83%(6)83% нашли этот документ полезным (6 голосов)

880 просмотров641 страницаlibro de calculo para aprendizaje sin maestro

Lectures on Freshman Calculus

© © All Rights Reserved

PDF или читайте онлайн в Scribd

libro de calculo para aprendizaje sin maestro

© All Rights Reserved

83%(6)83% нашли этот документ полезным (6 голосов)

880 просмотров641 страницаLectures on Freshman Calculus

libro de calculo para aprendizaje sin maestro

© All Rights Reserved

Вы находитесь на странице: 1из 641

An intuitive exposition of the basic techniques
for calculating with derivatives and integrals
a
LECTURES ON
FRESHMAN CALCULUS
ALLAN B. CRUSE / MILLIANNE GRANBERG / University of San Francisco
A
vv
ADDISON-WESLEY PUBLISHING COMPANY
Reading, Massachusetts / Menlo Park, California / London / Don Mills, OntarioTo our teacher,
colleague and friend,
OTTO PLAAT.PREFACE
These informal lectures are addressed to the college freshman who wants to know
what the calculus is about and how it may be used to solve basic problems in the
various quantitative disciplines. The lectures grew out of the authors’ introductory
course at the University of San Francisco during the years 1966-1970 in which
students were enrolled from virtually every department, including mathematics
and physics, the life sciences, Liberal Aris, business, and social-science. Invariably
our students had already studied high-school algebra and geometry, and accord-
ingly those studies became the prerequisites for reading the present text. In addi-
tion, some of our students had been introduced to trigonometry and analytic
geometry, and always these students found that this added background was very
helpful. However, almost none of our students had had any previous acquaintance _
with the calculus itself.
Perhaps what is unique about this introductory textbook is the extent to
which the authors have concerned themselves with motivation. This will be evident
in the selection and arrangement of the topics, in the quasi-historical spirit of the
exposition, and in the general level of mathematical sophistication that is pre-
sumed. Great pains have been taken to meet the freshman student on his home
ground, to minimize dreary preliminaries, to rely upon geometrical and physical
intuition, to place the questions considered in a meaningful context, and to make
the greatest possible contact with the student’s previous mathematical experience,
reminding him of what he already knows, at the same time pointing out to him
important gaps in his present knowledge.
Our approach to the calculus is concrete and computational, rather than
abstract and deductive. It is informal and applied, as opposed to rigorous and
theoretical. Fundamental concepts are introduced by means of the role they play
in solving basic problems. We have concentrated on those topics which will be
most attractive and meaningful to a student at the first-year level. In particular, we
have tried to keep in mind a clear distinction between calculus and analysis.
For us, the subject of calculus is properly devoted to calculations performed
with derivatives and integrals. It is concerned with techniques for computing
numerical solutions in problems that lie beyond the reach of traditional high-school
methods. This means, on the one hand, that we do not attempt to treat difficult
foundational questions (such as the definition of limit), and, on the other, that we
assiduously avoid time spent in “proving the obvious.” We strive instead to
engender a feeling for our subject’s relevance to ordinary experience by continually
extracting mathematical ideas from familiar everyday settings.‘The kinds of questions that analysis is concerned with are seldom appreciated
by the freshman student. They are simply too sophisticated: they presume too
much previous mathematical experience and too refined a style of thought. They
take for granted an intimate acquaintance with difficulties and paradoxes which the
freshman does not yet even recognize exist. They confuse him and distract him
when they appear as part of the calculus, and thereby lend credence to his pre-
conceived notion that the aims of mathematics are mysterious and inscrutable.
‘The problem here, of course, is simply that his background has not prepared him to
appreciate the kinds of inquiries to which analysis is devoted. That will come later.
For the prospective Math Major, then, these lectures content themselves with
laying the ground—mathematically and psychologically—for a subsequent course
in Real Analysis, complete with epsilons, deltas, and least upper bounds, but they
do not inject those ideas prematurely here. The intricacies of continuity, com-
» convergence, and the rest, and the study of their interrelations, can be
enjoyed only after one knows where these concepts come from and what they are
really about. We believe those ideas properly constitute a course in themselves,
nnd that the delight they provide should not be spoiled through exposure at too
early a mathematical age.
For the non-Major, these lectures will show how to calculate with derivatives
and integrals in a variety of settings, and will explain why the techniques work,
without belaboring logical details of interest mainly to the pure theoretician. The
fact that the many problems and examples seem heavily slanted toward geometry
and physics should not discourage students in other disciplines; it simply happens
that these two domains provide the simplest setting in which to exhibit the funda-
mental concepts, but once the meaning of those concepts has been clearly grasped,
they can easily be transferred to other fields, such as economics, biology, or sta-
tistics.
San Francisco, California A. B.C
Junuary 1971 M. G.SPECIAL ACKNOWLEDGMENTS
The authors gratefully acknowledge the assistance of the Committee on Research
at the University of San Francisco which supported the initial phase of this writing
project, and of the Addison-Wesley Publishing Company which assisted in bringing
it to an early completion. Special thanks are due to our colleagues, Dan Gallin
and Robert Wolf, who read portions of the manuscript and offered useful sugges-
tions, and to our student, Dennis Batey, who provided us with some computer-
drawn graphs. The critical comments of Lynn Loomis and Marion Smith also
proved to be of enormous help to the authors and are deeply appreciated. We owe a
substantial debt as well to the talented production staff at Addison-Wesley for
their cooperation and expertise in preparing our manuscript for the presses.CONTENTS
Chapter I INTRODUCTION TO THE DERIVATIVE
1 Two Problems of Optimization 3
2 The Problem of Tangents 12
3 Descartes’ Method of Equal Roots 21
4 Fermat’s Method of Limits ee eee Om
5 Solution of the Optimization Problems ea
6 Some Rules for Slope-Predicting Formulas 48
7 The Rocket Problem. i 58
8 Independent and Dependent. Variables 66
9 Domain and Range of a Function . 14
10 Local and Global Extreme Values . 85
11 Instantaneous Rates of Change 93
12 Further Differentiation Rules . 101
18 Leibnitz’s Notation and Non-Time Rates. 109
14 Continuation of the Rocket Problem . 117
15 The Concept of Continuity . 126
16 Derivatives of the Circular Functions ' 188
17 Antiderivatives : aes 147
. #18 An Application in Optics | | 155
*19 An Application in Business. . 162
*20 An Application in Numerical Analysis 170
Chapter IT INTRODUCTION TO THE INTEGRAL
21 Buffon’s Needle Problem 7 181
22_Eudoxus’ Method of Exhaustion 187
23 Summation and Sigma Notation 198
24 Curvilinear Areas and the Integral Sign 7 208
25 Some General Area-Predicting Formulas . 216
26 Barrow’s Theorem on Evaluating Integrals 227
27 The Integral as a Limit of Riemann Sums 240
28 Volumes of Geometric Solids : 251
29 The Calculation of Work 261
30 An Example with the Ellipse 268
*31 Hydrostatic Pressure 276
82 Infinite Sums and Improper | Integrals” 282
*optional lessonChapter HI
Chapter IV.
CALCULUS WITH COMPOSITE FUNCTIONS
33 Concavity and the Second-Derivative Test
Linearization and Approximation .
A Proof of the Error Formula .
36 ‘The Mean-Value Principle - oe
37 Composite Functions and the Chain Rule
38 Problems Involving Related Rates
39 Curves Described with Parametric Equations
40 Cauchy’s Formula and L’Hospital’s Rule
41 Differentials and Implicit Differentiation .
42 Problems involving Auxiliary Variables
43 The Average Value of a Continuous Variable
44 The Change of Variables Technique
45 Volumes by the Method of Shells .
46 The Indefinite Integral . ee
47 The Method of Integration by Parts .
“48 The Binomial Theorem and Wallis’ Formula
292
301
310
318
325
333
340
349
356
364
372
379
387
394
402
408
INTEGRATION OF ADDITIONAL ELEMENTARY FUNCTIONS
A Search for Some Missing Antiderivatives
54 Introduction to the Number e .
55 The Series of Reciprocal Factorials
56 Rational and Irrational Numbers .
57 Proof that e is Irrational
*58 Proof that 7 is Irrational :
59 The Concept of Inverse Functions
60 Some Noninvertible Functions...
G1 Inverses for the Trigonometric Functions
62 The Exponential Function and its Properties
“63 A Different Definition of Irrational Powers? .
G4 Logarithms to Other Bases .
65 Are Length and Trigonometric Substitution
“6G Surface Area and Completing the Square
67 The Method of Partial Fractions . .,
68 Extending the Method of Partial Fractions
69 Tntroduction to Differential Equations
Approximating Integrals by the Trapezoidal Rule
Proof of the Trapezoidal Rule’s Error Formula .
The Logarithmic Laws and Their Application
53 Integrations Using the Natural Logarithm Function
416
424
432
437
444
450
458
463
468
472
479
489
498
506
513
523
529
538
546
551
55670 The Separation of Variables Technique
71 The Study of Population Growth .
72 The Hanging Cable Problem
Appendixes
I Basic Ideas of Analytic Geometry .
II Basic Ideas of Plane —
Il Tables. . . 7 eee
Answers to the Exercises
Index
Vortens
we
563
570
578
585
599
611
614
635CHAPTER I
INTRODUCTION TO THE DERIVATIVELECTURE 1
TWO PROBLEMS OF OPTIMIZATION
The calculus was invented in order to solve certain kinds of problems which are
beyond the reach of ordinary algebraic methods. We begin our study of the
calculus with some examples of such problems.
The box problem
An open box is to be made from a 4 ft by 5 ft piece of cardboard by cutting out
squares of equal size from the four corners and bending up the flaps to form sides.
What are the dimensions of the box of largest volume which can be made in
this way?
Although we will need some additional mathematical ideas to finish solving
this problem, we can get started with nothing more than high school algebra.
Suppose we use the letter x to stand for the length of the sides of the cut-out
corner squares. Then the dimensions of the box which we get by bending up the
flaps can be expressed as follows:
Length = 5 — 2xft, Width = 4 — 2x ft, Height = x ft.
The box’s volume will be the product of these three dimensions:
Volume = (5 ~ 2x)(4 — 2x)x ft3.
\— 5-22 ft A
3$ Introduction to the Derivative
In these formulas we can think of x as a variable whose possible values are
subject to certain physical restrictions. For example, since x represents an actual
physical length, x cannot have a negative value. Also, since the piece of card-
hoard is only 4ft wide, it would be impossible to cut out corner squares having
sides more than 2 ft long. These restrictions imply that x must have a value be-
tween 0 and 2. In other words, x must satisfy the inequality
Osx <2
Yet within these restrictions there remains an infinity of possible choices for x.
Our problem is to discover which of these choices for x will enable us to produce
the box of maximum volume.
Small z gives Large 2 gives Medium 2 gives
shallow box narrow box more volume
Because the box’s volume depends upon the value of x and cannot be cal-
culated until this value is specified, it will be convenient for us to use the mathe-
matical symbolism
V(x)
to stand for the volume of the box which results from a corner cut of size x. This
symbolism, which consists of the letter V (for volume), followed by the letter x
inside parentheses, is pronounced “V of x.” Its meaning should not be confused
with multiplication of V by x, which would be written V+ x, or simply Vx. We
have already found a formula for calculating V(x), namely
Vix) = © — 2x)(4 ~ ox)x. :
For example, we see that a corner cut of size } ft will produce a box whose volume *
is 6 {t', We can express this fact by writing the equation V(j) = 6. Similarly,
the equations V(}) = 3}% and V(0) = 0 express the volumes of the boxes which
result from corner cuts of sizes } ft and 0 ft, respectively.
: We can illustrate how the box’s volume changes as x increases by calculating
V(x) for some specific values of x and displaying our results in a table.
TABLE
Oia elt le |
ve [of si fe lex ).6 | at le vs | 01 | Two Problems of Optimization 5
Among the values for x which appear in this table, we can see that the box
of largest volume is obtained when x = 3. Does this choice for x, then, give us
the solution to the box problem?
No, probably not, because our table includes only a sample of the many pos-
sible choices for x. For all we know, a value of x different from the ones we con-
sidered might produce a volume surpassing any our table shows. (As a matter
of fact, V(0.74) = 6.564096 is larger than every volume we tabulated above.)
Perhaps we should consider enlarging our table to include more of the pos-
sible values for x. But would that really help us to find the solution?
A few moments’ thought should convince us that to enlarge the table would
be a waste of time. For one thing, we could never include all possible values for
x, no matter how large a table we made. Consequently, there would always exist
the possibility that our table might omit the very value of x which produces the
maximum volume.
On the other hand, even if our enlarged table did include the correct choice
for x, we would have no way of recognizing that this choice was the right one.
For in order to know that a certain volume was the maximum, we would need
first to compare it with the volumes of the boxes produced by every single one
of the permissible choices for x. Since there is an infinity of such choices, this
kind of comparison would be impossible to carry out. Thus it becomes clear that
solving our box problem will require a different approach.
Suppose we consider drawing a graph of our volume formula, in order to help
us think geometrically about the problem. We can make a reasonably accurate
sketch of this graph if we first plot the number pairs from our table and then draw
a smooth curve passing through those points.
GRAPH
V(q)=(-22)(4—22)z
0
i
z
i
i
i
i
i
2
$15258
Looking at this graph, we can see that the box problem would be solyed if
only we could determine the coordinates of the highest point on the curve. The
first coordinate of the highest point would be exactly the choice for x which pro-
duces the greatest volume, while the sécond coordinate would give the numericalon to the Derivative
6 Introduce
value of that volume. But how can we distinguish the highest point on our graph
from the other points nearby? ;
Notice that the highest point on our graph occurs where the curve has just
stopped rising but has not yet started falling—in other words, where the direction
of the curve is momentarily horizontal.
Descending Horizontal
More precisely, the highest point on our curve is the one point where a tangent
Hine would be horizontal.
*
bs
The tangent line ix horizontal The tangent line is not horizontal
But how do we go about locating the point on our curve where a tangent line would
be horizontal? And what is more pertinent, how do we determine that point’s co-
ordinates? Our study of the calculus will shortly enable us to answer these questions
in a surprisingly simple way, so that solving the box problem and others like it
will become a snap!
Let us put the box problem aside temporarily and turn to a companion prob-
lom whose analysis is quite similar.
The oil can problem
One quart
Suppose you are assi
cireular cylinder, an
57.75 in*) of oil.
What dimensions
mount of metal for
igned the job of designing an oil can in the shape of a right
id you are told that the can must hold 1 qt (approximately
should you specify so that your can will require the least
is manufacture?1 | Two Problems of Optimization a
‘Two numbers must be specified in order to determine the design of the oil
can described in this problem, namely the can’s radius and its height. Suppose
we use the letters r and A to stand for these two dimensions, measured in inches.
Joel
=
We know from high school geometry that the volume of a right circular cyl-
inder having these dimensions can be expressed as follows:
Volume = ar7h.
We are told that our oil can must contain exactly 1 qt of oil. This means that
its volume must be 57.75 in’. Therefore, we must choose the dimensions r and h
in such a way that-the following equation will be satisfied:
57.75 = arth.
Accordingly, we are really free to specify only one of the can’s dimensions, be-
cause if we choose a certain value for the radius r, then the height h is automatically
determined by the equation above. Specifically, we must have
_ 51.15
“pr?
If our can’s top, bottom, and sides are made from metal of uniform thickness,
then it is reasonable to assume that the amount of metal used up in manufacturing
the can will be proportional to the can’s surface area. Since we want to use as
little metal as possible, we should specify the radius r so that the resulting oil can
will have the smallest possible surface area.
It is not hard to find a formula for our can’s surface area. The top of the
can, as well as the bottom, will be a circular disk of radius r. Consequently, the
area of the top will be 72, and the area of the bottom will be the same. To find
the area of the can’s sides, we imagine cutting open the can from top to bottom and
flattening out its sides to form a rectangle as shown in the illustration.
Sen
Cut the Flatten out Find the area of
can open its sides the rectangle& Introduction to the Derivative
‘The rectangle’s height will be the same as the height h of the can, ed the rec-
tangle's width will be equal to the can’s circumference, which is 2nr. The area of
this rectangle, namely 2rrh, will be the area of the can’s sides. Finally, adding the
areas of the can’s top, bottom, and sides, we obtain a formula for the can’s total
surface area:
Surface area = oar? + _ar?_ + _2arh .
areaof areaof area of
top bottom sides
Remembering that our can’s height must be chosen to satisfy the equation
h = 57.75/zr?, we can rewrite our formula for the can’s surface area in terms of
the letter r alone:
Surface area = Qnr? + Qnr(57.75/ar?),
or equivalently, the formula may be simplified as follows:
R(dnr® + 281) -
Surface area =
In this formula, we can think of r as a variable whose possible values are
subject to certain restrictions, due to our interpretation of r as the radius of a
1-qt cylindrical can. In particular, since it is impossible for a cylinder to have a
negative radius, we sce that r cannot have a negative value. Furthermore, even
though it is possible for us to imagine a cylinder with radius zero, such a cylinder
would certainly haye no room inside it and therefore could not hold a quart of oil!
Consequently, in our oil can problem, the variable r can be neither negative nor
zero. In other words, r must be a positive number. We may conveniently express
this restriction upon r by writing the following inequality:
r>0.
Despite this restriction there remain an unlimited number of possible values for
our can’s radius r.
Large F gives Medium r gives
wide can less surface area1 | Two Problems of Optimization 9
For convenience we will use the symbolism S(r) to stand for the surface area
of the 1-qt cylindrical can having radius r. We have already derived a formula
for calculating S(r):
(4x73
Ss) 3 (Axr’ + 231) |
Our next step is to make a table and graph illustrating the way in which the sur-
face area S(r) changes as the radius r increases. We use the approximation + =
3.14 for carrying out the necessary arithmetical computations.
¢ CRAP
TABLE '
rt str) = Hee eB
Fier 5
Hh fa 0 ha BO NH
Although the table probably does not contain the solution: to the ofl can
problem, it does at Jeast provide assistance in drawing a fairly accurate sketch
of the graph of the surface-area formula. What we need to find, in order to finish
solving the problem, are the coordinates of the lowest point on this graph. The
first coordinate of this lowest point will be the exact choice for the radius r which
produces the oil can of minimum surface area while the second coordinate will be
the numerical value of that surface area.
Here we discover a striking similarity between the box problem and the oil
can problem. In the previous casé we wanted to find a graph’s highest point,
while in this case we wani to find a graph’s lowest point. In each case the point
we seek occurs precisely where the graph’s tangent line would be horizontal.
Without the help of the calculus, we can go no further in solving these two
problems. However, from what we have seen in analyzing them, it should be easy
to understand why the beginning phase of our work will concern itself with the
general problem of finding tangent lines to curves at various points.
ASSIGNMENT 1
1. Suppose an open tray is to be made from an 8 in. by 15 in. sheet of tin by cutting
out squares of equal size from the four corners and bending up the flaps to form sides.
a) Letting x stand for the length of the sides of the cut-out corner squares, find a
formula for the volume of the tray in terms of x.10 Introduction to the Derivative
b) Write an inequality which expresses the physical restrictions on the value of x.
c) Sketch a graph of your volume formula V(x).
d) Estimate the dimensions of the tray of largest volume.
. Suppose you are assigned the job of designing an oil can in the shape of a right circular
cylinder, Assume that the metal used in making the top and bottom of the can will
cost 2¢/in? while the metal used in making the sides will cost only ig¢/in®.
a) Letting rand h stand for the radius and height of your can, respectively, write a
formula for the total cost of the metal used in manufacturing it.
bb) Acsuming your can is supposed to hold 4 qt of oil, re-express the cost of the metal
used in terms of the variable r alone.
Bp
Sketch a graph of your cost formula C(r).
Estimate the dimensions of the can of least cost.
)
dq)
3, A closed-top box is to be made from a 5 ft by 8 ft piece of cardboard by cutting out
squares of equal size as shown below and folding along the dashed lines, tucking the
two extra flaps inside.
8) Letting x stand for the length of the sides of th
5 e cut-out
for the volume of the closed-top box. hou squares, Gnd formule
b) Write an inequality which expresses the physical restrictions on the value of x.
¢) Sketch a graph of your volume formula Vix).
4) Estimate the dimensions of the box of largest volume.
: a Jones own adjoining 25 ft by 100 ft lots, and Smith has already built a
te _ he w i around his lot. Now Jones wants to build a rectangular play yard
ee Bs enclosing as large an area as possible, but he has only 80 ft of wire fencing.
Se, Jones can utilize part of Smith’s fence along one side of his play yard.
¥) Letting x stand for the
I le ‘ an
find a formule in cane eneth of the portion of Smith's fence which Jones shares,
of x for the area of Jones's play yard.10.
1 | Two Problems of Optimization 11
b) Write an inequality which expresses the physical restrictions on the value of x.
©) Sketch a graph of your area formula A(x).
a) Estimate the dimensions Jones should use for his play yard.
|. When we discussed the oil can problem in Lecture 1, we made a table showing various
values for the oil can’s radius r and corresponding surface area S(r). Verify that this
table does not contain the exact solution to the oil can problem.
. Estimate the coordinates of each point on the graph below where a tangent line would
be horizontal.
. A department store regularly runs a 5 in. by 12 in. advertisement in the daily news-
paper. The newspaper sells space by the square inch and requires advertisers to
include }in. margins along the top, bottom, and sides of their ads. Could the de-
partment store continue to spend the same amount of money but obtain more area
for printed text by altering the shape of its ad?
. An apple orchard now has 40 trees per acre, with an average yield of 500 apples per
tree. For each additional tree planted per acre, the average yield per tree is reduced
by about one dozen apples. Could a larger crop of apples be obtained by planting
a larger number of trees per acre?
. Suppose that 7’ and x are two variables which are related by the formula
T(x) = x? — 2x43
where the value of x is restricted by the condition 2 < x < 5. Show that T = 4 is
not the smallest possible value for T under these conditions.
Suppose that x and y are two variables which are related by the formula
y%) = (« ~ 2)? 43,
with no restrictions placed on the value of x. Explain why, under these conditions,
the smallest possible value for y is y = 3.10 Introduction to the Derivative
b) Write an inequality which expresses the physical restrictions on the value of x.
¢) Sketch a graph of your volume formula V(a).
a} Estimate the dimensions of the tray of largest volume.
2, Suppose you are assigned the job of designing an oil can in the shape of a right circular
Assume that the metal used in making the top and bottom of the can will
while the metal used in making the sides will cost only 1¢/in®.
a) Letting r and A stand for the radius and height of your can, respectively, write a
formula for the total cost of the metal used in manufacturing it.
bh) Assuming your can is supposed to hold 4 qt of oil, re-express the cost of the metal
used in terms of the variable r alone.
©) Sketch a graph of your cost formula C(r).
4) Estimate the dimensions of the can of least cost.
s
. A closed-top box is to be made from a 5 ft by 8 ft piece of cardboard by cutting out
squares of equal size as shown below and folding along the dashed lines, tucking the
two extra flaps inside.
Tht
n) Letting x stand for the length of the sid
les of the cut-out
for the volume of the closed-top box. Ce ee
b) Write an inequality which expresses the physical restrictions on the value of x.
©) Sketch a graph of your volume formula V(x).
a) Estimate the dimensions of the box of largest volume.
‘mith and Jones own adjoining 25 ft by 100 ft lots, and Smith has already built a
poe aie oa around his lot. Now Jones wants to build a rectangular play yard
. — = losing as large an area as possible, but he has only 80 ft of wire fencing.
cones can utilize part of Smith’s fence along one side of his play yard.
1
1 {Smith's
{lot
4
Jones's
lot
a} Letting x stand for the len;
find sth of the portion of Smith’s fence which Jones shares,
a formula in terms of x for the area of Jones’s play yard.e
10.
1 | Two Problems of Optimization =H
b) Write an inequality which expresses the physical restritions on the valus of x,
) SE :
@) Extimete the dimensions Jones should use for bis play yard.
ich a graph of your area formula A(x).
When we discussed the ofl can 1
values for the oil can’s radius r and corresponding surface area S(r). V
table does not contain the exact solution to the cil can problem.
Estimate the coordinates of each point on the graph below where a tangent line woul
be horizontal.
‘A department store regularly runs a 5 in. by 12 in. advertisement in the daily ness.
paper. The newspaper sells space by the square inch and requires advertisers ty
include }in. margins along the top, bottom, and sides of their ads. Could the de:
partment store continue to spend the same amount of money but obtain more areq
for printed text by altering the shape of its ad?
‘An apple orchard now has 40 trees per acre, with an average yield of 500 apples por
tree. For each additional tree planted per acre, the average yield per tree is reduced
by about one dozen apples. Could a larger crop of apples be obtained by planting
a larger number of trees per acre?
Suppose that T and x are two variables which are related by the formula
T@ =
where the value of x is restricted by the condition 2 < x < 5.
not the smallest possible value for T under these conditions.
— 2x +3
. Show that T = dig
Suppose that x and y are two variables which are related by the formula
y@) = (« — 27 43,
with no restrictions placed on the value of x. Explain why, under these conditions,
the smallest possible value for y is y = 3.3 to the Derivative
12 Introd
LECTURE 2
THE PROBLEM OF TANGENTS
i in optimization problems, namely
In Lecture 1 we found that in order to solve certain op’
the hox problem and the oil can problem, we must be able to locate the exact
point on a graph where, if a tangent line were drawn, it would be horizontal.
The highest point
\
‘The lowest point
We are already somewhat familiar with the concept of a tangent line to a
curve from our study of high school geometry where we investigated tangent lines
to circles. A tangent line to a circle is a line which touches the circle in just one
point known as the point of tangency. All other lines either miss the circle com-
pletely or else cross the circle in two different places.
a
Most lines miss the circle or cross it twice
In geometry we learn a very
: ‘ simple procedure for constructing a tangent line
to a circle using the ruler and
compass. This procedure consists of drawing the2 | The Problem of Tangents 13
radius from the center of the circle to the desired point of tangency and then
constructing a perpendicular to this radius. This construction is illustrated below.
Locate center C Draw the radius to ——-Construet the Jine through T
of the circle the given point 7 perpendicular to this radius
It is easy to prove that the perpendicular constructed by this procedure must
be tangent to the circle at the desired point T. The proof consists of showing that
this point T'is the only place where the constructed line touches the circle.
$ lies too far away from C
For suppose S is any point on the constructed line different from the point T.
Then the three points C, S, and T will be the vertices of a right triangle, with
line CS as the hypotenuse. Since the hypotenuse of any right triangle is always
longer than either leg, we see that CS will be longer than CT. But CT is a radius
of the given circle. Therefore, the point S cannot lie on the circle because it is too
far away from the circle’s center C. This proves that the point T is the only point
where the line and the circle meet, and shows, therefore, that the line is tangent
to the circle at T.
This simple construction shows that, in the case of a circle, the concept of a
tangent line is quite clear and well understood.
When we look at other types of curves, however, we quickly realize that it
is more difficult to know what we mean by a tangent line. Of course we do have
a fairly clear intuitive notion of what a tangent line is, and we feel confident that
we would have no trouble recognizing a tangent line if we saw one.
TN GN
This one is a tangent line This one isn’t14 Introduction to the Derivative
a tangent line to a circle as ‘“‘a line which touches the circle
to other curves. Often a line which touches
e have in mind by a tangent line at all!
But our description of
only once” is not generally applicable
a curve in only one point will not be what we
‘The line touches only once but we wouldn’t
call it a tangent line
On the other hand, a line may sometimes touch a curve in more than one point
and still be considered a tangent.
e line touches twice but we would still
it a tangent line
Furthermore, the simple procedure we have described for constructing a tangent
line to a circle clearly will not work with other curves because noncircular curves
have no “center.” The perpendiculars to their tangents through the points of
tangency will not all “radiate” from one central point.
The curve doesn’t have a “center”
a Bee at the general concept of a tangent line to a curve is more elusive
cog special concept of a tangent line to a circle.
se ena ee tangent. lines to curves has engaged the interest of
eva hout mathematical history, and various methods have been
ics — ae aula to certain special curves. As an example, the
pee G 7 et ers i of a way to construct tangent lines to the parabola.
ai : 7 @ paral bola is one of the simplest of curves. It consists of all
r in the plane which are equally distant from a fixed point and a fixed
those2 | The Problem of Tangents 15
line. The fixed point is known as the focus of the parabola, and the fixed line is
known as the directrix.
Focus
Direetrix
Parabola
Every point on the parabola is equally distant
from the focus and the direetrix
Actually, of course, there are many different parabolas, depending upon how
the focus and directrix are chosen, just as there are many different circles, depend-
ing on the choice of center and size of the radius. Any parabola, however, will
automatically be symmetric about the line which passes through its focus at right
angles to its directrix. This line of symmetry, called the axis of the parabola, must.
cross the parabola in exactly one point known as the vertex. The various parts
of the parabola, its focus, directrix, axis, and vertex, are shown in the diagram.
Axis
Vertex
‘The parabola
Directrix
‘The parabola, its focus, directrix, axis, and
vertex
We can describe the Greek method for drawing tangent lines to the parabola
in three steps. Suppose we let T denote a point on the parabola through which
we want to draw a tangent. We first drop a perpendicular from this point T to
the parabola’s axis, letting P denote the foot of this perpendicular.
Axis
.
Drop perpendicular from point 7° to axis1G Introduction to the Derivative
Next, we draw a circle through the point P having the vertex Vasits center. This
circle will intersect the parabola’s axis in a new point, different from P, which we
label Q.
Draw cirele through Phaving center V
, we draw the straight line passing through points Q and T. We can prove
that. this line QT touches the parabola only at the point T, and this should convince
us that line Q7 is the tangent we were looking for.
T
Line QT is the desired tangent
To carry out the proof, let us suppose that S denotes a point on the line QT dif-
ferent from T. It will not be hard to show that this point S must lie further from
the focus than from the directrix, and hence that S does not lie on the parabola.
s
We begin this demonstration by
‘i droppin i i
and T to the directrix of our parabol pping perpendiculars from the points S
dP tc ja, labeling their feet A and B, respectively.
. Clear _ the segment SB is longer than the segment SA because ‘the hres
A, B, and S are the vertices of a right triangle having SB as hypotenuse.
as we have already menti i i i
neces i ioned, the hypotenuse of a right triangle is always longer2 | The Problem of Tangents 17
What is not so clear is the fact that the segments SF and SB are actually
equal in length, and it will require some argument to make that apparent. But
once that has been shown, then it will follow that SF must be longer than SA.
Length(SF) = length(SB) > length(SA).
Consequently, S cannot lie on the parabola for the reason that the parabola, by
definition, consists only of points lying equidistant from its focus and directrix.
‘This leaves T as the only possible point which the parabola and the line QT may
have in common.
Thus the whole argument boils down to showing that S is equally distant
from B and F, a fact which we shall now establish by proving that the line QT is
the perpendicular bisector of the line segment BF. It is well known, of course,
that all points on a line segment’s perpendicular bisector lie equally distant from
that segment’s endpoints.
Letting C denote the point where the parabola’s axis meets its directrix, we see
that segments CV and VF have the same length.
Length(CV) = length(VF).
This is because V is a point on the parabola and hence is equidistant from the
focus and directrix. At the same time, we know that segments QV and VP have
the same length because that is how the point Q was constructed.
Length(VP) = length(QV).18 Introduction fo the Derivative
init y i i ddition, and observing that V
Combining the two preceding equations by at :
must lie between Q.and F, and also between C and P, we find that segments CP
and QF must be equal in length.
Length(CP) = length@F).
Of course, segments BT and CP, being opposite sides of a rectangle, will have
the same length.
Length(BT’) = length(CP).
And this pair of equations shows us that the figure QFT'B must be a parallelogram,
since it has two parallel sides of equal length.
Length(BT’) = length(QF ).
Moreover, since 7" lies on the parabola, segments BT and TF must have equal
length.
Length(TF) = length(BT ).
This means that our parallelogram QFTB is actually a rhombus!
Now it is a well-known (and easily proved) theorem of geometry that the
diagonals of a rhombus are perpendicular bisectors of one another, and by appeal-
ing to that theorem we can complete our present argument. The segment BF is
one of the diagonals of our rhombus QF TB, and line QT coincides with the other
diagonal. Thus, according to the theorem mentioned, line QT’ must be the per-
pendicular bisector of BF. As we explained earlier, this fact establishes the equality
of segments SB and SF and completes our proof that the line QT touches our
given parabola only at the desired point T.
‘The three-step procedure devised by the Greeks for constructing tangents to
a parabola will not work with other types of curves for it is based upon special
geometrical properties possessed by the parabola alone. A different procedure
will have to be followed if we want to construct tangents to another kind of curve.
For example, consider the oval-shaped curve known in geometry as the ellipse.
This curve consists of all points in the plane the sum of whose distances from two
fixed points is equal to a constant.
P
Length(PF,) + length(PF2) = constant2| The Problem of Tangents 19
Each of the fixed points is called a focus of the ellipse. The line which passes
through these foci is known as the major azis of the ellipse, and the perpendicular
bisector of the segment which joins the two foci is known as the minor axis.
It is easy to describe a method for constructing tangents to an ellipse, though
of course it is a method different from those used for the circle or for the parabola.
Suppose we let 7’ be a point on the ellipse where we desire to draw a tangent line,
and let #, and F2 denote the two foci of the ellipse. The tangent line through
Tis then constructed in three steps.
ey
First, draw lines from T to each focus. Next, construct the angle bisector
of the angle 4F;TF2. Finally, erect a perpendicular to this angle bisector at the
point 7. It is not hard to show that this perpendicular will be the desired tan-
gent line.
For suppose that S denotes any point different from T which lies on this
constructed perpendicular. We prove that S cannot lie on the ellipse by reasoning
as follows. Construct the circle with center at T which passes through the nearest
focus, say F'2. Then extend the line segment FT until it intersects that circle in
a point, call it Z, which lies outside the ellipse. The triangle formed by the points
T, E, and Fe will be isosceles since the sides TE and TF» are radii of the same
circle, and its apex angle »’ “circle,” and so forth. Students who have nev
studied these ideas should postpone reading Le i
; fi ‘ ied
Appendix I, where the essential concepts of analytic cues ta ne in
detail, Students who are already famili nian ey
proceed to Lecture 3, iar with analytic geometry may sald!
ASSIGNMENT 2
1. Prove th i
hat the dingonals of a thombus are Perpendicular bisectors of each other.
2. Prove y pail i
rove that any pair of tangent lines to @ parabola must have a point in common
yin "3 | Descartes’ Method of Equal Roots 21
3. The center of a certain circle is also the focus of a certain parabola. Prove that if
the parabola intersects the circle in two diametrically opposite points, then the
circle will be tangent to the parabola’s directrix.
4, Let P and Q be two points on a parabola. Prove that if the line PQ passes through
the parabola’s focus, then the tangent line at P will be perpendicular to the tangent
line at Q.
5. Prove that each line parallel to a parabola’s axis intersects that parabola in one and
only one point.
6, An hyperbola is defined as a plane curve consisting of all points the difference of whose
distances from two fixed points is equal to a constant. The two fixed points are
called the foci of the hyperbola. Prove that if T denotes a point lying on an hyperbola,
and if F and F2 denote the foci of that hyperbola, then the angle bisector of the
angle 2 FTF? will be the tangent line to the hyperbola at the point T.
Length (7F,)—length (7F,)=constant
7. Prove that if an ellipse and an hyperbola have the same two foci, then at any point
where the two curves intersect, their tangent lines will be perpendicular.
LECTURE 3
DESCARTES’ METHOD OF EQUAL ROOTS
In our last lecture we examined various methods for constructing tangent lines
to certain plane curves. But as we remarked at the end of that discussion, it was
not until after the invention of analytic geometry early in the seventeenth century
that any real progress was made toward achieving a unified approach to the
Problem of tangents. Essentially, what analytic geometry provided was a means
of fusing ordinary geometry with algebra so that problems in one domain could
be translated into corresponding problems in the other. The basis for this trans-22 Introduction to the Derivative
lation was the identification of points with ordered pairs of numbers, and of lines
or curves with appropriate algebraic equations. The exact details of this transla-
tion process are described in Appendix J, and we will assume that the student is
familiar with them.
One of the most basic constructions in high school geometry is the erection
of a perpendicular to a given line through a given point. When this construction
problem is expressed algebraically, the point would of course be given as an ordered
pair of numbers, and the line as a first-degree equation.
Example
Consider the line represented by the equation
6x — 2y +12 =0
and the point represented by the ordered pair (1, 2).
In analytic geometry finding i i i
eee a ee
In determining this equation, we should keep in mind three conditions that
must be fulfilled. Since it is supposed io represent a straight line, the equation
must be of first dagree in the two variables x and y. Since itis supposed to pass
: te 2), this equation must be “satisfied” when the values
ee on ea Finally, since it should be perpendicular to the
o the ani ae ce — epee be equal to the negative reciprocal
Sie we a hee 0) and (0, 6) clearly lie on the graph of the given
rievetore | ae eh , as may be checked easily by direct substitution.
oo : ue is given line may be computed by dividing the difference
nd coordinates of the points by the difference in their first coordinates:
6-0
Slope =3 | Descartes’ Method of Equal Roots 23.
An alternative way to determine this slope would be to “‘solve” the equation of the
line for y, since when this is done the slope will always appear as the coefficient of x:
y = 3x+6.
The negative reciprocal of this slope is, of course, —4.
So we are seeking a first-degree equation having slope —4 which is satisfied
by the number pair (1, 2). Such an equation can be written at once:
y—2=(-)@ - 1).
Gr—2yt12=0
y-2=—ie-1)
Now, having seen how we can handle a fairly simple kind of geometrical con-
struction algebraically, let us turn to some more sophisticated examples of the
use of analytic geometry. For example, let us consider how we would handle some
of the tangent constructions that we developed in Lecture 2.
Example
The equation x? + y? = 25 represents the circle of radius 5 centered at the origin,
and the ordered pair (—3, 4) represents a point which lies on this circle, since the
pair of numbers satisfies the circle’s equation.
Find the equation of the tangent line to the circle at this point.24 Introduction to the Derivative
From our previous study of tangent lines to circles, we know that this tangent
must be perpendicular to the line segment joining the given point (—3, 4) to the
circle’s center (0,0). But this segment’s slope is equal to —3, so the tangent line
will have a slope equal to the negative reciprocal of —3, namely 2. Once we know
a point the line passes through, and its slope, we can immediately write an equation
for the line:
y-4= fet).
‘This completes the construction.
In Lecture 2 we described the procedure used by the Greeks in constructing
tangents to a parabola. It should be possible to carry out this same construction
analytically. Suppose, for example, that we consider the parabola determined by
taking the point (0, 4) as the focus and the line y = —} as the directrix, since it
turns out that this particular parabola has an especially simple equation.
yoo} bari ye =
| (3)
To determine this parabola’s equation, we recall that according to the definition
of a parabola, exactly those points (x,y) will lie on it which are equally distant
from the focus (0,4) and the directrix y = -}. This condition is expressed by
the equation
+ - 3)? = |y - (-2)-
When both sides are squared, and the equation is solved for x, we obtain a simpler
version of this parabola’s equation, namely
Vix
yor
One of the points lying on thi: i i
ak first drop 4 perpendicular from the point of tengency (2, 4) to the parab-
2 Since our parabola’s axis happens, in this case, to coincide with the
asy to tell what the foot of this perpendicular will be, namely (0, 4),
ing 4 units above the vertex. The corresponding point 4 units below the
vertex is the point (0, — is ii
a ne (0, —4), and this, of course, is where the tangent line must3 | Descartes’ Method of Equal Roots 25
Knowing that the tangent line must pass through the two points (2, 4) and (0, —4),
we can calculate its slope:
4—(-4)
Slope = =4,
2-0
and then immediately write its equation:
y—4 = 4 — 2).
This finishes the construction.
Now that we have illustrated the use of analytic geometry in handling geo-
metrical constructions algebraically, we are ready to describe a method for finding
the equations of tangent lines to various kinds of curves. This method was de-
veloped by René Descartes, who is usually credited as the founder of analytic
geometry. We have just seen how tangent equations may be determined in the
case of a circle and in the case of a parabola. However, we used a different method
in each of these cases. What Descartes sought to devise was a uniform, or stan-
dardized procedure by which these tangent equations could be found, regardless
of the type of curve in question. Descartes’ procedure is known today as “the
method of equal roots.” Let us describe the method, and see to what extent it is
successful in resolving the general problem of tangents.
For simplicity, let us demonstrate Descartes’ method of equal roots first on
a parabola. In fact, if we keep the same parabola that we used in the example
above, y = x2, and the same point of tangency, (2, 4), then we will be able to
check thé result we get using Descartes’ method against the result we found
using the Greek procedure.
‘We begin by considering a typical line passing through the point (2, 4). This
line will have an equation of the form
y-—4 =m — 2),Introduction to the Derivative
ts
5
where the letter m stands for the line’s slope. Normally, we expect such a line
to intersect the parabola in two different points as shown in the graph.
yem(z—2)
In order to determine the coordinates (x, ¥) of the points of intersection of the
line with the parabola, we would “solve” their equations simultaneously:
=x,
= m(x — 2).
Solving two such equations simultaneously for x and y is a process very fam-
jliar to us from high school algebra. We would normally proceed by substituting
the first equation into the second in order to eliminate the variable y, getting
x -—4=m(x — 2),
or equivalently, writing all terms on the left-hand side,
— mx + (2m — 4) = 0.
This is just an ordinary quadratic equation in x, and so we may solve it by using
the “quadratic formula” well known from high school. According to the quadratic
formula, the two roots of this quadratic equation are given by
—(-m) + Vm)? — 4M Gm — 4)
2(1) :
x=
or more simply,
m+ Vm? — 8m + 16
—
x
or course, the plus or minus sign here reflects our expectation that the parabola
and the line will have two different points in common, the first coordinates of those
{wo points being given by the two solutions shown for x. However, it is possible
that the radical may be equal to zero.
Vm? ~ 8m + 16 = 0.
: ae happens, we say that the two roots of our quadratic equation are equal.
in this case there would be only one simultaneous solution to the equations of3 | Descartes’ Method of Equal Roots 27
the line and the parabola. Geometrically, this would mean that the line meets
the parabola in only one point.
Descartes realized that when the equations for the line and the parabola
have only one simultaneous solution, namely (2, 4), then the line would be tangent
to the parabola! By choosing the line’s slope m in such a way that the two roots
of the quadratic equation are made equal Descartes knew he would have found
the slope of the tangent line!
Obviously the two roots of our quadratic equation will be equal just in case
the expression under the radical sign equals zero:
m? — 8m + 16 = 0,
or, in other words, just in case m is chosen so that m = 4. Therefore, by choosing
m = 4, we get the equation of the parabola’s tangent line at the point (2, 4):
y—4 = Ae ~ 2).
Note that this answer agrees exactly with our earlier result obtained. by using
the Greek construction, and this fact should convince us that the method of equal
roots is sound.
The importance of Descartes’ method of equal roots is that it also works with
curves other than the parabola. We can use the same method to obtain an equa-
tion for the tangent line to a circle at a given point. For comparative purposes,
let us show this by looking again at the circle represented by the equation x® +
y® = 25, and let us redetermine its tangent line through the point (—3, 4), this
time, however, by using Descartes’ method. -
As before, we begin with the equation for a typical line passing through the
point (~8, 4):
y~4 = me 4+ 3).
Again, the letter m stands for the line’s slope. Generally speaking, such a line
will intersect our circle in two different points, and therefore the equations for the
line and circle will generally have two simultaneous solutions: ‘
25,
m(x + 3).
ety
y-4
wa
y
(-3,4)
y—4=m(243)
r23 Introduction to the Derivative
To colve the two equations simultaneously for x and y, we would proceed by
substituting the second equation, after solving it for y, into the first equation,
thereby eliminating y and getting
a? + [m(x + 3) + 4P = 25
or, after expanding the left-hand side and collecting like terms,
(1 + m2)x? + (6m? + 8m)x + (Om? + 24m — 9) =
‘This is a quadratic equation in the variable x, and its roots are obtained by apply-
ing the quadratic formula:
= Gm" + 8m) + Gm? + 8m)? — 40 + m?) Om? + 24m — 9)
20 + m2) '
or more simply,
_ ~ 8m? + dm) & VI16m? = 24m + 9
(+ om?) :
These two roots will be equal only if the radical is zero:
V16m? — 24m +9 = 0.
Since 16m? — 24m +9 = (4m — 3)", we see that the two roots will be equal
only if mt is chosen such that
With this choice for m, we will get an equation for the tangent line to our circle
x? + y? = 25 at the point (—3, 4).
y—4 = 3 + 3).
Observe that this is the same it
2 answer we obtained earli i
problem by another meno obtained earlier when we solved this
Descartes’ method of e i idi
of equal roots succeeds in providing a single, standardized
a for determining the equations of tangent lines to the circle, to the
m trabo eo indeed to any curve which, like the circle and parabola, can be repre-
se Pree esate equation. In this sense, Descartes’ method serves to
a 7 oe of rates at least partially. But there are still many curves
i s' method cannot be applied. As a typical example of where
artes’ method fails, w i i
pageariee mcthod fis we can consider the curve represented by the third-3 |. Descartes’ Method of Equal Roots 29
Let us see what happens if we try to use Descartes’ method to find the tangent
Tine to this curve at, say, the point (1, 1).
A typical line through the point (1, 1) will have an equation of the form
yrl=me-,
where m stands for the line’s slope. According to Descartes’ method, we will want
to select the value of m so that the two equations
y= x,
y-1l=m@—-1),
will have only one simultaneous solution, namely the number pair (1, 1).
Without bothering to actually try and solve these two equations, let us simply
think about what happens graphically. There will be many different lines through
the point (1, 1) which intersect the-curve y = x? only once, and yet none of these
will be the tangent line we are seeking! This is clear from looking at the graph,
where we see that the actual tangent line must touch the curve twice!
Many lines cross the curve only once but. the
tangent line touches the curve twice30 Introduction to the Derivative
Clearly in this situation it would be futile to try and discover the tangent line’s
slope by using Descartes’ method of equal roots, since any value of m which
Descartes’ method would yield will obviously be the slope of a nontangent!
Thus, Descartes’ method fails utterly when applied to the curve represented
by the equation y = x*. Asa matter of fact, the method fails for the same reasons
with all curves described by third-degree equations, and with a multitude of
others as well. Indeed, whenever the tangent line intersects the curve in more
than one point, it is obvious that we will be unable to determine that tangent’s
slope by applying the method of equal roots.
Incidentally, we should point out that one of the curves which caused us to
investigate tangents in the first place, namely the graph which accompanied the
box problem in Lecture 1, was represented by a third-degree equation:
Vix) = © — 2x)(4 — Qx)x = 20x — 16x? + 4x3.
So Descartes’ method would be of no help there.
In its day, Descartes’ method of equal roots represented a genuine step for-
ward in the search for a uniform method of drawing tangents to curves, but despite
its partial success, it fell short of the desired goal. In fact, its limitations are so
severe that the method is almost worthless to us today, except perhaps as an in-
teresting mathematical museum piece.
Yet despite Descartes’ failure to completely solve the problem of tangents,
his preliminary work on analytic geometry was brilliant. His ideas were soon put
to use by another seventeenth-century mathematician, Pierre Fermat, whose
approach to the problem of tangents was far more successful, as we shall see.
ASSIGNMENT 3
1. A circle centered at the point (—5, —1) passes through the point (0,11). Find the
equation of the line which is tangent to this circle at the point (0, 11).
2, A circle centered at the point (2, —3) is tangent to the line represented by the equa-
tion » = 4, Find an equation for this circle.
- Fina an cawation for the circle centered at the point (0,0) and tangent to the line
bx By ~ 6 = 0.
» Write un equation for the parabola whose vertex is the point (2,4) and whose di-
reetrix is the line y = 7.
- The point (3, 10) is the focus of a certain parabola, and the line y = —4 is the di«
rectrix of that parabola, Find the coordinates of the parabola’s vertex, and find an
equation for the parabola’s axis of symmetry.
6.
ind an equation for the
ie the Hee gon forthe parabola whose focus is the point (1, —8) and whose directrix3 | Descartes’ Method of Equal Roots 31
7. A parabola has as its vertex the point (h, #) and has as its directrix the line y =
h — p. Prove that the coordinates of every point (x,y) on the parabola must satisfy _
the equation
y a @ —hy +k.
8. Show that any equation of the form
y = Ax? + Bx +C,
where A + 0, represents a parabola whose focus is the point
(—B/2A,C — B?/4A + 1/44),
and whose directrix is the line
y= C- BAA — 1/4A.
9. Use Descartes’ method of equal roots to find the slope of the line tangent to the
parabola y = x2 at the point (3, 9). Write this line’s equation and sketch a graph
showing the line and the parabola. :
10. Use Descartes’ method of equal roots to find the slope of the line tangent to the
parabola y = x? at the point (—2, 4). Write this line’s equation and sketch a graph
of this line and parabola.
11. Use Descartes’ method of equal roots to find the slope of the tangent line to the
circle x2 + y? = 25 at the point (4, —3). Write this line’s equation and sketch a
graph of the line and circle.
12. Use Descartes’ method of equal roots to find the slope of the tangent line to the
curve » = 1/x at the point (3, 4). Write this line's equation and sketch a graph
of the line and curve.22 Introduction to the Derivative
LECTURE 4
FERMAT’S METHOD OF LIMITS
As we saw in our last lecture, Descartes’ so-called “method of equal roots” pro-
vided a uniform procedure for determining the equations of tangent lines to a
small number of different curves, including the circle and the parabola, but failed
to work in general. In particular, the method cannot be used with important
kinds of curves that arise in practical problems, such as the box problem discussed
in Lecture 1. Consequently, for purposes of the calculus, the method must be
discarded.
Fortunately, one of Descartes’ contemporaries soon took a fresh look at the
problem of tangents. This remarkable individual was Pierre Fermat, who because
of his own work in analytic geometry, number theory, and probability, has been
called the greatest French mathematician of the seventeenth century. What con-
cerns us here is that Fermat was successful, where Descartes had failed, in solving
the problem of tangents: he found a way to determine the equations of tangent
lines to any type of curve whatever.
Fermat based his method for finding tangents on the following idea. Suppose
a particular curve is given to us, and suppose T' denotes a point on that curve
where we desire to draw a tangent line. Let P denote a nearby point on the same
curve, cither to the right or to the left of T, and consider the line which passes ©
through these two points P and T.
Such a line is called a secant Line, following the terminology used in high school
geometry for a line which crosses a circle at two different places.
Now imagine that the point P is allowed to move along the curve toward
the point T. What do you think will happen to the secant line PT? Fermat
realized that as the point. P approached the point T, whether from the right or.
from the Ieft, the secant line PT would be moving into the tangent position.
‘The secant Hine approaches the tangent position4 | Fermat's Method of Limits 33
By itself, Fermat’s idea is not especially helpful, but allied with the tools of
Descartes’ analytic geometry, it provides the key to the problem of tangents.
To see exactly how this is possible, let us reconsider one of our earlier examples,
the parabola y = x”, and Jet us see how to determine the tangent line to this curve
at the point (2, 4) this time making use of Fermat’s idea.
Let T denote the desired point of tangency (2, 4), and let P denote a nearby
point on the parabola having coordinates (w, v).
We are looking for the tangent line to this curve through T, and we know its
equation can be written in the form
y—4=m(z —- 2),
where the letter m stands for the tangent line’s slope. All we need to do, then, is ©
determine this slope m. So let us consider the secant line PT, and try to see what
happens to its slope as the point P moves along the curve toward the point T.
v-4
w-2
Slope(PT) =
Because the point P lies on the parabola, its coordinates (w, v) will have to
satisfy the parabola’s equation y = x*. This means that v = w®, and so we may
rewrite the formula for slope(PT) as follows, eliminating the variable v, and
simplifying the resulting expression:
— 2)
Slope(PT) = =w+2
As the point P moves along the curve toward the point T, so that its first co-
ordinate w is approaching the value 2, the slope of the secant line PT’ will be
approaching the value 4. Therefore, we conclude that the slope m of our tangent
line must be equal to 4 because 4 is the limiting value, as w approaches 2, of the
formula for the secant line’s slope:
Slope(PT) = w + 2.34 Introduction to the Derivative
We cannot help remarking that this reasoning leads us to the very same tangent-
line equation that we got earlier using the method of equal roots:
y~4=4(x — 2).
In order to express more concisely the ideas used in this example, mathe-
maticians often employ the following shorthand notation to refer to the limiting
value of a formula:
a Ww +2) = 4.
This equation is read as follows: “the limiting value, as w approaches 2, of the
expression w + 2, is equal to 4.” It is understood that w may approach 2 from
either direction along the coordinate axis.
Now let us illustrate Fermat’s method in another example by finding the
tangent line to a circle at some point. In order to make a comparison, let us look
once again at the problem of this type which we considered in our last lecture,
namely, finding the tangent line 1o the circle x? + y? = 25 at the point (—3, 4).
We proceed as before, letting T denote the point (—3, 4) which is to be the point
of tangency, and letting P denote a nearby point on the circle having coordinates
(tw, v}.
y
PpypPars
To caleulate the slope of the tangent line, which is all we need to do in order to
write its equation, we must try to determine the limiting value of the slope of
the secant line PT, as P approaches T:
limit
Limit slope(PT),
where slope(PT) is of course calculated according to the formula
Slope(PT) = -—44 | Fermat's Method of Limits 35
Keeping in mind that the point P must move along the circle as it approaches T,
so that its coordinates (w, v) will always satisfy the circle’s equation x? + y? = 25,
we can make the substitution
v= V25 — w?,
thereby eliminating the variable v from our slope formula:
V25 — w? — 4
Slope(PT’) = SS
Our problem, then, is to determine the limiting value of this formula, as the value
of w approaches —3.
limit slope(PT) = limit ee.
Of course, as this equation stands now, it is very difficult to tell what the value
of the limit will be, because as w approaches ~3, both the numerator and the
denominator are approaching zero. But we can get around this difficulty if we
“cationalize” the fraction’s numerator, multiplying it by the conjugate expression
(V25 — w? + 4)
so as to eliminate the radical sign from the numerator, and of course multiplying
by the same factor in the denominator to keep things equal. This gives
0/25 — w? — 4) 0/95 — w? + 4)
(w + 3) (25 — w? + 4)
_ (25 — w?) ~ 16
© w+ 3)0/25 — wu? + 4)
a 9-w*
© w+ 80/5 — WF + 4)
___@-w@+w)
© w+ 2/25 — WF + 4)
= 3-w .
(/2 — uw? + 4)
When the secant line’s slope is re-expressed in this new form, it becomes very easy
for us to see what the limiting value will be.
Slope(PT) =
Oa We
Timit al (PT
slope(PT) = oes a)
3 — (—3) 6 3
(Ve — (-3 +4) 8436 Introduction to the Derivative
We conclude that the slope of the tangent line to the circle x ty’ 25 at the
point (—3, 4) must be j, and hence the equation for this tangent line will be
y—4=F@ +3).
‘This answer agrees exactly with the answer we obtained using different methods
in our last lecture.
Having gained some confidence in the correctness of Fermat’s method from
the preceding two examples, let us iry to apply it to the problem we were unable
to solve before, the problem of finding the tangent line to the graph of the third-
degree equation y = x* at the point (1, 1).
Once again we let T denote the point of tangency (1, 1), and we Jet P denote
some other point on the curve having coordinates, say, (w, v). We will try to
find the limit of the slope of the secant line PT, as the point P approaches the
point T.
Because the point P moves along the curve y = x3, its coordinates (w, v) must
satisfy the curve’s equation. Thus, the slope of the secant line PT’ will be given
by the formula
= Ua Dw? twt)
i w—1
Taking the limit as P approaches 7, we get
Slope(PT) = . agen race
limit slope(PT) = limit w+wt+i)=P+1i+1-
We conclude from this calculation that i
thi i
hence that its equation will be ® fangent Tine’s slope must be 9 and
y~1= 3-1.
Thus, Fi . : :
te ermat’s method for finding tangent lines works even where earlier methods
We have now seen, in three i
Ly instances, how to use Fermat’s method to cal-
Brats slope of the tangent line to a curve at a particular point. But there are
Tana Lae ee Points on any given curve, and naturally we would expect the
S ine to have a different slope at each point. For instance, we found that4 j Fermat's Method of Limits 37
the slope of the tangent line to the curve y = x at the point (2,4) was equal to
4, but a tangent line to this same curve at a different point would have another
slope.
Example
Find the slope of the tangent line to the curve y = x7 at the point (—3, 9).
r- 9
we (= 3)
r 29 (ut t 3 ee
s Himit = limit SE Me =D
Himit slope(PT) ~ limit
Ww = 3) © -3-
The algebraic steps involved in this example are almost identical to those
used with the curve y= x7 previously, and this similarity suggests that we may
be able to apply Fermat's method in a way that would give more comprehensive
information about this curve. Let us see if it would be possible to determine a
mathematical principle or formula which would predict the value that the tangent
slope will have at various points along the curve.
With this aim in mind, suppose we let (x, 9°) represent. the coordinates of an
arbitrary point 7’ lying on the curve y = x*, and let (w, v) be the coordinates of a
nearby point P also lying on the curv48 Introduction to the Derivative
Since it is clear that the tangent’s slope will depend upon which point T we have
in mind, let us use the notation n(x) (pronounced “‘m of x”) to indicate the slope of
the tangent line at that point on the curve whose first coordinate is x. Then,
letting the point P move along the path of the curve toward the point 7, we can
calculate the value of our tangent’s slope m(x) as follows:
a es
mm(z) = limit slope(PT) = limit
= limit! 2 = timit = 2D +
Tn WE tre —x
i
limit (w + x) = x +x = 2x.
What we have obtained here is a slope-predicting formula for tangents to the
curve y = x".
Note that this formula m(x) = 2x would have predicted the tangent slopes
in our two previous examples with this curve. In one case, the point of tangency
was (2, 4) and the tangent’s slope was
m(2) =2-2=4,
In the other case, the point of tangency was (—3, 9) and the slope of the tangent
was
m(—8) = 2(-3) = -6.
By using this slope-predicting formula, we can tell immediately what the slope of a
tangent to our curve y = x? will be at any point we happen to be interested in.
Examples
What is the slope of the tangent line to the curve y = x2 at each of the follow-
ing points?
a) point (13,169): m(13) = 2-13 = 26,
b) point, = (4,1): ml) = 2-1
¢) point (-7, 49): m(—7) = 2-(-7) = -14.
da) point — 0,0): m(0) = 2-0 = 0.
Certainly we expect that the slope-predicting formula m/(x) = 2x will only
apply to the particular curve y = x, and that a different curve, such as y = x%,
will have a different slope-predicting formula associated with it.
Le us see whether we can derive the slope-predicting formula for the curve
aa ae use the rotation M (x) to stand for the slope of the tangent line to
Q ‘ gree curve at the point whose first coordinate is x. The steps are the
same as before. We let (x, y) represent the coordinates of the stationary point T,
and let (tr, v) be the coordinates of the nearby point P which moves toward T.4 | Fermat’s Method of Limits 39
Then we can calculate the value of M(x) as follows:
M(x) = limit slope(PT) = limit 2
woz WX
aii = tig Ae? + wx + 2”)
Sens WH eae wx
= limit (w? + wx + x”) = (x? + x? + x*) = 82”.
war
Again, by using this slope-predicting formula for the curve y = x°,
M(x) = 3x”,
we can quickly tell what the tangent’s slope will be at any point on the curve.
For example, at the point (—2, —8), the slope will be M(—2) = 3(—2)? =
at the point (0, 0), the slope will be (0) = 0; at the point (, }), the slope will
and so on. Note that this curve y = ao never has a tan-
gent line with a negative slope.
Knowing the slope-predicting formula for a curve saves time and work, since
we do not have to stop and recalculate the limit of the slopes of the secant lines
each time we want to know the tangent’s slope at a particular point. Instead, we
can simply substitute the value of that point’s first coordinate into our slope-
predicting formula and obtain the slope immediately. For this reason, among
others, the study of slope-predicting formulas will be one of the central concerns
of our calculus course.
ASSIGNMENT 4
1. Using Fermat’s method, find the slope of the line tangent to the given curve at the
given point; then write the tangent line’s equation.
a) The circle x? + y? = 25 at the point (3, 4)
b) The circle x” + y? = 25 at the point (3, —4)
c) The curve y — x? at the point (1, 0)
4) The curve y = x? ~ 4x3 at the point (1, —3)
q40 Introduction to the Derivative
Find the equation of the line tangent to the curve y = x5 at the point (—1, -1);
then find the coordinates of the second point where this tangent line crosses the
curve, Sketch a graph showing the curve and tangent line.
4 Jat the point (1, 2).
Find the equation of the line tangent to the curve y =
Sketch a graph showing this curve and its tangent line.
|. Using Fermat's method, show that the line tangent to the parabola y = x? — 2x
at the point (1, —1) is a horizontal line. Sketch a graph showing this parabola and
tangent line.
5. Derive the slope-predicting formula for each of the following curves.
ay ext +5 b) y =x? +28
ox ad) y = Ve
oy ees
2
. The normal line to a curve at a particular point on that curve is the straight line
which lies at right angles to the curve’s tangent line at that point.
For each of the following curves, find the equation of the normal line to the curve
at the given point,
n) Normal line to curve y = x! at point (—2, 16)
1
b) Normal line to curve y = = at point (3, 3)
c) Normal line to curve y = + 3 at point (7, 52)
+ x3 at point (—1, 0)
. A parabola is tangent to the line 83x — y — 6 = 0 at the point (0,6) and passes
through the point (1,0). Find this parabola’s equation, assuming it has the form
y= Ax? + Be $C,
where A, B, and C represent constants to be determined.
a) Normal line to curve y =
»
. Pind the equations of the lines passing throuzh the point (0, —4) which are tangent
to the parahola y = x?.
. Determine the limiting values of the indicated formulas:
3
b) limit 728
5 wp Wd
©) limit (x* 4 5x? + xy a mit #2
2 zal cee
10. Find the indicated limits:
timit 2/10
x0 1/x + Tx
limit @
0 \t,
b)
dqIntroduction to the Derivative
s
8
to stand for the slope of the tangent line at the point on the graph of the volume
formula whose first coordinate is equal to x. That point, of course, has coordinates
(x, Vix)), and a nearby point will have coordinates of the form (w, V(w)).
As usual, the slope V’(x) of the tangent line at (x, V(x)) is obtained by taking
the limit, as w approaches x, of the slope of the secant line through (x, V())
and (w, V(w)).
V(x) = limit slope(PT) = limip Y= VO |
Poa? ease
The formula for the slope of the secant line may be considerably simplified by
using elementary principles of algebra:
Slope(PT) = Yeh ve)
_ (20w — 18? + 4w*) — (20x — 18x? + 4x°)
Fear ines
_ 20(w — x) — 18(w? — x”) + 4(w® — 2°)
w-x
= 20(w — x) — 18(w ~ zYw + 2) + 4 — x)(w? + we + 2°)
w-Kx
20 — 18(w + x) + 4(w? + wx + 2%),
Having simplified our formula for slope(PT'), we can calculate V’(x) very quickly.
V'(x) = limit slope(PT’)
Pot
= limit [20 - 18(w + x) + 40” + wx + x”)]
20 — 18(x + x) + 402? + x? + 2?)
20 — 36x + 12x7,
By using the slope-predicting formula which we have just derived, we can
complete the box problem which we had left unsolved in Lecture 1. We were
‘ying to determine which value of x would cause the value of the box’s volume
i& | Solution of the Optimization Problems 43
V(x) to reach its maximum. This meant we were trying to calculate the coordinates
of the highest point on the graph of the volume formula, V(x) = 20x — 18x? +
4x3, All we knew about the highest point was that the tangent line there should
have a slope equal to zero.
But now that we have a formula which gives the value of the tangent line’s
slope at every point, namely
V'(x) = 20 — 86x + 122,
we can find out which value of x will cause the slope V‘(x) to be zero by simply
solving the equation V'(x) = 0 for x:
20 ~ 36x + 12x? = 0.
This equation may be solved by using the quadratic formula:
x= 36 + Vv (36)? — 4(12)(20) ©
242)
Simplifying, we get.
: 92 V2
oo
6
Our equation V'(x) = 0 accordingly has two different solutions, namely
9 vai _9_ Va
Sop tge fe 2 ages
However, only one of these two solutions is physically significant, because,
as we recall, in the context of the box problem the variable x was restricted by
the condition that it must have a value between 0 ft and 2 ft.
O0.
This quadratic expression may of course be factored,
x— 2 +6 = (3 — x)(2 +2),
and we see that x must be chosen to satisfy the condition
(8 —x)(24+2)>0.
ow we know that a product of two numbers will be positive if and only if
h factors have the same sign. Either both factors must be positive,
38-x>0 and 2+x>0,
or both factors must be negative,
38-x<0 and 2+x<0.
Tn the first of these cases, the two inequalities may be rewritten as an equivalent
pair of conditions:
38>x and x> 2,
or, still more simply, as a single continued inequality:
—2<2<3.
In the second case, the two inequalities are equivalent to the expressions9 | Domain and Range of a Function 81
dust as there is a special mathematical name given to the set of permissible
values for a function’s independent variable, so there is a special name for the set
of values assumed by the dependent variable. This latter set of numbers is called
the function’s range.
Example
Determine the domain and range of the function
yol+vx-2.
Following the rule of maximum domain, we presume that this function’s
domain consists of all x-values satisfying the inequality x > 2, since this condition
is both necessary and sufficient to assure that the expression under the radical sign
will be nonnegative. The range of this function, then, will consist of all y-values
which satisfy the inequality y > 1. A graph of this function, with its domain and
range marked off on the x-axis and y-axis respectively, is shown following.
Domain
The numbers which belong to the domain of a function are often called argu-
ments of the function, while the numbers which belong to the range are called
values of the function. For instance, in the example above, the number 1 is a value
of the function since it is included among the numbers in the range, but it is not an
argument of the function because it does not belong to the function’s domain.
The number 2, on the other hand, is both an argument and a value, while the
number 0 is neither an argument nor a value.
In most examples of functions that we deal with in the calculus, the domain
and range are made up of intervals of numbers. An interval may be either finite
or infinite in length, and may either include or exclude its endpoints. A finite
interval which includes both its endpoints is called a closed interval. For example,
the numbers satisfying the inequality 0 < x < 2 form a closed interval, which
we denote by writing [0, 2]. A finite interval which excludes both its endpoints is
known as an open interval. The numbers which satisfy the inequality —2 <
= < +8 form an example of an open interval, which is denoted by (—2, 3). Thus,
square brackets are used to indicate inclusion of endpoints while ordinary paren-
theses are used to indicate that the endpoints are excluded.
é
a
Closed interval [a,b) Open interval (a,b)£2 Introduction to the Derivative
Mixed cases can also occur. An interval may include its left-hand endpoint, but
exclude its right-hand endpoint, or vice versa. Such intervals are called half open
(or half closed). For example, the symbol (3, 7] denotes the half-open interval con-
sisting of all numbers which satisfy 3 < x < 7.
This same system of symbolism is easily adapted to denote infinite intervals
as well. For example, the interval consisting of all positive numbers is written
(0, +). Of course, the symbol + does not stand for a number, but merely
serves to indicate that the interval starts from zero and includes all numbers to
the right. This interval is open since its only endpoint, zero, is not included. In
the same fashion the interval of all nonpositive numbers is written (— ©, 0]. It is
a closed interval because its endpoint zero is included. The interval of all numbers
is written (-%, +). The notation [— ©, +] is never used, because ~ ©
and -+ # are not proper numbers.
Example
Consider the function we used in studying the pasture problem in Lecture 5.
‘That fanction was described by the formula
A = 500x ~
where x was subject to the physical restrictions given in the inequality 0 < x < 500.
In this example, the independent variable is x and the dependent variable is A.
The domain, therefore, is the interval [0, 500], and since we saw that the greatest
possible area for the pasture is 62,500 ft?, the range for this function will be the
interval {0, 62,500}. A graph for this function showing the domain and range is
given. :
62,500
A=5002-2"
Range
Domain 500
Certain algebraic formulas describe functions whose domains and ranges are
composed of two or more separate intervals. The formula
yexth
=
is an example of this type. Its domain consists of all numbers x 0, and may
be written as the union of two intervals, as follows:
Domain = (— 2,0) U (0, +).9 | Domain and Range of a Function 83
The range for this function, although not quite so easily determined, also turns
out to be a union of two separate intervals.
Range = (—, —2] U[+2, +o).
The graph for this function follows.
Frequently, the derivative of the function is a helpful tool in determining the
function’s range, since it can be used to locate the “high” and “low” points on a
function’s graph. For instance, in our example above, where y = x + 1/x, the
formula for the derivative y’(x) will be
y'@) =1- G/x),
and we will have »’(x) = 0 if and only if x = +1. This determines the x-coor-
dinates of the graph’s “high” and “low” points. The corresponding y-coordinates
are, of course, y = <2, and they represent the extremities of the function’s range.
ASSIGNMENT 9
1. Determine the domain of each of the following functions:
b) y =8x—7
ad) y = V8 — bx
Y= GaHeFH
2. Determine the domain of each of the following functions:
a) y = Vi00 — a2 by = @+D/@ -1
oy =1/(@? +5) a y = Gr t+ D/@? = og4
o
2
a
2
2
10.
Introduction to the Derivative
Determine the domain and range of each function:
a)y = V9x—5 b) y = 1/@" +11)
dy =@- 2/8 - =) a) y = V49 — x?
|. Determine the domain and range of each function:
x
a) y = VE by= Fel
aL Sad
yaar +s a) y = 27a" +>
. Water is running into a cylindrical cistern whose base is a circle of radius 9 ft and
whose height is 21 ft. Express the volume of water in the cistern as a function of the
water's depth and determine this function’s domain and range.
. ‘The area of a proposed city park will be 1000 yd? laid out in the shape of a rectangle.
Express the perimeter of the park as a function of the length of its longer side, and
determine this function's domain and range.
A rectangular garden will be built next to a long brick wall, and will be enclosed on
three sides by a picket fence. If the perimeter of the picket fence is 100 ft, express
the aren of the garden as a function of the length of its brick-wall side, and determine
this function’s domain and range.
. A rectangular office building is to be erected on a triangular lot whose sides are 60 ft,
80 ft, and 100 ft, respectively. The building fronts on the lot’s longest side, as shown
in the accompanying plan, and is subject to a height limit of 40 ft.
L hater —P,
100 ft
Assuming that the height of the building will be kept as low as possible, and that the
building will contain 17,280 ft? of space inside, express the floor area of the building’s
main level as a function of the width of the building’s front, and determine the domain
and range of this function.
. Determine the domain of each function:
1
af) = = = Va? — ae
ae b) g(x) = V(x? — 16)(64 — x?)
©) AG) = VB @ j(@) = 5/x — x*)
Determine the range of each function:
Qe td Figeaae
a) f(z) = Parent b) a(x) = 4xV49 — 22
1
One) ee +5 d) j@) = Vx3 + 6x10 | Local and Global Extreme Values 85
LECTURE 10
LOCAL AND GLOBAL EXTREME VALUES
In our last lecture and on several earlier occasions, we have been interested in
determining the maximum or minimum value that is assumed by a given variable.
We could describe this kind of problem in general terms as the problem of finding
the extreme values of a function. These extrema correspond, of course, to high
and low points on a function’s graph, and one tool we have seen for locating them
is the function’s derivative.
For example, if the point (a, f(a)) is the highest, or lowest, that occurs on a
given function’s graph, then we expect that the value f’(a) of the derivative at
that point should be equal to zero. It was this feature of the derivative which
enabled us to solve our early examples of optimization problems, and more recently
to determine in certain cases a function’s range.
y y
@,f@))
y=f@)
y=fz)
Tafa)
I
=
a
The numbers which satisfy the equation f’(x) = 0 are called critical numbers for
. the function, and we expect to find the function’s extreme values at the points
where these critical numbers occur. In other words, we expect that if f(a) = 0,
then f(a) should be an extreme value of the function, either a maximum value or a
minimum value.
Unfortunately, this expectation is not always fulfilled; the subject of extreme
values cannot be dismissed so simply. As the following example will show, it is
perfectly possible for the derivative of a function to be zero at a point without
the function’s having its greatest or its smallest value there.
Example
Consider the function f(x) = x3 — 12x. Its critical numbers are found by’ solving
the equation /’ (x) = 0:
£'(@) = 8x? — 12 = B(x + 2)(e — 2) = 0.
Thus, 2 = —2andx = +2 are the critical values of x.86 Introduction to the Derivative
However, neither f(2) nor f(—2) is an extreme value of this function f(x), as
can be seen from a short table of values:
2 3
-9
4/5
65
5 —4|-3| -2| -1])0 1
9
Fa
(8) -16
16} 11 |o| -11 16
65 | —16
We see that the function assumes Jarger values than f(—2) = 16, and smaller
values than f(2) = —16. Actually, this particular function does not have either
‘a greatest value or a smallest value, because its range includes all values. A sketch
showing the main part of this function’s graph appears on the left below.
(2,-16)
| Even though the function f(x) = x3 — 12x does not have any extreme values
in an absolute sense, it is clear from the graph that the function does have extrema
in a relative sense. While f(—2) = 16 is not the largest value of the function if
we look at the entire graph, it is the largest value if we restrict our attention to a
small portion of the graph, and for this reason f(—2) is called a local maximum
of the function. Similarly, /(2) = —16 is the smallest value of f(x) if we only
consider values of x near x = 2, and consequently f(2) is called a local minimum.
If the domain of our function f(x) = x* — 12x were restricted to include only
those values of x which satisfy the inequality —3 < x < +3, then /(—2) would
become the maximum value of this restricted function, and {(2) would be its
minimum value. A graph of the restricted function is given on the right above.
Note that in restricting the function’s domain to the closed interval [—3, 3),
we have created two new “local” extrema. The function value f(—3) is now the
smallest value that the function assumes for values of x near x = —3, while /(3)
is the largest value of f(x) when xisnearx = +3. If we had restricted the function
differently, say to the open interval —3 1, the quantity (x — 1)° will be positive, and
so g(x) will be greater than g(1). On the other hand, when x < 1, then (x — 1)?
will be negative, and thus g(x) will be less than g(1). A table of values confirms
this reasoning:
128
-3 -2
—24
co 8
-5
0
2
1
3
2
4
3
11
4
30
5
67
x
-61
B(x)
As the following graph shows, this function g(x) has ‘values which steadily increase
as x increases, even though the rate of increase, so to speak, is momentarily zero
when x = 1.
yar 32743242
The lesson we learn from this example is that a zero derivative need not imply
the existence of even a local extremum.88 Introduction to the Derivative
Moreover, to make matters worse, it is possible for a function to have an
extremum, either local or global, without the derivative’s being equal to zero at
that point! This happens in the following example.
Example
Consider the function h(x) = x°°, Clearly, h(0) = 0 is a local minimum value—
in fact, a global minimum value—of the function. Yet h’(0) is not equal to zero
because h’(0) does not exist! The function is not differentiable at x = 0. This is
obvious from looking at the formula for the derivative which is given by
oa
3Vx
Clearly, x = 0 is not in the domain of h’(x). A glance at the graph shows that a
“cusp” occurs at the origin and that the tangent line becomes vertical there.
h(a) = 3x78 =
y
yaa
x
The derivative doesn’t exist at the origin
Numbers occurring in the domain of a function, but not in the domain of that
function’s derivative, are known as singular numbers. Thus, the number 0 is a
singular number for the function h(x) = x2,
We sce that, as a tool for locating the high and low points on a function’s
graph, the derivative definitely has its limitations! It does not tell us for sure
where to find the extreme values, or even the local extreme values. What, then,
does the derivative tell us for sure? Roughly speaking, as we shall see, it tells us
where the extreme values aren’t!
Suppose, for example, that f(a) is a positive number. This indicates that the
graph of the function f(x) is rising, as we look from left to right, at the point
where x = a, for the tangent line there has a positive slope, and of course the
graph of the function must “hug” its tangent line at the point of tangency.
y
fla)
If J'(@)>0 then f(a) cannot be
a local extreme value10 | Local and Global Extreme Values 89
Thus, for values of x near x = a, f(x) will be smaller than f(a) when x < a, and
f(x) will be greater than f(a) when x > a. Consequently, f(a) could be neither a
Jocal maximum nor a local minimum.
Similarly, when f’(a) is a negative number, it indicates that the graph of the
function slopes downward at the point where x = a, which means that the func-
tion values are decreasing there as we look from left to right. Thus, if x is near
x = a but on the lower side, then f(x) will be greater than f(a), while as soon as x
is on the upper side of x = a, f(x) will be smaller than f(a).
If f(@)<0 then f(a) cannot be
a local extreme value
Once again, f(a) could not be a local-extremum since it is surrounded by larger
function values on one side, and by smaller function values on the other.
Of course, we should hasten to point out that our reasoning here applies only
when the function’s domain includes numbers on both sides of x = a. It obviously
would not apply if the number @ marked an endpoint of the function’s domain.
In fact, it happens in many cases that local extreme values do occur at the end-
points of a function’s domain, even though the derivative has a positive, or nega-
tive, value there. An example is provided by our previous function f(x) = x? — 12x
restricted to the domain —3 < x < +3. We have already seen that f(—3) and
f(+8) are local extrema in this example, although /’(—3) and f’(+38) are both
positive.
Let us try to summarize our preceding observations. Basically, we find that
the derivative gives us negative information about extreme values, Wherever the
derivative has a positive value or a negative value at an interior point of the
function’s domain, a local extremum cannot occur.
Consequently, in searching for a function’s extreme values, or local extreme
values, only three kinds of points need to be checked:
1) points where the derivative is zero;
2) points where the derivative does not exist;
3) endpoints of the function’s domain.
Thus, the “negative information” provided by the derivative turns out to be
extremely helpful, for without it, we would be searching for extrema over an
infinite set of numbers, yet with its help we'can eliminate all but a few possibilities.
‘These possibilities may then be checked individually to see whether they actually
yield extrema.90 Introduction to the Derivative
Example
Consider the function f(x) = 6x°* — 4x, where the domain is restricted to the
interval -1 < x < 2. Find this function’s extreme values, global and local.
First we calculate the derivative: f’(x) = 4x-'8 — 4. The only critical
number is x = 1, The derivative is undefined when x = 0. The endpoints of the
domain are x = —1 and x = 2. Thus, we have only four points that need to be
checked for possible extrema.
Note that these four values of x divide the domain into three intervals, on
each of which the derivative has constant sign. When —1 < x < 0, then f’(x) is
negative; when 0 < x <1, then f’(x) is positive; and when 1 < x < 2, then
f'(x) is negative again. Thus, the function f(x) is decreasing over the interval
(—1, 0), increasing over the interval (0,1), and decreasing again over the in-
terval (J, 2). These observations tell us that f(—1) is a local maximum, f(0) is a
local minimum, (1) is a local maximum, and f(2) is a local minimum. Finally,
by comparing maxima, we see that f(—1) = 10 is the global maximum, and by
comparing minima, that {(0) = 0 is the global minimum. A sketch of the graph
appears below:
y
y=678—47, ~1¢2<2
1
oO?
Sign of f(z)
By analyzing the way in which the sign of the derivative changes as we pass from
one subinterval to the next, we can distinguish the local maxima from the local
minima.
_ But in some cases the derivative may have the same sign on both sides of a
critical point. This happens in the following example.
Example
Consider the function g(x) =
— 3x5 + 1, and determine its various extrema.
_ The derivative here is g(x) = 4x3 ~ x2 = (lx — 1)x2. Thus we see that
a(x) = 0 only when x = 0 or x = 3. These are the only two critical numbers
for the function, Since the derivative exists at every point, and since the domain
of the function itself is assumed to be the set of all real numbers, there are no10 | Local and Global Extreme Values 91
singular numbers and no endpoints. The two critical numbers, then, mark the only
possible places where extreme values can occur. They divide the x-axis into three
subintervals.
The sign of the derivative g'(x) = 4x3 — x? is negative on the interval (— , 0),
also negative on the next interval (0, 3), and positive on the third interval (38, +).
The signs of the derivative indicate that the function g(x) is decreasing on the
first two intervals, and increasing on the third. Thus, the point (0, g(0)) is not a
high or low point on the graph, but the point (8, g(3)) is a low point.
oz) = Pgxt— 4x1
(0,1)
B,-#)
The minimum value of the function g(x) is g(3) = —2. There is no maximum value.
The value g(0) turns out not to be a local extreme value, even through g’(0) = 0.
ASSIGNMENT 10
1. Sketch graphs of the following functions, labeling the local and global extrema:
a) f(x) =5—x1, ~1exd< 42 by g(x) = 2x3, ~-2 22 < 41
o) h(x) =xt-— 8x2, -8 ## Другое от пользователя: eduarvesan1597