LECTURE NOTES ON

ENGINEERING COMPUTING

Joseph M. Powers

Department of Aerospace and Mechanical Engineering

University of Notre Dame
Notre Dame, Indiana 46556-5637
USA

updated
19 January 2025, 1:49pm
2

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Contents

Preface 7

1 Overview 9
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 A simple program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.1 Downloading the program . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.2 Saving the program onto the AFS server . . . . . . . . . . . . . . . . . 15
1.2.3 Accessing the campus Linux cluster . . . . . . . . . . . . . . . . . . . 15
1.2.4 Compiling a Fortran program . . . . . . . . . . . . . . . . . . . . . . 18
1.2.5 Editing text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3 Speed test: Fortran versus MATLAB . . . . . . . . . . . . . . . . . . . . . . . 20
1.4 Some definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Introduction to engineering plotting with MATLAB . . . . . . . . . . . . . . . 22
1.5.1 A bad MATLAB plot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.5.2 A better MATLAB plot . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.3 Another bad MATLAB plot . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5.4 Another better MATLAB plot . . . . . . . . . . . . . . . . . . . . . . . 26
1.6 Introduction to LATEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.7 Introduction to HTML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.8 UNIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.9 Online compilers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2 Introduction to problem solving 35

3 Introduction to programming languages 37

4 Introduction to programming 39
4.1 Statement types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Example program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4 Another example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3
4 CONTENTS

5 Arithmetic 45
5.1 Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Integer division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3 Binary representation of integers . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4 Overflow and underflow limits . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.5 Assignment of undefined variables . . . . . . . . . . . . . . . . . . . . . . . . 51

6 Arrays 1 53
6.1 Dimension of an array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2 One-dimensional arrays from time series . . . . . . . . . . . . . . . . . . . . 54
6.2.1 Fixed array size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2.2 Variable array size via parameter declaration . . . . . . . . . . . . . 59
6.3 Simple writing to files and MATLAB plotting . . . . . . . . . . . . . . . . . . . 60

7 Arrays 2 63
7.1 Varying the array size at run time . . . . . . . . . . . . . . . . . . . . . . . . 63
7.2 Higher dimension arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.3 Two-dimensional arrays of variable size . . . . . . . . . . . . . . . . . . . . . 69
7.4 Non-standard do loop increment syntax . . . . . . . . . . . . . . . . . . . . . 70

8 Whole array features 73

8.1 Dot product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.1.1 Real vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.1.2 Complex vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
8.2 Matrix transpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.3 Matrix-vector multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
8.4 Matrix-matrix multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8.5 Addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.6 Element-by-element matrix “multiplication” . . . . . . . . . . . . . . . . . . 86

9 Output of results 89
9.1 Unformatted printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
9.2 Formatted printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
9.3 Formatted writing to files . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

10 Reading data 99
10.1 Unformatted reading from a file . . . . . . . . . . . . . . . . . . . . . . . . . 99
10.2 Formatted reading from a file . . . . . . . . . . . . . . . . . . . . . . . . . . 102

11 I/O Concepts 105

CC BY-NC-ND. 19 January 2025, J. M. Powers.

CONTENTS 5

12 Functions 107
12.1 Pre-defined functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
12.2 User-defined functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
12.2.1 Simple structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
12.2.2 Use of dummy variables . . . . . . . . . . . . . . . . . . . . . . . . . 112
12.2.3 Sharing data using module . . . . . . . . . . . . . . . . . . . . . . . . 113
12.2.4 Splitting functions into separate files . . . . . . . . . . . . . . . . . . 115
12.2.5 Creation of executable script files . . . . . . . . . . . . . . . . . . . . 117

13 Control structures 119

14 Characters 127

15 Complex 129

16 Logical 135

17 Introduction to derived types 137

18 An introduction to pointers 139

19 Introduction to subroutines 141

19.1 Simple subroutine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
19.2 Subroutines for solution of a quadratic equation . . . . . . . . . . . . . . . . 143
19.3 Subroutines for solving systems of ordinary differential equations . . . . . . . 145
19.3.1 One linear ordinary differential equation . . . . . . . . . . . . . . . . 146
19.3.2 Three non-linear ordinary differential equations . . . . . . . . . . . . 150

20 Subroutines: 2 161

21 Modules 163

22 Converting from Fortran 77 165

22.1 A hello world program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
22.2 A simple program using arrays . . . . . . . . . . . . . . . . . . . . . . . . . . 166
22.3 A program using subroutines . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

23 Python and Fortran 173

24 Introduction to C 175
24.1 A hello world program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
24.2 A program using the Euler method . . . . . . . . . . . . . . . . . . . . . . . 176

CC BY-NC-ND. 19 January 2025, J. M. Powers.

6 CONTENTS

25 Introduction to Microsoft Excel 179

25.1 A hello world program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
25.2 A program using the Euler method . . . . . . . . . . . . . . . . . . . . . . . 180

26 Introduction to VBA 183

26.1 A hello world program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
26.2 A program using the Euler method . . . . . . . . . . . . . . . . . . . . . . . 184
26.3 Another VBA Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
26.3.1 Creating a calculation macro . . . . . . . . . . . . . . . . . . . . . . . 187
26.3.2 Clearing data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
26.3.2.1 Manual clearing . . . . . . . . . . . . . . . . . . . . . . . . . 190
26.3.2.2 Automated clearing . . . . . . . . . . . . . . . . . . . . . . 191
26.3.3 Automated plotting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
26.3.4 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

27 Introduction to Mathematica 195

27.1 A hello world program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
27.2 A simple code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
27.3 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
27.4 Lists, vectors, and matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
27.5 Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
27.6 Some plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
27.7 Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
27.8 The Euler method, DSolve, and NDSolve . . . . . . . . . . . . . . . . . . . . 207

Bibliography 211

Index 213

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Preface

These are lecture notes for AME 20214, Introduction to Engineering Computing, a one-
hour sophomore-level undergraduate course taught in the Department of Aerospace and
Mechanical Engineering at the University of Notre Dame.
The key objective of the course is to introduce students to the UNIX operating system
and structured high performance computing with a compiled language, here chosen to be
Fortran 90 and its descendants, as well as some others, such as the interpreted languages
MATLAB, Python, Mathematica, VBA, and Microsoft Excel, and a very brief introduction to
the compiled language C. Thanks to Mr. Travis Brown for help with a C example program
and Ms. Laura Paquin for help with VBA example programs and description.
An important secondary objective is to introduce the student to the process of scientific
computing: the art of using a computer to solve problems of scientific and engineering rele-
vance. To achieve that objective, some attention is focused on numerically solving physically
motivated systems of ordinary differential equations. The most important of these will be
the forced mass-spring-damper system, which will be addressed with a variety of methods.
The notes draw often on the text specified for the course, Chivers’ and Sleightholme’s
Introduction to Programming with Fortran, Third Edition, Springer, London, 2015. The
notes will highlight aspects of this text, and augment it in places. The notes and course data
can be found at https://www3.nd.edu/∼powers/ame.20214. At this stage, anyone is free
to make copies for their own use. I would be happy to receive suggestions for improvement.
Joseph M. Powers
[email protected]
https://www3.nd.edu/∼powers
Notre Dame, Indiana; USA
CC BY: $ = 19 January 2025
\

The content of this book is licensed under Creative Commons Attribution-Noncommercial-No Derivative Works 3.0.

7
8 CONTENTS

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 1

Overview

Read C&S, Chapter 1.

1.1 Motivation
We are faced with a problem in constrained optimization. We seek to maximize student
understanding of

• mathematical methods to computationally solve problems arising in aerospace and

mechanical engineering,

• programming languages for efficient solution of large-scale mathematical problems aris-

ing in engineering, and

• operating systems beyond those found by default in consumer-market computers,

subject to the constraints

• finite time, embodied in the one credit hour allocated to this task,

• finite student skills, and

• finite instructor skills.

We might conjecture that the understanding of these topics, Υ, is a function of the mathe-
matical method µ, the language chosen, λ, the operating system, σ, the pedagogy ρ, chosen,
the time τ allotted for study, the student and instructor skill sets, κ1 and κ2 . In a mathe-
matically inspired language, we might say

Υ = f (µ, λ, σ, ρ, τ, κ1 , κ2 ). (1.1)

9
10 CHAPTER 1. OVERVIEW

Our constraints are for time τ , student skills κ1 and instructor skills κ2 . Our constrained
optimization problem might be stated as, select µ, λ, σ, and ρ so as to

Maximize: Υ = f (µ, λ, σ, ρ, τ, κ1, κ2 ), (1.2)

Subject to: τ = 1, (1.3)
κ1 = constant, (1.4)
κ2 = constant. (1.5)

We are presented with a wide variety of mathematical problems µ of engineering relevance

including

• linear systems of algebraic equations,

• non-linear systems of algebraic equations,

• linear systems of ordinary differential equations,

• non-linear systems of differential equations,

• partial differential equations, or

• optimization problems.

We are presented with a wide variety of computing languages to formulate these problems,
including

• Fortran (from “FORmula TRANslating system”),

• COBOL (from “Common Business-Oriented Language”),

• ALGOL (from “ALGOrithmic Language”),

• Ada,

• BASIC (from “Beginner’s All-purpose Symbolic Instruction Code”),

• VBA (from “Visual Basic for Applications”),

• Pascal,

• C,

• C++,

• C#,

• Java,

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.1. MOTIVATION 11

• MATLAB (from “MATrix LABoratory”),

• Mathematica, or

• Microsoft Excel.
We are presented with a wide variety of operating system environments in which to execute
our programs including
• Microsoft Windows,

• DOS,

• OS-X,

• UNIX, or

• Linux.
Following considerable consideration, coupled with some experimentation, we believe the
near-optimal solution, reflecting the severe constraint of τ = 1 credit hour is given by the
following: student understanding, Υ, of engineering mathematics, relevant languages, and
advanced operating systems is given by choosing
• λ = Fortran 2003,

• σ = Linux.
Near neighbors within the solution space are also perfectly acceptable, e.g. Fortran 90,
Fortran 95, Fortran 2008, UNIX, OS-X. If the constraints of the curriculum evolve, and
languages and operating systems evolve, the optimal solution may evolve!
All choices have advantages and disadvantages. Certainly one chooses modes of trans-
portation depending on a variety of needs. We depict rough and imperfect analogies be-
tween some modern transportation choices and modern computing languages in Fig. 1.1.1
For transportation purposes, one might imagine that
• The racing cycle has speed, power, and agility but requires a highly trained user to
achieve peak performance. It maps into C.

• The modern muscle car has all features expected by most mass-market drivers, draws
heavily on past experiences, but still may not be the right machine for all market
segments. It maps into Fortran.
1
images from
http://motorsportsnewswire.files.wordpress.com/2011/05/infineon 2011 brammo empulse rr.jpg
http://www.netcarshow.com/ford/2013-mustang boss 302/1600x1200/wallpaper 01.htm
http:///www.motortrend.com/news/2017-dodge-grand-caravan-first-drive-not-dead-yet/
http://www.motorzoeker.nl/public/motorfotos/large/vespa/5611 foto1 k25j3xt2j8.jpg
https://commons.wikimedia.org/wiki/File:Campagna T-Rex 14-R rear left.jpg

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12 CHAPTER 1. OVERVIEW

high performance 1960’s inspired modern

racing cycle: C? muscle car: Fortran?

well-engineered and versatile lightweight and practical unconventional but stable

family minivan: MATLAB? scooter: Excel? motorized reverse tricycle: VBA?

Figure 1.1: Loose transportation analogies with computing languages.

• The minivan is appropriate for a wide variety of drivers and passengers, is durable,
and optimized for safety; it will not compete well in a race. It maps into MATLAB.

• The scooter is affordable, maneuverable, and is appropriate for many urban environ-
ments; it is not right for the highway. It maps into Excel.

• The motorized reverse tricycle is stable for some highway use, but unconventional. It
maps into VBA.

Now focusing on our actual computing systems, we list some advantages and disadvan-
tages associated with our present choices. First for Fortran:

• Fortran advantages

– a compiled language which yields very fast binary executable code widely used for
large scale problems in computational science, e.g. the widely used Top 500 su-
percomputing benchmark problem is based on Fortran-implemented algorithms.
– used in many modern supercomputing applications, e.g. the Department of En-
ergy’s major initiative in scientific computing for the 21st century, Predictive
Science Academic Alliance Program (PSAAP),
– explicitly designed for mathematical problems common in engineering; pedagog-
ically straightforward implementation of standard notions such as sine, cosine,
exponential, vectors, and matrices,
– widely used for decades in engineering with a large data base of legacy code and
packaged software,2
2
see http://www.netlib.org/

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.1. MOTIVATION 13

– another language in the engineer’s multi-lingual arsenal for problem-solving,

– a language which opens the door to learning a wide variety of compiled languages,
such as C, which was designed for device drivers and operating systems (not sci-
entific computing), and
– forces care and precision to be used in code writing–thus reinforcing fundamental
engineering design principles for quality control which transcend programming.

• Fortran disadvantages

– forces care and precision to be used in code writing–which induces frustration

when the code fails to work properly,
– relative to some other languages, does not interact as easily with the operating
system and input/output data, thus preventing use in many hardware applica-
tions, e.g. robotics,
– no longer commonly used in the computer science community, especially the large
fraction of that community who no longer are concerned with scientific computing
of mathematics-based engineering problems based on differential equations, and
– not particularly relevant to the large user community who primarily run pre-
programmed specialty software packages for engineering analysis and design, but
do not generate original source code.

For Linux, we have

• Linux advantages

– widely used,
– allows user to more easily understand computer system architecture,
– freely available
– very similar to UNIX and OS-X.

• Linux disadvantages

– more effort to learn, relative to What-You-See-Is-What-You-Get (WYSIWYG)

systems such as Windows or the front end of OS-X.
– users must manage their own systems.

Whatever language and operating system we use, and this course will consider Fortran
with short excursions into C, MATLAB, Microsoft Excel, VBA, and Mathematica, the essential
job of nearly any programming language is executing so-called three R’s:
• Readin’

• ’Ritin’

CC BY-NC-ND. 19 January 2025, J. M. Powers.

14 CHAPTER 1. OVERVIEW

• ‘Rithmatic

A typical program will change the order (and we will now use proper spelling) and first

• Read input data from a file or the screen,

• Perform arithmetic operations on the data, and

• Write output data to a file or the screen.

Recognizing the various methods that various languages use to execute the three R’s is the
key to learning a new computer language. There are a small number of typical operations
in scientific computing. Once the student learns how to do an operation in one language, it
is much easier to translate these steps to any of the other languages. Thus, in some sense,
learning a computer language may be more important than learning any particular computer
language. That is because learning a language typically entails learning the more complex
generalities associated with all languages.

1.2 A simple program

Let us consider one of the simplest possible Fortran that one can imagine, a so-called
“hello world” program.

1.2.1 Downloading the program

The program, with file name helloworld.f90, is given here:

program test
print*,’hello world’
end program test

One can download this program from the web by clicking the following web link,
https://www3.nd.edu/∼powers/ame.20214/2e/helloworld.f90
The first line of the program is the program name, where the name is arbitrary. The second
line tells the program to print to the screen, indicated by a ∗, the words within the single
quotation marks. The program is ended by a program end statement. The file name which
ends in “.f90” is essential and tells the compiler that this file follows the syntax of Fortran
90 and beyond, beyond including Fortran 90, 95, 2003, and 2008. It should be noted
that our local system can in fact invoke the Fortran 2003 standard, but for nearly all
problems we will consider, the Fortran 90 standard will suffice.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.2. A SIMPLE PROGRAM 15

1.2.2 Saving the program onto the AFS server

After downloading, one should save the file onto the campus file server known as AFS and
choose to name the saved file as helloworld.f90. The letters AFS are an acronym for
“Andrew File System.” Files saved onto your AFS space can be accessed and edited from
a wide variety of campus and off-campus machines. In many ways storing files on AFS is
like storing them on the “cloud,” though AFS has a much longer history than the cloud.
Importantly, it is possible to compile and execute binary executable files from AFS.
There are a variety of ways to save the file. Most browsers have some variety of save as
option. If exercised, be sure to save the file as plain text with a .f90 extension in its name.
Or one can cut and paste the necessary text into a variety of text editors such as gedit,
described in more detail in Sec. 1.2.5.
Note that the AFS space is distinct from the more common campus space in netfile.
For purposes of running compiled programs which are binary executable, such as those
originating in Fortran or C source codes, netfile is completely unusable. In short, while
you could certainly save a text file within netfile, you could never compile or execute the
program.

1.2.3 Accessing the campus Linux cluster

To compile and execute the program, you will need to log onto a machine which has a
Fortran compiler.

It is strongly recommended to directly log in by actually sitting at a campus Linux

cluster machine when first learning the Linux operating system. Indeed, every-
thing that can be achieved sitting in front of the Linux machine can also be
achieved remotely via a laptop; however, many things can go wrong. After brief
experience in front of the actual machine, it is not a difficult step to use Linux
remotely from a laptop.

That said, many users do have sufficient experience to go directly to remote login from a
laptop, and that procedure is described here. One important machine for remote login is
the campus Linux machine darrow.cc.nd.edu. The machine darrow is useful because all
campus personnel can log in to it from any remote site, on or off campus. Here is what one
might do if logging in from the Terminal application of a personal Macintosh machine using
the so-called “secure shell” program ssh:

esc300037:~ powers$ ssh -x darrow.cc.nd.edu -l powers

[email protected]’s password:
Last login: Wed Aug 2 08:52:14 2017 from c-98-253-148-207.hsd1.in.comcast.net

[powers@darrow5 ~]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

16 CHAPTER 1. OVERVIEW

Here, the argument of the −l option is your netid. Mine is powers. The general form is
ssh -X darrow.cc.nd.edu -l netid
If darrow.cc.nd.edu is overloaded with too many users, a user may log in remotely to
any of a set of virtual Linux machines. Their names are

remote101.helios.nd.edu
remote102.helios.nd.edu
remote103.helios.nd.edu
remote104.helios.nd.edu
remote105.helios.nd.edu
remote106.helios.nd.edu
remote107.helios.nd.edu
remote108.helios.nd.edu
remote201.helios.nd.edu
remote202.helios.nd.edu
remote203.helios.nd.edu
remote204.helios.nd.edu
remote205.helios.nd.edu
remote206.helios.nd.edu
remote207.helios.nd.edu
remote208.helios.nd.edu

For example to log in remotely to one of the virtual Linux machines, one might execute the
command sequence

esc300037:~ powers$ ssh -X remote103.helios.nd.edu -l powers

The authenticity of host ’remote103.helios.nd.edu (129.74.50.104)’ can’t be established.
RSA key fingerprint is 80:12:f5:2e:91:c7:97:83:8b:54:6a:2d:10:14:6f:e2.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added ’remote103.helios.nd.edu,129.74.50.104’ (RSA) to the list of
*******************************************************************************
*******************************************************************************
** **
** This system is for the use of AUTHORIZED USERS ONLY. **
** **
** Anyone using this system expressly consents to such monitoring and is **
** advised that if such monitoring reveals possible evidence of criminal **
** activity, system personnel may provide the evidence of such monitoring **
** to law enforcement officials. **
** **
** See the University of Notre Dame Responsible Use Policy for information **

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.2. A SIMPLE PROGRAM 17

** regarding proper use of this machine (http://oit.nd.edu/policies). **

** **
*******************************************************************************
*******************************************************************************

Remote use of this system is not for running big processes. If you need to
run big processes for research, please get an account from Center of Research
Computing (http://crc.nd.edu).

Please email [email protected] for any question of suggestion.

[email protected]’s password:
Last login: Thu Nov 12 15:20:01 2015 from dbrt155l-00.cc.nd.edu
[powers@remote103 ~]$

Other campus Linux machines are useful as well, but the user must be sitting at the
machine. A practical advantage of these physical machines is they only allow a single user,
so they will not become overloaded. Several such Linux machines can be found in

• 149 Fitzpatrick Hall (Engineering Library),

• B-19 Fitzpatrick Hall, and

• 108-110, 212-213 Stinson-Remick Hall.

A list of available machine names and locations is found at

http://pug.helios.nd.edu:2080/clusterfree.html
For students with Macintosh laptops, one can download so-called X11, also known as
XQuartz, software to create an X Windows environment. One can try the following site
for download:
http://www.xquartz.org
Once downloaded, take care to open the terminal application from XQuartz rather than the
standard Macintosh terminal application. From the XQuartz terminal, one can invoke the
command ssh command to a Linux machine Some more information can be found at
https://en.wikipedia.org/wiki/XQuartz
For students with Windows laptops, one can download software to create an X Windows
terminal environment. Instructions can be found at
https://www3.nd.edu/∼powers/ame.20214/xming.pdf
When invoking the so-called xming software, be sure to examine the “Connection” menu on
the left and under SSH, X11, check enable X11 forwarding.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

18 CHAPTER 1. OVERVIEW

Another common way, which requires a little more effort, to bring UNIX functionality
into a Windows environment is to install the free, open-source UNIX-emulator, Cygwin, onto
a Windows machine.
https://www.cygwin.com

1.2.4 Compiling a Fortran program

On darrow.cc.nd.edu, the relevant virtual Linux machines, and all physical campus Linux
cluster machines reside the Fortran compilers

• ifort, and

• gfortran.

The compiler ifort is the so-called Intel Fortran compiler. The compiler gfortran is
an open source compiler. The ifort compiler works on most campus Linux machines, but
may not work for all on darrow.cc.nd.edu. The gfortran compiler should work on all
campus Linux machines. Focus here will be on the ifort compiler; though it is a useful
exercise to make sure your program compiles and executes on a variety of compilers, as there
can be subtle differences. All these compilers work in essentially the same way. For instance
to use ifort to compile helloworld.f90, we perform

[powers@remote101 ame.20214]$ ifort helloworld.f90

This generates a so-called “binary” executable file, which is not easily viewable, named
“a.out”. We execute the file a.out via the command and get

[powers@remote101 ame.20214]$ a.out

hello world.
[powers@remote101 ame.20214]$

The output of the program is printed to the screen. Though it is not at all important for this
problem, we actually compiled the program with a Fortran 90 compiler. We can compile
and execute the program in Fortran 2003 via the command sequence

[powers@remote101 ame.20214]$ ifort helloworld.f90 -stand f03

[powers@remote101 ame.20214]$ a.out
hello world.
[powers@remote101 ame.20214]$

Here, the so-called compiler option, -stand f03, enforced the compilation using the Fortran
2003 standard.
We can compile and execute the program with the gfortran compiler via the command
sequence

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.2. A SIMPLE PROGRAM 19

[powers@darrow1-p ~]$ gfortran helloworld.f90

[powers@darrow1-p ~]$ a.out
hello world
[powers@darrow1-p ~]$

1.2.5 Editing text

You will need to be able to edit text files. This is best achieved with a text editor. For
the relevant tasks, tools such as Microsoft Word are inappropriate, because they generally
do not format the files in plain text; instead they often add extraneous data to the files for
purposes of visual display. These visually appealing elements render the files unusable for
programming purposes.
Instead, you should employ any of a variety of standard text editors. Perhaps the most
straightforward is one known as gedit. It is available on the campus Linux cluster machines.
To edit the file helloworld.f90, one can invoke on a campus Linux cluster machine the
command

[powers@darrow1-p ~]$ gedit helloworld.f90 &

This command spawns a new window in which the file helloworld.f90 can be edited and
saved. The character & is a useful UNIX command which commands the system to allow the
terminal window to be used simultaneously with the program gedit. If one is using gedit
from a remote terminal or personal laptop machine, one must be sure to have logged in from
a so-called X terminal window, using the command ssh -X darrow.cc.nd.edu -l userid
in order for the edit window to display. This is straightforward on a Macintosh computer
through use of its terminal application found in the Utilities folder. On a Windows
laptop, one may wish to use the terminal available via the Cygwin package. This is not an
issue if the user is seated at the actual machine in use.
An incomplete list of common text editors available on the campus Linux cluster, each
with advantages and disadvantages is as follows:

• gedit

• Kate, invoked by kate on the Notre Dame cluster (but not on darrow)

• vi

• Emacs, invoked by emacs on the Notre Dame cluster.

One can also use ordinary text editors on personal computing machines and save the file as a
plain text file in the appropriate space. Note that one should not save a Fortran source code
file in the common “rich text format” (.rtf) and expect it to be able to compile correctly.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

20 CHAPTER 1. OVERVIEW

1.3 Speed test: Fortran versus MATLAB

Let us check the speed of an ordinary Fortran program versus its MATLAB equivalent. Speed
tests are notorious for being unfair, depending on the options exercised. In this case, we
will use options which are widely considered to be default for both programs, and we will
execute on the same machine. Both programs will initialize a large matrix, of dimension
n × n. We will call the matrix x and its elements will be denoted xij . We will initialize so
that each element takes on the value of the product of i and j: xij = ij, where i = 1, . . . , n,
and j = 1, . . . , n. Thus x11 = (1)(1) = 1, x12 = (1)(2) = 2, . . . , x25 = (2)(5) = 10, and so
on. If n = 5, we should thus form the matrix
   
(1)(1) (1)(2) (1)(3) (1)(4) (1)(5) 1 2 3 4 5
(2)(1) (2)(2) (2)(3) (2)(4) (2)(5) 2 4 6 8 10
   
x=  (3)(1) (3)(2) (3)(3) (3)(4) (3)(5) = 3 6 9 12 15 .
   (1.6)
(4)(1) (4)(2) (4)(3) (4)(4) (4)(5) 4 8 12 16 20
(5)(1) (5)(2) (5)(3) (5)(4) (5)(5) 5 10 15 20 25

When n is large, it will take a long time to execute the initialization. Let us perform
this process via Fortran and MATLAB on the linux machine, darrow.cc.nd.edu. We shall
take n = 5000, a modestly sized matrix for scientific computing. The Fortran program
matrix.f90 and its execution are as follows:

[powers@darrow1-p ame.20214]$ cat matrix.f90

program matrix ! beginning of the program
implicit none ! force all variables to be declared
integer :: i,j ! declare i and j to be integers
integer, parameter :: n = 5*10**3 ! give the dimension of the matrix
real :: x(n,n) ! declare the matrix x to be n x n and real
do i=1,n ! first do loop
do j=1,n ! second do loop
x(i,j) = real(i)*real(j) ! initialize the matrix
enddo ! end of do second loop
enddo ! end of first do loop
print*,n,x(n,n) ! print some values
end program matrix ! end of program
[powers@darrow1-p ame.20214]$ ifort matrix.f90
[powers@darrow1-p ame.20214]$ a.out
6000 3.6000000E+07
[powers@darrow1-p ame.20214]$

We shall worry about the details of the program later. Note for now that there are many
complexities to execute this seemingly simple task. This is typical of Fortran codes. We
can check the speed of the program by executing with a UNIX utility time. We find

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.4. SOME DEFINITIONS 21

[powers@darrow2-p ame.20214]$ /usr/bin/time -p a.out

6000 3.6000000E+07
real 0.46
user 0.40
sys 0.05
[powers@darrow2-p ame.20214]$
The execution of this particular code was achieved in

tFortran ∼ 1.42 s. (1.7)

The MATLAB code, available in matrix.m, is given by

tic
n = 5*10^3;
for i=1:n
for j=1:n
x(i,j) = i*j;
end
end
n
x(n,n)
toc
The MATLAB commands tic and toc are for timing the code. Relative to the Fortran code,
the MATLAB code has a similar, but simpler structure and syntax. However, the execution of
this particular code was achieved in a much longer time:

tMATLAB ∼ 12.724 s. (1.8)

It must be emphasized that these numbers are typical, but unreliable. The machine may
have been performing different background tasks. Also, one could use optimizing Fortran
compilers to improve Fortran execution time. The main reason for the fast execution of the
Fortran program is that it was compiled prior to execution, whereas the MATLAB program,
using a so-called interpreted language was not. All things equal, compiled languages are typ-
ically much faster than interpreted languages. However, in most cases, interpreted languages
have been designed to be more user-friendly. Note that it is possible to use compiled versions
of MATLAB to improve its execution time significantly.

1.4 Some definitions

Here are a few key concepts and terms associated with computer systems:
• Bits and bytes are the representation of data by combinations of 0 and 1. A byte is
usually 8 bits. Most common machines today are based on a 64-bit architecture.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

22 CHAPTER 1. OVERVIEW

• Central Processing Unit (CPU) is, in colloquial terms, the main brain of the computer.
• One of the more important memory devices is Random Access Memory (RAM). This
memory is usually transient in that it is only active when the machine is turned on.
It is not used for permanent storage of files. Permanent storage is achieved via hard
drives, memory sticks, etc.
• A terminal is a device which allows direct input and output communications between
a user and the CPU of a machine. Terminals come in many flavors, including virtual.
• Software is an imprecise term which describes programs which allow interaction be-
tween the user and the computer. High level software is often more user-friendly and
has several layers of abstraction between the user and CPU. Low level software has
fewer levels of abstraction, is often more difficult to understand, but often tightly linked
to the CPU architecture. Microsoft Excel is a high level software tool. MATLAB is
somewhat lower level, but still at a high level. Fortran is at a lower level still, but in
many ways a high level language. So called assembly language is much closer to the
machine, and is certainly a low level software tool.
• A compiler, such as ifort, is a tool to convert a high level language such as Fortran
into a fast executable binary low level language usable by the machine. Use of com-
piled languages is highly advantageous for large scale problems in scientific computing
because of the speed of execution. For small scale problems, non-compiled interpreted
languages, such as MATLAB, are advantageous because of their relative simplicity of use.
• A text editor allows the user to alter the source code. Text editors typically refer to
tools which give the user access to the entire ASCII (American Standard Code for
Information Interchange) text of the file and avoid use of hidden and/or binary text.

1.5 Introduction to engineering plotting with MATLAB

Well done plots are an important part of communication for engineers. One should take care
to try to adhere to some general plot design principles:
• Use font sizes which are close to those of the printed text. Most software tools today,
unfortunately, use default font sizes which are too small for printed text; occasionally
they are too large. Whatever the case, the text on your plot should have a font size
very similar to that of your main text.
• Label your axes; include units when appropriate.
• In axes labels and text, use fully formatted mathematical symbols. For example, if the
proper symbol is φ, use it instead of “phi.” Or if the proper symbol is
y1 , do not
substitute y 1 or y1 or y1. The same holds for units; use as a label ρ mkg3 , not “rho
(kg/m3).”

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.5. INTRODUCTION TO ENGINEERING PLOTTING WITH MATLAB 23

• If there is more than one curve, be sure the reader knows which curve corresponds to
which case. There are a variety of ways to achieve this. The best is probably to use
combinations of solid lines, dashed lines, dotted lines, etc. Different lines can also be
assigned different colors, but one should recognize that sometimes the user will print
files with black and white printers, obscuring the meaning. Also, some colors are hard
to discern on varieties of computer screens and printers.

• Generally, reserve point markers like small circles or dots for cases where

– The data comes from a finite, discrete set of experimental observations, or

– The data comes from a finite, discrete set of computational runs.

Often it is best not to “connect the dots” especially if there is no reason to believe that
another data point, if sampled, would lie exactly on the interpolating curve.

• Generally, reserve smooth curves with no small circles or dots for cases where

– The curve is representing a continuous mathematical function such as y = x2 +

2x + 1; Indeed, one must sample a discrete number of points to generate the plot,
but the underlying function is continuous.
– There is good reason to believe that the discrete data points are actually samples
of a continuous function. We may not know the function, but when we continue
to sample, we find the continuity assumption holds.

• Use linear, log-linear, and log-log scaled plots as appropriate for the data at hand.
Sometimes linear scales hide the true nature of the data, and sometimes they reveal it.
The same can be said for log-linear and log-log scales. Be sure to examine your data,
and choose the most revelatory scaling.

1.5.1 A bad MATLAB plot.

Let us say we have the following data to plot:

x y
0 1
2 4
3 8
4 7
7 10
10 19

MATLAB defaults typically produce faulty plots. The following MATLAB script is an example:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

24 CHAPTER 1. OVERVIEW

x = [0,1,3,4,7,10];
y = [2,4,8,7,10,19];
plot(x,y)
This script generates the plot of Fig. 1.2. Figure 1.2 has the following problems:

20

18

16

14

12

10

2
0 1 2 3 4 5 6 7 8 9 10

Figure 1.2: Badly formatted plot of some data.

• The numbers on the axes are much smaller than the font size of the surrounding text.
This problem is pervasive when MATLAB default values are used.
• The axes are not labeled.
• The raw data should be indicated with distinct points. Though optional, it is probably
best not to connect them.

1.5.2 A better MATLAB plot

A few simple changes to the MATLAB script can improve the plot dramatically:
x = [0,1,3,4,7,10];
y = [2,4,8,7,10,19];
plot(x,y,’o’),...
xlabel(’x’,’FontSize’,24),...
ylabel(’y’,’FontSize’,24);...
axis([0,12,0,25]),...
set(gca,’FontSize’,20)
This script generates the plot of Fig. 1.3. Here the axes were expanded, the data were plotted
as individual points, and the font size was increased.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.5. INTRODUCTION TO ENGINEERING PLOTTING WITH MATLAB 25

25

20

15
y

10

0
0 2 4 6 8 10 12
x

Figure 1.3: Properly formatted plot of some data.

1.5.3 Another bad MATLAB plot

Let us say we want to plot on a log-log scale the two continuous curves
y = x2 , y = x3 , x ∈ [0.01, 1000].
Shown next is a MATLAB script which generates a bad plot:
clear;
n = 1000;
xmin = 0.01;
xmax = 1000;
dx = (xmax-xmin)/(n-1);
for i=1:n
x(i) = xmin+(i-1)*dx;
y1(i) = x(i)^2;
y2(i) = x(i)^3;
end
loglog(x,y1,’bo’,x,y2,’bo’)
This script generates the plot of Fig. 1.4. Figure 1.4 has the following problems:
• The underlying curves are continuous, but only discrete points have been plotted.
• It is difficult to distinguish the two curves, especially for small x.
• The numbers on the axes are much smaller than the font size of the surrounding text.
• The axes are not labeled.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26 CHAPTER 1. OVERVIEW

10
10

8
10

6
10

4
10

2
10

0
10

−2
10

−4
10

−6
10
−2 −1 0 1 2 3
10 10 10 10 10 10

Figure 1.4: Badly formatted plot of the curves y = x2 , y = x3 .

1.5.4 Another better MATLAB plot

The following MATLAB script gives a better plot of the same two curves:

clear;
n = 1000;
xmin = 0.01;
xmax = 1000;
dx = (xmax-xmin)/(n-1);
for i=1:n
x(i) = xmin+(i-1)*dx;
y1(i) = x(i)^2;
y2(i) = x(i)^3;
end
loglog(x,y1,x,y2),...
xlabel(’x’,’FontSize’,24),...
ylabel(’y’,’FontSize’,24)
legend(’y=x^2’,’y=x^3’,’Location’,’NorthWest’)
set(gca,’FontSize’,20)

This script generates the plot of Fig. 1.5. Here the data were plotted as continuous curves,
the two curves were clearly distinguished and labeled, and the font sizes were increased to
be nearly the same as that of the surrounding text.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.6. INTRODUCTION TO LATEX 27

10
10
y=x2
y=x3
5
10

0
10
y

−5
10

−10
10 −2 0 2 4
10 10 10 10
x

Figure 1.5: Properly formatted log-log plot of the curves y = x2 , y = x3 .

1.6 Introduction to LATEX

Fortran is an example of a software application the runs in the UNIX environment. The text
formatter, LATEX, is another example. It is useful for the following reasons:

• It produces the best output of text, figures, and equations of any program I’ve seen.

• It is machine-independent. It runs on Linux, Macintosh (see TeXShop), and Windows

(see TeXnicCenter ) machines. You can e-mail ASCII versions of most relevant files.

• It is the tool of choice for many research scientists and engineers. Many journals accept
LATEX submissions, and many books are written in LATEX.

Some basic instructions are given next. You can be a little sloppy about spacing. It ad-
justs the text to look good. You can make the text smaller. You can make the text tiny. You can
link to web sites
Skip a line for a new paragraph. You can use italics (e.g. Computers are everywhere) or
bold. Greek letters are a snap: Ψ, ψ, Φ, φ. Equations within text are easy— A well known
equation for a line is y = mx + b. You can also set aside equations like so:

d2 x X
m = F, Newton’s second law. (1.9)
dt2
Eq. (1.9) is Newton’s second law. References3 are available. If you have an postscript file,
say f1.eps, in the same local directory, you can insert the file as a figure. Figure 1.6 gives

CC BY-NC-ND. 19 January 2025, J. M. Powers.

28 CHAPTER 1. OVERVIEW

y
103

101

10-1

10-3

10-5

10-7

x
0.01 0.1 1 10

Figure 1.6: Sample log-log plot of some data.

a log-log plot of some data. Figure 1.6 was first created with MATLAB, was saved as a .eps
file, then imported into Adobe Illustrator to adjust the fonts to match those of the main
text. In this case, it was necessary to download common LATEX fonts onto a local machine.
Running LATEX
You can create a LATEX file with any text editor (vi, Emacs, gedit, etc.). To get a
document, you need to run the LATEX application on the text file. The text file must have
the suffix “.tex” On a Linux cluster machine, this is done via the command

[powers@darrow2-p ame.20214]$ latex2pdf file.tex

This generates file.pdf. You should execute this command at least twice to ensure correct
figure and equation numbering.
Alternatively, you can use TeXShop on a Macintosh or TeXnicCenter on a Windows-based
machine.

1.7 Introduction to HTML

To build web pages, one often uses so-called HyperText Markup Language, HTML. HTML is in
many ways like a programming language, and one can do many things with it. Key to the
approach is having several publicly available files residing on a disk space which is accessible
to the public. On the Notre Dame system, that space exists by default on a server that can
be mapped to your local machine.
3
Lamport, L., 1986, LATEX: User’s Guide & Reference Manual, Addison-Wesley: Reading, Massachusetts.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.7. INTRODUCTION TO HTML 29

See course TAs for information on how to connect your individual machine to this file
space.
The key file to get a web site going must be named index.html and must reside in the
www folder. An example is given in the file index.html, listed next.

<html>
<head>
<title>My Home Page</title>
</head>

This is my home page.

<p>
<a href=https://www3.nd.edu/~powers/ame.20214/helloworld.f90>Link</a> to the file
helloworld.f90. Whatever file to which you link must reside in the proper
directory.

<p>
<a href=https://www.google.com>Link</a> to Google.

</html>

When this file is opened in a browser, one might expect to see something like the screen
shot of Fig. 1.7.

Figure 1.7: Screen shot of a web browser which opened the file index.html.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

30 CHAPTER 1. OVERVIEW

One can note a few points:

• Commands within HTML are begun with the structure <....> and ended with the
structure </.....>.

• Examples include

– <html>, </html>, opens and closes the file and identifies is as an html file. The
file itself must be named with the .html extension.
– <head>, </head>, begins and ends the header.
– <title>, </title>, begins and ends the title of the header.
– <a .....>, </a>, begins and ends a link to a page
– <p>, skip a line.

• There are many other commands which can be used, and one can consult one of many
references to find information.

• Many applications have WYSIWYG ways to create and store HTML files.

1.8 UNIX

We will focus on the UNIX operating system and its variants. A full understanding of this
extensive system is beyond the scope of this course, so we shall focus only on the fun-
damentals, which are not difficult. A good source for more help with UNIX is found at
https://www.tjhsst.edu/∼dhyatt/superap/unixcmd.html. Here, we simply summarize
a few simple commonly used commands relevant to the UNIX operating system.

• pwd print working directory. This tells you where you are in the file structure.

• . A shortcut for the working directory.

• .. A shortcut for one directory up the tree structure from the working directory.

• cd changes the directory to the so-called home directory.

• cd directory This changes directories to the directory named directory, which must be
a sub-directory of the current working directory.

• cd .. This changes directories to one up from the working directory.

• cd ../.. This changes directories to two up from the working directory.

• ls This lists the elements contained within the working directory.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.8. UNIX 31

• ls -lrt This lists the elements contained within the working directory in detailed
form, also giving the ordering by files most recently modified.

• rm file This removes the file named file from the directory.

• cp file1 file2 This copies the file named file1 into the file named file2. If file2 already
exists, it is overwritten by file1.

• mv file1 file2 This 1) moves the file named file1 into the file named file2, and 2) removes
the file named file1. If file2 already exists, it is overwritten by file1.

• mkdir directory This creates the directory named directory within the working direc-
tory.

• rmdir directory This removes the directory named directory. The directory must be
empty in order to remove it.

• cat file Concatenate (i.e. print to the screen) the contents of file.

• more file Print to the screen the contents of file one screen at a time and use the space
bar to move down in the file.

• less file Like more except one can use arrows or j, k keys to navigate up and down
in file.

• UNIX command & Execute the process UNIX command in the background so as to
keep the terminal window available for user input.

• UNIX shortcut * a wildcard character which stands in for arbitrary characters.

One should note that other operating systems such as OS-X or Microsoft Windows have
file system devices such as OS-X’s “Finder” which give a WYSIWYG view of the UNIX tree
structure of organizing files. The two approaches are entirely compatible, and in fact can be
applied to arrange the same files. One is simply more visual than the other.
A file can be identified by its local address as long as the user resides in the proper
directory. For example to view the file helloworld.f90, one first insures it lives within the
present directory, and then executes

[powers@darrow2-p ame.20214]$ cat helloworld.f90

If the user resides in a different directory and still wishes to view the same file without first
navigating to that directory, one can used the absolute address, for example

[powers@darrow2-p ame.20214]$ cat /afs/nd.edu/users/powers/www/ame.20214/helloworld.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

32 CHAPTER 1. OVERVIEW

A highly useful feature of most modern UNIX implementations is the arrow keys’ ability
to recall previous commands and edit them. It is difficult to display this without an actual
demonstration.
Here we illustrate some of the power of UNIX commands. The user may want to copy all
of the sample Fortran programs from the course home space to the user’s personal home
space. Since at Notre Dame both the course home space and the user’s home space are all
within AFS, and because the course home space is a public directory, copying is easy and can
be done in a few simple commands. The sequence of commands is described as follows:

1. Sign onto a UNIX-based machine.

2. Create a new directory within your home space.

3. Change directories into that new directory.

4. Copy all files with an .f90 extension into your new directory.

5. List those files to make sure they are present.

A sample set of UNIX commands which achieves this is as follows:

[powers@darrow2-p ~]$ mkdir programs

[powers@darrow2-p ~]$ cd programs
[powers@darrow2-p ~/programs]$ ls
[powers@darrow2-p ~/programs]$ cp ~powers/www/ame.20214/2e/*.f90 .
[powers@darrow2-p ~/programs]$ ls -lrt
total 10
-rw-r--r-- 1 powers campus 65 Sep 16 12:41 ch15a.f90
-rw-r--r-- 1 powers campus 894 Sep 16 12:41 ch13d.f90
-rw-r--r-- 1 powers campus 269 Sep 16 12:41 ch13c.f90
-rw-r--r-- 1 powers campus 213 Sep 16 12:41 ch13b.f90
-rw-r--r-- 1 powers campus 210 Sep 16 12:41 ch13a.f90
-rw-r--r-- 1 powers campus 156 Sep 16 12:41 ch12g.f90
-rw-r--r-- 1 powers campus 202 Sep 16 12:41 ch12f.f90
-rw-r--r-- 1 powers campus 217 Sep 16 12:41 ch12e.f90
-rw-r--r-- 1 powers campus 152 Sep 16 12:41 ch12d.f90
-rw-r--r-- 1 powers campus 152 Sep 16 12:41 ch12c.f90
[powers@darrow2-p ~/programs]$

Here mkdir made the new directory programs. We then moved into the directory programs,
via the cd command. We then listed the files within programs and found that it was empty.
We then copied all files with the .f90 extension into the existing directory, indicated by
the “dot”, [.], shorthand for the existing directory. We then re-listed the programs in the
directory and found ten new programs listed.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

1.9. ONLINE COMPILERS 33

1.9 Online compilers

There exist a variety of open-source web-based compilers of a variety of languages, including
Fortran. These can be useful for editing and debugging, mainly because the user interface
is more intuitive than UNIX. There may be challenges in porting input and output files,
depending on the web interface chosen. An example interface is found at
http://www.tutorialspoint.com/compile fortran online.php.
A screen shot of this web browser-based interface is shown in Fig. 1.8.

Figure 1.8: Screen shot of a web browser-based Fortran interface.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

34 CHAPTER 1. OVERVIEW

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 2

Introduction to problem solving

Read C&S, Chapter 2.

Here, we summarize a few of the key concepts of this chapter. Problems are often formu-
lated in words, which can be ambiguous. Some problems are amenable to mathematical and
computational analysis, and a challenging chore is to translate imprecise words into precise
computer code. The code to solve the posed problem is based upon a logical construct known
as an algorithm. An algorithm is much like a recipe for cooking. An important element of
most good scientific computing task involves what is known as object-oriented programming.
Such an approach relies upon identifying repetitive tasks and writing a single algorithm for
a task that is commonly executed as part of solving a larger problem. These building blocks,
known as objects can be highly refined and optimized since they will be repeated many times
in the course of a calculation. One might imagine developing a sub-algorithm for performing
matrix inversion within a larger problem that requires inversion of many matrices of different
types. Rather than write separate code for each matrix inversion, one writes one subroutine
to invert matrices of arbitrary dimension, provided of course they are invertible. Then that
subroutine may be called upon many times within the more general algorithm.

35
36 CHAPTER 2. INTRODUCTION TO PROBLEM SOLVING

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 3

Introduction to programming
languages

Read C&S, Chapter 3.

One can consult the text for a fine summary of the history of programming languages.

37
38 CHAPTER 3. INTRODUCTION TO PROGRAMMING LANGUAGES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 4

Introduction to programming

Read C&S, Chapter 4.

Here, we summarize some important introductory comments regarding programming in

Fortran. We adopt many notions and language directly from C&S.

4.1 Statement types

Our language will have a variety of statement types including

1. data description statements: These set the type of data, for example, real, integer,
character, each of which have different features.

2. control structures: These are instructions for how the program is to proceed, for ex-
ample do, if.

3. data processing statements: These instructions actually process the data. If the data are
numbers, a data processing operation might include addition, x=x+1. This is a typical
data processing statement, which is mathematical, but not ordinary mathematics! It
says to take the number x, add one to it, and replace the value of x. We think of it as
an assignment statement rather than a mathematical equation. So, if we start with x
with a value of 5 the processing statement x=x+1 maps 5 into 6.

4. input and output (I/O) statements: These statements allow the program to commu-
nicate with the outside world by reading in data and writing out data. An input
statement will often have a read and an output statement will often have a write or
print.

4.2 Example program

Let us examine C&S’s program ch0401.f90:

39
40 CHAPTER 4. INTRODUCTION TO PROGRAMMING

program ch0401
!
! This program reads in and prints out a name
!
implicit none
character*20 :: first_name
!
print *,’ type in your first name.’
print *,’ up to 20 characters’
read *,first_name
print *,first_name
!
end program ch0401

Some notes adopted directly from C&S:

• Lower case characters are used. In old versions of Fortran, upper case was often the
only option. Today, Fortran is simply case-insensitive, so it does not matter. Many
programmers find it æesthetically pleasing to exclusively use lower case. Our text
sporadically reverts to the older convention of upper case. The important thing for
you to recognize is that as far as a Fortran compiler is concerned the variables are case
insensitive; e.g. a variable named VaLuE is seen by the Fortran compiler as identical
to value.

• Each line is a distinct statement.

• The order of the statements is important. First things are done first. Last things are
done last.

• The opening statement is the program name and must appear. It is useful to give the
program a meaningful name. Here, we mean the first program of Chapter 4.

• The character “!” denotes a comment statement. They are ignored by the compiler, but
are very useful to the program author and readers who are trying to understand your
program. It is a good practice for students to insert a surplus of comment statements
in their programs.

• The statement implicit none is a highly useful statement that should be included in
all of your programs. It forces you to declare the type of all variables. And if you make
a typographical error in your actual program, you will likely have mis-typed a variable
name, and the compiler will identify this error for you. Otherwise, the compiler may
simply assign a value of 0 to your mis-typed variable, and this can cause your program
to yield results which are difficult to interpret.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

4.3. DATA TYPES 41

• The statement character*20 :: first name declares the variable first name to
be a character variable. Such a variable is one whose value is an ASCII character. It
is not a number. The character here is defined to be up to 20 characters in length.
Note that character*20 works in most older Fortran compilers, but generates a com-
piler WARNING when the Fortran 2003 standard is imposed with the ifort compiler.
Note that compiler warnings are not compiler errors; a program which compiles with
warnings will still execute. To meet the 2003 standard, one can use the command
character (len=20).

• The print statement sends output to the screen. It is an I/O statement; specifically, it
is an O statement. The ∗ indicates the output is unformatted. Later, we shall see how
to format output. The print statement must be followed by a comma. Items within
single right-leaning apostrophes are those which are actually printed to the screen.

• The read statement is an I/O statement also; specifically it is an I statement. That

is, it requests input from the screen. The ∗ indicates the input data is unformatted.

• After the data has been read, it is then outputted in an unformatted fashion to the
screen.

• The end program statement indicates the end of the program.

4.3 Data types

Some important data types in Fortran are listed next:

• integer: e.g. -461; depending on the computer architecture, e.g. 32 or 64-bit machines,
the upper and lower bounds of the accepted integers can be extended. For a 64-bit
machine, we might say integer (kind=8).

• real, a seven significant digit real number, e.g. 1.234567. This is equivalent to the
statement real (kind=4).

• complex, a seven significant digit complex number, e.g. 1.234567 + 7.654321i. This is
equivalent to the statement complex (kind=4).

• real (kind=8), a fifteen significant digit real number, e.g. 1.23456789012345.

• complex (kind=8), a fifteen significant digit complex number, e.g. 1.23456789012345+

5.43210987654321i.

• real (kind=16), a thirty-three significant digit real number, e.g.

1.23456789012345678901234567890123.
This does not work on all compilers. It does work for ifort.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

42 CHAPTER 4. INTRODUCTION TO PROGRAMMING

• complex (kind=16), a thirty-three significant digit complex number, e.g.

1.23456789012345678901234567890123 + 1.23456789012345678901234567890123i.
This does not work on all compilers. It does work for ifort.
• character (len=10), a ten character string, e.g. abcdefghij.
• logical, a logical variable which has value of either .true. or .false.

4.4 Another example

Let’s look at a modification of another program from C&S. We name this file ch0402m.f90,
with the m denoting a modification from C&S.

[powers@darrow1-p ame.20214]$ cat ch0402m.f90

program ch0402m
!
! This program reads in three numbers and sums
! and averages them.
!
implicit none
integer, parameter:: p=4
real (kind=8) :: n1,n2,n3
real (kind=p) :: average1 , total1
real (kind=2*p) :: average2 , total2
complex :: average3, total3
integer :: n=3

print*, ’ type in three numbers.’

print*, ’ separated by spaces or commas’
read*,n1,n2,n3
total1 = n1+n2+n3
total2 = n1+n2+n3
total3 = n1+n2+n3
average1 = total1/n
average2 = total2/n
average3 = total3/n
print*, ’single precision total of numbers is ’, total1
print*, ’double precision total of numbers is ’, total2
print*, ’complex total of numbers is ’, total3
print*, ’single precision average of numbers is ’,average1
print*, ’double precision average of numbers is ’,average2
print*, ’complex average of numbers is ’,average3

CC BY-NC-ND. 19 January 2025, J. M. Powers.

4.4. ANOTHER EXAMPLE 43

end program ch0402m

[powers@darrow1-p ame.20214]$ ifort ch0402m.f90
[powers@darrow1-p ame.20214]$ a.out
type in three numbers.
separated by spaces or commas
1.2 10.0 20.122
single precision total of numbers is 31.32200
double precision total of numbers is 31.3220000000000
complex total of numbers is (31.32200,0.0000000E+00)
single precision average of numbers is 10.44067
double precision average of numbers is 10.4406666666667
complex average of numbers is (10.44067,0.0000000E+00)
[powers@darrow1-p ame.20214]$
Let us examine some features of this program and its output:
• We have type declarations of real (kind=4), real (kind=8), complex, and integer.
If we do not specify a kind, default values are assumed. The type declarations must
be at the beginning of the program, before one does any processing of the data.
• One may choose whether or not to initialize variables within the data assignment
statement. Here, we chose to initialize n, average1, total1, average2, and total2.
We did not initialize average3 or total3.
• We often choose exponential notation in Fortran. For example
1.23 × 104 = 1.23e4. (4.1)
• The first statement which actually processes data is total1 = n1+n2+n3, which assigns
a value to total1.
• Order must be
– program statement,
– type declarations, e.g. real, integer,
– processing and I/O statements,
– end program statement.
• Comments may appear any where after program and before end program.
• Variable names may be up to 31 characters.
• Line length can be up to 132 characters
• To continue a line, use the & character.
• If implicit none is not used, then variables beginning with the characters a − h and
o − z are by default real, and variables beginning with i − n are by default integer.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

44 CHAPTER 4. INTRODUCTION TO PROGRAMMING

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 5

Arithmetic

Read C&S, Chapter 5.

Here, we will reinforce some notions introduced earlier regarding arithmetic on a digital
computer using Fortran. There are several important issues which arise due to precision
limits of the machine as well as specific syntax. Many notions are consistent with standard
mathematics, but some idiosyncrasies arise because of finite precision as well as the need to
have different data types for integers and real numbers.

5.1 Operations
There are five key operations available: To avoid surprises, one should use parentheses liber-

addition +
subtraction -
division /
multiplication *
exponentiation **

Table 5.1: Fortran arithmetic operators

ally to avoid confusion when applying operations. If parentheses are not used, Fortran has
a complicated set of priorities for which operation takes precedence. For example, consider
the mathematical expression
5.(1. + 3.)
x= . (5.1)
4.
The following is the preferred way in Fortran to perform this operation:
x = (5.*(1.+3.))/4.
The following is acceptable as well

45
46 CHAPTER 5. ARITHMETIC

x = 5*(1.+3.)/4.

Turning this around, one might ask what is the mathematical meaning of the Fortran
assignment

x = 1.+2./3.-4.*5.

In Fortran, multiplication and division take priority over addition and subtraction, so the
expression can be interpreted mathematically as

2. 55.
x = 1. + − (4.)(5.) = − = −18.33333..... (5.2)
3. 3.

5.2 Integer division

Some unusual problems can arise when dividing by integers as shown by the following code,
ch0502.f90:

[powers@darrow2-p 2e]$ cat ch0502.f90

program ch0502
implicit none
real :: a,b,c
integer :: i
a = 1.5
b = 2.0
c = a / b
i = a / b
print *,a,b
print *,c
print *,i
end program ch0502
[powers@darrow2-p 2e]$ ifort ch0502.f90
[powers@darrow2-p 2e]$ a.out
1.500000 2.000000
0.7500000
0
[powers@darrow2-p 2e]$

Ordinary mathematics would hold that c should have the same value as i. Both are
found by forming a/b. But their values are different. This is because we have declared c to
be real and i to be integer. The number i must be an integer, and the rule applied was
to round down to the nearest integer, which here was 0.
Another example is seen next in ch0503.f90:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

5.2. INTEGER DIVISION 47

[powers@darrow2-p 2e]$ cat ch0503.f90

program ch0503
implicit none
integer :: I,J,K
real :: Answer
I = 5
J = 2
K = 4
Answer = I / J * K
print *,I
print *,J
print *,K
print *,Answer
end program ch0503
[powers@darrow2-p 2e]$ ifort ch0503.f90
[powers@darrow2-p 2e]$ a.out
5
2
4
8.000000
[powers@darrow2-p 2e]$

Here, ordinary mathematics would tell us that

5
Answer = 4 = 10. (5.3)
2
Note that Answer should in fact be an integer. But our Fortran code returned the nearby
value of Answer = 8. This is because the code first divided 5 by 2, rounding to 2. It then
multiplied by 4 to get 8. Generally, this is undesirable. In such cases, one should take care
to see that all such calculations are done with REAL variable specifications.
Why then use an integer specification at all? There are several reasons:

• The integer representation uses less data storage.

• The integer representation is exact, which is often an advantage.

• The integer representation can lead some operations to execute more efficiently, e.g.
x2 when x is real is executed faster and with more accuracy as x ∗ ∗2 than x ∗ ∗2.0.

Let us look at one final example, found in ch5a.f90:

[powers@darrow2-p 2e]$ cat ch5a.f90

program ch5a ! beginning of the program
integer::i,j

CC BY-NC-ND. 19 January 2025, J. M. Powers.

48 CHAPTER 5. ARITHMETIC

i = 1
j = 3
print*,i/j
print*,real(i)/real(j)
print*,real(i,8)/real(j,8)
print*,real(i,16)/real(j,16)
end program ch5a ! end of program
[powers@darrow2-p 2e]$ ifort ch5a.f90
[powers@darrow2-p 2e]$ a.out
0
0.3333333
0.333333333333333
0.333333333333333333333333333333333
[powers@darrow2-p 2e]$

Note that i/j maps to 0 for i = 1 and j = 3. The command real(i) gives us a single
precision real version of i, and similarly for j. The commands real(i, 8) and real(i, 16)
give double and quad precision versions of i, and similarly for j.

5.3 Binary representation of integers

Numbers are actually represented on a computer as combinations of 0 and 1 in base 2. For
example, consider the base 2 number
x = 100111011110110. (5.4)
This is actually
x = |1 ×{z214} + |0 ×{z213} + |0 ×{z212} + |1 ×{z211} + |1 ×{z210} + |1 × 9
{z2} + 0| ×
8
{z2} + |1 ×
7
{z2}
16384 0 0 2048 1024 512 0 128
6 5 4 3 2 1 0
+ 1| ×
{z2} + |1 ×
{z2} + |1 ×
{z2} + |0 ×
{z2} + |1 ×
{z2} + |1 ×
{z2} + 0| ×
{z2}, (5.5)
64 32 16 0 4 2 0
= 16384 + 2048 + 1024 + 512 + 128 + 64 + 32 + 16 + 4 + 2, (5.6)
= 20214. (5.7)
As an aside, this number is represented in Roman numerals as XXCCXIV. Returning to
binary representation, the general form for an integer i is
q
X
i=s lk 2k−1 . (5.8)
k=1

Here, s is the sign of the integer (±), lk takes on the value of 0 or 1, and q is the maximum
number of terms our computer permits.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

5.4. OVERFLOW AND UNDERFLOW LIMITS 49

5.4 Overflow and underflow limits

With a 32-bit computer architecture, we have q = 31, and the largest integer that can be
represented is
231 − 1 = 2147483647 = 2.147483647 × 109
With the 64-bit architecture more common in modern computers, we have q = 63, and the
largest integer that can be represented in principle is

263 − 1 = 9223372036854775807 = 9.223372036854775807 × 1018

Testing the ifort compiler on darrow.cc.nd.edu however shows that the largest integer in
default operating mode is actually 2147483647, when using default compilation. However,
one can invoke 64 bit integers via the compilation command ifort -integer-size 64. For
example, we find for the program ch5d.f90

[powers@darrow1-p 2e]$ cat ch5d.f90

program test
implicit none
integer :: i
print*,huge(i)
end program test
[powers@darrow1-p 2e]$ ifort -integer-size 64 ch5d.f90
[powers@darrow1-p 2e]$ a.out
9223372036854775807

that we recover the maximum integer limit for a 64 bit architecture. There is an equivalent
syntax for gfortran to invoke 64 bit integers:

[powers@darrow1-p 2e]$ cat ch5d.f90

program test
implicit none
integer :: i
print*,huge(i)
end program test
[powers@darrow1-p 2e]$ gfortran -fdefault-integer-8 ch5d.f90
[powers@darrow1-p 2e]$ a.out
9223372036854775807

For real numbers, there are similar limits. While actual values can vary, C&S report
typical real numbers in a 32-bit architecture machine must be such that

|x| ∈ [0.3 × 10−38 , 1.7 × 1038 ].

CC BY-NC-ND. 19 January 2025, J. M. Powers.

50 CHAPTER 5. ARITHMETIC

For a 64-bit architecture machine, C&S report typical limits are

|x| ∈ [0.5 × 10−308 , 0.8 × 10308 ].

Numbers outside these bounds can generate what is known as overflow and underflow errors.
Note that 0. is not in itself an underflow. Sometimes a compiler will not report an underflow
error, but instead will map the result to zero.
These can actually be checked. On darrow.cc.nd.edu using the ifort compiler, the
ranges are in fact smaller. The program ch5b.f90 yields the following:
[powers@darrow2-p 2e]$ cat ch5b.f90
program ch5b ! beginning of the program
implicit none
real(16) :: x
x=1.e10
print*,’10**4930=’,x**493
print*,’10**-4930=’,x**-493
print*,’10**5000=’,x**500
print*,’10**-5000=’,x**-500
end program ch5b ! end of program
[powers@darrow2-p 2e]$ ifort ch5b.f90
[powers@darrow2-p 2e]$ a.out
10**4930= 1.000000000000000000000000000000001E+4930
10**-4930= 1.000000000000000000000000000000032E-4930
10**5000= Infinity
10**-5000= 0.000000000000000000000000000000000E+0000
[powers@darrow2-p 2e]$
The actual bounds of the intel compiler on darrow.cc.nd.edu in quad precision, found
using the commands tiny(x) and huge(x), are

|x| ∈ [3.362103143112093506262677817321753×10−4932, 1.189731495357231765085759326628007×104932].

Numbers outside this bound are reported as Infinity, which will induce difficulties. Under-
flow numbers are mapped to zero, which can also cause surprises and consequent difficulties,
especially if division by that number is required later in the calculation! Note in double
precision using ifort or gfortran on darrow.cc.nd.edu, we find instead

|x| ∈ [2.225073858507201 × 10−308 , 1.797693134862316 × 10308 ].

In single precision using ifort or gfortran on darrow.cc.nd.edu, we find

|x| ∈ [1.1754944 × 10−38 , 3.4028235 × 1038 ].

These are similar to those reported by C&S.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

5.5. ASSIGNMENT OF UNDEFINED VARIABLES 51

5.5 Assignment of undefined variables

Lastly, we note that variables can be defined and undefined in Fortran. The compiler will
not view the existence of undefined variables as a problem. However, if one attempts to use
an undefined variable, the program may respond in an unusual manner. Let us examine how
the program ch5c.f90 behaves under the ifort compiler.
A listing of the program, its compilation with ifort and execution are given next:

[powers@darrow2-p 2e]$ cat ch5c.f90

program ch5c
implicit none
integer (kind=8) :: n=1, m !n is initialized, m is not
real (kind=8) :: x=1., y !x is initialized, y is not
open (unit=10,file=’out.txt’) !open an output file
print*,’n= ’,n
print*,’m= ’,m
print*,’n*m= ’,n*m
write(10,*) n,m,n*m !write some output to the output file
print*,’x= ’,x
print*,’y= ’,y
print*,’x*y= ’,x*y
write(10,*) x,y,x*y !write some output to the output file
end program ch5c
[powers@darrow2-p 2e]$ ifort ch5c.f90
[powers@darrow2-p 2e]$ a.out
n= 1
m= 0
n*m= 0
x= 1.00000000000000
y= 0.000000000000000E+000
x*y= 0.000000000000000E+000
[powers@darrow2-p 2e]$ cat out.txt
1 0 0
1.00000000000000 0.000000000000000E+000 0.000000000000000E+000
[powers@darrow2-p 2e]$

Note we did not initialize the integer m or the real y. In both cases, ifort assigned them
a value of zero. This can cause surprises if it is not intended for the variables to take on a
value of zero.
As an aside, we also wrote the output to a file. To do so we had to open the file, using
the command

open (unit=10,file=’out.txt’)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

52 CHAPTER 5. ARITHMETIC

This opens a file which we have named out.txt. It will be a plain ASCII text file which we
could edit with any text editor, like gedit. To populate this file, we must write to it. We
have chosen to write twice to the file via the write commands

write(10,*) n,m,n*m
write(10,*) x,y,x*y

The number 10 is linked via the open statement to the file out.txt. After execution, we
find that out.txt has been created and populated with output.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 6

Arrays 1

Read C&S, Chapter 6.

6.1 Dimension of an array

We often write lists. Often those lists have structure. Sometimes the structure is one-
dimensional; other times it is two-dimensional. It can also be higher dimensional. These
lists can be structured into arrays. One-dimensional arrays are known as vectors. Note
that these vectors will usually not have our typical geometrical interpretation from ordinary
Euclidean geometry and Gibbs vector mathematics; however, modern mathematics shows
that the traditional concepts of Euclid and Gibbs can be extended. In short, a vector Gibbs
would be comfortable with might be
 
1
x = 5 , (6.1)
9
which could be interpreted as x = 1i + 5j + 9k, a vector in three-dimensional space. In
matrix notation, it has dimension 3 × 1 in the space of two-dimensional arrays. Gibbs would
call it a vector in a three-dimensional space because the first index is 3. We instead take the
array dimension to be the number of non-unity dimensions. That is, we might also think of
the vector of Eq. (6.1) as of dimension 3 × 1 × 1 × 1 in the space of four-dimensional arrays.
But since only one of the dimensions is non-unity, it is really a one-dimensional array, or a
vector. Thus, our so-called “one-dimensional” vectors could take the form
 
1
 5 
 
 9 
 
x=  −4 .
 (6.2)
 7 
 
 18 
−74

53
54 CHAPTER 6. ARRAYS 1

The vector of Eq. (6.2) is of dimension 7 × 1 and is also a one-dimensional array. Note that a
scalar φ might be thought of as a 1 × 1 array (or for that matter a 1 × 1 × 1 three-dimensional
array:


φ= 9 . (6.3)
A scalar can be thought of as a zero-dimensional array, since it has no non-unity dimensions.
Two-dimensional arrays are known as matrices. An example might be
 
1 6 8 9
A =  2 −4 2 10 . (6.4)
−9 0 1 5
This array has dimension 3 × 4. In four-dimensional space, it could be represented as
3 × 4 × 1 × 1 array. Because only two of the dimensions are non-unity, we take A to be a
two-dimensional array.

6.2 One-dimensional arrays from time series

Let us consider an important kind of one-dimensional array in engineering mathematics, a
time series obtained by sampling an equation. This will be useful for, among other things,
generating plots. Let us imagine for example that the temperature in South Bend on a
typical September day is known to be well modeled by the following equation:
 
2πt
T (t) = a − b sin , (6.5)
c
where T is the mean temperature in K, b is the amplitude of the temperature fluctuation in
K, c is the period of the oscillation in hours, and t is the time in hours. We could imagine
that a = 270 K, b = 10 K, c = 24 hr are reasonable values. Thus, we have
 
2πt
T (t) = (270 K) − (10 K) sin . (6.6)
24 hr

6.2.1 Fixed array size

Now we may want to plot this data. Nearly all plotting routines require a finite number of
sample points. It is easily verified that if we sample the temperature every two hours, we
will generate the set of data points, given in Table 6.1.
Now we can imagine a one-dimensional array, or vector, of values for both T and for t.
Here, we have 13 values of both t and T . We might call the discrete values ti and Ti , with
i = 1, . . . , 13. So for i = 5, we have
i = 5, (6.7)
t5 = 8 hr (6.8)
T5 = 261.34 K. (6.9)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

6.2. ONE-DIMENSIONAL ARRAYS FROM TIME SERIES 55

i t (hr) T (K)
1 0 270.00
2 2 265.00
3 4 234.00
4 6 260.00
5 8 261.34
6 10 265.00
7 12 270.00
8 14 275.00
9 16 278.66
10 18 280.00
11 20 278.66
12 22 275.00
13 24 270.00

Table 6.1: Temperature versus time at discrete times

CC BY-NC-ND. 19 January 2025, J. M. Powers.

56 CHAPTER 6. ARRAYS 1

We can write a Fortran program to generate this table, embodied in ch6a.f90:

[powers@darrow2-p ame.20214]$ cat ch6a.f90
program temperature
implicit none
integer (kind=8) ::i
real (kind=8), dimension(1:13):: t,temp
real (kind=8), parameter :: a=270., b=10., c = 24.
real (kind=8) :: pi
pi = 4.*atan(1.)
print*,’pi = ’, pi
do i=1,13
t(i) = 2.*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ ifort ch6a.f90
[powers@darrow2-p ame.20214]$ a.out
pi = 3.14159274101257
1 0.000000000000000E+000 270.000000000000
2 2.00000000000000 264.999999873816
3 4.00000000000000 261.339745816451
4 6.00000000000000 260.000000000000
5 8.00000000000000 261.339746253565
6 10.0000000000000 265.000000630920
7 12.0000000000000 270.000000874228
8 14.0000000000000 275.000000883287
9 16.0000000000000 278.660254620663
10 18.0000000000000 280.000000000000
11 20.0000000000000 278.660253309321
12 22.0000000000000 274.999998611977
13 24.0000000000000 269.999998251544
[powers@darrow2-p ame.20214]$
Let us note several features of this program:
• We are attempting to do double precision calculations, with kind = 8 for real numbers.
• We are tempted to use t for time and T for temperature. That would be a mistake,
since any Fortran compiler would interpret them as identical variables. So we use t
for time and temp for temperature.
• The dimension statement tells us that both t and temp are one-dimensional arrays,
whose first index is 1 and last index is 13. This highly general statement is useful

CC BY-NC-ND. 19 January 2025, J. M. Powers.

6.2. ONE-DIMENSIONAL ARRAYS FROM TIME SERIES 57

because it reminds us of our full range of options. Alternatively, we could have used
the simpler

real (kind=8), dimension(13):: t,temp

where the first index is by default 1. Or we could have used the simpler still

real (kind=8) :: t(13),temp(13)

The second form is common, and tells us both t and temp are one-dimensional arrays
of length 13.

• We set parameter values for a, b, and c. It is good practice to leave as much generality
as possible in your program.

• While we could have defined π, we chose to let the computer do it, since we know from
trigonometry that
π
= tan−1 1. (6.10)
4
Note that in Fortran, the function tan−1 is given by atan. We printed π to the screen
to make sure our formula was correct.

• We now have an all-important do loop to initialize values for ti = t(i) and Ti =

temp(i).

• It is good practice to indent items within the do loop so that it is clear to the reader
what is in and outside of the loop.

• We print values of t(i) and temp(i) to the screen.

• We note that we seem to have only achieved single precision results! Note that at
t = 2, we should have
 
2π(2)
T = 270 − 10 sin , (6.11)
24
π 
= 270 − 10 sin , (6.12)
 6
1
= 270 − 10 , (6.13)
2
= 265. (6.14)

The result should be nearly exactly 265, but we see precision errors in the last six
digits, namely temp(2) = 264.999999873816. Why? If we truly want double precision
accuracy, we must take care that every constant is truly double precision.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

58 CHAPTER 6. ARRAYS 1

The following modification, ch6b.f90, allows for single, double, and quad precision, which
we exercise here for double precision:
[powers@darrow1-p ame.20214]$ cat ch6b.f90
program temperature
implicit none
integer, parameter :: p=8
integer (kind=8) :: i
real (kind=p), dimension(1:13):: t,temp
real (kind=p), parameter :: a=270._p, b=10._p, c = 24._p
real (kind=p) :: pi
pi = 4._p*atan(1._p)
print*,’pi = ’, pi
do i=1,13
t(i) = 2._p*(i-1)
temp(i) = a - b*sin(2._p*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow1-p ame.20214]$ ifort ch6b.f90
[powers@darrow1-p ame.20214]$ a.out
pi = 3.14159265358979
1 0.000000000000000E+000 270.000000000000
2 2.00000000000000 265.000000000000
3 4.00000000000000 261.339745962156
4 6.00000000000000 260.000000000000
5 8.00000000000000 261.339745962156
6 10.0000000000000 265.000000000000
7 12.0000000000000 270.000000000000
8 14.0000000000000 275.000000000000
9 16.0000000000000 278.660254037844
10 18.0000000000000 280.000000000000
11 20.0000000000000 278.660254037844
12 22.0000000000000 275.000000000000
13 24.0000000000000 270.000000000000
[powers@darrow1-p ame.20214]$
Note:
• The parameter p sets the precision; here, we have p = 8, for double precision.
• Every numerical constant is defined in terms of its precision, e.g.
4._p
• The results now have true double precision accuracy.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

6.2. ONE-DIMENSIONAL ARRAYS FROM TIME SERIES 59

6.2.2 Variable array size via parameter declaration

Lastly, let us generalize the program for a variable number of sample points. This is actually
the preferred way to write the program. Here, we choose to stay in single precision and give
the program listing of ch6c.f90, compilation, and execution as

[powers@darrow2-p ame.20214]$ cat ch6c.f90

program temperature
implicit none
integer, parameter :: imax=7
integer (kind=4) :: i
real (kind=4), dimension(1:imax):: t,temp
real (kind=4), parameter :: a=270., b=10., c = 24., tmax = 24.
real (kind=4) :: pi, dt
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ ifort ch6c.f90
[powers@darrow2-p ame.20214]$ a.out
pi = 3.141593
1 0.0000000E+00 270.0000
2 4.000000 261.3398
3 8.000000 261.3398
4 12.00000 270.0000
5 16.00000 278.6602
6 20.00000 278.6602
7 24.00000 270.0000
[powers@darrow2-p ame.20214]$

Note

• The program structure is similar to the two previous.

• We now have imax as an adjustable parameter. Here, we chose a value of imax = 7.

• The array size is allowed to vary when the parameter statement gives the actual size
of the array.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

60 CHAPTER 6. ARRAYS 1

• We now have a variable time step dt which will have to be calculated. Mathematically,
the formula is
tmax
∆t = . (6.15)
imax − 1
So, if we just have one time step, imax = 2 and ∆t = tmax . If we have two times
steps, imax = 3, and ∆t = tmax /2. For our imax = 7, we have six time steps with
∆t = tmax /6 = (24 hr)/6 = 4 hr.

• We calculated dt, taking care to convert integers to reals when dividing. The com-
mand real(imax) simply converts imax to its single precision real number equivalent.
Recall if we had wanted a double precision real, we would have used the command
real(imax, 8).

• The parameter i in the do loop now ranges from i = 1 to i = imax.

6.3 Simple writing to files and MATLAB plotting

We often want to plot our data. An easy way to do this is to write our output data to a file,
and then import and plot that data with another application. We shall demonstrate this
with MATLAB. We first modify our program slightly so that it opens the file temperature.out
and writes to it. This is achieved in ch6d.f90:

[powers@darrow2-p ame.20214]$ cat ch6d.f90

program temperature
implicit none
integer, parameter :: imax=13
integer (kind=4) :: i
real (kind=4), dimension(1:imax):: t,temp
real (kind=4), parameter :: a=270., b=10., c = 24., tmax = 24.
real (kind=4) :: pi, dt
open(unit=20,file=’temperature.out’)
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
write(20,*)t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ ifort ch6d.f90
[powers@darrow2-p ame.20214]$ a.out

CC BY-NC-ND. 19 January 2025, J. M. Powers.

6.3. SIMPLE WRITING TO FILES AND MATLAB PLOTTING 61

pi = 3.141593
1 0.0000000E+00 270.0000
2 2.000000 265.0000
3 4.000000 261.3398
4 6.000000 260.0000
5 8.000000 261.3398
6 10.00000 265.0000
7 12.00000 270.0000
8 14.00000 275.0000
9 16.00000 278.6602
10 18.00000 280.0000
11 20.00000 278.6602
12 22.00000 275.0000
13 24.00000 270.0000
[powers@darrow2-p ame.20214]$
Now we next write and execute a simple MATLAB script, taking care to see it resides in the
same directory as our file temperature.out. The MATLAB script is given in ptemperature.m:
clear;
load(’temperature.out’)
t = temperature(:,1);
temp = temperature(:,2);
plot(t,temp,’-o’),...
xlabel(’t (hr)’,’FontSize’,24),...
ylabel(’T (K)’,’FontSize’,24);...
set(gca,’FontSize’,20)
The contents of temperature.out are loaded into MATLAB via the load command. The first
column is assigned to time t and the second column is assigned to temperature temp. We
then make a plot, taking care to include larger than default font sizes, as well as small circles
for the individual points, and axes labels with units. The command
’FontSize’, 24
within xlabel and ylabel, changed the font size for the axes labels. The command
set(gca,’FontSize’,20)
changed the font size for the numbers on the axes. The figure is shown in Fig. 6.1. While the
MATLAB-generated plot is not bad, it has some minor flaws. The exact solution is not shown,
only a straight line interpolation between points. It is possible to reset MATLAB defaults to
avoid these flaws. Alternatively, we can import the .eps file into another application, such
as Adobe Illustrator, and edit the plot with this highly potent tool. We show in Fig. 6.2
a plot of the continuous function of Eq. (6.6) and the discrete values at the sample points,
after further processing with Adobe Illustrator.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

62 CHAPTER 6. ARRAYS 1

280

275
T (K)

270

265

260
0 5 10 15 20 25
t (hr)

Figure 6.1: Temperature versus time; figure directly from MATLAB, saved in .eps format.

T (K)
280

275

270

265

5 10 15 20 t (hr)

Figure 6.2: Continuous and discrete values of T (t); plot improved with Adobe Illustrator.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 7

Arrays 2

Read C&S, Chapter 7.

7.1 Varying the array size at run time

A significant advantage of modern Fortran is its ability to have memory sizes which can be
set outside of the program. The key to this is the allocatable attribute in a declaration
statement. For example, let us modify a program from the previous chapter, ch6c.f90, so
that the user can input the value of the number of time steps from outside of the program.
Here is the structure of the modified program, named ch7a.f90.

[powers@darrow1-p ame.20214]$ cat ch7a.f90

program temperature
implicit none
integer :: imax
integer (kind=4) :: i
real (kind=4), dimension(:), allocatable :: t,temp ! modified
real (kind=4), parameter :: a=270., b=10., c = 24., tmax = 24.
real (kind=4) :: pi, dt
print*,’ Enter number of data sampling points’ ! new data entry prompt
read*,imax ! new data entry
allocate(t(1:imax)) ! new statement for array size
allocate(temp(1:imax)) ! new statement for array size
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)

63
64 CHAPTER 7. ARRAYS 2

end do
end program temperature
[powers@darrow1-p ame.20214]$ ifort ch7a.f90
[powers@darrow1-p ame.20214]$ a.out
Enter number of data sampling points
7
pi = 3.141593
1 0.0000000E+00 270.0000
2 4.000000 261.3398
3 8.000000 261.3398
4 12.00000 270.0000
5 16.00000 278.6602
6 20.00000 278.6602
7 24.00000 270.0000
[powers@darrow1-p ame.20214]$ a.out
Enter number of data sampling points
13
pi = 3.141593
1 0.0000000E+00 270.0000
2 2.000000 265.0000
3 4.000000 261.3398
4 6.000000 260.0000
5 8.000000 261.3398
6 10.00000 265.0000
7 12.00000 270.0000
8 14.00000 275.0000
9 16.00000 278.6602
10 18.00000 280.0000
11 20.00000 278.6602
12 22.00000 275.0000
13 24.00000 270.0000
[powers@darrow1-p ame.20214]$

Note the following:

• The variables t and temp have not been dimensioned, but have been declared as arrays
with no dimensions as of yet.

• The dimensions of the arrays, imax is read in from the screen and given as direct input
by the user.

• Once read in, the array dimensions of t and temp are assigned with the allocate
function.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

7.2. HIGHER DIMENSION ARRAYS 65

x3
f (3)

T33

T32
T31

T23 f (2)

T13 T22
T21
T12
T11 x2

f (1)

x1

Figure 7.1: Sketch of the components of a stress tensor on a material element.

• The program is compiled once and executed twice, each time for a different value of
imax. The first value is 7 and the second value is 13; no recompilation is necessary.

7.2 Higher dimension arrays

We often have need to consider arrays of higher dimension than unity. Most common is the
use of two-dimensional arrays, or matrices, which in general are of dimension n × m, where
n and m are integers. If the two-dimensional array is of dimension n × n, it is said to be
square.
Let us consider the mathematical operation of matrix-vector multiplication to illustrate
the use of two-dimensional arrays in Fortran. We can in fact consider a physically motivated
problem where matrices arise. In solid mechanics, we are often given the stress tensor, really
a 3 × 3 matrix to help describe the forces on a body.
We show in Fig. 7.1 a sketch of a cube in three space where the ordinary (x, y, z) coor-
dinates are replace by (x1 , x2 , x3 ). On each face of the cube is a vector f, which represents
a so-called surface traction force. The surface traction force has units of N/m2 and rep-

CC BY-NC-ND. 19 January 2025, J. M. Powers.

66 CHAPTER 7. ARRAYS 2

resents a force per area. Each of the surface tractions can be represented in terms of its
three-dimensional Cartesian components, e.g.

f (1) = T11 e1 + T12 e2 + T13 e3 , (7.1)

f (2) = T21 e1 + T22 e2 + T23 e3 , (7.2)
f (3) = T31 e1 + T32 e2 + T33 e3 . (7.3)

Here e1 , e2 and e3 are unit vectors in the 1, 2, and 3 directions, respectively. We see a
structure in these equations, and in fact are inclined to organize the values of Tij into a
two-dimensional array, defined as the stress tensor T:
 
T11 T12 T13
T = Tij = T21 T22 T23  (7.4)
T31 T32 T33

They are equivalent to the familiar i, j, and k. Now, detailed analysis reveals that the
surface traction force f on an arbitrarily inclined surface, with unit surface normal n is given
by the following:
f = TT · n. (7.5)
Here the superscript T denotes a matrix transpose. In component form, we could say
    
f1 T11 T21 T31 n1
f2  = T12 T22 T32  n2  (7.6)
f3 T13 T23 T33 n3

In summation form, this is equivalent to

3
X
fi = nj Tji . (7.7)
j=1

For example, for i = 1, we have

f1 = n1 T11 + n2 T21 + n3 T31 . (7.8)

Recalling that any vector v can be normalized by dividing by its magnitude, we can say
v v1 e1 + v2 e2 + v3 e3
n= = p 2 . (7.9)
|v| v1 + v22 + v32
Let us build a program which assuming a known value of a stress tensor, takes an arbitrary
vector v as input, converts that vector to a unit vector n, and then finds the associated surface
traction vector f. This will be aided greatly by the use of two-dimensional arrays. However,
before entering into the actual program, let us outline our algorithm:
• initialize the two-dimensional array for the stress tensor T.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

7.2. HIGHER DIMENSION ARRAYS 67

• enter three real numbers for the components of v.

• convert v into a unit normal vector n.

• perform the matrix-vector multiplication of Eq. (7.7).

• output the results.

For our program, we will choose the stress tensor to be

   
T11 T12 T13 1 5 3
T = T21 T22 T23  = −2 4 0 . (7.10)
T31 T32 T33 −1 2 3

Note that our stress tensor is asymmetric. Usually, stress tensors are symmetric, but if
so-called “body couples,” which can be induced by magnetic forces, are present, the stress
tensor may be asymmetric.
Note that if we select a unit vector aligned with one of the coordinate axes, say n =
(1, 0, 0)T , we find
    
T11 T21 T31 1 T11
(1) T
f =T ·n =  T12 T22 T32   0 = T12  ,
  (7.11)
T13 T23 T33 0 T13
    
1 −2 −1 1 1
=  5 4 2   0 = 5 .
  (7.12)
3 0 3 0 3

This unit normal vector in the 1 direction selected the components associated with the 1
face of the unit cube.
If we select a direction, say v = (1, 2, 3)T , we then would find
√ √
|v| = 12 + 22 + 32 = 14, (7.13)

and    
√1
0.2673
 √214 
n=  14  = 0.5345 . (7.14)
√3 0.8018
14

The surface traction force associated with the surface whose unit normal is that just given
is
      
f1 1 −2 −1 0.2673 −1.6036
f = f2
  =  5 4 2  0.5345 =  5.0780  . (7.15)
f3 3 0 3 0.8018 3.2071

Here is a Fortran program, ch7b.f90, and its execution for the two values just shown:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

68 CHAPTER 7. ARRAYS 2

[powers@remote101 ame.20214]$ cat ch7b.f90

program stress ! program to calculate the surface
! traction force, f, given the stress
! tensor, T, and a user-input
! direction vector, v
implicit none
integer :: i,j
real, dimension(1:3,1:3) :: T ! declaration for 2d array (matrix)
real, dimension(1:3) :: v,n,f ! declaration for 1d arrays (vectors)
real :: magv
T(1,1) = 1. ! set values of stress tensor T
T(1,2) = 5.
T(1,3) = 3.
T(2,1) = -2.
T(2,2) = 4.
T(2,3) = 0.
T(3,1) = -1.
T(3,2) = 2.
T(3,3) = 3.
print*, ’enter v(1), v(2), v(3)’ ! enter values for the directions
read*,v(1),v(2),v(3)
magv = sqrt(v(1)**2 + v(2)**2 + v(3)**2) ! calculate the magnitude of v
do i=1,3 ! calculate the normal vector n
n(i) = v(i)/magv
end do
print*,’n= ’,n ! print the normal vector, n
do i=1,3 ! perform the matrix multiplication
f(i)=0. ! initialize f(i) to zero
do j=1,3
f(i) = f(i) + n(j)*T(j,i) ! compute the sum
end do
end do
print*,’f = ’,f ! print the surface traction, f
end program stress
[powers@remote101 ame.20214]$ ifort ch7b.f90
[powers@remote101 ame.20214]$ a.out
enter v(1), v(2), v(3)
1 0 0
n= 1.000000 0.0000000E+00 0.0000000E+00
f = 1.000000 5.000000 3.000000
[powers@remote101 ame.20214]$ a.out
enter v(1), v(2), v(3)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

7.3. TWO-DIMENSIONAL ARRAYS OF VARIABLE SIZE 69

1 2 3
n= 0.2672612 0.5345225 0.8017837
f = -1.603567 5.077964 3.207135
[powers@remote101 ame.20214]$

Note the following:

• The two-dimensional array for the stress tensor is T. Its declaration statement is

real, dimension(1:3,1:3) :: T

Two completely equivalent statements are

real, dimension(3,3) :: T
real :: T(3,3)

• We used do loops to organize two important sets of repetitive calculations: 1) the

minorly repetitive calculation of the normal vector n, and 2) the more repetitive cal-
culation of the matrix-vector product TT · n. The second calculation employed two
nested do loops. Note that the critical component of the calculation,

f(i) = f(i) + n(j)*T(j,i)

P3
is the Fortran equivalent of the mathematical operation of Eq. (7.7), fi = j=1 nj Tji .

7.3 Two-dimensional arrays of variable size

We can have two-dimensional arrays of variable size. It is also possible to let that size be
read outside of the program. An example is given in ch7d.f90:

[powers@remote105 ame.20214]$ cat ch7d.f90

program twod
implicit none
integer :: imax,jmax
integer (kind=4) :: i,j
integer (kind=4), dimension(:,:), allocatable :: x !dimension variable array
print*,’ Enter the two lengths of the array’ !variable array size
read*,imax,jmax
allocate(x(1:imax,1:jmax))
do i=1,imax
do j=1,jmax
x(i,j) = i*j
enddo

CC BY-NC-ND. 19 January 2025, J. M. Powers.

70 CHAPTER 7. ARRAYS 2

enddo
do i=1,imax
print*,(x(i,j),j=1,jmax)
enddo
end program twod
[powers@remote105 ame.20214]$ ifort ch7d.f90
[powers@remote105 ame.20214]$ a.out
Enter the two lengths of the array
3 5
1 2 3 4 5
2 4 6 8 10
3 6 9 12 15
[powers@remote105 ame.20214]$

Note

• Here, we form the array of integers whose elements are given by xij = ij.

• The variable lengths of the two-dimensional array is indicated by dimension(:,:).

• We show results for a 3 × 5 array.

7.4 Non-standard do loop increment syntax

We last note that a do loop may start at an integer not equal to 1 and it may increment by
integers not equal to one. An example is given in ch7c.f90:

[powers@darrow2-p 2e]$ cat ch7c.f90

program do
implicit none
integer :: i,imin=-4,imax=10,inc=2
do i=imin,imax,inc
print*,’i = ’,i
end do
end program do
[powers@darrow2-p 2e]$ ifort ch7c.f90
[powers@darrow2-p 2e]$ a.out
i = -4
i = -2
i = 0
i = 2
i = 4
i = 6

CC BY-NC-ND. 19 January 2025, J. M. Powers.

7.4. NON-STANDARD DO LOOP INCREMENT SYNTAX 71

i = 8
i = 10
[powers@darrow2-p 2e]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

72 CHAPTER 7. ARRAYS 2

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 8

Whole array features

Read C&S, Chapter 8.

The Fortran language has some limited features in manipulating vectors and arrays
with simple global commands, known here as whole array features. These features are fairly
characterized as clumsy but somewhat useful. Some of the whole array features are inspired
by traditional linear algebra. In brief, there are Fortran short cuts for the

• vector dot product, dot product

• matrix transpose, transpose

• matrix-vector product, matmul and

• matrix-matrix product, matmul.

Note, there is no shorthand for the matrix inverse, or for eigenvalues or eigenvectors, among
other things. These important operations are often available as part of libraries external to
the Fortran language, but easily woven into a program.
The reader will soon learn that, relative to some competing languages, especially MATLAB,
Fortran has only limited effectiveness employing compact matrix notation for linear algebra
problems. Worse still, the Fortran implementation does not always follow common linear
algebra conventions. Nevertheless, those features that exist can be useful in writing compact
code. Moreover, whatever method is used for matrix operations, Fortran will provide a very
fast execution time, which is in fact what often matters most to the practicing engineer.

73
74 CHAPTER 8. WHOLE ARRAY FEATURES

8.1 Dot product

8.1.1 Real vectors
We often wish to take the dot product of two vectors. Mathematically, we might say for
vectors of length 3
 
T

v1 X3
u · v = u1 u2 u3  v2 = u1 v1 + u2 v2 + u3 v3 =
 ui vi . (8.1)
v3 i=1

In Fortran, this is achieved by the function dot product via the command

dot_product(u,v)

As long as u and v are real, the dot product operation commutes:

uT · v = vT · u, u, v ∈ R. (8.2)

An example is shown next of the dot product of two vectors u = (1, 2, 3)T and v = (4, 5, 6)T .
It is easily seen that
 

4
uT · v = 1 2 3 5 = (1)(4) + (2)(5) + (3)(6) = 32. (8.3)
6

In Fortran, this is embodied in the program ch8a.f90:

[powers@remote101 ame.20214]$ cat ch8a.f90

program dotproduct
implicit none
real :: u(3),v(3)
u(1) = 1.
u(2) = 2.
u(3) = 3.
v(1) = 4.
v(2) = 5.
v(3) = 6.
print*,’u = ’,u
print*,’v = ’,v
print*,’u.v =’,dot_product(u,v)
print*,’v.u =’,dot_product(v,u)
end program dotproduct
[powers@remote101 ame.20214]$ ifort ch8a.f90
[powers@remote101 ame.20214]$ a.out

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.1. DOT PRODUCT 75

u = 1.000000 2.000000 3.000000

v = 4.000000 5.000000 6.000000
u.v = 32.00000
v.u = 32.00000
[powers@remote101 ame.20214]$

We note

• We could have achieved the same result with a do loop and extra coding.

• The dot product commutes, as expected for real variables.

8.1.2 Complex vectors

If u and v are complex, it can be shown that

uT · v = vT · u, u, v ∈ C, (8.4)

where the overline indicates the complex conjugate, x + iy = x − iy. This unusual definition
is to guarantee that uT · u = |u|2 is a real positive scalar. Among other things, such a
property is necessary for any viable theory of quantum physics.
Consider if
 
1+i
u = 1 + 0i . (8.5)
0+i

Then, we see the square of the magnitude of u is

 
3
X
1+i
uT · u = |u|2 = uj uj = 1 − i 1 − 0i 0 − i 1 + 0i = 2 + 1 + 1 = 4. (8.6)
j=1 0+i

In Fortran, this is embodied in the program ch8b.f90:

[powers@remote201 ame.20214]$ cat ch8b.f90

program dotproduct !calculate the square of the magnitude of
!a complex vector via the dot product
implicit none
complex :: u(3) !declare the complex vector
u(1) = (1,1) !u(1) = 1+1i
u(2) = (1,0) !u(2) = 1+0i
u(3) = (0,1) !u(3) = 0+1i
print*,’u = ’,u
print*,’u.u =’,dot_product(u,u) !find the dot product

CC BY-NC-ND. 19 January 2025, J. M. Powers.

76 CHAPTER 8. WHOLE ARRAY FEATURES

end program dotproduct

[powers@remote201 ame.20214]$ ifort ch8b.f90
[powers@remote201 ame.20214]$ a.out
u = (1.000000,1.000000) (1.000000,0.0000000E+00) (0.0000000E+00,1.000000)
u.u = (4.000000,0.0000000E+00)
[powers@remote201 ame.20214]$
We note
• The complex numbers are expressed as ordered pairs, e.g. u(1) = (1,1).

• The complex numbers were entered as integers, but were converted automatically to
floating point numbers. We probably should have used decimal points to be careful,
but it turned out not to matter in this case.

• The dot product of a complex scalar into itself indeed gives a positive real number.

• In this degenerate case, the dot product in fact commutes, only because we are dotting
a vector into itself.

• The reader can easily verify that the dot product does not commute for two different
complex vectors.

8.2 Matrix transpose

The matrix transpose simply trades elements across the diagonal. Consider the matrix
 
1 2 3
A= . (8.7)
4 5 6

Its transpose is
 
1 4
AT = 2 5 . (8.8)
3 6

In index notation, we would say

ATij = Aji . (8.9)

We implement this in the code ch8g.f90:

[powers@darrow1-p ame.20214]$ cat ch8g.f90
program tpose !This program finds the transpose of a matrix.
implicit none
integer :: A(2,3),At(3,2),i,j

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.2. MATRIX TRANSPOSE 77

A(1,1) = 1 !Initialize A
A(1,2) = 2
A(1,3) = 3
A(2,1) = 4
A(2,2) = 5
A(2,3) = 6

print*,’raw form, A = ’,A

print*,’ ’
print*,’ structured form, A = ’
do i=1,2
print*,(A(i,j), j=1,3)
enddo

At=transpose(A)
print*,’ ’
print*,’raw form, At = ’,At
print*,’ ’
print*,’ structured form, At = ’
do i=1,3
print*,(At(i,j), j=1,2)
enddo
end program tpose
[powers@darrow1-p ame.20214]$ ifort ch8g.f90
[powers@darrow1-p ame.20214]$ a.out
raw form, A = 1 4 2 5 3
6

structured form, A =
1 2 3
4 5 6

raw form, At = 1 2 3 4 5
6

structured form, At =
1 4
2 5
3 6
[powers@darrow1-p ame.20214]$

Note:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

78 CHAPTER 8. WHOLE ARRAY FEATURES

• For variety, we chose integer arrays.

• A has dimensions 2 × 3, and AT has dimensions 3 × 2.

• The command transpose works just as it is defined in linear algebra.

• The default print command prints each element of the matrix as a simple one-
dimensional list of numbers, ordered column by column.

• It probably would have been better for Fortran to have been designed so the default
print command was row by row, which would have allowed more consistency with
standard matrix algebra.

• With a special structuring, we were able to print the matrices in a more standard form.
This involved commands like print*,(A(i,j),j=1,3), as well as the do loop in which
it was embedded.

• We declared A and At to be integers, entered them as integers, and they print as

integers.

8.3 Matrix-vector multiplication

We can form the product of a matrix and a vector, as we have seen in Ch. 7. Such an
operation does not commute. When T is a matrix and n is vector,

TT · n 6= nT · T, (8.10)

unless we have the special case in which T is symmetric. Much like the intrinsic function
dot product simplified the coding, the Fortran intrinsic function matmul simplifies matrix-
vector multiplication. This is illustrated with the code ch8c.f90, where we compute
3
X
T
f = T · n, fi = TijT nj , (8.11)
j=1
3
X
f T = nT · T, fi = nj Tji . (8.12)
j=1

[powers@remote101 ame.20214]$ cat ch8c.f90

program stress ! program to calculate the surface
! traction force, f, given the stress
! tensor, T, and a user-input
! direction vector, v
implicit none
integer :: i,j

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.3. MATRIX-VECTOR MULTIPLICATION 79

real, dimension(1:3,1:3) :: T,Tt ! declaration for 2d array (matrix)

real, dimension(1:3) :: v,n,f ! declaration for 1d arrays (vectors)
real :: magv
T(1,1) = 1. ! set values of stress tensor T
T(1,2) = 5.
T(1,3) = 3.
T(2,1) = -2.
T(2,2) = 4.
T(2,3) = 0.
T(3,1) = -1.
T(3,2) = 2.
T(3,3) = 3.
Tt = transpose(T) ! define the transpose of T
print*, ’enter v(1), v(2), v(3)’ ! enter values for the directions
read*,v(1),v(2),v(3)
magv = sqrt(v(1)**2 + v(2)**2 + v(3)**2) ! calculate the magnitude of v
do i=1,3 ! calculate the normal vector n
n(i) = v(i)/magv
end do
print*,’n= ’,n ! print the normal vector, n
do i=1,3 ! perform the matrix multiplication
f(i)=0. ! initialize f(i) to zero
do j=1,3
f(i) = f(i) + n(j)*T(j,i) ! compute the sum
end do
end do
print*,’f = ’,f ! print the surface traction, f
print*, ’f = transpose(n).T = ’, matmul(n,T)
print*, ’f = transpose(T).n = ’, matmul(Tt,n)
print*, ’T.n = ’, matmul(T,n)
print*,’n.n = ’, dot_product(n,n)
end program stress
[powers@remote101 ame.20214]$ ifort ch8c.f90
[powers@remote101 ame.20214]$ a.out
enter v(1), v(2), v(3)
1 2 3
n= 0.2672612 0.5345225 0.8017837
f = -1.603567 5.077964 3.207135
f = transpose(n).T = -1.603567 5.077964 3.207135
f = transpose(T).n = -1.603567 5.077964 3.207135
T.n = 5.345225 1.603567 3.207135
n.n = 0.9999999

CC BY-NC-ND. 19 January 2025, J. M. Powers.

80 CHAPTER 8. WHOLE ARRAY FEATURES

[powers@remote101 ame.20214]$

We note

• We used the useful intrinsic operation transpose to form the transpose of several
matrices. Note that both T and TT have dimension 3 × 3.

• The function matmul(n,T) yielded the same result as our more complex program for
f. This really was the equivalent of forming f T = nT · T.

• We performed the commutation matmul(T,n) and found it yielded a different result,

confirming that the matrix vector multiplication may not commute.

• We transposed T and formed matmul(Ttranspose,n) and recovered the correct f.

• We checked using dot product(n,n) that nT · n = 1, as expected.

• While doable and efficient to code, whole array matrix operations in Fortran are
complicated relative to some other codes such as MATLAB.

• Fortran does not recognize the difference between row vectors and column vectors;
each is simply a one-dimensional list. For this problem n and nT as well as f and f T
are treated identically. This is not the case in MATLAB.

8.4 Matrix-matrix multiplication

We can also use the Fortran intrinsic matmul to multiply two matrices. They may be square
or non-square. Let us consider A · B = C, where
 
  1 1 1 2
1 2 2
A= , B = 2 3 4 5 (8.13)
1 1 1
1 3 4 5

Note if we write in the dimensions of the arrays as subscripts, we see what the matrix product
will yield,
A2×3 · B3×4 = C2×4 . (8.14)
The adjacent 3’s cancel, leaving us with a 2 × 4 result. Standard matrix multiplication then
reveals that
 
  1 1 1 2  
1 2 2  7 13 17 22
2 3 4 5 = (8.15)
1 1 1 4 7 9 12
| {z } 1 3 4 5 | {z }
A | {z } C
B

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.4. MATRIX-MATRIX MULTIPLICATION 81

In index notation, we could say

3
X
Cij = Aik Bkj . (8.16)
k=1

This is embodied in the Fortran program ch8d.f90:

[powers@darrow1-p ame.20214]$ cat ch8d.f90
program multiply !This program multiplies two rectangular matrices.
implicit none
integer :: A(2,3),B(3,4),C(2,4),i,j
A(1,1) = 1
A(1,2) = 2
A(1,3) = 2
A(2,1) = 1
A(2,2) = 1
A(2,3) = 1
B(1,1) = 1
B(1,2) = 1
B(1,3) = 1
B(1,4) = 2
B(2,1) = 2
B(2,2) = 3
B(2,3) = 4
B(2,4) = 5
B(3,1) = 1
B(3,2) = 3
B(3,3) = 4
B(3,4) = 5
C=matmul(A,B)
print*,’C = ’,C
print*,’ ’
print*,’C = ’
do i=1,2
print*,(C(i,j), j=1,4)
enddo
end program multiply
[powers@darrow1-p ame.20214]$ ifort ch8d.f90
ch8d.f90(22): (col. 3) remark: LOOP WAS VECTORIZED.
[powers@darrow1-p ame.20214]$ a.out
C = 7 4 13 7 17 9
22 12

CC BY-NC-ND. 19 January 2025, J. M. Powers.

82 CHAPTER 8. WHOLE ARRAY FEATURES

C =
7 13 17 22
4 7 9 12
[powers@darrow1-p ame.20214]$
Note:
• We used the simplest possible declaration statement.
• We had to initialize each of the elements of A and B.
• The compiler noted the loop was vectorized. At present, this is not important.
• The output C was formed via matmul. The numbers have the correct value, but one
must take care to interpret them as to how they actually fit into the array C. The
unformatted output of C is given as a column-by-column list.
• We also printed C in a traditionally structured way via a do loop.
We illustrate an alternate and more compact way to initialize matrices and do the same
task as the previous program in the code of ch8e.f90:
[powers@darrow1-p ame.20214]$ cat ch8e.f90
program multiply !This program multiplies two rectangular matrices.
implicit none
integer :: A(2,3) = & !Note this ordering convention is backwards from
(/1, 1, & !that of standard matrix algebra...
2, 1, &
2, 1 /)
integer :: B(3,4) = & !Once again a backwards ordering convention
(/1, 2, 1, &
1, 3, 3, &
1, 4, 4, &
2, 5, 5/)
integer :: C(2,4)
C=matmul(A,B)
print*,C
end program multiply
[powers@darrow1-p ame.20214]$ ifort ch8e.f90
ch8e.f90(13): (col. 3) remark: LOOP WAS VECTORIZED.
[powers@darrow1-p ame.20214]$ a.out
7 4 13 7 17 9
22 12
[powers@darrow1-p ame.20214]$
We note that

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.4. MATRIX-MATRIX MULTIPLICATION 83

• We achieved precisely the same result as the previous version of the algorithm.

• The ordering convention is precisely backwards from conventional mathematics! We

certainly have A as a 2 × 3 matrix. But it is entered in as data in a way which appears
to be appropriate for a 3 × 2 matrix. This is unfortunate, and due to the adoption of
an alternate approach to rows and columns in Fortran. Thus one needs to be careful
when performing matrix operations.

• Fortran really only keeps track of one-dimensional lists, and we entered the data for
A and B as one-dimensional lists, but we used the continuation line character & to aid
in the visual display of the matrix structure on input.

• This technique for initializing two-dimensional arrays also works for one-dimensional
arrays.

We can also achieve the same result of a traditional matrix-matrix multiplication from a
detailed triply nested do loop, found in ch8f.f90:

[powers@darrow1-p ame.20214]$ cat ch8f.f90

program multiply !This program multiplies two rectangular matrices.
implicit none
integer :: A(2,3),B(3,4),C(2,4),i,j,k
A(1,1) = 1
A(1,2) = 2
A(1,3) = 2
A(2,1) = 1
A(2,2) = 1
A(2,3) = 1
B(1,1) = 1
B(1,2) = 1
B(1,3) = 1
B(1,4) = 2
B(2,1) = 2
B(2,2) = 3
B(2,3) = 4
B(2,4) = 5
B(3,1) = 1
B(3,2) = 3
B(3,3) = 4
B(3,4) = 5
do i=1,2
do j=1,4
C(i,j) = 0
do k=1,3

CC BY-NC-ND. 19 January 2025, J. M. Powers.

84 CHAPTER 8. WHOLE ARRAY FEATURES

C(i,j) = C(i,j)+A(i,k)*B(k,j)
enddo
enddo
enddo
print*,C
end program multiply
[powers@darrow1-p ame.20214]$ ifort ch8f.f90
[powers@darrow1-p ame.20214]$ a.out
7 4 13 7 17 9
22 12
[powers@darrow1-p ame.20214]$
Note that

• Triply nested do loops pose no particular problem.

• We had to insure that i ran from 1 to 2 and j from 1 to 4, consistent with the
dimensions of C. Then k ran from 1 to 3, consistent with the dimensions of A and B.

8.5 Addition
Two arrays of the same dimension can be added. As an example, we take
   
1 2 5 6
A= , B= . (8.17)
3 4 7 8

Then
   
1+5 2+6 6 8
C=A+B= = . (8.18)
3+7 4+8 10 12

In index notation this simply says

Cij = Aij + Bij . (8.19)

This works in Fortran in an obvious way, as seen in ch8h.f90:

[powers@darrow1-p ame.20214]$ cat ch8h.f90
program add !This program adds two matrices.
implicit none
integer :: A(2,2)=(/1,3,2,4/),B(2,2)=(/5,7,6,8/),C(2,2),i,j

print*,’ A = ’
do i=1,2
print*,(A(i,j), j=1,2)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.5. ADDITION 85

enddo

print*,’ B = ’
do i=1,2
print*,(B(i,j), j=1,2)
enddo

C = A + B

print*,’ C = A+B = ’
do i=1,2
print*,(C(i,j), j=1,2)
enddo

end program add

[powers@darrow1-p ame.20214]$ ifort ch8h.f90
ch8h.f90(15): (col. 1) remark: LOOP WAS VECTORIZED.
[powers@darrow1-p ame.20214]$ a.out
A =
1 2
3 4
B =
5 6
7 8
C = A+B =
6 8
10 12
[powers@darrow1-p ame.20214]$

Note:

• We used the compact, but non-intuitive initialization of A and B as a one-dimensional

list, column by column, starting with column 1, of each matrix.

• We printed each matrix using the nested do statements which allow the natural matrix
form and structure to be displayed.

• The matrix C is indeed composed of the sum of each element of A and B. The compact
statement C = A+ B was used instead of a set of instructions involving two do loops
to sum the matrices.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

86 CHAPTER 8. WHOLE ARRAY FEATURES

8.6 Element-by-element matrix “multiplication”

The command ∗ in Fortran when applied to matrices does not imply ordinary matrix mul-
tiplication. It simply invokes a term-by-term multiplication, which is occasionally of value.
Consider
   
1 2 3 7 8 9
A= , B= . (8.20)
4 5 6 10 11 12

Note that the formal matrix multiplication does not exist because of incompatible dimensions
of A and B for this operation. However, the element-by-element “multiplication” does exist
and is given by
   
(1)(7) (2)(8) (3)(9) 7 16 27
C= = . (8.21)
(4)(10) (5)(11) (6)(12) 40 55 72

This is embodied in the program ch8i.f90:

[powers@darrow1-p ame.20214]$ cat ch8i.f90

program multiply !This program "multiplies" two matrices.
!It is not ordinary matrix multiplication
implicit none
integer :: A(2,3)=(/1,4,2,5,3,6/),B(2,3)=(/7,10,8,11,9,12/),C(2,3),i,j

print*,’ A = ’
do i=1,2
print*,(A(i,j), j=1,3)
enddo

print*,’ B = ’
do i=1,2
print*,(B(i,j), j=1,3)
enddo

C = A*B

print*,’ C = A*B = ’
do i=1,2
print*,(C(i,j), j=1,3)
enddo

end program multiply

[powers@darrow1-p ame.20214]$ ifort ch8i.f90
[powers@darrow1-p ame.20214]$ a.out

CC BY-NC-ND. 19 January 2025, J. M. Powers.

8.6. ELEMENT-BY-ELEMENT MATRIX “MULTIPLICATION” 87

A =
1 2 3
4 5 6
B =
7 8 9
10 11 12
C = A*B =
7 16 27
40 55 72
[powers@darrow1-p ame.20214]$

Similarly the Fortran statement sqrt(A) takes the square root of every element of the matrix
A. This is not a traditional matrix function.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

88 CHAPTER 8. WHOLE ARRAY FEATURES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 9

Output of results

Read C&S, Chapter 9.

The print* statement we have used thus far sends raw output to the screen, and is a
good way. But it can be æsthetically unwieldy to the extent that the reader has unnecessary
difficulty interpreting the results. To alleviate this difficulty, one can use a variety of tools
to produced output with better format. And the key to this is the use of what is known as
a format statement.

9.1 Unformatted printing

Let us first consider some clumsy unformatted output, generated by the program ch9a.f90:
[powers@darrow1-p ame.20214]$ cat ch9a.f90
program unformatoutput
implicit none
integer :: i
real :: x
do i = 1,8
x = 10.**i
print*,i,x
end do
end program unformatoutput
[powers@darrow1-p ame.20214]$ ifort ch9a.f90
[powers@darrow1-p ame.20214]$ a.out
1 10.00000
2 100.0000
3 1000.000
4 10000.00
5 100000.0
6 1000000.

89
90 CHAPTER 9. OUTPUT OF RESULTS

7 1.0000000E+07
8 1.0000000E+08
[powers@darrow1-p ame.20214]$
Note:
• The decimal points are not aligned for the real numbers.

• The output changes suddenly to exponential notation from ordinary decimal notation.

9.2 Formatted printing

We gain more control over the output with use of the format command, as shown in
ch9b.f90:
[powers@darrow1-p ame.20214]$ cat ch9b.f90
program formatoutput
implicit none
integer :: i
real :: x
do i = 1,8
x = 10.**i
print 100,i,x
100 format(i3,3x,f12.2)
end do
end program formatoutput
[powers@darrow1-p ame.20214]$ ifort ch9b.f90
[powers@darrow1-p ame.20214]$ a.out
1 10.00
2 100.00
3 1000.00
4 10000.00
5 100000.00
6 1000000.00
7 10000000.00
8 100000000.00
[powers@darrow1-p ame.20214]$
Note:
• The decimal places are now aligned.

• There are only two digits to the right of the decimal point in all real numbers.

• There is no change to exponential notation.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

9.2. FORMATTED PRINTING 91

• This is realized due to the format statement 100 format(i3,3x,f12.2). Let us break
this down.

– 100: This is a label, and is called upon by the formatted print 100. We often
place the format statement near the print statement, but this is not required. It
is also possible to have many print statements draw upon one format statement.
– i3: This says to print the number as a three digit integer. If the integer has in
fact four or more entries, the output will be problematic, but the program will
execute.
– 3x: This says to skip three spaces.
– f12.2: The f format says to print a real number in non-exponential notation.
The 12 assigns twelve spaces to the number. These spaces include the +/− signs,
the decimal point, and all digits. The .2 says to include exactly two digits to the
right of the decimal point.

Let us see what happens when we only allow ten instead of twelve slots for the real
number, and simultaneously ask for three digits right of the decimal point. To do that, we
change f12.2 to f10.3. Let’s also change x to −x. Let’s also change the i3 format to i1, so
that we only print one digit of the integer. We will also let 50 values be computed. Results
are shown in ch9c.f90:

[powers@darrow1-p ame.20214]$ cat ch9c.f90

program formatoutput2
implicit none
integer :: i
real :: x
do i = 1,50
x = -10.**i
print 100,i,x
100 format(i1,3x,f10.3)
end do
end program formatoutput2
[powers@darrow1-p ame.20214]$ ifort ch9c.f90
[powers@darrow1-p ame.20214]$ a.out
1 -10.000
2 -100.000
3 -1000.000
4 -10000.000
5 **********
6 **********
7 **********
8 **********

CC BY-NC-ND. 19 January 2025, J. M. Powers.

92 CHAPTER 9. OUTPUT OF RESULTS

9 **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* **********
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity
* -Infinity

CC BY-NC-ND. 19 January 2025, J. M. Powers.

9.2. FORMATTED PRINTING 93

[powers@darrow1-p ame.20214]$

Note:

• The last value of x that is actually displayed is −10000.000. This number has entries
in 10 slots:

– the minus sign: 1

– digits to the left of the decimal point: 5
– the decimal point: 1
– digits to the right of the decimal point: 3
– Total slots= 10

• This is consistent with the f10.3 format.

• Numbers that cannot fit into this format are displayed as **********. Note there are
ten stars.

• The last integer that displays is the last single digit integer, 9.

• At a certain value, x is not just unprintable, it has encountered an overflow error,

indicated by -Infinity in the output.

We can change the format to print many large numbers. In ch9h.f90, we change to an
f50.3 format and get

[powers@darrow1-p ame.20214]$ cat ch9h.f90

program formatoutput2
implicit none
integer :: i
real :: x
do i = 1,50
x = -10.**i
print 100,i,x
100 format(i3,3x,f50.3)
end do
end program formatoutput2
[powers@darrow1-p ame.20214]$ ifort ch9h.f90
[powers@darrow1-p ame.20214]$ a.out
1 -10.000
2 -100.000
3 -1000.000
4 -10000.000
5 -100000.000

CC BY-NC-ND. 19 January 2025, J. M. Powers.

94 CHAPTER 9. OUTPUT OF RESULTS

6 -1000000.000
7 -10000000.000
8 -100000000.000
9 -1000000000.000
10 -10000000000.000
11 -99999997952.000
12 -999999995904.000
13 -9999999827968.000
14 -100000000376832.000
15 -999999986991104.000
16 -10000000272564224.000
17 -99999998430674944.000
18 -999999984306749440.000
19 -9999999980506447872.000
20 -100000002004087734272.000
21 -1000000020040877342720.000
22 -9999999778196308361216.000
23 -99999997781963083612160.000
24 -1000000013848427855085568.000
25 -9999999562023526247432192.000
26 -100000002537764290115403776.000
27 -999999988484154753734934528.000
28 -9999999442119689768320106496.000
29 -100000001504746621987668885504.000
30 -1000000015047466219876688855040.000
31 -9999999848243207295109594873856.000
32 -100000003318135351409612647563264.000
33 -999999994495727286427992885035008.000
34 -9999999790214767953607394487959552.000
35 -100000004091847875962975319375216640.000
36 -999999961690316245365415600208216064.000
37 -9999999933815812510711506376257961984.000
38 -99999996802856924650656260769173209090.000
39 -Infinity
40 -Infinity
41 -Infinity
42 -Infinity
43 -Infinity
44 -Infinity
45 -Infinity
46 -Infinity
47 -Infinity

CC BY-NC-ND. 19 January 2025, J. M. Powers.

9.2. FORMATTED PRINTING 95

48 -Infinity
49 -Infinity
50 -Infinity
[powers@darrow1-p ame.20214]$

Note that we only achieve seven digits of accuracy and that for large numbers, the output
values have a strong pseudo-random component in many digits. This is because we only
requested single precision accuracy.
Let us do the same chore in exponential notation, using the so-called e format. So, let
us replace f10.3 by e10.3 and examine the result in in ch9d.f90:

[powers@darrow1-p ame.20214]$ cat ch9d.f90

program formatoutput3
implicit none
integer :: i
real :: x
do i = 1,8
x = -10.**i
print 100,i,x
100 format(i3,3x,e10.3)
end do
end program formatoutput3
[powers@darrow1-p ame.20214]$ ifort ch9d.f90
[powers@darrow1-p ame.20214]$ a.out
1 -0.100E+02
2 -0.100E+03
3 -0.100E+04
4 -0.100E+05
5 -0.100E+06
6 -0.100E+07
7 -0.100E+08
8 -0.100E+09
[powers@darrow1-p ame.20214]$

Note:

• There is no longer a problem with data that cannot be displayed. This is an advantage
of the e format.

• The alignment is somewhat better.

Character variables can be printed to the screen in a formatted form using the a format.
An example is given in ch9e.f90:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

96 CHAPTER 9. OUTPUT OF RESULTS

[powers@darrow1-p ame.20214]$ cat ch9e.f90

program formatoutput4
implicit none
character (12) :: a,b,c,d,e
a = ’I’
b = ’am’
c = ’an’
d = ’engineering’
e = ’enthusiast.’
print*,a,b,c,d,e
print 100, a,b,c,d,e
100 format(a1,1x,a2,1x,a2,1x,a11,1x,a11)
end program formatoutput4
[powers@darrow1-p ame.20214]$ ifort ch9e.f90
[powers@darrow1-p ame.20214]$ a.out
I am an engineering enthusiast.
I am an engineering enthusiast.
[powers@darrow1-p ame.20214]$

Note:

• The variables a, b, c, d, e are declared character variables which are allowed up to 12

letters in length. They are assigned values as ASCII strings, and their assignment must
be enclosed within single quotation marks.

• The unformatted print * statement gives an unevenly formatted result.

• The formatted print 100 statement calls upon statement 100 to dictate its output
format to the screen.

9.3 Formatted writing to files

Often we will want to write our results to a file. There are many variants on this discussed
in the text. Here we give a simple example. To write to a file, we first must open that file,
where its name is declared. Then we are allowed to write to it. To write in a formatted
fashion is nearly identical to printing in a formatted fashion. Here we modify an earlier
program to implement the writing of formatted output to the file numbers; see ch9f.f90:

[powers@darrow1-p ame.20214]$ cat ch9f.f90

program formatoutput5
implicit none
integer :: i
real :: x

CC BY-NC-ND. 19 January 2025, J. M. Powers.

9.3. FORMATTED WRITING TO FILES 97

open(unit=20,file=’numbers’)
do i = 1,8
x = -10.**i
print 100,i,x
write(20,100) i,x
100 format(i3,3x,e10.3)
end do
end program formatoutput5
[powers@darrow1-p ame.20214]$ ifort ch9f.f90
[powers@darrow1-p ame.20214]$ a.out
1 -0.100E+02
2 -0.100E+03
3 -0.100E+04
4 -0.100E+05
5 -0.100E+06
6 -0.100E+07
7 -0.100E+08
8 -0.100E+09
[powers@darrow1-p ame.20214]$ cat numbers
1 -0.100E+02
2 -0.100E+03
3 -0.100E+04
4 -0.100E+05
5 -0.100E+06
6 -0.100E+07
7 -0.100E+08
8 -0.100E+09
[powers@darrow1-p ame.20214]$

Note:

• We associated the arbitrary integer 20 with the output file we chose to name numbers.

• We did not put any identifier like .txt or .dat on the file numbers. This is fine, but
some applications will not recognize it as a text file, while others will.

• We used the same format statement to print and to write the same data; conse-
quently, the output to the screen is identical to that in the file.

If we neglect to open a file, but write to “20” anyway, most compilers will be accommodating.
The ifort compiler will obligingly open a file and give it the name fort.20. It will then
write to this file, if requested. This is shown in ch9g.f90:

[powers@darrow1-p ame.20214]$ cat ch9g.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

98 CHAPTER 9. OUTPUT OF RESULTS

program formatoutput6
implicit none
integer :: i
real :: x
do i = 1,8
x = -10.**i
write(20,100) i,x
100 format(i3,3x,e10.3)
end do
end program formatoutput6
[powers@darrow1-p ame.20214]$ ifort ch9g.f90
[powers@darrow1-p ame.20214]$ a.out
[powers@darrow1-p ame.20214]$ cat fort.20
1 -0.100E+02
2 -0.100E+03
3 -0.100E+04
4 -0.100E+05
5 -0.100E+06
6 -0.100E+07
7 -0.100E+08
8 -0.100E+09
[powers@darrow1-p ame.20214]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 10

Reading data

Read C&S, Chapter 10.

The read* statement we have used thus far reads raw input from the screen, and is a good
way to read limited amounts of data. But it can be problematic if there is a large amount of
input data that could change from case to case. In such a case, we are well advised to read
the data in from a file. And there are two ways to read the data, unformatted and formatted.
Unformatted is the simplest, but is difficult for the user to comprehend. Formatting input
data requires more effort, but is easier for the user to understand.

10.1 Unformatted reading from a file

Let us say we have available a list of numbers in a file named ch10a.in:

[powers@darrow1-p ame.20214]$ cat ch10a.in

0. 0.
1.1232 2.23
2. 4.1
3.51 7.9
5. 10.
[powers@darrow1-p ame.20214]$

There are two columns of data, but the file lacks a clean structure. Let us consider a Fortran
program to read this data and print it to the screen in a formatted fashion. We will associate
the first column with a vector x(i) and the second column with a vector y(i). The program
and its execution are found in ch10a.f90:

[powers@darrow1-p ame.20214]$ cat ch10a.f90

program readdata
implicit none
real :: x(5),y(5)

99
100 CHAPTER 10. READING DATA

integer :: i
open(10,file=’ch10a.in’)
do i=1,5
read(10,*) x(i),y(i)
print 199, i,x(i),y(i)
end do
199 format(i4,2x,e12.6,2x,e12.6)
end program readdata
[powers@darrow1-p ame.20214]$ ifort ch10a.f90
[powers@darrow1-p ame.20214]$ a.out
1 0.000000E+00 0.000000E+00
2 0.112320E+01 0.223000E+01
3 0.200000E+01 0.410000E+01
4 0.351000E+01 0.790000E+01
5 0.500000E+01 0.100000E+02
[powers@darrow1-p ame.20214]$
Note:
• The number 10 is associated with the input file ch10a.in.

• We read from unit 10, which is now our input file.

• We print the data to the screen in a formatted fashion and see that it is in fact the
same as our input data.

• We needed to know that there were only five lines in the input file. This can pose a
problem for input data files of arbitrary length.
Let us say we try to consider 10 values of x and y, but we only have 5 values in the input
file. This induces a problem, as is evident in ch10b.f90:
[powers@remote202 ame.20214]$ cat ch10b.f90
program readdata
implicit none
real :: x(10),y(10)
integer :: i
open(10,file=’ch10a.in’)
do i=1,10
read(10,*) x(i),y(i)
print 199, i,x(i),y(i)
end do
199 format(i4,2x,e12.6,2x,e12.6)
end program readdata
[powers@remote202 ame.20214]$ ifort ch10b.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

10.1. UNFORMATTED READING FROM A FILE 101

[powers@remote202 ame.20214]$ a.out

1 0.000000E+00 0.000000E+00
2 0.112320E+01 0.223000E+01
3 0.200000E+01 0.410000E+01
4 0.351000E+01 0.790000E+01
5 0.500000E+01 0.100000E+02
forrtl: severe (24): end-of-file during read, unit 10,
file /afs/nd.edu/users/powers/www/ame.20214/2e/ch10a.in
Image PC Routine Line Source
a.out 0000000000456C9E Unknown Unknown Unknown
a.out 0000000000455E9A Unknown Unknown Unknown
a.out 000000000043D936 Unknown Unknown Unknown
a.out 000000000042FAA2 Unknown Unknown Unknown
a.out 000000000042F0BE Unknown Unknown Unknown
a.out 0000000000410160 Unknown Unknown Unknown
a.out 0000000000402DE2 Unknown Unknown Unknown
a.out 000000000040290E Unknown Unknown Unknown
libc.so.6 00000032CAE1ECDD Unknown Unknown Unknown
a.out 0000000000402829 Unknown Unknown Unknown
[powers@remote202 ame.20214]$
We changed the array dimensions to 10 and tried to read 10 values. Clearly 5 values were
read and printed to the screen, But an execution error arose when the unavailable sixth value
was attempted to be read.
One common, albeit clumsy, way around this issue is to guarantee that the input data
file has a last line of data which is an unusual value. And when the main code reads this
unusual value, it knows that it is at the end of the input data file. Here we have created
ch10b.in:
[powers@remote202 ame.20214]$ cat ch10b.in
0. 0.
1.1232 2.23
2. 4.1
3.51 7.9
5. 10.
-9999. -9999.
[powers@remote202 ame.20214]$
The last line has two unusual values −9999., twice. This number should be designed to be an
unrealistic value for the user’s application. Our modified code then has arrays dimensioned
to a large value so that we can handle potentially large input files. And it has a condition
to quit reading the data file. All this is seen in the listing and execution of ch10c.f90:
[powers@remote202 ame.20214]$ cat ch10c.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

102 CHAPTER 10. READING DATA

program readdata
implicit none
integer, parameter :: imax=1000
real :: x(imax),y(imax)
integer :: i
open(10,file=’ch10b.in’)
do
read(10,*) x(i),y(i)
if(x(i).lt.-9998.)exit
print 199, i,x(i),y(i)
end do
199 format(i4,2x,e12.6,2x,e12.6)
end program readdata
[powers@remote202 ame.20214]$ ifort ch10c.f90
[powers@remote202 ame.20214]$ a.out
0 0.000000E+00 0.000000E+00
0 0.112320E+01 0.223000E+01
0 0.200000E+01 0.410000E+01
0 0.351000E+01 0.790000E+01
0 0.500000E+01 0.100000E+02
[powers@remote202 ame.20214]$
Note:
• We are prepared for as many as imax = 1000 data points, having dimensioned each
vector to this value.

• Our do loop is now unbounded by a counter.

• We exit the do loop if we find a data point with a value of x(i) less than -9998.

• If our input data file does not have the appropriate end of file statement, we will likely
find an execution error.

10.2 Formatted reading from a file

Often it is helpful for the user for input data in a file to be formatted in a way that the user
can recognize the data. Let us rewrite the program of Sec. 7.1 so that all possible input data
is read from file ch10d.in.
[powers@darrow2-p ame.20214]$ cat ch10d.in
a (Kelvin) = 270.
b (Kelvin) = 10.
c (hour) = 24.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

10.2. FORMATTED READING FROM A FILE 103

tmax (hour) = 24.

number of samples = 10
123456789*123456789*123456789*
[powers@darrow2-p ame.20214]$

We see

• There is more room for descriptive words in the input file.

• The last line is a convenience for counting spaces. It will not be read and not influence
the program.

Use of this input file will allow us to compile the program once, and if we want to change
the input file, we can simply edit it, and re-run the binary executable a.out. The program
itself, its compilation, and execution are given as

[[powers@darrow2-p ame.20214]$ cat ch10d.f90

program temperature
implicit none
integer (kind=4) :: i,imax
real (kind=4), dimension(:), allocatable :: t,temp
real (kind=4) :: a,b,c,tmax,pi,dt
open(12,file=’ch10d.in’) !open the input file
read(12,100) a !read from the input file
read(12,100) b !read from the input file
read(12,100) c !read from the input file
read(12,100) tmax !read from the input file
read(12,101) imax !read from the input file
100 format(20x,f10.5) !format of the input data for reals
101 format(20x,i10) !format of the input data for integers
allocate(t(1:imax))
allocate(temp(1:imax))
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ ifort ch10d.f90
[powers@darrow2-p ame.20214]$ a.out
pi = 3.141593

CC BY-NC-ND. 19 January 2025, J. M. Powers.

104 CHAPTER 10. READING DATA

1 0.0000000E+00 270.0000
2 2.666667 263.5721
3 5.333333 260.1519
4 8.000000 261.3398
5 10.66667 266.5798
6 13.33333 273.4202
7 16.00000 278.6602
8 18.66667 279.8481
9 21.33333 276.4279
10 24.00000 270.0000
[powers@darrow2-p ame.20214]$

Note:

• We read in values for a, b, c, tmax, and imax from file ch10d.in

• We chose to read one value for each read; we could have restructured the reading so
that one read statement read several variables.

• The 100 format statement said to ignore the first 20 spaces in the line, and read as a
real variable that which occupies the next 10 spaces, with up to five decimal points.
This employed the f10.5 format. For exponential notation data, we could have chosen
e10.5, for example.

• The 101 format statement said to ignore the first 20 spaces in the line, and read as an
integer variable that which occupies the next 10 spaces.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 11

I/O Concepts

Read C&S, Chapter 11.

There is some additional information on Input/Output concepts in the text. The inter-
ested student should review the appropriate material.

105
106 CHAPTER 11. I/O CONCEPTS

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 12

Functions

Read C&S, Chapter 12.

Functions are a useful aspect of the Fortran language. Many intrinsic functions are
built into the language and simplify many mathematical operations. These are pre-defined
functions. It is also possible to build user-defined functions. Let us examine both types.

12.1 Pre-defined functions

Fortran has a large number of pre-defined functions whose general structure is of the form

y = function(x)

Here x is the input, y is the output, and function is the intrinsic Fortran function. One
example is the exponential function. The Fortran version of

y = ex , (12.1)

is

y = exp(x)

Here exp is the intrinsic function.

We list some of the more common Fortran functions in Table 12.1, similar to that given
in C&S. The text of C&S has a more complete discussion.
As might be expected, these functions may have limitations on their domain. Consider
the square root function applied to a real number and to a complex number in ch12a.f90:

[powers@darrow2-p ame.20214]$ cat ch12a.f90

program sqroot
real :: x = -1.
print*,sqrt(x)

107
108 CHAPTER 12. FUNCTIONS

Function Action Example

int conversion to integer j = int(x)

real conversion to real x = real(j)
cmplx conversion to complex z = cmplx(x)
abs absolute value x = abs(x)
sqrt square root (positive) x = sqrt(y)
exp exponential y = exp(x)
log natural logarithm x = log(y)
log10 common logarithm x = log10(y)
sin sine x = sin(y)
cos cosine x = cos(y)
tan tangent x = tan(y)
asin arcsine y = asin(x)
acos arccosine y = acos(x)
atan arctangent y = atan(x)
sinh hyperbolic sine x = sinh(y)
cosh hyperbolic cosine x = cosh(y)
tanh hyperbolic tangent x = tanh(y)
asinh inverse hyperbolic sine y = asinh(x)
acosh inverse hyperbolic cosine y = acosh(x)
atanh inverse hyperbolic tangent y = atanh(x)

Table 12.1: List of some common Fortran pre-defined functions.

end program sqroot

[powers@darrow2-p ame.20214]$ ifort ch12a.f90
[powers@darrow2-p ame.20214]$ a.out
NaN
[powers@darrow2-p ame.20214]$

Note:

• We set x to be real and assigned it a negative value.

• We then tried to take the square root of a negative real number. Because the square
root of a negative real number is not a real number, the program refuses to perform
the calculation and returns the value NaN, which stands for Not a Number.

Let us declare x to be complex and take its square root; see ch12b.f90:

[powers@darrow2-p ame.20214]$ cat ch12b.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12.2. USER-DEFINED FUNCTIONS 109

program sqroot
complex :: x = (-1.,0.)
print*,sqrt(x)
end program sqroot
[powers@darrow2-p ame.20214]$ ifort ch12b.f90
[powers@darrow2-p ame.20214]$ a.out
(0.0000000E+00,1.000000)
[powers@darrow2-p ame.20214]$

• A correct value of 0 + i is returned.

√
• The actual value of −1 is 0 ± i; we must recognize that only one of the roots is
returned by sqrt.

In a similar manner, functions such as sin−1 , cos−1 , or tan−1 are multi-valued, but Fortran
will only return one value. For example, while we know that
π
sin−1 (1) = + 2nπ, n = . . . , −2, −1, 0, 1, 2, . . . (12.2)
2
Fortran returns only

asin(1.) = 1.570796

which we see is a single precision representation of π/2. In fact, Fortran defines asin so
that its range is

sin−1 (x) = asin(x) ∈ [−π/2, π/2]. (12.3)

Note also that the domain of asin must be such that

x ∈ [−1, 1]. (12.4)

The user must take care that the appropriate value of the multi-valued inverse functions
is being used.

12.2 User-defined functions

We may use a type of subroutine, a user-defined function, to simplify repeated calculations.
Such user-defined functions are a part of a so-called object-oriented program. Object-oriented
programming is often closely associated with languages such as C++ or Java, which employ
the so-called class to serve a similar role as a subroutine. In Fortran the module is roughly
equivalent to the class of C++ or Java and brings much of the same functionality. Certainly
it is possible to write Fortran user-defined functions without the use of module, but the
modern user is advised to employ this feature for stronger compatibility across different

CC BY-NC-ND. 19 January 2025, J. M. Powers.

110 CHAPTER 12. FUNCTIONS

computing platforms, it is possible and useful to have Fortran-generated code interact with
C++ or Java code.
Let us say we have a generic function f (x) and wish to have a program to build a list of
values of x and f (x) for x ∈ [0, 1]. We want the main program to be general, and to have
the ability to plug in a function from somewhere else. We will take

y = f (x) = sin(2πx), x ∈ [0, 1]. (12.5)

We can explore this in the following set of example programs, each of which employs a
different program structure to achieve the same end.

12.2.1 Simple structure

The following gives a simple structure for evaluating a user-defined function, based on a
module. It is found in ch12c.f90:

[powers@remote102 ame.20214]$ cat ch12c.f90

!---------------------------------------------------------------------------------
module myfunction !This is a module declaration
contains
real function f(x) !This is the function subroutine
implicit none
real, intent(in) :: x !Declare x as a real input variable
real :: pi
pi = 4.*atan(1.)
f = sin(2.*pi*x)
end function f
end module myfunction
!
!---------------------------------------------------------------------------------
!
program fofx !This is the main program--it comes last
use myfunction !The main program draws upon myfunction
implicit none
real :: x,y,xmin=0.,xmax=1.,dx
integer :: i,n
print*,’enter number of points, n>1’
read*,n
dx = (xmax-xmin)/real(n-1.)
do i=1,n
x = xmin + dx*(i-1.)
y = f(x) !Here we call on the function subroutine
print 100,x,y

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12.2. USER-DEFINED FUNCTIONS 111

100 format(f12.6,2x,f12.6)
end do
end program fofx
!---------------------------------------------------------------------------------
[powers@remote102 ame.20214]$ ifort ch12c.f90
[powers@remote102 ame.20214]$ a.out
enter number of points, n>1
13
0.000000 0.000000
0.083333 0.500000
0.166667 0.866025
0.250000 1.000000
0.333333 0.866025
0.416667 0.500000
0.500000 0.000000
0.583333 -0.500000
0.666667 -0.866025
0.750000 -1.000000
0.833333 -0.866025
0.916667 -0.500000
1.000000 0.000000
[powers@remote102 ame.20214]$

Note:

• The function subroutine f(x) is embedded within the module named myfunction,
which appears first in the program.

• The module syntax requires that we identify that which it contains.

• For function subroutines, we exercise the option of reserving some variables for in-
put. Here our input variable is x, and it is reserved as input by the statement real,
intent(in) :: x.

• The function must have a type, here real.

• The action of the function is to take a value of x and map it into f and return that
value to the main program.

• The main program appears last. It calls upon the function subroutine with the state-
ment f(x).

• We must inform the main program that a function subroutine module here myfunction,
is to be employed. That is achieved by use myfunction, which appears before variable
declaration statements.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

112 CHAPTER 12. FUNCTIONS

• The user must have defined f(x), since it is not a pre-defined Fortran function.

• We can extend this syntax to evaluate functions of N variables of the form f (x1 , x2 , x3 , . . . , xN ).
That is to say, functions can take several inputs and map them to a single output.

12.2.2 Use of dummy variables

The previous example used x in the main program and in the function subroutine. The
function subroutine allows more flexibility. Let us consider our general function to be of the
form

f (t) = sin(2πt). (12.6)

Let us further define

y = f (x), (12.7)
x
z = f . (12.8)
2
Thus, with our general form for f , we expect

y = sin(2πx), (12.9)
z = sin(πx). (12.10)

This is illustrated with the program ch12f.f90

[powers@remote102 ame.20214]$ cat ch12f.f90
!---------------------------------------------------------------------------------
module myfunction !This is a module declaration
contains
real function f(t) !This is the function subroutine
implicit none
real, intent(in) :: t !Declare t as a real input variable
real :: pi
pi = 4.*atan(1.)
f = sin(2.*pi*t)
end function f
end module myfunction
!
!---------------------------------------------------------------------------------
!
program fofx !This is the main program--it comes last
use myfunction !The main program draws upon myfunction
implicit none
real :: x,y,z,xmin=0.,xmax=1.,dx

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12.2. USER-DEFINED FUNCTIONS 113

integer :: i,n
print*,’enter number of points, n>1’
read*,n
dx = (xmax-xmin)/real(n-1.)
do i=1,n
x = xmin + dx*(i-1.)
y = f(x) !Here we call on the function subroutine
z = f(x/2.) !Here we again call on the function subroutine
print 100,x,y,z
100 format(f12.6,2x,f12.6,2x,f12.6)
end do
end program fofx
!---------------------------------------------------------------------------------
[powers@remote102 ame.20214]$ ifort ch12f.f90
[powers@remote102 ame.20214]$ a.out
enter number of points, n>1
13
0.000000 0.000000 0.000000
0.083333 0.500000 0.258819
0.166667 0.866025 0.500000
0.250000 1.000000 0.707107
0.333333 0.866025 0.866025
0.416667 0.500000 0.965926
0.500000 0.000000 1.000000
0.583333 -0.500000 0.965926
0.666667 -0.866025 0.866025
0.750000 -1.000000 0.707107
0.833333 -0.866025 0.500000
0.916667 -0.500000 0.258819
1.000000 0.000000 0.000000
[powers@remote102 ame.20214]$

12.2.3 Sharing data using module

Sometimes we wish to share data between the main program and the function subroutine.
There are many ways to do this; a good way is to use the module option in a slightly different
fashion. Let us say that in our example program for this chapter that we would like both
the main program and the function to know what the value of π is. And we only want
to compute it once. We can introduce a module at the beginning of a program to build
shared data, and then import that data into the elements of the program that require it. An
example is shown in ch12g.f90:
[powers@remote102 ame.20214]$ cat ch12g.f90

CC BY-NC-ND. 19 January 2025, J. M. Powers.

114 CHAPTER 12. FUNCTIONS

!---------------------------------------------------------------------------------
module commondata !This sets up data which will be shared.
implicit none
real :: pi=4.*atan(1.)
end module commondata
!
!---------------------------------------------------------------------------------
!
module myfunction !This is a module declaration
contains
real function f(t) !This is the function subroutine
use commondata !Import the common data
implicit none
real, intent(in) :: t !Declare t as a real input variable
f = sin(2.*pi*t) !Employ the common data, pi
end function f
end module myfunction
!
!---------------------------------------------------------------------------------
!
program fofx !This is the main program--it comes last
use commondata !Import the common data
use myfunction !The main program draws upon myfunction
implicit none
real :: x,y,z,xmin=0.,xmax=1.,dx
integer :: i,n
print*, pi !print the common data
print*,’enter number of points, n>1’
read*,n
dx = (xmax-xmin)/real(n-1.)
do i=1,n
x = xmin + dx*(i-1.)
y = f(x) !Here we call on the function subroutine
z = f(x/2.) !Here we again call on the function subroutine
print 100,x,y,z
100 format(f12.6,2x,f12.6,2x,f12.6)
end do
end program fofx
!---------------------------------------------------------------------------------
[powers@remote102 ame.20214]$ ifort ch12g.f90
[powers@remote102 ame.20214]$ a.out
3.141593

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12.2. USER-DEFINED FUNCTIONS 115

enter number of points, n>1

13
0.000000 0.000000 0.000000
0.083333 0.500000 0.258819
0.166667 0.866025 0.500000
0.250000 1.000000 0.707107
0.333333 0.866025 0.866025
0.416667 0.500000 0.965926
0.500000 0.000000 1.000000
0.583333 -0.500000 0.965926
0.666667 -0.866025 0.866025
0.750000 -1.000000 0.707107
0.833333 -0.866025 0.500000
0.916667 -0.500000 0.258819
1.000000 0.000000 0.000000
[powers@remote102 ame.20214]$

Note that neither the main program nor the function subroutine constructed pi; moreover, pi
was not declared in either. Instead, pi was declared and initialized in module commondata,
and commondata was imported into the main program and the function via the command
use. The notion of the module is discussed further in Ch. 21.

12.2.4 Splitting functions into separate files

It is sometimes convenient to split the main program and function subroutines into separate
files. Then we can compile them one at a time if we want and combine them later. Let us
take the main program to be in the file ch12d.f90, given next:

[powers@remote102 ame.20214]$ cat ch12d.f90

program fofx !This is the main program--it comes last
use myfunction !The main program draws upon myfunction
implicit none
real :: x,y,xmin=0.,xmax=1.,dx
integer :: i,n
print*,’enter number of points, n>1’
read*,n
dx = (xmax-xmin)/real(n-1.)
do i=1,n
x = xmin + dx*(i-1.)
y = f(x) !Here we call on the function subroutine
print 100,x,y
100 format(f12.6,2x,f12.6)
end do

CC BY-NC-ND. 19 January 2025, J. M. Powers.

116 CHAPTER 12. FUNCTIONS

end program fofx

[powers@remote102 ame.20214]$

Then let us take the function subroutine to reside in ch12e.f90:

[powers@remote102 ame.20214]$ cat ch12e.f90

module myfunction !This is a module declaration
contains
real function f(x) !This is the function subroutine
implicit none
real, intent(in) :: x !Declare x as a real input variable
real :: pi
pi = 4.*atan(1.)
f = sin(2.*pi*x)
end function f
end module myfunction
[powers@remote102 ame.20214]$

Now, we can do a different kind of compilation of the function subroutine module first
via the command

[powers@remote102 ame.20214]$ ifort -c ch12e.f90

This -c option generates the binary file ch12e.o. It is binary, but it is not executable
on its own. To work, it must be linked to a driver program. We can achieve the linkage to
the driver program ch12d.f90 by compiling ch12d.f90 and informing it that it should be
linked to ch12e.o via the compilation command

[powers@darrow2-p ame.20214]$ ifort ch12d.f90 ch12e.o

This command generates the binary executable a.out. Note that if we did not link to
ch12e.o, the main program ch12d.f90 would not compile, because it was missing f(x).
Here is the execution of a.out for 7 points:

[powers@darrow2-p ame.20214]$ a.out

enter number of points, n>1
7
0.000000 0.000000
0.166667 0.866025
0.333333 0.866025
0.500000 0.000000
0.666667 -0.866025
0.833333 -0.866025
1.000000 0.000000
[powers@darrow2-p ame.20214]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

12.2. USER-DEFINED FUNCTIONS 117

Another option is to use the -c option on both .f90 files first and create two .o files,
then use the compiler to merge them to create a.out:

[powers@remote102 ame.20214]$ ifort -c ch12d.f90

[powers@remote102 ame.20214]$ ifort -c ch12e.f90
[powers@remote102 ame.20214]$ ifort ch12d.o ch12e.o
[powers@remote102 ame.20214]$ a.out
enter number of points, n>1
7
0.000000 0.000000
0.166667 0.866025
0.333333 0.866025
0.500000 0.000000
0.666667 -0.866025
0.833333 -0.866025
1.000000 0.000000
[powers@remote102 ame.20214]$

12.2.5 Creation of executable script files

One advantage of the UNIX system is that one can create files which actually contain a list of
commands and execute them at once. In the previous example, we had to compile ch12e.f90
with the -c option, then link and compile this with ch12d.f90. Then we had to execute
the binary executable with a.out. Let us create a single file runch12de which performs all
these tasks:

[powers@remote102 ame.20214]$ cat runch12de

ifort -c ch12d.f90
ifort -c ch12e.f90
ifort ch12d.o ch12e.o
a.out
[powers@remote102 ame.20214]$

The file has our chores combined into a single list. We next have to one time only declare
this file to be executable. In UNIX this is achieved via

[powers@darrow2-p ame.20214]$ chmod u+x runch12de

The so-called “change mode” command, chmod, changes permissions on files. We have
changed the permission for the user, u, so as to add, +, the ability to execute, x, the file. So
entering the new UNIX command runch12de compiles and executes, generating

[powers@darrow2-p ame.20214]$ runch12de

enter number of points, n>1

CC BY-NC-ND. 19 January 2025, J. M. Powers.

118 CHAPTER 12. FUNCTIONS

8
0.000000 0.000000
0.142857 0.781832
0.285714 0.974928
0.428571 0.433883
0.571429 -0.433884
0.714286 -0.974928
0.857143 -0.781831
1.000000 0.000000
[powers@darrow2-p ame.20214]$

A more powerful way to build executable programs is through the UNIX utility Make, which
we will not discuss further here, but offers many advantages to the interested user.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 13

Control structures

Read C&S, Chapter 13.

There are an important set of logical structures to control Fortran programs. We have
seen some of these before. The most critical are based on either

• if, then, endif: a binary choice, or

• do, enddo: an iteration.

For binary choices, we also have the constructs if, then, else if, endif. For iterations,
we have several constructs:

• do i=1,n,

• do while,

• do ...if, then exit, enddo,

• others.

A typical if statement will require a syntax for the condition to be met. For example,
the English sentence, “If today is Thursday, then attend AME 20214,” has a Fortran-like
structure

if (today is Thursday) then

attend AME 20214.

We need a syntax for English words like “is.” That syntax and others are listed in Table 13.1,
adapted from C&S: An example of an if-based structure is obtained if we seek the real roots
of the quadratic equation

ax2 + bx + c = 0, a, b, c ∈ R. (13.1)

119
120 CHAPTER 13. CONTROL STRUCTURES

Operator Meaning Type

== equal to relational
/= not equal to relational
>= greater than or equal to relational
<= less than or equal to relational
< less than relational
> greater than relational
.and. and logical
.or. or logical
.not. not logical

Table 13.1: Control structures in Fortran.

We know the solution is √

−b ± b2 − 4ac
x= . (13.2)
2a
Moreover, we know there are two real roots if

b2 − 4ac > 0, (13.3)

there are two repeated real roots if

b2 − 4ac = 0, (13.4)

and there are no real roots if

b2 − 4ac < 0. (13.5)

Given a, b, and c, we should be able to construct a Fortran program to identify the real
roots, if they exist, and to also tell us if no real roots exist. This is achieved in a small
modification of a program given by C&S, found in ch13a.f90:

[powers@remote202 ame.20214]$ cat ch13a.f90

PROGRAM ch13a
IMPLICIT NONE
REAL :: a , b , c , term , root1 , root2
!
! a b and c are the coefficients of the terms
! a*x**2+b*x+c
! find the roots of the quadratic, root1 and root2
!

CC BY-NC-ND. 19 January 2025, J. M. Powers.

 121

PRINT*,’ GIVE THE COEFFICIENTS a, b AND c’

READ*,a,b,c
term = b*b - 4.*a*c
! if term < 0, roots are complex
! if term = 0, roots are equal
! if term > 0, roots are real and different
IF(term < 0.0)THEN
PRINT*,’ ROOTS ARE COMPLEX’
ELSEIF(term > 0.0)THEN
term = SQRT(term)
root1 = (-b+term)/2./a
root2 = (-b-term)/2./a
PRINT*,’ ROOTS ARE ’,root1,’ AND ’,root2
ELSE
root1 = -b/2./a
PRINT*,’ TWO EQUAL ROOTS, EACH IS ’,root1
ENDIF
END PROGRAM ch13a
[powers@remote202 ame.20214]$ ifort ch13a.f90
[powers@remote202 ame.20214]$ a.out
GIVE THE COEFFICIENTS a, b AND c
1 -3 2
ROOTS ARE 2.000000 AND 1.000000
[powers@remote202 ame.20214]$
Note
• conditions lie inside the parentheses, e.g. term > 0.0.

• there is a logical structure to the if, then, elseif, then, else, endif.
We can also use logical constructs, such as .and. to effect. Let us say we want to plot a
so-called “top hat” function:

 0, x ∈ (−∞, 1),
f (x) = 1, x ∈ [1, 2], (13.6)

0, x ∈ [2, ∞).

The nature of the function f (x) is that it has explicit conditional dependency on x which
is appropriate for an if, then implementation in Fortran. We give such an implementation
in ch13b.f90:
[powers@remote102 ame.20214]$ cat ch13b.f90
!---------------------------------------------------------------------------------
module myfunction !This is a module declaration

CC BY-NC-ND. 19 January 2025, J. M. Powers.

122 CHAPTER 13. CONTROL STRUCTURES

contains
real function f(t) !This is the function subroutine
implicit none
real, intent(in) :: t !Declare t as a real input variable
if(t<1.0.or.t>2.0) then
f = 0.0
else
f = 1.0
endif
end function f
end module myfunction
!
!---------------------------------------------------------------------------------
!
program fofx !This is the main program--it comes last
use myfunction !The main program draws upon myfunction
implicit none
real :: x,y,xmin=-3.,xmax=3.,dx
integer :: i,n
print*,’enter number of points, n>1’
read*,n
dx = (xmax-xmin)/real(n-1.)
do i=1,n
x = xmin + dx*(i-1.)
y = f(x) !Here we call on the function subroutine
print 100,x,y
100 format(f12.6,2x,f12.6)
end do
end program fofx
!---------------------------------------------------------------------------------
[powers@remote102 ame.20214]$ ifort ch13b.f90
[powers@remote102 ame.20214]$ a.out
enter number of points, n>1
11
-3.000000 0.000000
-2.400000 0.000000
-1.800000 0.000000
-1.200000 0.000000
-0.600000 0.000000
0.000000 0.000000
0.600000 0.000000
1.200000 1.000000

CC BY-NC-ND. 19 January 2025, J. M. Powers.

 123

f (x)
1.0

0.8

0.6

0.4

0.2

x
-3 -2 -1 1 2 3

Figure 13.1: Plot of top hat function f (x) for x ∈ [−3, 3], discrete and exact values.

1.800000 1.000000
2.400000 0.000000
3.000000 0.000000
[powers@remote102 ame.20214]$

Note:

• We elected to use the function subroutine to evaluate f , here expressed generically as

f (t).

• Within the function subroutine, we used the conditional if, then, else, endif
structure.

• The conditional for the if statement involved the Boolean logical operator .or.

We give a plot of the exact representation of f (x) and the eleven discrete points we chose to
evaluate in Fig. 13.1.
We can illustrate a do while structure with the following problem. Use the Euler method
to solve the ordinary differential equation

dy
= λy, y(0) = yo , t ∈ [0, tstop ], (13.7)
dt
for λ = −1, yo = 1, tstop = 1. Euler’s method has us approximate the first derivative as

dy yn+1 − yn
∼ , (13.8)
dt ∆t

CC BY-NC-ND. 19 January 2025, J. M. Powers.

124 CHAPTER 13. CONTROL STRUCTURES

and evaluate the right hand side of the differential equation at the n time step. This gives
rise to
yn+1 − yn
= λyn , (13.9)
∆t
yn+1 = yn + λ∆tyn . (13.10)

We choose ∆t = 0.1. A Fortran implementation of this is given in ch13c.f90:

[powers@remote202 ame.20214]$ cat ch13c.f90

program euler
implicit none
integer :: n
real, parameter :: tstop=1., yo = 1., lambda = -1., dt=1.e-1
real :: y,ynew,t
t = 0.
y = yo
n=0
print*,t,y,yo*exp(lambda*t)
do while (t<tstop) !do while structure for loop
ynew = y+lambda*dt*y
n = n+1
t = dt*n
y = ynew
print*,t,y,yo*exp(lambda*t)
enddo
print*,’error ’,y-yo*exp(lambda*t)
end program euler
[powers@remote202 ame.20214]$ ifort ch13c.f90
[powers@remote202 ame.20214]$ a.out
0.0000000E+00 1.000000 1.000000
0.1000000 0.9000000 0.9048374
0.2000000 0.8100000 0.8187308
0.3000000 0.7290000 0.7408182
0.4000000 0.6561000 0.6703200
0.5000000 0.5904900 0.6065307
0.6000000 0.5314410 0.5488116
0.7000000 0.4782969 0.4965853
0.8000000 0.4304672 0.4493290
0.9000000 0.3874205 0.4065697
1.000000 0.3486784 0.3678795
error -1.9201010E-02
[powers@remote202 ame.20214]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

 125

y
1.0

0.8

0.6

0.4
t
0.2 0.4 0.6 0.8 1.0

Figure 13.2: Plot of the exact solution to dy/dt = −y, y(0) = 1 (y = exp(−t)), along with
its approximation from the Euler method with ∆t = 0.1.

We note that the do while structure has been implemented, and it is an obvious and conve-
nient program control structure for this problem. We give a plot of the exact representation
of y(x) and the eleven discrete points we chose to evaluate with the Euler method in Fig. 13.2.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

126 CHAPTER 13. CONTROL STRUCTURES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 14

Characters

Read C&S, Chapter 14.

One should read the text of C&S to learn further about characters within Fortran.

127
128 CHAPTER 14. CHARACTERS

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 15

Complex

Read C&S, Chapter 15.

Complex mathematics is a rich and varied subject. At its foundation lays the number i
where
√
i ≡ −1. (15.1)

If we consider numbers z to be complex, then we can think of them as

z = x + iy. (15.2)

Here x and y are real numbers. We say mathematically,

z ∈ C1 , x ∈ R1 , y ∈ R1 . (15.3)

where C and R denote the complex and real numbers, respectively, and the superscript 1
denotes scalars. The real part of z is defined as

ℜ(z) = ℜ(x + iy) = x. (15.4)

The Fortran equivalent is the function real:

ℜ(z) = real(z). (15.5)

The imaginary part of z is defined as

ℑ(z) = ℑ(x + iy) = y. (15.6)

The Fortran equivalent is the function aimag:

ℑ(z) = aimag(z). (15.7)

Many compilers also accept imag.

129
130 CHAPTER 15. COMPLEX

If z = x + iy, then the so-called complex conjugate, z, is defined as

z ≡ x − iy. (15.8)
The Fortran command for taking the complex conjugate of a complex variable z is
z = (x + iy) = conjg(z). (15.9)
Note that
zz = (x + iy)(x − iy) = x2 + ixy − ixy − (i)2 y 2 = x2 + y 2. (15.10)
We call this the square of magnitude of the complex number:
|z|2 = zz = x2 + y 2, |z|2 = R1 . (15.11)
The magnitude of a complex number is obviously a positive real number. We can also get
the magnitude of a complex number by the command abs(z).
For z = 3 + 4i, this is illustrated in ch15a.f90:
[powers@remote102 ame.20214]$ cat ch15a.f90
program cplex
complex :: z,zconj,zmag,zabs
real :: x,y
z = (3.,4.)
zconj = conjg(z)
zmag = sqrt(z*zconj)
zabs = abs(z)
x = real(z)
y = aimag(z)
print*,’z = x+iy = ’,z
print*,’zbar = ’,zconj
print*,’sqrt(x^2+y^2) = ’,zmag
print*,’|z| = ’,zabs
print*,’Re(z) = x = ’,x
print*,’Im(z) = y = ’,y
end program cplex
[powers@remote102 ame.20214]$ ifort ch15a.f90
[powers@remote102 ame.20214]$ a.out
z = x+iy = (3.000000,4.000000)
zbar = (3.000000,-4.000000)
sqrt(x^2+y^2) = (5.000000,0.0000000E+00)
|z| = (5.000000,0.0000000E+00)
Re(z) = x = 3.000000
Im(z) = y = 4.000000
[powers@remote102 ame.20214]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

 131

Note, we have given Fortran output for the mathematical operations

1. z = 3 + 4i,

2. z = 3 − 4i,
√ p √ √
3. |z| = zz = (3 + 4i)(3 − 4i) = 32 + 42 = 25 = 5 + 0i.

4. ℜ(z) = ℜ(3 + 4i) = 3.

5. ℑ(z) = ℑ(3 + 4i) = 4.

Obviously operations which have restricted domains
√ for real numbers have less restricted
domains for complex numbers; as an example z is defined for all z = x + iy with x ∈
(−∞, ∞), y ∈ (−∞, ∞). Moreover ln z is similarly defined for all z, as are all trigonometric
functions. Fortran is well equipped to handle these operations when variables are declared
to be complex. We give an example program in ch15b.f90:
[powers@remote202 ame.20214]$ cat ch15b.f90
program cplex
complex :: z1,z2,z3

z1 = (-2.,0.)
z2 = (0.,1.)
z3 = (1.,1.)

print*,’sqrt(-2) = ’, sqrt(z1)
print*,’ln(-2) = ’, log(z1)
print*,’sin(-2) = ’, sin(z1)

print*,’sqrt(i) = ’, sqrt(z2)
print*,’ln(i) = ’, log(z2)
print*,’sin(i) = ’, sin(z2)

print*,’sqrt(1+i) = ’, sqrt(z3)
print*,’ln(1+i) = ’, log(z3)
print*,’sin(1+i) = ’, sin(z3)

end program cplex

[powers@remote202 ame.20214]$ ifort ch15b.f90
[powers@remote202 ame.20214]$ a.out
sqrt(-2) = (0.0000000E+00,1.414214)
ln(-2) = (0.6931472,3.141593)
sin(-2) = (-0.9092974,0.0000000E+00)
sqrt(i) = (0.7071068,0.7071068)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

132 CHAPTER 15. COMPLEX

ln(i) = (0.0000000E+00,1.570796)
sin(i) = (0.0000000E+00,1.175201)
sqrt(1+i) = (1.098684,0.4550899)
ln(1+i) = (0.3465736,0.7853982)
sin(1+i) = (1.298458,0.6349639)
[powers@remote202 ame.20214]$
We see
1. for z = −2 + 0i that
√ √ √
(a) z = −2 = 2i,
(b) ln z = ln(−2) = ln(2) + iπ, and
(c) sin(z) = sin(−2) = − sin(2).
2. for z = 0 + i that
√ √ √ √
(a) z = i = 2 + 2i,
(b) ln z = ln(i) = i π2 , and
1 −e−1
(c) sin(z) = sin(i) = i e 2
.
3. for z = 1 + i that
√ √
(a) z = 1 + i = 21/4 (cos(π/8) + i sin(π/8)) = 1.09868 + 0.45509i,
1
(b) ln z = ln(1 + i) = 2
ln(2) + i π4 = 0.346574 + 0.785398i, and
(c) sin(z) = sin(1 + i) = 1.29846 + 0.634964i.
We lastly note that inverse trigonometric functions have a clean mathematical definition for
complex numbers. For instance, it can be shown that
 √ 
sin−1 z = −i ln iz + 1 − z 2 , z ∈ C1 , (15.12)

so that, for example

−1
 √  π √
sin (2) = −i ln (2 + 3)i = − ln(2 + 3)i = 1.5708 − 1.31696i. (15.13)
2
Note also that the inverse sine is not unique as the following are all solutions:
sin−1 (2) = 1.5708 + 2nπ − 1.31696i, n = . . . , −2, −1, 0, 1, 2, . . . (15.14)
Note to achieve real values of sin−1 x, we must restrict x so that x ∈ R1 . When we go outside
of the domain x ∈ [−1, 1], the range is extended to the complex numbers.
Now one might suppose Fortran’s asin could handle complex numbers, but that this is
only available in the Fortran 2008 standard, not yet invoked on our ifort compiler. This
is not a serious problem as the interested user can always define a function which serves the
same purpose. This is seen in ch15c.f90:

CC BY-NC-ND. 19 January 2025, J. M. Powers.

 133

[powers@remote202 ame.20214]$ cat ch15c.f90

program arcsine
implicit none
complex :: z,arcsin
z = (2.,0.)
print*, ’arcsin(’,z,’)= ’,arcsin(z) !call upon our function arcsin
end program arcsine

complex function arcsin(t)

implicit none
complex :: t,i = (0.,1.)
arcsin = -i*log(t*i+sqrt(1-t**2))
end function arcsin
[powers@remote202 ame.20214]$ ifort ch15c.f90
[powers@remote202 ame.20214]$ a.out
arcsin( (2.000000,0.0000000E+00) )= (1.570796,-1.316958)
[powers@remote202 ame.20214]$

Note

• We simply defined our own function subroutine arcsin and gave it the proper defini-
tion, consistent with that of mathematics.

• When exercised on z = 2 + 0i, our program returned the value, sin−1 (2) = 1.570796 −
1.316958i. This value is acceptable, though we must realize that in general others
could be as well.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

134 CHAPTER 15. COMPLEX

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 16

Logical

Read C&S, Chapter 16.

The interested reader can consult the text of C&S.

135
136 CHAPTER 16. LOGICAL

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 17

Introduction to derived types

Read C&S, Chapter 17.

The interested reader can consult C&S’s Chapter 17 for a discussion of derived types.

137
138 CHAPTER 17. INTRODUCTION TO DERIVED TYPES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 18

An introduction to pointers

Read C&S, Chapter 18.

The interested reader can consult C&S for a discussion of pointers.

139
140 CHAPTER 18. AN INTRODUCTION TO POINTERS

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 19

Introduction to subroutines

Read C&S, Chapter 19.

Many times we have a task that is composed of a sequence of steps that need to be re-
peated many times. For an efficient code, it is to our advantage to write a general subroutine
that can handle tasks which must be repeated.

19.1 Simple subroutine

Let us begin with a problem that is probably too simple to justify a subroutine, but which will
illustrate the syntax. Let us say that we wish to convert from Cartesian to polar coordinates
via
p
r = + x2 + y 2, (19.1)
y 
θ = arctan . (19.2)
x
and that we need to calculate r and θ for arbitrary x and y. Because there are two outputs,
r and θ, a single user-defined function subroutine is inappropriate. While we could employ
two user-defined functions for this, let us instead employ a single user-defined subroutine.
We show such a subroutine and its use in the program ch19a.f90.

[powers@darrow1-p ame.20214]$ cat ch19a.f90

!-----------------------------------------------------------------------------------
module xyrt !beginning of subroutine module
contains
subroutine crt(xx,yy,rr,ttheta) !beginning of actual subroutine
implicit none
real, intent(out) :: rr !useful optional statement reserving rr for output
real, intent(out) :: ttheta !useful optional statement reserving qq for output
real, intent(in) :: xx !useful optional statement reserving xx for input

141
142 CHAPTER 19. INTRODUCTION TO SUBROUTINES

real, intent(in) :: yy !useful optional statement reserving yy for input

rr = sqrt(xx**2+yy**2)
ttheta = atan(yy/xx)
end subroutine crt
end module xyrt
!
!-----------------------------------------------------------------------------------
!
program ch19a
use xyrt !import the module with the subroutine
implicit none
real:: x=3., y=4., r, theta
call crt(x,y,r,theta) !call the actual subroutine
print*,’x = ’, x
print*,’y = ’, y
print*,’r = ’, r
print*,’theta = ’, theta
end program ch19a
!-----------------------------------------------------------------------------------
[powers@darrow1-p ame.20214]$ ifort ch19a.f90
[powers@darrow1-p ame.20214]$ a.out
x = 3.000000
y = 4.000000
r = 5.000000
theta = 0.9272952
[powers@darrow1-p ame.20214]$

Note:

• The module containing the subroutine is placed first in the program structure.

• The subroutine uses a set of dummy variable names, xx, yy, rr, ttheta, which we
choose to be similar to those of the actual variables. We could have chosen them to be
identical. No matter which choice we made, those variables are local variables to the
subroutine. They are passed back to the main program where they take on the value
of the names assigned in the main program. So rr gets assigned in the subroutine,
and it is passed back to the main program, where it is assigned to the variable r. The
same holds for θ, represented as ttheta and theta.

• It is good, but optional, practice to define variables as either input or output variables
within the subroutine. Here, input variables were defined with intent(in) as input
variables only. This then requires that the subroutine not change their value. Output
variables were defined with intent(out) and were not given initial values.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.2. SUBROUTINES FOR SOLUTION OF A QUADRATIC EQUATION 143

• Though not used here, variables which are allowed values on input and can be changed
for output can be assigned intent(inout).

• The main program, which appears last, is executed first.

• The main program assigns values to x and y, but leaves r and theta unassigned
initially.

• The main program refers to the subroutine module at its beginning.

• The first real executable line of the main program is to call upon the subroutine, where
r gets its value.

• After execution of the subroutine, control returns to the main program.

• The main program then prints all variables to the screen.

19.2 Subroutines for solution of a quadratic equation

We next consider a subroutine to aid in solving for the real roots of the quadratic equation

ax2 + bx + c = 0. (19.3)

We know the solution to be √

−b ± b2 − 4ac
x= . (19.4)
2a
We give a program which is an adaptation of one found in C&S to solve this problem with
the aid of two subroutines, ch19b.f90:

[powers@remote102 ame.20214]$ cat ch19b.f90

module interact_module
contains
subroutine interact(a,b,c)
implicit none
real, intent (out) :: a
real, intent (out) :: b
real, intent (out) :: c

print *, ’ type in the coefficients a, b and c’

read (unit=*,fmt=*) a, b, c
end subroutine interact
end module interact_module

module solve_module

CC BY-NC-ND. 19 January 2025, J. M. Powers.

144 CHAPTER 19. INTRODUCTION TO SUBROUTINES

contains
subroutine solve(e,f,g,root1,root2,ifail)
implicit none
real, intent (in) :: e
real, intent (in) :: f
real, intent (in) :: g
real, intent (out) :: root1
real, intent (out) :: root2
integer, intent (inout) :: ifail
! local variables
real :: term
real :: a2

term = f*f - 4.*e*g

a2 = e*2.0
! if term < 0, roots are complex
if (term<0.0) then
ifail = 1
else
term = sqrt(term)
root1 = (-f+term)/a2
root2 = (-f-term)/a2
end if
end subroutine solve
end module solve_module

program roots
use interact_module
use solve_module
implicit none
! simple example of the use of a main program and two
! subroutines. one interacts with the user and the
! second solves a quadratic equation,
! based on the user input.
real :: p, q, r, root1, root2
integer :: ifail = 0

call interact(p,q,r)
call solve(p,q,r,root1,root2,ifail)
if (ifail==1) then
print *, ’ complex roots’
print *, ’ calculation abandoned’

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 145
else
print *, ’ roots are ’, root1, ’ ’, root2
end if
end program roots
[powers@remote102 ame.20214]$ ifort ch19b.f90
[powers@remote102 ame.20214]$ a.out
type in the coefficients a, b and c
1 -4 3
roots are 3.000000 1.000000
[powers@remote102 ame.20214]$

Note:

• For the variables a, b and c, the main program uses p, q, and r, while subroutine
interact uses a, b, and c, and subroutine solve uses e, f, and g.

• Data input is handled by subroutine interact.

• The output of subroutine interact is used as input for subroutine solve.

• The variable ifail is a flag which lets the main program know whether or not the
calculation was such that real roots were found. The variable ifail is initialized to
zero in the main program, and is an input to subroutine solve. If the discriminant
is negative, the roots are complex, and the subroutine sets ifail to 1. Note that
ifail is both an input and an output to the subroutine, and is declared as such with
intent(inout).

19.3 Subroutines for solving systems of ordinary dif-

ferential equations
Let us develop a program to solve a general system of N non-linear ordinary differential
equations using a first order Euler method. We briefly review the relevant mathematics.
We consider the following initial value problem:

dy1
= f1 (y1 , y2 , . . . , yN ), y1 (0) = y1o , (19.5)
dt
dy2
= f2 (y1 , y2 , . . . , yN ), y2 (0) = y2o , (19.6)
dt
..
. (19.7)
dyN
= fN (y1 , y2 , . . . , yN ), yN (0) = yN o . (19.8)
dt

CC BY-NC-ND. 19 January 2025, J. M. Powers.

146 CHAPTER 19. INTRODUCTION TO SUBROUTINES

Here f1 , f2 , . . . , fN are general non-linear algebraic functions. A simple finite difference

approximation of these equations gives

y1n+1 − y1n
= f1 (y1n , y2n , . . . , yN
n
), y11 = y1o , (19.9)
∆t
y2n+1 − y2n
= f2 (y1n , y2n , . . . , yN
n
), y21 = y2o , (19.10)
∆t
..
. (19.11)
n+1 n
yN − yN
= fN (y1n , y2n , . . . , yN
n
), 1
yN = yN o . (19.12)
∆t
Here, the superscripts are not exponents; instead, they are simply indices for counting. We
can solve for the new values at n + 1 in terms of the old at n directly:

y1n+1 = y1n + ∆tf1 (y1n , y2n , . . . , yN

n
), (19.13)
y2n+1 = y2n + ∆tf2 (y1n , y2n , . . . , yN
n
), (19.14)
..
. (19.15)
n+1 n
yN = yN + ∆tfN (y1n , y2n , . . . , yN n
). (19.16)

If we repeat this process many times for sufficiently small values of ∆t, we get a good
approximation to the continuous behavior of y1 (t), y2 (t), . . . , yN (t).
Now we wish to have a program with two desirable features:

• It can handle arbitrary functions f1 , f2 , . . . , fN .

• It can handle arbitrary integration algorithms, say Euler, or say the common Runge-
Kutta method. For example, the two-step Runge-Kutta method is as follows:

n+1/2 ∆t
yi = yin + fi (yjn ), i, j = 1, . . . , N, (19.17)
2
n+1/2
yin+1 = yin + ∆tfi (yj ), i, j = 1, . . . , N. (19.18)

19.3.1 One linear ordinary differential equation

Let us apply this strategy to a “system” comprised of N = 1 equations:
dy1
= λy1, y1 (0) = y1o . (19.19)
dt
We will take λ = −1, y1o = 1. The exact solution is given by

y1 (t) = e−t . (19.20)

Our general code for this problem is in ch19c.f90.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 147
[powers@darrow2-p ame.20214]$ cat ch19c.f90
!-------------------------------------------------------------------------
!
module commondata
integer, parameter :: neq=1 !set the number of equations
integer, parameter :: nt=10 !set the number of time steps
real, parameter :: tstop = 1. !set the stop time
real, dimension(1:neq) :: yic=(/1./) !set the initial conditions
end module commondata
!
!-------------------------------------------------------------------------
!
module rhs
! This subroutine in this module evaluates the right hand side of the
! system of neq ordinary differential equations.
contains
subroutine fcn(t,y,ydot)
use commondata
implicit none
real, intent(in) :: t
real, intent(in), dimension(1:neq) :: y
real, intent(out), dimension(1:neq) :: ydot
real, parameter :: lambda=-1.
ydot(1) = lambda*y(1)
end subroutine fcn
end module rhs
!
!-------------------------------------------------------------------------
!
module ode1module
! The subroutine in this module is a generic algorithm for
! the first order explicit Euler method
! for solving the system of neq ordinary
! differential equations evaluated by
! subroutine fcn.
contains
subroutine ode1(y,t,tout)
use commondata
use rhs
implicit none
integer :: i
real, intent(inout), dimension(1:neq) :: y

CC BY-NC-ND. 19 January 2025, J. M. Powers.

148 CHAPTER 19. INTRODUCTION TO SUBROUTINES

real, intent(inout) :: t
real, intent(in) :: tout
real, dimension(1:neq) :: ydot
real :: dt
dt = tout-t
call fcn(t,y,ydot) !estimate ydot at t
do i=1,neq
y(i) = y(i)+dt*ydot(i) !estimate y at t
enddo
t = tout !update t
end subroutine ode1
end module ode1module
!
!-------------------------------------------------------------------------
!
module ode2module
! The subroutine in this module is a generic algorithm for
! the second order Runge-Kutta method for
! solving the system of neq ordinary differential
! equations evaluated by subroutine fcn.
contains
subroutine ode2(y,t,tout)
use commondata
use rhs
implicit none
integer :: i
real, intent(inout), dimension(1:neq) :: y
real, dimension(1:neq) :: yi,ydot
real, intent(inout) :: t
real, intent(in) :: tout
real :: dt
dt = tout-t
call fcn(t,y,ydot) !estimate ydot at t
do i=1,neq
yi(i) = y(i)+dt*ydot(i)/2. !estimate y at t+dt/2
enddo
call fcn(t+dt/2.,yi,ydot) !estimate ydot at t+dt/2
do i=1,neq
y(i) = y(i) + dt*ydot(i) !estimate y at t
enddo
t = tout !update t
end subroutine ode2

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 149
end module ode2module
!
!-------------------------------------------------------------------------
!
program solveode
! This is the main driver program for solving
! a generic set of neq differential equations.
use commondata
use ode1module
use ode2module
implicit none
integer :: i
real, dimension(1:neq) :: y
real :: t,tout,dt
open(10,file = ’output.dat’)
t = 0. !initialize t
dt = tstop/real(nt) !determine dt
do i=1,neq
y(i) = yic(i) !assign y to its initial condition
enddo
print*,t,y !output results
write(10,*)t,y
do i=1,nt !main loop in time
tout = t + dt !set incremental end time
call ode1(y,t,tout) !update y and t
print*,t,y !output the results
write(10,*)t,y
enddo
print*,’error = ’,y(1) - exp(-t)
end program solveode
[powers@darrow2-p ame.20214]$ ifort ch19c.f90
[powers@darrow2-p ame.20214]$ a.out
0.0000000E+00 1.000000
0.1000000 0.9000000
0.2000000 0.8100000
0.3000000 0.7290000
0.4000000 0.6561000
0.5000000 0.5904900
0.6000000 0.5314410
0.7000000 0.4782969
0.8000001 0.4304672
0.9000001 0.3874205

CC BY-NC-ND. 19 January 2025, J. M. Powers.

150 CHAPTER 19. INTRODUCTION TO SUBROUTINES

1.000000 0.3486784
error = -1.9200981E-02
[powers@darrow2-p ame.20214]$
Note:
• There is a module for shared data, commondata.
• There are three subroutine modules:
– rhs which evaluates the right hand sides of the differential equations,
– ode1module which embodies the Euler method, and
– ode2module which embodies the second order Runge-Kutta method.
• The three subroutine modules each contain a subroutine.
• Some subroutines use modules. Here ode1 and ode2 both need to call upon fcn to
evaluate the right hand side of the function. Also the main driver program, solveode,
may call on either ode1 or ode2. It does not choose to call upon fcn.
• The main program calls upon either ode1 or ode2 to advance the solution from t to
tout. Both y and t enter the solver with old values and return with new values.
• The particularities of the mathematics problem to be solved is confined to the commondata
and rhs modules. The driver program as well as the Euler and Runge-Kutta solvers
are general.
• When we execute with the Euler solver ode1, we get an error at t = 1 of −1.9200981 ×
10−2 .
• If we do nothing but change to ode2, the error changes to 6.6155195 × 10−4. Thus, a
simple change to our algorithm gave us two orders of magnitude improvement in the
error.
We plot the exact solution along with the approximations of the Euler and two-step Runge-
Kutta methods in Fig. 19.1.

19.3.2 Three non-linear ordinary differential equations

Let us apply this method to a problem with some interesting dynamics. Consider the non-
linear forced mass-spring-damper system of
d2 y dy dy
m 2
+ b + k1 y + k2 y 3 = Fo cos νt, y(0) = yo , (0) = 0. (19.21)
dt dt dt
This is also known as a forced, damped, Duffing equation. Here y is the position, yo is the
initial position, m is the mass, b is the damping coefficient, k1 is the linear spring constant,

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 151
y1
1.0

0.8

0.6

0.4
t
0.2 0.4 0.6 0.8 1.0

Figure 19.1: Plot of y1 (t), solid line is the exact solution, blue dots give the finite difference
solution with the first order Euler method, red dots give the finite difference solution with
the second order Runge-Kutta method.

k2 is the non-linear spring constant, Fo is the amplitude of the forcing function, and ν is the
frequency of forcing. Let us do a useful scaling exercise, common in engineering. Our goal
is to achieve more universality in our results. Let us define a dimensionless distance y ∗ by
scaling y by its initial value yo :
y
y∗ = . (19.22)
yo

Thus, we might expect y ∗ to have a value around 1; we say that y ∗ = O(1), or of “order
one.” There are many choices of how to scale time. Let us make an obvious choice of

t∗ = νt. (19.23)

Since the frequency ν has units of Hz = 1/s and time t has units of seconds, t∗ will be
dimensionless. We make these substitutions into our system of equations, Eqs. (19.21), and
get

d2 y ∗ dy ∗ dy
myo ν 2 2 + by o ν + k1 yo y ∗ + k2 yo3 y ∗3 = Fo cos t∗ , yo y ∗(0) = yo , yo ν (0) = 0.
(19.24)
dt ∗ dt∗ dt
Now let us scale the second order differential equation by the constant myo ν 2 , and make
obvious reductions to the initial conditions to get

d2 y ∗ b dy ∗ k1 ∗ k2 yo2 ∗ 3 Fo dy
2 + + 2
y + 2
y = cos t∗ , y ∗ (0) = 1, (0) = 0. (19.25)
dt ∗ mν dt∗ mν mν myo ν 2 dt

CC BY-NC-ND. 19 January 2025, J. M. Powers.

152 CHAPTER 19. INTRODUCTION TO SUBROUTINES

Let us take
b k1 k2 yo2 Fo
δ≡ , β≡ , α= , f= . (19.26)
mν mν 2 mν 2 myo ν 2
Each of these system parameters, δ, β, α, and f are dimensionless. Physically, the represent
the ratio of various forces. For example, δ is the ratio of the damping force to the so-called
“inertia” of the system:
byo ν damping force b
δ= = = . (19.27)
myo ν 2 inertia mν
The term β is the ratio of the linear spring force to the inertial force; α is the ratio of the
non-linear spring force to the inertial force, and f is the ratio of the driving force to the
inertia. Thus our governing equation becomes

d2 y ∗ dy ∗ dy ∗
2 + δ + βy ∗ + αy ∗3 = f cos t∗ , y ∗ (0) = 1, (0) = 0. (19.28)
dt ∗ dt∗ dt∗
For convenience, let us drop the ∗ notation, and simply realize that all quantities are dimen-
sionless. So we have
d2 y dy dy
2
+ δ + βy + αy 3 = f cos t, y(0) = 1, (0) = 0. (19.29)
dt dt dt
Now let us take

y1 ≡ y, (19.30)
dy
y2 ≡ . (19.31)
dt
Thus, Eq. (19.29) can be written as
dy1
= y2 , y1 (0) = 1, (19.32)
dt
dy2
= −βy1 − δy2 − αy13 + f cos t, y2 (0) = 0. (19.33)
dt
Lastly, let us define t = s = y3 , so that dt/ds = dy3/ds = 1 with t(s = 0) = y3 (s = 0) = 0.
Then, replacing t by y3 , and multiplying right and left sides of our system by dt/ds = 1, we
get
dt dy1
= (1) (y2 ) , (19.34)
ds dt
dt dy2

= (1) −βy1 − δy2 − αy13 + f cos y3 , (19.35)
ds dt
dt dy3
= (1)1. (19.36)
ds dt

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 153
The chain rule then gives
dy1
= y2 (19.37)
ds
dy2
= −βy1 − δy2 − αy13 + f cos y3 , (19.38)
ds
dy3
= 1. (19.39)
ds
Since s = t, we can also say
dy1
= y2 , (19.40)
dt
dy2
= −βy1 − δy2 − αy13 + f cos y3 , (19.41)
dt
dy3
= 1. (19.42)
dt
This system has some interesting mathematical properties when we take the parameters of
α = 1, β = −1, δ = 0.22, f = 0.3. This is a highly unusual physical system, as the linear
spring constant is in fact negative!
To simulate this function, we simply modify our commondata rhs modules so that they
are as given in the complete ch19d.f90:
!-------------------------------------------------------------------------
!
module commondata
integer, parameter :: neq=3 !set the number of equations
integer, parameter :: nt=100000 !set the number of time steps
real, parameter :: tstop = 200. !set the stop time
real, dimension(1:neq) :: yic=(/1.,0.,0./) !set the initial conditions
end module commondata
!
!-------------------------------------------------------------------------
!
module rhs
! This subroutine in this module evaluates the right hand side of the
! system of neq ordinary differential equations.
contains
subroutine fcn(t,y,ydot)
use commondata
implicit none
real, intent(in) :: t
real, intent(in), dimension(1:neq) :: y
real, intent(out), dimension(1:neq) :: ydot

CC BY-NC-ND. 19 January 2025, J. M. Powers.

154 CHAPTER 19. INTRODUCTION TO SUBROUTINES

y1 0

−1

0 50 100 150 200 250

t

Figure 19.2: Plot of prediction of position versus time, y1 (t), for the forced, damped Duffing
equation for t ∈ [0, 200]; α = 1, β = −1, δ = 0.22, f = 0.3.

real, parameter :: beta=-1.,delta=0.22,alpha=1.,f=0.3

ydot(1) = y(2)
ydot(2) = -beta*y(1)-delta*y(2)-alpha*y(1)**3+f*cos(y(3))
ydot(3) = 1.
end subroutine fcn
end module rhs
!
!-------------------------------------------------------------------------
!

We also select y1 (0) = 1, y2 (0) = 0, and y3 (0) = 0, and integrate for t ∈ [0, 200] with
∆t = 0.02. Now y3 (t) = t is not very interesting. But y1 (t) = y(t) is interesting. We plot
the position (y1 ) versus time (y3 = t) in Fig. 19.2. We see there is an initial transient and
then a relaxation to a cyclic behavior for t > 100. We then plot the velocity (y2 ) versus
time (y3 = t) in Fig. 19.3. We again see there is an initial transient and then a relaxation
to a cyclic behavior for t > 100. We plot the position (y1 ) versus velocity (y2 ) in Fig. 19.4.
Figure 19.4 was for t ∈ [0, 200]. If we only plot the long time behavior, say t ∈ [100, 200], we
find an interesting result. For this late time domain, we plot the position versus velocity in
Fig. 19.5. We see that the solution has become periodic; that is, it truly repeats itself.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 155
0.8

0.4

y2 0

−0.4

−0.8
0 50 100 150 200 250
t

Figure 19.3: Plot of prediction of velocity versus time, y2 (t), for the forced, damped Duffing
equation for t ∈ [0, 200]; α = 1, β = −1, δ = 0.22, f = 0.3.

0.8

0.4

y2 0

−0.4

−0.8
−1 0 1
y1

Figure 19.4: Plot of prediction of velocity versus position, y2 (y1 ), for the forced, damped
Duffing equation, t ∈ [0, 200]; α = 1, β = −1, δ = 0.22, f = 0.3.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

156 CHAPTER 19. INTRODUCTION TO SUBROUTINES

0.8

0.4

y2 0

−0.4

−0.8
−1 0 1
y1

Figure 19.5: Plot of prediction of velocity versus position, y2 (y1 ), for the forced, damped
Duffing equation, t ∈ [100, 200]; α = 1, β = −1, δ = 0.22, f = 0.3.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 157
y1 y2

0.8

0.4
0.4

0
0

−0.4 −0.4

0 40 80 120 160 200 −0.4 0 0.4

t y1

Figure 19.6: Plot of prediction of position versus time, y1 (t), t ∈ [0, 200] and velocity versus
position, y2 (y1 ), t ∈ [100, 200], for the forced, damped Duffing equation, α = 1, β = 1,
δ = 0.22, f = 0.3.

Now the negative linear spring constant was highly unusual and might be difficult to
actually build in nature. Let us return to a more physical situation where the linear spring
constant is positive; we now take β = 1 and leave all other parameters the same. We give
relevant plots in Fig. 19.6. It is noted the the phase plane plot has unfolded and now has a
simpler ellipsoidal shape. It also appears there is a single frequency in play in the long time
limit.
Let us make the problem simpler still by removing the effects of the non-linear spring by
setting α = 0. We give relevant plots in Fig. 19.7. It is noted the the phase plane plot has a
circular shape. It also appears there is a single frequency in play in the long time limit. Even
though there is damping, the amplitude of oscillation is O(1) in the long time limit.√This is
because we are forcing the system at a frequency near its resonant frequency of ∼ β = 1.
Let us change the natural√frequency√ of the system by studying β = 10, thus inducing
a resonant frequency of ∼ β = 10 ∼ 3.16. We are still forcing the system with a
dimensionless frequency of unity. We give relevant plots in Fig. 19.8. It also appears there
is a single frequency, that of the driver, in play in the long time limit. It is also noted that
the phase plane plot retains a circular shape, but its amplitude is greatly reduced. This is
because we are forcing the system at a frequency not near its resonant frequency.
Lastly, we can highlight the object-oriented nature of our approach by segregating our
source code into separate files:
• commondata.f90,

CC BY-NC-ND. 19 January 2025, J. M. Powers.

158 CHAPTER 19. INTRODUCTION TO SUBROUTINES

y1 y2

1 1

0 0

−1 −1

0 40 80 120 160 200 −1 0 1

t y1

Figure 19.7: Plot of prediction of position versus time, y1 (t), t ∈ [0, 200] and velocity versus
position, y2 (y1 ), t ∈ [100, 200], for the forced, damped Duffing equation, α = 0, β = 1,
δ = 0.22, f = 0.3.

y1 y2

0.8

0.02
0.4

0.0 0.

−0.4
−0.02

−0.8

0 40 80 120 160 200 −0.02 0 0.02

t y1

Figure 19.8: Plot of prediction of position versus time, y1 (t), t ∈ [0, 200] and velocity versus
position, y2 (y1 ), t ∈ [100, 200], for the forced, damped Duffing equation, α = 0, β = 10,
δ = 0.22, f = 0.3.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

19.3. SUBROUTINES FOR SOLVING SYSTEMS OF ORDINARY DIFFERENTIAL
EQUATIONS 159
• rhs.f90,

• ode1.f90

• ode2.f90, and

• solveode.f90.

Then we can build an executable file runode, to compile and execute our programs:

[powers@remote105 ame.20214]$ cat runode

ifort -c commondata.f90
ifort -c rhs.f90
ifort -c ode1.f90
ifort -c ode2.f90
ifort -c solveode.f90
ifort commondata.o rhs.o ode1.o ode2.o solveode.o
a.out
[powers@remote105 ame.20214]$

We omit the execution step because it generates lengthy output.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

160 CHAPTER 19. INTRODUCTION TO SUBROUTINES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 20

Subroutines: 2

Read C&S, Chapter 20.

The interested reader can consult the text of C&S.

161
162 CHAPTER 20. SUBROUTINES: 2

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 21

Modules

Read C&S, Chapter 21.

We have already seen some use of the module feature. The interested reader can consult
the text of C&S for more details.

163
164 CHAPTER 21. MODULES

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 22

Converting from Fortran 77

Read C&S, Chapter 36.

There is a large body of legacy Fortran code written under the Fortran 77 standard.
Code written under this standard often has noticeable differences from that written under
the Fortran 90, 95, 2003, 2008 standards, along with many similarities. With a few
exceptions, the user can expect
• Code which compiles and executes with a Fortran 77 compiler will also compile and
execute, generating the same output, under a Fortran 90, 95, 2003, 2008 standard.

• Code which was written under the enhanced standards of Fortran 90, 95, 2003,
2008 most likely will not be able to be compiled with a Fortran 77 compiler.
As the Fortran 77 standard has fewer features and thus is more streamlined, for certain
applications, it will execute faster than programs written with the later standard, and for this
reason, some users prefer it. However, for many modern scientific computing environments,
especially those requiring dynamic memory allocation and parallelization, Fortran 77 is
simply unusable and thus not an option.

22.1 A hello world program

We give a “hello world” program for Fortran 77 in helloworld.f:
[powers@darrow2-p ame.20214]$ cat helloworld.f
program test
print*,’Hello world.’
end program test
[powers@darrow2-p ame.20214]$ f77 helloworld.f
[powers@darrow2-p ame.20214]$ a.out
Hello world.
[powers@darrow2-p ame.20214]$ ifort helloworld.f

165
166 CHAPTER 22. CONVERTING FROM FORTRAN 77

[powers@darrow2-p ame.20214]$ a.out

Hello world.
[powers@darrow2-p ame.20214]$
Note:
• The file identifier for a Fortran 77 standard program is .f.

• All Fortran 77 programs must have most text beginning in column 7 or greater. Ex-
ceptions are the numbers for format statements and for a few other statements. This
is a old feature motivated by the use of punchcards for early computer programs.

• The program compiles with the compiler f77 and executed successfully.

• The program compiled with the modern compiler ifort and executed successfully.

22.2 A simple program using arrays

Here we revisit an earlier program. We recall the earlier presented ch6c.f90:
[powers@darrow2-p ame.20214]$ cat ch6c.f90
program temperature
implicit none
integer, parameter :: imax=7
integer (kind=4) :: i
real (kind=4), dimension(1:imax):: t,temp
real (kind=4), parameter :: a=270., b=10., c = 24., tmax = 24.
real (kind=4) :: pi, dt
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ ifort ch6c.f90
[powers@darrow2-p ame.20214]$ a.out
pi = 3.141593
1 0.0000000E+00 270.0000
2 4.000000 261.3398
3 8.000000 261.3398
4 12.00000 270.0000

CC BY-NC-ND. 19 January 2025, J. M. Powers.

22.2. A SIMPLE PROGRAM USING ARRAYS 167

5 16.00000 278.6602
6 20.00000 278.6602
7 24.00000 270.0000
[powers@darrow2-p ame.20214]$

The Fortran 77 equivalent of this program, and its execution, is given in ch22a.f:

[powers@darrow2-p ame.20214]$ cat ch22a.f

program temperature
implicit none
integer*4 imax
parameter (imax=7)
integer*4 i
real*4 t(1:imax),temp(1:imax)
real*4 a,b,c,tmax
parameter (a=270.,b=10.,c=24.,tmax=24.)
real*4 pi, dt
pi = 4.*atan(1.)
print*,’pi = ’, pi
dt = tmax/(real(imax)-1.)
do i=1,imax
t(i) = dt*(i-1)
temp(i) = a - b*sin(2.*pi*t(i)/c)
print*,i,t(i),temp(i)
end do
end program temperature
[powers@darrow2-p ame.20214]$ f77 ch22a.f
[powers@darrow2-p ame.20214]$ a.out
pi = 3.14159274
1 0. 270.
2 4. 261.339752
3 8. 261.339752
4 12. 270.
5 16. 278.660248
6 20. 278.660248
7 24. 270.
[powers@darrow2-p ame.20214]$ ifort ch22a.f
[powers@darrow2-p ame.20214]$ a.out
pi = 3.141593
1 0.0000000E+00 270.0000
2 4.000000 261.3398
3 8.000000 261.3398
4 12.00000 270.0000

CC BY-NC-ND. 19 January 2025, J. M. Powers.

168 CHAPTER 22. CONVERTING FROM FORTRAN 77

5 16.00000 278.6602
6 20.00000 278.6602
7 24.00000 270.0000
[powers@darrow2-p ame.20214]$

We note the under the Fortran 77 standard

• Again, all lines start at column 7 or greater.

• We exercise the common option not to use :: in the variable declaration sections. We
could do this however.

• Variables must first be declared, then later assigned a value in a parameter statement.

• The kind statement is not used in Fortran 77. For (kind=4), we employ a *4. We
also could have simply used real or integer as appropriate.

• We cannot use dimension to set array dimensions.

• The output to the screen has weaker formatting.

Here are a few other key differences.

• Fortran 77 often uses double precision for kind=8.

• Some Fortran 77 compilers require a double precision version of intrinsic functions,

e.g. the double precision version of the exponential, exp is dexp. The double precision
version of sin is dsin.

• Modules are not used in Fortran 77. To pass data between subroutines and the main
program, common blocks are used. It is possible to use common blocks in Fortran 90,
95, 2003, 2008, but we will not focus on that here.

• Whole array functions such as matmul are not a part of Fortran 77.

22.3 A program using subroutines

Here we give the Fortran 77 version of an earlier program for solving generic systems of
ordinary differential equations, ch22c.f90. The Fortran 77 version is found in ch22b.f.
That program, its compilation, and execution are as follows:

[powers@remote105 ame.20214]$ cat ch22b.f

program solveode
! This is the main driver program for solving
! a generic set of neq differential equations.
implicit none

CC BY-NC-ND. 19 January 2025, J. M. Powers.

22.3. A PROGRAM USING SUBROUTINES 169

integer :: i,neq,nt
parameter (neq=1,nt=10)
real :: y(1:neq)
real :: t,tout,dt
real :: tstop,yic(1:neq)
parameter (tstop=1.)
data yic/1./ !data statement for array

open(10,file = ’output.dat’)
t = 0. !initialize t
dt = tstop/real(nt) !determine dt
do i=1,neq
y(i) = yic(i) !assign y to its initial condition
enddo
print*,t,y !output results
write(10,*)t,y
do i=1,nt !main loop in time
tout = t + dt !set incremental end time
call ode1(y,t,tout,neq) !update y and t
print*,t,y !output the results
write(10,*)t,y
enddo
print*,’error = ’,y(1) - exp(-t)
stop
end
!
!-------------------------------------------------------------------------
!
! This subroutine evaluates the right hand side of the
! system of neq ordinary differential equations.
!
subroutine fcn(t,y,ydot,neq)
implicit none
integer neq
real :: t
real :: y(1:neq)
real :: ydot(1:neq)
real :: lambda
parameter (lambda=-1.)
ydot(1) = lambda*y(1)
return
end

CC BY-NC-ND. 19 January 2025, J. M. Powers.

170 CHAPTER 22. CONVERTING FROM FORTRAN 77

!
!-------------------------------------------------------------------------
!
! The subroutine is a generic algorithm for
! the first order explicit Euler method
! for solving the system of neq ordinary
! differential equations evaluated by
! subroutine fcn.
!
subroutine ode1(y,t,tout,neq)
implicit none
integer :: i,neq
real :: y(1:neq)
real :: t
real :: tout
real :: ydot(1:neq)
real :: dt
dt = tout-t
call fcn(t,y,ydot,neq) !estimate ydot at t
do i=1,neq
y(i) = y(i)+dt*ydot(i) !estimate y at t
enddo
t = tout !update t
return
end
!
!-------------------------------------------------------------------------
!
! This subroutine is a generic algorithm for
! the second order Runge-Kutta method for
! solving the system of neq ordinary differential
! equations evaluated by subroutine fcn.
!
subroutine ode2(y,t,tout,neq)
implicit none
integer :: i,neq
real :: y(1:neq)
real :: yi(1:neq),ydot(1:neq)
real :: t
real :: tout
real :: dt
dt = tout-t

CC BY-NC-ND. 19 January 2025, J. M. Powers.

22.3. A PROGRAM USING SUBROUTINES 171

call fcn(t,y,ydot,neq) !estimate ydot at t

do i=1,neq
yi(i) = y(i)+dt*ydot(i)/2. !estimate y at t+dt/2
enddo
call fcn(t+dt/2.,yi,ydot,neq) !estimate ydot at t+dt/2
do i=1,neq
y(i) = y(i) + dt*ydot(i) !estimate y at t
enddo
t = tout !update t
return
end
!
!-------------------------------------------------------------------------
!
[powers@remote105 ame.20214]$ f77 ch22b.f
[powers@remote105 ame.20214]$ a.out
0. 1.
0.100000001 0.899999976
0.200000003 0.810000002
0.300000012 0.728999972
0.400000006 0.656099975
0.5 0.590489984
0.600000024 0.531440973
0.700000048 0.478296876
0.800000072 0.430467188
0.900000095 0.387420475
1.00000012 0.34867841
error = -0.0192009807
[powers@remote105 ame.20214]$

Note

• The module structure was not used.

• The main program appeared first.

• The parameter statement has a slightly different syntax.

• We used a data statement to initialize the array yic.

• We did not use a commondata structure for shared data. Sometimes this is useful, and
one can employ a so-called common block, not done here, when appropriate.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

172 CHAPTER 22. CONVERTING FROM FORTRAN 77

• The number of equations, neq, was passed as a parameter to most subroutines because
Fortran 77 is more restrictive with variable sized arrays.

• We could have separated our code into distinct codes for the main program and sub-
routines, compiled each under the −c option, and employed an executable script file.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 23

Python and Fortran

Python is a general open-source interpreted high-level programming language. It shares some

functionality with MATLAB. It is possible to use Python to interact with Fortran routines.
This section of the notes will give only the most rudimentary of introductions, focusing
entirely on how Python can be implemented with Fortran on local Notre Dame machines.
At present software to allow Python to interact with Fortran is only loaded on the
remote machines (e.g. remote102.helios.nd.edu).
Now let us consider a Fortran subroutine to give a six-term Taylor series approximation
of ex about x = 0. The exact Taylor series is
∞
x
X xn
e = . (23.1)
n=0
n!

The six-term approximation is

x2 x3 x4 x5
ex ∼ 1 + x + + + + . (23.2)
2 6 24 120
A function subroutine to accomplish this task is embodied in ch23a.f90:
[powers@remote102 ame.20214]$ cat ch23a.f90
real function expa(x)
implicit none
real, intent(in) :: x
expa = 1. + x + x**2/2. + x**3/6. + x**4/24. + x**5/120.
end function expa
[powers@remote102 ame.20214]$
The function takes an input x and returns an approximation of ex in the term expa.
Now let us operate on ch23a.f90 in such a fashion that it can by used by Python. This
is achieved with the Fortran to Python interface generator f2py. The generation is achieved
as follows:

173
174 CHAPTER 23. PYTHON AND FORTRAN

[powers@remote102 ame.20214]$ f2py -c -m ch23a ch23a.f90 --fcompiler=gfortran

The generation relies on the freeware compiler gfortran. It generates a binary output file
named ch23a.so. It also creates a large output to the screen, omitted here.
At this stage we can open Python, and use it to call upon our Fortran interface to
approximate ex . We will look at three approximations: e0 , e1 , and e2 . The sequence of
commands that achieves this is as follows:

[powers@remote102 ame.20214]$ python

Python 2.6.6 (r266:84292, May 1 2012, 13:52:17)
[GCC 4.4.6 20110731 (Red Hat 4.4.6-3)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import ch23a
>>> print ch23a.expa(0)
1.0
>>> print ch23a.expa(1)
2.71666693687
>>> print ch23a.expa(2)
7.26666688919
>>> exit()
[powers@remote102 ame.20214]$

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 24

Introduction to C

The C language is a commonly used programming language for which we give brief examples
here.

24.1 A hello world program

Here is a “hello world” program in C, helloworld.c
[powers@darrow1-p 2e]$ cat helloworld.c
#include <stdio.h>
int main()
{
printf("hello world\n");
}
[powers@darrow1-p 2e]$ gcc helloworld.c
[powers@darrow1-p 2e]$ a.out
hello world
[powers@darrow1-p 2e]$
Note:
• The .c identifier is appropriate for a C program.
• The symbol # invokes a so-called “pre-processor”.
• The #include statement inserts the object file into the program at that line.
• The code needs to appeal to an external standard input output library, stdio.h.
• The main program is encased by brackets, { and }.
• The appendage \n instructs the program to begin a new line.
• The compiler is invoked by the command gcc.

175
176 CHAPTER 24. INTRODUCTION TO C

24.2 A program using the Euler method

Let us now use C to solve one of our paradigm problems, dy/dt = λy, with y(0) = yo via the
Euler method. We take λ = −1 and yo = 1. The program is found in euler.c. Its listing,
compilation, and execution steps are given next.

[powers@darrow2-p ame.20214]$ cat euler.c

/*Name: euler.c
*Description: Approximates dy/dt = lambda*y; y(0) = y0
* with Euler’s method. Exact solution is
* y = y0*exp(lambda*t)
*/
//Include I/O and math libraries
#include <stdio.h>
#include <math.h>
//Define problem parameters
#define tstop 1.0f
#define y0 1.0f
#define lambda -1.0f
#define dt 0.1f
int main ( void ){
FILE *output; //output file syntax
output = fopen("output.txt","w"); //open output file
float y=y0, ynew, t = 0; //set initial time/value
int n = 0; //initialize counter
printf("%.8f\t%.8f\t%.8f\n", t, y, y0*expf(lambda*t)); //print initial values
fprintf(output,"%.8f\t%.8f\t%.8f\n", t, y, y0*expf(lambda*t));//write initial values
do{
ynew = y+lambda*dt*y; //approximate new solution
n = n+1; //update counter
t = dt*n; //update time
y = ynew; //update y
printf("%.8f\t%.8f\t%.8f\n", t, y, y0*expf(lambda*t));//print results
fprintf(output,"%.8f\t%.8f\t%.8f\n", t, y, y0*expf(lambda*t));//write results
} while(t < tstop); //loop end condition
printf("error %.8e\n", y-y0*expf(lambda*t)); //print final error
fclose(output); //close output file
return 0;
}
[powers@darrow2-p ame.20214]$ gcc -lm euler.c
[powers@darrow2-p ame.20214]$ a.out
0.00000000 1.00000000 1.00000000
0.10000000 0.89999998 0.90483743

CC BY-NC-ND. 19 January 2025, J. M. Powers.

24.2. A PROGRAM USING THE EULER METHOD 177

0.20000000 0.81000000 0.81873077

0.30000001 0.72899997 0.74081820
0.40000001 0.65609998 0.67032003
0.50000000 0.59048998 0.60653067
0.60000002 0.53144097 0.54881161
0.69999999 0.47829688 0.49658531
0.80000001 0.43046719 0.44932896
0.90000004 0.38742048 0.40656966
1.00000000 0.34867844 0.36787945
error -1.92010105e-02
[powers@darrow2-p ame.20214]$ cat output.txt
0.00000000 1.00000000 1.00000000
0.10000000 0.89999998 0.90483743
0.20000000 0.81000000 0.81873077
0.30000001 0.72899997 0.74081820
0.40000001 0.65609998 0.67032003
0.50000000 0.59048998 0.60653067
0.60000002 0.53144097 0.54881161
0.69999999 0.47829688 0.49658531
0.80000001 0.43046719 0.44932896
0.90000004 0.38742048 0.40656966
1.00000000 0.34867844 0.36787945
[powers@darrow2-p ame.20214]$

Note:

• The .c identifier is appropriate for a C program.

• A block of comment lines is bracketed by a pair of /*.

• The symbol # invokes a so-called “pre-processor”.

• The #include statement inserts the object file into the program at that line.

• The #define statement will substitute the second argument for the first in all occur-
rences of the first; it is equivalent to a Fortran parameter statement.

• An inline comment is initiated by the symbols //.

• The code needs to appeal to external libraries, a standard input output library, stdio.h,
and a mathematics library, math.h.

• A floating point number is identified by a f at its trailing edge.

• Integers do not have a trailing f.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

178 CHAPTER 24. INTRODUCTION TO C

• Formatted printing to the screen is complicated.

• We have also opened, written to, and closed the file output.txt.

• The compiler is invoked by the command gcc.

• The compiler option -lm invokes the math library.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 25

Introduction to Microsoft Excel

The Microsoft Excel software is a common tool for some straightforward tasks in engi-
neering computing. While it is most commonly used for business and finance tasks, with
some effort, it can be used as a a software tool to solve problems in scientific computing. It
is generally not as efficient, both in programming ease as well as computational efficiency,
as other tools that have been studied. But its near universal presence in the worldwide
business environment gives it the advantage of being able to run on nearly every modern
office computer. Occasionally, this advantage can outweigh its considerable disadvantages,
especially for small-scale problems.

25.1 A hello world program

The “hello world” program in helloworld.xlsx is trivial. One simply enters “hello world”
into a box. The actual display is shown in Fig. 25.1.

Figure 25.1: Microsoft Excel spreadsheet for “hello world.”

179
180 CHAPTER 25. INTRODUCTION TO MICROSOFT EXCEL

25.2 A program using the Euler method

We once again consider our model problem
dy
= −y, y(0) = 1. (25.1)
dt
We use the Euler method with ∆t = 0.1. The source code is in the file euler.xlsx. The
formulæ are entered in the spreadsheet, displayed in Fig. 25.2. This is not the default display

Figure 25.2: Microsoft Excel spreadsheet showing formulæ to solve dy/dt = −y, y(0) = 1
using the Euler method with ∆t = 0.1.

format of Microsoft Excel, but can be realized by exercising some of the menu-driven
options. The graph shown within Fig. 25.2 is a plot of y(t) for both the Euler method and
the exact solution. With considerable and non-obvious effort, the plot could be significantly
improved. The rudimentary outline for forming this plot on the Macintosh version of Excel
is as follows:

• Generate the appropriate numbers in columns A, B, and C with the appropriate

formulæ.

• With the computer mouse, highlight and select all of the numbers within columns A,
B, and C.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

25.2. A PROGRAM USING THE EULER METHOD 181

• From the Charts menu item, select Scatter, and from within Scatter, select Smooth
Marked Scatter. This will generate a plot.

• From the Charts menu item, select Chart Layout, and from within this, select Axis
Titles. This will give you options for entering text for axis titles.

• Explore many of the other options available for formatting the graphics, for example,
under Chart Layout, one may define a Chart Title.

The actual display that is given by default upon entry of the formulæ is shown in Fig. 25.3.

Figure 25.3: Microsoft Excel spreadsheet showing results of simulation of dy/dt = −y,
y(0) = 1 using the Euler method with ∆t = 0.1.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

182 CHAPTER 25. INTRODUCTION TO MICROSOFT EXCEL

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 26

Introduction to VBA

co-authored with Ms. Laura A. Paquin

The VBA language is a variant of the BASIC language specialized for execution within
the spreadsheet application Microsoft Excel. It is in widespread use in many engineering
environments.
Using a so-called “macro,” a program written or recorded within Microsoft Excel to
automate a process, the user can link a code to a keyboard shortcut or a “button” in order
to perform a task repeatedly. In order to enable macro-writing, one must first activate the
Developer tab in Microsoft Excel. For Macintosh users, this can be achieved by clicking
Preferences... in the Excel drop-down menu, then navigating to Ribbon and checking
Developer in the list of tabs available. For Windows users, this can be achieved by going
to Excel Options in the Office Button, navigating to Popular, and then checking Show
Developer tab in the Ribbon. Additionally, security settings may have to be modified
in order to allow the use of macros. These settings reside again in Excel Options within
the Trust Center tab. From here, you can navigate into Trust Center Settings... and
change the Macro Settings to enable all macros.

26.1 A hello world program

The “hello world” program in VBA helloworld.xlsm calls upon a pop-up box to display a
specified message. To build this from scratch, the user can first, in the Developer tab within
Excel, place a button on a spreadsheet by selecting Button. Put the button in any location
on your worksheet and then select New to begin writing the code in the Visual Basic editor.
For this program, only three lines of code are required:

Sub Button1_Click()
MsgBox ("hello world")
End Sub

183
184 CHAPTER 26. INTRODUCTION TO VBA

After closing the editor, one can activate the macro by clicking on the button in the spread-
sheet. A depiction of the action of the “hello world” in the Excel environment is shown in
Fig. 26.1.

Figure 26.1: Macintosh screen shot of Excel spreadsheet after executing a simple “hello
world” VBA program.

26.2 A program using the Euler method

We once again consider our model problem
dy
= −y, y(0) = 1. (26.1)
dt
We use the Euler method with ∆t = 0.1. The VBA source code for the simulation is in the
file VBA Euler.xlsm. The initial conditions and system parameters are entered in the Excel
spreadsheet, displayed in Fig. 26.2. The initial value of t = 0 is in cell F2, and that of
y(0) = 1 is in cell G2. The step size ∆t = 0.1 is in cell H2, and the final value of t = 1 is in
cell I2.
The macro code, embedded within the spreadsheet is given by
Sub Button1_Click()
Dim i As Integer ’iteration counter
Dim tstart As Double ’initial t value
Dim ystart As Double ’initial y value
Dim dt As Double ’incremental t value
Dim n As Double ’final iteration value

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26.2. A PROGRAM USING THE EULER METHOD 185

Figure 26.2: Microsoft Excel spreadsheet showing parameters to solve dy/dt = −y, y(0) =
1 with ∆t = 0.1.

Dim tend As Double ’final t value--end of domain

Dim t As Double
Dim y As Double

’Assigning User-Input Parameters

tstart = Cells(2, 6)
ystart = Cells(2, 7)
dt = Cells(2, 8)
tend = Cells(2, 9)

’Initializing
t = tstart
y = ystart
n = Round(((tend - tstart) / dt), 0) ’rounded to 0 decimal places

’Running Euler
For i = 1 To n
t = t + dt
y = y + (-1 * y * dt)
’Printing Values to Cells
Cells((i + 2), 6) = t
Cells((i + 2), 7) = y
Next i

CC BY-NC-ND. 19 January 2025, J. M. Powers.

186 CHAPTER 26. INTRODUCTION TO VBA

End Sub

By inspection, one can infer that the program

• Declares variables of the appropriate type,

• Identifies input variables from the Excel spreadsheet and reads them,

• Executes a loop in which is embodied the Euler method,

• Has comments following the symbol ’.

Following execution via clicking the “Run Euler” button, the spreadsheet has the display
shown in Fig. 26.3. The results are identical to those shown in Fig. 25.3.

Figure 26.3: Microsoft Excel spreadsheet showing approximate solution via the Euler
method of dy/dt = −y, y(0) = 1 with ∆t = 0.1.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26.3. ANOTHER VBA EXAMPLE 187

26.3 Another VBA Example

Programming with VBA in Microsoft Excel provides the user the opportunity to use the
worksheet layout to organize input and output. This section will demonstrate how to tackle
the second order differential equation,

d2 y dy
m + b + ky = Fo sin(νt), (26.2)
dt2 dt
with the Euler method encoded in VBA. The user may establish parameter values within the
code in the VBA workspace, or the program can call upon cells to initialize these values. For
this example, the program will draw upon a row of populated cells to obtain parameters for
the problem. This allows the user the ability to change parameters, like a damping coefficient
or forcing frequency, on the worksheet, and rerun the macro without opening the code.

26.3.1 Creating a calculation macro

One may begin by creating a Form Control Button as shown in Fig. 26.4. The user can

Figure 26.4: Creating a Form Control Button.

draw the button on the worksheet somewhere that will not interfere with the desired location
of input or output. After naming the macro (in this case MSD solve) as shown in Fig. 26.5,
clicking New will bring the user into the Visual Basic workspace where he/she will code
the program to be activated by the button.
Closing the workspace at any time will automatically save the script and return the user
to the spreadsheet. To access the VBA workspace again, the user may navigate into the
Developer tab and select Visual Basic as shown on the left of the screen in Fig. 26.6. The
code for this macro can be seen below.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

188 CHAPTER 26. INTRODUCTION TO VBA

Figure 26.5: Creating macro and VBA workspace: a) Drawing button, naming macro, and
beginning program, b) Visual Basic editor.

Sub MSD_solve()
Dim i, N As Integer
Dim m, k, b, Fo, nu, yo, vo, tstart, tend As Double ’Parameters
Dim y, v, t, ynew, vnew, tnew As Double ’Values to be calculated

’Initialize Variables
m = Cells(2, 2)
k = Cells(2, 3)
b = Cells(2, 4)
Fo = Cells(2, 5)
nu = Cells(2, 6)
yo = Cells(2, 7)
vo = Cells(2, 8)
tstart = Cells(2, 9)
tend = Cells(2, 10)
N = Cells(2, 11)

dt = (tend - tstart) / N

’At t=0
t = tstart
y = yo
v = vo
Cells(5, 3) = t ’print initial t
Cells(5, 4) = y ’print initial y

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26.3. ANOTHER VBA EXAMPLE 189

Figure 26.6: Navigating back into VBA workspace.

Cells(5, 5) = v ’print initial v

’Calculating solution
For i = 1 To N
tnew = t + dt
ynew = y + dt * v
vnew = v + dt * (-(k / m) * y - (b / m) * v + (Fo / m) * Sin(nu * t))
t = tnew
y = ynew
v = vnew
’print updated t, y, v
Cells(i + 5, 3) = t
Cells(i + 5, 4) = y
Cells(i + 5, 5) = v
Next i

End Sub

Note the following within the program:

• It dimensions every parameter and variable appropriately as a Double or an Integer.

• It establishes the value of these variables by calling on the contents of Cells(i, j).

• It iterates through a do loop using the syntax For i and Next i.

• It prints the values from successive iterations to the cells in a column j using the format

CC BY-NC-ND. 19 January 2025, J. M. Powers.

190 CHAPTER 26. INTRODUCTION TO VBA

Cells(i + 5, j), and it offsets the row index by 5 in order to begin output beneath
the title cells.

After exiting the editor, one may activate the program by clicking on the button. Results
for t ∈ (0, 20 s), N = 20, and the parameters listed in Table 26.1, are shown in Fig. 26.7.

Table 26.1: Mass-Spring-Damper System Parameters

m (kg) k (N/m) b (N s/m) Fo (N) ν (Hz) yo (m) vo (m/s)
3 9 2 10 2 1 0

Figure 26.7: Output generated by macro.

26.3.2 Clearing data

26.3.2.1 Manual clearing
While the macro will overwrite N + 1 cells in columns D and E every time the button is
pushed, it will not clear the cell contents in these columns before re-populating them. Thus,
one can imagine why having a Clear macro would be needed to erase leftover data in cases
when tend or N is decreased between runs. One can clear a block of cells by selecting the top
cell (in this case C5), pressing the the key combination Ctrl + Shift + ↓ and then right
clicking the selected cell block to choose Clear Contents, as shown in Fig. 26.8. To clear

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26.3. ANOTHER VBA EXAMPLE 191

Figure 26.8: Manually clearing cell block contents.

a cell block spanning more than one column, one can use Ctrl + Shift + ↓ + → after
selecting the top left cell.

26.3.2.2 Automated clearing

For this activity, the user will automate the clearing process by recording a macro which will
clear the desired block of data upon every click of the button. After running the calculation
macro, one can press Record Macro in the Developer tab. After naming the macro Clear
and pressing OK as shown in Fig. 26.9, the macro will record all the user’s following actions
until he/she presses Stop Recording.
For this function, the user should again select the top left cell in the output block, hold
Ctrl + Shift + ↓ + →, and then clear these cells. After pressing Stop Recording and
then navigating to Macros within the Developer tab, one can view his/her encoded actions
by selecting the new Clear macro and choosing Step Into. Upon viewing the script for the
new Clear subroutine in the VBA workspace, one should see a code of the form below.

Sub Clear()
’
’ Clear Macro
’
’

CC BY-NC-ND. 19 January 2025, J. M. Powers.

192 CHAPTER 26. INTRODUCTION TO VBA

Figure 26.9: Recording a Clear Data Macro, a) Navigating to Record Macro, b) Naming
macro.

Range("C5").Select
Range(Selection, Selection.End(xlDown)).Select
Range(Selection, Selection.End(xlToRight)).Select
Selection.ClearContents
End Sub

Brief inspection of this code will demonstrate the following:

• It selects the top left cell C5 pressed by the user.

• It extends the selection to the last cell of the populated cell block within the first
column, denoted Selection.End(xlDown).

• It extends the selection to the rightmost populated column with the populated cell
block with Selection.End(xlToRight).

• Once the entire cell block is included in the Selection, it depopulates it with ClearContents.

To embody this macro in a button, one can again choose a Form Control button in the
Developer tab. After placing the button on the worksheet, the user will assign the Clear
macro to the button as shown in Fig. 26.10. The two buttons now allow the user to perform
as many runs as desired for the Euler approximation of the mass-spring-damper system with
any specified time range and N+1 number of points.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

26.3. ANOTHER VBA EXAMPLE 193

Figure 26.10: Assigning Clear macro to button.

26.3.3 Automated plotting

One may automate a visual representation of output by creating a plot linked to the output
cell block. This example will demonstrate how the macro will automate plots of the position
and velocity of the system.
The user can create the plot by selecting the title cell reading time, using Ctrl + Shift
+ ↓ to select all time values in the column, holding Ctrl and selecting the position cell,
and then using Ctrl + Shift + ↓ again to select the entirety of the position data. Once
this data is selected, choosing Scatter from the Insert tab will display a scatter plot of the
position as a function of time.
Performing the same sequence and choosing the velocity title cell instead of the position
one will generate an analogous plot of velocity as a function of time. Figure 26.11 demon-
strates the results of this process.
These plots, synchronized with the cell blocks containing the time, position, and velocity
data, will update each time a macro is activated. Because of the Ctrl + Shift + ↓ com-
mand initially utilized to select the plot data, the scatter plots will update to include as many
entries as are generated with each run. For example, if the user increases or decreases tend or
N and then reruns the calculation macro, the number of points plotted in the scatter plots
will correspond to the new number of populated cells in the time, position, and velocity
blocks.
One may embellish the generated plots in the Chart Tools ribbon.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

194 CHAPTER 26. INTRODUCTION TO VBA

Figure 26.11: Automated scatter plots.

26.3.4 Miscellaneous
Other important items when programming in VBA:

• One can customize the button by right clicking on it, choosing Edit Text and/or
Format Control.

• One should save the Excel file as a .xlsm (macro-enabled) instead of .xlsx.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Chapter 27

Introduction to Mathematica

Visit Wolfram’s Mathematica reference guide.

The Mathematica software can be a useful tool for engineering computing. Among its
advantages are

• It can do exact symbolic mathematics in terms of variables.

• It can do numerical calculations with arbitrary precision.

• It can render high quality graphics.

• It is available on campus cluster machines, and can be downloaded from the OIT
website for use on personal computers, http://oit.nd.edu.

• It has a very particular syntax, forcing the user to employ precision in writing code.

Among its disadvantages are

• It has a very particular syntax, forcing the user to employ precision in writing code.

27.1 A hello world program

A particularly simple Mathematica version of a hello world code is given in helloworld.nb:
It is trivially constructed by entering “hello world” and selecting the drop-down menu option
Convert to Text Cell. A screen capture is given in Fig. 27.1.

27.2 A simple code

A particularly simple Mathematica code is given in ch26a.nb:

195
196 CHAPTER 27. INTRODUCTION TO MATHEMATICA

Figure 27.1: hello world in a Mathematica window.

In[1]:= 1 + 1

Out[1]= 2
Note:
• The input is 1 + 1.

• One employs “shift-return” from the keyboard and gets the output.

• Both input and output are integers.

If downloading this file with a web browser, you will likely find a long text file. You
should be able to save this file as something with a .nb extension. Most web browsers will
try to force you to do something else, but you should be able to override their default.

27.3 Precision
Mathematica has many ways to handle precision. Some of them are shown in ch26b.nb:

In[3]:= 1/3

Out[3]= 1/3

In[4]:= 1./3.

Out[4]= 0.333333

In[5]:= 1/3 // N

Out[5]= 0.333333

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.4. LISTS, VECTORS, AND MATRICES 197

In[6]:= N[1/3]

Out[6]= 0.333333

In[7]:= N[1/3, 20]

Out[7]= 0.33333333333333333333

In[9]:= N[1/3, 50]

Out[9]= 0.33333333333333333333333333333333333333333333333333

Note:

• When rational numbers, i.e. ratios of integers, are entered, rational numbers are out-
put.

• Finite precision numbers result when the decimal point is employed.

• The default number of digits output to the screen for a finite precision number is six.

• The command // N asks for a finite precision version of a number.

• Mathematica-defined functions are commenced with capital letters and encased in

square brackets; the Mathematica function N asks for the finite precision version of a
number.

• We can request as much precision as we desire. Twenty digits are requested for x by
N[x, 20]; fifty digits are requested by N[x, 50].

27.4 Lists, vectors, and matrices

Mathematica has straightforward ways to deal with lists of numbers. One-dimensional lists
can be thought of as vectors. Two-dimensional lists can be thought of as matrices. It is
possible to have higher dimensional lists. The code ch26c.nb has several examples of how
to deal with lists, including comments:

In[1]:= (* This is a comment *)

(* Assign the value 3 to variable x, "=" for assignment; then cube it *)

x = 3
x^3

CC BY-NC-ND. 19 January 2025, J. M. Powers.

198 CHAPTER 27. INTRODUCTION TO MATHEMATICA

Out[1]= 3

Out[2]= 27

In[3]:= (* Now assign x to be a vector, i.e. a list of numbers; {...} *)

x = {2, 4, 5}
(* Display x in matrix form *)
x // MatrixForm

Out[3]= {2, 4, 5}

Out[4]//MatrixForm=
 
2
4
5

In[5]:= (* Now assign A to be a matrix, i.e. a list of lists *)

A = {{1, 3, 2}, {4, 5, 2}, {0, 9, 1}}
(* Display A in matrix form *)
A // MatrixForm

Out[5]= {{1, 3, 2}, {4, 5, 2}, {0, 9, 1}}

Out[6]//MatrixForm=
 
1 3 2
4 5 2
0 9 1

In[7]:= (* Take the matrix-vector product b=A.x *)

b = A.x

Out[7]= {24, 38, 41}

In[8]:= (* Solve for x, knowing A and b via x = Inverse[A].b *)

x = Inverse[A].b

Out[8]= {2, 4, 5}

In[9]:= (* Solve for x if b = {1,2,3} *)

b = {1, 2, 3}
x = Inverse[A].b

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.4. LISTS, VECTORS, AND MATRICES 199

Out[9]= {1, 2, 3}

Out[10]= {5/47, 16/47, -(3/47)}

In[11]:= (* Select isolated components of x *)

x[[1]]
x[[2]]
x[[3]]

Out[11]= 5/47

Out[12]= 16/47

Out[13]= -(3/47)

In[14]:= (* Select an isolated component of A *)

A[[1, 1]]

Out[14]= 1

In[15]:= (* Create some structured lists *)

x = Table[0, {i, 1, 5}]
x = Table[i, {i, 1, 5}]
x = Table[i^2, {i, 1, 5}]
A = Table[Table[i*j, {j, 1, 3}], {i, 1, 3}]
A // MatrixForm

Out[15]= {0, 0, 0, 0, 0}

Out[16]= {1, 2, 3, 4, 5}

Out[17]= {1, 4, 9, 16, 25}

Out[18]= {{1, 2, 3}, {2, 4, 6}, {3, 6, 9}}

Out[19]//MatrixForm=
 
1 2 3
2 4 6
3 6 9

In[20]:= (* Give the LaTeX form of A *)

A // TeXForm

CC BY-NC-ND. 19 January 2025, J. M. Powers.

200 CHAPTER 27. INTRODUCTION TO MATHEMATICA

Out[20]//TeXForm=
\left(
\begin{array}{ccc}
1 & 2 & 3 \\
2 & 4 & 6 \\
3 & 6 & 9
\end{array}
\right)

27.5 Algebra
The code ch26d.nb gives some algebra operations:

(* This code will do some problems in algebra. *)

(* Expand a polynomial *)
Expand[(x + 1)^3]

Out[1]= 1 + 3 x + 3 x^2 + x^3

(* Factor a polynomial *)
Factor[x^2 + 4 x + 3]

Out[2]= (1 + x) (3 + x)

In[3]:= (* Solve an equation; "==" for equals *)

Solve[ x^2 + 5 x + 6 == 0, x]

Out[3]= {{x -> -3}, {x -> -2}}

(* Define an equation, then solve it; "=" for assignment *)

eq = x^2 + 5 x + 6 == 0
Solve[eq, x]
(* Result is a list of two rules; rules have -> *)

Out[4]= 6 + 5 x + x^2 == 0

Out[5]= {{x -> -3}, {x -> -2}}

In[6]:= (* Isolate the two solutions *)

(* /. denotes, "evaluate with" *)

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.6. SOME PLOTS 201

{s1, s2} = Solve[eq, x]

x1 = x /. s1
x2 = x /. s2

Out[6]= {{x -> -3}, {x -> -2}}

Out[7]= -3

Out[8]= -2

In[9]:= (* Solve a general quadratic equation and give the result in Fortran form *)
{s1, s2} = Solve[a*x^2 + b*x + c == 0, x]
x1 = x /. s1
x2 = x /. s2

Out[9]= {{x -> (-b - Sqrt[b^2 - 4 a c])/(2 a)}, {x -> (-b + Sqrt[b^2 - 4 a c])/(2 a)}}

Out[10]= (-b - Sqrt[b^2 - 4 a c])/(2 a)

Out[11]= (-b + Sqrt[b^2 - 4 a c])/(2 a)

In[12]:= x1 // FortranForm

Out[12]= (-b - Sqrt(b**2 - 4*a*c))/(2.*a)

In[13]:= x2 // FortranForm

Out[13]=(-b + Sqrt(b**2 - 4*a*c))/(2.*a)

27.6 Some plots

Let us plot
y = 2x2 + 1, x ∈ [−2, 2].
This is achieved with

Plot[2 x^2 + 1, {x, -2, 2}, AxesLabel -> {"x", "y"},

LabelStyle->{FontFamily->"Times"}]

The plot is in Fig. 27.2a. Had we not specified any FontFamily, we would have obtained
the default font.
We can make the fonts more readable and the axes labels italicized with the

CC BY-NC-ND. 19 January 2025, J. M. Powers.

202 CHAPTER 27. INTRODUCTION TO MATHEMATICA

y y

8 8

6 6

4 4

2 2

x x
-2 -1 1 2 -2 -1 1 2
a) b)

Figure 27.2: Plots of y = 2x2 + 1: a) small fonts, b) large fonts.

Plot[2 x^2 + 1, {x, -2, 2}, AxesLabel -> {x, y},

LabelStyle->{FontFamily->"Times"}, AxesStyle -> Directive[16]]

The plot is in Fig. 27.2b.

We can plot two curves by plotting a list containing the two functions to be plotted. Let
us plot
y = 2x2 + 1, y = x3 , x ∈ [−2, 2].
This is achieved by the command

Plot[{2 x^2 + 1, x^3}, {x, -2, 2}, AxesLabel -> {x, y},
LabelStyle->{FontFamily->"Times"}, AxesStyle -> Directive[16]]

The plot is in Fig. 27.3.

We can render log-log plots. Consider the function

y = 2x2 + 1, x ∈ [0.001, 10],

in a log-log plot. This is achieved by the command

LogLogPlot[2 x^2 + 1, {x, .001, 10}, AxesLabel -> {x, y},

LabelStyle->{FontFamily->"Times"}, AxesStyle -> Directive[16]]

The plot is in Fig. 27.4.

We can render plots of data points. Consider the data

(x, y) = (0, 0), (1, 1), (2, 2), (10, 10).

We can plot the data by the command

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.6. SOME PLOTS 203

x
-2 -1 1 2

-5

Figure 27.3: Plot of y = 2x2 + 1 and y = x3 .

Figure 27.4: Log-log plot of y = 2x2 + 1.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

204 CHAPTER 27. INTRODUCTION TO MATHEMATICA

y
10

x
2 4 6 8 10

Figure 27.5: List plot of the data (0, 0), (1, 1), (2, 2), (10, 10).

data = {{0, 0}, {1, 1}, {2, 2}, {10, 10}}

ListPlot[data, AxesLabel -> {x, y}, LabelStyle -> {FontFamily -> "Times"},
AxesStyle -> Directive[16], PlotStyle -> PointSize[0.015]]

The plot is in Fig. 27.5.

We can create two different types of plots and merge them via the following commands:
data = {{0, 0}, {1, 1}, {2, 2}, {10, 10}}
p1 = ListPlot[data, AxesLabel -> {x, y}, LabelStyle -> {FontFamily -> "Times"},
AxesStyle -> Directive[16], PlotStyle -> PointSize[0.015]];
p2 = Plot[x, {x, 0, 10}];
Show[p1, p2]
The plot is in Fig. 27.6.
We can make three-dimensional surface plots. Consider the function

z = x2 + y 2 , x ∈ [−4, 4], y ∈ [−4, 4].

Plot3D[x^2 + y^2, {x, -4, 4}, {y, -4, 4}, AxesLabel -> {x, y, z},
LabelStyle -> {FontFamily -> "Times"}, AxesStyle -> Directive[14]]
The plot is in Fig. 27.7.
We can save figures in a variety of formats, including .eps. This is achieved by using the
mouse to select the figure, then going to the menu under File, and choose Save Selection
As. These plots were generated directly from Mathematica. We could have also imported
them into Adobe Illustrator for further processing.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.6. SOME PLOTS 205

y
10

x
2 4 6 8 10

Figure 27.6: Plot of the data (0, 0), (1, 1), (2, 2), (10, 10) combined with a plot of y = x.

Figure 27.7: Plot of z = x2 + y 2.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

206 CHAPTER 27. INTRODUCTION TO MATHEMATICA

27.7 Calculus
A variety of operations from calculus are described in ch26e.nb:

In[65]:= y = 2 x^2 + 3
(* Indefinite integral *)
Integrate[y, x]
(* Definite integral *)
Integrate[y, {x, 0, 1}]

Out[65]= 3 + 2 x^2

Out[66]= 3 x + (2 x^3)/3

Out[67]= 11/3

In[68]:= (* First derivative *)

D[y, x]

Out[68]= 4 x

In[69]:= (* Second derivative *)

D[y, {x, 2}]

Out[69]= 4

In[121]:= (* Differential equation *)

Clear[y]
DSolve[{y’’[x] + y’[x] + 4 y[x] == 0}, y[x], x]
{sol} = DSolve[{y’’[x] + y’[x] + 4 y[x] == 0, y[0] == 1, y’[0] == 0},
y[x], x]
yy = y[x] /. sol
Plot[yy, {x, 0, 10}, PlotRange -> All, AxesLabel -> {x, y},
LabelStyle -> {FontFamily -> "Times"}, AxesStyle -> Directive[14]]

Out[122]= {{y[x] ->

E^(-x/2) C[2] Cos[(Sqrt[15] x)/2] +
E^(-x/2) C[1] Sin[(Sqrt[15] x)/2]}}

Out[123]= {{y[x] ->

1/15 E^(-x/
2) (15 Cos[(Sqrt[15] x)/2] + Sqrt[15] Sin[(Sqrt[15] x)/2])}}

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.8. THE EULER METHOD, DSOLVE, AND NDSOLVE 207

Figure 27.8: Plot of solution to y ′′ + y ′ + 4y = 0, y(0) = 1, y ′ (0) = 0.

Out[124]= 1/15 E^(-x/

2) (15 Cos[(Sqrt[15] x)/2] + Sqrt[15] Sin[(Sqrt[15] x)/2])

The plot is in Fig. 27.8.

27.8 The Euler method, DSolve, and NDSolve

We can solve differential equations with Mathematica. In ch26f.nb, we solve
dy
= λy, y(0) = 1, λ = −1, t ∈ [0, 1], (27.1)
dt
via three different methods:

• the exact solver DSolve,

• the Euler method, and

• the Mathematica numerical solver NDSolve.

Here is the code:

In[1]:= (* Solve dy/dt = -y, y(0) = 1 exactly *)

lambda = -1;
eqs = {y’[t] == lambda*y[t], y[0] == 1}
{sol} = DSolve[eqs, y[t], t]
yy = y[t] /. sol
p1 = Plot[yy, {t, 0, 1}, AxesLabel -> {t, y},

CC BY-NC-ND. 19 January 2025, J. M. Powers.

208 CHAPTER 27. INTRODUCTION TO MATHEMATICA

LabelStyle -> {FontFamily -> "Times"}, AxesStyle -> Directive[16]];

Out[2]= {Derivative[1][y][t] == -y[t], y[0] == 1}

Out[3]= {{y[t] -> E^-t}}

Out[4]= E^-t

In[6]:= (* Solve dy/dt = -y, y(0) = 1 with the Euler method *)

nt = 11;
p = 20
yyy = Table[0, {i, 1, nt}];
ttt = Table[0, {i, 1, nt}];
tstop = 1;
dt = N[tstop/(nt - 1), p];
yyy[[1]] = 1;
ttt[[1]] = 0;
Do[
{
yyy[[i + 1]] = yyy[[i]] + dt*lambda*yyy[[i]],
ttt[[i + 1]] = ttt[[i]] + dt
},
{i, 1, nt - 1}]
yt = Table[{ttt[[i]], yyy[[i]]}, {i, 1, nt}]
p2 = ListPlot[yt];
Show[p1, p2];

Out[7]= 20

Out[15]= {{0, 1}, {0.10000000000000000000,

0.90000000000000000000}, {0.20000000000000000000,
0.81000000000000000000}, {0.30000000000000000000,
0.72900000000000000000}, {0.40000000000000000000,
0.65610000000000000000}, {0.50000000000000000000,
0.59049000000000000000}, {0.60000000000000000000,
0.53144100000000000000}, {0.70000000000000000000,
0.47829690000000000000}, {0.80000000000000000000,
0.43046721000000000000}, {0.90000000000000000000,
0.38742048900000000000}, {1.0000000000000000000,
0.3486784401000000000}}

CC BY-NC-ND. 19 January 2025, J. M. Powers.

27.8. THE EULER METHOD, DSOLVE, AND NDSOLVE 209

Figure 27.9: Plot of exact and Euler solution to dy/dt = −y, y(0) = 1.

In[18]:= (* Solve dy/dt = -y, y(0) = 1 with Mathematica’s numerical solver *)

lambda = -1;
p = 20;
eqs = {y’[t] == lambda*y[t], y[0] == 1}
{sol} = NDSolve[eqs, y[t], {t, 0, 1}, WorkingPrecision -> p]
yy = y[t] /. sol
Plot[yy, {t, 0, 1}, AxesLabel -> {t, y},
LabelStyle -> {FontFamily -> "Times"}, AxesStyle -> Directive[16]];

Out[20]= {Derivative[1][y][t] == -y[t], y[0] == 1}

Out[21]= {{y[t] -> InterpolatingFunction[{{0,1.0000000000000000000}},<>][t]}}

Out[22]= InterpolatingFunction[{{0,1.0000000000000000000}},<>][t]

The plot of the exact and Euler solution is in Fig. 27.9. Some miscellaneous features include

• Use a semicolon, “;” to suppress output to the screen.

• Greek letters can be formatted by use of the esc key. For example for the symbol φ,
one enters esc f esc.

• C is reserved for use by Mathematica. So is N.

• Online help for commands is available in many ways. For example for more information
on NDSolve, enter ?NDSolve.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

210 CHAPTER 27. INTRODUCTION TO MATHEMATICA

• Occasionally, you will need to “reboot” Mathematica. This can happen when its cache
gets cluttered with old variables and you would like to refresh them. One can achieve
this by quitting the kernel. To do so, go to the Evaluation menu and select Quit
Kernel.

CC BY-NC-ND. 19 January 2025, J. M. Powers.

Bibliography

S. J. Chapman, 2018, Fortran for Scientists and Engineers, Fourth Edition, McGraw-Hill,
New York.
This useful text is written for first year students in engineering.

I. Chivers and J. Sleightholme, 2015, Introduction to Programming with Fortran with Cov-
erage of Fortran 90, 95, 2003, 2008 and 77, Third Edition, Springer, London.
This text gives an excellent introduction to the neophyte as well as useful information to the expert.
Practical details as well as underlying programming philosophy are included. Moreover, the authors
recognize that scientific computing is often the purpose of programming in Fortran, and include
useful examples directed at mathematical science problems which extend beyond programming.

N. S. Clerman and W. Spector, 2012, Modern Fortran: Style and Usage, Cambridge, New
York.
This is a modern text which does a good job explaining the usage of the Fortran language.

M. T. Heath, 2002, Scientific Computing: An Introductory Survey, Second Edition, McGraw-

Hill, New York.
This advanced undergraduate text focuses on algorithms for solving important mathematical problems
in science.

H. P. Langtangen, 2016, Python Scripting for Computational Science, Fifth Edition, Springer,
Berlin.
This text focuses on how to implement the Python scripting language for problems in scientific com-
puting. It includes a discussion of how to interface with lower level languages such as Fortran or
C.

A. Markus, 2012, Modern Fortran in Practice, Cambridge, New York.

This is an introductory text for learning the Fortran language.

W. E. Mayo and M. Cwiakala, 1995, Programming with Fortran 77, Schaum’s Outlines,
McGraw-Hill, New York.

211
 This is a readable guide which is very useful for learning programming in Fortran 77. While it does
not have all of the modern upgrades to the language, it is nevertheless a useful source, especially with
regard to legacy code.

W. E. Mayo and M. Cwiakala, 1995, Programming with Fortran 90, Schaum’s Outlines,
McGraw-Hill, New York.
This is a readable guide which is very useful for learning programming in Fortran 90.

M. Metcalf, J. Reid, and M. Cohen, 2011, Modern Fortran Explained, Oxford, Oxford.
This is a fine textbook which introduces the Fortran language.

W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling, 1992, Numerical

Recipes in Fortran 77: The Art of Scientific Computing, Second Edition, Cambridge,
New York.
This book is a tour de force of practical programming combined with applied mathematics to build
and execute a wide variety of useful algorithms to solve problems in scientific computing. It is highly
recommended.

W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling, 2007, Numerical

Recipes: The Art of Scientific Computing, Third Edition, Cambridge, New York.
This update to the second edition uses the C++ language.

G. Strang, 2007, Computational Science and Engineering, Wellesley-Cambridge, Wellesley,

MA.
This readable text presents material usually considered to be at the graduate level. That said, an
interested undergraduate can learn much from it, as the author has a style which lends itself to
students who are motivated by challenging material.

212
Index

32-bit architecture, 41, 49 COBOL, 10

64-bit architecture, 21, 41, 49, 50 comment statement, 40
compiler, 14, 15, 18, 22, 40, 41, 82, 97, 117,
abs, 108, 130 129, 132, 165, 166, 174, 175, 178
acos, 108 error, 41
acosh, 108 option, 18
Ada, 10 warning, 41
Adobe Illustrator, 28, 61, 204 complex, 129
AFS, 15, 32 complex, 41
aimag, 129 complex conjugate, 130
ALGOL, 10 complex vectors, 75
algorithm, 12, 35, 66, 83, 146, 150 conjg, 130
Andrew File System, 15 contains, 111
arithmetic, 45 cos, 108
ASCII, 22, 27, 41, 52, 96 cosh, 108
asin, 108, 132 cp, 31
asinh, 108 Cygwin, 18, 19
assembly language, 22
atan, 57, 108 derived types, 137
atanh, 108 do, 39
do while, 119, 123, 125
BASIC, 10, 183 DOS, 11
binary executable, 12, 15, 103, 116, 117 dot product, 73, 74
binary representation of integers, 48 Duffing equation, 150, 154
bit, 21, 41, 49 dummy variable, 112, 142
byte, 21
Emacs, 19, 28
C, 10, 13, 15, 175 Euler method, 123, 125, 145, 150, 176, 180,
C++, 10, 109, 110 184, 207
C#, 10 executable script file, 117
cat, 31 exp, 108
cd, 30
Central Processing Unit (CPU), 22 format, 41
character, 39, 42 formatted, 99
cloud, 15 Fortran, 10, 12–15, 18–22, 27, 32, 39–41, 43,
cmplx, 108 45–47, 51, 56, 57, 63, 65, 67, 69, 73–

213
 75, 78, 80, 81, 83, 84, 86, 87, 99, 107, MATLAB, 11, 20–22, 60, 61, 73, 80, 173
109, 110, 112, 119–121, 124, 127, 129– matmul, 73, 78, 80, 168
132, 165 matrix addition, 84
Fortran 2003, 11, 14, 18, 165 matrix-matrix multiplication, 80
Fortran 2008, 11, 132, 165 matrix-vector multiplication, 78
Fortran 77, 165–168 Microsoft Excel, 11, 22, 179, 183
Fortran 90, 11, 14, 18, 165 Microsoft Windows, 11
Fortran 95, 11, 165 mkdir, 31
function subroutine, 107, 111–113, 115, 116 module, 109–111, 113, 150, 163
more, 31
gedit, 19, 28, 52 mv, 31
gfortran, 18, 174
NaN, Not a Number, 108
hello world, 14, 165, 175, 179, 183, 195
netfile, 15
HTML, 28, 30
non-linear ordinary differential equations, 145,
if, 39 150
ifort, 18, 22, 41, 49, 51, 97, 132, 166
object-oriented programming, 35, 109
imag, 129
open, 96
implicit none, 40, 43
operations, 45
input output, I/O, 39, 41, 43
ordinary differential equations, 10, 123, 145
int, 108
OS-X, 11, 13
integer, 39, 41
output, 89
integer division, 46
intent, 111, 142 overflow, 49, 93
interpreted language, 21 Pascal, 10
Java, 10, 109, 110 pointers, 139
pre-defined functions, 107
Kate, 19 print, 39
kind, 41 pwd, 30
Python, 173
LATEX, 27, 200
less, 31 quadratic equation, 119, 143, 201
Linux, 11, 13, 15–19, 27, 28
load, 61 Random Access Memory (RAM), 22
local variable, 142 read, 39, 99
log, 108 real, 39, 41, 108, 129
log10, 108 rm, 31
logical, 42 rmdir, 31
ls, 30 Runge-Kutta method, 146, 150

Make, 118 secure shell, 15

mass-spring-damper system, 150 sin, 108
Mathematica, 11, 195 sinh, 108

214
software, 22
sqrt, 108
ssh, 15
subroutine, 35, 141–143, 150, 161, 172, 173
function, 107, 111–113, 115, 116

tan, 108
tanh, 108
Taylor series, 173
Terminal, 15
terminal, 17, 19, 22, 31
TeXnicCenter, 27, 28
TeXShop, 27, 28
text editor, 19, 22
transpose, 76
transpose, 73

undefined variables, 51
underflow, 49
unformatted, 41, 99
Linux, 13
UNIX, 11, 19, 20, 30–32, 117
user-defined functions, 109

VBA, 10, 183

vi, 19, 28

Windows, 13, 17, 19, 27, 28, 31, 183

write, 39, 52, 60
WYSIWYG, 13, 30, 31

X Windows, 17
X11, 17
xming, 17
XQuartz, 17

215