See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259815160

Ordinary Differential Equations by Vladimir I. Arnold

Article  in  SIAM Review · January 2007

DOI: 10.2307/20453964

CITATION READS
1 9,510

1 author:

Carmen Chicone
University of Missouri
173 PUBLICATIONS   3,258 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Solution multiplicities for chemical reactors View project

All content following this page was uploaded by Carmen Chicone on 10 December 2018.

The user has requested enhancement of the downloaded file.

 SIAM REVIEW

c 2007 Society for Industrial and Applied Mathematics
Vol. 49, No. 2, pp. 327–367

Book Reviews
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php

Edited by Robert E. O’Malley, Jr.

Featured Review: The Art of Computer Programming, Vol. 4. Generating All

Trees: History of Combinatorial Generation. By Donald E. Knuth. Addison-Wesley,
Upper Saddle River, NJ, 2006. $19.99. vi+120 pp., softcover. ISBN 0-321-33570-8.

Being just a humble mathematician, I have not the faintest idea whether mathematics
is the music of the heavenly spheres. Yet, it is fairly safe to bet that, had heavenly
spheres been studied by standard tools of science, had they been measured, weighed,
had their chemical composition been analyzed, had inquiry been made of their ulti-
mate structure, of their mechanics and dynamics, of their science, it would have been
formulated in the language of mathematics.
We all take for granted this universal effectiveness of mathematics, yet is it im-
portant to pause once in a while and ask ourselves what exactly renders mathematics
into such a universal tool. The first, obvious, answer is its rigor. Mathematics places
you into a straitjacket of exact thinking, forces you to define precisely and to infer
rigorously from your assumptions. It leaves at its best no room for woolliness, hand
waving, and “proof by authority” that are besetting other areas of scholarship.
Yet, rigor is just part of the answer. Another, perhaps less evident, secret behind
the surprising effectiveness of mathematics is its ability to express universal relation-
ships. At an elementary level this is clear, since the most basic universal relationships
are enumerative: we count objects and compare them. However, at a more advanced
level, we wish to take heed not just of the number of our “objects” but of their in-
ternal relationship and interconnectedness. At a phenomenological level numbers are
unequal to this task and we need to use graphs: collections of vertices and edges.
The most useful graphs are trees, where a single path of edges extends between
every two vertices. By singling out one vertex as a root, we establish partial ordering:
vertex a precedes vertex b if it features earlier along the (unique!) path extending
from the root. (In particular, the root “precedes” every other vertex.) A useful model
is a family tree, but such rooted trees extend the length and width of applications
of mathematics: from linguistics and genetics all the way to quantum field theory
(Feynman diagrams!).
It is perhaps not surprising that computer scientists (and, indeed, computer pro-
grammers) are the main beneficiaries of mathematical rigor: unlike protein-based
life-forms, computers simply will not settle for less. You can sometimes talk your way
out of a hole with your students or colleagues—certainly not recommended but this
is the real world!—but not with computers. Moreover, the importance of effective im-
plementation of ideas is clearly emphasized in an algorithmic setting, and this, again,
is best undertaken within the confines of formal reasoning and a language which is
rigorously descriptive of the inner structure of the underlying problem. Which brings
us to the slim, yet important, “fascicle” on the generation of trees.

Publishers are invited to send books for review to Book Reviews Editor, SIAM, 3600 University
City Science Center, Philadelphia, PA 19104-2688.
327
 328 BOOK REVIEWS

Donald Knuth (thrice-daily blessed by all TEX-using mathematicians!) has been

assembling and writing for the last 45 years a comprehensive compendium of computer
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php

programming. The project, which apparently started as a modest attempt to write

a textbook on compiler design, has mushroomed into three huge volumes already
available (and updated in several editions), a fourth volume (in reality, three separate
volumes) being written piecemeal, and three more volumes to come. The fourth
“multivolume,” dedicated to combinatorial algorithms, is published in small chunks
(fascicles), of which the current is the fourth. Its theme is the generation of all trees
in a structured manner.
Careful reader would have noticed a slight of hand in the last sentence. “All”
trees form an infinite set, and even Donald Knuth is not allowed algorithms running
in infinite time or at infinite speed. Thus, the idea is to restrict the subject of study
to a finite subset of trees, delimited by a feature (the simplest being just the number
of their vertices), and generate all its members in a fast and structured manner,
transparent to good computer programming practice.
The first third of the fascicle is concerned with algorithms: while previous fascicles
dealt with n-tuples, permutations, combinations, partitions, and set partitions, this
one is concerned with generating deterministic and random trees. Now, the basic
common denominator to all enumerations over combinatorial objects is that they are
computationally intensive and the number of operations increases rapidly with regard
to “size” (the number of vertices, the length of a word, etc.). This underlies the
importance of efficiency in the design of algorithm, a thread implicit in the exposition.
As a particularly appealing example I would single out the Christmas tree pattern:
We arrange

n  all 2n bit strings of length n (that is, binary numbers with n digits) into
n/2 rows and n + 1 columns by replacing every row “σ1 σ2 . . . σs ” of the order-n
pattern by the two rows
σ2 0 . . . σs 0
σ1 0 σ1 1 . . . σs−1 1 σs 1
(omitting the first row for s = 1). Thus,
n=1: 01
n=2: 10
00 01 11
n=3: 100 101
010 110
000 001 011 111
n=4: 1010
1000 1001 1011
1100
0100 0101 1101
0010 0110 1110
0000 0001 0011 0111 1111
A Christmas tree pattern emerges. . . . Inasmuch as this arrangement is probably
of little use beyond the didactic, other tree algorithms are of vital importance in
programming.
The list of tree algorithms is exhaustive from a purely computer-scientific point
of view. It would have been arguably interesting to consider trees and forests (words
 BOOK REVIEWS 329

composed of trees) that occur in other areas of mathematical and scientific research
and which reveal, often by means of a shuffle product, Hopf algebra, and dendriform
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php

algebra formalism, a range of truly surprising connections and hidden structures. My

own favorite is the recent set of ideas centered around the work of Alain Connes
and Dirk Kreimer, bringing together the “Butcher trees” and B-series from numeri-
cal analysis of Runge–Kutta methods with the understanding of the renormalization
group in quantum field theory. Yet, one should not forget that this is a book (or a
fascicle) in computer programming!
The second third of the book is concerned with the history of trees and tree
generation. It might be surprising to a mathematician that this youngest area of
scholarship, computer sciences, is exhibiting such an awareness of its historical roots,
but it is one of these pleasing surprises. Once we consider trees not just as abstract
combinatorial objects but as a means to express combinatorial relationship, they im-
pinge on issues that concerned scholars for millennia: from Chinese tile arrangements
to Indian and Greek poetic meters, from medieval Kabbalah to medieval music and
beyond. If the list appears esoteric, this is only because of our preconceptions. The
quest for structure, pattern, and symmetry underlies human culture and scholarship,
even if sometimes it is expressed in verses of the poets, rather than in mathematical
formalism. This portion of the fascicle might be an easy reading after the formal,
detailed algorithms of its first part, but it is delightful, fascinating, and it successfully
endeavors to connect modern reasoning to its ancient intellectual roots. Moreover, it
drives powerfully home the message how wide these roots are: not just Europe but
also the Islamic world, India, China, and Japan.
Those readers who became excessively mellow in the second part of the fascicle are
brought down to the formal earth in the closing third. It contains detailed solutions to
exercises that conclude Part I and are highly recommended to those virtuous readers
who have tried their best and, like this reviewer, often failed in their own efforts.
There is little need to dwell upon the fundamental value of the “Knuth project”
to computer sciences. Writing as an applied mathematician rather than a computer
scientist, this reviewer is persuaded also in the underlying broader message of the
importance and beauty of formal structures that bind together science and technology
with the language of mathematics. As a community, we owe thanks to Donald Knuth
for much more than just mathematical typesetting.
ARIEH ISERLES
University of Cambridge

A Course in Mathematical Biology: Quan-
titative Modeling with Mathematical and
Computational Methods. By Gerda de Vries,
Thomas Hillen, Mark Lewis, Johannes Müller,
and Birgitt Schönfisch. SIAM, Philadelphia, 2006.
$69.00. xii+309 pp., softcover. ISBN 0-89871-
612-8.
A variety of mathematical and statistical
topics are covered in A Course in Mathe-
matical Biology: Quantitative Modeling with
Mathematical and Computational Methods,
topics that are important in the formula-
tion and analysis of mathematical models
in biology. Six chapters are devoted to
difference equations, ordinary differential
equations, partial differential equations,
stochastic processes, cellular automata,
and parameter estimation (Chapters 2–7).
Mathematical and statistical methods are
introduced through some classical and well-
chosen biological models such as models for
population genetics, host-parasitoid, SIR
epidemic, spruce budworm, Fisher’s equa-
 330 BOOK REVIEWS
tion, forest succession, polymerase chain
reaction, and simple birth and death pro-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
cesses. An extensive list of 169 references
is provided for more detailed treatment of
the methods and proofs of some of the the-
orems. Exercises are provided at the end
of each of the six chapters. A tutorial on
Maple and suggestions for 25 group projects
are provided in Chapters 8 and 9, respec-
tively. The topics of the projects come from
epidemiology, population dynamics, spa-
tial spread, and physiology. They are not
the typical “toy” textbook problems that
have well-defined solutions but are realistic
open-ended problems that require students
to think creatively and employ a variety
of techniques to address the problem, e.g.,
Project 5: Cholera in South Africa 2000–
2001; Project 11: Fairy Rings; Project
18: Re-invasion of Otters to California’s
Coast; and Project 20: Modeling of Heart
Beats. Two completed projects are pre-
sented in detail in Chapter 10, Project 10:
Cell Competition and Project 14: Chemo-
tactic Paradox. The projects are designed
for group work, requiring students with
strengths in different areas to formulate and
analyze models that will address the prob-
lems posed in the project. For example, in
Project 18 on sea otters, students are asked
to plot the distance spread northward and
southward along the California coast and
the total range radius versus time based on
data from reference [3] and to research the
life history of sea otters to make a realistic
mathematical model for the spread process.
Given the small amount of space devoted
to each of the six topics, there is a wide
range of mathematical and statistical topics
introduced, e.g., stability theory for equilib-
rium solutions, bifurcation theory, traveling
wave solutions, branching process models,
maximum likelihood method, and Akaike
information criteria. Because the coverage
of the topics is abbreviated, there are neces-
sarily some gaps in the presentation of the
theory. The references cited by the authors
fill in these gaps and if the book is used for a
course, the references should be made avail-
able to the students as additional sources
of information. Students should also have
access to computer software such as Maple
and MATLAB. Some topics that are not
discussed in this book are the linear theory
of difference and differential equations and
the Routh–Hurwitz criteria for higher order
differential equations. Bifurcation theory
for difference equations is covered briefly in
terms of a logistic difference equation. The
discussions about continuous-time Markov
chains and cellular automata do not in-
clude computer programs to illustrate how
the graphics are generated.
In the preface of A Course in Mathe-
matical Biology: Quantitative Modeling with
Mathematical and Computational Methods,
it is stated that the “book stems from ped-
agogic material developed by the authors
for a 7–11 day workshop in mathemati-
cal biology.” This book would work very
well for a short course or a workshop. In
a classroom setting, it would work well
if the course were team-taught by faculty
in mathematics, statistics, and/or biology
with the objective of getting students to
work in interdisciplinary teams to study bi-
ological problems (as illustrated by the two
completed projects in Chapter 10). This
book would not work well for a traditional
course with lectures, homework, and ex-
ams. The book by Britton [1] is designed
for a more traditional mathematical biology
course. However, the intent of A Course in
Mathematical Biology: Quantitative Model-
ing with Mathematical and Computational
Methods is to introduce students to mathe-
matical biology in a nontraditional setting,
and to allow them to interact and to develop
model building and critical thinking skills.
This can be accomplished through the mod-
eling projects described in the text.
The topics presented in the book under
review are accessible to undergraduate stu-
dents who have had courses in calculus, lin-
ear algebra, ordinary and partial differential
equations, and probability theory. Greater
mathematical maturity is required of the
student for this book than for some other
books designed for undergraduate students
[2, 4, 5], but it also covers a broader range
of topics. Similar to the latter books, use
of computer algebra systems and computer
programming languages for numerical anal-
ysis and simulation are an essential part of
the course. An especially attractive feature
of this book is the presentation of a broad
range of mathematical and statistical tools
and techniques with the overarching goal of
 BOOK REVIEWS
using these tools and techniques to study
realistic biological problems. A course de-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
signed with this goal in mind would pro-
vide excellent training for undergraduate
students in interdisciplinary research.
REFERENCES
[1] N. F. Britton, Essential Mathematical Bi-
ology, Springer Undergrad. Math. Ser.,
Springer-Verlag, London, 2003.
[2] S. P. Ellner and J. Guckenheimer, Dy-
namic Models in Biology, Princeton
University Press, Princeton, NJ, 2006.
[3] J. A. Lubina and S. A. Levin, The spread
of a reinvading species: Range expan-
sion in the California sea otter, Am.
Nat., 131 (1988), pp. 526–543.
[4] D. D. Mooney and R. J. Swift, A Course
in Mathematical Modeling, Mathemat-
ical Association of America, Washing-
ton, D.C., 1999.
[5] E. K. Yeargers, R. W. Shonkwiler, and
J. V. Herod, An Introduction to the
Mathematics of Biology with Compu-
ter Algebra Models, Birkhäuser Boston,
Boston, MA, 1996.
LINDA J. S. ALLEN
Texas Tech University
Real Optimization with SAP APO. By Josef
Kallrath and Thomas I. Maindl. Springer-Verlag,
Berlin, 2006. $79.95. xxvi+321 pp., hardcover.
ISBN 3-540-22561-7.
Producing and bringing products to market
underlie the lifestyle of those in developed
economies. It is easy for the lay person to
overlook the complexities inherent in these
processes. Such complexities include, but
are not limited to, adjusting activities to en-
sure that demand is met while not exhaust-
ing raw materials, labor, or transportation
possibilities, and, if demand can be sat-
isfied, minimizing a cost or maximizing a
profit. Adding to the complexities are mul-
tiple or conflicting product-specific goals,
such as minimizing the changes over time
in use of labor on assembly lines. Advanced
software systems that integrate data acqui-
sition, user interfaces, and mathematical
ideas and algorithms for constraint satis-
faction and local and global optimization
331
are increasingly used. This book is about
such supply chain problems and software.
The comment on the back jacket accu-
rately describes readers who can clearly
benefit from this book: “The authors
address readers involved in optimization
projects in which SAP and, particularly,
SAP APO are implemented in companies.
These are the project designers, project
leaders, the IT personnel inside the com-
panies, but also operations research practi-
tioners, supply chain management consul-
tants, and decision makers in the area of tool
selection for optimization tasks.” However,
the authors, expert consultants in supply
chain issues but also accomplished astro-
physicists in academia, are in a position
to provide relevance as well as precise lan-
guage, mathematical perspective, and clar-
ity. This bridge between the academic and
commercial can benefit both users of pack-
ages for supply chain planning and opera-
tions and academically oriented members of
the mathematical optimization community.
The term “real optimization” in the title
can refer to two things. First, it can refer
to the process of finding an actual optimum
of an objective function, as contrasted with
merely producing a production or schedul-
ing plan that satisfies requirements. Sec-
ond, it can refer to the fact that the case
studies involve real companies with imple-
mentations tailored to and used in actual
supply chain planning and operation.
SAP is a well-known corporation that
provides general software and consulting for
data acquisition and management, as well as
software to compute feasible solutions and
optimal solutions to supply chain problems.
External companies and consulting services
also supply interfaces to SAP products.
“SAP APO” stands for “SAP Advanced
Planning and Optimizer.” Most of the book
focuses on optimization and supply chain is-
sues in the context of SAP products, which
is reasonable in view of those products’
widespread use. Readers with a more gen-
eral interest in applied optimization or sup-
ply chain problems can view the book’s use
of SAP APO as one large case study.
The book is organized into four parts:
an extensive introduction, detailed case
studies, a summary section, and three ap-
pendices.
 332 BOOK REVIEWS
Part I: The Introduction. Chapter 1 gives
an introduction to the business perspective
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
of supply chain data acquisition and organi-
zation, including standardized terminology
for supply chain and planning systems and
a brief introduction to the components and
structure of SAP APO.
Chapter 2 contains definitions of terms
related to data models and optimization,
and types of optimization problems (mixed
integer linear programs, nonlinear pro-
grams, etc.). Problem representation in
the data (the supply chain model as in
SAP APO) and the mathematical model
(in terms of variables, indices, objective,
and constraints) are contrasted. The au-
thors note that SAP allows definition of
a model in supply chain terminology, but
the automatic generation of objective and
constraints is proprietary.
Chapters 3 and 4, containing numer-
ous examples and screen shots, illustrate
setting up and dealing with the model in
SAP APO. For instance, the system can be
instructed to treat mixed integer models
as purely continuous models, etc. These
are practical ways of conquering combina-
torial complexity. The result is a feasible
(although not necessarily globally optimal)
solution obtained within a predictable com-
puting time; the point is that this solution
is both practical to obtain and valuable in
actual supply chain planning and operation.
A detailed derivation of a mathematical op-
timization problem from an example supply
chain model (although hidden from the user
in SAP APO) is also presented.
Part II: Detailed Case Studies. Chapters 5
through 9 contain case studies describing
specific consulting (or application develop-
ment) experiences, in candid detail. This
includes interaction with the business client
to determine goals and specifics of the appli-
cation, thoughts on where and how human
interaction with the software system best
occurs, how well the packaged components
in SAP APO fit the application, and when
customized external solutions are more ap-
propriate.
These case studies illustrate the impor-
tance of user interfaces and user interaction.
That is, deciding on an actual plan does not
only involve a single pass of formulating a
fixed model, running a global optimization,
then using the result to dictate the business
plan, but is an iterative process:
1. posing an initial model;
2. doing an optimization to compute an
initial plan;
3. adjusting objective function weights
and adding previously unidentified
constraints to define business priori-
ties and make the solution correspond
more closely to reality;
4. repeating steps 1, 2, and 3.
Indeed, an important early part of the pro-
cess is prioritizing different business ob-
jectives, which leads to different objective
functions. Heuristics such as “minimize as-
sembly line idle time” may be more practical
to use than overall goals such as “maxi-
mize profit,” and various overall goals may
conflict. Often more than one type of opti-
mization is involved, with an initial approx-
imation given, say, with a continuous linear
program solution, and subsequent refine-
ments done with mixed integer technology.
The book provides details.
The specific case studies, in Chapters 5
through 8, are from semiconductor man-
ufacturing, consumer products, the auto-
motive and chemical industries, and the
process industry. Chapter 9, written by the
external authors Rémi Lequette of ILOG
corporation and Axel Hecker of Mathesis
GmbH, describes models and implemen-
tations tailored to specific situations, for
which the intrinsic capabilities of SAP APO
do not suffice.
Parts III and IV: Concluding Considera-
tions and Appendices. The 20-page conclu-
sion contains a summary of what users can
expect from SAP APO, general thoughts
on design and implementation of business
supply chain software, and thoughts on the
relationship of business software to ideas
and techniques from the mathematical op-
timization community. It also contains gen-
eral thoughts on what users can and cannot
expect from SAP APO, how such users can
increase their benefits by doing “real opti-
mization” (as opposed to data management
and finding a feasible plan), and how SAP
might improve its “APO” product.
 BOOK REVIEWS
The first appendix, “The Hitchhiker’s
Guide to SAP APO,” gives a brief introduc-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
tion to the SAP APO components and SAP
APO’s hierarchical planning philosophy.
The second appendix, 20 pages long, con-
tains a succinct overview of “Mathematical
Foundations of Optimization.” The topics
include linear programming (with sketches
of the primal simplex algorithm and asso-
ciated techniques, the dual simplex algo-
rithm, and interior point methods), mixed
integer linear programming, and multicri-
teria optimization and goal programming.
(Mixed integer linear programming rep-
resents the predominant type of underly-
ing mathematical optimization problem en-
countered in the book.)
The book includes a glossary, 104 refer-
ences dealing with both mathematical and
business issues, biographies of the authors,
a seven-page index, and a list of acronyms
and abbreviations.
Concluding Recommendations. I recom-
mend this book for personnel in corpo-
rations and governmental entities contem-
plating improving supply chain operations
with SAP software. The book may also
be of more general interest to operations
researchers and mathematical optimization
professionals, especially those interested in
mixed integer linear programs, those want-
ing to keep their academic activities relevant
to real-world concerns, and those wishing to
gain experience in commercial applications.
R. BAKER KEARFOTT
University of Louisiana at Lafayette
Probability Theory with Applications.
Second Edition. By M. M. Rao and Randall J.
Swift. Springer-Verlag, New York, 2006. $89.95.
xvii+527 pp., hardcover. ISBN 0-387-27730-7.
This book is a thoroughly revised and up-
dated version of the first edition written by
M. M. Rao. It expresses the in many ways
original views of the authors on how the sub-
ject could be presented in graduate courses
for students of mathematical sciences. The
first edition was fairly successful and the re-
viewer used it for preparing his own lectures
on probability theory and recommended it
333
to his students. In the meantime about
20 years have passed and a second edition
was necessary to take into account new ad-
vances in theory and applications, in par-
ticular providing the basic mathematical
background for stochastic modeling in fi-
nance.
The reviewer was pleased to notice that
this second edition retains the positive fea-
tures of the first edition with regard to
style and mathematical presentation. Ide-
ally, readers of this book should be famil-
iar with real analysis and the Lebesgue
measure. This would considerably facili-
tate the understanding of Chapters 1 to
3, which are devoted to the foundations
of probability theory. In Chapter 1, spe-
cial emphasis is placed on the fundamen-
tal law of probability (better known as the
law of the subconscious statistician) and
key properties of characteristic functions,
whereas Chapters 2 and 3 give sophisti-
cated treatments of basic concepts such as
independence and conditional expectations
in conjunction with the laws of large num-
bers, the Glivenko–Cantelli theorem, ran-
dom walks and queuing models, martin-
gales, and the theorem of Radon–Nikodým.
Chapter 4 provides an excellent treatment
of probability distributions, characteristic
functions and their mutual relationships, as
well as applications to the convergence of
sums of random variables. Chapter 5 is the
“heart of the analytical theory of proba-
bility.” From a unifying point of view, it
deals with other topics such as infinite di-
visibility and Spitzer’s identity, as well as
central and other limit theorems. The er-
ror estimation for central limit theorems
is, unfortunately, restricted to the results
of Berry (1941) and Esseen (1945). Other
improved results due to Bikelis, Nagaev,
and in particular Michel (1981) and Paditz
(1989) are not even mentioned. It is a gen-
eral shortcoming of the book that in crucial
cases references to the state of the art are
missing, particularly in Chapters 6 to 8
dealing with martingales, more limit theo-
rems, and some basic concepts of stochastic
processes. Nevertheless, these chapters may
very well serve as a first reading and can
be strongly recommended to students and
other readers who need to upgrade their
knowledge of the mathematical theory of
 334 BOOK REVIEWS
probability to be able to follow modern ap-
plications.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
Compared to the first edition, the lists
of exercises stated after each chapter have
been considerably enhanced and adapted to
changes of content. For more sophisticated
exercises, the authors give rather detailed
instructions on how to solve them. Gener-
ally, lecturers will appreciate this book as a
valuable source for supplementing their own
collections of student-tailored problems. As
with the first edition, the reviewer will defi-
nitely make use of this book for developing
and revising his own lectures and encourage
serious students to use it for accompanying
reading.
FRANK BEICHELT
University of Witwatersrand
Johannesburg
Computational Contact Mechanics. Sec-
ond Edition. By Peter Wriggers. Springer-
Verlag, New York, 2006. $99.00. xii+518 pp.,
hardcover. ISBN 3-540-32608-1.
Most books on finite element methods fo-
cus primarily on element formulations and
treat only the simplest boundary condi-
tions. This is unfortunate given the im-
portance of contact mechanics in many
engineering problems and the mathemat-
ical challenge of imposing inequality con-
straints accurately. Computational Contact
Mechanics, by Peter Wriggers, is an excel-
lent book that fills this important gap.
While contact problems are intrinsically
nonlinear because of the contact inequal-
ity, the emphasis is on contact methods for
problems that have nonlinear, large defor-
mations, nonlinear material response, and
coupled thermomechanical problems. The
book is especially strong in four areas: (1)
its coverage of the mechanics of contact;
(2) the comprehensive treatment of differ-
ent methods for imposing inequality con-
straints; (3) formulations of large deforma-
tions; and (4) thermomechanical contact.
Most computational contact research fo-
cuses on frictionless contact or contact with
Coulomb friction. Compared to the numeri-
cal difficulties associated with realistic mod-
els, these models are the Poisson’s equation
of contact research: if a method can’t solve
problems with these models, it probably
can’t solve any contact problem. The book
covers constitutive equations for both nor-
mal contact and tangential contact, includ-
ing friction laws for metal forming, rubber
and polymers, concrete, and several others.
It discusses the computational challenges
of incorporating these models into contact
algorithms and demonstrates that a great
deal of research remains to be done.
The penalty method and Lagrange multi-
pliers are the two methods most commonly
used in contact algorithms. Wriggers covers
these methods plus direct constraint elim-
ination, Nitsche’s method, perturbed and
augmented Lagrange methods, the barrier
method, and the cross-constraint method.
An especially nice feature of the book is
that the discrete equations are derived and
written out explicitly for each method for a
simple model one-dimensional problem, and
their accuracy is compared. After devel-
oping discretizations for node-to-node con-
tact and for nonmatching meshes assuming
small deformations, almost 100 pages are
devoted to large deformation contact dis-
cretizations. Recent research on mortar dis-
cretizations and smooth contact discretiza-
tions is presented.
The typical treatments of thermome-
chanical contact are naive, ignoring the
interaction between the mechanical contact
and the heat conduction across the con-
tact interface. Wriggers demonstrates with
a simple one-dimensional problem that a
seemingly straightforward approach may
not be well posed. The requirements for a
well-posed problem and realistic heat con-
duction models for contact interfaces are
presented.
Additional topics that are covered thor-
oughly include global solution algorithms,
contact search and detection strategies,
beam contact, critical point analysis with
contact constraints, and adaptive finite el-
ement methods with contact.
The book is not simply a monograph re-
viewing the author’s distinguished research
career, but a true textbook covering the
research of the major players in computa-
tional contact mechanics over the last 25
years and offers an extensive list of refer-
 BOOK REVIEWS
ences. While Peter Wriggers is at the Uni-
versity of Hanover, the book is not a trans-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
lation and the writing is very clear. If I were
given the opportunity to teach an advanced
special topics course on finite element meth-
ods, I wouldn’t hesitate to use this book.
If it has a shortcoming as a textbook, it
is that it does not have homework prob-
lems. Although it is self-contained in the
sense that there are chapters that develop fi-
nite deformation continuum mechanics and
nonlinear finite element methods from the
basics, the pace is too fast to be a first
introduction to these topics. I recommend
having a background of at least one year at
the graduate level in continuum mechanics
and finite element methods before tackling
this book. For a researcher who has studied
these areas before and may be a little rusty,
the text provides an excellent review.
This book is not written in the theorem-
proof style of math textbooks, but in the
constructive style found in science and engi-
neering textbooks. As with shell elements,
the majority of accurate, robust contact
methods are not amenable to formal proofs
of convergence, and this book therefore re-
flects that fact. Those looking for a more
formal style of presentation would probably
prefer the 1988 SIAM book Contact Prob-
lems in Elasticity: A Study of Variational
Inequalities and Finite Element Methods by
Kikuchi and Oden, or Glowinski, Lions, and
Trémolières’ Numerical Analysis of Varia-
tional Inequalities (North-Holland, 1981).
DAVID J. BENSON
University of California
San Diego
Ordinary Differential Equations. By
Vladimir I. Arnold. Springer-Verlag, Berlin, 2006.
$64.95. iv+334 pp., softcover. ISBN 3-540-
34563-9.
Professor Arnold has expanded his clas-
sic book to include new material on ex-
ponential growth, predator-prey, the pen-
dulum, impulse response, symmetry groups
and group actions, perturbation and bifur-
cation, Sturm–Liouville theory, and first-
order partial differential equations. In ad-
dition, he has included a wealth of new
335
examples and exercises. The new edition is
highly recommended as a general reference
for the essential theory of ordinary differ-
ential equations and as a textbook for an
introductory course for serious undergrad-
uate or graduate students.
There is a standard criticism of Professor
Arnold’s books: They are beautiful treat-
ments of their subjects that are appreciated
by experts, but too many details are omit-
ted for students to learn the mathematics
required to prove the statements that he
so effortlessly justifies. While this point of
view has some merit in pedagogy, its logical
extreme (the mathematical cookbook at the
introductory level and the advanced presen-
tation of pure mathematics devoid of un-
derlying ideas and connections to the other
sciences) is strongly opposed by Arnold.
Indeed, there is a rebuttal in the preface:
The completion of the formation of
analysis as an independent scientific
discipline is connected with the name of
Leibniz (1646–1716). Another of Leib-
niz’ grand achievements is the broad
publicizing of analysis (his first publi-
cation is an article in 1684) and the
development of its algorithms to com-
plete automatization: he thus discov-
ered a method of teaching how to use
analysis (and teaching analysis itself)
to people who do not understand it
at all—a development that has to be
resisted even today.
Arnold’s book is certainly meant to teach
the subject of differential equations to those
who truly wish to understand it.
The first edition’s first chapter, “Ba-
sic Concepts,” has six sections (a total of
47 pages) with titles “Phase Spaces and
Phase Flows,” “Vector Fields on the Line,”
“Phase Flows on the Line,” “Vector Fields
and Phase Flows in the Plane,” “Nonau-
tonomous Equations,” and “The Tangent
Space.” The material in the first chapter
of this edition has expanded to 75 pages
with section titles “Phase Spaces,” “Vec-
tor Fields on the Line,” “Linear Equa-
tions,” “Phase Flows,” “The Action of
Diffeomorphisms on Vector Fields and Di-
rection Fields,” and “Symmetries.”
The new material includes a discussion
of some popular elementary topics found
 336 BOOK REVIEWS
in undergraduate textbooks: exponential
growth, logistic growth, harvesting, and
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
predator-prey. These concepts are seam-
lessly woven into the original narrative that
treated elementary mechanics: free fall,
small oscillations, the pendulum, the in-
verted pendulum, and the spherical pendu-
lum. Direct product planar systems, that
is, systems of the form
ẋ = f (x),
ẏ = g(y),
are introduced and used to illustrate the
typical phase portraits of planar rest points.
This leads to a discussion of separable
equations, and culminates in a proof of
the periodicity of the physical orbits of the
Lotka–Volterra system. The reader is soon
introduced to limit cycles, structural sta-
bility, and continuations (of rest points and
periodic orbits) via the implicit function
theorem.
A simple and elegant treatment of the
(Dirac) δ-function is given along with an
introduction to Green’s function for inho-
mogeneous linear equations.
The actions of diffeomorphisms on vec-
tor fields and flows (changes of coordinates
for differential equations) is discussed in
depth. This sets the stage for the intro-
duction of symmetry groups—Lie groups
as they first appeared—and their applica-
tion to the solution of homogeneous and
quasi-homogeneous differential equations.
The treatment here exemplifies the differ-
ence between the Arnold style and the cook-
book style of most elementary (and widely
used) texts on ordinary differential equa-
tions. The cookbooks give the change of
variables required to transform a homoge-
neous equation into a linear equation (which
can be solved explicitly), but Arnold reveals
the origin of this transformation as a natu-
ral result of an underlying symmetry.
Chapter 2, “Basic Theorems,” has been
expanded from 47 to 63 pages. It con-
tains the sections “Rectification Theorems”
(which includes the important theorems on
existence, uniqueness, and smooth depen-
dence on initial data and parameters), “Ap-
plications to Equations of Higher Order
than First,” “The Phase Curves of an Au-
tonomous System,” “The Derivative in the
Direction of a Vector Field and First Inte-
grals,” and “The Conservative System with
One Degree of Freedom,” together with
a new section called “First-Order Linear
and Quasi-linear Partial Differential Equa-
tions.” In this section, the essentials of the
theory of characteristics (in the simplest
case) are presented with the underlying
theme best expressed by a quote from its
introductory paragraph:
The essence of the connection between
a partial differential equation and a
characteristic equation is that the mo-
tion of a solid medium can be described
using both the ordinary differential
equations of motion of its particles and
the partial differential equations for a
field.
The last chapters, “Linear Systems,”
“Proofs of the Main Theorems,” and “Dif-
ferential Equations on Manifolds,” remain
mostly unrevised with the notable addition
of a section on Sturm’s theory: the inter-
lacing of zeros, the comparison theorem,
and the Sturm–Liouville boundary value
problem.
The appendix, “Sample Examination
Problems,” has been expanded to include
a sample examination from Moscow State
University and many new supplementary
problems.
Every serious student of ordinary differ-
ential equations should read this book. In
the U.S. system, it is an excellent text for an
introductory graduate course. While most
of the presentation is not out of the reach
of undergraduates, it will be unlikely to
replace the mathematics cookbook in the
standard sophomore introduction to ordi-
nary differential equations. Also, most un-
dergraduate students do not have time in
their schedules to take a second course on or-
dinary differential equations. On the other
hand, the U.S. system regularly produces
students who overcome their sophomore
introduction to differential equations by
“complete automatization” (perhaps with
some insight provided by Professor Arnold)
and eventually gain a complete understand-
ing of this wonderful and useful subject.
CARMEN CHICONE
University of Missouri
 BOOK REVIEWS
Modeling and Computation of Boundary-
Layer Flows. Second Edition. By Tuncer Ce-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
beci and Jean Cousteix. Springer-Verlag, New
York, 2005. $159.00. xvi+502 pp., hardcover.
ISBN 3-540-24459-X.
Boundary layer theory has a long his-
tory; its mathematical formulation dates
back to Ludwig Prandtl (1904) and is fa-
mously known to every postgraduate stu-
dent and researcher in fluid dynamics as the
boundary layer approximations of the three-
dimensional Navier–Stokes (NS) equations.
The history is indeed very rich in a variety
of problems of practical relevance for which,
in the days before modern CFD, solutions
had to be found, the main ones being the
fluid dynamic drag of a body and its coun-
terpart the fluid dynamic lift of a body. The
latter, of course, is of particular interest to
aeronautical engineers.
Although several books and treatises over
the years have provided fruitful descrip-
tions of this theory to various depths, the
book under review provides depths previ-
ously unmatched in a single volume (albeit
while dealing with flows of interest only to
aeronautical engineers). It is this aspect
that should immediately attract a potential
buyer of this book.
The book begins invitingly with case
studies of (a) skin-friction drag reduction
through shaping of airfoils with the aim
of delaying the onset of transition and
through boundary layer suction, and (b)
prediction of the maximum lift-coefficient of
multi-element wings. This telling introduc-
tion not only does justice to the authors’
long-standing contributions to aerodynam-
ics research but also creates curiosity in a
reader about the mathematical nuances of
the developments in this theory that are to
follow in the rest of the book.
In this 13-chapter book, Chapters 2 to
9 are devoted to incompressible flows and
Chapters 10 to 13 to compressible flows.
Laminar, transitional, and turbulent flows
are considered in two and three dimensions.
Both the wall boundary layers as well as free
shear layers are considered. These chapters
comprehensively illustrate how, when, and
why similarity, local nonsimilarity, integral,
and finite-difference (Keller’s box method
is preferred) methods are to be employed
337
to solve boundary layer equations. Transi-
tional boundary layers are analyzed through
linear stability (Orr–Sommerfeld equation)
as well as through application of eigenvalue
methods such as the en method. Where ap-
plicable, empirical correlations for the on-
set and end of the transition regime are
also given that can be profitably used in
integral as well as finite-difference solutions
to the entire laminar, transitional through
to turbulent development of a wall bound-
ary layer. Chapters 6, 7, and 12 deal with
turbulent flows where approaches to turbu-
lence modeling through zero-equation (or
mixing length) models as well as through
models based on transport equations (e.g.,
one-equation models based on νt -equation
and two-equation models based on k- and
k-ω variables) have been listed. These are
used to model isotropic turbulent viscosity
µt in the Stokes-like relation
 


∂ui ∂uj 2
− ρ u i u j = µt + − δij k,
∂xj ∂xi 3
where coplanarity between stress and strain
is indicated. While this approach is quite
satisfactory in most long-and-thin bound-
ary layer flows, there are more complex
situations where transport equations for



−ρui uj must be solved directly. Such stress
equation models have also been listed. Mod-
els of increasing complexity are required to
capture effects of surface curvature, rapidly
varying pressure gradients and suction, and
other effects where the viscous inner layer of
a turbulent boundary layer is influenced by
events in the outer layers. In this respect,
description of the great near-universality
of the inner and outer layers of a turbu-
lent boundary layer that was established by
(say) the late 1950s through phenomeno-
logical arguments is a most welcome fea-
ture. Another feature that will be partic-
ularly welcomed by students and teachers
is the large number of end-of-chapter prob-
lems whose solutions will be an enriching
experience. To aid solutions, a CD-ROM
containing computer programs is provided.
Chapters 9 and 13 are devoted to interac-
tive boundary layer theory (IBLT). Conven-
tionally, the boundary layer equations for a
wall flow are solved for a prescribed pressure
gradient that is evaluated from inviscid flow
theory. In real viscous flows, however, the
 338 BOOK REVIEWS
true pressure gradient deviates from that
derived from inviscid flow theory. This is
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
particularly so in the vicinity of the sep-
aration point and near the trailing edges
of plates and airfoils. The IBLT resolves
this problem by matching the displacement
thickness derived from viscous and inviscid
flow equations. This chapter describes the
required adaptations to the BPL2 program
along with several applications.
Chapters 11, 12, and 14 deal with lam-
inar, turbulent, and transitional compress-
ible flows. Here treatments of shock-shear
layer interactions and formulations of tur-
bulent transport equations allowing for den-
sity fluctuations and for surface roughness
are special features.
It is nice to see a fine book in print that
every fluid dynamics student or researcher
would wish to possess, particularly because
there are a large number of comparisons
between theoretical predictions and exper-
imental data. It is a fine source of all that
is relevant in boundary layer theory both
historically as well as in terms of contempo-
rary and future fluid dynamic applications.
Although not discussed explicitly, the meth-
ods of this book can be extended to heat
and mass transfer in boundary layers as
well as to internal (ducted) boundary layer
development. Overall, this is a must-buy
for applied mathematicians as well as for
engineers.
A. W. DATE
Indian Institute of Technology
Bombay
Two-Point Boundary Value Problems:
Lower and Upper Solutions. By Colette De
Coster and Patrick Habets. Elsevier B.V., Amster-
dam, 2006. $182.00. xii+502 pp., hardcover.
ISBN 0-444-52200-X.
The theory of lower and upper solutions
has a long history in connection with the
attempt to find solution methods for second
order boundary value problems (BVPs).
The purpose of this book is twofold: first,
to introduce the method for ordinary differ-
ential equations (ODEs), and second, to de-
scribe some of the more recent and involved
work in this area. I believe that it succeeds
admirably on both counts. Although the
methods presented are often motivated by
problems in partial differential equations,
the upper and lower solution method has
developed into a cohesive theory by itself.
The problem of finding a solution to a BVP
is replaced by the ostensibly simpler prob-
lem of finding lower and upper solutions.
How does one recognize that certain as-
sumptions will lead to obtaining lower and
upper solutions and then, if so, how does
one go about obtaining them?
The earliest attempts at establishing so-
lutions to an ODE homogeneous Dirichlet
problem, due to Picard in 1893, led to the
monotone iterative method. Some of these
ideas may also be found in work of Perron
(1915) and Müller (1926) (for systems). A
major breakthrough came with a paper by
Scorza Dragoni in 1931, who considered the
Dirichlet BVP
u

= f (t, u, u
), u(a) = A, u(b) = B,
in which the existence of solutions α, β ∈
C 2 ([a, b]) which satisfy α(t) ≤ β(t) on [a, b],
and in addition satisfy certain differential
inequalities, is assumed. The next improve-
ment or extension of the method came with
the work of Nagumo in 1937 in which (what
is now called) a Nagumo condition was in-
troduced. This allowed one to enlarge the
class of functions by the assumption that the
differential inequalities are satisfied along
the functions α and β, whereas in Scorza
Dragoni’s papers the assumptions were re-
quired for sets of values of u
. The authors
then go into great detail about various other
historical aspects in the development of the
theory. Many additional insightful remarks
and comments are made in this introduction
to the method, on topics such as maximal
and minimal solutions, generalized Nagumo
conditions, bounding functions, degree the-
ory, variational methods, existence of posi-
tive solutions, and more!
After this introduction, the authors con-
sider in Chapters 1 and 2 the periodic
BVP and the separated BVP, respectively.
They then develop the relationship with
degree theory and variational methods, in-
cluding the minimax method as well as the
monotone iterative method and its abstract
formulation, in three very long and care-
fully presented chapters. Chapter 6 treats
 BOOK REVIEWS
parametric multiplicity problems for the
Lienard and the Rayleigh equations (the
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
Ambrosetti–Prodi problem) and further ex-
tensions of them. Resonance and nonres-
onance problems are studied in Chapter
7, along with Landesman–Lazer conditions
and results based on the Fredholm alterna-
tive. The existence of positive solutions to
the Dirichlet problem



u + f (t, u) = 0, u(a) = 0 = u(b)
is considered in Chapter 8. Various types of
singular problems are considered in Chapter
9, and in Chapter 10 the lower and upper
solution method is applied to singular per-
turbations.
In addition to the very well presented
introduction, the authors include an ex-
tensive list of bibliographical notes to each
of the chapters and also provide an ap-
pendix which includes basic statements on
degree theory, variational methods, spectral
results, and the maximum/antimaximum
principles. There are more than 300 refer-
ences to the literature as well.
This is a valuable book because of the
exhaustive treatment of the lower and up-
per solution method in its many forms, as
well as the historical details and manifold
developments in the theory since its incep-
tion. Many illustrative examples are also
presented in which lower and upper solu-
tions are constructed. This should be a
very useful book for researchers in the area
of BVPs and applications.
LYNN ERBE
University of Nebraska–Lincoln
Metric Spaces. By Satish Shirali and Harkri-
shan Vasudeva. Springer-Verlag, London, 2006.
$44.95. viii+222 pp., softcover. ISBN 1-85233-
922-5.
This book introduces the fundamentals of
analysis in metric spaces. It’s written in
a very spare theorem-proof-example style;
has illustrative examples and exercises;
spends little time on discussion, develop-
ment of intuition, or substantial applica-
tions; begins by stating that the abstract
postulational method has a vital role in
modern mathematics; implicitly assumes
339
this is the way to teach mathematics. Use-
ful resource for writing lectures? Certainly.
Good way to read and learn about analysis?
You decide.
DONALD ESTEP
Colorado State University
Triangulations and Applications. By Oyvind
Hjelle and Morten Daehlen. Springer-Verlag,
Berlin, 2006. $64.95. xi+234 pp., hardcover.
ISBN 3-540-33260-X.
Let me start with what I like least about the
book: the title. I find it fairly misleading: in
the area of visualization (the book appears
in a series called Mathematics and Visu-
alization), the term triangulation typically
refers to a 3D mesh. The book, however,
is entirely restricted to triangulations of 2D
planar domains.
For those domains, however, it does pro-
vide a detailed and adequate treatment. A
brief overview of the contents:
• basic facts and algorithms about tri-
angles and triangulations;
• graphs and data structures;
• Delaunay triangulations and Voronoi
diagrams: theory and algorithms;
• data dependent triangulations;
• constrained triangulations;
• refinement and mesh generation;
• piecewise linear least squares approx-
imation;
• introduction to the companion soft-
ware.
Triangulations are well understood in the
field of CG (computational geometry) [1]—
one might ask what another text could add.
This book is definitely not a rehash of CG
concepts. CG is mainly concerned with the
order of complexity of geometric algorithms
or uniqueness of solutions. These concepts
are covered briefly here as well, but the
main approach is totally different, and in
a sense is complementary to the CG view
of triangulations. The overall flavor of the
book reflects the use of triangulations for
the description of planar objects and as the
domain for bivariate functions.
Triangulated planar objects are meshes;
important topics such as mesh editing or
 340 BOOK REVIEWS
mesh refinement are covered; mesh coars-
ening and mesh smoothing are not.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
One of the main strengths of the book is
the coverage of bivariate functions over tri-
angulations. These functions are piecewise
linear over the triangulation and are used
to approximate bivariate functions as well
as discrete data. This problem gives rise to
a kind of triangulation which is not con-
sidered in CG: the data dependent triangu-
lation. For this, triangles are not optimal
with respect to a Delaunay criterion but
rather they are constructed such that they
“line up” optimally with a given function.
The chapter on least squares approxi-
mations is excellent. It shows how to re-
alistically approximate huge data sets by
a manageable piecewise linear surface. A
mix of interpolation and approximation is
also covered which is important in several
applications.
The book is meant as a text for a graduate
class on triangulations. Such a class would
fit into the area of scientific computing or
numerical analysis. A course text should
have problems and exercises, and indeed
there are about eight problems at the end
of each chapter, with a good mix of theoret-
ical problems and programming exercises.
For the programming part, students (as
well as other readers) do not have to start
from scratch: a complete software pack-
age is available from a companion web site.
There one can find descriptions of the ba-
sic data structures as well as algorithms
for constrained and unconstrained Delau-
nay triangulations, a triangulation editor,
and various query tools. This reviewer did
not try it out, but knowing that it is based
on highly successful work by the Norwegian
SINTEF organization, there is little doubt
that it is a good product.
REFERENCE
[1] M. de Berg, M. van Krefeld, M. Over-
mars, and O. Schwarzkopf, Compu-
tational Geometry: Theory and Ap-
plications, 2nd ed., Springer-Verlag,
Berlin, 2000.
GERALD FARIN
Arizona State University
Large-Eddy Simulations of Turbulence. By
Marcel Lesieur, Olivier Metaı́s , and Pierre Comte.
Cambridge University Press, Cambridge, UK,
2005. $65.00. xii+219 pp., hardcover. ISBN
0-521-78124-8.
Hydrodynamic turbulence is often referred
to as the last open problem of classical me-
chanics. Despite nearly a century of efforts
by the best minds, many questions remain
open, and considerable progress has only
been achieved for the special case of homo-
geneous and isotropic turbulence in three
dimensions [15, 16, 4]. To a lesser extent,
there has also been some progress toward
understanding homogeneous and isotropic
turbulence in two dimensions [11, 20, 5].
On the theoretical front there has been very
limited success in understanding inhomoge-
neous turbulence, compressible turbulence,
and geophysical turbulence.
Our intuitive understanding of turbu-
lence goes back to Richardson [18], Kol-
mogorov [9, 8], and Batchelor [1], who con-
jectured that if energy is injected into the
system by forcing at large scales, then for
sufficiently large Reynolds number, hydro-
dynamic instability results in the transfer
of energy to smaller scales, and for suffi-
ciently small scales this energy is dissipated
by viscosity. In between the forcing length
scale 0 , where the energy comes in, and the
dissipation length scale η, where the energy
comes out, there is a range of scales, where
the energy is cascaded by local nonlinear
interactions from wavenumber to wavenum-
ber down the range. This region is called
the inertial range, and it was conjectured
by Kolmogorov that it is governed by uni-
versal statistical principles. The conjecture
was motivated by the notion that at length
scales in the inertial range, the system for-
gets how the energy gets there, and the
only dynamics in the inertial range is the
transfer of energy to smaller scales. This
conjecture leads to predictions about the
energy spectrum E(k) of turbulence as well
as the scaling properties of the statistical
moments of velocity differences.
Kolmogorov’s idea remained a conjecture
until 1962, when it was confirmed exper-
imentally by measurements of the veloc-
ity of deep oceanic currents [6]. It was in
that same year that Kolmogorov [10] and
 BOOK REVIEWS
Oboukhov [17] suggested that there may
be a small departure from his original pre-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
dictions, and further work over the next
decades showed that this was indeed true
[19]. So, one of the open questions that
has received a lot of theoretical attention
is, Why is the energy spectrum E(k) of
turbulence in the inertial range in such sig-
nificant agreement with the predictions of
Kolmogorov’s 1941 theory (K41), and what
is the origin of the deviations from that
theory?
Now, one can argue that we already have
a pretty good theory of hydrodynamic tur-
bulence: the Navier–Stokes equations! All
the physics of the Kolmogorov energy cas-
cade can come from the numerical solu-
tion of the Navier–Stokes equations! The
problem is that a numerical solution of the
Navier–Stokes equations by direct numeri-
cal simulation is not possible for Reynolds
numbers large enough to be of any practi-
cal interest. To the best of my knowledge,
the state of the art in direct numerical sim-
ulations is 40963 resolution [7], performed
by one of the largest supercomputers in
the world, the Earth Simulator in Japan.
Furthermore, for geophysical fluid dynam-
ics and engineering applications one has to
go beyond the Navier–Stokes equations and
consider more accurate models that include
temperature and density. A practical al-
ternative to direct numerical simulation is
large eddy simulation (LES), the topic of
the book under review. The essential idea
is that in practical situations we are inter-
ested in the effect of turbulence on fluid
motions at large scales, so we go ahead and
model the nonlinear dynamics at smaller
scales so that it is not necessary to resolve
them numerically.
The book begins in the preface with a
delightful historical introduction by Jim Ri-
ley. The first chapter gives a very concise
overview of the Kolmogorov theory as well
as a conceptual introduction to LES in gen-
eral. The second chapter gives a very inter-
esting discussion of vortex dynamics which
will be of interest to a wider audience of
turbulence researchers. Emphasis is given
to criteria that can be used to characterize
coherent structures in turbulence. In ad-
dition to researchers of three-dimensional
turbulence, this topic is also relevant to the
341
investigation of two-dimensional turbulence
as well as geophysical turbulence. Chapter
3 discusses the traditional LES models. An-
other chapter of general interest is Chap-
ter 4, which discusses spectral LES models
such as EDQNM and RNG. These models
are often used, not only for numerical simu-
lations, but also for theoretical arguments,
as in, for example, [14]. There’s also an in-
teresting discussion of the famous “bump”
in the inertial range energy spectrum, for
which a variety of explanations have been
suggested [2, 12]. The remaining chapters
discuss various more modern LES models
and a wide range of engineering and indus-
trial applications of these models. Chap-
ter 5 discusses models for inhomogeneous
turbulence, Chapter 6 discusses structure
function models, and Chapter 7 discusses
models for compressible turbulence. Com-
pressible turbulence is of unique interest
due to its applications in aerodynamics and
aerospace engineering. Finally, Chapter 8
discusses geophysical fluid dynamics, begin-
ning with a very nice conceptual introduc-
tion, and emphasizes the modeling of storm
formation.
For graduate students, I should note that
the book does not give a comprehensive
introduction to the theory of turbulence,
Chapter 1 notwithstanding. One should
also study other textbooks, starting from
Frisch [3] and Lesieur [13], to gain a deeper
appreciation of what is understood and
what is not understood in this very excit-
ing research area, before specializing into
numerics. The book also doesn’t discuss
the numerical implementation of the cor-
responding simulation codes. On the other
hand, the objective of the book, as stated in
the introduction, is to present in detail a va-
riety of LES models, and to give the reader
a thorough understanding of turbulence dy-
namics through numerical results obtained
by these models. The authors have done an
excellent job in achieving these objectives.
Furthermore, they have included material,
such as in Chapter 7, that hasn’t been re-
viewed previously in the literature, as well
as material of general interest to the com-
munity, such as in Chapters 2 and 4. With
its impressive bibliography, it will be an
invaluable resource for students who want
to study the literature of this field, as well
 342 BOOK REVIEWS
as researchers who are currently working in
LES.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
REFERENCES
[1] G. Batchelor, Kolmogorov’s theory of
locally isotropic turbulence, Proc.
Camb. Phil. Soc., 43 (1947), pp. 533–
559.
[2] G. Falkovich, Bottleneck phenomenon in
developed turbulence, Phys. Fluids, 6
(1994), pp. 1411–1414.
[3] U. Frisch, Turbulence: The legacy of
A. N. Kolmogorov, Cambridge Uni-
versity Press, Cambridge, UK, 1995.
[4] M. Giles, Anomalous scaling in homoge-
neous isotropic turbulence, J. Phys.
A, 34 (2001), pp. 4389–4435.
[5] E. Gkioulekas and K. K. Tung, Re-
cent developments in understand-
ing two-dimensional turbulence and
the Nastrom-Gage spectrum, J. Low
Temp. Phys., 145 (2006), pp. 25–57.
[6] H. Grant, R. Stewart, and A. Moil-
liet, Turbulence spectra from a tidal
channel, J. Fluid. Mech., 12 (1962),
pp. 241–263.
[7] Y. Kaneda, T. Ishihara, M. Yokokawa,
K. Itakura, and A. Uno, Energy dis-
sipation rate and energy spectrum in
high resolution direct numerical simu-
lations of turbulence in a periodic box,
Phys. Fluids, 15 (2003), pp. L21–L24.
[8] A. Kolmogorov, Dissipation of energy in
the locally isotropic turbulence, Dokl.
Akad. Nauk. SSSR, 32 (1941), pp. 16–
18 (in Russian). English translation
in Proc. Roy. Soc. London Ser. A, 434
(1991), pp. 15–17.
[9] A. Kolmogorov, The local structure of
turbulence in incompressible viscous
fluid for very large Reynolds num-
bers, Dokl. Akad. Nauk. SSSR, 30
(1941), pp. 301–305 (in Russian). En-
glish translation in Proc. Roy. Soc.
London Ser. A, 434 (1991), pp. 9–13.
[10] A. Kolmogorov, A refinement of pre-
vious hypotheses concerning the local
structure of turbulence in a viscous
incompressible fluid at high Reynolds
number, J. Fluid. Mech., 13 (1962),
pp. 82–95.
[11] R. Kraichnan and D. Montgomery,
Two dimensional turbulence, Rep.
Prog. Phys., 43 (1980), pp. 547–619.
[12] S. Kurien, M. Taylor, and T. Mat-
sumoto, Cascade time-scales of en-
ergy and helicity in homogeneous, iso-
tropic turbulence, Phys. Rev. E, 69
(2004), article 066313.
[13] M. Lesieur, Turbulence in fluids, Kluwer
Academic, Dordrecht, The Nether-
lands, 1990.
[14] V. S. L’vov, A. Pomyalov, and I. Pro-
caccia, Quasi-Gaussian statistics of
hydrodynamic turbulence in 3/4 + 
dimensions, Phys. Rev. Lett., 89
(2002), article 064501.
[15] V. S. L’vov and I. Procaccia, Hydro-
dynamic turbulence: A 19th century
problem with a challenge for the 21st
century, in Turbulence Modeling and
Vortex Dynamics (Istanbul, 1996),
O. Boratav, A. Eden, and A. Erzan,
eds., Lecture Notes in Phys. 491,
Springer-Verlag, Berlin, 1997, pp. 1–
16.
[16] V. L’vov and I. Procaccia, Analytic cal-
culation of the anomalous exponents
in turbulence: Using the fusion rules
to flush out a small parameter, Phys.
Rev. E, 62 (2000), pp. 8037–8057.
[17] A. Oboukhov, Some specific features of
the atmospheric turbulence, J. Fluid.
Mech., 13 (1962), pp. 77–81.
[18] L. Richardson, Weather prediction by
numerical processes, Cambridge Uni-
versity Press, Cambridge, UK, 1922.
[19] K. Sreenivasan and R. Antonia, The
phenomenology of small-scale turbu-
lence, Ann. Rev. Fluid Mech., 29
(1997), pp. 435–472.
[20] P. Tabeling, Two dimensional turbu-
lence: A physicist approach, Phys.
Rep., 362 (2002), pp. 1–62.
ELEFTHERIOS GKIOULEKAS
University of Central Florida
A Guide to MATLAB for Beginners and
Experienced Users. Second Edition. By
Brian R. Hunt, Ronald L. Lipsman, and Jonathan M.
Rosenberg. Cambridge University Press, New
York, 2006. $50.00. xv+309 pp., softcover.
ISBN 0-521-61565-8.
This is a readable introduction to using
MATLAB 7 (and the immediate prede-
cessor versions of MATLAB). Much of the
book is devoted to describing how to use the
various computational facilities provided by
MATLAB and only a small proportion to
the language itself and to its mathematical
software library. The authors assume from
the start that the reader has access to MAT-
LAB, Simulink, and the Symbolic Toolbox,
 BOOK REVIEWS
which is essentially the content of the MAT-
LAB and Simulink Student Version. There
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
are also references to and examples of the
use of some of MATLAB’s other toolboxes,
but these are less likely to be available in
the classroom. As promised in the book’s
title, it switches frequently between intro-
ductory and advanced material, sometimes
without warning.
While reading this book I compared it for
content with another well-known textbook/
manual on MATLAB 7 by Higham and
Higham (H & H) [1]. Of course, both
texts are already, marginally, out of date as
MATLAB is now at version R2006a, and
the next release, R2006b, will be available
before this review is published.
These two books could hardly be more
different. Indeed, they could be viewed as
complementary. H & H’s text is more nar-
rowly focused, is somewhat longer, and has
a much deeper coverage of the MATLAB
language and its mathematical software li-
brary. In the book under review (H, L &
R), the symbolic approach to solving prob-
lems often takes precedence over numerical
approaches, whereas in H & H discussion of
the symbolic toolbox is restricted to one (of
the 22) chapters. In H, L & R, a chapter is
devoted to Simulink and there is discussion
of it elsewhere in the text, whereas H & H do
not appear to mention it. H, L & R include
one chapter and some earlier discussion of
publish and related new commands intro-
duced in MATLAB 7, and another chapter
on the construction of GUIs; H & H mention
some of these topics briefly but essentially
only to say that they are outside the scope
of their discussion.
Features of the new book include three
practice sets with complete answers, a long,
and mainly straightforward, chapter on ap-
plications and modeling, a chapter on trou-
bleshooting and debugging, and a glossary
of MATLAB commands. The practice sets
give a flavor of the content and level of
the book: Algebra and Arithmetic; Cal-
culus, Graphics, and Linear Algebra; and
Developing your MATLAB Skills. The ap-
plications and modeling chapter digs a lit-
tle deeper, including showing how to use
both Simulink and the MATLAB built-in
function ode45 to solve a simple ordinary
differential equation. It is also shown how
343
to set up and solve finite difference equa-
tions for the heat equation with variable
conductivity and, alternatively, how to use
the built-in MATLAB function pdepe to
solve the same problem.
REFERENCE
[1] D. J. Higham and N. J. Higham, MAT-
LAB Guide, 2nd ed., SIAM, Philadel-
phia, 2005.
IAN GLADWELL
Southern Methodist University
Sturm–Liouville Theory: Past and Pre-
sent. Edited by Werner O. Amrein, Andreas M.
Hinz, and David B. Pearson. Birkhäuser, Basel,
2005. $89.95. xx+335 pp., hardcover. ISBN
3-7643-7066-1.
When I first saw this book on display at the
Joint Mathematics Meeting, I immediately
added it to my mental list of “must reads”
even though I did not stop to look at the con-
tents. So when the opportunity to review
it arose, I jumped at the chance since now
“must” would become “have-to.” I must
confess a little disappointment when I first
opened the book. I was expecting a mono-
graph detailing the history, significance, and
intertwining of the Sturm–Liouville story in
one continuous narrative, but instead found
a collection of essays on various aspects of
the problem. My disappointment was ex-
tremely short lived, however, for what we
have here is a collection of important pa-
pers showing the impact that Sturm, and
Sturm–Liouville theory in particular, has
had on the direction of mathematical re-
search in the last 150 plus years. It is the
fundamental papers by Sturm [10, 11, 12]
and his joint paper with Liouville [13] that
form the basis for what we refer to today
as Sturm–Liouville theory.
In its simplest form, the basic Sturm–
Liouville problem consists of the differential
equation
(1) (py
)
+ λqy = 0
together with some boundary conditions,
for example, the Dirichlet
(2) y(0) = 0 and y(L) = 0,
 344 BOOK REVIEWS
the Neumann



Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
(3) (py )(0) = 0 and (py )(L) = 0,
or some other type of boundary condition.
The problem then is to find values of the
parameter λ for which the problem (1) and
(2) (or (1) and (3)) has nontrivial solutions.
In commemoration of the 200th year of
the birth of Sturm, a conference took place
at the University of Geneva (Switzerland)
from September 15 to 19, 2003. The sixty
participants represented sixteen different
countries. The first part of the conference
was devoted to Sturm’s contributions to
mathematics and physics; the second part
focused on Sturm–Liouville theory and in-
cluded general presentations of a survey
nature as well as some on recent research in
the field. The book under review is based
on the second part of the conference and
contains six survey papers followed by six
articles devoted more to topics of current
research interest. A separate volume of
Sturm’s complete works is being prepared
for publication [8].
In “Sturm’s 1836 Oscillation Results,
Evolution of the Theory,” Hinton gives a
detailed history and discussion of the Sturm
comparison theorem and the origin of os-
cillation theory for the second order linear
equation
(4) (py
)
+ qy = 0,
including some of the basic results of Fite,
Wintner, and Leighton. The Sturm com-
parison theorem simply states that if (4) is
oscillatory,1
Q≥q and p ≥ P,
then the equation
(P y
)
+ Qy = 0
is oscillatory. Hinton also shows the im-
pact of Sturm’s oscillation and comparison
results on further directions in research.
While this survey at times seems to pro-
ceed at breath-taking speed, touching on
differential equations in the complex do-
main, higher order equations, systems, and
1 Recall that for a second order linear equa-
tion, if one solution oscillates, then all solutions
are oscillatory.
boundary conditions containing a param-
eter, Hinton tends to focus on just the
ground-breaking results, and gives sufficient
references to survey monographs so that in-
terested readers can explore some of these
ideas in detail on their own.2
Simon discusses Sturm theory for differ-
ence equations in his paper “Sturm Oscilla-
tion and Comparison Theorems,” and as he
points out, it has long been believed that
Sturm obtained his famous comparison re-
sults for difference equations as well as for
differential equations. In fact, some believe
that Sturm may have discovered these re-
sults for difference equations before doing
so for differential equations (see Reid [9]).
The first place that the difference equation
results appear in print is usually attributed
to Bôcher [2, 3] in 1898.
While Everitt in “Charles Sturm and the
Development of Sturm–Liouville Theory in
the Years 1900 to 1950” issues a disclaimer
that his paper not be “counted as a history
of Sturm–Liouville theory for the period
1900 to 1950” due to a lack of sufficient
references to many mathematicians in his
bibliography, what he presents is in fact an
eloquent description of the development of
Sturm–Liouville theory in that time period
and its influence on the direction and devel-
opment of modern analysis. In particular,
he documents the work of Weyl, Dixon,
Stone, and Titchmarsh during this period.
In each case, Everitt details concisely the
significant contributions that each of these
important figures made to the advancement
of Sturm–Liouville theory beginning with
the classic paper of Weyl [15], who intro-
duced the famous limit-point/limit-circle
classification of solutions.3
Weidmann (“Spectral Theory of Sturm–
Liouville Operators. Approximation by
2 One easily accessible source on oscillation
theory for linear equations, for example, is the
book by Swanson [14].
3 For a discussion of the limit-point/limit-
circle problem, the Weyl alternative, the exten-
sion of the limit-point/limit-circle problem to
nonlinear equations, and the relationship be-
tween the limit-point/limit-circle problem and
other asymptotic properties of solutions, see the
monograph by Bartušek, Došlá, and Graef [1].
 BOOK REVIEWS
Regular Problems”) gives a broad overview
of the self-adjointness and spectral theory
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
of Sturm–Liouville operators with particu-
lar attention to maximal and minimal oper-
ators, the Weyl alternative, and deficiency
indices. Of special interest here is the ap-
proximation of the spectral properties of
singular problems by using regular prob-
lems and appropriate limiting arguments.
In “Spectral Theory of Sturm–Liouville
Operators on Infinite Intervals: A Review of
Recent Developments,” Last gives a survey
of recent work on spectral theory of Sturm–
Liouville operators on infinite intervals over
the last thirty years in both the continuous
and the discrete cases. The discussion of the
location of the spectrum and their spectral
measures is especially of interest here.
Gilbert examines the connection between
asymptotic behavior of solutions of singular
Sturm–Liouville operators and the spectral
properties of corresponding self-adjoint op-
erators in her paper “Asymptotic Methods
in the Spectral Analysis of Sturm–Liouville
Operators.” Her approach makes use of the
notion of subordinacy of solutions.
The paper by Bennewitz and Everitt,
“The Titchmarsh–Weyl Eigenfunction Ex-
pansion Theorem for Sturm–Liouville Dif-
ferential Equations,” gives the histori-
cal background of the Titchmarsh–Weyl
eigenfunction expansion results for Sturm–
Liouville differential equations. They give
a new version of the expansion theorem
that includes both the continuous and the
discrete cases of the spectrum. They make
extensive use of the Titchmarsh–Weyl m-
coefficient.
The paper “Sturm’s Theorems on Zero
Sets in Nonlinear Parabolic Equations” by
Galaktionov and Harwin is particularly in-
teresting from the point of view that many
people who work on ordinary differential
equations may be unaware of Sturm’s re-
sults [12] on the zeros of solutions of partial
differential equations. As the authors point
out, this fact was essentially ignored un-
til a paper by Pólya [7] appeared in 1933.
The authors of this contribution, in par-
ticular, discuss the situation for parabolic
equations and survey results from the last
quarter century.
Chen, in “A Survey of Nonlinear Sturm–
Liouville Equations,” discusses nonlinear
345
Sturm–Liouville problems beginning with
Nehari’s 1960 paper [6]. The emphasis here
is on what are called “nodal sets” as they
relate to Sturm’s oscillation results for lin-
ear problems. Although it is not mentioned
here, in recent years the basic oscillation
and comparison results of Sturm have been
extended in a variety of directions to non-
linear problems, for example, to nonlinear
differential equations, nonlinear difference
equations, equation with delays and neutral
terms, and even to higher order nonlinear
differential and difference equations, where
the complexity of the notation due to the
generality of the equations tends to obscure
the results (see, for example, [5]). Most
recently, the comparison results have been
extended to nonlinear dynamic equations
on time scales.
The relationship between the bound-
ary conditions and the spectra of Sturm–
Liouville operators is discussed by del Río
in “Boundary Conditions and Spectra of
Sturm–Liouville Operators.” This relates
back to the work of Weyl [15], who showed
the stability of the essential spectrum with
respect to changes in the boundary condi-
tions.
“Uniqueness of the Matrix Sturm–
Liouville Equation Given a Part of the Mon-
odromy Matrix, and Borg Type Results” by
Malamud is a survey of recent results on in-
verse spectral theory based on the approach
used by Borg [4]. This is an important and
expanding area of research in which infor-
mation about the spectrum is used to obtain
the Sturm–Liouville equation itself. The in-
terest in this article is in matrix equations
rather than the scalar case.
The last paper in this volume, “A Cata-
logue of Sturm–Liouville Differential Equa-
tions” by Everitt, is unique in that it is
a collection of more than fifty examples
of Sturm–Liouville equations along with a
classification of their endpoints and infor-
mation about the location of their eigenval-
ues and the spectra of their corresponding
differential operators. What a wealth of in-
formation in an easily accessible format. In
many cases their connection to special func-
tions is also indicated. Even better, this
article begins with a very nicely written
concise summary of Sturm–Liouville the-
ory. This in itself is worth the price of the
 346 BOOK REVIEWS
monograph. In cases where the equation
is linked to a physical application, this is
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
noted as well. Known references to numer-
ical approximations of the eigenvalues are
also given.
As a final note, the elegant introduction
by David Pearson gives some background
on Sturm himself, his collaboration with
Liouville, and the beautiful connection be-
tween Sturm–Liouville theory and quantum
mechanics. It is highly worthwhile reading
all on its own.
This volume has something for every-
one who has any interest in the impact of
Charles François Sturm’s fundamental work
on current mathematical research.
REFERENCES
[1] M. Bartušek, Z. Došlá, and J. R.
Graef, The Nonlinear Limit-Point/
Limit-Circle Problem, Birkhäuser
Boston, Boston, MA, 2004.
[2] M. Bôcher, The theorems of oscillation
of Sturm and Klein (First Paper),
Bull. Amer. Math. Soc., 4 (1898), pp.
295–313.
[3] M. Bôcher, The theorems of oscillation
of Sturm and Klein (Second Paper),
Bull. Amer. Math. Soc., 4 (1898), pp.
365–376.
[4] G. Borg, Eine Umkehrung der Sturm-
Liouvilleschen Eigenwertaufgabe,
Acta Math., 78 (1946), pp. 1–96.
[5] J. R. Graef, A. Miciano-Cariño, and
C. Qian, A general comparison re-
sult for higher order nonlinear differ-
ence equations with deviating argu-
ments, J. Difference Equations Appl.,
8 (2002), pp. 1033–1052.
[6] Z. Nehari, Characteristic values associ-
ated with a class of nonlinear second-
order differential equations, Acta
Math., 105 (1961), pp. 141–175.
[7] G. Pólya, Qualitatives über Wärmeaus-
gleich, Z. Angew. Math. Mech., 13
(1933), pp. 125–128.
[8] J.-C. Pont, ed., The Collected Works of
Charles François Sturm, in prepara-
tion.
[9] W. T. Reid, A historical note on Sturmian
theory, J. Differential Equations, 20
(1976), pp. 316–320.
[10] C. Sturm, Extrait d’un Mémoire sur
l’intégration d’un système d’équa-
tions différentielles linéaires, Bull.
Sci. Math. Férussac, 12 (1829), pp.
313–322.
[11] C. Sturm, Mémoire sur les Équations
différentielles linéaires du second or-
dre, J. Math. Pures Appl., 1 (1836),
pp. 106–186.
[12] C. Sturm, Mémoire sur une classe
d’Équations à différences partielles,
J. Math. Pures Appl., 1 (1836), pp.
373–444.
[13] C. Sturm and J. Liouville, Extrait d’un
Mémoire sur le développement des
fonctions en séries dont les différents
termes sont assujettis à satisfaire
à une méme équation différentielle
linéaire, contenant un paramètre
variable, J. Math. Pures Appl., 2
(1837), pp. 220–223.
[14] C. A. Swanson, Comparison and Os-
cillation Theory of Linear Differen-
tial Equations, Academic Press, New
York, 1968.
[15] H. Weyl, Über gewöhnliche Differential-
gleichungen mit Singularitäten und
die zugehörige Entwicklung willkür-
licher Funktionen, Math. Ann., 68
(1910), pp. 220–269.
JOHN R. GRAEF
University of Tennessee at Chattanooga
Nonlinear Problems of Elasticity. Second
Edition. By Stuart S. Antman. Springer-Verlag,
New York, 2005. $89.95. xviii+831 pp., hard-
cover. ISBN 0-387-20880-1.
Nonlinear (finite) elasticity is a central
paradigm of continuum mechanics. Cor-
rectly formulated problems of the subject,
incorporating the exact geometry of de-
formation and nonlinear material response,
are typically difficult; there are many open
problems. The subject has immediate ap-
plicability to flexible structures and rubber-
like solids, and it forms a foundation for
theories of more exotic solid materials un-
dergoing finite deformations. In spite of its
classical roots (dating back to Euler and
Cauchy, for example), the nonlinear theory
was not finalized until the second half of the
twentieth century—a time when nonlinear
analysis also emerged as a mature mathe-
matical subject. The influence of elasticity
on analysis is well known [1, 2]. The ap-
plication of modern methods of analysis to
nonlinear elasticity is an ongoing, fertile en-
deavor. This brings us to the book under
review.
 BOOK REVIEWS
Antman’s impressive work is like no
other—both a comprehensive treatise on
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
nonlinear elasticity and a quintessential ex-
ample of applied nonlinear analysis. Shortly
after the appearance of the first edition,
two authoritative and enthusiastic review-
ers predicted it would become a standard
reference in this field [3, 4]. Ten years later
those predictions are now realized. A guid-
ing theme throughout is the demand for
consistent care in both the formulation and
the analysis of specific classes of problems.
The rigorous existence of solutions and the
determination of thresholds of instability
and bifurcation, classified according to non-
linear material response, are established for
numerous novel problems—in most cases
based upon the author’s own research. The
book is especially unique in its coverage of
nonlinearly elastic strings, rods, and shells
(of Cosserat type).
The second edition retains all the flavor
of the first—and more. The text has been
revised and updated. Several new sections
have been added, featuring new problems
for strings, planar rods, spatial rods, ax-
isymmetric shells, and 3-d solids, expanded
coverage of transverse symmetry of spatial
rods, expanded coverage of general theo-
ries of rods and general theories of shells,
and a new chapter on general results in 3-d
nonlinear elasticity.
This book is a “must” for researchers and
graduate students interested in nonlinear
continuum mechanics and applied analysis.
The work is scholarly and well written. The
depth and breadth of coverage is stagger-
ing, and it is not for the “faint of heart.”
Antman expresses this well in the preface:
“This book is directed toward scientists, en-
gineers, and mathematicians who wish to
see careful treatments of uncompromised
problems.”
REFERENCES
[1] C. Truesdell, The influence of elasticity
on analysis: The classic heritage, Bull.
Amer. Math. Soc. (N.S.), 9 (1983), pp.
293–310.
[2] S. S. Antman, The influence of elas-
ticity on analysis: Modern develop-
ments, Bull. Amer. Math. Soc. (N.S.),
9 (1983), pp. 267–291.
347
[3] M. Renardy, Review of Nonlinear Prob-
lems of Elasticity (S. S. Antman),
SIAM Rev., 37 (1995), p. 637.
[4] J. M. Ball, Review of Nonlinear Problems
of Elasticity (S. S. Antman), Bull.
Amer. Math. Soc. (N.S.), 33 (1996),
pp. 269–276.
TIMOTHY J. HEALEY
Cornell University
Knowledge Discovery in Proteomics. By
Igor Jurisica and Dennis Wigle. Chapman and
Hall/CRC, Boca Raton, FL, 2006. $89.95.
xxiii+318 pp., hardcover. ISBN 1-58488-439-
8.
In broad terms, researchers in the life sci-
ences are attempting to organize and under-
stand all the processes of living organisms
from the cellular to the whole organism
level. Such an aim requires the accumula-
tion of piles of data as well as novel methods
for organizing and managing all that data to
enable the extraction of useful information.
Advancements in high-throughput (HTP)
experimental techniques such as mass spec-
troscopy and the completion of major col-
laborative works like the human genome
project have provided an abundant sup-
ply of data. However, the development of
scalable methods for unifying the data and
sifting through it to make new discoveries
has been a slower process and has become
a main challenge being faced.
It is not just the amount of new data
available daily, but also the diversity of
that data that is difficult to deal with.
Biological systems are subject to various
environmental conditions, any number of
which can drastically alter the outcome of
an experiment when perturbed. Knowing
that certain proteins interact, for example,
is not sufficient for our understanding if we
do not also know important experimental
details such as stimuli, location, and time.
The issue is further complicated by difficul-
ties with automated access and querying of
diverse public databases, redundancies such
as multiple names for a single protein, and
missing and noisy data. When one thinks
of the large number of proteins within a sin-
gle cell, let alone a multicellular organism,
the scale of this problem quickly becomes
 348 BOOK REVIEWS
overwhelming and can give one the sense of
drinking from a fire hydrant.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
This is where knowledge discovery plays
an important role and motivates the timely
publication of the book by Jurisica and
Wigle. Knowledge discovery, as the authors
define it, is the process of finding useful pat-
terns in data with an emphasis on the end
product of the process, which is knowledge.
Naturally, this iterative process involves the
management of available knowledge in a
way that optimizes sharing and collabora-
tion. However, when attempting to merge
information it is difficult to determine what
information is useful and what should be
included. It is also difficult to plan ahead
because it is impossible to predict what
new methods and discoveries will be devel-
oped and lead to new data representation
requirements years from now. In Chapters
1 and 2, the authors provide an overview of
the knowledge discovery process with sev-
eral specific examples that emphasize the
usefulness of approaches that have been em-
ployed while pointing out limitations that
need to be addressed. By doing so, they
clearly express the need for defining stan-
dards for experimental data representation
that are both comprehensive and flexible,
and they motivate the discussion that fol-
lows in subsequent chapters.
The authors discuss several influential ex-
perimental and theoretical techniques em-
ployed in bioinformatics, proteomics, and
systems biology today. Chapter 3 provides
an introduction to mass spectrometry (MS)
and an overview of the state-of-the-art MS-
based experimental procedures. Some of the
data processing tools and search algorithms
used to manage and extract information
from global proteomic studies are also de-
scribed in Chapter 3. Chapters 4 and 5 go
into much more depth than the other chap-
ters and make up the majority of the content
of the book. Chapter 4 is an in-depth ex-
planation of the graph theory analysis of
protein-protein interactions (PPIs). It ex-
pounds the terminology, types of network
models, topological features, motifs, algo-
rithms, and public data sets used to infer
and study PPIs. Chapter 5 is a discussion of
HTP protein crystallization approaches and
the automated tools used to analyze and
extract information from the experimental
results. Issues with diverse chemical condi-
tions and bottlenecks in the crystallization
and analysis process are also elucidated.
Chapter 6 is a discussion of the approaches
and tools used to integrate diverse forms
of data from public databases as well as
the significant issues faced and the benefits
gained in that process. The book is con-
cluded in Chapter 7 with an overview of
systems biology and a description of how
knowledge discovery is involved in advanc-
ing various areas of research in this field.
The authors recognize the importance of
understanding how experimental and theo-
retical techniques interact in the knowledge
discovery process; and therefore they try
to strike a balance between these areas
in the book. Researchers involved in the
knowledge discovery process come from a
variety of different backgrounds including
biology, physics, mathematics, computer
science, and engineering. Therefore, in or-
der to coordinate everyone’s efforts, it is
necessary that researchers from all these
areas understand how their results and tools
will be used by others, and that they conse-
quently optimize for sharing and collabora-
tion. Throughout the text, impressive ex-
amples from the literature and the authors’
personal research experiences highlight this
key concept and point out where limitations
and major challenges still exist.
The book was written by eight contribut-
ing authors with varying degrees of expe-
rience, from Ph.D. candidates to faculty
members. As a result, there is a degree
of disconnect between chapters in terms of
writing styles as well as in the extent to
which each topic is addressed. That crit-
icism aside, this is a timely publication
and, for the most part, a successful effort
to depict many hot areas of research. I
would highly recommend the book to new
researchers, as it describes the limitations
of tools and techniques being used today
and draws attention to many open prob-
lems. I would also recommend the book to
more experienced researchers who have a
primarily theoretical or experimental bent,
as attaining the goal of unifying existing
knowledge will require a concerted effort
from everyone involved.
WILLIAM J. HEUETT
University of Colorado
 BOOK REVIEWS
Synthetic Differential Geometry. Second
Edition. By Anders Kock. Cambridge Univer-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
sity Press, Cambridge, UK, 2006. $55.00. xii+
233 pp., softcover. ISBN 0-521-68738-1.
Forty years ago it became apparent that
certain categorical methods, devised for
the study of algebraic and analytic spaces,
can also be applied to smooth (C-infinity)
spaces. The ramifications of this obser-
vation for continuum physics are still be-
ing worked out, but major advances were
recorded in the 1991 Springer book Models
for Smooth Infinitesimal Analysis by Ieke
Moerdijk and Gonzalo Reyes; undergrad-
uate textbooks on differentiable analysis
that explicitly make use of the resulting
insights were prepared by René Lavend-
homme (1996) and by John Bell (1998).
These three books, as well as the research
papers by Marta Bunge, Eduardo Dubuc,
Felipe Gago, and others, descend from the
first book to treat these developments, Syn-
thetic Differential Geometry (Cambridge,
1981) by Anders Kock; the volume under
review is a revised and augmented version
of that earlier book.
Several mathematicians have proposed
improvements of the category of finite-
dimensional smooth manifolds. Notably,
K.T. Chen and Alfred Frölicher both con-
tributed interesting proposals to volume
1174 of the Springer Lecture Notes in Math-
ematics, which records a 1982 meeting at
Buffalo. Frölicher’s admirably simple cat-
egory is the basis for the book The Con-
venient Setting of Global Analysis (AMS,
1997) by Andreas Kriegl and Peter Michor.
However, that category does not permit the
“representability” of tangent vectors and
differential forms. Chen’s category, on the
other hand, was explicitly designed to allow
representability of differential forms, but
does not include representability of tangent
vectors, nor the representability of smooth
function spaces. Remarkably, all these rep-
resentabilities and more do follow from the
“topos” formulation expounded in the book
under review. The practical significance of
representability is that many constructions
of a naively set-theoretic character are not
mere formulas, but correspond to actual
geometric objects, indeed objects with an
automatic smooth structure.
349
Not only diffeomorphisms, but paths,
functions, subspace inclusions, product pro-
jections, etc., are used in differential the-
ories, and composition of these leads ex-
plicitly to the idea of a category; but
various relevant categories arise, depending
on which objects are admitted and which
functorial constructions on them are pos-
sible. The variety of categories also arises
from the fact that useful levels of smooth-
ness (polynomial, analytic, or an intermedi-
ate level such as the one generated by bump
functions) all give rise to categories whose
general properties and geometric uses are
somewhat similar, so that a common non-
trivial theory is possible. We refer here to
the categories described by that theory as
“smooth toposes.”
Besides the usual cartesian products
(used by Galileo for the description of mo-
tion), two of the striking features of smooth
toposes are the representability (described
more fully two paragraphs below) of tan-
gent bundles and of function spaces. For
300 years mathematical scientists such as
Bernoulli, Euler, Lie, and Cartan have used
these functors and even their representabil-
ity, despite the fact that such representabil-
ity was inconsistent with the formal founda-
tions that were available. The issue in all the
classical cases is to broaden the definition
of the spaces sufficiently to permit the de-
sired representabilities, thereby streamlin-
ing constructions and computations with-
out obscuring the spaces and mappings of
interest.
By Yoneda’s lemma, representing objects
are unique, in particular the object T , which
represents the tangent bundle of any space
X as the function space X T ; T is first-order
infinitesimal in the sense that the kernel of
the bundle projection RT → R for the ring
R of reals consists of quantities of square 0.
Elementary considerations show that the
natural properties of T are “inconsistent
with classical logic,” not in any mystical
intuitive way, but just in the sense that
the lattice of subspaces is not Boolean. It
should be clear from studying this book that
it was the development of the geometry that
led to “logical” considerations such as non-
Boolean subspace lattices, double-negation
of points, explicit distinctions between the-
ories and models, etc.; researchers in this
 350 BOOK REVIEWS
field did not set out to impose a precon-
ceived version of logic onto the subject.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
In particular, the nilpotent infinitesimals
as representors for the geometric tangents
should not be confused with the infinitesi-
mals invoked in “nonstandard” systems of
logical formulas. Euler went so far as to
affirm that real numbers themselves are ra-
tios of infinitesimals; the fact that R is the
zero-preserving subspace of T T now gives
rigorous justification to his idea.
The representability of function spaces
in a smooth topos implies a unique notion
of smoothness for functionals. An example
is the smooth space of all V -valued distri-
butions of compact support on X, where
V is a space with an action of the multi-
plicative monoid R. This is just the sub-
space of the double function space V to
the power (R to the power X) consisting
of the R-homogeneous smooth functionals.
(As Kock and Reyes showed in their recent
work on the heat equation, noncompact dis-
tributions can also be handled.) But what
precisely is meant by representability of
function spaces? This special case of Daniel
Kan’s concept of an adjoint functor means
that given any three spaces B, I, S, there
are other spaces S I and S B for which maps
I × B → S correspond uniquely to maps
B → S I and also to maps I → S B ; if B,
I, S denote a body, a time interval, and a
space region, respectively, then we have the
well-known bijection between placements in
path space and paths in placement space.
These function spaces permit the calcula-
tion of the velocity field on B and the motion
of the center of mass of B, by composition
with basic functionals of differentiation and
integration.
Around 1980 it was realized that much
stronger representabilities follow automati-
cally in all the constructed examples of rel-
evant categories. It had already been clear
in 1967 that the category of first-order ODE
in a smooth topos would again be a smooth
topos (with all the representability and ex-
actness that provides); here a first-order
ODE (or vector field) on X is equivalent
to a pointed action T × X → X. Indeed
the pointed actions of any given pointed
space, with the natural morphisms of such
actions, form a category as good as the am-
bient category of spaces. But as expressed
by the property that T is an ATOM (“amaz-
ingly tiny objectified motion”), the stronger
representability implies the nonclassical re-
sult that the category of nth order ODE is
also a topos. Here n is a given map T → A
(where, for example, A = T 2 /2! is the rep-
resentor for the symmetric second tangent
bundle functor); an nth order ODE on X
is a given section of the induced X A → X T
(in the example, a given way of prolonging
any linear infinitesimal path to a quadratic
one); and solutions are morphisms in the re-
sulting category having time-like domains.
In their work on the wave equation and the
heat equation, Kock and Reyes have ex-
plicitly utilized this notion of second-order
equation as applied on infinite-dimensional
spaces of distributions.
In the above brief sketch we have omitted
many aspects, such as affine connections,
the Thom–Mather singularity classification,
the role of nonnilpotent germs, and the
Stokes and DeRham theorems for currents,
all of which have been revealed in a fresh
light by these synthetic considerations. The
“synthetic” method, developed from the ge-
ometric tradition of Sophus Lie, (1) starts
with an objective concept and (2) describes
it axiomatically, then (3) constructs models
for those axioms by using numerical coordi-
nates or even logical power sets, and finally
(4) uses both the axioms and the models
to further illuminate and make precise the
concept. An informed modification of the
axioms, or a refinement of the concept, may
prove necessary. For the objective concept
of spaces as arenas for smooth motion, one
discerns the role of all these synthetic stages
in Anders Kock’s book.
F. WILLIAM LAWVERE
SUNY Buffalo
Partial Differential Equations of Applied
Mathematics. Third Edition. By Erich Zau-
derer. Wiley-Interscience, Hoboken, NJ, 2006.
$115.00. xxviii+930 pp., hardcover. ISBN 0-
471-69073-2.
Among the many partial differential equa-
tion (PDE) textbooks for advanced under-
graduates and graduate students studying
applied mathematics or engineering, Erich
 BOOK REVIEWS
Zauderer’s text sets itself apart. On first
appearance, because of its tome-like struc-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
ture (930 pages in this new third edition),
it looks more like an intimidating reference
book, or handbook, than a textbook. But
a deeper examination reveals some impor-
tant topics and approaches that are not
common in most PDE books at this level.
(See SIAM Review [1] for a comparative
review of several PDE texts up to the year
2000.)
The book begins with a chapter on ran-
dom walks and PDEs. In this context of
probability, one visits diffusion, Brown-
ian motion, the Fokker–Planck equation,
Laplace’s equation, Green’s functions, first
order PDEs, and other hyperbolic equa-
tions associated with transition probabil-
ities. This is the only text I know that
begins this way, and this reviewer finds this
approach highly satisfying. Therefore, stu-
dents are immediately exposed to key ideas
in probability that lead to the derivation of
the standard equations we study in PDEs.
Later in the text one finds the common
derivations of the various equations, for ex-
ample, the heat equation based on an energy
balance law.
The second major feature of the text is
the coverage of perturbation and asymp-
totic methods for PDEs. This is the most
distinctive part of the text and it spans
about 170 pages (Chapters 9 and 10).
The chapter on perturbation methods in-
cludes material on regular and singular per-
turbation, boundary layers, and parabolic
approximation. The chapter on asymp-
totic methods includes equations with large
parameters, the eikonal and transport equa-
tions, scattering, the propagation of singu-
larities and discontinuities, and the asymp-
totic simplification of PDEs. Here one
will find dissipative wave equations, Burg-
ers’ equation, the Navier–Stokes equations,
KdV, and more. The important techniques
in these two chapters do not appear in any
other textbook, to my knowledge, with such
accessibility.
The chief difference between the second
and the new third editions is the inclusion
of topics on the numerical solution of PDEs.
Two new, final chapters deal with finite dif-
ference methods and finite element meth-
ods. Complementary to these chapters, at
351
the end of each of the first ten chapters
there is a section on Maple methods where
the author introduces a variety of Maple
procedures to deal with the problems intro-
duced in that chapter. The Maple programs
are available on a web site. Some scientists
will be distressed because Maple, rather
than MATLAB, is the computer algebra
system of choice. It is my impression that
many engineers prefer MATLAB for scien-
tific computation; most of my colleagues use
MATLAB for this purpose, leaving Maple
as an instructional package or as a resource
for symbolic computation. But I agree with
the author in stating that it is unwieldy to
do both—or even to include Mathematica.
Another large improvement over the pre-
vious edition is that exercises are now at the
end of each section, rather than amassed at
the end of chapters. Unlike the second edi-
tion, which has the solution to selected exer-
cises in the back of the text, the third edition
directs the reader to a web site for solutions.
The exercises are one of the strengths of this
book, and there are a number of new exer-
cises.
The other material in the text follows the
presentation in the second edition. To give a
flavor of the basic content, here are the chap-
ter topics: random walks and PDEs, first
order PDEs, classification and characteris-
tics, initial and boundary value problems
in bounded domains, integral transforms,
integral relations, Green’s functions, vari-
ational methods and maximum principles,
perturbation and asymptotic methods, and
numerical methods. It sounds like a list
of everything a budding engineer, scientist,
or applied mathematician needs to know.
Indeed, it nearly is!
Because of the text’s size, and the fact
that our own course is a one-semester, three
credit hour course, I have not adopted Za-
uderer as a textbook. However, I have of-
ten used it as a reference, particularly for
asymptotics, and as a source of exercises,
of which there are many.
If you can get beyond the massive struc-
ture of the book and you don’t mind sifting
out a selection of topics for your own course,
then Zauderer would be an excellent choice
for an upper division or graduate course in
PDEs for engineers, scientists, and applied
mathematics students. With some careful
 352 BOOK REVIEWS
choices, one might be able to use the book in
the standard “Fourier series and boundary
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
value problems” course taught to under-
graduates at most universities and many
smaller colleges. However, the exposition is
at a more mature level than that in some
of the other standard undergraduate texts;
so the instructor should be sure of the au-
dience. In summary, this third edition is a
noteworthy revision of a popular PDE book,
and it should be highly useful to students
and instructors alike as either a text or a
reference volume.
REFERENCE
[1] J. D. Logan, Featured Review: PDE
Books, Present and Future, SIAM
Rev., 42 (2000), pp. 515–522.
DAVID LOGAN
University of Nebraska
Lincoln
Yearning for the Impossible: The Surpris-
ing Truths of Mathematics. By John Still-
well. A K Peters, Wellesley, MA, 2006. $29.95.
xiv+230 pp., hardcover. ISBN 1-56881-254-X.
All too often, when browsing in the math
section at my local Barnes & Noble super-
store, I come across books with provocative
titles that suggest a good read awaits me in
an important branch of mathematics. Alas,
upon opening them up, what I find instead
is a potpourri of glib pseudohistory, curious
quotes culled from other, earlier books of
the same ilk, and, worst of all, a book writ-
ten all too obviously by a journalist with
little if any professional training in mathe-
matics. Don’t get me wrong—those authors
are often good writers, but it soon becomes
clear that they really don’t “understand”
mathematics. Such books are like orange
juice that has had all the pulp strained
out—it all goes down easily enough, sure,
but most of what makes it worth drinking
in the first place has been removed.
In his new book, Yearning for the Impos-
sible, Professor Stillwell has achieved what
many might well have come to believe to be
nearly impossible in mathematical exposi-
tion for the masses, good nutritional stuff
(with most of the pulp still present) that
is also enjoyable to read. The title comes
from Stillwell’s dual goals of first arguing
his belief that mathematics treads a narrow
path between the trivial and the impossi-
ble, and then illuminating with numerous
examples just how very narrow that path
can be. There are minor quibbles with ter-
minology that one could make (easy enough
to do with just about any book, I think),
such as on p. 6 calling 9 + 16 = 25 an equa-
tion rather than an identity. And I remain
not quite convinced that Hamilton’s discov-
ery of quaternions is quite as revolutionary
as mathematicians are prone to argue (but
I’m an engineer, so take that into account!).
Those are pretty lame complaints—I admit
it!—and they are about the “worst” I could
find.
Yearning for the Impossible succeeds, in
every topic treated, in bringing a fresh eye to
questions even mathematicians might think
have been mined in the past to boring ex-
haustion. Stillwell, however, shows there is
still a lot of gold to be found, if one only
thinks about things in a new way. In the
discussion of Chapter 1 on irrational num-
bers, for example, you’ll be amazed at how
much Stillwell extracts from beautiful argu-
ments about the square root of two. I think
his presentation of the nontermination of
the Euclidean algorithm as equivalent to
irrationality is, as a physicist might put it,
“sweet.” And the derivation of the volume
of a tetrahedron, in Chapter 4, is just plain
pretty (it provides a “practical” application
of the trick for summing geometric series).
In that and other chapters he tackles
imaginary numbers, the geometry of how
things look to the human eye, the infinitesi-
mal, the geometry of flat, curved, and peri-
odic spaces, the fourth dimension, and the
concept of infinity. In every one of these
treatments Stillwell brings new, unortho-
dox insights to his writing that will stimu-
late readers (from high schoolers to emeri-
tus professors) to think about old topics in
new, nonstale ways. The arguments in the
book are not always easy to follow on first
reading, but are logically correct and, ulti-
mately, convincing. Yearning for the Impos-
sible will be a treat for teachers, too, who are
looking for new ways to bring stimulating,
fresh examples into their courses. The book
 BOOK REVIEWS
is a great read (the publisher, A K Peters,
has done a first-rate job of packaging the
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
book, with an index that is actually useful
for finding things) and shows that Professor
Stillwell’s award of the MAA’s 2005 Chau-
venet Prize for expository writing was well
deserved indeed.
PAUL J. NAHIN
University of New Hampshire
Italian Mathematics Between the Two
World Wars. By Angelo Guerraggio and
Pietro Nastasi. Birkhäuser-Verlag, Basel, 2005.
$119.00. x+299 pp., hardcover. ISBN 3-7643-
6555-2.
This book appears in the Historical Stud-
ies Science Networks series of Birkhäuser-
Verlag. Since Poincaré in 1908 said that
the Circolo Matématico de Palermo was the
most important world mathematical orga-
nization, the contents of this book should
have considerable interest, even if Italian
mathematics actually came in third place in
European mathematics in the cited period,
after those of Germany and France. Also,
the works of the Italian school of algebraic
geometry were prominent in this period.
In addition, the names of several Italian
mathematicians are permanently inscribed
in textbooks, e.g., Volterra, Betti, Tonelli,
Fubini, Vivanti, Vitali, Enriques, and many
others.
This book begins (as all good books
should) “before the beginning” and its first
two chapters are devoted to this. The
“Risorgimento generation” (Betti, Brioschi,
Casorati, Beltrami, Cremona) had already
established a name for mathematics done on
the Italian peninsula by 1870 (when Italian
unification was completed and Rome be-
came the capital of the newly unified Italy).
Their successors (e.g., Veronese, Corrado
Segre, and others) brought algebraic ge-
ometry, “italienische Geometrie,” to full
flower. At the turn of the twentieth cen-
tury, though, Volterra was the most promi-
nent Italian mathematician, in analysis and
mathematical physics, and mathematicians
like Fubini and Vitali had grown beside him.
These men left their stamp on real analy-
sis, and were a counterweight in Italian
353
mathematics to the already prominent al-
gebraic geometry of Castelnouvo, Enriques,
and Severi. One should not forget either
Levi-Civita, who had correspondence with
Einstein that in April 1915 the latter called
outstandingly important for him.
Of course, in August 1914 World War I
broke out. This actually ended the “golden
age” of Italian mathematics. After all,
World War I was a turning point for all of
European (and American) society. Noth-
ing remained as it was heretofore. Italian
mathematicians, by and large, were in fa-
vor of Italian intervention on the side of
the Allies and against the Central Powers
though this did not happen until May 24,
1915. Vito Volterra (1860–1940) was per-
haps the most committed mathematician
urging Italian intervention on the side of
England, France, and Russia. This is clear
from his correspondence at the time. This is
perhaps the place to mention that although
all the letters and commentary in Italian are
translated in the text, none of the consider-
able number in French are, and while this
posed no difficulty to the reviewer, it may
well be one for some prospective readers.
Castelnuovo (1865–1952) and Enriques
(1871–1946) also had the same persuasion
as Volterra, though somewhat more muted.
Their collaboration, which would culminate
in a complete classification of algebraic sur-
faces in 1914, began in 1892. Also, Enriques
married Castelnuovo’s sister.
Chapter III, titled “Volterra’s Leader-
ship,” is the first truly post–World War
I chapter. Postwar there was the well-
known exclusion under French leadership
of members of the former Central Powers
from international scientific collaboration.
Volterra was the leader of Italian mathe-
matics during this period until his refusal
to take the fascist oath in 1931, though he
did not win all “battles,” even within his
own faculty.
In 1921 a chair became open in the math-
ematics faculty in Rome when an analyst
died. Volterra wanted an analyst to be suc-
ceeded by an analyst. In 1918 Levi-Civita
(1873–1941) had become a faculty mem-
ber at Rome, after earlier attempts to woo
him from Padua had failed. Levi-Civita
was a committed pacifist and socialist and
did not agree with Volterra. Among the
 354 BOOK REVIEWS
prime candidates were Leonida Tonelli, En-
riques, and Francesco Severi (1879–1961).
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
Severi (who had been Enriques’ assistant
earlier in Bologna—the book under review
says he had studied with Enriques, but
to this reviewer that is unclear), with the
support of his friend Levi-Civita, was suc-
cessful. Enriques also arrived shortly there-
after in Rome, when Castelnuovo volun-
tarily resigned the chair that he had held
for many years. At this time, Severi and
Levi-Civita were both socialists—and the
Socialist Party had been opposed to entry
into the war. Levi-Civita’s support of Severi
was based on his belief that he was the best
mathematician, from all aspects, who was
available. For future developments, it may
be noted here that Levi-Civita, Volterra,
Castelnuovo, and Enriques were all Jew-
ish, but Severi was not. During the 1920s
Volterra was instrumental in the founding
of both the Unione Matematica Italiana
(UMI) and the Consiglio Nazionale delle
Ricerche (CNR). These would later become
something of rivals, though initially the
CNR supported the UMI.
Of course, any discussion of Italy between
the wars must deal with the rise of fascism.
On October 27, 1922, the Fascist Party an-
nounced the “march on Rome”; the King
refused to sign the previously agreed upon
declaration of martial law; on October 29,
Mussolini was charged with forming a new
government. The leading fascist figure in
education was the Sicilian philosopher Gio-
vanni Gentile. The authors trace the early
mathematical experience of fascism through
him; in fact early on (in 1911) Benedetto
Croce and Gentile had had a philosophical
dispute with Enriques, who was also some-
thing of a philosopher, which, according to
the authors, had a negative effect on math-
ematicians wanting active participation in
the cultural life of Italy. Gentile did not
join the Fascist Party until May 1923 and
became the “Philosopher of Fascism.”
In 1925 Gentile issued a fascist manifesto;
the only mathematicians to support it were
Corrado Gini and Salvatore Pincherle. Gini,
of course, is still mentioned in statistics
books for “Gini’s coefficient,” and Pincherle
would be a major figure in organizing the
International Congress of Bologna in 1928.
In fact, the reviewer is surprised by the
scant attention given to the Congress in this
book. A countermanifesto to Gentile’s was
issued by Benedetto Croce. Croce, who had
been a friend and collaborator of Gentile,
had earlier broken with him, and was pre-
vailed upon to issue this countermanifesto.
This was signed by a number of significant
mathematicians including Volterra, Castel-
nuovo, Levi-Civita, and Severi, though not
Enriques. Croce became a representative of
the antifascist forces in Italy, though he was
always skeptical of democracy and “mass-
man.”
It is Severi who became the leader of Ital-
ian mathematics. The former socialist and
interventionist (like Mussolini) became a
fascist. Severi and Enriques, who had been
coworkers in the first decade of the twenti-
eth century, developed a very quarrelsome
relationship that was only formally made up
in 1929. Though the authors do not men-
tion it, Severi was apparently well known
for his quarrelsomeness. Severi as a politi-
cian and mathematician became (at first
through Gentile’s support—and his appar-
ent influence on Gentile) “one man alone in
the lead.” After the war he was threatened
with lynching and became a Catholic who
suggested that mathematics and physics
proved God’s existence. Throughout the
fascist period he was the master of math-
ematics (though apparently he did not fol-
low Gentile in support of the “Republic of
Salo”).
One figure not yet mentioned who de-
serves note is Mauro Picone (1885–1977),
who considered himself “a black shirt since
the very beginning.” In 1927 he founded
the Istituto Nazionale per le Applicazioni
del Calcolo (INAC) that would become the
envy of some German mathematicians (like
Gustav Doetsch). While it seems Picone
believed in applications, other foreign ob-
servers (like Ludwig Prandtl) did not feel
that his institute did “applied mathemat-
ics.” In 1929 it became part of the CNR,
whose president was Guglielmo Marconi.
The authors identify Picone as an applied
mathematician, but it is perhaps necessary
to indicate that this may be disputed.
The authors follow Italian mathematics
through the insistence on the fascist oath in
1931 and the racial laws of 1938. While Ital-
ian Jews had lived as normal citizens at least
 BOOK REVIEWS
since 1870 (and often earlier), until this was
canceled by the racial laws, it is undeniable
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
that not only in the distant past (“ghetto”
is, after all, an Italian word) but also in
the first decades of the twentieth century
anti-Semitic feelings occasionally revealed
themselves in the correspondence of Ital-
ian mathematicians. 1938 was the year for
anti-Semitic action in other countries as
well, most prominently Kristallnacht. Ital-
ian anti-Semitism did not go as deeply into
the culture as it did in Germany, Russia,
or Poland, but it was real. In any case, it
was never biological. Severi and Picone were
not the only fascist mathematicians, Enrico
Bomplani was also one, and he in fact wrote
a history of Italian mathematics omitting
the names, among others, of Segre, Castel-
nuovo, Volterra, Enriques, and Levi-Civita.
There were other fascist mathematicians as
well.
The authors’ book closes with the onset
of World War II, of which there are brief
mentions.
Actually it is not surprising that the 21
years between the wars is too short a period
for consideration (which is why Volterra so
dominates the early part of the book and
Severi and Enriques its second half). This
book provides much interesting information
not otherwise available in English, and the
authors have previously written on cognate
topics.
Unfortunately, the book cannot be rec-
ommended to English readers. The reason
for this is not just the occasional plodding
style of the authors, but the numerous er-
rors (not all minor) in English that litter
the book and are sometimes confusing. A
good proofreader or copy editor would have
caught these many errors, which seem even
worse in the translations from Italian than
in the original text. Of course, presum-
ably to save money, Birkhäuser (or was
it Springer, or perhaps Bertelsmann?) ap-
parently did not employ a proofreader or
copy editor of any sort. Sometimes these
errors are humorous, for example, on p. 57
there appears the phrase “more than 40% of
[French] students in Mathematics or in Sci-
ences were killed or blessed [sic] in France
during the war.” The French word for
wounded is “blessé” but the authors are here
citing an English-language author! The oc-
355
casional “é” and “con” make me think that
the book was originally written in Italian,
and then “translated” by the authors. Most
often these errors are confusing. The lack
of care which this book presents extends to
the pictures. The same picture of Leonida
Tonelli appears on p. 64 and p. 162. Severi’s
picture (the same one) appears on p. 62 and
p. 103 and these are both clearly cropped
from the picture of him and Picone (and
unnamed others) on p. 261. This careless-
ness begins on the page before the preface
(!) where the cover illustration is identified
as being of L. E. J. Brouwer (subject of
another book in the series) when actually
it is of three Italian mathematicians: Levi-
Civita, Enriques, and ? (I do not know).
What Birkhäuser should do (if they wish
to attract English-language readers) is with-
draw the book from circulation, have the
French passages translated (by an English
speaker!), have the whole book gone over
(by an English-speaking copy editor), and
then reissue it.
S. L. SEGAL
University of Rochester
Perturbations of Positive Semigroups
with Applications. By Jacek Banasiak and Luisa
Arlotti. Springer-Verlag, London, 2006. $89.95.
xiv+438 pp., hardcover. ISBN 1-85233-993-4.
In 1798, the British economist Thomas
Malthus suggested the linear differential
equation x
= αx as a simple model for
the growth of populations. Since such a
model only shows exponential growth or
decay which (in the case of growth) can
hardly continue for too long, in 1845 Pierre-
Francois Verhulst suggested the so-called
logistic equation x
= αx(1 − K x
), which
incorporates the idea that there is an envi-
ronmental limit to growth, the carrying ca-
pacity K. Both equations are special cases
of Bernoulli’s equation x
= αx − γxθ with
θ > 1, which is named after Jacob Bernoulli
and was solved by Leibniz in 1696, and
there are many other biologically mean-
ingful nonlinear modifications of x
= αx
[30, Chap. 5]. Nevertheless, these two basic
equations are indicative of the two main
different paths dynamical systems theory
 356 BOOK REVIEWS
would later take: (linear) operator semi-
groups and nonlinear semiflows (dynamical
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
systems). They are also indicative of this re-
view being written by somebody interested
in life science applications.
A semiflow, whether linear or nonlinear,
describes the development of a system in
time according to the following rule:
Consider three points in time, r < s < t.
If the system starts at x at time r and
reaches a state y at the time s and takes y
as a starting point at time s and evolves to
a state z at the time t, then z is the same
state as the one the system reaches when
evolving directly from time r to time t.
More formally, if Φ(t, r, x) denotes the
state the system will reach at time t when
it starts at state x at time r < t, we have
the functional relationship
(1)
Φ(t, r, x) = Φ(t, s, Φ(s, r, x)), r < s < t,
Φ(r, r, x) = x.
To be even more precise, we take a set of
possible states X and a nontrivial interval
I. It is then convenient to define the (not
necessarily everywhere defined) operators
Φ(t, s)(x) = Φ(t, s, x) for t, s ∈ I, t ≥ s,
and x ∈ X. The semiflow property (1) can
then conveniently be written as
(2)  been proved useful in nonlinear dynamical
Φ(t, s) ◦ Φ(s, r) = Φ(t, r) r ≤ s ≤ t,
Φ(s, s) = I r, s, t ∈ I.
Here the first equation includes the equal-
ity of the domains of definition, where for
two operators F and G the domain of their
composition is
D(G ◦ F ) = {x ∈ D(F ); F (x) ∈ D(G)}.
It is still useful to keep the formulation (1)
in mind because typically X is a topological
space. The most fruitful continuity assump-
tion is that Φ is continuous from ∆ × X to
X, where ∆ = {(t, r); r, t ∈ I, r ≤ t} and
∆ × X carries the product topology. In
the Malthus world of unrestricted growth,
X usually is a Banach space and the op-
erators Φ(t, s) are everywhere defined on
X and linear. The continuity of Φ then
equivalently translates to Φ(t, s) being a
strongly continuous family of bounded lin-
ear operators. This means that the realm
of (linear) functional analysis is entered,
whose powerful methods are responsible for
the special path linear semigroups (evo-
lution systems) have taken as compared
to general semiflows. It is in this realm
that the book by Banasiak and Arlotti,
from now on called [BA], resides. [BA] also
lives in a world without seasons as it re-
stricts itself to autonomous linear semi-
flows where Φ(t + s, s) = Φ(t, 0) =: S(t).
This is not so restrictive as it may ap-
pear at first glance, as there are procedures
which transform nonautonomous semiflows
into autonomous semiflows on a larger state
space. The easiest such transformation is
Ψ(t, (s, x)) = (t + s, Φ(t + s, s, x)), which
introduces an autonomous semiflow on the
state space I × X. This transformation is
not useful in a linear context as Ψ is not
linear in (s, x) even if the Φ(t, s) are lin-
ear maps. Even for nonlinear semiflows,
this transformation is of restricted value,
namely, mainly for so-called asymptoti-
cally autonomous semiflows [22]. Another
nonlinear transformation leads to skew-
product semiflows, where the state space
is very complicated as it contains semi-
flows, Ψ(t)(x, Φ) = (Φ(t, 0, x), Φt ), where
Φt is the shifted semiflow Φt (s, r, x) =
Φ(t + s, t + r, x). In spite of the unwieldy
state space, skew-product semiflows have
systems theory [12]. Again, linearity is not
preserved even if the maps Φ(t, s) are linear.
This is accomplished by so-called (Howland)
evolution semigroups (see [5] and the refer-
ences therein), where the new state space
is Lp (I, X), 1 ≤ p < ∞, and
[S(t)f ](r)
 
Φ(t, t − r)f (t − r) if t, t − r ∈ I,
= ,
0 otherwise
t ≥ 0, r ∈ I.
Before we describe what is specifically done
in [BA], let us summarize what is not cov-
ered (on purpose because the authors want
to follow their own interests and not re-
peat topics that are easily available in the
literature): spectral theory, large-time be-
havior (see the references in [BA, p. xii]
and [7, 36]), dual semigroups [4, 6, 7, 8, 33],
integrated semigroups [2, 29, 32], analytic
 BOOK REVIEWS 357

semigroups [1, 7], applications to Markov (iii) For every x ∈ D(A)+ , there ex-
processes [11, 28, 15], and applications to ists some x∗ ∈ X ∗ , x∗ ≥ 0, with
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php

functional differential equations [8]. [BA] x, x∗  = x and (A + B)x, x∗  ≤
has an impressive list of 172 references; I 0.
have tried to make the list of references be-
Then there exists an extension à of A+B,
low as complementary as possible and, in
D(Ã) ⊇ D(A), which generates a positive
order to limit its length, have only repeated
C0 -semigroup T of contractions. The resol-
references in [BA] if necessary to keep this
vent of à satisfies the strong series expan-
review self-contained.
sion,
The focus of [BA] is positive semigroups.
This reflects that the relevant states of most 
∞
biological and many physical problems are (λ − Ã)−1 x = (λ − A)−1 (B(λ − A)−1 )n x,
positive. More precisely, the state space X n=0
contains a closed positive convex cone X+ . (3)
x ∈ X, λ > 0.
A semigroup S = {S(t); t ≥ 0} is called
positive if S(t)X+ ⊆ X+ . From a nonlin-
The positivity of the perturbed semi-
ear point of view, a positive semigroup is
group, T , is not explicitly stated in [BA,
a monotone semiflow. Monotone semiflows
Thm. 5.2], and the subsequent material
have a powerful theory of their own [13, 27]
in [BA, sec. 5.1] makes the reader wonder
which, however, has not too much bearing
whether additional assumptions are needed
on positive semigroups because linearity al-
for positivity. However, by (3), Ã is re-
lows the derivation of even farther-reaching
solvent positive, which implies that T is
results [34]. More precisely, as correctly
positive by the exponential formula [BA,
described in the title, [BA] concentrates
(3.22)]. Actually T is the smallest positive
on the perturbation of positive semigroups
semigroup whose generator extends A + B
and its applications to birth and death
[BA, Prop. 5.7].
processes, fragmentation processes (pure or
Condition (iii) is central in applying
with growth or decay), and to kinetic the-
[BA, Thm. 5.2]. An important class of
ory. The book can be divided roughly into
Kantorovich–Banach spaces is that of ab-
three parts. Chapters 2, 3, and 4 describe
stract L spaces, i.e., Banach lattices where
(mostly known) topics from functional anal-
the norm is additive on the positive cone,
ysis and C0 -semigroup theory that will be
x + y = x + y if x, y ∈ X+ . Then
needed for the remaining parts. The central
there exists a unique positive φ ∈ X ∗ such
Chapters 5 and 6 generalize what the au-
that
thors call Kato’s perturbation theorem to
abstract L spaces and develop the theme of
(4) φx = x, φ = x, x ∈ X+ .
honest and dishonest semigroups. The third
part, actually more than half of the book, The following corollary holds (cf. [31]).
consists of applications. Corollary 2. Let X be an abstract L
One of the highlights of the central Chap- space. Let A be the generator of a posi-
ters 5 and 6 is Theorem 5.2, which I give tive C0 -semigroup S of contractions and let
in slightly modified form. In the follow- B : D(A) → X be A-bounded and satisfy
ing, “spr” denotes the spectral radius of a the following assumptions:
bounded linear operator.
Theorem 1. Let X be a Kantorovich– (i) B is positive.
Banach space (i.e., a Banach lattice with a (iii)
φ(A + B)x ≤ 0 for all x ∈ D(A)+ .
fully regular cone). Let A be the generator
of a positive C0 -semigroup S of contractions Then there exists an extension à of A+B,
and let B : D(A) → X be A-bounded and D(Ã) ⊇ D(A), which generates a positive
satisfy the following assumptions: C0 -semigroup, T , of contractions and satis-
fies (3).
(i) B is positive: Bx ∈ X+ for all
It is easy to see that condition (iii)
is also
x ∈ D(A)+ = X+ ∩ D(A).
necessary. Corollary 2 is stated and proved
(ii) lim sup spr(B(λ − A)−1 ) ≤ 1. in [BA, Cor. 5.17] for the important class
λ→∞
 358 BOOK REVIEWS
of abstract L spaces X = L1 (Ω, µ). In this
case x, φ = Ω xdµ. While every abstract
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
L space is isomorphic to some L1 (Ω, µ) [25,
Thm. II.8.5], this does not help in verifying
(iii)
. Another important example of an ab-
stract L space is M(Ω), the Banach space
of (signed) measures µ of finite variation on
some σ-algebra of subsets of Ω [15]. In this
case µ, φ = µ(Ω). An application of Corol-
lary 2 to M(Ω) can be found in [31, sec. 5].
Returning to Theorem 1, we cannot ex-
pect à = A + B unless lim supλ→∞ spr
B(λ − A)−1 < 1, for, in an abstract L
space, A + B is resolvent positive if and
only if lim supλ→∞ spr B(λ − A)−1 < 1 [35]
[BA, Thm. 5.10]. This raises the possibil-
ity that at least à = A + B if lim supλ→∞
spr B(λ − A)−1 = 1. As it turns out, this
does not necessarily hold even in the sce-
nario of Corollary 2. In order to show that
this question is not only of abstract inter-
est, let us introduce some more terminology
[BA, sec. 6.1]. A positive C0 -semigroup of
contractions is called a substochastic semi-
group. A positive C0 -semigroup T is called
a stochastic semigroup if T (t)x = x
for all x ∈ X+ . If X = M(Ω), a positive
semigroup is stochastic if and only if it pre-
serves the property of being a probability
measure, µ ≥ 0, µ(Ω) = 1, obviously an im-
portant feature for stochastic processes. A
stochastic semigroup on an abstract L space
is called strictly stochastic if φ(A + B) = 0.
Notice that strict stochasticity is stronger
than being substochastic and not stochas-
tic. As it turns out, Ã = A + B is equiv-
alent to the stochasticity of the semigroup
generated by Ã.
Theorem 3. Let X be an abstract L
space. Let A be the generator of a pos-
itive semigroup S of contractions and let
B : D(A) → X be A-bounded and positive.
Assume that φ(A + B) = 0. Then the fol-
lowing are equivalent for the extension à of
A + B and the substochastic semigroup T
generated by à according to Corollary 2:
(i) Ã is the closure of A + B.
(ii) T is a stochastic semigroup.
(iii) (B(λ − A)−1 )n → 0 strongly as
n → ∞.
Indeed, (i) =⇒ (ii) follows from the def-
inition of the closure and standard semi-
group theory (proof of [BA, Thm. 6.13]),
while (ii) =⇒ (iii) (proof of [BA, Thm. 6.8])
and (iii) =⇒ (i) follow from (3).
In [BA, sec. 6.2], Theorem 3 is extended
to strictly substochastic semigroups intro-
ducing the concepts of honest and dishonest
semigroups. While these concepts are cen-
tral to [BA], their discussion may lead too
far for this review. Stochastic semigroups
are an example of honest semigroups, while
the substochastic semigroup T in Theorem 3
is a dishonest semigroup if T is not stochas-
tic in spite of φ(A + B) = 0.
Notice that condition (iii) in Theorem 3 is
the only one which is stated in terms of the
original ingredients A and B. Differently
from what one might expect, (iii) does not
seem to play any major role in discussing
the honesty of the semigroups which arise
in the many applications that are discussed
in Chapters 7 to 11. Rather the authors use
extension techniques which are developed in
section 6.3 and rely on the special structure
of X = L1 (Ω, µ). The applications cover
some 240 pages and present a dazzling mul-
titude of problems: birth and death pro-
cesses (though not the general Kolmogorov
differential equations; see [14, 31, 18] and
the references therein), pure fragmentation
processes, fragmentation with growth and
decay, kinetic theory (Maxwell–Bolzmann
and semiconductor equations), as well as
elastic and inelastic collision models. It is
worth mentioning that birth and death pro-
cesses are not the only possible applications
in mathematical biology. Nonlinear versions
of the Kolmogorov differential equations
have been suggested to model prion prolif-
eration (which supposedly plays a role in
transmitted spongiform encephalopathies
like mad cow disease) [20, 23] as well as
spatially implicit metapopulation models
and host-macroparasite diseases [17, 19].
Fragmentation models with growth also ap-
ply to size-structured cell division models
[3, 9, 17, 21, 36] and again to prion prolif-
eration models [10, 16, 24, 26].
Obviously [BA] should be in the hand of
every researcher who is interested in any
of these applications. It also seems to be
the only book available that deals with rel-
atively bounded positive perturbations of
positive semigroups and with general ques-
tions of their stochasticity and honesty.
 BOOK REVIEWS
REFERENCES
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
[1] H. Amann, Linear and Quasilinear [15]
Parabolic Problems. Volume I. Ab-
stract Linear Theory, Birkhäuser,
Basel, 1995.
[2] W. Arendt, C. J. K. Batty, M. Hieber,
and F. Neubrander, Vector-valued [16]
Laplace Transforms and Cauchy
Problems, Birkhäuser, Basel, 2001.
[3] O. Arino, A survey of structured cell
population dynamics, Acta Biotheo- [17]
retica, 43 (1995), pp. 3–25.
[4] P. L. Butzer and H. Berens, Semi-
groups of Operators and Approxi-
mation, Springer, Berlin, Heidelberg, [18]
1967.
[5] C. Chicone and Y. Latushkin, Evolution
Semigroups in Dynamical Systems
and Differential Equations, AMS,
Providence, RI, 1999.
[6] Ph. Clément, O. Diekmann, M. Gyl- [19]
lenberg, H. J. A. M. Heijmans,
and H. R. Thieme, A Hille-Yosida
theorem for a class of weakly contin-
uous semigroups, Semigroup Forum,
38 (1989), pp. 157–177.
[7] Ph. Clément, H. J. A. M. Heijmans,
[20]
S. Angenent, C. J. van Duijn, and
B. de Pagter, One-Parameter Semi-
groups, North–Holland, Amsterdam,
1987.
[8] O. Diekmann, S. A. van Gils, S. M.
[21]
Verduyn Lunel, and H.-O. Wal-
ther, Delay Equations: Functional-,
Complex-, and Nonlinear Analysis,
Springer, New York, 1995.
[9] O. Diekmann, H. J. A. M. Heijmans,
[22]
and H. R. Thieme, On the stability
of the cell size distribution, J. Math.
Biology, 19 (1984), pp. 227–248.
[10] H. Engler, J. Prüss, and G. F. Webb,
Analysis of a model for the dynamics
of prions II, J. Math. Anal. Appl., [23]
324 (2006), pp. 98–117.
[11] W. Feller, An Introduction to Probabil-
ity Theory and Its Applications, Vol.
II, Wiley, New York, 1965. [24]
[12] J. K. Hale, Asymptotic Behavior of Dis-
sipative Systems, AMS, Providence,
RI, 1988.
[13] M. W. Hirsch and H. L. Smith, Mono-
tone dynamical systems, in Hand- [25]
book of Differential Equations, Vol. 2,
A. Cañada, P. Drábek, and A. Fonda,
eds., Elsevier, Amsterdam, 2005, pp. [26]
239–357.
[14] T. Kato, On the semi-groups generated by
Kolmogoroff ’s differential equations,
359
J. Math. Soc. Japan, 6 (1954), pp.
1–15.
T. Lant and H. R. Thieme, Markov tran-
sition functions and semigroups of
measures, Semigroup Forum, http://
dx.doi.org/10.1007/s00233-006-0636-
z.
P. Laurençot and C. Walker, Well-
posedness for a model of prion pro-
liferation dynamics, J. Evol. Equ., to
appear.
P. Magal and S. Ruan, eds., Struc-
tured Population Models in Biology
and Epidemiology, Springer, to ap-
pear.
M. Martcheva, H. R. Thieme, and T.
Dhirasakdanon, Kolmogorov’s dif-
ferential equations and positive semi-
groups on first moment sequence
spaces, J. Math. Biol., 53 (2006), pp.
642–671.
M. Martcheva and H. R. Thieme, In-
finite ODE systems modeling size-
structured metapopulations, macro-
parasitic diseases, and prion pro-
liferation, in Structured Population
Models in Biology and Epidemiology,
Springer, to appear.
J. Masel, V. A. A. Jansen, and M. A.
Nowak, Quantifying the kinetic pa-
rameters of prion replication, Bio-
phys. Chemistry, 77 (1999), pp. 139–
152.
J. A. J. Metz and O. Diekmann, The
Dynamics of Physiologically Struc-
tured Populations, Lecture Notes in
Biomath. 68, Springer, Berlin, Hei-
delberg, 1986.
K. Mischaikow, H. L. Smith, and
H. R. Thieme, Asymptotically au-
tonomous semiflows: Chain recur-
rence and Lyapunov functions, Trans.
Amer. Math. Soc., 347 (1995), pp.
1669–1685.
M. A. Nowak, D. C. Krakauer, A.
Klug, and R. M. May, Prion in-
fection dynamics, Integrative Biol., 1
(1998), pp. 3–15.
J. Prüss, L. Pujo-Menjouet, G. F.
Webb, and R. Zacher, Analysis of
a model for the dynamics of prions,
Discrete Contin. Dyn. Syst. Ser. B, 6
(2006), pp. 215–225.
H. H. Schaefer, Banach Lattices and
Positive Operators, Springer, Berlin,
Heidelberg, 1974.
G. Simonett and C. Walker, On the
solvability of a mathematical model
for prion proliferation, J. Math. Anal.
Appl., 324 (2006), pp. 580–603.
 360 BOOK REVIEWS
[27] H. L. Smith, Monotone Dynamical Sys-
tems: An Introduction to the Theory
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
of Competitive and Cooperative Sys-
tems, AMS, Providence, RI, 1995.
[28] K. Taira, Semigroups, Boundary Value
Problems and Markov Processes,
Springer, Berlin, Heidelberg, 2004.
[29] H. R. Thieme, Positive perturbations of
dual and integrated semigroups, Adv.
Math. Sci. Appl., 6 (1996), pp. 445–
507.
[30] H. R. Thieme, Mathematics in Population
Biology, Princeton University Press,
Princeton, NJ, 2003.
[31] H. R. Thieme and J. Voigt, Stochas-
tic semigroups: Their construction
by perturbation and approximation,
in Positivity IV: Theory and Appli-
cations, M. R. Weber and J. Voigt,
eds., Technical University of Dresden,
Dresden, 2006, pp. 135–146.
[32] H. R. Thieme and H. Voßeler, A Stielt-
jes type convolution for integrated
semigroups of strong bounded vari-
ation and Lp solutions to the abstract
Cauchy problem, J. Differential Inte-
gral Equations, 15 (2002), pp. 1171–
1218.
[33] J. van Neerven, The Adjoint of a Semi-
group of Linear Operators, Lecture
Notes in Math. 1529, Springer, Berlin,
Heidelberg, 1992.
[34] J. van Neerven The Asymptotic Be-
haviour of Semigroups of Linear Op-
erators, Birkhäuser, Basel, 1996.
[35] J. Voigt, On resolvent positive oper-
ators and C0 -semigroups on AL-
spaces, Semigroup Forum, 38 (1989),
pp. 263–266.
[36] I. I. Vrabie, C0 -Semigroups and Appli-
cations, Elsevier, Amsterdam, 2003.
[37] G. F. Webb, Population models struc-
tured by age, size, and spatial position,
in Structured Population Models in
Biology and Epidemiology, Springer,
to appear.
H. R. THIEME
Arizona State University
Extreme Value Theory: An Introduction.
By Laurens de Haan and Ana Ferreira. Springer-
Verlag, New York, 2006. $59.95. xviii+417 pp.,
hardcover. ISBN 0-387-23946-4.
As evidenced by the aftermath of hurricane
Katrina in September 2005, extreme val-
ues can have enormous consequences. This
book is in the long tradition of study of
extreme value theory in the Netherlands,
motivated in large part by the continuing
importance of this theory for the design
of dikes to keep out the North Sea. As
the authors say in their preface, “Approx-
imately 40% of the Netherlands is below
sea level.” The storm surge experienced in
Plaquemines parish near New Orleans dur-
ing hurricane Katrina was 4.3 meters high
(according to [10]). For comparison, the
catastrophic floods in the Netherlands in
1953 were caused by a 3.85 meter surge,
while historical accounts estimate a 4.0 me-
ter surge for the flood of 1570 as discussed in
[3]. For further discussion of the 1953 esti-
mate of the one-in-ten-thousand year storm
surge height for the Netherlands, given as
5.14 by the van Dantzig report, see [3]. (It
is unfortunate that [3] is not referenced in
the book under review.)
In writing this book on extreme value the-
ory, the authors begin by setting themselves
a very ambitious goal. They say (p. viii),
“The aim of this book is to give a thorough
account of the basic theory of extreme val-
ues, probabilistic and statistical, theoretical
and applied.” While they have succeeded
admirably concerning the probabilistic and
theoretical aspects of extreme value theory,
it seems to this reviewer that they fail to
meet the goals concerning the statistical
and applied aspects of extremes. But this
is, after all, a book about theory.
The book consists of 10 chapters, di-
vided into three parts: Part I, consisting of
Chapters 1–5, treats the probabilistic and
statistical sides of one-dimensional extreme
value theory. (This part occupies approxi-
mately half of the 400-plus pages.) Part II,
in Chapters 6–8 (about a fifth of the book),
treats finite-dimensional extreme value the-
ory, a topic of considerable research interest
over the past 20 years. Part III, given in
Chapters 9 and 10 (less than one sixth of the
book), deals with the situation in which the
data consist of continuous time stochastic
processes. The book also contains two ap-
pendices, one on Skorokhod’s theorem and
Vervaat’s lemma, and another on regular
variation and extensions thereof, an area to
which de Haan has made substantial con-
tributions.
A total of 93 exercises are included at
the ends of the various chapters, the bulk
 BOOK REVIEWS
of which (49) are connected with Chapters
1–3. The bibliography lists 105 papers and
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
books, 24 of which are authored or coau-
thored by de Haan. (This should be com-
pared with the approximately 370 entries
in [1] and 646 in [5].)
Two main themes are given full treatment
here for the first time in textbook form.
The first of these is the progression from
univariate extreme value theory to multi-
variate extreme value theory, and further to
function-valued extreme value theory. This
is a valuable feature of this book, since it is
an area in which considerable research and
progress have been made over the past 25
years. Multivariate extreme value theory is
also treated in Chapters 8 and 9 of the re-
cent book [1], which also makes an effort to
cover both the probabilistic and statistical
aspects of the theory. It is not covered in
[5], but it is a topic that is of current in-
terest in connection with a wide range of
applications, including environmental mon-
itoring and financial statistics; see, e.g., [8]
and [9].
The second theme is the systematic use
of the theory of “extended regular varia-
tion,” due in large part to the work of
the first author, and the associated theory
of “second order regular variation.” The
mathematics of these classes of functions
is developed in Appendix B and used sys-
tematically throughout the text. In this
connection it is somewhat curious that the
authors do not reference the 1987 book [2],
but this might be partly explained by the
fact that [6], which also appeared in 1987,
was also not referenced in [2]. Both [2] and
[6] are included in the comprehensive bibli-
ography of [5].
The authors do make an effort to treat
the statistical and applications sides of ex-
treme value theory. They devote Chapters 3
and 4 to statistical aspects of univariate ex-
treme value theory, and Chapters 7 and 8 to
statistical aspects of multivariate extreme
value theory. Moreover, they treat three
sets of data as “case studies”: the first in-
volves sea levels at Delfzijl on the northeast
coast of the Netherlands (with a univari-
ate data set consisting of 1873 points); the
second involves the S&P total return index
over the period January 1, 1980 to May
14, 2002 (a univariate data set consisting of
5835 points); while the third concerns the
361
life spans of people born in the Netherlands
during the period (1877–1881) and still alive
on January 1, 1971 (with 10391 points).
These case studies are developed further
and used to illustrate the theory in Chap-
ters 3 and 4 in a total of about 14 pages.
Unfortunately the data sets are apparently
not publicly available, so this precludes nu-
merical verification of the results obtained
or comparison with other methods. While
this material makes a start at delivering on
the authors’ stated goal of giving a “thor-
ough account of the basic theory of extreme
values. . . statistical. . . and applied,” I was
left somewhat disappointed that more had
not been done in this direction. There is no
treatment of profile likelihood methods for
finding confidence bounds (see, e.g., Chap-
ters 5 and 7 of [1]), essentially no treatment
of data dependent choices of k for estimation
of the extreme value index (see, e.g., [4], sec-
tion 4.7 of [1], and [7] for a different method
involving kernel smoothing), no treatment
of extreme value theory in the context of
time series models (see, e.g., Chapter 10 of
[1] or Chapter 7 of [5]), and no discussion
of regression analysis of extreme values (see
Chapter 7 of [1]). Although a reader or stu-
dent will get some taste of the statistical
theory of extreme values from this book, for
a thorough treatment of statistical methods
for extreme values and more complete treat-
ment of applications, the reader will need
turn to other books (for example, to [1] and
[5]) and the still growing research literature.
This book is a “must have” for experts
in extreme value theory, and would be an
excellent choice for a special topics course
on the probabilistic and theoretical sides of
extreme value theory, especially a course
emphasizing the multivariate aspects of ex-
tremes. For thorough treatments of the sta-
tistical aspects and applications the reader
will need to look elsewhere.
REFERENCES
[1] J. Beirlant, Y. Goegebeur, J. Segers,
and J. Teugels, Statistics of Ex-
tremes, Wiley Series in Probability
and Statistics, John Wiley, Chiches-
ter, UK, 2004.
[2] N. H. Bingham, C. M. Goldie, and
J. L. Teugels, Regular Variation,
Encyclopedia Math. Appl. 27, Cam-
 362 BOOK REVIEWS
bridge University Press, Cambridge,
UK, 1987.
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
[3] L. de Haan, Fighting the arch-enemy with
mathematics, Statist. Neerlandica, 44
(1990), pp. 45–68.
[4] G. Draisma, L. de Haan, L. Peng, and
T. T. Pereira, A bootstrap-based
method to achieve optimality in esti-
mating the extreme-value index, Ex-
tremes, 2 (1999), pp. 367–404.
[5] P. Embrechts, C. Klüppelberg, and
T. Mikosch, Modelling Extremal
Events, Appl. Math. (N.Y.) 33,
Springer-Verlag, Berlin, 1997.
[6] J. L. Geluk and L. de Haan, Regu-
lar Variation, Extensions and Taube-
rian Theorems, CWI Tract 40, Stich-
ting Mathematisch Centrum Cen-
trum voor Wiskunde en Informatica,
Amsterdam, The Netherlands, 1987.
[7] P. Groeneboom, H. P. Lopuhaä, and
P. P. de Wolf, Kernel-type estima-
tors for the extreme value index, Ann.
Statist., 31 (2003), pp. 1956–1995.
[8] N. D. Le and J. V. Zidek, Statistical
Analysis of Environmental Space-
Time Processes, Springer Ser. Sta-
tist., Springer-Verlag, New York,
2006.
[9] Y. Malevergne and D. Sornette,
Extreme Financial Risks, Springer-
Verlag, Berlin, 2006.
[10] Wikipedia, Hurricane Katrina, http://en.
wikipedia.org/wiki/HurricaneKatrina
(2006).
JON A. WELLNER
University of Washington
Interfacial Convection in Multilayer Sys-
tems. By A. Nepomnyashchy, I. Simanovskii, and
J. C. Legros. Springer-Verlag, New York, 2006.
$79.95. xii+306 pp., hardcover. ISBN 0-387-
22194-8.
This monograph presents the current state
of the art in the understanding of a class of
problems in the study of Rayleigh–Bénard
convection (RBC). RBC has been the focus
of extensive research in physics and math-
ematics for at least the past five decades.
In its simplest form, RBC describes the be-
havior of a uniform layer of fluid which is
heated from below. Instabilities stemming
from the thermal expansion of the fluid
can drive convective motion in the fluid,
which has been observed to take many in-
tricate forms. The richness of its dynamics,
patterns, and instabilities has made RBC a
paradigm for pattern formation. It has been
an important testbed for studies of stability,
bifurcations, and weakly nonlinear analysis.
Large numbers of research articles and sev-
eral books provide excellent introductions
to the nonlinear dynamics of such systems,
for example, [1].
The current book, however, is not an
introduction; rather, it is a systematic
overview of the dynamics and instabilities
that occur in more realistic and inevitably
more complicated forms of RBC. The first
chapter is a brisk fifteen pages that gives
the educated reader a good cross-section of
references to previous works in the course of
introducing the governing equations. The
beginning of the hierarchy of models ex-
tending RBC is Bénard convection, where
the top of the fluid layer is a moving bound-
ary to be determined as part of the solution.
If stresses created by temperature-induced
variations in the surface tension along this
interface are included, then the system
is called the Bénard–Marangoni problem.
Classic one-layer models neglect influences
of gas above the fluid layer on the dynamics
of the interface. In contrast, in two-layer
and multilayer models the motion of in-
terfaces separating immiscible layers of flu-
ids is determined by balancing forces from
the adjacent layers. One-layer models with
their competing Marangoni (thermocapil-
lary) and Rayleigh (buoyancy) instability
mechanisms are described on pages 4–10.
Page 10 and the following 230 pages give
the authors’ intensive examination of insta-
bilities in multilayer convection problems.
The opening sections in Chapter 2 begin
with a linear analysis of Rayleigh instabili-
ties in the basic two-layer model. Successive
sections incorporate more effects and cor-
responding instabilities into the model. In
the simplest cases, where the interface is
assumed fixed, stability curves are given
for many different parameter ranges and
descriptions of the notable instabilities (in-
cluding oscillatory instabilities and anticon-
vection) are accompanied by figures illus-
trating the streamlines and isotherms in
each situation. When Marangoni effects,
deformable interfaces, and surfactants are
 BOOK REVIEWS
included, full results have not been worked
out in all cases, but the authors provide a
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
good overview and references to the most re-
cent experimental and analytical/numerical
studies.
Chapter 3 gives the stability analysis of
the basic three-layer models, i.e., three lay-
ers of immiscible fluids with undeformable
interfaces sandwiched between two rigid
fixed-temperature plates. Notably, this
chapter describes how this system was
studied in experiments on the international
Spacelab in 1994. Chapter 4 describes the
further complications that arise when de-
formations of the interfaces are allowed
in this coupled system. Behaviors such
as finite-amplitude traveling waves on the
interfaces are obtained from analysis and
numerical simulations of the nonlinear am-
plitude equations derived here. Chapter 5
goes on to give stability results for two- and
three-layer models subject to additional
lateral temperature gradients. The book
concludes with Chapters 6 and 7, where
thin film models and the outlook for fur-
ther work on problems in biology and other
applications are discussed.
The authors make a good case for the
need to study these more complicated prob-
lems as several classes of instabilities caused
by the coupling of layers across interfaces
cannot occur in one-layer models. They
provide a detailed presentation with careful
considerations of the limitations of the mod-
els (such as the relation of deformable in-
terfaces to the Boussinesq approximation).
It is unavoidable that this book has some
overlap with some of the authors’ previous
work [2], but it is clear that it represents a
significant advance with more depth of anal-
ysis and a greatly extended set of models
given in a systematic presentation. While
this book is part of Springer’s Monographs
in Mathematics series, I believe it will be
warmly welcomed by physicists and engi-
neers. It is a rich and thorough guide to
the diverse and complex physical behaviors
that occur in RBC problems. These prob-
lems are of interest to physicists and applied
mathematicians alike, though readers may
be hard-pressed to try to directly extend
the results to other problems as most of the
analysis is done very specifically for RBC.
This book will serve as a valuable reference
363
for all researchers seeking to understand
the nature and complexities of this class of
problems.
REFERENCES
[1] P. Colinet, J. C. Legros, and M. G.
Velarde, Nonlinear Dynamics of
Surface-Tension-Driven Instabilities,
WILEY-VCH Verlag, Berlin, 2001.
[2] A. A. Nepomnyashchy, M. G. Velarde,
and P. Colinet, Interfacial Phe-
nomena and Convection, Chapman
Hall/CRC Monogr. Surv. Pure Appl.
Math. 124, Chapman & Hall/CRC,
Boca Raton, FL, 2002.
THOMAS P. WITELSKI
Duke University
The Generic Chaining. By Michel Tala-
grand. Springer-Verlag, Berlin, 2005. $99.00.
viii+222 pp., hardcover. ISBN 3-540-24518-9.
One of the topics that are at the same time
very challenging and of significant impor-
tance in probability theory, analysis, and
ergodic theory is the study of inequalities
involving suprema of families of random
variables (or, more generally, real-valued
measurable functions).
Let T be a nonempty set that has at least
two elements, and let (Xt )t∈T be a family of
distinct random variables defined on some
probability space (Ω, F, P ). If T is a finite
set, then clearly E supt∈T Xt is well defined;
if T is not finite, set
E sup Xt = sup E sup Xt | F ⊆ T, F finite .
t∈T t∈F
The main theme of the book under review
is a delicate but powerful procedure created
by the author in order to simplify the proof
that he obtained in [7] of a long-standing
conjecture of Fernique [4].
Assume that (Xt )t∈T is a centered Gaus-
sian process, and let d : T × T → R
be a distance on T defined by d(s, t) =
(E((Xs − Xt )2 ))1/2 for every s ∈ T and
t ∈ T.
A very old problem that has its origin
in Kolmogorov’s work is to find a strictly
positive function Γ(T, d) of the metric space
(T, d) (that is, a quantity that depends only
 364 BOOK REVIEWS

on (T, d)) which has the property that there sequence of partitions of T , and using these
exists a strictly positive universal constant partitions to define Γ(T, d) (for the inequal-
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php

c, c > 1, such that for every centered Gaus- ities (1)).

sian process (Xt )t∈T as above, the inequal- After discussing in detail the generic
ities chaining method and its use in obtaining
 
1
(1) Γ(T, d) ≤ E sup Xt ≤ cΓ(T, d)
c t∈T
hold true. Note that the problem makes
sense because if (Xt )t∈T is a centered Gaus-
sian process, then E(supt∈T Xt ) > 0.
Substantial progress in studying the
above-mentioned problem was made by
Dudley in [3], who obtained an upper bound
for E(supt∈T Xt ) in terms of (T, d) only;
the inequality is of significant independent
interest and is known today as Dudley’s
entropy bound. Then, in 1975, Fernique
(see [4]) published another upper bound for
E(supt∈T Xt ) also in terms of (T, d) only;
Fernique obtained the bound by defining
and using a certain probability measure
on the Borel σ-algebra of (T, d), measure
which is known as a majorizing measure,
and he conjectured that the measure (ac-
tually, his bound) can be used to define
Γ(T, d) that yields the desired upper and
lower bounds for E(supt∈T Xt ). In 1987
Talagrand (see [7]) published a proof of
Fernique’s conjecture, thus solving com-
pletely the long-standing problem of defin-
ing a function Γ(T, d) for which there exists
c such that (1) is valid. However, this is
not the end of the story. Soon after Ta-
lagrand’s achievement was acknowledged,
it was felt that Fernique’s majorizing mea-
sure was difficult to grasp and that a simpler
function Γ(T, d) should be obtained. Again,
the author of the book under review found
the solution. He devised a method for con-
structing a function Γ for which there exists
a universal constant c such that (1) holds,
a method which is simple enough to be
easy to understand, and at the same time
sophisticated and flexible enough to have
many other applications besides obtaining
the above-mentioned function Γ and con-
stant c. The book under review is about
this method and some of its applications,
and yes, you guessed right, the method is
called the generic chaining method; hence,
the title of the book.
Roughly speaking, the generic chaining
method consists of cleverly constructing a
the inequalities (1) for Gaussian processes,
the author discusses various applications of
the method. Some of the applications are
the following:
• obtaining various inequalities for p-
stable processes;
• an extension of a result of Ossiander
[6];
• applications to matchings that include
(among other results) an extension
of a theorem of Ajtai, Komlós, and
Tusnády [1];
• applications involving Bernoulli pro-
cesses;
• applications to infinitely divisible pro-
cesses involving an extension of the
generic chaining method to the case
in which, given a family (Xt )t∈T , one
has to consider a family of distances
on T (rather than just one distance);
• applications to functional analysis
that include a strengthening of a result
of Bourgain [2].
Even though the book under review ap-
peared in 2005, people have already started
to use the monograph’s approach to the
generic chaining (see Klartag and Mendel-
son’s paper [5]).
The results discussed in the book involve
a tremendous amount of creativity, and the
author has made every possible effort to ex-
plain how he arrived at the results he deals
with.
I believe that the monograph is of interest
to anyone who works even remotely with in-
dexed families of maps (be those maps con-
tinuous functions, measurable functions, or
measures) because, while discussing spe-
cific results, the author also offers a wealth
of general principles for dealing with such
families, principles that could be used in re-
lated approaches but in very different set-
tings. Thus, in my opinion, the book is
of interest to almost everyone who works
in probability, analysis, or ergodic theory
even if she/he has not done any research
involving the topics covered by the volume
 BOOK REVIEWS
under review. Actually, I agreed to review
the work, even though my research interests
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
are very far from the results discussed in
the monograph, exactly because I want to
emphasize the fact that the book is of in-
terest to a larger audience, and not only to
people who publish in the areas presented
in the work.
As we all know, there is a perception
in the mathematical community (especially
among the people who have written at least
one book) that one cannot find flawless
monographs in mathematics. In the work
under review, the author refers repeatedly
to an appendix that is assumed to be part
of the book, and the bad news is that there
is no appendix in the volume. However,
the good news is that the appendix can
be downloaded at http://www.math.ohio-
state.edu/˜talagran/appendix.pdf.
REFERENCES
[1] M. Ajtai, J. Komlós, and G. Tusnády,
On optimal matchings, Combinatorica,
4 (1984), pp. 259–264.
[2] J. Bourgain, Bounded orthogonal sets
and the Λp -problem, Acta Math., 162
(1989), pp. 227–246.
[3] R. M. Dudley, The sizes of compact sub-
sets of Hilbert space and continuity of
Gaussian processes, J. Funct. Anal., 1
(1967), pp. 290–330.
[4] X. Fernique, Regularité des trajectoires
des fonctions aléatoires gaussiennes,
Lecture Notes in Math. 480, Springer,
Berlin, 1975, pp. 1–96.
[5] B. Klartag and S. Mendelson, Empiri-
cal processes and random projections,
J. Funct. Anal., 225 (2005), pp. 229–
245.
[6] M. Ossiander, A central limit theorem un-
der metric entropy with L2 bracketing,
Ann. Probab., 15 (1987), pp. 897–919.
[7] M. Talagrand, Regularity of Gaussian
processes, Acta Math., 159 (1987), pp.
99–149.
RADU ZAHAROPOL
Mathematical Reviews
Karl Pearson: The Scientific Life in a Sta-
tistical Age. By Theodore M. Porter. Princeton
University Press, Princeton, NJ, 2004. $35.00.
viii+342 pp., hardcover. ISBN 0-691-11445-5.
365
This biography of Karl Pearson, a key fig-
ure in the development of modern statis-
tics, is much more about the person than
about his statistical accomplishments. Of
its ten chapters, only Chapters 8 and 9
contain more than brief mention of his sta-
tistical work, and a reader seeking details
regarding his statistical contributions will
find very few even in the chapters covering
his career as a statistician. This reflects the
author’s intent. The author convinced me
that Pearson’s life was presented as Pear-
son saw it: a unified story, a life lived whole
that couldn’t be separated into “student
of philosophy and history,” “poet,” “pro-
fessor of statistics,” “husband and father,”
etc. By the end of the book, I felt that I
knew this complex person well. In the next
few paragraphs, I summarize what I learned
about Pearson from the book to give read-
ers a sense of his complexity; references are
to pages in the book. This summary is no
substitute for reading the book itself!
Pearson was born in 1857 to an ambi-
tious lawyer father and a sensitive mother.
The book contains little about his child-
hood; he was “plagued by delicate health,”
“earnest and highly sensitive” (p. 20). As a
teenager he was tutored at Cambridge, with
an emphasis on mathematics, but entered
King’s College (Cambridge) where human-
istic studies were prominent and mathemat-
ical offerings limited. He read poetry and
philosophy and was impressed by Goethe,
Rousseau, Spinoza, and Hegel. However, he
also studied mathematics and physics, scor-
ing third in the Mathematical Tripos, a key
exam. He was “almost despondent about
the result” (p. 41), though his tutors and
others assured him that the top three were
close, and he would have scored higher if
he had not wisely decided to pursue a more
general course of study instead of focusing
only on the Tripos. At this point in his life,
he was conflicted about philosophy, reli-
gion (moving from Christianity towards ag-
nosticism), and career (law? mathematical
physics?). He spent a year studying in Ger-
many in hopes of resolving these conflicts.
His efforts after returning from Germany,
still in his twenties, focused on philosophy,
literature, history, and politics. Although a
restrained person, he embraced enthusiastic
fervor (channeled by reason) in his intellec-
 366 BOOK REVIEWS
tual life. At age 23, he wrote a fictionalized
autobiography, a romantic novel called The
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
New Werther. Pearson’s poetic bent, love of
nature, and respect for Goethe as not just a
writer but a moral teacher are all evident in
this novel. But critics were not impressed.
Pearson became a socialist, embracing so-
cialism based on science and reading Marx
with admiration. Marx declined an offer
from Pearson to translate Das Kapital into
English. Pearson was attracted to Catholi-
cism during this period but remained reli-
giously conflicted. He wrote “a passion play
for the nineteenth century” (p. 69; like his
novel, a critical failure) but also studied for
the bar exam (though he soon decided he
hated law). His research interests during
this period were divided between German
history and mathematical physics, and he
was under consideration for a position as
lecturer in German literature and history
at Cambridge when he was offered, and ac-
cepted, a permanent appointment as pro-
fessor of applied mathematics at University
College London in 1884. However, he con-
tinued to publish historical essays and gave
extension lectures in cultural history. He
also “lectured to workingmen’s associations
and to such groups as the Sunday Lec-
ture Society” (p. 109). He knew and cor-
responded with George Bernard Shaw and
other Fabian socialists and admired, though
never met, William Morris. However, “he
was too strong-willed and idiosyncratic to
join anyone else’s movement” (p. 109).
Pearson’s “interest in women’s issues and
curiosity about women’s feelings and per-
ceptions” led him to take the initiative in
establishing a “Men and Women’s Club”
in 1885. The club was to be “devoted to
research and frank discussion on issues con-
cerning women and relations between the
sexes” (p. 125). By 1889, when the club
dissolved, Pearson was better known for
his writings on the history and status of
women, which he considered his most im-
portant area of research, than for his math-
ematics. In spite of his attempt to make
the club a place for purely intellectual dis-
cussions between men and women, he fell
in love with Maria Sharpe, secretary of the
club, and married her in 1890.
However, he continued to apply his math-
ematical skills to physics, developing the-
ories both physical and metaphysical. In
1892, at age 35, he published a book, The
Grammar of Science, examining the foun-
dations of science. Perceptions and con-
ceptions were the basic elements of sci-
ence. Experience provided perceptions, but
the human mind produced the phenomena
and laws of nature through which these
perceptions were explained. The scientific
method involved the classification of facts
and search for generalizations, not only in
physics but in any investigation, e.g., of
social or economic questions. He soon re-
alized that probability and mathematical
statistics provided a model of rigorous rea-
soning that could be applied to this search
for generalizations, and he committed him-
self to the program of quantitative research
that would occupy him for the rest of his
life. I invite you to read about that part of
his life in the book that is the subject of
this review.
What I have failed to capture in the
above summary is captured very well in
Porter’s book. Pearson’s temperament and
self-image determined the course of his pro-
fessional and personal life and made it a
unified story. It was a story left to Porter
to tell. Pearson wrote Galton’s life in three
volumes but never wrote his own. Although
he kept journals, wrote many letters, and
even wrote an autobiographical novel at a
young age, he never wrote an autobiogra-
phy. Pearson “insisted on the integrity and
coherence of his career and vision” (p. 314)
and hoped to be remembered for it. Yet
he was the founder of an approach that
provided a routine, impersonal method for
arriving at scientific conclusions. So it is
not surprising that a fear he expressed to-
ward the end of his life has been realized.
His name is associated not with a person
but with a correlation coefficient and other
particular statistics and statistical meth-
ods. Hopefully Porter’s book changes that
for his readers.
Perhaps I enjoyed the book particularly
because I also studied philosophy before I
studied statistics but was always interested
in science. As was, I believe, the case for
Pearson, one of the appeals of a career in
statistics for me was that, as a statistician,
one can be involved in contributing to the
solutions of diverse scientific problems. My
 BOOK REVIEWS
philosophical, religious, and political back-
ground made Porter’s book easier reading
Downloaded 06/18/17 to 161.130.252.114. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php
for me than it may be for some who are
less familiar with his historical references.
However, the book contains many footnotes
and a long bibliography, so the reader can
find background information while or after
View publication stats
367
reading the story. There is a certain amount
of jumping back and forth between different
parts of Pearson’s life in many chapters, but
I did not find that problematic. It was just
part of Porter’s method of telling the story.
JUDITH ZEH
University of Washington