BASIC INSIGHTS IN VECTOR

CALCULUS
With a Supplement on Mathematical Understanding
Other World Scientific Titles by the Author

Invitation to Generalized Empirical Method:

In Philosophy and Science
ISBN: 978-981-3208-43-8

The (Pre-)Dawning of Functional Specialization in Physics

ISBN: 978-981-3209-09-1
BASIC INSIGHTS IN VECTOR
CALCULUS
With a Supplement on Mathematical
Understanding

Terrance Quinn
Middle Tennessee State University, USA

Zine Boudhraa
Montgomery College, Maryland, USA

Sanjay Rai
Montgomery College, Maryland, USA
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Control Number: 2020030528

British Library Cataloguing-in-Publication Data

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

BASIC INSIGHTS IN VECTOR CALCULUS

With a Supplement on Mathematical Understanding
Copyright © 2021 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is
not required from the publisher.

ISBN 978-981-122-256-6 (hardcover)

ISBN 978-981-122-257-3 (ebook for institutions)
ISBN 978-981-122-258-0 (ebook for individuals)

For any available supplementary material, please visit

https://www.worldscientific.com/worldscibooks/10.1142/11892#t=suppl
Desk Editor: Liu Yumeng

Printed in Singapore
Terrance Quinn: In memory of my parents, George and Bernice, and two of
my brothers, Patrick and John, all of whom encouraged me in my interest in
mathematics and science, from an early age.
Zineddine Boudhraa: In memory of my mother, Habiba, as well as the
memory of all of those who lost their battles to cancer.
Sanjay Rai: To my parents, Shri Sarva Deo Rai, my late father, and Smt.
Munni Rai, my mother, neither of whom had the opportunity for education,
but through their own experiences in life, understood the need and potential
of education in changing the lives of a generation.
We also dedicate this book to students and teachers, in the hope that it will be
enjoyable and helpful in your growth in mathematics.
 Preface

This book was developed from a series of lecture notes for teachers of
undergraduate vector calculus, and students who want to understand the
essentials of the classical theorems. With that said, we believe that graduate
students in mathematics and mathematical sciences generally will also find
the material helpful. In order to reach some level of mastery of, say,
differential forms, one needs to have source insights for the classical vector
calculus theorems. Promoting those source insights is the main purpose of
our book. To help set the stage, we begin by providing some context.

Figure P.1 Positively-oriented closed curve C.

P.1 General context

The vector calculus theorems emerged gradually through a cross-fertilization
of physics and mathematics. In each of the theorems, an integral of a vector
quantity along a “boundary” is found to be equal to a combination of
derivatives in the region or volume “interior” to the “boundary.” For instance,
Green’s theorem is for a vector field

defined on a region R in the (x, y) plane. The boundary of R is a “positively-

oriented” closed curve1 C (see Figure P.1). With appropriate hypotheses in
place,

where

The vector calculus theorems are standard content for undergraduate

multivariable calculus books presently on the market.2 There is, however, a
gap in the textbook literature. Filling that gap will meet a need for students in
both pure and applied mathematics. Note that the gap is not in topics treated,
but in certain learning outcomes.
Currently available texts generally do a good job at providing applications, as
well as examples for developing symbolic and computational techniques.
And in most books, there are proofs — or at least partial proofs — of the
vector calculus theorems.
In modern mathematics, proofs are needed. As you will find by adverting to
your own experience, however, prior to proof, there is basic insight. For
mathematicians and mathematics students, basic insight is a familiar event.3
It is an insight, through a diagram or symbolism, or some combination in
one’s senses. Basic insight is not a guess, but this is not to suggest that
guessing does not occur. Basic insight, however, is something more. It
sometimes comes only after considerable effort, such as a breakthrough to
what might turn out to be a key result in a Ph.D. thesis. Or, one may discover
a possible solution to a problem in a homework set. The emergence is not
routine. In basic insight, we grasp possibility.
If, however, one goes on to ask ‘Is it so?’, then one has shifted to a new kind
or “level” of inquiry. Again, it is a matter of adverting to your own
experience. Asking ‘Is it so?’ is asking about something that, to some extent,
one already understands, at least as a possibility. Is the “it” that one has
grasped correct? One has had an insight. It is not nothing. But does one’s
understanding hold up? Are there counter-examples that might invalidate
one’s conjecture, or cases not accounted for? Is correction or revision
needed? Is what one understands partly mathematical but also partly
descriptive of diagrams, such as when “outward unit normal” is only
descriptively defined in coordinate geometry?4 And so on. All of this is
familiar to the mathematics student and the professional. In modern
mathematics, asking “Is it so?” calls for proof. That is, one seeks “middle
terms”5 by which that which one has discovered as a possibility is understood
to cohere with other results that have already been understood to belong to an
(axiomatic) context.
Basic insight, then, is crucial. It is a beginning. It emerges from a ‘What is
it?’ inquiry-mode, wherein one grasps possibilities. Being asked to read a
proof prior to being invited to basic insight is being asked to read a solution
to a problem that has not yet had. And, without basic insight, the possibility
of competence with symbolic and computational techniques is significantly
undermined.
In calculus textbooks presently available, however, the emphasis tends to be
on symbolic technique and computation.6 This Preface is not the place for a
literature review. To help illustrate the problem, it will be enough to advert to
two common ways that “curl” is introduced.

P.2 Beginning with the coordinate definition of the 3-d vector

curl
Frequently, the discussion of “curl” begins by providing a formula for the 3d
“vector curl” for F = Pi + Qj + Rk. Assuming that all partial derivatives exist,
This is then followed by discussions suggesting that “curl” represents
rotations in, for example, a fluid flow along a surface. And so, when curl F =
0, a fluid is said to be “irrotational.” Examples follow where the vector curl is
computed for known vector fields. Justification for the association with fluid
flow is then provided, ultimately, by appealing to Stokes’ Theorem in three
dimensions. Examples can provide plausibility. Note, however, that Stokes’
Theorem is a generalization of Green’s theorem in the plane to 2-d surfaces
in (x, y, z) space. Appealing to the generalization directs student attention still
further away from the possibility of basic insight. What is the basic insight
that initially leads to the definition of the 3-d (or 2-d) coordinate definition of
curl?

P.3 Beginning with descriptions of imagined paddle wheels

and integral definitions of circulation
A different but common approach is to begin by describing merely imagined
fluid flow and then inviting the reader to think of how a paddle wheel might
rotate, or not, when immersed in the fluid flow at various angles.
Following thought experiments about water flow and paddle wheels, a
“circulation density vector” in the direction of a normal vector n is defined
by:

Note, however, that what happens to paddle wheels in water flow is only
settled by experiment. As it happens, if a handle with a paddle wheel is
placed horizontally, then in some cases the paddle wheel rotates, while in
other cases it does not. Whether or not a paddle wheel rotates also depends on
orientation of the handle relative to the direction of flow. The integral
definition given above comes from classical fluid dynamics and, with some
work, can lead to the 3-d vector curl formula given in Section P.2. Discussion
of “imagined push and pull of water flow” notwithstanding, without
computing the link from the integral to the vector form, the student is
directed away from the possibility of basic insight and here, too, is instead
moved along into a flow of exercises for symbolic techniques.
While descriptions of water flow can invite beginnings, it took physics
decades to formulate key quantities and vector formulas. There are sequences
of insights needed, to go from imagining and describing water motion to
understanding such motion in terms of vector fields, vectors normal to a
plane in which “rotational displacement” occurs, combinations of partial
derivatives and line integrals.

P.4 Do vector fields represent force or velocity? Or other?

In undergraduate multi-variable calculus books, line integrals are often
motivated by appealing to the definition of work, from classical physics. In
simple cases, Work = (Force) × (Distance). More generally, the work done by
a force vector field F along a curve C is defined to be the line integral F ·
dr.
On the other hand, flux and circulation integrals are usually motivated by
asking the reader to think of the vector field F as representing a velocity field.
Then, with no further advisories, formulas for velocity fields and for force
fields are combined in the same vector calculus theorems.
The vector calculus theorems are, of course, correct. This approach, however,
covers over fundamental insights and, for the unwary, introduces various
errors. In both classical physics and calculus, flows, velocities and forces are
not directly comparable. When mass is constant, force is the derivative of
(mass) × (velocity). In what way, then, do examples computing line integrals
of force vector fields ((mass) × (acceleration)) help a beginner understand
circulation when in fluid dynamics circulation is something else? In fluid
dynamics, circulation is an average mass-flow rate obtained from a velocity
field. As in Section P.3, circulation density is a subtly defined circulation per
unit area.

P.5 Historical context and pedagogy

In multi-variable calculus books, the approach for teaching “curl” tends to be
“too mathematical.” Why do we say that when the theorems in question are
the fundamental mathematical theorems of vector calculus? It is because
initial inquiry and understanding that led (and leads) to their discovery (e.g.,
the definition of “curl” in 2 dimensions) are about fluid motion and other
field quantities (electricity and magnetism). Indeed, the vector calculus
theorems were discovered by mathematical physicists.7 And not adverting to
that obfuscates main sources of understanding.
This raises questions about undergraduate curriculum structures. Nowadays,
students who take undergraduate courses in Calculus III or Vector Calculus
are from majors including, but not limited to, mathematics, engineering,
aerospace, geography, oceanography, mathematical statistics, biology, and
mathematics education. It would not be practical to ask that all students from
all majors who are wanting a working knowledge of vector calculus also take
courses in mathematical physics. This raises questions about questions. What
are questions and key insights needed to reach basic understanding of the
fundamental theorems of vector calculus? And, are they accessible to
students from a variety of majors?
This is not a “speculative” question. It is about inquiry and insights that are
already present in the tradition.
What we did, then, was sift through the historical development of classical
mathematical physics, with an eye on fluid dynamics and the vector
theorems. This included re-reading some of the older undergraduate
textbooks for physics students, with titles such as “Introduction to
Mathematical Physics.” There is no new mathematics or physics in this book.
Nor are we trying to teach mathematics based on “some model of how the
topic in question can be learned.”8 We are, instead, drawing on what actually
has gone forward in classical mathematics and mathematical physics.
Note, however, that we are not attempting to implement the “historical
method of teaching” that has been popular in some quarters in the last 20
years or so. This is not a reflection on that method. We are guiding students
toward particular insights that are part of the tradition, even if some of those
insights have not yet been adverted to and are found in diverse sources. What
is new, then, is not only an identification of insights, but an identification of
“genetic sequences” of accessible and, as it turns out, elementary insights.
And so we have found that undergraduates with only modest background
have sufficient background to climb into the context of the fundamental
theorems of vector calculus. The lessons of the book are organized
accordingly.
For that reason, we do not start with definitions of vector fields or their
derivatives, nor with line integrals. Partly taking our lead from early
textbooks in mathematical physics, the lessons begin by inviting attention to
what can be described, namely, fluids in motion as well as test particles in
fluids in motion. (Most students will not have access to apparatus for
studying fluid motion. But we hope that most students have an opportunity to
observe water flowing somewhere.) To avoid confusion caused by talking
about integrals for work and integrals for circulation in the same context, we
keep the focus throughout on streamlines and first derivatives called velocity
fields.9 In that context, various questions arise naturally. For instance, there is
the problem of understanding describable rotation in fluid flows. As in
physics, we can use first-order approximation to estimate differences in
locations of test-particles in different streamlines. There is no need for
metaphors such as “tendency to rotate.” Instead, the formula curl(a, b) = [bx –
ay] is discovered as a precise solution to a focused inquiry.
The background needed is two-variable calculus and an understanding of
tangent vectors of streamlines.
Once the student is comfortable in understanding the vector calculus
theorems for velocity vector fields, it is not too difficult to go back over
results and realize that most of the mathematical aspects of the development
do not rely on the physical significance of the vector field. In that way, then,
the student gets two other sets of applications, namely: (1) in physics, for a
vector field that is a force vector field; and (2) in mathematics, for all
mathematical vector fields for which hypotheses are satisfied.

P.6 Prerequisites, the lessons, and follow up

For the student reader, formal prerequisites include: a semester each of
single-variable differential and integral calculus; and some experience in
multi-variable calculus. Having some understanding of the multi-variable
chain rule will help. However, that result, as well as applications in
integration, are reviewed in Sections 1.9–1.13. For readers who already have
some linear algebra in their background, a few supplementary results are
provided at the end of Part 1. These shed additional light on the equivalence
of Green’s Theorem and the 2-d Divergence Theorem. We do that by looking
to the Jacobian (that is, the first-order approximation) of the velocity vector
field. Similar results are included at the end of Part 2, in 3 dimensions. In that
case, one looks to the 3 × 3 Jacobian.
Part 1 is for results in the (x, y) plane. Part 1 leads up to Green’s theorem and
(in the 2-d plane only, its equivalent) the 2-d divergence theorem. Part 2 of
the book is for three dimensions. It leads up to Stokes’ theorem and the 3-d
divergence theorem.
The sections are brief. We have included problems and exercises.
Basic insights in vector calculus will be helpful in many areas in the
undergraduate Science, Technology, Engineering and Mathematics programs.
These include fluid dynamics as well as more advanced mathematics in
axiomatic contexts.10
Follow-up material at the undergraduate level can be found in, for instance,
the following two books, both of which provide far more complete coverage
of topics than the present book: Harry F. Davis and Arthur David Snider,
Introduction to Vector Analysis, 7th ed. (Mt. Pleasant SC: Hawkes Learning,
2000); and Miroslav Lovrić, Vector Calculus (Hoboken NJ: John Wiley and
Sons. Inc., 2007). The book by Davis and Snider includes discussions of
some of the more advanced equations of mathematical physics. See, e.g.,
Appendix C, The Vector Equations of Classical Mechanics (339–351) and
Appendix D, The Vector Equations of Electromagnetism (352–371) and, of
course, the equation of continuity (123). They provide two mathematical
derivations of the 2-d curl (125–132). Readers of the present book will be in
good shape to read the proof of Stokes’ theorem (298) given in the book by
Snider and Davis. The book by Lovrić provides some mathematical proofs
for special cases. It has an especially rich supply of exercises and examples.
It also includes introductory level discussion of some of the equations from
classical physics. See, e.g., Fluid and Heat Flow (520–522), Vector Calculus
in Electromagnetism (553–566), and Vector Calculus in Fluid Flow (566–
576). The book by Lovrić also has a brief but convenient technique-oriented
introduction to differential forms, Differential Forms and Classical Vector
Integration Theorems (536–552).

P.7 Our experience

In American colleges and public universities, the third semester of a calculus
sequence normally includes multi-variable calculus. In that third semester,
many topics are included and there is usually not enough time to thoroughly
treat the fundamental theorems in both two dimensions and three dimensions.
When time is limited, we have found that students manage better when the
focus is mainly on the 2-dimensional case, treated at a more leisurely pace
and in some detail. This allows the teacher to promote basic insights needed
for Green’s Theorem and the 2-d Divergence Theorem. This avoids having to
rush through complex vector theorems, which can leave students with a
confused jumble of integral formulas. On the other hand, we have found that
by first promoting basic understanding of the vector calculus theorems in 2
dimensions, a few lectures at the end of a Calculus III semester with
examples and pointings to the 3-d cases are generally well-received.
In a full semester of Vector Calculus, we have found that by promoting basic
insight, students from diverse majors have been able to do end-of-semester
projects. In those projects, students are asked to do original work. We don’t
mean developing “original theorems.” Rather, in an area of their choice, with
each project pre-approved by the instructor, students are asked to work out
details of applications of the vector calculus theorems. We have found that
students tend to take happy ownership of such projects. It is delightful and
inspiring to witness the wide range of presentations that reveal students’
understanding, as well as ever further applications of the fundamental
theorems of vector calculus.

P.8 Acknowledgments
We are indebted to colleagues of Sanjay Rai and Zine Boudhraa at
Montgomery College, MD. Motivation for this book partly emerged in the
context of conversations regarding possible redesign of particular elements of
the undergraduate mathematics curriculum. Some of these discussions
regarded the teaching of single- and multi-variable calculus, and the
importance of understanding, and being in control of concepts. The
divergence and curl operators especially illustrated the need for insight
regarding what they convey about a vector field. We would like to thank our
students who help us grow as teachers. Sanjay Rai would like to thank his
professors from Allahabad University, Dalhousie University and the
University of Arkansas for their tutelage and inspiration. Grateful to all of his
teachers, Terrance Quinn would like to thank three, in particular, who
introduced him to the vector calculus theorems, in various contexts: Professor
Rooney (1925–2016), in Advanced Calculus (University of Toronto, Fall and
Winter of 1981/82). Rooney opened our eyes to riches of classical
mathematics; Professor Keith Johnson (Dalhousie University) helped us learn
about differential forms (Spring, 1987) from Michael Spivak’s deceptively
brief Calculus on Manifolds11; and Professor Lowell Jones who (in a
graduate course at SUNY at Stony Brook, Fall, 1988) shed light on the
significance of the vector calculus theorems in 20th century differential
geometry.12 Finally, we’d like to thank two anonymous reviewers for their
comments and suggestions which helped improve the book.

_________________________
1In this context, closed means that when parameterized in a permissible way (for instance, one-one
and differentiable) the initial and final points of r(t) = (x(t), y(t)), t0 ≤ t ≤ t1 are the same, that is, (x(t0),
y(t0)) = (x(t1), y(t1)). For the term positively-oriented, non-technically, one can imagine a
parameterization tracking counter-clockwise relative to a point in the region R bounded by the curve.
The “circle” on the integral symbol indicates that the curve for integration is closed. To compute the
line-integral , we use any permissible parameterization r(t) to get

The integral of curl(a, b) is the familiar 2-d integral from 2-variable integral calculus. In order for the
formulas to be well-defined, it is crucial that the integrals involved be invariant under permissible
changes of variables. That follows from the chain rule in finite dimensions. Details are brought out in
the book.
2One may explore the literature.
3For more detailed observations, see the Supplement.
4Rigorous definition of orientation requires a more advanced context. See Supplement, note 31, for an
indication of the importance of identifying “casual insights.”
5See Supplement, Section A.4, for more details on “proof” and “middle terms.”
6This includes “calculus reform.” See Sections P.2 and P.3. The books by Snider and Davis (2000)
and Lovrić (2007) both include some pedagogical strategies that are “on point.” See Section P.6.
7Regarding the historical origins of “Green’s theorem,” Katz observed that “[a]ll of the
mathematicians who stated and proved versions of this theorem were interested in it for specific
physical reasons” (Victor J. Katz, “The History of Stokes’ Theorem,” Mathematics Magazine, vol. 52,
no. 3 (May, 1979), pp. 146–156. See p. 149.)
8Ed Dubinksy, “Using a Theory of Learning in College Mathematics Courses,” MSOR Connections
(2000), 1.10.11120/msor.2001.01020010. This article was originally published in Newsletter 12,
TaLUM, Teaching and Learning Undergraduate Mathematics subgroup, 2001. It is available online:
www.math.wisc.edu/wil-son/Courses/Math903/UsingAPOS.pdf. Accessed July 17, 2019. See
Supplement, Part C, for discussion.
 9The approach taken is guided by the historical development in mathematics and mathematical
physics. It is the approach taken in modern differential geometry, where streamlines and velocity fields
are the basic elements. Circulation is an average weighted velocity along a curve. Force is a second
derivative. Work is an average weighted force along a curve.
10See Katz, 1979, “Generalization and unification,” pp. 150–156.
11Michael Spivak, Calculus on Manifolds. A Modern Approach to Classical theorems of Advanced
Calculus. Reading, MA: Addison-Wesley, 1965.
12The textbook used was the classic: Noel J. Hicks, Notes on Differential Geometry. New York: Van
Norstrand Publishing, 1965.
 Contents

Preface
P.1 General context
P.2 Beginning with the coordinate definition of the 3-d vector curl
P.3 Beginning with descriptions of imagined paddle wheels and
integral definitions of circulation
P.4 Do vector fields represent force or velocity? Or other?
P.5 Historical context and pedagogy
P.6 Prerequisites, the lessons, and follow up
P.7 Our experience
P.8 Acknowledgments

1. REVIEWING SOME CALCULUS ESSENTIALS

1.1 Rates of change in one variable
1.2 The fundamental theorem of calculus in one variable
1.3 Applications by which the theorem earns the name “fundamental”
1.4 Linear approximation and area
1.5 Intimations of Green’s theorem in one dimension
1.6 Dot-product or scalar product: a solution to a problem about
projection length, in coordinates
1.7 Areas and volumes, in coordinates
1.8 Vector cross-product: a mathematical solution to a problem about
 rotation, from classical physics
1.9 The chain rule in one variable
1.10 An elementary example of ratios in integration in one variable
1.11 The general case: applications of the chain rule in integration in one
variable
1.12 The chain rule in two dimensions and one independent variable
1.13 The chain rule in two dimensions and two independent variables
1.14 Implications for integration over a region in two variables

FLUID MOTION IN 2 DIMENSIONS

2. 2-D MASS-FLOW VELOCITY FIELDS

2.1 Surface vector fields for fluid flow
2.2 Water flows, streamlines and integral curves
2.3 Mass flow rates
2.4 Units for mass flow rates

3. CIRCULATION AND GREEN’S THEOREM

3.1 Circulation (mass flow rate) along a parallel line segment
3.2 Circulation (mass flow rate) of a constant velocity field along an
arbitrary line segment in 2-d
3.3 Circulation along a curve in 2-d
3.4 Invariance of the circulation integral
3.5 Abbreviated notation for the circulation integral
3.6 Adding circulations from different curves
3.7 Preparing for Green’s Theorem
3.8 Green’s Theorem
3.9 Rotation and circulation density
3.10 A general case
4. 2-D FLUX AND DIVERGENCE
4.1 2-dimensional flux
4.2 Complementarity of 2-d flux and 2-d circulation
4.3 Divergence in 2-d and rate of relative change in areas

5. UNIFYING 2-D RESULTS

5.1 Unifying results
5.2 The divergence and the Jacobian
5.3 The curl and the Jacobian

FLUID MOTION IN 3 DIMENSIONS

6. FLUX AND DIVERGENCE

6.1 Beginnings
6.2 Flux across a surface S
6.3 Flux across a surface that bounds a volume

7. STOKES’ THEOREM
7.1 Stokes’ Theorem: The Question
7.2 Stokes’ Theorem: First approach: elementary derivation of
“Green’s theorem” for a two-parameter surface

8. RELATIVE CHANGE IN VOLUMES AND IN INCREMENTS

8.1 Rate of change of relative change in volume
8.2 Rate of change of relative change of increments
8.3 Pure rotation (Jv)A represents an isometry
8.4 Flux density and relative change in volume

SUPPLEMENT: MATHEMATICAL UNDERSTANDING

Introduction to the Supplement

Part A Mathematical Understanding

A.1 A diagram
A.2 Some puzzles
A.3 Descriptive and explanatory understanding, and judgment in
mathematics
A.4 Descriptive definition and explanatory definition
A.5 Proofs
A.6 Algebra from arithmetic, “and so on”: that is, sequences of “higher
viewpoints”
A.7 Correlations, concepts and other fruit of understanding
A.8 The historical context for teachers (and scholars)

Part B A Few Implications for Teaching

Part C Observations Regarding Modern Mathematics Education

Index
 Chapter 1

REVIEWING SOME CALCULUS

ESSENTIALS

The 2-d and 3-d vector calculus theorems are extensions of the Fundamental
Theorem of Calculus in one variable that students encounter in a first course
in calculus. To understand the vector calculus theorems, one needs a good
understanding of the one variable case. Sections 1.1 and 1.2 review some key
ideas in single variable calculus, leading up to the fundamental theorem in
one variable. Section 1.3 is to help bring out why the fundamental theorem of
calculus is called “fundamental.” (Among other things, it led to the solution
of numerous previously unsolved problems. It also opened up new lines of
inquiry.) Sections 1.4 and 1.5 foreshadow how the fundamental theorem in
one variable is both clue and key for the vector calculus theorems. In the 2-d
and 3-d cases, the theorems involve areas and angles between vectors.
Sections 1.6 and 1.7 are for developing the dot product, and the vector
product in three dimensions. The dot product and the vector product are both
solutions to problems in classical coordinate geometry. The cross-product
rule is a technical result that will be needed. Sections 1.8, 1.9 and 1.10
discuss three main cases. Section 1.11 draws attention to implications in
integration over a region in the plane.
This chapter is for development in prerequisite material. If you already have a
good understanding of one variable calculus and other topics treated here,
you could go directly to the beginning of Part 1, Fluid Motion in 2
Dimensions. You could then re-visit topics treated in this chapter on a “need
to know basis.”

1.1 Rates of change in one variable

Galileo discovered the “law of falling bodies.” He began by slowing motion
to “free-roll,” by rolling steel balls down polished beams of wood, set at
various angles to the ground.

Figure 1.1 Galileo’s experiments. Rolling steel balls down wooden beams. Measure distances and
times.

He had a radically new idea: compare measured distances and measured

times.13 By changing the angle and length of the beams of wood, he studied
numerous cases. In tables of numerical data, he discovered that, in each case,
distance was (approximately) proportional to the square of time, s = kt2. The
constant of proportionality k depended on angle of the beam. Using
elementary geometry, Galileo extrapolated to the case of a “vertical beam”14
and obtained his now famous result, “the law of falling bodies.”
In today’s units, the law of falling bodies is s = 16t2 feet in t seconds.15 By
selecting an appropriate angle for the beam of wood, we treat the case k = 1,
that is, s = t2 feet in t seconds.
Recall that average speed is a ratio,

Straightforward computation shows that for s = t2, average speed in the fourth
second is larger than average speed in the third second, average speed in the
fifth second is larger than average speed in the fourth second, and so on. In
other words, free-fall is “accelerated motion.”
A question arises: What is the speed of a free-falling object falling near a
particular time, near, say, t = 3?

Figure 1.2 Distances, times and average speeds are conveniently represented in a graph. Average
speed for a time interval is slope of a secant. Slope of secant for t = 3 to t = 5 is .

Students who have been through a first course in calculus will be familiar
with the following computations.
For t = 3, and with the idea of improving accuracy in computing average
velocity, we use increasingly small time intervals.
In units of , for the time interval t = 3 to t = 3 + 1, average speed is

For the time interval t = 3 to , average speed is

For the time interval t = 3 to , average speed is

And so on!
That is, for the time interval t = 3 to t = 3 + h, average speed is

On the graph of s = t2, these are slopes of secants that get increasingly close
to being tangent to the graph at t = 3.
The algebra reveals that the smaller the time interval h, the closer [2·3+(h)] is
to 2 · 3, that is, the closer the average speed is to 2 · 3 = 6. Saying it another
way, as h approaches zero, the difference between the difference ratio
and 6 approaches zero.
Figure 1.3 For h small, the secant is close to being the tangent to the graph at (3, 9); and the slope of
the secant is close to being the slope of tangent line.

The quantity 2 · 3 = 6 is called the “instantaneous” rate of change at t = 3.

The adjective “instantaneous” has often led to confusion. If we restrict to
time t = 3, there is no average speed since for average speed we need two
times and two possibly different locations. The name “instantaneous” is,
however, a convenient name for the entire computation and rationale given
above.
By an insight, the instantaneous rate of change is abstracted from sets of
fractions of average rates of change.
On the graph, average rates are the slopes of secants; and the smaller the time
interval the closer the secant is to the being tangent to the graph. The
instantaneous rate of change 2 · 3 is, then, the slope of the tangent line of the
graph of s = t2 at the point (3, 9).
General Case: In a first Calculus course, one is asked to determine rates of
change. One begins with average rates of change, that is, difference ratios of
the form
When represented on a graph, difference ratios are slopes of secants.
Following the approach described above, one increasingly reduces
increments h. By examining difference ratios, one identifies the
“instantaneous rate of change” at t0 (when it exists, mathematically). For the
graph, that quantity is said to be the slope of the tangent line at (t0, f(t0)). That
tangent line is said to be the linear approximation to the graph, near the point
(t0, f(t0)). The equation of that tangent line is

Figure 1.4 Linear approximation to y = f(t) is the tangent line given by y = f(t0) + f′(t0)(t – t0). The
time difference t – t0 = h.

Notation: Using Leibniz’s notation, the instantaneous rate of change is

written .16 Other notation that is also now standard goes back to Newton
and Lagrange, a dot and a prime, respectively (convenient when, for instance,
the independent variable does not require special attention). And so, while
there are other notations, in calculus and its applications the following three
are common:

More generally, the derived quantity is called the derivative at t. Since the
derivative usually is computed for more than one value of is
also called the derivative function.
Derivatives and Anti-derivatives: Once the derivative is understood, we can
study the derivative of sums, products and quotients of functions. That way
we can build up a repertoire of derivative formulas. By the same token, we
can work out tables of “anti-derivatives.” For instance, . If, then,
we start with g(x) = x and ask for the function f(x) that satisfies , the
solution is .

Exercises

Exercise 1.1.1
Suppose a ramp is at an angle so that Galileo’s result for “free-roll” is s = t2
feet in t seconds. (If you do a course in Newtonian physics, you will be able
to compute the angle at which that occurs. For now, suppose that in Galileo’s
experiment, free-roll s = 16 sin (θ) t2. Find the angle for which s = t2.)

(a) Compute the average speed for the time interval t = 6 to t = 6.5.
(b) Compute the average speed for the time interval t = 6 to t = 6.1.
(c) Compute the average speed for the time interval t = 6 to t = 6 + h, where
h is any small positive number.
(d) Compute the “instantaneous” speed at t = 6? Explain.

Exercise 1.1.2
Do Exercise 1.1.1 for a more general case. That is, as is done in the book for
s = t2, use only “first principles” to find the “instantaneous speed” at any time
t during free-roll.
Exercise 1.1.3
Suppose that a metal shuttle attached to a spring slides along a smooth
surface (friction is negligible) and that, for a time, its motion approximately
satisfies s = t3 feet in t seconds. Compute a good estimate for the
“instantaneous speed” at time t.
Exercise 1.1.4
Suppose that . Using only first principles show that

Exercise 1.1.5
Suppose that f(x) = x3. Using only first principles show that

Exercise 1.1.6
Suppose that f(x) = cos x and g(x) = sin x, where x is in radians. Using only
first principles, show that

Hint: What are the addition formulas for cos (x + h) and sin (x + h)? Looking
to the unit circle, for h ≈ 0, h ≠ 0, what are geometric approximations for
and ? Remember what radians means, a unit distance. What is that unit, in
the case of a circle of radius 1?

1.2 The fundamental theorem of calculus in one variable

Computations such as those in Section 1.1 are well-known. Moreover, both
linear approximation and being able to find anti-derivatives are helpful in
solving many problems in mathematics and applications. But these are not
yet The Fundamental Theorem of Calculus.
To discover the fundamental theorem of calculus, we look to the problem of
determining the instantaneous rate of change of “area.”17
Consider, for instance, the area of a square whose length is x. The area is A(x)
= x2.
If we start with, say, a square of length x = 3, then its area is A(3) = 32 = 9. If
the length of that square is increased by dx = 1, then the area of the new
square is A(3 + dx) = A(4) = 42 = 16. The change in area is

If, however, we start with a square of length x = 10 and again increase length
by dx = 1, then

The problem is to determine how area of a square changes relative to change

in length.
The computations needed are described in Section 1.1. In that sense,
numerically, the answer is already known. But the new question is not merely
numerical. We ask about “rate of change of area relative to change in length
of area.” There is, then, a further insight to be had, an insight that (in this
case) pertains to the geometry of the square.
Before going on with the discussion about square areas, however, it may help
to pause over two simpler examples.
Imagine two rolls of paper. One of these is 24′′ wide; and the other is 40′′
wide.
Imagine the 24′′ wide roll already rolled out across a flat cutting board. If it is
rolled forward an additional 1′′, additional area obtained is 24′′ × 1′′ = 24
square inches.
If we do the same computation for the 40′′ wide roll and it too is rolled
forward an additional 1′′, then the additional area is 40′′ × 1′′ = 40 square
inches.
What is it about the rolls of paper that provides different answers?
For the 24′′ wide roll, the rate of change of area is
and for the 40′′ wide roll, the rate of change of area is

Figure 1.5 Two rolls of paper, of width 24′′ and 40′′ respectively. Rolling forward, the rate of change
of area is (24 sq. inches)/(inch advanced) and (40 sq. inches)/(inch advanced), respectively.

In other words, for a given roll of paper, the rate of change of area is the
width of the roll!
Let’s now go back to the question about square areas. After some progress in
differential calculus, we have that

But the new question asks for something new:

What is the rate of change of a square area relative to change of length?
For instance, for x = 3 to , there is an initial 3 × 3 square nested
inside a larger
square (see Figure 1.6).
In terms of the length of the initial 3 × 3 square, what is the ratio

Figure 1.6 [Area of square of length ] - [Area of square of length 3] = [Areas of two
rectangles] + [Area of corner piece].

By appealing to the diagram, the change in area consists of two identical

rectangles together with a corner piece. Each rectangle is of length 3 and
width 1/100. To get the average rate of change, we need to divide the sum of
these “leading edge areas” by the change in length that produces the new
square.
That is,

The general case is similar:

Let h be a small change in length of any square of length x.
Then, again by appealing to the diagram,

Do you see, in terms of the geometry, why there is a factor of 2?

By increasing the length of a square to produce a new square, the initial
square is expanded along two front-lines! The result is essentially the same as
for the rolls of paper discussed in the examples above.

Figure 1.7 [Area of square of length (3 + h)] – [Area of square of length 3] = [Areas of two rectangles]
+ [Area of corner piece].

That is:
Relative to change in length x, the instantaneous rate of change of a square
area A(x) = x2 is the length of the advancing front line. In the case of a square
of length x, that length happens to be 2x.
Without yet having a rigorous definition of area,18 the same question can be
asked about an area A(x) between the graph of a non-negative function y =
f(x) ≥ 0 and the x axis.
Notation: The symbolism for ‘area under a graph’ was introduced by Leibniz
(1646–1716):
For the interval a ≤ t ≤ x, the area A(x) is written

where t is a “dummy” variable.19 The Leibniz notation is to express the idea

that the total area under a graph is somehow a sum of thin slices of area, each
of which is of height f(t) and width dt. The symbol “ ” is an elongated “S”,
for “Sum.”

Figure 1.8 Leibnitz heuristics and notation: Area A(x) under graph is a sum of areas of the form
“f(t)dt”: .

The Question: For the area function , what is the rate of

change of the area, relative to changes in length x of the base? That is, what is

Caution: In some books, is defined to be the anti-derivative

of y = f(x). This makes the issue a tautology. But the definition does not cause
any problems. Why not? Because it is a consequence of the Fundamental
Theorem of Calculus.
Just as with the rolls of paper and the square, to get the key insight, we can
again appeal to a diagram.

Figure 1.9 A(x + h) – A(x) ≈ f(x) · h + [area of small triangle]. Small triangle is determined by the
tangent line at (x, f(x)).

From the diagram,

Appealing to the results of Section 1.1, area of the triangular piece is

approximately
And so,

Remember that x is fixed. As the change in length of the base approaches

zero, the average rate of change of area approaches f(x).
Geometrically, f(x) is the length of the front-line of the advancing area.
This the key insight for the Fundamental Theorem:

Theorem 1.1 (Fundamental Theorem of Calculus).

Let represent the area between the graph of the function y
= f(x) and the x-axis. Then the rate of change of is the
length of the advancing (or retreating) front-line f(x).

Using Leibniz’s notation, every symbol has a meaning when we write

Exercises

In Exercises 1.2.1–1.2.4, (a) find y as a function of x and (b) demonstrate the

Fundamental Theorem of Calculus by differentiating the result of part (a).

Exercise 1.2.1
Exercise 1.2.2

Exercise 1.2.3

Exercise 1.2.4

In Exercise 1.2.5–1.2.6, use the Fundamental Theorem of Calculus to find .

Exercise 1.2.5

Exercise 1.2.6

In Exercise 1.2.7–1.2.8, express the solution of the initial value problem as an

integral.

Exercise 1.2.7

Exercise 1.2.8

1.3 Applications by which the theorem earns the name

“fundamental”
To begin to see why the fundamental theorem of calculus is “fundamental,”
we look to the historical context.
In antiquity, only a few formulas for areas were known. Each area had its
own definition. The area of a rectangle is

By subtracting and adding congruent elements, the area of a parallelogram is

(defined to be) the area of a rectangle of the same base and height; and the
area of a triangle is one half that of a parallelogram of the same base and
height. Hence, the area of a triangle is

And so on, that is, formulas can be obtained for areas of figures, when the
figures can be constructed from rectangles, parallelograms and triangles.
Partly because they all are reducible to rectangles, the answers are mutually
compatible.
However, what is the area of a circle of radius r? This doesn’t fit into the
system so easily. We can make a beginning by observing that a planar circle
is defined by a center C and its radius r. We can then use that radius to
construct 4 squares to cover the circle. By insight into the diagram, the area
of four squares each of length r is greater than the area of the circle of radius
r.
Figure 1.10 Area of circle of radius r is less than 4r2.

On the other hand, with more refined geometric constructions (using

inscribed and circumscribed polygons20), the area of 3 squares of length r is
less than the area of the circle. And so we get the following:

Observe that the argument does not depend on the radius. And so we
conclude that there is a ratio, call it π, 3 < π < 4, such that the

There are many other figures in Euclidean geometry. Archimedes solved the
problem of “quadrature of the parabola.” He obtained a formula for the area
of a parabola without needing to introduce unknown constants like ‘π.’ By
subtraction, we then also get a formula for area between a parabola y = x2 and
the x-axis. To get his result, however, Archimedes had to develop the theory
of non-terminating geometric series. He needed and reached an
understanding of “limit of a sequence of geometric sums.”22 Note that all of
this was done about 2000 years before equivalent results were (re-)
discovered by European mathematicians.23
If there had been Nobel Prizes given out in in ancient Greece, we venture that
Archimedes would have deserved several, one of which would have been for
solving the problem of the quadrature of the parabola.
What about other areas? What are areas under the graphs of y = x3, y = x4, · ·
· ; under the graphs of trigonometric functions like y = sin x, y = cos x, y = tan
x; derived trigonometric functions like y = sec2 x and y = csc2 x; and
exponential functions like y = 2x? Does each type of type of function require
its own definition and perhaps its own Nobel Prize?

Figure 1.11 The Fundamental Theorem of Calculus provides the basis for computing areas. (a) A(x) =
ex. (b) A(x) = x5/5. (c) A(x) = ln(x + 0.5).

We begin to see the importance of the Fundamental Theorem of Calculus by

observing that it allows us to reproduce Archimedes’ solution, and solve
many other such problems, almost effortlessly.
Let A(x) be the area under the graph of the parabola given by y = f(x) = x2,
over the interval [0, x]. By the Fundamental Theorem of Calculus,

Appealing to our repertoire of derivatives and anti-derivatives, we get that

Exactly similar arguments provide areas under the graphs of the other
functions mentioned above, and many more. That is, in each case, the
possibility of obtaining an explicit formula for area depends on having
techniques for finding anti-derivatives.
Exercises

Exercise 1.3.1
Graph y = f(t) = t2, t ≥ 0.
The problem is to find the area under the graph of the parabola. As explained
in the text, Archimedes solved the problem by developing the theory of
convergent geometric series. Using derivatives, the solution to the problem is
elegant and extends to an indefinite range of other functions and applications.

(a) Draw the area under the graph of y = f(t) = t2, for the interval t = 0 to t =
5. What is the length of the vertical front-line?
(b) Without doing “calculus”, what is an approximation for the area swept
out, if the vertical front-line moves forward by Δt = 0.1?
(c) What is an approximation for the area swept out if the front-line moves
forward by “small” Δt = h > 0?
(d) What is the instantaneous rate of change of area at t = 5?
(e) Repeat (c)-(d) for the interval t = 0 to t = x. That is, using first principles,
what is the instantaneous rate of change of A(x), the area under the graph
of y = f(t) = t2, at t = x? In other words, what is ?
(f) Using (e), what is the formula for the area A(x) under the graph of y = t2,
t ≥ 0, at t = x? This reproduces Archimedes’ result.
Exercise 1.3.2
Repeat Exercise 1.3.1 for the areas under the graphs of the following
functions:

(a) y = t4, on the interval t ≥ 0

(b) y = cos t, on the interval 0 ≤ t ≤ π/2
(c) y = sin t, on the interval 0 ≤ t ≤ π/2
(d) y = sec2 t, on the interval 0 ≤ t < π/2
(e) , on the interval t ≥ 0
(f) y = et, on the interval t ≥ 0
1.4 Linear approximation and area
Let A(x) be the area under the graph of a function y = f(x) > 0, over the
interval [a, b].

Figure 1.12 Linear approximation of area A(x): A(x) ≈ A(xi) + A′(xi)(x − xi).

An area function A(x) is a function of x that can be graphed on an (x, y) plane.

The function has its own derivative function, A′(x). Assuming that the
derivative exists everywhere, at each reference point xi there is a linear
approximation, namely,

Equivalently, we can write

Figure 1.13 For each increment, , where for all i = 1, 2, · · · ,
n. Therefore the net difference A(b) – A(a) is approximated by the sum of approximations to
incremental differences. That is, .

Partition [a, b] into sub-intervals of equal length,

where xi – xi–1 = dx > 0 for i = 1, · · · n.

We get the following approximations:

Adding equations, the left-hand-side telescopes. Recall also that a = x0 and xn

= b.
So we get
If we let dx → 0, and if the limit exists, it follows that

Question: Does this mean that

So, does the Fundamental Theorem of Calculus really only depend on linear
approximation?
Discussion: The first two equalities,

cause no serious difficulty. The second is consistent with the meaning of

Leibniz’s notation. And, the first equality follows from linear approximation,
a refinement on techniques going back to ancient times. As we have just
done, using linear approximation,

approximates area under the graph of a function, a function that in the present
context happens to be y = A′(x).
In the Question, it is the third equality where we find what might seem to be
“sleight of hand.” Why can we write

It is here that we invoke the Fundamental Theorem of Calculus, namely, that

A′(x) = f(x).
With that identification, it is evident that the chain of equalities

is partly, but not merely, a consequence of linear approximation. That is, the
ingredient that provides the third equality is the fact that the derivative of area
under the graph of the function y = f(x) > 0 is the length of the front-line of
advancing area.

Exercises

Exercise 1.4.1
Using equation 1.2, find the formula for the area under the graph of y = f(x) =
x, over the interval 0 ≤ x ≤ b. If you draw the graph, you will find that you
already know the answer from classical geometry. The exercise here is to find
it using the new method. Hint: Use subintervals of width . Also, what is
1 + 2 + · · · + n?
Exercise 1.4.2
Using equation 1.2, find the formula for the area under the graph of y = f(x) =
x2, over the interval 0 ≤ x ≤ b. Hint: Use subintervals of width . Also,
what is 12 + 22 + · · · + n2? One way to find this sum is to construct the
following series of equations:

Adding, the left side is a telescoping series while the right side can be written
in terms of the unknown 12 + 22 + · · · + n2 along with two sums that are
already known.
Exercise 1.4.3
Generalize Exercises 1.4.1 and 1.4.2. That is, for k a positive integer, using
equation 1.2 of Section 1.4, find the formula for the area under the graph of y
= f(x) = xk, over the interval 0 ≤ x ≤ b. Hint: What is 1k + 2k + · · · +nk?
Imitate the computations used in Exercises 1.4.1 and 1.4.2. You will need the
binomial theorem.

1.5 Intimations of Green’s theorem in one dimension

Equations 1.1 and 1.2 foreshadow computations needed in developing
Green’s theorem, Stokes’ theorem, and the divergence theorem.
For the vector theorems treated in this book, contexts differ. However, in all
cases, a crucial computation involves essentially the same strategy. What
follows is a pared down and simplified illustration:
Suppose that g(x, y) is defined and differentiable on the rectangular region

and that, for each x, g(x, y) is a (linear mass density)24 function in y. This
means that g(x, y)dy approximates the mass of a length dy in the y direction,
based at (x, y).

Figure 1.14 Shaded verticals are to indicate that there is a linear mass-function in the y-direction.

Now consider the difference

with g defined as above.

By definition of g(x, y), this approximates the difference in mass of the two
vertical sides of the rectangular region R.
By linear approximation, we can write the following:

In other words,

And so: A difference of two weighted line elements in the y direction (each of
which is an integrand for a one-dimensional integral) is approximated by a
weighted area element, an integrand for a two-dimensional integral.
Moreover, the weight for the area element is a derivative of the weight for the
line element.
This computation does not provide the fundamental theorems of vector
calculus. But you might keep it in mind as foreshadowing for the
development of the vector calculus theorems.

Exercises

Exercise 1.5.1
Let g(x, y) = x2 y3. Using linear approximation, show

In Exercises 1.5.2–1.5.4, compute the first-order partial derivatives.

Exercise 1.5.2
f(x, y) = 2x3 + 3x2 y3 – y4 + 8
Exercise 1.5.3
f(x, y) = yexy + 10xy2
Exercise 1.5.4
Exercise 1.5.5
For z = x2 + 3yex/y, find an equation of the tangent plane at P = (0, 1, 3).
Exercise 1.5.6
The linearization of a function z = f(x, y) at a point (x0, y0) is the function L(x,
y): (a) that is linear in both x and y; (b) that agrees with f(x, y) at the reference
point (x0, y0); and (c) whose partial derivatives also agree at the reference
point. In other words, the linearization is of the form

and satsifies

Find the linearization L(x, y) of

at (2, 1) and use it to estimate f(2.1, 1.1).

Exercise 1.5.7
Suppose that a function z = f(x, y) is defined on a neighborhood containing
the reference point (x0, y0). Recall that the differential dz is the linear
approximation to f(x, y) – f(x0, y0), in terms of dx = (x − x0) and dy = (y − y0).
Find the differential dz for z = xy + exy.

1.6 Dot-product or scalar product: a solution to a problem

about projection length, in coordinates
You probably already know the formula for the “dot product,” learned
perhaps in Calculus III, linear algebra, or some other undergraduate course.
But what is the dot product? Much of geometry in undergraduate
mathematics is Euclidean geometry and yet the dot product does not appear
in Euclid’s geometry.
At a time when coordinate charts were in common use for making maps for
surveying and traveling, similar advances in mathematics were made by both
Fermat and Descartes. In Euclid’s geometry there are abstractly defined
figures, points, lengths and so on. In coordinate geometry, one provides
figures with locations and dimensions on a coordinate plane. It is still
Euclidean geometry. But as history shows, the new approach is enormously
fruitful. Among other things, it allows one to solve geometry problems by
doing algebra and to solve algebra problems by appealing to geometry. It is
in this context that the following question arises:
Question: Given two vectors v = (a, b) and w = (c, d) emanating from the
same point, in terms of the coordinates a, b, c, and d, what is the length of the
orthogonal projection of v onto the line that determined by vector w?

Figure 1.15 Determine the length of the orthogonal projection of v onto w in terms of coordinates a, b,
c and d.

Problem 1.1.

(i) Assuming the Pythagorean Theorem and appealing only to the axioms
of Euclidean geometry, develop the law of cosines. Hint: Extend the
diagram so that you can express the problem in terms of two nested
right angle triangles.
(ii) Apply the law of cosines to the case where two sides of a triangle are
given by coordinate vectors v = (a, b) and w = (c, d) emanating from
the origin. Depending on one’s choice of initial and final point, a third
side of the triangle is u = ±(c – a, d – b).
 (iii) Develop a formula for the magnitude of the projection of v = (a, b)
onto w = (c, d), in terms of the coordinates of v = (a, b) and w = (c, d).
Hint: See ii.
(iv) Observe that a key element of the projection formula obtained in iii. is
the combination ac + bd. This combination is called the dot product
and is written v · w = (a, b) · (c, d) = ac + bd.
(v) Recast your results so that you get the well-known formula for v · w in
terms of || v ||, || w || and cos θ, where θ is an angle between v = (a, b)
and w = (c, d).

Exercises

In Exercises 1.6.1–1.6.3, compute u · v, || u ||, and || v ||.

Exercise 1.6.1
u = (3, 1), v = (4, –2)
Exercise 1.6.2
u = (–1, 3, 1), v = (0, 7, 2)
Exercise 1.6.3
u = 6i – j + 2k, v = 4i + j – 3k
Exercise 1.6.4
Find a unit vector that points in the same direction as the vector u = i – 2j +
3k.
In Exercises 1.6.5–1.6.6, find the angle between u and v.
Exercise 1.6.5
u = i + 3j, v = –2i + 3j
Exercise 1.6.6
u = i + j + k, v = 2i + 5j – k
Exercise 1.6.7
Let v = (a, b) and e = (1, 0). From classical geometry, v · (1, 0) = a = ||v|| cos
θ is the projection of v = (a, b) onto the x-axis. Show that if u = (c, d) is any
unit vector, then v · u is the projection of v onto the u-axis.
In Exercises 1.6.8–1.6.9, calculate the projection of u onto v.
Exercise 1.6.8

Exercise 1.6.9
u = 2i – j + 3k, v = –i + j + 3k

1.7 Areas and volumes, in coordinates

The problem of rewriting classical geometry in terms of coordinates
continues.
Areas
Two vectors v = (a1, a2) and w = (b1, b2) determine sides of a parallelogram.
Question: What is the area of the parallelogram, in terms of the coordinates
of v and w?

Problem 1.2.
The classical formula for the area of a parallelogram is (base) × (height). To
compute height, we need the sine function. Starting with the classical
formula and using both the sine function and the Pythagorean formula,
show from first principles that the area of the parallelogram determined by
v = (a1, a2) and w = (b1, b2) is |a1b2 – b1a2|. Hint: Express the sine function
in terms of the cosine function and use Problem 1.1.

Figure 1.16 Determine the area of the parallelogram determined by v = (a1, a2) and w = (b1, b2), in
terms of coordinates a1, a2, b1 and b2.

The quantity a1b2 – b1a2 is called the determinant of the pair of vectors v and
w. The determinant can be written as follows:

Also common notation is using vertical lines:

Volumes
In (x, y, z) space, three vectors u = (a1, a2, a3), v = (b1, b2, b3) and w = (c1, c2,
c3) determine sides of a parallelepiped.

Figure 1.17 Determine the volume of a parallelepiped determined by u = (a1, a2, a3), v = (b1, b2, b3)
and w = (c1, c2, c3), in terms of the coordinates a1, a2, a3, b1, b2, b3, c1, c2, and c3.

Question: What is the volume of the parallelepiped, in terms of the

coordinates of the vectors u, v and w?
The classical formula is that the volume of a parallelepiped is

Invoke the previous problem as needed and so work out that, in terms of
coordinates, the volume of the parallelepiped is
The quantity

is called the determinant of the three vectors u = (a1, a2, a3), v = (b1, b2, b3)
and w = (c1, c2, c3).
Similar to the 2-d case, this is often written:

For students familiar with linear algebra

Evidently, the two cases of “determinant” are similar. We can begin to
explore that similarity by looking to underlying geometry.

Figure 1.18 Area of parallelogram determined by 2v and w is twice the area of that determined by v
and w.

If one edge of a parallelogram is multiplied by a factor of 2, say, then the area

is also multiplied by a factor of 2.
In the determinant formula,
The same sort of result holds for volume of a parallelepiped. For instance,
from classical geometry, the volume determined by three vectors

is four times the volume determined by u = (a1, a2, a3), v = (b1, b2, b3) and w
= (c1, c2, c3). And, if we substitute the coordinates of u = (a1, a2, a3), 4v =
(4b1, 4b2, 4b3) and w = (c1, c2, c3) into the 3-d determinant we get

Problem 1.3.
Make a diagram for the volumes discussed above.

If we switch the order of two sides then areas and volumes do not change.
However, the two vectors representing the sides change in relative
orientation.
Figure 1.19 In the diagram, v to w is a negative rotation, while w to v is a positive rotation.

This shows up in the determinant formula as a change in sign, but leaves the
absolute value invariant. In the 2-d case, the computation is as follows:

Two properties common to both the 2-d and 3-d determinant functions are:
they are linear in each vector component; and they alternate in sign when two
vectors are switched.
Differential forms generalize areas and volumes to higher dimensions. For
readers who plan to go on study differential forms, note that these two
properties are all that is needed to define a determinant function in n
dimensions. In other words, a generalized n-dimensional determinant
function (or n-dimensional volume function) is any real-valued function

with the following two properties:

(i) D is linear in each component (multi-linear); and

(ii) when two vectors are switched, the value changes in sign (alternating).
That is,

For any n, there are many such volume functions. As the following example
illustrates, it is a matter of choosing units.

Problem 1.4.
With e1 = (1, 0) and e2 = (0, 1), define .

(i) Show that the defining properties determine for all pairs of
vectors v = (a1, a2) and w = (b1, b2).
(ii) What is the relation between this and the classical
?

Problem 1.5.
Show that a generalized determinant function is completely determined by
its value on “the unit n-cube,” that is, by the value of D(e1, e2, · · · , en),
where e1 = (1, 0, 0, · · · , 0), e2 = (0, 1, 0, · · · , 0), · · ·, en = (0, 0, 0, · · · ,
1).

Exercises

In Exercises 1.7.1–1.7.2, compute u + v and 2u – 3v.

Exercise 1.7.1
u = (–1, 5), v = (2, –3)
Exercise 1.7.2
u = (1, –1, 3), v = (7, 0, –1)
Exercise 1.7.3
Find the area of the parallelogram determined by the vectors u = (–1, 5) and v
= (2, – 3).
Exercise 1.7.4
Find the volume of the parallelepiped determined by the vectors u = (2, –1,
3), v = (1, 3, 0), and w = (0, 6, 1).

1.8 Vector cross-product: a mathematical solution to a

 problem about rotation, from classical physics
Setting up the problem
A topic in undergraduate physics is rotational motion.
Imagine a rigid axle, and a rigid radial arm attached and orthogonal to the
axle. The arm can only rotate in the plane Π that is orthogonal to the axle.
The radial arm has a length and direction in (x, y, z) space and so can be
represented by a “vector”, namely, r = (a1, a2, a3). Suppose F = (b1, b2, b3)
represents a force “applied” at the end-point of the radial arm. We can
assume that the force vector is also in the plane Π. (If F is not in Π then,
since the structure is assumed to be rigid, and in classical physics it is found
experimentally that components of acceleration add linearly, we could use the
projection of F onto the plane Π. Here, we don’t need that generality.)

Figure 1.20 Rigid radial arm attached and orthogonal to an axle. Applied force F = (b1, b2, b3) at end-
point of r = (a1, a2, a3) and coplanar with plane of rotation of r about axle.

The problem is to determine the angular acceleration of the radial arm caused
by the applied force F = (b1, b2, b3).
The structure is rigid. Within the constraints, only rotational motion is
possible. By taking the component of F = (b1, b2, b3) in the plane Π that is
orthogonal to r = (a1, a2, a3), and using the law of the lever, we can
distinguish three cases:
(1) Angular acceleration is maximum when F = (b1, b2, b3) is orthogonal to
r = (a1, a2, a3).
(2) Angular acceleration is zero when F = (b1, b2, b3) is parallel to r = (a1,
a2, a3) (pointing either toward or away from the axis of rotation).
(3) Angular acceleration is between (i) and (ii), when F = (b1, b2, b3) is at
other angles relative to r = (a1, a2, a3). More precisely, using the law of
the lever, angular acceleration is proportional to the area of the
parallelogram determined by the vectors F = (b1, b2, b3) and r = (a1, a2,
a3).

Problem 1.6.
Make diagrams for cases (1), (2) and (3).

Angular vectors
In experiments, it is found that if two forces F1 and F2 are applied, angular
accelerations add. However, angular accelerations are not vectors whose
components represent displacement of the end-point of the radial arm. There
is the question, then, of how to represent angular accelerations by quantities
that add.
In physics, rotation is about an axle which is normal to the plane of motion.
And, in coordinate geometry, a plane Π is represented by a normal vector.
Bringing these two ideas together, “an angular vector” is defined to be a
vector along the axis of rotation normal to the plane Π in which rotational
displacement occurs. Positive and negative angular rotations in Π about the
axle are represented by vectors pointing “upward” and “downward” along the
axis (or axle) of rotation, respectively.25
This is well-known in classical physics and leads to the following
mathematical problem:
To more easily allow for other applications, let’s first replace the symbols F
and r by u and v.
Given u = (a1, a2, a3), v = (b1, b2, b3), find a vector c = (c1, c2, c3) that has the
following three properties:
(1) c = (c1, c2, c3) is orthogonal to u = (a1, a2, a3);
(2) c = (c1, c2, c3) is orthogonal to v = (b1, b2, b3); and
(3) || c || = || (c1, c2, c3) ||= (area of parallelogram determined by u and v).
The new vector is called the cross-product or vector product of u and v and is
denoted by the u × v.

Problem 1.7.
Solve the system of equations (1), (2) and (3) and obtain that

Figure 1.21 Given u = (a1, a2, a3) and v = (b1, b2, b3), the cross-product c is defined to be the vector
that is orthogonal to both u and v, whose length is equal to the area of the parallelogram determined by
u and v, and whose direction is determined by the “right-hand rule”.

Exercises

In Exercises 1.8.1–1.8.2, compute the cross product of u and v.

Exercise 1.8.1
u = (–3, 1, 2), v = (2, 0, –1)
Exercise 1.8.2
u = 2i – j + k, v = i + 4k
Exercise 1.8.3
Find a unit vector that is orthogonal to both j + k and – i + j – k.
Exercise 1.8.4
Assume that a × b = (0, 2, 1), a × c = (–1, 1, 2) and b × c = (1, 3, –2).
Calculate a × (2b + c) and (a + b) × (3a + b – c).
Exercise 1.8.5
Find the area of the parallelogram determined by the three points (1, 2, 3), (1,
2, 4) and (1, 1, 1).
Exercise 1.8.6
Find the area of the parallelogram determined by the three points(1, 2, 3), (2,
4, 6) and (3, 6, 9). What does this mean about the relative location of these
points? Can you prove the result using vectors?
Exercise 1.8.7
Find a vector that is orthogonal to the plane determined by the three points (1,
0, 0), (0, 1, 0) and (0, 0, 1).

1.9 The chain rule in one variable

The chain rule is used throughout calculus, in all dimensions. It is a feature of
linear approximation.

One dimension, without calculus

Example
The rule can be discovered in elementary measurement problems.
Consider the two ratios:

and
The question is whether or not it possible to use these given ratios to
determine the ratio

Of course, the ratio inches to feet is familiar and known. But the question
here (a special case of the chain rule) is to use the given ratios — inches to
yards, and yards to feet — to determine the ratio inches to feet.
Why is the question plausible?
The first ratio relates inches and yards. The second ratio relates yards and
feet. There is, then, a link between the two ratios, a common unit, yards.
The trick is to take advantage of that common unit.
One way to do that is as follows:

The common unit “yard” cancels and the arithmetic provides a ratio of
“inches to feet,” that is, , or in the symbolism of proportions, a 12 : 1
ratio.
Example
Suppose that a car gets highway mileage of

and that on a particular trip the car is consuming gasoline at a rate of

How fast is the car traveling?

Again, an ad hoc solution is not difficult. But does not the approach of the
previous example suggest the possibility of more a systematic solution?
We are looking for the ratio .

We are given two ratios with units and .

Clearly, the product of these two ratios will be in the required units, .
In other words, the ratio is obtained as follows:

More generally, ratios change. But that is precisely what calculus is designed
to handle.
Example
Let y = x2, x = x(t) = 5t + 4 and consider the composition

We have

The problem is to determine , by using the two given ratios.

As in the previous example, one could solve this in an ad hoc way. For
instance, in the present context, we could expand

and then apply derivative formulas to the resulting quadratic function. The
question here, however, is different.
While we are only using elementary functions, the problem anticipates the
case where functions involved are more sophisticated. They could be high
degree polynomials, trigonometric functions, and so on.
And so the question is: Can we use the rates of change of the component
functions to determine the rate of change of a composition function?
Since the composition function is built out of known component functions,
does it not stand to reason that this might be possible? Part of the value of
Leibniz’s symbolism is that it reminds us that while the derivative of a
function is not a ratio, it is a limit of ratios. And, indeed, using linear
approximation, that limit can be used to estimate change in function values.
And so, taking advantage of the Leibniz notation, might we not anticipate
that

But again, the derivative symbols are limits of ratios. So, let’s dig into these
limits in order to get a better grip on this.
To get at , we need to estimate how a small increment Δt impacts the
composition

A value for t determines x; and a value for x determines y.

(1) For a small increment Δt, there is a small increment in x = x(t) = 5t + 4:

(2) For a small increment Δx, there is a small increment in y = f(x) = x2:

As we did in the last two examples, we can bring these ratios together.
Algebraically, that means substituting (1) into (2). Hence,
We now have something to prove that can be sorted out by appealing to first
principles, and the definition of “limit.”
The general case is sorted out in exactly the same way. And so we get the
result anticipated.

Theorem 1.2 (The chain rule in one dimension).

If y = f(x), x = x(t), and y(t) := y(x(t)), then

Exercises

Exercise 1.9.1
A car gets 30 miles per gallon on the highway. On a particular trip, the car is
using up 2 gallons per hour. From the discussion in the text, the speed of the
car is 60 m.p.h. Use the chain rule to find the same result.
Exercise 1.9.2
For and x = (t + 1)2, find .

Exercise 1.9.3
For and x = t + sin t, find .

Exercise 1.9.4
For , find .

Exercise 1.9.5
For , find .

1.10 An elementary example of ratios in integration in one

variable
Before getting into the subtleties of applying the chain rule in integration,
let’s explore an example where integration is accomplished through a single
change of scale. The example may at first seem overly elementary. But it
carries many aspects of the problem. At the same time, if one is to understand
the chain rule and its applications to integration, one needs to be able to
explain how this all plays out in even the simplest of examples.
Example
A carpenter who builds furniture buys wood from a local mill. The mill deals
with large quantities and sells by the foot. The “2 by 2′′ (lengths of wood
whose cross-section is 2′′ × 2′′) is selling at a price of 6 . (Note that we
write the dollar sign ‘$’ after the number 6. This notation is nonstandard in
banking and finance. However, writing it that way better fits the
mathematical symbolism.)

Figure 1.22 Lengths of 2′′ × 2′′ wood sell at 6 $ per foot.

As it happens, units that are more convenient for the carpenter (both for
building furniture and for accounting) are . In those units, the selling price
of the wood is

Question: The carpenter needs to buy 120′′ of wood. How do we compute

the cost?
Again, getting the answer is not a problem. For instance, since we now have
the cost per inch, we could simply multiply

Or, we could convert the length of 120′′ into feet. Then,

Not surprisingly, the answer is the same in both cases. For, whether we use
inches or feet, it is the same length of wood and the cost is given in the same
currency.
The problem posed here, however, is not just to find the answer but to
understand how the answer is reached. Given the cost in units of

and given that we need to purchase so many inches of wood (e.g., 120′′), how
do we re-express all of that in inches and

An elementary solution for the case when ratios are constant

Whether we use feet or inches, the cost per unit length is constant and so, just
as in elementary school, the problem can be solved by a strategic product of
fractions.
The systematic solution for converting is:

Notice in the second (middle) expression that, because of introducing inches,

the “number” for the length is multiplied a factor of 12 while the number for
increments is multiplied by a factor of . In other words, the two multipliers
cancel, leaving the cost the same.
Cases where ratios and scales are non-constant
The problem is to add, to get a total cost for a length of wood. There are
lengths (intervals) and (as will be shown) there is a cost density function C(x)
defined on the interval.
In the example, just discussed that cost density function happens to be
constant. And so the change of scale from feet to inches is the same change
of scale all along. But when buying timber, cost density functions need not be
constant.
For instance, for the construction of a specially designed A-frame hall, the
cross-section of large wooden beams may need to gradually expand toward
their base. This is to meet both design and engineering requirements. In this
scenario, wider cross-sections have more volume and so cost more.

Figure 1.23 Tapered beam of wood. Cost of cross-section depends on width.

Or again, because the range of forces is so large in earthquakes, to measure

intensity it is convenient to use exponents to distinguish cases. Measuring
devices are calibrated to provide not Newtons but a weighted logarithm of
Newtons known as the Richter scale. In fact, logarithmic scales are used in
many applications, including in the study of sound frequencies, for the which
a standard scale is decibels.
Our task is to be able to handle all such cases, that is, where C(x) is not
necessarily constant; and also where scale may vary not by some constant
ratio but according to some function x = x(t).
The key is that for, for small intervals, we can use linear approximation and
then add results. And so we can expect that the general solution will be an
integral. However, in anticipation of handling the general case, let’s first
return to the example where the carpenter is buying wood.
Temporary return to the elementary carpentry example
Let’s treat each aspect of the problem in turn. It is a matter of writing the
elementary solution already given, but now with an eye for parts of the
solution that in the general case might vary.
Let x be feet. For the mill, the cost density function near x is approximated by

In this example, for a small segment, this ratio is

In other words, as we already knew, the cost-density function is the constant

function

The length of wood to be purchased is 10 feet.

Partitioning the length of wood into increments of equal length Δx, the total
cost is
Or, in Leibniz notation,

To meet the needs of the carpenter, we can redo this by changing the length
scale from feet to inches.
Let t = inches. The relationship between inches and feet (between x and t) is t
= 12x.
In order to calculate the integral in the new scale, we need to first convert all
terms of the integral to the new scale.

(1) The new interval is t(0) = 12(0) = 0 and t(10) = 12(10) = 120.
(2) The function C(x) is constant and so, numerically, C (x(t)) = 6 and units
are . The units here are not right. Remember, we are foot
converting to inches. However, there is no cause for concern. Much as in
the solution to the elementary case, once we put everything together,
units will “self-adjust.”
(3) Since t (inches) = 12x (feet), an increment dt in inches correlates with an
increment dx in feet by the relation dt = 12dx. In other words,

Substituting into the integral that was given in terms of feet, and tracking
units,

If we suppress explicit identification of units, the algebraic structure is more

visible. That is,
In other words, from (1), the change of variables leads to a change in the
interval. From (2), we see what the cost function looks like in the new
variable. From (3), a multiplying factor compensates for the change in (1).
Observe, also, that the net effect of the (internal) change of variables is that
the $-value of the integral is invariant.

1.11 The general case: applications of the chain rule in

integration in one variable
The main idea is that by reversing the chain rule we get an integration
formula.
Example

Consider the integral

One may spot that the integrand can be expressed as a derivative of a

composition.
By the chain rule,

and

By the Fundamental Theorem of Calculus, this last integral is

What if an anti-derivative is not quite so obvious? One approach is called the
“ method of substitution”:
That is, we use Leibniz notation for the chain rule and compute accordingly:
if y = y(x) and x = x(t) then

In the example just given, on a hunch that the integrand might be brought
under control, let x = t5 + t. Then,

Leibniz notation is then convenient for expressing the corresponding linear

approximation:

By substitution, the integral can be expressed as

where the interval of integration is yet to be determined.

For the interval of integration, observe that when t = 1, x = 2 and when t = 2,
x = 34. Hence, the integral is

The argument in the general case is the same. By the chain rule,
By the Fundamental Theorem of Calculus in one variable,

Hence, the integral

As in the example above, the “method of substitution” uses Leibniz notation

to compute the new integrand. It is useful when, for instance, a strategic
change of variables produces an integrand that is a derivative of the form
.
To see how this works for an integral

we introduce a change of variables x = x(t).

As in the Example above,

(1) If x(c) = α and x(d) = b, then the new interval of integration is c ≤ t ≤ d.

(2) In terms of the new variable, the new function is obtained by using the
normal rules of function composition, that is, y = y(x) = y(x(t)).
(3) All of this is an application of the chain rule. In other words, the
“method of substitution” is the symbolic technique that takes advantage
of the Leibniz abbreviation already used in the development of the chain
rule.
As in the example above where g(t) = 3(t5 + t)2(5t4 + 1), we need to compute
an integral of the form g(t)dt. The strategy is in the hope that a change of
variables might bring this into tractable form. We might try some x(t).
By linear approximation, an increment dt along the t axis corresponds to an
increment

along the x axis.

If the integrand g(t) is then identified to be of the form

then by the chain rule we get

The new integrand might then be further analyzed. In some cases, such as in
the example above, for a known function f(x). In that case, by the
Fundamental Theorem of Calculus, we conclude with a numerical solution,
namely, f(x(d)) – f(x(c)).

Exercises

Exercise 1.11.1
Evaluate the definite integral: .

Exercise 1.11.2
Evaluate the definite integral: .

Exercise 1.11.3
Evaluate the definite integral: .

Exercise 1.11.4
Evaluate the definite integral: .

In Exercises 1.11.5–1.11.6, find the given area.

Exercise 1.11.5
The area of the region bounded by and the x-axis for 1 ≤ x ≤ e.

Exercise 1.11.6
The area of the region bounded by and the x-axis for 0 ≤ x ≤ 1.

1.12 The chain rule in two dimensions and one independent

variable
Suppose that z = f(x, y) and (x(t), y(t)) is a curve in the (x, y) plane.

Figure 1.24 The curve z(t) = f(x(t), y(t)) has a tangent line and slope at (x(t), y(t)).

The problem is to determine the slope of the tangent line of the curve z =
f(x(t), y(t)) in (x, y, z) space. In other words, what is where z = f(x, y) and
each of x and y are functions of t, x = x(t) and y = y(t)?
Figure 1.25 To compute net change, use Δx, Δy and partial derivatives of z = f(x, y) to compute Δzx-
direction and Δzy-direction. Then add.

Using the one-dimensional result together with the diagram, tracking change
in one direction at a time, the result follows:

(1) For a small increment Δt, there are small increments

and

(2) For the small increments in (1), there are two small increments in z = f(x,
y):
 and

The total increment Δz in z = f(x, y) is, then, approximately

Dividing by Δt, and taking the limit, the result follows. That is, if z = f(x(t),
y(t)), then

Exercises

In Exercises 1.12.1–1.12.3, use the Chain Rule to find .

Exercise 1.12.1

Exercise 1.12.2
z = sin x cos y, x = 3t, y = t + 4
Exercise 1.12.3
z = ln (x2 + y), x = e–3t, y = te3t
Exercise 1.12.4
Suppose that a curve in space tracks along the surface of the graph of z = x2 +
y2, and that the projection of the curve onto the x-y plane is given by the
curve r(t) = (cos t, sin t, 0). Find the slope of the tangent line to the curve in
three dimensions. Graph your results and explain.
Exercise 1.12.5
Suppose that a curve in space tracks along the surface of the graph of z = x2 +
y2, and that the projection of the curve onto the x-y plane is given by the
curve r(t) = (4 cos t, 3 sin t, 0). Find the slope of the tangent line to the curve
in three dimensions. Graph your results and explain.

1.13 The chain rule in two dimensions and two independent

variables
In mathematics and its applications, it is often useful to be able to convert
results from one coordinate system to another. A familiar example is between
(r, θ) polar coordinates and (x, y) rectangular coordinates. But the problem
also comes up in maps and atlases. For instance, locations given relative to
coordinates lined up with a local river system and watershed might need to be
translated to a map of a larger region based on a north-south and east-west
grid.
Example
Let’s begin by looking to the familiar case of polar coordinates and
rectangular coordinates. Both sets of coordinates can be used to determine all
points of the Euclidean plane. The plane is endowed with the Pythagorean
distance function.
Figure 1.26 Polar and Cartesian coordinates for Euclidean plane.

A first set of coordinate curves consists of mutually perpendicular lines,

providing what is known as (x, y) coordinate lines. An origin O is selected.
Centered at the selected origin O, another set of coordinate curves can be
introduced, namely, radial arms emanating from the origin O together with
circles centered at the origin O.
For present purposes, we can restrict our attention to points that are interior to
the first quadrant of the (x, y) plane, that is, where x > 0 and y > 0.
Using these two different sets of coordinate curves, each point P on the plane
can be represented in two different ways:

(1) by a pair (x, y), where x and y refer to distances along rectangular
coordinate lines relative to the selected origin; and
(2) by a pair (r, θ) , where r is the distance from the origin to the P and θ is
the angle (in radians) from the line y = 0 to the radial line containing P.
There are translation equations going both ways:

(1) Given (r, θ), the corresponding rectangular coordinates are

(2) Given (x, y), the corresponding polar coordinates are

To see the “two-variable chain rule” in action, let’s look to the transformation
given by equation (1) immediately above.
Suppose that two points P and Q on the plane are relatively close to each
other. (We can do this because the plane has the Pythagorean distance
function.) We suppose also that the polar coordinates of both P and Q are
known, as are the increments dr and dθ.
As in examples of the chain rule given above, the problem is to determine the
corresponding increments in the variables (x, y). This is a valid question
because x = x(r, θ) and y = y(r, θ).
Given P = P(x, y) = P(r, θ) and Q = Q(r + dr, θ + dθ), what are the
corresponding increments dx and dy by which to (linearly) approximate Q =
Q(x + dx, y + dy) in rectangular coordinates?

Figure 1.27 Given (r, θ) and (x, y), × = r cos θ, y = r sin θ and (r + dr, θ + dθ), what is the new point in
Cartesian coordinates?
Example
Let’s start with the x coordinate. From P to Q there is a change in both r and
θ, namely, dr and dθ. As described in previous examples, through linear
approximation the increment in x is a sum of an increment determined by a
change in r and an increment determined by a change in θ. That is,

In the same way,

Since points in the plane are determined by pairs of coordinates, this is

written more compactly, as follows:
For a pair of increments in polar coordinates dr and dθ, the corresponding
increments in rectangular coordinates are given by

Definition 1.1. The matrix is an abbreviation for

the first-order linear approximation of relative rates of change and is also

called the Jacobian matrix for equation (1).

Problem 1.8.
Show that in the case of polar coordinates,
The general case
To get the general result, observe that the computations above did not depend
on the fact that r and θ were polar coordinates. The key is that there is a
functional dependence of the form x = x(r, θ) and y = y(r, θ).
For the general case, we write x = x(u, v) and y = y(u, v); and represent
coordinates (u, v) by their own coordinate plane.
The problem is: Given the pair of increments , what is the corresponding
pair of increments ?
As with the special case of polar coordinates, we use linear approximation
given by the Jacobian matrix
Figure 1.28 Given (u, v) and (x, y), x = x(u, v), y = y(u, v) and (u + du, v + dv), what is the new point in
Cartesian coordinates?

The chain rule for two dimensions and two independent variables is

Exercises

Exercise 1.13.1
Show that the mapping (x, y) → (r, θ) is one-one and onto.
In Exercises 1.13.2–1.13.4, use the Chain Rule to find and .

Exercise 1.13.2
z = x3 + xy, x = s + t, y = st
Exercise 1.13.3
z = x + xy + 4, x = s/t, y = s – 3t
Exercise 1.13.4
z = yex/y, x = st, y = t/s
In Exercises 1.13.5–1.13.6, find the Jacobian matrix .

Exercise 1.13.5
x = 2u – 3v and y = 5u + 7v
Exercise 1.13.6
x = 3u + uv and

1.14 Implications for integration over a region in two

variables
Variables are as stated in Section 1.13, in the general case. As in the one-
variable chain rule, there are applications in two-variable integration.

Figure 1.29 The linear approximation maps a rectangle in (u, v) plane to a parallelogram in (x, y)
plane.

In (u, v) coordinates, an element of area is of the form du × dv. This is a small

rectangle, one side of which is the vector
and the other side of which is the vector

What is the corresponding area element measured in the (x, y) coordinate

system?

Problem 1.9.
By linear approximation to the transformation

and the chain rule, the rectangle

is linearly mapped (by the Jacobian) to a parallelogram in (x, y) coordinates,

with sides

From our previous work on area functions for the Euclidean plane, the area of
this parallelogram is

This provides the area element and we get the well-known “change of
variable” formula for integrals:
If the region S(u, v) in (u, v) corresponds to R(x, y) in (x, y), then
Example
In polar coordinates,

and so

Exercises

In the following exercises, draw the coordinate lines u = u0 and v = v0.

Exercise 1.14.1
This exercise concerns the iterated integral

(a) Evaluate this integral and sketch the region D of integration in the xy-
plane.
(b) Let u = x + y and v = x – y. Find the region S in the uv-plane that
corresponds to D.
(c) Use the change of variables formula to evaluate the integral by using the
substitution u = x + y, v = x – y.
Exercise 1.14.2
This exercise concerns the iterated integral

(a) Evaluate this integral and sketch the region D of integration in the xy-
plane.
(b) Let u = x + 3y and v = x – y. Find the region S in the uv-plane that
corresponds to D.
(c) Use the change of variables formula to evaluate the integral by using the
substitution u = x + 3y, v = x – y.
Exercise 1.14.3
Evaluate ∫∫(3x − 2y)dA, where D is the region enclosed by the ellipse 9x2 +
16y2 = 144. Change the variables using and .
Exercise 1.14.4
Transform the given integral to one in polar coordinates then evaluate the
polar integral.

(a)

(b)

_________________________
13How did he measure “time”? Galileo describes using both a pendulum and a water clock.
Discovering a correlation between measured distances and times was the beginning of modern physics.
Regarding Galileo’s work, there remain unanswered questions. See, e.g. Lindberg, David C. “Galileo’s
Experiments on Falling Bodies,” Isis 56, no. 3 (1965): 352–54. http://www.jstor.org/stable/228111.
14In modern terminology, the constant k is proportional to sin(angle between beam and ground). This

reaches a maximum when the angle is 90o.

15This is approximate, ignores wind resistance and supposes “other things being equal.”
16Using lower case d’s, Leibniz’s notation reminds one that the derivative is obtained from difference
ratios of the form .
17See note 7.
18See note 7.
19The Leibniz notation expresses the idea of “summing.” It is consistent with methods of
approximation going back to antiquity. Later, building on the work of Cauchy and others, Riemann
(1826–1866) worked out a definition of area between the graph of a function and the x-axis. In
contemporary calculus textbooks, we usually find the Fundamental Theorem of Calculus presented
following an introduction to Riemann integration. Unfortunately, that approach distracts students from
being able to get the key insight that leads to the Fundamental Theorem of Calculus, a result that
historically was known decades prior to Riemann’s work. There are distinct questions. The
Fundamental Theorem of Calculus answers the question: What is the rate of change of area?
Riemann’s work provided an answer to the question: What is area? Derivation of the Fundamental
Theorem of Calculus from the definition of the Riemann integral answers the question: Is Riemann’s
definition consistent with the earlier result about rate of change of area?
20See note 9.
21Using regular 96-gons, Archimedes obtained the sharp estimate . Accurate estimates
were also obtained in other cultures including, for instance, in astronomical calculations in the
Shatapatha Brahmana, written in the 6th century B.C.E., in India.
22T. L. Heath, “Quadrature of the Parabola,” in The Works of Archimedes (Cambridge: Cambridge
University Press, 1897), Book I, 233–252,
https://archive.org/details/worksofarchimede029517mbp/page/n421.
23Cauchy (1789–1857) was one of the European mathematicians who broke through to a definition.
See Victor J. Katz, A History of Mathematics, An Introduction, 2nd ed. (Reading, MA: Addison Wesley
Longman, Inc., 1998), 706–729.
24These will be introduced in the main text, in Parts 1 and 2.
25For now, this is partly descriptive. But appealing to diagrams we get the right formulas.
 PART 1
FLUID MOTION IN 2 DIMENSIONS
 Chapter 2

2-D MASS-FLOW VELOCITY

FIELDS

2.1 Surface vector fields for fluid flow

Have you ever watched and wondered about the motion of a stream, or been
intrigued by a vortex of water draining from a sink? Perhaps you’ve watched
ground water emerging from a spring (which in French is called “une
source”).
In applied science programs, water flow patterns are produced and studied
with chutes and by mechanically introducing sources and drains (or “sinks”).
A relatively simple case is water flowing slowly, at a constant speed in a
shallow chute in a laboratory.
A water flow has depth and samples of water have mass that can be measured
with weigh scales. There is, however, a problem here that is not encountered
in Newtonian physics. In Newtonian physics, one studies motion of an
isolated object with a center of mass. This is then “idealized” and
mathematized when one speaks of “point-particles” with mass and velocity.
By contrast, water flow is a “medium in motion.”
However, we can take advantage of classical physics. Introduce a small
fragment of cork, or anything that is small and floats along with the water
flow. This is sometimes called a “test particle.” If we follow a moving test
particle, it has velocities in the classical sense. Note also that, in practice, test
particles can be introduced at all locations on the surface of a water flow.

Figure 2.1 Four test particles with velocities at locations (x1, y1), (x2, y2), (x3, y3), (x4, y4).

For this book, we are assuming no “turbulence”26 and that the surface
remains (more or less) level. In fact, there are usually waves of at least some
amplitude. However, at low speeds, and if measurements need not be
extremely refined, then vertical motion can be eliminated from
approximations. Note also that by using submerged test particles and
transparent walls in a laboratory chute, it is found that velocities below and
near the surface are approximately equal to surface velocities. For now, then,
we focus only on horizontal components of motion of a shallow water flow.
Note, however, that the velocities of the surface are also understood to
represent the velocities of water below and near the surface.
To get started, then, what we need is a 2-component velocity vector field for
the horizontal velocity of a test particle at location (x, y) on the surface at
time t. That is, the velocity vector is of the form

Remember, however, that velocity is also the velocity of test particles below
and near the surface. That is, locally, the velocity is independent of depth z
which leads to the notation
And actual water has breadth and depth.
As in classical physics, we make use of the idealization wherein we speak of
“points” that are “locations without breadth or depth.”

Problem 2.1.
In this book, we will not be studying the case where there is time-
dependence. Instead, we will focus on cases where v is of the form

The velocity field is constant in time t but need not be constant in space
coordinates (x, y).

(1) Can you imagine elementary cases where the velocity vector field is
independent of time but is not constant in its space coordinates? Hint:
Imagine a chute or ramp down which water flows due to gravity, and
the chute then levelling out further downstream.
(2) Can you produce explicit velocity fields

to match the phenomena described in (1)? Hint: Think of Galileo’s

result about free-fall.
(3) While we will not be inquiring into the case of time-dependence, for
context, it can be helpful to be aware of a few cases. For examples,
think of water flow in a chute where the water flow is driven by an
oscillator. Or, again, if a chute is engineered to do so, the angle of
descent could change in time.
(4) Can you produce explicit vector fields

to match the phenomena described in (3)?

Problem 2.2.
By using small time increments and first-order terms, work out that speed
 of a surface near (x, y) is given by

Hint: With projections, treat one component at a time and use the Euclidean
distance function on the 2-dimensional surface.

Exercises
In Exercises 2.1.1–2.1.10, graph the velocity field, for x ≥ 0. Try this without
using a graphing calculator or graphing program. Determine the speed and
describe the fluid motion.
Exercise 2.1.1
v(x, y) = (3, 0)
Exercise 2.1.2
v(x, y) = (x, 0)
Exercise 2.1.3
v(x, y) = (3x, x)
Exercise 2.1.4
v(x, y) = (x, y)
Exercise 2.1.5
v(x, y) = (−y, x)
Exercise 2.1.6
v(x, y) = (3x+ y, x + y)
Exercise 2.1.7
v(x, y) = (1, cos x)
Exercise 2.1.8
v(x, y) = (x, cos x)
Exercise 2.1.9
v(x, y) = (cos2 x, sin x)
Exercise 2.1.10
v(x, y) = (sin2 x, cos2 x)
2.2 Water flows, streamlines and integral curves
Test particles follow trajectories called streamlines. The velocity vector of a
test particle is tangent to the streamline. On the other hand, given a velocity
field v and an initial point, a curve r(u) = (x(u), y(u)) is a streamline or
integral curve when at every point along the curve its velocity vector is v.
That is, we require

For details, see a Calculus III textbook.

Figure 2.2 Velocity vector (x′(u), y′(u)) of a parameterized streamline.

.

Exercises
In Exercises 2.2.1–2.2.8, compute the integral curve for the given velocity
field.

Exercise 2.2.1
v(x, y) = (3, 0)

Exercise 2.2.2
v(x, y) = (x, 0)

Exercise 2.2.3
v(x, y) = (3x, x)

Exercise 2.2.4
v(x, y) = (x, y)

Exercise 2.2.5
v(x, y) = (−y, x)

Exercise 2.2.6
v(x, y) = (3x + y, x + y)

Exercise 2.2.7
v(x, y) = (1 , cos x)

Exercise 2.2.8
v(x, y) = (x, cos x)

2.3 Mass flow rates

Imagine a 1 cm cross-section (in the direction of the y-axis) of a shallow
water flow moving in the x-direction with a constant speed of, say, 10 cm per
second.
Figure 2.3 Velocity with x-component = 10 cm/sec and linear mass density μ g/cm. Total mass-flow
10 · μ g in 1 second.

Keep in mind that our inquiry begins with water and other fluid flows. And
so, we can take advantage of what is known in elementary physics. A key
quantity in physics is momentum, namely, “(mass) × (velocity).” Classical
physics investigates the motion of objects with centers of mass, but we are
investigating water flow. We can adapt the definition of momentum to water
flow.
In the situation just described, what is the rate at which mass flows in the x-
direction across a 1 cm cross-section parallel to the y-axis? In other words,
how many grams of water flow in the x-direction, per second, when the
velocity of the flow is 10 cm per second in the x-direction and the cross-
section (y direction) is 1 cm?
In one second, 10 · 1 square cms of water pass the 1 cm cross-section. In a
water flow, below each square cm of the surface is a volume of water with so
many grams per cubic cm. Remember that, in our notation, we have
suppressed the depth component, but depth is there. This means that we can
introduce a linear mass density term, μ. In other words, μ represents the
number of grams per linear cm. This can be defined and measured because,
while the vector field has only two components, we are referring to volumes
of water in 3 dimensions, and volumes have mass.
In our example, then, the mass-flow per second that passes a 1 cm cross-
section is

2.4 Units for mass flow rates

Let’s pause here to keep track of units. This will be crucial in applications as
well as in developing mathematical results.
The term μ is “grams per cm in the x-direction”; speed is “cms per second”;
and length is “cm in the x-direction”. So, the units of ‘10·μ’ are “(cm/sec) ·
(grams/cm)”. In other words, the quantity 10 · μ is a rate, “grams per second”
(with the understanding that rate is of a 3-d flow of water moving in the x-
direction).
 Problem 2.3.
In fluid dynamics, μ is a function of both space and time coordinates. For
the present context, however, we suppose that μ is constant. We have been
talking about water flow. For water, μ is approximately 1 g/cm, since the
gram was originally defined as the mass of one cubic centimetre of water at
its maximum density at 4°C. However, our inquiry here anticipates the
study of other fluids too, for which μ is not unity. Nevertheless, with
constant μ, it is with no loss of generality that we develop our results for μ =
1. Why not? If, for instance, μ = 8, how might we adjust computations so
that a new μ, μ′ say, is unity? Hint: volume = (length)3. Might we replace
cm with a new scale? If so, what might that new scale be? How should it be
related to cm?

Problem 2.4.
If time is some fraction of a second, if the cross-section in the y-direction is
some fraction of a cm, or if the depth is something different than 1 cm (see
hypotheses given in the second paragraph, above), then mass-flow rates
need to be adjusted. Explore different cases. What are the appropriate
formulas?

_________________________
26For the present context, we are not providing a definition of turbulence. That would be done in a
follow-up course in fluid dynamics.
 Chapter 3

CIRCULATION AND GREEN’S

THEOREM

3.1 Circulation (mass flow rate) along a parallel line segment

In studying fluid flow, an engineer may need to know the total mass flow
rate, not only near an initial reference line (like the y-axis), but along a
length. Suppose the flow and cross-section are as described above. Extend
your view to include a 5-cm length in the positive x-direction.

Figure 3.1 Water flow along a 5 cm × 1 cm area.

Note that, numerically, the surface area of a 5 × 1 surface also is 5, but units
are cm2.
If we first look at the 1 × 1 cm2 immediately adjacent to the y-axis in the x-
direction, then in one second, 10 cm flows through that square. But the
segment along the x-axis that we are imagining is 5 cm in length and 1 cm
wide.
Imagine the flow originating well to left (from somewhere in the negative x-
direction) and continuing well to the right (in the positive x-direction).

Figure 3.2 Velocity with x-component = 10 cm/sec and linear mass density μ g/cm. Total mass-flow
10 · μ g in 1 second.

Through each of the five 1 × 1 cm2 along that length, there is a flow rate of
10 centimeters per second. So, with a 1-cm cross section, the total mass-flow
rate through the entire length in the x-direction is

This rate is called the circulation (of mass) along the 5-cm length.
(Note that we are not replacing 5 · 10 by 50. This is intentional, in order to
help us keep track of units and sources of terms. This will be helpful as we go
on to other cases.)

Problem 3.1.
In the example described, what is the total mass in motion, over a 3 second
time period?

The purpose of Problem 3.1 is to emphasize that units for circulation are
rates, a total of “grams per second along a curve.” If an engineer needs to
determine not mass flow rate along a curve, but total mass in motion along a
curve over a period of time, then they would need to also compute (or
approximate) a time integral.

3.2 Circulation (mass flow rate) of a constant velocity field

along an arbitrary line segment in 2-d
Suppose velocity field v = (a, b) is constant. In other words, both a and b are
constant two-variable functions in the (x, y) plane.

Figure 3.3 Component of v along a line segment of length L.

Suppose that a line segment of length L has a unit tangent vector T = (c, d),
(that is c2 + d2 = 1).
Note: The fact that components of velocity vectors are relevant is a physical
result. That is, in applications, velocity and momenta vectors add component-
wise.
And so, making use of the dot product, we can compute the mass flow rate
along the length of L. For then the component of v = (a, b) in the direction of
L (parallel to L) is v · T. Hence, the mass-flow rate along all of L is

3.3 Circulation along a curve in 2-d

For simplicity, we restrict attention to curves C that are one-one and
differentiate. (See a Calculus III textbook for examples.)

Figure 3.4 Element of circulation along a curve C is d(circulation)|C = μ((v · T)ds).

Comment: As you might expect, the key is repeated use of linear

approximation.
Suppose that a velocity field v = (a(x, y), b(x, y)) and a curve C is given by
coordinate functions r(u) = (x(w), y(w)).
The problem then is to determine the total rate of mass flow parallel to the
curve C and along its entire length.
Prior to dealing with details of parameterization, let’s begin with some basic
geometry.
An element of arc-length is ds and T is the unit tangent vector along the
curve.
The corresponding element of circulation (i.e., to first-order approximation,
the total mass flow rate in the direction of curve element of length ds) is

This leads to the following definition:

Definition 3.1. Total circulation along C is the limit of sums of elements of

circulation. That is, the circulation of v along C is

By reviewing “curves” in a Calculus III textbook, this integral is computed

by using a parameterization for the curve C. See Figure 3.4, above. In some
cases, that parameter can be arc-length s. However, we need a symbol other
than s to allow for all cases. It also would be good to have a symbol different
from t. That is because, in fluid dynamics, t is used to represent ‘time.’
However, in the present context, the curve C is a fixed curve. It is the fluid,
not C, that is in motion.
To avoid ambiguity, we use a different but also common symbol for a
parameter for curves, namely, u.
Suppose that we have a parameterization for the curve C. In order to compute
the integral in Definition 3.1, let’s first sort out the integrand.
As you will remember from Calculus III, using first-order approximation we
get

where the derivative is with respect to the parameter u.

A unit tangent vector to the curve is
Along the curve

Substituting these into the defining integral, we get that the total circulation
of v along the curve C is

Exercises

In Exercises 3.3.1–3.3.13, use the formula in Section 3.3 to compute the total
circulation for each velocity v(x, y) along the curve C. Graph and explain
your results.
Exercise 3.3.1
v(x, y) = (3, 0), C is the line-segment along the x-axis that starts at (0, 0) and
ends at (5, 0).
Exercise 3.3.2
v(x, y) = (3, 0), C is the line-segment along the x-axis that starts at (0, 0) and
ends at (0, 5).
Exercise 3.3.3
v(x, y) = (x, 0), C is the line-segment along the x-axis that starts at (0, 0) and
ends at (5, 0).
Exercise 3.3.4
v(x, y) = (x, ex+y), C is the line-segment along the x-axis that starts at (0, 0)
and ends at (5, 0).
Exercise 3.3.5
v(x, y) = (x, b(x, y)), where b(x, y) is any function for which the integrals exist
(e.g., continuous), and C is the line-segment along the x-axis that starts at (0,
0) and ends at (5, 0).
Exercise 3.3.6
v(x, y) = (3, 0), C is the boundary of the square with vertices A(0, 0), B(1, 0),
C(1, 1), D(0, 1), parameterized counter-clockwise and starts at A(0, 0).
Exercise 3.3.7
v(x, y) = (x, 0), C is the boundary of the square with vertices A(0, 0), B(1, 0),
C(1, 1), D(0, 1), parameterized counter-clockwise and starts at A(0, 0).
Exercise 3.3.8
v(x, y) = (x, y), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized counter-clockwise.
Exercise 3.3.9
v(x, y) = (x, x), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized clockwise.
Exercise 3.3.10
v(x, y) = (−y, x), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized counter-clockwise.
Exercise 3.3.11
v(x, y) = (−y, x), C is the circle of radius r > 0 centered at the origin (0, 0),
initial point A(1, 0), parameterized counter-clockwise.
Exercise 3.3.12
v(x, y) = (3x, x), C is the boundary of the triangle with vertices P(0, 0),
Q(10/3, 0), R(3, 0), initial point P(0, 0), parameterized counter-clockwise.
Exercise 3.3.13
v(x, y) = (3x + y, x + y), C is the boundary of the triangle with vertices P(0,
0), Q(10/3, 0), R(3, 0), initial point P(0, 0), parameterized counterclockwise.

3.4 Invariance of the circulation integral

Evaluation of this integral depends on having a parameterization for the curve
C. You might wonder, what happens when we use different parametrizations?
To explore that question, let’s first look to a simpler case, where the curve is
along the x-axis and a(x(u), y(u)) = 0. In that special case, the integral is of
the form

Does this remind you of something from an earlier mathematics course?

The parameterization of the curve is the “inside” function y(u). From
Calculus I, think of the chain rule, and its applications to integration called
the “method of substitution.” That method works because

In other words, the chain rule is what allows us to evaluate an integral using a
parameter that happens to be convenient. So long as the criteria of the chain
rule are met, the integral has the same value, namely,

and y(u0) and y(u1) are, by definition, the same for all parameterizations,
because these represent the y-coordinates of the end points of the curve C in
the ambient coordinate plane.

Problem 3.2.
A good exercise is sort this out in the general case, without appealing to a
textbook. It is a good way to shore up one’s understanding of the chain rule.
In other words, show, by computation, that the circulation integral does not
depend on parameterization.

3.5 Abbreviated notation for the circulation integral

Thanks to the invariance just discussed, we obtain the following
(unambiguous) abbreviation for circulation:
Circulation of v along C is
Another convenient abbreviation is obtained by recalling that

and that (dx, dy) is an increment in the plane that can be written as dr. That is,
r := (x, y) and dr := (dx, dy).
Using this more concise notation, we can also write circulation of v along C
is

In order to compute circulation, we need a parameterization of the curve. But

in these last two formulas, the parameterization is nowhere in sight. Why is
that? At this stage, the abbreviation means: introduce whatever
parameterization is convenient. This is valid because as in Problem 3.2,
circulation is (within a class of parameterizations) independent of
parameterization.
With parameter u and Leibniz’s notation, dx becomes

and dy becomes

The other terms are a(x(u), y(u)) and b(x(u), y(u)).

With these substitutions, one then computes a real integral, just as in Calculus
I.
Observation: In a more advanced course, the formula for circulation is a
“function” of two “variables,” namely, C and v. Why is that? It is because
circulation is independent of parameterization. Clearly, however, it depends
on the curve C and the vector field v, both defined on the plane.27

3.6 Adding circulations from different curves

Suppose that, in order to collect water for towns A and B, a hydro-
engineering firm plans to install a pipe system across a river (see Figure 3.5).

Figure 3.5 Two pipes collecting river water that flows due east.

Let’s suppose that the river flows west to east. Town A is on an island in the
middle of the river and town B is on the north shore, somewhat east, and so
downstream, of the island. A first pipe will connect a location O on the
southern shoreline to town A. That will provide the island-town with water. A
second pipe will connect the island to the northern shoreline. That pipe will
provide water for town B. There will be collection holes in both pipes into
which river water flows. We are ignoring numerous aspects of the
engineering problem. Still, the rate at which water flows into and then along a
collection pipe, while not identical to, is proportional to a circulation integral.
The part of the problem that is new here is that we will need to compute two
integrals. Let v be the vector field for the river velocity, C1 is the curve across
the river that the first pipe will follow from the southern shore to the island;
and C2 is the curve across the river that the second pipe will follow,
connecting the island to the northern shore.
At a moment in time, we consider the sum of all mass-flow rates at all
locations along both pipes. So, the total mass-flow rate of water along the two
pipes will be the sum of two circulation integrals. However, some care is
needed in choosing parameterizations.
Suppose that, for the pipe connecting the southern shore to the island, the
circulation integral is parameterized so that the initial point is on the southern
shore and the final point is on the island.
Then, one contribution to total circulation is

The circulation integral for the second pipe is

We already know that circulation is independent of parameterization. Might

that also work for two line integrals?
Note, however, that invariance of the circulation integral has hypotheses.
Parameterizations are required to have the same initial end points. That
invariance does not tell us how to combine two independently defined
circulations along different curves with initial and end points determined by
an application.
Clues can be had by looking to familiar integrals. In Calculus II, an integral
over an interval [c, d] is an integral over a curve, namely, a parameterized
line segment where the initial point is c and the final point is d.
What happens to the integral if we reverse that, so that the initial point of the
integral is d and the final point is c. In other words, how are the following
two integrals related,

You may remember that they are “the same” but opposite in sign.28
If you trace back through how to compute circulation, you will find the same
result. That is, if we switch end points, the sign is reversed, in the original
circulation. To see that, introduce a parameterization and you will have a
Calculus II problem.
What does that tell us about the engineering problem for the two towns?
If the engineering firm uses the island as the end point of both C1 and C2,
then one of the circulations will be negative in sign to the other. We leave it
as an exercise for the reader to explore why that would cause problems. One
obvious problem is that adding two circulations where the connecting curves

Figure 3.6 Elementary integration. Reversing order of integration changes sign of integral.

are “oppositely oriented” can result in a computed zero for circulation. Of

course, that would not match the situation of an actual river flow along two
pipes.
How, then, should we combine the two circulation integrals, the one for C1
and the other for C2 . We already have the invariance result. We can choose
parameterizations for each curve, individually.
But we need to combine them in a way that we avoid the problem just
described and so that the sum provides quantities needed in hydro-
engineering.
There is only one degree of freedom.
That is, we need the end of point of C1 to be the initial point of C2.
Doing that, we obtain the following definition for adding curves and
circulations.

Definition 3.2. Suppose that the end point of a curve C1 is the initial point
of curve C2 . Then, the sum of the two curves is denoted by C = C1 + C2.
By definition, the “sum” means: join the two curves, as described above, to
get a new “larger” curve. The circulation along the sum of curves is then
defined to be

Note: This result extends to adding any finite number of curves. It follows
that

so long as the end point of Ci is the initial point of Ci+1, for i = 1, · · · , m–1.

Problem 3.3.
Prove that under the hypotheses, the definition for adding circulation
integrals is independent of parameterizations.

Exercises

Exercise 3.6.1
Let v(x, y) = (x, 0) and C be the boundary of the square with vertices P(0, 0),
Q(1, 0), R(1, 1), S(0, 1), parameterized counter-clockwise and starts at P(0,
0). Using the four sides of the square as curves, write C as a sum C = C1 + C2
+ C3 + C4. Show that

Exercise 3.6.2
Let C1 be the unit circle centered at the origin (0, 0), initial point A(1, 0),
parameterized counter-clockwise, and let C2 be the unit circle centered at the
origin (0, 0), initial point A(1, 0), parameterized clockwise.

(a) Describe C1 + C2.

(b) Let v(x, y) = (a(x, y), b(x, y)) be a velocity field. Show that

3.7 Preparing for Green’s Theorem

Our inquiry began with preliminary descriptions of fluid, such as water, in
motion. An example is a “steady-state,” that is, when water flows with
constant velocity in a shallow chute or canal. Other examples include: flow of
a shallow river where water moves faster in the center of the river while near
the shoreline water speed slows to almost zero; the spiral motion of water in a
whirlpool; and the outward flow of water from a source.
In all of these cases, surfaces can be seen to rotate, contract and dilate, as the
case may be.
Simplified versions of such patterns are captured in the following
mathematical examples.
In each case, we compute an approximation to circulation about a small
rectangle of dimensions dx × dy.
By appealing to elementary calculus, we will get a general result. That result
will provide us with the ingredients needed for Green’s Theorem. For the
moment, however, let’s focus on the following examples.
Example
In this example, let v = (2, 0).

Figure 3.7 Vector field v = (2, 0) and rectangle dx × dy centered at (0, 0) with sided I, II, III and IV.

Circulation is a mass-flow rate along a curve. However, let’s explore the idea
of “circulation about a point” which, in this case, is (0, 0).
Sides I and III are parallel to the x-axis, while sides II and IV are parallel to
the y-axis. The net mass flow rate is, then, a combination of circulations
along I and III, respectively. However, what combination should that be?
We need to compute along curves that are appropriately connected. Start with
I parameterized in the positive x-direction. Tracking around the rectangle to
compute circulation along all segments that surround the origin, we get

with the stipulation that these are parameterized so that the end point of I is
the initial point of II, the end point of II is the initial point of III and the end
point of III is the initial point of IV.

Problem 3.4.
Show that this yields the following sum:

Why are there zeros? It is because the vector field is orthogonal to increments
II and IV, respectively. This is consistent with the graphed vector field.
There is no circulation about the origin because the flow is constant in one
direction. This is only descriptive, but it is a clue.
Example
In this example, let v = (y, 0).

Figure 3.8 Vector field v = (y, 0). Rectangle in the first quadrant with vertices A(x, y), B(x + dx, y), C(x
+ dx, y + dy) and D(x, y + dy).

Draw a rectangle in the first quadrant of the (x, y) plane with the following
vertices: A(x, y), B(x + dx, y), C(x + dx, y + dy) and D(x, y + dy).
Appropriately parameterized, AB is I, BC is II, DC is III and DA is IV.
Let’s now compute circulation along the boundary of the rectangle, using the
curve I + II + III + IV, as in the diagram.

Figure 3.9 Boundary of rectangle parameterized: I + II + III + IV, where I = AB, II = BC, III = CD
and IV = DA.

The computation gives

What does the negative sign mean?
In the diagram, the vertices of IV are A and D.
The vector field v = (y, 0) has only horizontal components.
For a small time increment, dt > 0 say, A and D move to A′ and D′,
respectively.
These are determined by following “test points” along streamlines. So, AD
and A′D′ are along “lines of sight” obtained when following A and D to A′
and D′, respectively. And, we can compare line segments AD and A′D′.
Remark regarding rotation: In the last example, relative to the y-axis, the
angle of the line of sight between the two vertices changes. More precisely, a
negative (i.e., clockwise) angle dθ is introduced.
Is it a coincidence that circulation and dθ are both negative?29
For now, let’s leave the observation as needing further inquiry.

Figure 3.10 Small time increment dt > 0, A flows to A′, and D flows to D′. The vector field is v = (y, 0)
implies |DD′| > |AA′|.

Example
For this example, v = (−y, x) is the familiar vector field whose streamlines
orbit the origin (0, 0).
As in previous examples, we can introduce a small rectangle which, in this
example, will be centered about (0, 0).
Figure 3.11 Vector field v = (−y, x). Small rectangle dx × dy centered at (0, 0).

The rectangle of dimensions dx × dy does not easily fit into circular

streamlines. Instead of using the corners of the rectangle as reference points,
as in the previous examples, let’s use the mid-points of each side of the
rectangle. This is equally valid since all of our computations are by way of
approximation. The lengths of each side will be dx and dy, respectively.
The vertices of the rectangle need to be:

The mid-points of sides I, II, III and IV are then

The vector field is defined by v(x, y) = (–y, x). Hence,

Therefore, the approximation to circulation around the rectangle ABCD is

Remark regarding rotation: As in the previous example, for a small time
increment, dt > 0 say, A and B move to A′ and B′. Notice that once again, for
a small time increment, algebraic signs of circulation and an angular change
are the same.
Example
For this example, we model a source with v = (x, y). Streamlines emanate
from the origin.
Again, construct a rectangle ABCD, with dimensions as in the previous
example, centered about the origin. Show that, in this case, the circulation is
zero. Can you anticipate from the diagram why this is so?
Now that we have explored a few examples, let’s work on the general case.

Figure 3.12 Vector field v = (x, y).

Circulation about a rectangle with dimensions dx × dy and consequences

of linear approximation
The vector field is v(a(x, y), b(x, y)). The vertices of the rectangle ABCD are
To approximate circulation around the rectangle, we compute:

Grouping terms, we obtain

The next step is elementary and is the key to discovering Green’s Theorem. It
leads to the main ingredient for the theorem. See Section 3.8.
Grouping terms, there are differences in function values. And linear
approximation tells us how to approximate those differences.
In two variables, linear approximation involves two partial derivatives and a
reference point. In the present application, the reference point is (x, y).

Figure 3.13 v(x, y) = (a(x, y), b(x, y)), evaluated at vertices A(x, y), B(x+dx, y), C(x + dx, y + dy) and
D(x, y + dy).
But in the two differences

only one of the terms involves the reference point (x, y). A common trick
(which survives into higher analysis) is to compare by getting from one
location to the next in steps that, strategically, “go through” the reference
point (x, y). For instance, one can do this by following the edges of the
rectangle.
In the difference

the “final point” is (x + dx, y) while the “initial point” is (x, y + dy). To
compare these points, first compare the function values at (x + dx, y) and (x,
y); and then compare the function values at (x, y) and (x, y + dy).
This means that we compute

Grouping terms we get,

This gives

Remember that this is an approximation only for the first term in brackets in
the expression

We will need to sort out the second bracket too. But for the moment, let’s
continue with our focus on the first square bracket, and substitute what we
just computed.
The first square bracket becomes
In the present context, then, circulation is found to be an integral relative to
an area of a small rectangle, relative to which the term

(This needs to be made precise. It can be shown that the term

does not contribute to the integral. In a more advanced course, we would

need a limit argument.)
What do we have so far?
With linear approximation,

Problem 3.5.
Work out the details and show that, in our linear approximation, the other
square bracket gives a similar result, namely:

If we combine the two sets of computations, what is the result? (See (2)
below.)

We now have two ways to approximate circulation of v(x, y) along the

positively oriented boundary of a small rectangle dx × dy

(1) A direct approximation is

(2) If we use linear approximation for differences in function values in 1
(ignoring higher order terms involving dxdx, dyddy, dxdxdy, dxdydy, and
so on, coming from Taylor’s theorem) we obtain a different
approximation to circulation along the boundary of the small rectangle,
namely,

Higher order terms do not enter into the final integrals.

Written as a single equation (ignoring higher order terms), we have two
ways of computing circulation around the small rectangle dx × dy. That
is,

This equality is the key to Green’s Theorem.

Before getting to Green’s theorem, however, an observation is already
possible.
Circulation around the small rectangle dx × dy is proportional to the area
of the small rectangle, namely, the product dxdy, where the
proportionality factor is

In other words, dividing by dxdy, the term bx − ay is circulation density per

unit area (with the above construction implicit).

3.8 Green’s Theorem

From Section 3.7, using linear approximation, circulation around a small
rectangle is equal to the area of the rectangle times the circulation density
function, namely, bx − ay.
In practice, we need to know circulation around not merely a small
rectangle, but a region R whose positively oriented boundary ∂R is a
closed curve C.
Recall that a curve C is closed when its initial point is the same as its final
point. To rigorously define positive orientation for a curve would go beyond
the context of this introduction. For present purposes, a parameterization is
positive when it tracks counter-clockwise relative to an interior point of the
region R.

Figure 3.14 Closed curve, P = r(u0) = r(u1), with positive orientation, that is, parameterization tracks
boundary curve C = ∂R counter-clockwise, with dθ > 0.

Question: Can our result about a small rectangle be extended to a positively

oriented boundary curve C of a region R?
We can take advantage of our result from the previous subsection:

In other words,
The term

can be extended to the region R. This is because it is a two variable function

multiplied by an area element dxdy. In other words, it is an integrand for an
integral over the region R.
This is our clue.
By definition of an integral in the plane, we introduce a grid of small
rectangles that partition the region R. We restrict to that part of the partition
that lies entirely within the region R. We can add up the results and then take
the limit, where the limit is as dimensions of the small rectangles converge to
zero. All of this is made precise in Calculus III and in more advanced
courses.
To explore this in more detail, let’s see what happens if we look at four sub-
rectangles in a grid.
There are four summands of the form

where each (xij, yij) is, say, the bottom left corner of each rectangle in the
sum.
Where does circulation come into play?
Remember that in their original form (before linear approximation), each of
those summands is an approximation to

And, if we add up area elements, then we get a corresponding sum of

circulations.
Figure 3.15 Four sub-rectangles of a partition of R.

Problem 3.6.
As the diagram suggests, contributions to circulation along common edges
internal to the four rectangles all cancel. That is because they are each
computed twice, with opposite signs.
Hence . . . . !
Do you see it?
The sum of circulations around the four rectangles reduces to the
circulation along the outer edge of the large rectangle that is the union of
the four small rectangles.
Hence . . . . !!
And, we can keep going!!!

That is, if we add up circulations around all rectangles in the partition within
the region R, common internal terms cancel. And this leaves circulation along
the closed curve that is determined by the union of the outer edges of
rectangles of the partition that lie within R.
In the limit, these outer edges converge to the boundary curve of the region
and the sums therefore converge to the contour integral along the closed
curve,
Figure 3.16 (See Figure 3.15.) Interior circulations cancel. Remaining terms are along boundary of
grid that is interior to the region R.

On the other hand, linear approximation of each summand means that we also
end up with sums that converge to an area integral, namely,

The two computations are two different ways of computing the same
quantity.
And so we get the following:

This equality is called Green’s Theorem.

Below, we go on to a different question that also yields the quantity bx − ay.
In that context we will find out that “curl” is a fitting name for the quantity bx
– ay which, as above, is also seen to be the circulation density.
This means that Green’s Theorem can also be written

where ∂R stands for the positively oriented boundary of R, which in the

formula above is also called C and, as is customary, the “circle” on the
integral symbol indicates that C is closed.
Examples

(1) For v = (2, 0), circulation density of v = 0 + 0 = 0.

(2) For v = (y, 0), circulation density of v = –1.
(3) For v = (−y, x), circulation density of v = 1 − (−1) = 2.
(4) For v = (x, y), circulation density of v = 1 − 1 = 0.

Exercises

In Exercises 3.8.1–3.8.11, compute both sides of Green’s Theorem, for each

velocity v(x, y) along the curve C = ∂R that is the boundary of the region R.
Exercise 3.8.1
v(x, y) = (3, 0), C is the boundary of the square with vertices A(0, 0), B(1, 0),
C(1, 1), D(0, 1), parameterized counter-clockwise and starts at A(0, 0).
Exercise 3.8.2
v(x, y) = (x, 0), C is the boundary of the square with vertices A(0, 0), B(1, 0),
C(1, 1), D(0, 1), parameterized counter-clockwise and starts at A(0, 0).
Exercise 3.8.3
v(x, y) = (x, y), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized counter-clockwise.
Exercise 3.8.4
v(x, y) = (x, x), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized clockwise.
Exercise 3.8.5
v(x, y) = (–y, x), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized counter-clockwise.
Exercise 3.8.6
v(x, y) = (–y, x), C is the circle of radius r > 0 centered at the origin (0, 0),
initial point A(1, 0), parameterized counter-clockwise.
Exercise 3.8.7
v(x, y) = (3x, x), C is the boundary of the triangle with vertices P(0, 0),
Q(10/3, 0), R(3, 0), initial point P(0, 0), parameterized counter-clockwise.
Exercise 3.8.8
v(x, y) = (3x + y, x + y), C is the boundary of the triangle with vertices P(0,
0), Q(10/3, 0), R(3, 0), initial point P(0, 0), parameterized counterclockwise.
Exercise 3.8.9
v(x, y) = (12x–4y, 5x + 7y), C is the boundary of the region bounded by the
graphs of the functions y = x4 and y = 16.
Exercise 3.8.10
v(x, y) = (3xy2, 5x2y), C is the boundary of the triangle with vertices A(0, 0),
B(1, 0), C(1, 1), ordered counter-clockwise.
Exercise 3.8.11
v(x, y) = (–y, x), C is the unit circle centered at the origin (0, 0), initial point
A(1, 0), parameterized clockwise.
Exercise 3.8.12
To generalized Exercise 3.8.5, suppose that v(x, y) = (a(x, y), b(x, y)) satisfies
curl v(x, y) = 1. Assuming all integrals are defined and finite, use Green’s
Theorem to compute the total circulation along ∂R in terms of a quantity
associated with the bounded region R.
Exercise 3.8.13
Let and R be the unit disk centered at (0, 0). The
circulation of v(x, y) along ∂R is –2π. But curl v = 0. Does this contradict
Green’s Theorem? Hint: Implicit in the development of Green’s Theorem are
hypotheses about domains.
Exercise 3.8.14
Let . Compute the circulation along the circle of
radius r > 0 centered at the origin (0, 0).
Exercise 3.8.15
Suppose that v(x, y) = (a(x, y), b(x, y)) = Δf = (fx, fy). Show that, for a region
R and boundary ∂R, both sides of Green’s theorem are zero. Graph and think
about streamlines which, in this case, are also called “gradient flows.” These
are important in physics. For instance, except when there are singularities,
air-flow of a weather system follows the gradient of the level curves of air-
pressure. The same comment applies to fluid motion, generally.
Exercise 3.8.16
Let and f(x, y) = arctan . Show that

For R the unit disk centered at (0, 0), compute the circulation along ∂R,
counter-clockwise. Does this contradict the previous exercise? Graph and
think about streamlines.

3.9 Rotation and circulation density

From the examples, we have accumulating evidence that the algebraic sign of
the quantity bx – ay correlates with some kind of rotation occurring in
streamline trajectories.
However, what is it that rotates? How are to define and compute
“rotation” for streamlines?
Let’s explore this a bit.
Example
Let v = (y, 0) be the vector field.
At time t = 0, start with reference points A = (0, 1) and B = (0, 2).
In this case, the formula for the vector field is explicit and linear and so
approximation arguments are not needed.
Figure 3.17 Vector field v = (y, 0). Increment AB flows to increment A′B′. Relative change in angle is
dθ < 0.

We suppose a small time interval dt. Let’s see what happens to A = (0, 1) and
B = (0, 2) along their respective integral curves, reaching A′ and B′,
respectively; and then compare AB with A′B′.
At A = (0, 1), the velocity vector is v = (1, 0) and at B = (0, 2), the velocity
vector is v = (2, 0).
So for small time interval dt we get

This means that

But recall that AB = (0, 1). So, for instance, for small dt > 0, the
corresponding angle differential between AB and A′B′ is dθ < 0, clockwise, or
negative.
Remember that circulation density of v is –1 < 0.
Example
Let v = (–y, x) be the vector field. At time t = 0, start with reference points A
= (1, 1) and B = (2, 2). In other words, A and B are along the same radius
emanating from the origin (0, 0).
The formula for the vector field is linear and so linear approximation methods
are redundant.
We start with the line segment AB = (1, 1) and we suppose a small time
interval dt.
Figure 3.18 Vector field v = (–y, x). Two initial points are A = (1, 1) and B = (2, 2).

At A = (1, 1), the velocity vector is v = (–1, 1) and at B = (2, 2), the velocity
vector is v = (–2, 2).
So for small time interval dt we get

This means that

But recall that AB = (1, 1). For small dt > 0, the corresponding angle between
AB and A′B′, is dθ counterclockwise, or positive.
As is evident from the geometry,

For small angle differentials, sine of an angle differential is close to the angle
differential (all angles, of course, in radians). Hence, for the small time
interval, we get that
We get circulation density of v = (–y, x) is 1 – (–1) = 2 > 0.

Figure 3.19 Vector field v = (–y, x). Increment AB flows to increment A′B′. Relative change in angle is
dθ > 0.

Problem 3.7.
From the geometry sin .
But for small angle differentials, the sine is close to the angle (all angles in
radians). And so, for small time interval, we get that dθ ≈ dt > 0.

Evidence continues to accumulate that rotation in some way correlates

with circulation density function.

3.10 A general case

Question: In the general case, is there a relationship between bx – ay and
“rotation” identified by comparing A′B′ to AB (where A reaches A′ and B
reaches B′ along their respective streamlines, in a small time interval dt > 0)?
More simply expressed, the question can read:
Figure 3.20 A, B, flow to A′, B′ in small time increment dt > 0. In linear approximation, A′(x, y) ≈ A(x,
y) + v(x, y)dt and B′(x + dx, y + dy) ≈ B(x + dx, y + dy) + v(x + dx, y + dy)dt.

How do we compare two increments? Note that the increments in question

are at different locations and their respective coordinates are relative to the
coordinate system located at those different locations.
You may recall from Calculus III that comparing increments and tangent
vectors obtained at different locations can have its subtleties. The canonical
example is the surface of the sphere. The curvature of the sphere needs to be
taken into account if one is to suitably “‘transport” and compare tangent
vectors obtained at different locations.
In our case, all of our computations are on the plane which has zero
curvature. This means that we can simply compare coordinates. Implicitly,
however, comparing AB and A′B′ does require transporting and comparing in
a common frame. With that technicality aside, we can bring precision to
our inquiry with linear approximation.
And so, while several computations follow, they are routine in the sense that
we are computing and assembling quantities needed in order to follow
through with linear approximation.
The question is about what happens locally, that is, local to a point (x, y).
Unlike the examples above, we now suppose that two initial points A and B
are relatively close to each other.
And so we write A = (x, y) and B = (x + dx, y + dy), with line segment or
“increment” AB = (dx, dy).
For the time being, AB = (dx, dy) is fixed, an initial increment.
In order to compute A′B′ (by linear approximation) we need to first find the
coordinates of A′(x, y) and B′(x + dx, y + dy).
These are points along respective streamlines of A(x, y) and B(x + dx, y + dy),
in small time dt.
Just as in the examples, in order to linearly approximate A′(x, y) and B′(x+dx,
y+dy), we need the velocity vectors at both A(x, y) and B(x+dx, y+dy),
respectively.
The vector v(x, y) at A(x, y) is

The vector v(x + dx, y + dy) at B(x + dx, y + dy) is

To approximate A′(x, y) and B′(x + dx, y + dy) we suppose a small time

interval dt. Under the hypotheses of linear approximation

and

Next, we compute and compare the two line segments, AB and A′B′. We have
AB = (dx, dy) and from the above computations,
Remember that the objective is to compare AB and A′B′.

Figure 3.21 is congruent and parallel to A′B′. Compare AB and .

The line segment AB = (dx, dy) is a small increment connecting two nearby
points A and B. Based on our computations so far, we have that AB = (dx, dy)
flows to a new line segment A′B′.
It can be helpful to display this in matrix notation:

Or, with the understanding that A′B′ and AB are increments,

where I represents the 2 × 2 identity matrix.

Alternatively, the relative strain is:
The problem is now tractable. The question of relative rotation becomes a
precise question of whether or not the linear transformation

has rotational components.

Note that the symbolism ‘Jv’ is for the “Jacobian of v”, the matrix of first-
order derivatives of the vector field v(x, y) = (a(x, y), b(x, y)).

We continue to take our lead from classical fluid dynamics

Description of flows reveals two distinct types of motion: “rotational
motion,” such as when velocities form streamlines that circle a drain; and
motion where velocities do not significantly change direction, where motion
“is mainly not rotational,” such as when water flows along a straight canal.
How might we distinguish these more precisely? We appeal to centuries of
successful physics. The Jacobian is, among other things,31 a linear
approximation to the flow velocity field. The question arises, then, as to
whether or not we can decompose the Jacobian of the velocity field in a way
that corresponds with observable differences in the two types of flow just
described.
We will need a result from undergraduate linear algebra, the usefulness of
which emerges in the next section.
A matrix M can be written uniquely as a sum of a symmetric matrix S and an
anti-symmetric matrix A.

Definition 3.3. A matrix S is symmetric if and only if Stranspose = S. A

matrix A is anti-symmetric if and only if Atranspose = –A.

Note: In this book, all matrices are real. Also, henceforth, the superscript
‘transpose’ is abbreviated to ‘t’.

Problem 3.8.
The decomposition is unique in the sense that if
 is a sum of a symmetric and an anti-symmetric matrix, then

Applying this decomposition to the Jacobian, we get

where subscripts S and A have the obvious meanings.

Explicitly, this gives

Why is this decomposition helpful for our inquiry?

Let’s begin with the symmetric part of Jv.
Since

is symmetric, there is an orthogonal basis relative to which (Jv)S can be

diagonalized. For the velocity field representing a fluid flow, that basis
sometimes is called the basis for the “principle coordinates” of the flow.

Problem 3.9.
In the case of 2 × 2 matrices, prove the diagonalization result for symmetric
matrices. (In higher dimensions, we need the real spectral theorem. That
would require more background than is being supposed for this book.)
Figure 3.22 Using, principle coordinates of (Jv)S, in linear approximation, rectangles flow to
rectangles.

To see why the principle coordinate system is useful, identify a small

rectangular area (rectangular relative to the principle coordinate system). If
we view the flow of vertices in time relative to these axes (that is, if with a
common parameter we view the small rectangle as its vertices flow along
their respective streamlines/integral curves) then, while sides may change in
length through dilation or contraction relative to the principle axes, for small
time increments, the rectangle flows to rectangles.
Dilation and contraction rates are the eigenvalues of (Jv)S. This will become
clear in sections to follow.
The anti-symmetric part of Jv tells us something different.
The anti-symmetric part is

Question: Does this type of matrix look familiar?

Example
Suppose that
Remember that the transformation matrix represents a system of ordinary
differential equations for the relative change in increments AB to A′B′. To
compare, we need a common frame, (u, v) say. Then, by definition, we obtain
a system of ordinary differential equations:

There are methods for solving such systems of equations, studied in

undergraduate courses in differential equations. Or, with an ad hoc approach,
observe that

solves the initial-value problem B′(0) = B, with β0 = ||B′(0)||. In other words,

the motion is rotation about the origin, in vector coordinates.

Can you now jump to the general case?

If

then angular velocity of A′B′ relative to AB is

This means that we now have a precisely identified geometric meaning for
the quantity bx – ay.
That is,
One way to see that is to observe that

is the unique solution to the initial-value problem B′(0) = B, with β0 = ||B′(0)||.

For this reason, the anti-symmetric (transformation) matrix

is also called the pure rotation part of the (transformation) matrix

Summarizing results about (Jv)A and discovering the name “curl”

Up to first order approximation,

where

Dividing by dt, we obtain

More formally, let v(x, y) = (a(x, y), b(x, y)).

The pure rotation component of the linear approximation of A′B′ relative to
AB is

and the term

Hence, the term bx − ay appropriately is called the (scalar32) “curl” of the 2-d
vector field v(x, y) = (a(x, y), b(x, y)). And so we define (scalar)

Definition 3.4. In fluid dynamics, a 2-d flow is said to be irrotational if

curl v = bx − ay = 0.

We have solved two problems, but we are getting the same answer in
both cases

(1) The quantity (bx – ba) is circulation density per unit area.
(Recall that “circulation” is defined to be a line integral of (components
of) mass-flow rates along a curve. In the present context, circulation
density per unit area is a limit of line integrals, each of which is along a
closed curve with (x, y) in the region interior to the curve.)
(2) Near the same reference point (x, y) as in 1., the term (bx – ay) is twice
the angular velocity of A′B′ relative to AB. Because of this, it is
appropriate to call (bx – ay) the (scalar) “curl” of the velocity field v(x, y)
= (a(x, y), b(x, y)).
Question: Why are the two answers the same? Sorting that out will require
digging into subtleties. We will take this up at the end of Part 1.
Let’s return to some of our familiar velocity vector fields, to see our results in
action.
Example
Let v = (–y, x). Then curl v = curl(–y, x) = bx – ay = 2.
Calculating the matrix quantities, we get

Here,

is a pure rotation with unit angular velocity, while the angular velocity
obtained from the anti-symmetric

is 2.
Example
Let v = (x, y). This is a vector field where the origin is a source and
streamlines are directed along radial arms emanating from the origin.
Looking at the graph of the vector field, one might correctly anticipate that
relative rotation is minimal or perhaps zero.
Then curl v = curl (x, y) = bx – ay = 1 – 1 = 0. Calculating the matrix
quantities, we get

For this vector field, up to first order approximation, relatively close

streamlines are parallel and in sync.

Exercises

Exercise 3.10.1
Let v(x, y) = (1, 1).
(a) Graph the velocity field and its streamlines.
(b) Find explicit formulas for the streamlines, in the form r(u) = (x(u), y(u)),
where u is a real variable.
(c) Compute curl v(x, y). Explain.
Exercise 3.10.2
Let v(x, y) = (x, x).

(a) Graph the velocity field and its streamlines.

(b) Set up appropriate differential equations and find explicit formulas for
the streamlines, in the form r(u) = (x(u), y(u)), where u is a real variable.
(c) Compute curl v(x, y).
(d) The streamlines are parallel but there is a non-zero rotational component.
Explain. Hint: You might start by comparing v(1, 0) and v(0, 1).
Remember that the curl v(x, y) is a first-order approximation.
(e) Compute (Jv)S and (Jv)A.
Exercise 3.10.3
Let v(x, y) = (x, y).

(a) Graph the velocity field and its streamlines.

(b) Set up appropriate differential equations and find explicit formulas for
the streamlines, in the form r(u) = (x(u), y(u)), where u is a real variable.
(c) Compute curl v(x, y).
(d) The streamlines are not parallel but the rotational component is zero.
Explain. Hint: You might start by comparing v(1, 0) and v(0, 1).
Remember that the curl v(x, y) is a first-order approximation.
(e) Compute (Jv)S and (Jv)A.

_________________________
27In a more advanced context (“higher viewpoint” — see Part A of the Supplement), terms of the form
w = adx + bdy are called “1-forms” and C is a “one-dimensional manifold.” Higher dimensional objects
like surfaces, volumes, and so on are “two-dimensional manifolds,” “three-dimensional manifolds,” and
so on. The “k-forms” formalize the notion of “line element,” “area element” and “volume element,”
respectively. A one-form w = adx + bdy provides “line elements” along a curve C in the sense that, at
each point p on C, it defines a linear functional on the space of tangent vectors at p. Computation
reduces to what we have described above. That is, one needs to introduce a parameterization and
integrate in the usual way.
28Hint: An original notion of integral is “area under a curve, over an interval.” When intervals change,
how do areas add or subtract?
29One way to sort out what the relationship is to invoke Green’s Theorem. But that would be making
use of a result that in this book has not yet been established.
30Steps need proof including rigorous limit arguments. See Preface.
31It is also the matrix of components of acceleration of the velocity flow. Further aspects of the
Jacobian would be studied in a course in fluid dynamics.
32A “vector curl” in 3-d is obtained in Part 2 of this book.
 Chapter 4

2-D FLUX AND DIVERGENCE

4.1 2-dimensional flux

As in previous sections, suppose v(x, y) = a(x, y), b(x, y) is a 2-component
vector field defined in the plane. Green’s theorem emerges from inquiring
about mass-flow rate along a curve which, for consistency, is required to be
positively oriented. (If we look to a small rectangle dx × dy, linear
approximation gives that the mass-flow rate along the positively oriented
boundary of the rectangle is (bx – ay) × dxdy.)
We now look to a complementary problem that also is inspired by physics.
Question: What is the net mass-flow rate across a small rectangular region,
in other words, what is the “mass-flow rate out - mass-flow rate in”?

As in previous sections, the small rectangle dx × dy has four sides, I, II, III
and IV.
We estimate “mass-flow rate out minus mass-flow rate in.”
To do that, we use the components of v = (a, b) that are orthogonal to each
side, and subtract appropriately.
That is, for the small rectangle, the net mass-flow rate is
This is similar to the computation for mass-flow rate along a curve which, as
you may recall, was a line integral.
Question: But is this last equation a line integral?

Figure 4.1 Net mass-flow rate through region dx × dy is net mass-flow rate in y-direction minus net
mass-flow rate in x-direction. Use normal components to sides I, II, III and IV, respectively. That is,
mass-flow rate is [b(x + dx, y + dy)dx – b(x, y)dx] + [a(x + dx, y)dx – a(x, y + dy)dx].

Clue: Factors can be obtained by identifying components of the vector field

that are orthogonal to sides I, II, III and IV, as the case may be.
Problem to solve: The computation so far is ad hoc, obtained by appealing to
a diagram for a region rather than by appealing to a defined line integral.
To see how we might rewrite this (physically significant quantity) as a line
integral, let’s begin by organizing terms so that, as hinted at in the clue
above, sides are in the right order to potentially give a positively oriented
boundary curve for the rectangle dx × dy.
That is, we order the sides as follows:
I followed by II, II followed by III and then III followed by IV, and so IV
back to the initial point.
By looking at the figure, the net mass-flow rate across the small region is,
then,
Think about geometry and vectors. In particular, remember the clue given
above, regarding orthogonality.
The first term –b(x, y) is the dot product of v(x, y) = (a(x, y), b(x, y)) with the
vector N = (0, –1) that is normal to side I and pointing downward.
The second term a(x + dx, y) is the dot product of v(x, y) with the vector N =
(1, 0) that is normal to side II and pointing to the right in the positive x-
direction. And so on.

Problem 4.1.
Provide a similar identification for the other two terms.

Observation: In each case we get the right terms, so long as we compute the
projection (or dot product) onto what, relative to the positively oriented
boundary is the “outward directed unit normal vector N.”
Therefore, the net mass-flow rate across the region dx × dy can indeed be
written as a line integral in the following way:
The rectangle has sides I, II, III and IV. Link them up to obtain a positively
oriented boundary curve for the region

where addition of curves is defined in Section 3.6.

Let N be the outward directed unit normal vector along each side. Then an
approximation to the net mass-flow rate through the small rectangular region
is

A technical issue regarding the integral for net mass-flow rate across a
small rectangular region
Can the integral that we just defined be expressed as a line integral of the
form
where f and g are functions defined on the plane?
To compute

we need terms for the integrand.

These are

Observe that ||(y′, – x′)|| = ||(x′, y′)||.

Substituting, we get

In other words, the net mass-flow rate is a line integral of the form

That is, by setting f = –b and g = a, we get

Problem 4.2.
The 2-d flux integral is invariant under change of parameterization. Provide
details.

Continuing with the main discussion

For the small region, the net mass-flow rate is
 Problem 4.3.
Just as we did in our inquiry that led to Green’s theorem, by linear
approximation (ignoring high order difference terms dxdx, dydy, dxdxdy,
dydydx, · · · which in the limit do not contribute to integrals), this
combination of mass-flow rates terms can be written as
[rate across III – rate across I]+[rate across II – rate across IV]
= [ax + by]dxdy.

Consequently, near a point (x, y), the net mass-flow rate through the small
rectangular region dx × dy is (ax + by)(dxdy).
Dividing by the area term, we get that the net mass-flow rate per unit area is
ax + by.
Just as in Green’s theorem, we obtain a density function which in this case is:

The quantity ax + by is also called the divergence of the vector field and is
often written as div v.
Just as in the development of Green’s theorem, by linear approximation we
get two ways of approximating the same quantity. And the resulting theorem
can be left as an exercise.

Problem 4.4. (2-d divergence theorem)

Provide the main details and, with appropriate hypotheses in place (e.g.,
that the boundary curve ∂R of the region R be positively oriented), show
that having the two ways of representing the same quantity leads to the
following integral formula:

If we include other standard ways of writing these integrals, we get the

following “line up”:
In other words, under suitable hypotheses, for a vector field v, the flux across
the boundary of the region is equal to the integral of the divergence over the
interior of the region.
Note: This theorem is a 2-d analogue for the classical divergence theorem in
3-d that ‘in physics’ also is called Gauss’ theorem.

Exercises

In Exercises 4.1.1–4.1.7, for each vector field v(x, y) = (a(x, y), b(x, y)),
where possible, compute the flux across the boundary ∂R of the region R,
oriented counter-clockwise: (i) directly by integrating along ∂R; and (ii) by
using the 2-d divergence theorem (Problem 4.4). Graph the vector field and
its streamlines and, if possible, find explicit formulas for the streamlines by
solving an appropriate system of differential equations. Graph and explain
your results.
Exercise 4.1.1
v(x, y) = (1, 0), R is the square with vertices A(0, 0), B(1, 0), C(1, 1), D(0, 1).
Exercise 4.1.2
v(x, y) = (1, 0), R is the triangle with vertices A(0, 0), B(1, 1), C(0, 1).
Exercise 4.1.3
v(x, y) = (x, y), R is the unit circle centered at the origin.
Exercise 4.1.4
v(x, y) = (1, 1), R is the unit circle centered at the origin.
Exercise 4.1.5
v(x, y) = (x, x), R is the square with vertices A(0, 0), B(1, 0), C(1, 1), D(0, 1).
Exercise 4.1.6
v(x, y) = (x, x), R is the square with vertices A(1, 0), B(2, 1), C(1, 2), D(0, 1).
Exercise 4.1.7
v(x, y) = (–y, x), R is the unit circle centered at the origin.
Exercise 4.1.8
Let and R be the unit disk centered at (0, 0). Show
that both sides of the divergence theorem are zero. But why does this not
provide an illustration of the divergence theorem? Hint: Consider the
domains.
Exercise 4.1.9
This exercise relates to Exercises 4.1.3–4.1.4:

R a region whose boundary ∂R is given by a simple closed curve r(u) = (x(u),

y(u)), parameterized to track counter-clockwise. Show that

4.2 Complementarity of 2-d flux and 2-d circulation

If you are getting comfortable handling line integrals, perhaps you have
spotted that we can get the 2-d divergence theorem from Green’s theorem,
without having to argue from first principles. Although, looking for that
alternate approach usually presupposes some prior grasp of the divergence
theorem. For then one is on the lookout for a derivation of what one already
knows within the axiomatic system.

Problem 4.5. (The 2-d divergence theorem from Green’s theorem; and
vice versa)
We have two integral formulas:
 Green’s theorem:
and the
2-d divergence theorem: .
The formulas are obtained by answering different questions about a vector
field v = (a, b). The questions and answers are different. At the same time,
there are evident similarities. Are the two formulas related? Or, in symbols,

Clue: Under our hypotheses, the vector field v = (a, b) was arbitrary.
To follow up on that clue, let’s use different symbols for each formula.
Rewriting, we get
Green’s theorem:

and
2-d divergence theorem:

Evidently, a linkage between the formulas is established by setting –f = a, e =

b.
That is, given the ordered pair (a, b), construct the new ordered pair (b, –a).

Problem 4.6. Show that this substitution allows us to derive Green’s

Theorem from the 2-d divergence theorem; and vice versa.

Question: What is the geometry of this linkage?

The vector field w = (b, –a) is everywhere orthogonal to v = (a, b). Hence, a
circulation integral of v = (a, b) along a curve is a flux integral of w = v⊥ =
(b, –a) along the same curve; and vice versa.

Figure 4.2 For v = (a, b), v⊥ = (b, –a).

Example
For v = (2, 0), div v = div(2, 0) = 0 + 0 = 0.
Example
For v = (y, 0), there is a source term in the first component that increases with
y.
The divergence is div (y, 0) = 1 + 0 = 1. In other words, as y increases, so
does the divergence.
Example
Let v = (–y, x) be the familiar vector field whose streamlines orbit the origin
(0, 0). As you might expect, the divergence is zero. Explicitly, div(–y, x) = 0
+ 0.
Example
Imagine a round fountain with a conical surface, and a steady source of water
that flows from its top. Water flows down the conical surface in all
directions. It has (approximately) constant acceleration due to gravity until at
the bottom of the cone’s edge, water falls into a surrounding channel that
collects water for recirculation through the fountain.
Figure 4.3 Water fountain and aerial view of water-flow velocity field.

Problem 4.7. Explain why the depth of water decreases as it spreads out
along the surface of the cone. Hints: Polar coordinates, increments in
radius, and water volume. From physics, assume that water is not
destroyed.

With an aerial view (that is, focusing on x and y coordinates and ignoring
height of cone and total depth of water at locations along the surface of the
cone), streamlines flow radially, away from the origin; and the magnitude of
the vectors (speed) increase with distance from the origin.
A simplistic mathematical model that has similar features is v(x, y) = (x, y).
For this vector field, divergence is div (x, y) = 1 + 1 = 2.
The flux density per unit area is 2.
Let’s review what this means in this case.
Figure 4.4 Approximating difference in flux using difference area.

Start with a reference point (x, y) and suppose that C = ∂R is a positively

oriented closed curve that is boundary to a small region R containing the
point (x, y). Replace R by a slightly (linear approximation) larger area R′, one
that includes R (see Figure 4.4).
There is a difference in area dA.
What is d(flux), the difference in the flux integral for C′ = ∂R′?
Since div (x, y) = 1 + 1 = 2 is flux density per unit area, d(flux) approximately
2dA.
Discovering that the quantity div v = ax + by is also a “rate of relative
change in area”33
The 2-d divergence theorem relates flux along a closed curve (whose interior
contains a fixed reference point) to a density function per unit area. But as
can be observed in fluid motion, there is also motion of “elements of area.’:
Figure 4.5 Vertices AB1CB2 flow to

We can, for instance, imagine test particles at vertices of a rectangle. As some

examples already reveal, in some cases “surface areas of rectangles expand”
while, in other cases, surface areas decrease (for instance, when a flow is
forced into a narrow channel).
This leads to the following question:
Question: For a given velocity vector field, how do areas change as vertices
move along their respective streamlines?
Remark: Are you noticing a similarity here with the line of questioning that
led to a second derivation of the circulation density function, bx – ax? In that
case, a question came up about the relative change of increments, A′B′ and
AB, in time. The present question is similar. Now, however, we ask about
relative change in areas, in time.
Figure 4.6 Rectangle determined by Vertices A, B1, B2 yields rectangle determined by .

Preliminaries: Three reference points A, B1 and B2 determine a

parallelogram with sides AB1 and AB2 and area [Area]. (They might be
collinear but draw the picture as though they are not.) If we track the motions
of these points along their respective streamlines to A′, and , we get a
new parallelogram, one with sides and and area [Area]′.
The question now is to compare the two areas,

as a function of time. This means that we need to use a common integral

curve parameter for all three streamlines.

Warmup problems

Warmup 1:
Let v(x, y) = (x2, 0). In this example, there is acceleration in the first
component of the velocity field, relative to change in the first coordinate. Let
A = (1, 0), B (2, 0) and B2(1, 1). In this case, initial [Area] = 1.
Motion is in the direction of the x-axis. Suppose that the initial value of the
parameter for the integral curves is t = 0 and that time increases.
At t = 0, v(1, 0) = (1, 0), v(2, 0) = (4, 0), and v(1, 1) = (1, 0).
For a small time increment dt > 0, A = (1, 0) and B2 = (1, 1) both advance in
the x direction by 1.

Figure 4.7 [Area]′ – [Area] = (1 + 3dt) – 1 = 3dt.

And so A′ = (1 + dt, 0) and . But at v(2, 0) = (4, 0). This means

that B1 advances to (in linear approximation).
The original square has been both translated and dilated in the x direction.
Explicitly,

Therefore,

In other words, the area is an increasing function.

Warmup 2:
In the previous example, points were separated by integer lengths. And, the
change in areas is completely determined by the first coordinate function of
the vector field, namely, a(x, y) = x2.
Before tackling the general problem, let’s try a generic one-dimensional
problem, v = v(x).
In this case, the question is about how interval lengths L change in time along
the x-axis.
We suppose a small time increment dt > 0.
Initial points are A = x and B = x + dx, with L = dx. The increment dx > 0 is
fixed.
The velocity (vectors) at these points are v(x) and

Figure 4.8 A flows to A′, B flows to B′. Compare lengths of L = ||AB|| and L′ = L||A′B′||.

By linear approximation, in a small time increment dt > 0, these points flow

to

and

Because

these are

and

This gives

But L = dx, which means that

Dividing by dt we get

What does this mean?

For the argument, the increment L = dx > 0 is fixed.
It tells us that as we follow a path of intervals whose end points are
determined by integration of the one-dimensional vector field, the lengths of
the intervals change in time at a rate of vxdx, namely, the one-dimensional
divergence of the velocity field v = v(x) times the length of the initial interval
L = dx > 0.
What happens if dx = 0? The right hand side of equation 4.1 is zero because
then [vxdx] = [vx · 0] = 0.
What about the left side of the equation? What is happening in the geometry?
When dx = 0, the initial interval is of length zero. Then A = B and A′ = B′,
which means that all lengths L = 0 = L′.

4.3 Divergence in 2-d and rate of relative change in areas

Let’s now work out the general case in 2 dimensions.
Suppose that v(x, y) = (a(x, y), b(x, y)) is a velocity vector field in the plane.
As above, we start with three initial points A, B1 and B2, where A = (x, y), B1
= (x + dx, y) and B2 = (x, y + dy).
These three points determine a parallelogram (in fact, a rectangle lined up
with the axes) with sides AB1 and AB2 with area [Area] = dxdy.
Suppose a common small time increment dt > 0.
Using that common parameter, we approximate new locations along
streamlines, A′, and .
These three points also determine a parallelogram, with area [Area]′. See
Figure 4.6. The problem is to compute
In order to approximate A′, and , we first need to know the velocity
vectors at each of A = (x, y), B1 = (x + dx, y) and B2 = (x, y + dy).
At A = (x, y),

At B1 = (x + dx, y),

At B2 = (x, y + dy),

Therefore,

and

Also,

We can now compute the sides of the new parallelogram:

and

Writing the pair of vectors in matrix form,

Problem 4.8.
Using the absolute value of the determinant, compute the area determined
by and . Show that

This reduces to

For a discussion about why the determinant function provides area, see
Section 1.7.

Recall that [Area] = dxdy. And so, up to linear approximation,

That is,

Note: This is not a rate of change of area. The solution is a rate of change of
relative change in area, determined by the flow of vertices along streamlines
(integral curves) that are simultaneously parameterized.34

Definition 4.1. In fluid dynamics, if the rate of relative change of area is

zero then mass is neither destroyed nor created. For that reason, a flow for
 which

is called incompressible.

Concluding remarks regarding relative change in area

Start with a small element of area and examine its progression for small time,
relative to the original small element of area. Consider the relative change of
area, with respect to time. In the limit, we obtain ax + by.
Recall that ax + by is called the divergence of the vector field, where that
name originates from estimating net mass-flow rates (flux) through a small
region whose boundary is a positively oriented curve.
Just as in our inquiry about circulation, two different questions have the
same answer:
At a point (x, y):

(1) The flux density per unit area is

(2) The time derivative of relative change in area is

Question: Why are these answers the same? In classical fluid dynamics this
is equivalent to a conservation law for a type of fluid. In Chapter 5, we go
into some of the mathematical details.

Summary of results about “curl” and “divergence” in 2-d

Suppose a velocity vector field v(x, y) = (a(x, y), b(x, y)).

The Jacobian of the vector field is
The Jacobian can be written uniquely as a sum of symmetric and
antisymmetric parts. That is,

where the symmetric part is

and the ant-symmetric part is

At a point (x, y):

Circulation:

(1) At a fixed time t, using a rectangle dx × dy, the circulation density

function per unit area is bx – aα. (By linear approximation)
(2) By summing both sides of (1), and taking limits as partitions are refined,
we get Green’s Theorem in the plane, that is:

Or, in words, with modest hypotheses, the circulation of v = (a, b) along

the closed boundary of R is equal to the area integral of curl v over the
region R.
(3) Following increments along streamlines, the quantity (bx – ay) is twice
the angular velocity of A′B′ relative to AB (using a common integral
curve parameter). For that reason, (bx – ay) is called the “curl” of the
vector field.
(4) In terms of the Jacobian Jv, curl is a difference of off-diagonal elements.
Alternatively, it is the off-diagonal element of the pure-rotational
 (antisymmetric) component (Jv)A of the Jacobian Jv.

Flux:

(1) At a fixed time t, using a rectangle dx × dy, the flux density function per
unit area is div v = div (a, b) = ax + by. (Linear approximation)
(2) By summing both sides of (1), and taking limits as partitions are refined,
we get the 2-d divergence theorem, that is:

Or, in words, with modest hypotheses, the flux of v = (a, b) along the
closed boundary of R is equal to the area integral of div v, over the
region R.
(3) Following areas as vertices flow along streamlines, the rate of relative
change in area is div v = div (a, b) = ax + by (using common integral
curve parameter).
(4) In terms of the Jacobian Jv, the divergence

_________________________
33In fluid dynamics, this is called “strain” in 2-d motion.
34See note 8.
 Chapter 5

UNIFYING 2-D RESULTS

5.1 Unifying results

The following discussion is for readers with some background in
undergraduate linear algebra.
In the summary given in Chapter 4, similarities and linkages invite further
inquiry.

(1) Green’s theorem and the 2-d divergence theorem tell us different things
about a vector field. Mathematically, however, the two theorems are
equivalent.
(2) In Green’s theorem, circulation density (bx – ay) is also equal to

(as defined above).

(3) In the 2-d divergence theorem, flux density ax + by is also equal to the
rate of relative change in areas (as defined above).
(4) The flux density is trace (Jv)S; and circulation density is “signed off-
diagonal trace” of (Jv)A.

What would be nice is a unified account of these similarities and linkages.

Writing both flux density and circulation density in terms of Ja-cobians
Flux density is trace (Jv)S = trace (Jv). Might circulation density bx – ay be
written in a similar way?
Circulation density is an “off-diagonal trace” of (Jv)A(x, y). Like flux density,
is circulation density also the trace of some matrix?
Certainly, it is, if we “manually” construct the new matrix

That is,

Why does that work?

Recall the equivalence of Green’s theorem and the 2-d divergence theorem.
“[A] circulation integral of v = (a, b) along a curve is a flux integral of w
= v⊥ = (b, –a) along the same curve; and vice versa.”
More precisely, the circulation density of v = (a, b) along a curve is the flux
density of v⊥ = (b, –a); and vice versa.
That means that the circulation density of v is the

Bringing the two results together,

Next, let’s have a look at equalities observed between density functions and
rates of change obtained from geometric constructions.
In the 2-d divergence theorem, flux density ax + by is equal to the rate of
relative change in areas.
The two quantities are computed and defined differently.
The flux density is a limit of ratios of line integrals to interior areas, with no
explicit reference to streamlines.
But the rate (of change) of relative change in areas is obtained by tracking
vertices of rectangles along respective streamlines, with a common
parameter.
Let’s explore a few examples.
Example
Suppose that the coordinate system is already aligned with the vector field. In
other words, with (x, y) coordinate lines, we suppose that the vector field is of
the form v(x, y) = (a(x, y), 0).

Figure 5.1 A vector field with y-component zero; v = (a(x, y),0).

Let’s compute both quantities, that is, both the ratio of flux to area and the
ratio of relative change in areas.
We can start with a rectangle whose vertices are listed in counter-clockwise
order:

For the rectangular region dx × dy with vertices given in counterclockwise

order starting with (x, y), the element of flux is given by the following
computation:
Along the upper and lower boundaries of the region, the vector field is
orthogonal to the normal vectors of the boundary curves. As the last equation
gives explicitly, the element of flux is, therefore,

Problem 5.1.
Show that, as before, by linear approximation — ignoring high order error
terms — the sum reduces to ax(x, y)dxdy.

Dividing by the area dxdy we get that flux density is ax(x, y) (which is ax + by
because in these coordinates, b ≡ 0).
In the same example, let’s look at the rate of change of the relative change in
areas.
As in the general derivation of the result given earlier, we need to compute an
approximation to

per unit time.

In the present example, motion is parallel to x axis.

Problem 5.2.
Show that the rate of change of relative area reduces to ax(x, y).

Notice that the coordinate axes are lined up with streamlines. Since motion is
parallel to the x-axis, the only contribution to flux density is ax(x, y); and the
rate of change of relative change of areas is ax(x, y).
 Problem 5.3.
Investigate the geometry in the somewhat more general case where

Problem 5.4.
Investigate the case where

with

That is, suppose that the Jacobian represents a pure rotation. Using only
elementary geometry, show that in the case of pure rotation, areas are
preserved. That is, show that

Hint: Remember that the components of the Jacobian of v are coefficients

for an ordinary system of linear differential equations and initial-value
problem. See Section 3.10. Observe also that trace(Jv) = 0.

5.2 The divergence and the Jacobian

The examples in Section 5.1 draw attention to the fact that, in certain cases, it
is clear from geometry that the divergence is the same as the rate of change of
the relative change in areas along streamlines.
The main clue is the following:
For any two square matrices A and B,

If we apply this to Jacobian matrices, we get

This means that, in addition to the canonical decomposition of the Jacobian
into its symmetric and anti-symmetric parts, the 2-d divergence along
streamlines is independent of the 2-d scalar curl.
Notice, however, that the decomposition into symmetric and antisymmetric
parts includes ay and bx in both summands:

Note also that the structures are linear.

Let M2 stand for the linear vector space of real 2 × 2 matrices.
While the representation is not unique, every matrix in M2 is the Jacobian of
some vector field v.
So, define the transformations S, A : M2 → M2 by

and
We already have that M2 = rgS ⊕ rgA. (See Section 3.10.)
Explicitly,

and

Observe also that (A ∘ S) = 0 and (S ∘ A) = 0.

The divergence of a vector field v is then given by the following composition
of linear maps:

Using the abbreviation ‘ker’ to stand for the kernel (or null space) of a linear
map, observe that for

we have

and

Furthermore, rg(A) ⊂ ker(trace ∘ S) is a one-dimensional subspace of the 3-

dimensional kernel of the linear map (trace ∘ S).
Note also that
and so a more explicit identification of ker(trace ∘ S) is:

What does the analysis of linear structure mean for a fluid with a 2-d velocity
field v?
If div v = 0 then, as defined in Section 4.3, the fluid is incompressible.
From the linear structure described above, an incompressible fluid can have a
velocity field that is partly rotational.
If div v = 0, then eigenvalues of the symmetric part (Jv)S are either both zero,
or are equal but of opposite sign.
If they are both zero, then there is neither contraction nor dilation relative to
the principle coordinates.
If they are non-zero but equal and of opposite sign, then there is contraction
relative to one of the principle axes matched by a dilation relative to the other
principle axis. See Figure 3.22.

5.3 The curl and the Jacobian

Let’s now sort why the (scalar) curl, or circulation density, is a “signed off-
diagonal trace” of (Jv)A.

Problem 5.5.
By going back to first principles, investigate 2-d (scalar) curl and geometry
in the case where the velocity field is of the form v = (a(x, y),0).

Let’s do it together.
We are inquiring about curl. There is, however, a correspondence between
curl and divergence, namely, curl v = div v⊥. Where does that lead?
Given the vector field v = (a, b), then v⊥ = (b, –a) also is a vector field with
its own Jacobian, that is,
The Jacobian of v⊥ has its own symmetric and anti-symmetric parts:

and

It follows that

The computation also reveals that the 2-d scalar curl is unaffected by 2-d
divergence along streamlines. Again, analysis of linear structures gives
additional results.
Observe that

is obtained from

by switching two rows and multiplying one row by (–1).

Hence (Jv⊥) = (QP)(Jv), where

Or, (Jv⊥) = T(Jv) = (QP)(Jv).

The transformation T : M2 → M2 defined by

is an isomorphism of M2:

(i) If (QP)C1 = (QP)C2, then

hence T is one-one.
(ii) Given C ∈ M2, then T((QP)−1C) = C, hence T is onto.

Problem 5.6.
Recall that if curl v = 0, then the vector field is irrotational. Because div=
trace(Jv⊥) = trace(T(v)), just as was done in Section 5.2, analysis of the
kernel of appropriate linear maps provides further results.

Orthogonal trajectories of v are determined by v⊥.

If v is the velocity field for a fluid then the orthogonal trajectories of v do
not represent mass-flow. What do they represent? If a common parameter is
used for integral curves of v⊥ then, physically, an orthogonal trajectory is a
front-line or leading edge of advancing fluid whose streamlines are
determined by v. In important applications, v = ∇f. In that case, there is
additional structure. This is explored somewhat in the exercises that follow.
Figure 5.2 Orthogonal trajectories of v = (a, b) are integral curves of v⊥ = (b, –a).

Exercises

Exercise 5.3.1
Let f(x, y) = x2 – y. Find the level curves f(x, y) = c, where c is a constant. Let
v(x, y) = ∇f = (2x, –1) and v⊥(x, y) = (–1, –2x). Show that the level curves of
f are the streamlines of v⊥(x, y) and that the orthogonal trajectories to the
level curves are the streamlines of v(x, y) = ∇f. Graph both sets of curves and
explain.
Exercise 5.3.2
Let r(u) be a flow line of a gradient vector field v = ∇f. Show that g(u) =
f(r(u)) is an increasing function of u. Thus a particle moving along a flow line
of the gradient field v = ∇f will move from lower to higher values of the
potential function f.

_________________________
35For example trace
 PART 2
FLUID MOTION IN 3 DIMENSIONS
 Chapter 6

FLUX AND DIVERGENCE

6.1 Beginnings
As in the 2-d case, we start with a velocity vector field. Now, however, we
look to the three-dimensional case. The velocity field is of the form v = (a(x,
y, z), b(x, y, z), c(x, y, z)) in an (x, y, z) coordinate system. We continue to
assume that velocity fields are independent of time. We also continue to
assume that all functions are continuously differentiable everywhere. For
brevity, we write v = (a, b, c), with the understanding that the coordinate
functions are of the form a = a(x, y, z), b = b(x, y, z) and c = c(x, y, z).

Problem 6.1.
What are some examples?
Figure 6.1 Mass per unit area is used to compute mass of volumes.

As in the 2-d case, the velocity field can be integrated to give integral curves,
trajectories of “test particles.” Also as in the 2-d case, initially, density
functions are empirically verifiable. For instance, for a layer of known depth
there is a mass-density per unit area. It is called “per unit area” with the
understanding that masses are of volumes.
For a tube of known diameter, there can be a length density function. Again,
mass is of increments of volume, in this case along a tube.

Figure 6.2 Volume has mass. Mass per unit length is used to compute mass of volumes.

With these subtleties understood, we now ask about mass-flow rates across
surfaces.
Suppose that an area density function μ is defined on surface S in (x, y, z)
space.
Mass-density is a property of a fluid. As in Part 1, we restrict to the case
where that function is constant. If that constant is not unity then, by an
appropriate adjustment of units, it becomes unity. Henceforth, we assume that
μ = 1. Under these hypotheses, this means that, numerically, momentum μ =
μ(a, b, c) = (a, b, c). Although, units of momentum are mass × velocity.

Problem 6.2.
Work out details for the change of scale needed if, for instance, mass-
density per unit area is gm per square cm.
Figure 6.3 Velocity field passes through surface area.

6.2 Flux across a surface S

We now assume that surface mass-density function μ is unity, per square unit
of area. Let v = (a(x, y, z), b(x, y, z), c(x, y, z)) be the mass-flow velocity
vector at (x, y, z) and suppose that dS is the area of a small “patch” or
increment of a surface S.
Question: How shall we define and compute the mass-flow rate across an
increment of surface area? As in 2-d, a key is that, in physics, momenta add.
It is that law of physics that justifies resolving a velocity vector field into
components relative to a reference frame.
Let dS be the area of a small “patch” at a point (x, y, z) on the surface S.
(Using linear approximation, this will be defined more carefully in sections
below.) Let n be a unit vector normal to the tangent plane of S at (x, y, z). For
a vector field v = (a, b, c), the increment of flux is the increment of mass flow
orthogonal to S at (x, y, z). That is, d(flux) = (v · n)dS.
In other words, the increment of flux across that patch of surface at is (the
component of v = (a, b, c) that crosses the surface orthogonally) × (the
increment of surface area).
Figure 6.4 Velocity field v(x, y, z), patch of surface area dS.

Now that we have “increments of flux” defined, much as in the 2-d case, this
allows for the following:

Definition 6.1. For v = (a, b, c) and suppose that n is a unit normal vector
normal field defined along the surface S. Relative to a unit normal vector
field n, the flux (mass-flow rate) across the surface S is

The definition includes the word “relative to n”. As you may already have
observed, at a point on a surface in (x, y, z) space, there are two possible unit
vectors. We will need to say more about that, shortly.
As in the 2-d case, this is an average, or integral. In the 3-d case, it is an
integral of mass-flow rates across the surface S in (x, y, z) space.
Figure 6.5 Increment of flux is increment of mass-flow orthogonal to surface, that is, d(flux) = (v ·
n)dS.

Exercises

In Exercises 6.2.1–6.2.8, find the flux for the given vector field v across the
surface S.
Exercise 6.2.1
v = (z, 5, 2y), S is the part of the plane x + y + z = 1 in the first octant with the
upward orientation
Exercise 6.2.2
v = (xy, yz, zx), S is the part of the paraboloid z = 2 – x2 – y2 that lies above
the square 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, and has upward orientation
Exercise 6.2.3
v = (y, xz, x), S is the hemisphere oriented in the direction of
the positive y-axis
Exercise 6.2.4
v = (x, y, z4), S is the part of the cone under the plane z = 1 with
the upward orientation
Exercise 6.2.5
v = (x, y, 2z), S is the part of the plane z = 3x + 2 that lies within the cylinder
x2 + y2 = 4 with the upward orientation
Exercise 6.2.6
v = (xz, x, y), S is the part of the part of the sphere x2 + y2 + z2 = 81 in the first
octant with the upward orientation
Exercise 6.2.7
v = (3x, y, 5), S is the part of the paraboloid z = 9 – x2 – y2 that lies above the
xy-plane with the upward orientation
Exercise 6.2.8
v = (xz, x, y), S is the hemisphere oriented in the direction of
the positive y-axis

6.3 Flux across a surface that bounds a volume

Example
Let S2 = {(x, y, z)|x2 + y2 + z2 = 1}, the surface of the sphere of radius 1
centered at the origin (0, 0, 0). Let v = (a, b, c) be a velocity field.
The problem is to compute

How is the definition of flux to be applied?

To compute the integral, we need to make use of parameterizations for S2 of
the form

For the surface of the sphere, however, there is no single two-variable

differentiable parameterization that covers the entire surface. Still, we can
patch a few together, so long as taken together all points of the surface are
accounted for.
However, to see that there is a potential problem here, let be the “north
pole” hemisphere and be the hemisphere of points where x > 0. A
familiar parameterization for is

In the same way, a similar parameterization for is

Except for coordinate lines missing, is a quadrant of the surface of

the sphere.

Figure 6.6 A quadrant of the surface of the unit sphere. Two parameterizations partly overlap.

Question: Which parameterization do we use for ?

If it doesn’t matter, then the integral can be computed patch by patch, taking
care not to “double count,” or “triple count,” etc., on regions where
parameterized subsets of the surface overlap.
This raises the following more precisely formulated question:
Question: In the general case, what happens to an integral for flux when
there is a change of variables?
As in the 2-d case, it is an elementary but crucial question. If you are familiar
with multi-variable integration, you might correctly expect that the integral is
invariant.

Problem 6.3.
Just as in the 2-d case, invariance reduces to the fact that there is a chain
rule for a change variables in a surface integral. The proof is “in the books”
but it is an excellent exercise to sort through the details for oneself, namely,
that

is invariant under a change of parameterization. This means that flux is

well-defined.

Preparing to continue with the example v · n dS for the surface of the

sphere S2
As sample parameterizations given above already indicate, computations can
be somewhat involved. So, before continuing with the example, let’s pause in
order to explore a similar but more elementary problem.
Figure 6.7 A 3-d coordinate box dx × dy × dz based at x0, y0, z0.

Let S(x0, y0, z0) be the six-sided surface of a 3-d “box” with sides parallel to the
coordinate planes, with lengths dx, dy and dz, and corner point (x0, y0, z0) (see
Figure 6.7).
Much as in the 2-d case, in 3-d, one way to use approximate “net flux” near a
point is by using linear approximation and summing differences in flux
across opposite (parallel) sides. Using that approach, net mass-flow through
the 3-d box dx × dy × dz is approximated by

Problem 6.4.
Retaining only first-order terms, the “net flux” indicated is

Note: We wrote “net flux” instead of ‘flux’ because we were appealing to the
diagram and did not use normal unit vector field n as is required by the
definition of flux.
This raises a question: Can the sum of terms just obtained be expressed in
terms of a normal vector field n defined on S(x0, y0, z0)?

Problem 6.5.

(1) The identification is made by making n the “outward facing unit

normal” on each face. The six cases are
n = (±1, 0, 0), (0, ±1, 0) and (0, 0, ±1).
(2) Note that n = (±1, 0, 0), (0, ±1, 0) and (0, 0, ±1) as described is not
differentiable. We can ignore the problematic points. What points are
they? Why, for present purposes, can we ignore them?

And so we get that if E(x0, y0, z0) = dx × dy × dz, the coordinate box based at
(x0, y0, z0), then by using linear approximation, the flux over the surface of
the box is the divergence of the vector field times dxdydz, the volume of the
box. That is,

Continuing with the example v · n dS

The result just obtained is for a small coordinate box.

Can we use this to build up to a result about the surface of the sphere?
We can introduce a 3-d partition for the volume interior to the sphere, a
partition consisting of boxes Eijk, each of dimension dx × dy × dz and volume
dxdydz. Summing over boxes Eijk interior to the 3-d volume E (the unit ball)
gives an approximation to the volume integral of the divergence of the vector
field.
But, what do those volume sums have to do with flux along the surface S2?
Hint: See Figure 6.8. Each box has its own unit normal vector field. Along a
common face, how do unit normal vectors compare? How do surface
integrals combine when obtained from a common face?

Figure 6.8 Hint for Problem 6.6.

Problem 6.6.

(1) Show that if we add equation 6.1 for two adjoining boxes in a partition,
common surface integrals cancel.
(2) Repeat this result for four adjoining boxes.
(3) Extend this result to the entire triply indexed sum of boxes Eijk based at
(xi, yj, zk) whose six-sided surfaces are labelled Sijk.

The results of Problem 6.6 lead to the following two ways of evaluating the
same limit:
By linear approximation, for each partition of the volume, finite sums can be
written as
with

From the exercise, however, cancellation of surface integrals along common

faces means that the first sum also can be written as

And so it follows that

The divergence theorem in 3-d

Various aspects of the example can be generalized.
We first need to say something about the meaning of “positive orientation”
and “outward facing unit normal.” In undergraduate calculus textbooks,36
this is handled by appealing to diagrams and convention. For instance, by
looking to diagrams and graphs of functions, one identifies an “outward unit
normal” for an ellipsoid. But, in the textbooks, this is done by appealing to a
diagram rather than a definition. Moreover, a normal vector field does not
always exist. (The canonical example is the Möbius strip.) When a normal
vector field does exist, it is not unique (For w = 0, n · w = 0 if and only if
(–n) · w = 0).
A choice of unit normal vector field on a surface is called an “orientation.”
We won’t say more on this issue here. We are drawing attention to the fact
that we are not ready to rigorously define “outward unit normal.” For now,
we need to rely on insight into diagrams and understanding of particular
cases. We hope that you are beginning to see there is a question here. But to
handle the question of “orientation” would take us well beyond the context of
these introductory notes.
In the same way, we won’t get into attempting to rigorously define “simple
solid region” in (x, y, z). Preliminary definitions can be found in Stewart’s
book,37 as well as in many other Calculus III textbooks.
With these caveats, the example brings us to the divergence theorem in 3-
dimesions.

Theorem 6.1. Let v = (a, b, c). Suppose that E is a simple solid region in
(x, y, z) and S is its boundary surface. This is written S = ∂E. Then the flux
integral of v over the boundary S = ∂E is equal to the volume integral of
the divergence of v over the volume E.
That is,

Problem 6.7.
Extend the method of the previous example to establish this result. In a
more advanced course, rigorous limit arguments would be needed.

The solution of a technical problem that hints at the need of a higher

viewpoint for surfaces and integrals
From Part 1, we have a similar integral for flux in the plane. That is, the flux
of w across C is
Figure 6.9 Because of linear approximation, surface integrals over boxes of the form dx × dy × dz are
volume integrals over the interiors of dx × dy × dz. Respective sums converge to the same limit,
because of inner cancellations of oppositely oriented common surfaces.

where, along C, N is a unit vector orthogonal to the curve (tangent vector)

and is a “outwardly directed.” (In the present 3-d context, the symbol ‘v’ is
already in use. So for this discussion we write w = (e(x, y), f(x, y)) for a 2-d
vector field.)
Initially, the flux integral depends on a vector N that is defined by appealing
to an ambient space. The question of change of variables also came up.
Thanks to the chain rule, both issues were resolved:
(1) In Section 4.1, it was shown the integral can be written as a line integral
without reference to a “normal vector N”; and
(2) If two parameterizations have the same orientation along a curve, then the
integral is invariant under change of parameterization. (Two
parameterizations of a curve have the same orientation if at each point their
tangent vectors are proportional by a strictly positive scalar function defined
along the curve.)
We recall some main details of the calculation for (1), in the 2-d case:
To compute w · N ds, first obtain terms for the integrand.
These are:

Observe that ||(y′, – x′) ||=|| (x, y)||.

Substituting into the integral, we get

In other words, the net mass-flow rate of w = (e, f) along C is a line integral
of the vector field, without having to appeal to an additional vector defined in
the ambient 2-d space.
Question: What about the 3-d case?
We have defined 3-d flux as v · n dS.

(1) Can we write this in a way that does not require construction of a
positively oriented normal vector n?
(2) Is the integral invariant under (some) changes of parameterization?

In the 3-d case there are additional terms. But, as it turns out, we can mimic
the 2-d argument.
As in the 2-d case, we begin by computing terms for the integrand. A local
parameterization is needed, of form

For review, see a Calculus III textbook.

By definition, the term dS is an element of area of the tangent plane of S.
The tangent plane at a point has two (basis) vectors,
The parallelogram determined by these two vectors has area ||Tu × Tv||. The
parallelogram determined by

Figure 6.10 A parameterization of a 2-d surface in (x, y, z) space: r : (u, v) → (x(u, v), y(u, v), z(u, v)).

has area

A normal unit vector is .

Bringing these computations together, we get

Working this out explicitly gives

Observe that the components of Tu × Tv are (signed) areas obtained by

projecting the parallelogram determined by Tu and Tv onto coordinate planes.
To see that, observe that the absolute value of the first component is a cross-
product:

Figure 6.11 Basis vectors Tu, Tv for tangent plane of S at r(u, v).

We also need to multiply by dudv.

As in the 2-d case, we can invoke the chain rule and obtain

(where dy and dz are the coordinate intervals for sides of the parallelogram in
(x, y, z) coordinates).
The other components are computed similarly.
That is, the second component of Tu × Tv is –dxdz and the third component is
dxdy.
Therefore,
This implies that the flux is

That is, as hoped, we obtain an expression for flux that does not explicitly
involve a normal vector n.

Figure 6.12 By definition, projections of Tu and Tv onto (y, z)-plane are (0, yu, zu) and (0, yv, zv).
Area of projected parallelogram is therefore |yuzv – yvzu|; which is the absolute value of the first
component of Tu × Tv. Similar computations hold for the other two projections.

How is this formula used? It provides a template for evaluation of flux.

A “template”? What that means is that in each case we introduce a
parameterization and for the terms dydz, dxdz, dxdy, and use the chain rule.

Problem 6.8.
Go through the details to show that the flux is invariant under change of
parameterization.
Observation: The formula

depends on the surface S and the vector field v = (a, b, c) defined along S.
The flux integral is, therefore, a function of “two variables,” where the
“variables” are surfaces and vector fields, respectively. This hints of a higher
viewpoint needed to handle integrals on surfaces. For a two-parameter
surface, “integrands” are of the form

These integrands are called “2-forms” and the surfaces are called
“manifolds.” The definition of manifold embraces the fact that
parameterizations are not unique and that, as observed above, some surfaces
(e.g., S2) require more than one parameterization to cover an entire surface.

Exercises

In Exercises 6.3.1–6.3.4, use the Divergence Theorem to calculate v · dS.

Exercise 6.3.1
v = (3xy2, xez, z3), S is the surface of the solid bounded by y2 + z2 = 1, x = –3,
and x = 1.
Exercise 6.3.2
v = (x2y, xy2, 2xyz), S is the surface of the solid bounded by x = 0, y = 0, z =
0, and x + 4y + z = 4.
Exercise 6.3.3
v = (–x4, x3z2, –4xy2z), S is the surface of the solid bounded by x2 + y2 = 1, z
= x + 10, and z = 0.
Exercise 6.3.4
v = (4x3, 4y3z, 3z4), S is the unit sphere.

_________________________
36See, e.g., James Stewart, Calculus: Early Transcendentals, 6th ed., (Belmont, CA: Thomson Higher
Education, 2008): 1067.
37See note 36.
 Chapter 7

STOKES’ THEOREM

7.1 Stokes’ Theorem: The Question

Figure 7.1 w = ProjSv.

Suppose that S is a two-parameter surface in (x, y, z) space with boundary

curve ∂S = C, and that v = (a(x, y, z), b(x, y, z), c(x, y, z)) is a vector field
defined on a volume containing S.
In physical applications, questions arise about the component of v parallel to
S.
Key point: The question is not about v and the surface S but is, rather, about
w = ProjSv and the surface S.
More precisely, the inquiry is about the projection of v onto (the tangent
plane) at each point of S.
This is similar to the problem solved by Green’s Theorem, in the sense that
we have a vector field w defined on a surface with boundary curve ∂S = C.

Figure 7.2 Given v(x, y, z) in (x, y, z)-space, w = ProjSv is a vector field along a 2-d surface S, with
boundary curve C = ∂S.

Can a circulation along the boundary ∂S = C be computed as an integral of a

circulation density on the surface S? Is there an extension of Green’s integral
formula for a region in the plane to an integral formula for a region that is a
surface in (x, y, z) space?
Question: In other words, is there a “Green’s Theorem” for two-dimensional
surfaces in (x, y, z) space?
The curl operator in the (x, y) plane cannot be applied directly to surfaces in
(x, y, z) space, surfaces that, in general, can be curved and pitched at angles
relative to the (x, y) plane. We will somehow need to appeal to the fact that,
locally, the surface S can be parameterized by a function of the form

(1) One approach to the problem follows what we did in the development of
Green’s Theorem. That is, begin with the definition of circulation.
As in the 2-d case, linear approximation can be used to replace a combination
of line integrals, with surface integrals. Note that since there are three
components and two parameters, the computations become somewhat
involved but they do not require new ideas as such.
(2) Another approach is to take advantage of (*). Then apply Green’s
Theorem to quantities in the (u, v)-plane.
With some guidance provided, both approaches are good homework
exercises and are illuminating in different ways.
By making use of projections, the first approach calls forth techniques that
eventually are needed in modern differential geometry and applications. The
second approach implicitly incorporates the projection in (1) and reveals
coherence of operations in vector calculus.
Below, we provide some details on both sets of computations.

7.2 Stokes’ Theorem: First approach: elementary derivation

of “Green’s theorem” for a two-parameter surface
A first step is to compute the projection of v = (a, b, c) onto the tangent plane
of S.38 See Figure 7.1.
Two basis vectors tangent to the surface are:

and

We can use the cross-product to get a vector field normal to the tangent
plane. That is, set

(We use the capital letter ‘N’ rather than lower case ‘n’ because the nor
vector just computed is not, in general, a unit vector as described in Sec 6.2.)
Explicitly,

The components for N = Tu × Tv will be used below. For now, let’s keep our
focus on the geometry of vectors.
Using the dot product, ProjN(v), the projection of v onto N, can be computed
directly.
The vector field that we need (ProjS(v), the projection of v onto the surface S)
then is defined by the equation

In other words,

Using the fact that v · N = ||v|| ||N|| cos θ,

Therefore, we get that

Let’s rename the projection vector w = ProjS(v).

The vector field w is defined on, and tangent to, the surface S.
The inquiry, then, is about the circulation of w along C = ∂S, the boundary of
the 2-d surface S ⊂ ((x, y, z) space).
As we did in 2-d, we can first look to a simpler case. That is, let’s look to the
curve on the surface determined by a positively oriented boundary of a
coordinate interval.
In the present setting, a coordinate interval is du × dv.
By linear approximation, we get the boundary of a parallelogram on the
surface in (x, y, z) space, two of whose sides are duTu and dvTv.
By definition of the cross-product, the element of surface area dS is ||duTu ×
dvTv||. We have an element of the surface, namely, a parallelogram tangent to
the surface, with sides duTu and dvTv. From undergraduate coordinate
geometry, that planar surface element is represented by a normal vector,
namely, the surface element vector

As in Part 1, there are four sides, I, II, III and IV, each of which has its own
initial and final points in (x, y, z) space. These are obtained by tracking
counter-clockwise around the boundary of coordinate interval du × dv and
consequently producing the boundary of the parallelogram in (x, y, z) space.
On sides I, II, III and IV, respectively, the contributions to circulation are:

Figure 7.3 Parallelogram with sides duTu and dvTv.

side I:
side II:

side III:

side IV:

Next, collect like-terms and use linear approximation to estimate differences.

Problem 7.1.
Show that, by ignoring high order error terms, circulation around the
positively oriented boundary of the coordinate interval du × dv mapped to
(x, y, z) space is

The vector

is called the “(3-d) vector curl.”

That is,

Note: Computing the vector curl in two ways:

1. As a computational device, the components curl (a, b, c) can be obtained in
cyclic order from the following matrix:

For the first component, suppress the a and ∂x column and use the next two
columns to compute curl (y, z)(b, c).
For the second component, suppress the b and ∂y column and use the next
two columns to compute curl (z, x)(c, a).
For the third component, suppress the c and ∂z column and use the next two
columns to compute curl (x, y)(a, b).
The subscript notation for the pairs of variables is, we hope, clear.
This way of computing the 3-d curl vector emphasizes the fact that the
components of the 3-d vector curl are 2-d scalar curls.
More precisely, the components of curl (a, b, c) are the scalar curls of the
projections of v(a, b, c) onto the (y, z) plane, the (z, x) plane and the (x, y)
planes, respectively.
Note also that this ordering and naming of coordinate planes matches the
derivation above. To see that, write (x y z x y z) and move from left to right in
cyclic ordering. That is, first suppress x to get (y, z); then suppress y to get (z,
x); and finally suppress z to get (x, y).
These aspects of curl (a, b, c) = ((cy – bz), (az – cx), (bx –ay)) will be further
explored in sections below.
2. Another common way to compute curl (a, b, c) is to use symbolism
derived from determinants. That is,

where i = (1, 0, 0), j = (0, 1, 0), k = (0, 0, 1) and the determinant formula is
applied symbolically. Observe that each row has different types of terms.
 Problem 7.2 (Stokes’ Theorem).
Fill in needed details. Introduce a partition of the (u, v) parameter space
with coordinate intervals du × dv. Observe that circulation around the
boundary of S can now be computed in two ways (two different limits) and
that we therefore get Stokes’ theorem, namely, with all terms defined as
above,

Note: Just as in the 2-d case, the key is to use linear approximation to replace
differences of functions by their derivatives multiplied by appropriate
coordinate interval lengths.

Problem 7.3 (A technical issue).

The computations in 1. rely on parameters (u, v). You may remember that
more than one parameterization may be needed to cover a surface. Show
that integrals in the formula

are invariant under change of parameterization (u, v) → (s, t).

Suppose a region R1 in the (u, v) plane and a parameterization

maps R1 to S1 ; and a region R2 in the (s, t) plane and a parameterization

maps R2 to S2. The regions S1 and S2 and their boundaries ∂S1 = C1 and dS2
= C2 may overlap. Provide details on why the invariance of the integrals
makes it possible to apply Stokes’ theorem to S1 ∪ S2 with boundary ∂(S1
∪ S2).
Figure 7.4 See Problem 7.3.

Without loss of generality, then, let’s suppose that a single parameterization

is all that is needed. Otherwise, we would patch together a finite number of
parameterizations.39 Formalizing this needs finite unions and intersections of
S1, S2, · · · , Sm.
Stokes’ Theorem: Second approach: applying Green’s theorem in the (u,
v) parameter plane
Question: From the first approach, we can be assured that Green’s theorem
lifts to surfaces in (x, y, z) space. That lifted version of Green’s Theorem is
called Stokes’ Theorem. However, can we avoid having to go back to first
principles? Might we instead use what we know from the 2-dimensional
case? That is, can we lift Green’s theorem to a surface in 3 dimensions more
directly?
The question can be put symbolically. Doing so brings focus to our search for
middle terms.

Let’s follow our hunch. We can begin with a direct computation:

where, as above,

Here, ∂R is a curve in the (u, v) plane that gets mapped to C in (x, y, z) space.
Green’s theorem applies directly to the (u, v) vector field (v · Tu, v · Tv).
That is,

Here, R(u, v) and ∂R(u, v) are in the (u, v) plane.

We are attempting to make use of the parameters (u, v), so let’s explore that
dependence more carefully.
Figure 7.5 Parameterization r maps R in (u, v)-space to 2-d surface S in (x, y, z)-space. Express
integrals defined on S in terms of parameters (u, v). Apply Green’s Theorem in (u, v)-space.

In the (u, v) plane, the (scalar)

From Calculus III, the middle terms cancel giving

Problem 7.4.
 We know what we’re hunting for. So, all that we need to do is verify that
the integrand

It then follows that

Finally, then, we get again Stokes’ theorem.40

The joint resolution of curl (v) and dS
Two vectors in play are

and

They emerge together, by computing the circulation density of v = (a, b, c)

around a surface element determined by duTu and dvTv, a sub-parallelogram
of the parallelogram determined by the two tangent vectors Tu and Tv. See
Figure 7.3.
By definition of the cross-product (Chapter 1), a vector that is orthogonal to
the parallelogram determined by the two tangent vectors duTu and dvTv, and
whose length is equal to it, is N = Tu × Tv. Let’s now compare the
coordinates of curl v and N = Tu × Tv.
The components of

are the 2-d scalar curls of the projections of v = (a, b, c) onto the (y, z) plane,
the (z, x) plane and the (x, y) plane.
The components of

are the signed areas of the projection of Tu × Tv onto the (y, z) plane, the (z,
x) plane and the (x, y) plane, respectively.
We have, then, two vectors, three projections, and four areas. What are the
four areas? They are the area A of the parallelogram determined by Tu and Tv
together with A(y,z), A(z,x) and A(x,z), the areas of the three projections of the
parallelogram of area A onto the (y, z) plane, the (z, x) plane and the (x, y)
plane, respectively. See Figure 6.12.
Is there a relation between the areas A, A(y,z), A(z,x) and A(x,z)?
If we bisect the surface area of the parallelogram determined by Tu and Tv
along its diagonal, we obtain a triangle. That triangle is of area A. It can be
projected onto the (y, z) plane, the (z, x) plane and the (x, y) plane, providing
triangles of areas A(y,z), A(z,x) and A(x,z), respectively. An elementary
computation shows that

from which it follows that

This implies that the integrand in Stokes’ theorem can be written as follows:

where
Comment: This shows explicitly that circulation density on a 2-d surface in
3-d depends on how the 3-d vector curl v is pitched relative to the normal
vector of the surface. This is not surprising, however, since Stokes’ theorem
is about the projection of v onto S. In particular, if v is parallel to N = Tu ×
Tv, then the projection is zero, as are the integrals.
The calculation also shows that the standard Euclidean structure of the 3-d
vector dS is also a Euclidean structure of triples of projections of 2-d areas.

Figure 7.6 The 3-d vector curl v relative to the normal vector dS = Tu × Tv.

Exercises

Exercise 7.2.1
Let S1 be the portion of the paraboloid z = 4 – x2 – y2 above the xy-plane and
S2 be the hemisphere x2 + y2 + z2 = 4, z > 0. Suppose v(x, y, z) is a vector
field whose components have continuous partial derivatives. Explain why
In Exercises 7.2.2–7.2.4, verify Stokes’ Theorem for the given vector field v
and surface S.
Exercise 7.2.2
v = (y, x, xz), S is the part of the plane x + y + z = 1 that lies in the first octant,
oriented upward
Exercise 7.2.3
v = (y – z, x – z, y – x), S is the hemisphere , oriented upward
Exercise 7.2.4
v = (x2z, xy, xz2), S is the portion of the paraboloid z = 9 – x2 – y2 above the
xy-plane, oriented upward
Exercise 7.2.5
Use Stokes’ Theorem to evaluate v · dr, where v = (x2 + 2xyz, x2y2, x2y3)
and C is the intersection of x2 + y2 = 1 and z = x – y oriented clockwise
Exercise 7.2.6
Let , R is the unit circle x2 + y2 ≤ 1, z = 0.
Compute both sides of the formula in Stokes’ Theorem. Does this example
contradict Stokes’ Theorem? Hint: Implicit in the development of the
theorem are hypotheses about domains. Also implicit is the continuity of
partial derivatives of the components of v in xyz-space. These issues are
accounted for in axiomatically rigorous presentations of the theorem.
Exercise 7.2.7
In electromagnetism, current through a surface generates a magnetic field
along the curve that is boundary to the surface; and vice versa. Let B(x, y, z)
represent the magnetic field corresponding to a steady current J(x, y, z). For a
surface S with ∂S = C, the integral form of Ampère’s law is:
Use Stokes’ Theorem to obtain the differential form of Ampere’s law, that is,
curl B = μ0J.

_________________________
38You will find all of this done in Calculus III textbooks.
39Why can we assume that the number is finite? It is because the examples of surfaces given are all
“compact.” The definition of compact is provided in more advanced texts.
40A similar derivation that uses vector product identities is: Harry F. Davis and Arthur David Snider,
Introduction to Vector Analysis, 7th ed. (Mt. Pleasant, SC: Hawkes Learning, 2000): 298.
 Chapter 8

RELATIVE CHANGE IN
VOLUMES AND IN
INCREMENTS

8.1 Rate of change of relative change in volume

Let v(x, y, z) = (a(x, y, z), b(x, y, z), c(x, y, z)).
In (x, y, z) space, four relatively close reference points A, B1, B2 and B3
determine a small parallelepiped with sides AB1, AB2 and AB3 and volume
[Volume].
If we (linearly) approximate the motions of all four points along their
respective streamlines to A′, , and , we get a new parallelepiped with
sides A′ , A′ and A′ and volume [Volume]′.
Figure 8.1 Vertices A1, B1, B2, B3 of box in (x, y, z)-space flow to vertices of
parallelepiped in (x, y, z)-space.

Problem 8.1.
Using only computations from elementary geometry of vectors, and then
linear approximation, show that the rate of relative change in volume is

Definition 8.1. Since the rate is determined along streamlines, just as for 2-
d flows treated in Part 1, a 3-d fluid flow is called incompressible when

Problem 8.2.

Explore cases where . For instance, show that a rect-angle

lined up with the coordinate axes flows to rectangles also lined up with
coordinate axes.

8.2 Rate of change of relative change of increments

As in Section 3.10, we can explore the rate of relative change of increments.
Suppose that A and B are relatively close initial points in (x, y, z) space.
Suppose that, up to linear approximation, A′ and B′ are the points to which A
and B flow along their respective streamlines, in small times dt > 0. The
problem is to compare A′B′ with AB. That is, the problem is to approximate
the relative strain A′B′ – AB.

Problem 8.3.
Show that, just as in the 2-case, linear approximation gives us that

Hence, the relative strain vector is given by

Figure 8.2 Initial points A(x, y, z) and B(x + dx, y + dy, z + dz) flow to A′, B′. Use v(x, y, z) and v(x+dx,
y+dy, z+dz) and common time increment dt > 0 to approximate A′ and B′.

With Jv the Jacobian, that is

we get

Problem 8.4.
As we did in the 2-d case, in a unique way, we can decompose the Jacobian
into a sum of the form

where (Jv)S is symmetric and (Jv)A is anti-symmetric.

Recall that a matrix M is symmetric if and only if Mt = M; and a matrix M
is anti-symmetric if and only if Mt = –M.
As in the 2-d case, the superscript ‘transpose’ is abbreviated by ‘t’.
Show that

Observation:

and
As in 2-d, the real symmetric matrix (Jv)S can be diagonalized relative to
“principal coordinates.” See an undergraduate linear algebra book that treats
the real spectral theorem. The matrix (Jv)S is called the linear approximation
for pure strain and (Jv)A is called the linear approximation for pure rotation.
The reason for the name “strain” is that relative to the principle coordinates,
(Jv)S gives the linear approximation to relative change in increments along
coordinate axes. Why the name “rotation” is appropriate is brought out in
Section 8.3. (Recall that Section 3.9 treats the same problem in two
dimensions.)

8.3 Pure rotation (Jv)A represents an isometry

Notice that, except for a constant factor of , the components of (Jv)A are the
same as the components of the derived vector field

In Section 7.2, ((cy – bz), (az – cx), (bx – ay)) was obtained not by estimating
relative changes in increments A′B′ – AB but by computing ratios of integrals
to obtain circulation density along a surface, per unit area.
This leads to the following question:
Question: Why is the same vector obtained as the answer to the two different
questions?
As we discussed above, there are three projections of v onto coordinate
planes, each of which has its own scalar curl. We have already sorted out the
connection between 2-d circulation and 2-d rotation. The problem here is to
assemble 2-d results in the 3-d setting.
The matrix (Jv)A is the anti-symmetric part of the Jacobian of a velocity field,
the components of which represent a system of differential equations for
streamlines. For now, let′s ignore the symmetric part. We will deal with that
shortly.
For (Jv)A, the system of differential equations is:

We treat each (two-dimensional) coordinate plane in turn.

Let (Jv)A(yz) be the restriction of (Jv)A to the (y, z). This is implemented by an
orthogonal projection along the x axis. That is,

This represents pure rotation in the (y, z) plane, with angular velocity

Up to linear approximation, the projection of the flow is of the form

where r(y, z) is the projection of the initial length r = || AB || = || (α, β, γ) || onto

the (y, z) plane.
Similarly, we get two other coordinate plane projections of the flow:
where r(x, y) is the projection of the initial length r = || AB || = || (α, β, γ) || onto
the (x, y) plane; and

where r(z, x) is the projection of the initial length r =|| AB ||=|| (α, β, γ) || onto
the (z, x) plane.
Each of the projections are rotational in 2-d.
What does that mean about the given flow in 3-d? Is the 3-d streamline also
rotational?
The projections are known and so

Adding equations, we get that

But r =|| AB || = || (α, β, γ) ||. Computing the right-hand side explicitly, we get

Therefore, in the linear approximation,

It follows that the flow for (Jv)A is an isometry in 3-d.

Problem 8.5.
 The vector field curl v is independent of the relative strain.

(i) As noted at the beginning of this section, except for a constant factor
of , the components of (Jv)A are the same as the components of the
derived vector field

(ii) With and , not only is M2 = rgS ⊕ rgA,

but A ∘ S = 0, S ∘ A = 0, S ∘ S = S and A ∘ A = A.
(iii) The Ker A = rgS and the ker S = rgA.
(iv) There is a linear isomorphism A(Jv) → curl v.
(v) Show that the 3-d vector curl is independent of S(Jv).

8.4 Flux density and relative change in volume

Problem 8.6.
We have two ways of defining the divergence of a vector field in 3-d: 1. as
a flux density; and 2. as rate of relative change in volume.

(i) Explore cases where the Jacobian is of the form .

(ii) In the general case, trace (S(Jv)) = div v.

(iii) Show that the divergence is sum of the eigenvalues of S(J(x,y,z)v). Why
does this make sense for a fluid flow? Hint: You will need to invoke a
theorem from linear algebra.
(iv) For the general case, use the properties of the trace function to show
that the 3-d divergence is unaffected by the symmetric part of the
Jacobian.

Problem 8.7.
By direct computation of derivatives, show that for any vector field w, we
get div (curl (w)) = 0. Hence, if v is of the form v = curl (w) for some
 vector field w, then the flow v is incompressible.

Problem 8.8.
Suppose that v is the velocity field for a mass-flow in (x, y, z) space. For a
vector field in 2-d, we identify “orthogonal trajectories,” that is, “integral
curves” that are everywhere orthogonal to the original vector field. (See the
last paragraph of Section 5.3.) In 3-d, there is a similar construction. In 2-d,
by using a common parameter, the orthogonal trajectories are moving front-
lines in a mass-flow. In 3-d, we have a moving front-surface. We need the
equation for the surface that is everywhere orthogonal to the mass flow
given by v. In other words, the defining differential equation for such a
surface is of the form v · f = 0. At each point (x, y, z), there is a two-
dimensional vector space that solves the equation. The problem then is to
inquire into the possibility of integrating the system to produce surfaces that
are everywhere orthogonal to v.

Figure 8.3 Initial points A(x, y, z) and B(x + dx, y + dy, z + dz) flow to A′, B′. Use v(x, y, z) and v(x+dx,
y+dy, z+dz) and common time increment dt > 0 to approximate A′ and B′.

The need to study such systems of (differential) equations arises naturally in

the present context. But theorems needed go well beyond basic insights in
vector calculus.
 PART 3
SUPPLEMENT: MATHEMATICAL
UNDERSTANDING
 Introduction to the Supplement

The main text was on vector calculus. Our intention is that this Supplement
will be of wider application. The purpose is to help teachers and students
make beginnings in describing (instances of your own) mathematical
development. As will become evident, this will be useful for all levels of
pedagogy.
Part of the novelty of our approach here is that, through examples and
directed questions, we invite you to grow in understanding your own
mathematical understanding and that way provide yourself with essentials for
growing as a teacher.
The meaning of the italicized statement will emerge and grow by doing
exercises such as those provided in Part A. What we are referring to will be a
personal achievement. For the teacher, however, it is rarely an altogether
private achievement. We say that because, of course, as teachers, what we
think about mathematical development (which includes one’s own
mathematical development) factors into what we attempt to promote in our
students.
Thought on teaching and learning mathematics goes back to antiquity. For
instance, although in a different context and not called “mathematics
education,” there were Plato’s reflections about the slave-boy who is helped
to solve the problem of doubling the area of a square (Meno). However, the
discipline now called “mathematics education” is mainly a 20th century
development.41 Good work has been done. For, instance, advances have been
made in “experiential learning” and “engaged learning.” However, looking at
the entire discipline, results have been mixed and opinion varies regarding
the extent to which methods have been effective, or not.42 In Part C, we
provide a few comments regarding these issues.
What we would like to do first, however (in Part A), is invite you to a few
relatively novel43 exercises in mathematics. There is no “conceptual model”
or “representational system.” Nor do we draw on or appeal to student test
scores, classroom observations, student task orientation, student cooperation,
student behavior or other evidence that mathematical learning might or might
not be happening for someone else. It’s not that such results will not
eventually contribute to progress in understanding mathematical development
in history. The focus that we invite here is more elementary. We invite you to
“go to source.” The mathematics will be familiar. Part of what is new is that
you are invited to a precise “puzzling about your own puzzling in
mathematics.” The focus, then, is you, and me, or rather, at least initially,
you-about-you and me-about-me. The invitation is to make initial progress in
being able to advert to,44 focus on, inquire about and discern orderings of
distinct events in our own inquiry and understanding, in instances, when we
are doing mathematics.45
We introduce a key diagram (Figure A.1) which could — in a sense — be
said to be a “model.” But it is not a “conceptual model” or “speculative
framework.” Think more of something analogous to what the physics
community has been getting to, namely, a “best-to-date standard model”
which, for the most part, was discovered in and continues to be verified (or
not) in instances, in experience. Or again, in chemistry, one could say that
the periodic table is a “model.” However, the periodic table emerged from
and has been established through centuries of ongoing experimental work
investigating chemical dynamics. In a somewhat similar way, we think that
you will begin to see that Figure A.1 also is no mere model. By turning
attention to our own experience, in exercises in Part A, Figure A.1 emerges
and is verified. It too is a “table of elements,” that is, “elements of dynamics
of knowing.” Note that Part A is only a first few “experiments.” We leave it
to interested readers to go further, to begin exploring the significance of
Figure A.1 in other instances in your own mathematical development (and
more).46
What all of this may have to do with improving teaching in mathematics will
be touched on briefly in part B. Part C draws attention to a few anomalies in
contemporary mathematics education — in both content and method. A
challenge for the education community will be to take advantage of good
work that has been done but to also make progress in getting a handle on the
various anomalies and misdirects. To do so effectively will not be easy. It
will need global collaboration. For this Supplement, it will be enough to tease
a few key issues “into view.”

_________________________
41Alan H. Schoenfeld, “Research in Mathematics Education,” Review of Research in Mathematics
Education 40, issue 1 (2016): 497-528. Philip s. Jones, ed., A History of Mathematics Education in the
United States and Canada, Thirty-Second Yearbook. National Council of Teachers of Mathematics.
Reston, VA: National Council of Teachers of Mathematics, 1970 (second printing 2002).
42Dawn Leslie and Heather Mendick, eds., Debates in Mathematics Education. Abingdon, Oxon:
Routledge, 2014.
43The approach is not original but will be new for the science of mathematics pedagogy. An advanced
source is: Bernard Lonergan, Insight: A Study of Human Understanding, eds. Frederick E. Crowe and
Robert M. Doran, vol. 3 in The Collected Works of Bernard Lonergan. Toronto: University of Toronto
Press, 1992. Introductory level works are: (1) John Benton and Terrance Quinn, Journeyism, 2018,
https://bentonfuturology.com/journeyism/; (2) John Benton, Alessandra Drage and Philip
McShane, Introducing Critical Thinking. Halifax, Canada: Axial Publishing, 2005 (Reprint 2006).
(This book has been translated into Spanish (Madrid: Plaza y Valdés, 2011); and (3) Philip McShane,
Wealth of Self and Wealth of Nations: Self-Axis of the Great Ascent, Hicksville, NY: Exposition Press,
1975. (This is available for free at http://www.philipmcshane.org/wp-
content/themes/philip/online_publications/books/wealth.pdf.) Compared to the present
Supplement, references (1), (2) and (3) are far broader in their coverage. The Supplement focuses on
mathematical development and pushes into further examples.
44The (intransitive) verb ‘advert’ is convenient. We are using it in its first meaning in the Merrian-
Webster: “advert, intransitive verb. 1: to turn the mind or attention – used with to.”
45Occasionally, we find it convenient to use a single expression to refer to the whole inquiry: “self-
attention.” This is meant in the precise sense that will begin to emerge by doing the exercises in Part A.
The turn “to attend to one’s own understanding” can seem strange if one has been trained to focus
mainly on models.
46See references in note 42.
 PART A

Mathematical Understanding

A.1 A diagram
Much in the way an introductory level chemistry text may include a
simplified periodic table in the front cover, we begin by providing a
simplified version47 of a key diagram, a “table of elements of knowing.” The
diagram will be developed as we go.

Figure A.1 Dynamics of knowing

A.2 Some puzzles

A.2.1 A first puzzle

A first puzzle is given by the following:

An ellipsis “· · · ” means that the pattern is to be continued. An insight is
needed. Let’s concentrate on the second ellipsis, the “· · · ” that follows ‘44.’
What?-ing and direct insight
Before getting too far into solving the puzzle, let’s pause in a first effort to
notice something about what we are doing.
Are you wondering about “the array”? What is it?’, where the it’ is the array
that you have in sight.
If Yes, then you are “What?-ing”, you are in what we could call a “What is
it? inquiry-poise.” Why do I say “inquiry-poise” and not “inquiry”? Of
course, both are correct. It is for emphasis only that we are adjusting familiar
vocabulary somewhat. One wonders. In other words, ‘what?-ing’ is a
transitive verb, something that we do and what?-ing is a holistic poise.
Again, we are not suggesting that we necessarily utter words such as What is
it?’. In wondering about a seen sequence of symbols we are focused on
symbols. Symbols are in and of our senses, are in and of our sense-ability.
The temporary neologism, then, is to help draw your attention, your self-
attention, to the fact that inquiry is inquiry of a whole person and where, in
the present example, focus is on something in-and-of one’s sense-ability.
Writing down what we have just described, we get part of Figure A.1:

Figure A.2 Wonder about data of sense.

Perhaps you have already discovered a (possible) solution and can now
continue the sequence.
If so, something happened. There was a change “in” you. A name for that
“all-of-you-event” is insight.
Let’s call this a direct insight.
Why do we need the adjective ‘direct’? Partly, it is to distinguish this insight
from what turns out to be a second kind of insight that can emerge in a
follow-up inquiry-poise discussed below.48
For this sequence-puzzle, if you’ve had a direct insight, then you’ve gotten
hold of something, a possibility.
From that (act of) insight, automatically as it were — but with no implication
here of machinery or technology — there is a “procession” in you.
Why do we say “procession”? Focusing on an image in one’s sense-ability,
being “lit up” by insight “into” an image in one’s sense-ability, not only does
a light “go on” in us in that concrete image-focus, but light immediately also
“goes on in us.” There is in us, as it were, “light from light” in the sense that
following insight, a solution “emerges in us.” There is the famous story of
Archimedes, his calling out “Eureka!” As the Greek verb says, when we
discover the solution to the sequence puzzle, it is not just an insight.
Forthwith, and forth and with, there is also an “I’ve got it.” We are not
playing a grammar game here but are inviting your attention to an event.
There is an ‘it’ in “I’ve got it.” There is something that is not merely what we
had in our sense-ability. There is something that emerges “in us” from our
insight-into-image, something that in many respects is “transferable.” For
instance, following direct insight, we can turn our attention elsewhere but
then later return to the sequence-puzzle. The same solution then comes to
mind but without labor.
Why is it without further labor? Because we already get the point. We already
“got it.” It will be convenient to introduce a name. For historical reasons, let’s
call the “something that emerges in us,” the “it” of “I’ve got it,” “inner
formulation.”49
Note that the adjective “inner” is not to suggest that what we get hold of is
somehow “spatially inside” (versus a “spatially outside”). The adjective is
merely a metaphor. Inner formulation is a fact. It is what we have in as much
as we have, in fact, discovered a possible solution. Note that inner
formulation is to be distinguished from “formulation” in the sense of
providing a “formula.” Although, evidently, in some cases, a formula can
precede insight, and also be an expression of inner formulation.
An expression for what we have just described is given by the bottom two
rows of Figure A.1:
Figure A.3 Wonder about data50 of sense, insight and inner formulation

Is?-ing and reflective insight

Does the diagram in Figure A.3 express elements that emerge in our inquiry-
poise?
Asking that question reveals that in addition to wondering What is it?’ we
also sometimes ask a further question. That is, we also wonder Is it so?’,
where the it’ of Is it so?’ is, again, our possible solution, our inner
formulation that emerged in us from our What is it? inquiry-poise.
To be sure, our Is it so? inquiry-poise does not always emerge! Whether
because of haste or temperament or circumstances, sometimes we don’t get to
‘Is it so?’ But is it not (self-)51 evident that we do — at least sometimes —
rise to an Is it so?’ inquiry poise? Is it not so?
As we find in experience, then, in addition to a ‘What is it?’ inquiry-poise we
also sometimes engage in a further inquiry-poise. We have gotten hold of
something through direct insight. We might also reflect on our (potential)
resolution to our What is it? inquiry-poise. To be in keeping with tradition,
let’s call this further inquiry reflective inquiry.
Are direct and reflective inquiry so different?
At a dinner party, a colleague asked a mutual friend, “What would you like to
drink?” For fun, she answered, “Yes.” “Is that so?” She answered, “Tea.” It
was funny at the time.
The purpose of our story is to help bring out that it is (self-) evident that
wondering ‘What is it?’ and wondering ‘Is it so?’ are (radically52) different
questions.
In the sequence puzzle, de facto, the question What is that pattern?’ is not
resolved by answering Yes.’
Nor do we resolve an Is it so?’ question by providing a possible solution to
the puzzle.
In other words, it is not merely Yes or No or Probably Yes, or Probably No.
We move to an inner Yes (or inner No) regarding a possible solution already
obtained through direct insight.
In an Is it so?’ inquiry-poise, what is it that we are we looking for?
Continuing with the sequence puzzle, one might, for instance, re-examine the
pattern in one’s sense-ability, to confirm whether or not one’s solution holds
up, that is, whether or not one can indeed account for all terms of the
sequence. Or, perhaps you feel that you were distracted at the time and don’t
trust your solution. You might then give the whole sequence-puzzle a fresh
look. And so on.
And so on? Notice that our Is it so?’ inquiry-poise is open ended. You have
discovered a possible solution. You have that solution in mind. But ‘Is it so?’
All that you have done so far is up for grabs. Does your solution hold up
under scrutiny? Were you hasty and do you need to start over? Eventually, if
one is to not simply guess, something more is needed. What is that “more”
that is needed? At what point are we “good to go” (or not) with our possible
solution?
In an Is it so?’ inquiry-poise, we may have another insight. We may reach an
understanding to the effect that: Yes, (probably) I have enough to go
on. “We reach an understanding that (probably) we have sufficient grounds to
assent to our solution.” (Note that in this context the word probably’ is an
adjective, not a term from empirical probability theory. Why introduce the
adverb? Again, reflection is open ended. We can always ask more questions.)
This can be as spontaneous and rapid as hearing sounds and “immediately”
knowing that children are playing soccer in a field across the road. But in
mathematics, modern rigor asks us to, as much as possible, spell out what
precisely we understand and to identify grounds for what we understand.53
In the case of the sequence puzzle the pattern is what we might call a “pattern
in symbolism as symbolism.”54 In attempting to resolve our Is it so?’ inquiry,
we again can advert to sense-ability. Now, though, it is not adverting to an
image in a What is it?’ inquiry-poise. Our new lean is to discern whether or
not our already discovered solution accounts for all terms of the puzzle. It is a
personal achievement: one returns to one’s puzzle to sort out whether or not
one can now, in fact, account for all terms of the puzzle.
If you are starting to become somewhat familiar with the more complete
focus (that is, of not only doing mathematics but pausing to discern what you
are doing when you are doing mathematics), then you may have already
noticed that following reflective insight there is a second procession. The
second procession is to an inner event called “judgement,” an inner “Yes, it is
so” (or “No, it is not so,” or Probably Yes, or Probably No, where, again, in
this context, probably’ is an adverb for the quality of one’s judgment). As in
procession that follows direct insight, following reflective insight, procession
is “automatic” as it were—it occurs. Following reflective insight, procession
is to a distinct (but not separable) inner event. In some cultures, occurrence of
assent (or its opposite) is obvious because it is spontaneously revealed by a
familiar movement of the chin. However, becoming acquainted with people
from different parts of the world soon reveals that what that movement is
varies among cultures.
Until now, we have held back the solution to the sequence puzzle A.2.1. We
did not want to deprive you of the pleasure of discovery.
The sequence 1, 2, 3, · · · is partitioned. Some of the numbers are above a
line and some are below. Why? Perhaps surprisingly, a clue is that problem
can be solved by school children. In case you haven’t yet solved the problem,
we provide the answer below. But in case you wish to keep working on it, we
put the solution in a footnote to make it easier to not peek.55
Yes, children can solve the problem. But we are inviting you to do something
more, something that children are not asked to do (at least not yet, at this time
in history). We are inviting you to make beginnings in discerning aspects of
what you are doing, what you have done, what you have achieved. You might
notice, for instance, that in this case the key insight is “understanding
differences in symbols as symbols,” that is, getting hold of a pattern in one’s
sense-ability. If we were to write the (“natural”) numbers in Roman
Numerals, base 2 or (like the Babylonians) base 60, this puzzle would be
destroyed.
There are, however, sequence puzzles that call for other kinds of insight.

A.2.2 A Famous Sequence

There is the famous sequence, 1, 1, 2, 3, 5, 8, · · · .56
If you have already gotten hold of the pattern, notice that, by contrast with
the puzzle given in Section A.2.1, what is key here is not understanding what
symbols happen to look like, “as symbols.”57 And so, for instance, essentially
the same puzzle can be expressed in Roman numerals, base 2 and base 60.
In Roman Numerals, the puzzle can be written as: I, I, II, III, V, VIII, XIII, ....
In a base 2, the puzzle can be written as: 1, 1, 10, 11, 101, 1000, 1101, · · · .
If you have already studied some of the early history of mathematics, you
will know that the sequence could also be expressed in cuneiform, using base
60.
An understanding of the sequence can also be expressed implicitly with
visibly different formulas, three examples of which are:

A.3 Descriptive and explanatory understanding, and

judgment in mathematics
Descriptive and explanatory understanding

In example (A.1), our understanding is descriptive. For the sequence-puzzle

A.2.1, descriptive refers to the fact that what we grasp is “a pattern in our
sense-ability.”
By contrast, in example (A.2), our understanding is explanatory. Rather than
“a pattern in sense-ability,” the key insight needed is to discover a pattern of
mutually defined terms, and operations.59
As is evident from these two examples, both descriptive and explanatory
understanding are part of and contribute to our mathematical development.
For another example, you may recall a familiar sequence from calculus that is
often called the ‘power rule’:
y = x, y′ = 1; y = x2, y′ = 2x; y = x3, y′ = 3x2; and so on.
The symbolic pattern can be grasped without needing to understand anything
at all about differentiation. In that case, what is grasped is a “symbolic pattern
as a symbolic pattern,” “a pattern in our sense-ability.” But the mathematical
meaning of these symbols is reached through explanatory understanding.
Evidently, the observation applies generally. We can understand how to use
symbolism in linear algebra; differential equations; and on into the most
remote realms of contemporary mathematics. In each case, descriptive
understanding is an achievement and by it we understand, for instance, how
to use symbols and diagrams. However, there is also the possibility of
understanding what symbols and diagrams mean which, in mathematics, is
explanatory understanding.60

Judgment
Note that in (A.1) and (A.2) we find the same core pattern expressed by
Figure A.1. We begin with inquiry about an “image” in sense-ability, and “a
focus inquiry.” In both cases, a What is it?’ inquiry-poise is resolved through
direct insight, following from which there is a procession “in us” to “inner
formulation.”
At the same time, there are significant differences in what is being
understood. As you might expect, grounds for judgment also turn out to be
different.
In example (A.1), direct insight is of a “pattern grasped in our sense-ability.”
Is our solution correct? Is it so? By adverting again to patterns in our sense-
ability, we find and discover sufficient grounds (or not) for assenting to our
possible solution. There is then a procession to “Yes, it is so,” or perhaps,
“No it is not so.”
In example (A.2), however, what we discover is a pattern of mutually defined
terms, and operations that, as it happens, can be both presented and expressed
in many ways.61 In that case, sufficient grounds for assenting to our possible
solution are not “patterns in sense-ability” but whether or not one can,
indeed, successfully continue the sequence by appealing to our directly
grasped (possible) pattern of mutually defined terms, and operations.
Figure A.4 Reflective inquiry and reflective insight in elementary sequence puzzle A.2.1

A.4 Descriptive definition and explanatory definition

Let’s revisit the ‘power rule’ sequence: “y = x, y′ = 1; y = x2, y′ = 2x; y = x3, y′
= 3x2; and so on.”
In Section A.3, we invited you to observe that in mathematics we enjoy at
least two main “genera” of direct insight namely, descriptive and explanatory.
Through both, we reach a possible resolution of a What is it?’ inquiry-poise.
But for the power rule, there is only sequence. How do these two resolutions
compare? What are we getting at with two answers regarding one sequence?
All along, here, it is a matter of describing what we are doing, of “catching
oneself in the act of doing mathematics.” From descriptive insight, one
understands how to continue the sequence of symbols y = x, y′ = 1; y = x2, y′
= 2x; y = x3, y′ = 3x2. But what do the symbols mean? As in Section A.3, a
different understanding is possible. For the power rule sequence, one may
also reach a basic understanding of “derivative.”62
By descriptive insight, then, we understand how to use symbols and
diagrams. This makes it possible to “define” symbols and diagrams in the
same that way words are “defined” in a typical dictionary. In other words,
thanks to descriptive understanding, we can “define” symbols and other
mathematical words and diagrams by indicating how to use them. Naturally
enough, we call this nominal definition.
By explanatory insight, however, we go on to say what the mathematical
symbols, words and diagrams mean. So, there is also explanatory definition.
We end this section with one more example. To keep the Supplement short,
we leave each sentence as an exercise: Defining an ellipse as a “perfectly
symmetrical oval” is a nominal definition. Explanatory definition can be
reached from the clue that an ellipse is “a circle with two centers.” (It will
help to make a diagram for drawing an ellipse. This can include an imagined
string, two imagined pins and one imagined pencil or pen.) The key insight
needed can lead one to (explanatorily) define an ellipse to be the locus of all
coplanar points satisfying the equation F1P + F2P = constant. If one has yet
not had an insight into “one’s image of a two-center circle,” then one’s
“definition” F1P+F2P = constant is nominal. Note, however, that in that case
nominal definition is at some remove from an imagined “perfectly
symmetrical oval” named ellipse.’ For, in that case, nominal understanding is
not of how to use ellipse’ as a name for an imagined shape but of how to use
the name ellipse’ in conjunction with a (combination of) symbols ‘F1P + F2P
= constant.’

A.5 Proofs
My beginnings in calculus, by Terrance Quinn
In September of 1980, in the first weeks of my first-year calculus course at
the University of Toronto, the lecturer63 wrote a theorem on the blackboard
about continuity of a function y = f(x). He also provided a proof that went
something like this: “Suppose that ϵ > 0 and let . · · · ” He went on to fill
two boards with inequalities and ended with words similar to: “· · · from
which we can conclude that |f(x) – f(y)| < ϵ. Since ϵ > 0 was arbitrary, the
result follows. ”64
I trusted our excellent teacher and had no doubt that “the result did follow.”
But during that lecture I also knew that I didn’t yet see why. That evening, I
started in on trying to unpack the proof. I found some of the inequalities
within the proof (“steps” in the proof) to be a bit tricky. Before too long,
however, I could follow the proof “step by step,” inequality by inequality. I
could see that, yes, each step of the proof worked. But I remember that it was
also obvious to me that I still didn’t really have it. What I mean is, I was
aware that even though I could check all of the steps, I didn’t see how I might
have produced a similar proof myself. Up to that point, in my mathematical
education, I had been naively thinking that I already had a good
understanding of calculus. In some respects, I did. However, the first weeks
of that first semester of calculus were a shock to me and — as I learned —
for some (not all) of my classmates. Eventually, along with other “survivors,”
I started to find my way. For me, the “epsilon-delta business” crystallized in a
problem that involved a damped sine function defined over the real line. In
hindsight, I see now that my insight in that case was similar to Archimedes’
insight regarding the remainder term of a series. Problem by problem and
insight by insight, I crawled my way into modern axiomatics for calculus,
a.k.a. “elementary real analysis.”65
In that first-year class, a common type of problem was of the form: “Prove
(or disprove) X (where X was some statement).” How to begin? I soon
learned that, for me, unless it was “asily done”(in other words, unless I
already understood), it was often best to not start with the general statement.
It usually worked better for me — that is, I worked better — by starting with
examples. Of course, I kept the X in mind and oscillated back and forth
between examples and thinking about the X statement. But inevitably, the
break for me would come while working on an example. That started to take
some the mystery out of “proof writing.” In other words, “being able to write
a proof is not separate from understanding.”
I noticed that the order in which the write-up proceeded usually turned out to
be, to some extent, in a kind of reverse order to what I might call “the order
of inquiry and discovery.” I now realize that was no accident. I have been
recounting some of my early struggles in a challenging and exciting first-year
calculus course. But to begin homing in on key and core issues, it is better to
now shift to a more elementary example, essential features of which are
already described by another author.
Understanding and syllogism

In a circle of, say, unit radius, two diameters, perpendicular to

each other, are drawn. [You will need to make a diagram.] From an
arbitrary point P on the circumference two perpendiculars PR and PS
are drawn to the two diameters. The problem is, What is the ratio of
RS to the radius? . . . Joining R and S will be an evident thing to do;
but it may take a pedagogue to adequately dispose the phantasm by
the drawing of another line. The line to draw is the line joining the
 center to the point P, say OP. Eureka! With the insight there emerges
the solution, the relation between RS and the radius.
Note now that the solution can be formulated or thrown into
syllogistic form, and this will help you get some light on features of
the syllogism that are often misrepresented. We have, therefore, the
syllogism:

In this light we may note important characteristics of the procedure. We

started not with two premises, but with the conclusion in the form

Our search, through diagram, was for a middle term, and the middle term was
supplied as soon as one adverted to the significance of OP. Only then can the
syllogism be constructed. To coin an expression for this constructing, one
might say that the insight is crystallized into a syllogism. This does not mean
of, of course, that somehow insight has been pinned down on a page. What
has happened is that we have given the insight symbolic expression. Giving
all the relevant insights symbolic expression is by no means always an easy
thing to do, even when it can be done. Modern geometers have found fault
with Euclid in this matter. There are insights in the Elements which are not
explicitly acknowledged either in axioms or in the theorems, yet which were
not uncrystallizable.66
Let me note further that our simple puzzle and solution is a paradigm of how
Euclid and company may well have proceeded. They did not proceed step by
step down the page of a modern textbook, from the stated theorem to the fully
constructed diagram, to the step by step deduction beneath.67
We may return now to a brief consideration of the simple symbolic
expression of understanding which is the (mathematical) syllogism.
[In mathematics, the syllogism is a] help toward understanding.
The syllogism is not some mysterious replacement for understanding.
One may look at the syllogism as a proof of the conclusion, but this can only
mean that the structure facilitates a grasp of the implication of the conclusion
in the premises.
One might note, too, that such structures facilitate the checking, the Is-
question, relating to that grasp.68

A.6 Algebra from arithmetic, “and so on”: that is, sequences

of “higher viewpoints”
When were children, we learned to count.
“Five baskets of apples and 7 baskets of apples is 12 baskets of apples.”
Before long, however, through insight, we learned to ‘abstract.’69
For instance, in this example, one might go on to understand that whatever
we happen to be counting, we can count “how many”:
7 · 3’s + 5 · 3’s = (7 + 5) · 3’s = 12 · 3’s; and
5 · 9’s + 8 · 9’s = (5 + 8) · 9’s = 13 · 9’s.
Abstracting further,
7 · 3 + 5 · 3 = (7 + 5) · 3 = 12 · 3; and
5 · 9 + 8 · 9 = (5 + 8) · 9 = 13 · 9;
and so on.
And so on?
By insight we are beginning to transition “beyond” arithmetic.
One can solve many arithmetic problems without breaking into, or up into,
algebra. That is both historical and biographical. In moving into algebra,
however, we do not lose arithmetic. We still understand, for instance, that 5 +
7 = 12. But now we can also write and understand that 5z + 7z = (5 + 7)z =
12z, whatever the number z happens to be.
By insight, one understands and holds together an indefinite range of
instances in arithmetic.
Moving further into the new context:
5 · 3 + 7 · 3 = (5 + 7) · 3
7 · 3 + 5 · 3 = (7 + 5) · 3 = 12 · 3
5 · 9 + 8 · 9 = (5 + 8) · 9 = 13 · 9
···
?
(direct insight)
from which one can write: x · z + y · z = (x + y) · z.
As discussed in Section A.3, that insight may be nominal, as when a student
grasps “a pattern in symbolism as symbolism.” But one might also
understand that x, y and z represent any possible numbers, and that the
expression x · z + y · z = (x + y) · z represents a pattern of operations. In that
case, we not only know how to use the symbols x, y and z, we also have
gotten hold of a correlation of mutually defined terms, and operations, by
which one can explain what the expression means. Moreover, with that
understanding, we can now just as easily write a · c + b · c = (a + b) · c, U · V
+ W · V = (U + W) · V, (U times V) plus (W times V) = (U plus W) times V,
and so on.
Transitioning from arithmetic to algebra is sometimes called transitioning to a
“higher viewpoint.” In the case of arithmetic and algebra, both “viewpoints”
are systems. The word higher’ is a spatial metaphor, but it serves nicely. For,
as revealed in instances, by understanding in algebra one has control over
indefinite ranges of arithmetic computations.
A key point: Self-notice that images by which we reach the higher
viewpoint of algebra are diagrams and symbols from lower viewpoints
such as arithmetic and elementary geometry.
We are not asserting this key point as a conclusion from a “theoretical
framework” or model for mathematical development. It is an observation
reached by adverting to instances to one’s own experience.70
In mathematics, does transitioning from lower to higher viewpoints happen
often?
In order to make beginnings toward being able to answer this Is-question one
would need to have (considerable) experience in mathematics. One would
also need to attend to what one has been doing. The question calls for and
depends on growth in mathematics as well as growth in the kind of
“balanced” attention being invited in this Supplement. In other words, there is
the further challenge of adverting to, noticing and discerning as precisely as
possible what one is doing when one is doing mathematics.71

A.7 Correlations, concepts and other fruit of understanding

Correlations
In coordinate (Cartesian) geometry, the “parent parabola” is defined by the
equation y = x2. Other parabolas are obtained by adjusting the scale of one or
both of the axes, by translation, and by rotation.
We start with this definition because it will be familiar to many teachers. It is
what is often given in high school and first year undergraduate books.
The present challenge is to make a beginning in “unpacking” (what turn out
to be “layerings” of) insights that often are neither expressed nor adverted to.
Let’s begin with our focus on y = x2, an algebraic formula. In that context,
what are the x and the y? By using a pair of symbols, we refer to an indefinite
range of pairs of numbers, some of which are (0, 0), (-1, 1), (1, 1), (-2, 4), (2,
4), (-3, 9), (3, 9), and so on. In each pair, what is x? It represents every
number that squares to give a corresponding number y. What is y? It
represents every number that is the square of some x. The x and the y are
mutually defined by a correlation that can be expressed by y = x2’. Note that
in this context, terms, operations and equality are algebraic. But in that
correlating, we also understand and correlate an indefinite range of pairs of
numbers where in each case, terms, operations and equality are arithmetic.
That’s a start. However, we haven’t finished “unpacking.” Why not?
Coordinate geometry is not merely algebra. In coordinate geometry, the x and
the y are coordinates. In other words, the x and y refer to lengths along axes
that, by hypothesis, have scales. So, let’s look more closely to our underlying
(or perhaps latent) understanding of an x-axis.
One can imagine a distinguished straight line in the Euclidean plane72 and
call it the ‘x-axis.’ In coordinate geometry, we also have a scale along that
distinguished line (and, indeed, along all lines in the same Euclidean plane).
And so we can write x = 0, x = 1, x = 2, x = 22/7, x = e, x = π, and so on. In
coordinate geometry, these numbers refer to lengths along the distinguished
axis.
This may seem obvious enough. But is it?
Let’s take one of these, x = 2, say. In this context, to what do we refer with
the number 2?
Here, x = 2 is a “distance” along the x-axis. How is that distance determined?
You might remember that the task here is not to invent something new but to
discern elements in what we already do. Distance is given in terms of a unit
distance along the x-axis.
OK, but what is a unit distance ?
In physical measurement, a unit distance is whatever happens to be
convenient. This might be the length of one’s thumb (an inch, or in French,
‘un pouce,’ which means thumb’), a handy length of wood, the distance from
the tip of one’s middle finger to one’s elbow (cubit), a “standardized cubit’
such as the length of a piece copper used by the Sumerians in the third
millennium B.C.E., and so on. In modern times, a ‘meter’ is the length of a
bar of platinum held in Paris, France.
In mathematics, distance is not a physical length. Still, there is nothing to
stop us from imagining and thinking about a unit length,

Physical lengths can and do change. Two people with different arm lengths
will yield different cubits. A piece of wood may dry out and its length may
then change. Or again, suppose we have two metal rulers with “cms” etched
along their sides. In other words, “1 cm” is the unit length. If one of the rulers
is used at different locations on a hot wood stove then “1 cm” at one location
can be longer than “1 cm” at another, when compared with the second ruler
that is kept from direct contact with the hot stove. But if there is no fire in the
stove, and if no finer measurements are needed then, for practical purposes,
no problems arise. That is, the unit length is “invariant.” hrough the ages,
builders, carpenters, architects and engineers have relied on such invariance
when designing and building structures — from wooden shacks to pharaohs’
pyramids, from modern homes to city skyscrapers.
In classical geometry, we “ake this idea and run with it.” We suppose that an
imagined unit length is invariant in the sense that an imagined unit length at
one location is the same as an imagined unit length at another.
Let’s now return our focus to the x-axis. Let Ox be a (distinguished) point on
the x-axis. The point Ox is called “origin.” Let X be another point on the same
axis, a point that is, say, two units’ “to the right” of Ox. Notice our ongoing
reliance on imagination, e.g., “to the right,” and “to the left.”
We have an imagined x-axis which is an imagined distinguished line in the
Euclidean plane. By insight, we suppose that the line extends indefinitely in
both directions:

We imagine two points on this imagined x-axis, Ox and X:

In the imagined length between these two imagined locations, one can
imagine fitting copies of the imagined unit length, “ _____ ” and “ _____ ”,
distinct line segments that, by hypothesis, are otherwise identical. In
particular, we suppose that they have the same length.
A handy diagram for this is:

So, in our understanding, we correlate an imagined line segment with copies

of an imagined unit length:

But we also correlate an imagined “ _______ ” with the number ‘2’,

a diagram for which is:

In coordinate geometry, “x = 2” means that we take both of these correlations

“together,” a diagram for which is:
In other words, “together” means that we need a further insight by which we
obtain, grasp, or reach a further correlation. That is, we correlate two
correlations.
What we find then is that, in coordinate geometry, asserting that x = 2 is an
expression of an insight by which one holds together a “layering” of insights:
one correlates two correlations. Similarly, x = 3 is an expression for
correlating two correlations; and so on for all values of x; and similarly for all
values of y.
We started with the expression y = x2. If we only focus on numbers, then our
understanding is an algebraic correlation. In coordinate geometry, however,
the x’s and y’s are lengths along two distinguished (and perpendicular) lines
in the Euclidean plane, a plane in which there is, by hypothesis, an invariant
unit length. Unpacking key insights, the x-values represent an aggregate of
correlations of correlations; as do the y-values. In coordinate geometry, then,
y = x2 is an expression of a correlation of correlations of correlations.

Concepts
In Euclidean geometry, we speak of ‘points,’ ‘lines,’ ‘line segments,’
‘length,’ invariance,’ and so on, as well as the entire Euclidean plane.’
What are these?
What, for instance, is a point?
We can imagine a small dot on paper (or, say, on a computer screen). An
imagined dot has breadth and so we can imagine two smaller dots within an
imagined dot. Still, “the point” is “clear,” is it not? That is, attempting to “pin
down” a location unambiguously, we need to keep making an imagined dot
smaller. So long as a dot has any breadth or depth, further dots can be
imagined within the imagined dot.
By insight, then, we define a point on the plane to be “a location that has no
breadth and no depth.”
Or, in translation from Book I of Euclid’s Elements: “A point is that which
has no part.”
A point then, (self-) evidently is the fruit of insight! It is also a concept.73
In a similar way, that is, by adverting to experience, one can find that lines,
planes, invariant length, and so on, are also the fruit of insight and also are
concepts.
Continuing in this way, what becomes (self-) evident is that while it is
true that images and patterns in our sense-ability on which we focus
inquiry can be said to be “primitive,” points, lines and other concepts are
the fruit of understanding and, in particular, are neither reducible to
imagination nor are they primitive elements in mathematics.

A.8 The historical context for teachers (and scholars)

How do we go on from elementary and preliminary exercises? It is for each
teacher and scholar in mathematics to work out how far you need to go. “Just
as in any subject, one masters the essentials by varying the inciden-tals.”74 In
so far as one has the time and the proclivity, one can identify key diagrams
and symbolic expressions, helpfully directed questions, key insights, skills,
and sequences of such in one’s own mathematical development.
You might observe that, unless one is breaking new ground, when
understanding mathematics one is also sharing in key insights from the
historical development of the field. Note that this is not a statement regarding
“the historical method” for teaching. This is an observation about learning
mathematics, for instance, algebra, or calculus. How else is one to know and
teach “completing the square” that was discovered in various times and
places in antiquity, or “the calculus” that was discovered by Newton and
Leibniz, and know these in ways that allow one to teach them effectively,
unless one has not (at least partially) identified key questions and insights? At
the same time, (self-) evidently, it is always our own inquiry and our own
insights that we experience. Our own development is our main source of
data75 on mathematical development.
It was Bernard Lonergan who first brought attention to the need and
possibility of the balanced method to which we are inviting mathematics
teachers. And so we end Part A of this Supplement with a quotation from
Lonergan on historical understanding [of mathematical development].
The original quotation is from a discussion of historical understanding of a
discipline.
 The history of [mathematics] is in fact the history of its
development. But this development, which would be the theme of a
history [of mathematics], is not something simple and straightforward
but something that has occurred in a long series of various stages,
errors, detours, and corrections. To the extent that the one studying
this movement learns about this developmental process, one already
possesses within oneself an instance of that development which took
place perhaps over several centuries. This can happen only if one
understands both the subject and the way in which he or she learned
about it. Only then will one understand which elements in the
historical developmental process had to be understood before others,
which were the causes of progress in understanding and which held it
back, which elements really belong to that particular science and
which do not, and which elements contained errors. Only then will
one be able to tell at what point in the history of the subject there
emerged new visions of the whole and the first true system occurred,
and when transitions took place from an earlier to a later systematic
ordering; which systematization was simply an expansion of the
former and which was radically new; what progressive
transformations the whole subject matter underwent; how everything
that was previously explained by the old systematization is now also
explained by the new one, as well as many other things that the old
one did not explain [as in the discoveries in mathematics, for
example, by Euclid, Archimedes, Apollonius, Pappus, Al-Khwarizmi,
Descartes, Fermat, Newton, Leibniz, Cauchy, Fourier, Riemann,
Galois, the great analysts, and others]. Only then will one be able to
understand what factors favored progress, what hindered it, why, and
so forth.
Clearly, therefore, [master teachers of mathematics] have to have
a thorough knowledge and understanding of the whole subject. And it
is not enough that they understand it in any way at all, but they must
have a systematic understanding of it. For the precept, when applied
to [the history of mathematics], means that they must understand
successive systems that have progressively developed over time.
This systematic understanding of a development ought to make
use of an analogy with the development that takes place in the
mind of [a teacher] who is learning the subject, and this interior
 development within the mind of the [teacher] ought to parallel the
historical process by which [mathematics] itself developed.76

_________________________
47A more nuanced version is available in Bernard J. F. Lonergan, “Appendix A, Two Diagrams,” in
Phenomenology and Logic: The Boston College Lectures on Mathematical Logic and Existentialism,
eds. Frederick E. Crowe, Robert M. Doran, vol. 18 in The Collected Works of Bernard Lonergan,
Toronto: University of Toronto Press, 2001, 319–323. A diagram for the “Dynamics of Doing” is the
second diagram in Lonergan, “Appendix A, Two Diagrams.” Evidently, our “dynamics of doing” are
similar to our dynamics of knowing. For decision, however, inquiry is in a different mode. Our
dynamics of doing subsumes our dynamics of knowing. The interweaving of the two modes of inquiry
are brought out in student exercises in John Benton and Terrance Quinn, “The Dynamics of Doing,”
Journeyism 16 https://bentonfuturology.com/journeyism16/ and Jour-neyism 17,
https://bentonfuturology.com/journeyism17/. The dynamics of doing are not an immediate focus of
this Supplement. However, it is worth noting that progress in adverting to and describing dynamics of
doing also is needed. Among other things, such progress will help resolve numerous contemporary
issues in mathematics education where, so far, dynamics of the two modes are not adverted to and
consequently, not yet adequately distinguished. For instance, it is sometimes said that a teacher’s job is
to help a student “decide” which concept or solution to accept. The plausibility of such notions emerges
from the mistaken model that mathematical understanding is a matter of connecting concepts. The fact
that that view is mistaken is revealed — becomes (self-) evident — by doing the exercises given
throughout Part A. For context in mathematics education, see Part C.
48See below: Is?-ing and reflective insight.
49Lonergan, Phenomenology and Logic, Dynamics of Knowing, “Appendix A,” 322. See also
McShane, Wealth of Self, 15.
50See, e.g., Benton and Quinn, “The Dynamics of Knowing: The First Four Boxes,”
https://bentonfuturology.com/journeyism10/; and McShane, Benton, Drage and Mc-Shane, ch.
16, “What-Questions,” 67–69.
51See note 45.
52I am using the first meaning of the word ‘radical,’ referring to ‘root.’
53That is not easy. See, e.g., note 69.
54If you haven’t already solved the puzzle, that is a clue. See also the personal anecdote at the end of
this section.
55Answer to sequence puzzle A.2.1: Symbols that are written only with straight edges go above the
line, while symbols that include curved edges go below the line.
56The Fibonacci sequence is a sequence of numbers of fertile pairs of male and female rabbits. Start
with the hypotheses given by Fibonacci (about fertile male and female pairs, gestation periods and the
numbers of progeny). The problem is within reach of the contemporary senior high school or
undergraduate student. It is interesting and well worth doing. Discovering the solution for oneself one
takes a step in a climb toward the modern theory of recursive sequences.
57Long before Fibonacci, the same sequence is found in ancient Sanskrit texts, written in a base 60
number system. See, e.g., Tia Ghose, “What is the Fibonacci Sequence?” LiveScience (October 24,
2018): https://www.livescience.com/37470-fibonacci-sequence.html; and Keith Devlin, Finding
Fibonacci: The Quest to Rediscover the Forgotten Mathematical Genius Who Changed the World.
Princeton, NJ: Princeton University Press, 2017.
58From a “higher viewpoint,” we could also identify the solution as follows: “The sequence in (A.2) is
the (unique) solution to a second-order homogeneous recurrence relation with constant coefficients
whose (non-reduced) characteristic equation is t2 – t2 – 1 = 0, and whose first two initial-values are 1
and 1.” See Section A.6.
59More generally, descriptive understanding is “grasping patterns in experience, as experience.” And
so to discern events and orderings of events in our dynamics of knowing is a beginning but it, too, is
descriptive. Progress in explanatory understanding of our dynamics of knowing will be future growth
for the academic community. Aspects of that future progress are already partly in evidence. And so,
eventually, it will include an “integration” of human biophysics, biochemistry, human zoology and
psychology, cogni-tional theory and more. To glimpse something of the challenge, precise densely
expressed heuristics can be found in Lonergan, Insight, 489 (add the word ‘self’ to the paragraph that
begins “[(Self-) S]tudy of an organism begins ...”).
60For a discussion in elementary contexts, see Terrance J. Quinn, “On Two Types of Learning (in
mathematics) and Implications for Teaching,” On Learning Problems in Mathematics, Research
Council on Mathematics Learning, Mathematical Association of America, FOCUS, Fall 2004, 31–43.
61See, e.g., note 60.
62In calculus, there are two key basic insights, basic in the sense of prior to axiomatics. A satisfactory
definition of limit did not emerge for several decades after the initial discoveries of calculus. But both
Newton and Leibniz each had those two key insights (and, of course, much more). Once shared with
the scientific communities, those two key insights soon made possible the solution of problems from
antiquity and also opened up vast ranges of new lines of inquiry and development in mathematical
sciences and engineering. See Terrance J. Quinn, “Getting Started in Calculus,” Problems, Resources
and Issues in Mathematics Undergraduate Studies (PRIMUS), vol. 13, issue 1 (March 2003): 55–74.
63I am referring to Prof. Edward Bierstone who, as I understand it, is still at the University of Toronto.
See https://www.math.toronto.edu/bierston/. Accessed July 17, 2019.
64The textbook in the 1980–81 academic year was the edition then available of Michael Spivak,
Calculus. Cambridge: Cambridge University Press.
65Historically, axiomatics for calculus were obtained after several decades of collaboration following
the initial breakthroughs of Newton and Leibniz.
66It concerns drawing a line across the interior of a triangle. One identifies the point where the line
crosses the opposite side. Why is there a point of intersection? Felix Klein discusses the problem: “Of
course, no one would doubt that, intuitively, but in the framework of axiomatic deduction we need a
special axiom, the so-called betweenness axiom for the plane. This axiom states that if a line enters a
triangle through a side, it must leave it through the other side — a trivial fact of our space perception
which requires emphasis as such, because it is independent of the other axioms. . . . If we omit this, as
Euclid does, we cannot reach the ideal of a pure logical control of geometry. We must continually refer
to the figure” (Felix Klein, Elementary Mathematics from an Advanced Standpoint: Geometry
(Mineola, NY: Dover Publications, 2004), 88. This is an unabridged republication of an earlier Dover
reprint (1949) of the translation first published by The Macmillan Company, New York, in 1939). As is
well known, there are non-Euclidean geometries for which the axiom does not hold. For further
discussion, see McShane, Wealth of Self, 68–9.
67“One should note, too, the significance of this for the teaching of geometry. Too often pupils [are
asked to] begin, not with the thrill of a puzzle but with the top of the page, and at most get a vague line-
by-line comprehension of the theorem. Memory is burdened, and examinations consists in filling out
theorems - from the bottom and top of the page! — and passing over the riders” (McShane, Wealth of
Self, 69). This provides foreshadowing to Part B.
68Philip McShane, Wealth of Self, 67-9.
69Here, ‘abstract’ is a transitive verb for what we do in thought: “abstracter/,” to draw away, remove,
derivative of abstract (or borrowed directly from Latin abstractus)” https://www.merriam-
webster.com/dictionary/abstract.
70As you might already be seeing, this has implications for teaching. See Part B.
71The task is further described in Section A.8.
72What is a straight line in Euclidean geometry? (Here, we avoid the extensive literature regarding
“primitive concepts,” “primitive,” and “vicious circles in logic.” Such problems need to be handled, but
not here.) In Book I of the Elements, Euclid provides what evidently are nominal definitions: “Def. 1.2.
A line is a breadthless length.” “Def. 1.4. A straight line lies equally with respect to the points on
itself.” Explanatory definition is obtained when a straight angle is defined to be the sum of two 90-
degree angles. In that way, we go beyond merely describing what we imagine. In that case, a straight
angle is defined by correlating terms in the system.
73“concept: noun, (Entry 1 of 2) 1: something conceived in the mind: thought, notion, 2: an abstract or
generic idea generalized from particular instances” https://www.merriam-
webster.com/dictionary/concept.
74Lonergan, Insight, 56.
75See note 50.
76Bernard Lonergan, Early Works on Theological Method 2, vol. 23 in the Collected Works of Bernard
Lonergan, translated by Michael G. Shields, edited by Robert M. Doran and H. Daniel Monsour
(Toronto: University of Toronto Press, 2013): 175–177. For a somewhat dated but still excellent source
on mathematical development, see, Eric Temple Bell, The Development of Mathematics. New York:
McGraw-Hill, 1940. Second Edition: New York, McGraw-Hill, 1945. Reprint: Dover Publications,
1992. University of Toronto Press, 2013): 175–177. For a somewhat dated but still excellent source on
mathematical development, see, Eric Temple Bell, The Development of Mathematics. (New York:
McGraw-Hill, 1940. Second Edition: New York, McGraw-Hill, 1945. Reprint: Dover Publications,
1992).
 PART B

A Few Implications for Teaching

It is for brevity that we merely state a few results. To bring these out
pedagogically would need a few chapters. However, we hope that providing
these few results will serve as a guide and invitation to further reflection. If
you have worked through the exercises in Part A (or similar ones), you may
get to some (or all) of these yourself. As pointed to in Section A.8, further
details and implications will be discovered through one’s ongoing growth in
mathematics and in self-attention in mathematics.77 We do not comment on
curricula of any departments or colleges involved in the ongoing challenging
and creative work of meeting students’ needs and program needs in diverse
circumstances.
Part of the invitation is for a teacher to make progress in identifying and
distinguishing nominal and explanatory understandings, in one’s own
understanding.
As experience reveals, nominal understanding includes understanding by
which one reaches competence with symbolic techniques and computations.
The task of pedagogy invites teachers to know as much as possible about
where, in precise terms, the content of a course fits in the historical
development of the field. See Section A.8.
By the same token, in order to be able to help a student make progress in the
field, it will help if a teacher learns as much as possible about where a
student’s understanding is in relation to the historical development of the
field.
As expressed by Figure A.1, by adverting to one’s experience in doing
mathematics, it becomes evident that understanding emerges through inquiry
“into” images in one’s sense-ability. Consequently, in a given mathematical
context, it will help if a teacher is familiar with diagrams and symbolisms that
help raise and direct student inquiry toward reaching key insights appropriate
to a context.
To help a student reach a higher viewpoint, it will help if a teacher can
provide diagrams and symbolisms from lower viewpoints by which a student
(or group of students) can reach key insights needed to begin breaking into
the higher viewpoint. See Section A.6.
For a century and more, there has been ongoing discussion about whether or
not teachers should “use history” to teach mathematics. An advantage of the
“historical method” is that it has the potential for helping students raise
questions and reach key insights in the field. But of course not all of the
searchings from history can be included. A selection needs to be made and
tailored to an audience and to a curriculum. In that sense, a teacher needs to
“make history better than it was.” More precisely, a teacher can make
progress in identifying specific developmental sequences.78 In that way, for
instance, in two leisurely 45 minute classroom sessions one can help senior
high school mathematics students and freshmen level general calculus
students reach the two key insights had by both Newton and Leibniz from
which the entire body of calculus is developed.79
Pedagogy called engaged learning has been a fruitful idea. The importance
and relevance of modern psychology and neuroscience (so, e.g., behaviour,
environmental considerations, learning style, group work, and so on) to
pedagogy cannot be denied. But might we not also ask, engaged in what? We
are back at the need and possibility of identifying helpfully directed diagrams
and symbols by which to subtly guide student inquiry toward specific and
key mathematical insights.
Only a small percentage of students go on to become mathematics majors
who, in addition to needing basic80 key insights, also go on to (sequences of)
axiomatic systems. But students in applied mathematics, engineering,
applications and general service courses also need to be helped so that
inquiry be directed toward the emergence of initial or basic key insights.
Why? Without basic insights (which, as experience reveals, precede the
emergence of axiomatic system), understanding is mainly nominal. Even if a
student is only wanting to be able to do routine computations in a practical
career, if understanding is merely nominal, they will be unable to proceed
whenever boundary conditions do not sufficiently mesh with what is only
nominally familiar.
Modern textbooks speak of “understanding concepts” and “conceptual
understanding.”81 It is said (or assumed) that “concepts are primitive
elements” and that mathematical understanding is a matter of “connecting
concepts” (that one has prior to understanding). It is also now popular to
assert analogies between structures of computer programs and human
understanding. There is an Is it so?’ question here. By adverting to instances
in our experience (see, e.g., Section A.7), it is (self-) evident that while there
are “primitive elements” in mathematics (for instance, in Euclidean
geometry), it is not concepts that are primitive. And the experiential basis of
alleged analogies with computer programming are not found in mathematical
understanding. They are obtained, rather, by correlating structures of
computer programs with structural features of hypothetical conceptual
models of mathematical understanding. On the other hand, adverting to
experience in mathematical understanding reveals that points, lines, invariant
unit length, and other familiar concepts emerge from insight. In other words,
concepts are the fruit of understanding.
Relative to arithmetic, algebra is a higher viewpoint. As already described,
one reaches a higher viewpoint by abstracting82 from instances in a lower
viewpoint.
This sheds light for us on the problem of giving students calculators and other
computational technologies too soon. If a student does not have a good
understanding and technical competence in arithmetic, then they are lacking
most if not all of the symbolic data, diagrams and understanding needed in
order to break through to the higher viewpoint of algebra. Similarly, if a
student lacks nominal and explanatory understanding in algebra and
coordinate geometry, they will be lacking most if not all experience needed in
order to break through to the higher viewpoint of calculus. Again, there are
transitions from algebra to abstract algebra, from calculus to function theory,
and so on.
Evidently, building up an axiomatic system relies on understanding that is
beyond initial or basic understanding (see, e.g., Section A.5). This helps
reveal a common misdirect, that is, when a chapter or lesson begins with
“preliminary concepts,” “axioms” and/or “axiomatically correct definitions.”
Certainly, that approach is dull and is a way to undermine engaged learning.
Attempting to start with axioms and preliminary concepts is, in fact,
attempting to start with answers to questions from a further context. The
approach does not bring the student into an inquiry zone that promotes
emergence of initial key insights. But this is another face of the
misconception in mathematics education that is often called “conceptual
understanding.”83
Whether implicit or explicit, a teacher’s view of mathematical development
influences how and what we teach. Whatever one’s context in teaching, and
whether private or shared with colleagues, there is the possibility of reflecting
on one’s own experience in mathematics and on what one hopes one’s
students also will learn. In other words, as invited in Part A, there is the
possibility of growing in being able to advert to and precisely distinguish
elements in one’s own experience. If one is already a successful teacher, the
invitation is to accurately identify sources of one’s success and that way
become an even more effective teacher. If one is struggling as a junior
teacher, the invitation can include the task of identifying gaps in one’s
mathematical understanding, inconsistencies in what one understands, and
elements of one’s own understanding (specific combinations of nominal and
explanatory, in and across diverse contexts), and to make efforts to bring
one’s learning and teaching into better harmony with the historical field and
the needs of one’s students.

_________________________
77See note 45. “balanced method.”
78See note 76.
79See note 62.
80See note 62.
81See Part C.
82See note 69.
83See three paragraphs above. On the other hand, ‘conceptual understanding’ is just a name. Many
excellent teachers use the term in a positive sense, namely, when referring to student understanding that
goes beyond mere technical competence. Implicitly, they are referring to both nominal and explanatory
understanding in mathematics. The name “conceptual understanding” however is, in fact, inherited
from the philosophical literature and in that context refers to what does not occur in mathematics. That
is, as experience reveals, we do not understand concepts but, rather, concepts emerge from
understanding.
 PART C

Observations Regarding Modern

Mathematics Education

For decades, emphases in mathematics education have been on “theoretical

frameworks.”84 Attempts to justify a theoretical framework are made by
working out features of, and in, a framework; designing lessons based on
such derivations; and then appealing to observations and statistical analyses
of, for instance, student scores, student behavior, reported feelings, and
numerous other “aggregate-events” remote to source events that are
mathematical inquiry and understanding itself.85 Unfortunately, this approach
has been promoting fundamentally mistaken views of mathematical
understanding.
We provide only a few remarks. We are not attempting to give a scholarly
discussion of contemporary mathematics education.86 Our strategy here
mainly is to tease at a few strands of the problem. Among other things, this
will help teachers begin see that it is possible to personally assess models that
one may be asked to implement.
Let’s start by looking to the Introduction of the well-known and influential
paper by Dubinsky and McDonald (2001).87

We do not think that a theory of learning is a statement of truth

and although it may or may not be an approximation to what is really
happening when an individual tries to learn one or another concept in
 mathematics, this is not our focus. Rather we concentrate on how a
theory of learning mathematics can help us understand the learning
process by providing explanations of phenomena that we can observe
in students who are trying to construct their understandings of
mathematical concepts and by suggesting directions for pedagogy that
can help in this learning process.88

Note that, in the first sentence, the authors express a lack of concern for
whether or not their theory is “a statement of truth.” They then remove
themselves from the task of inquiring into “what is really happening when an
individual tries to learn”: “this is not our focus.” Their focus is, instead,
“phenomena that we can observe in students.”89 The last sentence of the
paragraph reveals that in their inquiry into learning, they presuppose a theory
of learning. That this, they presuppose a version of constructivism which, in
this context, is to the effect that students “learn one or another concept in
mathematics”; that they “construct their understanding of mathematical
concepts”; (later in the paper) that when learning mathematics one
“[perceives] mathematical problem situations”; and that, when learning, “an
individual is developing her or his understanding of a concept.”
If you have done exercises such as those in Part A (or beyond) and have
made beginnings in understanding your own understanding in mathematics,
do not the claims made in the paper cry out for correction?90
Perhaps, though, the resulting pedagogy stands. So, let’s look at what
Dubinsky and McDonald suggest for teaching cosets:

Pedagogy is then designed to help the students make these mental

constructions and relate them to the mathematical concept of coset. In
our work, we have used cooperative learning and implementing
mathematical concepts on the computer in a programming language
which supports many mathematical constructs in a syntax very similar
to standard mathematical notation.

Evens and odds, “clock arithmetic” and other similar arithmetic and
geometric groupings are learned by children. In that way, students learn much
of what is needed in order to prepare the way for reaching a higher viewpoint
that includes an understanding of cosets. In our own mathematical
development, we learned mathematics, and did not make use of syntax
“similar to standard mathematical notation.” Indeed, in the historical
development of cosets, founders did not appeal to computer programming
language. Is it not evident that syntax (of computer programming) “very
similar to standard mathematical notation” is, in fact, a distraction from the
mathematical problem?
Why did Dubinksy and McDonald advocate that approach to teaching cosets?
Why attempt to use something similar to mathematics to teach mathematics
instead of helping students learn mathematics itself? Indeed, why might that
be relevant (or not) when “what is really happening” is, as they suggested,
not the focus and not known? They explain it as follows:
Their design instruction focused, not directly on mathematics, but on some
model of how the topic in question can be learned.91
Dubinsky and McDonald had a remarkable dedication to the cause of
mathematics education. As scientists, however, they chose to not focus on
“what is really happening” but rather on “phenomena that we can observe in
students.” In that approach, they also “took themselves out of the equation”
and so did not avail of experience in mathematical understanding.
Constructivist models allegedly account for mathematical understanding.
And so, to reveal the presence of flaws, all that is needed is a
counterexample. But by doing exercises in Part A (or similar ones), one can
obtain numerous counterexamples across a range of instances in elementary
mathematics.

Concluding comments for Part C

As already alluded to, conceptual models of human understanding are not
new. Part of a larger tradition called “conceptualism,” roots of contemporary
constructivism go back at least as far as claims made by Duns Scotus (1266–
1308).92
Aspects of engaged learning models are proving to be significant and are
partly grounded in progress in human neuroscience and human psychology.
But in instances, where a particular mathematical result is needed in what,
specifically, is a student to be engaged? And so, a model might be partly
valid but also call for development. On the other hand, there are “theoretical
frameworks” that are mainly speculative.
How does one tell the difference and so make positive use (or not) of models
encountered?
One can always ask:
Does the model explain instances, in detail, in my experience?
If a model of mathematical learning does not bear out in one’s own instances
of mathematical learning, then (self-) evidently the model is in some way
flawed. Note that what is novel93 here and most important is not this or that
model. Neither are we suggesting that teachers or other readers believe what
we have only very briefly touched on this Supplement. Rather, we are
inviting teachers to a way by which one can check for oneself and so obtain
“ground to stand on,” one’s own experience. We have found, and for
instance, we hope that you are finding, that by adverting to one’s own
experience (by doing exercises like those given in Part A, and more), one can
make progress in identifying and distinguishing orderings of key and core
events in sequences of instances in one’s own mathematical development.
Among those who engage in such exercises, there will be differences in
mathematical background and in descriptions of elements in one’s inquiry
and understanding.94 So, we can expect that there will also be differences in
views regarding learning and teaching mathematics. Like in any serious
science, such differences will not be resolved by logical debate about models.
However, to the best of our ability we can attempt to spell out and discuss
aspects of our own experience, inquiry and growth in mathematics.

_________________________
84Schoenfeld, 2016. For a survey of “theoretical frameworks used in the field of Calculus education,”
see David Bressoud, I. Ghedamsi, V. Martinez-Luaces, G. Torner, G. Teaching and Learning of
Calculus, ICME-13 Topical Surveys, Springer Open (2016) 1–37.
https://www.springer.com/gp/book/9783319329741.
85See, e.g., Schoenfeld, 2016.
86That will be a community achievement, coming from a new kind of collaboration on a global scale.
87E. Dubinsky and M. A. McDonald, “APOS: A constructivist theory of learning in undergraduate
mathematics education research,” in D. Holton et al. (Eds.), The Teaching and learning of Mathematics
at University Level: An ICMI Study (Dordrecht, Netherlands: Kluwer Academic Publishers, 2001):
273–280.
88Dubinsky and McDonald, 2001.
89If one’s focus is on “phenomena that we can observe in students” then inquiry will not be about
mathematical insight but, rather, patterns in “phenomena what we can observe in students” who do not
yet understand. Note, too, that if(as in APOS and its applications) those phenomena are not
explanatorily defined, statistical analysis has little or at most preliminary explanatory significance.
90Regarding concepts, see Section A.7. The confusion is historical and not unique to the work of
Dubinsky and McDonald.
91Ed Dubinksy, “Using a Theory of Learning in College Mathematics Courses,” MSOR Connections
(2000), 1.10.11120/msor.2001.01020010. This article was originally published in Newsletter 12,
TaLUM, Teaching and Learning Undergraduate Mathematics subgroup, 2001. It is available online:
http://www.math.wisc.edu/wil-son/Courses/Math903/UsingAPOS.pdf. Accessed July 17, 2019. The
tradition continues. The literature is extensive. For a point of entry into the literature see, e.g., L.
Benton, C. Hoyles, I. Kalas and R. Noss, “Designing for learning mathematics through computer
programming: A case study of puplis engaging with place value,” International Journal of Child-
Computer Interaction, vol. 16 (2018), 68–76.
92In Scotus’ speculative theory of knowing, “[i]ntellectual intuitive cognition does not require
phantasms; the cognized object somehow just causes the intellectual act by which its existence is made
present to the intellect. As Robert Pasnau notes, intellectual intuitive cognition is in effect a ‘form of
extra-sensory perception’ (Pasnau [2002])” (Williams, Thomas, “John Duns Scotus,” The Stanford
Encyclopedia of Philosophy (Spring 2016 Edition), Edward N. Zalta (ed.),
https://plato.stanford.edu/archives/spr2016/entries/duns-scotus/. See
https://plato.stanford.edu/entries/duns-scotus/#MatForBodSou.
93The method is not new but will be new in mathematics education. See note 43.
94Although, the basic structure given in Figure A.1 is invariant. If one claims that a different structure
accounts for one’s understanding, is it because one understands something about an understanding that
one does not have; or that one judges what one does not understand? In the terminology of philosophy
of science, there results what is called ‘performance-contradiction.’ Philosophical terminology is not
the main issue here. As in the text above, a fundamental question always is: “Does a model explain
instances, in detail, in my experience?” Is this me?
 Index

Acceleration, 34, 108

angular, 35
Angle, 1, 67
straight, 209
Anti-derivative, 7, 13
Approximation, 12
linear, 6, 8, 20, 22, 38, 56, 91, 157, 182
Archimedes, 17, 195, 204
Area
rate of change of area, 9, 15
rate of relative change in area, 125, 129, 136
Average speed, 2

Base, 199
Boundary
circulation along, 166, 171
curve, 95
positively oriented, 99
surface, 158

Chain rule, 38
applications to integration, 46
in one dimension, 41
in two dimensions, 50
Circulation, 73, 74, 81
2-dimensional, 121
3-dimensional, 176
along a curve, 76
circulation integral, 79, 82, 83
Coordinate geometry, 1, 26, 36, 209
Coordinates, 26, 28
polar, 53
principle coordinates, 109
rectangular, 55
Counterexamples, 224
Cross-product, 1, 34, 36
Curl, 99, 111, 132, 141, 142, 166, 175, 176
Curve
boundary, 95, 116
closed, 95
positively oriented, 95
streamline, 68

Decomposition, 108
of the Jacobian, 139
Definition
of flux, 150
of incompressible flow, 131
of irrotational flow, 112
of Jacobian, 56
of symmetric matrix, 108
of total circulation, 77
sum of two curves, 85
Density
circulation, 95, 101, 103, 112, 125, 135, 141, 166, 175, 176, 182
rotation, 101
Derivative, 7
function, 7
of area under the graph, 23
Determinant
2-d, 29, 33
3-d, 30, 33
Divergence, 115, 119, 139, 147
2-d, 129
3-d, 157, 185
Divergence Theorem
2-d, 119, 125, 135
3-d, 119, 157
Dot product, 1, 26, 27

Euclid, 26

Fluid dynamics, 66, 71, 77, 108, 112, 125, 131

Fluid motion, 1
Flux, 115, 125, 147
2-dimensional, 115, 121
3-dimensional, 155, 160
density, 124, 135, 136
flux integral, 118, 122, 136, 163
Fundamental theorem of calculus, 1, 8, 15

Galileo, 2, 67
Graph, 4–6
area under, 12
Green’s Theorem, 23, 86, 88, 94, 98, 115, 121, 135
for surfaces, 166
in the plane, 132
 vs Stokes’Theorem, 173

Height, 13

Incompressible, 131, 141, 180, 185

Increment, 2, 40, 46, 51, 55, 81
patch, 149
relative change, 179
Insight, 5, 9, 14, 15, 194, 195, 197, 212
basic, 186, 202
direct, 201
reflective, 198
Instantaneous, 5, 7, 12
Integral, 13, 25
change of variable, 59
circulation integral, 79
integral curve, 68, 186
Integral curve, 68, 110, 126, 133, 148, 186
Interior
area, 136
of a closed curve, 112, 125
point, 54
Invariance, 79, 80, 83, 154, 210, 212
Irrotational, 112, 143
Isometry, 182, 184

Jacobian, 56, 133, 135, 139, 141

Length, 10, 11, 23, 24, 26, 75, 210, 211

Limit, 18, 40, 41, 136
Line element, 81
Line segment, 73, 75, 83, 106, 211
Linear algebra, 26, 30, 135

Mass, 24, 131

density, 70, 148
flow, 65
flow rate, 70, 73, 87
units, 71
Mathematics, 26, 53
education, 189, 191, 193, 221
history, 214, 218
insight, 213
judgment, 201
Method of substitution, 47, 48
Middle terms, 173, 174

Nobel Prize, 18
Normal, 36
field, 150, 155, 156, 158, 167
vector, 36, 117, 137, 159, 162, 176
Oriented
oppositely, 84
positively, 94, 99, 115, 170

Parallel, 35, 70, 76, 113, 165, 176

Parallelogram
area of, 16, 29
Parameter, 77, 126, 143
parameterization, 76, 79, 81, 83, 85, 118, 159, 164
for sphere, 152
positive, 95
Physics
classical, 34, 70
modern, 2
Newtonian, 65
Polar coordinates, 53
Positively Oriented, 94, 95, 99, 116
boundary, 95
closed curve, 124
normal curve, 160
Projection, 26, 27, 165, 176, 183
Proof, 203

Rate
net mass-flow rate, 116, 117
of change, 2, 5
Rectangular coordinates, 53
Region, 24, 58
partition, 96
simple solid, 158
Rotation, 36, 88, 101, 108
pure, 111, 182
rotational, 34, 141, 184
rotational displacement, 36

Scalar product, 26
Stokes’ Theorem, 165, 167, 173
Strain, 125
pure, 182
relative, 107, 180, 185
Streamlines, 68, 89, 101, 108, 113, 124, 133, 136, 142, 183
Sum, 12, 13, 55
of two curves, 85
Surface, 65, 81, 149, 152
2-dimensional, 67, 168, 176
boundary, 158
integral, 166
orientation, 158
Symmetric, 109, 139, 141, 142, 185
anti-symmetric matrix, 108, 111, 113, 183
symmetric matrix, 108, 132, 182
Telescopes, 22
Trace, 133, 135, 136, 139, 141

Variable
change of variable, 59
independent, 7
Vector calculus theorems, 1, 25
Velocity
angular, 111, 133
field, 65, 75, 108, 141
vector, 66, 68, 106, 149
Volume, 28, 152
3-d, 71, 156
rate of change, 179
rate of relative change, 180
relative change, 179, 185
Volumes, 28

Water flows, 67, 68, 86, 108, 123