...
'EdgeColor', 'k') % sphere resized, moved and recolored
axis equal
xlabel('x'),ylabel('y'),zlabel('z')
 Three-Dimensional Plots 73

Figure 4.19 Spheres using axis equal. (A) default sphere, (B) sphere(50) and (C) sphere
(10).

Figure 4.20 Drawing different spheres with [x,y,z]=sphere(n).

The command cylinder works similarly to the sphere command and will generate
a unit circular cylinder: a cylinder of radius 1 and height 1 with the center of circle at
(0, 0) in the xy-plane (see Fig. 4.21(A)). The command cylinder(f) will create a
cylinder using f as the profile curve; in other words, f is the radius of the cylinder at
equally spaced heights along the cylinder (see Fig. 4.21(B)). Thus f could be one value
of a vector of values. The command cylinder is equivalent to cylinder(1). The
command cylinder(f,n) will create a cylinder using f as the profile curve but using
n equally spaced points around its circumference (see Fig. 4.21(C)). Thus cylinder is
equivalent to cylinder(1,20).
As in the case with the sphere command, using [x,y,z]=cylinder(f,n) al-
lows for more control over modifying the cylinders generated using transformations
on (x, y, z) (see Fig. 4.22).
74 Programming Mathematics Using MATLAB®

Figure 4.21 Cylinders using axis equal. (A) default cylinder, (B) cylinder(0.5) and (C)
cylinder(2,50).

[x,y,z]=cylinder(1,50);
mesh(x,y,z)
hold on
surf(x-1,y+2,2*z)
surf(2*x-1,2*y+1,0.5*z+3,'FaceColor',[0.3,0.3,0.3])
hold off
axis equal
xlabel('x'),ylabel('y'),zlabel('z')

As mentioned above, the value(s) of f can be a vector of values, or as in the case

below determined by values of a function (see Fig. 4.23).

t=linspace(0,2*pi);
f=2*sin(t) + t + 2;
[x,y,z]=cylinder(f,50);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
%%
v1=[0,10];
v2=0:10;
v3=0:0.5:10;
v4=v3.^2;
[x,y,z]=cylinder(v2,50);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')

Experiment with the above code to see the difference between using v1, v2 (shown),
v3, and v4.
If you would like to have the cylinder horizontal instead, it may be easiest to define
the matrices as usual but switch the variables in the plotting command (see Fig. 4.24).
You can then adjust the cylinder’s length, width, and center via transformations similar
to what is done in Fig. 4.22. These adjustments will be explored in the exercises.
 Three-Dimensional Plots 75

Figure 4.22 Drawing different cylinders with [x,y,z]=cylinder(f,n).

Figure 4.23 Cylinders with non-constant profile f. (A) cylinder(f,50), (B) cylinder
(0:10,50).

t=linspace(0,2*pi);
f=2*sin(t)+t + 2;
[x,y,z]=cylinder(f,50);
mesh(3*z+2,x,y) % horizontal, shifted cylinder
xlabel('x'),ylabel('y'),zlabel('z')

4.5.2 Cylindrical coordinates

To plot a function given in cylindrical coordinates, one can always use the commands
to convert to Cartesion coordinates after using the meshgrid command on θ and r or
one can use the [x,y,z]=pol2cart(theta,r,z) command. Both ways are shown in
Example 4.5.1 and in Fig. 4.25.
76 Programming Mathematics Using MATLAB®

Figure 4.24 Creating a horizontal cylinder.

Figure 4.25 Cylindrical coordinates example. (A) Using conversion equations, (B) Using pol2cart.

Example 4.5.1. Graph the surface z = r that is given in cylindrical coordinates for
r ∈ [−2, 2] and θ ∈ [0, 2π].

%% Using Conversion Equations

tdomain=linspace(0,2*pi); rdomain = linspace(-2,2);
[t,r]=meshgrid(tdomain,rdomain);
x = r.*cos(t);
y = r.*sin(t);
z = r;
mesh(x,y,z)
 Three-Dimensional Plots 77

xlabel('x'),ylabel('y'),zlabel('z')
axis equal
%% Using pol2cart
tdomain2=linspace(0,2*pi); rdomain2 = linspace(-2,2);
[t2,r2]=meshgrid(tdomain2,rdomain2);
z2=r2;
[x2,y2,z2]=pol2cart(t2,r2,z2);
mesh(x2,y2,z2)
xlabel('x'),ylabel('y'),zlabel('z')
axis equal

4.5.3 Spherical coordinates

Using spherical coordinates can be useful; unfortunately the definitions of θ and φ differ
from many calculus texts. In some calculus texts you have the triple (r , θ, φ) where θ is
the azimuth angle as with the view command and φ is the zenith angle, or the angle
measured from the positive z-axis. In physics, the definitions of θ and φ are reversed. In
MATLAB, θ is defined as the azimuth angle but φ is measured as the angle of elevation
as in the view command discussed in Section 4.3 and shown in Fig. 4.14.
As with cylindrical coordinates, one may use the conversion equations or the com-
mand sph2cart, keeping in mind the definition MATLAB has for φ (see Fig. 4.26).

%% using calculus definition of phi

% and usual conversion equations
theta=linspace(0,2*pi);
phi=linspace(0,pi/4);
[t,p]=meshgrid(theta, phi);
rho = 3+0*t; % the 0*t ensures rho is correct size
x = rho.*cos(t).*sin(p);
y = rho.*sin(t).*sin(p);
z = rho.*cos(p);
mesh(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
axis equal
%% using MATLAB's definition of phi
% and the adjusted conversions
clf
theta=linspace(0,2*pi);
phi=linspace(0,pi/4);
[t,p]=meshgrid(theta, phi);
rho = 3+0*t;
x = rho.*cos(t).*cos(p);
y = rho.*sin(t).*cos(p);
z = rho.*sin(p);
mesh(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
78 Programming Mathematics Using MATLAB®

axis equal
%% using sph2cart
clf
theta=linspace(0,2*pi);
phi=linspace(0,pi/4);
[t,p]=meshgrid(theta, phi);
rho = 3+0*t;
[x,y,z]=sph2cart(t,p,rho);
mesh(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
axis equal

Note that there are other commands such as cart2sph and cart2pol but these are
not discussed here.

4.6. Exercises
1. A helix is a three-dimensional curve (shape of a spring) that can be modeled by
the following equations:

x = a cos t,
y = a sin t,
z = bt,

where a is the radius of the helical path and b is a constant that determines the
“tightness” of the path.
(a) Plot a basic helix (a = b = 1) for five complete turns (be careful: what should
be your domain for t equal)?
The DNA molecule has the shape of a double helix, where one helix can be
modeled by the above equations. The radius of each helix of the DNA molecule
is about 10 angstroms (1 Å = 10−8 cm). Each helix rises about 34 Å during each
complete turn [26].

Figure 4.26 Spherical coordinates. (A) Calculus definition of φ , (B) MATLAB definition of φ and (C)
Using sph2cart.
 Three-Dimensional Plots 79

(b) Figure out what a and b should equal for the single DNA helix. Work on
paper (if any) should be turned in and the answers should be text on the
webpage. Have the units in angstroms and state the EXACT VALUES.
(c) Plot five complete turns of the DNA helix (what should be the domain for
t equal?).
(d) Plot the helix for 100 complete turns. BE CAREFUL! In order for this to
be shown correctly (i.e., showing all of the turns), you will have to define
your domain correctly.
(e) Estimate the full length of ONE COMPLETE TURN of the DNA helix
using the arc length formula

 d  2  2  2
dx dy dz
L= + + dt.
c dt dt dt

Figure out what c and d should be in the integral and calculate L on paper
using calculus, showing all work. Turn in your work on paper and state the
EXACT VALUE of your answer as text on the webpage (in angstroms).
Also use MATLAB to give a numerical approximation of this answer in
angstroms.
(f) In one cell in the human body, it has been said that DNA makes about
2.9 × 108 complete turns. Using your (exact value) answer in part (1e), use
MATLAB to give a numerical approximation of the FULL LENTH of the
DNA helix, in meters.
(g) It has been said that there are about 1013 cells in the human body. Use this,
along with your answer in part (1f) to estimate the length of all of the DNA
in a human body. The average distance between the Earth and Mars is 225
million km [22]. Would the stretched out DNA in a human body reach
Mars? If so, how many roundtrips would there be from the Earth to Mars
and back?
2. Graph the following 3D surfaces. For each of these, have the titles specify the
problem number and part, and if the surface has a special name include that as
well. Label the axes.
(a) z = 1 − |x + y| − |y − x|. Use fsurf.
(b) z = x3 − 3xy2 , x, y ∈ [−15, 15] × [−15, 15] (monkey saddle). Use mesh.
sin(x2 + y2 )
(c) z = , x, y ∈ [−5, 5] × [−5, 5]. Use surf and have the z-axis go
(x2 + y2 )
from −2 to 2.
(d) z = sgn(xy) sgn(1 − 100x2 + 100y2 ), x, y ∈ [−10, 10] × [−10, 10]. Use surf
and 50 points each for the domains of x and y.
(e) z = sgn(xy) sgn(1 − 100x2 + 100y2 ), x, y ∈ [−10, 10] × [−10, 10]. Use mesh
and 200 points each for the domains of x and y.
80 Programming Mathematics Using MATLAB®

(f) z = sin x + sin y, x, y ∈ [0, 6π] × [0, 20]. Use axis equal and your choice of
mesh or surf.
(g) Graph the level curves (contour lines) for z = sin(x − y), x, y ∈ [−10, 10] ×
[−10, 10].
(h) Graph the level curves (contour lines) for z = 1 − |x + y| − |y − x|, x, y ∈
[−15, 15] × [−15, 15].
3. Graph the following 3D surfaces. For each of these, have the titles specify the
problem number and part, and have the axes labeled.
(a) z = x3 y − xy3 , x, y ∈ [−5, 5] × [−5, 5] (dog saddle). Use surf.
(b) z = cos(x − y), x, y ∈ [−2π, 2π] × [−2π, 2π]. Use mesh.
(c) z = ex cos(y), x ∈ [0, 3], y ∈ [−3π, 3π]. Your choice of mesh or surf.
(d) z = cos x − cos y, x, y ∈ [−10, 10] × [−10, 10] our choice of mesh or surf.
(e) Graph the level curves (contour lines) for z = cos(x − y), x, y ∈ [−10, 10] ×
[−10, 10].
4. Consider the 3D parametric surface that is known as the Möbius strip.
 
m(u, v) = (7 + v cos(u)) cos(2u), (7 + v cos(u)) sin(2u), (v sin(u)

for (u, v) ∈ [0, 2π] × [−1, 1]. All graphs should have their domains defined appro-
priately so there are enough points, but not too many points. All axes should be
labeled appropriately.
(a) Graph the strip m using mesh. Make the z-axis from −2 to 2.
(b) Note that if we fix one of the variables u or v, the result is a vector function
or space curve. Add graphs of the “edges” of the strip by creating two space
curves; one when v = 1 and the other when v = −1. Plot the first curve
black and the second in white. How many edges does the strip have?
(c) Add graphs of space curves when v = 0 and v = 0.5, the first as blue and the
second as red.
(d) Graph just the space curves from parts (b) and (c).
(e) Consider a general Möbius strip
 
m(u, v) = (a + bv cos(u)) cos(2u), (a + bv cos(u)) sin(2u), (bv sin(u)

for (u, v) ∈ [0, 2π] × [−1, 1]. The strip above is when a = 7 and b = 1. Graph
the strip for different values of a and b. What changes about the strip when
we change those values? What happens if we change the domain for v to
[−5, 5]?
5. Graph the following 3D surfaces. For each of these, have the titles specify the
problem number and part, and have the axes labeled.
(a) z = sin(x − y), x, y ∈ [−π, π] × [−2π, 2π]. Use surf and the default number
of elements when defining your domain.
 Three-Dimensional Plots 81

(b) z = sin(x − y), x, y ∈ [−π, π ] × [−2π, 2π ]. Use surf and when defining your
domain, use fewer than the default to get a decent, yet not jagged, graph.
(c) Graph the level curves (contour lines) for z = sin(x − y), x, y ∈ [−10, 10] ×
[−10, 10].
6. Graph the function
   
2π x 2π y
f (x, y) = x cos cos , x ∈ [0, 150], y ∈ [0, 100].
50 50
Create the plots using mesh, surf and create a level curve (contour plot) of f (x, y)
and display all three of them side-by-side using subplot. Explain on paper what
the difference between mesh and surf are, which you prefer, and why. Make sure
each of the axes are labeled and each plot has an appropriate title.
7. Graph the function f and a contour plot of the function
   
2π x 3π y
f (x, y) = sin sin
60 60
in the domain x ∈ [0, 100] and y ∈ [0, 100]. Graph the plots side-by-side.
8. Use the sphere and surf commands to create a sphere of radius 4, centered at
the point (−2, 3, 5). Also, make the sphere a color of your choice.
9. Use the cylinder and mesh commands to create a cylinder of radius 5, height 7
and with axis at the vertical line with x = −2 and y = 3.
10. Make the above cylinder horizontal.
11. Create a solid of revolution such as one would see in integral calculus. First, graph
f (x) = x|sin(x)| + 4 from x = 1 to x = 8. We will create a solid of revolution by
rotating it about the x-axis. Similar to Fig. 4.24, use the cylinder command and
mesh to create this solid so that it is horizontal and the x-values go from 1 to 8.
Label the axes appropriately.
12. Consider the surface xeyz = 1 at the point (1, −1).
(a) Find the tangent plane and normal line at that point.
(b) Graph the surface, tangent plane, and normal line. For the surface and plane,
the domains should be x, y ∈ [0.5, 1.5] × [−1.5, −0.5]. The domain for the
line should be t ∈ [−1, 1]. Make sure your domains are defined appropri-
ately to be able to see the picture, and modify the aspect ratio to see the
orthogonality of the line with the surface. Have the line in black, the surface
z = f (x, y) be one color, and the plane be another. The use of the command
alpha and/or view may be useful to get a decent figure.
CHAPTER 5

Functions
5.1. The lookfor and help commands
The command lookfor is a useful command when you think there may be a command
with a certain name, or that performs a certain function.

>> lookfor transpose

ctranspose - {\S}\textquotesingle{\S} Complex conjugate
transpose.
transpose - .{\S}\textquotesingle{\S} Transpose.
tfqmr - Transpose Free Quasi-Minimal Residual Method.
finargflip - Transpose array arguments to conform to size
conventions.
tsAlignSizes - If the time vector is aligned to differing
dimensions, a 'transpose' is

>> lookfor 'dot product'

dot - Vector dot product.

Once you know the command name, you can use the help command as discussed
previously.

>> help dot

dot Vector dot product.
C = dot(A,B) returns the scalar product of the vectors A and B.
A and B must be vectors of the same length. When A and B are both
column vectors, dot(A,B) is the same as A'*B.

dot(A,B), for N-D arrays A and B, returns the scalar product

along the first non-singleton dimension of A and B. A and B must
have the same size.

dot(A,B,DIM) returns the scalar product of A and B in the

dimension DIM.

Class support for inputs A,B:

float: double, single

See also cross.

Reference page in Help browser

doc dot

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00010-7 All rights reserved. 83
84 Programming Mathematics Using MATLAB®

MATLAB® help and lookfor commands will search both the native MATLAB
functions but also functions within the current folder or directory.

5.2. File format

There are many reasons why one would want to write their own functions. MATLAB
functions are M-files (extension .m) but with a certain format.

function header
% H1 LINE
% HELP LINE(S)
BODY
end % end line not always necessary, but useful for clarity

The first line of the file (header) MUST be of the form

function outputvariable = functionname(inputvariable)
To avoid confusion, the function name should be the name of the M-file (without
the .m).
The H1 line is the line that is used with the lookfor command. The command
searches the H1 lines for the keyword(s) given, and if found it prints the H1 line. Note
that if the keyword is part of the function name, it will not show the function unless the
function name appears on the H1 line.
For example, consider the following two functions, fexample1 and fexample2.

function y = fexample1(x)
% Evaluates x^(1/3).
y = x^(1/3);
end

function y = fexample2(x)
% FEXAMPLE2(X) Evaluates x^(1/3).
% X can be a number, vector or matrix.
y = x.^(1/3);
end

If we use the lookfor command, it will not give use fexample1 since the H1
line does not contain the function name, while both would be returned when we
lookfor evaluates.

>> lookfor fexample

fexample2 - (X) Evaluates x^(1/3).

>> lookfor evaluates

fexample1 - Evaluates x^(1/3).
 Functions 85

fexample2 - (X) Evaluates x^(1/3).

evalmcw - Evaluates a list of functions in a editable text
uicontrol.

HELP LINES are SUBSEQUENT COMMENT LINES without any breaks. The
lines, in addition to the H1 lines, appear as the text when using help and should be a
more detailed description of the function and its usage including what the input(s) and
output(s) are. Using the MATLAB format, any reference to the FUNCTIONNAME
and INPUT(S), OUTPUT(S) should be capitalized in the comments for formatting
purposes though in reality they are not capitalized when used. Note that by formatting
your help lines this way, in recent versions of MATLAB the capitalized function name
appears in lowercase bold in the command window.

>> help fexample1

Evaluates x^(1/3).

>> help fexample2

`fexample2(X)` Evaluates X^(1/3).
X can be a number, vector or matrix.

To create paragraphs within help, have a line with only the comment symbol. An
actually blank line will break the help comment block and the next comment is then
not considered part of the help.

function A = fexample3(x,y)
% FEXAMPLE3(X,Y) evaluates X^Y for the input numbers or vectors X, Y.
%
% See also FEXAMPLE1, FEXAMPLE2.

% LAO
A = x.^y;
end

>> help fexample3

fexample3(X,Y) evaluates X^Y
for the inputs (numbers or matrices) X and Y.

See also fexample1, fexample2.

What follows is the main block of the function that includes commands and com-
ments. The function file can be with our without an end. The end is necessary for
nested functions, or multiple functions in one file.
86 Programming Mathematics Using MATLAB®

5.3. Function examples

This section discusses examples of functions; the best way to learn how to write func-
tions is to use these functions as templates to do the exercises at the end of this chapter.

5.3.1 Basic function examples

The following examples were shown in Section 5.2 but are repeated here:

function y = fexample1(x)
% Evaluates x^(1/3).

y = x^(1/3);
end

The writing of fexample1 could be improved. The HELP lines could include the
function name, and it could be written so x could be a vector or matrix (if so desired).
The improvement is seen in fexample2 below.

function y = fexample2(x)
% FEXAMPLE2(X) Evaluates X^(1/3).
% X can be a number, vector or matrix.

y = x.^(1/3);
end

5.3.2 More function examples – multiple inputs

Multiple inputs of a function are listed with a comma separating the variable names. The
function fexample3 below shows how to create fancier HELP lines than fexample1
and fexample2. Notice also that the function requires TWO inputs. When there are
multiple inputs and/or outputs, it is especially important to make the HELP lines useful
and clear for the order of the variables.

function A = fexample3(x,y)
% FEXAMPLE3(X,Y) evaluates X^Y
% for the inputs (numbers or matrices) X and Y.
%
% See also FEXAMPLE1, FEXAMPLE2.

% LAO
A = x.^y;
end

The ability of having optional inputs is discussed in Section 6.8.

 Functions 87

5.3.3 Multiple outputs

Multiple outputs can be handled in several ways, and one way may be better than the
other depending on your usage of that function. See quadratic1.m and quadratic2.m
for some examples. In one, you need to know there are multiple outputs and store them
to see both of them. In the other, the multiple outputs are such that both are output.

function [r1,r2] = quadratic1(a, b, c)

% QUADRATIC1 calculates the quadratic formula on coefficients A, B, and C.
% [R1, R2] = QUADRATIC1(A, B, C) to compute
% the two roots, R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0

% LAO, demonstrating multiple outputs

d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

function [R] = quadratic2(a, b, c)

% QUADRATIC2 calculates the quadratic formula on coefficients A, B, and C.
% [R] = QUADRATIC2(A, B, C) to compute the two roots,
% R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0.
% Returns a column vector R = [R1; R2]

% LAO, demonstrating multiple outputs

d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

R=[r1; r2];

Note that the above functions do not have end. As mentioned above, this is not nec-
essary if there are not nested or multiple functions within the same file. One may
argue that it is good practice to still include them. Also note that in quadratic2, the
one output variable name is in brackets. This is not necessary; the functionality would
be the same if the output variable name appeared without the brackets (as it does in
fexample1.
The real difference in these functions is how the output variable(s) are defined and
thus how we must use the functions correctly.

>> quadratic1(1,3,2)
ans =
-2
88 Programming Mathematics Using MATLAB®

>> quadratic2(1,3,2)
ans =
-2
-1

Note for quadratic1, only the first output variable is displayed and stored in ans,
while for quadratic2, both solutions are displayed and stored in ans as a (column)
vector. We could also store the output as a row vector as in quadratic2b.

function [R] = quadratic2b(a, b, c)

% QUADRATIC2B calculates the quadratic formula on coefficients A, B, and C.
% [R] = QUADRATIC2B(A, B, C) to compute the two roots,
% R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0.
% Returns a row vector R = [R1, R2]

% LAO, demonstrating multiple outputs

d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

R=[r1, r2];

>> quadratic2b(1,3,2)
ans =
-2 -1

The issue with the way these functions are written is that the user needs to know
how the outputs are going to be displayed in order to use them correctly for anything
useful. See the example below.

>> [x1,x2]=quadratic1(1,3,2)
x1 =
-2
x2 =
-1
>> [x1,x2]=quadratic2(1,3,2)
Error using quadratic2
Too many output arguments.
>> X=quadratic1(1,3,2)
X =
-2
>> [X]=quadratic1(1,3,2)
X =
-2
>> X=quadratic2(1,3,2)
 Functions 89

X =
-2
-1

Examples of how to improve these functions are shown in the following chapter
once conditional statements are discussed.

5.3.4 Bad examples

There are at least two things wrong with the following function. Can you find the
errors?

function y = badfunction1(x)
% FCNEX1 Evaluates the cube root of a number X.

% LAO

x^(1/3)

Below we show a few commands with badfunction1.

>> help badfunction1

FCNEX1 Evaluates the cube root of a number X.

>> badfunction1(8)
ans =
2

Can you find anything wrong with badfunction2?

function Y = badfunction2(x)
% BADFUNCTION2 Evaluates the cube root of a number X.
% LAO

Y = x^(1/3)
end

Again, you may be able to figure out the errors by seeing the function in use.

>> help badfunction2

badfunction2 Evaluates the cube root of a number X.
LAO

>> badfunction2(8)
Y =
2
ans =
2
90 Programming Mathematics Using MATLAB®

5.4. Exercises
√
x5 2x + 3
1. Write a function named ch5prob1 for the function f (x) = 2 . The input of
(x + 1)3
the function should be x and the output should be the calculated value f (x). Write
the function so that x can be a number, vector, or matrix just as the sin function
works.
2. Consider the function
 
1 −1· x−μ 2
f (x) = √ e 2 σ
.
σ 2π
(a) Create a function called npdf that takes as input x, μ, and σ . It then calculates
y = f (x), making sure that the calculations can be done if x is a number,
vector or matrix, just as the sin function works (σ and μ are numbers and it
is assumed the user will do this correctly; no error checking is done on this).
(b) For μ = 0 and σ = 1, graph y = f (x) for x ∈ [−5, 5].
(c) Graph the following two plots on the same figure. The first should be of
y = f (x) for x ∈ [130, 170], μ = 150.24, and σ = 8.44. The second should be
for y = f (x) for x ∈ [130, 170], μ = 153.07, and σ = 9.24. Have a legend for
this, with the first plot having the title “Verbal Reasoning” and the second
plot having the title “Quantitative Reasoning”. The overall title should be
“GRE Test Scores July 1, 2015 to June 30, 2018” [5, p. 18].
3. Consider the function
   − ν+2 1
 ν+2 1 t2
f (t) = √   1+
νπ  ν2 ν

where ν > 0 and (x) is the Gamma function (gamma(x) in MATLAB)

 ∞
(x) = tx−1 e−t dt.
0

(a) Create a function called tpdf that takes as input t and ν and computes y =
f (t) making sure that the calculations can be done if t is a number, vector or
matrix, just as the sin function works (no error checking is done on ν ).
(b) Perform the command help tpdf.
(c) Use your function to graph y = f (t) for t ∈ [−4, 4] and ν = 1.
(d) Use your function to graph the following three plots on the same figure.
They should all be for t ∈ [−4, 4]. The first plot should be with ν = 1, the
second with ν = 2, and the third with ν = 5. Have all be different colors and
have a legend for the plots. The Greek letter ν can be written in text as “\nu”
in MATLAB.
 Functions 91

4. In Section 5.3.3 starting on page 87 there are four m-files: quadratic1.m,

quadratic2.m, quadratic2b.m and quadratic3.m. All do basically the same
thing but in slightly different ways. Save these functions to run them for the fol-
lowing problems.
(a) Run quadratic1 using the quadratic equations x2 + 2x + 1 and x2 − 2x +
5 by typing in just the function name with the appropriate inputs on the
command line. What are your outputs, respectively?
(b) Now run quadratic1 on the same equations, but this time by typing in:
A = quadratic1... and also [r1, r2] = quadratic1... How is the output
different?
(c) Now run do parts (a) and (b) but on quadratic2 and quadratic2b. How
does these differ from usingquadratic1? Look at the files to see exactly how
these differences occur. (Do not describe the differences that you see in the
program code, but the differences in the output.)
(d) Do parts (a) and (b) on quadratic3. Describe how this is different than
quadratic1, quadratic2, and quadratic2b. (Again, do not describe the
differences in the code, but the differences in the output.)
5. Using quadratic3.m, what can you say about the “correctness” of your output for
the following equations? (Figure out with paper and pencil what the answers should
be, and compare with the answers given by MATLAB.)
x 15
(a) x2 + − = 0,
2 2
(b) x + 9.15x + 15.05 = 0,
2

(c) x2 − 100,000,000x + 1 = 0.
6. Recall that polar coordinates (r , θ ) can be converted to Cartesian coordinates (x, y)
with the equations
x = r cos θ,
y = r sin θ.

Consider the family of curves given by r = 1 + c sin(nθ ), where c is any real number
and n is a positive integer. Write a function file called pcurves that takes as input
n, c and θ where θ can be either a number or vector, n is a positive integer and c
is any real number (no need to have error checks). The outputs should be the x
and y values. Then, using your function, graph the curves (in the Cartesian (x, y)
coordinate system) using subplot and varying n and c to answer the following
questions. Be sure to label your graphs appropriately.
(a) How do the graphs change as n increases?
(b) How do the graphs change as c changes?
7. Elementary row functions For the following functions, make sure the help lines
make it clear not only what the function does, but what the inputs are and in what
order they should be.
92 Programming Mathematics Using MATLAB®

(a) Create a function called swap that performs the first elementary row function
on a matrix. The inputs should be A,i,j and the function should return a
matrix B that is the same as A except the ith and jth rows are swapped.
(b) Create a function called multrow that performs the second elementary row
function on a matrix. The inputs should be A,r,i and the function should
return a matrix B that is the same as A except the ith row of A is multiplied
by r.
(c) Create a function called addrow that performs the third elementary row
function on a matrix. The inputs should be A,r,i,j and the function should
return a matrix B that is the same as A except r times the ith row of A is added
to the jth row.
8. You want to create multiple versions of homework and exam problems quickly.
For example, you want to create homework problems with different coefficients of
polynomials and you will use MATLAB to generate these coefficients. MATLAB
has ways to create random numbers, but we want specific constraints and we may
modify these constraints for different problems, so we will create functions to use
repeatedly. The help lines for these functions should be very clear on how to use
these functions. In Exercise 10 in Chapter 6 we will expand on these functions.
(a) Create a function randInt2 that takes as inputs 2 integers, a and b. The
function generates a random integer from a to b. No error checking will be
done on the inputs a and b; we will assume the user inputs integers a < b.
(Error checks will be created in Exercise 10 in Chapter 6.)
(b) Create a function randList that takes as input a vector of numbers (do not
do any error checking on this; assume a vector is correctly input by the user).
The function will return a random element from the vector (list).
CHAPTER 6

Control Flow
6.1. Relational and logical operators
Relational and logical operators allow one to compare two values, variables, etc. The
operators take on the form expression1 OPERATOR expression2 and evaluate to a logical
data type; either true (1) or false (0).

Relational operators
You can use relational operators on vectors to compare to other vector(s) of the same
size. The comparison is made component-wise and returns a vector of the same size
with each entry either true or false (see Table 6.1).

>> 1 < 2
ans =
logical
1
>> 1 > 2
ans =
logical
0
>> 6/3 == 2
ans =
logical
1
>> 1 == sin(pi)
ans =
logical
0
>> A=randi([0,5],1,5), B=randi([-5,5],1,5)
A =
4 1 3 4 5

Table 6.1 Relational operators.

< less than
> greater than
<= less than or equal
>= greater than or equal
== equal
~= not equal

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00011-9 All rights reserved. 93
94 Programming Mathematics Using MATLAB®

B =
5 1 -4 -4 -3
>> A <= B
ans =
1×5 logical array
1 1 0 0 0
>> A == B
ans =
1×5 logical array
0 1 0 0 0
>> A ~= B
ans =
1×5 logical array
1 0 1 1 1

Likewise, you can use relational operators to compare matrices of the same size.

>> X=randi([-5,5],2,3), Y = randi([-5,5],2,3)

X =
1 5 3
1 -2 3
Y =
-1 -5 0
1 -5 3
>> X > Y
ans =
2×3 logical array
1 1 1
0 1 0

Logical operators
Logical operators can be used to create compound statements that evaluate to either
true or false (see Table 6.2).
Eager versions will evaluate both expressions no matter what. The short-circuit
versions are only good on scalars and will only evaluate the second expression

Table 6.2 Logical operators.

&, && AND (eager, short-circuit)
|, || OR (eager, short-circuit)
~ NOT
xor exclusive OR
all TRUE if all elements of array are TRUE
any TRUE if any of elements of array are TRUE
 Control Flow 95

if needed. Pros and cons to using each version are more apparent in complicated
programs. For a discussion on the topic, search “short-circuit evaluation” on Wikipedia.

>> z=-exp(1), w = pi
z =
-2.7183
w =
3.1416
>> z > 0 && w > 0
ans =
logical
0
>> z > 0 || w > 0
ans =
logical
1
>> z > 0 & w > 0
ans =
logical
0
>> z > 0 | w > 0
ans =
logical
1

The eager versions of the logical operators produce matrices of 0s and 1s if one of
the arguments is a matrix.

>> x = [-1 0 pi]; y = [1 2/3 -sqrt(2)];

>> x > 0 & y > 0
ans =
1×3 logical array
0 0 0
>> x > 0 && y > 0
Operands to the || and && operators must be convertible to logical scalar values.
>> X=randi([-5,5],2,3), Y=randi([-5,5],2,3)
X =
-4 5 3
5 0 -4
Y =
-1 3 2
5 5 -5
>> X > Y
ans =
2×3 logical array
0 1 1
0 0 1
>> (X > Y) & (Y < 1)
96 Programming Mathematics Using MATLAB®

ans =
2×3 logical array
0 0 0
0 0 1
>> (X > Y) | (Y < 1)
ans =
2×3 logical array
1 1 1
0 0 1
>> xor(X > 0, Y > 0)
ans =
2×3 logical array
0 0 0
0 1 1

The any and all operators are useful, and work on vectors or on the columns of a
matrix.

>> x=randi(10,1,5), y=randi(10,1,5)

x =
5 5 4 10 4
y =
2 8 4 3 5
>> any(y > x)
ans =
logical
1
>> all(y > x)
ans =
logical
0
>> all(X > 5)
ans =
1×3 logical array
0 0 0
>> all(Y >= 3)
ans =
1×3 logical array
0 1 0
>> any(Y < 0)
ans =
1×3 logical array
1 0 1

Note how any and all work column-wise or over all elements on matrices.

>> any(Y == 5, 2)
ans =
 Control Flow 97

2×1 logical array

0
1
>> any(X == 2, 'all')
ans =
logical
0
>> any(X == 0, 'all')
ans =
logical
1

Also note where the operators fall under the order of precedence (see Table 6.3).

6.2. If statements
Flow control in the form of if statements is vital for programming. MATLAB® has if,
if-else, and if-elseif-else statements. These statements can be nested. Notice
that there is no begin in these statements, but there must be an end.
If statements have the following pattern:

if SOME TEST EXPRESSION

MATLAB command(s)
end

a=5; b=-2;
if b ~= 0 && (a == 5 || a == 3)
yy = a/b;
zz=a+b;
disp('hi!')
end

Table 6.3 Precedence order of operators.

1 Parentheses
2 Exponents and transpose (')
3 Negation
4 Multiplication and division
5 Addition and subtraction
6 Colon operator
7 Relational operators
8 Logical AND (&, &&)
9 Logical OR (|, ||)
10 Assignment (=)
98 Programming Mathematics Using MATLAB®

The following function ispos uses an if-else statement.

function y = ispos(n)
% ISPOS(N) returns true if N is positive, false otherwise.
if n > 0
y = true;
else
y = false;
end
end

Example 6.2.1. Use an if-else statement to define a function for the following
piecewise function:

2x − 5 x > 3,
f (x) =
x2 − 8 x ≤ 3.

function f = pwexample(x)
% PWEXAMPLE(X) demonstrates an if-else statement for piecewise function.

if x > 3
f = 2*x - 5;
else
f = x^2 - 8;
end

NOTE: the above example will not work if x is a vector or matrix, even if component-
wise calculations are used. See Example 6.4.1 on how to accommodate this.

Consider the following nested statement.

if x < 0
s = -1;
else
if x > 0
s = 1;
else
s = 0;
end
end

Nested statements such as the one above are common in other languages. In MAT-
LAB as in some other languages, one can instead take advantage of if-elseif-else
statements to accomplish the same thing but more efficiently.
 Control Flow 99

if x < 0
s = -1;
elseif x > 0
s = 1;
else
s = 0;
end

6.3. Switch/case
The switch command is an alternative to using nested statements or if-else-elseif state-
ments in some cases. Switch tests against the equality of an expression against certain
value(s). Switch statements work best against a discrete set of values, while in the case of
the piecewise function example with inequalities, one would still need an if-statement.

z = input('Enter an integer: ');

switch mod(z,2)
case 0
disp('you entered an even integer')
case 1
disp('you entered an odd integer')
otherwise
disp('you did not enter an integer')
end

switch x
case {-1, 0, 1}
y = 3;
case {-2, 2}
y = 5;
otherwise
y = 7;
end

6.4. Use of characteristic functions

In mathematics, characteristic functions or indicator functions are functions of the form
χ : A → {0, 1} where A is a set within a larger set U. For any x ∈ U, χ (x) = 1 if x ∈ A,
and χ (x) = 0 if x ∈/ A. For example, if U = R and A = [0, ∞), then χ (2.5) = 1 and
χ (−3) = 0.
100 Programming Mathematics Using MATLAB®

Because the logical true is also 1, and logical false is 0 within MATLAB, we can easily
create characteristic functions that can be implemented to create piecewise functions
without if-statements, among other things. Consider the following example.

Example 6.4.1. Rewrite the piecewise function in Example 6.2.1 with the use of a
characteristic function.

function f = pwexample2(x)
% PWEXAMPLE3(X) demonstrates an characteristic function to create a
% piecewise function.
% This function will work if X is a vector or matrix.

f = (2*x - 5).*(x > 3) + (x.^2 - 8).*(x <= 3);

end

6.5. For loops

For loops are extremely useful when you want to cycle through data and perform tasks,
or when there is an iterative process. They have the following format:

for VARIABLE = EXPRESSION

STATEMENT(S) OR COMMAND(S)
end

The expression is usually a vector defined with the colon operator or explicitly defined.
It can also be a matrix, but note that implementation is done one column at a time.
Consider the following function, myscatter, which creates n random points (x, y)
and plots each one.

function myscatter(n)
% MYSCATTER Example of using a for loop to create a scatter plot.
% MYSCATTER(N) creates a plot of N points that are randomly generated.
x=rand(1);
y=rand(1);
plot(x,y, '*', 'MarkerSize',8)
hold on
for k=2:n
x=rand(1);
y=rand(1);
plot(x,y, '*', 'MarkerSize',8)
end
hold off
end
 Control Flow 101

Note that this is not the most efficient way of doing this within MATLAB. First, there
is already a scatter plot command and secondly, we can make the above code more
efficient by what is called VECTORIZING THE CODE, which we will discuss later.
Another common use of a for loop is for calculating sums.

function s = geomseries(r, n)
% GEOMSERIES computes the sum of the geometric sequence
% Y = GEOMSERIES(R,N) is the sum of R^k from k=0 to k=N.
s = 0;
for k=0:n
s = s + r^k;
end
end

The following example shows how to use nested for-loops to work with matrices.

Example 6.5.1. Improve the previous function in Example 6.2.1 so that it also works
if x is a vector or matrix with the use of if-statements and nested for-loops.

2x − 5 x > 3,
f (x) =
x2 − 8 x ≤ 3.

function f = pwexample3(x)

f = 0*x;
[rows,cols] = size(x);
for j = 1:rows
for k = 1:cols
if x(j,k) > 3
f(j,k) = 2*x(j,k) - 5;
else
f(j,k) = x(j,k)^2 - 8;
end
end
end

We saw in Example 6.4.1 that this longer code is not necessary with the use of
characteristic functions, but there are many examples why nested for loops would be
necessary. For example, if the index of the matrix entry is needed, or in the case of
double (triple, etc.) summations.
The colon operator can be used to have the for loop increment backwards, or you
can explicitly state the values as seen in the examples below.
102 Programming Mathematics Using MATLAB®

for k = 10:-1:1
disp(k)
pause(1)
end
disp('Blast off!')
%% different increment
for k = 2:2:8
disp(k)
pause(0.5)
end
disp('Who do we appreciate?')
%% using defined vector
x = [2 4 6 8];
for k = x
disp(k)
pause(0.5)
end
disp('Who do we appreciate?')

The following example shows how the for loop works with matrices. Run the code
to see the results. Note that the results within the loop are displayed for demonstration
purposes.

x=[2 6; 4 8];
for k=x
disp(k)
pause(0.5)
end
disp('Who do we appreciate?')
%% showing the matrix implementation
x
for k=x
k
sumk = sum(k)
maxk = max(k)
end

6.6. While loops

While loops can be used to accomplish similar and different tasks as a for loop. The
format of a while loop is as follows.

while EXPRESSION
STATEMENT(S) OR COMMAND(S)
end
 Control Flow 103

The statements or commands within the loop are executed as long as the expression
is true. This can mean that until something with expression changes, the loop may
become an infinite loop.
You can use while loops to generate better data, or continue to ask the user until a
valid answer is submitted.

a = input('Enter a non-zero integer: ');

while (a == 0) | (mod(a,1) ~= 0)
a = input('Enter a non-zero integer: ');
end
b = randi([-10,10]);
c = randi([-3,3]);
while (b^2 - 4*a*c <= 0)
b = randi([-10,10]);
c = randi([-3,3]);
end

While loops are also useful in working with tolerance checks. In general, the dif-
ference between while loops and for loops is with for loops, the number of times to
loop through is known, and with while loops, you want to keep doing something until
something changes.

6.7. Useful commands break, continue, return, and error

The break command terminates the loop and the first statement after the loop’s end is
executed. If break is used within nested loops, the inner loop is executed and control
is passed to the loop at the next higher level.
The continue command allows the call of the command between the continue
statement and the end for that loop to be bypassed, and the next increment of the loop
is then executed. Note the differences when the following code is run.

%% break example
clc
for k=1:5
if k==3
break
disp('ah ha!')
end
disp(k)
end
%% continue example
clc
for k=1:5
if k==3
continue
104 Programming Mathematics Using MATLAB®

disp('ah ha!')
end
disp(k)
end

The return command is useful for script or function files, in which the command
acts like an end to the function, or “end of file” of the script file.
The error command is useful in which you can develop meaningful error messages
to the user. One can accomplish a similar task using disp or even fprintf (more on
that command later) along with return, but error is more efficient and the error
message is automatically displayed in red similar to MATLAB errors.

n = input('Enter a non-zero number: ')

if n == 0
error('You entered zero. Try again.')
end

6.8. Optional inputs and outputs of functions

One can work with optional inputs and outputs using nargin, varargin, nargout,
and varargout variables. The nargin variable is a number that gives you how many
inputs were given to the function, while nargout is the number of output arguments
requested by the user. These variables allow us to use functions in multiple ways. The
use of the nargout variable is best explained with the SIZE command in MATLAB:

>> A
A =
1 2 3
4 5 6
>> size(A) % no output specified
ans =
2 3
>> y = size(A) % one output specified
y =
2 3
>> [r, c] = size(A) % two outputs specified
r =
2
c =
3

The following function quadratic4 was written to display or store the outputs
similar to the size command but for the quadratic function.
 Control Flow 105

function [r1,r2] = quadratic4(a, b, c)

% QUADRATIC4 uses the quadratic formula.
% [R1, R2] = QUADRATIC4(A, B, C) to compute
% the two roots, R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0.
d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

if nargout < 2
r1 = [r1, r2];
end
end

The function quadratic5 demonstrates the use of nargin in addition to nargout.

This function allows for either two or three input variables for the coefficients of a
quadratic equation. If two coefficients are input, the leading coefficient a is set equal
to 1.

function [r1,r2] = quadratic5(n1,n2,n3)

% QUADRATIC5(A,B,C) or QUADRATIC5(B,C) uses the quadratic formula for the given
coefficients.
% [R1, R2] = QUADRATIC5(A, B, C) to compute
% the two roots, R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0.
% [R1, R2] = QUADRATIC5(B, C) then A = 1.
if nargin == 2
a = 1;
b = n1;
c = n2;
else
a = n1;
b = n2;
c = n3;
end

d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

if nargout < 2
r1 = [r1, r2];
end
end
106 Programming Mathematics Using MATLAB®

The variables varargin and varargout are cell arrays of variable input or output
arguments. The quadratic6 function demonstrates the use of varargin to function
similarly to quadratic5.

function [r1,r2] = quadratic6(varargin)

% QUADRATIC6(A,B,C) or QUADRATIC6(B,C) uses the quadratic formula for the given
coefficients.
% [R1, R2] = QUADRATIC6(A, B, C) to compute
% the two roots, R1 and R2 of the quadratic equation
% Ax^2 + Bx + C = 0.
% [R1, R2] = QUADRATIC6(B, C) then A = 1.
if nargin == 2
a = 1;
b = varargin{1};
c = varargin{2};
else
a = varargin{1};
b = varargin{2};
c = varargin{3};
end

d = b^2-4*a*c;

r1 = (-b-sqrt(d))/(2*a);
r2 = (-b+sqrt(d))/(2*a);

if nargout < 2
r1 = [r1, r2];
end
end

The function mySquare demonstrates another nice use for varargin. This function
allows for optional LineSpec (line style, marker symbol, and color).

function mySquare(s,x,y,varargin)
% MYSQUARE plots a square with length S with lower left corner at X,Y
% MYSQUARE(S,X,Y) or MYSQUARE(S,X,Y, LINESPEC)

plot([x,x+s,x+s,x,x],[y,y,y+s,y+s,y],varargin{:})

The function myRandN is another demonstration of using both nargin and

varargin.

function y=myRandN(mu,sigma,varargin)
% MYEXAMPLE demonstrates using nargin and vargin
switch nargin
case 2
 Control Flow 107

y=randn*sigma + mu;
case 3
y=sigma*randn(varargin{1}) + mu;
case 4
y = sigma*randn(varargin{1},varargin{2}) + mu;
end
end

6.9. Exercises
√
x3 2x + 3
1. Write a function named hw6prob1 for the function f (x) = that does
(x4 + 1)3
the following:
• The input of the function should be x and the output should be the computed
value y = f (x).
• Write the function so that x can be a number, vector or matrix.
• Using the any and/or all logical operators, first test within the function to
make sure that all of the values of x are valid for R-valued calculations. If not,
an appropriate error message is displayed using the error command.
• You should test your calculations with at least 4 different inputs. Write down
the input tests you used, along with the answers that your function gives you.
2. Create a function called isZ that will take one input and output a logical true
or false (value of either 0 or 1 specifying false or true, respectively). If the input
is an integer (NOT TALKING ABOUT DATA TYPE INTEGER – meaning
it is not π , or not 3/2, etc.), then it will return a logical true, otherwise it will
return a logical false. Note that this should still work if the input is a vector or
matrix, and would return a vector or matrix of true/false based on whether each
entry/component is an integer or not. Run tests of the function. This function
should be used on any subsequent problem that needs to check if a variable or
input is an integer.
3. Create a function called isEven that will have one input that is supposed to be
an integer. First check whether the input is an integer; if not, use the error
command to display an appropriate error message. The function should return a
logical true if the input is even, and a logical false if it is odd. Run tests of the
function.
4. Finish the function plotVec that is found on the text website. This function has
one or more inputs and uses varargin. The first input is P, which is a column
vector with two or three rows (check for it: if not, return an error). The function
will take the column vector and plot the vector as a line segment from the origin
to the given point (from the input P). Any other input arguments are optional and
specify how the vector will be drawn (color, line width, marker type, etc.). The
108 Programming Mathematics Using MATLAB®

function will not have an output (it will only create a plot). Hint: remember the
difference between plot and plot3! Run it once for a 2D point and another time
for a 3D opint, using your choice of plot specifications (or none) on each.
BONUS: create a function plotVec2 that will expand on plotVec. This function
will take a MATRIX of 2D or 3D points and connect them all, in order. Thus it
could create a polygon from the matrix in which each column is a vertex of the
polygon. Discuss the limitations or difficulties this may have in coding it and/or
implementation.
5. Finish the function plotPlane that is found on the text website. This function
has two or more inputs and uses varagin. The first input is M which is a matrix
that has exactly three rows and at least two columns (check for it; if not, return
an error). The second input ax determines the domain for the plane; the domain
will be from -ax and ax (Hint: remember meshgrid!) Any other optional inputs
will specify EdgeColor, etc. to plot the plane using the mesh command. Use the
function to plot a plane using M =randi; create M, display it and then use your
function to plot the plane.
6. Create a new function called parity that does the following:
⎧
⎪
⎪
⎨1 x is even,
f (x) = 0 x is odd,
⎪
⎪
⎩ −1 x is not an integer.

We will be doing our own checking if x is an integer even if MATLAB has a

function for this. Make sure x can be a number, vector, or matrix, and that the
function then returns a corresponding number, vector, or matrix.
(a) Calculate f (x) for x = [2, −3, −4, 5, 0, 1/2].
(b) Calculate f (x) for
 
0 −1
x= .
π 100
7. Create a new function called npdf2 that does the same as npdf in Problem 2 in
Chapter 5, but IN ADDITION:
• The function npdf2 has OPTIONAL inputs σ and μ. If only one input is
given (i.e., x is the only input), then σ = 1 and μ = 0. If two inputs are given,
they are x and σ and μ is taken to be equal to 0. If three inputs are given, they
are x, σ , and μ. If the optional inputs are given, the following “error checks”
are done.
• Checks to make sure the inputs σ , and μ are numbers rather than vectors or
matrices. Here we will create our own check (even though MATLAB may
have functions that can do these checks). If a vector or matrix was input, an
appropriate error message is displayed.
 Control Flow 109

• Checks to make sure σ > 0; if not, an appropriate error message is displayed.

8. Create a SCRIPT FILE called pball.m that will use the input command to take
ask for the user, “How many Power Ball tickets?” (n). Check that n is a number
(rather than a vector or matrix) and a positive integer and if not, keep asking the
user “Try again. How many Power Ball tickets?” until the input is correct. Then
using input again, ask, “Power Play option? (y/n)” (PP). Likewise, if the answer
to the “Power Play” option is not “Y”, “y”, “N”, or “n” keep asking the user,
“Try again. Power Play option? (y/n)” You will use the switch/case command
for the answer to the “Power Play” question. If PP is “Y” or “y”, each Powerball
ticket costs $3. If PP is “N” or “n”, each Powerball ticket costs $2. The file then
displays a matrix with n rows in which each row is a Powerball ticket (as specified
in Exercise 9(d) of Chapter 2 [The first five entries of each ticket are the “white
balls” and the last entry of each ticket is the “Powerball” for the drawing. For
each ticket, the first five numbers should be random integers from 1 to 69 with
no repeats and not necessarily in order and the last number should be a random
integer from 1 to 26.] but can now be done in an easier way). The amount owed
for the Powerball tickets is also displayed. You do not need to get fancy with
displaying the tickets and price; we will work on that later.
Generate randi(10) Power Ball tickets and your choice as to whether or not the
Power Play option is selected.
9. A circle of radius r and center (h, k) can be parameterized with the equations

x = r cos(θ ) + h
y = r sin(θ ) + k

for θ ∈ [0, 2π ]. Create a function myCircle that takes as inputs a positive number
r, a 2D point (vector) C, and an optional positive integer n. Error checking should
be done on r, C (and n if input) and appropriate messages displayed using the
error command. This function will output vectors x and y of size n (with default
value of n = 100) that are the calculated parametric equations for a circle with
radius r with center C (h, k). Have the output be similar to size or quadratic4
(page 105) in that both are output regardless of how the user runs the function.
(a) Use your function to plot the unit circle and another circle with radius 3
and center (−2, 4).
(b) Create another graph where you plot the unit circle but with no input for
n, n = 4, n = 5, and n = 9.
The command axis equal may come in handy for these plots!
10. We will expand on the Problem 8 in Chapter 5.
(a) Create a function randInt.m that takes as inputs two OR three arguments.
If the two arguments a and b are input, then the function generates a random
110 Programming Mathematics Using MATLAB®

integer from a to b. If the three arguments a, b, and c are input, the function
generates a random number from a to b with an increment of c. So if I
input randInt(-5,5) it would give me a random integer (NOT data type
integer!) from −5 to 5. But if I input randInt(-5,5,.5) it would give me
a random number from the list −5, −4.5, −4, −3.5, . . . , 4, 4.5, 5. ERROR
CHECK: check to make sure a < b and (if input) c > 0. If either of these
errors occur, use the error command to stop the function and output an
appropriate error message.
(b) Create a function nonzeroRand.m that takes as inputs two OR three argu-
ments. If the two arguments a and b are input, then the function generates a
random NON-ZERO INTEGER from a to b. If the three arguments a, b,
and c are input, the function generates a random NON-ZERO NUMBER
from a to b with an increment of c. You will use a WHILE loop within this
function. ERROR CHECK: check to make sure a < b and (if input) c > 0.
If either of these occur, use the error command to stop the function and
output an appropriate error message.
11. You are creating a multiple choice problem in which students will be asked to
solve ax + b = cx + d for x, where a, b, c, and d are certain random integers. You
have decided that the wrong answers given in the problem are:

b−d a−c b+d b+d

w1 = , w2 = , w3 = , and w4 = .
a−c d−b a−c a+c

Initially, the problem is set such that a is a random integer between 2 and 6, b is
a non-zero random integer between −4 and 4, c is a random integer between 3
and 7, and d = a + b − c − 1. The problem is set that if c = a or a = c + 1, a new c is
generated until both c = a and a = c + 1. The problem is also set that if b = d, a new
b is generated (and thus d) until b = d. (Think: why would we want this check?
Unfortunately, you notice that there are times that some coefficient combinations
make it so there are duplicate answers listed in the multiple choice problem. You
already realize there are issues when the denominators and/or numerators equal
zero but those are already taken care of; you need to find the other bad combina-
tions. Instead of trying to figure out algebraically how to define the coefficients so
this does not happen, you write a MATLAB script called algProblem.m to help
pinpoint what combination(s) of coefficients will and will not work.)
(a) What is the correct answer to the problem in terms of a, b, c, and d? Show
all work on paper.
(b) Create a script file algProblem that does the following:
• Initialize vectors A, B, and C to contain all of the possibilities of values
for a, b, and c, respectively.
 Control Flow 111

• Initialize a counter variable (name of your choice under MATLAB

guidelines) to equal 0. This variable will keep track of how many “bad”
coefficient combinations there are.
• Create nested for loops, the first based on the possibilities for a (the
elements in A), the second based on the possibilities for c, and the third
based on the possibilities for b (this is purposely “out of order”). These
for loops will cycle through all of the possible combination of coeffi-
cients and perform the checks.
• Within the second for loop, check to see if any of the denominators and
numerators of the answers that are based on a and c equal zero. If so, the
continue command is used to move onto the next iteration of this for
loop (and thus the next combo possibility for a and c). This combination
is not “counted” since other code within the problem already disallows
this combination.
• Within the third for loop define d based on the current values of a, b,
and c and the formula given above.
– A check is made to see if any of the denominators and numerators
of the answers (both correct and wrong!) that are based on b and d
equal zero. If so, again the continue command is used to move onto
the next iteration of the loop (and thus the next set of coefficients)
since similar to above this combination is already disallowed and not
“counted” as a bad combination.
– Define variables for the five answers to the multiple choice question
(the correct and four wrong answers).
– We want to make sure all five answers are distinct values. How many
pairwise comparisons need to be done? There are multiple ways to
do this check. One way is to create a logical vector that compares
each pair of answers with each other. Thus this vector should contain
only ones and zeros. For example, the first entry of the vector equals
true (1) if the w1 = w2 and false (0) otherwise.
– Using the any command on the vector, if any of the values in this
logical vector equals true (1) (i.e., any pair of answers is equal), then
the current combination of coefficients is “bad.” The counter from
above is incremented, the coefficients are output to the command
window along with the five calculated answers, in the order of: cor-
rect, w1, w2, w3, and w4. Thus when the script is run, you can see
by the end how many “bad” combinations there are, and you can
scroll through and see not only the coefficients but the five answers
of the multiple choice problem. Try to make the output as clear and
concise as possible; you may want to create a vector abcd that has
112 Programming Mathematics Using MATLAB®

all four values of the coefficients and output that vector; also create a
vector Answers that has the multiple choice answers and output that
vector.
(c) How many bad combinations of coefficients are there? What are they? State
your answer on your paper and list the bad combinations of coefficients
clearly. No need to show your work on this one; I should be able to run
your script to see the work. Do not run the script file within the homework
file.
12. Create a new function called mysgn that is a variation on the signum function:
⎧
⎪
⎪
⎨1 x > 0,
sgn(x) = 0 x = 0,
⎪
⎪
⎩−1 x < 0.

In this case, mysgn will have the following criteria:

• Has one input x which can be a number, vector, or matrix, and will have as
output y ⎧ that is a (1 × 4) row vector y.
⎪
⎪
⎨1 x or all elements are > 0,
• y(1, 1) = −1 x or all elements are < 0,
⎪
⎪
⎩0 otherwise,

1 x or any elements are > 0,
• y(1, 2) =
0 otherwise,

−1 x or any elements are < 0,
• y(1, 3) =
0 otherwise,

0 x or any elements are = 0,
• y(1, 4) =
1 otherwise.
• This function should run like the size function in MATLAB; if all four out-
puts are specified, it assigns y(1, 1) to the first specified output, and y(1, 2) to
the second specified output, and if one or no outputs are specified, it gives
a vector with containing the answers. Any other number of outputs specified
returns an error with a meaningful message.
√
(a) Calculate u = mysgn(x), v = mysgn(a), w = mysgn(b) for x = − 2,
a = π , and b = 0.
(b) Calculate [a,b,c,d]= mysgn(x) for x = -10:2:10.
(c) Calculate mysgn(x) for x = 0:2:10.
(d) Calculate mysgn(x) for x = 1:2:10.
(e) Calculate mysgn for A = eye(3).
(f) Calculate mysgn for B = ones(3).
 Control Flow 113

(g) Calculate mysgn for -3∗A.∗B.

13. Notice what happens when we enter (-8)^(1/3) into MATLAB. Create a func-
tion rootn that gives us the real-valued third root of a negative integer. The
√
function rootn will calculate n x and also does the following for a number x:
√
• The inputs are n and x of which we are going to compute n x. The help lines
should indicate what the inputs are and in what order.
• The function checks that x is a number rather than a vector or matrix. If not,
an error message is displayed and the function should stop.
• Another check is that n is a natural number. If not, an appropriate error mes-
sage should be displayed and the function should stop.
• If n is even, it checks to make sure that x is nonnegative. If not, an error
message is displayed and the function should stop.
• If n is odd, it computes the root as expected. In other words, it should give
√
−2 as the answer for 3 −8.
NOTE that we are creating our own version of the MATLAB function nthroot.
CHAPTER 7

Miscellaneous Commands and Code

Improvement
7.1. Miscellaneous commands
7.1.1 The fprintf command
The command fprintf allows us to specify more formatting, including new lines and
tabs to display text and/or variables. The command disp can be used to do this, but it
does not allow new lines, tabs, or special formatting.

>> n=5; x=1/2;

>> y=geomseries(x,n);
>> fprintf('The estimate of the geometric series when r = %f and n = %i is %f',x,n,y)
The estimate of the geometric series when r = 0.500000 and n = 5 is 1.968750>>

Notice if you copy the above fprintf command and run it, the command prompt
and cursor is at the end of the text, which is difficult to see. To get a new line, use “\n”.
A tab is “\t”.

>> fprintf('\n\nThe estimate of the geometric series when r = %f and n = %i\n\t is %f

\n\n',x,n,y)

The estimate of the geometric series when r = 0.500000 and n = 5

is 1.968750

Notice the “placeholders” for where the variables go in the text. These are using
formats.

>> fprintf('\n Here is 2^10: %f\n\n',2^10)

Here is 2^10: 1024.000000

>> fprintf('\n Here is 2^10: %e\n\n',2^10)

Here is 2^10: 1.024000e+03

>> fprintf('\n Here is 2^10: %E\n\n',2^10)

Here is 2^10: 1.024000E+03

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00012-0 All rights reserved. 115
116 Programming Mathematics Using MATLAB®

What if you wanted to display only a certain number of decimal places?

>> fprintf('\n Here is e: %4.2f\n\n',exp(1))

Here is e: 2.72

>> fprintf('\n Here is e: %2.4f\n\n',exp(1))

Here is e: 2.7183

Further explanations of the above formatting can be found in the MATLAB® doc-
umentation.
The formatting in the fprintf does not pay attention to the formatting you have
set outside of the command (see the section below for a table of formats):

>> format bank

>> exp(1)

ans =

2.72

>> fprintf('\n Here is e: %f\n\n',exp(1))

Here is e: 2.718282

You can get fancy with justification:

>> fprintf('\nsome text:%3.0f\nsome text:%3.0f\n\n',7,2^10)

some text: 7
some text:1024

>> fprintf('%-2.0f\n%-2.0f\n\n',7,2^10)
7
1024

>> fprintf('\nsome text:%-6.0f\nsome text:%-6.0f\n\n',7,2^10)

some text:7
some text:1024

Another nice format is using “%g” which will use either %f or %e, whichever is the
shortest:

>> fprintf('%f %f\n\n', 2^10, 2^100)

1024.000000 1267650600228229401496703205376.000000
 Miscellaneous Commands and Code Improvement 117

>> fprintf('%e %e\n\n', 2^10, 2^100)

1.024000e+03 1.267651e+30

>> fprintf('%g %g\n\n', 2^10, 2^100)

1024 1.26765e+30

7.1.2 The sprintf command

The command sprintf is the same as fprintf except that its output is a string, which
can be useful for labeling plots (see Fig. 7.1).

a=10;
b=34/(2*pi);
plottitle=sprintf('Helix with a=%i and b approximately %3.5f',a,b);
t=linspace(0,100,750);
x=10*cos(t);
y=10*sin(t);
z=b*t;
plot3(x,y,z)
title(plottitle)

It is also useful for error messages.

>> n=101;
>> errormsg = sprintf('We need n to be even. You entered n = %i',n);
>> if mod(n,2)~=0
error(errormsg)
end
We need n to be even. You entered n = 101

Figure 7.1 Example using sprintf with title.

118 Programming Mathematics Using MATLAB®

7.1.3 Formats revisited

In Section 1.2.3 using format was discussed. Recall that the default format is to dis-
play four decimal places (which is actually format short). Above we had switched to
format bank. To get back to the default format, use just format.

>> x=10*pi

x =

31.42

>> format
>> x

x =

31.4159

See Table 1.4 for other formats one can use.

Notice that long e, short g, etc. can be one word:

>> format compact

>> format longe
>> x
x =
3.141592653589793e+01
>> format +
>> x
x =
+
>> -x
ans =
-
>> format shorteng
>> 1000*x
ans =
31.4159e+003
>> pi/1575
ans =
1.9947e-003
>> pi/22222
ans =
141.3731e-006

7.1.4 The save/load commands

One can save workspace by using the command save, such as
 Miscellaneous Commands and Code Improvement 119

>> save('saveddata1.mat')

Then one can load that file at any other time to retrieve the data.

>> load('saveddata1.mat')

7.1.5 The tic/toc commands

The command tic starts a timer, while toc ends it and stores or displays the amount of
time used.

>> tic
>> n=0
n =
0.0000e+000
>> for k=1:10000
n=n+n^2;
end
>> toc
Elapsed time is 12.720630 seconds.

Notice the above is also the time it took to type those commands into the command
window.

7.1.6 The fill command

The fill(x,y,color) command fills the area between the points (x, y) given with
the given color with a black edge, connecting automatically the first point with the last
point. Fig. 7.2(A) below shows using fill on a sine wave and Fig. 7.2(B) fills a unit
circle.

% First example
color1 = [0,0.4470,0.7410]; % default first color starting in version 2014a
x=linspace(-3,3);
fill(x,sin(pi*x),color1)
% Second example
t=linspace(0,2*pi);
xc = cos(t); yc = sin(t);
fill(xc,yc,color1)

If x and y are defined with x=linspace(-2,2); y=x.^2; you get Fig. 7.3(A)
using fill(x,y,color1). For regions such as “area under the curve”, you will need
to adjust the points. Using fill([-2,x,2],[0,y,0],color1) on the same x and y
leads to Fig. 7.3(B).
120 Programming Mathematics Using MATLAB®

Figure 7.2 Examples of fill command. (A) Filling y = sin(π x ), (B) Filling a circle.

Figure 7.3 Adjusting to get the area under the curve. (A) Fill y = x 2 , (B) Adding points to adjust.

Here follows another example of “area under the curve” that created Fig. 7.4.

color2=[0.8500, 0.3250, 0.0980]; % default 2nd color since v2014a

x=linspace(-pi,pi);
y=cos(x)+2;
fill([x(1), x,x(length(x))],[0, y,0],color2)
axis equal

For basic regions between two functions, one can do the following to create Fig. 7.5.
Notice the use of the useful command fliplr(x) which creates a vector that has the
values of x but in reverse order.
 Miscellaneous Commands and Code Improvement 121

Figure 7.4 Another area under curve.

Figure 7.5 Basic region between two curves.

x=linspace(-pi,pi);
y=cos(x);
x2=fliplr(x);
y2=-sin(x2)-3;
fill([x,x2],[y,y2],color2)

Here is another example using fill (see Fig 7.6).

t = pi/8: pi/4: pi/8 + 2*pi;

xo=cos(t); yo=sin(t);
fill(xo,yo,'r')
hold on
plot(xo,yo,'w','LineWidth',10)
plot(xo,yo,'k')
122 Programming Mathematics Using MATLAB®

hold off
axis equal
ax=1.2;
axis([-ax,ax,-ax,ax])

Figure 7.6 Using fill for stop sign example.

7.1.7 The command alpha

The order of plot commands, including fill, can make a difference on the appearance
in the figure window. If you run the following code within MATLAB, since the fill
comes after plotting the sine wave, it “covers up” the sine wave as shown in Fig. 7.7(A).

plot(xsine,ysine)
hold on
fill(xtriangle,ytriangle,'y')
hold off

If we reorder the commands as in the following code, the sine wave is then plotted
“on top” of the filled triangle, as shown in Fig. 7.7(B).
For certain plotting commands, the alpha(n) command can create transparency.
The value n is a number between 0 and 1, with 1 being opaque, and 0 being completely
transparent. (Experiment!) Thus a combination of the order of the commands and the
use of alpha can give you more control over the look of the figure (see Fig. 7.8).

plot(xsine,ysine,'k')
hold on
 Miscellaneous Commands and Code Improvement 123

fill(xtriangle,ytriangle,'g')
alpha(0.3)
hold off
axis equal

Figure 7.7 The order of plot and fill commands matter. (A) Plot first, then fill, (B) Fill first, then plot.

Figure 7.8 Using alpha to create transparency.

Note that the alpha command will apply to all commands that appear before it; thus
you can control which objects get transparency by the order of the alpha and plotting
commands, as shown in Fig. 7.9 below.

fill(xtriangle,ytriangle,'g');
hold on
fill(xcircle,ycircle,'r');
alpha(0.2)
124 Programming Mathematics Using MATLAB®

fill(xsine,ysine,'b');
hold off
axis equal

You can also define different alpha to specific commands, or apply alpha to only one
as the following code demonstrated in Fig. 7.10.
f1=fill(xtriangle,ytriangle,'g');
hold on
f2=fill(xcircle,ycircle,'r');
f3=fill(xsine,ysine,'b');
alpha(f3,0.4)
alpha(f2,0.7)
hold off
axis equal

Figure 7.9 Order of alpha and other commands matter.

Figure 7.10 Order of alpha and other commands matter.

 Miscellaneous Commands and Code Improvement 125

7.1.8 The syms, diff, int, and subs commands

The following subsection uses commands that require the Symbolic Math Toolbox. This
may not be installed on your version of MATLAB.
There are many different commands and ways to create symbolic variables and func-
tions within MATLAB. I am only giving a few examples here that are useful for some
of the exercises.

>> syms x
>> f(x) = sin(pi*x)
f(x) =
sin(pi*x)
>> f(1)
ans =
0
>> A=f([0, 1/6, 1/4, 1/3, 1/2])
A =
[ 0, 1/2, 2^(1/2)/2, 3^(1/2)/2, 1]
>> double(A) % this will give numeric approximations to A
ans =
0 0.5000 0.7071 0.8660 1.0000

If you already have a string defined that we want to turn into a symbolic function,
one can use the str2sym command for MATLAB R2017b or later. Otherwise one
should use the evalin command. Other ways generate warnings and errors. Note that
component-wise operations are not needed.

>> fstring = 'x^2 - 2*x + 1'

fstring =
'x^2 - 2*x + 1'
>> syms x
>> f(x) = str2sym(fstring)
f(x) =
x^2 - 2*x + 1
>> f(3)
ans =
4
>> f([0:2])
ans =
[ 1, 0, 1]
>> f([0 1;2 3])
ans =
[ 1, 0]
[ 1, 4]

Symbolic variables can be declared and manipulated algebraically.

126 Programming Mathematics Using MATLAB®

>> syms x y
>> g(x)=x^3*sin(x);
>> t=linspace(-10,10);
>> plot(t,g(t))
>> expand((x+y)^3)
ans =
x^3 + 3*x^2*y + 3*x*y^2 + y^3
>> factor(x^2 + x*y -2*y^2)
ans =
[ x - y, x + 2*y]

The diff command can be used to differentiate a symbolic function.

>> syms x
>> f(x)=sin(pi*x) + exp(x);
>> g(x)=25*x^4;
>> df = diff(f)
df(x) =
exp(x) + pi*cos(pi*x)
>> dg = diff(g,x)
dg(x) =
100*x^3
>> dg(-1)
ans =
-100
>> syms t
>> h(x,t) = t^2 - 5*t*x + x^3
h(x, t) =
t^2 - 5*t*x + x^3
>> dhdt = diff(h,t)
dhdt(x, t) =
2*t - 5*x

Similarly, the int command can be used for integrals, both definite and indefinite.

>> f,g,h
f(x) =
sin(pi*x) + exp(x)
g(x) =
25*x^4
h(x, t) =
t^2 - 5*t*x + x^3
>> int(f)
ans(x) =
exp(x) - cos(pi*x)/pi
>> int(g,1,2)
ans =
155
 Miscellaneous Commands and Code Improvement 127

>> inthdt = int(h,t)

inthdt(x, t) =
(t*(2*t^2 - 15*t*x + 6*x^3))/6
>> expand(inthdt)
ans(x, t) =
t^3/3 - (5*t^2*x)/2 + t*x^3

7.1.9 Commands for polynomials

Polynomials in MATLAB can be “stored” with a vector of their coefficients, with the
leading coefficient first and the constant term last. Remember that in this case
you must have 0s for the terms of that degree that are not a part of the polynomial.
Thus for the polynomial p1 (x) = π x2 − 2x + 3 one can define within MATLAB
the vector p1=[pi -2 3]; and for the polynomial p2 (x) = 4x + 1/2 one can define
p2=[4 1/2]. In order to work with both of them at the same time, the vectors should
be the same size. Thus it would be better to define p2=[0 4 1/2]. We can easily
add/subtract polynomials, since when we do this mathematically we just add/subtract
the coefficients of like terms. When the coefficients are stored in vectors of the same
size, this is equivalent to adding/subtracting the vectors.

>> p1=[pi -2 3];

>> p2=[0 4 1/2];
>> p1+p2
ans =
3.1416 2.0000 3.5000
>> p1-4*p2
ans =
3.1416 -18.0000 1.0000

The command conv (short for convolution) will do the appropriate algebra on
vectors that is equivalent to multiplication of polynomials. Note that for conv the vec-
tors do not need to be of the same size. The command deconv will perform polynomial
division.

>> conv([3 2],[1 2])

ans =
3 8 4
>> conv([1 -2], [1 2])
ans =
1 0 -4
>> conv([1 -2 1], [1 -1])
ans =
1 -3 3 -1
>> deconv([1 0 -9], [1, 3])
128 Programming Mathematics Using MATLAB®

ans =
1 -3

The polyder and polyint commands will return vectors that are the coefficients
for the derivatives and integrals, respectively, of the polynomials/vectors.

>> polyder([3 2 -4 3]) % 3x^3 + 2x^2 -4x + 3

ans =
9 4 -4
>> polyint([1 4 -3]) % x^2 + 2x - 3
ans =
0.3333 2.0000 -3.0000 0

The command polyval evaluates the given polynomial at the indicated value.

>> polyval([1 -2 3],3) % x^2 - 2x + 3

ans =
6

The roots command should be self-explanatory, but realize that it will also return
complex roots.

>> sln1 = roots([1 2 -3])

sln1 =
-3.0000
1.0000
>> sln2 = roots([1 2 1])
sln2 =
-1
-1
>> sln3 = roots([1 0 1])
ans =
0.0000 + 1.0000i
0.0000 - 1.0000i

The command polyfit(x,y,n) will try and fit (using least-squares) the input data
(x, y) to a polynomial of degree n as the following example demonstrates.

Example 7.1.1. Fit the points (0, 0), (−1, 1), and (2, −2) to a line. State the line and
plot the data points and line.
We can use polyfit with n = 1 for a line.

x=[0, -1, 2]; y= [0, 1, -1];

p=polyfit(x,y,1)
xplot=linspace(-2,3);
yplot=polyval(p,xplot);
plot(x,y,'*')
 Miscellaneous Commands and Code Improvement 129

hold on
plot(xplot,yplot,'k')
hold off

The above code returns

p =
-0.642857142857143 0.214285714285714

From this output, we see that the least-squares fit is

p(x) = (−0.642857142857143)x + 0.214285714285714,

and this line and these points are shown in Fig. 7.11.

Fig. 7.12 demonstrates the results of polyfit using polynomials of different degrees
on x=0.5; y=x.∗cos(x).∗exp(x).

7.2. Code improvement

There are many ways one can improve code to make the programs run more efficiently
and faster. One way is to reduce the number of unnecessary variables or steps. Some-
times, it is easier to write a program creating intermediate variables but then you can go
back and put those calculations or comparisons within the same line. Another way is to
fine tune the data types used in variables. For example, if you know that a variable will
be an integer between −128 and 127, you can specify the variable to be stored as int8
to save memory and execution time. These methods are often covered in a computer

Figure 7.11 Polyfit example for line.

130 Programming Mathematics Using MATLAB®

Figure 7.12 Polyfit example for different polynomial degrees.

science course and we will not go into details about this here. Many of our exercises in
this course will not warrant this type of detail.
This section will focus on two basic methods of code improvement that may not be
available in other languages: vectorization of code, and preallocation.

7.2.1 Vectorization of code

The method of vectorizing code is to capitalize on MATLAB’s many features and com-
mands with vectors and matrices. In many languages, a loop (usually a for loop) is
needed to accomplish a task/operation on a vector, or over a set number of times. in
MATLAB, sometimes one can create a vector and use the task/operation once on the
entire vector. Note that there are situations in which you may still need to use loops on
vectors and matrices. This may be explained best with an example.

Example 7.2.1. Calculate the geometric series the traditional way and using vector-
ization of code and compare using the tic and toc commands.

function [t1, t2] = geomseries2(r, n)

% GEOMSERIES2(R,N) uses the geometric series to demonstrate the vectorization of
code.
% [T1, T2] = GEOMSERIES2(R,N) will calculate two times using tic/toc.
% T1 is the time taken to calculate the summation of
% R^k from k = 0 to N using the traditional way.
% T2 is the time taken to calculate the summation of
% R^k from k = 0 to N using vectorization of code.
 Miscellaneous Commands and Code Improvement 131

% Traditional way
tic
s1=0;
for k=0:n
s1=s1+r^k;
end
t1 = toc;
% Vectorization of code
tic
k=0:n;
s2=sum(r.^k);
t2 = toc;
end

Now look at the difference in times. For various values of n, we now run the
following code.

[t1,t2]=geomseries2(0.5,n);
fprintf('\nThe two times for n = \%i are: \n\ttime1 = \%2.5f and \n\ttime2 = \%2.5f\n
\n', n, t1, t2);

The output of the above code follows.

The two times for n = 10000 are:

time1 = 0.01474 and
time2 = 0.00045

The two times for n = 100000 are:

time1 = 0.10631 and
time2 = 0.00269

The two times for n = 1000000 are:

time1 = 1.06955 and
time2 = 0.02689

>>

7.2.2 Preallocation
One can make programs more efficient by preallocating the size of large vectors or
matrices rather than appending. For example, if you have a recursive algorithm, you
can preallocate a vector or matrix for the number of entries needed to complete the
algorithm. Again, explanation will be by example.
132 Programming Mathematics Using MATLAB®

Example 7.2.2. Demonstrate the efficiency of preallocation by calculating the first n

Fibonacci numbers.

function F = Preallocation(n)
% PREALLOCATION(N) demonstrates the efficiency of preallocation.
% PREALLOCATION(N) calculates the first N Fibonacci numbers.

% Traditional way.
F1=[1 1];
tic
for k=3:n
nextOne = F1(k-1) + F1(k-2);
F1 = [F1 nextOne];
end
toc
% Preallocation
tic
F2=ones(1,n);
for k=3:n
F2(k) = F2(k-1) + F2(k-2);
end
toc
F = F2;

The comparison can be made by looking at the following output.

>> n=1000, Preallocation(n);

n =
1000
Elapsed time is 0.002641 seconds.
Elapsed time is 0.000031 seconds.
>> n=10000, Preallocation(n);
n =
10000
Elapsed time is 0.025771 seconds.
Elapsed time is 0.000274 seconds.
>> n=1000000, Preallocation(n);
n =
1000000
Elapsed time is 2.020418 seconds.
Elapsed time is 0.018352 seconds.
CHAPTER 8

Transformations and Fern Fractals

8.1. Linear transformations
Linear transformations have many special properties in addition to the properties from
the definition below. Linear transformations preserve lines, as opposed to nonlinear
transformations that may transform a line segment into a parabolic shape, or an ellipse,
etc.

Definition 8.1.1. Linear transformations are functions or map-

pings from a vector space V to a vector space W are transformations
(functions) T : V → W such that
• T (u + v) = T (u) + T (v) for all u, v ∈ V .
• T (cu) = cT (u) for all u ∈ V , c ∈ R.

It is discussed in linear algebra that linear transformations can be associated (uniquely)

with matrices. We will be looking at linear transformations from R2 to R2 , so these
transformations correspond to 2 × 2 matrices (with R-entries). Thus T (x) ↔ Ax, where
A is a 2 × 2 matrix and x is a 2 × 1 matrix or column vector.
Special examples of linear transformations are rotations, translations, and scaling
transformations (dilations/contractions).

• Scaling transformation (see Fig. 8.1) is a transformation T : R2 → R2 defined by

T (x) = cx where c ∈ (0, ∞). Using properties of scalar multiplication on vectors, we
see that the image is now c times longer (or shorter) than the input but is in the
same direction. Thus
– If c > 1, T is a dilation by a factor of c.
– If c ∈ (0, 1), T is a contraction by a factor of c.
We can put T into matrix form:
      
x c 0 x cx
T = =
y 0 c y cy

• Rotation by an angle θ about the origin O where the rotation is measured from
the positive x-axis (see Fig. 8.2) can be represented as:
    
x1 cos θ − sin θ x1
T = .
x2 sin θ cos θ x2

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00014-4 All rights reserved. 135
136 Programming Mathematics Using MATLAB®

Figure 8.1 Scaling example.

Figure 8.2 Rotation example with θ = π8 .

• Reflection about a line  through the origin, where θ is the angle from the positive
x-axis to  is given by
   
cos 2θ sin 2θ a b
A= = , a2 + b2 = 1.
sin 2θ − cos 2θ b −a
 Transformations and Fern Fractals 137

Figure 8.3 Reflection example about the y-axis.

   
x −x
Example 8.1.1. T = reflects the point (vector) in R2 about the
y y
y-axis (see Fig. 8.3). Using θ = π/2, the matrix for this transformation is given by
the equation

    
x −1 0 x
T =
y 0 1 y

• Composition of transformations Consider two linear transformations T and S,

both from R2 to R2 . Suppose A is the matrix corresponding to T and B is the
matrix corresponding to S; i.e.,

T (x) = Ax and S(x) = Bx ∀ x ∈ R2

Then the composition of the transformations T and S, T ◦ S, is given by the matrix

AB;
(T ◦ S)(x) = T (S(x)) = T (Bx) = A(Bx) = ABx.

If you think about some transformations that are performed after each other, the or-
der in which they are done will matter. Likewise, since matrix multiplication is not
commutative, the order matters (see Fig. 8.4).
The above are examples in which the matrices have particular patterns or formats
to generate particular transformations or geometric results. Any matrix will yield a
138 Programming Mathematics Using MATLAB®

Figure 8.4 Composition example. (A) Rotation, then reflection, then dilation, (B) Dilation, then reflec-
tion, then rotation.

Figure 8.5 Random linear transformation.

linear transformation; the big difference being when transforming a polygon as in our
examples, the angles between sides are not necessarily preserved and lengths of the sides
may not be scaled equally. For example, consider the linear transformation obtained
from the matrix M (see Fig. 8.5):
 
− 12 −2
M= 3
2 −2

The reflection and rotation transformations are examples of special transformations

known as orthogonal transformations. Orthogonal transformations are linear trans-
 Transformations and Fern Fractals 139

formations T : Rn → Rn such that

T (x) = x.

These transformations preserve the lengths of the vectors and the inner or dot products.
Thus in Euclidean vector spaces, the angles between vectors are preserved in addition
to the lengths of the vectors. This means that under these linear transformations, right
triangles transform to right triangles of the same size. Contraction and dilation trans-
formations also preserve angles, but the lengths of the vectors change so right triangles
transform to similar right triangles. Other transformations will transform right triangles
to triangles, but the transformed triangles are not necessarily right triangles and the sides
are not necessarily the same length or even same proportion.

8.2. Affine transformations

Affine transformations are not exactly linear transformations, but are similar in that
they can be written in terms of matrices and vectors.

Definition 8.2.1. Affine transformations are functions or mappings

from one vector space V to another vector space W of the form

T (x) = Mx + v

where M is a given matrix and v is a given vector.

One can think of affine transformations as linear transformations (Mx) then shifted
by the vector v. Similar to linear transformations, affine transformations preserve straight
lines: thus triangles remain triangles, quadrilaterals remain quadrilaterals, etc.
The example below shows examples of two affine transformations.

Example 8.2.1. The first affine transformation shown in Fig. 8.6 is

1   
3 0 3
x+ .
0 − 13 4

The second affine transformation shown in Fig. 8.6 is

 1  
1 4 3
x+ .
1 3 4
4 4
140 Programming Mathematics Using MATLAB®

Figure 8.6 Affine transformation examples.

8.3. Fern fractals

“Clouds are not spheres, mountains are not cones, coastlines are not circles, and bark is
not smooth, nor does lightning travel in a straight line” – Benoit Mandelbrot [16]
According to many stories, Mandelbrot started thinking about “fractals” when trying
to measure the coastline of Britain. He coined the term “fractal” from Latin “fractus”
meaning “broken” or “fractured”.
Fractals have two main distinctions:
1. Self-similarity (pattern repeats itself at magnification).
2. Non-integer dimension (Hausdorff dimension) which will not be discussed at this
time.
Fractals can involve many different areas of mathematics such as complex analysis, ge-
ometry, dynamical systems, linear algebra and the visualization of them is considered by
many to be beautiful.
Fern fractals are examples of iterated function systems (IFSs) and are generated
by affine transformations. We will be looking at the Barnsley fern fractal [2], [19]. The
basic premise of basic fern fractals is as follows:
• Four affine transformations each have a probability of being chosen.
1. The transformation with the highest probability generates successively smaller
leaflets.
2. Two transformations have equal probabilities. One of them generates the largest
left-hand leaflet, the other generates the largest right-hand leaflet.
3. The transformation with the smallest probability generates the stem.
 Transformations and Fern Fractals 141

• Start at the origin, one of the four transformations is chosen based on its probability.
The next point is calculated based on this transformation calculated on the current
point.
• Repeat this process with each new point calculated.
For a specific IFS, the image created will resemble ferns. In order to work with the
probabilities and chosen transformations, we will split up the interval [0, 1] based on
probabilities. For example, if there are three transformations, the first having probability
0.25, the second 0.5, and the third 0.25, then we will select which transformation is
chosen by using rand to generate a number between 0 and 1. If the number is between
0 and 0.25, the first transformation is selected. If the number is between 0.25 and 0.75,
the second transformation is selected. Lastly, if the number is between 0.75 and 1, the
third transformation is selected. Which transformations get the endpoints of 0.25 and
0.75 does not really matter.

8.4. Exercises
1. Set up
(a) We will define the RGB code for four colors are as follows:

Name RGB Code Variable name

My Green 0-104-87 mygreen
My Gray 200-200-200 mygray
My Gold 250-227-125 mygold
My Red 179-8-56 myred

Save the MyColors.mat file found in on the text website to your folder/direc-
tory and use the load('MyColors') command to load the vectors that can
be used as colors.
(b) One of the shapes we will use in these exercises will be an ellipse made with
100 points using the parametric equations x = 2 cos(t) and y = 3 sin(t) for an
appropriate domain for t. Set up the vectors of these values (variable names of
your choice!) to plot the ellipse in this exercise and in subsequent exercises.
To test, use the plot command to plot the ellipse in mygreen with a Line
Width of 2.
2. Visualizing linear transformations
For this exercise we will be visualizing some examples of linear transformations.
The process will be the same or similar for all of them, so copy/paste will be your
friend. Remember that using the plot(x,y) command, Matlab® “connects the
dots” of the points given with x and y as vectors of the x- and y-coordinates of the
points. We will use this idea to draw shapes and the transformed shapes. One of the
shapes will be the ellipse discussed in #1b. Note that once the x and y vectors for
142 Programming Mathematics Using MATLAB®

the ellipse are created in #1b, as long as those vector names are unique you do not
need to recreate them in subsequent exercises. The other shapes will be made using
your function myCircle2 that is a (fixed) copy of the function from Exercise 9 in
Chapter 6.
Our shapes will be plotted by forming a vector comprised of the x-values and an-
other comprised of the y-values of the points that make up our shapes. Each of our
transformations will be associated with a 2 × 2 matrix A. What the matrices should
be for reflections, rotations, scaling, and compositions are in the lecture/class notes.
To visualize the transformation associated with a matrix A, we want to calculate
Av for each of the points v (which is actually a column vector) that make up our
shapes and store the answer in a new set of points. But this would be tedious and/or
inefficient to do this for each point. Instead, we can do this all at once to all of the
points by using matrix calculations. This is where it is easiest to have the original
points in a 2 × n matrix V (n is determined by the number of points that make up
our shape). If we set up the matrix V so that the first row contains the x-values of
the points, and the second row contains the y-values, then each column of V is one
point. The beauty of matrix multiplication is that T = AV will be all of the points
that make up the transformed shape. For each of the linear transformations we will
do the following.
• Formulate the vectors x, y, and matrix V that are composed of the vertices or
points of the specified original shape. Plot the original shape in mygreen with
Line Width thicker than the default.
• Figure out what the appropriate matrix A would be for the transformation,
and display that matrix. In other words, when defining the matrix for each
transformation, do not suppress the output so we can see the values in the
matrix when it is published. It may be better to have a unique name for each
transformation matrix, instead of A for each problem.
• Calculate T = AV and plot the transformed shape in the stated color with Line
Width thicker than the default.
• Plot the two stated points (markers) of the original shape in mygreen, using the
“*” marker and square marker, respectively. Plot the two transformed markers
using the same shapes, but in the same color as the transformed shape. Thus
you should be able to see where the markers went under the transformation.
• Use the command axis equal and set the axes appropriately so the edges and
points of the shapes do not touch the edge of the figure. You may have to
experiment with the order of axis equal and the command(s) to set
the axes.
• Use a legend with the titles “Original” and “Transformed” for the 2 shapes,
and “1st marker” and “2nd marker” for the specified points. Note that the
order of your plot commands will affect the order the titles should be in the
 Transformations and Fern Fractals 143

legend! You may want to change the location of the legend with 'Location',
'Southeast' (or some other location specification) so you can see all of the
figures and legend. (This location may need to change for different transforma-
tions below.) For example,
legend('Sample title1', 'Sample title2', 'Location', 'Southeast').
Perform the above process for each of the following transformations and shapes.
Have the text that is in bold be the title of your figure (first example, title('
Reflection Example'))
(a) Reflection example with θ = π/6. Original shape: the ellipse. Transformed
shape in myred. First marker is the first point, the second marker is the 25th
point. Also draw the reflection line of the angle θ (easiest to use trig!) as a
black, dashed line.
(b) Scaling examples with c = 1.5 and c = 1/2. Original shape: a hexagon us-
ing your circle2 function. Transformed shapes are in mygray and mygold,
respectively. First marker is the first point, the second marker is the second
point. Plot all three shapes: the original and two transformed, on the same
figure.
(c) Rotation example with θ =?. Original shape: a pentagon. Notice that using
your circle2 function to create the pentagon makes the shape “off kilter”
or tipped. We want to orient the pentagon so the bottom side is perfectly
horizontal, so we need to rotate it by an appropriate angle θ . Figure out what
θ should equal and use this angle for your rotation example. Transformed
shape should be in myred. First marker is the first point, the second marker
is the second point.
(d) Two composition examples Original shape: a triangle using your
circle2. Transformed shapes are in myred and mygold, respectively. First
marker is the first point, the second marker is the second point. The first
transformation is a reflection about the y-axis (what is θ ?) followed by a
rotation with θ = π/4. The second transformation is the rotation and then
reflection. Plot all three shapes on the same figure.
(e) Two random transformations Original shape: the ellipse. Transformed
shapes are in mygray and mygold. First marker is the first point, the second
marker is the tenth point. The matrix for these random linear transformations
are 2 × 2 matrices with random integer entries from −5 to 5.
(f) Two random transformations Original shape: the square. Transformed
shapes are in mygray and mygold. First marker is the first point, the second
marker is the second point. The matrix for these random linear transforma-
tions are 2 × 2 matrices with random integer entries from −5 to 5.
3. Visualizing affine transformations For this problem we will be doing the fol-
lowing:
144 Programming Mathematics Using MATLAB®

• Save the affineData.mat file found on the text website to your folder/directory
and use the load('affineData') command to load the matrices and vectors
for the affine transformations (listed below).
• Draw the hexagon using your circle2 in black.
• Calculate the transformed hexagon using the affine transformations below (T1
for the first transformation, T2 for the second, etc.).
• Plot the transformed hexagons in mygreen, myred, and mygold, respectively.
• Use the command axis equal and define the axes to an appropriate view so
the hexagons do not touch the edge of the figure.
• Have a legend identifying the original, first transformation, etc.
• No axis labels are needed for these graphs.
• The title of the graph should be “Affine Transformations.”
Affine transformations Tk (x) = Mk x + vk , (k = 1, 2, 3) where
       
0 1 1.5 3 1 −1.25
M1 = , v1 = , M2 = , v2 = ,
3 1 0 0 1 −1.25
   
−3 0 −1
M3 = , v3 = .
2 3 1

(NOTE: your points/vectors x should be a column vectors for this notation).

4. Fern fractals We will create functions fern1, fern2, and fern3 that have as input
a positive integer n and optional input variables. The function should check for the
validity of n, if not, return an error. You may use your previously written functions
to help with this error check. Optional input variables will be plot specifiers (color,
marker size, marker type, etc.). The functions will use the affine transformations
defined by the iterated function system (IFS) below using n iterations to generate a
figure of a fern fractal. The output of the function will be the value from the toc
command. Each affine transformation Tk (k = 1, 2, 3, 4) involves the six parameters
a, b, c , d, e, f :
   
a b e
Tk (x) = Mk x + vk = x+
c d f
where each transformation has probability pk of being performed. The IFS for the
Barnsley fern [2, p. 86] we will be plotting is in Table 8.1.
The three functions will use slightly different algorithms. The first one is a basic
one in which you may find the algorithm in a basic linear algebra text [27] or on
Wikipedia for plotting these fractals in other languages.
(a) For fern1, here is the algorithm:
i. Use the command tic.
 Transformations and Fern Fractals 145

Table 8.1 IFS for Barnsley fern fractal.

T a b c d e f p
1 0 0 0 4 0 0 1
25 100
2 17 1 1
− 25 17 0 8 17
20 25 20 5 20
3 1 − 13 23 11 0 8 7
5 50 100 50 5 100
4 3
− 20 7 13 6 0 11 7
25 50 25 25 100

ii. Load the variables found in fernIFS.mat that can be found on the text
website to have the matrices, vectors, and probabilities that are needed to
do the following
  tasks.
0
iii. Let x = .
0
iv. If n is the only input, plot the point x as a point using “*” marker. Oth-
erwise, plot the point with the additional inputs using varargin.
v. Use the random number generator rand to select one of the affine trans-
formations Tk according to the given probabilities.
vi. Redefine x to be the new x = Tk (x) = Mk x + vk .
vii. Plot the new point x as you did with the original x.
viii. Repeat steps (vi) through (viii) so a total of n new points are plotted
(where should the hold on and hold off commands be? How many
times should your loop iterate?).
ix. The last command of the function should be storing toc into your output
variable to capture the time it took to generate and plot the fern fractal.
(b) For fern2, we will adjust the process of fern1 by vectorizing the code.
i. Use the command tic.
ii. Load the variables in fernIFS.mat found on the text website to have the
matrices, vectors, and probabilities that are needed to do the following
tasks.  
0
iii. Let x = and set X = x.
0
iv. Use the random number generator rand to create a vector of n random
numbers to determine all of the affine transformations.
v. Repeat the following so a total of n new points are in the matrix X:
A. Select one of the affine transformations Tk according to the given
probabilities using the appropriate element from the vector of ran-
dom numbers created above.
B. Redefine x = Tk (x) = Mk x + vk . Concatenate X with the new x (add
another column to X that is the new x).
146 Programming Mathematics Using MATLAB®

vi. If n is the only input, plot the points in the matrix X as a point using
“*” marker. Otherwise, plot the points with the additional inputs using
varargin. (Note that the first row of X contains the x-coordinates, the
second row contains the y-coordinates.)
vii. The last command of the function should be storing toc into your output
variable to capture the time it took to generate and plot the fern fractal.
(c) For fern3, we will adjust the process of fern2 by preallocation.
i. Use the command tic.
ii. Load the variables in fernIFS.mat found on the text website to have the
matrices, vectors, and probabilities that are needed to do the following
tasks.
iii. Define x as a matrix with two rows and n + 1 columns of zeros.
iv. Use the random number generator rand to create a vector of n random
numbers to determine all of the affine transformations.
v. For columns 2 through n + 1 of x, use the appropriate element from
the vector of random numbers created above to select one of the affine
transformations Tk according to the given probabilities and let that column
of x equal Tk (previous column of x).
vi. If n is the only input, plot the points in the matrix x as a point using
“*” marker. Otherwise, plot the points with the additional inputs using
varargin. (Note that the first row of x contains the x-coordinates, the
second row contains the y-coordinates.)
vii. The last command of the function should be storing toc into your output
variable to capture the time it took to generate and plot the fern fractal.
5. Display the fern using fern1, fern2, and fern3 with n = 2000. Before each fern
command, use the command clf, and after each fern command, use the commands
axis equal and axis off. For each of the ferns, use the sprintf command to
create titles that say which fern function, the value of n, and how much time elapsed
to create the fern fractal.
6. Display the fern using fern1, fern2, and fern3 with n = 5000, and specify the
square marker, color mygreen, and have the marker size be smaller than the default.
Before each fern command, use the command clf, and after each fern command,
use the commands axis equal and axis off. For each of the ferns, use the
sprintf command to create titles that say which fern function, the value of n,
and how much time elapsed to create the fern fractal.
7. Display the fern using fern2 and fern3 with n = 50,000 and specify a marker
and a color of your choice (do not use the default color). Use other specifiers if
you want. Before each fern command, use the command clf, and after each fern
command, use the commands axis equal and axis off. For each of the ferns,
use the sprintf command to create titles that say which fern function, the value
of n, and how much time elapsed to create the fern fractal.
CHAPTER 9

Complex Numbers and Fractals

9.1. Complex numbers
Complex numbers (C) are actually points in R2 with additional structure. The point
(a, b) in R2 would be the point z = a + bi in C. The a is the real part or z, and b is
the imaginary part of z. The x-axis is now called the real axis, and the y-axis is now
called the imaginary axis (see Fig. 9.1).

9.1.1 Adding complex numbers

As in R2 , one can add and subtract these numbers “element-wise”. The result is the
same as adding vectors (using triangle or parallelogram law) (see Fig. 9.2).

9.1.2 Multiplication by a real numbers (scalars)

Just as in vectors, multiplying a complex number by a scalar (real number) c changes the
norm/magnitude/length by a factor or c. Recall that multiplying by a negative number
makes the vector point in the opposite direction (see Fig. 9.3).

9.1.3 Multiplication and de Moivre’s theorem/formula

The biggest difference between vectors in R2 and complex numbers is the ability to
multiply. We learned that there is no way to multiply vectors. But we can multiply
complex numbers. If z = a + bi and w = c + di, then to find zw we just “foil it out” and

Figure 9.1 Complex numbers.

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00015-6 All rights reserved. 147
148 Programming Mathematics Using MATLAB®

Figure 9.2 Adding complex numbers.

Figure 9.3 Multiplying complex numbers by a scalar.

remember that i2 = −1 to get

zw = (ac − bd) + i(ad + bc ).

This still does not show us how to understand the importance of using complex
numbers for multiplication rather than vectors. Representing complex numbers in a
different way shows us the meaning and use of multiplication.
 Complex Numbers and Fractals 149

As we learned in Calculus II, there are two ways one can identify points on the
2D plane: Cartesian coordinates (x, y) and polar coordinates (r , θ ) using the conversion
formulas:

x = r cos θ,
y = r sin θ,
x + y2 = r 2 ,
2

y
= tan θ.
x
With this idea, using z = x + iy one has the polar representation of a complex
number:
z = x + iy = r cos θ + ir sin θ

where r is the length, norm, or magnitude of z (r = |z| = x2 + y2 ) and θ is the
argument of z (the angle of rotation from the positive real-axis to z).

Using trigonometry and polar representations, we can visualize what happens when
we multiply complex numbers. Using trigonometric identities, one can prove de
Moivre’s formula/theorem:

Theorem 9.1.1 (de Moivre’s theorem). When z = r cos θ + ir sin θ ,

and n is any natural number,

zn = r n cos(nθ ) + ir n sin(nθ ).

This means that when we take a complex number and raise it to a positive integer
power n, the result is a complex number in which the length is taken to that power n
and the argument is multiplied n times (thus rotated n times).
√
Consider the graph below. For the number z = 2 + 2i, we have r = |z| = 2 2, and
π π √
θ = . Thus z2 has length r 2 = 8, and argument 2θ = . Also, z3 has length r 3 = 8 8,
4 4
3π
and argument 3θ = (see Fig. 9.4).
4
If we let z = r cos θ + i sin θ and w = s cos ψ + is sin ψ , then using trigonometric iden-
tities we get
zw = rs cos(θ + ψ) + irs sin(θ + ψ).
Thus the lengths are multiplied and the arguments (angles of rotation) are
added.
150 Programming Mathematics Using MATLAB®

Figure 9.4 Powers of complex numbers.

Figure 9.5 Multiplying complex numbers.

√
Consider
√
the complex numbers pictured in Fig. √ 9.5: z1 = 2 cos(π/4) + i2 sin(π/4) =
2 + i 2 and z2 = 6 cos(π/3) + i6 sin(π/3) = 3 + (3 3)i. If we multiply them using the
polar representation, we get
√ √ √ √
z3 =12 cos(7π/12) + i12 sin(7π/12). = (3 3 − 3 6) + i(3 6 + 3 2) ≈ −3.1058
+ 11.5911i.

9.1.4 Plotting complex numbers in MATLAB®

One nice thing about complex numbers is it makes plotting points “easier” in MAT-
LAB. Say we wanted to plot and/or connect the two points (2, 3) and (6, 5). As points
in R2 , we would have to do something like the following:
 Complex Numbers and Fractals 151

x1=[2;3]; x2=[6;5];
x=[x1 x2];
plot(x(1,:),x(2,:),'*')

Using complex numbers instead, the same picture could be created by

a=2+3i; b=6+5i;
plot([a b],'*')

Similarly, if we wanted to connect the points, we can do the following:

a=2+3i; b=6+5i;
plot([a b])

A word of caution: if you have a purely real number and want it plotted as a complex
number, then you either have to have it within a vector of other complex numbers or
plot the real and imaginary part separately. For example, x=[2,i]; plot(x,'∗') will
plot the 2 correctly, but if you have x=2; instead, or even x=2+0∗i, you will not get the
desired results.

9.1.5 Creating line segments with complex numbers

If we wanted to look at the line segment (vector) connecting points A(2, 3) and B(6, 5),
in multivariable calculus we learn that the coordinates of the position vector (vector
originating at the origin) will be the coordinates of B minus the coordinates of A.
In our example, it would have the coordinates (4, 2). If instead we take the complex
number version of the points a = 2 + 3i, and b = 6 + 5i, then if we look at b − a we get
b − a = 4 + 2i. This vector, whether it is written in R2 or in C, has the same length and
direction as the vector connecting point A to B, as shown in Fig. 9.6.
If we wanted to move that line segment to start at another point, say, 1 + i, the line
segment would be between the points 1 + i and b − a + 1 + i. If instead we wanted to
b−a
take the vector b − a and scale it by, say, 1/3, then you just create the vector (see
3
Fig. 9.7).
We could have this scaled vector start at a, or we could have this scaled vector start
at b (see Fig. 9.8).
We can take that scaled vector and rotate it. Recall above that multiplying two
complex numbers will add the arguments (angles of rotation in polar representation)
and their lengths will be multiplied (see Section 9.1.3). Suppose we were to take the
angle θ = − π2 and create a complex number z = cos θ + i sin θ . What would this number
z equal? The length of z is 1, so any point that we multiply z by will have the same
length, but the angle will be rotated by θ = − π2 . This is shown in Fig. 9.9.
152 Programming Mathematics Using MATLAB®

Figure 9.6 Line segment b − a.

Figure 9.7 Scaling and shifting. (A) Shifting the line segment, (B) Scaling a line segment.

Figure 9.8 Scaling and shifting again. (A) Scaled vector starting at a, (B) Scaled vector starting at b.
 Complex Numbers and Fractals 153

Figure 9.9 Rotating a scaled line segment.

Figure 9.10 The classic Chaos Game board.

9.2. The Chaos Game

The Chaos Game is an interesting way to generate fractals. The classic Chaos Game is
to use an equilateral triangle, and color each vertex a different color (e.g., red, yellow,
and blue) (see Fig. 9.10). Color a standard, six-sided die so two faces are red, two are
yellow, and two are blue.
Now choose a random point, called the seed. Classically it would be a point inside
the triangle, but it does not matter, because after several turns, the points will be inside
the triangle. Now roll the die. Move half the distance from the seed towards the vertex
with the same color as what was rolled. Mark that point. Roll again, from the point you
just marked, move half the distance towards the vertex with the same color as what was
rolled. Mark that point. Repeat. (See Fig. 9.11.)
154 Programming Mathematics Using MATLAB®

Figure 9.11 A few turns of the Chaos Game (with turns marked for demonstration).

Figure 9.12 The Koch Curve for n = 0, 1, 2, 3.

After many iterations, the resulting picture may surprise you!

Modifications of this game is to use different polygons, and different distances to-
wards the vertices to create different pictures.

9.3. Line replacement fractals

9.3.1 Snowflake fractals
The Koch Snowflake is a famous/familiar fractal. Several iterations of the Koch Curve
are shown in Fig. 9.12. The snowflake is formed by repeating the Koch Curve around
an equilateral triangle (see Fig. 9.13).
There are variations on this snowflake. One can have the base shape be a hexagon
instead of a triangle (see Fig. 9.14).
 Complex Numbers and Fractals 155

Figure 9.13 Several Iterations of the Koch Snowflake.

Figure 9.14 A variation of the snowflake.

Figure 9.15 Several iterations of the Gosper Curve.

9.3.2 Gosper Island

Gosper Island is a fractal that can be considered another variation of the Koch Snowflake.
It begins with a line segment and modifies the line segment at each iteration (see
Figs. 9.15 and 9.16).
156 Programming Mathematics Using MATLAB®

Figure 9.16 Another view of the Gosper Curve.

Figure 9.17 The Gosper Curve explained.

How one gets the line segments for the Gosper Curve is by taking the points (com-
plex numbers) a and b and calculating the points x1 and x2 and connecting the points
a, x1 , x2 , and b. The length of each subsequent line segment is the length of b√−7a . The
√ 
point x1 is rotated from the segment b − a by an angle of θ1 = tan−1 53 . The point x2
is rotated from the segment b − a by an angle of θ2 = θ1 − π (see Fig. 9.17). Then the
Gosper Curve is repeated around a hexagon Figs. 9.18 and 9.19).

9.4. Geometric series

Geometric series are useful in many ways. Geometric series will help us to calculate
things regarding snowflake fractals in the exercises.

Definition 9.4.1. A geometric series is a series of the form

∞

ar k = a + ar + ar 2 + ar 3 + · · · + ar n + · · · .
k=0

The following theorem is provided without proof and can be found in any calculus
book.
 Complex Numbers and Fractals 157

Figure 9.18 Several iterations of the Gosper Island.

Figure 9.19 Gosper Island, n = 6.

Theorem 9.4.1. The geometric series

∞

ar k = a + ar + ar 2 + ar 3 + · · · + ar n + · · ·
k=0

converges if and only if |r | < 1. In this case,

∞
 a
ar k = .
k=0
1−r
158 Programming Mathematics Using MATLAB®

9.5. Exercises
1. Chaos Game We will simulate the Chaos Game and use complex numbers. Our
seed will be a complex number generated by the rand command for the real and
imaginary values of the seed. You will create a function file called chaosGame that
has one or more inputs. The first input is a positive integer n that will determine
the number of “rolls of the die” there will be in the Chaos Game, or more inputs.
The additional inputs, which are optional, will be marker specifications for drawing
the points. The output of the function will be the time elapsed (generated by the
commands tic and toc) to play the game. Your function file will do the following.
(i) Capture the start time with tic.
(ii) Perform the error check on the function input n and use the error com-
mand with appropriate message.
(iii) Generate the three vertices of the equilateral triangle as cos θ + i sin θ for θ
from π/2 to 2π + π/2, with an increment of 2π/3. The triangle must be so
the bottom base is horizontal. Draw the triangle in a solid black line.
(iv) Store the default colors using the commands
colors = lines(8); colors(8,:)=[0,0,0];
(v) Draw the vertices using the “o” marker and color them using three of the
default colors stored in colors. To make them more visible, specify the
MarkerFaceColor and MarkerEdgeColor to be the appropriate color. You
may want the MarkerSize to be larger than the default. (Hint: a for loop
may be useful!)
(vi) Preallocate a vector to the appropriate size to contain the “seed” and the n
points generated by the n rolls of the die.
(vii) Generate the seed to be in the first element of this vector by using the rand
command. This seed should be a complex number. (How would we get a
random complex number using the rand command?) Note that this seed
point need not be inside the triangle.
(viii) Now play the Chaos game for n turns: Use randi to simulate a roll of
a standard die. The next element of your vector will be the point that is
halfway between the current element and the vertex chosen by the “roll of
the die”. Note that you must make it so each vertex has an equal chance of
being chosen by a roll of the die.
(ix) After all of the turns are completed (thus the vector has been filled with the
complex numbers obtained from the turns of the Chaos Game), either plot
the points in the vector as black periods ('k.') or use the varargin to plot
the points.
(x) The output of the function is the time elapsed as determined by toc.
 Complex Numbers and Fractals 159

(a) Play the game for n = 10 with no other input. Use axis off and axis
equal commands. Put a meaningful title on the figure that has the number
of iterations and time elapsed.
(b) Do the same as above but for n = 200, have the marker be “*” and set the
MarkerSize to 3.
(c) Do the same as above but for n = 30,000 and specifying a black “o” with
MarkerSize of 1.
2. Chaos Game 2 We will simulate a modified version of the Chaos game and use
complex numbers. We will use a square instead of a triangle and have eight “ver-
tices” that are the corners and midpoints of the sides. Our seed will be a complex
number generated by the rand command for the real and imaginary values of the
seed. You will create a function file called chaosGameSquare.m that has one or
more inputs. The first input is a positive integer n that will determine the number
of “rolls of the die” there will be in the Chaos Game. The additional inputs, which
are optional, will be marker specifications for drawing the points. The output of
the function will be the time elapsed (generated by the commands tic and toc) to
play the game. Your function file will do the following.
(i) Capture the start time with tic.
(ii) Perform the error check on the function input n and use the error com-
mand with appropriate message.
(iii) Store the vertices of the square in a vector: 1, 1 + i, i, −1 + i, −1, −1 − i, −i,
1 − i (you may want to add extra 1 at the end to make it easier to draw...)
Draw the square in a solid black line.
(iv) Store the default colors using the commands
colors = lines(8); colors(8,:)=[0,0,0];
(v) Draw the vertices using the “o” marker and color them using the de-
fault colors stored in colors. To make them more visible, specify the
MarkerFaceColor and MarkerEdgeColor to be the appropriate color. You
may want the MarkerSize to be larger than the default. (Hint: a for loop
may be useful!)
(vi) Preallocate a vector to the appropriate size to contain the “seed” and the n
points generated by the game.
(vii) Generate the seed to be in the first element of this vector by using the rand
command. This seed should be a complex number. (How would we get a
random complex number using the rand command?)
(viii) Generate n rolls of an eight-sided die in a vector rolls.
(ix) Now play the Chaos Game for n turns: Using the rolls vector to determine
the chosen vertex, the next element of your vector of points will be gone
from the previous element to 1/3 of the distance between the previous
element and the chosen vertex.
160 Programming Mathematics Using MATLAB®

(x) After all of the turns are completed (thus the vector has been filled with the
complex numbers obtained from the turns of the Chaos Game), either plot
the points in the vector as periods ('.') or use the varargin to plot the
points.
(xi) The output of the function is the time elapsed as determined by toc.
Once the function is created, play the game as follows.
(a) Play the game for n = 50 with no other inputs. Use axis off and axis
equal commands. Put a meaningful title on the figure that has the number
of iterations and time elapsed (using sprintf).
(b) Do the same as above but for n = 1000.
(c) Do the same as above but for n = 50,000 and specifying a black period with
MarkerSize of 2.
3. Chaos Game 3 We will simulate a modified version of the Chaos game and use
complex numbers. We will use a pentagon instead of a triangle. The function will
be chaosGamePent. We will do as in Exercise 2 except for the initial shape and
vertices, the die will be a five-sided die, and in step 2ix, the next element will be the
point that is 3/8 the distance between the current element and the vertex chosen
by the “roll of the die.” Note that the definition of the pentagon and its vertices
should be such that it lies flat on the bottom of the figure. This can be done with
the equation sin θ + i cos θ for θ = k(2π/5) for k = 1, 2, . . . , 5.
(a) Play the game for n = 100 with no other inputs. Use axis off and axis
equal commands. Put a meaningful title on the figure that has the number
of iterations and time elapsed (using sprintf).
(b) Do the same as above but for n = 1000.
(c) Do the same as above but for n = 30,000 and specifying a period with Mark-
erSize of 1.
4. Chaos Game 4 We will simulate a modified version of the Chaos game and use
complex numbers. We will use a hexagon instead of a triangle. The function will be
chaosGameHex. We will do as in Exercise 2 except for the initial shape and vertices,
the die will be a standard die, and in step 2ix, the next element will be the point
that is 1/3 the distance between the current element and the vertex chosen by the
“roll of the die.” Note that the definition of the hexagon and its vertices should be
such that it lies flat on the bottom of the figure. This can be done with the equation
cos θ + i sin θ for θ = k(π/3) for k = 1, 2, . . . , 6.
(a) Play the game for n = 100 with no other inputs. Use axis off and axis
equal commands. Put a meaningful title on the figure that has the number
of iterations and time elapsed (using sprintf).
(b) Do the same as above but for n = 1000.
(c) Do the same as above but for n = 30,000 and specifying a period with Mark-
erSize of 1.
 Complex Numbers and Fractals 161

5. Koch Snowflake We will create a function kochSnowflake that will have a nested
function kochpoints that will draw the Koch Snowflake fractal at the nth iter-
ation. The input for kochSnowflake will be a nonnegative integer n (return an
appropriate error message if n is not a nonnegative integer). The output will be the
time elapsed using the tic and toc commands.
Pseudocode for function kochSnowflake
• Capture the start time.
• Check that n is a nonnegative integer; if not, give appropriate error message
using the error command.
• Create a vector v of vertices of the equilateral triangle; calculated by cos(kθ ) +
i sin(kθ ) for k = 1, . . . 4 (so they connect) for the appropriate θ or by using your
mycircle2 function. The triangle must be so the bottom base is horizontal.
• If n = 0, plot the triangle. Calculate the time elapsed and use the return com-
mand.
• Initialize a vector S to be the first vertex of the triangle (stored in v) and create
a loop that will use the nested function kochpoints on the last element of S,
the next vertex of the triangle (stored in v) for n iterations. If we looped through
using the nested function kochpoints on each of the vertices of the hexagon,
there would be repeated points. Thus for each of the three sides of the triangle,
we replace S with S minus the last element and concatenate it with the points
from the nested function kochpoints using the last element of S and the next
vertex (stored in v) using the given n.
• Plot the sets of vertices S.
• Capture the time elapsed for the output of the function.
Pseudocode for nested function kochpoints: has inputs a, b, and n and output
vector V .
• Define the output V to have the elements a and b.
• Repeat the following n times:
– Define an empty vector B.
– From 1 to [(number of elements in V ) − 1]
∗ let a = the current element of V and b = the next element of V .
∗ Calculate x1 to be the point that is 1/3 of the way from a towards b.
∗ Calculate x2 to be the point that is the midpoint between a and b MINUS
√
3/2i(b − a)/3.
∗ Calculate x3 to be the point 2/3 of the way from a towards b.
(OR USE ANY DERIVATIONS FOR x1, x2, and x3 we did in class or
you come up with on your own.) As long as it gives the proper picture, it
is fine. Keep in mind that your derivation should have as few calculations
as possible in order to speed up the code. To speed up the code even
more, if you are using the same calculations in several of these formulas,
162 Programming Mathematics Using MATLAB®

perform the calculation first, storing it as some variable, and then use the
variable within the calculations.
∗ Concatenate B with the elements a, x1, x2, and x3.
– Now we redefine V . The new V will be B concatenated with the last ele-
ment of the current V .
Once the function is created, we will visualize the snowflake as follows.
(a) Use subplot to have two rows, two columns of the snowflake plotted for
n = 0, 1, 2, 3 (capture the time elapsed for each). Hint: it may be easiest to
have a for loop for this.
• For each snowflake/subplot, capture the time elapsed.
• For each snowflake/subplot, use the axis equal and axis off com-
mands.
• For each snowflake/subplot, create a title with “n = , y s” where y is
the time elapsed to create that subplot. For example, you may see a title
such as “n = 3, 0.3456 s” because the third iteration of the snowflake
took 0.3456 s (made up numbers).
(b) Create another plot (NOT A SUBPLOT) that for n = 5 with a similar title
as above that gives the number of iterations and time elapses, using axis
equal and axis off.
(c) Calculations For the snowflake, calculate the following, where iteration
n = 0 is the triangle. Show all work on paper (turned in).
i. Create a chart that has for each iteration n the number of sides/line seg-
ments, the number of new triangles that replaces part of the line segment
at that iteration (this √will be 0 for n = 0), the length of the side of the
triangle at this stage ( 3 for n = 0), and the area of the triangle. Do this
for n = 0, n = 1, n = 2, n = 3, and n = k for some k ≥ 1.
ii. The area An enclosed by the snowflake at iteration n. Also find the limit
A = lim An , if it exists.
n→∞
iii. The perimeter Pn of the snowflake at iteration n. Also find the limit
P = lim Pn , if it exists.
n→∞
6. Fractal Snowflake We will create a function snowflake that will have a nested
function snowflakepoints that will draw the Snowflake fractal at the nth it-
eration. The input for snowflake will be a nonnegative integer n (return an
appropriate error message if n is not a nonnegative integer). Any additional in-
puts will be plot specifications. The output will be the time elapsed using the tic
and toc commands.
Pseudocode for function snowflake:
• Capture the start time.
• Check that n is a nonnegative integer; if not, give appropriate error message
using the error command.
 Complex Numbers and Fractals 163

• Create a vector v of vertices of a hexagon; calculated by cos(kθ ) + i sin(kθ ) for

k = 1, . . . , 7 (so they connect) for the appropriate θ or by using your mycircle2
function. The hexagon must be so the bottom base is horizontal.
• If n = 0, plot the hexagon. Calculate the time elapsed and use the return
command to end the function.
• Initialize a vector S to be the first vertex of the hexagon (stored in v) and
create a loop that will use the nested function snowflakepoints on the last
element of S and the current vertex of the hexagon stored in v for n iterations.
If we looped through using the nested function snowflakepoints on each
of the vertices of the hexagon, there would be repeated points. Thus for each
of the sides of the hexagon, we replace S with S minus the last element and
concatenate it with the points from the nested function snowflakepoints
using the last element of S and the next vertex (stored in v) using the given n.
• Plot the sets of vertices S, using varargin for the optional inputs if needed.
• Capture the time elapsed for the output of the function.
Pseudocode for nested function snowflakepoints: has inputs a, b, and n and
output vector V .
• Define the output V to have the elements a and b.
• Repeat the following n times:
– Define an empty vector B.
– From 1 to [(number of elements in V ) − 1]
∗ Let a = the current element of V and b = the next element of V .
∗ Calculate x1 to be the point that is 1/3 of the way from a towards b.
∗  Calculate x2 to be the point that is the (midpoint between a and b) PLUS
√ 
i 3 (b − a)
.
2 3
∗ Calculate x3 to be the point 2/3 of the way from a towards b.
(OR USE ANY DERIVATIONS FOR x1, x2, and x3 we did in class or
you come up with on your own.) As long as it gives the proper picture, it
is fine. Keep in mind that your derivation should have as few calculations
as possible in order to speed up the code. To speed up the code even
more, if you are using the same calculations in several of these formulas,
perform the calculation first, storing it as some variable, and then use the
variable within the calculations.
∗ Concatenate B with the elements a, x1, x2, and x3.
– Now we redefine V . The new V will be B concatenated with the last ele-
ment of the current V .
Once the function is created, we will visualize the snowflake as follows.
164 Programming Mathematics Using MATLAB®

(a) Use subplot to have 2 rows, 2 columns of the snowflake plotted for n =
0, 1, 2, 3 (capture the time elapsed for each). Hint: it may be easiest to have
a for loop for this.
• For each snowflake/subplot, capture the time elapsed.
• For each snowflake/subplot, use the axis equal and axis off com-
mands.
• For each snowflake/subplot, create a title with “n = , y s” where y is
the time elapsed to create that subplot. For example, you may see a title
such as “n = 3, 0.3456 s” because the third iteration of the snowflake
took 0.3456 seconds (made up numbers).
(b) Create another plot (NOT A SUBPLOT) that for n = 5 and optional color
“Deep Sky Blue” that has rgb(0, 191, 255) (remember how you need to
convert to a vector for MATLAB to recognize the RGB color). Have a
similar title as above that gives the number of iterations and time elapses,
and also using axis equal and axis off. Set the background color to
white using set(gcf,'Color','w').
(c) EXTRA CREDIT: figure out how to make an animated gif of the iterations
looping from n = 0 to n = 5 (no title).
7. Gosper Island We will create a function gosper that will have a nested function
gosperpoints that will draw the Gosper Island fractal at the nth iteration. The in-
put for gosper will be a nonnegative integer n (return an appropriate error message
if n is not a nonnegative integer). The output will be the time elapsed using the tic
and toc commands.
Pseudocode for gosper.
• Capture start time.
• Check that n is a nonnegative integer; if not, give appropriate error message.
• Create a vector v of vertices of a hexagon; calculated by cos(kθ ) + i sin(kθ ) for
k = 1, . . . 7 (so they connect) for the appropriate θ or by using your mycircle2
function.
• Initialize a vector G to be an empty vector.
• For each of the six sides of the hexagon (sides between points v(k) and v(k + 1),
use the nested function gosperpointsLO, to get the new vertices.
• Plot the sets of vertices G.
• Capture the time elapsed.
Pseudocode for gosperpoints
√
: has inputs a, b, and n and output V .
−1
• Let θ1 = tan ( 3/5), and θ2 = θ1 − π .
• Calculate the two rotations r1 = cos(θ1 ) + i sin(θ1 ), and r2 = cos(θ2 ) + i sin(θ2 ).
• Define the output V to have the elements a and b.
• Repeat the following n times:
– Define an empty vector B;
 Complex Numbers and Fractals 165

– From 1 to (number of elements in V − 1)

∗ let a = the current element of V and b = the next element of V .
∗ Calculate x = b√−7a .
∗ Concatenate B with the elements a, r1 x + a, r2 x + b.
– Redefine V to be B concatenated with the last element of the current V .
Once the function is created, we will visualize Gosper Island as follows.
(a) Use subplot to have two rows, three columns of the Gosper Island plotted
for n = 0, 1, 2, 3, 4, and 5 (capture the time elapsed for each).
(b) For each subplot, use the axis equal and axis off commands.
(c) For each subplot, create a title with “n = x, y s” where x is the value of n
for that subplot, and y is the time elapsed to create that subplot.
(d) Calculations For the Gosper Island, calculate the following (stage n = 0 is
the hexagon). Show all work on paper (turned in).
i. The number of points pn that make up one curve at each state n (do not
count repeats that may be in your code).
ii. The island is made up of six curves connected. Find the number of points
xn that make up the entire island at stage n (do not count repeats that may
be in your code).
iii. The area An inside the island at stage n. Also find the limit A =
limn→∞ An , if it exists.
iv. The perimeter Pn of the island at stage n. Also find the limit P =
limn→∞ Pn , if it exists.
CHAPTER 10

Series and Taylor Polynomials

10.1. Review of series
The material will be more understandable after a review of sequences and series from
calculus, for example Chapter 11 from [24]. Recall from Section 9.4 that a geometric
series is a series of the form
∞
 ∞

rn = r n−1
n=0 n=1
1
and that a geometric series of this form converges to when |r | < 1 and is divergent
1−r
otherwise.
From now on, the subscripts and superscripts on the summation notation will be
 ∞

left off. In other words, the notation an will stand for an .
n=0
Another useful type of series is the alternating series.
Definition 10.1.1. An alternating series is a series of the form
 
(−1)n bn or (−1)n+1 bn

where bn ≥ 0 for all n.

Example 10.1.1. A famous example of an alternating series is the alternating har-
monic series:
∞
 1 1 1 1
(−1)n+1 = 1 − + − + ···
n=1
n 2 3 4

Theorem 10.1.1 (Alternating

 series estimation theorem). Consider
n
an alternating series (−1) bn that satisfies
(a) 0 ≤ bn+1 ≤ bn for all n, and
(b) lim bn = 0
n→∞
Then the alternating series converges. In addition, if we say
 
n
s= (−1)n bn and sn = (−1)k bk , then the remainder Rn = s − sn
k=0
is estimated by
|Rn | = |s − sn | ≤ bn+1 .

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00016-8 All rights reserved. 167
168 Programming Mathematics Using MATLAB®

The proof of this is not given here. Notice that many times (and indeed, even in
lecture), this is split up into two theorems; the first with the statement giving the suffi-
cient conditions for an alternating series to converge, and the second has the statement
about the remainder. For brevity’s sake it is combined here into one theorem.

Example 10.1.2. The alternating harmonic series converges by the alternating series
estimation theorem, as bn = 1n is a decreasing sequence that converges to 0. In addition,
if we looked at
1 1 1 1 47
s5 = 1 − + − + = = 0.783̄,
2 3 4 5 60
then |R5 | ≤ b6 = 16 . Thus the actual sum s of the series is somewhere in the interval

47 1 37 57
± , or between and , or between 0.616̄ and 0.95.
60 6 60 60
 
Recall that a series an is absolutely convergent if |an | converges. An impor-
tant theorem tells us that any absolutely convergent series converges, but not necessarily
vice versa. This gives us the term conditionally convergent; a series is conditionally
 
convergent if an converges but |an | diverges.

Example 10.1.3. The alternating harmonic series is conditionally convergent.

Two useful tests for convergence are the ratio and root tests. There are, of course,
others but for now we will remind you of these two without proof.

Theorem 10.1.2 (Ratio test). Consider the series an and let
 
 an+1 
lim   = L.
n→∞ a  n

(a) If L < 1, the series an is absolutely convergent.

(b) If L > 1 (including L = ∞), the series an diverges.
(c) If L = 1, the ratio test is not applicable.


Theorem 10.1.3 (Root test). Consider the series an and let

lim n
|an | = L .
n→∞

(a) If L < 1, the series an is absolutely convergent.

(b) If L > 1 (including L = ∞), the series an diverges.
(c) If L = 1, the ratio test is not applicable.
 Series and Taylor Polynomials 169

∞
 (−1)n
Example 10.1.4. Consider the series .
n=0
n!
We have
 
 an+1  1
lim   = lim (n+1)!
n→∞  a  n n→∞ 1
n!
n!
= lim
n→∞ (n + 1)!
1
= lim = 0.
n→∞ n+1

Since this limit equals zero, which is certainly less than one, the series is absolutely
convergent by the ratio test.

All of the above are useful, but they are really used in the context of power series, and
in particular, Taylor series.

10.2. Power series

Definition 10.2.1. Recall the definition of a power series centered

at x0 ∈ R:
∞

an (x − x0 )n = a0 + a1 (x − x0 ) + a2 (x − x0 )2 + a3 (x − x0 )3 + · · ·
n=0

The real numbers a0 , a1 , a2 , . . . are called the coefficients of the

power series.

The main questions about power series are: “For what values of x does the series
converge?” and “What does the series equal?” Knowing these answers allow us to see
why we care about power series. In other words, when does the infinite sum make
sense, and given a value of x, when does it equal some real number, even if it is difficult
to find that exact real number? Notice that every power series converges at x = x0 . But
if that were the only case, then power series would not be useful at all. So we want
power series that converge for additional values of x.
The goal is to REPRESENT A FUNCTION, ESPECIALLY A COMPLICATED
FUNCTION, AS A POWER SERIES. This is tremendously useful for approximations
and calculations and many other results beyond the scope of this course.
170 Programming Mathematics Using MATLAB®

∞

Theorem 10.2.1. Consider the power series an (x − x0 )n .
n=0
(a) If the power series converges at c = x0 , then it converges for all
x such that |x − x0 | < |c − x0 |.
(b) If the power series diverges at d ∈ R, it diverges for all x such
that |x − x0 | > |d − x0 |.

Proof. Suppose there exists c = x0 such that the power series converges. We will show
that, for every x such that |x − x0 | < |c − x0 |, the series converges.

Let x ∈ R such that |x − x0 | ≤ r < |c − x0 |. We know that ∞ 0 an (c − x0 ) is convergent,
n

so
lim an (c − x0 )n = 0.
n→∞

In particular, there is a constant M such that |an (c − x0 )n | ≤ M for all n. Thus

 n  n
|x − x0 | r
|an | |x − x0 | = |an | |x − x0 |
n n
≤M .
|c − x0 | |c − x0 |
  n
Since |c−rx0 | < 1, we have then that M |c−rx0 | is a convergent geometric series, so

we have an (x − x0 )n is absolutely convergent.
Now suppose that there exist d ∈ R such that the power series diverges. Suppose
there exists w ∈ R with |d − x0 | < |w − x0 | and the power series converges at x = w. By
above, the series must converge at d, which is a contradiction. Thus there cannot exist
any such w.

The above theorem is useful, but is mostly used to prove the following theorem.

∞

Theorem 10.2.2. For a given power series an (x − x0 )n , there are
n=0
only three possibilities:
(a) The series converges only when x = x0 .
(b) The series converges for all x ∈ R.
(c) There exist R > 0 such that the series converges if |x − a| < R
and diverges if |x − a| > R.

Proof. Consider Theorem 10.2.1. If no such c = x0 exists, we have case 1 of the theorem.
If no such d exists, we have case 2 of the theorem.
Now suppose we have both a point c = x0 where the series converges and a point d
where the series diverges. Let C = |c − x0 | and D = |d − x0 | and consider the intervals
IC = (x0 − C , x0 + C ) and ID = [x0 − D, x0 + D]. (Draw a picture!)
 Series and Taylor Polynomials 171

Let S = {x ∈ R | series converges}. This set is non-empty since we have all x ∈ S for
all x ∈ IC by Theorem 10.2.1. By the same theorem, x ∈/ S for all x ∈/ ID . Thus x0 + D is
an upper bound for S since for all x > x0 + D, the series diverges so is not in S. By the
completeness axiom of real numbers, S has a “least upper bound”, which we will call
R. We know that R > 0. This tells us if |x − x0 | < R, x ∈ S, and if |x − x0 | > R, x ∈/ S,
and the result is proven.
In the third case, what happens when |x − a| = R depends on the series and you
have to check. In other words, some series converge at the endpoints, others diverge,
and others converge at one endpoint and diverge at the other.
The number R is called the radius of convergence and in the first case of Theo-
rem 10.2.2, we say R = 0. In the second case of Theorem 10.2.2 we say R = ∞. In the
third case, you have four possibilities for the interval of convergence:

(x0 − R, x0 + R), [x0 − R, x0 + R], (x0 − R, x0 + R], [x0 − R, x0 + R).

This theorem is important because it tells us that a power series will never only
converge at a few points, or only on the integers, etc.
 (−1)n xn
Example 10.2.1. Consider the power series . We see that
n!
  |x|n+1
 an+1 
lim   = lim (n+1)!
n→∞ a  n n→∞ |x|n
n!
n!|x|
= lim
n→∞ (n + 1)!
|x|
= lim = 0.
n→∞ n + 1

Thus the series is absolutely convergent for any x ∈ R by the ratio test. Thus the radius
of convergence is R = ∞ and the interval of convergence is I = (−∞, ∞).

Example 10.2.2. Compare the radii and intervals of convergence of the following
series. The work is shown for one of the series. Can you see how to get the answers for
the others?
∞

(a) (x − 7)n . R = 1, I = (6, 8).
n=0
∞
 (x − 7)n
(b) , R = 1, I = [6, 8).
n=1
n
  |x−7|n+1
 an+1 
lim   = lim n+1
n→∞ a  n n→∞ |x−7|n
n
172 Programming Mathematics Using MATLAB®

n|x − 7|
= lim = |x − 7|.
n→∞ n + 1

By the ratio test, the power series converges when |x − 7| < 1, and diverges when
|x − 7| > 1 and so R = 1. But when |x − 7| = 1, i.e., when x = 6 and x = 8, the
ratio test is not applicable so we look at each of these cases separately. When x =
 (x − 7)n  (−1)n
6, the power series = which is the alternating harmonic
n n
series, which converges. When x = 8, the power series is the harmonic series,
which diverges. Thus I = [6, 8).
∞
(x − 7)n
(c) , R = 1, I = [6, 8].
n=1
n2
∞

(d) n!(x − 7)n , R = 0, I = {7}.
n=0
∞
 (x − 7)n
(e) , R = ∞, I = (−∞, ∞).
n=0
n!

Definition 10.2.2. A function f is analytic at x = a if there exists a

power series expansion centered at a with radius of convergence R > 0
that converges to f (x) for x ∈ (a − R, a + R).

The next theorem is very useful because it allows us to get series for more functions,
and possibly use Taylor polynomials (see Section 10.3 below) for approximating these
functions.
Theorem 10.2.3. If f (x) is analytic (can be represented as
a power

series) with f (x) = an (x − a)n and R > 0, then f  (x) and f (x) dx
can also be represented as a power series with radius of convergence
also R. In addition, for all x ∈ (a − R, a + R),
∞

f  (x) = nan (x − a)n−1 = a1 + 2a2 (x − a) + 3a3 (x − a)2 + · · ·
n=1

and

∞
 (x − a)n+1 (x − a)2
f (x) dx = C + an = C + a0 (x − a) + a1 + ···
n=0
n+1 2

Note that this theorem says

d   d
an (x − a)n = an (x − a)n
dx dx
 Series and Taylor Polynomials 173

and

 

an (x − a) dx =
n
an (x − a)n dx.

IMPORTANT: the theorem tells us the radius of convergence stays the same, yet the
actual interval of convergence may change as demonstrated in the following example.
∞
 (x − 7)n
Example 10.2.3. Consider the series from Example 10.2.2 above. We
n=1
n
∞
 (x − 7)n
got R = 1, I = [6, 8). Let f (x) = . Theorem 10.2.3 tells us that, for all x ∈
n=1
n
(6, 8), the derivative exists and is

∞
 ∞
 ∞

n(x − 7)n−1
f  (x) = = (x − 7)n−1 = (x − 7)n .
n=1
n n=1 n=0

Thus R = 1 for the series for f  (x). Upon further investigation, we see that the interval
of convergence for this series for f  (x) is now (6, 8), which is different from the interval
of convergence for f (x).

10.3. Taylor polynomials and Taylor series

The rest of the theorems in this chapter will not be proven. Proofs can be found in
calculus and/or analysis texts.

Definition 10.3.1. For a given function f (x) and a ∈ R, the Taylor

series of f (x) about x = a is defined as:
∞
 f (k) (a) f  (a)
(x − a)k = f (a) + f  (a)(x − a) + (x − a)2 + · · ·
k=0
k! 2!

Here we are assuming that f is infinitely differentiable at a. We are also

using the convention 0! = 1, and f (0) (x) = f (x).

Taylor series are useful in their own right, but are most useful when looking at the
partial sums of the series, known as Taylor polynomials.
174 Programming Mathematics Using MATLAB®

Figure 10.1 Taylor polynomials for y = ex .

Definition 10.3.2. For a given function f (x) and a ∈ R, the Taylor

polynomial of degree n about x = a is defined as

n
f (k) (a)
Tn (x) = (x − a)k
k=0
k!
f  (a) f (n) (a)
= f (a) + f  (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!
Here we are assuming that f is differentiable n times at a. Then we
let Rn (x) = f (x) − Tn (x) be the remainder of the Taylor polyno-
mial. Taylor series with center a = 0 are commonly called Maclaurin
series.

Example 10.3.1. The Maclaurin series for f (x) = ex becomes

 xn
ex =
n!

since f (n) (0) = 1 for all n. Fig. 10.1 above shows how Taylor polynomials are decent
approximations for the exponential function near x = 0, and the higher degree of Taylor
polynomial used, the better this approximation becomes, even for values further away
from x = 0.

The next theorem relates general power series to Taylor series.

 Series and Taylor Polynomials 175

Theorem 10.3.1 (Taylor series theorem). Suppose that a power series


an (x − a)n has a radius of convergence 0 < R ≤ ∞ (R > 0). Then,
the function f (x) that this series represents is infinitely differentiable
on (a − R, a + R) and its coefficients an are given by

f (n) (a)
an = , n = 0, 1, 2, . . .
n!

Theorem 10.3.2. Using the notation from above, if

f (x) = Tn (x) + Rn (x), and lim Rn (x) = 0

n→∞

for |x − a| < R, then f is equal to the sum of the Taylor series:

∞
 f (k) (a)
f (x) = (x − a)k
k=0
k!
f  (a) f (n) (a)
= f (a) + f  (a)(x − a) + (x − a)2 + · · · + (x − a)n + · · · .
2! n!
The key is to get an estimate of Rn (x) in order for these series to be useful.

Theorem 10.3.3 (Taylor’s theorem). Suppose that, for any n ∈ N, f (x)

has n + 1 derivatives in a neighborhood of x = a. If

n
f (k) (a)
Tn (x) = (x − a)k
k=0
k!
f  (a) f (n) (a)
= f (a) + f  (a)(x − a) + (x − a)2 + · · · + (x − a)n
2! n!
then f (x) = Tn (x) + Rn (x), where

f (n+1) (c )
Rn (x) = (x − a)n+1
(n + 1)!

for some c between x and a. If x = a, then c = a.

Note: this formula for the remainder is known as Lagrange’s form of the remainder.
There are other forms of the remainder. Here are two other famous forms:
f (n+1) (c )
Rn (x) = (x − c )n (x − a) Cauchy’s form,
n

!
1 x (n+1)
Rn (x) = f (t)(x − t)n dt Integral form.
n! a
176 Programming Mathematics Using MATLAB®

Here is another way one can estimate the error when using Taylor polynomials to
estimate the function represented by a Taylor series:
Theorem 10.3.4 (Taylor’s inequality). Suppose f (x) is represented as
a power series centered at a with R > 0. If |f (n+1) (x)| ≤ M for all x
with |x − a| ≤ δ , then the remainder Rn (x) of the Taylor series satisfies
M
|Rn (x)| ≤ |x − a|n+1 for |x − a| ≤ δ.
(n + 1)!

If you read the above theorems carefully, you may notice that it is NOT quite the
case that “f is analytic iff f is infinitely differentiable. Here is a counterexample:
Example 10.3.2. Center the series at a = 0 for f (x):

e−1/x x > 0,
f (x) =
0 x ≤ 0.

It can be shown (can you show it?) that, while f is infinitely differentiable, even at 0,
there is no power series (with R > 0) centered at 0 (look at Lagrange’s form of the
remainder for x > 0 close to 0).

The good news is that many functions that we are familiar with CAN be represented
by a Taylor or Maclaurin series.
Common Maclaurin series (look up others in your calculus book):
∞
 xn
ex = R = ∞,
n=0
n!
∞
 (−1)n x2n
cos x = R = ∞,
n=0
(2n)!
∞
 (−1)n x2n+1
sin x = R = ∞,
n=0
(2n + 1)!
 (−1)n x2n+1
tan−1 x = R = 1,
2n + 1
∞

1
= xn R = 1.
1−x n=0

One can use these series to get an estimate for e:


n
1
e≈ .
k!
k=0
 Series and Taylor Polynomials 177

The following examples show the usefulness of Theorem 10.2.3 when applied to
Maclaurin and Taylor series.

Example 10.3.3. Use Maclaurin series to prove the following statements.

d
(a) Prove ex = ex .
dx
d
(b) Prove that cos x is an even function and that cos x = − sin x.
dx
d
(c) Prove that sin x is an odd function and that sin x = cos x.
dx

Example 10.3.4. What is a Maclaurin series for ln(1 − x) centered at 0? What is its
radius of convergence?
∞
 1
Example 10.3.5. Consider the geometric series xn =, for |x| < 1.
n=0
1 − x
1
(a) Use the geometric series above to find a series for . What is the radius of
1 + x2
convergence?
(b) Use Theorem 10.2.3 to find a series for tan−1 x. What is the radius of conver-
gence? Interval of convergence?
∞
(−1)k π
(c) Prove the Leibniz series identity: = .
k=0
2k + 1 4

Example 10.3.6. Using the Maclaurin series for ln(1 + x), show that
∞
 (−1)k+1
ln 2 = .
k=1
k

10.4. Exercises
1. Create a function isnnInt that will have one input, n, and the output will be a
logical true/false as to whether n is a nonnegative integer (but not necessarily of
data type integer; data type double is still allowed). Note that no error messages
are displayed with this function; it returns either true or false.
2. This problem looks at the Leibniz series
∞

π (−1)k
= .
4 2k + 1
k=0

The nth partial sum of the series can be used to estimate π/4 and thus other values
with π (this is a crude estimation; other/better estimations are a topic for another
178 Programming Mathematics Using MATLAB®

course). Notice that it is an alternating series. Your leibniz function will have the
following features.
• The function has one input, n, which should be a nonnegative integer.
Use your function isnnInt to check it; if not, an appropriate error message is
displayed using the error command.
• The function will have two outputs. The first output is the nth partial sum of
the Leibniz series:

n
(−1)k π
≈ .
(2k + 1) 4
k=0

A loop can be used for this calculation but it may be fastest if you use vector-
ized code instead. The second output will be the estimated error of this partial
sum using the alternating series estimation theorem.
3. The next function nleibniz2 will not calculate partial sums, but instead tell you
what n should equal used in order to use the leibniz function to approximate π/4
to a certain degree of accuracy (“tolerance level”).
• The nleibniz2 function has one input: a “tolerance level” ε for using the partial
sum.
• Your function should check that ε > 0; if not, an appropriate error message is
displayed using the error command.
• The function figures out the minimum value of n (obtained from the alter-
nating series estimation theorem) for the nth partial sum to have an error less
than or equal to ε. The output of the function is this value n.
4. Use your function leibniz to estimate π and its error. (Careful! You need to
adjust your answers from the answer you get from your function for this!). Note
that even though MATLAB® has an accurate estimation of π , do not use this
to estimate the error. Use n = 3, n = 9, n = 99, and n = 9999. Use the fprintf
command to make your answers appear like below, with the same number of
decimal places shown (it may be more convenient to have more than one fprintf
command). Note that the numbers below are made up for display purposes.
Using the partial sum of the Leibniz series:
n Estimate Error estimate Between
3 3.123456789 1.123456789 2.00000000 and 4.24691358
9 3.143456789 0.123456789 3.02000000 and 3.26691358
99 3.141556789 0.012345678 3.12921111 and 3.15390247
9999 3.141596789 0.001234567 3.14036222 and 3.14283136

5. Using nleibniz2, what value of n should be used to get the estimate of π accurate
to six decimal places? Careful! What would the error tolerance need to equal to
be within that many decimal places, and how do you adjust between error for π/4
and error for π ?
 Series and Taylor Polynomials 179

6. This problem looks at the alternating harmonic series

∞
 (−1)k+1
= ln(2).
k=1
k

In calculus or analysis, using Taylor (Maclaurin) series on f (x) = ln(1 + x) it can be

shown that the Taylor series converges when x = 1 and you get the above equation.
Therefore, the nth partial sum of the series can be used to estimate ln(2). Your ln2
function will do the following:
• The function will have one input n which should be a positive integer. Check
for it using isnnInt; if not, an appropriate error message is displayed using
the error command.
• The function will have two outputs. The first output is the nth partial sum of
the series:
n
(−1)k+1
.
k=1
k
A loop can be used for this calculation but it may be fastest if you use vector-
ized code instead. The second output will be the estimated error of this partial
sum using the alternating series estimation theorem.
7. The next function nln2 will not calculate partial sums, but instead tell you the
minimum n should equal in order to use the ln2 function to approximate ln(2) to
a certain degree of accuracy (“tolerance level”).
• The nln2 function has one input: a “tolerance level” ε for using the partial
sum.
• Your function should check that ε > 0; if not, an appropriate error message is
displayed using the error command.
• The function figures out the minimum value of n (determined by the alter-
nating series estimation theorem) for the nth partial sum to have an error less
than or equal to ε. The output of the function is this value n.
8. Use your functions nln2 and ln2 to estimate ln(8) accurate to four decimal places.
Careful! What would the error tolerance need to equal to be within that many
decimal places, and how do you adjust the error for ln(2) to get the estimate for
ln(8)? Use the value of n you get from nln2 in your function ln2 to estimate ln(8).
9. This problem is looking at Taylor/Maclaurin series and Taylor polynomials for
f (x) = cos(x).
∞
(−1)k x2k
cos(x) =
(2k)!
k=0

Create a function called mycos.m that calculates the nth-degree Taylor polynomial
for f (x) = cos(x) using the above Maclaurin series for cos(x). The function should
have two inputs; x and n. Make sure that n is a nonnegative integer using your
180 Programming Mathematics Using MATLAB®

function isnnInt; if not, an appropriate error message is displayed using the error
command. The calculations should be such that x can be a number, vector, or
matrix. The input n is the highest degree term that will appear in the poly-
nomial. Thus if n = 8, the degree of the polynomial will be no bigger than 8.
Notice that if we enter n = 9, we should get the same as n = 8. Use the warning
command if an odd value of n is entered. Careful: if we want n = 8, what should k
equal for the last term in the sum? If we input n = 9, what should happen? Using
floor, ceil, or fix may be useful here. The output of the function will be the
calculated value of the Taylor polynomial Tn (x).
Plot the Taylor polynomials for x ∈ [−10, 10] for n = 4, n = j, n = k, and n = 24
using the above function mycos along with y = cos(x) on the same figure with the
following specifications.
• The values of j and k will be random integers between 6 and 12 determined
using the randi command. Use a while loop to make sure k = j.
• Have y = cos(x) dotted and in black, and have the others in different colors
and/or shapes (dotted, dashed, etc.). Use your own judgment on LineWidth,
etc. to make the graphs clear.
• Have an appropriate legend and title. (sprintf may be useful to put the values
of n in the legend!)
• The vertical axis should only be between −2 and 2.
10. This problem is looking at Taylor/Maclaurin series and Taylor polynomials for
f (x) = tan−1 (x).
∞
 ∞
(−1)(k−1) x2k−1 or  (−1)k x2k+1
tan−1 (x) = = , x ∈ [−1, 1].
k=1
2k − 1 k=0
2k + 1

Create a function called myatan.m that calculates the nth-degree Taylor polynomial
for f (x) = tan−1 (x) using the above Maclaurin series for tan−1 (x). The function
should have two inputs; x and n. Make sure that n is a nonnegative integer using
your function isnnInt; if not, an appropriate error message is displayed using the
error command). The calculations should be such that x can be a number, vector,
or matrix. The input n is the highest degree term that will appear in the
polynomial. Thus if n = 8, the degree of the polynomial will be no bigger than
8. Notice that if we enter n = 8, we should get the same as n = 7. Use the warning
command if an odd value of n is entered. Careful: if we want n = 7, what should k
equal for the last term in the sum? If we input n = 8, what should happen? Using
floor, ceil, or fix may be useful here. The output of the function will be the
calculated value of the Taylor polynomial Tn (x).
Plot the Taylor polynomials for x ∈ [−1.5, 1.5] for n = 3, 5, 13, and 52 along with
y = tan−1 (x) (atan) using the above function myatan on the same figure with the
following specifications.
 Series and Taylor Polynomials 181

• Have y = tan−1 (x) dotted and in black, and have the others in different colors
and/or shapes (dotted, dashed, etc.). Use your own judgment on LineWidth,
etc. to make the graphs clear.
• Have an appropriate legend and title. (sprintf may be useful to put the values
of n in the legend!)
• The vertical axis should only be between −2 and 2.
11. Taylor series and polynomials for g(x) = tan−1 (x3 ).
(a) Using the Maclaurin series for tan−1 (x), find the Maclaurin series for
tan−1 (x3 ) and its radius of convergence, showing any work needed on paper.
State the formula along with its radius of convergence.
(b) Integrate the series you get in part (a) for a > 0 to get a series for the fol-
lowing integral:

a
a
g(x) dx = tan−1 (x3 ) dx,
0 0
also find the radius of convergence for the series for this integral, showing
all work on paper. State the formula for the series along with its the radius
of convergence.
(c) Create a function called atanint.m that calculates the nth-degree Taylor
polynomial for the series in part (b) for a given a and a given n. The func-
tion should check that a > 0, and that n is a nonnegative integer (use your
function isnnInt); if not, appropriate error messages would be displayed us-
ing the error command. Just as in myatan above, the input n should be
the highest degree that will appear in the partial sum (Taylor polynomial).
Notice that the series is an alternating series. Using the alternating series
estimation theorem, figure out what the estimated error would be for the
partial sum with input n, and take that as the second output. The second
output will be given whether or not it is asked for (such as the MATLAB
command/function size).
(d) Use properties of integrals, the fact that g(x) is either an even or odd func-
tion, and your function atanint if necessary to estimate the following (use
format long for these answers). Show all work on paper.

1/4
i. tan−1 (x3 ) dx (use n = 5 and n = 20),

0 1/4
ii. tan−1 (x3 ) dx (use n = 5 and n = 20),

−31//44
iii. tan−1 (x3 ) dx (use n = 5 and n = 20).
−1/4
Make sure your answers are clearly labeled.
CHAPTER 11

Numerical Integration
11.1. Approximating integrals/numerical integration
Our goal is to discuss basic ways to approximate the integral
 b
f (x) dx.
a

The idea is that we view the integral as area “under the curve” (between the curve
and the x-axis, with negative area if the curve is below the x-axis). We approximate
the actual area (integral) by approximating the area with areas of regions that are easier
to calculate. The key for any of numerical integration techniques is to subdivide the
interval [a, b] into n subintervals. In this document we are making our lives easier by
dividing the interval into n subintervals of equal width:

b−a
x = .
n
Fancier methods would use different width subintervals depending on how the function
behaves on the domain, etc. Then n regions are formed based on the subdivisions and
the values of f (x), and the areas of these regions are calculated and totaled. The idea is
that, as n → ∞ or, equivalently, as the width x → 0, these approximations approach
 b
the value of f (x) dx (see Fig. 11.1).
a
The subdivision of the interval [a, b] is such that a = x0 and b = xn to get

a = x0 < x1 < x2 < · · · < xn−1 < xn = b.

The formula for the points xk is

xk = a + kx, k = 0, 1, . . . n.

11.2. Riemann sums


Riemann sums are used to approximate ab f (x) dx by using the areas of rectangles or
trapezoids for the approximating areas. Each rectangle/trapezoid has width x. How
we choose the height of the rectangles gives us different methods of approximation, and
there is also the trapezoidal method.

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00017-X All rights reserved. 183
184 Programming Mathematics Using MATLAB®

Figure 11.1 Approximating areas.

Figure 11.2 Sample points for approximating rectangles

For the kth subinterval [x , x ], we choose a sample point x∗k ∈ [xk−1 , xk ]. The height
 ∗  k−1 k
of the kth rectangle is f xk . Thus the general form of using rectangles for approximating
b
a f (x) dx is
 
n
b  ∗
f (x) dx ≈ f xk x.
a k=1

The three most common methods using rectangles are by using the left endpoint,
right endpoint, or midpoint of the subinterval to choose x∗k , and thus the height of the
rectangle (see Fig. 11.2 above).
 Numerical Integration 185

Figure 11.3 Trapezoidal rule.

If you look closely at Fig. 11.1, you will notice that the rectangles were drawn using
the midpoint method.
The trapezoidal rule uses trapezoids instead of rectanglesfor the regions  
to approx- 
imate the area. The vertices of the kth trapezoid are (xk−1 , 0) , xk−1 , f (xk−1 , xk , f (xk ) ,
(xk , 0). The area of the kth trapezoid is
x  
Ak = f (xk−1 ) + f (xk ) .
2
Why? Because it is actually the average of the left and right endpoint rules (see Fig. 11.3
above).
Thus the formula for using the trapezoidal rule with n rectangles is

x  
n
b 
f (x) dx ≈ Tn = f (xk−1 ) + f (xk )
a 2 k=1
x  
= f (x0 ) + 2f (x1 ) + 2f (x2 ) + · · · + 2f (xn−1 ) + f (xn )
2
Fig. 11.4 shows all four Riemann sums.

11.3. Error bounds

We have error bounds for the midpoint and trapezoidal rules:
Theorem 11.3.1 (Error bounds). Suppose |f  (x)| ≤ m for all x ∈
 b
[a, b]. Then using the midpoint rule to estimate the integral f (x) dx
a
will have an error εM with
m(b − a)3
|εM | ≤
24n2
and using the trapezoidal rule will have an error εT with
m(b − a)3
|εT | ≤ .
12n2
186 Programming Mathematics Using MATLAB®

Figure 11.4 Common Riemann sums.

We will not prove these error bounds in these notes; that is for a different course.
Notice that the error bounds get smaller faster for the midpoint rule than the trapezoidal
rule (which may seem counterintuitive).

11.4. Simpson’s rule

For the Simpson rule of approximating integrals, areas under quadratics are used to
 b
approximate the area under the curve, A = f (x) dx. The Simpson rule starts by sub-
a
b−a
dividing the interval [a, b] into n equal subintervals of width x = , where n is
n
even. Then we take pairs of consecutive subintervals and estimate the area under the curve
with a quadratic (see Fig. 11.5). Thus our first approximating quadratic is estimating the
area from x0 to x2 , then the second approximating quadratic is estimating the area from
x2 to x4 , etc., until the n/2th approximating quadratic is estimating the area from xn−2
to xn . Notice that our formulas for x0 , x1 , xk , etc. are as in the previous rules.
It is done this way because in order to come up with a quadratic

A1 x2 + B1 x + C1 ,

one needs three points on the quadratic. Having these pairs of subintervals allows for
three data points at each approximation. Let us investigate the first estimating quadratic;
we want to estimate the area under the curve y = f (x) from x0 to x2 . We have three
data points: (x0 , f (x0 )), (x1 , f (x1 )), and (x2 , f (x2 )). For ease of notation, let us have
y0 = f (x0 ), etc.
 Numerical Integration 187

Figure 11.5 Simpson’s rule.

Figure 11.6 Translating areas.

In order to make finding the quadratic A1 x2 + B1 x + C1 easier, let us shift the x’s to
be from −h to h. So x0 → −h, x1 → 0, and x2 → h, with h = x (see Fig. 11.6).
Notice that the area under the curve does not change and we now have the data
points (−h, y0 ), (0, y1 ), and (h, y2 ) to figure out the approximating quadratic and/or
approximating area.
To figure out A1 x2 + B1 x + C1 :
x = 0 =⇒ C1 = y1 , (11.1)
x = −h =⇒ A1 h2 − B1 h + C1 = y0 , (11.2)
x = h =⇒ A1 h2 + B1 h + C1 = y2 . (11.3)

Adding Eqs. (11.2) and (11.3), we see that

y0 + y2 = 2A1 h2 + 2C1 . (11.4)
188 Programming Mathematics Using MATLAB®

If we were actually going to figure out the coefficients of the approximating

quadratic, we have already figured C1 in (11.1), and we can use that in (11.4) to figure
out A1 , and then use either (11.2) or (11.3) to find B1 .
But we actually do not need to figure out all of the coefficients because all we need
is to figure out the area under the quadratic, i.e.,
 h
A1 x2 + B1 x + C1 dx.
−h

Using facts about integrals of even and odd functions, we get

 h  h
A1 x2 + B1 x + C1 dx = 2 A1 x2 + C1 dx
−h 0
h
2 
= A1 x3 + 2C1 x
3 0
2
= A1 h + 2C1 h
3
3
h 
= 2A1 h2 + 6C1 . (11.5)
3
By combining (11.1) with (11.4), we get

y0 + 4y1 + y2 = 2A1 h2 + 6C1 . (11.6)

Thus
 h
h 
A1 x2 + B1 x + C1 dx = y0 + 4y1 + y2 .
−h 3
Now if we shift the x back, the area under the curve remains the same and we get
 x2
h 
A1 x2 + B1 x + C1 dx = y0 + 4y1 + y2 .
x0 3

Likewise
 x4
h 
A2 x2 + B2 x + C2 dx = y2 + 4y3 + y4 .
x2 3
Thus the general formula for the Simpson rule is
 b
h 
f (x) dx ≈ y0 + 4y1 + 2y2 + 4y3 + 2y4 + · · · + 2yn−2 + 4yn−1 + yn .
a 3

Notice the pattern of the coefficients: 1, 4, 2, 4, 2, . . . , 2, 4, 1.

There is also an error estimate for the Simpson rule (not proven here).
 Numerical Integration 189

Theorem 11.4.1 (Simpson’s rule error estimate). If |f (4) (x)| ≤ m for

 b
x ∈ [a, b], then using Simpson’s rule to estimate f (x) dx with n
a
subintervals will have an error εS where
m(b − a)5
|εS | ≤ .
180n4

11.5. Exercises
Note that all ERROR CHECKS within your functions should use the error command
and display a meaningful error message.
1. Create a function RSum that takes as input a function fstring, a, b, n, sumType,
and color. The commands syms, str2sym, subs, and double may be useful or
necessary. The switch selection statement may also come in handy for this assign-
ment.
• The input fstring will be a string for the function f (x) that we want to in-
tegrate numerically. The string is in single quotes using standard MATLAB®
notation WITHOUT component-wise calculations. No error checking will be
done on this.
• The inputs a and b are the lower and upper bounds, respectively for the integra-
tion. An error check will check that a < b; if not, an appropriate error message
will be displayed.
• The input n will be the number of rectangles/subintervals that will be made for
the numerical integration (Riemann Sum). An error check will check that n is
a positive integer; if not, an appropriate error message will be displayed.
• The input sumType will determine which type of Riemann sum will be calcu-
lated and plotted. It will be a string in single quotes and will be either 'right',
'left', 'mid' or 'trap'; if not, an appropriate error message will be dis-
played.
• The input color will specify which color will be used to fill in the rectangles.
It could be a vector or a string name for a color. No error checking will be
done on this input.
The function will have 1 output; the calculated Riemann sum (using the right
Riemann sum if sumType = 'right', the left Riemann sum if sumType = 'left',
the midpoint rule if sumType = 'mid', and the trapezoidal rule if sumType =
'trap'). The value that is the output should NOT be symbolic—thus you may
need to use the double command.
The function will also plot the rule (depending on the value of sumType) of y = f (x)
from x = a to x = b using n subdivisions. You will use the fill command. Both the
edges of the rectangles/trapezoids and the function y = f (x) will be plotted in black.
190 Programming Mathematics Using MATLAB®

There will be no titles, axis labels, or other axis commands within the function; the
user can add these outside of the function.
2. For f (x) = sin4 (π x) + 2x, figure out the following by hand on paper and/or using
MATLAB to help with some calculations. Any calculations not done by hand should
be shown in MATLAB (“show your work”).
 3
(a) Approximate f (x) dx using a left Riemann sum and n = 4.
1 3
(b) Approximate f (x) dx using a right Riemann sum and n = 4.
1 3
(c) Approximate f (x) dx using a midpoint rule and n = 4.
1 3
(d) Approximate f (x) dx using a trapezoidal rule and n = 4.
1
3. (a) Use the subplot command and your function to show all four Riemann
 3
sums with n = 4 on the same figure using your RSum function for f (x) dx
1
using f (x) = sin (π x) + 2x. This will be a 2 × 2 figure where the top row
4

will be the left and right Riemann sums and the second row will be the
midpoint rule and trapezoidal rule. Let the left Riemann sum be in MyGreen
([0, 0.4078, 0.3412]), the right Riemann sum in yellow, the midpoint rule in
red with alpha(0.3), and the trapezoidal rule in gray ([0.75, 0.75, 0.75]).
Make sure you add titles specifying which rule is which.
(b) Compare the answers you get from your function with your answers in the
previous problem. Are they what you expected?
4. Use the subplot command to plot the one of the rules (your choice for the color
and rule but use the same for all four within these subplots) using your RSum function
 3
on a figure with 2 × 2 subplots: n = 4, n = 8, n = 30, n = 75 for (sin4 (π x)+ 2x) dx.
1
5. Create a function SRule that takes as input a function fstring, a, b, and n.
• The input fstring is a string for the function f (x) in single quotes using stan-
dard MATLAB notation WITHOUT component-wise operations, just as in
RSum. No error check is done on this input.
• The inputs a and b are the lower and upper bounds, respectively for the integra-
tion. An error check will check that a <b; if not, an appropriate error message
will be displayed.
• The input n will be the number of rectangles/subintervals that will be made for
the Simpson rule. An error check will check that n is a positive EVEN integer;
if not, an appropriate error message will be displayed.
The commands syms, str2sym, sum, subs, and double will be useful or necessary.
The output of the function will be the approximation of the integral of f (x) from a
 Numerical Integration 191

to b using the Simpson rule on the given n. Just as in the RSum function, it should
return a numerical approximation so you may need to use the double command.
6. Create a function SPlot.m that has the same inputs (and error checks) as SRule
in the above exercise. This function will plot the Simpson rule of y = f (x), from
x = a to x = b using n subdivisions to create a figure similar to Fig. 11.5. You will
use the syms, str2sym, polyfit, and the polyval commands. The approximating
quadratics will be plotted without a color specified (so that the graphs cycle through
the colors), and the curve y = f (x) will be in black. Have the domains in the plots for
the approximating quadratics be 0.05 beyond the xk used in the approximations. For
example, if the first quadratic is approximating the curve from 1 to 1.5, then have
the plot of the quadratic be from 0.95 to 1.55. You will also plot the subdivisions in
black; these will be vertical lines from (xk , 0) to (xk , yk ) for each of x0 , x1 , . . . , xn .
 3
7. We will investigate Simpson’s rule to estimate f (x) dx for f (x) = sin4 (π x) + 2x,
1
above with n = 2 and n = 4. For the subdivisions, figure out what the FIRST ap-
proximating quadratics would be. This should be done on paper and turned in,
showing all work, explaining it clearly, and using exact values. Any calculations us-
ing technology should be done in MATLAB. The answers for the approximating
quadratics (and the subintervals they are for) should be written clearly. Use the in-
terval [−h, h] and figure out the coefficients A, B, and C based on the values of
yk , yk+1 , and yk+2 from the notes. That quadratic is based on the middle x-value
equaling 0. Use a horizontal shift so that the middle value is now at xk+1 .
(a) Figure out the FIRST approximating quadratic for n = 2.
(b) Figure out the FIRST approximating quadratic for n = 4.
8. Check your answers in the above problem using the command polyfit and even
poly2sym, clearly labeling your answers.
(a) Check for n = 2.
(b) Check for n = 4.  3
9. Use your SRule function to approximate (sin4 (π x) + 2x) dx with the Simpson
1
rule using n = 2, n = 4, n = 8, and n = 16. Also, use subplot as in previous exercises
with your SPlot to visualize the Simpson rule for these four numerical approxima-
tions.
CHAPTER 12

The Gram–Schmidt Process

12.1. General vector spaces and subspaces
12.1.1 Vector spaces

Definition 12.1.1. A vector space is a set V of elements called vec-

tors that have operations called vector addition and scalar multiplica-
tion defined so that the following conditions hold for any u, u, w ∈ V
and scalars a, and b.
• Closure properties
c-1. u + v ∈ V
c-2. au ∈ V
• Addition properties
a-1. u + v = v + u (commutativity)
a-2. u + (v + w) = (u + v) + w (associativity)
a-3. ∃ zero vector 0 ∈ V such that (additive identity)
u+0=u ∀u∈V
a-4. ∀ u ∈ V , ∃ −u ∈ V such that (additive inverse)
u + (−u) = 0
• (Scalar) multiplication properties
m-1. a(bu) = (ab)u (associativity)
m-2. a(u + v) = au + av (distributive)
m-3. (a + b)u = au + bu (distributive)
m-4. 1u = u ∀ u ∈ V

Scalars are commonly the set of real numbers (called a real vector space) or the set
of complex numbers (called a complex vector space).
Examples of vector spaces
1. Rn : Euclidean vector space of Rn
2. Mmn : the set of m × n matrices.
3. The set of all functions with domain R:
• Addition is defined pointwise: f + g = h is the function
h(x) = (f + g)(x) = f (x) + g(x).
• af is defined as the function af (x).

Programming Mathematics Using MATLAB® Copyright © 2021 Elsevier Inc.

https://doi.org/10.1016/B978-0-12-817799-0.00018-1 All rights reserved. 193
194 Programming Mathematics Using MATLAB®

Note: you could define a different domain instead of R. You could also change it to
be all continuous functions on domain D, all differentiable functions with domain D,
etc.
4. Pn : the set of all real-polynomials of degree ≤ n.
5. C n [a, b]: the set of all functions whose nth derivatives exist and are continuous on
[a, b] (thus the function, first derivative, second derivative, . . . , nth derivative are all
continuous on [a, b]).
6. The set of all functions that satisfy the differential equation 3y − y + y = 0.

Theorem 12.1.1 (Properties of vectors). Let V be a vector space,

v ∈ V , 0 ∈ V and 0 the zero scalar. Then
1. 0v = 0, and c 0 = 0 for any scalar c.
2. (−1)v = −v.
3. If c v = 0, then either c = 0 or v = 0.

12.1.2 Subspaces

Definition 12.1.2. Let V be a vector space and U be a nonempty

subset of V . If U is a vector space under the same addition and scalar
multiplication, then U is called a subspace of V .
U is a subspace if it is closed under addition and scalar multiplica-
tion. All other vector space properties in the definition are inherited
from V .

Example 12.1.1. Which of the following are subspaces of Mnn ?

(a) The subset of all symmetric matrices.
(b) The subset of all matrices that are not symmetric.
(c) The subset of invertible matrices.
(d) All diagonal matrices.

Example 12.1.2. Which subsets W are subspaces of the vector space V ?

(a) V = P3 , W = P2 .
(b) V = P3 , W = {p(x) | p(x) = ax2 + bx + 1, a, b ∈ R}.
(c) V = set of all functions with domain being R,
W = {f (x) ∈ V | f (0) = 0}.
(d) V = set of all functions with domain being R,
W = {f (x) ∈ V | f (0) = 1}.
(e) V = set of all functions

with domain being R,
W = {f (x) ∈ V | ab f (x) dx = 0} for some fixed a, b ∈ R.
 The Gram–Schmidt Process 195

12.2. Linear combinations of vectors

Definition 12.2.1. Let V be a vector space and v1 , v2 , . . . , vm ∈ V .

We say v ∈ V is a linear combination of v1 , v2 , . . . , vm if there exist
scalars a1 , a2 , . . . am such that

v = a1 v1 + a2 v2 + · · · + am vm .

Example 12.2.1. Is the first vector a linear combination of the other vectors? If so,
write the linear combination.
(a) x = (−11, 9, −3); v1 =(1, −1,1), v2 =(2, 1,4), v3 =(−2, 3, 
1).
1 −25 2 0 0 1 1 5
(b) x = ; v1 = , v2 = , v3 = .
8 −13 1 −1 3 4 −1 2
(c) f (x) = 2x2 + x − 3; g(x) = x2 − x + 1, h(x) = x2 + 2x − 2.

Definition 12.2.2. Let V be a vector space and v1 , v2 , . . . , vm ∈ V .

These vectors span V if every vector V can be written as a linear
combination of v1 , v2 , . . . , vm .

Example 12.2.2. The following polynomials span V = P2 :

f1 = 1, f2 = x, f3 = x2 + x, f4 = x2 − 1.

Example 12.2.3. The following matrices span V = M22 :

       
1 0 1 1 0 0 0 0
A1 = , A2 = , A3 = , A4 = .
0 0 0 0 1 0 0 1

Theorem 12.2.1. Let S = {v1 , v2 , . . . vm } ⊂ V . Let U be the set con-

sisting of all linear combinations of vectors from S. Then U is a
subspace of V spanned by the vectors in S, and we say U is the vector
space generated by S, or is the span of S. We denote this by

U = span S = span{v1 , v2 , . . . vm }.

Example 12.2.4. Determine whether the vector v is in the span of S.

(a) v = (1, 4, −3), S = {(1, 0, 1), (1, 1, 0), (3, 1, 2)} (yes).
196 Programming Mathematics Using MATLAB®

(b) v = (1, 1, 2), S = {(0, 1, 0), (3, 5, 6), (1, 2, 1)} (no).

Example 12.2.5. Give examples of three other functions in the subspace span S,
where
S = {2x + 1, 3x2 + x − 3}.

12.3. Linear independence and bases

12.3.1 Linear independence

Definition 12.3.1. Let V be a vector space.

1. S = {v1 , v2 , . . . , vm } ⊂ V is linearly dependent if there exists a
nontrivial solution to

c1 v1 + c2 v2 + · · · + cm vm = 0. (*)

2. S is LINEARLY INDEPENDENT if the only solution to (*) is the

trivial solution
c1 = c2 = · · · = cm = 0.

Example 12.3.1. Are the following sets of vectors linearly independent or linearly
dependent?
(a) {1, x, x2 } ⊂ P2 .
{x2 + x3
(b)  } ⊂P3 . 
1 0 0 0
(c) , ⊂ M22 .
0 1 0 1
(d) {f1 , f2 , f3 , f4 } ⊂ P3 where

f1 = 1, f2 = x, f3 = x3 + x, and f4 = x2 − 1.

(e) {f1 , f2 , f3 } ⊂ P1 where

f1 = 1, f2 = x, and f3 = 2x − 1.

Theorem 12.3.1. Let V be a vector space and V = span{v1 , v2 ,

. . . vm }. Each vector in V can be expressed uniquely as a linear com-
bination of these vectors iff the vectors are linearly independent.
 The Gram–Schmidt Process 197

Proof. (⇐=) Assume the set of vectors S = {v1 , v2 , . . . vm } are LI. Let v ∈ V . Since V =
span S, by definition of span we can write v as a linear combo of these vectors. Suppose
we have two ways of writing v as a linear combination:
 
v= ak vk , v= bk vk .

Then we have

(ak − bk )vk = 0.

But since these vectors are LI, ak − bk = 0 for all k. Thus ak = bk for all k, and there’s
only one way of expressing v as a linear combo of the vectors in S.
( =⇒ ) Let v ∈ V be arbitrary and we are given that v can be written uniquely as a
linear combo of vectors in S. Since 0 ∈ V , then 0 can be written uniquely as a linear
combination of vectors in S. We know

0= 0vk ,

so this must be the only solution, thus the vectors in S are linear independent.

12.3.2 Bases
Definition 12.3.2. Let V be a vector space. Let S = {v1 , v2 , . . . , vm } ⊂
V be such that
1. S is a linearly independent set of vectors, and
2. V = span S.
Then S is called a BASIS for V .
Thus if S is a basis for V , each vector in V can be expressed
uniquely as a linear combination of vectors in S.

Example 12.3.2. The standard basis for Rn is

e1 = (1, 0, . . . , 0), e2 = (0, 1, 0, . . . , 0), . . . , en = (0, . . . , 0, 1).

Example 12.3.3. A basis for V = M22 is

       
1 0 0 1 0 0 0 0
A1 = , A2 = , A3 = , A4 = .
0 0 0 0 1 0 0 1

This is the standard basis for M22 .

198 Programming Mathematics Using MATLAB®

Example 12.3.4. The standard basis for V = P3 is

f1 = 1, f2 = x, f3 = x2 , f4 = x3 .

Theorem 12.3.2. Let {v1 , v2 , . . . , vn } be a basis for vector space V . If

{w1 , w2 , . . . , wm } is a set of more than n vectors in V , then this set is
linearly dependent.


Proof. Let ck wk = 0. Since {v1 , v2 , . . . , vn } is a basis for V , each of the vectors wk can
be written as a linear combo of these basis vectors.

w1 = a11 v1 + a12 v2 + ...

Thus we get

ck (ak1 v1 + ak2 v2 + · · · + akn vn ) = 0.
If we rearrange, we get

(c1 a11 + c2 a21 + · · · cm am1 )v1 + · · · = 0.

Thus a11 c1 + a21 c2 + · · · + am1 cm = 0, etc. Notice that we have m variables and n equations,
but n < m so at least one of the variables is free. Therefore there are nontrivial solutions
for the ck so the set is LD.

Corollary 12.3.1. All bases for a vector space V have the same num-
ber of vectors.

Definition 12.3.3. The number of vectors in the basis of a vector

space V is called the DIMENSION OF V and is denoted by dim(V ).

Example 12.3.5. dim(M22 ) = 4, dim(P2 ) = 3.

Some vector spaces have infinite dimension!

Example 12.3.6. P = set of all polynomials and C 2 [0, 1], etc. are infinite dimensional
vector spaces.
 The Gram–Schmidt Process 199

Example 12.3.7. What are the dimensions of the following spaces?

(a) M44 .
(b) V = set of all diagonal 4 × 4 matrices.
(c) W = set of all upper triangular 4 × 4 matrices.

Theorem 12.3.3. 1. If V = {0}, then dim(V ) = 0

2. If dim(V ) = 1, V ⊂ Rn , (n = 2, 3), then V is a line through the
origin.
3. If dim(V ) = 2, V ⊂ Rn , (n = 2, 3), then V is a plane through the
origin.

Theorem 12.3.4. Let V be a vector space and dim(V ) = n.

Let S = {v1 , v2 , . . . , vn } ⊂ V .
1. If S is a linearly independent set, then S is a basis for V .
2. If span(S) = V , then S is a basis for V .

⎡ √ ⎤ ⎡ √ ⎤ ⎡ ⎤
1/ 2 −1/ 2 0
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Example 12.3.8. ⎣ 0√ ⎦ , ⎣ √ 0 ⎦ , ⎣1⎦ are linearly independent in R3 (check
1/ 2 1 2 0
it!) so the set of these vectors form a basis for R3 .

Theorem 12.3.5. Let V be a vector space of dimension n. Let

S = {v1 , v2 , . . . , vm } ⊂ V . Suppose S is a linearly independent set
with m < n. Then there exist vectors vm+1 , vm+2 , . . . , vn such that

S ∪ {vm+1 , vm+2 , . . . , vn }

form a basis for V .

   
1 0 0 0
Example 12.3.9. , form a linearly independent set in the space of
0 0 0 1
lower triangular matrices. We need one more to form a basis:
 
0 0
.
5 0
200 Programming Mathematics Using MATLAB®

12.4. Rank

Definition 12.4.1. Let A be an m × n matrix. The rows of A may

be viewed as row vectors R = {r1 , . . . , rm }, and the column vectors
C = {c1 , . . . cn }. Each row vector has n components, so R ⊂ Rn , and
each column vector has m components, so C ⊂ Rm . The span(R) is a
subspace of Rn and is called the ROW SPACE OF A, and span(C ) is a
subspace of Rm and is called the COLUMN SPACE OF A.

Theorem 12.4.1. The row space and column space of a matrix A

have the same dimension (even if these subspaces live in different vec-
tor spaces).

Definition 12.4.2. The dimension of the row space and the column
space of a matrix A is called the RANK OF A, denoted by rank(A).

rank(A) = dim(row space of A) = dim(col space of A).

How do we find the rank and dimension of these spaces?

Theorem 12.4.2. Let A and B be row equivalent matrices. Then A

and B have the same row space, and rank(A) = rank(B).

Proof. Since A and B are row equivalent, the rows of B can be obtained from the rows
of A through elementary row operations, and vice versa. Thus each row of B is a linear
combination of the rows of A. Thus the row vectors of B are in the row space of A, and
vice versa. Thus the row spaces are the same.

Theorem 12.4.3. (Really a corollary) Let A be a matrix in reduced

echelon form. The nonzero row vectors of A are a basis for the rows-
pace of A. Thus the number of nonzero rows in A are rank(A).

Theorem 12.4.4. (Again, really a corollary) Let E = rref(A). The

nonzero row vectors of E form a basis for the row space of A. The
number of nonzero rows in E is rank(A).
 The Gram–Schmidt Process 201

Corollary 12.4.1. To find a basis for the column space of A, put AT

in reduced echelon form.

Theorem 12.4.5. Let Ax = y be an m × n system and let B be the

augmented matrix [A y].
1. If rank(A) = rank(B) = r and r = n, then the solution is unique.
2. If rank(A) = rank(B) = r and r < n, then the solution has infinitely
many solutions.
3. If rank(A) = rank(B), then there are no solutions.

Theorem 12.4.6. (Summary) Let A be an n × n matrix. TFAE

1. det(A) = 0.
2. A is nonsingular.
3. A is invertible.
4. A is row equivalent to In .
5. The system of equations Ax = b has a unique solution (no matter
what the value of b is).
6. rank(A) = n.
7. The column vectors of A are linearly independent.
8. The column vectors of A span Rn .
9. The column vectors of A form a basis for Rn .

12.5. Orthonormal vectors and the Gram–Schmidt process

12.5.1 Orthogonal and orthonormal vectors
Recall the following definitions.

Definition 12.5.1. The DOT PRODUCT or INNER PRODUCT or

SCALAR PRODUCT of u = (u1 , u2 , . . . , un ) and v = (v1 , v2 , . . . , vn ) in
Rn , is

n
u · v = u1 v1 + u2 v2 + · · · + un vn = uk vk .
k=1
202 Programming Mathematics Using MATLAB®

Definition 12.5.2. The NORM / MAGNITUDE / LENGTH of a vector

u = (u1 , u2 , . . . , un ) ∈ Rn is

u = u21 + u22 + · · · + u2n = (u · u)1/2 .

Any vector u = 0 can be normalized; i.e., a unit vector v can be found in the same
direction as u:
1
v= u.
u

Definition 12.5.3. A pair of vectors u, v in Rn is ORTHOGONAL if

u · v = 0.

(In R2 and R3 , orthogonal is synonymous with perpendicular.)

Definition 12.5.4. A set of vectors S ⊂ V is said to be an

ORTHOGONAL SET if every pair of vectors in the set is orthogo-
nal. The set S is an ORTHONORMAL SET if is an orthogonal set and
every vector is a unit vector.

⎧ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎫
⎪
⎨ 1 1 1 ⎪⎬
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Example 12.5.1. S = u1 = ⎣−1⎦ , u2 = ⎣2⎦ , u3 = ⎣ 0 ⎦ ,
⎪
⎩ ⎪
1 1 −1 ⎭

u1 · u2 = 0,
u1 · u3 = 0,
u2 · u3 = 0,

=⇒ S is an orthogonal set.

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
√1 √1 0
⎢ 2
⎥ ⎢ 2
⎥ ⎢ ⎥
Example 12.5.2. The vectors v1 = ⎢ 1 ⎥
⎣− √ 2 ⎦ , v2 = ⎢ 1 ⎥
⎣ √2 ⎦ , v3 = ⎢ ⎥
⎣ 0 ⎦ form an
0 0 −1
orthonormal set.

Theorem 12.5.1. An orthogonal set is linearly independent.

 The Gram–Schmidt Process 203

Proof. Let S = {v1 , v2 , . . . , vm } be an orthogonal set of nonzero vectors in V . Consider

the equation
c1 v1 + c2 v2 + · · · + cm vm = 0.
Take the equation and take the dot product with a vector vk ∈ S:

(c1 v1 + c2 v2 + · · · + cm vm ) · vk = 0 · vk ,
c1 v1 · vk + c2 v2 · vk + · · · + cm vm · vk = 0;

since the vectors are orthogonal, we get

ck vk · vk = 0.

Since we know vk = 0, we get ck = 0. This can be done for each k = 1, 2, . . . m and we

see that the set S is linearly independent.

Definition 12.5.5. An orthogonal set that is a basis is an

ORTHOGONAL BASIS . An orthonormal set that is a basis is an
ORTHONORMAL BASIS .

Example 12.5.3. The set S from Example 12.5.1 are vectors in R3 . They are orthog-
onal; thus they are LI. We know dim(R3 ) = 3 thus S is a basis for R3 and since they
√
are
orthogonal, S is an orthogonal basis for R3 . Is S an orthonormal basis? u1 = 3 so
no. (But one could form one from S... remember how?)

Example 12.5.4. The standard bases for Rn , Mmn , and Pn are all orthonormal. The
vectors in Example 12.5.2 above are another example of an orthonormal basis for R3 .

 One can always “normalize” an orthogonal basis to get an orthonormal one.

The following theorem hints at one importance of orthonormal bases.

Theorem 12.5.2. Let {v1 , v2 , vn } be an orthonormal basis for a vector

space V . Let u ∈ V . Then u can be written as a linear combination of
the vectors in the basis in the following way:

u = (u · v1 )v1 + (u · v2 )v2 + · · · + (u · vn )vn .

204 Programming Mathematics Using MATLAB®

Proof. We know that we can find coefficients c1 , c2 , . . . , cn such that

c1 v1 + c2 v2 + · · · + cn vn = u.

Now take the dot product of the equation with vk :

(c1 v1 + c2 v2 + · · · + cn vn ) = u · vk ,
···
ck vk · vk = u · vk ,

since the set is orthonormal, vk · vk = 1 so we have

ck = u · vk .

Example 12.5.5. Using the above orthonormal basis for R3 from Example 12.5.2,
find the coefficients for (−1, 2, −3).

12.5.2 The Gram–Schmidt process

The Gram–Schmidt process is a way of finding an orthogonal basis from a given basis.

Theorem 12.5.3. Let W be a p-dimensional subspace of Rn and let

{w1 , w2 , . . . , wp } be any basis for W . Then {u1 , u2 , . . . , up } is an or-
thogonal basis for W where

u1 = w1 ,
u1 · w2
u2 = w2 − u1 ,
u1 2
u1 · w3 u2 · w3
u3 = w3 − u1 − u2 ,
u1 2 u2 2
..
.
i−1
 uk · wi
ui = wi − uk .
uk 2
k=1

Proof. The proof is in linear algebra books—read it. Basically the numbers in front of
the uk render each vector orthogonal to the previous ones.
 The Gram–Schmidt Process 205

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 2 1
⎢0⎥ ⎢1⎥ ⎢−1⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Example 12.5.6. w1 = ⎢ ⎥ , w2 = ⎢ ⎥ , w3 = ⎢ ⎥,
⎣1⎦ ⎣0⎦ ⎣0⎦
2 2 1
⎡ ⎤
1
⎢0⎥
⎢ ⎥
u1 = w1 = ⎢ ⎥ ,
⎣1⎦
2
u1 · w2 2 + 0 + 0 + 4 6
= = = 1,
u1 2 1+1+4 6
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
2 1 1
⎢1⎥ ⎢0⎥ ⎢ 1 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
u2 = w2 − (1)u1 = ⎢ ⎥ − ⎢ ⎥ = ⎢ ⎥ ,
⎣0⎦ ⎣1⎦ ⎣−1⎦
2 2 0
u1 · w3 1 + 0 + 0 + 2 1
= = ,
u1 2 6 2
u2 · w3 1 − 1 + 0 + 0
= = 0,
u2 2 3
⎡ ⎤ ⎡ ⎤
1
1 2
⎢ ⎥ ⎢ ⎥
1 ⎢−1⎥ ⎢ 0 ⎥
u3 = w3 − u1 − 0u2 = ⎢ ⎥ − ⎢ 1 ⎥
2 ⎣ 0 ⎦ ⎣2⎦
1 1
⎡ ⎤
1
2
⎢ −1 ⎥
⎢ ⎥
= ⎢ 1⎥,
⎣− 2 ⎦
0

⎧⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎫
⎪
⎪ 1 1 1 ⎪
⎪
⎪
⎨⎢0⎥ ⎢1⎥ ⎢ −1 ⎥⎪
2
⎬
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥, ⎢ ⎥, ⎢ 1⎥ .
⎪⎣1⎦
⎪ ⎣−1⎦ ⎪
⎣− 2 ⎦⎪
⎪
⎩ 2 ⎪
0 0 ⎭
√ √ √
What if one wanted an orthonormal basis? u1 = 6, u2 = 3, u3 = 2
6
...

Why do we want orthogonal or orthonormal bases?

Suppose B = {b1 , b2 , . . . , bp } is a basis for a subspace W . Then, for any x ∈ W ,

x = a1 b1 + a2 b2 + · · · + ap bp
206 Programming Mathematics Using MATLAB®

where these ai are unique. These ai are called the COORDINATES of x with respect to
the basis B.
⎧⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎫
⎪
⎨ 2 1 1 ⎪⎬
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Example 12.5.7. B = ⎣ 1 ⎦, ⎣2⎦ , ⎣−1⎦ is a basis for R3 but not an or-
⎪
⎩ −1 ⎪
0 1 ⎭
thogonal basis. ⎡ ⎤
3
⎢ ⎥
Express x = ⎣ 2 ⎦ in terms of B. We would have to solve the system ab1 + bb2 +
−1
cb3 = x:
  rref 7 1 1
b1 b2 b3 x −−→ a = , b = , c = .
6 2 6

But if B2 = {b1 , b2 , . . . , bp } was an orthogonal basis, it would be easier to solve for

the coordinates:

x = a1 b1 + a2 b2 + · · · + ap bp ,
bi x = · · · = a1 bTi b1 + a2 bTi b2 + · · · + ai bTi bi
T
+ · · · + ap bTi bp ,
a3 · 0,
=⇒ ai bTi bi = ai bi 2 = bTi x
bTi x bi · x
=⇒ ai = = .
bi 2 bi 2
Thus if a basis is orthonormal, the coefficients with respect to that basis are

ai = bTi x = bi · x.
⎡ ⎤ ⎧⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎫
−1 ⎪
⎨ 1 −1 0 ⎪
⎬
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Example 12.5.8. x = ⎣ 1 ⎦, B = ⎣0⎦ , ⎣ 0 ⎦ , ⎣1⎦ ; B is an orthogonal basis.
⎪
⎩ 1 ⎪
2 1 0 ⎭
Write x in terms of B,
b 1 · x −1 + 2 1
a1 = = = ,
b1 2 2 2
b2 · x 1 + 2 3
a2 = = = ,
b2 2 2 2
b2 · x 1
a3 = = = 1,
b3 2 1

1 3
x = b1 + b2 + b3 .
2 2
 The Gram–Schmidt Process 207

Figure 12.1 A basis for R2 .

Besides finding “coordinates” more easily, there are many other reasons why one
would want orthogonal or orthonormal bases for spaces and subspaces.

Example 12.5.9. In this example we will visualize the difference between bases in
R2 . Let
   
1 4
b1 = and b2 = .
1 3
These vectors are linearly independent, so they form a basis for R2 .
If we graph these two vectors, we clearly see that the vectors are not orthogonal and
are of different lengths (see Fig. 12.1).
Using the Gram–Schmidt process, we get an orthogonal basis of
   1

1 2
u1 = and u2 = .
1 − 12

If we graph these vectors, we clearly see these are now orthogonal (see Fig. 12.2).
If we normalize these vectors, we get the following orthonormal basis for R2 :
   
√1 √1
2 2
v1 = and v2 = .
√1
2
− √12

Graphing these vectors along with the unit circle we clearly see these vectors are not
only orthogonal but also of length 1 (see Fig. 12.3).
208 Programming Mathematics Using MATLAB®

Figure 12.2 Orthogonal basis vectors for R2 .

Figure 12.3 Orthonormal basis vectors for R2 .

Example 12.5.10. In this example we will visualize the difference between bases for
R3 . Let
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
4 1 1
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
b1 = ⎣ 1 ⎦ , b2 = ⎣1⎦ , and b3 = ⎣−1⎦ .
−1 0 1

These vectors are linearly independent, so they form a basis for R3 .

If we graph these three vectors and adjust the view, we see that the vectors are not
orthogonal to each other, and they are of different lengths (see Fig. 12.4).
 The Gram–Schmidt Process 209

Figure 12.4 Basis vectors for R3 .

Figure 12.5 Orthogonal basis vectors.

Using the Gram–Schmidt process, we get an orthogonal basis of

⎡ ⎤ ⎡ 1⎤ ⎡ ⎤
4 −9 5
11
⎢ ⎥ ⎢ 13 ⎥ ⎢ 5⎥
u1 = ⎢ ⎥
⎣ 1 ⎦, u2 = ⎣ 18 ⎥
⎢
⎦, u3 = ⎢ ⎥
⎣− 11 ⎦ .
−1 5
18
15
11

If we graph these and change the view, we can see these are orthogonal vectors (see
Fig. 12.5).
If we take these orthogonal basis vectors and normalize them, we get the following
orthonormal basis. The unit sphere is also shown so we can see that the vectors are
indeed of length 1 (see Fig. 12.6).
210 Programming Mathematics Using MATLAB®

Figure 12.6 Orthonormal basis vectors.

12.6. Answers to example problems

Example 12.1.1

(a) Yes (b) No (c) No (d) Yes

Example 12.1.2

(a) Yes (b) No (c) Yes (d) No (e) Yes

Example 12.2.1
(a) Yes. x = v1 − 2v2 + 4v3 .
(b) Yes. x = 3v1 + 0v2 − 5v3 .
(c) No.

Example 12.2.4

(a) Yes. (b) No.

Example 12.2.5 (Answers may vary)

1. f1 (x) = 3x2 + 3x − 2,
2. f2 (x) = 3x2 − x − 4,
3. f3 (x) = 6x2 − 7.
 The Gram–Schmidt Process 211

Example 12.3.1

(a) LI (b) LI (c) LI (d) LI (e) LD

Example 12.3.7

(a) dim(M44 ) = 16, (b) dim(V ) = 4, (c) dim(W ) = 10.

⎡ ⎤
−1
⎢ ⎥ −3 1
Example 12.5.5 ⎣ 2 ⎦ = √ v1 + √ v2 + 3v3 , thus the coefficients for this vector
−3
2 2
 
3 1
with respect to the basis in Example 12.5.2 are − √ , √ , 3 .
2 2

12.7. Exercises
Note that all ERROR CHECKS within your functions should use the error command
and display a meaningful error message.
1. Finish the function gs.m that is found on the text website. It will have as input a
m × n matrix in which the columns of the matrix are linearly independent; thus
the columns of the matrix form a basis for the column space W ⊆ Rm . (We will
not check for this linearly independence—we will assume the user has figured this
out correctly.) The function will return a matrix in which the columns form an
orthogonal basis for the column space W . The function will use the Gram–Schmidt
process to create the vectors (output matrix).
2. Finish the function gsON.m that is found on the text website. It will take as input a
m × n matrix in which the columns of the matrix are linearly independent as in the
above problem. (Again, we will not check for this linearly independence—we will
assume the user has figured this out correctly.) The function will return a matrix in
which the columns form an orthonormal basis for the column space W . This function
may use the function gs.m.
3. For this problem, you will show an example of a basis for R2 using the Gram–
Schmidt process.
(a) Use the randi command to generate a 2 × 2 matrix with integers from −100
to 100. If the determinant of this matrix equals 0, generate another one until
you get a matrix with nonzero determinant (the use of while may be useful
here). The columns of this matrix form a basis for R2 and will be the in-
put matrix for gs and gsON. Make the matrix used output to the command
window.
(b) Use your gs function to generate an orthogonal basis from this basis.
(c) Use your gsON function to generate an orthonormal basis from this basis.
212 Programming Mathematics Using MATLAB®

(d) Using subplots, generate a 2 × 2 matrix of plots of these vectors using the
plotVec function from Exercise 4 in Chapter 6. The first graph will have
your original 2 column vectors plotted. The second graph will have the or-
thogonal basis vectors plotted, and the third graph will have the orthonormal
basis vectors plotted. The fourth graph will have the orthonormal basis vec-
tors plotted along with the unit circle (in blue).
For the first two graphs the range on the axes should be from −100 to 100,
and the third and fourth graphs from −1 to 1. The first graph the vectors
should be in blue, the second in red and the third/fourth in black or another
color of your choosing. Make the LineWidth 2 on all of them, have the
grid on and use the axis equal command in all of the subplots.
4. For this problem, you will show an example of a basis for R3 using the Gram–
Schmidt process.
(a) Use the randi command to generate a 3 × 3 matrix with integers from −100
to 100. If the determinant of this matrix equals 0, generate another one un-
til you get a matrix with nonzero determinant (the use of while may be
useful here). The columns of this matrix form a basis for R3 and will be the
input matrix for gs and gsON. Have the matrix used output to the command
window.
(b) Use your gs function to generate an orthogonal basis from this basis.
(c) Use your gsON function to generate an orthonormal basis from this basis.
(d) Using subplots, generate a 2 × 2 matrix of plots of these vectors using the
plotVec function from Exercise 4 in Chapter 6. The first graph will have
your original 3 column vectors plotted. The second graph will have the or-
thogonal basis vectors plotted, and the third and fourth graphs will have the
orthonormal basis vectors plotted. The fourth graph will also have the unit
sphere plotted using the commands sphere, mesh and specify a color for the
EdgeColor. Also, in order to see the vectors inside the sphere, have the com-
mand hidden off after the mesh command and you may want to set the
alpha value JUST FOR THE MESH COMMAND FOR THE SPHERE
set to a number close to 0 (transparent).
For the first two graphs the range on the axes should be from −100 to 100,
and the third from −1 to 1. The first graph the vectors should be in blue,
the second in red and the third/fourth in black. Make the LineWidth 2 on
all of them, have the grid on and use the axis equal command in all of the
subplots.
5. For this problem we will generate two non-colinear vectors in 3D, plot the plane
containing these vectors, the vectors, and the orthogonal and orthonormal basis
vectors for this plane. Because of the subplot above, you may want to start off this
section of code with clf. Generate a random 3 × 2 matrix with values between
 The Gram–Schmidt Process 213

−2 and 2 (integers or real numbers—your choice). If the cross product of the two
columns equals the zero column vector, keep generating a new random matrix
until the cross product is nonzero. Plot the plane that contains these two vectors
using plotPlane (from Exercise 5 in Chapter 6), with domain from −2 to 2, and
EdgeColor a gray color. Using gs and gsON, find the orthogonal basis vectors and
orthonormal basis vectors for this same plane. Plot the original vectors in a blue
color, the orthogonal vectors in a reddish color and dotted, and the orthonormal
vectors in black. Experiment with the LineWidth so you can see the vectors.
NOTE: For all of the above 3D graphs, you may use the view command so you can
see all of the vectors and you compare between the original and orthogonal, but each
time you rerun it the best view needed may be different because of the different random
vectors generated. Do not worry if the 3D view on any published page or turned in
work is not ideal.
APPENDIX A

Publishing and Live Scripts

A.1. Live scripts
New features of Live Scripts are being added with almost every new update of
MATLAB® . Thus this will not be a detailed explanation of Live Scripts.
Live Scripts have extensions “.mlx” and as mentioned above, can only be viewed and
edited within MATLAB. One can create a Live Script in several ways. You can convert
an already written script to a Live Script by saving the file and selecting the format
“MATLAB Live Code Files (.mlx).” You can also start a new Live Script by selecting
“New > Live Script” within the Editor tab.
A nice feature of Live Scripts is that the code and output or figure are displayed
side-by-side by default, while a published document has a vertical layout. Thus one can
easily modify the code and see the result. Another wonderful feature of Live Scripts is
the ability to add “Controls” such as sliders, drop down menus, etc. The Live Editor has
the ability to have code interspersed with formatted text, similar in style to a published
script file. One can Export the Live Script to several formats, similar to the options of
publishing script files and the look of the resulting document is also similar. One should
note that when you export a Live Script with controls, what is shown is the currently
selected control. See the text website for examples of Live Scripts.

A.2. Basic scripts or M-files

The m-files and the published pages for these examples can be found at on the text
website.
Here I will explain some of the syntax needed to publish m-files to webpages
(HTML files). If you are doing this for an assignment or project, the specific direc-
tions of where to save these files, etc. are given separately.
First, you must understand the difference between a BASIC SCRIPT FILE (m-file)
and a PUBLISHABLE SCRIPT FILE. A script file is just an ASCII (American Standard
Code for Information Interchange, i.e., basic text) file with an extension of .m. Within
that file are MATLAB commands as if you had typed them in at the command prompt
within the Command Window. You can put comments (and should, especially for long
files) to better read/debug the file. Comments start with “%”. Here is an example:

Link 1: basicMFile.m
% Example of Basic Script file

215
216 Publishing and Live Scripts

% Lisa Oberbroeckling, 2019

clc
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
% next problem
A=[1 2 3;4 5 6;7 8 9];
B=[A(1,:); -4*A(1,:) + A(2,:); A(3,:)]
B2=[B(1,:); B(2,:); -7*B(1,:) + B(3,:)]
C=[B2(1,:); -1/3*B2(2,:); B2(3,:)]

A.3. Publishing M-files

There are several different ways you can publish your m-files to HTML files using MAT-
LAB. One way is to enter the following on the command line within the Command
Window: publish('filename.m').
You can also type: publish('filename.m', 'html'). You would do this if you
have changed the publishing settings to be “latex”, for example.
The other way of publishing, which is more common, is to do it within MATLAB’s
Editor Window and press the “Publish” button (or “Save and Publish” button if changes
have been made without saving).
This will create a folder named “html” in the current working directory (if there is
not a folder of that name already there), and put the filename.html file and any other
files necessary for the webpage (PNG files for images, for instance) in the “html” folder.
If we publish the above file, it would not be very pretty. See Link 1 on the course
website.
Notice that the output that is shown on the webpage is out of order of the commands
given; the output of the next problem appears BEFORE the plot. Thus we want to
format the comments in a special way so that when it is published, the MATLAB
commands and output appear in order. This is the topic of the next section.

A.4. Using sections

In order to format our m-file to make it better for publishing, we want to break up
the commands in the m-file into SECTIONS. Each SECTION is broken up by “%%”.
When you do this within the Editor Window, you will notice lines appearing between
each section. Sections are useful not only for publishing, but for running and debugging
scripts. Sections are used in publishing to signify different sections of the page (like for
homework assignments to have different sections for each problem). Sections also deter-
mine how/where output for lines of code are displayed. For more detailed information,
see the MATLAB Documentation on Publish and Share MATLAB Code [17].
 Publishing and Live Scripts 217

A.4.1 Using sections for publishing

If you have text following the “%%” on the same line, this also creates a SECTION
with that text as the section title. In addition, a bulleted list is created with those linked
section titles at the top of the webpage. For the following we just added “%%” to two
lines (lines 1 and 7); compare with basicMFile.m on page 215.

Link 2: publishMFile1.m
%% Example of Basic Script file
% Lisa Oberbroeckling, 2019
clc
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
%% second problem
A=[1 2 3;4 5 6;7 8 9];
B=[A(1,:); -4*A(1,:) + A(2,:); A(3,:)]
B2=[B(1,:); B(2,:); -7*B(1,:) + B(3,:)]
C=[B2(1,:); -1/3*B2(2,:); B2(3,:)]

The above is better than the published page without sections, but can be better. We
may want the page to start with a title. This is done by adding a line “%%” after our title
(and another comment line(s) for other introductory text, like my name).

LINK 3: publishMFile2.m
%% Example of Basic Script file
% Lisa Oberbroeckling, 2019
%%
clc
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
%% second problem
A=[1 2 3;4 5 6;7 8 9];
B=[A(1,:); -4*A(1,:) + A(2,:); A(3,:)]
B2=[B(1,:); B(2,:); -7*B(1,:) + B(3,:)]
C=[B2(1,:); -1/3*B2(2,:); B2(3,:)]

If you look at the published webpage, you will notice that we have a section link and
title for the “second problem” but not for the first. So we probably want to change line
3 to include a section title:
218 Publishing and Live Scripts

LINK 4: publishMFile3.m
%% Example of Basic Script file
% Lisa Oberbroeckling, 2019
%% first problem
clc
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
%% second problem
A=[1 2 3;4 5 6;7 8 9];
B=[A(1,:); -4*A(1,:) + A(2,:); A(3,:)]
B2=[B(1,:); B(2,:); -7*B(1,:) + B(3,:)]
C=[B2(1,:); -1/3*B2(2,:); B2(3,:)]

When publishing m-files, each time a new section is started, MATLAB displays the
output created by the commands of the previous section. The difference between
publishMFile3.html and publishMFile4.html is where the output is displayed for the
second problem.

LINK 5: publishMFile4.m (partial view)

%% second problem
% problem 2a
A=[1 2 3;4 5 6;7 8 9];
B=[A(1,:); -4*A(1,:) + A(2,:); A(3,:)]
%%%
% problem 2b
B2=[B(1,:); B(2,:); -7*B(1,:) + B(3,:)]
%%
% problem 2c
C=[B2(1,:); -1/3*B2(2,:); B2(3,:)]
%%
% problem 2d
x=linspace(-10,10);
y=exp(x); % can comment after a command, too
plot(x,y)
title('Another Example')

Another important place to insert a section break is when you want to have text follow-
ing MATLAB commands, but within the same section. If you just include comments
after the MATLAB commands, they will be formatted as comments within the dis-
played code, not as text. Instead, insert a section break (without a section title) and then
the comment block that will be the text:
 Publishing and Live Scripts 219

LINK 5: publishMFile4.m (partial view)

%% third problem
A=[1 2 3;4 5 6;7 8 9;eye(3)]
% If I don't have a cell break above this comment, this
% text just appears as comments within the command lines,
% and not text on the webpage.
%%
% Instead, have a cell break or a section break

Note that you can also have section titles without having section breaks. This is done
by having the line start with %%%” along with the section text. This will have the text
and/or MATLAB commands be within the sections, but the output of those commands
at the next section break, which may not have the desired effect.

LINK 5: publishMFile4.m (partial view)

%%% Next section
% this section does not have a cell break. This may
% or may not be useful depending on how you want the
% output displayed on the published webpage. It works
% here because this is the last section and cell.
x=linspace(0.0001,10);
y=log(x);
plot(x,y)

Using sections is especially important for m-files with multiple plots. Remember, MAT-
LAB only shows the last plotting command (like plot, ..., mesh, surf, etc.). You can
have multiple commands appear on the same figure by using the hold on and hold
off commands. But if you want to display multiple figures (not in the same window),
you have to either use the pause command or the figure command. This first example
uses the pause command:

LINK 6: publishMFile5.m
%% Example of Basic Script with pause
% Lisa Oberbroeckling, 2019
%%
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
hold on
y=cos(x);
plot(x,y,'r')
hold off
220 Publishing and Live Scripts

title('First Plot')
pause
[x,y]=meshgrid(linspace(-10,10));
z=sin(x).*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Second Plot')

If you run the file publishMFile5.m, the first figure will appear and then MATLAB
will be paused. The second will appear after any key is pressed. When this is published,
even the publishing will be on pause after the first figure is created until you press a key.
But if you look at the webpage, only the last figure is actually shown on the webpage.
As discussed above, when publishing, MATLAB runs each section as a block and then
displays any output. At the end of the section the only output that MATLAB sees
as being created by the section of commands is the second figure. The second figure
replaces the first figure, so it is not shown on the webpage. Thus, we need to create
a section for each figure we want on the webpage. When we publish the m-file, we
have to remember to “press any key” for the publishing can continue, which is really
annoying so you may want to take the pause command out or comment it out. In the
following example we created a new section without a section title.

LINK 7: publishMFile5b.m
%% Example of Basic Script with pause
% Lisa Oberbroeckling, 2019
%%
close all
clc
% first plot
x=linspace(-pi,pi);
y=sin(x);
plot(x,y)
hold on
y=cos(x);
plot(x,y,'r')
hold off
title('First Plot')
% pause
%%
[x,y]=meshgrid(linspace(-10,10));
z=sin(x).*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Second Plot')
 Publishing and Live Scripts 221

The next group of files use the figure command.

LINK 8: publishMFile6.m
%% Example of Basic Script with figure
% Lisa Oberbroeckling, 2019
%%
close all
clc
% first plot
figure(1)
x=linspace(-pi,pi);
y=sin(x);
hold on
y=cos(x);
plot(x,y,'r')
hold off
plot(x,y, 'r')
title('First Plot')
% Second plot
figure(2)
[x,y]=meshgrid(linspace(-10,10));
z=sin(x).*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Second Plot')
% third plot
figure(3)
[x,y]=meshgrid(linspace(-10,10));
z=x.*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Third Plot')
% fourth plot
figure(4)
[x,y]=meshgrid(linspace(-10,10));
z=x.*y;
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Fourth Plot')

When the above file is published, it puts each figure side-by-side on one line. Depending
on how many figures you have this may not have the desired effect, so you may want to
have each figure within its own section instead.

LINK 9: publishMFile6b.m
%% Example of Basic Script with figure
% Lisa Oberbroeckling, 2019
222 Publishing and Live Scripts

%%
close all
clc
%% first plot
figure(1)
x=linspace(-pi,pi);
y=sin(x);
hold on
y=cos(x);
plot(x,y,'r')
hold off
plot(x,y, 'r')
title('First Plot')
%% Second plot
figure(2)
[x,y]=meshgrid(linspace(-10,10));
z=sin(x).*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Second Plot')
%% third plot
figure(3)
[x,y]=meshgrid(linspace(-10,10));
z=x.*cos(y);
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Third Plot')
%% fourth plot
figure(4)
[x,y]=meshgrid(linspace(-10,10));
z=x.*y;
mesh(x,y,z)
xlabel('x'),ylabel('y'),zlabel('z')
title('Fourth Plot')

Note that you can also use sections for running/debugging code. This topic is only
covered briefly.

A.4.2 Using sections for running/debugging files

Using sections is for running portions of your code is especially useful. You can separate
out self-contained portions of your code and just run that piece. This is especially useful
for long homework assignments within one file; just run on problem at a time to see if
runs as expected. You can run the code within a section block several ways.
1. While the cursor is within the section you want to run, use the keystrokes of
Ctrl-Enter (Cmd-return on a Mac). Using the keystrokes of Ctrl-Shift-Enter (Cmd-
return-enter) will run the current section and put the cursor within the next section.
 Publishing and Live Scripts 223

2. Within the editor, select the “Run Section” or “Run and Advance” buttons on the
Editor Tab.
WARNING: when you run (evaluate) a section, the file may not be saved automatically
as it is when you run the entire m-file!

A.5. Formatting text

You can format your m-file by clicking on the “Publish” tab. By clicking on one of
the items, MATLAB will insert text into your m-file for that purpose. This includes
inserting text for things already discussed above, like inserting a section break, title, etc.
There are also buttons on the toolbar within the editor window for some of these items.
You also have the ability to customize your toolbar to add others.
Going to the menu and/or using the buttons on the toolbar can take extra time after
awhile, so it is also useful to know how to just type in the formatting.

A.5.1 Basic text formatting

In order to create a new line or new paragraph of text, have a blank comment line in
between the lines.

Link 10: publishMFile7.m (partial view)

%
% Text can be *bold*, _italic,_ and/or |monospaced|.
% One can also combine these formats like:
%
% _*BOLD, ITALIC TEXT*_
%

Text can be formatted to be bold, italic, monospaced, or combinations such as bold and
italic.

Link 10: publishMFile7.m (partial view)

%
%% Unordered (Bulleted) List
%
% * first item
% * second item blah blah
%
224 Publishing and Live Scripts

A.5.2 Lists
One can have an unordered, or bulleted list.

Link 10: publishMFile7.m (partial view)

% without a section title. You also must have a blank comment line to end the list.
Here's another example.
%%
%
% * item number 1
% * item number 2

Keep in mind that you must have a section break before the list, with or without a
section title. You also must have a blank comment line to end the list. Here is another
example.

Link 10: publishMFile7.m (partial view)

%
% # first item
% # second item blah blah
%
% As in the bulleted list, one must have a cell break, with or without a

You can also have an ordered (numbered list) using the same formatting as above, but
with “#” instead of “∗” for each list item.

Link 10: publishMFile7.m (partial view)

% section title. Second example without section title:
%%
%
% # blah blah blah blah blah blah blah blah blah blah blah blah blah blah blah blah
blah blah blah blah
% # yadda yadda yadda yadda yadda yadda yadda yadda yadda yadda yadda yadda yadda
yadda yadda yadda yadda yadda yadda yadda

As in the bulleted list, one must have a section break, with or without a section title.
Second example without section title:

Link 10: publishMFile7.m (partial view)

% You can have the links display the URL or display other text.
 Publishing and Live Scripts 225

%
% Here's an example of using the URL as
% the text of the link: <http://www.mathworks.com>
%

A.5.3 HTML links

You can have the links display the URL or display other text. URL as the link:

Link 10: publishMFile7.m (partial view)

% *LINKING THE M-FILE* It may be nice (required!) to link the M-File that
% was published to create the webpage.

You can have any text for the link:

Link 10: publishMFile7.m (partial view)

% <../publishMFile7.m M-File for this page>.
%

LINKING THE M-FILE: It may be nice (required!) to link the m-file that was
published to create the webpage. The easiest way to do it is like it is done below rather
than using the entire URL.

Link 10: publishMFile7.m (partial view)

% the file. The editor might automatically wrap the text if the URL
% and/or the text for the link is long. If this is the case, go back and
% make it one line.
%

The “../” before the filename means to go back one folder from where the HTML file
is located, which is where the m-file is located.
IMPORTANT: the text within the “<” and “>” must be on the same line within
the file. The editor might automatically wrap the text if the URL and/or the text for
the link is long. If this is the case, you must go back and make it one line. Otherwise,
the link will be broken!

A.5.4 Inserting images

Any figures that the MATLAB code creates will automatically be saved as PNG files
and inserted on the webpage. You can also include images like the following example.
226 Publishing and Live Scripts

Link 10: publishMFile7.m (partial view)

% into which the HTML file the M-file produced is located.
%% Preformatted Text
%

The image must be on its own line; no text can appear before or after it for the image
to be shown correctly on the page. Also note that the above is assuming the file for the
image is located in the “html” folder into which the HTML file the M-file produced is
located.

A.5.5 Pre-formatted text

Pre-formatted text is a way to display text on the webpage EXACTLY as it appears
in the editor, including extra spaces, line breaks in exactly the same place, etc. This is
commonly used for displaying lines of code in programming.
IMPORTANT: notice that the code below does not appear any different from plain
comments. In order for it to be pre-formatted, you must use the menu item to insert
the lines and change it. Here is what the menu item inserts:

Link 10: publishMFile7.m (partial view)

%
% preformatted text
% displayed
% exactly

Then you change the words “PREFORMATTED” AND “TEXT” and add lines if
necessary:

Link 10: publishMFile7.m (partial view)

% ------------------------------------------------------
% function y = myexample(x)
% % MYEXAMPLE(X) is a function for example purposed only
% %
% y = x.^2;
% ------------------------------------------------------
%
%
%% Inserting HTML Code
%
% <html>
% <table border="1" cellspacing="0" cellpadding="3"><tr><td>one</td><td>two</td></tr
></table>
% </html>
 Publishing and Live Scripts 227

A.5.6 Inserting HTML code

You can insert other HTML code (such as tables) into your document.

Link 10: publishMFile7.m (partial view)

%
% One can insert basic LaTeX commands and equations. One can have LaTeX
% code with the paragraph like
% $$ e^{\pi i} + 1 = 0 $$
% or on a separate line.

A.5.7 Inserting LATEX equations

Basic LATEX equations can be displayed on the webpage. Only basic mathematics can be
displayed when publishing to HTML mode, and some symbols do not display correctly.
For example, the not equals “=” symbol did not work in earlier versions, then it worked
in version 2012a, and so on. Technically speaking, when these equations are published
to an HTML file, each equation is saved as a PNG image (in the “html” folder or same
folder the HTML file is saved) and that image file is displayed on the webpage.

Link 10: publishMFile7.m (partial view)

%
%
%
% Notice that the "not equals" symbol may or may not work, depending on the version
of MATLAB!
% $$ 0 \ne 1 $$
%% M-file that created this page
% <../publishMFile7.m publishMFile7.m>

These images make the formulas appear blurry and/or small, and as mentioned above,
not every symbol will display correctly. Also, it could make the webpage slower to load
and would not be as accessible. Thus it may be better to use MathJax if creating a
webpage. If there is a need for a lot of LATEX or more complicated math typesetting,
consider publishing it as a LATEX file.
APPENDIX B

Final Projects
B.1. Ciphers
Create function(s) that create a substitution cipher encoder and decoder and a transpo-
sition cipher encoder and decoder, plus functions that demonstrate these ciphers. Some
useful MATLAB® functions: randperm, double, char, and reshape.

B.1.1 Substitution cipher

The encoder function will create a basic substitution cipher on an “extended alphabet”.
The encoder function will have two inputs: the first is string (entered in single quotes)
that is a filename for a text file in which the text may include numbers, characters, and
spaces. The second input will be a flag (either 1 or 0) as to whether the messages are
displayed.
You will use the command fileread(filename) to open the file and store the text
as one long vector of strings. Any characters that have an ASCII code of 32–126 are
allowed: see http://www.asciitable.com/. (ASCII stands for “American Standard Code
for Information Interchange.”) The encoder function will then encode the text using a
random substitution cipher. The function will return a string that is the encoded text
along with a matrix of characters that will be the key. The top row of the matrix will
be the original alphabet (represented by the ASCII code), and the second row will
be the substituted alphabet (represented by the ASCII code). Also, if the flag is true for
displaying the messages, the encoder function will display both the original and encoded
message to the screens. The display will not be as one long string; instead, the messages
will be formatted to be split across several lines so that no more than 50 characters are on
one line. If the flag is false, no output is displayed to the screen (unless the user chooses
to see the output of the function).
The decoder function will have three inputs; the key (which would be the matrix
that the encoder function gives you), the filename of the encoded text, and the flag for
displaying the messages to the screen. Using the key, the decoder function will return a
string of text that is the decoded message. If the flag is true for displaying the messages,
then both the encoded message and decoded message are displayed to the screen as in
the encoder function. If the flag is false, no output is displayed to the screen (unless the
user chooses to see the output of the function).
IMPORTANT: do not choose a specific substitution cipher—have MATLAB ran-
domly choose it so that each time you run the encoding function, you may get a
different encoded message (thus the need for the key to decode it).

229
230 Final Projects

To demonstrate the above functions, create another function that will create tables
that show the substitution cipher. It will as input take the “key” and create tables of the
character and what that character looks like in the encoded message. Have one table be
the characters (may split up into multiple tables for visual purposes), another table the
numbers, and one or two tables the letters (split between upper and lower? you will be
the judge). Thus your output may look like this (example cipher shown):
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
-----------------------------------------------------
D E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ]

a b c d e f g h i j k l m n o p q r s t u v w x y z
-----------------------------------------------------
d e f g h i j k l m n o p q r s t u v w x y z { | }

0 1 2 3 4 5 6 7 8 9
---------------------
3 4 5 6 7 8 9 : ; <

! " # $ % & ’ ( ) * + , - . / : ; < = > ? @ [ \ ] ^ _ ‘ { | } ~

-------------------------------------------------------------------
# $ % & ’ ( ) * + , - . / 0 1 2 = > ? @ A B C ^ _ ‘ a b c ~ ! "

This would signify that for every “A” in the original message, it is encoded as “D”,
every space becomes the character “#”, etc. Make these tables visually pleasing in the
output window (using fprintf, etc.).

B.1.2 Columnar transposition cipher

For the columnar transposition cipher, the encoding function has as input filename of
the original text (similar to above: characters, spaces, numbers, etc. are allowed), the
number of columns to use, and a flag for displaying the messages (similar as above). The
number of rows in the “matrix” is determined by the number of characters in the given
text; if there are not enough characters to completely fill the matrix, dummy (random)
characters are put at the end of the text to complete the matrix. The output will be the
encoded text and a key which is a vector that tells the order of the permuted columns.
The decoder function will have as input the filename that contains the encoded text, the
key (vector), and a flag for displaying the messages. The output is the decoded message.
 Final Projects 231

Another function will demonstrate the columnar transposition. It will take as input
the original message and the key. The function will display the original message, the
matrix of characters, the permuted matrix of characters, and the encoded message.
These displays should be formatted such that if the messages are long, then no more
than 50 characters are displayed on one line.
Your report should include a basic history of these types of ciphers, examples, pros
and cons to using these types of ciphers, and other variations of the ciphers.

B.2. Game of Pig

You will create a script file in which the user will use to play the game of Pig against
the computer. The script file will use the input command.
Rules of the game
Object: get to 100 points before your opponent.
A player’s turn involves rolling a standard six-sided die.
• If anything other than a 1 is rolled, the number rolled is added to the player’s
subtotal. The player can choose to roll again, or stop their turn and the subtotal is
added to the player’s total.
• If a 1 is rolled, the player’s turn is over and no points are added to the player’s total.
A running total of both player’s scores are always displayed.
For the user’s turn, your script will roll the die and display the number rolled. If the
number is 2 through 6, you will ask whether the user wants to roll again and a subtotal
is displayed, etc.
For the computer’s turn, if the number 2 through 6 is rolled, the computer will
have a certain algorithm that will determine whether to roll again. There may be an
initial probability p of rolling again or different probabilities under various conditions
of total scores and subscores, etc. Also, at each subsequent roll, the computer should
be less likely to roll again than before. You can choose your algorithm and value of
p and how the probability is reduced with each subsequent roll (as long as it fulfills
the requirements). In your final report you will explain how you chose the algorithm
and probability value(s). You do not need to get fancy with figuring out the “optimal
strategy” to figure out these numbers (that would be material for another class).
At the beginning of each turn, whether it is the user’s turn or the computer’s, the
screen is cleared and the both players’ totals are displayed (and continue to be displayed).
For each roll of the die, the roll is displayed, and the current subtotal (if applicable).
Thus you see each of the computer’s rolls within a turn. The values displayed should be
clearly labeled and easy to read (use of fprintf will be helpful here).
Your report should include some possible variations and possible improvements to
your game.
232 Final Projects

B.3. Linearization and Newton’s method

One application of differentiation is “linearization;” using the tangent line to a function
to approximate values of the function near the point at which the tangent line is calcu-
lated. Finding the linearization of f (x) at x = a is equivalent to finding the tangent line
to f (x) at x = a. You will create a function to help demonstrate linearization and to use
it to approximate values.

B.3.1 Linearization
Create a function called myLinear in which the function takes a string that is the
function f (x), and a value a. Using the commands syms, str2sym, subs, and diff,
your function will calculate the tangent line to y = f (x) at the point x = a. The function
will return the linearization L (x) = f  (a)(x − a) + f (a) (as a symbolic function).
In a separate script file that you may publish, you will use your function to do the
following problems.
√
1. Consider the function f (x) = 3 1 + x.
(a) Use your myLinear function to find the linearization of f (x) at the point
x = 0. Careful! You may have to use a different function than (1+x)^(1/3)
or nthroot(1+x,3) in order to make it work (surd?). Display the √
answer.
√
(b) Use myLinear function to approximate the values of 3 0.95 and 3 1.1.
(c) Graph the function y = f (x), the tangent line at x√= 0, and the points cor-
√
responding to the approximations of 3 0.95 and 3 1.1 on the same graph.
Create several graphs, zooming in to see the difference between the graph
and the tangent line at these points. Make sure you create a legend to make
things clearer.
2. Let f (x) = (x − 1)2 , g(x) = e−2x , and h(x) = 1 + ln(1 − 2x).
(a) Use your myLinear function to find the linearizations of f , g, and h at a = 0.
What do you notice? Why did this happen?
(b) Graph f , g, h, and the tangent lines on one graph. Create a legend in or-
der to tell which is which. For which function is the linearization a better
approximation (and for approximately what x-values)? Create several graphs,
zooming in, to support your answers.

B.3.2 Newton’s method

Another application of derivatives is Newton’s method for finding roots.
Create a function newton that uses Newton’s method to find an approximate solu-
tion to the equation f (x) = 0, for a given function f . The input should be:
• the function f (given as a string with x as a variable),
• an initial guess x0 for the solution,
 Final Projects 233

• the desired accuracy (ERROR CHECK: this number should be positive; if not, an
appropriate error message should be displayed using the error command),
• and the maximum number of iterations allowed (so it will stop if accuracy cannot
be reached) (ERROR CHECK: this number should be a positive integer; if not,
an appropriate error message should be displayed using the error command).
The function uses the Symbolic Math Toolbox (syms, diff, etc.) to convert the given
string for the function to a symbolic function and to calculate the derivative of the given
function f . Your HELP lines should make clear what the inputs are and the order. The
output will be the approximation to the solution of f (x) = 0. Your code will iterate
until the absolute value of the difference between the last two iterations is less than the
desired tolerance/accuracy OR the maximum number of iterations has been reached.
In either case, an appropriate message should be printed on the screen so the
user knows if desired accuracy has been reached or not. The message should include
how many iterations were completed. Use either disp or fprintf, and/or sprintf for
this message; experiment with this. The output of your function is the LAST iterate. In
other words, if x7 was calculated to determine that x6 is accurate enough, still output x7
and state that 7 iterations were calculated. If the derivative ever equals 0 at any iteration,
an appropriate error message should be displayed and the function will stop (use the
error command).
The following problems will demonstrate/use Newton’s method. The code for these
problems will be in the same script file as the problems on Linearization.
3. Consider the function f (x) = x4 − x − 1.
(a) Use your newton function to find x2 using x1 = 1 to find an approximation
to a root of f (x).
(b) Graph y = f (x), the tangent line at x1 = 1, and the tangent line at x2 to see
how the roots of each subsequent tangent line gets closer to the root. Make
sure you have a legend and appropriate axes to be able to see √ everything.
4. Use your function newton to find all roots to the equation earctan(x) = x3 + 1 correct
to eight decimal places (make sure all decimal places are displayed). In order to do
this, first create an appropriate plot to figure out what you are going to use for your
initial approximations of x1 for each root.
5. To demonstrate the importance of that first guess, consider the function f (x) =
x3 − x − 1.
(a) Use your function newton to find a root of the equation
√
correct to six decimal
places using an initial approximation of x1 = 1/ 3. State how many iterations
were needed or why Newton’s method did not work in this case.
(b) Use your function newton to find a root of the equation correct to six decimal
places using an initial approximation of x1 = 0.57. State how many iterations
were needed or why Newton’s method did not work in this case.
234 Final Projects

(c) Use your function newton to find a root of the equation correct to six decimal
places using an initial approximation of x1 = 0.6. State how many iterations
were needed or why Newton’s method did not work in this case.
(d) Use your function newton to find a root of the equation correct to six decimal
places using an initial approximation of x1 = 1. State how many iterations
were needed or why Newton’s method did not work in this case. √
(e) Graph the function y = f (x), and the tangent lines at x = 1/ 3, x = 0.57,
x = 0.6, and x = 1. Does the graph explain your answers to the above? Make
sure your graph has a legend and appropriate axes (or create a second graph
with appropriate axes) to see what is going on and support your claims.
NOTE: if any of the above return an error, you may need to comment out that
code so the rest of your script can run afterwards (keep the code in there to “show
your work” and mention something in the comments/text of your report).

B.4. Disk and Shell method

Create demonstrations of the Disk and Shell methods for volumes of revolution.
For the disk method, we will consider the solid of revolution by revolving the region
formed by y = f (x), y = 0, x = a, and x = b about the line y = k. For the shell method,
we will consider the solid of revolution by revolving the same region about the line
x = k. To make it easier, we will assume that the lines of revolution lie completely
outside or on the border of the revolved region.
Your function(s) will have as inputs the function f (x) (entered as a string), a, b, k and
the positive integer n. Your function should check that a < b and that either x or y is
given and that n is a positive integer. Your function(s) will do the following:
• Create a figure that shows the region and the line of revolution with the region
shaded/colored and an appropriate title.
• Form n rectangles of width x = (b − a)/n and height of each rectangle is f (x∗k ),
where x∗k is the midpoint of the kth subinterval. These rectangles along with the
original region will be graphed in a second figure.
• Revolve these rectangles to form the approximating disks (or approximating shells),
and plot these disks (shells) in 3D. You will also calculate the volume of these
approximating disks (shells) and display the answer both in the title of the plot
and as output of the function.
• Show the volume of revolution in 3D.
In all of the figures created, axes should be labeled, appropriate titles created, etc. They
should not appear jagged so appropriate domains/views will need to be defined.
Your report should show the use of these for various functions and/or values of n,
and different lines of revolution.
Some useful MATLAB commands are sphere and cylinder discussed in Sec-
tion 4.5.1.
 Final Projects 235

B.5. Power ball data

Practice exploring data with Powerball numbers. There are many things that can be
explored about the Powerball numbers that have been drawn. The rules of the Powerball
game have changed, and thus any of the statistics and graphs should reflect these changes.
An explanation of the these changes should be included in the report. Visualization of
the data is preferred and should be clear with titles, legends, etc. Many times the way
the data are given to us need to be cleaned up and modified before it is in a usable
format. Anything that needs to be done to accomplish this should be documented and
explained in the report.
Use the PowerballNumbers.txt data file to load the Powerball numbers drawn with
their dates. Then explore away! A common numerical summary of the columns (num-
bers drawn). Frequency, histograms, min, max, average, median, mode, etc. These may
be done based on the white ball order and overall. What are the differences or gaps
between the numbers in any drawing (average, min, max)? Is there anything interest-
ing when you look at the sums of the numbers drawn? Were any winning drawings
repeated? What about the white ball only combo? If so, how many? Were any draw-
ings in which the Powerball was the same number as one of the white balls? If so, how
many times? Is there any number not been selected? Have there been any consecutive
numbers drawn? If so, how many? Any time three consective numbers drawn? Etc.
Can you generate your own random Powerball tickets? How many times does it take
to generate a drawing that has already been a winner? How about you pick a specific
drawing date (like near your birthday...) and see how long it may take to generate that
winning drawing?
Use the PBJackpots.xlsx file to load the dates of winning jackpots and their amounts.
Matching the dates with the other data, is there anything different about the numbers
that end up being jackpots and the numbers that don’t? What about the summaries
about the size of the jackpots? The last rule change was supposed to achieve higher
jackpots. Does this appear to be true? If you graph the size of the jackpots over time, is
there a trend on spikes, etc.?
APPENDIX C

Linear Algebra Projects

C.1. Matrix calculations and linear systems
C.1.1 First handout
The following is the basic information given to the students for the introductory project
followed by problems that have been a part of this project. Additional information of
how the project is saved and submitted is omitted. The submissions can be given in the
form of a diary of the commands, a script file, published script file (to a webpage, PDF,
etc.) or a Live Script.

Entering a matrix

Rows of a matrix are separated by semicolons (;) and entries within a row are
separated by either commas or spaces. Brackets enclose the matrix. For example, you
could use either of the following commands to define the same matrix.
A=[1,2,-1,2,0;2,1,1,-1,0;3,-1,-2,3,0]
OR
A=[1 2 -1 2 0;2 1 1 -1 0;3 -1 -2 3 0].

Semicolons

Semicolons “;” at the end of a command suppress output to the command window.
See the difference with the above commands with and without a semicolon at the end
of the line.

Using your own functions

There are three functions that were written for some of the exercises below. You have
the ability to write your own, and may be asked to write functions later or in another
course. In order for these “hand-written” functions to run in MATLAB® ,
they need to be in the folder/directory you are working in. So you need to
pay attention to the “Current Folder” on the MATLAB screen above the editor and
command window and/or to the left of the screen. You will need to save these function
files in your own folder where you will be saving, running, and possibly publishing,
your SCRIPT or LIVE SCRIPT file from. There are some naming conventions to
your functions. For example, we cannot name a function sin.m since there is already a

237
238 Linear Algebra Projects

“sin” function in MATLAB. When MATLAB sees what appears to be a command to

run a function, MATLAB always checks for the function file of that name in its own
directories first, then the “Current Folder”. Thus our sin.m function file will never be
“found” to run. And if you are in a different folder from where you saved your function
file, it will not be found and an error will appear.

RREF

To get a matrix already entered and named A in reduced row echelon form, just
type
rref(A)
Notice that the answer may not give the answer in exact values. You could have
MATLAB do this by using the SYMBOLIC MATH TOOLBOX. (Most versions of
MATLAB will come with this toolbox.) This is done using the function sym. For
example, type
sym(rref(A))
Notice that the resulting output is now in exact value form. REMEMBER THIS;
we will be using exact values on multiple exercises.
Another way is to use the rats function or change the format of answers to format
rat. THIS IS ONLY USEFUL IF YOUR ANSWERS AND MATRICES WILL
ONLY HAVE INTEGER OR RATIONAL NUMBERS IN THEM. For example, if
you switch to format rat, notice what happens when you enter pi. You can display
your matrix fraction, rather than decimal form by entering the following:
format rat
rref(A)
Or, without changing the format, you could enter:
rats(rref(A))
To get back to the default format, enter format or format short. To test, you can
always enter pi and see if you get 3.1416.

HELP in MATLAB

If you know a command or function name, you can always get help for it by typing
help command. For example, you can type help rref. Also, if the function files are
copied over the current directory you can type help addrow to get some help for it.
For help with MATLAB defined functions and commands you can also use the Help
menu item at the top.
 Linear Algebra Projects 239

C.1.2 Exercises
These are exercises I’ve included on the first “introductory” project at various times.
I have not included all of them at once. Which exercises have been chosen and how
many are chosen depends on the textbook used and the timing of the assignment.
1. Consider the linear system below. Use MATLAB to solve the system by putting
it in RREF form but do each step separately (i.e., DON’T use the RREF com-
mand). Use the three functions swap, mult, and/or addrow found on the text
website. NOTE: In order to be able to use the functions within MATLAB, you
MUST copy the three files to your Current Folder within MATLAB.

x1 − x2 − 2x3 = −5,
6x1 − 5x2 − 7x3 = −3,
−2x1 − 6x3 = −44.

You may need to rerun your script file after each step to see what the next step
should be. In the end, you should have all steps needed to get it into RREF form.
Also, you will need to have exact values (thus the rats command may need to be
used—at every stage or at the end—you will be the judge).
2. Consider the linear system below. Use MATLAB to solve the system by putting
the augmented matrix in RREF form. You may use the rref command for this,
but make sure that exact answers are given.

x1 + 3x3 + 3x4 = −24,

x2 − 4x3 − 2x4 = 29,
3x1 − 3x2 + 24x3 + 15x4 = −171,
−x2 + 4x3 + 7x4 = −54.

3. Consider the linear system below. Use MATLAB to solve the system by putting
the augmented matrix in RREF form. You may use the rref command for this,
but make sure that exact answers are given.

x1 + x2 = −2,
x2 + x3 = 2,
x3 + x4 = 3,
x1 + x4 = −1.

4. Consider the linear system below. Use MATLAB to solve the system by putting
the augmented matrix in RREF form. You may use the rref command for this,
but make sure that exact answers are given.

4x1 − 3x2 + 3x3 + 4x4 =0,

240 Linear Algebra Projects

−x1 + x2 + 2x3 + 3x4 =5,

3x1 − 2x2 + 5x3 + 7x4 =5,
−3x1 + 3x2 + 6x3 + 9x4 =15.

5. Suppose a quadratic polynomial f (x) = ax2 + bx + c goes through the points (−1, 3),
(0, 2), and (2, 24). Using the points, write a system of linear equations to solve for
the unknown coefficients of f (x). Use MATLAB to solve this system and thus find
the polynomial by putting the augmented matrix in RREF form. You may use
the rref command for this, but make sure that exact answers are given.
6. Suppose a cubic polynomial f (x) is such that f (−1) = −6, f  (−1) = 2, f  (−1) =
−4, and f  (−1) = 6. Using the information about the derivatives, write a linear
system to solve for the unknown coefficients of f (x). Use MATLAB to solve this
system and thus find the polynomial by putting the augmented matrix in RREF
form. You may use the rref command for this, but make sure that exact answers
are given.
7. Suppose a quartic (degree 4) polynomial f (x) goes through the points (−2, −33),
(−1, 3), (0, 7), (1, 15), and (2, 15). Using the points, write a system of linear equa-
tions to solve for the unknown coefficients of f (x). Use MATLAB to solve this
system and thus find the polynomial by putting the augmented matrix in RREF
form. You may use the rref command for this, but make sure that exact answers
are given.
8. Consider the following matrices:
⎡ ⎤ ⎡ ⎤
10 10 10 10 0 0 1 0
⎢ 9 8 7 6 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 0 1 0 0 ⎥
A=⎢ ⎥, B=⎢ ⎥.
⎣ 5 4 3 2 ⎦ ⎣ 1 0 0 0 ⎦
1 1 1 1 0 0 0 1

Use MATLAB to calculate AB and BA.

9. In Exercise 8 above you calculated AB and BA. The matrix B is a special matrix
in that multiplication with this matrix gives a very specific result. Look closely at
your answers for AB and BA.
(a) Multiplying on the right by B (so looking at AB) results in an elementary
operation performed on the matrix A. Which elementary operation is it?
i. SwapRows. Two rows are swapped.
ii. SwapCols. Two columns are swapped.
iii. MultRow. A row is multiplied by a non-zero constant.
iv. MultCol. A column is multiplied by a non-zero constant.
v. AddRow. A multiple of one row is added to another row.
vi. AddCol. A multiple of one column is added to another column.
 Linear Algebra Projects 241

(b) Multiplying on the left by B (so looking at BA) results in an elementary

operation performed on the matrix A. Which elementary operation is it?
i. SwapRows. Two rows are swapped.
ii. SwapCols. Two columns are swapped.
iii. MultRow. A row is multiplied by a non-zero constant.
iv. MultCol. A column is multiplied by a non-zero constant.
v. AddRow. A multiple of one row is added to another row.
vi. AddCol. A multiple of one column is added to another column.
(c) Write a brief description specifically of what the elementary operations are
(which rows or columns? multiplied by what?). BE SPECIFIC!
10. Consider the following matrices:
⎡ ⎤ ⎡ ⎤
1 1 1 1 1 0 0 0
⎢ ⎥ ⎢ ⎥
⎢ 2 3 4 5 ⎥ ⎢ 2 1 0 0 ⎥
C=⎢ ⎥, D=⎢ ⎥.
⎣ 6 7 8 9 ⎦ ⎣ 0 0 1 0 ⎦
10 10 10 10 0 0 0 1

Use MATLAB to calculate CD and DC.

11. In Exercise 10 above you calculated CD and DC for the following matrices:
⎡ ⎤ ⎡ ⎤
1 1 1 1 1 0 0 0
⎢ ⎥ ⎢ ⎥
⎢ 2 3 4 5 ⎥ ⎢ 2 1 0 0 ⎥
C=⎢ ⎥, D=⎢ ⎥.
⎣ 6 7 8 9 ⎦ ⎣ 0 0 1 0 ⎦
10 10 10 10 0 0 0 1

The matrix D is a special matrix in that multiplication with this matrix gives a
very specific result. Look closely at your answers for CD and DC.
(a) Multiplying on the right by D (so looking at CD) results in an elementary
operation performed on the matrix C. Which elementary operation is it?
i. SwapRows. Two rows swapped.
ii. SwapCols. Two columns swapped.
iii. MultRow. A row is multiplied by a non-zero constant.
iv. MultCol. A column is multiplied by a non-zero constant.
v. AddRow. A multiple of one row is added to another row.
vi. AddCol. A multiple of one column is added to another column.
(b) Multiplying on the left by D (so looking at DC) results in an elementary
operation performed on the matrix C. Which elementary operation is it?
i. SwapRows. Two rows are swapped.
ii. SwapCols. Two columns are swapped.
iii. MultRow. A row is multiplied by a non-zero constant.
iv. MultCol. A column is multiplied by a non-zero constant.
242 Linear Algebra Projects

v. AddRow. A multiple of one row is added to another row.

vi. AddCol. A multiple of one column is added to another column.
(c) Write a brief description specifically what the elementary operations are
(which rows or columns? multiplied by what?). BE SPECIFIC!
For Problems 12 to 18, consider the chemical reaction given where a, b, c, and c
are unknown positive integers. The number of atoms of each element must be the
same before and after the reaction for the reaction to be balanced. For example,
because the number of oxygen atoms must remain the same, for Exercise 12 below
we have the equation
2a + b = 2c + 3d.
For each element in the reaction, we get an equation that together form a linear
system representing the chemical reaction. Many times there are not enough equa-
tions to get an exact solution; thus there may be infinitely many solutions overall
for the variables. Typically, one gives the solution that gives the smallest possible
positive integers. Balance the reactions using the smallest possible valid positive
integers for your answers.
(a) Write the system of equations to balance the chemical reaction.
(b) Write the augmented matrix used to solve this system of equations.
(c) Use the rref command on the augmented matrix within MATLAB to
find solution(s) to the system. You may want to use sym(rref(A)) so the
answers appear in exact form rather than numerical approximations. Write
the general solution to the system. Then, find the solution for balancing the
equation; that is, state the solution that uses the smallest positive integers
possible to balance the equation.
(d) Write the system of equations in the form of Ax = 0, stating what A and x
equal for this system. Use the “divide into” operator within MATLAB A\b
to solve the system Ax = b for x. Does this give you the desired answer?
(e) Use the commands null(A) and null(sym(A)).
12. The chemical reaction of nitrogen dioxide and water into nitrous and nitric acids:

aNO2 + bH2 O → cHNO2 + dHNO3 .

13. The chemical reaction of phosphorus pentoxide and calcium fluoride into phos-
phorus pentafluoride and tricalcium phosphate:

aP4 O10 + bCaF2 → cPF5 + dCa3 (PO4 )2 .

14. The chemical reaction of aluminum and water into aluminum hydroxide and hy-
drogen:
aAl + bH2 O → cAl(OH )3 + dH2 .
 Linear Algebra Projects 243

15. The chemical reaction of ammonia and oxygen into nitric oxide and water:

aNH3 + bO2 → cNO + dH2 O.

16. The chemical reaction hydrazine and dinitrogen tetroxide into nitrogen and water:

aN2 H4 + bN2 O4 → cN2 + dH2 O.

17. The chemical reaction of methane and oxygen into carbon dioxide and water:

aCH4 + bO2 → cCO2 + dH2 O.

18. The chemical reaction of ethane, carbon dioxide, and water into ethanol:

aC2 H6 + bCO2 + cH2 O → dC2 H5 OH .

C.2. The Hill cipher

The Hill cipher cannot be decoded using the same techniques as the basic substitution
or transposition ciphers. Those techniques include counting the number of times letters
are used to narrow down the code for common letters in the English language, etc. The
Hill cipher is such that the same letter may appear differently in each occurrence in the
message. Thus messages encoded using the Hill cipher can be very difficult to decode
by hand without knowing a “key”. We will not discuss the pros and cons of the Hill
cipher.
The Hill cipher takes strings of letters in a message, translates them into numbers
(A = 0, B = 1, . . ., Z = 25) and puts these numbers in a column of size m, creating
a matrix of size m × n (m = the number of columns) and multiplies them by an m × m
matrix called a key. If the message to encode does not fill up a column, you can use
“dummy letters” to complete the column. Once multiplied by the key, you end up with
a new m × n matrix that you then change back into your encoded message. This is done
by reducing the numbers MODULO 26 and translating the numbers back into letters.

2 3
Example C.2.1. Use the key K = to encode “Linear Algebra” using the
1 4
letter “Z” for any “dummy letters” that are needed.
Notice we have a 2 × 2 key so we will be splitting the letters up into pairs, so the
columns will be of size m = 2. We have an odd number of letters so we will use Z = 25
for the “dummy letter” at the end and we will have seven pairs (so n = 7).

LINEAR ALGEBRA = 11 8 13 4 0 17 0 11 6 4 1 17 0
244 Linear Algebra Projects

Thus the 2 × 7 matrix becomes


11 13 0 0 6 1 0
M= .
8 4 17 11 4 17 25

Multiply the key with this matrix:

  
2 3 11 13 0 0 6 1 0 46 38 51 33 24 53 75
KM = =
1 4 8 4 17 11 4 17 25 43 29 68 44 22 69 100
= E.

Now we need to translate this new 2 × 7 matrix back into letters. This is where
MODULAR arithmetic comes in. We have used modular arithmetic without even
realizing it; translating the 24-hour clock into the 12-hour clock is arithmetic modulo
12. We may also do this when giving change to a cashier to get only quarters back.
Here is a more formal mathematical definition: Let a, b, and n be integers. We say a is
equivalent to b modulo n, denoted a ≡ b (mod n), if a − b is an integer multiple of n.
When dealing with positive integers, the easiest way to think about it is, “What is the
remainder when we divide a by n? That is b.” So for instance, when I divide 16 by 12, the
remainder is 4 so 16 ≡ 4 (mod 12). When I divide 88 by 25, I get a remainder of 13 so
88 ≡ 13 (mod 25). For modulo 26, 26 ≡ 0 (mod 26), 27 ≡ 1 (mod 26), 28 ≡ 2 (mod 26)
and so on.
THERE IS A COMMAND IN MATLAB FOR MODULAR ARITHMETIC;
THIS WILL BE USEFUL!
Thus for the matrix E above we have

20 12 25 7 24 1 23
E≡ (mod 26),
17 3 16 18 22 17 22

which then transforms into the letters URMDZQHSYWBRXW. Thus “LINEAR ALGEBRA”
gets encoded into “URMDZQHSYWBRXW”. Notice how the letter “A” is encoded
differently each time in our message.

DECRYPTION:

Obviously we would want to be able to decrypt a Hill cipher. This would involve
solving KM = E where M is unknown but E is the encoded message we are given and
K was the original encryption key matrix. Thus, if K −1 exists and we knew what it
equaled, we would get M = K −1 E. So we would want K to be an invertible matrix,
or a nonsingular one. We also want K −1 to have entries that are integers, modulo 26.
 Linear Algebra Projects 245

There are other issues because of the modular arithmetic that we will not get into right
now. So how to do this? 
2 3
For the above key K = , we get
1 4

−1
4
− 35
K = 5 .
− 15 2
5

We do not want fractions so what are these fractions equivalent to modulo 26? The
easiest way to show you is by example. Remember 26 ≡ 0 (mod 26) so 26
5 ≡ 5 (mod 26).
0

Thus
4 26 30
+ = =6
5 5 5
and
−3 + 26 + 26 + 26 75
= = 15.
5 5
Thus 45 ≡ 6 (mod 26) and − 35 ≡ 15 (mod 26). So you get a decryption matrix D ≡
K −1 (mod 26) and for our example,

6 15
D= .
5 16

Notice (check it!), if we would be given the encoded message E above, we get

DE ≡ M (mod 26).

TIP: If converting fractions modulo 26 by hand, write out a bunch of multiples of 26.
By writing out a bunch of multiples, I saw that 3 · 26 = 78 so if I subtracted 3, I get a
nice multiple of 5, etc.
Notice that we are not encoding numbers, spaces or punctuation—this can be done
as well by adding to our “alphabet”. For example, we could have a space = 26, “.” =
27, and “?” = 28 and we would be working modulo 29.

C.2.1 Useful commands

In computers, ASCII codes represent text. For the upper-case English alphabet, the
codes are from 65–90. Thus the ASCII code for A is 65 and the ASCII code for Z is 90.
You can look up ASCII codes for other text and symbols but these are the only ones
we are using for this project. We are going to use the following commands to convert
between letters and numbers. There are also some useful commands to convert vectors
to matrices, etc. that will be useful to us that is discussed below.
246 Linear Algebra Projects

NOTE: Not all of the details of these commands are discussed here. Only an ex-
planation of what we need to know about the commands for our use for this project is
provided. Feel free to explore these commands further on your own!

DOUBLE
The command we need to convert our strings to ASCII codes is double. Here is an
example:

>> x='ABCD'
x =
'ABCD'
>> y=double(x)
y =
65 66 67 68

CHAR
The command to convert numbers that are ASCII codes to text is char. Here is an
example.

>> x=0:25; % creates a vector of the numbers from 0 to 25. Output is suppressed.
>> y=x+65 % adds 65 to every element in the vector so the is now from 65 to 90.
y =
Columns 1 through 18
65 66 67 68 69 70 71 72 73 74 75 76 77 78
79 80 81 82
Columns 19 through 26
83 84 85 86 87 88 89 90
>> alphabet=char(y) % alphabet is now a string that should all uppercase.
alphabet =
'ABCDEFGHIJKLMNOPQRSTUVWXYZ'

RESHAPE
We want to “reshape” a vector X to a matrix of size m × n. Then we use the command
reshape(X,m,n):

>> ex1 = reshape(alphabet, 2, 13)

ex1 =
2{\texttimes}13 char array
'ACEGIKMOQSUWY'
'BDFHJLNPRTVXZ'
>> A=[1 2 3 4]
A =
 Linear Algebra Projects 247

1 2 3 4
>> reshape(A,2,2)
ans =
1 3
2 4

MOD
We want to see what numbers are equivalent modulo 26. The command mod(x,26)
will do this. Let us look at some examples. For example, we know from clocks that 14
is equivalent to 2 modulo 12:

>> mod(14,12)
ans =
2

We can also use the mod command if we want the last four digits of a number. Use
mod(x,10000):

>> mod(987654321,10000)
ans =
4321

For us, we want matrices modulo 26. Luckily, the command works on vectors and
matrices.

>> x = [12, 30; 46 53]

x =
12 30
46 53
>> y=mod(x,26)
y =
12 4
20 1

Since 30, 46, and 53 are equivalent to 4, 20, and 1, respectively, modulo 26 we see
that y now has the values 12, 4, 20, and 1.

Using MATLAB for the Hill cipher example

What follows is an explanation of how to use MATLAB to do the work for us on the
Hill Cipher handout.
Our key is the matrix K:

K = [2 3;1 4];
248 Linear Algebra Projects

The numbers for our message are LINEARALGEBRA = 11 8 13 4 0 17 0 11 6 4

1 17 0. We could have figured this out on paper (which is error-prone) or we can use
MATLAB. For the project, we want to use MATLAB as much as possible.

>> msg='LINEARALGEBRA'
msg =
'LINEARALGEBRA'
>> msgNumber=double(msg)
msgNumber =
76 73 78 69 65 82 65 76 71 69 66 82 65

Notice that msgNumber is now a vector of the ASCII codes for our message. For the
Hill Cipher we want the numbers to be from 0 to 25 instead of 65 to 90. Thus we will
subtract 65 from our msgNumber.

>> msgNumber=msgNumber-65
msgNumber =
11 8 13 4 0 17 0 11 6 4 1 17 0

Now we need to get these numbers into the correct format. Notice that there are
13 letters in this message, we want 2 rows so we would have seven columns and thus
we need one “dummy letter” of Z = 25 at the end. I could add this at the end of
my msgNumber vector (if you know how to do this) or now that I’ve figured it out, I
just redo the above commands. In your script file you would just correct the original
commands you had.

>> msg='LINEARALGEBRAZ'
msg =
'LINEARALGEBRAZ'
>> msgNumber = double(msg)-65
msgNumber =
11 8 13 4 0 17 0 11 6 4 1 17 0 25

Notice that these numbers are the same as on the handout. But from these numbers
we need to create the 2 × 7 matrix M. Thus we use the reshape command:

>> msgNumberMtx=reshape(msgNumber,2,7)
msgNumberMtx =
11 13 0 0 6 1 0
8 4 17 11 4 17 25

Now we do the encoding. When we do the matrix multiplication we see that we

do not get numbers between 0 and 25 so we use the mod command (discussed above).
Then we transfer the matrix back to a vector and then letters. Notice that the encoded
message is the same as on the handout. I did not suppress any output so you could see
 Linear Algebra Projects 249

what each command does. You may want to suppress the output from some or all of
the intermediate steps so that you only see the end result.

>> encodedMsg=K*msgNumberMtx
encodedMsg =
46 38 51 33 24 53 75
43 29 68 44 22 69 100
>> encodedMsg=mod(encodedMsg,26)
encodedMsg =
20 12 25 7 24 1 23
17 3 16 18 22 17 22
>> encodedMsgNum=reshape(encodedMsg,1,14)
encodedMsgNum =
20 17 12 3 25 16 7 18 24 22 1 17 23 22
>> encodedMsgNum=encodedMsgNum+65
encodedMsgNum =
85 82 77 68 90 81 72 83 89 87 66 82 88 87
>> codedMsg=char(encodedMsgNum)
codedMsg =
'URMDZQHSYWBRXW'

Finding the inverse of the key

There are several ways one can find an inverse of a matrix in MATLAB. We could do it
in MATLAB the way we do it my hand or by using the command inv. We are safe to
use the rats command since we know our original matrix has only integer entries. It
would NOT be safe if our original matrix had irrational entries.

>> D=inv(K)
D =
0.8000 -0.6000
-0.2000 0.4000
>> rats(D) % this is for display purposes only
ans =
2{\texttimes}28 char array
' 4/5 -3/5 '
' -1/5 2/5 '

Now we need to figure out what these fractions are modulo 26.

>> a=26/5; % Didn't want to type 26/5 over and over

>> D(1,1)+a % Ran once, saw it was an integer between 0 and 26 so done
ans =
6
>> D(1,1)=ans % Reassigning the 1,1 entry of D to be the above answer
D =
6.0000 -0.6000
250 Linear Algebra Projects

-0.2000 0.4000
>> D(1,2) + a + a + a % Re-ran this, adding "a" each time until get desired integer
ans =
15
>> D(1,2)= ans % Reassigning the 1,2 entry of D to be the previous answer
D =
6.0000 15.0000
-0.2000 0.4000
>> D(2,1)+a
ans =
5
>> D(2,1)=ans
D =
6.0000 15.0000
5.0000 0.4000
>> D(2,2)+a+a+a
ans =
16
>> D(2,2)=ans
D =
6 15
5 16

Now we have the inverse matrix, modulo 26 as in the handout. Let us check against
our encoded message. Output of the intermediate steps is suppressed.

>> codedMsg
codedMsg =
'URMDZQHSYWBRXW'
>> numCode2 = double(codedMsg);
>> numCode2=numCode2-65; % convert to numbers from 0 to 25
>> E2=reshape(numCode2,2,7) % reshape to 2x7 matrix
E2 =
20 12 25 7 24 1 23
17 3 16 18 22 17 22

Notice E2 is the same as E in the above example. Now decode and then convert to
letters:

>> chkMsg=D*E2;
>> chkMsg=mod(chkMsg,26);
>> chkMsg=reshape(chkMsg,1,14);
>> chkMsg=chkMsg+65;
>> char(chkMsg)
ans =
'LINEARALGEBRAZ'
 Linear Algebra Projects 251

We may have been told that the dummy letter(s) that would be used would be Z, or
we see that the entire decoded message does not make sense as given and recognize that
the original message was “LINEAR ALGEBRA”.

C.2.2 Exercises
Complete these problems within MATLAB, showing all work, even calculations done,
within MATLAB.
1. Consider the Hill cipher with the key

2 5
.
1 4

(a) Encode the following text, adding Z if necessary, stripping spaces and punc-
tuation:
Step on no pets.
(b) Calculate the “inverse matrix” (decode matrix) with entries that are integers
from 0 to 25 (integers modulo 26).
(c) Using the decode matrix above, decode the following message, stripping off
any extraneous “dummy letters” that may have been put at the end. In other
words, strip off any extraneous “dummy letters” so that the decoded message
is an actual word or phrase in English.
SIEBZY
2. Consider the Hill cipher key in Exercise 1 above and its decode matrix.
(a) Use the key to encode “Name no one man,” adding Q if necessary, stripping
spaces and punctuation.
(b) Use the decode matrix to decode “IRGDTF” stripping off any extraneous
“dummy letters” that may have been put at the end.
3. Consider the Hill cipher with the key

1 5
.
2 7

(a) Encode the following text, adding X if necessary, stripping spaces and punc-
tuation:
Madam, I’m Adam.
(b) Calculate the “inverse matrix” (decode matrix) with entries that are integers
from 0 to 25 (integers modulo 26).
(c) Using the decode matrix above, decode the following message, stripping off
any extraneous “dummy letters” that may have been put at the end. In other
words, strip off any extraneous “dummy letters” so that the decoded message
is an actual word or phrase in English.
FYPSNE
252 Linear Algebra Projects

4. Consider the Hill cipher key in Exercise 3 above and its decode matrix.
(a) Use the key to encode “Was it a rat I saw?,” adding Q if necessary, stripping
spaces and punctuation.
(b) Use the decode matrix to decode “KUYWFN” stripping off any extraneous
“dummy letters” that may have been put at the end.

C.3. Least-squares solutions

C.3.1 Brief overview
This should supplement a section in a linear algebra text on least-squares solutions as
many details are not provided here.
When the system Ax = b has no solution, we may want to find an approximation
to a solution. This is usually when we have too many equations, and the system is
overdetermined. Since Ax − b = 0 for any x, we want to look at approximate solutions

x and A
x − b = e, the “error” vector. If we make the length of e as small as possible,
we’ve minimized the error and have a “best approximation” to a solution to Ax = b.
This is what a least-squares solution is; the vector

x that makes

e = A

x − b

as small as possible.
Without getting into the details, we find the least-squares solution

x to Ax = b by
solving the system A Ax = A b.
T T

Example C.3.1. Consider the system

5x + 6y = 3,
6x + 7y = 1,
x + y = −5.

>> A=[5 6;6 7;1 1]; b=[3;1;-5];

>> M=[A b]
M =
5 6 3
6 7 1
1 1 -5
>> rref(M)
ans =
1 0 0
0 1 0
0 0 1
 Linear Algebra Projects 253

We see above that the system has no solutions. But the least-squares solution is the
solution to AT Ax = AT b:

>> B=A'*A; c=A'*b;

>> N=[B c]
N =
62 73 16
73 86 20
>> rref(N)
ans =
1 0 -28
0 1 24

So the least-squares solution to Ax = b is


−28

x= .
24

C.3.2 Curve fitting

A common use for least-squares solutions is curve fitting.

Example C.3.2. Suppose we want a parabola (quadratic function) that best fits these
data points:
(−1, 1) (3, 0) (0, 1) (−2, −2) (2, 3)
By looking at f (x) = ax2 + bx + c we get the following equations:

f (−1) = 1 =⇒ a−b +c = 1,
f (3) = 0 =⇒ 9a + 3b +c = 0,
f (0) = 1 =⇒ c = 1,
f (−2) = −2 =⇒ 4a − 2b +c = −2,
f (2) = 3 =⇒ 4a + 2b +c = 3.

These equations form a linear system Ax = b:

>> A=[1 -1 1;9 3 1;0 0 1;4 -2 1;4 2 1];

>> b=[1;0;1;-2;3];
>> M=[A b]
M =
1 -1 1 1
9 3 1 0
0 0 1 1
4 -2 1 -2
254 Linear Algebra Projects

4 2 1 3
>> rref(M)
ans =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
0 0 0 0

We see above that there is no solution. We find a least-squares solution:

>> rref([A'*A A'*b])

ans =
1.0000 0 0 -0.5000
0 1.0000 0 1.0000
0 0 1.0000 2.0000
>> rats(rref([A'*A A'*b]))
ans =
3{\texttimes}56 char array
' 1 0 0 -1/2 '
' 0 1 0 1 '
' 0 0 1 2 '

⎡ ⎤
− 12
⎢ ⎥
So the least-squares solution is

x=⎣ 1 ⎦ and thus the quadratic function that best

2
1
fits the data is f (x) = − x2 + x + 2 (see Fig. C.1).
2

C.3.3 Exercises
All of the work to answer the following exercises must be done in MATLAB and appear
in your file. Even basic calculations that you would normally need scratch paper or your
calculator for should be done in MATLAB.
1. Consider the following system:
⎡ ⎤ ⎡ ⎤
2 −2 15
⎢ ⎥ ⎢ ⎥
⎣ −2 2 ⎦ x = ⎣ −9 ⎦ .
5 5 10

(a) Use MATLAB to try and solve the system and notice why we may need
least-squares.
(b) Use MATLAB to find the least-squares solution

x of the system.
 Linear Algebra Projects 255

Figure C.1 Data with curve fitting.

2. Consider the following system:

⎡ ⎤ ⎡ ⎤
1 1 −1 7
⎢ 1 ⎥ ⎢ ⎥
⎢ 1 1 ⎥ ⎢ 7 ⎥
⎢ ⎥x = ⎢ ⎥.
⎣ 1 −1 1 ⎦ ⎣ −7 ⎦
1 −1 −1 5

(a) Use MATLAB to try and solve the system and notice why we may need
least-squares.
(b) Use MATLAB to find the least-squares solution
x of the system.
3. Consider the linear function f (x) = mx + b with the data points (−2, 8), (0, 4), and
(2, 12).
(a) Come up with a linear system to solve for m and b.
(b) Use MATLAB to try and solve the system and notice why we may need
least-squares.
(c) Use MATLAB to find the least-squares solution
x of the system.
(d) BONUS: plot the data points and the line y = mx + b on the same figure in
MATLAB.
4. Consider the quadratic polynomial f (x) = ax2 + bx + c with the data points (0, −1),
(1, 0), (2, −7), and (3, −2).
(a) Come up with a linear system to solve for a, b, and c.
(b) Use MATLAB to try and solve the system and notice why we may need
least-squares.
(c) Use MATLAB to find the least-squares solution
x of the system.
256 Linear Algebra Projects

(d) BONUS: plot the data points and the line y = f (x) on the same figure in
MATLAB.

C.4. Markov matrices

C.4.1 Brief overview
One use for eigenvalues and eigenvectors is with Markov matrices. There are many
texts on this subject; for example, Section 2.8 in Ref. [27] discusses Markov matrices in
more detail with several examples. There is a lot to study about Markov matrices and
Markov Chains; we will only introduce the idea.
In our application we have a Markov matrix without going into detail of the
definition of a Markov matrix. A Markov matrix is also called a stochastic matrix, a
transition matrix, a probability matrix, or a substitution matrix. The type of Markov
matrix we will be looking at is a matrix in which:
• All entries of the matrix are nonnegative.
• The entries of each column add up to 1.
In our examples, the entries of the Markov matrix are probabilities that something
will transition from one state to another. Thus the entry pij is the probability that of
transitioning from state i to state j in one “generation.”
Consider the Markov matrix

p p
P = 11 12
p21 p22

where one “generation” is one day. Then the value of p11 is the probability someone
in group A will stay in group A the next day, p12 is the probability someone in group
A will go to group B, p21 is the probability that someone in group B will go to group
A the next day, and finally p22 is the probability that someone in group B will stay in
group B the next day.
In this project we will look at two examples and investigate the relationship between
the Markov matrices, eigenvalues, and eigenvectors. We will also have to remember
how to calculate limits from calculus. It may be useful to remind yourself of the limit of
a geometric sequence.
There will be work shown in your MATLAB file and maybe some work on paper to
do these problems. There are examples of how to work with the characteristic function
and/or find eigenvalues on the text website.

C.4.2 Exercises
All of the work to answer the exercises below must be done in MATLAB and appear in
your file. Even basic calculations that you would normally need scratch paper or your
 Linear Algebra Projects 257

calculator for should be done in MATLAB. USING CALCULATORS, MATH-

EMATICA, MAPLE, WOLFRAM ALPHA, ETC. IS NOT ALLOWED. Any
written work should be handed in.
1. At the end of every month a manager rates the performance of each member of
their staff as poor, average, or excellent. If a worker was rated poor at the end of
one month, then the probability that the next month the worker will be rated poor
is 0.15, average is 0.7 and excellent is 0.15. If a worker was rated average at the end
of one month then probability that the next month the worker will be rated poor
and average is 0.2 and 0.5, respectively. If a worker was rated excellent at the end of
one month then the probability that the next month the worker will be rated poor
is 0.05 and average is 0.75.
Complete the construction of the Markov matrix A below for the above informa-
tion. The first column of the matrix are the probabilities of being rated poor at the
end of one month and then rated the next month as poor, average and excellent,
respectively. Thus the first column of the matrix looks like
⎡ ⎤
0.15
⎢ ⎥
⎣ 0.7 ⎦ .
0.15

The second column of A has the probabilities of being rated average at the end
of one month and then being rated the next month as poor, average and excellent,
respectively and so on. Recall that probabilities should add up to 1, that is that at the
end of any given month, the probabilities of being rated poor, average, or excellent
should add up to 1.
Find A. If you need calculations done to figure out the entries, those should be
done in MATLAB. At the minimum, the only thing needed to be done in the
MATLAB is to enter the matrix, named A.
2. Use the Markov matrix A from the previous exercise.
(a) Suppose that initially, 10% of the workers were rated poor and 75% of the
workers were rated as average. Complete the entries for the initial state vector
x0 , stating the values as percentages (thus 16 rather than 0.16 for 16%).
(b) Use MATLAB to find the percentage of workers will be rated poor after 5
months. Round your answer to two decimal places.
3. Use the Markov matrix from Exercise 1.
(a) Find the characteristic polynomial p(x) of that matrix:
(b) Given that a Markov matrix always has an eigenvalue of λ = 1, find the other 2
eigenvalues with λ2 < λ3 . USING CALCULATORS, MATHEMATICA,
MAPLE, WOLFRAM ALPHA, ETC. IS NOT ALLOWED. State the
exact values of the eigenvectors.
258 Linear Algebra Projects

4. Suppose you had the following Markov matrix instead of the one given in the
previous exercises.
⎡ ⎤
0.2 0.3 0.1
⎢ ⎥
⎣ 0.7 0.4 0.7 ⎦ .
0.1 0.3 0.2
3 1
The eigenvalues for this Markov matrix are λ1 = 1, λ2 = − , and λ3 = , with
10 10
corresponding eigenvectors
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 1 −1
⎢ 7 ⎥ ⎢ ⎥ ⎢ ⎥
u 1 = ⎣ 3 ⎦ , u 2 = ⎣ −2 ⎦ , u 3 = ⎣ 0 ⎦ ,
1 1 1

respectively. Notice that the matrix is NOT defective. Recall the theorem in linear
algebra that says these eigenvectors are linearly independent. Thus these vectors
form a basis for R3 so if our initial state vector was
⎡ ⎤
15
⎢ ⎥
x0 = ⎣ 75 ⎦ ,
10

we can write this initial state vector as a linear combination of the three eigenvec-
tors. In other words, we can write

x0 = a1 u1 + a2 u2 + a3 u3 .

(a) Use MATLAB to solve for the coefficients a1 , a2 , and a3 .

(b) Write the exact values for these coefficients.
(c) Since we have a difference equation xk = Ak x0 , we can write

xk = Ak (a1 u1 + a2 u2 + a3 u3 ).

Simplify the right hand side of this equation to be in terms of the coefficients
above, eigenvalues, and eigenvectors. Use the exact values for the coefficients
and eigenvalues but keep u1 , u2 , u3 in the equation.
(d) Write the resulting equation.
(e) The above equation can make calculations easier, and thus faster. Based on
this equation, use MATLAB to find the percentage that was rated as average
after 30 months, both the exact value and numerical approximation rounded
to four decimal places.
 Linear Algebra Projects 259

(f) Based on the equation from the second part (with the exact values), figure
out what the lim xk equals using exact values. What is your answer rounded
k→∞
to four decimal places? (use MATLAB)
(g) Based on the meaning behind the Markov matrix and vectors x0 and xk ,
explain what this limit means for this particular application.
APPENDIX D

Multivariable Calculus Projects

D.1. Lines and planes
Copy the template script file to your own folder. Change the equations, etc. to be the
equations for your problems. You may need to adjust the DOMAINS to get better
pictures! For # 2 and # 3: you are creating a 3D picture of these lines, planes, etc.
Then you may use the function AnimateView found on the text website to create an
animated GIF file of “spinning” the figure around.
1. Consider the line L (t) = 2 + t, 7 + 5t.
(a) What is the point where the line L intersects the x-axis? What is the value of
t for that point?
(b) What is the point where the line L intersects the y-axis? What is the value of
t for that point?
(c) At what point(s) does the line L intersect the parabola y = x2 ? What is the
value of t for the point(s)?
Plot the line L, the points of intersection and the parabola y = x2 on the same figure
in MATLAB® . Create an appropriate legend and title.
2. Consider the two lines L1 : −2t, 1 + 2t, 3t and L2 : −5 + s, 2 + 3s, 2 + 4s.
Find the point of intersection of the two lines. Plot the two lines and the point of
intersection on the same figure in MATLAB.
BONUS: create an animated GIF that by using the AnimateView function found
on the text website.
3. Consider the plane −0.5x − 2.5y + z = 1.
(a) Find the parametric equations for the line through the point P = (5, 3.75,
2.75) that is perpendicular to the plane where the parameter t = 0 should
correspond to the point P. The direction vector of the line should be the
same as the standard normal vector of the plane.
(b) Find the point Q where the line intersects the yz-plane.
Plot the plane and line in MATLAB on the same figure. Also mark points P and Q
on the figure.
BONUS: create an animated GIF that by using the AnimateView function found
on the text website.

261
262 Multivariable Calculus Projects

D.2. Vector functions

D.2.1 2D example plots
In 2D, vector functions are not different from the parametric equations you saw in single
variable calculus. To plot them, you must first establish your domain for t. The easiest
way to do this is with the command linspace.

t = linspace(0,2*pi);

IMPORTANT: notice the semi-colon at the end of these lines; if there is none, all
of those values for t (and x, y, etc. below) will appear in your command window. Now
we enter the equations for x and y, using component-wise calculations.

x = 3*cos(t);
y = 2*sin(t);

Now we can plot the x and y values. For 2D plots, it is simple:

plot(x,y)
title('2D Vector Function Example 1')

The above  equations did not need component-wise calculations. This example
sin(2t) cos(2t)
does: r(t) = ,
4 + t2 4 + t2
t = linspace(0,10);
x = sin(2*t)./(4 + t.^2);
y = cos(2*t)./(4 + t.^2);
plot(x,y)
title('2D Vector Function Example 2')

D.2.2 3D example plot

This example involves component-wise calculations, and the domain needed to be
adjusted to show the entire trefoil knot. Notice for a 3D space curve, the command is
plot3 instead of plot. It is always good to label the axes for perspective.

t=linspace(0,4*pi);
x=(2+cos(1.5*t)).*cos(t);
y=(2+cos(1.5*t)).*sin(t);
z=sin(1.5*t);
plot3(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
title('3D Vector Function/Space Curve Example')
 Multivariable Calculus Projects 263

D.2.3 Bad domain example

The following is an example in which the domain does not have enough points and it
creates a jagged curve (try it yourself!) that should be smooth.

t=linspace(0,4*pi);
x=cos(6*t);
y=sin(6*t);
z=t;
plot3(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Bad Domain Example')

The problem above was because by default the linspace(a,b) command creates
one hundred values between a and b for MATLAB use to connect to plot the graph.
By having linspace(a,b,n), you are specifying n values between a and b. So below,
we have t equal a vector of 500 values between 0 and 4π with which to create the x, y,
and z values to build points to connect into a graph.

t=linspace(0,4*pi,500);
x=cos(6*t);
y=sin(6*t);
z=t;
plot3(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Better Domain Example')

D.2.4 Adjusting the view

You can adjust the view. Using this command may involve experimentation with the
values and rerunning the script.

t=linspace(0,4*pi);
x=(2+cos(1.5*t)).*cos(t);
y=(2+cos(1.5*t)).*sin(t);
z=sin(1.5*t);
plot3(x,y,z)
view(0,90)
xlabel('x'), ylabel('y'), zlabel('z')
title('Adjusting View on 3D Graph')

D.2.5 Sphere command

The sphere command will set up variables x, y, and z to form a unit sphere using the
surf or mesh commands. Without going into details, here are examples.
264 Multivariable Calculus Projects

[x,y,z]=sphere(100);
mesh(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere Command: Unit Sphere')
axis equal
axis([-5 5 -5 5 -5 5])

How to get one of a different radius (r = 3) than 1:

[x,y,z]=sphere(100);
mesh(3*x,3*y,3*z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere with different radius')
axis equal
axis([-5 5 -5 5 -5 5])

How to get a sphere with a different center C (2, −1, 3) from the origin:

[x,y,z]=sphere(100);
mesh(x+2,y-1,z+3)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere with center not at the origin')
axis equal
axis([-5 5 -5 5 -5 5])

QUESTION: How could you get ANY sphere; a different center from the origin
and radius other than 1?
Here they are side-by-side for comparison:

subplot(131)
mesh(x,y,z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere Command: Unit Sphere')
axis equal
axis([-5 5 -5 5 -5 5])
subplot(132)
mesh(3*x,3*y,3*z)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere with different radius')
axis equal
axis([-5 5 -5 5 -5 5])
subplot(133)
mesh(x+2,y-1,z+3)
xlabel('x'), ylabel('y'), zlabel('z')
title('Sphere with center not at the origin')
axis equal
axis([-5 5 -5 5 -5 5])
 Multivariable Calculus Projects 265

D.2.6 Multiple plots on one figure

You must use the hold on and hold off commands around the additional plot/plot3
/mesh, etc. commands for all to show up within the same figure.

t = linspace(-pi,3*pi);
x = 2*cos(t);
y = sin(t);
z = t;
plot3(x,y,z, 'k')
xlabel('x'), ylabel('y'), zlabel('z')
title('Multiple Graphs in One Example')
hold on % use this to add more plots to current figure
t2 = linspace(-1,1);
x2=-2*t2;
y2=1 + 0*t2;
z2=pi/2 + t2;
plot3(x2,y2,z2)
hold off % make sure you have this at the end

D.2.7 Exercises
1. Consider the vector function with parametric equations

x = cos(t) 4 − 0.25 cos2(10t) + 5,

y = sin(t) 4 − 0.25 cos2(10t) + 9,
z = 0.5 cos(10t) − 8.

(a) Show that, for any t, the space curve from these parametric equations lies on
a sphere by finding the equation of the sphere, showing all work on paper.
Hint: using these equations for x, y, and z, can you get something in the
form of (x − h)2 + (y − k)2 + (z − l)2 = r 2 ? Write the equation of the sphere.
What is the center and radius of the sphere?
(b) Within MATLAB, plot the parametric equations to make the space curve.
Make sure your domain is defined nicely so you get the entire graph and it is
not a jagged curve.
(c) Within MATLAB, create a second plot of the space curve and the sphere
(using the sphere command) so they appear in the same figure. Make the
space curve black and thicker than the default.
2. Consider the vector function with parametric equations

x = cos(π t),
y = sin(π t),
266 Multivariable Calculus Projects

z = 3 sin(π t).

(a) Find the vector form of the equation of the tangent line to the curve at t = 1.
(b) Find the vector form of the equation of the tangent line to the curve at
t = 1/2.
(c) Find the point of intersection of these two lines.
(d) Within MATLAB, plot the parametric equations, the tangent lines and the
point. Make sure your domain is defined nicely so you get the entire graph
and it is not a jagged space curve. Make the space curve in black, the first
tangent line in blue, the second in red, and mark the point of intersection
with a black x. 
3. Consider the surface f (x, y) = 4x2 y + 15 + 2xe1−y ln(x).
(a) Find the equation of the tangent plane to when x = 1 and y = 1. Show your
work on paper.
(b) Within MATLAB, graph the surface z = f (x, y) and the tangent plane, mark-
ing the point (1, 1, f (1, 1)). Make the surface yellow, the tangent plane or-
ange, and the point black. Make the domain for x to be from 0.000001 to 3
and the domain for y from 0 to 3.
BONUS: Use the AnimateView2 function found on the text website to cre-
ate an animated gif picture.
(c) Notice that the above question could have been phrased “find the linear
approximation L (x, y) of the function f (x, y) = ... at ...” and the answer would
have been the same.
Use your answer above to approximate f (1.01, 0.99) by calculating this
approximation within MATLAB. Name the approximation calculation
fApprox within MATLAB.
(d) Have MATLAB calculate the actual value of f (1.01, 0.99) (answer given to
four decimal places by default). Name this calculation of f fActual within
MATLAB.

D.3. Applications of double integrals

D.3.1 Calculating integrals and viewing regions
For these exercises, we will use the Symbolic Math Toolbox within MATLAB to cal-
culate integrals.
 2  1
For example, to calculate (x2 + y2 ) dy dx we can enter
−2 0

>> syms x y
>> int1 = int(x^2 + y^2, y, 0, 1)
int1 =
 Multivariable Calculus Projects 267

x^2 + 1/3
>> int2 = int(int1, x, -2, 2)
int2 =
20/3

Consider the tetrahedron bounded by x = 0, y = 0, z = 0, and z = 10 − 4x − 2y. The

following code allows the solid to be viewed in MATLAB.
First, we can see view the projection of the solid in the xy-plane.

x=linspace(0,2.5);
y=5-2*x;
fill([x,0],[y,0], [0.75, 0.75, 0.75])
axis([0,2.5,0,5.5]) % adjust the axis([xmin, xmax, ymin, ymax]) if necessary
title('Projection of Tetrahedron onto xy-plane')

The following code allows us to view the solid.

xdomain = linspace(0,2.5,81); % adjust if necessary

ydomain = linspace(0,5,81); % adjust if necessary
[x,y]=meshgrid(xdomain, ydomain);
z=10-4*x-2*y;
%%%
% BONUS IF CAN MODIFY BELOW APPROPRIATELY TO GET YOUR CORRECT 3-D REGION
% IF NOT, DELETE THIS SECTION
L=size(z);
for m=1:L(1)
for n=1:L(2)
if (y(m,n) > (5-2*x(m,n)))
y(m,n) = 5-2*x(m,n);
end
end
end
z=10-4*x-2*y;
% END OF BONUS SECTION
%%%
colormap('Summer') % defines color
C=0*z - 100; % color for tetrahedron
meshz(x,y,z,C)
hold on
mesh(x,y,0*x,'EdgeColor',loygray)
hold off
xlabel('x'),ylabel('y'),zlabel('z')
view(65,20) % adjust if necessary
268 Multivariable Calculus Projects

D.3.2 Exercises
Use MATLAB to calculate the integrals needed to answer the following questions. Show
all work on paper any calculations needed to set them up. Sketch the regions involved.
1. Find the mass and center of mass of the lamina that occupies the region D bounded
√
by the parabolas y = 81 x and x = 9y2 and has density function ρ(x, y) = x. Plot
1 2

the figure using MATLAB.

2. A lamina occupies the region inside the circle x2 + y2 = 4y but outside the circle x2 +
y2 = 4. Find the center of mass if the density at any point is inversely proportional to
its distance from the origin. (Hint: it may be useful to convert to polar coordinates.)
3. Consider the function

4
25 xy 0 ≤ x ≤ 5, 0 ≤ y ≤ 1,
f (x, y) =
0 otherwise.

(a) Verify the function f (x, y) is a joint density function.

(b) If X and Y are random variables whose joint density function is the function
 
f above, find P X ≥ 52 .
 
(c) Find P X ≥ 52 , Y ≤ 12 .
(d) Find the expected values of X and Y .
4. Consider the mass of the solid tetrahedron bounded by the xy-plane, the yz-plane,
the xz-plane, and the plane x/6 + y/5 + z/30 = 1.
(a) Set up a triple integral to find the mass of the solid tetrahedron if the density
function is given by ρ(x, y, z) = x + y. Write an iterated integral in this form
to find the mass of the solid:
  B D F
m= f (x, y, z) dV = dz dy dx
A C E
R

specifying the integrand and limits of integration A, B, C, D, E, and F.

(b) Use MATLAB to calculate the mass m.
(c) Use MATLAB to calculate the center of mass (your answer should be a point).
(d) BONUS: Use MATLAB to draw the projection of the solid onto the
xy-plane.
(e) BONUS: Use MATLAB to draw the 3D solid.
5. A family of surfaces ρ = 1 + 15 sin(mθ )s sin(nφ), where m and n are positive integers
are known as “bumpy spheres.” They can be used to model tumors. We will work
with the bumpy sphere
1
ρ =1+ sin(7θ ) sin(8φ).
5
 Multivariable Calculus Projects 269

(a) Use MATLAB to graph the bumpy sphere using spherical coordinates and
then converting to rectangular coordinates to get x, y, z defined. This can
be done by setting up the proper domains for θ and φ using meshgrid,
calculating ρ and then creating x, y, and z using the sph2cart command.
Use mesh, making the EdgeColor the color of your choice.
(b) Write on paper the integral that would be used to calculate the volume of
this bumpy sphere.
(c) Use MATLAB to calculate the exact value of the volume of the bumpy
sphere.
(d) Write the answer MATLAB gives for the volume on paper, writing it in cor-
rect notation.
√
For example, if MATLAB gives the answer as pi∗sqrt(3)/2,
write π 2 3 . 
6. Consider the integral f (x, y, z)dV where E is the solid bounded by z = 0, x =
E
0, z = y − 8x and y = 24.
(a) Use MATLAB to draw the 3D solid.
(b) It may be helpful to sketch on paper the projections of the solid onto the
2D coordinate planes to help change the order of the iterated integrals. You
can get help for these sketches by adjusting the view of the 3D solid within
MATLAB. 
(c) Express the integral f (x, y, z)dV as an iterated integral in six different
E
ways, where E is the solid bounded by z = 0, x = 0, z = y − 8x and y = 24.
(d) Choose one of the above six iterated integrals to have MATLAB compute
if f (x, y, z) = xyz using the Symbolic Math Toolbox. Which one did you
choose and what is the answer?
References

[1] Stormy Attaway, MATLAB: A Practical Introduction to Programming and Problem Solving, third
edition, Elsevier: Butterworth-Heinemann, Amsterdam, 2013.
[2] Michael Barnsley, Fractals Everywhere, second edition, Morgan Kaufmann, 2000.
[3] Niraj Chokshi, How Powerball manipulated the odds to create a $1.5 billion jackpot, Washington
Post, January 13, 2016, https://www.washingtonpost.com/news/post-nation/wp/2016/01/13/how-
powerball-manipulated-the-odds-to-make-a-1-5-billion-jackpot-happen/.
[4] Robert L. Devaney, The Chaos Game, http://math.bu.edu/DYSYS/chaos-game/node1.html, 1995.
[5] Educational Testing Service, GRE graduate record examinations guide to the use of scores, http://
www.ets.org/gre/guide, 2019.
[6] Temple H. Fay, The butterfly curve, The American Mathematical Monthly 96 (5) (1989) 442–443,
https://doi.org/10.2307/2325155.
[7] C.D. Fryar, Q. Gu, C.L. Ogden, K.M. Flegal, Anthropometric reference data for children and adults:
United States, 2011–2014, National Center for Health Statistics, Vital Health Statistics 3 (39) (2016),
https://www.cdc.gov/nchs/data/series/sr_03/sr03_039.pdf.
[8] Golden Gate Bridge Highway & Transportation District, FAQ, https://www.goldengate.org/faq/#59.
[9] Golden Gate Bridge Highway & Transportation District, School Projects, https://www.goldengate.
org/bridge/history-research/educational-resources/school-projects/, 2006–2018.
[10] Lester S. Hill, Cryptography in an algebraic alphabet, The American Mathematical Monthly 36 (6)
(Jun. - Jul. 1929) 306–312.
[11] Lester S. Hill, Concerning certain linear transformation apparatus of cryptography, The American
Mathematical Monthly 38 (3) (Mar. 1931) 135–154.
[12] Guinness World Records Limited, Fastest Lacrosse Shot, https://www.guinnessworldrecords.com/
world-records/fastest-lacrosse-shot?fb_comment_id=605553549549656_647775638660780, 2019.
[13] Desmond J. Higham, Nicholas J. Higham, MATLAB Guide, second edition, Society for Industrial
and Applied Mathematics, Philadelphia, 2005.
[14] Ben Joffe, Functions 3D: examples, https://www.benjoffe.com/code/tools/functions3d/examples,
2019.
[15] Lee W. Johnson, R. Dean Riess, Jimmy T. Arnold, Introduction to Linear Algebra, fifth edition,
Addison-Wesley, Boston, 2002.
[16] Benoit B. Mandelbrot, The Fractal Geometry of Nature, 1983.
[17] The MathWorks, Inc., MATLAB Documentation, https://www.mathworks.com/help/matlab/,
2019.
[18] Todd Neller, The Game of Pig, Gettysburg College Department of Computer Science, http://cs.
gettysburg.edu/projects/pig/piggame.html. (Accessed 30 September 2019).
[19] David Nicholls, Fractal ferns, https://www.dcnicholls.com/byzantium/ferns/fractal.html, 1998.
[20] Robert Osserman, Mathematics of the Gateway Arch, Notices of the American Mathematical Society
57 (2) (February 2010) 220–229.
[21] Rapid Tables, CMYK to RGB color conversion, https://www.rapidtables.com/convert/color/cmyk-
to-rgb.html.
[22] SPACE.com Staff, What Is the Distance Between Earth and Mars?, Future US, Inc., https://www.
space.com/14729-spacekids-distance-earth-mars.html, March 1, 2012.
[23] E.M. Standish, Keplerian Elements for Approximate Positions of the Major Planets, Solar System
Dynamics Group, JPL/Caltech, https://ssd.jpl.nasa.gov/txt/aprx_pos_planets.pdf.
[24] James Stewart, Single Variable Calculus: Early Transcendentals, eighth edition, Cengage Learning,
Boston, MA, USA, 2016.

271
272 References

[25] James Stewart, Multivariable Calculus: Early Transcendentals, eighth edition, Cengage Learning,
Boston, MA, USA, 2016.
[26] J.D. Watson, F.H. Crick, Molecular structure of nucleic acids; a structure for deoxyribose nucleic acid,
Nature 171 (4356) (April 1953) 737–738.
[27] Gareth Williams, Linear Algebra with Applications, seventh edition, Jones & Bartlett Publishing, 2009.
Index

A D
Alternating series, 167, 168, 172, 178, 179, 181 Default
Angle, 31, 68, 77, 135, 136, 138, 139, 143, 149, colormap, 68
151, 156 colors, 42, 44, 45, 146, 158, 159
Approximating quadratic, 186–188, 191 LineWidth, 43
ASCII files, 11 MarkerSize, 43
Augmented matrix, 201 view, 69
Axes, 36, 38, 39, 53, 54, 57, 79, 81, 142, 144, 190, width, 52
212 Deleting elements, 20
Azimuth angle, 77 Dice rolls in MATLAB® , 55
Differentiable functions, 194
DNA helix, 79
B Domain, 33, 34, 52, 53, 59, 62, 63, 65, 72, 78–81,
Barnsley fern, 144 108, 141, 183, 191, 194, 213
Barnsley fern fractal, 140 definitions, 34
Batch files, 11 issues, 64
Butterfly curve, 56 Double helix, 78

C E
Catenary curve, 52 Eager versions, 94, 95
Chaos game, 153, 158–160 Ellipse, 36, 37, 135, 141–143
Characteristic functions, 99–101 Equilateral triangle, 153, 154, 158, 161
Code, 33, 45, 48, 55, 65, 66, 68, 74, 91, 101–103, Error, 103, 107, 108, 112, 144, 158, 159, 176, 178,
111, 122, 124, 129, 131, 145, 161, 163, 179, 189, 190
165, 212 bounds, 185, 186
improvement, 129, 130 checking, 90, 92, 109, 189
RGB, 45, 141 checks, 91, 92, 108, 110, 144, 189, 191, 211
vectorization, 130 estimate, 188
Colon operator, 29, 34, 65, 100, 101 message, 104, 107–110, 113, 117, 161, 162, 164,
Color, 33, 43–45, 54, 59, 66, 81, 106, 107, 119, 177–181, 189, 190, 211
141, 142, 144, 146, 153, 158, 159, 189, tolerance, 178, 179
212 Euclidean vector spaces, 139, 193
Colormap, 68 Exponential function, 174
Command
alpha, 122 F
window, 3, 4, 7, 10, 11, 60, 85, 111, 119, 211, Fern, 146
212 fractal, 140, 144–146
Conditionally convergent, 168 File format, 84
Contour plot, 65, 81 File function, 85, 104, 158, 159
Convergent geometric series, 170 Fill command, 119
Curve, 33, 40, 49, 50, 52–54, 56, 80, 119, 120, Format, 4, 8, 9, 13, 14, 17, 84, 102, 115, 116, 118
165, 183, 186–188, 191 Fprintf command, 115
Cylinder, 74, 81 Fractal snowflake, 162

273
274 Index

Function, 90 Marker size, 144, 146

examples, 86 Marker specifications, 158, 159
file, 85, 104, 158, 159 MATLAB®
in MATLAB® , 5, 112 editor window, 11
outputs, 104 errors, 104
Functionality, 87 version, 42–44
Matrix, 5, 17, 18, 22, 24, 25, 28–30, 62, 63, 65,
G 74, 94–96, 101, 102, 108, 130, 131, 135,
Game, 30, 154, 158–160 137, 139, 142, 144–146
Gamma function, 90 multiplication, 23, 142
Gosper curve, 156 Meshgrid command, 63
Gosper island, 155, 164, 165 Midpoint rule, 185, 186, 189, 190
Graphs, 53, 66, 70, 72, 80, 91, 144, 180, 181, 212, Minimum value, 30, 178, 179
213 Multiple
inputs, 86
H outputs, 87
Harmonic series, 172 plots, 40–42, 66
Helix, 78, 79 Multivariable calculus, 65, 70, 151
Hexagon, 143, 144, 154, 160, 161, 163–165
N
I Native MATLAB® functions, 84
Imaginary part, 147, 151 Negative increment value, 16
Increment, 15, 16, 31, 34, 110, 158 Nested function, 85, 161–164
Indicator functions, 99 Nested loops, 111
Infinite loop, 103 Nonlinear transformations, 135
Inputs, 33, 59, 63, 91, 92, 104, 107–109, 113,
158–161, 163, 164, 189–191 O
Inputs multiple, 86 Orthonormal
Integral calculus, 81 basis vectors, 212, 213
Iterated function system (IFS), 140, 144 set, 202, 203
vectors, 201, 213
K Output, 9, 29–31, 33, 55, 59, 86–88, 91, 104, 112,
Koch curve, 154 142
Koch snowflake, 154, 155, 161
P
L Parametric curves, 49
Legend, 40, 53, 54, 56, 57, 90, 142–144, 180, 181 Parametric equations, 49, 50, 54, 59, 70, 109, 141
Leibniz function, 178 Partial sum, 173, 177–179, 181
Leibniz series, 177, 178 Pentagon, 143, 160
Level curves, 65, 66, 80, 81 Piecewise function, 98–100
Linear transformations, 135, 137, 139, 141–143 Plain text files, 10, 11
Logical operators, 93–95, 107 Plot
Loop, 101–103, 110, 111, 130, 158, 159, 161–163, commands, 40, 41, 101, 122, 142
178–180 specifications, 162
Plotting, 16, 36, 39, 54, 62, 65, 144, 150
M commands, 33, 41, 48, 74, 122, 123
Maclaurin series, 174, 176, 177, 179–181 functions, 33
Marker, 33, 42, 43, 57, 142–146, 158, 159 points, 150
 Index 275

Polar Snowflake fractals, 154, 156, 162

curves, 50 Space curves, 59, 80
plots, 50, 56 Sprintf command, 117
representation, 149–151 Submatrix, 18
Position vector, 151 Subplots, 46–49, 162, 165, 190, 212
Power series, 169–172, 174–176 Symbolic variables, 125
expansion, 172
Powerball ticket, 109 T
Preallocation, 130–132, 146 Taylor
Precedence rules, 6, 12 polynomial, 173, 174, 176, 179–181
series, 169, 173–177, 179, 181
R Tic/toc commands, 119
Random Tolerance level, 178, 179
generator functions, 31 Transformations, 73, 74, 135, 137–139, 141, 142
integer, 31, 92, 109, 110, 180 Trapezoidal rule, 185, 186, 189, 190
matrix, 28, 213
number, 25–28, 92, 110, 145, 146
number generator, 25, 55, 145, 146
V
Variables
point, 100, 153
vectors, 25, 213 assignments, 9–11
Ratio test, 168, 169, 171, 172 input, 106
Relational operators, 93, 94 names, 3, 9, 86, 141
Remainder function, 5 Vectorized code, 178, 179
RGB Vectorizing code, 101, 130
code, 45, 141 Vectors
colors, 45, 52 addition, 193
value, 44 elements, 18
vector, 52 linear combo, 197
Riemann sum, 183, 185, 189, 190 linearly independent, 196
names, 142
S orthogonal, 204
Script files, 10, 11, 54, 104, 110, 112 orthonormal, 201
Series, 156, 167–173, 175–177, 179, 181 random, 25, 213
Taylor, 169, 173–177, 181 space, 135, 139, 193–200, 203
Shapes, 141–143, 180, 181 Vertices, 34, 142, 154, 158–161, 163, 185
Signum function, 112
Simpson rule, 186, 188, 190, 191 Z
Snowflake, 154, 162, 163 Zenith angle, 77