SIAM REVIEW c 2004 Society for Industrial and Applied Mathematics

Vol. 46, No. 3, pp. 549595

Book Reviews
Edited by Robert E. OMalley, Jr.
Featured Review: Selected Books on Advanced Engineering Mathematics
Advanced Engineering Mathematics with MATLAB. Second Edition. By Dean
G. Duffy. Chapman & Hall/CRC, Boca Raton, FL, 2003. $99.95. xvi+818 pp., hardcover.
ISBN 1-58488-349-9.
Advanced Mathematics for Engineering and Science. By C. F. Chan Man Fong, D. De
Kee, and P. N. Kaloni. World Scientic, River Edge, NJ, 2003. $58.00. xiv+880 pp., softcover.
ISBN 981-238-292-5.
Mathematical Methods for Scientists and Engineers. By Donald A. McQuarrie. Univer-
sity Science Books, Sausalito, CA, 2003. $90.00. xiv+1161 pp., hardcover. ISBN 1-891389-
24-6.
Advanced Mathematics and Mechanics Using MATLAB. Third Edition. By Howard
B. Wilson, Louis H. Turcotte, and David Halpern. Chapman & Hall/CRC, Boca Raton, FL, 2003.
$89.95. xiv+678 pp., hardcover. ISBN 1-58488-262-X.
Engineers generally complete their required mathematics courses in their rst two
years of college. Many departments of mathematical sciences also oer elective classes
to teach more mathematics to juniors and seniors in these majors. The same or similar
courses help prepare rst- and second-year graduate students for masters- or doctoral-
level study of engineering and science.
In the last year alone, quite a few mathematicians, scientists, and engineers have
written textbooks designed for such courses. Reviewing some of them provides a
springboard to re-evaluate
1. what subjects to include in a course on advanced engineering mathematics
(AEM);
2. how to teach them, vis-`a-vis organization of topics, level of detail, amount of
mathematical rigor, degree of application, use of computer algebra systems
(CAS); and
3. which textbook to use, if any.
The answers to these questions depend crucially on the purpose of and clientele
for the course. One should cultivate mathematical sophistication in engineers and
scientists by emphasizing how the subjects form a coherent structure. Alternatively,
if AEM means a collection of various tools, the instructor should convey what broad
purposes it serves. To tailor the content to his or her particular audience and to
take advantage of all of the best resources available, the professor should interrelate
material from various textbooks specializing in single subjects. The students can then
view an AEM text as a reference for an assemblage of topics.
Publishers are invited to send books for review to Book Reviews Editor, SIAM, 3600 University
City Science Center, Philadelphia, PA 19104-2688.
549
550 BOOK REVIEWS
Table 1 Curriculum proposals.
Duy A Laplace transform / separation of variables in PDEs /
vector calculus or linear algebra
Duy B SturmLiouville problems / separation of variables in PDEs /
vector calculus or linear algebra
Duy C Complex variables / Fourier series /
Fourier, Laplace, z-, and Hilbert transforms
Chan Man Fong et al. Series solutions and special functions / vector and
tensor analysis / PDEs / numerical methods
Kreyszig, 1999 Sequence of courses: (1) ODEs, (2) linear algebra and vector
analysis, (3) complex analysis, (4) numerical methods
Kreyszig, 1962 Sequence of courses: (1) ODEs, (2) vectors and matrices
(3) Fourier series and PDEs, (4) complex analysis
Hildebrand, 1949 Laplace transform / numerical ODEs / series solutions and
special functions / boundary-value problems and orthogonal
functions / vector analysis / PDEs / complex variables
Gross A Complex numbers and functions / linear analysis / systems of
ODEs / divergence and Stokess theorems
Gross B Series solutions and special functions / method of dominant
balance / SturmLiouville theory / Fourier series and
transform / PDEs
1. What Should We Teach in AEM?
The question of what to cover in a course on AEM diers from the question of what
to include in a textbook on AEM. Indeed, some texts contain much elementary mate-
rial from calculus and ordinary dierential equations (ODEs), as well as an array of
advanced topics far too extensive to address in a single course. The book may benet
the student more as an enduring reference than as a guide to careful study during one
or two semesters.
Two of the volumes under review suggest curricula explicitly. Dean G. Duy
gives the more detailed guidance, proposing three courses in AEM for which his book
can serve. (See Table 1.) His Course A assumes familiarity with basic dierential
equations, perhaps in the context of an expanded calculus class. This version of AEM
includes the Laplace transform and separation of variables for the heat, wave, and/or
Laplace equation. His more advanced Course B has a course on dierential equations
as a prerequisite. Course B teaches (or reteaches) separation of variables from the
point of view of SturmLiouville theory and eigenfunction expansions. Both Courses
A and B conclude with a chapter on either vector calculus or linear algebra. The
latter ends with a ve-page section on systems of linear dierential equations.
Duy has used the text himself recently in a Course C for electrical, communi-
cations, and systems engineers. It opens with a unit on complex variables, followed
by Fourier series and Fourier, Laplace, and z-transforms. In this second edition, he
has added a new chapter on the Hilbert transform as a concluding unit for just this
audience. Two of the transform chapters distinguish the contents of Duys book
from the contents of the other three under review, as Table 2 indicates.
Duy has added another two chapters to this new edition, as well. He says they
permit the book to serve as a text for a course in dierential equations. Also unlike
the 1998 version, this edition integrates MATLAB routines into the exposition.
BOOK REVIEWS 551
Table 2 Chapters included in only one of the four texts.
Duy Z-transform, Hilbert transform
Chan Man Fong et al. Tensor analysis, special topics (in advanced classical
mechanics, plus statistical mechanics)
McQuarrie Functions dened as integrals, probability, statistics
Wilson et al. Use of MATLAB, bending analysis of beams of general
cross section, nonlinear optimization applications
C. F. Chan Man Fong, D. De Kee, and P. N. Kaloni suggest an undergraduate
course that starts with a unit on series solutions and special functions, followed by
topics in vector and tensor analysis, partial dierential equations (PDEs), and numer-
ical methods, including numerical ODEs. They do not include in this outline their
chapters on numerical PDEs, complex variables, or calculus of variations. The book
also contains a review of calculus and ODEs, as well as a collection of short sections on
special topics (in advanced classical mechanics and statistical mechanics). A couple
of the subjects do not appear in the other texts under review, as Table 2 shows. In
addition, Chan Man Fong et al. devote a consolidated section (about 20% of their
book) to numerical methods, whereas Duy interweaves numerical overviews into sec-
tions on dierential equations and transforms. Chan Man Fong et al. also include
some more advanced topics in PDEs, like Mellin transforms, similarity solutions, and
a systematic treatment of Greens functions, unlike the authors of the other books
under review.
Donald A. McQuarrie wrote his book for students who have had 1
1
2
or 2 years
of calculus and little else. He introduces a wide variety of topics, hoping to inspire
the student to later deeper study, suggesting references at the end of each chapter.
Indeed, of the texts under review, only his lacks the word advanced in the title.
The volume opens with over 100 pages of calculus, four times as many as Chan Man
Fong et al. McQuarrie also provides a chapter that gives an elementary introduction
to vectors (before covering the divergence theorem and Stokess theorem, as do the
other authors, in a later chapter). He leaves to the instructors discretion which of
the remaining 19 chapterseach around 50 pages longto cover. Table 2 lists those
chapters that include material not included in the other three texts. Unlike the others,
McQuarries book teaches no numerical methods.
Speaking of numerics, the book by Howard B. Wilson, Louis H. Turcotte, and
David Halpern has a very dierent avor from the other three. As the title suggests,
simulations of mechanical systems receive a great deal of attention. The authors rec-
ommend their text as a reference or supplementary text in computationally oriented
courses emphasizing applications. Students in such courses must have backgrounds
in Euclidean geometry, Newtonian mechanics, and some mathematics beyond calcu-
lus. As a secondary text, the book does not impose a particular curriculum. Also, it
emphasizes examples over mathematics and methods. Still, it does rather come to a
consensus with the other authors on which topics to include under the title advanced
mathematics.
As one would expect, however, it treats those topics totally dierently. For ex-
ample, the chapter entitled Summary of Concepts from Linear Algebra spends a
total of two pages on an introduction, vectors, norms, linear independence, and rank.
Two subsequent sections include about fteen pages each of examples of least squares
552 BOOK REVIEWS
Table 3 Topics commonly included in AEM texts.
Most AEM texts Many AEM texts
Linear algebra Numerical methods
Vector calculus Calculus of variations
ODEs Probability
PDEs Statistics
Complex variables Optimization
ts and eigenvalue problems, respectively. After a few pages on computing natural
frequencies for a rectangular membrane, the chapter closes with only two pages on
column space, null space, orthonormal bases, and singular-value decomposition.
This third edition has integrated time-dependent solutions of linear PDEs and
corresponding animations into the same chapter outline as the previous edition. Per-
haps David Halpern, who did not coauthor the previous two editions, contributed this
new material. See Table 2 for a list of chapters particular to this book.
Upon confronting her 15 pounds of new books on AEM, some applied mathe-
maticians asked this reviewer, Why use something other than Kreyszig? Looking
at this question from the curriculum standpoint, one observes that Erwin Kreyszigs
text [8] contains the topics in Duys Courses A and B, as listed in Table 1. His book
does not address some of the other entries in that table, however. It lacks the z- and
Hilbert transforms of Duys Course C and the tensor analysis that Chan Man Fong
et al. advocate.
In the preface to the eighth edition of his AEM text, Kreyszig asserted that matrix
algebra, numerical methods, statistics, and graph theory play larger and larger roles
in engineering applications. He does not suggest how to integrate these elements into
a course on AEM, however. Instead, he proposes the four-semester sequence in Table
1, namely, (1) ODEs, (2) linear algebra and vector analysis, (3) complex analysis, and
(4) numerical methods.
As another alternative to a multitopic course on AEM, Kreyszig suggests several
one-semester courses with sharper focus, e.g., (1) numerical linear algebra, (2) (linear)
optimization, and (3) graphs and combinatorial optimization. He also includes a
class on Fourier series and PDEs in the list. As Table 1 shows, Kreyszigs rst
edition included that course in the four-semester sequence. Later, numerical methods
replaced it.
Like Kreyszigs rst edition in 1962, many popular AEM texts have crystallized
around the subjects presented in Table 3, namely, linear algebra, vector calculus,
ODEs, PDEs, and complex variables. Some, like Kreyszigs current edition, also
include numerical methods, probability, statistics, and optimization. The textbooks
under review largely t this framework, laid out in Francis B. Hildebrands Advanced
Calculus for Engineers [6] in 1949 (except that the latter includes no linear algebra).
Hildebrand chose the textbooks title to represent areas of mathematics based
upon a sound working knowledge of elementary calculus and . . . important in a num-
ber of elds of application (p. v). In fact, the second edition (1976), still in print,
bears the title Advanced Calculus for Applications. Hildebrand presents the con-
tents as a course curriculum, suggesting one may cover complex variables at the
end or at the beginning and that some may consider the chapter on linear ODEs
as a review. He promises an integrated treatment with PDEs of mathematical
physics as the end point and the remaining chapters, including the Laplace transform,
BOOK REVIEWS 553
numerical ODEs, series solutions and special functions, boundary-value problems and
orthogonal functions, and vector analysis, as steps upon the way. (See Table 1.)
The tables of contents of textbooks and the ideas listed in Table 1 can only
provide starting points for the design of a successful curriculum in AEM. To teach an
appropriate course, one must know ones audience. In 1990, Garfunkel and Young [4]
estimated that 170,000 students in the United States enroll annually in mathematics
courses beyond calculus outside of mathematics departments. Their study shows that
mathematics faculty and curricula are often seen as at best irrelevant and at worst
counterproductive. One hopes that mathematicians today communicate with the
markets their courses serve to ensure that our colleagues outside mathematics view
AEM as not only relevant, but also valuable, as the instructors must intend.
This reviewer has recently engaged in such discussions with engineers and scien-
tists inside and outside of her home institution and inside and outside of academia.
She suggests that doing similarly will help other mathematicians carry out the fol-
lowing steps:
1. Identify the Purpose of AEM. For example, if the course targets undergradu-
ates going directly to industry, some practicing engineers suggest that topics
include complex numbers and functions, numerical root-nding, linear al-
gebra and numerical linear algebra, qualitative and numerical methods for
solving dierential equations, and/or the real and complex Fourier series and
Fourier transforms. One could integrate technological tools widely used in the
workplace, like graphing calculators and MATLAB. This reviewer has some
of these elements in her Course A at The University of Akron. (See Table 1.)
If, on the other hand, the course prepares students for graduate study, par-
ticularly of mechanical or chemical engineering, then consider teaching series
solutions and special functions, SturmLiouville theory, and PDEs (a subset
of Hildebrands curriculum in Table 1). This reviewers Course B follows this
outline. (See Table 1.) One of the former undergraduates in the class sent
an unsolicited e-mail saying, It is quite a luxury to be this well prepared
for a class [conduction heat transfer] in his graduate program. Colleagues
have received similar feedback when teaching Course B, suggesting that a
pioneering AEM curriculum from 1949 remains relevant today.
2. Identify the Students Backgrounds. For undergraduates, set as the prereq-
uisites the requirements for their majors. If graduate students have passed
mathematics qualifying exams, identify the content. Find out whether the
students have access to and know how to use a CAS or at least a graphing
calculator.
This reviewers Courses A and B (Table 1) reect that the engineering and sci-
ence undergraduates who enroll have previously taken introductory courses in
dierential equations and linear algebra. They have strong academic records
and often minor in applied mathematics. Engineering graduate students
in the process of preparing for mathematics qualifying exams also take her
courses, cross-listed as both undergraduate and graduate courses.
3. Extend the Students Knowledge of Prerequisite Fields. One could easily t a
one-semester course (or more) under this heading.
This reviewer chooses to unite familiar material from linear algebra and ele-
mentary ODEs via an integrated discussion of the following:
1. Vector spaces
2. Linear operators and linear operator equations
3. Basis and dimension
554 BOOK REVIEWS
4. Inner products, orthogonality, induced norms, and orthonormality
5. Eigenvalues and eigenvectors
Table 1 summarizes these as linear analysis, in the spirit of Donald L. Krei-
der et al., who wrote their book [7] for courses in advanced engineering math-
ematics. The topics above constitute a highly condensed and interwoven
treatment of material distributed in six of their chapters.
1
However, this re-
viewers Course A addresses the items in the list above in the context of vector
spaces over the complex numbers, having opened the course with introductory
material in complex variables (Table 1).
A sophomore-level course in ODEs usually includes only a couple of the fol-
lowing units:
1. Laplace transform
2. Series solutions and special functions
3. Linear and nonlinear systems and phase plane analysis
4. Fourier series and separation of variables for solving PDEs
5. SturmLiouville problems, eigenfunction expansions, and separation of
variables for solving PDEs
6. Further methods for solving PDEs
7. Numerical methods
8. Asymptotic methods
Teach some of the remaining items. This reviewer covers all of these areas in
an introduction to dierential equations, together with Courses A and B in
AEM. (For the latter, see Table 1.)
Consider picking up vector analysis where the calculus course ended. This
reviewers Course A covers the divergence theorem and Stokess theorem and
uses them to reformulate integral conservation laws in dierential form. (See
Table 1.)
4. Introduce the Students to Subjects New to Them. Include linear algebra if they
havent taken it. The Mathematics Working Group of the European Society
for Engineering Education (SEFI) specically recommends linear transforms
and least-squares tting [10] as part of a core curriculum for engineers.
The group also considers introductory probability and statistics as core areas,
as well as such discrete math topics as logic, sets, induction, recursion, and
graphs [10].
Choose from among countless other candidates. Table 3 includes some. This
reviewer teaches complex numbers and functions in her Course A (Table 1).
Cover topics of particular interest to the audience, e.g., discrete transforms
for electrical engineers, residue theory for control theorists, or tensor analysis,
group theory, or Greens functions for physicists.
2. How Should We Teach AEM?
Kreysig [8] wrote:
It would make no sense to overload students with all kinds of little things
that might be of occasional use. Instead it is important that students
become familiar with ways to think mathematically, recognize the need
for applying mathematical methods to engineering problems, realize that
1
The chapters cover (1) real vector spaces, (2) linear transformations and matrices, (3) the general
theory of linear dierential equations, (7) Euclidean spaces, (8) convergence in Euclidean spaces, and
(12) boundary-value problems for ODEs.
BOOK REVIEWS 555
mathematics is a systematic science built on relatively few basic concepts
and involving powerful unifying principles, and get a rm grasp for the
interrelation between theory, computing, and experiment.
One can debate whether mathematics in fact contains relatively few basic concepts,
but nobody can deny that AEM courses often contain too many narrow methods, ob-
scuring the powerful unifying principles. However, in the same preface quoted above,
Kreyszig described the chapters in his book as quite independent, accommodating a
variety of courses. The same holds true for most AEM texts, so bringing out the
connections falls heavily on the instructor.
This reviewer tells students that the word advanced appears in the course title
not only because they will study subjects beyond their basic requirements, but also
because they will acquire a more sophisticated understanding of material they have
studied already. They should see the connections among subjects they previously
viewed as disjoint. Such an appreciation will serve them well in future encounters
with mathematics in their elds.
For example, the one-semester Course A that this reviewer proposes (Table 1)
introduces complex numbers and functions to lead into vector spaces over the com-
plex numbers. The unit on linear analysis outlined above (section 1) integrates the
students prior knowledge of linear algebra and elementary ODEs. Eigenvalues pro-
vide a point of transition to exact and qualitative treatments of linear and nonlinear
ODE systems, including phase plane analysis of nonlinear equations. The discussion
of direction elds will serve as a new application of the vector elds the students rst
encountered in multivariable calculus. They can then review the operations of diver-
gence and curl on vector elds, together with the gradient. Because third-semester
calculus often cannot cover them thoroughly, the divergence theorem and Stokess
theorem come next. Using them to derive PDE conservation laws ends the class with
a preview of further work to do in PDEs.
This reviewer does some of the further work in Course B (Table 1), designed to
share Hildebrands framework discussed above (and listed in Table 1). In particular,
the heat, wave, and Laplace equations constitute the goal, with the ODE topics as
steps toward that end. Series solutions lead to the derivation of Legendre polynomials
and Bessel functions. To complement expansions about ordinary and regular singular
points, the curriculum also includes the method of balance at irregular singular points,
in particular at innity. Carl M. Bender and Steven A. Orszag [2] covered these topics
in a thorough chapter on local analysis.
In the next unit, SturmLiouville theory contrasts with local analysis; one can
emphasize that a partial sum of an eigenfunction expansion of the solution to an in-
homogeneous SturmLiouville problem approximates the solution well on the whole
problem domain. The Fourier series becomes a special case of eigenfunction expan-
sion. The course next introduces the Fourier transform. The last third of the course
involves solving PDEs via separation of variables, the Laplace and Fourier transforms,
dAlemberts method, and the method of characteristics.
The Courses A and B that this reviewer teaches (Table 1) have the titles AEM
I and AEM II. Part II does not depend on Part I. In fact, Hildebrands book shows
that Course B need not rely even on linear algebra. One can make many connections
informally with the linear analysis of Part I, however, to enrich the experience of
students taking both semesters.
AEM should give a big (probably rough) picture of the wide topics (such as those
in Table 3) that it addresses, even if it can only include a small slice in detail. Students
556 BOOK REVIEWS
need to develop the ability to tackle basic problems and interpret solutions (sometimes
qualitatively), see the limitations of the methods they study, and have some idea of
how to transcend these limitations.
For example, students learning about systems of ODEs can reinforce their per-
spectives on ODEs as a whole, e.g., that many linear problems yield to known methods
(with solutions to constant-coecient problems especially accessible) and that solu-
tions to inhomogeneous problems build on the homogeneous. They will easily classify
dierential equations and give examples of corresponding physical problems. They
will know that one nth-order equation can be re-expressed as a system of n rst-
order equations. Students will use phase plane analysis to show qualitative behaviors
clearly, recognizing that in many physical systems varying a parameter can produce
an abrupt change in equilibrium solution behavior. They will know when linearization
gives reliable information and how much error the process accrues. They will know
what numerical methods are, when to turn to them, and that software can execute
some of them quite directly.
In connecting dierent subjects and giving a wide perspective, as this reviewer
advocates, one must guard against covering material too theoretically. Certainly AEM
should use proof very judiciously. Contrary to an instructors plans, a high degree of
rigor can interfere both with unifying dierent subjects and with providing a broad
perspective. Hildebrand [6] took the following approach:
It is not infrequently true that the natural way of discovering a particular
technique or relationship is by no means the most ecient way of estab-
lishing its validity. The compromise adopted here, in such cases, usually
consists in showing in as direct a way as possible that the desired result
is plausible, and in then stating conditions under which the result can be
rigorously established. Thus, the engineering reader may proceed in the
direction along which he might have discovered a given fact for himself. . . .
In Garfunkel and Youngs study, those outside mathematics departments commonly
complained that [m]athematics faculty teach mathematics as an art with full abstrac-
tion, not as a tool [4]. They wanted applications, in contrast to overly generalized
presentations that are not (usually) presented as being useful or interesting to a prac-
tical person [4].
One must use physical context to motivate students. On the other hand, one must
not devote so much time and eort to one setting as to make a technique appear too
narrowly applicable. Such an approach would contribute to the impression of many of
Garfunkel and Youngs respondents that [m]athematics courses do not give students
the knowledge or the mathematical maturity for further work [4]. P. C. Cretchley,
A. J. Roberts, and C. J. Harman [3] note that where applications are used to motivate
learning, hasty treatment of enough material only for the task in hand does not invite
the longer-term higher-order reection that ensures deeper secure learning. They
say, Given the dangers of fragmentation of content and shallow treatment of core
topics, we argue for coordinated consolidation of core skills [3].
In disagreement with this reviewer, one could argue that consolidating core skills
in a eld of mathematics requires taking an entire course in it, and that coherence
comes from taking such courses sequentially or in some carefully prescribed pattern.
However, this paradigm applies only to the mathematical sciences major. Indeed,
many departments teach their own advanced mathematics courses expressly because
(as quoted by Garfunkel and Young [4]) [t]here is not room. . . for every student
to take the separate courses in dierential equations (ordinary and partial), vector
BOOK REVIEWS 557
and tensor analysis, complex variables, Fourier series, probability. . . . [A]ll these are
covered in our one-year course by omitting the detailed proofs and generalizations.
In such a setting, another way to unify the topics is through the eld of application
itself. After identifying the purpose of the course, as outlined in the previous section,
this approach may prove appealing.
Regardless, one must integrate applications into the course. Although the name
AEM shows that the subject matter is mathematics, the word engineering appears,
too. The instructor must communicate the reason: how the material arises in phys-
ical problems. A biomedical engineer in this reviewers class said he enjoyed the
realization that certain techniques apply not only to certain problems in a given eld,
but arise in many elds. One can make this point without spending time on the
details.
Returning to the quotation that opened this section, one would like to show
students the relationships among theory, computation, and experiment. Including
experiment in a course on AEM may prove a tall order, but one can easily emphasize
physical interpretations of models and solutions. Hildebrand recognized the impor-
tance of computation in his 1949 book [6], including a chapter on numerical methods
for solving ODEs. Such material constitutes a vital part of the curriculum today:
Engineers use complicated models that generally demand numerical treatment. The
SEFI report (p. 48) noted considerable benet from the integration of numerical
methods with their analytical counterparts [10]. It noted that a few steps using a
simple (even simplistic) method like Eulers can teach much about the nature of
numerical solutions and that much of this would probably be overlooked if all the
students did was learn the appropriate syntax to invoke a dierential equation solver
in a commercial software package [10].
Like applications, the CAS in AEM should provide a clearer and more motivating
way of addressing certain topics, not serve as an end in itself. As an example of an
appropriate use of technology, showing direction elds with Maple gives much imme-
diate and useful information about solutions to initial-value problems, particularly in
nonlinear systems of ODEs.
3. From Which Textbook Should We Teach AEM?
First, consider the four texts under review. Duy has highlighted integral and dis-
crete transforms. An instructor whose syllabus reects Course C in Table 1 should
use Duys textbook over other AEM texts, which commonly omit z- and Hilbert
transforms.
In their book, Chan Man Fong et al. pay attention to the needs of physicists who
take AEM. (See Table 2.) Most AEM texts will not appeal to a professor who shares
their interest in tensor analysis, but many established volumes on mathematics for
physics t the bill, such as that of George Arfken (and now his coauthor Hans Weber)
[1]. If the teacher wants that chapter under the same cover as a fairly extensive catalog
of numerical methods, the pool of candidates shrinks, and he or she should consider
Chan Man Fong et al.
Recall that McQuarrie aims to give a rst exposure to many topics, to pique
the students interest for further study. Although the material matches the other
textbooks quite closely, the book has a somewhat more informal tone. The chatty
approach tends to embed denitions in the text. The author sets apart some theorems
(not generally labeled as such) in italics, while he includes others within the discussion
like any other sentence. This works well for the casual read, bringing out a few key
558 BOOK REVIEWS
points. Perhaps some arts-and-sciences readers will enjoy the style of exposition more
than engineering students. The latter group often favors a thorough accounting of
boxed facts.
Instructors who incorporate Mathematica into their classes will appreciate that
McQuarrie sprinkles Mathematica statements here and there. The problems that
require a CAS, however, explicitly permit the use of whichever one the student prefers.
(Some of these are just as easy to solve by hand, for example, nding the gradient of
a simple function.)
The biographies, which the authors wife Carole wrote, make clear connections to
the technical content. Also, unlike biographies in many mathematics textbooks, they
include plenty of facts likely to hold the typical student readers interest.
McQuarries 22 chapters have a nice organization overall. For instance, the early
introduction to complex variables has advantages. However, some scrambling occurs,
as well. For example, the author expresses too much willingness to ll the chapter
on matrices and eigenvalue problems with applications from linear systems of ODEs,
which the subsequent chapter addressesin a sixth section (with section 1 covering
ODEs of rst order and rst degree).
A rst edition will tend to have errors (as will likely any edition of a thousand-
plus pages). An egregious one jumped out at this reviewer in an example on page
495: The list of eigenvectors includes the zero vector, leading to the conclusion that
the double eigenvalue does not have linearly independent eigenvectors.
Material in the book by Wilson et al. can serve as a valuable resource for the
AEM course that provides a brief exposure to numerics. It can also supplement the
more heavily numerical course. In the former case, one could focus, for example,
on the chapter on integration of nonlinear initial-value problems. The instructor
could conveniently insert the chapter, with its accessible succinct sections, into the
curriculum for a qualitative or somewhat more detailed overview. He or she could
also connect it easily (perhaps loosely) with other topics.
The rst section heading contains the somewhat peculiar terminology nonlinear
matrix dierential equations and opens with the usual (for this book) evangelical
statements on user-friendly software. The student may notice that most of the text-
books cited date from decades ago, but the authors make the point that products like
MATLAB make the methods much easier to use.
The chapters brief introduction makes for good reading for engineering students,
bringing out many key broad ideas in applied mathematics and numerical methods: It
addresses appropriate levels of complexity in mathematical models and the sensitivity
of nonlinear systems to small changes in physical parameters. It explains nicely that
analytical techniques cannot produce solutions to most dierential equations and that
many approximate methods derive from series expansions.
The two-page summary of RungeKutta methods gives a nice birds eye view to
the student who has studied the subject in a little more detail and to the novice who
reads carefully and patiently. (Both groups will notice a typographical error, which
also appeared in the previous edition: Reducing the step-size by h [sic] reduces the
truncation error by about a factor of (1/2)
3
= 1/8.)
The text emphasizes the importance of understanding the error inherent in any
numerical method. A short section derives a stability condition quite intuitively for
an nth-order RungeKutta method on the example y

(t) = y for complex. It shows

a little MATLAB program to plot the stability boundary in the complex plane, using
an intrinsic function that nds complex roots, and gives a clear interpretation of the
resulting graphs.
BOOK REVIEWS 559
The chapter on integration of nonlinear initial-value problems also discusses the
trade-os and limitations of reducing step size and gives a good and strictly descriptive
explanation of variable step-size algorithms.
The remainder of the chapter discusses examples of forced oscillation. These
include an inverted pendulum, dynamics of a spinning top, motion of a projectile,
and others.
Wilson et al. include the MATLAB programs in their entirety. Good explanations
bracket many of them, and they include their purposes in comments. Nevertheless, key
commands remain uncommented. The reasonably well trained eye can decipher them,
but some will certainly remain opaque even to the thorough student. Downloading and
running many of these routines will produce output from a black box. The philosophy
of the methods and the interpretation of the results can serve a pedagogical purpose,
but the students might gain only limited insight from experimenting with the codes
themselves.
Having pondered some pros and cons associated with these new books, many
readers will remain committed to their preferred books on AEM. A few will like them
enough to teach multiple courses from them, resulting in nancial savings for the
students. (When this reviewer used Peter V. ONeils book [11], she learned that one
of her students had used an earlier edition for a two-semester sequence on dierential
equations as a freshman in mechanical engineering at National Cheng Kung University
in Taiwan. He liked picking the book up again in AEM.) For example, the book costs
for Kreyszigs four semesters in Table 1 would average $31.99 per term, much less
than the average cost for a book in the single subject.
Generally, however, teaching multiple courses from the same book poses a variety
of challenges, rst and foremost that most students dont have the luxury of enrolling
in three courses beyond dierential equations. Those students will pay too much for
their very heavy dierential equations textbook (e.g., $127.95 and ve pounds for
Kreyszig).
One colleague told this reviewer that publishers should ship AEM texts to physical
education departments, where students could carry them to earn credit in weight
lifting. Luckily, professors can request a pruned version of a text. A special division
of the publishing company excludes chapters irrelevant to a course and includes the
instructors own material.
For instance, this reviewers Course B (Table 1) could use Kreszigs chapter on
series solutions and special functions, as well as the two chapters on Fourier analysis
and PDEs, cutting 20 chapters superuous to the curriculum. The savings would allow
ample room to add sections on inhomogeneous SturmLiouville and some treatment
of asymptotic methods.
Copyrights needed for any new components aect the price of such a custom
product. Most often the book costs less than the original. However, departments
need to commit to purchasing a minimum order of several hundred copies over a
couple of years.
Readers who have not yet identied an AEM text they especially like should
realize that most texts will address the subjects in Table 3. Few do so in a way that
diers radically from the others.
The occasional AEM text has a special way of interrelating subject matter. For
example, Gilbert Strangs book [13], which serves a one-year course in applied math-
ematics and advanced calculus and engineering mathematics (p. viii), integrates
the continuous with the discrete, emphasizing linear algebra and numerical methods
throughout and showcasing the nite element method, the fast Fourier transform,
560 BOOK REVIEWS
combinatorics, and optimization. It de-emphasizes series solutions and avoid[s] a
total and fatal immersion into vector calculuswhich has too frequently replaced
applied mathematics, and taken all the fun out of it (p. viii).
Bender and Orszags text [2] corresponds to a course on qualitative methods for
ODEs and dierence equations for advanced undergraduates and beginning graduate
students. Originally published in 1978 by McGraw-Hill and republished in 1999, the
book presents an excellent organization of a subject that can often seem like a hodge-
podge of surprising tricks. The four parts include fundamentals (ODEs and dierence
equations), local analysis (approximate solutions of linear dierential equations, non-
linear dierential equations, and dierence equations, plus asymptotic expansion of
integrals), perturbation methods, and global analysis (boundary-layer theory, WKB
theory, and multiple-scale analysis).
V. V. Mitin, D. A. Romanov, and M. P. Polis organized their text [9] around
the concept of mappings, their properties and manifestations, claiming to avoid
what they see as the overemphasis on operators and on discrete mathematics. As
such, their AEM book opens unconventionally with chapters on set theory, relations
and mappings, mathematical logic, and algebraic structures. After discussing linear
mappings and matrices, they address metrics and topological properties, then Banach
and Hilbert spaces. The nal chapters cover orthonormal bases, operator equations,
Fourier and Laplace transforms, and PDEs.
Nearly all AEM texts, however, will still require picking which ideas to cover
and how to relate them. Instructors will often nd a weakness in the way an author
addresses at least one of the desired topics, tempting them to choose material not
from among AEM chapters, but from among chapters in single-subject books with
applied avors. Everyone has his or her favorites on the elds listed in Table 3.
Frequently, AEM chapters on the same subjects read like poor cousins, right down to
less interesting problems and applications.
To take an example at random, Michael D. Greenberg [5] included a mostly solid
set of 14 exercises on generalized vector spaces. However, he has covered enough
material to have included more variety: like proving that the set of two-by-two ma-
trices or the set of solutions to some linear homogeneous dierential equation is a
vector space. (One problem errs in asking the student to prove that the set of nth-
degree polynomials is a vector space.) Only three of the problems deal with inner
products. He also gives no practical context for vector spaces, which engineering
students always nd mysterious and abstract. He does mention that he will revisit
them (which he does nicely) many chapters later when he covers Fourier methods
and PDEs. On the other hand, in his linear algebra textbook [14], Gareth Williams
has a good mix of accessible problems on inner products, in addition to a compelling
elementary discussion of the representation of space-time as a vector in R
4
. He
describes the quest for an inner product that would produce the appropriate geom-
etry for capturing experiments in relativity (and explains Minkowskis pseudo-inner
product).
Although the multiple-textbook scheme seems like it would cause even more in-
jury to the students pocketbook and biceps, most institutions can combine book
chapters into a single course packet and sell copies, given enough lead time for copy-
right clearance. Schaums outline [12] then provides a cheap, light, and reasonably
thorough companion for reference. This reviewer plans to use this approach the next
time she teaches AEM.
BOOK REVIEWS 561
REFERENCES
[1] G. Arfken and H. Weber, Mathematical Methods for Physicists, 5th ed., Academic Press,
Boston, 2001.
[2] C. M. Bender and S. A. Orszag, Advanced Mathematical Methods for Scientists and Engi-
neers: Asymptotic Methods and Perturbation Theory, Springer-Verlag, Berlin, 1999.
[3] P. C. Cretchley, A. J. Roberts, and C. J. Harman, Engineering mathematics: Time for a
core curriculum?, in Proceedings of the Sixth Engineering Mathematics and Applications
Conference, R. L. May and W. F. Blyth, eds., 2003, pp. 2530.
[4] S. A. Garfunkel and G. S. Young, Mathematics outside of mathematics departments, Notices
Amer. Math. Soc., 37 (1990), pp. 408411.
[5] M. D. Greenberg, Advanced Engineering Mathematics, 2nd ed., Prentice-Hall, Upper Saddle
River, NJ, 1998.
[6] F. B. Hildebrand, Advanced Calculus for Engineers, Prentice-Hall, New York, 1949.
[7] D. L. Kreider, R. G. Kuller, D. R. Ostberg, and F. W. Perkins, Linear Analysis, Addi-
son-Wesley, Reading, MA, 1966.
[8] E. Kreyszig, Advanced Engineering Mathematics, 8th ed., John Wiley, New York, 1999.
[9] V. V. Mitin, D. A. Romanov, and M. P. Polis, Modern Advanced Mathematics for Engineers,
Wiley-Interscience, New York, 2001.
[10] L. Mustoe and D. Lawson, Mathematics for the European Engineer: A Curriculum for the
Twenty-First Century, Report, SEFI Mathematics Working Group, Brussels, Belgium,
2002.
[11] P. V. ONeil, Advanced Engineering Mathematics, 5th ed., Brooks/Cole, Pacic Grove, CA,
2003.
[12] M. R. Spiegel, Advanced Mathematics for Engineers and Scientists, 5th ed., McGraw-Hill,
New York, 1971.
[13] G. Strang, Introduction to Applied Mathematics, Wellesley-Cambridge, Cambridge, MA, 1986.
[14] G. Williams, Linear Algebra with Applications, 4th ed., Jones and Bartlett, Sudbury, MA,
2001.
LAURA K. GROSS
The University of Akron
An Introduction to Numerical Analysis.
By E. S uli and D. Mayers. Cambridge Uni-
versity Press, Cambridge, UK, 2003. $35.00.
x+433 pp., softcover. ISBN 0-521-81026-4.
The present book has grown out of printed
notes that have accompanied lectures given
to undergraduate mathematicians at Ox-
ford University over many years. Quoting
from the back cover, Based on a success-
ful course at Oxford University, this book
covers a wide range of . . . problems from the
approximation of functions and integrals to
the approximate solution of algebraic, tran-
scendental dierential and integral equa-
tions. Throughout the book, particular at-
tention is paid to the essential qualities of
a numerical algorithm-stability, accuracy,
reliability and eciency.
The authors go further than simply pro-
viding recipes for solving computational
problems. They carefully analyse the rea-
sons why methods might fail to give ac-
curate answers, or why one method might
return an answer in seconds while another
would take billions of years. This book is
ideal as a text for students in the second
year of a university mathematics course.
It combines practicality regarding applica-
tions with consistently high standards of
rigor.
The reviewer nds that the authors have,
indeed, largely met these aims. The book is
well written and oers a good level of rigor
for the intended readership. The expected
level of background would be more in keep-
ing with that of upper level undergraduates
of a typical U.S. university. The exercises
are sensible and tractable. An email address
is given for instructors to obtain a LaTeXle
for solutions to the exercises. The intention
of this book is not to be a handbook on the
topic of numerical analysis, but rather to
give mathematics students the concepts for
numerically solving a number of mathemat-
ical problems. Some knowledge of matrix