100% found this document useful (4 votes)
48 views83 pages

Mathematical Modeling the Life Sciences: Numerical Recipes in Python and MATLAB 1st Edition N. G. Cogan instant download

The document presents 'Mathematical Modeling the Life Sciences: Numerical Recipes in Python and MATLAB' by N. G. Cogan, which aims to bridge the gap between numerical solutions and modeling techniques in biological applications. It emphasizes minimal mathematics, focusing on implementation, application, and sensitivity analysis, with practical codes provided in Python and MATLAB. The text is designed to enhance understanding of mathematical modeling in life sciences through historical context and practical examples.

Uploaded by

datuszoppojw
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (4 votes)
48 views83 pages

Mathematical Modeling the Life Sciences: Numerical Recipes in Python and MATLAB 1st Edition N. G. Cogan instant download

The document presents 'Mathematical Modeling the Life Sciences: Numerical Recipes in Python and MATLAB' by N. G. Cogan, which aims to bridge the gap between numerical solutions and modeling techniques in biological applications. It emphasizes minimal mathematics, focusing on implementation, application, and sensitivity analysis, with practical codes provided in Python and MATLAB. The text is designed to enhance understanding of mathematical modeling in life sciences through historical context and practical examples.

Uploaded by

datuszoppojw
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 83

Mathematical Modeling the Life Sciences:

Numerical Recipes in Python and MATLAB 1st


Edition N. G. Cogan install download

https://ebookmeta.com/product/mathematical-modeling-the-life-
sciences-numerical-recipes-in-python-and-matlab-1st-edition-n-g-
cogan/

Download more ebook from https://ebookmeta.com


We believe these products will be a great fit for you. Click
the link to download now, or visit ebookmeta.com
to discover even more!

MATLAB for Engineering and the Life Sciences 2nd


Edition Joe Tranquillo

https://ebookmeta.com/product/matlab-for-engineering-and-the-
life-sciences-2nd-edition-joe-tranquillo/

Simulation with Python: Develop Simulation and Modeling


in Natural Sciences, Engineering, and Social Sciences
1st Edition Rongpeng Li

https://ebookmeta.com/product/simulation-with-python-develop-
simulation-and-modeling-in-natural-sciences-engineering-and-
social-sciences-1st-edition-rongpeng-li/

Mathematical Methods for Life Sciences 1st Edition


Cinzia Bisi

https://ebookmeta.com/product/mathematical-methods-for-life-
sciences-1st-edition-cinzia-bisi/

Incubus 3 1st Edition Brandon Varnell

https://ebookmeta.com/product/incubus-3-1st-edition-brandon-
varnell/
Literary Publishing in America, 1790-1850 William
Charvat

https://ebookmeta.com/product/literary-publishing-in-
america-1790-1850-william-charvat/

Making Revolution My Life in the Black Panther Party


1st Edition Don Cox

https://ebookmeta.com/product/making-revolution-my-life-in-the-
black-panther-party-1st-edition-don-cox/

Pattern Recognition and Data Analysis with Applications


Deepak Gupta

https://ebookmeta.com/product/pattern-recognition-and-data-
analysis-with-applications-deepak-gupta/

One Last Breath 1st Edition K T Galloway

https://ebookmeta.com/product/one-last-breath-1st-edition-k-t-
galloway/

Steel Reinforced Concrete Structures Assessment and


Repair of Corrosion 3rd Edition Mohamed Abdallah El-
Reedy

https://ebookmeta.com/product/steel-reinforced-concrete-
structures-assessment-and-repair-of-corrosion-3rd-edition-
mohamed-abdallah-el-reedy/
Comics and Adaptation 1st Edition Benoît Mitaine Editor
David Roche Editor Isabelle Schmitt Pitiot Editor
Aarnoud Rommens David Roche

https://ebookmeta.com/product/comics-and-adaptation-1st-edition-
benoit-mitaine-editor-david-roche-editor-isabelle-schmitt-pitiot-
editor-aarnoud-rommens-david-roche/
Mathematical
Modeling the Life
Sciences

The purpose of this unique textbook is to bridge the gap between the need for numerical
solutions to modeling techniques through computer simulations to develop skill in em-
ploying sensitivity analysis to biological and life sciences applications.

The underpinning mathematics is minimalized. The focus is on the consequences, imple-


mentation, and application. Historical context motivates the models. An understanding of
the earliest models provides insight into more complicated ones.

While the text avoids getting mired in the details of numerical analysis, it demonstrates
how to use numerical methods and provides core codes that can be readily altered to fit a
variety of situations.

Numerical scripts in both Python and MATLAB® are included. Python is compiled in
Jupyter Notebook to aid classroom use. Additionally, codes are organized and available
online.

One of the most important skills requiring the use of computer simulations is sensitivity
analysis. Sensitivity analysis is increasingly used in biomathematics. There are numerous
pitfalls to using sensitivity analysis and therefore a need for exposure to worked examples
in order to successfully transfer their use from mathematicians to biologists.

The interconnections between mathematics and the life sciences have an extensive his-
tory. This book offers a new approach to using mathematics to model applications using
computers, to employ numerical methods, and takes students a step further into the realm
of sensitivity analysis. With some guidance and practice, the reader will have a new and
incredibly powerful tool to use.

Nicholas G. Cogan is Professor of Mathematics at Florida State University. He began


studying mathematical biology in undergraduate school and received his Ph.D. from
the University of Utah under James P. Keener. He routinely works with microbiolo-
gists, environmental engineers, clinicians, and other scientists outside of mathematics.
He has taught for twenty years at the undergraduate level and his research focuses on
mathematical modeling in the life sciences. He is the author of over fifty articles using
mathematics with biology.
Textbooks in Mathematics
Series editors:

Al Boggess, Kenneth H. Rosen

An Introduction to Complex Analysis and the Laplace Transform


Vladimir Eiderman

An Invitation to Abstract Algebra


Steven J. Rosenberg
Numerical Analysis and Scientific Computation
Jeffery J. Leader

Introduction to Linear Algebra


Computation, Application and Theory
Mark J. DeBonis

The Elements of Advanced Mathematics, Fifth Edition


Steven G. Krantz

Differential Equations
Theory, Technique, and Practice, Third Edition
Steven G. Krantz

Real Analysis and Foundations, Fifth Edition


Steven G. Krantz

Geometry and Its Applications, Third Edition


Walter J. Meyer

Transition to Advanced Mathematics


Danilo R. Diedrichs and Stephen Lovett

Modeling Change and Uncertainty


Machine Learning and Other Techniques
William P. Fox and Robert E. Burks

Abstract Algebra
A First Course, Second Edition
Stephen Lovett

Multiplicative Differential Calculus


Svetlin Georgiev, Khaled Zennir

Applied Differential Equations


The Primary Course
Vladimir A. Dobrushkin

Introduction to Computational Mathematics: An Outline


William C. Bauldry

Mathematical Modeling the Life Sciences


Numerical Recipes in Python and MATLAB®
N. G. Cogan

https://www.routledge.com/Textbooks-in-Mathematics/book-series/CANDHTEXBOOMTH
Mathematical
Modeling the Life
Sciences
Numerical Recipes in Python
and MATLAB®

N. G. Cogan
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks
does not warrant the accuracy of the text or exercises in this book. This book’s use or discussion of
MATLAB® software or related products does not constitute endorsement or sponsorship by The
MathWorks of a particular pedagogical approach or particular use of the MATLAB® software.

First edition published 2023


by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press


4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

CRC Press is an imprint of Taylor & Francis Group, LLC

© 2023 Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and pub-
lisher cannot assume responsibility for the validity of all materials or the consequences of their use.
The authors and publishers have attempted to trace the copyright holders of all material reproduced
in this publication and apologize to copyright holders if permission to publish in this form has not
been obtained. If any copyright material has not been acknowledged please write and let us know so
we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.
com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA
01923, 978-750-8400. For works that are not available on CCC please contact mpkbookspermis-
[email protected]

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are
used only for identification and explanation without intent to infringe.

ISBN: 9780367554934 (hbk)


ISBN: 9781032328263 (pbk)
ISBN: 9781003316930 (ebk)

DOI: 10.1201/9781003316930

Publisher’s note: This book has been prepared from camera-ready copy provided by the authors.
I am deeply grateful to my mentors. Clyde Martin invited me into the
field of mathematical biology when I was an unruly undergraduate.
Jim Keener taught me how to look at problems without fear. Lisa
Fauci taught me how to think deeply about a problem, care about the
details, and how to mentor. Yousuff Hussaini taught me to think about
the big picture and look for connections across disciplines and topics.
I have also been blessed with a wonderful group of colleagues and
students at Florida State University and cannot be grateful enough
for this community.

However

There is no doubt that my family is the most important thing in my


life. My wife has supported me, pushed me, and loved me. My children
have challenged me, inspired me, and encouraged me. I could not
have done this without them.
Contents

Foreword xi

1 Introduction 1
1.1 What Is a Model? . . . . . . . . . . . . . . . . . . . 1
1.2 Projectile Motion . . . . . . . . . . . . . . . . . . . 2
1.3 Problems . . . . . . . . . . . . . . . . . . . . . . . 8

2 Mathematical Background 9
2.1 Mathematical Preliminaries . . . . . . . . . . . . . 9
2.1.1 Linear . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Nonlinear Equations . . . . . . . . . . . . . 13
2.2 Linearization . . . . . . . . . . . . . . . . . . . . . 15
2.3 Qualitative Analysis . . . . . . . . . . . . . . . . . 18
2.4 Problems . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Appendix: Planar Example . . . . . . . . . . . . . . 23

3 Introduction to the Numerical Methods 27


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Best Practices in Coding . . . . . . . . . . . . . . . 29
3.2.1 Folder Structure . . . . . . . . . . . . . . . . 29
3.2.2 Naming Conventions . . . . . . . . . . . . . 30
3.2.3 Code Structure . . . . . . . . . . . . . . . . 30
3.2.4 Comments . . . . . . . . . . . . . . . . . . . 32
3.3 Getting the Programs Running . . . . . . . . . . . . 32
3.3.1 Python . . . . . . . . . . . . . . . . . . . . . 32
3.3.2 MATLAB® . . . . . . . . . . . . . . . . . . 34
3.4 Initial Programs . . . . . . . . . . . . . . . . . . . . 34
3.4.1 Differential Equations in Python . . . . . . . 37
3.4.2 Differential Equations in MATLAB® . . . . . 38
3.5 Problems . . . . . . . . . . . . . . . . . . . . . . . 40

vii
viii Contents

3.6 Appendix: Sample Scripts . . . . . . . . . . . . . . 44


3.6.1 Python . . . . . . . . . . . . . . . . . . . . . 44
3.6.2 MATLAB® . . . . . . . . . . . . . . . . . . 45

4 Ecology 47
4.1 Historical Background . . . . . . . . . . . . . . . . 47
4.2 Single Species Models . . . . . . . . . . . . . . . . 50
4.2.1 The Exponential Model . . . . . . . . . . . . 51
4.2.2 The Logistic Model . . . . . . . . . . . . . . 53
4.2.3 Analysis . . . . . . . . . . . . . . . . . . . . 53
4.2.4 Predator/Prey – Lotka-Volterra . . . . . . . . 54
4.2.5 Analysis . . . . . . . . . . . . . . . . . . . . 56
4.2.6 Sensitivity: One at a Time, Scatterplots . . . 59
4.3 Competitive Exclusion . . . . . . . . . . . . . . . . 62
4.3.1 Model . . . . . . . . . . . . . . . . . . . . . 63
4.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 65
4.3.3 Sensitivity: Linear Regression . . . . . . . . 68
4.4 State of the Art and Caveats . . . . . . . . . . . . . 71
4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . 73

5 Within-host Disease Models 77


5.1 Historical Background . . . . . . . . . . . . . . . . 77
5.2 Pathological: Tumor . . . . . . . . . . . . . . . . . 81
5.2.1 Model . . . . . . . . . . . . . . . . . . . . . 82
5.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 84
5.2.3 Sensitivity: Direct Estimation . . . . . . . . . 87
5.3 Viral: Acute Infection . . . . . . . . . . . . . . . . . 92
5.3.1 Model . . . . . . . . . . . . . . . . . . . . . 92
5.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 94
5.3.3 Sensitivity Analysis: Feature Sensitivity . . . 96
5.4 Chronic: Tuberculosis . . . . . . . . . . . . . . . . 99
5.4.1 Model . . . . . . . . . . . . . . . . . . . . . 99
5.4.2 Analysis . . . . . . . . . . . . . . . . . . . . 101
5.4.3 Sensitivity: Relative Change . . . . . . . . . 103
5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . 104
5.6 Appendix . . . . . . . . . . . . . . . . . . . . . . . 108
Contents ix

6 Between Host-Disease Models 111


6.1 Historical Background . . . . . . . . . . . . . . . . 111
6.2 Two Compartment Models . . . . . . . . . . . . . . 115
6.2.1 Model . . . . . . . . . . . . . . . . . . . . . 115
6.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 117
6.2.3 Sensitivity Analysis: Spider Plot . . . . . . . 118
6.3 Classical SIR . . . . . . . . . . . . . . . . . . . . . 121
6.3.1 Model . . . . . . . . . . . . . . . . . . . . . 121
6.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 122
6.3.3 Sensitivity Analysis: Tornado Plots . . . . . . 125
6.4 Waning Antigens . . . . . . . . . . . . . . . . . . . 126
6.4.1 Model: SIRS . . . . . . . . . . . . . . . . . 127
6.4.2 Analysis . . . . . . . . . . . . . . . . . . . . 129
6.4.3 Sensitivity Analysis: Cobweb Diagrams . . . 129
6.5 Caveats and State of the Art . . . . . . . . . . . . . 131
6.6 Problems . . . . . . . . . . . . . . . . . . . . . . . 132

7 Microbiology 135
7.1 Historical Background . . . . . . . . . . . . . . . . 135
7.2 Bacterial Growth: Chemostat . . . . . . . . . . . . . 138
7.2.1 Model . . . . . . . . . . . . . . . . . . . . . 140
7.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 141
7.2.3 Sensitivity Analysis: Correlation Coefficient,
Pearson’s Moment Correlation . . . . . . . . 142
7.3 Multiple State Models: Free/attached . . . . . . . . 146
7.3.1 Model: Freter . . . . . . . . . . . . . . . . . 148
7.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 150
7.3.3 Sensitivity Analysis: Correlation Coefficient,
Spearman . . . . . . . . . . . . . . . . . . . 151
7.4 Cooperators, Cheaters, and Competitions . . . . . . 154
7.4.1 Model . . . . . . . . . . . . . . . . . . . . . 155
7.4.2 Analysis . . . . . . . . . . . . . . . . . . . . 156
7.4.3 Sensitivity Analysis: Sensitivity in Time and
Partial Rank Correlation Coefficient (PRCC) 157
7.5 State of the Art and Caveats . . . . . . . . . . . . . 161
7.6 Problems . . . . . . . . . . . . . . . . . . . . . . . 162
x Contents

8 Circulation and Cardiac Physiology 167


8.1 Historical Background . . . . . . . . . . . . . . . . 167
8.2 Blood Circulation Models . . . . . . . . . . . . . . 173
8.2.1 Model: Algebraic . . . . . . . . . . . . . . . 174
8.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 176
8.2.3 Sensitivity Analysis: Sampling Methods . . . 176
8.3 Cardiac Physiology . . . . . . . . . . . . . . . . . . 177
8.3.1 Model: Noble . . . . . . . . . . . . . . . . . 178
8.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 181
8.3.3 Sensitivity Analysis: Morris Screening . . . . 183
8.4 State of the Art and Caveats . . . . . . . . . . . . . 186
8.5 Problems . . . . . . . . . . . . . . . . . . . . . . . 187

9 Neuroscience 189
9.1 Historical Background . . . . . . . . . . . . . . . . 189
9.2 Action Potential . . . . . . . . . . . . . . . . . . . . 192
9.2.1 Model: Hodgkin-Huxley . . . . . . . . . . . 194
9.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 196
9.2.3 Sensitivity Analysis: ANOVA – Sobol’ . . . . 196
9.3 Fitzhugh-Nagumo . . . . . . . . . . . . . . . . . . 201
9.3.1 Model . . . . . . . . . . . . . . . . . . . . . 202
9.3.2 Analysis . . . . . . . . . . . . . . . . . . . . 203
9.3.3 Sensitivity: Moment Independent . . . . . . . 203
9.4 State of the Art and Caveats . . . . . . . . . . . . . 208
9.5 Problems . . . . . . . . . . . . . . . . . . . . . . . 209

10 Genetics 211
10.1 Historical Background . . . . . . . . . . . . . . . . 211
10.2 Heredity . . . . . . . . . . . . . . . . . . . . . . . . 215
10.2.1 Mathematics . . . . . . . . . . . . . . . . . 216
10.2.2 Analysis . . . . . . . . . . . . . . . . . . . . 217
10.2.3 Sensitivity: Factorial Design . . . . . . . . . 218
10.3 State of the Art and Caveats . . . . . . . . . . . . . 219
10.4 Problems . . . . . . . . . . . . . . . . . . . . . . . 220

Bibliography 223

Index 229
Foreword

This book was originally conceived as a book by a mathematician but


intended for biologists – or at least readers whose primary interest is in
biological insights. This seems an impossible task since there are in-
herent differences in the two disciplines. Biology is primarily an obser-
vational science while mathematics is primarily a constructive science.
However, mathematical biology as a subdiscipline of both mathemat-
ics and biology has moved from a small community into an incredibly
active and diverse field. This is because both mathematics and biology
benefit from the close communication.
But this leads us to ask what pieces of information or skills are re-
quired? How much mathematics does a biologist need to really know?
How much biology does a mathematician really need to know? Our an-
swer is “Enough to do what is necessary”. With that as our philosophy,
we have tried to minimize the underpinning of the mathematics and
focus on the consequences and implementation/application. It really is
important to understand where some of the analysis comes from – es-
pecially if you ever want to do anything that is even slightly beyond
the scope of this book. But, we have also provided numerical codes
that are reasonably generic. In our view, most of dangers of numeri-
cal packages is pushing them beyond what they are intended for, so
we have tried to include some problems that show you danger signs
to look out for. Similarly, we do not claim to be inclusive or complete
regarding any of the biological topics. We have tried to motivate the
models from a historical context since an understanding of the earliest
models and biological questions provide insight into the more compli-
cated models in the current literature.
The main important theme in this book that differentiates it from
other mathematical biology texts, is sensitivity analysis. This was in-
troduced more than 100 years ago and it has become more widely used.
But very few people know where sensitivity measures come from or

xi
xii Foreword

what methods can be used. Even fewer people have access to robust
methods that are simple to implement for a variety of problems. Most
packages are either not flexible enough or so packed that it is very
daunting for a non-expert to think about using them. Hopefully with
some guidance and practice, the reader will have a new and incredibly
powerful tool to use.
The book is aimed at undergraduates, but then almost all the books
I used in graduate school claimed the same thing. I have used the
material presented here with both graduates and undergraduates. I
feel that the material is too cursory for a decent graduate course in
mathematics – there are many other mathematical biology textbooks
that are more complete. But, even in a graduate curriculum in mathe-
matics, there is a need for practice in implementation.
Our guess is that the notation is a bit daunting for a typical un-
dergraduate biology major. However with a single course in calculus
and a bit of priming in differential equations, the material is accessible.
In fact, this is one of the aims of the presentation – we believe that a
practicing biologist should have a course in calculus and really needs
some understanding of differential equations. But not what is typically
taught in a mathematics department. Most real biological processes
avoid the issues that are the heart of the mathematics of differential
equations. Dynamical systems (i.e. differential equations where time
is the underlying independent variable) are reasonably well behaved
unless the defining terms are complicated (or more specifically if there
are discontinuities, or certain nonlinearities). But these are typically
not found in biologically motivated problems and are definitely not the
place to start! On the other hand, almost all interesting behaviors in
the real world are nonlinear. A first course in differential equations of-
ten spends the majority of the time on linear equations. So how do we
bridge this gap between the relatively simple and the relatively useful?
Our main tools are numerical. Along with demonstrating sensitivity
analysis, we show many examples of manipulating differential equa-
tions and numerical methods. With this as a foundation, there are many
very practical problems that are within reach.
If a student has no exposure to differential equations, some study
in Chapter 2 will be required. This introduces the main tools that are
needed: Steady-states, stability, qualitative analysis, and time-course
Foreword xiii

plots. One of the most important goals in the book is to allow, en-
courage, and in some sense require that the students can use numerical
methods to calculate much of this. We have provided scripts that are
relatively generic and can produce all of these. Of course, there is no
such thing as a truly black-box script and the user has to be responsi-
ble for understanding that numerical methods can miss steady-states,
produce erroneous estimates and are not able to check the users signs!
The rule of GIGO (Garbage In, Garbage Out) is not just a rule of thumb
but a governing principle. But the methods described and the example
models provides a working template that will be useful to a wide vari-
ety of scientists.
The organization of the book is relatively linear in the topic of sen-
sitivity analysis. The first few chapters deal with one-at-a-time meth-
ods and those that are primarily visual or geometric. For these, we have
included numerical scripts that show the logic and implementation.
The purpose of these files is to show the flow of information and
they are not designed for efficiency. In the later chapters, we introduce
more sophisticated methods (e.g. Screening methods, PRCC, Sobol’).
We have chosen to show how to use other packages for these rather
than show the native code. Partly, the implementation requires infor-
mation that is far beyond the scope of this class to understand – for ex-
ample almost all implementations of Sobol’ indices use Monte Carlo
integration. A second reason for showing this is to show that it is pos-
sible to use very sophisticated methods as long as the student has suf-
ficient background. PRCC, Sobol’, and screening methods are used in
high-level, cutting-edge research – and it is within reach of the stu-
dents.
On the other hand, the biological topics are completely inter-
changeable. We have attempted to use examples that are relevant for
sensitivity methods. For example, it is not useful to use a screening
method for a small number of parameters. But an instructor should be
able to pull biological material from any chapter. The numerical scripts
are reasonably generic and after a few weeks of exposure I have found
students are able to manipulate the codes very easily.
There is some required background, mostly from the mathemat-
ical side. A sequence in calculus and some exposure to differential
equations is really all that is needed. The first three chapters describe
xiv Foreword

the philosophy of modeling which is typically not well articulated in


textbooks. This is followed by some mathematical requirements and
advice on numerical practices. When I have taught from this material,
this comprises about two weeks of a semester – then the interesting
material is accessible.
We use both MATLAB® and Python for the numerical portions.
They both have strengths and weaknesses but it is very good to see that
they are quite similar. Being able to implement, run and read code in
both is quite useful. We will eventually include R and Julia since these
are becoming wide-spread. In my opinion, there is little reason to learn
uninterpreted languages like Fortran and C for most people who focus
on modeling. The models change too quickly for any payoff in terms
of efficiency.
Numerical methods can be complemented by symbolic methods
like Mathematica and Maple. In my opinion these packages are not
well suited for these sorts of methods – however, it is useful to be
aware of them.
Finally, I hope that the material is interesting – I am fascinated by
the historical setting for both the science and the mathematics. The
beginning of each chapter covers some of the relevant history. I hope it
interests the reader enough to dig deeper into where these topics came
from.

N. G. Cogan
Tallahassee, Fl
1
Introduction

Modeling is a central concept in mathematical biology since there are


not fundamental “laws” to rely on. Therefore it is important to have an
idea of a model in mind. By definition, models are approximations of
reality. Good models are capable of providing insight into biological
hypotheses and built to be refined and adjusted.

All models are wrong, but


some are useful.

Attributed to George E. P. Box

Models should be as simple as


possible, but no simpler.

Adaptation of Albert Einstein

1.1 What Is a Model?


Models are the translation of a real world problem into a statement
or question in mathematical language. Models connect the lab scien-
tist, clinician, or practitioner to the mathematician/quantitative scien-
tist. Bridging this gap requires an understanding of both languages, as
well as an appreciation of the gap between them.
Like translations, there can be good and bad translations – those
that are accurate, those that capture the spirit of the primary language
and those that open up insights. So what makes a good model? This
depends a lot of the context. Some qualities that make good models
DOI: 10.1201/9781003316930-1 1
2 Introduction

include accuracy (how true to the science is the model), flexibility (if
you change the experiment slightly, do you need a brand new model?),
simplicity (is the model tractable?), practicality (does it include things
that are not measurable?).

1.2 Projectile Motion


There are many relationships that are really models. For example,
D = rt models the how the distance travels depends on the rate and
time. The ideal gas law relates pressure, volume, and temperature,
PV = nRT . Einstein derived a model describing the relationship be-
tween energy and mass, e = Mc2 . All of these are models that provide
insight into different processes. There are distinctions between these
models. It is hard to imagine how D = rt could not be an accurate
model (although there is an underlying assumption that the rate is con-
stant) while the ideal gas law becomes less and less accurate when the
gas is close to a phase transition, say near the condensation point. The
last relationship depends on assumptions underlying much of modern
physics. So models can come in a range of subjective interpretation.
We will detail one model that is familiar to any student in calculus
– projectile motion. We assume that an object is fired directly upward
from a certain height above ground. The object goes up, stops moving,
reverses direction and returns down. A description of the height of the
object as a function of time is y(t) = − 21 gt 2 + v0t + y0 . Gravitational
acceleration is denoted g, the initial velocity and positions are v0 and
y0 , respectively. There is a choice in frame of reference where we are
assuming that gravity only acts in one direction (the object is dropped
or thrown vertically with no horizontal displacement) and that gravity
causes acceleration in the “negative” direction. If the object starts with
a positive height, y0 and is released the height will decrease.
This is often presented as “projectile motion” and as though it is
always true. We could think of this function as the model – we can
make predictions about the time it takes to return to y = 0 or what is
the maximum height the object reaches. However, viewing this rela-
tionship that way obscures a broader modeling method – How do we
Projectile Motion 3

adjust this for multiple objects? Is this accurate in all velocity regimes?
What about projectile motion in different environments?
In calculus you might have seen this as a question about motion un-
der constant acceleration. Acceleration is the rate of change of velocity,
v(t) and Newton argued that acceleration due to gravity is constant so
that,
dv
= −g.
dt
Since g is a constant, we can integrate both sides,

v(t) = −gt + c0 .

Now we know how the velocity changes over time and velocity is
the rate of change of position,
dy
= −gt + c0 .
dt
Since g and c0 are constant, we can integrate both sides,
1
y(t) = − gt 2 + c0t + c1 . (1.1)
2
You can almost see the correspondence between y = − 21 gt 2 + v0t +
y0 and y(t) = − 12 gt 2 + c0t + c1 , but we have not yet determined the
integration constants. We have to provide some information to do this.
If we denote the initial position – the position at time 0 as y(0) = y0 ,
and plug this into 1.1, we find that c1 = y0 .
So what is c0 ? One way to determine it is if we define the initial
velocity, v(0) = v0 . Then the velocity (the derivative of the position,
y(t)) is,
dy(t)
= −gt + c0 . (1.2)
dt
Evaluating this at t = 0, we find that c0 = v0 .
So the “model” y = − 21 gt 2 + v0t + y0 is really the solution to the
model,
d2y
= −g. (1.3)
dt 2
4 Introduction

This is a differential equation, and if we prescribe information


y(0) = y0 and dydt (0) = v0 , we can solve this completely.
But where did the model come from? Part of the point of this book
is to learn how to invent these sorts of models, so we won’t go into
too much detail here. However, this is a nice example that has roots
to Newton and the time that calculus was created. One of Newton’s
most valuable contribution is the idea of Laws – that are general prin-
ciples that lead to mathematical statements. One of his laws of motion
(the second) states that f = ma. That is, there is a balance between
forces applied on an object and the acceleration of the objects mass.
So Equation 1.3 is really a translation of Newton’s second law (with a
bit of algebra),
ma = f ,
2
d y(t) f
2
= ,
dt m
= −g.
This is a model of the position of an object under constant accelera-
tion. From a modeling standpoint, we started with a very broad “law” –
Newton’s second law – and then translated this in the case of assuming
only (constant) gravitational attraction. This model can be completely
solved for predictions. The next tasks depend on many things including
the goals of the model, how the predictions compare to observations;
however, for all good models, there are testable predictions.
In this case, we could predict the maximum height of an object that
is thrown, the time at which the object reaches that height, when the
object hits the ground, etc. But what if we talk to an experimentalist
and we are asked what the terminal velocity is? This is the constant
speed that an object reaches when falling. But a look at this model
indicates that there is no such thing. In fact, this model assumes the
velocity increases like the square of the time.
As modelers, we have to ask what is wrong – our prediction does
not match observations. A little exploration uncovers the trouble. There
are many other forces that act on the object. There is air-resistance,
coriolis force, even forces at the atomic scales. We have to consider
which to include and which to exclude. Do we need to include rel-
ativistic effects? This is a commonly accepted model of motion that
appears at odds with Newton’s model.
Projectile Motion 5

There is no complete answer for this. In principle all forces need to


be accounted for, but this is like a 1-1 scale map. It does not really help
matters. But we can work at estimating which forces matter the most.
In this case, unless the velocity gets very large, or the mass of the object
is tiny or huge, the main forces that matter when considering terminal
velocity is drag. We can then develop a model for drag and go back
to our experimental colleague and compare our new, revised predic-
tion with their observations. This loop between observations leading
to models leading to predictions leading to revised observations is one
of the most successful ways for science to progress. Modeling is a cru-
cial step to this leading to quantitive questions (how fast does an object
have to move before constant acceleration deviates from observations
by more than 5%) that drive experimental design and insight.
In this book, we will also consider a less standard view of models.
In many situations, we know that there are many interactions. Consider
a gene network schematic shown in Figure 1.1. Clearly there is a lot
going on. For complicated networks, the models get extremely large –
think thousands of equations with tens of thousands of parameters.
How can we possibly assess such a model?
We will show how to use sensitivity analysis to reduce the model
and make simplifications that greatly aid the connection between mod-
els and experiments. This has a different philosophy than used to un-
derstand projectile motion. In that situation, we had a simple model
that may or may not reflect observations. If the predictions are inaccu-
rate, we look for additional terms. We could think of this as a bottom
up approach where you start small and add to the model as need. A top
down approach takes a large amount of current knowledge and tries to
pinpoint the most relevant parts for specific predictions. Both stand-
points are valid and useful and we will move back and forth between
them but it is very useful to know the differences between them.
An additional issue that mathematical biologists have to grapple
with is the lack of universal laws in biology. There are very few prin-
ciples that can be viewed in a way that is similar to Newton’s laws of
motion. Instead, we have to work a bit harder to develop principles that
can be translated into theoretical models. Some principles are broad –
we use conservation laws in many of the chapters. Some are known
from other disciples – the law of mass action has been well developed
6 Introduction

Figure 1.1: Example of a network diagram.

in chemistry. Other principles that guide models is the philosophy that


mathematics is portable and that some aspects of models are in some
way generic. Biological “switches” often follow the form of Fitzhugh-
Nagumo like dynamics. In all of these cases it is very helpful to have a
library of models that we know. Therefore many of the examples in the
chapters begin with historically important models that set a framework
to work within.
For most of the models developed in this book, we will start with
simple models and add to them as needed. This has certain specific
limitations that we should all be aware of. The main limitation that we
will acknowledge is the massive amount of mathematics that can be
brought to bear on biological questions. Topics ranging from statistics,
calculus, differential equations, machine learning, topology, abstract
algebra, graph theory, number theory, geometry, numerical analysis
and others have all been used to obtain important insight into biolog-
ical processes. There is no one book that will cover the diversity of
topics. Even so, there is no one researcher who uses all of these meth-
ods. For about the past two hundred years, methods from calculus have
been extremely effective in translating hypotheses concerning the nat-
ural world into mathematical language which can be used to analyze,
Projectile Motion 7

predict and quantify observations. We will focus almost exclusively on


differential equation models – for historical reasons they are the most
prevalent but they are also quite flexible.
There are also many, many good books with biological models.
These range from undergraduate-level to specialized graduate level
texts. It is highly recommended to explore these texts. We list a few
here but this is absolutely not a definitive list.

Undergraduate Level
1. The classic book by Edelstein-Keshet, Mathematical Models in Bi-
ology [14]. This includes discrete models and continuous ordinary
and partial differential equations models.
2. Modeling aimed primarily at biologist’s by Day and Otto, A Biol-
ogist’s Guide to Mathematical Modeling in Ecology and Evolution
[53]. A welcoming book focusing primarily on ecology an evolu-
tion.
3. A comprehensive book at the calculus level including statistics, ba-
sic programming, introductory models and differential equations:
Mathematics for the Life Sciences by Bodine, Lenhart, and Gross
[6].
4. Essential Mathematical Biology by Britton [9]. This book covers a
lot of ground with a variety of models, methods and examples.
Graduate Level
1. Mathematical Physiology by Keener an Sneyd [36] is full of de-
tailed models of an incredible number of biological topics focused
on human physiology.
2. J.D. Murray’s classic Mathematical Biology. The early, one volume
version is quite good but there are more applications and modern
models in the two volume editions [49, 50].
3. Lee Segal wrote many modeling books and was one of the origi-
nal, modern mathematical biologists see [59] and, although not re-
stricted to biology, the classic modeling with CC Lin is chock full
of methods and examples [42].
8 Introduction

Other authors who have made major contributions to pedagogy


(that is teaching, rather than research) include Linda Allen, Avner
Friedman, Stacey Smith?, Claudia Neuhauser, Frank Hoppensteadt,
Martin Novak, Suzanne Lenhart, and many more.

1.3 Problems
Problems 1.1 Show that if an object is accelerating due to gravity,
there is no maximum velocity.

Problems 1.2 Suppose an object is thrown in the air from a height of


zero. What is the maximal height that is reached if the initial velocity
is 3 ft/s?

Problems 1.3 Drag force or air resistance is one of the forces that is
neglected in simple projectile motion. One difficulty in modeling air
resistance is that direction of the force depends on the velocity. For a
projectile fired into the air, the drag force acts in the same direction as
gravity when the projectile is moving upward and in the opposite when
the object is moving downward.

(a) Show that the force F = −v2 sgn(v) is consistent with this where
sgn(v) is 1 if v > 0 or −1 if v < 0.

(b) Consider a ball dropped from rest at height, h. Find the solution to
the position as a function of time including air resistance. What is
the terminal velocity? Does the concept make sense?

(c) How fast does an object need to be traveling before air resistance
is 10% of the force of gravity?
2
Mathematical Background

This chapter provides an overview of the main tools for analytic stud-
ies. We briefly review differential equations including linear, constant
coefficient equations, separable equations, linearization, and qualita-
tive analysis.

2.1 Mathematical Preliminaries


Differential equations are relations between unknown  function, y(t),
d n y d (n−1) y dy
and its derivatives, F dt n , dt (n−1) , ..., dt , y(t), f (t) . As in algebra, one
of the goals is to determine the unknown  n (n−1) that satisfies some constraint

this relationship – for example F dt n , dt (n−1)y , ..., dy
d y d
dt , y(t), f (t) = 0. A
few examples are,
dy
− f (t) = 0,
dt
 2 2
d y d 2 y dy
+ + sin(t) = 0,
dt 2 dt 2 dt
 3 
d y dy
3
+ y = et .
dt dt
One could naively think of the goal of determining y(t) as some
sort of integration. For the first example above,
dy
− f (t) = 0,
dtZ
dy
Z
dt = f (t)dt,
dt Z
y(t) = f (t)dt.

DOI: 10.1201/9781003316930-2 9
10 Mathematical Background

Sometimes we can undo the derivatives. The solution to the equation


dy R
dt − f (t) = 0 is y(t) = f (t)dt. Recalling from calculus that until we
have specified the domain, these integrals are indefinite and only deter-
mined up to a constant. For an equation that has a highest derivative n
(referred to as an nth order differential equation), we find n integration
constants that need to be determined.
The theory of differential equations is very well developed, es-
pecially for the types of models that we will mainly be considering,
namely initial value problems. For these types of differential equations,
the integration constants are determined by providing information at
a specific time, Information is typically the value of the function and
enough derivatives to determine the constants (remember the initial po-
sition and velocity in the projectile motion section). The most widely
used alternative to this is to provide information about the unknown
function at multiple points. These equations are typically referred to as
boundary value problems and are more technically challenging.
For initial value problems, broadly speaking as long as all the func-
tions involved are well-behaved (are able to be differentiated a suitable
number of times and have no singularities), one can show that there is
a unique solution to the equation. We will focus exclusively on these
and for the most part initial value problems have unique solutions as
long as we prescribe initial values. We can then turn to finding these
solutions. To do this, it is often useful to classify initial value problems
since certain methods are only useful for certain classes of equations.
The reasons to cover analytic methods (that is methods that you can
do on paper and completely understand) are two-fold. First, our analy-
sis will lead to insight reality-checks for our models. Second, since we
will be doing a lot of of numerical simulations, having certain behav-
iors in hand will help us interpret our simulations.
Differential equations can be differentiated into two broad cate-
gories – linear or nonlinear depending on whether the relationship
(n) (n−1)
F( ddt (n)y , ddt (n−1)y , ..., dy
dt , y(t), f (t)) is linear, with no terms with products
of terms containing y(t), or not. Solution techniques and questions
about uniqueness of solutions depend a lot on the classification.
Mathematical Preliminaries 11

2.1.1 Linear
Linear initial value problems can be written as,
dny d (n−1) y dy
an (t) n
+ a n−1 (t) + ... + a1 (t) + a0 (t)y(t) = f (t)
dt dt (n−1) dt
y(0) = y0
dy
(0) = y1
dt
..
.
d n−1 y
(0) = yn−1
dt n−1
If the right-hand-side is zero, this is a homogenous equation and
y(t) = 0 is a candidate solution, depending on the initial conditions
provided.
There are many techniques for analyzing these, but we will focus
on one more restrictions. If the coefficients are constant, there is a com-
pletely algorithmic way to understand the solution. For a concrete ex-
ample, we can consider a second order example,
d2y dy
a 2
+ b + cy = 0 (2.1)
dt dt
y(0) = y0
dy
(0) = y1 .
dt
We will come up with a solution by noticing that we already know
a function whose derivatives are related to that function. This is the
definition of the exponential, eλt . So we can guess this as a solution
and substitute this into Equation 2.2. This gives us, aλ 2 eλt + bλ eλt +
ceλt = 0. Since the exponential is never zero, we can simplify this to
find the characteristic equation,
aλ 2 + bλ + c = 0.
This is an algebraic equation for λ , so we know that there are two
values of λ that satisfy the equation,

−b ± b2 − 4ac
λ± = .
2a
12 Mathematical Background

These are referred to as eigenvalues, which is a term that may be


familiar if you have had linear algebra. The most important thing to
notice at this point is that y(t) = Aeλt is a solution for each of these
values of λ and for any value of A. Linear combinations of both of
the solutions leads to a solution since the differential equation is linear.
In general any solution to the differential Equation 2.2 can be written
y(t) = Aeλ+t + Beλ−t . The initial conditions determine A and B.
There are certain things to be aware of. First, what if the values
of λ± are complex? How do we understand eλ+t in this case? What
happens if b2 − 4ac = 0? For a short version that is applicable to this
material, we note that if the eigenvalues are real and distinct, the so-
lutions are exponential and the solutions either exponential increase
or decrease to zero. If the eigenvalues are complex conjugate pairs,
λ± = R ± iI the solutions can be written as combinations of eRt sin(It)
and eRt cos(It). These solutions oscillate with amplitudes that grow or
decay according to whether R is positive, negative or zero.
There is also the case where the eigenvalues are real and repeated.
In a differential equations course you learn methods to deal with this.
In this book it is less useful to address this. This is what is often re-
ferred to as a non-generic case. That is, it requires a precise relation-
ship in the parameters. The point of view that comes from sensitivity
and as a practical applied mathematician is that parameters are not ab-
solute, fixed values. They may vary between individual experiments,
between individual species, between times of the year or other small
differences. As such, it tends to be less important to understand things
that require exact values of parameters and we focus on things that
occur in wider parameter spaces.
One more useful piece of information is that any higher order dif-
ferential equation can be written as a system of lower-order differen-
tial equations. If we define a new variable, u = dy dt , we can see that
d2y
du
dt = = − ba dy
dt 2
c
dt − a y. Equation 2.2 can be written differently,
dy
= u
dt
du b dy c b c
= − − y = − w − y.
dt a dt a a a
This can be written very succinctly as
dy
= My, (2.2)
dt
Mathematical Preliminaries 13

where,
 
0 1
M=
− a − ba
c

We define y =< y(t), u(t) >.


It turns out that the eigenvalues of this matrix are equivalent to the
roots of the characteristic Equation 2.2. We will see that our numerical
methods are typically written in this form and, in fact, many of the
models come in this form.
That works well for a very restricted class of equations – lin-
ear, constant coefficient, homogenous initial value problems. However,
most processes in nature are nonlinear. This means that we have to have
some way to analyze nonlinear equations. To do this, we will introduce
two analytic methods that will be supplemented with direct numerical
simulations. The first method illustrates one of the most useful insights
in mathematics, namely if we look close enough, most nonlinear pro-
cesses can be approximated by linear processes. This is used in all ar-
eas of mathematics including topology, algebra, differential equations,
etc. Any student has seen this in calculus where nonlinear functions are
approximated by linear functions using Taylors’ theorem. We can do
this for differential equations. The second tool is to look at the qual-
itative behavior of solutions by considering how the derivatives of y
control the rate of change of aspects of the graph of y. We will do this
in the context of first order equations in this chapter, but in subsequent
chapters we will explore the same processes for higher order systems.

2.1.2 Nonlinear Equations


Nonlinear equations are much more difficult to solve analytically.
There are several classes that are generally solvable and are focused
on in a course in differential equations.
It is typical to write the implicit relationship between y(t) and
n (n−1)
its derivatives, F( ddt ny , ddt (n−1)y , ..., dy
dt , y(t), f (t)) = 0, in terms of the
highest derivative,
!
dny d (n−1) y dy
= G , ..., , y(t),t .
dt n dt (n−1) dt
dny
Where G encodes all of the steps needed to isolate dt n .
14 Mathematical Background

To be very clear, we can start with first-order equations which can


be written as,
dy
= G(y,t). (2.3)
dt
These cannot be solved by integrating both sides since the unknown
y occurs on both sides. There are some cases where we can almost do
this though. If G(y,t) = G1 (y)G2 (t) we can rewrite the equation,
dy
= G(y,t) = G1 (y)G2 (t),
dt
G1 (y)dy = G2 (t)dt.
This is a slight abuse of notation but can be made formally cor-
rect. Then integrating both sides with respect to the arguments gives
an implicit solution,
Z Z
G1 (y)dy = G2 (t)dt.

Separation of Variables
Consider the differential equation,
dy t
= ,
dt y
y(0) = 1.

We can solve this using separation of variables:


dy t
= ,
dt y
ydy = tdt,
Z Z
ydy = tdt,
y2 = t 2 + c,

which gives an implicit solution for y.


The initial condition, y(0) = 1, implies
√ that c = 1 and we can
write an explicit solution y(t) = ± t 2 + 1.
Linearization 15

It should be noted, however, that this does not work in general.


Even when it does work, the solution is often only written implicitly
and may be quite complicated and unwieldy.

2.2 Linearization
Linearization is one of the major concepts in mathematics. Almost all
branches use linearization to approximate the behavior of nonlinear
processes. Applied mathematics relies on linearization to a wide ex-
tent. We will begin with a brief review from calculus since this is often
the first place we see the formal idea of linearization.
Start with a function of an independent variable t, say f (t). Taylors’
theorem states that under some conditions on f , we can write f (t) as
a polynomial, P(t) = ∑i αit i . Moreover, there is an interval about any
point in the domain of f (where f obeys certain restrictions), where
f (t) = P(t). This is a remarkable result that provides insight into how
to integrate and differentiate a range of functions, since we know that
integration and differentiation of polynomials has a predictable pat-
terns. But how do we determine P(t)? Taylors’ theorem states that,
near any point, t, we can write f (t) as,
d2 f
df 2 (a)
f (t) = f (a) + (a)(t − a) + dt (t − a)2 +
dt 2!
dn f
n (a)
... + dt (t − a)n ) + error.
n!
Therefore the coefficients of the polynomials are related to the deriva-
tives of f . It also means that as long as t is close to t, (t − a)i is very
small. So we could approximate f ,
df
f (t) ≈ f (a) + (a)(t − a),
dt
where the error term can usually be estimated using the mean value
theorem. Graphically, this means that near any point, a nice enough
function can be thought of as a line.
16 Mathematical Background

This idea can be used to approximate nonlinear differential equa-


tions by linear equations. To introduce the topic, we will start with the
simplest case of a scalar equation of the form,
dy
= f (y), (2.4)
dt
y(0) = y0 ,

although this idea can be generalized to other forms of differential


equations. One of the most important restrictions that we are requir-
ing is that the right-hand-side cannot depend explicitly on time. These
equations are referred to as autonomous equations and are the main
focus of this text. There are other methods that are used for non-
autonomous equations and can be found in textbooks on differential
equations (for example [7]).
In this case, the goal is to determine the behavior of the solution.
In calculus, we need a specific place to linearize near. Linearization is
inherently a local argument and cannot in general be used everywhere.
In calculus we linearize at a point. In differential equations, we lin-
earize around a known solution. This seems counter-intuitive at first,
since the goal was to find a solution in the first place. However, it is of-
ten simple to find some special solutions to differential equations. For
example, we can look for solutions that do not depend on time. These
are steady-state solutions and are constants that satisfy the differential
equation. To find steady-state solutions, ȳ, we have to solve,
f (ȳ) = 0.

There may be no steady-states, in which case we cannot proceed


with the linearization and we have to do something else such as direct
numerical simulations. Otherwise, we have at least one solution, ȳ. We
then want to find out how the solution behaves near this steady-state.
We define the solution we are studying as,
y(t) = ȳ + εY (t). (2.5)

The function Y (t) is close to ȳ as long as εY (t) is “small”. We will not


be perfectly precise here about how small is small enough but the idea
is that understanding the dynamics of Y (t) can provide information
about y(t).
Linearization 17

To derive an equation for Y (t), we put the solution in Equation 2.5


into Equation 2.4,
dy
= f (y),
dt
d ȳ + εY (t)
= f (ȳ + εY (t)),
dt
dY (t)
ε = f (ȳ + εY (t)),
dt
= f (ȳ) + ε f 0 (ȳ)Y (t) + error.

We have used the fact that ȳ does not depend on time to simplify the
left-hand-side and Taylors’ theorem to approximate f (y) with a linear
approximation near ȳ. Notice that this provides an equation for Y (t),
dY (t)
= f 0 (ȳ)Y (t). (2.6)
dt
Since ȳ is constant and this equation is linear, we know exactly how
0
Y (t) behaves since the solution is Y (t) = ke f (ȳ)t . If f 0 (ȳ) > 0, Y (t)
increases and y(t) moves away from ȳ. On the other hand, if f 0 (ȳ) <=,
Y (t) decreases and the solution moves towards ȳ. This says something
specific about the steady-state solution, ȳ. If f 0 (ȳ) > 0, we refer to
the steady-state as unstable. If f 0 (ȳ) < 0, ȳ is stable. If f 0 (ȳ) = 0, the
linearization fails and we have to use different arguments to understand
the behavior.
We can generalize this idea. If we have a system of n nonlinear
differential equations,
dy1
= f1 (y1 , y2 , ..., yn ),
dt
dy2
= f2 (y1 , y2 , ..., yn ),
dt
... ...
dyn
= fn (y1 , y2 , ..., yn )
dt
we define the steady-state as a vector ȳ = (ȳ1 , ȳ2 , ..., ȳn ). We look near
ȳ,

y = ȳ + εY(t),
18 Mathematical Background

by inserting this into the differential equations to find,


dY
= JY, (2.7)
dt
We use the notation J since the matrix we have obtained is referred to
as the Jacobian. The Jacobian is a useful concept, and for us one of the
uses is to shorten some of the calculations. The Jacobian,
 
∂f ∂ f1 ∂ f1
 ∂ x11 ∂ x2 ... xn 
 
 
 ∂ f2 ∂ f2
... ∂ f2 
 ∂ x1 ∂ x2 xn 
J=
 .
 (2.8)
 .. .. ... .. 
 . . 

 
 
∂ fn ∂ fn ∂ fn
∂ x1 ∂ x2 ... xn

We show the details for planar systems in the appendix.

2.3 Qualitative Analysis


The last thing that we will review here is a method for understanding
the broad behavior of the solutions of differential equations. This does
not provide useful information for quantifiable predictions. However,
it is quite useful for modeling as it can give insight into the general
behavior of solutions to models. This often helps diagnose issues with
the model that prevent reasonable physical interpretation. Just as lin-
earization does not provide all information (but restricted to starting
near a known solution), neither does qualitative analysis.
We will see that qualitative analysis is almost always restricted to
scalar or planar systems and it is useful to look at each one separately.
We start with scalar equations,
dy
= f (y).
dt
We will again restrict our discussion to autonomous equations and
note that we are not imposing initial conditions. Qualitative analysis
provides information for the behavior for all initial conditions.
Qualitative Analysis 19

Figure 2.1: Phase line showing how f (y) determines whether y increases or
decreases and the qualitative sketch of the solutions, y(t) for different initial
conditions.

We can consider the graph of the function f (y) (see Figure 2.1).
Where f crosses the y-axis, f (y) = 0 which means dy dt is zero there.
That is all roots of f (y) are steady-state solutions. As long as f is
nice enough (continuous and differentiable), if the sign of f changes
between y1 and y2 , there is a root. That means that dy
dt has to be of one
sign between the roots of f . That means that the solution y is either
increasing or decreasing on intervals that do not contain a root. We can
sketch the direction that the solution moves on the “phase-line” (see
Figure 2.1). This can be used to sketch the solution of the equation as
a function of time.
The argument for planar curves is similar but a bit more involved.
Consider the system of equations,
dx
= f (x, y),
dt
dy
= g(x, y).
dt
20 Mathematical Background

Figure 2.2: Phase plane showing the x and y-nullclines. The direction of
increase/decrease in the solution x and y are shown. The same phase plane is
shown with a sketch of the trajectories.

The curves f (x, y) and g(x, y) can be drawn in the (x, y)-plane.
Above f , x must be increasing. While below f , x is decreasing. Simi-
larly, the evolution of y depends on which side of g we are. We can also
see that the solution curve must be tangent to the vector field ( f , g). So
it is possible to see how trajectories move in the phase-plane – where
we think of the solution as a parameterized curve (x(t), y(t)) (see Fig-
ure 2.2).

2.4 Problems
Problems 2.1 Classify the following differential equations as linear
or nonlinear. Also indicate the order of the differential equation
dy
(a) dt + 2y = sin(t)
2
   2 
(b) t ddt 2y + dy
dt = 1
3
(1+t )
3t
− (1+t 2) y

 4
d2y dy
(c) − dt 2 dt =4
Problems 21
5
(d) 4 ddt 5y + cos(t) = 0
3
(e) y ddt 3y − t 2 dy
dt + y = 0
2
(f) t 5 ddt 2y + t 2 dy
dt + y = sin(t)

Problems 2.2 A relation or operator, F(x) is linear if two properties


hold: F(x + y) = F(x) + F(y) and F(cx) = xF(x).
(a) Show that the derivative operator is linear.

(b) Suppose that y1 (t) and y2 (t) both solve the equation,

d2y dy
2
+ sin(t) + cos(t)y = 10,
dt dt
dy
(0) = 1,
dt
y(0) = 1.

Show that y1 (t) + y2 (t) is also a solution.

(c) Show that a linear combination of any two solutions to the differ-
ential equation,

d 2 y dy
+ y2 = 0,
dt 2 dt
dy
(0) = 10,
dt
y(0) = 10.

is not a solution to the equation. (Hint: There is no reason to try


and solve this equation but see if y = ay1 + by2 is a solution if y1
and y2 are.)

Problems 2.3 Euler’s formula states that eix = cos(x) + i sin(x).


(a) Use Euler’s identity to show that eR±iI = eR cos(I) ± ieR sin(t)

(b) Find two linear combinations of y1 (t) = eR cos(I) + ieR sin(t) and
y2 (t) = eR cos(I) − ieR sin(t) that involve either cos(t) or sin(t).
22 Mathematical Background

Problems 2.4 The following steps show how to relate the solution to
a second order, linear, constant coefficient differential equation with
complex eigenvalues to real-valued solutions.
2
Consider the differential equation ddt 2y + 4 dy
dt + 5y = 0.

(a) Find the eigenvalues.


(b) Use Euler’s formula to write the two solutions in terms of cos(αt)
and sin(αt).
(c) Show that the real and complex part of the solutions are themselves
solutions.
Problems 2.5 The eigenvalues of a 2×2 matrix,
 
a b
A=
b d
Can be found by finding the roots of the determinant of A − λ I,
where I is the 2×2 identity matrix,
(a) Find the eigenvalues of
 
2 1
A=
−3 1

(b) Show that the eigenvalues of the matrix associated with Equation
2.2 are the same as the roots of the characteristic polynomial of
d2y
dt 2
= − ba dy c
dt − a y.

Problems 2.6 Use separation of variables to solve,


(a)
dy
= y(y + 2),
dt
y(0) = 1,

(b)
dy
= y(t + 3),
dt
y(0) = 1,
Appendix: Planar Example 23

Problems 2.7 (a) Sketch the graph of a nonlinear function f (x), along
with the linearization at a point, x0 .
(b) Use the sketch to show how the linearization provides an approxi-
mation of the function value at x0 + δ x.
Problems 2.8 Find and classify the steady-states of the following dif-
ferential equations:
dy
(a) dt = y(1 − y)
dy
(b) dt = −1 + x2
dy
(c) dt = sin(x).
Problems 2.9 Suppose we have the planar system of differential equa-
tions,

dx
= x − xy,
dt
dy
= −y + xy.
dt
(a) Find all equilibiria.
(b) Find the Jacobian at the equilibria.
(c) Classify the equilibria.

2.5 Appendix: Planar Example


Many of the examples in this book are planar systems where lineariza-
tion is very well described. We will briefly cover this and show some
of the linear algebra details in this context.
Consider a planar system of equations,
dx
= f (x, y),
dt
dy
= g(x, y). (2.9)
dt
24 Mathematical Background

We assume that we have found a steady-state, (x̄, ȳ) so that


f (x̄, ȳ) = g(x̄, ȳ) = 0. We consider the behavior near this solution,

x(t) = x̄ + εX(t),
y(t) = ȳ + εY (t).

Plugging this into the differential equations, using Taylors’ theo-


rem for multivariable functions and dropping the nonlinear terms, we
find,

dx
= f (x, y),
dt
d x̄ + εX(t)
= f (x̄ + εX(t), ȳ + εY (t)),
dt
dX(t) ∂ f (x̄, ȳ) ∂ f (x̄, ȳ)
ε = f (x̄, ȳ) + ε X +ε Y,
dt ∂x ∂y
dX(t) ∂ f (x̄, ȳ) ∂ f (x̄, ȳ)
= X+ Y,
dt ∂x ∂y
for the x-component. Similarly,
d ȳ + εY (t)
= g(x̄ + εX(t), ȳ + εY (t)),
dt
dY (t) ∂ g(x̄, ȳ) ∂ g(x̄, ȳ)
ε = g(x̄, ȳ) + ε X +ε Y,
dt ∂x ∂y
dY (t) ∂ g(x̄, ȳ) ∂ g(x̄, ȳ)
= X+ Y,
dt ∂x ∂y
for the y-component.
The Jacobian matrix is,

" #
∂ f (x̄,ȳ) ∂ g(x̄,ȳ)
J= ∂x ∂x .
∂ g(x̄,ȳ) ∂ g(x̄,ȳ)
∂x ∂y

For short-hand, we can write this


 
a b
J= .
c d
Appendix: Planar Example 25

To find the eigenvalues of J we find the roots of the determinant of


J − λ I,
 
a−λ b
= 0,
c d −λ
(a − λ )(d − λ ) − bc = 0,
2
λ − (a + d)λ + ad − bc = 0,
λ 2 − TrJλ + DetJ = 0.

Where TrJ is the trace of J (the sum of the diagonals) and DetJ is
the determinant of J. Therefore, the eigenvalues of the Jacobian are,

TrJ ± TrJ 2 − 4DetJ
λ +± = .
2
The components of the Jacobian completely determine whether the
real part of the eigenvalues is positive, negative or zero.
3
Introduction to the Numerical Methods

This chapter provides an overview of the numerical platforms that we


use throughout the book. We provide some motivation for the two that
we have chosen and describe how to start using the programs.

3.1 Introduction
The models that we will use in this book are primarily differential
equations. This presents a challenge: How do we understand the mod-
els, use them to make useful predictions or provide insight into the bi-
ology without several additional years of mathematical training? Typ-
ical undergraduate sequences in calculus, linear algebra, differential
equations, and scientific computing are useful for understanding the
nuances of models. Nonlinearities often put models outside the reach
of an undergraduate differential equations class while scientific com-
puting or numerical methods classes often focus on the underlying
methods, rather than the applications.
In this chapter, we discuss the two main computer programs/lan-
guages that we will use throughout the course. The book will inten-
tionally provide parallel numerical scripts in Python and MATLAB® .
Both of these are able to visualize solutions, numerically approximate
solutions to a range of equations, and can be used to automate tasks,
like sampling, which will be important throughout the book. So why
these two packages? Years ago students would have to learn Fortran
to be able to solve equations numerically – adding several semesters
of course work. Both of these languages are widely used and it is our
philosophy that students should have some exposure to all of these lan-
guages and then focus on the one that is most comfortable. There are
many other languages/platforms that are in use. Languages based on

DOI: 10.1201/9781003316930-3 27
28 Introduction to the Numerical Methods

C-like languages including R are widely used. There are open MPI lan-
guages that are able to call both C++ and Fortran libraries such as Julia.
This book is relatively agnostic about the language of choice. MAT-
LAB is well developed and has an excellent support system. Python is
very flexible and has many workgroups that are designed for helping.
They each have some drawbacks as well. MATLAB has a reputation as
a “slow” language because it is interpreted. However, there are many
ways to speed things up with more experience. Python has a specific
culture that can sometimes be overly complicated and difficult to fol-
low since it is so flexible and does not have a centralized structure. R
is a relatively complex language to start with and has a steep learning
curve. All of these platforms will run into mathematical problems that
take an unreasonable amount of time to complete. There should be no
examples in this book where this will happen but keep in mind there is
no perfect language and many, many people studying how to run more
and more complicated problems. The recommendation here is to work
with codes in both MATLAB and Python. Eventually, each person will
find what is most comfortable.
For those that have never done any programming it might seem
strange to use the word “comfortable” – aren’t all language uncomfort-
able? This is definitely true at the beginning. After a bit of work, there
are nuances that make certain languages/programs easier for different
people. For example, MATLAB m-files do not read indentation. So
you can type commands wherever you want. Python is indent-aware
and will throw an error (i.e. break and not run) if the indents do not
match. This means that the typical Python files look neater and more
organized than a typical working m-file. At the same time, MATLAB
is more self-contained and does not need any extra libraries to be read
in so an m-file written for 2018 MATLAB will work on any machine
with 2018 MATLAB. In Python and R, libraries have to be loaded to
“teach” the computers Python what certain commands mean. This can
occasionally cause issues when sharing files.
All of these are accessible and we will start with essentially no
prior knowledge. This chapter is to provide a basic outline of how to
access the platform (MATLAB or Python), how to write a program,
how to edit a program, and how to run a program. We will consider a
few best practices in writing numerical codes in general and in each
platform specifically.
Best Practices in Coding 29

One important note about the codes provided in this book. For
many chapters we have opted for simpler, less efficient, “home-built”
code rather than implementing large scale packages. There are two rea-
sons for this. First, it is much easier to get used to writing and editing
scripts with simpler examples – not to mention developing a better
understanding of the actual methods used. Second, with all packages
there is an issue of upkeep and verification. One cannot assume that
the provided code does not have bugs or errors considering companies
such as Apple, Google, and Microsoft constantly develop patches to
fix codes that cost millions of person hours to develop. Similarly, codes
developed for specific projects are not always updates which means it
is possible to lose functionality permanently if you don’t know how
it was written. Some exceptions here are in the later chapters where
the implementation is beyond what is expected by readers. The book
as chosen a few, well-known, reasonably well-supported implementa-
tions; however these are by no means the only methods available. The
reader should feel free to explore decide what is the cutting-edge im-
plementation.

3.2 Best Practices in Coding


Writing codes is very much like writing in any language. You would
like the writing to be clear, concise, accurate, and readable. One ex-
tra consideration is the user interaction with most of these programs.
When writing a poem, it is unusual for the author to expect that a
reader will add to it or change specific lines or words. Good coding
expects this and invites this. So we have to make sure that the files are
organized without extraneous parts or things that don’t work. In the
codes developed in this book, we have made a conscious choice to in-
clude some commands that are not the most efficient to ensure they are
clear.

3.2.1 Folder Structure


If you open your computer and the desktop is full of files, images, and
assorted notes, I would encourage you to start learning file structures.
In this book, there will be many different files used. Some will be used
30 Introduction to the Numerical Methods

for many applications, some will only be used for a single model. Your
life will be immeasurably easier if you learn to sort these so you can
find them easily and not accidentally delete or overwrite files you are
working on. The simplest structure that has some redundancy, but is
ordered in a sensible manner might be a folder for the specific course.
Within that folder, you have folders for each chapter. Within each chap-
ter folder you have folders for MATLAB and Python and an assign-
ment folder. Within each of the numerical folders you have folders for
tasks in the assignment (for example Question 1). There is a high like-
lihood of redundancy with this since Question 1 might use the same
script as Question 2. However, there is much less chance that you ac-
cidentally over-write your code answering Question 1 when you start
working on Question 2. Typically, for long-lived projects, there are
specific times when the file structure is revised. But this should be done
as a completely separate task than building the files. It is our expecta-
tion that most students using this book will not need to revise their file
structure.
One disadvantage to this is that when running scripts, there is the
need to change folders. Fortunately in all of our examples, the play
button will query this and automatically change folders for you. But it
is something to be aware of when sharing codes with other people.

3.2.2 Naming Conventions


All names should be explanatory. This includes folder names, file
names, variable names, function names, and code-block names. If you
have a file in a folder called temp1, it might be junk or it might be
the final that you never got around to moving to the correct location.
Likewise, a file called stuff.py requires a person to open it to see
what it is. It is very confusing if you open a file that is supposed to
plot a function f (x,t) and x is never used and there is some param-
eter called g that has a value that is never explained. All of these
names should be understandable with very little reader orientation.
Consider variable names like gamma for a parameter γ. File names
like pred_prey_chapter_1 are easy to understand.

3.2.3 Code Structure


All codes consist of blocks that are typically identified by some com-
menting convention. In MATLAB it might look like,
Discovering Diverse Content Through
Random Scribd Documents
Egypt at least suggest that devotional language as well as proverbs
(cf. I Kings 4:30) were common knowledge among the two peoples.

The two compositions are basically different, however, in that the


Biblical psalmist acknowledges Yahweh, the God of creation and
providence, as a spiritual being associated with natural phenomena
only as their creator, whereas Aton, Akhenaton’s “sole god” is
identified with the disk of the sun. While Akhenaton seems to have
spiritualized the Egyptian sun worship, he never divorced himself
completely from it. Biblical monotheism asserts that the One God
made “lights in the firmament of heaven to give light upon the earth”
(Gen. 1:15). The Hymn of the Aton reaches a high point in the
devotional literature of Egypt, but its monotheism was radically
different from that presented on the pages of Scripture.

44
VI
THE AFFAIRS OF EMPIRE

The Amarna tablets enable us to see evidences of the decline of


Egyptian power and prestige during the latter years of the reign of
Amenhotep III and throughout the reign of Amenhotep IV
(Akhenaton). About forty of them record correspondence between
the rulers of Egypt and the rulers of the major powers of the Amarna
Age. We find letters from the Kassite kings Kadashman-Enlil I and
Bumaburiash II of Babylon, from Ashuruballit I of Assyria, from
Tushratta of Mitanni, from the Hittite king Suppiluliumas, and from an
unnamed king of Alashia (Cyprus).

The Kings of Mitanni

In upper Mesopotamia, Egypt had an ally in Mitanni, a kingdom


comprised largely of Hurrians (Biblical Horites), with an Indo-Aryan
ruling class. The Mitannian kingdom was established about 1500 B.C.
and at the height of its power reached from Nuzi and Arrapkha in
Assyria to Alalakh in Syria. Its capital, Wassukkanni, was on the
upper Habur River. A major threat both to Mitanni and, ultimately, to
Egypt, came from the rising Hittite Empire with its capital at Hatusa
(modern Bogazkoy) on the great bend of the Halys River in Asia
Minor. Babylon had suffered an eclipse since the empire of
Hammurabi and during the Amarna age it was ruled by a mountain
people known as Kassites. Assyria, north of Babylon, had been
subject to Mitanni until the Hittite conquest of Mitanni gave the
Assyrians an opportunity to free themselves and develop an
independent state.
The kings of Mitanni sent daughters to grace the harems of the
Pharaohs of Egypt, and desired gold in exchange. The Amarna tablets
include seven letters from Tushratta of Mitanni to Amenhotep 45
III, one to his widow Queen Tiy, and three to Amenhotep IV
(Akhenaton). In a typical communication he writes to Amenhotep III:

Let my brother send gold in very great quantity, without measure—


that is what my brother should send me—and let my brother send
more gold than (he sent) to my father, for in my brother’s land gold
is as plentiful as dust. May the gods so direct that, although now
gold is so plentiful in my brother’s land, he may have gold ten
times more plentiful than now. Let not the gold which I desire
trouble the heart of my brother, and let not my brother grieve my
heart. So let my brother send me gold, without measure, in very
[28]
great quantity.

Although Tushratta’s lust for gold may not have been appreciated in
Thebes, Egypt valued her Mitannian allies who served as a check on
the ambitious Hittites, thereby helping Egypt maintain control over
Syria and Palestine. Not only were princesses from Mitanni welcome
in Egypt, but the Mitannian gods might be of help to the Egyptians.
During the illness of Amenhotep III, Tushratta sent a statue of the
goddess Ishtar from Nineveh to bless the ailing Pharaoh:

Thus saith Ishtar, mistress of all the lands, “To Egypt, to the land
which I love I will go; I will return.” Verily I have now sent her, and
she is gone. Indeed in the time of my father the mistress (Ishtar)
went to that land, and inasmuch as she was revered when she
formerly resided there, so now may my brother tenfold more than
formerly honor her. May my brother honor her and joyfully send
her back, and may she return. May Ishtar, mistress of heaven,
protect my brother and me one hundred thousand years, and may
[29]
our lady give to us both great joy.

Tushratta looked to Egypt for help against Suppiluliumas and the


emerging Hittite Empire, but neither gold nor troops came.
Akhenaton was singing the praises of Aton in Akhetaton when the
Hittites succeeded in entering and sacking the Mitannian capital (ca.
1370 B.C.) and Tushratta was slain by one of his own sons. The
former king’s exiled brother and rival, Artatama, seems to have
seized control in the confusion that followed. Tushratta’s son
Mattiwaza (who may have been his father’s murderer) fled to the
Hittites, and Shutarna, son of Artatama, sought the friendship of the
Assyrians in his bid for the throne.

Emerging Assyria

The Mitannian Empire was at an end. Mattiwaza gained Hittite


support and took the throne of Mitanni with the aid of 46
Suppiluliumas’ army. He remained a vassal of the Hittites,
however, and his marriage to a daughter of Suppiluliumas further
strengthened Hittite power in Mitanni. At the same time Ashuruballit
of Assyria took advantage of the situation to seize the portion of
Mitanni nearest to him. Assyria had been controlled alternately by
Mitanni from the west and by Babylon from the south, but
Ashuruballit determined to head an independent state.

To strengthen his position, Ashuruballit sent messengers to


Akhenaton with a present of two white horses and a silver chariot.
[30]
The letter accompanying these gifts asks nothing in return, but a
[31]
second letter mentions a new palace that Ashuruballit is building,
with the suggestion, “If thou art very friendly disposed, then send
much gold.”

Evidently the Assyrian embassy was received with courtesy in the


court of Akhenaton, for Burnaburiash of Babylon was unhappy at the
thought that Egypt would deal with a people who had been subject
to Babylon. In anger he dispatched a letter:
To Niphururia (i.e., Akhenaton), king of Egypt, say. Thus says
Burnaburiash, king of Karduniash (i.e., Babylon), brother. I am well.
May it be well with you, your house, your wives, your sons, your
land, your chief men, your horses, your chariots. Since my fathers
and your fathers established friendly relations with one another
they sent rich presents to one another, and they refused not any
good request one of the other. Now my brother has sent (only) two
minas of gold as a present. But now, if gold is plentiful send me as
much as your fathers, but if it is scarce send half of what your
fathers did. Why did you send (only) two minas of gold. Now, since
my work on the House of God is great, and I have undertaken its
accomplishment vigorously, send much gold. And you, whatsover
you desire from my land, write and it shall be brought to you. In
the time of Kurigalzu, my father, the Canaanites as one man wrote
to him saying, “We will go against the border of the land (i.e.,
Egypt), and we will stage an invasion, and we will make an alliance
with you.” But my father wrote to them saying, “Cease speaking of
an alliance with me. If you are hostile against the king of Egypt,
my brother, and ally yourself with another, I will come and will
plunder you for he is in alliance with me.” My father did not listen
to them for the sake of your father. Now, concerning the Assyrians,
my subjects, have I not written to you? If you love me you will not
do business with them. Let them accomplish nothing. As a present
I have sent to you three minas of beautiful lapis-lazuli, and five
[32]
span of horses for five wooden chariots.

47
The Hittite Challenge

The Egyptians, however, did not give serious attention to their Asian
Empire until the reigns of Seti I (1318-1299 B.C.) and Rameses II
(1299-1232 B.C.). During the half century following the fall of Mitanni
the Hittites met no serious opposition in their desire to control the
whole of northern Mesopotamia and the Mediterranean coastal region
of Syria. Not only was a daughter of Suppiluliumas married to the
ruler of the vassal kingdom of Mitanni, but a son, Telepinus, was
installed as ruler of Aleppo in Syria. The Amorites who had been
vassals to Egypt were glad to welcome Hittite aid in establishing their
independence. North Syria, or Amurru, became a hotbed of anti-
Egyptian feeling, and Amorite princes served as unconscious pawns
of Suppiluliumas in weakening Egyptian control in the area and
preparing the way for Hittite domination.

Most of the Amarna letters were written by princes of city states in


Syria, Phoenicia, and Canaan who acknowledged sovereignty but
expressed the fear that rival princes or alien peoples might gain the
upper hand. Some of the letters represent factional disputes among
leaders both of whom protest their loyalty to Egypt. Among the rulers
in Syria and Palestine we find letters from Akizzi of Qatna, Abdi-
Ashirta and his son Aziru of Amurru (i.e., the Amorites), Rib-Addi of
Byblos, Ammunira of Beirut, Zimrida of Sidon, Abimilki of Tyre, and
Abdi-Khepa of Jerusalem. Other letters were sent from Hazor, Akko,
Megiddo, Gezer, Ashelon, and Lachish. In some instances minor
Egyptian officials wrote the letters.

Rib-Addi of Byblos

Rib-Addi of Byblos was threatened by Abdi-Ashirta and his son Aziru,


Amorite rulers who followed the Hittite “party line” in seeking to
remove northern Syria from the Egyptian sphere of influence. Gebal,
or Gubla, was the ancient name of Byblos, a city which carried on
commerce with Egypt as early as 3000 B.C. Excavations have
produced a cylinder from the Thinite period of Egyptian history when
the earliest Pharaohs occupied the throne. Vases discovered at Byblos
bear the names of ancient Pharaohs including Mycerinus (twenty-
seventh century B.C.), Unis and Pepi (twenty-fourth century B.C.).
From the port of Byblos, Egypt imported cedars and spruce for use in
building ships. Ships of Byblos also carried jars of oil, spices, wine,
and leather.
Although Rib-Addi wished to continue to serve as a loyal vassal 48
of Egypt, the chaos of the political situation pressed heavily
upon him. The Amarna collection includes fifty-three letters which he
addressed to Amenhotep II and Akhenaton warning them of the
difficulty of the situation. In one of them he notes:

Abdi-Ashirta has written to the warriors, “Assemble yourselves in


the house of Ninib, and we will fall upon Byblos.”... Thus have they
formed a conspiracy with one another, and thus have I great fear
that there is no man to rescue me out of their hand. Like birds that
lie in a net, so am I in Byblos. Why do you hold yourself back in
respect to your land? Behold, thus I have written to the palace,
and you have paid no attention to my word.... What shall I do in
[33]
my solitude? Behold, thus I ask day and night.

Abdi-Ashirta and his son Aziru also wrote letters to Egypt affirming
their loyalty. Two of them were addressed by Aziru to Dudu, an
officer of the Egyptian court with a Semitic name. Evidently Dudu
occupied a place in the court of Akhenaton comparable to that which
Joseph occupied some years before (Gen. 41:37-57). Aziru wrote:

To Dudu, my lord, my father.

So says Aziru, your son your servant:

At the feet of my father I will fall down. May my father be well.


Dudu, behold I have performed the wish of the king, my lord, and
whatever may be the wish of the king, my lord, let him write and I
will perform it. Further: Behold, you are there, my father, and
whatever is the wish of Dudu, my father, write, and I will indeed
perform it. Behold you are my father and my lord, and I am your
son. The lands of Amurru are your lands, and my house is your
house, and all your wish, write, and indeed I will perform your
wish. And indeed you are sitting before the king, my lord. Enemies
have spoken slanderous words to my father before the king, my
lord. But do not admit them. May you sit before the king, my lord,
when I arise, and do not admit against me slanderous words. I am
a servant of the king, my lord, and I will never depart from the
words of the king, my lord, and from the words of Dudu my father.
But if the king, my lord, does not love me but hates me, what shall
[34]
I then say?

There can be no doubt that Rib-Addi, and not Aziru, was truly loyal to
Egypt, for Aziru had made an alliance with Egypt’s enemies, the
Hittites. Writing to the Egyptian Pharaoh, Rib-Addi complained:

I have written to the king, my lord: “They have taken all my cities.
The son of Abdi-Ashirta is their lord. Byblos is the only city 49
that belongs to me.” And have I not sent my messenger to
the king, my lord? But you did not send soldiers, and you did not
permit my messenger to leave.... Send him with auxiliaries. If the
king hates his city, then I will abandon it, and if he permits me, an
old man, to depart, then send your man that he may protect it.
Why was nothing given to me from the palace? I have heard of the
Hati people (i.e. the Hittites) that they burn the lands with fire. I
have repeatedly written but no answer has come to me. All lands
of the king, my lord, are conquered, and my lord holds himself
back from them, and behold, now they bring soldiers from the Hati
[35]
lands to conquer Byblos. Therefore care for thy city.

Rib-Addi stood alone in defending Byblos, but the force of the enemy
proved too much. He fled to Beirut where he found refuge in the
palace of its prince, Ammunira, with whom he was related by
marriage. Byblos fell, but Rib-Addi still hoped for aid from Egypt to
win it back. From Beirut he wrote:

If the king holds back in respect to the city, all the cities of Canaan
(Kinahni) will be lost to him.... But let the king, my lord, quickly
[36]
send archers that they may take the city at once.
Finally Egypt did act. Aziru was apprehended and taken to Egypt
where he was evidently detained for some time. A son of Aziru
addressed a letter to the Egyptian official Dudu begging him for his
father’s release:

Aziru is your servant. Do not detain him there. Send him back
quickly that he may protect the lands of the king, our Lord.

He goes on to mention “the Sudu people” who were taking


advantage of Aziru’s absence to further their own ends:

And all lands and all Sudu-people have said, “Aziru does not come
from Egypt.” And now the Sudu steal out of the lands and exalt
themselves against me, saying, “Your father dwells in Egypt, so we
[37]
will make hostility with you.”

We do not know what happened to Rib-Addi. By his own testimony


he was an old man, but whether he died of natural causes or was a
casualty in the battles of his generation we do not know. There
seems to be a poetic justice in the fate of Aziru. As he had terrorized
the Phoenician countryside, so his Amorite lands were terrorized by
other tribes which were seeking a place for themselves in a time of
general chaos. While Suppiluliumas backed Aziru, the Hittites were
really only using him as a pawn to weaken Egyptian control in Asia
and prepare the way for Hittite domination.

50
Lab‘ayu of Shechem

The troublemaker in the region around Shechem was a man named


Lab‘ayu who, in league with the ‘Apiru people, sought to control the
central hill country of Canaan. Like Aziru, farther north, Lab‘ayu sent
letters to Egypt affirming his loyalty:

Behold, I am a faithful servant of the king, and I have not


committed a crime, and I have not sinned, and I do not refuse my
tribute, and I do not refuse the demand of my deputy. Behold, I
have been slandered and evil entreated but the king, my lord, has
[38]
not made known to me my crime.

Biridya of Megiddo, however, saw things differently:

To the king, my lord and my sun, say: Thus says Biridya, the
faithful servant of the king. At the feet of the king, my lord and my
sun, seven and seven times I fall. Let the king know that ever since
the archers returned, Lab‘ayu has carried on hostilities against me,
and we are not able to pluck the wool, and we are not able to go
outside the gate in the presence of Lab‘ayu since he has learned
that you have not given archers, and now his face is set to take
Megiddo, but let the king protect his city lest Lab‘ayu seize it.
Indeed, the city is destroyed by death from pestilence and disease.
Let the king give one hundred garrison troops lest Lab‘ayu seize it.
Verily there is no other purpose in Lab‘ayu. He seeks to destroy
[39]
Megiddo.

Biridya succeeded in capturing Lab‘ayu, and plans were made to send


him to Egypt. He was turned over to Zurata of Akko, an ally of
Biridya, who was to send him to Egypt by ship. Zurata, however,
[40]
accepted a bribe and released Lab‘ayu. His freedom was brief for
Lab‘ayu was murdered before he could reach home, but lawlessness
continued as the sons of Lab‘ayu continued to terrorize the
[41]
countryside.

Abdi-Khepa of Jerusalem

At least seven letters were addressed to the Pharaoh by Abdi-Khepa


of Jerusalem, asking help in resisting the encroachments of a people
known as ‘Apiru. He notes:
As sure as there is a ship in the midst of the sea, the mighty arm of
the king conquers. Nahrim and Kapasi, but now the ‘Apiru are
taking the cities of the king. There is not a single governor
[42]
remaining to the king, my lord. All have perished.

Abdi-Khepa tends to classify all his enemies as ‘Apiru, a word which


in such contexts is practically synonymous with outlaw, or bandit.
Things are so bad, Abdi-Khepa states, that the tribute which he 51
sent to Egypt was captured by these marauders on the plain of
[43]
Ajalon. This may have happened, but reports concerning Abdi-
Khepa himself suggest that it would not have been beneath his
dignity to concoct such a story to avoid paying tribute.

A neighboring king, Suwardata, thought to have been ruler of


Hebron, complained:

And the king, my lord, should know that Abdi-Khepa has taken my
city out of my hand. Further, let the king, my lord, ask if I have
taken a man, or even an ox or an ass from him.... Further, Lab‘ayu
who had taken our cities is dead, but verily Abdi-Khepa is another
[44]
Lab‘ayu and he takes our cities.

On other occasions, however, Suwardata and Abdi-Khepa were allied


against a common foe, the ‘Apiru. Suwardata wrote:

The king, my lord, should know that the ‘Apiru have arisen in the
land which the god of the king, my lord, has given me, and they
have attacked it, and the king my lord, should know that all my
brothers have abandoned me. I and Abdi-Khepa alone are left to
fight against the ‘Apiru, and Zurata, the prince of Akko, and
Indurata the prince of Achsaph, were the ones who hastened to
[45]
my help.

Amarna Age Palestine


The petty kings in Canaan were permitted their own armed forces
comprising chariots, owned by the aristocracy, and footmen drawn
from the peasant classes. Egypt did not interfere in local rivalries as
long as her revenues continued to come and her commissioners were
able to carry on the royal projects. When a local ruler had a
grievance against his fellows, he could plead his case showing that
the interests of Egypt would be best served by enabling him to defeat
his rivals. This usually meant a request for troops—particularly
bowmen. Egypt tolerated the perpetual squabbles of her subject
states, and it may even have been a policy to allow such quarrels
rather than to permit one state to gain enough power that it could
forge an empire of its own.

Many of the strongholds held by the rulers of Canaanite city states


had been fortified in Hyksos times. Egyptian control, however, was
maintained through commissioners appointed by the Pharaoh to
collect taxes and supervise the compulsory labor groups which
worked on roads, tended the Lebanon forest preserves, or worked in
the Valley of Esdraelon where wheat was grown for the royal 52
court. Under strong Pharaohs, the interests of the Empire were
carefully guarded, but the Amarna Age was a period during which
Egyptian prestige was in eclipse and local rivalries became
increasingly bitter. Only in extreme instances did Egypt interfere, and
then it was usually too late to rectify matters.

A brief letter from Biridya of Megiddo indicates that forced labor


(corvée) was expected of the subject states. Many, however, sensing
the loss in Egyptian power failed to provide laborers for the royal
projects:

To the king, my lord and my sun, say: Thus Biridya, the true
servant of the king, my lord and my sun, seven times and seven
times I fall. Let the king be informed concerning his servant and
concerning his city. Behold I am working in the town of Shunama,
and I bring men of the corvée, but behold the governors who are
with me do not as I do; they do not work in the town of Shunama;
and they do not bring men for the corvée, but I alone bring men
for the corvée from the town of Yapu. They come from Shunama
and likewise from the town of Nuribda. So let the king be informed
[46]
concerning his city.

The ‘Apiru

The identity of the ‘Apiru (also written in cuneiform SA GAZ) has


puzzled scholars since the discovery of the Amarna tablets. Some
categorically affirmed that the ‘Apiru are identical with the Biblical
Hebrews, or Israelites, and that the Amarna tablets reflect the
Canaanite version of events described in the Biblical book of Joshua.
[47]
Most scholars now agree that the ‘Apiru cannot be identified with
the Biblical Hebrews, although many suggest that the peoples are
related. A strong argument against identification comes from the fact
that ‘Apiru appear in a wide variety of places of which there is no hint
in the Biblical narrative. They appear in Sumer during the Ur III
dynasty (ca. 2050 B.C.), in Larsa during the reigns of Warad-Sin and
Rim-Sin (ca. 1770-1698 B.C.), in Hammurabi’s Babylon (ca. 1728-
1686 B.C.), in Mari during the reign of Zimri Lim (ca. 1730-1700
B.C.). They are mentioned in the large bodies of texts from Nuzi,
Ugarit, and Bogazkoy. None of these references bear any relationship
to the people of Israel.

In the Mari tablets the ‘Apiru are described as a semi-nomadic people


settled in the area between the Habur and the Balikh rivers, 53
north of the Euphrates. The tablets from Alalakh mention that
King Idrimi lived seven years among ‘Apiru soldiers. Studies in the
personal names of individuals designated ‘Apiru in the Amarna and
the Nuzi tablets have shown that they do not belong to any one
ethnic group, although West Semitic names are most common in the
Amarna texts.
There is considerable evidence that the ‘Apiru were regarded as a
social rather than an ethnic group. At Bogazkoy they are listed among
the social classes and appear to have been classified between
freemen and slaves. Wherever they appear they have one common
trait—they are beyond the jurisdiction of the established authority.
They frequently appear as a landless people who enter into
dependent status as agricultural workers or soldiers in exchange for
maintenance. The ‘Apiru of the Amarna tablets are never described as
invaders. They are people within the land who occupy areas not
controlled by the larger towns. In a time of weak central government
they sought to profit from the general confusion by challenging the
city-states. Whatever their ethnic origins, they were doubtless joined
by a variety of peoples from the oppressed elements of the
population. To the rulers of Canaan, the ‘Apiru were lawless bandits,
a menace to society. Although ‘Apiru is a much more inclusive term
than Israel, the citizens of the city-states of Canaan probably thought
of Joshua’s army much as they regarded the ‘Apiru of the Amarna
Age.

Although the place names of the Amarna texts are parallel to those of
the Old Testament, the personal names are totally different. In
Joshua we read of Adoni-zedek, not Abdi-Khepa, as king of
Jerusalem, and a number of other kings are named for the period of
the conquest (cf. Josh. 10:3). Meredith G. Kline, who holds to the
early date of the Exodus (1440 B.C.) has suggested that the conquest
of Canaan by Joshua precedes the Amarna Age and that the ‘Apiru of
the Amarna letters may actually be the forces of Cushan Rishathaim,
Israel’s first oppressor during the time of the Judges. He concludes
that the ‘Apiru are not to be associated with Israel, but rather must
be regarded as oppressors—the first of a series of such oppressors
[48]
described in the Book of Judges.

Most contemporary scholars date the conquest of Canaan after the


Amarna Age, suggesting some time around 1280 B.C., as the 54
[49]
probable date of the Exodus. This would place the Amarna
Age in the period between Joseph and Moses. Aside from the fact
that Israel was in Egypt during this time, and that they lost the
favored position which they enjoyed in the days of Joseph, Scripture
passes over this period with complete silence.

While we may not be able to pinpoint the exact chronology, the


description of events in Canaan during the Amarna Age lends
perspective to Biblical history during the years before the Monarchy.
Local and tribal loyalties were more meaningful than imperial
government, and centralized government was looked upon with
suspicion (cf. Judg. 9:7-15).

55
VII
TRADE AND COMMERCE DURING THE AMARNA
AGE

By the Amarna Age the Mediterranean had become a highway for the
ships of Egypt, Crete, Cyprus, Ugarit, the Phoenician cities, and even
distant Mycenae. Land routes around the Fertile Crescent saw a
steady stream of caravans bearing tribute to kings and items of trade
for commoners. Horses and lapis lazuli were carried westward from
Babylon, and its king Burnaburiash hoped for large quantities of
Egyptian gold. Caravans were subject to attack, and Burnaburiash
made it clear that it was the duty of Akhenaton to punish such
offenders:

Canaan (Kinahhi) is your land, and its kings are your servants. In
your land I have been violently dealt with. Blind them (i.e., the
raiders) and make good the money which they have stolen. Kill the
people who murdered my servants and avenge their blood, for if
you do not kill these people they will return, and my caravans, or
even your messengers they will murder, and messengers between
us will be intercepted, and if that happens, the inhabitants of the
[50]
land will fall away from you.

The king of Alashia (Cyprus) sent copper to Egypt, requesting silver


[51]
and gold in exchange. Iron, which in Hyksos times had twice the
value of gold, became more plentiful during the Amarna Age.
[52]
Tushratta of Mitanni sent iron to Egypt. Iron, however, was not in
common use in Israel until the time of David (I Chron. 22:3; 29:2).
During the days of Saul, the Philistines had a monopoly on iron in
Canaan:

Now there was no smith to be found throughout all the land of


Israel, for the Philistines said, “Lest the Hebrews make themselves
swords or spears”; but every one of the Israelites went down to
the Philistines to sharpen his plowshare, his mattock, his axe 56
or his sickle, and the charge was a pim for the plowshares
and for the mattocks, and a third of a shekel for sharpening the
axes and setting the goads (I Sam. 13:19-21).

Minoan Crete

The great sea power of the eastern Mediterranean prior to the


Amarna Age was Minoan Crete. The Cretans traded with Egypt from
the earliest history of the two peoples. In addition to the direct route
across the Mediterranean, the Minoans made use of an indirect trade
route along the southwestern and southern shores of Asia Minor, and
then southward by way of Cyprus to Egypt. The Egyptian word Keftiu
(Hebrew Caphtorim, Gen. 10:14; Deut. 2:23; Amos 9:7) may be used
of the peoples of southern Asia Minor as well as the inhabitants of
Crete and its adjacent islands. The Philistines trace their ancestry to
the Caphtorim (Amos 9:7), accounting for a non-Semitic element in
southern Canaan.

Cretan trade with Egypt is depicted in the tomb of Rekhmire,


lieutenant governor of Upper Egypt under Thothmes III (ca. 1490-
[53]
1435 B.C.). Here a prince of the Keftiu is depicted with gifts for
the rulers of Egypt. Cretan power came to an abrupt end, however,
some time around the end of the fifteenth century B.C., when
Knossus, the capital, and other centers of Minoan culture were
destroyed. The cause is not known, but Mycenaeans from mainland
Greece may have been responsible, at least in part, for the fall of
Knossus.
Early in the fourteenth century B.C., Mycenae became the cultural
and political center of the Aegean world. Trade with Egypt brought to
the Mycenaeans the ivory that appears frequently in their tombs.
Scarabs discovered at Mycenae bear the names of Amenhotep III and
his wife, Tiy.

The Phoenicians

It was not the Mycenaeans, however, but the Phoenicians who


succeeded the Minoans as the seafarers and the traders of the
eastern Mediterranean. A tomb painting from Thebes shows
Phoenician merchant ships tied up at docks along the Nile with their
crews selling merchandise in the Egyptian bazaar. Amarna letters
speak of Tyrian sailors and the wealth of their home port. Ships of
Arvad also carry merchandise to Egypt. Phoenician control of the
eastern Mediterranean was not challenged until Rome fought a 57
series of wars with Carthage, which began as a Phoenician
colony. We know the conflicts as the Punic (i.e., Phoenician) wars.

Commerce was not without its dangers. Roving bands of pirates from
Lycia in Asia Minor infested the eastern Mediterranean and even
landed on the coast of the Egyptian Delta. Amenhotep III found it
necessary to organize a police force to patrol the Delta coast and
keep the mouths of the Delta closed to all but lawful ships. The police
manned customs houses and collected duty on all merchandise that
was not consigned to the king. The land routes into Egypt were also
policed, and admission was only granted to those with legitimate
business.

58
VIII
THE ART OF AMARNA

Akhenaton’s influence in art, like his religious beliefs, had antecedents


and it would be improper to give him the credit—or blame—for all the
art forms which found expression during the period of his reign.
Nevertheless, under his inspiration we meet a new type of
naturalism, almost an expressionism, coming to full flower. W.
Stevenson Smith notes, “Men of ability ... fell in with the ideas of
Amenhotep IV and after a few tentative efforts, developed a new
[54]
style with remarkable speed.” Bas reliefs show that Akhenaton was
personally interested in art. He appears on visits to the sculptors’
workshops in the company of Nofretete. Akhenaton’s views of art are
reflected in the royal monuments of his reign, the stelae that were
erected to mark the boundaries of Akhetaton, and in the tombs
prepared for government officials in the eastern desert.

Sunken Relief

One change in the Amarna Age art was purely mechanical. Sunken
relief replaces the traditional raised relief in the ornamentation of the
rock tombs. Davies comments on the technique:

The rock in which they are hewn is far from having the uniform
good quality which would invite bas-reliefs of the usual kind. Nor
was Akhenaton willing, it appears, to employ the flat painting on
plastered walls which was so much in vogue, and which the artists
of Akhetaton also employed at times with good effect. The idea of
modelling in plaster was conceived or adopted; and since figures in
plaster-relief would have been liable to easy injury, the outline was
sunk so far below the general surface as to bring the parts in
highest relief just to its level Nor was this the only measure taken
to ensure durability. The whole design was first cut roughly 59
in sunk-relief in the stone itself. Then a fine plaster was
spread over it, covering all the inequalities and yet having the
support of all points of a solid stone core. While the plaster was
still soft, it was moulded with a blunt tool into the form and
features which the artist desired. Finally the whole was painted, all
the outlines being additionally marked out in red, frequently with
such deviations as to leave the copyist in dilemma between the
[55]
painted and the moulded lines.

Amarna Style Head. A relief showing the characteristic art of the


Amarna Age.
Queen Nofretete. The painted limestone bust shows the queen
wearing a conical blue headdress encircled by a band to which
the uraeus, symbol of royalty, is attached. The bust was found in
the studio of the sculptor Thutmose at Akhetaton.

Realism

Akhenaton’s chief contribution to art, however, was anything but


mechanical. Under his prodding, the artists at Akhetaton developed a
realism—and even a distortion—which contrasts with the
conservative, stylizing tendencies of earlier Egyptian art. The chief
sculptor Bek describes himself on a stele as one “whom his majesty
himself taught.”

Arthur Weigall suggests that the innovations which Akhenaton


brought into the art of his day were, in fact, a self-conscious return to
earlier art forms. Young Akhenaton, Weigall assumes, would have
discovered that the sun god Re-Harakhti was much more ancient
than Amon of Thebes, and that ancient art forms differed from those
in use during the Theban supremacy. In reverting to the 60
religious views of the Heliopolitan priesthood, Akhenaton would
[56]
also have chosen to effect a renaissance of earlier art forms.

Others have speculated on the possibility of Minoan influence on


Akhenaton’s art, noting that the Minoans adopted a naturalism which
parallels that of Amarna, although Knossus was sacked some time
during the reign of Amenhotep III. Barring the migration of Minoan
artists to Akhetaton (which is rather unlikely), it may be best to see in
the Amarna art forms a development based upon changes which
were already being felt in art circles in Egypt. John A. Wilson notes
that the older stylized art forms were on their way out as early as the
reign of Thutmose III, and that the earlier tradition ended by the
[57]
time of Hatshepsut.

The naturalism of Akhenaton, however, goes far beyond his


predecessors. The Pharaoh is not depicted in the splendid isolation of
a god-king, but in the informal pose of a husband and father.
Akhenaton habitually appears in the company of his wife, Nofretete,
and their daughters, of whom ultimately there were six. A stele
depicts Akhenaton kissing an infant while a second child sits on the
queen’s knee. Another shows a banquet scene with the king gnawing
on a large piece of meat while his wife is eating roast fowl with her
hands.

Caricature

Not only informality, but actual caricature tended to mark the art of
Akhetaton. The natural deformities of the king were more than
faithfully reproduced—they were exaggerated. The elongated skull,
long thin neck, pointed chin, obtruding stomach, and abnormally
large hips and thighs of the king may have been emphasized by
artists who felt that any characteristic of a son of Aton deserves
special attention.

The way in which people reacted to the king’s wishes may be seen in
a child’s toy depicted at Akhetaton. A tomb painting “shows a model
chariot drawn by monkeys. In the chariot is another monkey urging
along his steeds (his receding forehead is terribly like the king’s), by
him a monkey princess prods the rump of the horse-monkeys which
are jibbing and refusing to budge an inch in spite of a monkey groom
[58]
who is dragging at their bridles for dear life.” Such caricature
would indicate that the “image” of Pharaoh as a son of Aton 61
has been popularly dispelled, and with it much of his power
over his subjects.

Transitional Art Forms

The transition from the pre-Amarna art forms to those encouraged by


Akhenaton may be observed in the tomb of the vizier Ramose in the
Theban necropolis. Ramose first had a portrait of young Akhenaton
carved in his tomb in the conventional style, but later he added a
second portrait in the new style. The latter depicts Akhenaton
standing with Nofretete beneath the rays of the sun, bestowing
golden necklets upon their faithful vizier. Officials of the royal harem
and a number of servants look on. Akhenaton and his courtiers have
the physical characteristics which became conventional in Amarna
art.

The Development of Amarna Art

The most violent break with the older convention came in the early
years of Akhenaton’s reign. Before the move to Akhetaton, the
Theban hillside was dotted with tombs decorated with the newer art
forms and bearing inscriptions praising the Pharaoh. With the move
to Amarna, the art conventions matured. Artists developed their own
distinctive tastes and at times modified the prevailing tendencies.

The painted stucco pavement which Petrie discovered in 1891


expresses the love of nature which the Aton cult encouraged. It
depicts a pool surrounded by clumps of flowers in which birds are
sporting and calves playing. Frescoes from the Green Room of the
North Palace, excavated by Francis Newton in 1924, represent the
luxuriance of a papyrus thicket full of beautiful birds, brightened up
here and there by blue lotuses.

Some of the finest specimens of ancient Egyptian art have come from
the workshop of the sculptor Thutmose, discovered by Ludwig
Borchard during the German expedition at Amarna prior to World War
I. In preparing a series of heads of members of the royal family,
Thutmose chose to refine rather than to stress their physical
peculiarities. Thutmose based his work on keen observation,
augmented by casts taken from life when he wanted to record the
characteristic features of his subject. Among his masterpieces are the
famous painted limestone bust of Nofretete—perhaps the best known
piece of Egyptian art; and an unfinished portrait of the queen now in
the Cairo Museum.
62

IX
THE END OF AN ERA

The high hopes of Akhenaton’s early years met an untimely end. The
Asiatic provinces of Egypt fell away to the Hittites or to local
Canaanite princes who had little sympathy with the Empire. Although
there is no evidence of revolt in Egypt itself, Akhenaton’s alienation
from the older priesthood must have resulted in dislocations of the
economy, and difficulties in the smooth running of government.

Smenkhkare

There is some evidence to suggest that Nofretete lost favor with her
husband and moved to a new palace in the northern sector of
Akhetaton. The king gave high honor to his eldest daughter
Meritaton, whose husband, Smenkhkare became his successor on the
throne of Egypt. We have no records indicating events in the earliest
years of Smenkhkare’s reign but in the third year he is known to have
gone to Thebes. The reason for the visit can only be conjectured, but
it may have been a gesture to appease the Amon priesthood which
was still firmly entrenched there.

Our sources fail us again, but neither Akhenaton nor Smenkhkare are
mentioned after ca. 1350 B.C. Whether they died natural deaths, or
perished at the hands of assassins, can only be guessed. We are not
even sure if Smenkhkare was co-regent with his father-in-law or if
Akhenaton had died before he came to the throne. At most
Smenkhkare reigned but four years. If his trip to Thebes was made to
bring about a reconciliation with the Theban priesthood, it seems to
have failed completely.

63
Tutankhaton-Tutankhamon

Smenkhkare’s successor, Tutankhaton, was married to


Ankhesenpaton, the third daughter of Akhenaton and Nofretete.
Under Tutankhaton the capital was moved back to Thebes, and the
Amarna revolt was at an end. His name, meaning “the living image of
Aton,” was changed to Tutankhamon, “the living image of Amon,” and
Amon was restored to his place as chief deity of Egypt.
Ankhesenpaton’s name was changed to Ankhesenpamon for the
same reason. Although Tuntankhamon was one of Egypt’s lesser
kings, the discovery of his tomb by Howard Carter in 1923 has made
him the best known Pharaoh of Egyptian history to most westerners.

Tutankhamon’s return to the worship of Amon was a conscious


repudiation of the Aton cult. He actually ascribes the calamities that
befell Egypt in the years of Akhenaton to the anger of Amon:
The Throne of Tutankhamon. The throne dates to the time before
Tutankhamon renounced Atonism. His name appears as
Tutankhaton in the inlay, but in the gold work where it could
more easily be altered it has been changed to Tutankhamon. The
back of the throne pictures the king and his wife under the sun
disk (Aton).
Tutankhamon and His God. A black granite statue depicts the god
Amon (large figure) with Pharaoh Tutankhamon, who renounced
the Aton faith of Akhenaton and returned to Thebes, the center of
the Amon priesthood.

The temples of the gods and goddesses ... had gone to 64


pieces. Their shrines had become desolate and had become
overgrown mounds.... The land was topsy turvy and the gods
turned their backs upon this land. If one prayed to a god to seek
counsel from him, he would never come (at all). If one made
supplication to a goddess, similarly she would never come at all.
Their hearts were hurt (?) so they destroyed that which had been
[59]
made.

Following Tutankhamon’s early death we meet a story of intrigue and


international politics which involves his widow. Ankhesenpamon,
fearful of the future of herself and her country, wrote to the Hittite
king, Suppiluliumas, asking that one of his sons be sent to Egypt to
become her husband:

My husband died, and I have no son. People say that you have
many sons. If you were to send me one of your sons, he might
become my husband. I am loath to take a servant of mine and
[60]
make him my husband.

The Hittite king, suspecting something amiss, sent a servant to check


on matters in Egypt. When the envoy reached Thebes, the widowed
queen asked:

Why do you say, “They may try to deceive me.” If I had a son
would I write to a foreign country in a manner which is humiliating
to myself and my country. You do not trust me and tell me such a
thing. He who was my husband died and I have no sons. Shall I
perhaps take one of my servants and make him my husband? I
have not written to any other country. I have written only to you.
People say that you have many sons. Give me one of your sons,
[61]
and he is my husband and king in the land of Egypt.
Suppiluliumas was convinced of the good faith of the young widow
and sent a son to Egypt, but the young man never reached Thebes.
Along the way he was murdered by Egyptians who resented the
thought of a foreigner as their ruler. The result was a period of war
between the Hittites and Egypt. Another son of Suppiluliumas made a
record of the affair:

When my father gave them (the Egyptians) one of his sons (to take
over the kingship), they killed him as they led him there. My father
let his anger run away with him; he went to war against Egypt and
[62]
attacked Egypt.

The battle is not mentioned in the Egyptian annals. Probably it was


brief and indecisive, for the Hittites could not afford to throw a major
army into such a campaign. The rising power of Assyria was a threat
to Hittite control in the north, and she had to be ready to 65
protect her northern provinces. Had the Hittites launched a
major campaign against Egypt it is doubtful if she could have
survived.
Horemhab. Granite statue of the commander of Tutankhaton’s
armies, later a Pharaoh in his own right.

Eye and Horemhab


The rule of Egypt fell to the aged vizier Eye, who had been a
counselor and friend of Akhenaton. After four years Eye was
succeeded by Horemhab (ca. 1340-1310 B.C.), an energetic ruler
who sought to restore Egypt’s fortunes abroad and erase the memory
of the Amarna revolt at home. As a young general, Horemhab had
espoused the cause of Akhenaton, but as a Pharaoh he sought to
obliterate the records of the Amarna kings with as great enthusiasm
as Akhenaton had sought to eliminate the name of Amon. Later
orthodox king lists omit the names of Akhenaton, Smenkhkare,
Tutankhamon, and Eye, placing the name of Horemhab immediately
after Amenhotep III.

Although the Amon priests of Thebes seemed to be more firmly


entrenched than ever after the accession of Horemhab, the calendar
could not be pushed back completely. Egyptian art and 66
literature retained some of the naturalism of the Amarna
movement. There were effects in the religious world, too, for
although Atonism was not pure monotheism, it exhibited tendencies
in that direction which persisted in the Egyptian thought. God is
frequently addressed in the singular, although under different names,
in the hymns of the later periods of Egyptian history.

67

You might also like