DOCUMENT .

WESUME

SE 022:993"

ED- 143.516'.
.

AUTHOR
TITLE

And Others,AN
Herri9t, Sarah T.
Caltulus of Elemeniaty Functions, Part JI. Student
Text. ReVised Edition.'
Stanford Univ., Calif: School- Mathematics Study
-

INSTITUTION

Group. Iir ,
SPOILS AGENCY
PUB:DATE t

Foundation, Washington, D.C..

69

4467p.; For related dbcuments, gee- tE 022'991-994; Not

available in hard ccpy due' to marginal ldgibility of

oriOnal documen
EDRS !PRICE'

DESCRIPTORS.I

IDENTIFIERS

MF-$0.83 Plus Postage. ,HC Not Availdble from EARS..

*Algebra; *'Calculus; *Instructional Materials;
Mathematics; Number Conte -As; Secondary Education;
.*Secondary School Eathematids; *Textbooks
*School Mathematics Study Gr)bup.
44

ABSTRACT

This course is intended for students who have a

thordugh knowledge of 'college preparatory mathematics, including
algebra, axiomatic geometry, trigoitMetry, and analytic geometry.
This text, Part II, contains material designed to follow, Part I.
Chapters included in this text are': (6) Derivatives of Exponential
and Related Functions; (7) Area and the Integral; (8) Differentiation
Theory and Technique; and (9) Integration Theory and Technlque.Further
)
4pehdices include: (3) Mathematical Induction;
e,Integral
for
Monotone
M
Functions;
(5)
Tk
Techniques of Ih-tegration';
More about
(8)
(7)
Continuity
Theorems;
Inequalities
and
Limits;
f6)
Integrals; and (9) Logarithm and Exponential Functions as Solttions
to Differential Equations.

ci

-.o
,e

Ave*************************4t*********************,144**************
*
Documents acquired by ERIC include many inform4 unpublished
effort
*
* mterials not available from other sources. ERIC mates every
*
A to obtain the best copy available. Nevertheless, items of Marginal
* reproducibility are often encountered and this affects the quality *,

*
* of the microfiche and hardcopy. reproductions ERIC makes available
*
* via the ERIC Document Reproduction Service (EERS):-. EDRS is not
* responsible for the quality of the original document. Reproductions *
*
.* supplied by .EDRS are the best that can be made frOm'the original.
*****t*****p********S*****************4!*****4=************************

4.

.1

.1

St11091: \c
MATHEMATICS--

STUDY GROUP
\*
z

.C4L,CUL'OS' OF
ELEMENTA.RY FUNCTIONS. A
Part II.
Student

(RevisedEdition)
L.

',

U S DE PAR TMENT OF HEALTH

EDUCATION & WELFARE

NATIONAL INSTITUTE OF
.' EDUCATION

n.'11E PEPS!"

,t4

OP,

S'A;FD Z--)t) NO'

,,EN

r)cc r A, NA' 0,NA, .14

p
r
P"-,
A'

Liv

0,

SfISG

BEEN r's PQ0"

f.)
r
IA r,ON raw r,ti

44.3

DOC'JME N

DUC'E 0 FxArr v

THIS

utpir

To

HA>

iril

t I,'

r1,,

E.^,rEt;
E-6i1,

THE

F'Pif

;1)11-fiA( TOriti
.

7a.,Ars-;,-;e

a
A

r.

N,
411

'CALC-ULUS OF
-ELEMENTARY FUNCTIONS

S,

Part II

4.

Student Text'
(Revised Edition)

The follmyhig is:a list of all those who prfrairArted ifi the prePar afion of this
volume:

'

Sarah T. Herriot
Desmend'T. 'Jenkins

C..7 Leeds, III.

George P Richardson
Donald E. Richmond
Paul C. Shields

Gunn High School, Palo Alto, Calif.

Palo Alto Senior ftliih School,Caly.
Simon's Rock, Great Barrington, Mass.
SMSGStanford University, Calif.
Williams College (Emeritus), Williamstovp,Mais.
Wayne State` University, Detroit, Mick,

'

.
.5

3.
/

,
.

Financial support for the School Mathematics Study Group has been
provided by the National :Science Foundation.'
6

r
.

i. ,-

Permission to make verbatim use of material in this book must be secured.

from the Director of SMSG. Such perinission will be granted except in

tireusucil co cum.stances. Publications mcorporating SMSG materials must

inclUde both at:AC/a:ow* nent of the SMSG

copyright'
(Yale Univer.
.
the
case
may
be)
and
a
disclaimer'of
SMSG.
sity or Stanford Universit
endorsement. Exclusive license will not be granted save in exception-al
circumstances, and then only 1)5, specific action of the Advisory Board of
SMSG.
=

6c.

'^110'

.r
1968 and 1969 by The Board of Trustees of the Leland Stanford Junior University
All right reserved
Printed in the United States of America

4
1
Ne.

"44

ORD TO Fr.Fif
,

.t I wectudied polynomial,.circvlar;

In Chapters 1, Q, and., u,

exponentia , logarithm c, and power f.meti..ins.

many pro erties of the graph

As 'we saw in Chapters 2 and

c,'te obtained from the know-

ledge of the derivative .;re= the fur.otio, :inc, tip., value of LhC deriiative

can be interpreted os

'

t'v tanse,r.t line at a ppirit.

- Je.-e

nomial, and circular 4h.

Foe' poly-

to ,.d 'derivIltivec, which wEe

..,ne we we'sa

then used to dete-1,

d corcr,wity, velocitwi and

45.

T.n

The 5e
compos

.3/1

o:

inetloe

We study fuXil,,lon
studied fur.ctIon,-,

t:e

!nd

differ-

ential ca,lcul.::whic.

.11

,t the -apEt patk.oloLlool of . )e;

flinctions.
0

In Chapter

funeidn.

zd: sf :leo under the glaph of a-

The !e,

of the elope o:

si :12,
l_ne

n,j,_

between these t,,o ooncept,. w:., Jne

'first noted

e,-,:.

ar..41ated to that

te.

T.e discvc.-,, of the relationship

_ ,--re t

.o.:

treAtroughs an mathematics,
1, to -h)a that th.

('

arei 4

9g

boullIded by the graph of

given ny

result ie

F(h)

ot

ve'ftioel

,hd

,3
f.

pfound :el.tion.hip'known

2t.--

as ,,the Fund-rnt',1

Chapter

t-t%

to

tr.od

s,

, Tcnti tire; 91-ze,

Lgncept:
(that i., tehe

thre

vide a furth,er

LeL=Jiol of func,

ror

tions:
IntegratLri
'

whic

prb-

further in:Chapter ?,

r:

*-

tatior of,tht,

of 'Ir.-.1-rr

solids of revolution,

mates for Taylor atT-,roximtion...

it'

volume,: of .

and b ai-1,cutsion of remainder esti-

The appendices are intended to fill logical gabs'in the's intuitive

dozvelopment df the text and to extend the material of the text, concluding

,with Appendix 9 in which logarithmic and exponential functions are viewed as

It is shown how the ,expression

so'utioks of simple differential equations.

i

of_ the

logarithm as an integral can be used to 6btain the properties of r---.

logarithmic and exponential functions.

..

NI

4
V

Or'

."

tar

i.4
43

/
/
,c

14
lh

.,

ik

1.-

.
.41. .

.
N. .
.

....
1

40

6
i

4
9/ 4

1'

iv
../

...

TABLE OF-CONTENTS

Part 'II
Chapte,r 6.
s

t 'C)
at k0,a

6-2$

Thee Tangent atkan Arbitrary Point

.
Applications of Exponential Functions
The Derivative of a Logarithmic Function,
.Tayl.`or AprtrAimations to the Function. x -- ex
.
,,
The Powr Formula
Approximations to Logarithmic and Root Functions

615.

The Tangent Line -E7,---) the Graph of

\, 6-4.
6:-6.

41

6-1.
6-3.

,,

n 1 es

o-7,

7-3.
7-4.
7 -5.

, 7 -6.

x 1--",:ax

1+07
,..

,,

AREA AND TI

)407

417

s,

.,

4,21

-.

1:25
i

... .

.7

Chapter 7.
7-1.

.Page

.
.
.15ER1rVA5IIVES OF EXPONENTIAL..AND RELATED FUNCTIONS

f.
.

4311.;
: I% 4 2
.4.r.

1L`,,-.0

INTEGRAL

11 ..o.-

Area Under a Graph . .

The area Theorem .
The Fundamental TheoremoP.,Calculus
.

f+61

474.
484

Properties of Integrals

1$97

Signed Area.
Integration -Formulas

Se

517(

Chapter 8. .DIFFERENTIATION THEORY AND TECHNIQUE

.. ":
Differentiabil,ity

53

8-1.
8-2.
8-3.
8-4.
8;5.
45:6.

8-7:
8-8.
1 8-9.

Continuous Functions

.1

:534
540

The Mean Value Theorem

Theorem

Applications of the 'Mean Val* Theorem

Sums and Multplee . .
.
.
Products
Composite Functions'
/The Chain 'Rule

546

550
'554
565
573
577

The General Power and Reciprocal Rules

8-10. Tiie Quotient Rule
8-11. Inverse Fianctions

603

8-12. Implicitly Defined Functions-

613

590
596

tt

diapter 9.

INTEGRATION THEORY

ANDTECAQ*1

9;1. 'The MethOd of.Sustitution4

.

ge
Page

.t

..

---,

"619
619

..

Ilk

The Average Ordinate, or Mean Value, of. a Function

9-2.

-'

635

'61+0

9-3.

Volumes e t Solids of Revolution , ....

9-1i.

Atimation'of Definite Integrals.

Taylor Approximations

14':IATITEMATI4CAL

....

652

1,.

INDUCTION, . .......

679

'1,

679

A3 1.'.rne Pli,.ciple of Mathematical Induction

*

Aj-2',,Sums

,r.d

S'..,-, NotatDon.'

^'-'

693

705

AP2ENDI1: L. .71=HER TECH:LIQUES OF INT;ZRATION

GIA:41-..7..ior-s. of'Circular Functions

A4 -1

.
6

705

'

AL-2. Ir.tegraior. by- 'Parts

:.

.....

,.

'72'

A--,
4 Defir.lte Integrhls

713

722

A4-3. Integration of Rational Functions

..

*HEN=

UNCTIONS

THE INTEGRAL FOR MONOTONE

737
737

A5 -1. Ir.troduction

A--2.

Aria'

739

I*

A5-3. Integration by SlImniation Techniques .

111e Concept of\I,!tegrl.

it

Integrals of- Mpnotond Yunctions!`:

745
750
760,,

A -5. Elelentary Properties of -Integrals

s

APPENDIX C. .-INEO,U4LITZ.E3 AND tIMITS

773

kL-1. At,solute.V lte and Inequality

773

Definition of Lit

778

Functiohr

Tedhnique

787
-798

The04.ert:

.o

CONTINUITY THEOREMS
so
4
A7-1. Completeide:,s of the Real NumberSystem.

811

42F.ENDIX

'

The Separation Axioal

Vk7.4-__. The Extreme Val& and Intermediate Value Theorems .for Continuous
Functions
,

A7-3. The Mean Value Thebrem

A7-4. Applications of The Mean V1ue Theorem

rs'
.1/

665

.
.

ENDIX 3,

811
817

826

835

APPENDIX 8.

Page

'841

MORE ABOUT 'INTEGRALS

.r.'\40-1.* Existence of the Integral

4.

A8-2. The Integral oP a Continuous Function

...__

841'

_alio__ .

AIPPENDTX-9.

LOGARITHM ANDEXPONENTIALFUNCTIONS AS SOLUTIONS TO

DI;TERENTIAL EQUATIONS'

857

A9 -l. The, Logarithm as Integral ....... ..

/74

857

863

A9-2..the Exponential Functiont

I.
240

Chapter 6
..
k
DERIVATIVES OFIEXPONENTIAL AND RELATED FUNCTIONS
,

,;!
.
.
.
.
o
o
,
Thp derivative ofa polynomial function is again'a polynomial function.
.

again aff9i'rcular func-

Furthermore; the derivative of a circular function

tion.

This kind of repetitive property appears in a very strong form for

.

expoqential functions, for the slope of the tangent,line at a point on the

graph of an exponental function is proportional to the ordihate of the ,point.

,r.

The constant of propqrtionality is the slope.of the tangent,line at thelpoint'

where the graph crosses,the y-axis- Thenumber

for which the constant of proportionality is

'

....

the derivative of

x -)ex

is defined as the base

from which it follows that

1,

is the'same function

x -)g

x
.

These results,*.1,

e-tablished in the first two sections of thil' chapter as consequences of thy'

'..

laws of exponents and the assumption that

x -)2x

has a derivative at

x .... 'O.'

r 5 as,- inverses of exponential

Logarithm functions' were defined in Chap'

This inverse relation enables us to differentiate a logarithm

'functions.

function by a folding process (Section 6-5).

Using the fact that a power

function can be expressed in terms of exponential and logarithm funCtions we

are then able to find a formuil for the derivative of'a power"' function (Sec-

tiop 6-6). The concept of polynomial approximation, first discussed for

circular furctions in Chapter 1+, is then extended to exponential, logarithm,
and root functidq.s (Section 6:7).
a
6-1.

at -.(0,a0)

c.'

Now we wish to find the slope cirthe tangent.line to the graph, of

,,7

x -)ax

The Tangent Line to the Graph'of

at some arbitrary point on this curve.

Ou

ocedure for poZynomials

and'for the circular functions was to first find the

uation or slope of the

tangent line at ,the point where the'curve cfosses the

ertical axis and then

translate to obtain the corresponding results elsewhere.

,also be-followed here.

the'line X best

This procedure will.

In1oir previous discussiohs we found the'tangent as

we then showed that the slope of the tangent at a'point

it

is obtainable as the limit OT' slopes of lines conneWng the point under'con.)kideritionto nearby points. ,We_shall follow this Litter limft proCess here.

it

407

'

4
1....

Ig

\'

nro

'

To be concrete we first consider the problem of finding the slope of thq,

.tangent-to.the graph.of

x --)2x

A(0,20)

the line connecting

tangent line at A(0,20 ) .

Bkh,2

to

h.
)

If, Ihi

x = 0.

at.the point,ohere

is:small

will approximate the slope of the

(Se Figure 6-la)

= 2

s.

tr,

(h,2 y

/
//I

,,

N
.$

., 1

Figurg'6-la

44.

If ,B 4 close to.%A,
the slope of

T,

the slope of

the tangent at

AB

app ciprmates

(0,1).

The line AB

has slope
2 h. - 1

v
We want to find the number this ratio approximates when

1;08

.
..

1-1

Ihl.

11..small.

..

6-1

y.

Tdble 6-1
.

oh-

Values of t- =--

h '(correct to

small

..

-.2.h

2h

1.07177

,05.

1.03526

1
a

.0717.7

.718.

.03526

.7,05

.0069556

.696

.0034717

.694

-.01

h
h

.10

places)

1.066,9556

.005

1,0034717

.00l

1,0006934

',,

.:0006934

.6-)

.681

-.03406

-.05

.96594

-.01

.9930925

-.005

-.001

.-.006907.5-

h.691..

.9965402

:'-..0034598

.69f,

.9993071

-.0006929

.693

Table 6 -1 indicates some of the values of (1) for `small

h.

It iappears from

the table that

.

if

is small, then

Ihl

1r

(2)

- 1

where, to three places,

While ty25is approximation to k

is

0.693.

is correct, 7e need more than ataLle of

vllues, no 'matter how complete, to be certain.

Unfortunately, 'We have no

simple algebraic device for deterMi.ning the limit of this ratio as

.

approaches zero.

at

(0,1)

We'are assuming that the graph of

x -32

has a tangent
2

and that the slope of (this p4enQ s approximated by

1
.

.'We shall assume that (2) is true and qoncent'rateitin the consequences of this

assumption.
If (2)

of *x -42x

is true we have that the slope of the tangent line to .the graph
at

(0,20)

is

k.

At

(0,20)

= (01)

the equation of the

* tangent ls
y

(3)

--t:

Ixl* close to zero we' have

For

(4)

1tx.

.7,..

4.,

x
:"..

1 + kx.

-...

119

.-

t'

6 -i

Now consider the function

a....

x -)a

a / 1.

a > 0,

where

In Chapter 5 we'saw that we can express

A.

as a power of

2.

a = 2.

If

we

can write

ax

(5)

If we.asume that

replace

is small,, then we can

krki

is so small that

ax', in (4) and. use (5) to obtain

by

Ixr

a
a

( 6)

',T.

1 + k(ax)

In other words, the line with equation

4

= 1 +.(ka)x
6

is the tangent to the graph of

x --)ax

at'tlie point .0,1).

The coefficient

x gis the slope of this line, so the slope of,the tangen-Vto

of

at

x --esa

x = 0
.

at

%a.

is

the tangent line to the gl.aph of

4 = 22,

For example, since

at

X = 0,

has the equation

x = 0

= 2

Also, since

-1
,

1 + 2kx,.

the tangent to the graph of

VT
has the equasti9n
k

y =

The respective slopes of ties

lines are

2k

and

k
2

In our di$cussion of the circular functions'we sgw that we could select

our' scale (using.ra4ians, 'rather than degree measure) so that the slope of the

y .,sin x

tangent to

at; x = 0

turned out to be

1.

Similarly here we shell

obtain considerable simplification in our formulas if we choose- a in (6) 'so

1/k
1
then our
Thus if a = 2
that ka = 1. With ka = 1 we have a =
result (6) gives'
,ax T 1. +.x,

that 1,s, the slOide of the tangent to

....vmnowns4.4,4,,64-mPS n

-1

if

ix',

x -)ax,

is -small.;

at

x = 0

'`e

is

10.

=6-1

The' number' 2

1.1k

is so important that a special letter is assigned to it,

We can approximdte

namely, e.

e.. 2

by

lik

.where.

This gives the, approximation

$

x
(7)

If

'l + x

.1bel

is.

(0,1),

if

ex

at

- 1 6

;. 1.

h
The,use of

AI

(8)

is small.

lxi

A
is small then the -slope ofthe tangent to the graph of

in this sense olay be traced to the Swis.s mathematician

L fonard Euler (1707 -'1783).''Nlost of Euler's mathematical life was spent in

His work is still being collected'and at present

St. Petersbuxg,.Russia.

The number

numbers more than 80 volumes.

number

ranks in importance with the

n- and is, curiously enough, closely related to

ifwe use

to approximate

0.693

m.

we obtain

' 1x443

so' that

21.443

1/k

e =

2(20.4)(20.04)(20

4031

2: 2(1.320)(1.028)(1.002)
AP

: 2.72.

will obviously improve this approximatiod.

Closer approximations to

In

recent years, high speed computers have been used to obtain the decimal expan-,
sion of
15

correct to

2500

For the record, we note that the first

places.

are given by

places

4284

2.71828.
2.72

For our purposes either

The number

or

59045....

2.718 'will.be good' enough.

has been shown to be irrational, just as is *Z.

In fact,

a 'much stronger result has teen established, namely it has been shown that

is not the root of a polynomial equation wit

is true for IC.
2
x - 2 ='0.) .

(The number V-2"

rational coefficients.' The same

is such a root; e.g., it is a root of

a.

There is an important method for approximating

)

(1o)

e,

given as follows

(1 +

)P

n"

for

large.

411IP

14

*s

:rrale

m
immi f T 1 WE
INIMMIIIIM
1""WW115111
1111111
iiii.45 :._
IMPANIMIfflii

NE
II

iiiiin ME

Aniffdihilit'
.I MEMlir

ikwimPg
oakumiMFihl
. a .H
Ili

a
-EMS

EPEEMMiliti
gadp....qmpapil

2 MEMIMillii

MN"

M
TOMIPMFM

1 11 1111 ROI 1.1 or.11

E
EMEMEEE EEIMEN0 ME
ang
IE EEERNMM
I'
PIPPIIMMIE
NEMER MEM
EEE
wirirmili
ME
MMERPIIMMEri
1
..niudil En_
M EMMEN bri
E
EU& niEbiriE
,t I

HUME

EE

't

SEEM
NOE EE
MOE

EliiiiMagi

IN

ril

camommimpaghiN

EMEME.
F;
EISEN
Mrdr

mu

Ellti NEMOR
mu a

6h
,Exercises 6-1
1.

Given the function

for

x -tax

(a)

ai

a = 8,

(0,1) to the graph of the func-

Find the slopes of the tangent at

tion for'each value of

in terms of

(i)

k,

h.

where

,k

for small H..%

% h

\
as an IpprOximate value, using

(ii)
(b)

0.093.

Find the equations of the tangents, for which the sloped yere

obtained in part (a)

On one set, of axes for each value of

(c)

given abOve, sketch the

graph of
%

(i)

4the function;

(ii) the tangent obtained in part (b).

2.

Given
(a)-

(1.8)5.
h

Using the table for va4ues of

a power of

(i)

expresa (1.8)5

(ii)

approximate the value of

a's

4.

2;

(1.8)5

from

2a(i).

(b)

Using the table for

and ex

ex

express

(ii)

approximate the value,of

(1.8)5

,3."Folleit,the instructions of Number 2 for

0.

4.

Follow the instructipns of Numbgr.2 for

5.. Obtain bounds for

e;

as a power of

(1.8)5

(i)

(1.01)1,

from

2b (i).

(049)5.
.8
(1.02)

iff'ing the table for Va.ues of

2h

as

ollows:
4

(a)

)1

Write

(2 1)
1Cia

(b) ,Evaluatel 2
bounds'

6.

Obtain bounds for

.1

as an inequality
100

0.00-

<0.:o1)

a2 100
< k2 )

100a
1

and 2
(1.01)100.
_(0.5)-12,

2
,

thereby obtaining upper acid lower

using the table for

following a procedure similar to that of Number 5

ex

k6

and

e-x,, ,and

.,o

oto

y =, e

Consult the' ketch to write

.

mc;-bi Lihe

the slope

xi

L.

h.

4h,e,)

x
a

(b)

e,

ft

(c)

8..

Whiph is larger

t6 one 6ecimal place if hc.01.

Improve upon the result of part(q tq show that

As. the limit of

h / 0.

Use your expression,from part (b) and binomial expansion to give an

approximation for

(d)

m0 :1. and

Write an expresOon fo;.:

(1 +

1000

1001

1 n

or

grow,large withOUtbound.

as welet, n

1001

may be defined

10
7

B(h,eh)

the slope of

'Show that at ;x = 0

the tangent ;to the graph- of the

function
"^4

3x

3x

x --> e,

is clo e to

when

ihl

is clOse

toe o; byecompleting the folloW-'ing

4A

ables.
1.

.4
Y

II

,;

3h

3h

..

.20'

.
,

:15

05

'

.01

.0$

I.

t
4,14

Or'

3h

6-1

.10.

x = 0

Stow that at

the slope

the tangent to the graph of

.1 .

the functfion

B(h,eY2)

1.2

x/2

to 7I when

lose

is

111.1

A(0,1)

.clo e to zero, by completing these

tabl s.

t
h

7b

.6

h/2

h/2

59
.4o '

39

.2d

Ii.

.10\

.o6

.0? \
1

11.0 Show tht at

x = 0

the slope
.

of the t ngent to the graph of

1.6

the function
f

is clqs,e to

-2

when 1111

1.4

e-2x

is

close to ,zero, i9y comple,ting

the following tabled.

.0

.1P

A(0,1),,

NMI

111M
ik

III
.

.4

sa

415

rM

.2

B.0126(21

_18

h
,

-2h

2h

-2h

- 1

-2h

- 1

.
.

,.20
.

.15
I

.10
.05 ,

..
%

.02

'
,

.,01

.005
12.

In one of the ploblems of Number 1 we round the slope of the

equation of the tangent at the point

(0,1)

to the graph of the func-

r.

tion
f

-4 8x.

In terms of an inequality, approxikate this slope to fOursignificant

figures after filling in the following tables.
(In the first table, -h

approaches zero through positive values from

the right, and, in the second table,

approaches zerothrough nega-

tive values from the left.)

3h

3h

3h

- 1

3h
h

.
.

.20

.15
.

to, .10

.
.

.05

.01
.006

.0006

t..

.!,

-.20
-.15

-.10

-.01'.

-.bo6.

-.0006

416

19

- 1

6-2
,

6-2., The. Tangent at an Arbitrary. Point

In the previous section we obtained the result

l.

(1)

a.'

;--:

1.+ (ka)x,- for

r

a = 2

k. is the limit of.

and

e
1

With

small,

Where

!xi

as

h .approaches

0.

11

.
,

we have the' simpler result'

za,..-.,e

ex z 1.+ x.
We shall now 1.thow, that. the tangent, line to the graph of

D
t

sex

P(a,ea

+ ea(x ; a)

y =

(2)

4 that the slope of.th-e graph at

o

is

, :the same as the ordinate

(bee Figure 6-2a.)

.cf

Figure 6-2a

Graph_of

with tangents

= ex

'\

417

20

and

T2

6-2
result

\48' put

'

x = a

(x - a)

in

ex.

L
Then

(3)

ex = ea+(x-,a)

J.

. a

= e

If

is close to

is close to

a, x, --a

r`

X-8
e

2: 1 +

)
0

(X -

and hence,
a)

').

)_._

..

.0.

Substituting this result in (3),, gives

ex r.

The tangent to the graph of

+ ea(x - a).
at

(a,ea)

has the equatiOn

y= ea +e (x- a).
At the point

(a e
x

(5)
traph of

,
)

77.

the slope of tAe tangent to the

, ea,. /

is

IL

'

As in our previous discussion, the resulting "slope, function is called

the derivative.
,

value at

,x

That 1.s, the dertivatiye of

is the function whose

,,-,) e

..,,,

),%-w

tis the slope of the tangent line at

(xiex).

We restate

(5)

using derivative terminology.

.-..:

':.

If

A
:

'x

then the derivative

is 'given -by

(6)

f' c -) ex.
",
'''

In particular,

ex

(7)

is a solution lio- the differential

f'

f.

equatiorT4-

'

LEXample 6-2a.. Find the equation of the:tangent to the graph of

at the point

For

(3,e )

f : ,x

ex

we have the derivative

f' :

so that
e3

f7(3)

"e3.

f : x

The tangent to the graph of

has the equation

y

e3 + e'(x

148

21
t

3)a

1". at

with. slope

ex

..

4,

raNNEI

1............m.
.=.."=......::

..4.,

0 ,1114 ;NOM NON&

lnir7anliMIC...474"" .."112"411"4:1..2.''2 '"12.11 HiS

ORO

WAVE: WEENY

11 MINOS als '

.........................
linIMO.
.
wameos000
A mmeIrso=.....==:=.7.......s
-L-IMW"".....7gr".""...r. M".".......= Etilr....=.11=1=r."
.1.1
.....
... =p

MHO

... ni.....us . I ......

&&&

111111.

...

MIN

hr

1..

.1E. en ......

..................

..

. .......................

taM
C

war:-

IMRE.:

.. ... ...

..

XX

Oa

L'

SMUIVE4

WEB

r,;943.

tee

ETD 1411

"

IMAMS MMIIII
11111,
011

.............
..... HRH

M..

dm

::: X

11:1,
m

..

. .....

::::::77:::::

t"."7::::::.:2111:1::::::::...

..........

.......... ...

X XXX:::: X Xi .............

Vi i:
zurnumunrx:::::::.... ..:::::::::::::::::::::::::...num
.s.:xx::::::::::::::igh ..... ::::::::::::::::::!:::::::::::::::
... ........... - .........

ii:
ax.....

::::::::::::MMI::::

.......
..alprl ...NOM .......... :
NU

..............

ff

pl ....

MN

iX

MTN=

Lam

............ .....................

.14

101 .......OW.
..... 11

- Owens
!ME
ranMESEEISM::g BEM Li
ifn=r1=1"411:2:03ES
BIBLIR
1 *elm...Elle
...
AWE
..... WE =gri'''''.6"
MA =AMP
sus
lism
: D.r...f r.1 A%
.,= " 'i:WAIII BIEWILNIMIW:1==:=10
=9 SMISr""n.".%1
n ....... =MO /ENMr
........ 'MEM
1
EM.
t
......="1
...............nt ..... Wyo.., ona001111.1111
MIMS
:INZ
....
mow. N..
''''
`'
.. ..... - ISE
'7,4:2XXFSSX:XV4.1 lir. ; =MUNI X.INT22:11ZREF"..
f.inr""'''''
:
. ..1 . .,.16....: n.
J11111::::::1111111
;T1111119:=1:11 ...mirra91111:11:1:11N1
...... tr. es-__11
TarirdriNFUM
sus 71:s.....
es
.
air
'11
XXI
:51
tl&

:meal

11PE"...
-:--1

dielnat IMRE 1.1

-- a r: ........
...... ..............
"....."-----LiF.:271i1::31::::
- xx!.xxxg,". gar

-"1 'tleingint.-ran.x.: ..s..T4nNIF:XX/.

llhspagFirii;

IIPAJK:1111RdirfillrajElLEBSSP

Naha MIRY IMMO".31fimin

g ej.Cr..."".''Vq3C:rang
ini.1f....I -1.-v-"

Eil. ....BIUX:

tura-i-1.---un'tz.---.11"--

111.1:12.... 1:. ...:1

1-.
PM: ....
BligisTxtER: r-----4MilidAidjihiliaBbliiifflei=.3.,_3:00,_..."--1,nEFF
.will

"'

:24 ..

i m . X MX Xrd..
px.:::annealiBMPL
.....th.:...ummit
..... .....,...... ..
6n:a
gorampog
ralFatirp.pdrarstl:TX:parar":.;73.1;:rairderilar14....L..ird_

.. .. ....

...

1111:4"".
1111AMMUNNEIE1MIUM:111911"M ......
.......... . urem ...... !
3
-.on
r..-terax--en.-

.....

i.... ........ ..
wpm ..=:12=e.

m--r.-,nm..--:xxxxxxxxxxsx; --- mr.

In
LISIMILISM REHM

...... soma owe

rilrlin=M11 =iffia 77:7:=:= MEMCnr'''..'".:5:152

EMIPMI:XIMISI
XXI 1....11... gm.=

3:

Write an equation of the tiangent to the graph of

et,each point

,I)

(x,e) .given in Number 3.

4.

(a)

Thpough the point

(b) PlOraw a line

.draw a line

(3,4 )

which is symmetric to

with slope

L1
L

m =

2
.

with respect to the

y-axis.
(c).

What point on

(d)

What 'is the slope of* L'o?

,(e)

L0

corresponds to the point

Consider the,general case:

with slope = m,

L2

the y-axis. 'What point on

. L1?

line

andline

at itsthe slope of

on

Li?

drawn through point

symmetric to

L2

(3,4)

(r,$)

L1 .with respect, to

corresponds to point

on

(4,$)

L2?
.

(a). Plot the points

1.6.'

(b)

6.

x = ;2.0, -1.8, ..., 0.2:0.4;

for, which
'.

Through eacY of these points draw the graph of a line having slope
m = e

.(p)

(x,ex)

Show that these lines suggest the shape of the graph of

x'

4.

(a)' For each point plotted in Number 5( a) locatt the corresponding point
.1

which is symmetric with respect:..to thej,-axis; then through these

points draw lines symmetric to those of Number 4(b) with respect to

the y-axis.

Chow'that each point located in Number 6(a) lies on the graph of

g :
(c)

x -4e

-x,

Compare the slopes )f the lines drawn in Number 6(a) with those of
NuMber 5Zb).

7.

(a)

On one.s4..- of coordinate axes draw the graphs of

g : x )e

f-: x *ex 'and

-x
t

(b)"

Compare the slopes of the graphs drawn in (a) at

(c)

Compare the slope O'Nthe graph of

at

x = 0, +1, -1.

x = h 'with

g(h).

I.

gio

11

6.

Appiicationli of Exponential Functions.

6-3.

Exponential functions arise.in practice in the studyrof groNtla or decay.f

ther applications

We discuss compound Interest in this section and give

in the exer4ses.

P dollars-is)hvested at an annual

Compound interest.. Suppose that

rate of interest of

and iat tle end of each year inter-

per cent or

100 '

est is compounded, or added to,the principal.

years the total amour* o

Afte..2 t

on hand is given by ,
t

At = P(1 + y55)

However, the interest may be compounded semiannually, quarterly, or

a year.

If interest is, added to the principal- p

r 0

interest is

5755,71.

per period,

Hence, the amount

and
4

times

e rate of

times pe.r..year,
'\th.

tg nuthber, of'periodt. id

years is
years) is

r Int'
(1)

+ 10On'

nt

nt.

nt. periods (that is, after -t

Aht, after

The more often you compound interest, the more complicated tne'calculption be.

comes.

On the other hand, if we let

in (1) get larger indefinitely, we

approach the theoretical situation in Itihich interest is compounded continu-

ously; we shall see that the resultobtained Will enable.us to find easily a
very satisfactory approximation for the amount of money on band at the end of
a reasonable period of time-.

To study this idea, let,Tiv - h

At =

+ h)

( )

ttie value of

514

rt/100h

n,

Then, (1) bec

1(40h

h)1/hirt/100

For large

n -

sd that

approaches zero and the right side of'(2)

,

approxifnates

A = Pe

rt/100
3

4/0
/

the theoretical amount that would be obtained if interest were comp&nded con -'

tinuouslyat

per cent.

Thus

A = Pe

rt/100
a

421

6-3

, f
.

Example 6-3a.

pare the amount after

the amount after

We have

$100

If

11/4

is invested at

percent for

years, com-

10

15 yeIrs when interest is compounded continuously with

years it' 1,nterest is compounded only annually

10

P = 100, r = 4,

t = 10 (years).

and

If interest is compounded

continuously, (3) gives

A = 1 00 e 0.4

/,

which is aloproximately

149.

To compute interest compounded annually we substitute the above values of

P, r,

and

in (i).

This gives

--Alo

no(

o4)

lo
.

MTImay use a table of-common logarithms to estimate

A10
The results,

$149

and

4148,

100(1.48) = 148.

A10;

thus

differ by a, surprisingly small amount.

4

4
A
C

st,

422

025

6-3 *.

N
4
t

Exercises 611
When hrsisOn Jack-wa

1.

education.

borniHMr. Toffey invested '11000

for JIck's college

Intergst is compounded continuously at a rate.of

per cent.

How much money will' Mr. Toffgy have for Jack's education on Jack's
--' eighteenth birethday?

Using

2.
.

2 vg

e0

.693

find how many years it takes to double a sum of money

invested at- 3

per cent compounded continuously.

3. lack Toffey (of No. 1) earns a scholarship and elects to wait andto
withdraw his father':, investment when it has doubled.

e"

be when he withdrq0s the

How old will Jack

$2000?
a

Determine ho'- many year: it 4111 t4k0 to double a sum of money invested

at
(a)

pdr,centcompounded continuously;

(b)

per cent compounded continuously.

5;0 The 9uantity

(1 +

1 n

can be interpreted as the Value at the end of

one year of a deposit of one dollar left to acquire interest at an

annual interest rate of

compounded

_00%

times a year.

If the ft.;

terest is compounded continuously, that'is,,if.the interest 10 caLculated

,;s the limit in which the number

of -Iiiterest periods approaches

infinity, the value of the prihcipal st the end of ope year will be ,e

-dollars,

$2.72.

A California savings and loan association Qffers an interest rate of

(a)

".4.85%

1f

compounded cont;.nuously. 'What is the equivEilent annual

interest rate for more' left on deposit one year?

Uow long does it take for an amount of, money at the same interest

(b)

rate
6.

At

(4.85%

compounded continuously) to doubly itself?

kilometers above-sda level, the pressure in millimeters of mercury

is g1ven by the formula

/
where

P.

A,

P = P e

-o.11445h

is the pressure at sea level.

If

.4. is the pressure

180

millimetert of mercury?

423

264

PO =`760,

at what height

6-3

7.

'

A law frequently applied to the healing of wounds is expressed by the

formula
-nr

Q=Q0 e'

is the original area of the wound, Q is the area that

0
0
remains unhealed after n days, and r is the so-called rate of
where

healing.

If

find the time required for a wound to be

r = 0.12,

half-healed.

8.

If light of intensity
its intensity

I0

falls perpendicularly on'a block of glass,

at a depth of =x
= IOe

feet is
-kx

If one, third of the light is absorbed by

intensity
1

10

feet below the surface?

feet of glass, what is the

At what depth is the intensity'

1?
u,

14

J.
i

424

27

The Derivative of a Logarithmic Function

6-4.

The graph of the logarithmic function

a / 1,

a > 0,

x -4 loge x,

can be obtained by folding the graph of

4

x -44

over the line

Just as in the previous section we can 1.1se this folding

y = x.

x log,ds.

Trpcess to find the deriyatiVe of

case when

We discus fir,t

importan:

a r e.

Suppose

is a

( c,d)

point on
f

thu graph of;,

x

log,

loge c = d.

o that

Hence,
=

(1)

lies on th4graph of

(d,c)

so, that

ed

ex.

11

The tangent line

where

to graph of

is, the derivative of

gt

g
g..

gt

at tke point

(d,c)

has slope

Since

x,)ex

the slope of the tangent

to the graph of

we have

gt(d) =.ed,

(d,c).

he process,,of folding over the line given by y = x

to the to entjline
point

(See Figure 6-4a.)

425

28,

at'

carries

to the graph of the logarithmic function

1.d

gt.(d),

L1

at the

in-

Tangent-1,0e, L

y = x

folds over the line given by

onto tangent line

The slope of \L is the reciprocal Of the slope of

the slope of

L =

-so that

r
.

The slope of
point where

is the value of the derivative of

',1x = c.

x )loge x

at the

Tjs
426.

29

L.

6-4

To express this in terms of

we use (1) to replace

c,

ft(c).=
In general we can say,

by

c;

obtaining

f_sylr

if

x -4 19ge x

(2):

then

ft

x -4

1.

"

We can rewrite.S2) as

D(loge x).-=

The tangent line to the-graph of _f at the point (c,d)' has slope' ,

Hence the equation of
and passes through (c,d) = (c, loge
ft(c)'

the tangentline is
y = loge c +

If

is close to

- c).

1j..(x

the tangent line serves to approximate the curve and

.

we have
loge 5: tt.loge c + la(x -

(4)

The derivative of the general logarithm function loga

gn be
Or

obtained

It is also a simple

by aoprocess similar to that which we used to,derive (3).

consequence of a relation,derived earlier, namely

x

log

loga'x

log

a >,0,

if

a rl.

fi
In fact
p log
.

x =
a

1
log

e
-Os

427

30
N4-4

6.4
Example 6-4d.

"Find the equation of the tangent line to the graph,of

A

X -4 loge x;

at

Knowing that

p- oint where

log

we see that

e = 1,

e.

4
(e,l)

lies on the graph,of

.A.

x -41og

x.

Since

tangent at:..(e,l)

D log

-,

x =

the slope at

has the%equatiOn

(e,1)

is

e'

Hence the

,-,

1,

A
y = 1 + -e-x
- e).

The 'function

x -)log

is referred- to in Most 'advanced works as.the

x
e

logarithmic function and denoted ',Amply by log without sub.;cript.

logarithms (1.3garithms with,base

Common

10) are still useful for hand computation

but with .the advent:yr machine computation, they have lost much of their once

great importance.-"`The logarithms used in anagrsisare almost invariabl

logarithms with base

and.are xe2erred to as "natural" logarithms.

In most elementary texts

logl-''Mean8

in most professional literature

log x

logio x

means

and

loge x;

means floge x,

in this text we 'shall

try to avoid ambiguity by specifying the base of a logarithm unless the' context makes the base Perfectly clear.

John Napier (1550-1617) is justly regarded as the inventor of the

lo

rithmic function. Although the basic idea was definitely "in the air" of

his times, he was the first to publish a table of a logarithmic function

(1614) and his idea's

bout logai.ithmswere more iksightful and efficient for

the construction of tables than those of his contemporaries.

Napierian,

logarithms, usually thought to be logarithms to the Lase

are in fact

e,

given by

Napierian

log x = 10

logife (217).
10

Henry Briggs (1561-1631)'was largely responstb.le for the introducti-c\fiof

logarithms -with base

10

for the purposes of computation.

A table of natural logarithms (logarithms

(logarithms to the base
in,the accompanying Booklet of Tables (Table 6).

e) is

We can use this table to

compute logarithms not contained in it, if we apply the properties of logarithm

functions.

ir-

6-4
example 6-4k

Find.44.
1.44.
..
log

1.44 = (1.2)?

Since.

11.2)2 = 2 log

1.44 = log

1.2

2('0.1823)
I.

0.3646.

We can also perform computations using these properties a-nd t4e,Table.

Example 6-4c.

Compute

lq. approximately,

loge lq = log
1
z

31 /2 =

-1

log

-..

1.0986

.5493.

Since. loge 1.7 Z 0.5306

t.8.

and

loge 1:8 z 0.5878. I is between

and

1.7

Interpolating,

"

11;

r
.6

6-4

1.

,\

Exercises 6-4

Using the table of Natural logarithms find the approximate numerical

value for each of the following:

2.

(a)

loge(1.96)

[Hint:

1.96 = (10)2]

(b)

loge(2.03).

[Hint:

2.03 = (2.9)(7)]

(c)

loge(0'.52)

in two wags:

(i)

loge

(0.52)

=loge

(ii)

loge (0.52)

loge

(a)

loge (0.052)

(e)

loge (

(3.9)
7.5

7521200

39,000,000'

Using the tables for natural logarithm's find the approximate numTcal
-value for each of'the following:

3.

(a)

/27,

(c) '(94)2/3

(D)

3 Fri

(d)

For some

loge 2

.6931,

approximated-by

Using, only this formula and the table value,

find the following logarithms:

t

4.

loge x

we have by (5)

close to
- c).

loge c

(100)1/2

loge (2.03)'

(a)

loge (2.01)

(c)

(b)

logy (1.g6)

(d) .loge (1.91 +)

Using the results of Number 3,-find an approximate value for each of

the follOwing:
(a)

(2.01)5/3

()

(2.03).6

(b)

61/1.90

(d)

(1,94)1'1

is the x-intercept of the following?

5.

x 2;R 1g 3x
x.--) log

(iii)

x -) loge3

2x
e

x -) loge x

loge

.4

430

33

"+,

:.

.. .........

6-4
(b)

Given:

The _ x-intercept'aiu'et-- be-

As

(1.1)

(c)

6.

(a)

what interval?

--k(constant) > i.

e k

(i)

The x-intercept must be in what, interval?

(ii)

As

in

gets 'very large, -what does the. x-intercept approach?

log

Given:

> r.

loge kx, 1t(constant)

gets very large, what does the x-intercept approach?

For a given absqssa, -what is the vertical distance between each

of the followingi

--'4-

--

(i)

) (ii)

(0

loge 2x

and

x..

x .-->. loge

3x

and

x 4

4x

and

2:

.(iii)

loge

(iv)

loge (k + 1)x, and

loge x

lolit-X
e

-loge 3x
x

gets very large, that effect does

In Number 6(a)(iv) above_, as

(k > 1)

lode kit

this have on the vertical distance?

(a)

For a given abscissa, what is the vertical. distance .between each

of the following?
x

and

x 4 loge

loge -f

and

4 loge

X -4 1S2ge

(ii)
.e".

(b)

log e 3

(iii)

loge -5

and

(iv)

x
x -4 loge it-

and

x -4 loge

In Number 7(a)(iv) above, as

T
x

(k.> 1

k + 1

gets very large, what effect does

this have on the vertical distance?,

8.

(a)

Find the derivative of the following functions by using (14) and the
property,

log ab -4. log a + log b.

[Hint

Refiember that the

derivative. of a constant is zero.)

(0

(ii)

(iii)' x
(b)

loge 2X

4 log e z-2
I

4 loge

Sx

(iv)

4 log eA3

(v)

4 loge

kx,

k > 0

(vi)

4 log

i ,

k > 0

.'

Find the slope of each, of the curves represented in part

Number

8 at the point where" x "=e ..

.

431
r"*-^

A-

(a) Of

6-4
ti

+) Find the coordinates of the

(d)

Find ehe equation of the tangent line to each-of thecurvts at' x = e.

What are, the.y-intercepts of each tangent?

,,(e)
,

->

(ii)

Show
and

hat-the y-intercepts of the tangents to

(f)

Sketch
axes,

or the regron:

Q < x < 3.5,

and its tangent at

x,

x = e;,

and its tangent at

x =,6;

x -4loge 7 ,

and its tangent at

x = e.

x- end

y-intercepts of logarithm curves

xi and

y-intercepts of tangent lines

parallelism of tangents
vertical distance between tangents
vertical distance between logarithm curves.

9 Using the law of Toga

log a

= b log a.

Find the Wivative of the following

x -)1e5g

(iii)

loge Tc

(iv)

log

(b)

x -)log

x3

Sh?.R/thpt

(1)

D log

(ii)

D log

n
n..
x =
e

.x

96T =
e

nx
,

(iii)

(iv)

loge(cx + d)n

D loge nr7-czic)

nc
cx + d
,

n(cx

d)

J+

43..?
I)

Aik

-3 < y < 2:

Indicate, (where possible,)

(ii)

kx

-q

:4 loge 2x;

(i)

log

arefully. the following on one graph using the same'set of

(a)

:9

A'

s : x -4log

"1\

(k > 1) are symmetric with respect to the

x +ebloge

or

x = e.

on each curve above where

3b

3iT,c

and

6-4
:

X10.

Using the results of Number 9 above find, t'he de riutive of the following
functions.

1,111en,fox.mula, aoo not seem to apply,

[Hint:

of logarithms:

log ab = log a + log b,

log a

= b log at

log

*log (5x + 1)3

4.

remenr the laws

log - = log a - log b,

= -16g a.]
s

x -)log e(4x`n 167)

-) loge' x(1 - 2x)

(e)

x -) loge (loge 9x]

(f)

x 4, loge (sin

(g)i'x -) loge

2x
x - 1
2

) loge x2(3x - 1)

(h)

x -) loge

s.

fr-T-7
1 - x

Find the equation of the only tangent to the graph of .t - loge x

that
.

passes through the origin.

Comp

your equation with"the resLzlt of

,Example 6-4a.

D.

t
6-5.

.Taylor Approximations to the Function-

x )e

The derivative of

x )e

tives of

..are also

x 4 ex.

is
x

x -.)e

Id other words, iS f(x)'= e

ex = ft(x) = f"(x)

(1)*

Thus the second and higher derive_

x'
,

then

= f(n)(x) =

Just as-we,did for the sine function we -now seek to fixed polynomials with the
same derivatives as

More specifically, we wish to find a polynomial

x >e x.

such that

(2) "

(a

the degree of

(b) ;

p(0) =.1

(c)

the values of the first

x

does not exceed

e0

are the same for

degivatives,at p
x = 0..

For example, consider the case for which "n = 3.

p(x) = a + bx + cx

+ dx

and

We put

We have

p'(x) = b

2cx + 3dx

2
,

plx) = 2c +
4

p'"(x) = 6d;
so that'

p(0) = a,' p'(0) = b, p"(0) = 2b,

Suppose

x )e,,

so that

f(0) = 1,

Hence, if

f'(0) = 1,

f"(0) = 1,

f'"(0) = 1.

satisfies (2),then
1 = a,

so that

p'"(0) = 6d.

1 = b,

1:= 6d;

1 = 2c,

is necessarily given by

'In general, we have

(3)

p(x) = 1
x

2! 4- 3!

as the unique, polynomial which satisfies (2).

*
x
the Taylor approximations to e
.

Brook Taylor - English 1685-1731.

43/4

37

x
4.

n!

These polynomials are called

40

411.

el

MEET;

IM

Ea

....

i::

r-Ipil

.... ... : . .... ......

... ..
r .........

t:,

...

--::::::::::=.71:: .....

: .:. ...... ....3

' ..... "'IA

n=..clict- u incr.:::::=1:=
-: --cram
a ',F,Alaaeji.. aaaaaaaaa-TurdaiREWPWry2=7.112:Ercit.....E..2:::::11
.". - a 1
HT :14' ni,
rag JaPainlliffihefigiffillli."42. :
Itl
1410441;miloimi-itsk
IHnenucc......ccs.:41
111MEMgrunia-drLti.-Mcl"-itairig Eigifi Egi.;;
Pi mmo; Elmo 'ENE115:161FISPALv;:1 PiliElli. 11;;;:::77..:".177::::=A:71"
ram '

KrFssrwaffingtemilsELIP_E2S-11_.b_

01.4EIE._.n1:=uhnii!q5 k......
..:_ismr..E....:E.F..

AINERAskiiikulat.........prile:EPEL:PrerA-41,___us.:::::,,,,r::___.._ ................,..... ...

1.1:. i.
*. ;- . sp.-

s..isnlitkililiklirq 1iidlilligifituRRELEN=2.11L:viii:i. -

111

d Mr.

Prea:

Ulligirliiiilligffidrr...013

LF. UR... b"ri ":9:334'.."11ii Mil

311:21PIIMEMSEMPlim
..
Eir-uninlichlinp..-:;fflill..:...: :If.:4"
...
141:.:11111:::11:71VnitauffirutIONIT:1111,...11P-10:::*ra:
i s ....f"
.
... :Lc"a.--

ffileiThi.c..::

...SBA

.....

d'AIE;7:314.7.7" ry.:3:101
:=1101:0 OOOOOOOOOOOOOO

.....

In Chapter 9 we shall use area principles to establish the result:

x t

= 1

n
x
+ -- +
n.

n'

.14here the ,mminder term R

satisfies the inequality

M
R

(5)

'-filitieTrftgra:imple, if

<

n +1
x

then

0 < x ,< 1,

x
e

0 < x < M.

if

+ 1)1

= 1

21

3:

+ R

3'

i4

- where

e x

< 77- ,

4.

._

x
e

.. And

2
3
X
X
= 1 + x + 7.7 +
+
d.

'

4'

its

51'

.).

+ R
!,

31...,

<

where
.

.r
.

'

Fdrmtiol,as(4) and (5) are useful in constructing exponential tables.

x

observe that it is necessary to find

only for

0 < x

.Larger powers
0113

can Oe calculated from knowledge of these. For example, if we know

2.13
then we cdifind e
by using the relation

7 e2

'2.13
e

0.13
.

Negative powers can be obtained by taking reciprocals.

Thus

-1.3

Suppose we wish to construct tables of

We first choose

decimal place s.

cannot affect the first Iwo places.

ex

We

for'.0.< x < 1,

1.3

rrect to two

large enough so that the error term

We observe that OQ

[0,11,

1
e

< 3,

R
so

that formula (5) gives

<

n +1
e x
<
(n + 1):
(n + 1):

We can therefore estimate correctly to two decimal places if we choose

.large that

Rewriting we get

we can choose

n = 5

so

< 0.005.

(n

+ 3

or

1)1 . 105

600 < (n + 1)1.4.Since

and then know that using the formula

436

39

6: = 720;
g

ex xi

1 + x +

x
2!

4x xx
4

't

will give answers correct to two decimal places for

Example. Find

e0 .1

correct to three decimal places.

so large that on

Yr

e0.1 xn+1

Rn

+ 1):

e'l< el < 3,

We know that

M = 0.1.

We first estimate (5) with

need only choose

0'< x < 1.

so we

(0,11

10411+1) < .0005..

(n + 1):

We have
10714

= 0:125 x

104

< .0005.

Thus we know that, correct to three decimal places,

_

0.1
e

P4

(0.1)
2:

1 + 0.1 +
1

= 1

(0.1)3

3:

1.105.

200 + 756.

We,can also.use (4) and (5) to obtain limits as

various expressions involving

The next example illustrates this method.

ex.

-1

Find the limit of

Example 6-5b.

approaches zero of

(1 - ex)(1 - cos X)

approaches zero.

as, x

x3

We shall do this first in a rough way.

2

Since

x
2:

1 - --

cos x

1 + x% and

'

'2

ex)(1

(1

x.

cos x)

(1 - ex)(1 - cos x)*

Hence,

P 2

x3
and the-requif'ed limit is

1
-

2.

More precisely we can take account'of the errors made in using the

approximations to

and

cos x

if.we use the rema]indera

437

40-

.and R

in

e
ex
= 1 + x +

2n
cos

+,R2

= 1 -

Then
2
ex)(1 - cos x) = (ix .f R1)(E2- -.R2)

x3

R2 - 22

3. [1..

Since

<

2
ex

for

2:

on

,. 2

then

and

The result
e

approach
.,

11
,

(4)

2x

x3 j

< r-

lim

and

..

x -.)0

> x no matter how large the exponent

(1 - ex)(1 - cos x)

can also be used to shOw that if x

FKom (0- := for any

s:

R_R
i 2

Akt x

x2

R R

0 < R

and

[0;11

44114

R1 +

may be (k

x3
is large enough,

a positive integer).

x > 0,.

ex >x
.-n:

Let

tit= k + 1.

Then

....

k+1
--rx

x
e

n.

and

This means that, when

x..>

4:4

ex

> 1
'

xk

Th

that, is,

>,xk.

fe,

;4438
,

4:1
.

6-5

Exercises 6-5
1

1.

Write the first four terms of a polynomial approximation for each of the
following.
(a)
'(b)
(.c)

ex

(d)

cos x

-ex

(e)

-cos x

1 - e,X

(f)

1 - cos x

-- For Numbers 2 through 5 consider thei,graphif each function.

Write the poly-

nomial function, which serves as the best

(a)
.(b)

(c)

linear

quadratic

cubic

approximation to'the graph of the function near the y-axi.

x -4y = a0 + alx + a2x2 + a3x3 + a4x

3.

g: x

4.

-4sy = sin x

G :x
6.

= cos x
= e

Do You suppose that there are polynomial functions that can serve to
approximate the graph of
= loge x

f r x

at the y-axis?
7.

Compu'e

Explain;

21 places.r

'0 .01

Obtain the value of

each term to six,,places, contin!ing until you reach terms which have only
How
zeros in the first six places, add, and round off to five places.
that even though the remaining
many terms did y?u-rteed to uSe? No
terms are individually less than 0.000001, they might accumulate to
moo

give a very lhrge sum; inthis particular case, they do not.

8. . Qbtain,..en,ap

e0.2

/*ion to
(e0.1N2

by computing successively

eO.4

(.'S

=.(9.2)(e0.8).

Use the estimate

0.1

1.105

of Example 6-5a.

439

42

r.;

9 ; Suppose
2

,P(x)=1+x+x
+
n
.that

(x) = pn(x) -

p17.1

41 (x)

< pn(x)

(b)

Show that

(0

Deduce from (b) that

)1i

x > O.

if

pn(x) < ex

Observe that at

(Hint:

1.7

x = 0

x >,0.

if

both functions start`the same.

Then determine what affect the slopes have upon the graphs when
x > O.)
10.

and

Suppose. 'c > 1

330

x,

show that: g(X) >153(x).

x > 0

(a)

If

(b)

Show that if

(c)

(i)

If

cx 3

L!

+ x

g(x)

0 < x < 3 (c
then

x = 2,

then

1),

c < 2 <-3(c

g4 4x) > g(x)

1)

for x > 0.

*What is the smallest'

integer which satisfies this conclusion?

s
(ii)

If

show that

ex

g(2) > f(2).

(d) 'By an argument involving the comparison of the slopes of

show that
11.

g(x) > f(x)

for

0 < x <

and

./

Let

2
p

(x) =l+x+x +
2

gn(x) = 1 + x +

0n-1
+

x
(n - 1)!

where

c > 1.

Show that
pn(x) < e x < gn(x)

for

0 < x <

n(c - 1)
c

(Hint:
12.

See Nos. 9, 10, above.)

Using thea...lunctions

pn(x)

and

qn(x)

as well 'as the results- of

Numbei- 11, deduce that

ex' - pn(x) <

(c - 1))J1
if'

n!

13.

0 < x <n(c - 1)

What degree must the Taylor approximation be to give

eorreet to two decimal 'places?
2
,e
< 9.)

three decimal places?

449,

43

c > 1.

ex

for

IX I 4(.2 .

r
(Use the estimate

6-5
What degree must the Taylor approximation be to give

14.

correct to four decimal places?

Ix' < 0.5,

Find

15,

0.001

for

ex

(Use the-estimate

correct to five decimal places.

Do the same for

e().5 < 2.)

e-0'091.

.es

-11f.1 (a)

Replace

by

cx

to obtain approximations to

.ecx

of degrett

<5.
(b)

Find-a polynomial approximation to

of degree < 8

Find the limit of each of the following expressions as

17.

(1 - e-x )sin x
(a)

(b)

ex - cos x
x

cos x

- e

(c)

sin ;3
Find

18.

'lib.

el

x -41

CZ>

C
4

oas

ft.

9
9

441

41

approaches

0.

6-6

r-S\ The Power Formula

The result for the derivative of.x -)e

enables us to find the derio-

tives of the so-called power functionsf

where

is any real number, ration91 or irrational.

that if

xr,

r = n,

a positive integer, then

nx

f': x

It is remarkable that -f'

'number

r,

We know from Chapter 2

n-1
.

is given Fy tk-corresponding'formula for any real.,

so that
r -1

We shall prove this important result:

9

If

x -,x-

f'

-1.

then

rxr

We start with the remark that for any positive number

loge z
= z.

If, inparticular,

z - xr
.

log

= x

loge xr = r* loge x

Since

r log
xr = e

'(2)

For ,x

need' some number, say

log

Multiplying by

r,

b,

x
."

we have the hest linear approximation,

x z loge b +

(x - b).

we get
r loge x

r loge b +

- b

A
442

Oil

ff

6-61 \
From (2)1 we have
log
x

Thus,

b +11(x - b)

.2r log b

,
e

-r-(x

- b)

r log b
Since

according"to the'elaw of exponents.

'11

log
=

= b ,

we can

( -b)

xr z br eb

'write

near '0

Now for Z.-(x - b)

b

1:(x - b)

J
and. thererore

'

(x - b)

1 +

is approximately

xr

br

br(f +

br-l(x

br

Th

y = b

ofjy = x

+ rb

r,

r-1

is the equation of the tangent line .tto the graph

r-1
This is the
The slope of the tangent is rb

(x -.b) b)

(b,b

at

r,
). .

value of the derivative at

b.

Welhave)%therbfore, established (1) for the case 'x > 0;

that is, we

at

have shown

(1)

if

x -4xr,

then

f': x --)rx

r-1
.

This is the case which is mos t important' in prastice:' The formula (1) is
*

also correct when

For x < 0,
n

'x = 0

if

r > 1.

is undefined unless

non-negative integers,

n odd.

is rational with

r =mn , m

and

In this case, (1) holds but we shall not

_...

prove this statement here.

;.
Example 6-6a.

Find the derivative tf

We can write
f(x) =

and 60 (1) to obtain the derivative

443

1
1

1,

LI 6

'iZ"---)21

defined for

x # 0.

ft: x

-2

= -

T-4

'

x2

valid for any

x / 0.

x -4 -

Note that the. derivative of

is always negative; that is, any

..

tangent to its graph has negative slope.. Intuitively it is clear from this
that

1
--4

x # 0.

- is a decreasing functioh for all

The derivative in this case can also be obtained by using simple algebra
1,

The line. connecting

(x

, X)
-)

.1
(x + h
-has slope
' x + h
%

to

t..

:.
C.
r

h
h--

-4
hxth'
3:7-'

I.

1x - (x +
h(x + h)x

h
h

h)xJ

+
-1

(x + h)\
1

This difference quotient approaches

Example
f :

x -4

x3/2

6-6b.

at the point where

= 4.

------L.a

-3

f'(x) . i x2

- 2

x = 14,

0.

Find the equation of the tangent to the_graph of

Formula (1) gives

If

approaches,

as

then

12
/

..

:
il

f(4) = 43/2 = (117)3 . 8

4.,

and
f

4)11/(

The equation of the tangent to the graph, of

y=

(x - 4).

)44411.i.

117 = 3.
at

,,(4,$) is

6-6

x -)x

Find the derivative of

Example 6 -6c1.

Formula (1) gives the derivative

2 /2ak(

Since Iff,:> 0,

x > 0.

this is valid for

xj1/2,.

1.;

Find the derivative Of

'Example 6-6d.

Since

lc'.

have from (1)

112

-1/2

;3772

valid when

x > 0.

This result may also be obtained from the definition'of the derivative

1/T-TTE - Vi

;7

h -)0

if we transform the difference quotiep

by multiplying by

4717 +

h + 16c

Then

h) - x

4747-171.'-

h(Vx + h + 7)
1

The limit is

1- 1T-177 + Vi

1
IfTc

216-c.

We can generalize the Power Formula tp enable us to

(a)

(b)

ultiply by anyconstant

dhange

to

x - a,

k,

where

is a constant.

We shall show that

pkx ra

(3)

r-1

= krx

and

Ik(x

'PO

di

kr(xL a)r-1.

.a.

Hereafter, we Alan rifer to (4) as the

have used this term for the,special case,

ower Formula.

k = 1

and

a = 0.

To establish (3), we let

f(x) = kg(X).

Then

f(x + h) - f(x)
h

kg(x + h) - kex)

oh
g(x + h) - g(x)
h

44,5

48

Previ

sly, we

40.

to approach

Allowing '11

we have the result

0,

Df(x) = krg(1) .

(5)'

we have.established

If.in particular= g(x) ='xr,

a.

Dkicr = krxr-1.

To establish (4), we let

h

a)r

f : x,-kar.

and

The graph of

amount

a.

rt.
is the result of translating the graphof

by the

(See Figure 6-6a)

y = f(x)

y = h(x)

Figure16-6a
-The result of translating the'gral:th of" f.

At the point
point

and

Q(x - a ,f(

P(x,h(x))
- a)),

the tangent; T has the slope

the tangent

T1

has the slope

T are parallel, the two slopes are equal.

(6)

Since

104x) = f' (x - a).

f'(x)

kExr-1.

f'(x - a) = kr(x - a)r-1.

Hence,' finally,

446

49.

h'(x).

At the

f'(x - a):

Therefore,

Since

6-6

Exercises 6-6
1.

Find the derivatives of the following functions.

(a)

x -)-x3l2

(b)

(0

4
x -)
3
...

.(g)

2x

li

5 x2/5
(c)

(h), x -4 20(1:147)

(2L)1/10

(a)

(1)

x -3 2 31'z

[Hints

Simplify first!
rstl]

IfEc
x
2.

For what values of

3.

For what values of

are the above functions (No. 1) defined?

.

4.

x 'are the derivatives of the above funct,ions (No. 1)

defined?.
4.

Find the slope of the curves -(described by the (Uhctions of No. 1) at

x = 1,

and at

x = 2..
a.

5.

Which of the functions in Number 1 are defined at

x = 0?

6.

Which of the derivatives found in Number 1 are defined at

7.

,Find the,derivative of the following functions:

(a)

Ix

(b)

3 )17-77.

x = 0?

AP`

(e)

-.) I2x 1.7 = /2-

X-7474

t
(f)

,4717f
x _,___-..

)9---F
- 1

(g)

b, c, d

positive constants
N..

sf77----Fdc)

--,\
' 8.

For what values of

are the above (No. 7) functiohs defined?

9.

For what values of

are the derivatives of the above functiOns (No. 7).

4

defined?
,

448

51

6-6

10.

Given:

x -)21177

4
Then find

fC(-8), f'(-3), f'(2).

(a)

Find

(b)

(c)

f'

(d)

For what interval of

(e)

Find the equation of the tangent to tht curve at x= 0.

f:

x.

is defined for what interval of

9
is defined for what inItedal of
x

decreasing?

increasing?

is

x.

x = 0.

(f) Sketch the curve, and the tangent at

3/7
11.

i
Given:

(a)

Find

(b)

When is

(c

,Find the &illation of the tangent at

(d)

A tangent to the curve is parallel to 'x + y,= 2.' Find the equa-

f'.

increasing?

decreasing?

x = 1,'

tion of this tangent line.

(f)

12.

Sketch the -graph of the curl?, and the tangent line at

Given:

x = 1.

x -'x - - .

For what values of

O

If so, what is its equation?

x = 0?

(e)Is there a tangent line at

is

increasing? .decreasing?

I
lx1

What happ"ens to the curve when

getp larger?

Find the equation of the tangent(s) parallel to.the line-5x -

(m# b,

= mx +

If the curve is tangent to

constants) at some

point on the &ve, find the values which m can assume.

(e)

Sketch the graph.

(a)

Find the first three derivatives

of the polynomial

p', p", p'"

function

x2 0 xit
p

4"

4-

r.

Evaluate

(c)

Guess the derivative of

2
f

1 +

and

p'"(0}/.

3
+

x8

x9

x9 JO

+8:: +91 +91 +3.o!

5-i

p(0), p'(0), p"(0),

(b)

x6 x7

x5

a3!

x
q.

449.

5-2
S

= 0.

6-7.

Approximations to Logarithmic and Root Functions

As we try to list pOiYTIOtial functions to approximate logarithmic and

power functions a new situation arises:

the functions we try to approximate

or their derivatives may not be defined'at

Our usual procedure of

x = 0..

-\,,

first considering the graph of a function at the y-axis may be inappropriate.

We can avoid this problem by considering approximations to such a function at
other points, or we can find the appropriate Taylor approximations to a translated

unction.

Approximations to

"At

x = 0

loge (1 + x)

the function

x -olog

is not)defined, so we shall consider

e.

S
'

tee translated function

x -0 loge (1 +

Thig4process gives the subsequent derivatives:

f(" : x.4-2 x 3(1 +

x)-4 _
.

f (5)

31

'(1 +

x -42.x 3 c 4(1 + x)
(1 +

/ f

(k)

(-1)k-1(k - 1)t,

(- 1)k -1(k - 1)!(1 + ;)-k _

(1 + x)k
..)

where 'k

is an integer greater tan or equal to

1.

We let

x = 0

of thee to obtain the values

I

ft(0) =

f(0) = 0,

fm(0) = 2t,

(4)

f"(0) = -1,

1,

(0) = 3t,

(5)

(0)

4t,

and in general
(1)

Suppose

(k)

(0) = (-1)

k-1

k > 1.

(k

is a fixed positive integer !and that

P (x) = a 0 + a 1 x +.82x2 +

+ anxn.

As in Section 6-6 the values

(0),

p'(0),

45o

53

...,

are given by

in each

677

Pn(0) = a0,

41(0) = al,

pr(0) = 6

p;;(0) Nx.2a

a3
fi

and in general

(2)

If

( k)
(0) =
n

IC

is to be the Taylor approximation to

Pn

x )log

must have
pn(0) ='f(0),

p;;(0) = f"(0),

41(0) = f*(0),

so/that
a

a_e = -

= 1,

.= 0,

In general we equate (1) with (2) to obtabn for

1,

ai4

k > 1
C

kt a

k-1
= (-1)(k - 1)!

so that

(1)k1
ifo k > 1.

'

Therefore,
2

(4)

(x) = x - L

x3*

1 xn

x4
--r+

;..

We can use (4)'to give the Taylor polynomials for

x -31pg e.(1 + x)

error

the

In Section 9 -5 we shall show that for. x > 0

to any prescribed accuracy.

__

satisfies the inequality

1)
IRnI

1(5)

xn +1
n + 1
t/

If' n

is large and

0 < x <:1

Xn +1
177717

is. very small.

expect the error estimate to be small in the interval

high degree polynomial approximation.

For

x > 1,

Thqp, we can

0 < x < 1

powers of

So)

if we use

becbme very

.large so that the error estimate gives a large error., (Of course,,we cannot

then conclude that

Rn .is large, only that the estimate of R

It is, however, true that

n

isverrolargaywhen

is large.)

is large.

is larger than

and
4

451

54CO

,
;4'

'

i"

.,:.

Example 6-7a.
og

Use Taylor approximations of fifth degree to estimate

<----

4,

,With

n = 5,
.

the Taylor approximation for

x3

log (1 +.x) z xx'> 0 tile error

R'

(1 + x)

is

015

+ ----. - -,-

22

and for

x '1

4. x

satisfies'

,D

<

We let

x = 1

to obtain
log

2 Zn. -

22

10.th error at most

-- =

This is not very good.

To guarantee accuracy

to within 0.005

we could use (5) 'to.show that we must choosy

to be at

199

least

Example 6-7b.

Use Taylor approximations of third degree to estimate

loge 1.1.

For

x = 0.1

n = 3

and

log

(0.4 1)4

with error at most

(0.1)

1.1 2: 0.1

(0.1) 3 , .09533

'-'"

so that the estimate is correct to

.000025,

'

places. *:

Approximations to, 1117,777c.

At

x = 0

the derivative or they function

we consider the translated function

x.--))/7

1/17--7.

is not defined, so

The power formula (1)

gives the iUjeetsive derivatives

1.

x
x

frn
f
4'.

./
-12

+ x)

2)(1 + x)-3/2
2)(1

(4)

1,

x) -5/2

pc p(l+-x)-7/2.,
55 452

6-7
1

in

To slIggest:a pattern we rewritethese

the form

1 -2

!tit

f(11):

so

--)2(2

; 1)(1 + x)2

11
x --)e

..

. 1

^ 1)(-2-. - 2)(1 + xi

-3

/4

->2(2

1
1

- 2 )

1 )

3)(1 +

x)2

that,' in general, for k > 1,

f(k)

1 1

- 1)

t2

...

(k -

1)Y1 +

These give the voalues,

f(0) = 1
f' (0) =

f"(0) = (- 1)
ful(0) =
f(4)(0)

ph

- ,1)

1 1
= -2(-2

-1

(2 -

2)

2)(2 -'

3)

and, in general, for 1> 1,

-1.

(6)

(k)

11

(0) = -f(-f

1)

...

(-12L

k + 1).,

Suppose
4it

+ a1x + a x2 +

p(x) = a

so that, as we found

in (2):
p

-A

Equating

p(OY=

(k)

(0) = k! a

=o,

fl(0),

f(0), p'(0) =
a

r\

+ a xn

1,

2,

...

p(n)(0)= f(n)(0)

#= 1

:\PLA'Et.11=' 221

11
-4
2 2

1, )

2!

1,*(t -

1)(1 -

24
al4

2)

=-7

83 =

2 2

1)(12

0(12

41

3)

=-

1204,

453

56

gives

a-

6-7
n

2,

and; in general, for

...
-

(7)

- k t 1)

k!

These give the coefficients of the Taylor approximations to

we have.

n = 4

For example, if

x -.

p(x) = 1 +

IT X

as the polynomiaLWhich agrees with

5
- To
x4

1._3

1 x2 +

3.7-6

The remainder

itx

,ponding Taylor approximation to

at

x = 0.

be the corres-

p(x)

we let

n,

17-77i

derivatives agree ,with the first four derivatives of

befordr for each positive integer

and whose first four

x = 0

at

if-717-

110

is then

:given by
- p(x).

ye,
R

Estimates for

We content ours el va s

are usually somewhat complicated.

n

-Nt

4,-

With stating one result:

4

.4
(8)

IRn1 < ia

n+1

if

xn+1,

,0 4

x <

where
.

..

el
Example 6-7c.

'n)

1(1 - 1) ...

-7
a

n+1

(n,+ 1)!

..,
,

Use the Taylor approxiMatitwi,th" n =

-,

'

117

ci

`2

2 2'

We have,-

4*

'-'9"1.

- ,

,_

where

<

1851x5,

(1 - 1)' ...
2
,5!%

Setting 'x:=

gives

454;

lar

0 < x < 1..

- 4)
2

_2-5g

to estimate

12d

32

_16

6 -7

.2241

with!error;

0.001.

Thus, correct to two decimal places

374 1.22.

-J

PP

(455

.
.

58

-6-7
4

r
O

Exercises 6-7

Using 6) show that

1.

['loge 2 - lori(1) I < 01005

be in order that

How large must

n > 199.

if

.e

pn(1)1 < 5 x 10-10?

loge 1.2

Estimate

2.

3. Haw large must

4.

be in order to use the Taylor approximation- ta'find

correct to one decimal place.' (Hint:

loge 0.9

correct to two decimal' places.

loge 0.9 = lloge

-log(e (1 +

= 5

log; 3.

to estimate

(a)

Use the 'Taylor approximation with

(b)

What doed (5) giiIe as the maximum error in this case?

(c)

Compare yckur result with the value / of

(d)

Now use

(e)

What do you'think happens to

in the tables.

loge 3

n = 6, 7, 8, 9.
loge 3 _1pn(2)

as

becomes large?

Find

5..

loge(1 + x)
':. (a)

lim
Z -40

(sit: i)(log-41.+ x))

6.

lim

(b)

,(1 -; cos x)

Find the Taylor approximation of degree

estimate

R5

for

0 < x < 1.

8.

iihat is the maximum error?

-/g.

X ''-) s,T4777 with

= 4 to estimate

Repeat Aor _n = 5.

(See No. 6.)-!:4! 5

to- estimate 4'

..

Use the Taylor approximation to

,.

lb

..,

n.
Usq the Taylor approximation to' x -'.A71" x withWhat is jhe maximum error? Repeat for n = 4.

ir7

Use (8) to

..'

f'lla.

9.

----T-- y.

'7. asithe Ta/.approximation to

_

to. x --t,

x -P A17 with

n = 4

to estimate

Compare your result with the estimate

411P

TP 0.707,
,

59.

4#56

4"

6-7
10.

Find

(a)

cos x - if-477
log (1 + x)

lim

lim

(b)

x -40
11.

x,
-

47)7

(sin'X)

Find the Taylor approximation of deb,' e three to each of the following:

(a)

(b)

x.,(1

)05/3

C.3

a
O

457
e.-

60
r.

Chapter 7

AREA AND THE INTEGRAL

Thit; chapter begins a discussion of the concept of area of a region

bounded by the graph of a function.

Atfirst glance, the idea 9r area appears

to be entirely unrelated to our discussions of'derivatives in Chapters 2, 4,

and 6.

Upon closer inspection. however, we shall discover that these two ideas

must be related.

Suppose A(x)

represents the area of the shaded region shown

in the following figure.

`,

As we move
Changes.

x. along the horizontal axis, the axea

A measure.of the rate of change in A(x)

A(x)
is

of the shaded region

A'(x),

the value of

the derivative of the area function atx. The change in area is also related
4'

to the he "ght- of the graph of f

At

sider ;for example, the case when

f(x)

1_

x,

that 16, to the value'f(x).

Con -

is large.

this region has

area "A.(

+ h) - A(x)
= f(x)

If we move a small amount, say

units, to the right, the area

-increases fairly quickly,' so that the additional area

fairly large.

A(x + h)- A(x)

If, towever,. f(x)' is close'to the x-axis

459 6 1

is

ti

this region has

A(x + h) - A(x)

area

y = f(x)

'then the additional area A(X(+ h) - A(x)

will be fairly small.

These considerations lead us to. suspect that there must be some relation-

ship-between the rate of chaige of the area function

of

f,

that is

A't(x)

In this chapter we ahall

f(x).

must be related to

and the values

x -41.(x)

show that for the functions of interest to us in this text, the derivative
At

of the'area function is

that is,

f;

A1(x) =-: f(x).

Of course, it is not immediately obvious what the area bounded by a graph

should be, particUlarly if

Therefore,

is not a constant or linear function.

in the'first section, after considering constant and linear cases, we deal with
an approtcimatiod procedure for obtaining thp area of a region bounded by the
graph of-a nonlinear, function (Section 7-1).

A proof of the relation

is given in Section 7-2, and extended in Se.ctiOn 7-3 to estab-

Ao(e) = f(x)

lish the so-called Fundamental Theorem of Calculus; with the geometric inter-

-.pretation:that thearea bounded by the graph of

lines at

and

antiderIvative of

is given by the difference

b
f

(that is,

'

= f).

f,

the x-axis and vertical

F(b) - F(a)

where

is any

Further notation and properties are

-,introdirced in Section 74 and the results are extended'to signed area in

.=

Section 7 -5.

The 2ina3- section discusses the--usehof*antiderivativP ft:it-mules in calcu-

lating areas.

Furtfier,antidifferentiation methods are discussed in Chapter 9

and Appendix 4.

62

460 ;

7-1

7-1.

Area Under a Graph,

We first attack the.general problem of finding the a ea of a region

located in the first quadrant, bounded by_the graph of a n

egative function

f, .the x-axis, the y-axis and a second vertical line, ac in Figure 7 -la.

shall not specify the value of the coordinate

lihe cuts the x-axis.

at which the second Vertical

This will allow usto find general formulas rather than

We shall denote the desired area by

particular numb'ellf.

We

A(x).

Figure 7-la

Area under a graph

Frequently the first step a mathqmatician tokes in attacking a new pralem is to investigate a few special cases of the problem.

He often finds this

initial investigation very helpful in setting his mind working towards a general solution.

In.this spirit we begin withkthe simplest of polynomial func-

tions and, examine the area under the graph of the constant function,
f

where

is a fixed positive number.

c,

This case is very easy td handle. -In

fact, since we know that the area of a rectangle is equal to the product of
its base. ands its height, we see that the desired

areas

A(x) = cx.
.

'

(See Figure 7 -lb.)

.x

Figure 7-lb
The area of the shaded ;region is
461

cx,

7-1

Note that the "area function"

A 1 x -4 cx
.

is a linear function w/se derivative

is

At
.

x ) c.
0
.'
The next case we examine is that of a linear function
f

S.,

x .4 mx + b..

The area we wish to find is that of the shaded region in Figure 7 -lc.

= f(x) = nix + t

Figure 721c

Area under

This case is also ea

We, recall that the area o

nax + b

to handle since the shided region is a, trapezoid.

1

a trapezoid is

-2-

the sum of the parallel based

times the height.

In Figure 7-le the trar4zoid is.lying on its side, its

f(0)

"bases" have lengths

tie desired area is

and

and its "height" 0 x.

f(x)

Therefore,

A(x) =

f(0) + f(x) .
(m,- 0 + b) + (mX + b)

mx + 2b
2

_ MK + bx.
We .observe that the derivative

of the ."area functionl'

462

64

mx 2

+ b x

is the linear functiOn

f

mx + b .

After the constant functions and the licearfunctions,' the next simplest
polynomial functions are the quadratic functions.

Even though these functions

rseem to be but a step removed from the, linear functions, we shall see thit
they introduce an entirely new order of complexity.

The reason for this '1s

that tkoe graphs-Of quadratic functions are curves, and we have no formulas

for palculating areas of regions bounded by curves (except, of course, when

the curves are circlet*.

Hence, it will be wise to move more slowly, and

first stiftN very special case--say the function

x -.)x

2
.

(See Figure

7-1d.)

Figure 7-1d

Area under

x2

If it were possible to cut/ the region up into a finite number oPrectangu-

,
lar or triangular parts we'could add the areas of the parts to obtain the total
area.

Of course, we cannot do this.

to approximate the area.

The best we can do with such a method is

We scan cover the region with rectangles and obtain as

the sum per their awes a value that is somewhat larger than the one we seek.
On the other hand, we can pack rectangles into the region without overlapping,
and obtain in the sum of their areas a value that is somewhat too small.

In

this way we may at least hope to arrive at an approximate value that we might
be able to use in constructing our area function.
Our procedure is to subdivide the line segment from

to

into a

large number of equal parts, then to use theubintervals as bases of rec.

tangles interior.and exterior to the region.

To illustrate this procedure

we examine a case where the number of subdivigions is small.

4'63

G. 5

_f
4

7:1

--)

uppo
intervals.

we divide the<ine segment from 0 to

equal sub -

unto' 5

Each of these subintervals will be the base.of an interior reor

tangleithe largest rectangle that can be drawn under the curve with this
subinterval as base (Figure 7 -le).

Each of these subintervals will also be-

the base of an exterior rectangle,-the smallest rectangle that can be drawn

above the'curve with t4is rectangle as base (Figure 7-1f).

2x

3x "4x
'55

.Figure 7-le

Figure,7-1f

Area approximated by

Area approximated by

interior rectangles.

exterior rectangles.

We see from these-figures that our desired area

A(x)

satisfies6the two

inequalities
(1)

A(x) > the sum of the areas of the interior rectangles,

(2)

A(x) < the/sum of the areas of the exterior rectangles.

Let

:7

calculate the sums-of the areas of the interior aAd exterior rectangles.

If we split the segment from

0 to x

into

equal parts, the length of

4,

each part will be

5-

and the endpoints of-the parts will

x

(3)

0,

where
is

4x

- 3x

5x

S,

f(a),
From Figure 7-1g we see that the heightjof ari interior rectangle is
height of an exterior rectangle
a is the left endpoint of its base; the

f(b),

where

is the right endpoint

its base.

66
O

Heights of interior and exterior rectangles.

'Using thesubaivisions (3) we know th\t the heights of the (five*) interior rectangles are

f(0), f(n

5/,

foxl

f(2x)
5

f(11x\.

5/,

the hetgfits of the corresponding exterior vectangles are

.

f(22-(5), v21-'5 ), f(5), f(-5-5x).

Multiplying each of these heights by the common base Itngth J,
the Area' of the corresponding regtangles.

rectangles is

The sum-of the area ofithe-interior

y'

*(D)
.

f0)
5

f('221)
5"

f(I-4)

f(122`-)1.

The sum of the, areas of the exterior rectangles, is

14f0)
5

Since

we obtain

x2

f(2'1)

f(2)!) tf(1A

(^

f(22)l.

we have

The,leftmost "rectangular region" has zercvprea.

ib

x2

f(0) = 0,A354

2x

9x

I3x.

ft-0 = 75

25 ,

A
?

4x.
16)?
f(-5-) =

and

fk

25x
25

5x.
-

The sum of the areasof the interior rectangles)

2

x[

f5

5 1.25

4x

.21.
40-

9k
25

l6x2]
25

4.

25J

25

6x3

L..

,dooN
.pro.

The sum ot the as of the" exterior rectangles

9

3 r 1

25

5 1.25

4. 16

25.,

25

25

1 1x3

25

Our desired area

A(x)

lies between these two quantities; that is,

0
25
This i

however, w

< A(x) <

certainly not .a.very accurate estimate of our desired area. , If

use a larger number of subdivisions we may hope to improve our.

estimate.

To obtain a general estimation formula, we let

subdivisions of the segment from

X

to ,x.

denote the number of',,

The length of each part will be

and the endpoints will be

(n -

0,

I
The-heightsof the interior rectangles will be
f(0)
,

f(n),

1.'(-27-)1(),

r((n"-tr)..

The heights of the exterior rectangles will:be

n' ' f(2)

n

f(),
f(=-21)
n,

The sums of the areas of)the-interior snd exteridr redta4iXes

respectively
4
'466

68
ts,

7-1

f(i5) +. f(4

f(0)

X.

,Alts,s\

f : x --x

'Since

f2X.
f(-nr)].

we have.

'0
.

..../

-. 2
t.(2S)
2L.

.11(0)

4x

,(2x)

'\-1E1

2 '

e7

2
-.

and, in-: general

'

2 2

k x,

f(_kx)

0, 1, 2 ,

;. k

n '

, n

-The interior sum (4) can then be rewritten as

2

4x

[o +- x

-2

.2

.n

/+ (n

45

-1)2x2]

3
x_10A+
, -i-i
J.

n2.

nil`

To Simplifthis we Use the formula

+

(n - /)

for the fi

4 +

+ (n -:

- 1)

1,

1)2]

sqftres
.77c1-.4' I\

6n79

1
-

1
.

We can thus rewrite the ImteriorsAm 14) as

x3
3

x3

2n

1
A similar process applied to the exterior sum (5),gives the sum of the
areas of the e*xterior'rectangles.

Our desired area .A(x)

it_

(6)4

See Appendix

(..-.-

x3

x3

2n

6n2

lies between these two quantities; that is,

x
x3
2n

x33

x3

3.

6n

x3 +- x
x3
2n

6n2

3.

467

69
4

7-1

.rery large compared to

If

n.

This must be true for each positive integer

is fixed and

is

each of the terms

x3

x3

7 2n /

2n

x3
and -7
6n
.

must be very cloSe to zero.

can have is

A(x)

the area

We summarize:

if

This process suggests that the only value that

x3

3
x --x

is the area of the region-

and A(x)

,bounded by the x-axis, the y-axis, the graph of

and the vertical line

units to the right of the origin, then

X -)

Note that the derivative of the area function is

3x
f

2
= x

2
;

A' = f.

that is,

This samd relationshipA! = f

linear functions.

was true in the case of constant and

We might conjecture tbat it is always true.

7-2 we shall show that it is indeed true

In Section.

or a wide class of functions

f,

a.class which includes most of the functions of interest to us ip this boWs.1

AP,

+0.

1.

'.

0'

04.

r.

468

kExercises 7?1

a4

1.

he coordinate axes*,

We showed in thi-, -,ection tOt the region ,bounded

-

x2,

and a vertical line at

This inequality

of the interior and the* exterior r/ ctangles:

A
(a)

(1:

x3

4.,

1
-

4.'1,) <./A.N)

2n

% ?

between the sum

as an area which

x,

<--

.0 CI

6n-

,kri

(6)

was

6n

2).
I

It follows that

1/
1

3 (1
3

- 7.
4211

-it) < A(1) : 13(1.-,-

6n

'

12 \

1
4-

2n
6n- i

when

Express this rela.tiln

n =, 5

n = 100
(b)

From

(6)

I-

we know that

23

(1

< A(2) <

6n2

2n

23(Z
\ 3

2n

J-

r'

Using directly the results of part (a),

with minimum computa-

tion, express this relationship when

(C)

(i)

n = 5

(ii)

n .100

I.Xing

function,

3
1
,for
x
x --).:T
f

the area function associated with the

>x2,

find the area in'the first quadrant of the

region bounded by the coordinate axes,

line at
!line

I(ii)

x; 7
x

31:T

469

71

y = x2,

-and:the vertical

..,

2.

If'f :7X 74)(3

A(x)..is the-

and

area of the region

depicted: in the

sketch to the right,

show-that the area

function is

4
,

using the meth6d of

equal subintervals

of

this section for

n.

finding the area

function of
[Hint:

x2.

The sum of the

cubes of.the.first

n - 1

((n

integers is

1)n\ 2.1
)
2

/ x

(n -1)
x
n

Equal ,sub - intervals of

(a)

First, show that the sum of the areas of the interior rectangles is
x

IT

(b)

Second,find the sum of the areas of the exterior rectangles,.

showing that
)(4

x.4

12)

7.- (1

and as

n -400.,

< A(x) <

x -)Tx

(1

,+

2)
n-

'

4
7

()

Next, using the 'inequality of part (b) above, and letting

find an expression for

(d)
'!.

(i)

n =5

(ii)

n = 100.'

A(1),

From the expressions found for

x = 1,

when

A(11'

minimum computation, an expressio

for

in part (e) above,,-find, with

A(2), when

n = 100

ro

7 -1?

(e)

Using

furtAion,

x --)

for the area function associated with the

>x3,

find the -8...r.e,a in the first quadrant of the

region bounded by the coordinate axes,

y = x3,

and the vertical

line at
X = 0.4

(i).

(ii) 51
3;

Find the area of the region in the

first quadrant bounded by
y = x

y = 1,

and

[Hint

y =,1

x = 0,

3
.

and y = x3

inter-'
-4\

sect at

(1,1).

The shaded area

y = 1

eqn$11 fi the -area under

minus the area under

y = x3

(betw4en

x = 1).]

xc= 0

and
*WE

.4.

Find the area of the region in the

first quadranffrbounded by
and

y = x

[Hint:

y = x

=x

Find the intersection

points; find the area under each

curve between intersection points;

find the difference between these

areas.]

5.

-7

Sketch y = x3

and

2,

[-

< x <

Fn a similar manner to that of Number 3 and Number 4, -.find the area

between the two cunves.

O

4r15

7-1
t
the ,41rea in quadrqnt one

bounded by the quarter circle

(witn cestAr
radius

at origin and

2); the line

- 2y 4- 4 = 0,

and the

vertical line tangeat to the

circle.
.

[Hint:

14

Find intersection

points;, find area

of quarter

circle by, geometry;, subtract

areas . I

7.

Find area of region bounded by

y

0,

and

y =6 - 4
[Hint:

[Ise symmetry.:.
V

4.

(a) 1or the function

x-,

we developed .in this section

inequality for the area function:

x3

Cl

ArX)

71-1

< 3 11 +

2n

12.

2n
..-

Stiow that if we average these sums of areas of interior and

exterior for

Now estimate

A (x)

we.have

5,

17

A(x)

50

x3:
.

.(b)

we

tn .PI
\5

5y 5 ' 5
resulting. trapezoids.
,

f(:)) , OS.

..

for the same function. by connecting II (0,f(0))

f(21)

to

74

172

(2x

4,121

,,

, .,

and, summing the

ti

is
(c)

As a third estimate, sum

rectangles with equal widths along the

the x-axis, and heights erected at the midpoint of each interval;

i

e
'

'?

7.,

the'width of .each rectangle would be

x ,L

,would ,be

175;

and tie heights

(a) :11'nich of these three estimates above is the closest to the exact
1

area of -- x3

iso

N
14

473

73

7-2.

'

The Area Theorem'

In Section 7-1 we found some formulas for

he area of the region in the

first quadrant bounded by the graph of a function

and a second 'Vertical line,

f,

the x-axis, the y-axis

units to the right tf the origin, such as that

-shown in, Figure 7-2a.

Figure 7-2a
f."

Ares Under a Graph

we called the "area function."

which

we obtained a function 9x -4A(x),

k(x)

Calling the indicated area

The results obtained in Section .7-1 can'be

tabulated as follows:

Area function

Function

A'

....

Derivative of area function

It is impossible to miss the similarity between the first and third

V

columns ofNlis table.

Sincqethese two,col&ns are identA ical except for

3

headiiig we are ppactkally compelled to suspect that there must be some relationship between

and the derivative

A'

A.

of its area function

jecture:`,
If

then

We con .

is thd1area function associated with a function

A' = f.

%VP

474

f,

7-2

We shall Owe this( result with the following assumptions on

(a)

f- is up

Freasing filytion; that is,

The graph of

0 < c < d.

if

f(c) < f(d)

(2)
(b)

has no "gaps" for

Condition (b) means .ehat'if

f(x).

f(x + h)

lim

x > 0,

x > 0.

When condition

(b) is satWied we say that

To prove (1) we must

is continuous for

x > 0.

that

si(oW

A(x + h) - A(x)

lim

h
that is, that ithe slope of the line through P(x,A(x)) and Q(x + h,A(x + h))
4 approaches 0. Since the indicated limit is just
f(x). As
atil-oaches
A'(x),

which' is the slope of the tangent iCie at 'P(x,A(x)),. -we shall then

know that A' (x) .f( ).

(See Figure 7=2b.)

y = A(x)

the slope of the tangent

at P is the limit of the
. slope of PQ as h
approaches 0.

Figure 7-2b

Function

Grpph of the A:

Let us'first suppose that

like that shown in Figure 7-2c.

h > 0,

'so that the graph of

The two quantities A(x) and

is something
A(x + h)

are

.11.

the areas of the regions bounded.by the y-axis, the x-axis, the graph of
and the vertical lines which are respectively

'and

units to the

x + h

475

77
4

right of the origin.

Hence, the difference

A(x + h)

- A(x)

represents the area of the shaded region shown in Figure 7-2c.

xi+ h

Figure's7,-2C'

4(x + h) - A(x) = Area ofthe shaded region

Since we have assumed that

is. increasing, the shaded region. of Figure

7-20 includes the smaller rectangle

tangle
f(x)

RSWV.

and

f(

and is included in the larger rec-

TUWV

These rectangles have base length

+ h).

and the respective heights

Thus

,hf(x) < area of shaded region < hf

h);.

that is;

hf(x) < A(x + h)

- A(x) < hf(x + h).

This inequality used the assumption that

f(x)

A(x)

A(x + h)

(4)

From (3)

if

lk approaches_ 0

is positive-end

then

approaches

Hence, if

0
approaches

f(x).

0
We can, of course, replace the assumption that

the assumption that

ex).

approaches

so that, indeed
is will give_ the same result if h <-b,
- A(x
if the assumptions
f(x)z that is, At = f,

Comparable argam
A x + h

(2) 'hold.

< f(x + h).

f(x + h)

+ h) - A(x)

If we divideby h \we oitain

h > 0.

is decreasing.

78

is increasing by

This will. invert the inequality signs

.in- (4) but will not change the conclusion.

476

III-the above Proof we used the fact.tha(

A(x + h)

A(x)

is the'area of the shaded region shown in Figure 7-2c.

if the lower limit is taken to be any number

let A(x)

a < x.

This will also be true

In other words

we can

represent the area of ,the shaded region shown in Figuie 7-2d.

The

diffei.ence

A(x + h) - A(x)

will to the area of the darkly shaded region shown in, Figure 772e.

Figure 7 -2e

Figure 7-2A
-

44;

Assuming that

is increasing for

x > a

we could, repeat the foregoing

arguments to conclude that

f(x)

A(x + h) 1 A(x)

< f(x + h),

if

h > 0

> f(x,d-h),

if

b < q.

and

f(x) >

A(x + h) - A(x)
.

If we assume that the graph of

f(x)

lim
h

as

is continuous, then

and

11

approaches

0.

A(x + h) - A.(x)

Hence,

A' = f.

ti

47_

f(x + h)

approaches

7-2
This fact that the derivative of the area function is

will 'be referred

to as.the Area'T heorem.

is nonnegative and increasing on,the

Suppose

AREA THEOREM.
interval

a < x < b and that the graph of

Fob each

has no "gaps."

in this interval, if
A(x)

is the area bounded-by the y-axis, the graph of

at

and

and ordinates

then

x" (a <75-c <la}

AF(x) = f(x).

is assumed to be decreasing.on the

The same result will hold if

In the appendices it will be shown that the thebrem remains true

interval.

under more general conditions.

The Area Theorem doean't yet tell us how to find the area function
It'only tells us that the derivative> At

x -+A(x).

must baif.

for example, the problem of finding the area function A

it

Consider,
x -,x3.

We know that the derivative of

4
x

is the function

x )4x 3

so if We divide by

4t X4 is

then the derivative of

,X3.

/\le

We call

1 4
x --4 x

x -4x

an antiderivative of

Thus a good candidate for

is

x-)Tx4

.A

Note, however, that the derivative 0

1

is also

x,

x3.

In fact, if
x

10

is any constant then the derivative of

+ C

is

14:
4

This is also sometimes known as the Fundamental Theoremof Calculus, a

subsequent theorem which can be established analytically without area argwepts.
.

80

478

72
0.4(1

.
4

so that any function of thtype

is a candidate,fqi

+ C

-A

A. For-

*,

A.

tunatelY, there are, no other possibilities for

This, is a 'ccn.-.,eguence of

the following theorenf.

then there is, a constant

b,

;'(x) = F'(x), a'C x

If

.THE CONSTANT DIFFERENCE THEOREM.

such that

4C

C,

0(x) = F(x)

.1

a < x < b. d.

A more complete proorwill be found

We shall give an intuitive a;rgument.

8,

--in"-Chapter

If

'Proof:

the graph' of

G' (x) = F'(xl,

the same slope at each

on the interval

ra,bi.

either the graphs are the same, `G(x) = F(S))

c)`

grapj; of u.<3

and

haVe

This can happen only, if

or if one graph can be obtained

by raising or lowering the other a certain amount

constant

<

'0(x) = Ft(x) + C

4.00,
(See Figure '7-2f.)

fer.some

Ir

t).

Figure 7-21
When

x = a,
;(a) = F( a) +

Therefore, Ithe Constant is

)

C = 0(a)

= Va.) .

11,

14.79

ogee!

a,

pip

gxa mp le: 7-2a.

It

Al' = f.

Find' A(x) ' .if -14'

.

We kncrthat
that

F : x

.1 4

.;,.

x aw xv.
,.

,.

then ;

The tArea Theorem tells us

From'the Constant Difference Ttieprein, since A' = F',

must be a constant

, such
.

that

there

'

/44(x) = F(x) + C.

To determine
say

=' 0.

we need to know A(x)

C,

Since F(0) = A(d)

.x

and

C = 0.

,ar

-F(x)

one value of

x,

=o

o=o+c

Therefbre,
,40

A(x) =

1 X4.
4

Example 7-2b.

Find the area between t'/graph of

x-a3is and the lines

x = 1

'V

and

x -4 x2 + 2x,

th-e

v2.

A'( ) = x2 + 2x.

= x7.+ X2

F' (x)

2"

x + 2x.

By the Constant ,Differenft Theorem

A(x) = x2 1+ 2x +C
for. some constant

C.

Since A(1) = 0

0=1+2+C
and

C =

-3

and

A(x) = x2
r

Then

A='A(2)

= 4,+ 4 - 3 = 5
A

We-need a natation. for the area

the graph of

-1-2x - 3.
is the required area.

of) the region bounded by the'x-axis,

and the two vertical lines given by

x = a

and

(See Figure 712g.)

ti

480
I

8 2-

X = b.

7-2

'40

-Figure 7-2g

Area under a graph.

The usual symbol for this is

b
f(x)dx
a

suggested by the procedure (described in the previous section) bf approximating

sums for finding areas.
summation.

The

rectangle and the

f(x)

The symbol

"

"

(a modified letter .'S) indicates

is meant to suggest the ordinate of an outer or inner

"dx"-- (am indiviiible symbol) the-difference in the

x%S". at

the ends' of the base of a rectangle.

4

The symbol

f(x)dx

may be read,"The integral of

We shall sometimes vite this Nintegral more briefly

f.

44
0

from

to

b."

7-2

.
.r

Exercises 7-2

In Section 7-1 we obtained the estimates

x3
3

,3

5(3 'x3
+ 7 <
2n
6n

for each positive integer

A(x) <
3

X3

X3

Tri

6n2

where,

n,

A(x) =

f :_x

f;

-4x 2.

. Averagthese to' obtain the general estimate

A(x) tt

x3

6n

4
Use this estimate for A(x)

in order to calcu,late approximations of the

following quantities when* n = 10.

(a) .A(2)
(b)

A(2.1)

(c)

A(2.1) - A(2)
0.1

(d )

(e)

2.

A(x

A(x)

Suppose
.(a)

(b)

for general positive

Let 2h approach, 0

x2 +1!1;

lim
h '-4 0
lira
m

x,

in (d) and use this to est:if:nate

Find

1+12.

"..

f(x)dx
1

--' -h-40''

l+h
f(x)-d.x

1
14-h

(c ) Did you need to calculate

f(x)dx
1

and (b)?

Explain.

t_

482

in Orde.r

A'(x).

.
7-2
x

3. Suppose AN) =

f,

where

x -4

2'

<4

(a)

What is A(2)?

(b)

What

(c)

Did you need

find an antiderivatie for

t'o

o answer

in order

Or 4)?

(a)

4.

is At(3)?

such that

Find two'idistinct functions

g'

is the function

-43x2.

How are your functions related to each other?

5.

Find the area bounded by the coordinate.axes, the line

grapfi of the function

r; ..a

6.

and the

where

x -4 x2

( a )

(b)

f : x -42x + 1

(c)' f

f,

x.= 2,

+ x

4x

x .-4

x2 + 1.

(a)

Sketch the graph of

(b)

Mark the region bounded by this grkph, the coordinate axes,

x

and the

1.. Find the area of th1s region.

yourograph, the coordinate axes, and the

(c4* Mark the region bounded by
Find the area of this region,.

ling x = 2.
(d)

and the lines

'Mark the region bounded by your graph, the x=axis,
-How is this region related to 411e regions you
2.
x = 1 and

-(7

marked in (b) and (c)?

Find its area.

graph of
'Sketch the graph and find the area bounded by the
f

x -416 -

x2.,

the x-axis, and lines

x = 2 'wand

= 3.

graph of
Sl,4etch the graph and find" the area bounded by the
the x -axis, and the lines x = 1 and x = 2.

-44x3 - x,.

8.

For

-4 (x - 1)2

show how the .interval

so/that on each subinterval

Sake a

<3

can be, subdivide

is always increasing or arways decreasitig

sketch.

4830
C)?
4

0 < x

,
7-3

The Fundamental Theorem of-Calculus

7-3.

The following theorem summarizes the method for fiudtng area functions
This theorem is generally refer;ed to as

explained' in the previous section.

for cal -

the Fundaittentel...in.40.rem_slf..CaleatiirS'crftlfrVigra1:;?c,
culating areas by using. antiderivatives.

THE FUNDAMENTAL THEOREM OF CALCULUS.

If

is nonnegative,

twincreasing and its graph has no gaps on the interval

and if F

is any function whose derivative is

'interval,

a < x < b,

on this

x'
f.= F(x) - F(a), a < x < b.

The area function

Proof.

+
.

A(x) =

f
a

ii a function whose derivative is

f t'rom the Area Theorem).

A(a) = 0.

and

Since the functions

have the same derivative,

Constant Difference Theorem implies that there is a constant

A(xp )

= F(x) + C,

Then

Furthermore,

f,

the

such that

a < x < b.

C = A(a) - F(a).

Since

A(a) = a

= -F(a)
and

A(x) =F(x) - F(a).

x

That is

cf =

- F(a),

a
I

Remark.

1L,

This' theorem will still be true if

is atsuAd to be 'decrees-.

j/ ing on the interval, for the Area Theorem will remain true and the above proof
can be repeated verbatim. The theorem ,is easily, extended to{ the ca e when the
interval can be subdivided into smaller inte'rva s, on each of which) .VIFicreaT
ses or decreases. For example, suaose.thati F = f and that f
increases

for a < x < c'

and decreases for

tg. < x < b.

See Figure 7-3a;)

*
It relat s differentiation and integ ation.

484

86

.o

Now

7-3
*b

f = A + B =y

f +

f.'

.e P

.0..-

1e apply the Fundamental

orem to e

h term id obtain

cc

F(c) -

Fka),

F(b)

f =

,,`

When we add the two integrals the term

drops out.

F(c)

,a
We have

,S"
'...re

l'b

If

= F(c)

- F(a) + F(b) - F(c)

= F(b)

- F(a).

Figure 7-3a
I,

Area of Shaded \Region = Area of A + Area of B

When
integrals

we deal with.speeific functions we shall use the longer notation for

,

Thus, instead of writing;,

4

f where

ex,

we shall write pimply

I.
x.

(1

dx.

Pr
/

x,

In this example

F(x)

ex

tells us that
.A = e

1
-

e0 = e - I.

485

87

and. the Fundamental The rem

Since we can describe f equally well by

t -4 et

we could replace (1) by

et dt.

A=

(2)

Because

De

e`',

Fit)

102

and by the -Fundamental Theorem

A = el - e = e - 1
exactly as before. Because tIle result does not depend on the letter used,
the 'letter x in (11 is called a dummy variable.

t t'

Example 7 -?a.

->

dt. T'ne derivative of

Find A(x)

Hence t're derivative of

t'

is f

undarnental Tfeorefi,

t t
I

is

By the

-4

A(X) = F(X) - F(2) 1 V

F x'
r 772
Example 7-3b.

32
.

cos x dr.

sin x is.a function whose derivative is f.

The interval can i.)e subdivided into two subintervals (namely - < x < 0

F': x

TI"le sine function

0 < x < ;) so that f increases on the first subiniterval and decreases

on the second interval (see Figure -3b). We can, therefore, apply_ the remark
following the Fundam'arital TheOrem to conclude that

and

F(7-1)

= sin
-

it)

(.7.1)

6 = 2.
ti

r.
,11011$

\'!\

7-3

IT

Tr

increasing

decreasina

.0

A function=-F

whose derivative is

integraa'of

(or

is._called an antiderivative

It is also common to use the notation.

f.

for

F(x)

F(b) - F(a).

The Fundamental Theorem of Calculus may be stated in the form:

cb

(3)

'

f(x)dx = F(x)
a

= F(b) - F(a),
a

where fF

is an antiderivative

of

For example, since the derivative of

x

25

we say that

x3
3
I

-I

is an antiderivative of ex

a3
2

b3
3

and write

31
3

Ft = f

487

, 89
e

5
11.!

'1,

//
4

Example 7 -3c.

Find

dt

t -4 t5.

First we find an antiderivative of

ials reduce

by one,

Differentiation of*polynom

ntidIfferentiation should raise the degree

the degree by one, so

If'we recall that the: func ion

a

'has the deriVAiv$rft )ot 5 ,

we can see that

t6

is an antiderivative of

5
t

Therefore, we have
4

y4

t5 - dt =

Example 7Ei.

40951

Find the area of the region between the x-axis and one

arch of the sine curve given by

..o"

Y = sin x.

We want t? find (Figure 7-40.

sin xdx.
c

y = sin x

Figure 7 -3.c

sin x dx = area of shaded region.

O "
P

The derivative of the cosine function is the negative of the sine func'tion so that
/'

90 )'.COS X
488
ro

///

7-3

.is an antiderivative of

We have

--!sinx.

sin x dx = -cos x

it +

= -cos

cos, 0

7(11) + 1 = 2.

Example 7-3e.

Find

(x2 + 23c+ 4)dx.

c0

of

We could find an antiderivative

x --3-x

(3).

directly ind use

+ 2x 4- 4

.An alternative approach (which amounts to the same

thing) is to remember that the integral of a sum is the sum of, the integrals,

so that we can,wri)e
3

(x2 + 2x

4--

11)dx =

2
x dx +

4dx.

The funttions
x

2x

x2, x

x -414

and

have the respective antiderivatli.Les

, x

x2

4x;

end

so we have,,
3

(x -+ 2x + 4)dx
-1

+ 4x
0

()

Q3)
-

+x2

(e

02)
4

= 30.

Example 7-3f.

y = & and

Describe the area of the

egi'on between the graphs of

y = & a's the difference of two integrals and evaluat

'

489
e

91

7-3
A

,The area of region

in Figure 7-3d is
1

dx -

dx.

Figure T-3d

Area of A = Area of B - Area of C

Q

of

To find antiderivatives
= x1/3'

and12
'

x12.

ATc-

3E

and

we first write

and then recall the power formula

DSca = axa-1.

Here differentiation amounts to multiplying by the exponent and 'reducing the

exponent by
by

Then antidifferentiation amounts to raising the exponent

1.

Thus, we,.-have

1 and dividing`by the new exp'onent'.

3
4

4/3-

as respective anAdeLvatives

of x

desired area is

-ea rid

2 3/2
x
3

-x

3)/3-c

and 40x

x4/3

Il

12

rt

Evaluate

Example, 7-1g.

!sin xiclx:

-11/2
'

c
*am

.490

.1,

\*
#.

92

Therefore, OUT

0
7-3

Ih Figure 7:-3e we indicate (by shading) the region whose area is the
integral we wish to, evaluate:

Figure .7-3e,

We know that the area of region

of

should; suspect that the total area

(fro in Example 7-3d) and we

x.ego?ons

and

-A

Is

We can con-

3.

firm this. suspicion and gain additional' eXperience using antiderivatives.

By

definition of absolute value we have

ris

sin 'x,

for 7 sin )e > 0

'sin x,

for ' sin 'x < 0

x -4. 1 An Xi' --=,

,-._

We express our integral as the sum of two integrals:.

.....

.e

le

I sin x clx =

.,

t- t

,..

.(4)

,..

;'
......,

Co

,.,

kin xlilx +
;

kin xjdx

-,

...,,,_

..

it

sin ..i dx

.../2

.
:?/

Antideriptives
-sin x and

.:.,

'

le.

p _.'n x,'c)dx
,C-f
)

=
illi.

,,,r

'%

7in x

.e

are, respectiivelyr
to

.,

and

x --) cos x

x --) -Cas x.

Theifoe, we have

T.
>,

',

,.

i'-.1-,

'sin xjdx = cos x

os
,

'N'{'

"x),' ..
Q

'''

10

,!_

S.tCOS, V) -( -cos 0)

= COS 0 -COS( -.X

.....

.1

,V'

.!

O + (-(-1)i1,44-1)
:,....

.
:
.

..-

,,

//I

,...;.

.. 191

-,

P. ii
...;

i
,
O.

4'
i.

,'",
I

'

'

t..

,..

'

ij

1
'

'''1)

',.: t

-.

r,,

-*

t' ,

,"
1

II

773,
\

Example 7-3h.

Evauate

f(x)dx

if

f(x)

(2x - 1)

for

0 < x < 11

for

1 < x < 2

The area, of the shaded region in Figure 7-3f is given by the integralswe

Note the break in

wish-to evaluate.

the graph.of

In order

x = 1.

at

able, to apply the Fundamental

to

Theore

of Calculus, we first break

our in

rval into subintervals over

Which the gra

has no gaps:

of 'f

f(x)drF

1;7 dx +

(2x,- 1)2dx.

Antiderivatives for
.2
(2x - 1)

and

x :4 11.7.17

are respectively,

and

3-{-3 x3/2
3

-)4(-P$(2x

1.)3.

(CheCk by differentiation andesee

We theriPre

Exefcises 7-3, No. 5).

have
.

2
f(x)dx

_,2i
V2if 3/2
3

312

-0

+.

1).

(2x

)+T(3 3

312.

3.
-1)

+ 13
Figure '7 -3f

,492

94

7.-3
Exercised
L.

I1ind each of the following integrals.

2

.(a)

(x2 + x + 3)dx

(S)

CO
(x

(b)

--2

(c)

+ x + 3)dx

ex

(k)

dx

1.-1'

x2

+- x ,+ 3.)dx .

(2)

-2,

.r

(ex + 1)dx

J -i

/3.'
(d)

xn dx
J

Cos x' dx

+ x)dx

.--4,

j.2

. 2

ix" dx

(e)

(5x

(n)

0
1
(f)

+ 3x2 + 1)d)1-4:

1
,-

(ix +

4,
gx)dx

rt

7 (o)

1'1/16 -+

(sin x + ,cos,x)dx

it6

1111/3 x
(g)

3x12
dx
111/2

-Cr-

(h)

(p)

*.c

-1

(e_

+ sin x)dx

t,

;(5x- 6 + x2%)dx*

(q)

.f

-2

(x2 + 2x

+ 5,1

dx

_. .

2 .
(r) .

-'-' dx

(1)

1.1

3:

tan z dx

10 -

x.

t,

2.

Sketch the regions bounded by the x-axis, the curve

vertical lines

(b)

f
f

(c)

-)x3

-x = a

and

+ 2x +

1,

x -4 ex,

8 = -1,

x -4ex +

x2 ,.

in x

(e)

4
x.+ cos x,

f)

P.: x -4x

(a)

a =

1, b = 3

a = -1,

y = f(x)

Then find the areas

x = b.

and the

.'

1
k

(d)

(g)
.

$
4

-10;

-) 3fg
x ;

Cos x,

.1

a = 0,

a = -I,

b = 121-

2 '
-

= 1,
a

ti

7-3

vertical liner; then find it.:

and the given

y = f(x)

Sketch the regioki bounded by the -)Axis,

,ire,a....".

le

z =

vertical line.,

f : x --) !xi;

(a)

ometiv.)

(Check your result by elementary

(b)

4x

y
(c)

':

' 1

.(d)

at

vertical line

x --, !co, xl;

x ..

-.

,-

x = 7

- ,- sin xi;

./

x - -1,, x - 3

vert,ical linesat

r4x31;

x --4

x - 4

--,-!,

vertic1 lines at

x = -y,

x = 2y

(e)

-f: x --)

(a)

4.

;1_

, il;

Eval_uatt

(x-

and

3./7c)

w_ 1

0) = 1, t(1) = -1.

wl-ere

+,loge

7(r) - (1(x

(b)

31/7 + 50).'

x = 4

x,. 0,

yeftical lines at

N.

Find

(c)

G(x)- 11
0

Nat 15

if

G(x)

F(x)

5`.

(d)

F-nd an antiderative for each of tlie following functions.

f

(x - 1)3

2'

(10

f 3%

3x

g 4 ,X

Sx' - 12x.

x >

4-

6x - 1

'

1)3'

in thprform a(x - 0) 1.10

{Hint:
(b)

F1 = G'?

F with

Compare tin(; fanctions

and

with

g.

Compa're the

antiderivatives.
6.

Find an antiderivative for each of the following functions*

f
g

-) 3(x +

1)3

t'4

Cl

7.

Find

(3x + 14)5 dx

(a)

by first carrying ob

(b)

by Ming the method ifounin Number

e indicated multiplication,

494

90

6.

7 -3

b
.

3 dt?

1.1e following integrals are the same as

Which of

t
a

dy

(c)

a
.s

b
y 3 dt

(b)

d)
(d)

(t

a;

dt

a+1

a
.

9.

.)

b+1

Eyaluate,the following integrals using aelihe of symmetry appropriate

3

to the problem.

[et.g.,

7t/6

3 x2 dx =
3
0

dz: = 2

-3

(a)

.40

J(3

= 18.]

coo x dx

n/6

(1 ' 6x1di
-2

fc)

sin x dx

(d)

'r

10.. Find the.area,of the region bounded by the x-axis, the given curves
(Sketch:fi;St.)

and tie given vertical lines.

-x3

0 <)x

-x2 + 2,

vertical lines et

x = 4

< 1 /

cs

I < x

and

.x = 4
,

0 <x <3
vertical lin9s at

x <0

x = -

...

r.

1 and
2

x > 3

at
e

'

2
4

(Sketch

In Prcblems 11-12 deduce part (b) frOm the solutions toppart (a)..
each first.)
t

11.

(a)

(i)

.(8 - x

'Find

2)

dx

(ii)

2
X dx

f0

(b)

Find:the area of the regiOn bpunded above by y ='8 - x, below

2
by y = .x , tp the left by the wertical line x= -1,' and to the
...

ht by the vertical line

.

1.

x ----

,--'

495

'

",..-.'

97

...

t
f

Barn

7-3
ti

12.

(a)

(i) 'Find

(b)

x2 )dx;

(8
0

x2 dx

(ii)

A'

Find the are a Of the region bounded by ,y = 8 - x2

y = x2.

and

13,

(a.)

Find the solution of Number 11(b) dire tly Without using part (a)
of'Number 11.
..

(b)

,.

Find the solution of NuMber 12(b) directly Vittlout using Part '(a)
of Nthather 12.
v..

14.

Find the area bounded by

y'o=

sin x,

y = coa x,

x fib,

1/(Sketch

and

.4*

ti

,o

(4,

9136

x =

7 -4
.

7-4.

Properties of Integralb

We have seen that the integral

b

f(x)dx)

(or
a

can be interp'reie4 as the area; of the region beloW the graph df

/

the x -axis and between the ver4ic61 lines

x = a

and

above

f,

Certain pro

V.

pertieb of integrals ate immediately suggested by this area interpretation.

Since the area of a region like that shown in Figure 7-ka'ahkild be a
nonnegative number; we have the result'
b

If fx) > 0

for

a 5 x < b,

then

> 0
f _

Figure 74a

'

11

b'

Area under a graph.,

.

lai-ger regioh.
Also the area' of a?egion should not exceed the area of$any
.Aeful formulation of this idea is'the following:
.

"(2)

If:711?(4

< g(x), for a < x < b,

41e1-1

,
V

/I

f <
a

g.
a-

See Figure 7-4.)

a.

497 -

99

4.

Figure 7-4c

eFigure 7-14b
1

The mrea under

f_ does not

.,

m(b

exceed the area under

- a)

<

- a)

An applicatitn of the inequality in (2\ -gives bounds.'for the'area in terms of

bounds for
--)M
1

:-'

f.

Supi5ose

f(x) < M

is a corist4t and

for

we _can apply (2) to obtain

.:

f <
a

M(b

.!

With

a < x < b.

- a)..

<

to obtain m(b -

m <

Similar argt.mients can be dpplied if

f.

a
(See Figure 7-4c.)

In summary:

then

a < x,<*b

if -m < f(x) < M- for

(3)

f < M(b -

0 m(b - a) <
a

A line has no width and hence, hag te

area.

Thus, if we take b = a,
..

we should expect, the area to be zero, that

,a

f =

(11)

//
/

1.

This is consistent

Ti t

h the res.ult

(3)

for

f 'Ae'takm

..-

0 = m x 0

<ca f <M \x

.
.

r.

CI -

q =, O.

b = a

wed obtain

-"!
If we multiply ordinates by a. constant factor then we/expect the area to
,

be changed by a corresponding factor. 0ne useful conseg ence Ofthis:

If

(5)

g(x) = af(50,

iseF

A < x < b,Y

re

is ,a

\
iv

positive constant, then

a
a

-( See Figure 7-4d )

of

Figure, 7 -4d

The area under

%
f.

g tis a ttimes the area pander

If one region is the union of two non overlapping regions we'expect the
14

area ofthe firs't region to be the sum of the Leas 'of the subregions.

This

adaftivity prY.nciple'has two useful consequences, (6) and (7).

.

(6)

lies between

'

.4t

b,

then

'

cb

f,=

and

f +

f;

0
that_ is, if we cut'the'region under

the sum of thes two resulting areas:

(See Figure 7-4e.)

by a vertical line, then the 4.(rea is

'

Figure 7-4e
ti

The area of the region under the graph of

between
%

of regions

and
A

and

is the sum'Of the areas

B.

7-4" '

,.

.....

.: A seciond.useful formulation of additivity is obtained for the sum of two

The sum f + g is defined as,the 'function whose veLe at x is
=t

graphs.

1.647+ g(x);,

that is, the;graph of .f + g." is obtained .by adding. the ordinatas

of the:g aphg of

and

We nave

g.

(7)

cb'
a
.

(See Figure 7,-.400

The area of the region under the. grapl of6

plus.

thearea of the-region,under the.graph of

the sa:4'of the region under the graph gf

f'+ g:

is

Each o These principles can.1De deduced froM the Furidamental Theorem.

.

We,proVeseveral of them here, leaving the otheri as exercises.'

Example 7-:4a,

Prove that

f(x) '>.0

.,

.0"

for .a < x: < b

then

f > 0.
a

1 'Let
Since

'11,(x)

> 0

F1(x) > 0
and

44.

Fi(x) = f(x).

r- Increases on the interval

for a-< x.< b

for a< x < b -ia,b1.' Hende/.

F(b) >'F(a).

.*\
Then

.%
:

b
-

Jo

F(b).- F(a) > 0.

500

02

:1
e

Example 7-111D.

trove .that

if 'g(xi) = af(x)

(5)

fof a < x < b

where a' is a positive constant, then

Ve

r.!

rb

b
c

1(x) = g(x)'

F'(x) = f(x). (XF(x))' = af(7) = gx).

and:.
Then.

Since G arid, aF have *the tame derivatives

G.(x), = a'F.(x) + C.
4,

Now

g = G(b) '-G(a)

= TaF.(b) +
.

= ocr:F(b)

- r,:aF(a,)

C]

- F(a)1

f.
Note:' This prodf is equally valid

if,

a is a negative constant.

.
1,

501
.
\

103

r.

txercises'7-4
that is

1: Proye
4

.%

.,"

= 0:
a

.1

g ,is equivalent to
a

r
and then using (1).

,(i3

3.

f <

Prove (2) bytisin'g the fact that

2.

.1.

.0

Prove that.

f = f(x)

Let

Hint,:

F'

ar.d apply the Fundamental ,

= f

a
Theo?em.

4:

and

f + g

Prove (7) bY showing that

.

deri7atiVe and that they are equal at

5.
.

Suppose f
'10

x2, g

(a)

Graph each.

(b)

Show that

(c)

Verify

2x '+

ex)

.f(x)

'

for

f <

x
'f +

ca

x = a.

have the game

S'a

Use Number

Hint:

3.

3.

0 < x

< 3.
,

6.

Over the indicated interval for the fo

tion; find the maximum

'fplu'e

(M)

of th

owing functiotts:

graph the func-

function;, find the minimum

(m)1

.o

value of the funciion-t and, using' these, express with an inequality the

lower and upper bounds of t:he in:teg1 expression for the area. ,(Hint:
See Figure 7 -2c.]
6

-(a)

(b)''

f
f

1,

0 <

x .-4.x2 - 2x

+ 3,

and

x<

0 < x

<3

.t,.

.7.

For

3x '-r 2

g =

find

f 10
g =
4.

5,

f.

104 5d2

and verify that

7-4

;12
-

g : x -) -2(x - h) + 20.

and

+ 20

For f : x

8.

(a)

f(7

CO

l'4

:7

f*+

f = '

and verify that

g.

H 0

Tilus

c7

Find,-JO
3

f and

Graph

g(10.

show that

f(3) = g(0),

jF, nd a,suitable translation such that

..c-0

1,0

f.

3
?it

9.

3x + 5, g

For

x tand

: -x

verify

-41

that

g '4- 5

.-- 3

h.

a " :a
.

19.

Find each of the following integrals) after first graphing. the given

(a)

.r(x2

'

4'.

function oven the interval.

x)dx

-N.

4x+ 5)dx'

(b)
1

ft'
fx2

(c)
.

2x + 31dx_

1
;

1
2
(T x. + 7

(d)

''

- i)dx:

11..

Suppose ,f

whei:e

x' -4 px 1+ qx + r

i'

conaants. '
.

(a) .4ta

F-:

(b)

r .are noaegatjive

and

p, q

x. -) 2 x-3-4:

Show that if

a2 x2 4- rx

F' =f.

')--'

0 < a < b

and show that

then

cis

J0

A.

(Hint:

00

-a
,

_ .

503

105

7-4
12..

showC4at

In Exercises 7-1, *umber 2 'it was

7x

f -

Suppose

1+

,io.r

px3 .q.x2 + rx'

for

f,

x >O.

where T,

s,

Find

s 'are nonnega-

Suppose also that

teive cbtastants.

x4
3

x3 4. x2 + sx.

..,

Sow that

(a)

G' = g.

.
b

CO

g = G(b)

0 <a <,_ b 'th4 ri,

. ,

Show that if
.6

.- 13.

In Number 11 p'ut

(1(50 =

-F(x) e+ 1000

and show% tlet

F.

114.0;-.2.I'..d

A.

-- 15.

!X -

'

111.
1..

0(;).

f? gra p- is, of course,

',alp ul.)

..._

Fintthe area and graph of t' e region bAnded by

'and

f = G(b)

x2 dx.

-10

16.

"4int:

-3

'Find

2.rdx.

1,

.1)

,)..

, -.% ,

S.

- G(a)..

''; n.

(I-11.nt:

Translate and grapy

Find*ithe area of the region bounded by

ti.e

y = 2'x - 5)2 -,2

area_into t'e first quadrant.)

y = -(x

,-

1)2 +1

and

y = x.

.11

fref
5C.*

7-5

7-5. Signed Area

only

f(x)dx

ar

Until now we haye.discussed the integral

and the interval from

'in cases for which 'a < b

.b

could bea'subdivided

Was nonnegative, always increasing

sO that in each subinterval the function

to

/'( T always decreasng).and its,graph had no gaps.

a > b

o include situations'for which

We now extend our discussion

or for'which the graph of

may con-

,:

tain portions below the x-axis, preserving, if possible, the result

irb

if

- F(a)

f.(x)dx = F(To)

F'\

= f.

b ',

""'

....,.

fCxIdx as zigned area.

This can be accomplished by suitably interpreting

.--,

First conlieer the casegfor which' f .is nonpositiveon the interval

'tive
f

-F

and

a < x < b,

F' = f fn this case

so that

.-

is nonnegative and has antideriva-

-f

7$

-f(x)dx-= -F(x)

= -F(b1 + F(p)

'

71

'

This can-be tn erpreted as tWarta of the shaded region of Figure 7-511.

that this is the seine as the area of the shades region of Figure 7 -5a.
.

Iv

ot
ti

y = -f(x)

a
')

Figure 7-5a

=f(x)

Figure 7-5b.

If the. Fundamental Theorem is to h8ld we'shauld have

r

f(x)dx = F(b) - F(a).

a

Referring to (1), we seep that this requires that

b
."

E-f(Xl]dxc

f4x)dx = a.

Fes.

.4o.
o5

. .

3.0 7
L

Note

..

7 -50- .'

).

#1#

that is

f(x)dx

mdst be defined as the negative saf the area of the shaded

region of Figure 7-5a.

)
= f(Z)

Figure 7.75c,

Now suppose the graph of

that

::e-looks like that shown in Figure 7-5c and

W;,have

is an anti:derivative of -f.

4###,

area of Al

-f(x)dx = F(-c
a,

- F(a)'

c,.)

area of A2

= ,f.

/-f(x)dx

01
1'.
.
.

area clf A

F(c1)

F(02)

,..,..

1:

'

.f(x)dx = F(1;)- - F(c2)

=
3

c
2

Now note that,,,

, F(b)

F(c2) - F(aD

F(a) =" F(b), - F(c1),+ F(b1) - F(c2)

[F(c1 )r- F(a)] - tF(c
,

F(c

[F(b)

-(area of A1) - (area of A2) + (area of A3).

Ivjer words, if we wiLh

tg be

F(b) - F(al

- then we must have

,

f = (area of A1) - (area of A2) + (area of A

a

'1

).

F(c )]

In summary, if

and if we. di'ine

a < b, F' = f,

b
by

f = F(b)

(2)
a

then

will be the total area of the regions bounded by the Hgracl, of

A

which lie ;above the interval mints ti e total erea of t'l e ,regions bounded by
,

..

the graph' df

which lie below the interval. Ti s is called the signed area

..
.

determined by

on the interval from

,,f

to

a
.4

..>

b.

..

It is als8 common practice, to remove the restriction tl at

by

a, < b,

defining

.
a

f = -

b < a.

if

The ,fundamental re1'ation (2) will still hold,

.

ipr4

and

b '< a

'if

,c;

,.,,,>
? ',0%*

411,11*

then

tea

b
%

F' .= f

14

-Lz..

Fjb)

F(a.5 .

b .

The properties of the symbol

..

discussed in Section

..

ti

"-4

also hold
,4

for signed area:

b
(f + g)

(3)

b
f

4-

g;

(4)

where a
e.

f,

...,..

is any real number;

.
..,,

i"

(5)

I'

'

b'
s

f +

f =

f,

where

a, b, c

are

my

real numbers.

.
4.4

Notice, in fact, that (4) now holds without the restriction that
negative and

(5)

dbesn'A require that

Of courses if

and

f(x) > 0

< c

for

.
b.

f(i)dx.-> 0,
a

One conseqiiencel of this is the fact that

,909

cx

< b

then

a be non-

7-5

'

e(x)dx <

(lit)

g(X)dx

it'

a < x < b

.f(x):< g(x).'

and
i.

7.-For-We-then_have __g(x)

f(x) > 0, so that.

Adding

b*
f(x)dx

to both sides, we obtain (6).

1

(
ExaMPle 7-5a.

sin x dx.

Find

This integral can be interpreted as the signed area of the total shaded

region shownin Figure 7-d.

Since the regions above and below'the x -axis are

1

ke4

I
Figure 7-5d
'

y = sinx

the same, we should expect that the signed area is

(2) should cOrrobotate our expectation,

0.

The defining r0.ation'

In this case
&

F.: x -;-cos x

is an antiderivative of

x 74 sin x,

so (2) giyes
%

1.

.v.

sin x dx = -cos x

=''( -cos 10

-(4)) =

=. -( -1))

9.

508

110

0.

cos(-v))

134,

N
7-5- .

Example 71D.

Sketch the graph Of

61
Find '

-41

-2 < x < 3.

for

x2

f(x)dx,

[-f(01dx,

Pif =

.a hd

-2

[-f(x)]d

C =

/Use the fundamental. relation (2) to show that

The desired graph is shown in Figure 7-5e.

40.

66.

g/

Figure 7-5e.
y

x2.

,t6

-,<-4/
f

/
vir

7-5

t'

Thd function. F': x -4x -

is a antiderivative for

x3

(as

We have

Ft = f).

easily checked by showing that

ti

-1

-1

-1,

C- area of
3

1,

(1 - x )dx =.x -

f(x)dx = .

1-1

'3

..i.

..i.

= area of A

=
3

2'

[-f(x)idx =

(x? - 1)dx =

f.

,3

]1

ao

= 7 = area of A3;

P-(--

1
.

he fundamental relation ,()-givei

x3

- 47(-2) = x -

f(x)dx = F(3)

2
f-3

'20

-2

which is the same as

- (area of A3).= -

-(area of Al) + (area of A2)

o

Examplp 7-5c.

Fad

x7 dx

and

i1

dx.

We have
1

2
x

dx = -

dx =

Example 7-5d.

1
3

Find the area of the region enclosed by the graphs of thee

two functions
2
f

and

6x +

X' -g -x2 + 7x - 11.

A 's etch of the region whose areais sought is given in Figure 7-5f.
o.

510
,

112

7-5v

IP

(2,
1

,?t",

c.
-5f

We shbil show th,ot the desired

rea is given tv

'?/2

(g(x)

(41)dx.

First we note that

9/2
a

g(x)dx = -(area of A 1 )

(7)

rea of A

+ (area4f A3),

where

A
l'

and
2

are the regions indicated in Figure 7-5g,

A
3

ti

-I

/
/

(2,1)

/ \\

I
/

A3

t
%

I.

%./

1
..-

1'

Figure 7-5h

Figure 7-5g
efl

"1'13
,/

7-5'
Then we observe th'at

(8)

f(X)dx = -(area of A

.2

//

A4

Where region

- (area of

area of A3),-

is indicated in Figure 7-5h.

(8)

Subtracting
9/2

from

(7),

we ot4tain

.9/2

g(x)dx 1.2

f(x)dx = (area of

2 c

(.area of Ali),

....)

whichlis the area we seek.

.

g(x)dx

1 9/2

9/2

Sipce

9/2

,;.

f(x)dx =

61 2

(g( x) -f(x))dx,
2

--',),---

9/2

'`.

we eatabliAt that.

(g(x) - f(x))dx.. determines

he area of the region

2
between the graphs of
9/2

g and- f.

(dal - f(x)) dx

A simplae calculati n now gives

9/2

(-2x2 ,4-

= nl

13

- 18) dx

= -

13
2

2
-

1,0

17/ 2

12

?",

512

11.4

F,

'+

rf

'Exercises 7-5
1., (a)

Sketch the graph of the function

2

x -ax

x.< 2.

- lc,
.

(b)

Evaluate

- 1)dx.

4(

(0

Find the'area of the region'bounded by the x-axis and the graphH

0 the function,

- 1,

and

2.

(a)

x'= 0

between the vertical liries at

iw

x = 2.

Sketch the graph of the function

x3,

i x

lxl < 1.

(b). Evaluate

(c)

x3 dx.

Find the area,of the region between the graph of the function,
x -4 x

and the x-axis,, where

Ix; < 1. '

b
(d)

Find

b (b > 0) if

x- tix =
c-

.c- 2
-

dx.

Sketch.

1
3.

(a)

Evaluate

(b)

Eva uate
xisi-

(c)

-1

S etch and then find the area bounded by the x-axis,

d

fid

Ca

ixl

y = Ixl.

and the curve'

046

1.

you identify the curve?

= y

-tch and then find the area of the region bounded by

5.

6.

= 1,

Sket h andthen find the area df the region.bounded by the coordinate

-eie

'xi

y = x.

ketc4 and then find the area bounded by the x-axis,

(d)

4.

x dx.

and

- 4

2
.

ketch and then find the area of the region bounded by

y = x

3
,

y = -2x

1,

between the vertical lines

2
t

x = 0

and

x = 1.

513

115
a
1

-C

7-5
)

er

and

indicated

x + y = 2,,

in the figure above.

ts:

Find the -,.rea of the re,--ion. b.ounded by

the required region intq smaller regions

For'the. 'fir6t method

which can be evalda'ted as follows :

4-

dx

Is

'-(-47:)dX

013-c

QI

-TX-) 1dX

II;

+ 2) 1C1X

AII

IV

mailer region with their respective integrals.

'hi

!..

(t)

(-x +2)dx

regions $,h eh

idi"ng the required region into different smaller

re evluated as follows:
.11

-y 4.

) dX

1: - ( -ITO

-:

...

4-

)dX,, -

I-

4Z

F4[ ' ( -X +

2)

la]

J2
A

h their respective integrals.

4
('.c)

Show that th,- expre tons: of area

in

part (a) apd part (b) may be

to the, follow ing stPtement.

4
f;c. dx

( -x + 2) + &idx

,,ro$A point oat the relatiomihip of this expression for the area
and

the

figarF repre,enti% the area?

Could you have arrived at this

exp. e.,s Lon without going through the smaller sub-regions .of parts
(a) and..('.)?
(d)'

From tne expression for the area

in

region indicated in the figure.

5l4

16

part (,P) find the area of -the

,af

7-5

13;

(a)

Express an .integral represent-

ing the area of each of the

'

(DO MT

following regions:
EVALUATE.).

Region I:

(1)

bounded by the
X-Fixis and

)-

.y = 2x - x

lounddd by

Region II:

(ii)

y = 0, x = -1,
and

y = 2x - x

(b)

(iii)

Region III% bOundeld

(iv)

'Region IV:

bounded by

y = 0, x =.3,

y = 2x - x

and

-3, x = -1,

y = 0, y

and

x = 3.

Combine the. integrals of pa'i't (a) and show that the area of the

y.= -3 care expressed by'

regioh bounded by y = 2x -.x2 .ancl

the integral,

(2x - x

A =

'

+ 3)dx.

(c)

Find the area- of the region described in part (.b).

1.1k

y the graphs -Lethe functions

00a bounded only .by

..

x ,-) cos x

Ai''

',,

:' x -) -sin x

_Az_

if

is restricted to'the closed intei'7NT

-n < x < no

Sketch the

curves in this interval.

fl.
',

r 37.c/gr

;-'
(b)

(i)

cos x dx.

Evaluate

....

..__.:7_0

,---

----0-."'

3n/4
(ii)

',/,

(-sin x)dx1

Evaluate
1 -n/4

;
\

IL

ts

(iii)

Evaluate

r 3/(cos
i

\IN

x -.sin x>dx

-,.''

-n/4

e-

,..

'(iv)

Interpat parts (i), (it),

and

(iii) geometrically.

.%515

ti

C hi

7-5
11.

,..,.

10.

(';)

Use a .geometric argument to find

.',

fa,
- (b)

if

as

an odd Function

a's

f
f(x))a

Show that

.'

.0

(1'.e:, f(-x) = -tf(x)).

. I,

4,

0.

a'

if- f

is ,an even function_ (1..e.,

f(-x) =

(c)

11.

Show that i,f

3x)sin

'(x3

Evaluate

F' = f, G! = g,

dx.-

h ,

f(x)'<'g(x)

and

for

- -

a < x < b

then

F(b) - F(a) <G(b) - G(a).

Verify co5).

12.

f = F(b)

(Hint:

F(a).)

.._

il*
Suppose

13.

'

where

F(x), =

x
4 e,.

f :-x

x
(a)

What is

(b)

.
Find an expression for

(c)

ijse part (b) to find

(d)

In general, suppose

(a)

Find the area bounded by the x -axis and the curve

F(1)?

.
...

F(x).

...

,..
F' (x) .

x. .
b

114.

..!

(b)

7. .'

g.

'

Sketch.

Can you find

,..
(Hint:

-...x2

x3.
-, "7

Note analogy to pprt (a).)

N-.

,Find the area bounded by the y-axis ,and the furve

Sketch.

Gt(x)?

_Z

G(x)

= y2 - y 3 . ;
o ..

Y
,p

7-6,

776,, Integration Formulas

We -have seen.that the integral

f(x)dx

can be evaluated, if we can

1-a

find a fActit144 F

such-that

for then we have

Ft

c-A4`

f(x)dx = F(b)

F(a)f.

In general we find antiderivatives ty one or a combination of methods.

A method may consist of recalling a differentiat9on formula, judicious guessIn this section we review some of

ing, or using tables of antiderivatives.

the basic formulas used previously, give some additional formulas and discuss
the use of tables.

TecOr-ques for 2xtendi8g tie scope of our formulas will

be discussed in Charter 9, where we also discuss methods for obtaining approximate values for integrals.

Other integration methods are discussed in the

appendices.
'The common notation for an Etntiderivatfve of

If(i)dx,

whfCh is also called the indefinite integral of

f.

This symbol is cl4te

'similar to
4.

b-

f()dx
a

the integral of

1'.

a 'to

from

b. 'The symbol

-N

ff(X)dx

4defines a function, namely, a functiorahose derivative is

f.

The second

.symbot

f(x)dx

4
\N....
.

represents a number, which can be interpreted as the signed area determined

by

between

and

b.

517

.1 1 9

7.6

Integration fOrmulas are obtained by reversing the differentiation proces

4

for

DF(x) = f(x) .

means that

f(x)dx = F(x)

For example,

D 3 = x

2c2pci3Visince

is-any_constant, we have

Of course,.if- C

v3
s,"

+ C) = x

2
;

*4

more precisely we have

;,./

2
X

x3
UX =
3

,,

...

Diff rence Theorem (Theorem 7-3b) that all

'In' fact, we linow,from the Constani
.
4

antiderivatives of .-3,x
g

have -t]

'

.,
,

x3

;_-----

.-:

'

x -3. + C

w ere

is a constant.

'

In some books this facti's stressed by writing

f(x)dx =F(x) + C,
44,

where

'is a constant and

For"conveniei.ceiwe fblloW the .

DF(x)% =:f(x).

simple practice of ignoring this constant

in our formulas, each integre

tion formula giving only one function wh&se derivative is

f.

Other's are

obtained by -adding constants to' our antiderivatives.

The Power Formula

4
d

Recall that if

is any reel number the

Dxa = axa-l.

If

a / 0,

we can write.
=

D(-}1".

xa-4,

/1

120

518

'

7-6
1

so that ; x

.fs a function whose derivative is

arl

This tells us

that .

sr
a,
x

a / 0.

if

;0
For convenience we replace.

by

where

p + 1

p # -1,' to obtain the fb,rmula

'

is any real number except

+1)/1

xP dx = 17q. ,

p / -1.

In other words, an, antiderivative of a power function

is obtained by raiins th
Suppose
,

the formula

= -1,

exp(5Feivt "ty,

-1,

and divid;e1g by the..rfew exponent.

then our function

In

Section 6-6 we obtained

x"IN

loge x =

1
,

x > 0.

f(

This gives thesintegration

dx = loge x,

;4(

'x > O.

.cl

Circular and Exponential, 'Functions

From the 'formulas

ein x = cos x;

b cos- x = -sin x,

wa obtain the integration formulas

S

Cos x dx = sin x;
cos
x

Since

Isin x dx = -cos x.

we have the formula

ilex dx -

ex.

di

51121

653

It is a simple matter to extend these formulas to the case

For' example? we know

ex + d.

replaqed- by the lirtea,r- expression

is

at

a)

D sin (cx + d')" -='c,cos (cx

that

tr

(cx + d).

c.cos (cx + d)dx

we can write

rf

Pos (cx

sin (cx + d).

d5:dx..

Analogousdifferentiation formalas_were discussed in Volume One for poly.

)

ngmial,

In Chapte,; 9 we shall discuss

exponential and ?.ogarit1amiC. functions.

..

Here, we state the general

the formulas resulting from nquIinear substitutions.

result .foT linear replaeementA

410

irf(x)dx :F(x)

If

and

c / 0,

'4

then

F(4x + d).

d)dx =

j'f(cx

For easy refernce.w e summarize current resu lts' in Table 7-6.

Table 7-6
/

Some ,Ihtegration Formulas

a+1

(1)1,

xa dx =
.

x+

a / -1

i, s

1
-; dx = loge x,

(2)

(3)

f cos x

x > 0

dx - sin x,,

(4)

i(

;in x dx . -cos

'ex dx = ex

Vc

(6)

fox + d)dx =

C,

Fkcx
F( cx + d)

*Pdg.

441S

520

e,

.122

.448.

Example

7-6a.

1.

T3/

Find

j1

x2

The power formula (1), with

1

Z
4.

444

dx.

a = -2,

dx =

-2

giyes

dx =

;-

3/2

3/2.
1

dx =

-1

-11)

=3.

4
Example

7-6b.

Find

1)c dx.

The power formula (1.), with

I lc dx

a _.

f x1/2

gives.

dx =

23/2
x

so that
.

dx

2x

3/2

4
=

3(4 V2 2

3/2
).

o.
2

16 -

4,12-

'Example

7-6c.

Find

(sin x - 3 cos 2x)dx.

We have, from

and. (3),

.s'sin'
x

Relilacirig

''

x dx = -cos.'x

2x

in

and

I cos

the latter and using

3c,dx, = sin x.

(6),

we-have
J.

w;

cos 2x dx =

22: sin

'tx.t.

Therefore,-.we .conclude-

41,
4

1
1

7-6

(sin x

3 cos 2x)d).c

-3

sin x dx
0

= -cos x In
0

,__J-

sin 2x

.--2-'2

, -(cos n - cos 0] -

= -1-1

cos 2x dx

r;

1] -

sin n 2n -

0,- i

sin 0)

0] = 2.
`.

),

-1

2 e% dx.

Find

Example 7-6d.

J -10

We use

(5)

to obtain
-1

/-1

ex dx 3 2 ex

=2

2 ,ex dx

-10

:10

= alp

-1

2e

-10
.

Example 7-6e.

2x dx.

Find

We first convert to base

Now we use

(5)

c = loge 2.

where

2x = e x,

o obtain
i

f ex dx =
We replace

by

cx,

(6)

so that

cx
e

ex '.

gives'

ilx =

cx

where

c = loge 2.

we have

Converting to base

.:*

dx

It

0X7

log

#,4
1:424,t 22

2'

ire

7-6
so that
1

21

2 'Ax

(lo'be

.2

Example 7-6f.

(x + 1) 3ax.

Find
-1

We card evaluate this integral in !Lwo ways,

we expand to obtain

e
(x + 1)3

x3 + 3X: + 3x

so' that

o
(x

3X"

1) -dx

4-

'X

111d>

-1-

We apply the power formula (1) to each term to obtin''

(x + 1)3dx

x3 +

3x
+-

x)

_1

1r0
1

F1
'4

- 1 +

11

Alternatively we qan recognize tht the powel' formula (1) gives

1

x- dx = 7 x

f
and the linean substitution formula (6} gives

r(x + 1)3 dx =

+-'1)4:

(x

Therefore, we conclude that

,

,....

.4

(x + 1)3 dx,,=(x + 1)
-1

-I.

..:1,-,

The second method ts c

_

thinly quicker.

1
..523

12J

Example 7 -6g." Find

2
sin vx dx.

We have not yet obtained a differentiation formula which xesults in the

square of the sine function.

We use the fact that

1 - cos 2vx

sin2 vx =

thus, we have
1

22

sin2 vx dx =
sin

cos 2vx.

)dx-

0'

cos

1 dx - g

24tx. dx.

To evaluate this second integral, we combirie the cosine formula

(3) with

the linear substitution result (6) to obtain

cos(2vx)dx = -- sin(2vx).
210.
)
1.

We can write

or .

..c

cos 2vx dx ='

2v

sin 2vx
1

-0

= JL(sin 2i -sin 0)

= 0.

2v

Since the second integral is

0,

we conclude that

sin

2.

vx dx = g

1,dx - 0= 7 x

I.

=,7 .

Ss

Op

EXample 7-6h.

Show that the area of the shaded region of Figure 7 -6a

twice that of the shaded region of Figure 7-6b.

a'

-r

12.6524

1s

7 -6

g:

Figure 7L.6a

Figure 7-61)

et a

dx

1
dx.
x

0 =

and

We wish to show that

= 2.

'''''

Formula (2) giv s

dx = log. x,

4s'

a =

dx = loge x

= log

s.

4 - loge .1 = loge 4;

1 x

0 =

1x
Thus, we'conclude that

dx.=' loge x 11 = loge 2 - loge-i =.1.og. 2:

.,.

f
.log
e

log

4
2

log
log

2. loge 2

log
.

2.

2.

The Use of Tables

A longer 40D.Ip of integrals

'More differentiation methods

'tables.

s given in a separate booklet (Table 7).. As

e developed, we shall see how to construct these

The following examples make use of these tables.

N.

-525
I

127

Example

Find

7 -6i.

xe

dx.

Formula 16 of the tables gives

xe dx =

- ex

so that
1

xex dx = (xex

Alm

ex)
10

(lel :el..) ..(6e0

e0)

Example 7 -6j.

xe3x dx.

Find
0

Forniula 16 a the tattles gives

3x

xex dx = xx - ex. We replace x

by

anti use (6) to obtain

dx =

s3xe

3xe

3x - 3x
e );

so that
1

x e3x

1
. .ax

(3xe33c - e3x)

23-(

e3) - 2:(0e - e) .=1-e3

3
3

+ 1

log (1 + x)dx.

Example 7-6k. Find

0

We use Formula 7 o'f the booklet tables;

loge x dfc = x loge x

x.

Replace x by 1 y x and use (6) frbm this chapter to obtain

1

[(x 1)1ogx
+.:( + 1)

(1 + x)dx

- (x + 1)]

(2 loge 2 - 2) - (1 loge 1 - 1)
(.7

loge 2 - 1.

52'6'

128
4

0.

7-6'.
O

Example 7762._

sink x\c.

in

Formula 28 at the booklet tables gives' that

..

n- 1

sin nn-1
-1 x cos x

sin n x dx =

With

sin

x dx.

we have

n = 4,

-sin 3 x cos x

sin4 x dx =

sin

x dx.

To find this second integral we can use a trigonometi'fc identity (as in

Exampg 7-6g) or we can use Formula 28 againewith

sin

x dx =

sin x cos x

-sin x cos x
2

n = 2 to obtain.

1 dx

1
x.
2

Therefore, we have

I 71'

sin

N,.

,-sin
x dx - k

-g

3xr
711

3'sin x cos-x

x cos x

'

Since

sin it

= sin (-y) = 0,
3x

this becomes

7-.. -g

(1g)) =

Example 7-6m.

A-1418.,

Find

-x

dx.

.0

,...

r
i
.

The tables give no formula forte , e-x

dx.

There is a good reason for

4
.

it is known that there is no elementary, function whose derivative is

this:

x --/e-x

r
.

Our integral, therefore, canIibe.found by using the Fundamental

Theorem of Calculus and we must resort to some approximation method'in order

to estimate this integral.

We shall have more to -say about this in Section

,-

9-4.

'01.

527

1.2

II

--s

Exercises 7-6
.

For problems 1 -15 find the Following indefinite integrals.

(x2 + 1)dx

1.

f
+ x + x4)dx

2.

(1 ; x)dx,

5.

goo

6.

[Hints

(x > 0)

Write as

fractiSns_.11

I tin 3x dx

7.

cos(2x - 5)dx

8:

(-sin 2x )dx

CS

9.

I [-cos(3x-J)]dx
e

J,

10.

11.

cos 3x dx

C
2 in x cos x dx

[Hint:

Use trigonometric identity.]

.

12.

(3 ;in 2x - 6 cos 3x)dx

13.

f e2x dx

141 1 ex/3 dx

'15.

(ex + e-x)2d*

[Hint:

Remove parenthesis.]

528

131)

,11.,

For problems 6-25 find the following indefinite integrals, (iis.ng tables
...

Iffien neceqsary).

16.

ir x

ex dx

x4 loge x dx

21.

Jr_.,

x3 ex dx

-17.

18

.1 4

,x ...e dx

-I
2

O.

I.

24.:I

loge x dx

x,

x3 sin x dx

's 23.

,'

x2 sin x dx

i* . 22.-

e x sin iix dx

y.

x3 log

20.

x dx

3x
cos -- dx"

25.

For problems 26:31, sketch a graph Of the relevant regiori and find the value
of the indicated integral.

26.

+1

(x

e-x

dx

\-

-1

sin x)dx

28.

-x

dx

log

31.

x
e

lie

dx

.yr)-c

For problems 32-33, the following.instructions are to be followed (linear

substitution: translatlon)..

In this section we were given an area represented

by:

A.=

(x +

l)
)3

dx.

-1
P

By replacement of
,..1(

by x - 1),

+ 1

and

by

x,

appropriate .chadge

of limits, we Aincl, an equivalent. exkpression

After the linear substitution,

for the area.

we have

LA

x3 dx

and

529

13,u

"I

ta

evaluating the two equiyalwit forms of

the area:
0

10

A =

_1(x +.1.)3dx 4(x +1)4.

= r
-1

1
x3

ds
ALTS.

d _x =

4
X

we see that they tre, indeed;

. the same.

In the following two problems, follow the format ab ve:

Sketch the area

defined by the integral, make at appropriate linear substitution, sketch the

equivtlent area, and evaluate each.
4

32.

A =

3 (x - 2)

dx

33( A =

x(x - 1.0

cbc

Fot problem 34:35, follow the instructions of problems 32 and 33, except in

this easethe linear substitution is a scale change instead of a translation.

Drag two graphs asbefore.

nj2
34.

A =

35.

A =

sin 2x dx

15 dx
..f.

36.

(a)

Show that if

x < 0, then
D log

(-x) =

e
f

(Hint:

Sketch

x 7) loge (x),

x > 0

and

x < O.)
(b)

Use part (a) to find

-1

1 dx
rx

-3

and sketCh the area.

4

1530

132

loge (-x),

7 -6

e4n you apply the Fundamental Theorem to find

(a)

do so.

If so,

1 1 ".dx?

'

0 x

If not, state reasons.

-70:0

3------

tfi

1
1

lim
n

dx = 00
*".

Use part (b) to discuss what area, if any, you think sholkd be

(c)

assigned to the region bounded by

axes and the yertical line

y =

x / 0,

the

x = f.'

What answer seemcs reasonable to you for

(d)

and

.4.

With what

dx?
/

properties of areais your answer consistent? inconsistent?

ti

531

133

,..

Chapter 8

DIFFERENTIATION THEORY AND TECHNIQUE

o

In Chapters 2, 4, and 6 we shorted that the derivative of a polynomial

,

function was also a polynomial function (of one lover degree) and established
it
I

for certain transcendental 'functions the formulas:

D(cos x) = -sin x

D(sin x) = cos x
D(ex) = ex

.-

D(log

These a,re the.baslc.differentiation formulas.

x) =

1
.

Our primary purpose in this

.

chapter is to obtain formulas for differentiating variouealgebraic.combina.

tions of these functions and to use these aerivatives to discuss graphs and
motion.

The first section of this chapter includes a review of the, terminology of

derivatives, as well as an introduction to the relationship between continuity

and differentiability.

Various geometric properties of graphs of continuous

'

functions are illustrated in Sectiqn 8-2, where the Intermediate Value Theorem
and related theorems on maximum and minimum values of functions over intervals
are introduced to establish the connection between derivatives and the shape
of the graph of a function.

The Mean Value Theorem and applications are dis-

cussed in Sections 8-3 and 8-4.

Rolle's Theorem is left

to Exercis
7*.

a. special case of the Mean Value Theorem,

8-4, Numbera.

Derivatives of sums,

multiples and products' are discussed in Sections 8-5 and-8-6.

Functions which

are composites of simpler functions are discussed in Section 8,-7 and the

important "chain rule" for differentiating such functions is given in Section

8-8.

Special cases of the chain rule, which enable is to differentiate powers,

reciprocals and quotients are described in Sections 8-9 and 8-TO.

A general. ,*

discussion of the,"folding" process used in Chapters 5 and 6 .Eo define and

differentiate ,root and logarithmic functions is contained in Section 8-11.

These results are applied, in particulaF, to the inverse trigonometric func,

tions.

The final section of this chapter gives a special techniqUe for differ-

entiating functions Teich are defined implicitly by relations.

533'

134

Differeritiabil

Let as recall the

We have often found the derivative of a function.

definition
f(x + h) --f(x)
h

lim

f'(x)

h -) 0

at the point

represents the slope of the tangent to the graph of-

f'(x)

As you know,

(x,f(x)).

If

Example 8 -la.

aK

A few examples will freshen our memories.

we have

lim

f'(x) =
e

lim

h -'0

tie0

lim

x2

h)2

(x

2xh + h
h

2,

2x + h

h 40
41,

= 2x.
4

In different notation; we write

MO.

2 = 2x,

which we can read "the derivative of

2
x

is

2x."

Example 8 -lb.
f

f '(x) =

x + h

x-40
1

Since

X +'h

the numerator can be 14ritten as

(h /-0)

becomes

1
x(x. + h)

- (x + h) -

1
x

x(x +
h
x(x +

and the difference quotient

Taking the limit as

approaches zero, we

obtain
(x /

f'(x) = -

o)
.

and conclude that

.1

(0)
x

_Lx2.

(x /

o).

8Example

lim

f' (x) =

We transform the difference quotient by multiplying by,

o7-471-7 +

IT4-7 +
.40
`which is

in disguise......Tkren

x + h)

h( AT4. h

+ ITO
1

11(47717 + ,r)0

, ..

If we let

)./77+

approach zero we obtsiri

'x

v(x) =

(h

h +

,..

1
1

+fix

(x # 0);

2,6"c

otherwise state

Di- 121;
Let

Example 8-1d.

sin x cos h + cos x sin h,- sin x

lim

0
,cos

lim

-=

- sin x

sin (x

limy
h -4 0

f'(x)

Thep

sin x.

f : x

10;

- 1%

+ cos

/sin h%
k

------

\h -) 0

= sin x

.6

+ cos x

= cos x,

where we'use the fact that

;os h - 1

lim
h

- 0

lim

and

We' can write our result as

# o),

-6-c

sin h
h

1.

D sidx = cos x(

535,

t36
ti

,Does a function f
is defined?

Since

have a derivative for all values'of

for which

represents the slope-of the tangent at

f'(x)

f..

(x,f(x))

Can there be points dri the graph of

the question is.this:

at which either

(We remember that the slope

there is no tangent at all or a vertical tangent?

of a vertical line is undefined.)

It is not hard to see that the answer is

that there can be such points,

or example, the graph of

f

x -4'167

has a vertical tangent at the origin

and therefore

has no derivat ive

when

x = 0.

This appears from the

expreetion for

D= 1

x / p.

2,67

--- fails to exist when

Since

ue say that

x = 0,
0

is differentiable if

2i& C.

but not differentiable .at

..x > 0

x = 0.

A more interesting example'is furnished by the absolute value function

0

a.

-4x. 1x1.

Recall that
x

if

x > 0

tx

if

x < O.'

.,

.1xi. =

foot.

Its graph consiste*of two,half.

lines, one of which bisects the arst

quadrant and the other the second Oa'

(Flgure 8-1a).

Hence, there is

a corner at the origin.

Is there a tangent to .the graph

at the origin?
exist?
,

For

That is, does

x = 0,

fif0)

the difference

quotient is
to +

101

Figure'8 -la

.h

h > 0;

Ihi = h

and

X -4
= 1.

536,

137

8-1
= -h = -1. The slope
h
are
The situation is exactly the same whether Q ,and 0'

41-0

of

is

PQ'

close to

p:

If h < 0, 1111 = -h and

R is .1.
14

The slope o

Ll.

or not.

he difference

If there is to be a tangent it the origin

quotient must epproach a single limit whether

approaches zero from the

In this case, therefore, there can be no tangent.

right or the left.

Inspec-

tion of the graph makes this result reasonablehere isno single line
In general, a func-

P which filithe graph closely on both side

.through

tion'will faq.'to be differentiable at,any point where its graph has a "corner."
Consider the function

whose

AJvii

y1

values are given by

1

if x> 0,

f(x) =

x <0.

if

The graph of

is.sketChed in Figure

Note the "jump" at

8 -lb.

x = 0.

41e

-1

say that Tis discontinuous (that is, '"

not continuous) at

0,

has a discontinuity at

off' that

At such

x = 0.

Figure 8-lb

a point there cannot be a derivative.

Tp see this, consider what happens if we join
h / 0.

is large or small.

whether 'h
'

f(h) = 1 and

747

therefore, the slope of

If

(h,1).

is

Q.

to

P(0,1)

Q(h,f(h)) ,whire

The slope of
Q

h < 0, f(h), = -1,

PQ

is zero,

(h,-1),

is

and

PQ is
-1 -1

-h

z,

to be successively

If we take- h

slops 20, 206, k00,

h

...

-0.1, -0.01, -O.

Clearly,

-'fl.

we obtain the

es beyond all. bounds as

JP

Therefore,

0 through negative values.

approaches

incre

1; ...,

fl(0)

does, not-exist.

We generalize this result

In order that

ft():_shIllexlst, it is necessary that

be -84Wntinuous at

How can we show this?

we,say that
,

there i

a.

So far we have not said, exactly what we mean when

is continuous at

x = a.

We have ,been content to say that

no "gap" in the graph at, x = a.

adopt the following definitidh.

537

X38

We can now be more precise and

8-1
A function

said to be continuous

at

if

(k)

is defined

f(a)

f(a + h) = f(a)

lim

(2)

x = a

f .must have a value when

That is,

approaches

f(a), -11.14at be approached as

0,

and moreover this value,

that is, as

xr approaches

a. .tetAts illUstrate.

.,

If

Example 8-1e.

1
f

is not continuous at

x -43-c-,

enough to establish the conclusion.

This is

there is no such number.)

[f(0)4 does not exist:

is not a number.

0 'because

1111

However, for good measure we'see that

f(0 + h) =

0 + h

does,not

approach lay limit.

Example 8-1f.

Let
1, x > Df(x)
e

-1, x <0,

so that the graph is that shown in Figure 8 -lb.

it should tell us'that

defined?

f(0)

(1)

Is

(2)

Does

is not continuous at

Yes,

If our definition is any good,

0.

Let us apply the tests.

f(0) -='1.

ii

if

f(0 + h)

approach

as .h

h < 0 f(0 + h) = f(h) = -1

may, be chosen, f(h) = -1

approaches

0?

No; in fact,

and no matter how close to zero

is no clober to

Now that-we know what it means to say that

than

1 L (-1) = 2.

is continuous at

a,

we

are in a positidn to justify the statement (1), which we 14epeat fOr convenience.
In order that

(1)

ft(a)

befcontinuous at

To say that t
limit 'ft(a)

as

exists means that

approaches zero.

(f(a + h)
h

shall exist, it Is necessary that

a.

a,

f(a))

f(a + h) - f(a)

approaches a

Then
approaches

That is,

083

0 'f'(a) = 0.

8-1

approaches

f(a + h) - f(a}

and hence;

f(a + h)

, ,

-But-thra. means that

. approaches , f(a) .

is continuous at

a.

co

Exercises 8-1
for which the following functions are not differeriti-,

Find any values of

Give reasons and sketch the graphs.

able.
;

: t -4 lx

2. -f

11'

1
x --)

x + 2

3.

f : x --) 'sin xl.

4.

5.

6.

7.

Let

(a)

Find

(b)

Find

x32

X2/3
x > O.

x -a 4.in x- ,

(2-\
\

a positive_ integer,

f (-51;-))

))

/
2

.'

,.

,.

1,

(e)
3..1
Find f(13
)
2*

...

'
...

0.1.

(12111)

(721

.
t

f(ii so that

Is there any way to define'

lim

f(h) = f(0)?

tr tr

.-.....w.-

tz

ti

5140

8.2
,3

8.2 Continuous Functions

In the previous section, we showed that a function

at any a for which

Since a polynomial function has a deriva-

exists.

f'(a)

tive at each value of

must by continuous

polynomial functions are continuous everywhere.

x,

If

ts a rational function
1

where p

and

are polynomial futctions, we know that

for any value of

disContinuous.
f'(x)

tive

q(x)(

for which

q(x) = 0.

,such an

x,

Therefore, we can conclude that

ii not defined

iSIherefore

1f

We shall learn that for all other values of

exists.

the deriva-

x,

is continuous when

0.

For example,
never zero.

However;

is discontinuous.

x + 3
2
x + 2

is continuous everywhere since

x + 3
x
1

is conttkUous.except at

x, + g

x = 1

is

where it

Yi

The function
x

sin x
x
since

f(0)

If we define

f(0)

is discontinuous at

es

does not exist.

tole 0

2rt

(see Figure 8-2a) the func-

tion is still discontinuous since

(from Section 4-2)

lim

sin h

- 1 # f(0).

Figure 8-2a

is discontinuous at

In fact, -f
f

unless we take

is continuous everywhere.

f(0) = 1.

In this case,

There are two important theoreks'about functions that are continuous at

all points on an interval
interval,

(a,b)- which includes both endpoints, that is, an

.

a < x < b.

The Intermediate Value Theorem

THEOREM 8-2a.

f(b) =.B

If

and if

is continuous on

f
C

at least one number

(a,b)

with

is a number between' A--and

c

such that

f(a) = A

and

theb th6re is

f(c) = C.

540

141'

8-2

A
'A special case of this theorem is

the Location Theorem in which

C = 0

opposite ` signs and

ti

and

,polynomial function;

is a

have

(see Chapter

1).

We illustrate the theorem for a

We shall not prove the

few examples.

. .theorem here.

f(x) -

Let

Example 8-2a.

Figure 8-2b

x- + 2
a = 0, b. = 2: and

with

f(0) = i = A

and

should be some

Here

C = 1.

5
f(2) = ;. =,B.

for which

Since

Is between

there

acid

To see what it is, we note that

f(c) = C = 1.
J

c 1-3

*o

2
c

irhen

.+

)1

+ 2

that, is, when

+ 3,

.1

2
c

-c

1 = 0.

The possibilities are

;

c =

1+ 15

and

.1

.; 2

The first lies betweehi a

Example 8-2b. Let

as before.
-3

and

That

5.

hnd, b,

that is, between

f(x) = 2L1-1-3:

1
x
f(a) = f(0) = -3

We have

`2

Doe6 there exist a number

with
and

and '2,

as desired.

C = 1

a = 0
f(2) =

c, (eci< c < b)

C = 1

is between

such that

it possible that

This- equation is equiyalent to

/
which is

131-valent to

Hence, there is no possible solution

which is false.

The7deorem assumes that

surprise us.

[a,b] = [0,2]

is,contintous on

10

there is a discontinuity at

Howeve , for our

This should not

c.

l
' sin x
)

f(c) = -

is between

C
1

Of course,

,-f.

A = -1

Ifwe define

does not apply.

-f

C = -

so that the theorem

a = 0

by

7c

it is impossible to solve

is discontinuous at

a = 0,b =

, x = 0

B = 0

and

with

-1

Alth64h

f(x) =

Let

Example 8-2c:

NO"'

sin x
f t x -4

In fact,

we-have better luck:

C =

and choose

ft

)c.,1
0

r 1

is between A = 1-sand

and

B =.f(2) = 0;

sin c

does have a solution since

1
2

is continuous

(0, 7,4].

A second theorem about functions continuous on an interval

Ta,b] 1.

guarantees the..existence of maximum and minimum values.

THEOREM 8-2b.
number

on

If f is continuous on

and at least -one number

d, (a < d < b)

Here (1) means that for all

that far all

on

(a,b],

where

is a minimum,

f(x)

(2)

is`a maximum,

f(x)

(1)

where

(a < c <-b)

[a,'b],,

m.

[a,b],

x ion

figure

f(x) < f(c)

and (t) means

f(d) < f(x).

A maximum or minimum value ttly.ocurSaetl.i-een

point.

there is at least one

[a,b]

and

or at an end--

illustrate some of the pOssibilitigsl

.41542

43
1'0' 4,

8-2

4
yl

r
Figure 8-2d

Figure 8-2c

We have the following theorem.

If a maximum or minimum occurs between

THEOREM 8-2c.

(that isif it ii
then

f'(x).= O.

not at an endpoint) an4 if

f'(x)

.a.

ir

For definiteness let us consider the maximum

f(c)

where

f,(e) -06,

and

exists there,

there woad be higher points nearby on the right,

there would be higher points nearby on the left.

,e < c <1).

144
AI*

If` fs(c) < 6,

Since both of tbe7e'posgibi-

lities must be excluded, the &Ily remaintng possibility 1.S7f1(C) = O.

543

If

........._

81-2

The argument for the minimum value

is similar.

f(d)

:Re givethree examples ofthe use of this theorem.

Example 8-2d.
[-2,2].

f(2) =

f(-1) = 1

and

and

x ='1

at

x3 - x + 2

x = -1,

and

At the endpoints, we have

4

The minimum-value

f(x) =

for

f'(x) = x2 - 1 = 0

Since

Find

we should find

f(-2) =

and

131

and the maximum value

-s-

on the interval.

M = -3-

Each

occurs at an interior point and also at an endpoint.

If the inter/al were
In this case,

m = -4

we would have

[ -3,3],

and

M = 8

f(-3) =

and

f(3) = 8.

and both the maximum value and the minimum

value occur at the endpoints.

sin x

x #0,
'Example 8-2e.

As we know,

Let

is a minimum value
If we change
is now

Example 8-2f.

point where

)c

and let,

is nou continuous-on the whole interval

theorem does not apply.

f(0) = 1

f(x) =

In fa-et,

m = 0
f

at

v.

so that

f'(x) / 0

[a,b] = [0,v].

[a,b].

Hence, the

at all interior points

There

There is no maximum value.

f(0) = 1,

becomes continuous on

[0,v].

M.

Let

f'(x) =.0.

and let

Ix'

[a,b] = [ -1,2].

Turning to the endpoints

we might be tempted to say that

m = 1

and M = 2.

the minimum value.'. It occurs at a point where

544

143

f(-1) = 1

Actually

There is no
and

f(2) = 2,

110) = 0

f'(x)' -does not exist.

is

Exercises 8-2
,

Apply the Intermediate Value Theoll68-2a yhere possible.

If the theorem does

.; not apply, 'explain why not.

-1, b = 1, C = 0

-a

x3 - 3x

2.

Ix'

a = -1, b = 2, C.=

3.

x3 - 3x

a =*-1,%b = 1, C =

4.

5.

6.

f ; x

7.

: X -4

sin x

a= 0, b =

C = 2

- 3x

[-1,1]'

- 11

[ 0,2]'

10.

f : x

x3. - 3x

11.

''

12.

13.

i .:0x -,

C22,2i

for each of the following funqtions on the interval indicated.

-4x3

x < 0

9.

a = -1, b = 1, C =

and
f

0, b =

x > 0

0;

8.

a=

x
1,

Find

a = -1, b =

x - 4 sin x

'

1]1i4

[0 ,

2].

.x >0
,4

,
0,

1.41. ]

x <0

ow\
e

13:

8-3

The Moan Value Theorem

8 -3.

-The slope of the chord

.

with the points P(11)

and

Q(24).

PQ .is F7 57 - 3.
2.

(x,x ). has

The tangent to the graph at

As we follow the arc from

2x.

the slope

x
4 - 1

Consider the graph of

to

this slope changes from

Q,

passing through the value

4,

x =

At

when

,)therefore,

R(

to

the

slope of the tangent is exactly equal to

the slope of the chord and the tangent
is parallel-to the chord.
Consider

We",can generaliz-e-this idea.

f,

the graph of any diffevntiable function

P(a,f(a))

and let

As we go from

points on it.

to ,Q

assume that somewhere between

Figure 8-3a

be two

anT Q(b,f(b))

along the arc it seems reasonable to

Q

and

the tangent is parallel to the chord.

Let us consider other examples.'

:Example 8-3a.
= 4.

At

(x,x3),

-x3 with P(0,0)

Q(2,8),

and

which equals

f'(x) = 3x2,,

The tangent at

1/7 z 2-(1.73) z 1.15.

to ,PQ.

If

(Of course,is = - -?-

is outside the interval

when
8
31q

[0,2]

the slope,of
x2 = 11' and

is parallel
from

tq

.)

Example 8-3b.- With-the same function

P(-1,-1) _and

Q(1,1),

(PQ) = 1.

the slope

3x

pnd

x = - -1

Thus we find two efferent points

x -, x3

and the points

Now

= 1

or x=

R, R'

at which the tangent is parallel

to the chord. ,(See Figure 8-3b.)

r647

8-3

dt.

Example a(PQ) =

Slope

f'(x) = --

and

on the gr0h between

which

Ix!

Is there any place

3'

j.,6t

P = (-1,1)

and let

Q = (2,2).

and

for

The answer is *No."

ft(x) =\1

If x > o,
f'(x)"= -1.

At

and if

x < 0,

x = 0,, there is no

tangent.

This example shows that the principle,.thatwe are investigating. may not

)11,

fails to be

) hold if the function

differentia6 at some paint between

and

Figure 8-3c

Q.

.,*

ExaMple 8-3d.

R = !".

Since

Let

f' (x) =

x --) lq. with

, x / 0,

P(0,0), Q(4,2)

we 'have

x = 1.

We note that

1
___
= 1_ when
24-c

21/x

that is, when

Theslope of
.

is not differentiable at

which is atone end of the chord ....However,

.f

is continuous at

bse

NI/

547:148_

A- = 1,
P(10,0)

P.

at

The graph of *f

le 8-3e.

is a semi-circle with center

-.x2

Any' eh-grit is perpendicular to the radius.

(0,0)\and radius ,1.

Hence,

the slope at ankpoint.is they negative

reciprocal of

The deriVative is therefore

f': x

Notice that

and ft(Z) ,fail

'(-1)

fow

The tangent is,vertical in

to exist.
each case.

The function

.inuous at

P(-1,0)
.(1

and

Q.

and 4Q(1,0),

at which

is con-

',

Figure

If we chocise

slope

(RD = 0.

Is there a point

f'(x) = 0? .0f course, since

8 -sd

between

and

is such a point.

(0,1)

dy to state the theorem suggested by these examples.

We are now
(re

(..-

THEOREM Li.

If

b, (a < x < b)

f isOifferentiable for each

and if

(a < c < b)

x = a

f- is continuous at

then there is at leastRne number

.x

between

be4een

and

x = b,

and
and

-a

b,

such that
ft(c) -

f(b
.

-f(b) - f(a)
b - a
on the interval fi'om a to

This is usually called the Mean Value Theorem because

is the average or mean'alue of

f'(x)

'b.

We shall notprove the Men Value Theorem, but.shall use-it in the next
section to dralA certain important conclusions.

.548

149

8-3

Ti;

Exercises La

1.GiVen
'

2.

Q(1,2). 'Find the

and

P(0,0)

+ x- arid the points

x -4:5

point w.herethe tangent isTrallel to the chord ,PQ.

Where must we choose

the chord PQ

on the graph of

is parallel to the 1: ngent at

P = (0,0)

-so thq4owitn

(2,4)?

Where is the tangent to

parallel to the chord from

(-1,1)

(2;16)?
4.

Suppose ethat you drive fromSacramento (elevation 200

5480

Station Camp Ground (elevation

The map distance between

miles. Was there some time during the

100

'the two piints is exactly

feet).

feet) to Loggers-

trip when you tlere_on a portion of road that had a slope of exactly 1?
..0"4

Give your reason.

5.

Suppose you drive from New York to Chicago, sometimes stopping and oth
times driving as fast as

speed is

the trip when yo

6.

Two cities are

200

to the sther in
(a)

miles per hOur.

70

Is there some time during

Give reasons.

'50 miles per hour?

miles apart.

Starting from one you drive continually

hours, then stop.

'Is there some place on the trip where your speedometer reads

50?

Give reasons:
(b)

Is there some place on the trip where your ac eleration 14.@,s_, 0?

ii

Give reasons.
i
x

8:

is, there a pOint where the tangent is parallel to the

f.: x -4

7. 'Given
chord

PQ

GiVen

1q1,1), Q(2

where
1

PQ where

I'

i)? " If so,, find it

i4 there a point where

x -47c-

1;

f'(x)

is equal to the slope of

4

P( -1, -1) , Q(1,1Tj Explain.

,.

9. Let f': x ) 1x

if

x > 0

ib thert a point between

and let

P :and 'Q

f(0) = 0.

at which

With

P(0,0) 'and, Q(1,1)

f'(x) = slopeil(RD?

Explain.'

a.

54, 9

1 50

8 -4
8-4.

Applicatioms of the Mean Value' Theorem

If we up,e the Mean Value Theorem (813) we can now prove'certhin results
that we hav&;opreviously.taken for, granted.

THEOREM 8=4a.
(a <,x < b)
f(x)

If

f'(x) > 0

and if

f(x)

when

is between

is continuous at

increases uniformly on the interval

(We include the requirement that

and, b

a _and

then

b,

a < x < b.

be continuouat

and b

in

order ttrbe able to apply the Mean Value Theorem.)

By this die mean that for all nrers

xi

and

x2

such that

a < xi < x2 < b,

f(xl) < f(x2).

X'

i.c

(See Figure 8-4a, 14here

'r.

...

xl

thay coincide with

Proof.

or

with

Figure 8-4a

b.)

According to the Mean Value Theorem

f(x

- f(x

f'(c),

- x

(x. < c < x

).

Since this means that

c < b

it follows that

f'(c) > O.

Hence,

f(x2) -f(xi) = fi(c)(x2 - x ) > 0

and

f(x2) >, f(xl);

that

f(xi) <',f(x2).

In the same way we can easily prove the following theorem.

TlitOREM 8-41).

If

contivous at

interval

f'(x) < 0 when \a < x < b

and

then

b,

and if ,f

is

decreases uniformly on the

a < x '< b.

What can 1ft conclude if

,

thiS case, for all 4x1

alld

f'(x) = 0
x2

such

x between

for all

hat\'a < x

< x

apd

b? In

< b,

t5i

/
8-4

f (X1)

X2

(x2)

and

f (C) = 0

Xi.

0,

f(X1)

f(x2)

f(x1).
...

xi

Since this is true for all

and. larger

16 a constant fUnction on this interval.

THEOREM 8-4c.

If

continuous at

f'(x) .0 for
and

b,

(on the interval

[a,b],

a < x < b

then .f

and if

is

is a constant function on

fa,b].

What can we conclude if we know that the derivative

(decreases) uniforMly.on an interval

f'

increases

[alb)?
,

Let

be any-number between

and

a
.

b'

(a < c < b).

f(x) - f(c)
x - c

where
But

c < d < x

(see Figure

111,1(d) > f'(c).

Hence,

x > c

If

--41p).

f(x) - f(c)

x-c
1

Then

f(x) - f(c), > ft(c) (x - c)

Figure

and

8-10

f(x) > f(c) + f'(c) (x - c).

This means that to the right of

the graph of\, f

lies above the tangent

4

at !C. (Figure 8 -4c).'

r
551 ,152

Figure 8-4c

x <c

if

Similarly,

f(c) - f(x) _
c - x
Hence,

< e < c:

`-:

,f(c) - f(x)

<fi(c)

'C - X

f(c) - f(x) < fi(c) (c

f(x)> f(c) 7.1"(c) (c - x)

Again toe grlph of

f1(c) (x - c},
f(x) > f(c)
4-rlies Bove 'the 'tzilgent at c.

If a function.

has the property that its grap

finally

41.1

s above

gent whok point of Contact lies within an inter61

A

point of eontact), then we say that

If we

fa,b)

(except at the

is convex-od. (a,b0,

Apiece "above" by "below" in this stateme, "convex" is replaced*

by "concave."

t
THEORM 8,4M

If

f' iq.ncreapA (ieCreanes) un4forml4O6 an

Anter44 ia,b4--f--is donvex4onctive) on,-l[a,b]i

.

"?

552

1 5 3.
24.

).

8-4

Exercises 8-4
-

What does the Mean Value.Theorem become if

1.
.

equal to zero?

a < x <:b.

f(a)

and

The result.is called Rolle's Theorem.

se that for a function

_2. __Sup

we know that

f" > 0

are.both

f(b)

N.
on an interval

Show from the theorems of this section that

is convex

on the interval.

IPA

f' (x) > 0

fs(b). = 0,

fOr

a < x <b- and p.(x) <-0

for

b < x < c

while

usethe theorems of this section.to draw an appropriate

conclusion.
ti

4.

Suppose that

f'(x) = g'(x)

that 'f(x) - g(x) = C

Hint:

Assume that

for all

on an interval

on this interval where

[f(x) - g(x)]' = fqx.) - g"(x)

[a,b].

and use Theom

This is the important *Constant Difference Theorem of Section 7-3.

.

5.53.

Show

is a constant.

V'

8-5
a

8-5.

Stints and Multiples

,

The remaining sections of-this chapter-discuss methods for differentiatidg

In this section we examine sums and

various combinations of known functions.

-multiples of functions.

:/Functions which are the sum of other functions have been previously en-

For example, the graph of

countered many times.

3 cos gt + 4 sin gt

//,

was obtained in Chapter 3 by adding the corresponding'ordinates (Figure 8-5a)

of the two function's
u

at each value of

t -)3 cos Trt

and

t -44 sin rut

t.

p,u,v

a.

(0.29, 5)

(2,3)
IS

' /

(1.79,0)

(0.79,6
1

)(

'

(1,-3)
,

1.5, -4)

(1.29, -5J

rr.

Figure 8-5a

217

8-5

Here we say that

is the sum of the two functions

and

and write

f = u
This means' that for each

f(t), 140

the values

t,

and

are related by

v(t)

fit) =Il(.t) tev(t).

The differencerof ,two functions is defined analogously; for example,

f = u - v

if, for each

x,

the values

and

f( x) , u( x)

are related by

v(x)

- v(x).

f ( x) = u( x)

To be more concrete, if
f

2 sin 3x

3 cos 3x

,r

I
we can write

where

f = u - v,

The functiat

in the sense that the vlalues

u(x) = 2g(x).

3 cos 3x.

N 2 sin 3x
:

The graph of, u

u(x)

sirr 3x

x -4

is a multiple of the function

sin 3x

and

g(x)

are related by the equation

is obtained froM the graph of

ing the corresponding ordilnate of the graph of

by

2.

by multiply-

(See Figure

8-5b,)

I)

Jor

555-

-$

1. 5

A-

81.5

.U

`LA

-1
2g( t )

u( t

\
-2

Figure 8-5b
where

,.

Id

x -4

ax;

2 sin

x -4 sin

5x.

The basic rules for derivatives of sumsO-id multiples are easily obtained
d simply stated:

f = u + v,

if

(1)

thaX f' =u' + v';

and, for any constant,.....,,a,

if

( 2)

For

where

f = u

iample, if
:

-43

f' = ag'.

f = ag,` -then

cos ht

and

'4 4 ?It vt,

then
.

v ;

556

57

615
t,

that is, for each.

fl(t) =

=.-3v sin vt-1- hv cos nt.'

We also made use of (2).

use-of the fact that

For example, that

u'(t) F. -3v sin vt

makes

D(3 sin vt) = 3D(sin vt).-

We can use the concept of approximation along the tangent line to the

graph of afunction to show that (1) and (2)'hold." For example, suppose
v, .where

f = u

and 'v

are each differentiable at

linear
approximation to the graphs of u Act
.

a.

For the best

respectively we have-

u(x)

u(a) + i(a)(x - a),

v(x)

v(a) + v'(a)(x - a),

(3)
,

if

is close to

a.

Adding, we have

u(x) + v(x) m u(a) + v(a).4- (u'(a) + v'(a))(x - a).

(4)

Now we use the assumption that

to obtain

f ..-tt + 4

.1

...

f(x) z f(a) + (u'(a) + v'(a))(x - a).

.

For

x / a

we subtract

f(a)

from both sides and divide by

T(x) - f(a)
x - a
We take the limit as ':44

to get,

u'(a) + v'(a).

tt

approaches

x - a

to

obtain

lim
x-4 a

f(xx
)

f(a)
- a

(a),1

o-

We conclude that

'fl(a) = u'(a) + v'(a).

Theteasier intuitive argument which establishes that if

fl = ag'

was given in Chapter

f = ag then

6.

Wecan combine results (1) and (2),to tlifferentiate T.= u - v,

can write
where

f = u + w,

w = (-1)v,

sp that
f' = u' + w'

and

w',.= (71)0 =

.Thus, as we should expect:

(5)

fl = ut

v'

if

f = u - v.

5571 5 8

for we

8-5

Find the derivative of'

Example 8-5a:

x -)x - sin x

and diS.-dug

-2g < x < 2r.

its graph in the interval

We can let

and

x -4x

so that

x,

Since,

f = u - v.

ft . 1 - vL and

from (5),

cp

vt

1,

-4 cos x,

we have the result

fqx),_= 1 - cos x.
.

For all

This tells us that

since .dos x < 1.

x, fl(x) > 0,

increasing function for all

Furthermore, the graq of

x.

(-2x,f(-.27)), (0,f(0))

tangent at each of the points

is an

has a horizontal
since'

and '(2r,f(2,))

= f1(0) = V(20 = 0.

f'(-2r)

Since

Let us differentiate again.

f' = 10

we can ripply (5) with

,f, u

and

f', 10

replaced by

and

'NO

to obtain

the result
f"' = u"

Making use of (6), we have

x -40 and

u"

-sin x

v"

11

so that
ti

f"

x -4 sin x.

The function

f"

is nonnegative in -..the intervals

-2g <.x.< 75,

(7),

and

0 5 x <

and nbnpositive in the intervals

< x < 0

(8)

and

< 2g.

Thus, the graph of 'fisis convex in the intervals of (7) and concave in the

intervals of (8).

The graph of

(Figure 8-5c) is obtained by making use of this infor.

`1.

oration and plotting a few points.

44:

WO'

558

159

8-5
I

y:f

re

321-

8-5c

j,, Figure

y = x - sin x

Suppose a part.cle moves along a, horizontal line so that

its distanceqrom tht origin at time

t > 0

is given by

8 = t +

the- motion, %,

,
559

160

8-5

If
-

than

is close to

then

0,

If

which is very large.

.T.

so that

Geometrically

t.

the graph of; s = t + i approaches

t >TO

t,

becomes lar 41/4.1tIn other words, the vertical line given by

as

s = t

and approaches the line given by

approaches

the s-ails as

is very small,.

is very large, then T

is nearly equal to but slightly larger than

these observations mean that for

is nearly equal to and slightly larger

is an

t = 0

s = t + - as

raph of

-asymptote for the

iven,by

approaches

while the*line

grows large without bound,

is an asymptote for the graph as

S .F t

0,

through positive. values,

- can be obtained using the sum formula (1).

The derivative of
We have
'aft

D(t +

and

Dt = 1

SinCe

1
lit'

1.

,1%

- Et + Elk-) = Et + Dt.t

-1.
).

= -lt

-2

= -

we conclude that -s--

=1 -

D(t +

t=

The value of the derivative

< 0

if

t < 1

and

s' > 0

in the inteTial 0 <'t <41

if

1 -

t -)sl

t > 1,

is the velocity at time

the function

and ipereases in the interval

0

the valve of the derivative is

and

+tt > 1.

is the minimum vane of

Since

t.

decreases

When
s.

t = 1,
This

to
means that the particle moves toward the origin as t increases from; 0
is closest to the origin when t = 1 and then moves steadily away Trom
1,

the origin.

(5)
The second derivative is obtained by using the difference formula

and the power formula:,

D(t -2

= t 1

=
t

bound.

tt

is close -to

0,

and approaches

The second -Miyative

2
t -)
3

560

161

is positive), is very

Thus, the.acceleration is always positive (since

large when

as

grows large without

tells us that-the graph of

Figure 8-5dd

8-5
4

Exercises 8-5
"1.

Find the derivatives of each of the following

e2x

ex
(a)

y =-11--/1)- 3x-2/5

(b)

y =x2 + 2 sin xi

(c)

Y'= (3x2'+ l)(x4 + 1)

y =

'

(g)

+ cos x

3e-x

y = x + log

log

x2
.

x
e

xe +oex

(h)

= (1 - 2x)(-7 +

(d)

Tc.)

x
2.

0 < x < 1
,

3.

(a)

1
+ - , u

and

I X

for
v :x -)x

What is the equation of the tangent line to each at the

point whearZ

Sketch graphs of

x =

1
?

How'sare these tangent lines related?

At what points on the graph of

y = sin x -

1/S. cos x

is the tangent line horizontal?

(b)

At what points on'the graph of

= 2x - 2x

is the tangent line perpendicular to the lig whose equation is

y = 3r
(c)

2?

Suppose the tangent lines, to the g.cfs:tphs of

are parallel and nonvertical at the

y = 5f(x)
x = a.

int where

and

y = 7f(x)

ShoW' that

these tangent lines must be horizontal.

(d)--NShowethat-if, u, ,and

graphs 9f

x -)u(x)

are differentiable at
3v(x)

same slope at the point where

and

x = a

then

x = a

x -,u(x) - 11v(X)
v

Show that if

and b are constants then

D(au + bv) = a Du + b

562

163

have the

hEis a horizontal tan-

gent at '(a,y(a)).
4.

and.the

lb

Analyze

5.

'increase-dec.ease,

ir

convexity-concavity, and

(ii)

asymptotes of any)
z.

for each of the following functidns on the interval givey.

(a)

(b)

..(c)

0 < x < 2g

x -) x - cos x,
- 2x,

t -) t2 +

Sketch graphs.

0 < x < 1

0 Qt
0 < x < 2

(d)

(a.)

Show that

(b)

Find

: .x

.r

b
6.

7.

f(r)dx= -f(x).

1.2

dt.

e-'

Show that the acceleration,pf a particle whose equation of motion is

2

s(t) = 2 cos t + t8.

is always nonnegative.

Suppose you know only that the rules of this section hold and that
n
n-1
Doc
= nx
Can you find the derivative of a polynomial?
.

aly

"9.

Consider

$1-

Sketch the graph of

Define

10.

- xl.

21

g(x)

g.

explicitly in terms of linear functions for all real

(c)

For what values of

(a)

1 + x+ x 2< ex < 1 + x +..x

is the derivative not defined?

(Hint:

f(x) =

Put

0 < x< 2

22

ex

-+ --

and find the minimum of

f.

Proceed in e,similar manner for the right -hand side).

6

(b)

Show that if
u(x) < V(x)

(c)

fbr

Show that if
x > a

the

show'that

u(a) < v(a)

x > a.

u(a) <y(p),
u(x) <.v(x)

ut(x) < vt(x)

and irt(x) < vt(x)

(Hint;

for

Consider'

u*(a) < vi(a)

x > a.

for

and

(Hint:

x > a

then 111,

u.)

u"(x)<,v"(x)
Use (b) twice:

f6r

First

when. a < x.)-*

0,

,
0

563

1 (3 4
1

ti

8-5. o';.'
.4.
J

1.1

'

(g)

-..

res

(b)"
a.

1e
---

,,A y". - 3y1 +.6y =

-,,,,1.

then so' is

0,

3u +\ 8v.

y .=

...

e-x

ex

Show' that

y'14..- y.

the equat

also a solution to

r-=

and

If a

y = ct( ex

and

- e

are each solutions to

,

are constants is

13

f.3( ex -

v(x) 4,1ax + b,

u(x)

'___--X

e-x

y"

x
e

OP"

Suppose

are dol,utions to the equation

7 = v

and

y = u

Shot./ that -if

'where

anct

are constantp

1,tb

IAThat'is

(bi)

Show that

(c)

vi(x)

u'(x)

(a)

u"

= fir..

Prove the following conver'se:

Suppose' u

and

are continuous at

a linear

u - v

Use tShe Constant Difference Theorem twice.)

function.' (Hint:

"13.

then

u" =

\x = a:

Is

= 2u

3v

also

continuous at

14.

Suppose
x = a-.

a?

f = u + v
Must

.0

and

and
v

als

is differentiabI
be differentiab

and thus continuous at

nd thus continuous at,

give an example..

564

1.65
4

.8-6.

froduets
Each value of the functiorf

I
is just the product of the corresponding values of the two functions
u : x

that is,tfor each

and

x -4 ex;

x,

f(x) = u(x)v(x).

This relationship can be used to obtain the graph of

and

v,

for the ordinate of a point on the graph or

corresponding ordinates of the graphs of

and

v.

from, thegraphs of .1.1.

is the product of the'

(See Figure 8-6a.)

= u(x)v(x)

y =

v(4
x

ft

y = u(x)

-2\

-2

0.14

-0.28

-1

=.1-jp,.38

-0.38

-0.5 -0.9(0.61

-0.30

0
-.0.5
1

1
/

6/

Figure 8-6a

y = xe

u(x) v(x) f(x) 7 u(x)v(x)

0.5 1.6 p
1

2.7

0.8
2.7

S -6

In general, We sgy that ,the funct4on

is theprbduct of the two funcid

and

tions

an

write

ddit

f = uv
if for each

the values

f(x),

d(x)

are related by

v(x)

'f(x) = u(x)v(x):

(1)

f = uv

A formulaAr the derivative of

u

and

and

and

in-terms of the derivatives of

Suppose

can be obtained by using tangent line approximations.

are each differehtiable at

x = a

so that, if we take

close to

we have the best linear approximations

u(;) '.I1 u(a)

y(x)

+ 1.0(a)(x

a)

v(a) + vt(e)(x', a).

For the product 'we get

u(a)v(a) +[u(a)0(a) +v(a)1.0(a))(x -a) +u'(a)v'(a)(x -a)2.

u(x)v(x)

Since

we can rewrite this as

f.=

f(x)\ a f(a) + [u(a)v'(a) + v(a)u'(a)](x

so that, for

a) + 0(a)0(a)lx - a2

x / a

f(xx

f(a)
-

[u(a)vt(a) + v(a)0(a)] + 0(a)0(a)(x - a).

,

It follows that
lim f(x)
f(a)
x - a
x -)a

u(a)v:(a) ; v(00(a).

Thus,, we obtain the product rule:

zk"
ft(a)

(2)

u(a)vi(a) -+ v(a)ut(a)

This formula is sometimes written in the form

(uV)1 = uv' + vu'
Or

D(uv) = uDv + vru;

(4)

or expressed in words:

The derivative of the produet of two functions is the

first tithes the derivative of the second plus the second

(5),

times the derivative of the first.

io

566

167

a,

4
r .
.

For example,

x -4 x

,e

vi(x) =

and

11!(x) = l'

and

3:

Since.

is the product of

x -4 loge

the product rule gives

X,+
+ (loge x)

f' (x) = x

1 = 1 + loge x.

'

As another example we consider the functiom

i4

3x

sin 2x,

which is the product of

.0 :

x -4 e

3x

and

sin 2x..

The .productlfrule gives

(2 cos 2x) + (sin 2x)(3p5,x).

f'(x) = e3x*

Example .8-6a.

focate the intervals of increase and decrease, convexity

and concavity for the-graph of the function

-;

The function

x -4 xex.

is tha. product 'of

u : x

4 x

and

ex,.

so that

f'(x) = U(x)vi(x) + v(x)11160

= x ex + ex .1
%

(x.+ l)e

This will be positive for

graph of

x
.

x < =1

x > -1 'and negative for

so that the

fall's until it reaches '( -1,- .1e.".), and rides Ater that point;

The function

f'

(x + 1)ex

is the product of.

,e

: -x

x + 1

and

so the product rule gives

,;;.

567
J

168

x
.e

r(x) = u(x)vt(x) + v(x)0(x)

ex

1)ex

= (x + 2)e

.We conclude from this that the graph of

convex for

x > 2.

We should also note that as

these conclusions.
Af(x)'= xe

the graph of,

approaches
f

as

Example 8-6b.

is concave for

x < -2

-and

An extension of our sketch (Figure 8-6a) should reflect

moves far to the left

that is, the negative x-axis is an asymptote for

0;

grows large without bound through negative values.

Show that if

then

eax sin bx,

f"(x) - 2af'(x) + (32 + b2)f(x) = 0.

The product rule gives

eax

Disin bx) + (sin bx) D(eax)

eax(b cos bX) + (sin bx)(e6x)

= eax [b cos bx + a sin bx].

Again we use the product rare (as well as the sum rule) to 'obtain
eax

= e

ax

r
D[b cos bx + a sin bx] + [b cos bx + a sin bx]D(eax) *-

.2

L-b

sin bx + ab cos bx] + [b cos bx + a sin bx]aeax

eax((a2

b2)sin bx 4=2ab cos bx].

Therefore,
f "(x)

- 2af'(x)

b2)f(x).. eax((a2

(a2

b2)sin bx + 2ab cos bx]

2aeax.[b cos bx + a sin bx].

eax[(a2

.2.

.?

eaX (es<0; -2a2

+. e

ax

)Siti bx]
+ a2 +b2)sin bx]

[ (2ab 2 2ab)cos bx]

= 0.

e.

568

169
a

% 8-6
A

$ IV

Example 8-6c.

,,

zero of

f.

Suppose

A a.

is a polynqMial function and that

Show that them4tiplicity of. a -is water than

is a
-. \

if and.only

is a zero of

if

f'.

If the multiplicity of
f x) ;

exceeds

a)2

(x

'then

is a factor of.'

that` is

..f(x) = (x - a)2q(x),

where

q As a polynomial function.

Applying the product tule we have

2(x - a),

f'(x) = (x - a)2q'(x) + q(x)

so that indeed

Ple

f'(a) = 0.

If the multiplicity of

is

then.

1,

where

f(x) = (x - a)g(x),

The product ?hie gives

g(a) / 0.

'(x) + 1

g(x)

so that
A

f'(a) = g(a) / 0.

In other words, if the multiplicity of

zero of

is

fl.

*me

569

170-

then a

cannot also be a

re.
c\s

Exercises 8-6

Let

1.

= a
1,

+ m (x - a)
1

x ->x2 oat

grapfi of

be the equation of the tangent line to the .

and

(a,a2)

=,a

tion of the'tangent line to the graph of

Find

(b)

Form the product of ;he expressions for

x ->x3

(a)

+ m (x - a),

the equa-

(a,a3):

at

al' ml, !12' m2.

term involving
(x - a)

(x -

a
a)

uv = f

and

and omit the

The resulting expression is linear in

and hence defines a line.

tangent line to

x -)x-5-

Show that this line is the

at the point

(a,a5).

,,

Finethe d rivative of

(a)

f,

/Vhere

x(2x 1()

(b)

(4x - 2)(4 - 2x),

(c)

(x2 + x + 1)(X2 - x + 1)

f(x)
(m)

(n)

equals
x

log

1/2

(x -

ft

1.)

x
(o)

ic

e -x.

e-t'dt

Is

(a)

xix sin t"

(ax + b)3

(P),

dt

1
!

(e).

.(f)

(q)

x e

(5x + 2)

(r)

(loge x)(4x2 + 2x) (cos.

x
-.

x.

"3--,`--:)

sin x

2x)-

.
..

(g)

x ex

(s)

x7/2,

IL,
.

4
i,

>O

.(t)

1
- --

2 sin x cos x
x ex fbge(2x + 1)(sin-x)

(u)

x2

(v)

x log

(j), 3x (x

- 5)
-

(k)

i:7-c- cos 2x

(Q)

e3x sin (x + 1)

'..

(w)

x + 1)

e ex

9
or

8-6

L.

1)2)

(4) 'D(X(
r^

(e)

(f).

( g)

012
E'cr
'D

ofc

D(ex sih(1 - 2x)Y

( i)

D(v loge x)

(,3)

15( x

'1(k)
(2)

2x--)2

,D(x

(h)

, 2
Dkx cos X
,:

D(sin x,

log

-1/A

'')(1/2
x

- 2,r +
1-Cc

4.

(a)

f(x).= [u(x)]

Suppose

Show that

f'(x) ='2u(x)ut(x).

(Hint:

Use the Product Rule.)

5.

'Show that

(c)

Show that

(d)

Make. a conjecture about

D[u(x)]4

4(u(x))30(x).
Dfu(x)

Use the results of Number 4 to find, y'

2

D(u(ic)-1341 3(u(x).]2.uqx).

(b.)

(a)

y = sin

(b)

y =

(c)

y = (loge x)2

cos3 (4x)

if

(e)

y = (x2

(f)

y =

q)2

sin3 (-2x -

(f

1)

sin t2

dt)4

(d)

6.

y'=

(ex)4

Combine the 'method of Num1r 4 with the Product Rule to find'

x2(x24. n, 2
+1)3(x2 - x * 1)

(b)

y= x

(c)

y = (ax2 + bx + c)(dx2 +lex + f)

(d)

y = (ccs2

(e)

x)qn_ 2x
et

'ex sin"- (ax + b)

5Th

cdY_Ic

0-6
(

x
(f)

y =

dt)`

et

0
(g)

7.

y = x3(loge (x + 1)13

For each ofohe following functions, find the intervals of increase(or

decrease) and convexity (or concavity).

Sketch graphs over the intervals,

indicated.

(a) ly = x loge x, '0 < X < e

(b)

8.

0 < x < e2

loge x,

y =

(c)

y = sin 3 x,

(d)

y = x2 1p

0 < x < 2o
0 < )C <

Show that each of the following is an increasing function

x > 0

ex,

(a)

x -4

(b)

e
x -4

(c)

x-)

, x >

ex
)

x>a>
0
.

(d)'x-4xsin
9.

Show that if
g(a) # 0,

10.

,greater than _2

is.a zero of11.

(a)

where

f(x) = (X - a)2g(x)

then

Show that if

0 <x<1-'2(-

f/(a) = 0.

is a zero of the polynomial functi n

f"(a) = 0

f"(a) =

then

greater tgan

of

Show that if

is differ- tiable and

y = eax cos bx

then

of multiplicity

m st it be true that

2?

y" - 2ay' +

Y = 0.

4f*

'12.

(b)

Show that if

(a)

Show that

y = x2ex + 2xex *then

- 3y"

3e,

y = 0.

*1

(uv)" = uv"
(b)

II

Use (a) to find the second derivative of

f

( c)

2u'

What

x2 cos x.

(uv) "' ?

(d). Does (c) lead 'you to a conjecture about the nth derivative of

572

173

uv?

Composite Functions

8-7.

The function

is not a polynomial, circular, power,

.477-7.

f :_x

exponential or logarithm function; nor is it a sum or product of such functions'.

can give a blue as to.how to treat such a func-

The verbal description of

Verbally, the rule for

tion.

(1)

is

squared plus one."

"the square root of the quantity

x2 +1.,

In other words, first calculate the quantity

The operation defined by

root Of the result.

operations, finding

+ 1

and then ,ake the square

is.composed of two simpler

and taking square roots.

In this and the next

4

two sections we discuss functions which are compositions of other functions.

,The statement (1) can be trdhslated into a symbol* form which will display the fact that

x -4x

f : x

-)g71.

is composed of the two operations,

g(x) = u.= x

Let

and taking square roots.

+ 1

and

+ 1

h(u) = 47,

so that

f(x) = h(g(x)).
f(x)

To evaluate

we first evaluate

example, if x = 3,

g(x),

then evaluate

h(g(x)).

For

thdA

u = g"(3) = 32 + 1 = 10

and

f(3) = h(g(3)) = ti(10) = 45.

In general, we say that a function
tions

h -and:

g,

is a composition of the two funcg(x)

is defined, so are

ifwhenever f(x)

and

'11(g(x));

and then
f(x) = h(g00)..

The idea of composition has been previously used implicitly.

For example,,,_

the function
f

(2x27N43-r7

is a composition of the functions

h :

-) sin u

and

x -)u = 2x + 3;

that .is',,

f(x) = h(g(x)).
.

;le

Also.,:use has been made of the fJct an' ; the general exponential fu4cti-on

41v

573 1 7 4
;

x
f

a = ea.

is agcomposite'function since we can write

x -4/Et

and

u,

then
= h(g(x)) =

If

eax,.

Facility with composite functions depends upon ability to write compliSome examples and

cated expressionsias,composites of simpler expressions.

practjce exercises are provided to help you develop skill at doing this.

function

If we let

root of

x,"

the

is a composite of the sine and t e squareroot functions.

17

as the composite of simpler functions.

is usually read<the sine of the squ

sin 17

'Since

x -sin 17

Express

Example 8-7a.

u = g(x) = 17 and

h(u) = sin u,

we have

sin 17 = h(g(x)).

fib

-:-.

Express x

Example 8113.

-1

,.

--

as the composite of two simpler func-

tiohs in two ways.

The expression

2/3

can be read as

"the cube root of the square of

(2)

x"

'

or

"the square of the cube rootof

'(3)

Put

g(x) = x

= u

and

yu = v.
h(u)'= 3/-

x."

In symbolic form (2) becomes

= h(u) = h(g(x))',

(4)

while (3) becomes

(5)

)g3 = ev) = g(h(x))

In other words, in this case, it doesn

matter whether we square first

and_ then take the cube root,-or take the cube'groot and then square.
,

It should',

however, be noted that generally the order o

compositiOn is important.

the Example 8-7a we had

) = 17 = u

wher

since = h(g(x)),

.Reversing the order of comp sition? we have

g(h(x)) =

which is 'ce;tainly, not th-

in 17

ame as

4
574

"

175

and

h(u) = sin u.

In

.It should be observed that there are other ways of expressing x -4X2/3
For example',

as a .Composite.

x2/3
f(g(x)),

( 6)

where

g(x),

and

(x - 1)1/3

(x3 + 1)2/3,

f(x)

singe
x2/3..

f(g(x)) = [(x - 1).4. 1]2/3

.',

Exercises 8-7
.

1.

Express each of' the following as a composite of two functions which are
A,

polynomials', exponentials,,logarithms, power, sine or cosine functions.

(a)

A---7x2

(g)

x -)(2x2 - 2x + 1)-1/2

(b)

ex2

(h)

x -) loge, (sin x)2

( c)

x.

cos' (3--

(d)

x.,_;3e2 sin x,

1 + x

2.

_)ecosx
2
3x)

(e)

X.-410g;

)7717

(f)

x -4(21-

),

(k)

x -)2

(x 01)2

Express each of the following as the composition of three or more simpler

functions.
(a)

x -4 loge

-(b)

x -4

18x

+ 5x + 21

1 -+/ 7 cos x

(c)

x -4cos(sin(cos x))

(d)

x -4 (x-+ 1)3/5

(e)

x '4

(f)

x -4

11 - (log; x)
1

e2x

6*.

3.

jbcpress

lx1

as a composite of the function

function.

575,,

/
1 '7

I.

x -4x

and some other

PI

8-7
<t\''.<;

4.

(a)

Show that the coMposite ofitwo linear functions is linear.

>

(b)

Exhibit the comNsie of two quadratic functions.

'

of this composition?
(c)

What-,isthe degree

Is the composite of two polynomial functions a polynomial f.linction?

I

If'dO, what is its degree?

If

Suppose
by

and

X ;-4x

4'

f,

what is

x -) u(u(x))

f(3)?
114.

the function'dtfined

f,

Find an expression for

x -)

f(x) =',u(u(x)).

6.

(a)
-

Show that composition of power functions is a commutative operation,

that is, if

x -)xa

and

then '11(v(x)) = v(u(i4).

x -)xl)

(b)

Is the result of part (a) true for

(c)

Is the result of part (a) true for exponential functions

u

x -) a

and

x -)b

cos x

and. v

x.

(a, b.> 1)

.0.-

(d)

Is the result of pert (a).true for ,u

v

7.

x -) ex

-)1Oge x?

Express the following as a compoVetion of two functions

2
(a)..

t2/?.

dt

-2

Cl
(b)

et dt

stn x
x

-t
e

dt

0
8.

What is the .domain of the function'

(loge x

2
?

/
56

4,

30

,q.

14^

ie"'

.%

0,

c'
ci

..8. he

I..
Chtgin Rule

--,

'
.

Suppose

as a coNosite of4twp '1" rictions, g

we can .eVtes; .

8-8Q

and

can then be expressed in

whose derivatives axe known., The derivative of.

le.

f'-..

and .t.

,., terms of the derivatives of l' g

;(x) = g(Lc))

If

f'(x) =igl(h(X))hl(x),

then

This,resuf

is usually known As thp chain rule..

We have used the chain

h is a lirltar function..

rule for particular functionsin the base where

Fbr exaMplesuppose
4

+ b)

f(X) = g(h(x))

where

u -0 sin u

x -) a,

x -) ax + b = u.

and

Since

g'

and

u -) cos u

the chain rule (1) gives

fqx) = g'(h(x))h'(x)

=[coS(ax +b)a
= a cos(ax + by
which agrees with our previous result.

The general result for linearsubstitution is as follows.

f(x) = g(ax + b)

Let

The chain rule

h(x).= ax + b.

Suppose

gives,

f'(x) = g'(ax + b)10(x)

( ax + b)

that replacement of

multiplies the derivatively

by

ax + b

in 'a general function

a.

A .spetial case of 'the chain rule was used in Section 6 -7 to differentiate

a power function.

g': u

and

Suppw f

h : x -) r log

We.can write

x--> xr.

x = u.

f(x) = g(h(x)),: where

The derivatives of

and

are

1101

given by
gl

u -) eu

and

h'

571
_

_.

78,
-<!=.2:

4. .0
...

The chain rule gives

AP

111/4

ft(X) = gt(h(Xl)h*(x)

=1 gt(r loge i)

r log x

= x

= e

\*.

r
x

r x

r-1
e

`tat us now prove the chain rule by generalizing the tangent approximation
arguments,used in Section 6-7.

Suppose that

is related to

and

by

composition
f(x)'= g(h(x)).
If

,can write

is .ft4.fferentiabIeset

'a

and

is'differentiable at

h(a);

we

o2-

+111(a)(x - a),

(2)

for

close to

a,

and4414*-*
(3).

g(u) = g(h(a)) +51(h(a))(u - h(a)),

In particular, if .x. is otose to

zero so that

is close to

h(x)

We can replace

by

close to

ii(a)%

the second term of (2)' is clope to

h(a).

in (3) to obtain

h(x)

g(h(x))

for

g(h(a)) +.0(h(a)).(h(Z71-E-(a))..' -I-

P
which will, hold if

is close tb

again, this time to replace

ht(a)(x - a).
- a).

so wecan rewrite (4) as

nd diviideby
f(x)1;

x - a

f(dY

to obta n

:41gt(h(a))111(a).-..1

Theref re,'
lim

f`aiv _ it(h(a))ht(a),

x - a

whten'establishes the chain rule:

`II

fl(a) = gt(h(a))h1.(a)

578
p

.r

1119

We now use (2)

Thus, we have

f(a).+ gi(h(a))h*(a)(x 2a)

f(x)
f(a)

by

h(x) ci h(a)).

g(h(a)) + gt(h(a))ht(d)(

'By assumption- f(x) = g(h(x))

then, subtra

(so that

h(x). - h(a)

g(h(x))

(4)

8-8
,

The Leibniz notation

lr

for the derivative provides a convenient mnemonic

dx

y = g(h(x)); that is

Suppose

device for the chain rule.

u = h(x).

y = g(u) 'where

du
dx

flE

We can then /rite

du '

The chain rule can then be

expressed

1E

1Z

du
dx

du

dx

A,
Find the derivative of

Example 8 -8a.

-+/7;3..

..

k)

Ext

= x

g(x)

h(u) . iti

and

+ 1 = u
.1.

so that

4
=

Recall that ht(u)

_1

h(g(x))
The chain rule tells us that

gf(x) = 2x.

and that

2V
,D(47+;)

= ht(g(x))gt(x)
2x

2g 47,

fl

'

Example 8 -8b.

D(esirilx)

Find

To express

sin x

as acomposite of functions with known derivatives,A

O

put

u = h(x) = Sin x,

g(u) = eu

1'

0 that
e

stn x

=g(h(x))

and
e

hqm) = cos x,: e(u) = e

e chain rule gives

,

Dk e

sin x

= g/(h(x))
x

579.

I80

1-11(x

cos

8-8

Example 8-tc.

Fosr

-4(x2+,x + 1)1,

find

ft(-1).
O

We could expand and then differentiate.% Obviously, such a procedure would

be quite lengthy.

Instead we let

+x+.1 = u

h(x) =

and

g(u) = u

10
,

so

that

f(X) ='g(h(x)).,

Me have

ht(x) = 2x + 1,

g1(u) = 100u99,

so that (by thechaih rule).

ft(x) = 100(x2 + x + 1)99

Thus

'(2x

+ 1)'

ft(-1) = -100.

Example 8-8d.

Use the chin rule to show that

thds verifying integrtion formula

D(loge (cos xg-: -tan x,

of the Table of Integrals:

12
.

tan x dx- = -1o0 cos X).

i^1

Put

h(x) = u = cos x,

log

-r

N.

g(u) = loge u,. so that

/

(cos xl
x) = gkhkx)) ,

and hence
A

C9D(log

es

(cos x)) = g.'(h(x))h'(x)

sin x)

sin x4

cos x

-tan x.

Find

Exam le 8-8e.

5.-/Z

dx

if

sin (X2)

,=-

'

We let-u = x
du
dx

,
= ex,

du

cos u,

whence y

v = 1 +

and

gi

and

'

We have

dv
4

1
1 4: gin uf

dy.

dx

.dv

dv
du

Ole obtain

du
dx

There-

fore,

1,

580
.

;;L----AL

8-8

1) 0
k- .7)(c s u)(2x)
(

2) (cos u) (2x)

(1 +

u)

2x cos (x2)
'(1 + sin (x2))2

Example-81R:

Analyze the graph of

y = xe

2
-x
.

4.

The product rule gi

2

D xe-x

= xD e

-x

Dx

_= xD e-X-

+ e-x .

Applying the.chain rule to

-x
,

we get.

( 5)

,2 D e-x

(-2x) = -21-x2 :

.= e-x

F6).

We note that

y'

will have the same sign as'

Q

.-2x2+ 1 - -2(x

The graph falls until it reaches

1

1
(--, -- e

ig

-1/2 )" !

( 1

then rises to

then falls.

IE
1/4

Tospinalyze convexity we find the'second

eri.vativ

Apply the product

rule tto/(6)- to obtain

y" =D

[(-2x2

1)e1

1)15(e-x2 2 )

= (-2x2

Now use (5) and the fact'that

D(-2x2 + 1) = -4x' to

581

'182

D(-2x2 + 1).

btain

y" =

.-

(--2:4- 1)

(-2Xe-X ) + e-X
a

, = (20(3 - 6X)eirX
.
The-zstcond derivative

1 2.

or

..,

< x < 0

14)(x

-x

> 1,,

and concave efql

if

is large then

x1

4*

I./
We know

is large.

1x1

if

-x

we have

2
< e

ix

Then noting that'

x -) ex

since

-x

ixie-lxi

that

approaches
ix i

We can shdw' that

so that the x-axis ,is an asymptote.

= 0,

14

< x,

01'..

xe

)77

0 < x <

:IN

has the same sign as

y"

The graph is convex for -

<-

(-4X)

4x3 7 .6x = 4x(x --

..,

if

< - ix 1

is an increasing function..

....

Therefore, we have
2
14x

ixe-x

See (Figure

ix

8-8Et for the graph of

if

0,

y = xe

-1/2

1
e

I.

Figure

8 -8a

IS e-3/4)

'-7- , --f-x2
,y = xe'

4 .4 e-3/4)

concave

convex

dcmcave

-/S

is large.

!x1

con/ex

"t

8-8

Related-Rates
0

In Section 2-8 we discussed the distance, velocity, speed, and acceleration of a particle moving in a straight line.
depends on the time
f

t -)s = f(t).

The distance traveled

t) ,according to some law which defines a-function'

We have thought of the velocity

ds

ft(t)

as the

rate of change of distance with respect to tithe.

As we know from Section 4-4, We,are not limited to particles inoving in

a straight line.

Furthermore, 14 can consider their. relative motions, as we

did in Section 4-4, with point

moving along the x-axis as point

moved

around a circle.

Example 8-8g.

If a helicopter rises vertically from the surface pf the

earth at the constant speed of

20

to the horizon increasing after

perfect sphere with

4000

mi./hr., how fast is its line-of-sight

minutes?

(Assume thatthe earth is a

radius.)

fr

Since the line-Of=sight is tangent to the earth-at the hoizon,. it'is

perpendicular to the radius of the earth there.
height of the helicopter'is

;
where

represents th

20i

At time

(/n hours) the

(miles), and so by the Pythagorean Theoi'em,:o

= A20t + 4000)2

- 40002,1

lengkh of the line -of -sight to th

horizon.

Differ-

entia ing with respect to ti'e,

V

ds
1
= ,[(20t + 4
dt
e

et

40002)Y

(2( of + 4000)(20))
1
'

20(20t + 4000)

_
4

After

f(20t + Lq 0)

t .

6 =minutes,

2 1/2

40CC

>

(hours)," and

0
o

ds1
7:171

t=

0(2 + 4000)
1 '..
[(2 + 4006)2
40062)172
.

: 80040

(160C4)172

80040

77'.

583
.1
1

181

;,

8-8

ne-of-sight is increasing at a rate of approximately 633

Theref.ve,

mi./min.

10.5

r4

For the last

Example 8-81-1,

minutes of its flight prior to splash-

down, the moonship Columbiaedescended at an average rate of

20 mi./hr.,

The aircraft carrier Hornet was steaming directly toward the

approximately.

di
point of splashdown at the constant rate of

30

If the carrier was

mi./hr.

miles from the point of splashdown at 9:47 a.m. PDT July 25, 1969, how

fast was the distance between the carrier and the Columbia decreasing at
minute before splashdown?

a.m., -1
Let

represent the time elapss41 after the point

to splashdown.

20t

If

and the di%tance traveled by the carrier is

alttde

Columbia is at an

and the carrier is

minutes prior

is measured in hours, the distance Columbia falls is

of

30t: At

t = 0

mile (the distance it falls in

the
3

minutes),

miles away from the point of ,ulashdown, sb at time,

1 - 20t

9 - 3 Ot
.

Columbia is
,is

(1 - 20t)

miles abtve the point of splashdown and the carrier

The distance between them et time

miles-away.

(9 - 30t)

=41

is

20t)2 + (9 - 34002

Hence,
ds
ars

20t)2 + (9

2t(-

-20(1 4 200 ... 30(

g

((1 7,20t)

4-(9

30t)21-]12(2(1 - 20t)(-21

- 300

ct)ail /2
-

.
-1

One minute before spla4idown

t =

so

584
4

185

2 ( 9- - 3 Ot ) ( - 3 0 ]

t.

-20(1

t=

20
- 35)
-

30(9

20.2

f(1 - 7.5 )

30.2.172
30) 1,

(9

fp \

-31

30
- 35)

20
ri

8-8

-246.6
641.1/2.

z -3o.8.

Hence, the distance between the Columbia and the carrier is decrea
approximate rate of

30.9

mi./hr. at 9:49 a.m., one minute before

11.

585

186
I

18
1'

Exerc4es,8-8
1.

Find the derivatives of each of the following by making an appropriate

substitution:
x

(a)

X -4

(b)

x -4e

x -4(2x2 - 2x

(h)

x:4 loge (sin

(i)

(i)

(k)

2
.

cos (x3

(c)

(g)

- 3x)

cos 2 x
e

2 sin x
(d)

+X

(e),

loge

(f)

x -4

(2 --3x2)1,'

x-1

2(x+1)2

4,

And the derivatives of each of the following functiogs by making one

2.

or more substitutions.

(a)

1 + cos, x

$.

11 - (log; x)2

(b)

(c)

X -4

1 +.e

\(d)

2x

X -4 cos(sin(cos x))

Find the derivatives of each of the following functions by using the

3.

chain rule, along with the sum and product rules.

(x2 +.1)-1/2

1)1/2

(x2

/x2
(b)

X -4

- [x
x2
,

+ 2x + 1)

4 ex

2 1/2r 2
ix

2 -1/2
+ a 14

-1/2

si=c

x -4 sin (e
x

- a

+ a2

xt2x

X -4 X

x)

sin x'

loge

(Ix- cos, z)

sr

,586 .

187

..--

8-8
log
(h)

x + cos x

4-

sin x cos x 16ge

(1)

1-)

(j)

cos

(log

x) + sin

(16g

x)
e

.11

g(x)

if

4. .(a)

Show that If

f(x) =

h(t)dt

then

f'(x) = h(g(x))gf(x).

b(b)

Deduce from (a) that if

F(x)" =

then

F'(x) = -2x f(x2).

x
x

(6)

Verify (a by evaluating

sin t at

c -IT

and then calculating its

derivative.
5.

Find the derivative.s of each of the following functions

x

t2/3 at

(a)

(b)

dt

sin
x2
(c)

2
dt

0'
.

6.

(a)

Find the derivative of f

,41
x log x

, x >

(Hint:

Write

= e

x
:

What is the minimum value"of

f.

40

Find the second derivative' ofe-llqrnifhow that the graph of

is

convex.
.-

Determine interval

7.

of incrase-decrease
e
and convexityi-concavity
b

sk tch a graph.
,
.

.*

x
:

---- - [x(x

x -)

(b)

(c)'

2
-

11-1T

- 1

..
1

1x
1 + x

log
e

1 - x

-1 < x < 1

Then

;---

8.

Find the equation of the tangent line to the curve a t the point indicated:
put".

(a)

y = xe

'2
-x

,x=0

, x=
x = 11

(b)=

x2)32, x = frr

(c)

y = sin(n

(d)

y ,,loge (--77-. x2)

.
,

x = .--

x
. -

-(f)

40

y = (ex)1T, x = e
1

such that_for all

A, B, t, D

g(x)' = (Az

If

10.

of constants

The notation
y

of

determine the value of constants

f(x) = (Ax-+ Bain x + (Cx + D)cos x,

If

9.

dx

at x

x,

f' (x) =

Bx + C)sin x + (Ex-

4.

Let

.11.

V
This notation is used in the following problem's.

a 14

12.

Let

_. _

y = i'(x)

x = t

and

in x

y =
_

:'

__

Find

+ .E

Express _

x = h(t).

and

and-

dt

',

tpl
e
.

al
dx
X.=2'

.'

443

"::

y =

.f(x), x' c h(t), . xo

.

.
4:51-1(t0).

L\.

in.terms of t.

V;.

"` t=t,.,`

13 .* Let

x.

g'(x) = x2

x,

?is sometimes. used for the value of the derivative

!x =a

F_

deterMine the value

suchthat: for all

B, C, D, E, F

+ Ex

.'...

:-

Show thiit/

'''-v

..

.11

,
I

dx

114-

cl:t

az

X=X

dx
dt

I t=t

0-

t=t

f'

''''r

-,

'1

. 'Find the follo4ng:

+ D sin xi

(a)

D sin x

(b)

D(.x2 + -sin a' sin

x=v

x=11/14

,,

x=57(/3

it)

588

!'

1$9 i

i;

8-8

(c)

A.(x2

a2)I

(L.1 = -D)

lx=a
(d)

D(f(a)sin x + f(x)sin a + f(x)sin x

)1

0
A

x=a

A spherical balloon if being filled with helium at the rate of

in.3/min.

value of
16.

How fast is the radius increasing when

5

100

has reached the

inches?

A car crosses a railroad track moving perpendicular to the track at the

rate of 40 Mi./hr:

One guarter hour later a train crosses the same

intersection moving 72

mi./hr. along the track.

How fist are the car,

and train separating one hour after the car passed, the ijitersection?'
17.

A 'small rocket is shot s raidht up frog a point

observer'.

feet away from any

75

If the roc et,t/=avels at the constant rate of

how rapidly will,it be receding from the observer

100 ft./sec.,

seconds later?

.11

.04

Sa.
4

8-9
The General Powei and Reciprocal Rules

8-9.

'

A 'speciLal case of the chain rule, known as the geheral power rule, occurs

so frequently that it is worth discussing separately.

can be expresped as

Suppose the values of the function

f(x) '= (h(x)) r

where

where- h

= g(h(R)),

If

is differentiable at' x

.t

f(x)

is a function.

is a fixed real number and

-4 h(x) = u

and

g :

is defined (that is, if -8.

r(h(x))r.1

and if

In other words, ,111

u), then the chain rule gives

is'differentiable at

ft(x) = gt(h(x))ht(x)

Since

u - ru

g'

we can write this as

4 f'(x) = r(h(x))--'he(x)

(1)

This is the general poller rule.

Using the

D notation it can be

expressed as
r-1

r'

Da ,= ru

(2)

Du.

For example, suppose

f

x -* sin3 x

h : x -.sin x.

where

f(x) = (h(x))3,
The power formula (1) gives .
a

f'(x) = 3(h(x))2h1(x)
= 3 sin

x cos x.
.

an example of the case when the exponent

the function
1.6

+ 1 = x

: x -4; x

+ 11

e power forthula gives

.

590
I

1 9 lc.

is not an in eger, con-

8-9

D[(x2 + 1)1/2]
1.(x2

.) -1/2

D(x2

1)-1/,

2(x2

1)

2x

)7-7-7
11

As an example of

he case when

is a }negative integer,-conider the

fUnction

41

t.70

x -)

- (log!kx)

(log

-2
.

x)
e

4
-

The power formula then gives

fi(x) = D(Q-9ge

7'-2tloge x)-3
e

x)

-2

x(loge3
xi
r

casiwhen

. The

tion.

r = =l

is so important that it deserves sp4Cialj.considera-

Suppbsp theevalues of the function

cap'be expresse6 as'.

4f

f(x) =

A.

where g

is a fun6tion.

oe
...

We can then write

f(x) =

Ig(x))-1

'

and apply the power formula to obtain

f'(x) = D((g(x))'iliff. -(g(x))-2ag(;))
=.-(g(x))-2 g'(x)
7.

(g(x))2

s willvhold,p ovided

g(x) PO and

differentiable at

4.

x.,.Inl

...i

words; the derivative

f the reciprocal ,of 'a fuicti n is

de!ivati've of the function times the reciprocal of t

' Using D

re of t e function.
0.

notaiion4 we summariie:'?

-D g(x)

.7;71

e,negative of thee:

[g(x)_l2,

. t.7

59

1 .9 2
1
119,01t

sr

8 9.

3.

We shall refr to this as the recinrocal rule.'

For example, Suppose
o

x2 + 2
.

47

T6 reciprocal rifle gives

a
2

'

(2
x + 2)
x(22

x) =

r-F--"-)
X -14-2

4'

(x

-2x

..

2
+ 2),;

..:

A ldifferentiation formula for the secant function can be fouild using the
reciprocal rule. ,The, secant function is defined by

The expression
e..os

anodd multiple of
.values`

'sec:

cos x

cos x = 0,

is not defined

9lat 3.7s,'if

is

Thus the secant' function is defined only for th

which are not odd multiples of

-s

derivative

The reciprocal rule gives

D(cos x)

D(sec x) =

cos x

cosy.

(-sin x)

cost x
r
.

. 4

,,.
s

sin 30b

cost x

tan X =

, Since

in

xs x

and

sec x -

this result is usually expressed as

cos x

o.

01-

D(sec x)

P.

s.ec x tan x.

A corresponding f9rmuia for the cosecant function is given it the exercises.

,

.o.

...

.
...

592

193
I*

,,-

8-9

Exercises 8-9'0

the power frgniUla to find, the dexlvative of .each of the' followingl

(a)

x -)1T17.1 x

(e) *51

331 -'x)2
(f)

(g)

(h)

4)4/3

(1

0
ILO

co s10 2v

,12

+ 1 dt)

0
oir

dy
dx

Use; .the'reiprocal rule to find

if ,

y - ' 1x2

(a)

(d)

+ log

y =

1
(

f.

(0

-(f)

1 + e2x

,'Y
a'''-

.,35,-=

..

y.= sin x 41cos x)-1

y = sin3/2 (2x),

q (a)

=
4-

y'=

(1?)

4-

dt)

f = p

0
Mr

r
s.=

.(e)

For each of the following

I

state where defined,

(i)

t *

find the intervals of increase- decrease,

(ii)

,.

(iii)

convexity- concavity,

(iv)

asyniptotes ( if, any) , and

(v)

sketch.
1

(a).*.sy =

1 + x
(b)

y= ;117r-T-:).7

,
....

.
.

Find an equation for the topgent linp to each of the.follOwing curves at

/

..11

the indicated point.

/---4x

Tc

qc)

x)-1

11

y 2)
0 \ 1 - x

(b)

',,

.593

194, 0

Show that each of the following is 'an increasing function

5.

'
x

(a)

ex.

10

> 0?

1,

Find expressions for the'derivati.ves if

,.

csc

(b)

-cos

sin x

cot x

(41)', y

in x
cos x

an x

y =

(c)

(a), y .= sg)x,.=

(sin x)(cos

,cos x
-

sin x
and

Use the results of (a), (b), -(c)

(d)

to obtain the following:

D(trIn 3x)

kel

(f) :Ditan 2x
D(sec2 .x2)

Disersc x) i

(i)

1-

ervls
In what ititi

7.

is the

scant function increasing?. *convex?

--

its graph.

8.

(a)

/5:471.10

.D( sec x csc x)

(i)

in terms of

sec x

and

csc x

(ii)

in terms of.

tan x -;pd

cot x

..

(iii)'' in terms of

csc A and

..
e'

cot 2x

Find

(b)

*6...--___--

'

(c)

D(tan, x cot x)'

(id)

D(ain x cscx)

(iii')'

D(cos x sec x)

Find

4.1i)
e ,
,

=r,
4

.0D(sin x cot x)

(..1.)

.'

0.

D(cos x tan x)'

,

Sketch

8-9
et

Show that

If

(taxni(k+1)

x)

sect x,

_,tan

-..

, J

(b)

D(t csck x) = -csck x cot x,

(c)

D(cot

2
.

101_ (a)

x) = D1/4,csc

..

x)

Use the product and rdc.ip`rapal-rules to show that

P

.0

...--

2
(b)

k #,0

Find D (x

(uN
`vi

UV' - utv
v2

x)
I

.
'

p
/1

e*

595.1.96r.

8,10

"

8 -10._ .Ths. Quatient Rule

By combining the product rule and the'reciprocgl rule we can obtain a rule

for dbfferentiating Tiotients of functions. Suppose the values of the function

f - can be

exiSte''sfieci as
P(X)

cluc4.1\,.

f(x)

p and

where

common to write

,q(x),/ 0) .

are functions (Nand, oficourse,

f =

and call

the quotient oe

and

lit is:theA
q.

SinCeigeo
ti

can write
p(x)

'f(x)

.
-

.t.

the -function

and

p ; and the reciprocal of

is just the product of

If

1,

then the product rule gives

q(x) / 0,

and

'are'differe tiable at

q.

f' (x) = D(P(x)

1

= p(x) D(n7r) +

.
The reciprocal rule gives,

D p(x)

P'

Tic

-q'(x)

777

C\-

,1
tilx/

(q(x))2

C
-

4*.

..

to that

ft(x)

(1312S.11
Wx.))2

pt(x)

4.

clCxj

p(x)q'(x) + q(x111(x4
2

(q(x))
w

),This

(1.)

i,s

u uaily written in the form

q(x)Pi(x).- p(x)q1M

'fr(x)

(q(x)2)2

and is refrred to as the'quotient rule.

With

D' notation it can be Written

'`
as

(2)

o,

596

alk

.-1

8-lo
1

4.

11n words, the clerivatiVe of a.quotient is, the denominator times.the derivetive of the numerator minus the, numerator,'"times the derivative =of the denomitater, all over, the square, of the deriominator.

'

Uge the 'quctient rule I find the derivatiy,e of thetangent

fUnctionsr and discuss .its -graph in the intertal - 2 < x <' 2
Example 8-a0a.

Thitangent function can be expresled as

..

-tan.

,.

1.

s in x
cos x .

This function is defined for thos_e x for which cos,x / 0; that is, the
tangent function is deeknedI only when x is .not an odd multiple' of-' 2v
.

Tbe....9uoiient rule gives the 'derivative'. '1.

.
:
..,
.
D(iza
xl
cos
.x
p(sin
x)
- sin x D(cos x).
D(tan i) = `cos x,c.,
2',x
.
_ cos

11

.,

+ sin

cos2
cos

(3)

cogrx(cos x).- (sin )1)(-sq.n ,x)-'1/4,

2
,
cos x
, 7 ,0
,

...

Since sec*,x

,.

'

cos

this is,u$uell.y expressed as

C OS X

D(tan x)+

sec

x.
-'

,,

The function

that ,

is not zero in the interval

x , COS X

sec x > 0,

if

<x

-,

so

g< x < 2

i`

reforepthe tangent f4nction ,isen imoreasing functl.on in the interval. In

-fact the tangent function is stAttlyincreasingYon tAig interval.
Let us4denote the second derivative
of y = tap x by
'A.
A
.
'
y" = D( see2 4 = 2 sec .x 15(_s7 x)

..

y "..

.z...

'

1.

X7`
,

.-

I.

=a

sec

o= 2 o4 s a2'

o
.

x (sec x tan X)
)( :tan )1

t
.*

fo

%0

o- 'S

'?

''

o4

".

,-

,,l'

'

o to ,

'..

'' a ''

597_,
i

..

ie ..., aj.../.4
when w'e us'erthe pbwer
rule and the fact-t--(Secpon. 8.9; (1t, that
,
''''''
pcsec x)..1- sec x tan.x.
:
-

have

1,", 8

,
.

'I

8 -Y9

derivative yr = sect x tan,x 14111 be negative for '

-154x.< 0 find, positive fort 0 < x < ; thatis,"*.the graph of the tangent
funtion'is cohcave. in the left interval and-convex in thr-Night. interval;'
The ;'iec and

10011 x,

cos x

approaches

approaches zero

while

sin x approaches
,

Y.

1. -Thus the ling giVen by. , x =

is an asymptote and,. y = tallx baaoraes

approaches -1 from the left;.. sfrailarly`we could

large without 1potgid as /
ague that y = tan x grown large without-bound_throughaegative talues ps'
from the right. A graph. Of 'the tangent function in 1.1e
x approaphe s .
...
;
'interval, - 7.Ir. x < 7rr is. given in Figure 8 -10a.
.

..)

.,

Y.

Figtre 8 - 1 0 a

= tan x
.

598

I.

I-1

\..

8-10
...

Rational function's, the:C.1.s, quotients oS polynomials', can lie differ-

Such a function ls discusdd in the follow-

eiAiatad Tsing the quotiedts/lUle.

.ing, exampie.', 4*
.

140

.
Z:'

..

'

&ample 8-10b.

pisUcss,the graph of the function

4

'3

+ x
X

- 1

"

This function is not defined when x = + 1.

'

..

near zero.

Thus

,r,

from

+1

f(x) ,

be omes large and

+1 'from the left; tbat.is,'

approaches

while the denominator is negative and

1,

't.

14(x)1

As

91

the -1 4 the n;merator.dpproaches

negative

as

iipproaches,

'grows large without bound through negative

f
Similv.argiimentS'show, that f(x), grows large without bound .13hrOugh

vaiues.

poSitive 'values as

''(xl

+1

approaches

from tl:ie right.

.

,1-Suppose "x. approaches

-1

?rom the eft.

The numerator approaches, -1,

while the denominator is positive. and approaches

-1

front the left,

f(x),

0.

Thus as

approaches

grows urge without bound through negative values.

To discuss the behavi r when

1*

is, large We 'rewrite the- expresVion. for

f('x)" as

1 +

1
-

x3

X
1
1

7'.

If

is laege,'the expression in the parenthesis Is nearly

1x1

behaVes4ke x

,Note 'that

f 'is Continuollexcept

approaches r,

a,

while-the denominator approaches

theria's "i

1.

Thus f(x)

for large vadlues,'positive or negative.

x

For

example,

the numerator,apprdaches
- 1.

Thus f(x)

if

a3 +'a 2

1.,.

approaches,

.,42

a-

2
, .!

,e,

- 1

Ja3 4. a

f(a).. This'is illustrative of the fact that a rational,function

ti

1'

%'41.s coptinigius e)tcept at the"zeros of .its denominator.

We,now determine the intervals of increase and decrease.

The quotient

rule 'gives:

9,

I
4

84,1.04.

.
1) -1,(x3 + x2 - 1)I(x2

(x2 - 1)tr(3 + x2

ft(x)

,'2.
kx

41

(x

(x 27- 1)

x2

(xl +

7:

1)

1)2\.

I)(2x)

1)2

2 .
(Infaet, the dprivatiye of a

The .derivative, ft. 16 'Trational function.

'

.ss

ft(x) -

+ 1)2

..

from whiah,we see that thd sign of

(x .

- ')(x

!k2(x
ix

In factored form, we have

s a rational function.)

rational functiOn is a

ft

f- is: rising when x,< -I

)(x 4- V). 'We concluae'that,the gropph of

,.

,_

or Y.> V3 .and ls 41),ing when_ :,,43 <x < -1, -1'<x < 0, o< x < 1
s ..
1 ,< x < V-5..
,

..

'

is determtned'bythe sign of

or
.

.1

L.

p`91

-1

(IS

3:15
/

I.)

16.

11

-1I t'' 0
1,

..

(-)/s,
A

+1.)ir1 .

cs.

\.

Figure 8-10b

GraphJ?f y

+x2

- 1

,600
0

\.
3

e \.

1:

8-lo

e Exercises 8-10.
Evaluate
D(:)-c.-21-1)

(a)-

(i)
D(1 sl1----t),a)ri x)'.

-4

(11,)

D\c. ex' 2)

x*

"1

1 + x
-

( c)

D (

(k)

(i) D(Cos x sec, x)

2 - x2

5^

(4)

(3 4.* 2x2)',

(d)

4- X

x loge

5471

:1)(, +

ex -ex
x
-x
e

InrDF(1

(e)

+ 1.1(1

+ log

(g)

(0) D

yg/

,x)]
e

1 +.x2

log.'
e

7T-

1,

(h)

.41

X2

1.

4"

2
D( cot x) = -c.sc x.

Shbw that

---

'

3.'

-1

Dispuss the graphsrof etach of the

Sketch.

'following;

as

in

Example 6-10a, b..

.

Y= 2
,x - 1

(P/

II(

+1

-2x

(c) y =

.., FFind

.r.'1.14 see

(a)

x dx
4

(p)

sec" x

tan x

dx

lit43

^....
,A.

,..

.-

--,

lik.

__

601. ,..

202.

i.

,-

1
..

-;-

..,

..

8-3.o

5.

Let y = f(t),

= g(t),

t = h(x),

in terms 6f

(a) .Using Leibnizian notation,.

dy
dt ,

dt

7-3.t.

dt I.

dx

.(b)

Using (a) expres

dz

.dx

in terms of

ft, g',

and

X Xo

'4

.r
ti

602

h'.

and
'

*-

8-11.

Inverse Functions

Let us review our dt4cussions of Section 5-1 and 641 where we defined the

The funcion.

square root function and found its derivative.

-44,4,;

.'6

x --.,x2, "x'

_,..

is a 'strictly increasing function and its gr.aph meets eachhorizOntal line

y = c,

Kiveriby

0.

In other words

g(X1) < g(x2)

< x

0,< .x

if

1 -

116k
...

and each nonnegative number

is in the -range of

c = 4(d).

g; i./.,

The

function

is defined for each nonnegative real number

f(c)

that is

IC

f,

by

g(d)_

is the nonnegative real number

fines, unction
2

if

c.

,c ---'d-.

there is a unique

since for each, o ?P0

This follows from the fact that

such that

This de-

d > 0 .such that

is strictlrincreasing.

The graph of
y

is obtaiiied by folding the gr4h

over the line

that, is

(c,d)

lies on thq-graph of

lies on the graph of 'g.

if'ind only if

.(See Figure .8-11a.)

Figure 8.11a

.603'

20

(d,c)

c.

8711

The tangent to the graph of

.
.i.

----)

""'` N.,A,.,
. -s ,

g(d) +

,,,

gi(d.,),Cx : d) = e+

2d(x _ d).

must alSo be ipbSitive and this line folds overthe line

then .d

c > 0

y=

is given by the equation

(d,c)

'at

iny the line whose equation

y = x

given by

'

y = d +

2d

...

This'is-the tangent to the gfaph of

at the point .(c;d). iRepla,cing

by

...

at

we see that the tangent to the graph of f

I.',E,

has the equation

(c,d)

d.

.st

"A

..lifiL

+ 1---,(Z - C) .

,y = -/

1.'

21:C.

The coefficient of

;s the derivative of

t(c)

(2)

so that

at 4o,

> o,

1/.

?his same method was used to dafint

f

x -loge x, x > 0

#
'

in termsofthe function

1'2

and to obtain the derivative formula

1

=,

Suppose

In this s.gction we discussa general form of the folding process.

is defined for those numbers

t he function 4g

the case'g:x-4 g

be tife,entire real number line (as

case

2
:

x -) x

in an'intervp1

tintuous at each point of I and that

(as in the

Suppose further that

x > 0), or a line segment.

which may

I,

x;4atray
A

is con-

is strictly increasing; that i6,,,

A

g(x1) < g(x2)' if .P.R.;*

(a)'

and

x2

are in

If we fold the graph of g oveY the line given by

graph of a function

f.

xi < x2.

and

then we obtainwthe

y = x,

-The function is called the inverse of

and is

defined by
if

f(c)

t at is,'

f(c)

g(d)

range of

is den:led for those numbers

(meaning

-that

ber

.g = g(d)c

forsOme

in the domain of- f

in

I).

This defines d' function singe for a num.

there is exa43.y one number

in

suc

hat

e(d) = c.- This follows from the assumptioh,(3) thfrt

is strictly increasing.

That the domain of f .is an interval is a consequence ofthe assumption that

4

604

200
4

8-11

"
g

is continuous.

It the-'appendices, it will be shown that the inverse -f

is

.continuous atb each pbint of.tts domain.

The graphs of

f. and

are related

by

the,. condition

t
(c,d)
(4)

It

lies`on,the graph of

lies on the gtaph of

over the lime.given by

(d,c)

g;

that is, the graph of-the inverse

,g,

f .if'and Only if

can be obtained by folding the eaph of

it"

y = x.

The folaGg procese used-to finci"the derivative of the 'square .spot func
tion also works '14 the general case.

tinuous function -g

and that

Suppose

g(d) > 0.

f' is the. inveirs.v Of Vie conr

The tangent to the graph of

at

(d;c)

has-thre 'equation

y = .(d) +

- d).

This folds over, the line given by

into the line whose equat'ion is

y =

y = d

(x

c),
.

....-

``__the equation -of the tangent line to the graph of the interte. f
r

\c,d).

The value

is the coeffice
i nt o f

ft(c)

at the point

x,

I.

,f1(c) =
.

03-dy

if

gi(d) ? 0.
-

/,
.-

To obtain

formula for

ft(c)

in terms of

C,

we replace

by

f(c),

to obtain the inverse function rule,:

(5)

ft(c)

if

gt(r(c) ) > 0.

The geomefically'intuitixo folding process can be justified by rigorous

arguments;

In the appendices it is shown that limit concepts give the same

results; that is,

lim

- f(c)

f(c +
h

h )b

is indeed equal to

Definitions
and derivativesibf trie inverse circular functions can be ob.__
tat:led using this process.

05

,
."" a

0.4

06

a
8-11

r
The Arcsine Function
H.

it is strictly
we restrict the sine function to an interval in ftich

inc4s4g 'then the methods we. have been using. can be applied to obtain an
The
< x <
inverse function. It is conventional, to use the interval
Its inverse
x -)sin Z is strictly 'increasing on this interval.
Kunction g
is usually called the arcsine (or inverse sine) function,, and
.

functi,pn

.denOt'ed by arcsin.

The ranieNof
f

-1 < x < 1.

is defined for
d,

-1 < x < 1

is the Interval

7,

so that

x -)arcsin x
arcsin

C,

Its value at

c,

is that real number

sugh that
- - < d < 2 - .

and

sin d = c

In other words,

(6)

if and only- if

.1(c) =

sin d = c.

and

Idl <

For
example,
_
sin' 0 = 0; sin(- 1)8= .4

)-

sin 1

-11-

'

e,
so that
4

- T ; arcsin 1 = 2 - .

arcsin 0 = 0; arcsin
The graph of
g

x -)sin x_

x -)arcsin x

over.:the line given by

can be obtained by folding the graph of

y = x,

as shown in Figure 8-11b.

y = arcsin x,/ y = x
41

'/

y = sin x

31

4-

Figure 8-11b
606

..
Using the inverse function rule (5), we can express the derivative of the e
f -in terms Of the sine function

arcsine i'unction

f'(c) -

gt

1h this case

cos x,

if

NS

g (f(c))

g.

We have

g'(f(c)) > 0.

so that

gt(f(c))= cos(arcsin e)
and we have

f'(c)

1
-

cos(arcsin c)

cos(arcdin c) > O.

if
'

Referring to Figure 8 -lib we see that

cos(arCsin c) =

c 72

and hence we haye

f'(c)

'if

1cl < 1;

)1- 7-77

tHat is,

Figure 8,-11b

(7)

26king the Chain Rule into account, we write the more general result:

D(arcsin u) _

Du
,

lul < 1.

The graphhe arcsin function has a vertical tangent at

vt

x =+ 1.

This seems reasodalbe if we recairsh%faot that the sills function has a,

No=

horizontal tangent at, x =

The integtation,,formula corresponding to (7) is-1

A-

Thus for

lal <

y
2

and

dx.,= arcsin x,

1x1 < 1.

x2

-1b1 <

y
2

the Fundamental Theorem gives

s's

6o7

20.8
4

8*11

ei

are&

b --arcsin a .

dx.
2

..
.

,.'

I.

by

Replacing Nb

t,

by

0, ,and using the f ct that

arcsin 0 = 0,

we

haye
Q

(9)

arcsin t' =
.

I.

.c

dx;

Itl <

0 irg---/ - x
,

The Arctangent FunctiOn

Ats

The function

defined by,
y

g(x) = tan x;

,
/6

7,.7 < 'f <

'is strictly increasing and continuous., i,arth411;nve-, the range.of

,

'g

is the

entire rea'l line; that is, if c is a ny req. number,: then there is a number
mo
x.
,
x
-between - - and
,
such that g(d) ='c: The inverse function f,
2

known as 'the arctangent function, is accordingly defined for all real numbered,
c

follows:
f4c) . iki!'ctan c

(10)

x
2

Graph of

..;

and

Y = arctan x

it

is the real number

between

,such that 'tan d . c.,

and

y'= tan x pre sketched in Figure 8-11C.

arc 'tan

......

......

(c, d)

c )

1,

Figure '48-11c

698 ,

09

The inverse function formula (5). gives

gt(f(c))

since

D tan x = sec

sec (arctan'c)
A

x.

Referring-to Aure 8-14d, i,./e see that

4-

81%

Figure 8-11d

,
2,
s. ec. k

arctan' c) = 1 + c

and hence

ft(c) =

2
C

.1

In summary, w'e have

'his fraction is always positfire.

D(arctan x) =

'

1 + x

and the correspOnding integral form

S.

(12)

2
x

dx =arctan x.

ti
Taking the Chain Rule-into account, we write the more general result:
4

bu

D arctan u

c'

8-11
Exercises 8-11
,Determine the domain and range and draw the graph of the function

(b)

x -4 sin (arcsin x)/,'

x --)arcsin (cOs`x)/....

(c)
4

2.

arcsin (siri.x)

(a)

(d)

(e)

cos (arAin x
arc-tan (tan

xh

Derive the formula

rccos

-1

1 - X

3.

De x ive

each of the follo ing formulas.

(a).

arccot x -

(b)

1+
1

Darcsec x

/-5"
A

(c)

-1

D arccsc

lx I 1x2

4. ,F,Ivaluate:

fa)
(b)

D (x

s.

'

(c)

5.

arcsin x t arccos

-D

Find

arcsin x)

6.

Find

D (arcsin x)3

(e)

D (.

(f)

D (II:

arcsin h

lim

+ arcsin x )

arctan

, x

arctan x

h -) 0

of

(d)

(H

t:

x)

+ arctan x

What is,the definition of the derivative

f(x) = arcqn x, at

0 ?)

IL if.
dx

(a)

y = arcsin x2

(b)

y = arctan (3x -1-"

ti

y = e
(d)

y = e

arcsin x
2x 'arcsin

8-11 ,

4'
Elialuate

7.

1
(a)

0 I + x

dx

r y/6

------- dt

-y/4 F-7
111.

Find

8.

ir

given by

'2

0 1
x

is

F(x0

'if

F'-(x)

dt

+1 t."

)27-7J
,

..

sin X

dt

.2

1 + t

-.,Q
:S

lim

What is

dt?

il0 1 + t

Show that each of the folldwing functions

10..

find the derivative of

(a)

g:x

(b). g

x
1 + x
1

x >

has an inverse i and

-1)

x -)xlx! (a sketch is ,helpful.)

Show that if

11.

f.

'the domain of

iS the inverse of4-g' 'then

g. 'Assuming that

and

f(g(x)) = x

for all

in*

are differentiable apply the

chain rule to obta'na formula for'the derivative of

f.

Is this the same

as the rule (5)?

'.,

Suppose

12.

fl

and

g be the functi

are the respective inverses of

2

,defined by

g(x) =

and

i(g2(x)).
g.

(a), Find an exp ession for the inverse of

\-

(b)

Use this m

hod to find the inverse

(c)

What is th

derivative of the function

611

Let

of a -)(3x + 2)

x >

2
3

of'part (b)?

4'
,

8-11

invtrce of

811PPO-''.,

1.!

dx I'

as
dt

o(x),, x = f(y)

Show that

'

dy

(The symbol

,Put

g.

dy
dx

x=f(a)

means the value of the derivative of

s,

considered
.

tax

This is the basis .for,,the

as a function of t at the point where t = a) .
dx
(IL)-)
mnemonic expression of the inverse rule:
=
P. A`
'dx1
dy
.

1 4 .

The notation of N.,a;Ier i< gi es a method for finding derivatives..

example if

x,

are :

(L=
dx

then

x =sin .y

so

odx

=cosy

For'

and' hence vf

cod y

co 3 ( arcAlri 4

tl.il method to f:_nd the der'ivative of

se

(),
(b)

. 'irctln x
:

loc

(c)

(d)

y =

y =

x
n

'

x-

/
l

,
c

4.

i.,,,

8-12. Implicitly Defined blinctions

.

,,

A function which is described

in tetra's of .rational orations on, andt
i
compositions 'and inverses Si , known functions is daid to be deftned. explicitly.
/
4
g:b
' For example, if
..-

'''-

y = f(x)

.(1)

.'

e.

'..

i25

x2,

5,

f is defined explicitly.
It often happens that- a function is, lefixed. indirectly or im plicitly.
Thus

2
y

(2)
.

with the restrj.ctionithat 'y

'

defines the same function .as

Q,

. /

..

x -9

21511-.
,

cr

+ y2

If we adt no restriction, the graph of

,;

(2) is the &circle. with radius 5 and

center at (0,0). Only th& upper half

of this circle is tie graph

y
2177(2. (The lower half is the
c:4'

graph of y =

-1-25

x2);
,

We T1..1, of course, find ;fi-(x)

from (1,).

In fact,
-

-2x

(3)

V.25

-x

x2
.

Hoyever, we can also find the 'Slope of the graph frofia.(2) without solving for

A First of all,'
really MX)]

we note that this is

so that by the Chain Rule, its derivcive _at --x is
Dx

= -2x. When we come to

y2

2 (x) fi(x).
Hence, we have

2x + 2f(x) fi(x)= D(25) = 0

and; therefore,

f;(x)

472:c - frx (.f(x)

-13. 2 14-

,Z

or

_" .

'

'

8-1?

Usually we simplify the notation and write

2x + 2yy' = 0

.r

Y / 0)

and

._._,leaving the result in terms of

on the upper semicircle, (4) is'equivaleht to (3);

the answer in the- form (3).

y = X25 - x

Of course, 'since

y.

Ifwe wish the slope at

Often, however, we leave

say, (a point

(3,4.);

which is surely,on the upper semArcle), we obtain

3

Note thee the tangent is perpendicular to the radius whose

slope is

-Which agrees with our geometrical knowledg.

.
k
There are many casesin which it woul4 be difficult.f. not impossible to

solve explicitiyafor y in terms of

"'Example. 8-1.2a.

x.

4.

Given

.
x

xy
y-3: F.. !"

---

..

'

...-,

..

with the point

-

.(1,1)

onits grail:

We

an find the dope

out difficulty but would find it very 'aard :to:no explicitly.

:

y'

there with-

We have
.

..

2
3x -4-3y y' = xy' + y.

Hence,

'

(3Y

- x)Yt = Y - 3x
57- 3x

2
:

and
3y2 - x
2

3y

0:

Yi = 3 - 1

1.

614 2 1 5

:VP

Ex&unple {3-12b.

We find

y2

g.J.ven

at the point (1,2),./

to find .574

'

x3 yt + jx 2y + x 2yyt + y 2

ogg

Thn
.

tyt
11b..

y2),

(x3 +, 2xy)3r1* = -(3x2y

YI = -(3x2Y Y2)_.
.x3 + 2xy

At

(1,2),

yt = -10 = -2.'

t;

possible tosblze or y by the quadratic formula.

Thus

+ 24x

-X3

1.

2x

Which sign must we choos 1 so that y = 2 when

y',

x = 12 We forbear t

4since from here on /the direct method becomek too painful.

Implik differentiation, often simiplies the calcu7ations iheolved in

Problems about reiatLT rates (gection 8-8).
...

Example 8-12c.

of-sigy to the horizon, sand

s2
.

h
+

Let

When

```40002 = (h +4000).

)
dt

10 : h = 2,

Then

wet obtain

to

2s ds = 2(h + 4000 dh
dt

an4

:2

s
$O

the height .of the helicopter.

1ifferentiating implicitly with respect to

(1)

.
.
'be the length of the line-

Re.call Example 8-8a.

40022

40002 = 16002

s = V16002 z

ari= 20, the upward rate of the helicopter.

2(126.5) d1-1

Substituting in (1), we obtain

= 2(4002) 20,

t =15

615

216

40coe'. 20

and "

e
Example 8-12d.

633

46.5.

CIt"

Let

Recall Example 8--81-i--

ithe Columbia falls in

hours, and let

represent the distance

represent the distance traveled

by the carrier in theslme amount of time.

Then

'

d h` +,x2.
A

Differentiating implicitly-with respectto

ds
at

-20 and
pectively.

dx

-30,

dh'

at. + x

at

the velocities' of Columbia and the carrier, res-

t. increases.

One minute before splashdown,

s =

hr

ds

at

and
3-0

42

+ ){.

+ x- at

41.

_ 3

u=

are

minute)

x= 9 4

(the
h distance the Columbia falls in one

'

ti

and

h = 7

Negative signs are included do indicate that

decreasing as.

and

IT have

2h dh- + 2x lx-

2s

or

t,

(0 ) +

8(-30)

1
)57.7

30

'

30.8.

'

Hence, the distance between the Columbia and the carrier was decreasing at
'

the rate of

30.8

mi./hr. at 9:49 EP.M.

616

217

V
-0

8-12

Exercises 8-12
'4

i'..

For pos,itive
r = E' (p, q

y = x

if

x,

integers);then

where
y

= x

is a rational number, say'

..

Assuming the existence of the

derivative

derive the formula

Dy,

,.

Dy = rx -1

entiatia Ind the differentiation formula

2.

For each of the following, find. y'

Dx

using implicit differ-

n-1
,

for integral

without Solving for

n.

as'a func-

'tion of -x.
(a)

5x2 + y2

(b)

'2x2 -y2

,(c)

y2

3x2 * 6y = 12

For 'soh of the fRilowing,use implicit di'fferentlation to kind

(a)

Dy_

y --,, x

y41x
..

x- 14-- 0

+ y, - 2'xy = 0

(d)

3.

:-. .12

'5.

(b)

x y + xy

(c)

xmyn = 10' On, n, integers)

(d)

7:Tc7 + x = y

:-.-

.
,

1.>

41

17

-1

it ,

/-

For each equation, find the.slope o,1 the--eurv;eprisented, at the hated

point.

(a)

2x 2 + 3xy + y

(b),

x3 + y2x2 '+ y3

(c)

x2 - xJ

-+ x - 2y + 1 = 0

'

x cos y = 3x

(-24)

1 = 0, at the point '(1;11_,)

6y2 = 2

at the point

- 5

at, the point

(4,1)

at the point,..(V2-

ta

VID

8 -12

45.

ve represented at the point

For each equation, find the slope og/the c

,

or pointsyhere

(a)

x3'- 3axy + y'-' =b

(b)

xm + ym = 2
2

+ y

(c)

6. Tylq' .y4
(a)

Give a geometric eXplanation for these results.

..--.7A

= 2axy

by implicit differentiation.

;+"-a sin 57,.+ b cos x_= 0

x, cos y +,y sin x = 0

(b) *

.2

(0) -iinxy = sin x + sin y

y) = y

(a)

cse(x

(e)

x tan y - Y tan x = 1

(f)

y sin k = x :tan y

(g)

xy + sin y = 5

1
7.

If

then the equation

0 < x < a,

1/2

+ y

1/2

= a

1/2

defines

y as a

,'

function of

x.

solvIng for

Assuming the existence of.the derivative, show without

f,

8.

A gpherioal balloon is being filled_liith helium at the rate of

the value of .5
Wa

[Use:qmplicit differentiation.]

inches?

r is leaking out of a conical tank at thd rate of

30

The tank is

ft. across at the4op and

10

the water level dropping when the depth reaches

(The volume of a cane is

(a).

feet?

ft.

ft., upper base

If whter,isiollred in at the rate of

fast is the water level rising when the water is

11.

How fast is

,-,

,.....ilosceles trapezoid with altitudes 2

1

ft,. deep.

10 feet long has a'cross section the same shape as an

A trough

base

ft.3/min.

(arena of base).)

7(altitude)
-,

10.

Find

if

clY

dx

+ y

5
1

ft. and lOwer

ft.3/min,, how
ft. deep? '

= 2xy + 1.

c e

(b)

Sketch the graph of

(c)

Sketch the graph of

+ y

ix,- Y1

= 2xy + 1.
= 1.

No.

12e

100

How fast is'the radius increasing when it has reached

cubic inches/min.

9.

is always negative:

ft(4)

that

Work'EXercises 8-8, Number 17 using implicit differentiation.

%.

618

0 1

Chapter 9

INTEGRATION THEORY AND TECHNIQUE

We now-return to our study of "integration, begun in Chapter

that the area bounded biithe graph of a function

lines Ertma

tive of

and

b, was'given by. F(b) - F(a),

f,

where

the x-axis and vertical

F ,is an antideriva-

(the Fundamental Theorem of Calculus, Section 7-3).

,-

7 We saw

Various ele-

mentary,antidifferentiation formulas and the use of tables were discussed in

the final-section of Chapter 7.

In.. the first section of this chapter we pre-

sent a method for extending' the scope of these formulas and tables. ThisC

method is known as the method of substitutionandqs, in fact, the

entiation for

of the chain rule.' By appropriate substitution many unfamiliar

integrals can.be converted into forms previously, encountered or listed in the

tables.

More about the'method of substitution and otheim: athods of integra-

tion is contained in Appenatk 4.

The FUndeliental Theory enables'us to calqulate areas (when antideriva;

.,

tives'can be found).

F(b) - F(a)

where
_

this chapter.

There are other interpretatiOns of the difference

F" = f. ,One of these interpretations is disCussed in
11

We show how the concept of average value-of a functibn is

related to integration (Sacs on 9-2).

Then we show how the average value.

interpretation can be used tb calculate vblumes of solids of revolution

(Sectfon 9-3):
Numerical methods for approximating integrals are discussed in Section
Th se methods are useful, i:articularlY in conjunction with high speed
computer , inestimating,integrals when antiderivatiNies cannot be found.

The

final sec ion of this chapter shows how we. Can- btain_Taylovapproximations
with error estimates by integrating inecatitiitie

9-1.

The Method of Substitution,The scope of our integration tables can be greatly extended by using the

method of substitution.

This methOd often enables us to transform unfamiliar

integrals into familiarones.

rule.

It is ba4ed upon VII integral form of, the chain

01.0" '

410

In terms of antiderivatives, we have learned to symbolize the derivative

statement
1(u)
du

619

20
8

9-1..
by writing

-f(u)du.

'F(u) =

(1)

If

is a function of
d F(u)
dx

.(2)

the chain rule shou.s that

x,

du
dx,

d F(u)
du

du
dx

f(u)
`

%%which similarly'justifies the statement that

f (u)

F(u)

(3)

dx

Together, (2) 4nd (3) show that

4'

\'

f(u)du iiff(u)c-cti dr,

(4)

re

0
c.

if u

x.

is a, function of

., This equality (4) vastly increases the number of antidexivativciW-e may

determine.

It often happens that we are confronted by a rather complicated

integral, in terms of

x,

say,-which becomes substahtially &implified,and

familiar if we can express it in terms of a suitable variable

.functi,on of

u which is a

x.

For example, suppose we seek.to'determine the antiderivative

i'2x cos x

dx.

If we let
u =

x2,

du
= 2x,
dx

and
then

-17 2x cos x

Vcording to

, with

cos u

bix =

f(u) = cop u,

du ,,

;de may conclude that

,

du

cos u r dx =

f
62o

cos u du ,

9-1

and We should recognize this antiderivatiye as

Hence,

sin tr.

2x cos x2 dx =

and upon substituting back

u = x

cos u du = sin u

we find

2x cos x

dx = sin x

2
,

as desired.'
ay

The leibniz notation,

is more than a convenient device for

remdmberiAg the chain rule and the substitution rule (4).

laticians in practice to deal with the "numeratorr'
nator,"

dx,

as if

LI
dx

were a common fraction.

It prompts mathe-

dy, and the:denomi

For example, equation (4)

suggests that operationally the symbol

du'

dx

dx R

may be- replaced by the symbol

du

when 1.4,0 perform substitutions to integrate a functioh

"du," "dy," etc., are called differentials.

t.

The symbols

"dx,"

In practice they short cut the

thinking required to evaluate integrals by the method of:substitution,

as

the'examples below inniCate.

To find suitable substitutibris to reduce an integral to a known form is

no easy task and, in fact, May not be ivssible (see Example 9_la, below).
Practice is required to ob-tain,skill.at making appropriate substitutions.

Example 9-1a.

Put

u = x

Find

xex

110.

dx.

du
= 2x
dx

so that

and hence,

du = x dx.

Upon writing'
2

2
dx =

xe

(x dx)

J
we can make the replacements

621

222,
2

u = x

and

du . x dx

to obtairt

\
x
e

dx =

(x d.x)

eu(.1.1 du)

l ir

du

iy

Now replace

u by

to obtain

L')

x2

irxe x2 dx =

The formal substitutions (5) and the equation

x e

dx =

re
are- ist

shorthand: for the statements

x

lir u-du

2x dx = f e

where

lr e udu

dx

u = x2.,

Example 9-lb. ' Find

t,

sin (.2x + 3)dx.

Put
u
u = 2x + 3 , - =2.
dx

Make the substitutions

1

u = 2x + 3,

du = dx

e.

622
I

223

sin(2x + 3)dx =

(sin u);-- du

cos u

.1

derivative by

(6)

'

cos(2x + 3),

In general, we have seen that replacingtx

dertyative-by a.--hus_replacement of

If

_ -

,
.

--/

.0..

sin u du

ff"

ax + b ,multiplies the

by

multiplies the anti-

by ..ax + b

',

2t.---F--that is

F(x)

Example 9-%e.

then

f.(x)dx,

Find

tan x dx.

tan x =

sin x
cos x

Flax + b) =

ff(ax + b)dx

$ince

D cos x = -sin x

it seems appropriate to try the substitution

u = cos x.

Then t = -sin x,

and we have

-du = sin x dx

so that

Itan

x dx = J7

cos x

(sin x dx)

f
Cl

du

= -loge u,

if ,u > 0

= -loge(cos x),
e

if

cos x > 0.

The result
Jrtan x dx = -loge(cos x)

is formula 12 in the Table of Integrals. ,(See Exercises 9 -1, No. 9, for a

justification of the absolute values sign.)

623

224

,..

0:

9:1

1\

Put u

loge

Find

Example 9-1d.

x loge

x,s: so

dx.
X,,f

that

du

that is;

du =

dx.

Thus
1

ix

dx

loge. x

x(x dx)

loge'

1 du

= loge u,

if

u > 0
log_ x > 0.

loge(loge x), 2',0if

Example 9-1e.

; ,

sin2'x cos x dx.

Find

We try
u =

sin x,

so that

ILI = cos
dx

x.

theltsubsfitutions

u =-sin x,

du = cos x dx

thus gives

sin

x cos x dx = fu2 du

u3
"7"
= sin
3

r_

r 1,
Example 9-1f.

Find

(1,

- x2)5

One way to do this is to carry out the indicated Multiplications and

calculate

624

225

9=1

5x9

(x - 5x3 + 10x5.- 10x7 +

0

Let us, instead; try the Substitution

2

u = 1 - x

'

du
= -2x,
dx

so that

x dx = - - du

and we have

f(1'- x2)5 x dx =

u5( -

-u

du)

(1 :

12

Hence,
1

(1 - x
12

(1 - x )x dx -

Note that replacing

by

2 6

f
4'

and

du = -

1 - 0)6
12

x2

TE

u: and that

for

rE

In other words, we can express the limits of integration

complete our cilculation in terms of

'Find the area bounded by

.

x = 0, x = 1, .y = 0 * and y = xlx2 + 1.'

In integral notation our problem is to find

1

+ 1dx.

x.

2
u = x +

terms of

and

The next example al o illustrates

this fact.-

15:ample 9-1g.

1
12

in the expression

u =
gives the respective values

and

- 1)6

du
dx = 2x,

so that

625

226

du =x dx

9-1

N.

when x= 1.

when, x= 0 and u= 2

u= 1

Substitution ok thetse gives

1
X

2 uip

+ 1 dx

du) =
,

1/2

du

1 2

=.i:-.3u

3/2% i 2

3/2

--(a
3

1 ' 3/2 I 2

-7-:

3/2

-)

= -(212

1)

,t

'Find

Example 9-1h:

dt.

10
Let
u = -t

du

= -2t,

...

-ff du .-- t dt

...

----,,

Go that u -= -1 when

and u = -100

t = 1

wheft

t\\ 1.0

::--------

\'

1
.3

t
1.

-t

dt =

-t

-1

.
,

-- --

-100

10

10

(-u)eu( .: .3ff du)

(t dt) ) =

ueu du.

"?-
ar

The Tables give

ueudu = ueu

-t2

1
,*,

tae

dt. =,:(iteu - eu)

00

10
1

= ff[(-1)e-3-

e-1)

((-100)e-100

e-100)]

=-101-e-100
2

(,

626

1. 1 + 3x 2

Find

Example 9-1i.

dx.

+.x3

tat,

bz.

If we put

u =x +

du

then

x3,

= 1 + 3x

9.x

so that
t.

1 + 3x

1-

dx =

x +

2
+ 3x )dx)

x + x3
a.

du

tt,

= loge u,, if

A3

= loge (x

u > 0,'

-x

),

> 0.

.A

ft..

x2
' Example 9-1j.

Find

x6

.0a

16-tj'<

1:15c'''

Let us try the substitution

u =

du

6
,

= ,6x5.

We can tiAn trite

x = (1 - u)
x

1/6
,

so that

and

i1/3

= (1 - u)

5/6
x5 = (1 - u)-'
1

5/6

= dx.

Hence

i'lll

- x

dx J7.

1(1
u

u)1/3
1/2

;(1 - u )

-5/6
)du .

r'.

This lattir integral appears to be quite complicated, so. let us try mother
r.

substtUtiOn.
Put

u = x

du

3
,

ax

= 3x

2
du = x. dx.

dx,
3

- ,

This gives,

627

-4;te-t.;

a3
1

dx =
0

(1
u2

du)

a3

=
-

arcsin u

aresin'ea

3
V,

lo

dx.

Find

Example 9-1k.

Ii77}.7

Try the stptitution

/'

u =
o

'

du = dx

+ x,

to obtain
2

x2
0

IT

u - Q.) 2

dx =

yl

2 u2

a; + 1

-1/2.

du =

)du

/2

+ ,u

5/2

k5, u

'u3/2

_1/2)

2
1

14

16

15

15
For example, put

Other substitutions'are also useful.

du
dx

u = VT77

so that

(u2

2udu = dx,

Cl

x2

1,1

Ir-.

= 2
.C.

or.

(u

2111-77

and

1)2

dx

- 1)
.
u

(2u du)

= 2.1 Va(u

- 1

2u2 + 1)du

2(5.u5

-.-

?.3- u3

14 vE
15

628

229

16

Find

Example 9-12.

e-x

We first try
2

u = x

du = 2x dx

ti
0

obtaining

e-x

-u

dx = e du.
21:

The latter seems to be more complicated.

Searching tables of intvrals,leads

us nowhere for we find neither expression in our tablds.

substitutions, such as

u = AT,

We could tr5Pother

or even make wilderfstabs, such as

u = sin x.

In fact, no matter what substitution we trv:weshall get nowhere, for it

2
-x
was shown by Liouville in 1835 that the integral of. e,
cannot be expressed
as a finite combination or composition Of polynomials, cirtular.fnnctions,

exponential functions, q,logarithms.

te*

.
p

629

230

9-1
Exercises 9-1
vo
1.

Use, the indicated substitution to find each of the following:

they appear,

a, b,

and

dx;

u = x

x2
(a)

(b)

x37/1

(a + bl-X-)13

s[

x4 dx;

(Wherever

represent non-zero constants.)

.c

3 + a3

u.= 1 -

dx, b

A 0;

(d)

dx

fe)

u = x - 1

ax;

u = x2 + a2

'

+ a

(f) sis

aA

clx,

0;

'(cos x)s sin

(g)

u= 7= sin x

(h)

b + ce

(i)

cbc, c A 0;

r sec x dx;

u = b + cex

u ='sec x + tan x
1

2.

Find each of the following integrals by making an appropriate linear

substitution.,
(a)

2x

dx

(e-)

(f)

.
(b)

(1

1 ;)10

...

3x

dx

1
(32;(

1(1 - 5x)3

dx

-,..5

(c)

(d)

-.

(g)

---ix
a2

4./5

(h)

tan(1- x - 3)dx

-TT. dx

ti

.4

sin ax dx-

630

31

'

9 -1
3.

Find each =of the following integrals by making an ap propriate substitution.

sin x

3x2) x dx

(a) j- (4

( 1)

f (a

dx

cos x )2
a

(b)-jr cOs5 x sin x1 dx

dx

(i)

(4x3 .

(c)

r din

,(

dx

2x cos 2x lb:

404.

(0.4777

(e)

j- x

(k)

+ 4x2 dx

ir

.(loge x)
(r)

dx

(2)

gx4

dx

(m)

[Hint:

sin

x cos3 x dx

cos3

x = cos xcos x

= (1 - sin

cos ,r2x-dx

x)cos x.)
[Hint: :Re-'

sin3 4x cos8 4x dx

(n)
11

write

sin3 4x. )
.

Write

4.

EValuate each of. the following

1/2

(a)

2 f2x

dx

(g)

sin x dx

(h)

1)

1/2

cos4 x

(b)

2 log- x

r
(

dx

-1/2

Tr/ 3

cos lor, dX

dx

(1)
1
1

(d)

(2x +1)17 dx

(e)

dx

-1/2

(k)

f fr...7-1-dx

x3
e

dx(

(r)

dx

(t)

I r.;

631

232

x sin (2k2)dx

5.

Sometimes it is useful to reduce An integral to a known integral by

making two or more separate eub8titutIons.
2e

dx

e2x

we might put

p = e

du = ex dx
2ex

audthenputv=4i-1,2dv

4 + u

du

4 + u

= du

to obtain

dx =

+ e2x

For.example to find.

to obtain

4 + 4v`
= arctan
e

uv

1 + v

koarctan
= arctan (

J7
u
2

).

Find:

a
cos.

(a)

dx

sin2 x

2 + x
dx

(c)

AT + x
.

.17

6.

Find each of the following by making appropriate substitutions and then

using rf table of integrals.

(a)

f2

x e
ea.

(b)

][ x2 sin (x - 1)dx

2x

dx

f(f)

5c3 e-4x

x ex2 sin 2x2 dx

ti

x--sin ojx. dx

(g)

(d) is :ecos3 (4)dx

(h)

,(c)

loge (x_ + 1)3x

(sin x) Ibge S cos x)dx

632

23-3

9-1

x
jel

(i)

sin x

f
2 cot

7.

sin( x + 1)coe(21 + 2)dx.

(2)

xA1

,dx

x + cos x- 3

1 e :2

Even though

dx

dx

dx

,x

loge 502 + 1

cannot be expressed in terms of elementary func..;2

tions, approximate valued of the definite integral

dx

can be

found (using for example, the method s)of Section 9-4).

can then be.evaluated by appropriate substitutions,

ir1

Related integrals

Suppose

e"4? dx = Cc.

0
Show that
r.-0

dx = a

(a)

(X..1)2
4

2
(c)

e-x

(b)

dx = 2a

(d)

dx'=

-x

dx = 2a

0 1X

-1
(a)

-1

-1

Use the substitution

x = sin u

J_

to find

x2 dx.

Find

Let

+ x2

dx.

x = tan u.

ft

*41k

A complete discussion of this,method of substitution is given in Appendix

4.

633.

234.

7; 9-1
. 9.

4a)

From Chapter 7 we kfiaoll that if x.> 0,

1
3. 4'

dx = loge x.

Shol4 that if

x < 0,

theri

dx = log

by substituting
(b)

t = -x

Lx1

in place of

In the Tam of Integrals the result of part (a) is given as formula

2:

'

e
--"

dx = loge lx1

,
1

This formula can be meaningfully applied to calculate

fis
only if

and b

have the same sign.

No. 37.)

634

Why?

dx

(See Exercises 7-6,

9-2

9.2.

The Average Ordinate, or Mean Value, of a Function

As.iwe have seen, one possible interpretation of

is the (signed)

area

A _shown in Figure 9-2a.

Figure 9-2a
The average value of

f(x)

on the interval

[a,b]

is thought of as the
b

height of the rectangle with base

(b - a)

whose area equal's

f.

Figure 9-2b, it is denoted

f(x)ay.

Figure 92b
Thus, we define

()

f(x)ay.

by

(b - a);* 1(..x )ay.

635

2a6

f'

In

9-2
,
If

then

F' a f,

f =.F(b) - F(a.),
a

and equation (1) )ecomes.

Fk)
lab -

(a)
- a.

f(X )er10

F(b) - F(a)
has been encountered previously as the
b - a
Oterval, a < x < b.
of change of the function F on

The ratio

For example, if
fixed point at time

4resents:the distance of a body from a

s = F(t)
t,

.,

then

ep

416)

- a

F' = f
is,,

is the velocity,gf

ative

ty function of the Motion; that

can then be interpreted as'th4-

F'(t) = f(t)

The A,

a < t < b.

,is the average velocity in the time -intent

he160dy at time

t.

Thus the integral

,'

1
,

b - al

b 'I'

-(!,,,iv,

b - a

i_.,--

motira:is the average velocitytof the..7,.

time inter4a1

'tn.-the

'c

..,trt

,,

- F(a)

',.(b)

=.

a < t < b.

In gAneral, no matter what:theinterpretation for a particular function'

s

.,,

f,

,,

....-

the number

:.

1,10

la-

(2 )

is called the average value of

.(2).iAery useful.

This interpretation of

the interval.

or

In the next section we will see how the concept 61 average

Averagihg ideas also lead

value is related to volumes of solids of revolution.

.

us to useful methods for approximating integrals (see Section 9.4).

.

Example 9-2a.

SuPPose an automobile travels between two points, 100

,. miles apart, traveling at an average speed of -30

first half hour, then at an average of

of (hetrip.

What must

miles per hour for the

miles per host for 'the remainder

be in order that the trip shall take 'wg hours?

I

Let

f(t)

denote the velocity of the automobile at time

t.

While.we

A.
o

do not know
0 s t < 2

explicitly we do know its average value on the interval

and on each of the subintervals

and f < t < 2

0 < t,< 75/,

;':

'These are respectively:

"

636

237

9-2

I
2

100

g=
C

=-'5o

o
;1/2

f = 30-

and
2

f = C =

f -

.1'

0,

1/2

'2

fs'= 100

Since

f = 15,

and'

o
C =

170

2,

-00
0. - 15) =

= 56..67

3
s

Hence,Ithe speed we must dverage in the lett 1

bouts in 'order to average

nours

is approximately 56.67

50 miihr.,fd; the

mi., /hr.

tip
Suppose

Example 9 -2b,

function

graph of

f,

g, x = 0,

x -9c.

What must

x = 0,

x = n

x = n

x -) sin x
c

y = 0

an('

and that

is the '9onstant

be in order that the area bgended by the

is the salve as the area bounded by

and y = 0?
0

The situation is illustrated in Figure 9-2c.

the height

the
the shaded rectangle

the shelled ewe under the curve

OABC

Our problen) is to determine,

so that its area is the same as

y = sin x,

= sin x
,

Y =c
4'

A
(Tr)

'10 '"0,1

'ige 9-2

637 4

23.8

The area of the shaded region under the curve y = sin x

is

'71

sin x dx = -cos x

= -cos g + cos 0 = 2

0-,

while the area of the rectangle is

dx = c(g ,- 0) = gc,

10

and-therefore, ,gc = 2,

that is

c = g

The number

is just-the average
'4

2
g,

slue ofl

f:

l
1
f.

0
PP

ta

638 039

a.

s'

9-2

Exercises-9-2
1.

f. and find its'average

For each of the following sketch the graph of

value do the indicated interval.
x ->3x2 +

(a)

(b)

f ,x->

< 0

01

7,

0 < x < 1

4 + e

2
(c)

(d)

s - se

-1 < s < 1

.2

1 ,
(e)

< t < 4

f\)( x

1 < x < 0

3x +

'

Find the average tralue of the siner function on each of, the following

2.

intervals.

0 <x <n

- (a)
(b)

<x
_
_ <

(c)

1 + 77r < x < 1 + 9n

(d.)

c _
< x < c + 2n,

where

is any

number.

Show that if

3.

average value of

on any interval of length

pendent of the location of the interval.

Ex

is a constant,

(See Exercise

Find the average value of the slope of the tangent to the graph of

4.

Le t

-1 < x < 3

in the interval

+ 1

x -flc

5.

then the

is periodic and integrable with period' a,

represent the averagd value of a function

on the interval

av

For

[0,1].

x' ->x2,

show that
(fav)2

(f2)ay.

6._ Suppose a particle moves so that its acceleration.

,,

1
a(t) = t3 + .4.--

time

J.;

What is its average acceleration_ in the time ihterv'l

It

< t <
7.

*,

Shbi hat if

is linear then

[average value of

on

p < x <

f(p) + f(q)

o
4

639

9-3
tr

9-3.
,

Volumes of Solids of Revolution

The Fundamental Theorem of Calculus enables one to calculate areas by

finding antiderivatives.

Such techniques can be extended to enable balpula-

tion of are length of curves and,volumeS and surface areas of solid figures
I

to be made.

coUrses.

A full.L.eatment of'these topics will be left to subsequent

In order

tiation techniques

o give you an indication of the wide use of irit*Elifferen-

e shall discuss in this section the problem of finding

volumes of solids qf revolution.

Sdppose the region bounded by

y = f(X), ,x = a,

x = b

and y = 0

is

revolved about the x-axis, as shown in Figure 9-3a and b, obtaining a .solid of
revolution, as shown in Figure 9-3c.

Figure 9=3a

y = f(x)
4

Figure 0131)

640.

241

9-3

-A

.Figure 9-3c-

The volume of this solid can be determined by

procedure similar to the

Let V(et) be the Iliolume of the solid

one used to estOlish the Area Theorem.

obtained by revolving the region boupdea,by

y = f(X),
about the x-axis.

x = a: x = rind y =00

(V(t) is the volume of the shaded pOrtion of Figure 9-3d.)

Figure 9-3d
661
4f 4.

42

1.

+
TIV.8 defines the volume function,

t < b.

V : t -,V(t),

Let us also assume that

that

is increasing on the interval

is continuous and-nonnegative at each point of this interval.

using elementary properties of volume we shall thaw that the derivative

of

and

a < x < b

By
V'

is given- by

(1)

a <t <Th.

Vt(t) = ;(f(t))2,

The Fundamental Theorem of Calculus will then give us

b

V(b) - V(a) =

a(f(t)), dt.
a

for, V

is a function whose derivative is

t .4x(f(t))

Since

V(a) = 0,

we

Obtain from this the desired volume formula:

b
V(b) =

(2)

'

s,

,
It(r(t))

dt

Let us now prove (1).

Suppose

is fixed and that

V(t + h) - V(t)

(3)

,is the volume of the shaded region shown'in Figure 9-3e.

Figure 9 -3e

V(t + h) - V(t) = volumpO6f shaded region

)

J,

642

243

h > 0.

The 'quantity.

e-

The function

9-3

f is assumed to be (Creasing so that

(
for

f(t) < f(x) < f(t + h)

t,< x e t'+ h.

f.

Hence; the shaded solid in Figure 9-3e is inclUded in the cylinder

,centered on the x-axis with radius

f(t + 14 and length1.

CI.

(See Figure

9-3f)- Furthermore, the shaded solid'of

Figure 9,3g

Figure 9-3f

&terior Cylinder

Interior Cylinder

C1

C2

Figure 9-3e includes the cylinder ,C2

f(t)

and length

h.

centered on the x-axis with radius

(See Figure 9-3g.)

Recalling th0 the volume of'a

cylinder is
g X
O

(radiuslength

we have
volume Cl = g(f(t + h))2h
a

and
.% 2
volume C2 = g(fkt)) h.

Since the shaded region of,Figure 9-3e has volume

C

and is incltded in

C1,

V(t + h) - V(t),

we have

volume C2,< V(t + h) - V(t) < volume C2

643

2 44

includes

..

4.
1

9-3

that ,is
.

rc(f(t))2h < V(t + h) --V(t) < y(f(t + h))2h.

10
h

was assumed, to be positive we pan divide through by

Ito,(0)2

v(t + h)1- v(t)

(1:)

to obtain

\,.

< yr(f(t + h))2,

'
.

.
.

As

approaches- 0, f(t'+ h)

y(t + h) - v(t)
h

If

4t)

approaches

approaches

,,2

n(ftt))

so that
'

as

approaches

0.

is taken td be negative, the inequality (4) will be g.e'versed but the

conclusion (5) rem9ins the same.

This establishes
that, indeed,
.,\

v'(t) = n(f(t))2
and completes the proof of (1).
Remark:

The same result (2) will hold if

is assumed to be deCreasing

or if it is assumed that the interval can be subdivided into subintervals so

.*

that on each subinterval

is always increasing or

ways decreasing (see

'

the Remark in*Section 7-3 after the proof of the Fun amental Theorem)..!-. The

result can also be Atablished using only the assumption that

is continuous.

Before examining some examples, let us intrpret.the formula (2) in terms

of the concept of average value.

Consider a cross-section of the solid of

Figure 9-3d, perpendicular to the x-axis, cutting'the x-axis at

as the shaded region

of Figure 9-3h.

Figure 9-3h
A cross-section R.

644

245

(t.#0),

such

9-3
.

.....1
The region Al

is circular and has radius.

tharea

For each, t, a < t < b,

so its area is

f(t)

gcfkt))

C(t)

of ghe cross-section throUgh

(t,0)

therefore has aree

t.

C(t) = g(f(t))2.

0*

This defines the cross- sectional area'function

C,

C:t-) C(t) = g(f(0)2.

The_average value of

on the interval

a < t < b

is

b
a

1"

C(t)dt =

g(f(t))

dt

The voluffie of the solid of-Figre 9-3d is thus

b

,.

,
I.

(b - a) x

C(t )dt

g(f(t ))2dt,

1a

that is

(6)

volume = (length) x (average cross - sectional area).

In other words the cylinder formula

2

volume = la h = (cross,-sectional area) x (length)

can be extended to give the volume, of a solid of revolution merely by replacing
the cross- sectional. area (which is constant for a cylinder) by the average

Cross-sectional area.

This gives a convenient device for reconstructing the

'formula (2).

EScample 9-3g.

Find the volume of the solid of revolutiOn obtained by

revolving the regiod bounded by

jy = sin x,

x = 0," -x = g,

y ='"0
.

` ,about the x -axis.

,

To find the cross-sectional area function

sectIOn perpendicular to the x -axis through

region

is a circle with radius. sin t.

C

Thus,.

t g-)g sin-2 t.

616
.1

4.

C,

(' ,0).

lilt

be a crost-

(See Figure 9-3i.)

The

of

The average value of

on the, interval' 0 < t < n

n.
2
n sin t dt

rc

C(t)dt
0'

71

so tir desired volume is

'2"N

rg
1

(length) X (average cross-sectional area)

TC J

sin

ti

n sing t dt

t dt.

To calculate this integral one can use the tables or recall that
.

sin

1 - cos 2t

t _

so that

sin

t dt =

n
2

(1 - cos 2t)dt
0

sin 2t \
'
2

n it

2(t

1:
,.-

2
rc

'

Atth ia our,desired volume.

- ,

Example 9-3b.

Find the.volume

revolving the region bounded by

(i)

(ii)

of

the solidtof revolution obtainld by

x = 0, x

0; y = x2

about

11,e

the y-axis.
.1
4

E44-6

47

9-3
In each case we shall find the cross - sectional area function and apply (6).

(i) Revolution about the x-axis s.

A cross-section perpendicular to the x-axis at

has radius

(t,0)

t2

(see Figure 9-3j) and hence, the cross-sectional area function is:
C

,tt

The desired volume is

'

(length) x,(average cross-sectional area) = 2 x

C(t)dt

0
.

Y..:

4 -

y = x

= It

t14. dt

32 it

note scale
x

4
'

/ y = x

change

'

48.

9-3
4%-

Revolution about the y-axis-

(ii)

Revolution of the region about, the y-axis gives the hollowed out cylinder

A;

Figure 9-3i

.1

Sr

fop' v lew

648

t:4 9

!4,

9-5
In'thid case we take cross-sections perpendicular to the y-axis through
The,crois-section is a circular ring with inner radius
.1.

(See Figure 9-3k and 9-32).

12

(0,t).

it"_ and outer radius

The area of this cross - section is

A(A-)2-= A(i

C..' The average value of the area for

0 < t < 4

t)9= C(t).
is

C(t)dt .

(1 - t)dt

f
2
A

77-

Hence, the desired volume is

.47
(length) x (average cross-sectional area) = 4

= 2A.

Y4

`b

649

ft

9-3

(Note:
a c c

It is essential inproblems of this type for your solution to be

ompatitch,,1

1;

4,-0

- Exercises 9-3

-), ye,

'

Find the volume of the solid generated by revolving about the x-axis

the region blow the graph of each of the following functions and above
the indicated interval.
( a )

(b )

(c)

(d)

x -)3x,

0 < x < 2

x < 1

1,x1,1

-(x - 1)2 + i,

(e)

(f)

(g)

(h)

f : x -) tan 3c,

-)

4_

-1 < x < 2

0 <'x''<- 3

0 < x <

Use the procedure of this secti.on to find_the volume

v.'

cone of altitude

3.,

1 < x < 5

)176-g7,

_<x <2

2.

<

and base of radius

ofa

right circular

r.
4

Obtain thi formula for the volume of a sphere of radius-

by `first

showing ,that the sphere is a solid of revolutiag.

4..

Find thA volumk of theielliPsoid generated by revolving the ellipse

'2
2
x
-7 + y = 1 about
its manor axis.
(Assume a > b.) ..,
_

5.

Find the volume 'of the Segment of a sphereNzf

parallel planes if the bases of the segment ar

from the

6.

cri:..tzif

7.

y2 7

at distance

and

A cylindrical hole of radius. 1

a. solid. sphere of radius

about the' x-axis.

inch is drilled out along .a diameter of

inches.

evolving the region bounded

y = x

, and the line

j bounded'by twv-,,,

and are on the same side.I

Find the - volume of the solid obtained by

by .the parabola

erius

Find the` volume of the material

cut out.

8.

Pind 'the volume of the portion of g sphere 9f radius *r

remaining after

cylindrical hole is drilled out along its,diameter if the length oP

a

the h le is

N.

Check your answer by cOnsidering some special: cases.,

65d

'
/

) re 0 1
4.

v
.

9-3

9.

Find the volume of the solid of revolution obtained by revolving the

region bounded by x = 0, x = 2, y -= 0, y = x2

about

c,

11,,

(a)

the line

y = 4.

,(b))

-ths4 line

y-= -2.

(c)

the line

(d)

the line

x = 4'.

-.1

2.

tO

4
1

.s

do

r.

*0

..

.0

t
e .

..a.,

651
o
(3 rp z

t..

b,

Estimation of Definite Integrals

9=4.

The Fundamental Theorem of Calculus tells us that if

then the integral

F(b) - F(a),

is

if

is continuous
f.

is raft antfderivative of

Thus the problem of calculating areas (or average values, izolumes'of-solids of

,

'revolution, etc.) cabe solved if we can find an aritiderivatike for

The

f.

problem of finding sL,,::11 an L,ntiderivative. in terms of elementary functions may

not be solvable (see Exaiple 9-1.0.' Even it the problem is solvable the fO*M
of the antiderivative may be incc.uvenient

Various,methpds haysibeen developed

for estimating the integral

With the advent of high -speed computers

f.

obtaining approximate solutions

these methods have Lome valuable means for
o
such me thods will bf discussed in this

to.area and related problems.

section.

is increasing and continuous.on the interval

Let us suppose that :f

""b

.
,

a <

and that we see

to estimate

...-^r
,

'

be cohtaineh in the rectangle

ABCD

f.

The region bounded by

and will contain the rectangle

(see Figure 9-4a), so that

("b

area ABC'DT <

<

f < area ABCf.)

,a
=

ABC'Di

will

9 -it

The region ABCD

while the region ABM'

b - a

so that

b = a,

f(a)' and length

haS height

and length

f(b)

has height

f(a)(b - a)

f < f(b)(b - s).

'<

fl a

If we take the average of the numbers

= f(b)(b - a)

and

a = f(a)(b-- a)

we expect that this will give a better approximation to

than either of

s,

the numbers a and

ThisJtverage is

S.

leads, thereaere, to the approximatiqn,

This leads,
f(a) + f(b) i
kb - a).
2

(1)
.

This approximation will, in general, not be very good.

exact if

It is, however,

is linear, for if

cx + d,

. then

c(b

f _

- a
2

2
)

a)rca + d + cb + (13

+ d(b

f(a) + f(b)
2

The estimate (1) is just the area of the trapezoid

of Figure -4a, that

ABED

is, thekestimate is the integral of the linear function obt;tinedAy connecting

.(a;f(a))

to

with a straight lipe.

(b,f(b))

Thp ystimate (1) does not, of course, require that

can ,be used for more general functions.

be increasing and

To Obtain' better approximations to

f we can subdivide the interval

the integral

a < x < b

into small

subintervals, calculate the average (1) in each of these subintervals and add
these together.

Let us find a formula for this lapproximation in the case of

equal subdivisions.

Suppose

is a positive integer, and let the points

an_1( divide the interval

a2,
-

each of length

b f.
a
,

[a,b]

into

as shown in Figure 9 -4b..

v,

653

.,25/1

equal\sub-intervals

924

a
2

an-2

an

-1

Figure 9 -kb
,

Then we have

a2

al,

f =
.

f +

s,...

an

a3

f +

f + ... +

f.

a.
,..

a4

al

.Cf

n-1

ai+1

Each of tice integrals

is then approximated using (1), that is

a.

rai+1

f(a.

+ f(a. 1)

(a

- a ), i = 0,

i+1

, n'-

By adding these estimates We obtain

b

f(a0) + f(a31)

(a2) + f(al)

(a2

_tai} '+

f.

al)

f(a3) + f(a2)

Since each subinterval

i+1

Thus we can factor out

<

<

f(s n-1)4 f(a n)

a2) +

(a3

b
i

b % a_

J. =
'

and obtain

has length

;1/

2,

a -'

(a

- a
n-1

it follows that

1.

b - a

f 0 -1!----((f(a0) + f(al)

+ (f(a1) + f(a2))'+ (f(a2) + f( a3)

'

+ (f(an.1) + f(an)))'
The terms

f(a0)

and

f(ard

appear once, while each of the'

ms

f(a

f:(an.f) appears twice, that is:

f(a2),
t

(3)

b-a.(f(a0)
-T17-

2f(a-)
+.*
e

2ffal)

2f(an-1)

This approximation formula is known as the Trapezoidal Rule.

It approxi-

mates the integral by the sum of the areas,of the trapezoids obtained by
connecting

(ai,f(ai))

to

by straight lines, as shown in

(a1.4.1,f(ai4.1))

Figure 9-4c.

i+1

Figure

An obvious question is, "How much error is.involved in using

I2n8(f(a0) + 2f(al) + 2f(a2) +

fb

to approximate

+ 2f(an_i) + f(an))

-.t

f?"

It can be

that-the error is at most

Ito
)

M(b - a)

12n

where M

is a bound for" the 'Second derivative on the interval, that 'is,'

(x ) I < M, a < x '< b

See Calculus, SMSG, p. 831

*".'!7

65

.4

256

Example 9-4a.

- dx, correct to one decimal place

loge 2 =

Estimate

'be *using the Trapezoidal Rule.

loge 2

dx

,n = 2):

zff2) [1

-Px3

f '

-P -

M = 2

so the second derivative is

1 < x < 2 'so its

This function is 'decreasing' on the interval

loge 2

the maximum error in the es 'cimate

PP

ti

_M(b -

a)3

b = 2, a = 1, n = 2

Using (l) -with

= 2.

maximum on the interval is:

and

f": x

17
1
+ --)
=77.
2

,..
is

+ 2( -)
3

The first derivative of

2.

- -3

-P-x1

The formula (3,) gives (with

into two equal parts by setting

1 < x < 2

First divide the interval

2 -

13

24

is

-27
12 n

In other words,

16

t<

Since

-16

-,- >

24

0.66

18

--r- =

and

24

18
24

'kr

0.75,

this tells us that

0.66 -<

log

2 <0.75.

-.

1,,

Rounded off to one decimal place

0.8

2 <

log

24.

so we need to choose

loge 2

tould therefore, be either

0.7

larger to ob.tain assurance as to the first

decimal place iii. loge 2.,

AO"

c
Let us try

.
*--=

3,

which gives the points

4
1, a1 = 3
1
'

a2 = 52

3,

and the estimate

2

loge 2 =
,>

- dx
1 x

2 - 1

e X40

+2()

3) + 1 )

2(5---

'21
30

656

257.

rk

or

9-4
4fr

The maximum error is obtained from. (4), with

b = 2, al= 1, n = 3

and

M = 2:

M(b - 4)3

2
12

12n2

13
9

_so that
21

Since

721- 54.1 =

184

> 0.68

..

7 < loge
1

and

21

2 <

21

35

.1"

-I,

1
__
. 194m<

we have

0.72

0.68 <loge 2 < 0.72

that is, correct to one decimal place,

loge 2 = 0.7.

'''1,-

Sitni:Isonts Rule

Consider the Trape oidal Rule in the case when

n = 2:

f Al b- [f(a) 1. 211241)

f(b)).

a
.

Divide through by

b - a

and write with denominator

3

(5)

b - a

fa

f(a)

3:

f(a 423)

f(b)

Thi s relation expresses the average value of

over the interval

as a.14'S'ighted average of the values of

at the endpoints and

a < x < b

midpoint of the interval, the weights bein g

7,

exact if

a + b

is linear on the intervals

[a,

7,

This approximation is

7 -) and

+" b

[a7---, b],

but it

is not necessarily exact fo r non-linear functions.

This raises the possibility

that .ppme other choice,of weights might give a better estimate of the average
value of 47. In other words., we may be able tow choose a1, a2,

and' a3:to

othat
.

a11

a2 + a3 = 3
3

and

a1f(a) + a2f(-14-D) + a3f(b)

is a better approximation to the average value of

on the interval than is

(5).

657

5
a

9-4'
a

.,at us 'see if we can choose weights al' a2' a3

so that a

+ a2 + a3

and the approximation

b
(6)

f =

b = s
'is,

a quadrdtic function, say

is exact if

. ,

4. b) +-a f(b)

a 110 + a2 f(

, 1

--,

cx

+ dx + e.

In this case
b

cx

+ dx
---+ ex

2
1

c(b3

7=.

a3) +

e(b - a) I

1(b2_ o2
2

(b-a)(3(b2+ab+a2) + 2(b+a)+e3

e)+4(c(a--.}--.b ;2 4,d(ab,+p,

1240[ (ca2

'

b -

Thus, when

at

4(a +2

) + (cb2,,+ db +

'

to

b)

e.4_,_

is a quadratic:

..1..

1
b --.

__a + b.
fka). + 'erk-7)
.

f-

1 ...., . l
-ff IA n)

.,

-.,

a, = 77, a, = 2, a3 = .f

In other words, the choice of weighfs

will make

the approximation (6) exact for quadratic functions.

The resulting approxi-

niatpon'

b -

_,

4r(

(7)

,
a + b
b) + fkon

is known as-Simpson's"Rule.

The approximation can be improved by subdividing the interval 'a .X <-15e

into smallsubintervals, calculating the weighted average (7) in each sub,

%.

interval and adding these together , To obtain the general formula in the case
of equalisubintervals, suppose

a
2'

'

of length

be the endpoints of

2n

Let the points

is a positive integer.

(ab1,

equal subintervalS of

, a0
each

b - a
,

1)ective midpoints.

and let the points

(See Figure 9-4d.)

a3,

:a2n-1

be the reS-

658

C
1

9-4

14

2n-4

n-

a7

2n-1

M.

"Figure y{-4d

We have:
a

f.
a

a2

0
a

2n

aL

f +

f =

2n-2

2i+2

Each of the integrals

is then approximated using

(7 ),

.a21

21+2
b - a

f a 7ri(f(a21)

4f(a2i+1), + f(a2i+2)), i

n -1:'

0, 1, 2,

a
2i.

Add these together for

i = 0, 1; 2,

., n - 1

f a 11,-.61[(f(a0)-+ 4f(ai) + f(a2)) +

Maid

and factor out,

EFT

(f(a2).+14-f(a3) + f(a4))

f(a6))

4f(a5)

b - a

'"

(f(a2n-2)

+ 4f(a2n_1) + f(a2n))]

Which can be regrouped to give

O

(8)

a Mao)

4f(a1) + 2f(a2) +

-..

2f(a2n_2)

4f(a3) +2f(alft!
4f(a2n_l)'

f(a2n)]

,
NOte that the'coefficients of

f(a0)

and

cients of the remaining endpoint values

each

f(a2n)

are each-- 1,

f(a2n 2)

f(a2), f(a24),

and the coefficients of the midpoint values

the coeff,i,

f(ai),

are

f(a3),
A

659

260

.o.

7.,

-1

(2n -1are

each

the e stiMate.

The number

4.

the number of points used in

The,formula- (8) is the general form of Simpson's Rule.

-,

It can be shown

that thaverror in using Simpson's Rule is at most

(9)

--

where

hf. ..#1s

is

7.(x5.1

a)3

a bound for the fourth derivative

< Mi, a < x <

f(4)

on the interval, that

Comparing this with the trapezoidal error (4):,

*d'

see that if

enough the factor

is

1
7.4.

in (9) is much

12n2

smaller than

1
.2

in (4) dndherice,:the approximation using Simpson's Rule

will usually be better than the approximation using the Trapezoidal Rule.

se SiMPson's Rule with

Example 9-4b.

n = 1

"loge 2 =

we have

With It.=, 1,

ab = 1,

(7) (which is (8) when

n = 1) gives
2

-log
e

--

-7-74

I1 7

1.25.

b.

25

3s
The derivatives Of

f : x

-.

are

1
: x

"

2 '

-' -6

,f

( 4)

See Calcul,ts,

SMSG, pp. 833-4.

'660,

and' n . 2

to estimate

9-4

ti

.The function
is

4 -)(1)

MI1 = 24

f(4

= 24.

is

decreasing on the interval. 1 < x < 2

so its maximum

Now use the error estimate (9) with b =.2, a = 1, n 1 1,

to obtain the maximum error

24(2 - 1)5

1
120

24

180
Thus, we know that

1
120

.25'

loge 2

496

120

jb

Calculation gives:
25

0.688

120
25

76,

120

506
720 < 0.703

so that

0.688 <

loge 2 < 0.703

so that,' correct to One decimal-place

Noticte that by using the values of

loge 2 = 0.7.

at

points and Simpson's Rule

we obtained one decimal place accuracy, while the Trapezoidal Rule would not
guarantee this for
The case

n = 2

estimate (9) (with

points (see Example 9-4a).

will substantially improve the accuracy, for the errn.....,

b = 2, a = 1, n =.2, Mi = 24)

24(2 -

18o

(4)4

1929

In this case
a '

- 1
.

a1 =

_ 3
a2

=22

a3

a4 =2
and Simpson's -Rule

(8) gives

fj2
O

- 77

gives

' 1

9-4
i

)1

'

log

.4

1 tx = 2'7 [1 + 4(,) + ,2()'.

..1,x
6 X 2
----35'

2' =

3/1747

'

1747

.40

+4]

.
.

2520

IP 210 /

on

so that
.

11*

1747

2520

1920

Let us use the estimates

2 <

< log
e

1747

2520

Iry

1920

1920

< .0006

.1747
0.6932

2520

0.6933

from which we obtain

if

0.6926 <. loge 2 < 0.6939.

Thus, correct to two decimal places

loge 2 = 0.69.

To obtain the same.

accuracy with the Trapezoidal Rule we need to use the value of

points of the interval!

c4-

1,

at

14

9-4
o

Exercises 9-4

1.;

Estimate

4. 1
1

.7,

041 + t

,,.

using the Trapezoidal Rule with

dt

(4.s., three p6ints)

h = 2

'(a)
(b)

Also-, compare your result with the

Estimate the error in each case.

.,

`known value of

0 1 += t

72

2
2.

Estimate

using tables for

dx

0
n = 2' and

with

dt.

-x
e

and the'Trapezoidai Rule

Estimate the error in each case.,

n = 4.

3? Estimate
0 1
(a)

n = 1

(b)

n = 3

dt

using Simpson's Rule with

three points)

Estimate the error in each case and compare with the known value of the
integral.
2

2
4.

Estimate

and
5.

n = 2

-x

dx

n = I

(three points)

Estimate the error in each case.

(five points).

Show that Simpson's Rule is exact for cubics, that is

b

'\ if

b - a

1
:

'-)Ax 3

Suppose

+ Bx

---6 --(f(a)

+ 4f(ap)) + f(b)q

+ Ox + D.

(Hint:

B = e,= D- = 0

for the case

6,

using Simpson's Rule with

It is enough to establish this

since it is known to b

II

true for quadratics-)

Till the OPrapezoidallTae give too

---

is a convex function.

_._.r;_______..

f?

large or too small an satimate-foi.a

'

--7. --The letter n

appears in (3) and (8).

How is each use of

related, to
.

the number of points psed in the estimate?

,

...

8.

How large should

be taken in Simpsobs.Rule to give

0 1.+ t
accurately to five decimal places?
b

663

264'

dt

9 -4

9.

Use either Simpson's Rule or the Trapezoidal Rule to estimate' loge 3,

correct to four decimal places.

Use the Reciprocals Table to aid in

computation.

664
c) 4.4;

9-5

sip.-

9-5.

Taylor Approximations
In (6) of Section 7-5 we noted that

(1)

'I-1410dt <
a

g(t)d t

if

f(t) < g(,O,

a < t<_ b.

a
1

, This inequality cpn be used to obtain the Taylor approximations for a given

function with remainder estimates.

.

We first illustrate.this process for the exponential function

on the interval

[0,M],

,that is, for

0 < x < M.

x -)e

On this interval

1 < ex .

j In (1) we take

f(t) = 1

and

g(t) = et

with wit = 0

Figure 9-5a.)

Figure 9-5a

Then

1 dt <

c
Carrying out the integrations, we obtain
1

665

tdt:

and

b = x.

(She

9-5
x.< e

- 1

or
x < e

x
.

Now we apply (1) to these resulting functions, again with

b = x,.

and we obtain

et dt.

(1'4, t)dt <

,o

(See Figure 9-5b.)

ti

Figure 97513 /
4
2

Thus

and

ex - 1,
2".3"

'2

Repeating this proces

< e

x
.

we obtain successively

2
x

X3

ric.,

< ex
'
41-x

+,2!

3!

666

207
I

a = 0 and

All of those approximations to

too small.

ex

approximations, we use the fact that for

ex

(recall that

x -)ex

We use (1) with

xn
< ex.
n! -

.+

3.

. 2!

eM

is an increasing fun tion).

f(t) = et, g(t) = eM,

and

as before,

(See Figure 9-5c.)

Figure 9-5c
We have

x
dt <

that
x
e

Or

To obtain upper

0 < x < M,

x
e

- 1 < eM x
< 1 + eM .x.

667

2G8

dt, )

a = 0, b = x.

!,

;,

With

iitlt) = et

1.

and

*.;:"

g(t) = 1 + . em

t,

(7.) ,gives (see Figure 9-5.)

Figure 9-5d

et dt <

=10

fti or

t)dt

(1 + eM

7-,44ex-1<x+e M x2

thEit is
x:

+ x + e

,e

M x2

If we continue in this fashion we have

(3)

ex <-1, + x

fr' x

n1

M .n+1
x

(n + 1):

From (2) and (3;)

e. = 1 + x +

xn
--r
+ . . . + --r
+
n.
2.

x2

668

2 69

Rn (x)

9=5

0 <Al (x) <

n

%'e

ised this result in SecScion

chose the interval .10,1)

to estimate values of

so that

e xn+1

(n + 1)!

where we

M = 1

R (x) <

The procedure used with

x
e

can be applied to other functions.

The essential idea is to start with information about the derivatives of the
function on an interval,

[0,M),

derivatives

...,

f', f",

In the case of

say.

are the same As

x -4e

x
,

all

itself, so that to say,,

for example, that

(4)

< f(x) <13

10,M)

on

is the same as to say that

a < f

(5)

(n)

(x) < 0

on

[0,M).

The gene5alization that .1.4e require is not (4) but (5).

To illustrate the general procedure let us take

n = 4.

fore, with

a < f

(4)

(x) < 0

on

[0,M).

ye work first with the left inequality

a < f
and integrate from

(x)

x, (0 < x < M).

to

I- x

a dt <

rx

Then

f(4)(t)dt

J 0

ax < frn(x) - fol(0).

and
Hence,

(4)

Integrate agdln from

fm(?\;..co..7< ful(x).

to

(f4*(0) + at)dt <

fm(t)dt
0

ko,obtain

669

We begin, there-

9 -5
2

< f"(x) - f"(0).

f'n (0)x +
+,

and

e
0

< f"(x).

f"(0}-+ fn1(0)x +
a

Continuing
2

If'(0)' + f"(0)x + f"(0)

+34_<fqx),
-

2!

ii4sand finally,

(6)

+ ax4 < f(x).

f(0) + ft,(0)x + fiTc + fn.

If we work With
f

'4

(4)kx)
,
%

< p

in the same way we pbtain

(7)

f(x) < f(0) +."(0)x +

Hence, from (6) and (7)

.,4

+ Vet (0)

f(0) + fT(0)x +'... + fin10).)3-iT + R3(x)

f(x)

(8)

L.:

It ereax

IT
< R3(x) < ax

The bolynomial

p3(x) = f(0) + fT(0)x +

is

the Taylor approximation to

"

f"2!

fnl3(0 )3

of degree not exceeding three as it satis-

fies

p3(9).= f(0),

pt (0)
3

= f'(0), p"(0)
= f"(0), and p7(0) = fin (0).
3

The 'inequalities (1(61 and (7)

can then be written as

4

for

< f(x) - p (x) <

4

In

0 < x < M.

general, if
f(n+1)(x)

0 < x <,M

la

670

A- v

271

9-5

(9)

then

pn

is the

ne

2!

Taylor approximant to

nth

axn +1
< f(x) - p (x) <

(10)

TR- + 7 7 1- -

For nonpositive

Ox

n+ 1
; 0 < x < M.

(n +1)!
.

< K

For example, if

< M

for

n:

and

analogous results can be obtained.

Ifn+le
(x)

and

f(n)( 0,

f"( 0)x2

f,(0)

pn(x) = f(0)

is the Taylor approximation (14) then

r

(U)

in+1

0 < Ix' < M.

for

If(x) - pn(x)I < K

Of course, all these results assume that

(n+1)

satisfies the conditions of "

the Fundamental Theorem of Calcu/t's.(see Section 7-3).

Let us look at another example.

Example 9-5a.

Find the third degree TayloXPprokimatipn to

f : x -,11-71 and an error estimate for

Writing

with

f(x) = (1 + x)1/2

p = 1 + x

we o

0 < x <_ 1.

and using the power rule

in the successive derivatives:

fqx) = 2(1 + x)-

1/2
0

= -

f"' (x)

xY3`

+ x)-52

x(4)(x) = - 14(1 + x)-7/2

In particular

f"(0) =-

f(0) = 1, f'(0).=

so the third Taylor approxi

(12)

p (x) = 1 +

)01/2

f : x -4(1

nt to

x -

f":(0),.

x3
.

.1+

1 _3

+ Tux.

671

72,

(D1.1a

= ca.P-1Du)

Abcording to (10) the error in

Since

(0 (x) = -

is determined by .
3(x)
TE(1. + x)-7/2.'

is decreasing- on the interval

f

(4),

> f

1`

0 < x <

(4) (x)

> f

(4),k0j.

Substituting we calculate

'
f

i C.,

(4) (1)-= . 15
.,.(1
10

( 1

jg

14 < f(4)(X) <

SO

-15iff

'

+ 1)-7/2

07/2

and

"'

0.-

= ..

a 1111

.,;.

, 0<X<

1.

We oonclude from (10) that

x
- 7 Tr
<

(13)

1 2
(1 + x) /

151 x4

/30 <

.00.0.

go*

In particular, if

x = 0.2

p 3(0.2) = 1 +

1(o
2

2)

40

'1.0955,

IT..2 Is approximately

which indicates that

approximation may be found by substituting

15

ko.2)

- 7.

0 <`x < 1.

1.0955;

x = 0.2

< 117g - p(0.2) <-

The error in this

in/ (13):

151.2"

(0.2)4

3t '

which works out to be

-.00006 <

- p(0.2) < -.000011

from which we conclude that

1.09544

< 11-.7 < 1.09549,

4

or

PS

1.0954,

correct to

decimal-place's.

The Logarithm and Arcteingent 1-Unctions

The above methods can,,be applied to give the Taylor approximations to

x

with remainder estimatet.

loge (1

and

)C

x -is arctan x

These results can be obtained in sharper form by

noting that

r/..

672

2 3
'I

#'
AI ,..

e ...q

44) :

1120.

ci

,,

F A:

,2,

..
a

ol.

(15).

arctan x.=

4
of ck

I
o

1
--ndt

0 1 + t`

r-4.
and. Sinding suitable expressions 'for

1 + t

and

From the fgrmula

1 + t,

for the sun of a geometric prOgressiop,

1 '+ x + x2 +x3 + .

1.te" obtain

..

+ xn-1

xn

4.

'

1 -x

,1 -x

'1 -x

..

..,

ti

(16)

IP

1
,

-1+xtx +x
2

n-1

+ .., + x -

Ot

(16) becomes the desired expansion of

x = -t,

If

If' x = -t

2,

t2

1 - t

1-x
;

(_1)n

(_1)n-lt n-1

t3

1 + t

(17)

(16)7'pecomes the desired expansion of

1 + t

(18)

t2

(_1)n-lt2n-2,+

tli- t6

tn

2 ;

(...1.)n

Using (17),

-t_72n

1+t

we obtain for

x > -1.,

..

.,

loge (1 + x )

..C.

..6

x
'

(1 - t + t

- t

,
,n
,n-1 n-1,
t
)dt + k-1)

+... + k-1)

t 2-

= t - 2- 4.

=x(18)

t 3

t4

-T,

-7

n-1

+ ... +

-i

n
77

-1) n.

C x

J'

7 4. 7 -4- + .. :

we similarly obtain

673

tn
1 -+ t dt
lo

n-.L. x

. dt

1 + t

. a
A

dt, x > ;-1

(17.

53?

9-5

.,,

'.'r

J .0

'0'

Q.;
,.

loge (1

'

and.

::

4'

'I,

..

.7=)

Ft.

i.

1 4- t (It.

9-5

7
I

arctan );/=

(1

- t2

+ t

-'t

t2n

(..1)n-lt2n-2)at

dt

1 + te"

P
X3

=x ---+

X5

2n

-1)

2
d

2n-

We conclude tha t

loge(1 + x)

(19 )

x3

x4

(:_i)n-1

-4-- +,

where
(20)

,R

to
dt,
1 + t

=(-:1)n

and
arctan x = x -

2n

3
x5
+
3
5

Ix

1 x

2+
Rn
n

t2n

Rn = (-1)

(22)

dt.
t

0_

In Section 6 -9 it was shown thatif

f

1.0 g ( 1 +

and
2

x -

-n

( -

then

f(n)(0

pn(0), p(0) = pn(0), f"(0) = p;!,(0),

so that

(23)giv
approxi

nth degree:Ta lor approximant to

an explicit f
,f(-X ) .

For example, if

a for

rap

R
n

Hence, (19) a,nd

invo ved in using

pn(x)

can be easily es imated frgm (20).

I

and we put

0 < x < 1,

then

he error

f.

t >

to
1

'

to

9-5

(,since

1 + t > 1

so that

t > 0)

if

0 < t <

if

< .tn

xn+1

tn+1

- n+1

n+l

dt,

and- hence, _-

g <

dt =

IRni =

(23)

'

Therefore, the errorin using

0 <
- x <_ 1

is. at most

to approximate

p (x)
n

log (1 + x),
e

for

n+ 1

n + 1

.V This will be small if

is large.
--

Other intervals for

it will be shown that-if

are considered in the exercises.

becomes large, but in fact

if

= co

log (1 + x)

p (x) differ substantially from

n
f

the approximations

x > 1,

>-1.

co

Hence, for

lim
n

will not approach -0 as

then .R

x > 1,

In partfcular,

when n is large.

'the methods are easily adapted to show that (21) and (22) give the Taylor
approximations to the arctangen,Stand anexplicit formula for the error.

&ample 9-5b.

in (19)
in
to' estimate

Use

Formula (19) gives

loge -2.

6
log

2 = 1 -

wheie
1

t5

R5 = Usin

dt.

(23) we have

Thus, the estimate

1

loge

I
4-

."

5 .= 0.783
I.

675
O

is within

1
;.

of being correct.

to use (19) tp estimate

accuracy..

loge,2

This is not very good, in fact, if we 1A,sh

we must choose, n

very large to obtain much

Clearly, Simpson's Rule is a much more useful methoeforapProxi-

mating values of

loge..

,72

.C9

,*
r

:9-5

Exercises 9-5
1.

Start with the inequality

from

to

x >,0

x,

(a)

-x <'sin x < x

(b)

, xy < 1

(c)

x3

< x - sin x <

3! -

(d)

y
2
y3
3!

(e)

x5

Vos

and by repeated integration

obtain

'

2.

cos x <

-1 < cos x < 1,

x - (1 --L.\ < x
IT
2:'
x3

< sin x - (x - 7T) <

x5

Establish the inequalities of Number 1 for

(Hint:

x < 0.

Rather than

repeat the integrations use the odd and even function ideas.)

f
3.

Find the third degree Taylor approximation to

estimate for

4.

0.

0 < x.< 1.

in the interval

x -4.1177-7;:

-1 < x < 0.

rak

Do the same for the interval

(b)

Consider the function f

and an error

Estimate the error in the third degree estimate for

VI)

5.

VT-Ti

-.5 < x < 0.

1

-51 + x

Show that the formula (17) gives the Taylor approximation to

(a)

[Z.e.,_ p

n-1

(x) =1 - x + x2

.Taylor approximation of
Assume that the error

(b)

(-1)11-lxn-17?.

...

f.]

IR

n-2)

(n -

<

Find a statement for

11 + xl

x = 10 what iecthe error using

If

is the

f.

ost

D (10)

to approximate

-5

--

ir ?

(c)
,a)
.

6.

I
---- differ froth
'

How' does

1 + x

n-1

(X)

x > 1 land IT is large?,

if\
.

Find the Pn-ltx Y 'Taylor approximation,with an explicit remainder

formula for

x 7)---2 + x

----

[ Hint
Hin

2. + x

2_

-or.

of

7.

For what values

.

will the remainder approach

Da Number 6 for the function

] .

f.

as

n ) co?

x -41oge(2

677,

,2178

.,

8." Re 4111 that

tan a t tan (3

Aan
(a)

1 - tan b tan

Show that
1

= arctan 2 + arctan

.)7

(b)

Find

correct to two decimal places by using

n,

(21) ar
.

) and formulas

How many terms do you need to use

(22).

...

(a) SO t. (11
at

40
.

(b)

.411,

Tt

7 . 4 arctan m- - arctan

1.

Use (a) to find

How many.terms

correct to two decimal places.

It

do you need to use?

Using. this method_ n_ has been claculated on high spded computers to

more than
(a)

100,000

decimal places.

Show thal/;.,

,e1

loge 2 = -7 log

(b)

How many terms o

need to 'use (a)

9 + 2 log 24. + 3 log 81

e 00
e 10
e 25

the Taylor approximatiolito

o calculate

place6?

Find the Taylor approXrMants to

X -4 loge
with remainder estimate for

1x 1

loge(1 +

loge 2 correct to

< 1.

279

do you

decimal

0.p.

Appendix 2

MATHEMATICAL INDUCTION

A3-1.

The Principle of Mather.atice." iuction

The ability to form general hypotheses 'n the light of a li,nited numbpr
of facts is one of the most important signs c6f creativeness in a mathematician.

Equally imoortant is the ability to prove these guesses.

The best way to show

how to guess at ajgeneral principle fro,-.71irlited observations is to'gil.-e

examples.

Example

(n>.c:er

2UTM.E

l_tegers:

=
1'
3

5 -'-9 =

Notice theft, in each case the sut.1 is the .s.1.,,az'e'of tne nunher of

Conjecture:
.(This is true.

The s= of'the first

1 <t00,

Conjecture:

odd positive integers is

Car, you show it?)

.Eample A3-lb.
'

Consider the folimiing inequalities:

2 < 100,

3 <l00-,

4 < 100,

5<

All positive integers are less tnan

100,

100.

etc.

(False, of

course.)
! Example A3-lc.

Consider the numbel:7 of complex zeros, inciuding tne

repetitions, for Polynomials Of various degrees.

679

A3-1
.

Zero'degree:

/ 0).

(a

no zeros

0,

-a

First degree:

a x + a -5"
1
0

.:.

x = -- .

odczero at

a x

Second degree:

two zero

+ a1 'x + a0.

at

if.-

-a
x -

4' 4?

8'6112
.

2a

di

4-

Every polynomial of degree

Conjeciure:

zeros when repetitions are counted.

-n

has exactly

complex

(True.)

Otserve the operations necessary to compute the roots

Example A3-1d.

from the coeff:.,

Exarlole

The zero

Conjecture:

of a polynomial of degree

can be given in..

terms of the it.fficients.ty a formula which involves only additick subtraction, Multiplication, division, and the: extraction of roots,

Take any even number except

Example A3 -le.

'

(False.)

2' and try to express

the sum of as few primes As possible:

C

. 2 - 2,

6 = 3 + 3,
12

+ 5,

10 = 5 + 5,

= 7 + 7,. etc.

5 +

2 -can be expressed as the sum,of

Every even number but

Conjecture:

two pries.

8 =

(As yet, no one has been able to prove or disprove this conjec-

ture.)

Canmon to all these examples is the fact that we are trying to assert'
the members of a sequence of things:

something about

the sequence of odd

integers, the_seq,ence of positive integers, the sequence of.degrees of poly.

nomials, the sequence of even numbers greater than

2%%,Thesequential/atar-

acter of the problems natura.1y leads to thei ea of sequential proof.

know something-is true fo

I.P.We

the first .few members of t e sequenc, can we u e

that result to Prove its truth for the next member of the sequence?
done that, can we now carry the proof on to one more

ember?

Havi

Can we repe

the, process inlpfinitely?

0
Let us t y the idea of sequential proof on E ample A3-1a.
.;

that for the,first

odd intoegers

(1)

+ 3 it 5 +

1, 3, 5r ,..; 2k - 1,

+ (2k ::1) =

Suppose we know

IL

k2,

680

r
'4,

8
e

A3-1

can we prove,that upon adding the next higher odd number

From (1) we have at on

the 'next higher square?

(2k 4- 1)

by adding

we obtain

2k + 1

on both

(1

+ 3 + 5 +

+ (2k - 1))),+ (2k\ 1)

,

+ (2k + 1)

(k +

It is clear that if the conjecture bf Example A3-la is true at any stage then it
is true at the next stage.

Since it is true, for the first stage, it must be

true for the second stage, therefore true for the third stage, hence the

fourth, the 4fth

p

and so on forever.

%, Example A3-1f.

In many good toy shops there is a puzzle which consist,, of

three pegs end a se, of graduated discs as depict .1 in Figure A:-1a.

posed is to transfer the pile of discs from one pe

following rules:

to ano-the

The problem

under the

1.. Only one disc at a time may be transferr4from one p g to ano4her.

2.

No disc may ever be placed over a smAr.ler dr.ksc.

Figure A3 -1a

Two questions prise njtUltally;

t e

Is it possible tip execute th

If it is possi le

restricti ns?

complete lhe trarisfer of the discs?

how many moves'does i

If it were not for the idea o

proof, one might have di ficulty in attacking these qu4tions.

)1

task unc]er

take to
seqAentiaL-

L\
.

As it is, we obse

e tliat there is no problem in transferring one di,sc.,

\-.

i.

If we have to transfer two discs, we transfer one; leaving

peg free for

.

:the second disci we then transfer the second, disc and cover with he first.
_._

F
,.......

Is

A 3 -1

If we have to transfer three discs, we transfer the top two, as above.

Tbis leaves a peg for the third disc to whiOh it is the, molted, and the first
a

to discs are then transferred to cover the third disc.

The pattern has now emerged.

can transfer
leaving the

k + 1

discs, we

If we know how to'transfer. k

in the following way.

Firsts we transfer

discs

(k + 1)-th disc free to move to a new peg; we-move the (k + 1)-th

disc and then transfer the

We see then that it

discs again to cover it.

is possible,to move any- number of graduated.discs from one peg tg,another with-

out violating the rules.(1) and (2), since knowing how tomove one disc,
have a.rule which tells us hol, to transfer two, and then how to transfer, three,
and so on.

To determine the smallest number of moves it takes to transfer a pile-of

.

discs, we o'fterve that no disc can be moved unless all the discs above it have

Let us designate by

been transferreA leavinga free .peg to which to move it.

mk the minimum number of moves needed to transfer

k,

discs.

To move the

..':

(k + 1)-th disc, we first need

mk

moves to t

sfer the,discs above it to

anoth er peg.
To move the

After that we can transfer the

+ 1)-th dtscr,to the free peg.

(k + 2)-th disc (or.to conclude the game if the

is last) we%must now cover the

(k + 1)-th disc with,the'pAceding

.

this transfer of the

(k,+ 1)-th disc

discs cannot be accomplished ,in less than

We see then that the minimum number of moves for

nik+.1

k + 1

discs;

discs is

mk

moves.

2111k + *1 .

Thins is a recursive expression for the minimum number of moves,rthIltis,

if the minimum is known for a certain number of'discs, we can calculate the
minimum for one more dipc.

In this wa.4, we have defined the minimum number of

sequential moves: by adding one dia we-increase the necessary number of moves
to one more than twice the preceding number.

It'takes one move to Tye* one

disc, therefore itytakes three moves to move two discs, and=swtn.

Let us make a littletable (Table A311a)..

1/4

A3-1
Table A3-la
k

mks

15

31

6'

127

i,3'

r-

k = number of discs
mk = minimUm number of moves

Upon adding a disc we yooug.hly double the number of moves.

to compare tha number of* moves with the powers of two:

and we guess that

32, 64, 128, ...;

some valA k,

k
1/11; = 2

If this is true for

'

ank +

k+1

22k

k+1

= 2
= 2

and this is the value of

true for

1, 2, 4, S, 16,

we can, easily see that it must 1)e trE, for th, next, for we

have

for mk

- 1.

This leads us,

it valid when
2, 3, 4,

for

2n - 1,

k = 1,

k +1

- 1) - 1
- 2 +
- 1,

n = k + 1.

161know that the formula

but now we can prove in se4uence that it is

and so on.

According to persistent rdMr, t4re is a puzzle of this kind in a most

holy monastery hidden deep,in the Himalayas.
.

The puzzle consists of

64

discs of pure beatign gold and the pegs are diamond needles.

The story relates

that the game.of transferring the discs has been played night'and day by the
monks since the beginning of the world, and had yet to be concluded.
gas been said that when 'the
will come't,o an end.

64

It

discs are completely transferred, the4w

The physicists say the earth is about four billion

old, give or take a billion or two.

id

ears.

Assuming that the monks move one disc.

very second and play in the minimum number, t moves, is there anyecause for
anic?

(Cf. Pall, W. W., Mathematical Recre tions.

New York:' Macmillan Co.;

1

947; P. 303 ff.)

The.principle of sequehtial_proof, stated explicitly, is this (First

Principle of Mathematical Induction):
of assertions, and let

Let 'A

A
2'

-A

... ',

be a sequence
'

be the hypothesis'that all of thede are true.

hypothesis H will be accepted as proved if

1.

There fs aveneral proof-to show that if any assert171'Ak

then the next assertion Ak+1

A
-7-

I,

true;
.rue;

The

0
is true,

is true.

There is a-special proof to show that *.S

2,

If there are only a finite number of assertions n-thesequence, say ten,

itly to have a com-

then we need only carry-mit the chain of ten proofs expl

plete proof.

firk

If the assertions continue in sequence endlessly, as in Example

:1, then we cannot possibly verify directly every link in the chain of proof.
It is.gust for this reason--in effect that we can handle an infinite chain of
proof without specifically examining every link--that the concept of sequential
It is, in fact, at the heart of the logical deyel-

,pro1)f:becomes so valuable

opmeq of. mathematics.

4104

Through an unfortunate association of concepts this method of sequential

proof has been named7Mathematical induction."' Induction, in its common

1

English sense, is the guessing of general' propositions from a, number of

...

"Mathe-

This is the way one arrives at assertions to prove.

observed facts.

matical induction" is actually a method of deductiop or proof and not a proce1.

dure,of guessing, although to use it'ile-ordinarily must 'have some guess to

Thrs usage has been in the language for a long time, and we would:gain

test.

It us'keep it then, and remember that mathemati-

'Inothing by changing it now.

cal usage is special and often does not resemble in any respedt the usage of
common English.
A3...1a., above, the assertion

+ 3 + 5 +

is

An'

(211 - 1)

2
.

odd

k2)

tubers is

if the-eUm of the fiTst

is true (G} at

We--provedZ i-rstrthat if__Ak

1;+ 1

then Al isrtrde, -so that the sum of the first

Second, We'obsei'Ved that:di

f
These two steps complete the_proof.

is true:

(k + 1)2.

odd ru
numbers is

:";....

.'

'I = 12.`

,,,--

`4116

w, a method of proving a hypothesis about a list

Vathematical inductiois

...

or sequence of aspertions. -Unfortunately it ddesii't ..ell' us how to make the

hypothesis inthe.first place.

In the example just considered, it was easy to

,,

-$1...,

xi- pdd numblrs

guest from a few specific instances that Ahe Sum of the figt
.

'

is

---(

but the next problem (Example Af,;1g- Luay not be Sofoblriods.

4.:

%.4

EXamplL A3 -lg.
integers,

positive

es of the first .,.n

Consider the maimfok t1 1'S

...
,:.1.e

1.

M';, Ai

1
.

+ 2

+ 32 +

.s..,0'''
A.:

:..,,

,,

-%

------

We find that when

n = 1, the sum is

1;
..

684

tk..

...>

whent n144:2,- the sum is .55L

!..t:

when

1.

A3-1
.

the sum is

n = 3,

14; ''and so on.

.Let us make a table of the first few

values (Table A-1b).

Table A3 1b

sum

5-

14

30

55

91

140

204

Though somb mathematicians might be immediately able to see a forAa

. that will give us the sum, must of us would have to admit that the situation
is obscure.

Wt must look around for some trick to help us discover the pat.

tern which is surely there; what we do will therefore be a personal, individual matter.

It i

a astake to-think that only one approach is possible.

Sometimes exferience is aiiseful guide.

similar problems?

Do we,know the solutions to any

Well, we have here the sum of a sequence, and Example A3a

also dealt with the sum'of a sequence: the sum of the first
is

2
.

Consider the sum of the first

odd numbers

integers themselves (not their

\1

squares)--what is
1 + 2 + 3 +

+ n?

This seems to be a related problem,.and we can solve it witll ease.

5

form,an arithmetic progression in which the first term is

difference isalso

1;

So we have

and,the common

t' he sum, by the usual formula, is therefore

2
'

The terms

+ 1) F

2.

n.4

1 +.3 + 5+ ... + (2n.- 1) = n2

Ji

= 2n

1 + 2 + 3 +

2
+

n;
t

Is there any Nit-tern here which' might kp with our present problem?

..

---....4N

These two formulas have

one
common feature:- both are quadratic poly.
.
noMials in

n.

Might not the formula we want here also be. a polynomial?

It

ill

seems unlikely that a quadratic polynomial could do the'iob in this more

complicated.iiroblem, but how about one of higher degree?

Let's try a cubic:

....

..

.., :assume that there is'a formula,

*
.., 4

2
1

+ 2

+ ... + n

= an 3 + bn

2
,,-4- cn + d,

.s:

110

A3-3.

where

a, b, c,

Substituting

4 -successively in this formula, we get

and

n = 1, 2, 3,

are numbers yet to be determined.

aYld

= a + b + c +,d

1
1

+ 2

+ 2

. 8a +

+ 3

'32

12+ 2 2

+ 2c + d

= 27a + 9b + 3c + d

+ 4

64a + 16b ;- 4c+ d.

'Solvingwe find
a =

c =

= v.

We therefore conjecture lat

12

22

+ n

1
= T

1\tn

+ .1./ an

4
A

This then is our assertion

We have

Ak:

Add (k + 1)

now let us prove it.

k2 =
+ k

22 +
+ 2

k(k + 1)(2k 4. 1).

to both sides, factor, and,simplify:

.+ 2

+ ..: + k

1
+ (k + 1) 2 = 7 kkk-+ 1)(2k + lY + (k + 1)2
u

= (k + l)[a k(2k.+ 1) + (k + 1)]

=

1
Tk

+ 1)(k + 2)(2k + 3).,

and this last equation is just .Ak4.1,

.-

Mbregver,

Al,.

which states
.

is true:

....)0,

A,
"

As true; anc .A

which is therefore true if

(1)(2)(3)

n.

is therefore true for each positive, integer

n

There is another formulation of the principle of mathematical induction

This form involv s the assumption in the sequeh-.

which is extremely' useful.

tial step that every assertion up to a.cert :4. n point is true, rather than just
t

.\

.127

A3-1

-.

the one assertion immediately preceding.

Spepifically, we have the following

'(Second Principle of Mathematical Induction):

Again let

be a sequence of assertions, and let

are true.
1.

'

3/

:..
'

11"'6e the hypothesis that all of`these

The hypotAis H will be acected as proved if

There is a general pr.00f to sh w that if ew.pry prtceding assertion

Alt.,

l'

2.

2'

l'

There i

is true, then the next asrtion Ak

a special proof to show- that

Al

is true.

is true.

It is not hard to show that eith4r one pf the two principles of main4

matica2 induction can be' derived fro. the other.

The demonstration of this is

0

left as an.exercise.

The value of this second principle of'mathematical induction is that it

permits the treatmen

of manyproblems whicli would be ',I-ite difficult to

'handle,directly;on the basis of,the first prinEiple

SuCh problems usually

,

present a pre` complicated appearance than the kin

/-", an attack by the first principle.

which yi eld directly to

y

.
6
,

Example A3-1h.

Every nonempty set

of natural numbers (whether finite

S-

oCinfinite) ,contains a,least element.

Prbof.' The induction is bgsed on the fact tgat .S
number.

The'assertion Ak,,is that if

is in

least element. -.

the assertion

Al

ins' that if

contains

contains a least number.

'

contains a

.4

Initial Step:

contains some natural

then

S,

1,

then it

),

This 1.4,6ertainly true, since

l' Is the smalleli

r,

natural number and so is smaller than any other member of

Sequential Step:

and including

k.

We assume

Now lei ,S

An

S.

is true for all natural numbers tip to

be a set.&ntaining

k t r. *herb are two

possibilities:.
1.

contains a natural number

less than or equal to

k.

less than

It follows that

k + 1,. Ih that case.

q

is

contains a least.element.
?

2.

contains no natural num

least%

r less than

k + 1.

o, .

This example is 'valuable because it is a t.lair

.,

In-that cav, k + 1

induction equivalent to the other two,

..

An amusink example of a "proof" by this princi le `is given by $eckenbach .in

"
the American IMathematical MonthiSt,
ol. 52; 1945 .

,.

of an obvious one to be sure. ,

al!

'

'inciple of math tilatical

ir

687

283

414

Every.naturalnumber is interesting:'

THEOREM.

Argument. Consider the set

of all uninteresting natural numbers.

*7

This set contains a least element. Ahat an interesting number, the smallest ,
contains an interesting number

So

in the'set of uninteresting numbers:

(Contradiction,)

ftet all.

troutle with this "proof%of bourse is th'at we.have no definition of

"interesting"' cine'mans interest

is another

man's boredom..
.

A One of the most impor*t usesof mathematical induction is in definition

/

by recursion, that is, in defining a sequence of things as f011eows:

..

a defini-

tion is eivemfor the initial object Of the sequence, and a rule is supplied
A
7

so that if any term is known the rule provides a definition for the succeeding
one. '

,11.
.

\',

following way:.

(a / 0. recur

For exal le, we colild.have defined

I.
'0

Initial Step:

= 2

.r

ak+1 . a

Sequential Step

k
a

.(k

. 0, 1, 2, 3 , ...)

Here is, another usefuldefinition by recursion: 'let

n

product of the first

positive integers.

n:

We can define

n:

denote the
recursively")

*,

as follows:
1

Initial Step:

1: = 1
.

Sequential

(k + 1 ) :

(k +.1)(ki) Ck = 1, 2 , 3, ...1

Such definitions are convenient in proofs by mathematical induction.

Here is an example which involves the two definitions we la1337e'just given.
.

000.,00n

For'all positive integral: Vanes

Example A3-1i.

proof by matheMaiical induction is direct.

Initial Step:

Sequential Step:

The

t '

.
...

2n-I <

We flaile the following steps.

,o,

2 0 = 1 < 11 = 1

n,

S
Assu ming that the assertion is true at the k-tli step,

we seek to prove it for the

(k +"1)-th step.

By definition, we have
.

4k + 1): =
Prom the hypothesis,

> 2k-1,

(k + 1)(k:).

and consequently;

t
k-1 > 2
(k.+ 1): = (k 4.- 1)(M) > (k + 1\2
/

688

299

2k-1'. 2k

%.

A3-1
.

since

Ii > 1

is epositive integer).

(k

We conclude that

(k + 1)! > 2k.

e.

lahe proof is complete.

,-

L.

Y.Before ye conclude these remarks on mathematical.inductioft, a word o

caution.

For.a complete proof by mathematical inductionit.is important'to

show;he truthof both the initial step and.the sequent141 step of the induCtion principle tieing used.

There are many examples'of mathemgtical induction

gone 'aywire because one of these steps fails. liere are two examples.

0-

Example A3-1j.

Assertion:

All natural numbers are even-,

Argument: .Fox the proof we utilize the second principle of mathematical

induction and te.ke for

A_

the assertion that all natural numbers less than

or equal to

are even.

Now con

be any natural number with

i < kJ

er the natura4 number k + 1.

The pumber

,1

N.

'can easily be shown to be a natural number with

j < k,

bath

and

are.even; and hence

such that

i,

i + j = k +1

i <k and

But if

j < k.

k + 1 = i

Let

j,

the sump of two

evennuMbers,'and must itself be-evenr.

Find 9e hole in this argument.

t
o

Example A3-1k.

Assertion:
Argument:

a _and

All, girls are the same.*

Given girls designated by

I
a

are the samet,,Consider any set .S1

Clearly, if

and

b,, let

mean that

containing just one girl,

a and b denote girls in Sl,'then -a = b.

true for any set of

a = b

k girls that they are all the same-

Now suppose it is
Let

be a set

k+1.

-containing k + 1,
girls,

gl, g2,

the

girls

girls garg2,
gk,

g2, g3,

By hypothesis the

are all the same, but by the.same arguffient_SO are

gk, gk.1.1.

It follows that

1314 g2 =

gA = gk+1' We conclude that all girls of asset containing any positive,

integral number of them are the same.
Since thereis only a positive integral
,
nuMbet'of girls in the whole world, the assertion is proved.
_

Find the flaw in this argument.

We
original
that 411
to write

are not trying to express an overly blase/attitude abbut girls.

The
of tins eXi\mple.(aitrfbuted to the famous' logician Tarski) had it 1000,,
positive integers are the same; however, isn't it more interesting
about girlsT

6829Q

A3 -1
Exercises'A3-/

kr-

+ n =, n(n..+ 1).

1 + <2 + 3 +

1. ,g'rove by mathematical inductign'that

By mathematical iriduction prove the familiar result, er ing th

2.\2.

an arihmetic,progreSsion to. n

terms!

(a + d) + (a + 2d)

[2a + (n :. 1)d]

.
,

,.

... + (a + (n - 1)d) =

4-,

sum of

By mathematical iliduction,prove the familiar result, giving the sulk of a

3.'

geo,metric

progression to

a + ar + ar

terms:

+ ... + ar

n-1

a(r n

1)

- 1

Prove the following four statements by mathemayical induction.

12+ 5? )4-, .5-. +,

-4%

+ (2n. - 1)2 = 12: (4n3

. .

n),

2n <,2n

5.

6.'

If p > -1,

7.

1 + 2

(1 + p)n

then, for every positiAre integer 'n,

2+

+ n

22 t

(n -

2n-1- = 1 4-

1)2n

ing by the second principle of mathdmatical induction.

Prove the foil

ue

the number

For all natural-numgers :n7

8.

n + 1

either is a'prime or can

be factored into pritAt.

'.

Fpr each natural number

.9.

..,....

.,,,,

41

greater thanone,. let

n'

with the propep that for at least one pair,of natural numbers
r

with
.

D 4- 21 =

, U

nn

When

= U

..

U
p .1q

for all

na

is some given real number.

we define-11i =At where' a

n = 1,

16,;- q

...

-f

be a real number

n.'

,
,

,104

Attempt to prove 8 and 9 fiom'the first principle to see whatdifficulties.

.

arise.
.

Inithe next three problems, first disover a formula for the sum, and then.

.,

1.'

prove by mathematical induction that you are correct.

4,

11.
.

'1

1
1
---- + ----- +
3

I
.

1
'.3 ..

1
n(rr+ 1)
.

sa,

690

29I

..

..;..

A3-1
,

1?

13 + 23 + 33

+ .4:

n3.

(Hint:, Compare the sums yoU get berekwith

Examples A3-36 and A3-1g in ,,the text, or, alternatively,. assume, that the

required result is a polynomial of degree. 4.)

43.

2 + 2 . ,3 + 3

Example A3 -lg
1

14

in

4 + ... + n(n + 1).

(Hint:

Comp-ire this with

the
th t ex t )
.

Prove for all positive integers

I.

n,

(1 +

+ 175)(1 +

2n + 1)
2

;.(3.
'

'

In

15.

Prove that

2x

(1 + x)(1 + x )(1 + x

4)

)
'

1 - x2
I - x

+1..

.4

16.

Prove that

17.

Any infinite straight lihe separates, the 'plane into two parts; two

n(n

+ 5)

is divisible by

for all integral '7i.

intersecting Atraight lines separate ,the plane into four parts; and.

three non-concurrent lines, of which no two are parallel, separaVp'the

plane into seven parts.' Determine the number of parts into whic'h 'the

plane is separated by

n . straight lines of which no three meet in a

single common point), and no two .e.e parallel;

.

If concurrence is permitted?

then prove your result.

hen parallelism is permittdd?

Al are perniitted?*

If bcr

Consider the sequence of fractions

1

g i

j7

I '

'

Pn

17
12 '

'

qn

44

where each fractioni is obtained from the preceding by the rule

pn = pn-1r + 2q

gn

n-1

Pn-1

Pt
* Show that for
1/..

sufficiently large, the difference between

can be mat as small As desired.

and
n
1.
Show also'that the.. approximation

to I is improved at each successive stage of the sequence and th.t

the error alternates in sign.

Prove 'also that

and

Can -Tar-obtain a more general result

18.

'

tively prime, that it, the fraction --

is in lowest terms.

gPn
n

691

2 (,-)2

are rela, '

..CL:tq(n)"denote thg sum

be anypoIynomial'of degree'

'19. 'Let p

9*

(1)

p(3) ;

q(n) i p(l)

+ p(p),

m + 1

of degree

Prove that there is. a polyn,mal' q

satisfying'(1).

p,,,
/201.

Let the function

Initial Step:

Sequential Stec.

In part

SimOarl

f(

/Initial Step:

Sequential Step

efined recursively as follows:

= 3
f(n)

f(n + 1) = 3

lar, we have
g(n)

be

f(3) = 333

27

is defined by

g(1) = 9
,g(n.4- 1) = 96(n)*

Find the minima. value m for each

21.

etc.

-J.

Prove,pr',all naturaloumbers

that

n,

(1
n

n-1

(Hint:

Try, to expjess x

n-1
n-2
n-i
-y,x,y,
etc:)

''

/
1.: /

-J

iP

i30

(1
n r2 y5

,/.5)n

,
,

Ivan ihteger.

f(m) > g(n).

such that

- y

/5)n

in terms of

- A3- 2.. 'Sums and Sum Notation

(i)/ SUnt Notation

In,t he preceding section we made frequent use of extended.sums in which

'

..

(1)

1 .1 + ,,g3

5 +. .....+n(2n - 1).

For example, consider the stem

the terms exhibit a repetitive structure.

).

.
.

a4
4c We gTopt a,concise notation which indicates the repetition instead of spelling
.
\.it Girt... In this notation the sum (1) is written
'

.,.

n
:

4,1] k(2k - 1).

k=1
)1

where' k

'Tilts symbol means, "the sum of all terms, df the form k(2k - 1)

takes on the integer values from

n. inclusive."

The. Greek/capital ,.,_.

"Z" (sigma) corresponds to the iioman., *"S'ir/and is intended to suggest the word

"sum:"
Ilt

14

d an

The notation

be used more generally to express the sum of any quanti-

4,

ties 0

where

_takes on consecutive integral 'values; we may begin with

m and end with any integer n

-----

any. integer

2 0k = 0 m + 0m+1 + 0m+2
p

+0n

./

n >m. Thus

k-;--re

/'

where

(Note the trivial special cage,

n = m,

a "sue of one te.rm: E 0k = Om.)

11
k=m
..,,

. EXample'A3 -2a:
height
,

hk
-

If each of the regions

and width

wk,

R.

in (1) i
is a rectangle With

the sum of the areas may we written

n

..

...a.

w h. +wh
+wh3 3+ ... +wh'
= 2:,w'kh.".
x
n n
2 2
. 1 1

k=1

.Here are other typical examples:

3*
k

%._

'1

'

".

.
z,
1 +'
k=0

0,..

'1
2
_ 1+0
..i.
.1+1+3.--7-7
..,,0

-,

1.1-9
et
it

-/

=5;1

693

,i
.

4'44.

,
#

E,j+3, a 5 + 6 + 7 + 8 = 26.
.j=.-2

,combination 'of n functions :

7n

Dj A'

',

j=1

/.

)1"

j Cx) . = a1f 1-(x) + a2f. (x) + ... ,t

A polynomial of degree no greater than

.m.
..E
e x

c
,*

411

'

=e 0

1=0

+,1

x+ex
2
v.

..

'',..1.

m:

./4 e
1

...

+exm

111--

Example A3.-2b.' A simple but important sum is

:n

,quantity

is the sum of

'# "

c,
E
j=1

constant, that is, a quantity independent of the index

,

-----..":

..

f (x),

n n

is 'a

where ,c

of summation.

The

terms'eadh orwhich is c; it theregoe

J =1)

\ has the value nc.

In any summation/the values of the terms and'the total

by the choice of the index letter; thus.

ale

not affected

'II'

j=m

k=m :,.

i
:conenienee.
Example
___,--

1.

A3 -2c.
,

(a)-

E ai.,..a0

/'
+ al + a 2' =

Eap -1=E a2-'n

j=0 '

n=0

1)=1
.

(b)

= aP
E it..n.-i
i

+ 'an 0

ain-1

E (an.-I
j=0

i=0

Summation. is .41 linear

bftlow.

..

t` suit our own

the index letter'aid Its initial valu
-r /

We are free to- choose

..

ir

process; the ...

proof is left as. the first excercise
a*

-.

;..,/

694
)1

h. 0

'

A72
t\
Exgrcises A3-2a

11

(afk
.

+ Pgik). =

Write each

g,

k=1
o

'(a) .E m(m - 1)(m

2k
=1

m=2

10
* (t)

..,

thp following' bums in expanded .Form and evaluate:

'5

(a)

f->k

.k=1

k=1.

2.

..

10

1E 32

(e)

2i

i=0

'1=5

-,4

3,

+2- -.12)

.(r2

(f)

r:(4 - r);

r=0

tbefb

are true and which are false?

rig st ateme

usions.
10

='7

(a).

28
a

.=3
14((ri

- m) + 1

i=m

io

(c). E.k2.= '10

k=1

r=3.

1000.

(d)

.1000
k2

=5+

k=1

..

(e

, E k3

rt

n3

k=1

.., 1)3

j=-2

10

10

trr=1

10

(g)

10.

k3

(E
m=1

is.

.695

''
296

Justify

i
,

,.

n-1

(h)

E icr- i)in - i) ...E ici-

ti..

'(----\.:.

...Ix. _

1=2.

mI

(0 -E f(am_k) . E f(.,)
k=o
%

(t)

k=0
,

,P

--p

E E
.

Ak

;- `

ki,An_k

(Ak - A '

nk ).=

k=0

,/IE A
2

In

m-k

- 2m

,.

k A'

m-k

k=0 ,
.

2:14 f(kN(b - a)
'ni- n

Evaluate.

k=0

'

,-,

4.

--'--.

,,,k=0).
,/

,,

k Ak =

k=0

.m
(k)

,n

k=0

if - f(x) = x2, a F. 0, Jo = 1,

'and

k=1
(a)
,(b)

'n = 2
n

(c)Nn = 8
into

Subdivide the interval 10,1]

.5

equal parts.

interval obtain upper and lower bounds for

In each sub.'.

Using signa notation

use these upper and lower bounds to obtain expressions fdr upper and
_lower estimates of the area under .the curve

g = x2

[0,11:

on

If
.

you can evaluate these sunis without ,reading 1R1sewliere, do so..

(b.)

Write out the sum of the first

with first term

terms of an arithmetic progressiOn

and coupon. dicference

d.

Express the same sum

in sigma notation.
(b)

In sigma notation, 4rite the expression for the sum of the first
terms of a, geometric progression with first term

ratio

7..

(a)

and common

r.

consider a fundtion f defined. by

f(n) =

f(n)

Find
(b)

t(r - 1)(r - .2)(r -

for

n =

Give an example of

-3)(; -

(r

-,5) + r

2,

function

(similar -to that

it

(a)) such that

.

g(n) = 1
g(),,0

n = 12
O.

',I

696

291

8...Write each o;Ithe ,following sums in expanded fbrm and evaluate.

4

EE

(a)

rr(n -

n=1 :r=1

EE

,(b)

n=1

(rn

r=1.

n
9. ,The double sl'aml

is a shorthand notation for

0 j = 0

E,.{F(

0) + F(

, 1 )

;In)

or
F(0,0) + F0,1) s+

+ F(0,n)

(+F(1,0).:4- F(1,1) +

+ F(m,0) + F.(m,l) + ... + F(m,

....

EE'

In particular

. j

!-.

1 + 1

2 + 1

i=1 ,)=1

3 +

'2

1 + 2

'.

,
,
_..

+ 2

3 =.18.
m,

Evaluate:

ECE: k

* (a)

'

i=1

A.

EE

-(b)

k(c) EEmEi.(i,J)

4. =1 j=1.

j)
(d)

1=1' j=1

i=1 j=1

..

10.

Show thitt

(a)

k(k - 1)

k = 1

it

0, 1.

'

ib

'

ip

Evaluate.

(b)

1000

k(k - 1)

AL,/

N7 -'

4
l

k=2
.

.,
$

11,,

If ,S(n) = IAE:. f(1),

determine

f(m) . in ;berms of the sum functidn '

.

. i= 1-

, '14

97

,: . .
Determine

12.

ef(m)

iiiithe following summation formulae:

-..

(a) 1

Ef(il

.7

(e)

1=1.

(b)

n =

1=1

,.

nx=Ef.(i)

cos

f(i)

sin (an + b,1

=E

.,

tc)

=Ef(i)
-

=.1:

nl

(g)

f(i)

1=1

i=1

bn

ant

(d)

,,

fo..),..

1=1

i=1

'

(f)

1'

E-f(i)

c =

1=1
I

13;

We define

Binomial Theorem:

(111.)

h!

,.

Also

sush tliat 0 < r < n,

where

- r):n!

and

0! = 1

(nr)

r,

are integers

-n

.
}.._

= 0

if

r > n.

Show that

n
(a)

(on) k (i;) = 1

(b)

(111,)

= rn

r)-

01\

(n

;.)

1-1

`r + 11

(c)- istablish the Binomial Theorem

,.,n

(x + f)n =

E(111.)xil-1.
,r=0

yr

= x +

n-1

n-1'

n = 0, 1, 2, ..:,' by mathematical i auction.

14.

Using the Binomial Theorem,, giye the expansions for the following?.

(a)

15.

(3c

Y)3

(x

y)3

(c).

(2x
(x,

'3y)3

2y)5

Evaluate the following swig.

,

'(a)

1 triN

0'4- (n)
4
1

...

'

`ni

=E (11\
`zi-

r=p
4

n
2
4.

(..1)n

,,

=,(..1)r (n)

(b)

r=6

698'

29Q

.
,

...
,..,

.,40r

-.

r=0

-:--

using 15(a).
Rrji)

for

lin

(n

..:, n

..''

denotes ajolynomial of degree

k . 0, 1, 2,

flind

and'

r=0

r.,0

'"'

178

':E:(111,) 4

16..
aim E'r(111,1 .by.firstshowing
.
--.:./

...

(n

such that :-Pr;(x)

2x

1).

S..11Plination

Exercises A3,.1, No. ld illustrates a particularly useful Summation tech-

nique,'i.e., representation as a telescoping sum.

It was possible to write

1000
1
k(k -

k=2 /

1) - T'7.

in the form

2 + 7:7 4.

+1000

1'
999

1000
1.

12

1)
1

1)

(1

k'

,4

(.1:

`2

2'
,

'1

3'

(75g -,,999'

`9g9

1000)'

Each quantity subtracted in one parenthesis is,added baNk in the next, so that
,p,

thef*st twoterms telescope from'a,tum Of four numbers to a sum of two num-,

.1;64.). the first' three terms telescope from a sum of six numbers to a sumof

two mulOers, etc.

Finally,' the entire summation telescopes (or collapses)

1

into ,:a14n.nrof two

number in the last term.

n
44.%

(1)

...,t

',

numbersthe first number in the 'first term and the second.

Syablically, atelescopibg_sumhas the form

,
.

f(k) - flk -

1))

f(n)

f(m - 1).

k =m

In the above example, we have, m.= 2,

n = 1000,

sum telescopes to

E05

f(1000) - f(<L)

and

+ 1 = ?g

- i

f(k)

/generality.

f(k) = k,

f(k).

AlsO, we let

then
n

(2)

E1

k=1

k=1

'

695

Ott

thq

Let

that.

:.

We now use (1) "to-establish %short dictionary o1 summation

donsidering different functions

so

m q 1

formulaeby

without loss Of

f(k) = k2,

let

No

7!rhie result is nothing rum.

talen

n
Vk2I- (k

1)2

E'(2k

- I)

k=1

k=1

n,

k=1-

k=1

= n2

..

; or, equiyalently,

I
'

(3)

k = -t-i.,(n + 1) .

-.

k=1,

general arithmetic

By liklearly combining (2)/ and (3), we obtain the sum of

progression-

'

1:1

ni'n(n + 1)1
2

..

on .

'

k=1

-,,,4
,

,'''n

...

To obtain'the sum
i

2,

.,

- we let. f(k) = kJ

Then,

.,

.
1

k=1

:.

--,..;'''
'

3k2 + 3k 4:1)
=2.

k=1
s

.n
,

k2; -

k=1

3Ek+El=n3.
k=1

k-.4.

Using (2) and (3), ide,obtain

3n(n + 1)

{n3
3

n(n

1)-(2a+ 1)
6
;

k=1

We now man establish a sequential method of obtaining sums of the form

/

P(k)

Whose terms are values

P(k) .of a polynomial function. _Because a

polynomial

a linear combination of poWers

and summation is a linear process,

it'is sufficient to give a sequential method for

1: kr,

a n#nregative

'k=1
.integer.

Choosing

,r4

r+1

f(k) = k

tiff,' formula (1) gives us

in

Dkr+1

)r+1}
=

k=1

Using the Binomial Theorlm, we 'obtain

kr+1

(x+

(k

700

+ P(k)

A3-2
b

where

is a polynomial of degree .r

B(1y)

Thus, the suis

1.

can

be expressed in terms of sums 6f lower aegree.

for

r = 0, 1,

sum for any

and

Since we-already hare the sum

we can repeat the method sequentially,o Otain the

2,

r (compare with Exercises, A3-17, No. 19).

vt.!,

We _can enlarge our summation table by choosing other functional forMs

f(k),

d.g.,

siif(ak + b).

n'

By (1),

11: {sin(ak + b) - sin(a(k

(5)

1)+1=

sin(an + b)

sin b..
to

k=1

'Using the identity

--,

sin A- sin B= 2 sin

cos

A'

+ B
2

'

: in Equation 6y, we obtain

sin

Ecos(ak+ b

6)

= 6os(b + "")
sin

k=1
-a

If' b

(6)

2'

an
2

a
2

reduces to
all

'CO
If

(8)

cos ak = cos

'

((a.

+ 1) n) s'in"

k=1

sin

T
a
-2-

(6)reduces to

b = a +

an
sin-1T

sin ak = sin(b + an)

2

a
sin in

ByichOsing,other functions
:formulae.

f(k),,

we can enlarge our list of.summation

We leave this for exercises.

-

"

,eN

0
o

y
/4

(
r,77`.

1,.7

tb 7

701

...

Exercises A3-2b
,

Write the following 'sums in telesCoping form,

I
116

1.
'

14

Eiu(k) , u(k r.-:1..

i:le.,

in the form

and evaluate

'n

(a) . Dk(k

+ 1)1

,(e)

E k3'
k=1

lord

n
k(2k

(b)

(e)

k k
Ea
k=1

1)

----N

2k(2k + 1,

(g)

-,,

k:

I.

k=1

k=1
'

+ 3..)(k'+ 2).

'

...

.--....,

4, -..,,,,

..,,,

k=1

k=l

.
n

.
4

Using

2.

Eiu(k)

- u(.

'''.------.--",-.......

....,,)}.

u(n) =:u(0),

estatlisli a short dictionary

k=1

..

of summation formulae .by considering the'forlOwing functions

(.a.)

(a + kd)(*a + (k + 1)(1)

(b)

The reciprocal of

(C)

rk

(d).

krk

rat)

kirk

+ (k

u:

li,p)41)

N.

lr

'

(f)

p
3:

2
(k.)

(h)

arctan k

(i)

k sin k

1.

Simplify:

- 1)x

sin 3x +
clot x + cos 3x,+

sin x

cos t(ink.

a.-)xj

;s

702
I

at

ow

ti

"'".''c-.

-,

4,

Another methbd for summing,

Z P(k)

(.? ;a polynomial) can be obtained by

N.

using a special case of problem 2a, i.e.,.

n

1.

1
I

E/(,

+ 1)1(k)(k - 1)

...

(k - r + 1)- '(k)(k - 1)(k..-

7,,:./(k.- 0)

,,,,'

It-.-.1

-"'

(n _

= (n.+ 1)(n)(n - 1)

d.."'

1),

.."

t
,

n
or 2: k(k - 1)

419

An,+ 1)(

1) -

(k -

(n - r + 1)

1)

it + 1

k=1
First, we show how to represent,any polynomial

P(k)

of

th

egree in

the form
a k(k

C(k.- 1)
(i)

P(k) = a0 + a

than .a0

If .k =

It-e-r than

2:

-+

(k - r + 1)

1)

,,
rt.

k = 1,

P(0) ;
al = P(1)
if
/
= P(2),7 2P(1) +'P(0) '''In general, it/can be shown that
.-...

P(0) ;

if

then

(ii)

p(m - 2) - ... 4-(71)mP(0),

ain = P(m)- (11)P(m - 1) +

= 0, 1,

r.

e polynomials of degree

Since both sides of (i)

for" m= 0,

it must be an identity.

ANow sum

(k).

1
5.

Using Prob.

'nethe following sums:

n
(a),

_Ek
101
n

o.

(a)

Establish Equation (ii) of Number 4.

(b)

Show that

,.is zero-for i> r.

.

703

4,..

304-

and (i) is satisfied

,.

'Appendix 4

FURTHER 'TECHNIQUES -OF INTEGRATION

A4-1.- SubstltutiOnsoof Circular FunctiodK

'

Although it is not always possible to, integrate a given function

terms of elementary functions, thereare important broad classet of explicitly

integrable functions.

All powers- and,hende),...'clearly; all polynomials are

explicitly integrable.

It.is not so clear but it is true that alf rational

functions-are explicitly integrable (see SeCtion A4 -3).

It follows that all

integials which can be transformed by substitution into integrals of rational

In this section we shall show that an

functions -are explicitly integrable.

integral of any rational combination of

.

there

0'1

Q( x) = Ax2

Ex' + -C,

can be transformed into an integral of a rational combination of circular

functions;:and further that an integral of a rational cdmbination of circultr
functions can be transformed into an integral of a rational function.
400.

We should consider the substitution of a circular function Nhenever an

x

,integrand is a combination of
a .+ x2,

x2 - a2, (a > 0)

/2

and one of the expressions a - x

dx

suggestive of the Pythagoreari expression for

one of the sides of aright triangle in terms of the other two.

Consider

t:

+.1 =

'

a/2
42

We utilize the substitution

x =asine

'

-x =acos e

721)

le

dx = a,cos.0 de:

(See Figure' A4-1a.)`-Observing-that foe x

= -E21

e = g, we obtain by the

Substitution rule
Y /6

1,,

y/6

a cos e
a cos of

de =

705

'1

305

:
-

Figure' A4-la

Example A44-lb.

For the integral

dx
.1

we

2/2
0

(x2 + a )

mploy the substitution (see Figure A4-1/b)

x.
.
= a tan.e
oa

dx =

cos

2,

2
142 + x

de

cos e
e+

Thus we obtain

a
FIE; tire -.A4-lb
s.

I =

cos3

a3

'sin e
2

Example A4-lc

4.1

=2

cos e de

cos

+c=

.a + x2

The integration

= f

c2h2

dx
t3,2

./

706

36

performed with theN aid of the substitution (see Figure A4 -lc)

x

a sn
i Q.
E

dx -

= a sec e,

cos e

de = a sec' e tan e de

cos2 e

x2 - a

= a taxi e.

Figure A4 -lc
We have'

e),

j(oS

= ---j[cos

a sin e

'a tan e'

2.

sin e

de

cos 2e

2
a

1T-78

72

a2x-

,Congider the integral

Exsunple

1
-

dx .

a2

Using the substitution of EXamlile A4-1c we obtain

1

(a sift e

a tan e
'cos

) de--

cos e

de .

We can write
1
cos

cos 19

...

cos e

cos e r

cos O.
'2e

L',-

sin

1 + sin el

s i n e

With this much, as a hint we leave the integration as an exercise.

(See also

Section A4-3.)

Yen

take

0 < e <

A
2

for

x > 0
,

\707

and

<

<

for

x < 0.

MECUM A4 -la.

An integral of any rationarcombination of x

and

IQ(x) iwhere

+c

Q(x) = Ax2 + Bx

(1.)

x = f(e), where

can be -transformed by a substitution

is a

circular function, into an- integral of e rational combiliation,

'

sin B

anl cos 19...

Concerned witb4integrals.of

0( x

, 1/Q,(x,

..4a.imlnaali;ieaF transformati

the standard fo

s of Examines A4 la

form

Q(x) 4 given by (1).

0, is e rational expression and

where

the

to replace

For the proof

Q(x)

by one of

ly, c.

We "complete the squareto obtain

x +

(3)

4)2

RC

312 VI *;
4A2/J.,
tt.

b=--e=
ITT,& and
2A

We set
4A

x = u

in (3)., and

separate the pro lem in .o three oases:

daze

If A'< 0

and - - -n
\A

<0

we have

)+Jk

i4(x).,= c/87.77
Since. dk =

the'substitiltion

1,=

(4)

Now,employing the substitutfon

X=u
b ,

- b

yields

87:72)du.

u = a sin B

of Example A4-1a, We transform

the integral into the fprm

'(5)

I =a 0(a siP6 - b , c a cos 0).cogede , 0 = arcsin

x+b

gince 4 involves only rational operations, we have established the theorem

in this case.

708

A4-I
r

Case 11114.

If , A

r - ---f

and'
o

<

the . substitution-

0,

10.*

,,

x +b=u=atan 0,,
as...in, EX4mple.A4 -lb, confirms the theoremfol: this case.

Case

If A > 0

and

the substitution

33

4A2 )\ f

x + b = u =

cos 0

'

as in Examples A4-14c, yield's the desired result.

/'

'

The integral (2) can be also transformed into an. integral of a rational
combination of

sink t

where
.

and

cosh t

by an appropriate transformation ,x = T(,t)

nc ion.

1,

Theiproof isleft,as an exercise.

..

,ItEOREM,A4-1b.
and

cos x

An integral of a rational combinatfon of

function by a suitable substitution.

Proof.

*e

We consider integrals of the form

(8)

where

sin

can be .Wansformed into an integral of tl.rational

J*(sin x ,'co's x)dx

is a rational expression.

rational expressions in
,

.(9)

t = tan

We, observe that ,sin x

and . cot x

are
04.

;.namely,

4
sin x -

2t

' 2

'

, cos x =

1 + t7

1 - t

dt,

FUrthe'rmore,
N.1%

(10)

dx'= d(2 arctan t) -

1 + t
Contequently we may transform the .integral (8) into the integral of a-rational

function by employing the substitution

(1.1.)

= 2.arctan t;

709 '

9re

go/

..
in"the form

thus, entering (9).alid (10) in (8) we obtain, the

4.:

r f

(2)

42

1 + t
0

1 - t

tr

1 4

t2))41)

dt'
,

'

Theorems 10-3a and 10-3b do not necessarily Point the way to the simplest

method of integration for a functibn of one of the types considereg,

ere; they

simply indicate.,a line'of approach which is sure to work but wv lead to`
.

enormous complication,

Often some special deiice leads tp the solu.gon 'far

more ,S-imply an directly./ '.

Exercises A4-1'

./

x3,44

(h)

(c)

x2

x2)5

- x

being.positive.

x+

and

Integgate the folloiiing functions, the numbers

1.

(d)

x2 1. a2

(e)
lX +

+ ax + b
x2 :f

,12
a-) x

42x +x2
,b .
--7-2"

.1^'

(f)

4.

a'
Let 'R(x,y)

denote a rational function in

and

Reduce the

following integrals to integrals of rational fun ctions.

'

-,'

(a) 17--li(X

le; +

b)d4,

a # 0.

,,
(b)

R(x

+ b
,

riac

dx,

an integer,

710

310

7..

ad - be

O.

3.

Usinglthe result of Number 2, integrate

.40

14W.T7T + ax +
.
,

Reduce to rational.fOrp
1 +.x
1/1-7T

5.

1 + x

jbspress as elementary functions

I

(b)

(c)

r,

dx

'f
dx
cos 2x

,
4'

hC

6.

(e)

The integral

P(x1

dx,

ax2 2bx--;
degree

a A 0

and

where

P(x);. is a polynomial. of

can be reduced to a ratpnal trigonometric

form as described in the text:. It can he also reduc ed to the.,

1

integration of
Itax2

degree (n
P(x)

VL2

Q of

+ 2bat c

and.dOnstant

1)

k.

- D(Q(x)/ax2 + 2bx + c:)+

+.21)x

Show how to find

(b)

, namely for some polynomial

Aax'

,Using (a), integrate

and

t5

t3 + t
,

/37.7
(c)

Calculate tNntegial of (b) by using_trigonotetrIc substitutions,

and. compare the merits of the two methods.
1

711

7.

Integrate

(a). sin x
1

b)

cc

Wit

(by a method other than that of Example '4-1d),

F..

t.

712
C-

'c3

A4-2.

Integration

hy

Parts

Thebasic. formula.

The !'thod, Of integration by part's is used to

integrate certain kinds of prodIL"-s.

;he method corresponds to the formula

for thederivative of tt product.

THEOREM A4-2a.. If .f

and

a common interval containUg

(1)

have continuous derivdtives over

4

and b

then

b'
f.f(x)131(x)dk = (f(b)g(b)- f(a)g(a)1,-

fqx)exIdx.
.a

The'theorem follows directly from the product rule ((4) of Section 8-4),
and the Fundamental Theorem of Calculus.
0

In Leibnizian notation, for u = f(x) du = f'(x)dx

v = g(x),

and

we obthn fOI11.1!clqliAte_integralcurITFS-688ing to (1)r,_

dv = g'(x)dx

.(2)'

cu dv = uv -

Sv

du,

Integration by means of (21 is called integration IL parts.

.

.Example A4-2a.

To integrate

x -4 log

especially si'Mple,perivative a5d.set

then, we take * Nt

Vserve that

= loge

and

log

dv = 1

has an

e
dx.

For

v,

i., Consequently, from (2)

L

loge x dx = x loge x

dx
4,x

=x log e

the formula we have already obtained.

ep

In application, (2) is us d as'above for the integralOf a product where

..

" the,product of the integral o

on6 factor and the deAvative of the other is

formally integrable.

The 1,eibnizian notation in (2) was introdu6ed

explic t formula
.a fu ction of

ay

that

v,

al,

shorthand

for,

But the notation suggests that we might interpret

and

as the inverse function of

u.

the
u

as

This idea yields

luminating geometrical interpretation'of integration by parts. ,Stippose

u = f(x)

and 'v = g(x)

where f

and

713

3.'3
er

have inverses.

'Then' we can

40,)

write

u =0(v)

and

where

v = *(u)

and

(The proof ''

are inversest

Set

is left to'Exercises A4-2, Nb. 2).

v

= g(b).

1
Now suppose

We hali u1:1 0(vi)

and

u0 = f(a), u1'

and, inversely,

--2 f(b)

vi '= *(ui)

v0 . g(a)

and

i = 1, 2.

for

Then, from the

are increasing and nonnegative.

familiar interpietation of integrtl as area (see Figure A4 -2a) we immediately

hE;Ve

-r-

Figure"A4 -2a
.

,:
-......--

4,

,-/

11.

tray +
v0

u0

.t

from which we at once obtain

v du + u v
0 0'

.1

u 6. = [uivi -

Ili

uouo] -

v du.

u0

vo

..

From the Substitution Rule we immediately recognize this equation as a form

..,

of (I).

.it like geometrical argument gives,the,same result when

and

are decreasing.

in general, this interpretation of integration by parts gives the formal

'integral. of any function which has a formally integrable inverse.
4.4e

"S.

741

04.

J4

A4-4
Example A4 -2b.

Consider

S xn arcsin x

(n

Since the arcsin has a simple algebraic derivative we set

integrali n j -1).

u = arcsin

,x n+1
dv = kndx and take

Forthe domain '0 < x < /-1

n'

..= aresin

'that

124-1,/(n

+ 1)v

v =

and

n,,+ 1

n+1 .
'sin
u.

we have

From Theorem 114-lb we know

can be transformed into the integral of a rational function.

As We shall see (Section 114-4),reiional fun ctions-ve always formally integrable..
5 v,,clu

It follows that

that

sinn+1 u

Eircsin.x

xi/

is formally integrable with respect \to

is formally integrable with respect to

X.

and hence

Reduction to

thd integral of a rational function'is not necessarily the most efficient

way
to carry out theie integrations, but integration by parts can be used
more
'
effeCtively In other, ways to execute the integrations.
a
i
..
.

The idea of Emile A4 -2b1

U = f(x)dv = xr1 dx ,estalliehes the,

,for

formal integrability of xnf(x)

where

is any inverse circular

function,and in view of Example A4-2a, if

f(x) = log x.
a

-'Example A4-2c.'C,Onsidet
.

IXr

log x dx,

:1"

(r real).

_ .

Since log x

has EC simple derivative, we set u = log x,

dv = xrcix.

we take

xr+1
v =
r + 1

to obtdin
r+1

xr

_,,_, .....,...

x dx =

r + 1

log x

r + 1 Jxr dx

xr+1

-14-1-

=Nrr--+1013.1C
1

(r + 1)

2
.

If r = -1,

we may take

v =log x to obtain
x
x

dx = (log4c)

dx,

'which yield's

715

!".1k:J

315

J.

If

,..44.

-4'1---,--:.------....

...

\.2

))

fe klog x)

fil3g
x

-..-,,

dx=

+ C

,
-..

a te-Sialt which4is obtained more directly frOm the sUbstiSution `log x = t.

A 4...
------------1,,

,The- method of Example A4-2G, *Dor

u = fx)

and

the formal integrability of any faction of the form

..-

dv ='x

exhibits

x,

when

xn ffx)

n # -1,

where

is any rational combination of

ft(x)

.
.

..

and

IQ(x)

and

'Q(x)

is a

Integration by parts expresses the given integral in

quadratic polynomial.

,..70

n+, 1

terms of the integral of

,_

ft(x)

).

4'

which may be transformed

into the
.

.....

FroE the assumed-integrability

integral of 4, rational function by Theorem A4-1a.

of.rational fuuctions,the result follows.-

It follows as a Slight genera/a-

..

tion that

P(x)f(x)

is formally integrable for any polynomial function

From this argument we observe again that if

circular, function, then

h(x) = 0(xoiqx)),
x

is a logarithmit or inverse

is formally integrable.

x f(x)

a rationed combination of

P.

and.

In addition, for

l(x),

the expressions

Log h(x)

and

derivatives of

xn arctan h(x)

log

Example A4-2d.

and

arctan

and are all formally integrable since the

and are rational 'unctions.

COnsider the integral

x
jrx e

We integrate by parts.

Set

dx

u = x dv =.exdx

.1.:( ex

"dx = xe

= xe

and

- e

Then by (2)

ex dx

- : l

v.= ex.

Integration by parts may be used toITroduce & simplificatiot; rather than

a final complete integration as drk'Exanple A4-2c when

, '3 1 6

r = -1.

Example

Consider

II =

.I ebxj sin ax dx

..,

For

u = sin ax,

dv = e

bx
'

v =

dx,

we obtain

a I bx

bx

I = -

bx

sin ax - -

cos ax dx

sbx

whee
40.
O

bx

cos ax dx

resents the same difficulties of formal integration as I.

same technique, we can express

in terms of

an equation which can be solved for

-bx

v =

Now take

I.

HoWever, by the

and hopefuUy.may obtain

u = cos ax

and

in (2) to obtain

J =

1 bx
a jr bx
e
cos ax + 1-;
e
sin ax-dx
b

1
b

.-

.e = - e

bx

Yea

cos ax + - I
b

Entering the expression for

for

above in the expression Stir

, we obtain

"
*

I -

2
a

I' and solving

....'

bx ,

kb sin ax - a coq-ax).:

2i

+b

f'

.11
.*

(ii)

Recurrence relations.

The idea here is'to express an integral of

the general fcirni

f (x) dx
n

in terms of

irf

n-k

717

317

(x) dx,

Consider

Example A4-21?
,

t,

to

al

.40

Set

u = (1 - e ,

v =

dx

dv '71= x

J. xr{ 1 -

.r+1
x

r+1,

kl - x
r + 1

+.

r +a4,

'

4,

dx

r # -1) .

Then

f r+1,
r + 1

> 0

dx

kl - x)

r+1

where,.for

the result yields, Correctly,

= 0,

x
r + 1

In -

,Now, observe

_that
a.

r+1

n-1

- x)

x)n-11;_

= -x [(1 - x)n -

whence,

r+1

(1 2 x)n
+
r + 1
r + 1

This equation" may then be solveSrfor

In

[I ns-1

- I ].
n

in terms of

I__

r+1

x)n
(1
n + r + 1

n + r + 1

n-1 ,

or
r

.n

x (1 - x )

dx

k>11

r+1,

l - x
n +,r + 1

n.

r,

n 4 r + 1

Now tills formula may be applied recursively to express

in terms of I

-2

'I

n+r+ 1

x.

x kl

n-1,

in terms of

4.

n-3'

r (1

etc.4 to yield

4
n(1 - x)n-1
n + r

n(n = 1) (1 = x)n-2

(n + r)(n + r

ll

n(n - 1) ,.. 1
(n + r)(n + r = 1) 4..

iy
+'1)

Sometjmes it is necessary to prepare for int gration by parts by some

preliminary rearrandment, as we show in the foll

718

'318

7-4,

ing useful example.

Example AY2g.

Cor Eder

cos

We write. posnx ='cOsn-1 x cos x; . set

= sin x,

x dx

u'= cosn-1 x,

dv =,cosiX

to obtain

= dose

-1

x sin x +'(n

n.-

.'-2
n

cos

x sir,` x dx

, n-1

x sin x + (n - 1)

.1 -cos n-2

x (1 - cos

x)dx.

_Thus,

0
,

= cos

Solving for ,I

n:

n-1

xsin x + (n - 1)(In_

We have
'11-1

cos

x sin k
,

n - 1
n

n-?

ke-- -F.---.

Since the subscript is lowered by

at each step we observe for n

that the rec rsive.reduction of the integral terminates at

-..

dx = x,

=,f

cos x dx = sin Y.

even

n = 0 with

and for u odd, at

n =''l

with
*

Often the principle'use of a recurrence relation is not to obtain the

$
fornfii 4.ntegral in terms of elementary functions (which May not be possible)

butto obtain the original integral in terms of a simpler integral.

Example A4-2h.

Consider
' 2

I '=

""---P"

r. From u = x

e-x

n-1

I,

dv = x e

yx2 dx,

v.= 2

- xe+
1
-2

n-1 -x

dx.

we obta4n

2
(n - 1) .1 n-2 -x
dx
2

or

719j

xe

1
.:-

2 7.;

n, 1

n-1 -x

'+ 7-- In-2.

is odd, the recurrence relation 'Oyes', In

If

in terms of elementa

functions and

but

x
1
Il = - 7 e

is elementary and

If

1,

formally integrable in terms of elementary funOtions.

is

is even,

hen-

the integration of

is reduced to theintegration of

''

2.

-x

dx.
.

used\.
This integral ia,not-elementary.,.Ho eber, it. is well known and much

In terms of the error function

erf (the area undvr the normal

curve) given by

erf, x =

1
f17

2-

e
J.0

we have
I

= r-t ere

(-12-.)

numerical1
The common tables of the error function enable us to work,withlt
just as convenienti

as the circular functions.

Exerciseq A14-2
1.

Integrate the following.

arc cos x/m

(a)

x sin 3x

(3)

1/.7771771

(b)

5x

x
3

"

(.c)- x

(k)

x sin

(i)

-2x
.

x,

sin x
:---

.0
(d)

(m)

x2 arosin

,bx

(n)

cos3 2x

(o)

sin' x

(p)

sin (log ax)

arctanJ

(q)'

x tang

x arc tan x

(r)

(arcsinx)2'

(s)

sin ax cqs bx.

IC log ax

(e)
(4')

log

ri

,,s(g) rc cos 7x
(h)

.720

is

as,

t.

Support the geometrical inteiretation bf_integration by parts by

9
showing for u = f(x) and( v . g(x) where f and g

have inversesl,

that ..,uv. 0(v)

and .v = 111(u)

where ,0

and

are inverse functions.

3.

Verify as alleged after Example A4-2b that the method of, the example
.,.

does demonstrSte the reducibility of

rational finction if

xn f(x)dx

to the integral of a

is any inverse Circular function, or if

is

the. logarithmic function-r---------7'

b.

Establish'recurrence relations fOr

m and

ach of the following

(a)

sinn x dx

(b)

log
m

(c)' . i sin

xn ea4 dx

x dx

x cos

Sxn arc sin x dx

x dx.'

4,

sin

(h)

in each case

are positiventegers).

dx
x

arc tan x dx

(i)

Cos x dx

(Note the difference between

odd and n

even).

si

*.

721

321

yo

ew
?""'."

A
A4-3

Integration of Rational'Eunctions

;
The problems of. formal intfgration in the preceding sections of this
appendi,, x were often recast in the form of the problem of integrating a rational

For a rational function there always exists a formal integral in _

The formal integral is obtained by'reducing
terms of eleientary functions.
the rational function to a sum of a polynomial function and functions defined
function.

by the elementary forms

r
(x - c)n

px + q

(b > 0).

[(x - a)2 + b2ln

eitfier,frOm the
It can lie proved that such a reduction is possible,
df-a
FundamentalTheorempof Algebra which requires the theory of functions

complex variable, or directly by new algebraic tecbniques.

In either case

of thistext.
a complete proof would take us outside the frame
polynomial and
The reduction of a rational function into the sum of a
into partial fractions.
terms cf.-the form (1) and (2) is called a decomposition
We give one sirpple example?
Example A4 -3a.

A common case is give; by the rational expression

.(

- a x-

(x - a)(x - b)

x -

a!

b.

a);

From the decomposition (3) we immediately obtain the integr

1

'1

1(x

log(x - b) - log

)I)

177=E7'

- a) (x - b)

log

(i--77.4).

Let
,to put

be any rational function.

R(x)

By long division it i5 always possible

in the form

R(x) = S(x) +

Px)
(
Qtx).

where S,T, Q are polynomials and the degree of P

Q.

is less than that of

Since the polynomial. S is immediately integrable, we may omit it frpm

consideration.

It follows from the Fundamental Theorem 'Of Algebra (Appendix

2) that every polynomial

Q(x)

with Teal coefficients hase unique f8ctorizal-

tion of the form

722

A4-3

cl)ni(X ..0.,.((x2jti)24.b12;111.r/x._an2.1.b.211112L\
2,
..

`'(4) Q(x) = A (x

where the ek

are the distinct real roots of

Q,

andsa_ t ib ,
K
k

*-7

maginary roots

(b

-,.

Now suppose that

of

C,

0.

P(x)

R(

Q57
Q

the distinct
_

,..

and that P and

Q,

R(x)

> 0).

where the degree of

hae,no common factors.

P. is less than that

Then we aasert,that

is the sum of e)"Aor ssions ofwo standard forms;

for each real root

an e')cpreesion-of-t-he.form.
ra,

( 5)

x1,*
-

..

(x

.
'

where n

",' roots

the multiplicity of

a t ib

an expression of the

P2x,

(x - a)2 + b2

[(x

,form

41

pmx + qm

q2

a2

/ 0)

fareach pair of conjugate imaginary

pe + qi
(t)

(r

(X - C)n

212 +

[(x = a)2 + b?]m"

gM2./ 9)

where,

in

is theirlcommon multiplic't

without proof.

We merely use this format as a /glide

In each particular case it can be verified directly that,the

decomposition obtained is correct. Since we have obtained and verified

,

correctness of the partial fraction decomposition we have reduced thd in e.

gration problem to that of integrating the simple form (1) and (2).
-

BefOre we etbark,on the problem of integration let us see what is

involved in the algebraic probleM of obtaining the partial fraction decomposiv.

tion. The first pvieblem is to obtain -the'roots df the polynomial Q(x). In
general the roots of a polynomial cannot be obtained from the coefficients by
a formula involving only rational operations and rational powers.

There are

such formulas for the roots of polynomials of third and fourth-degreev-but----these formulas are generally,useless. - For example, thd formula-for the roots
ti.

of epolyndmiel of third degree may involve complex quantities even when all
three 400ts are real. For 'computational purposes it would be sufficient to
estimate the roots numerically, but it is usually easier to estimate the
integral directly (see Chapter 9).

Nonetheless, the method of decomposition

is valuable because often the factorization of

Q(x)

is given by

ditions of the problem and often the factorization is easily obtained.

723

;3 2 3

con;

Next, we= turn our attention to the igroblem of obta ning the partial

fraction decompogiiion' on.;the denominator is given iin factored form.

decomFinal we consider the problem of obtaining the partial fraction

.

position of

P(x)

P(x)
Q(x)

wherefthe Tootslof

rk -degree of P
.constants A

(x

)(X - c ),... (x 1
2

ih

c)

Q are all real and simple (ofmultiplicity

Q., From the foregoing, there exist.'

is less than that%of

(k = 1, 2,

such that

n)

Al
x - cl

F(x)
Q(x)'

1) and the

n
.

Por

we obtain on multiplicqtion

. .
I

F(x)(x Al
where

S(x)

ci) = T(x)

) (

Q(x)

is the sum of all thepartial fractions but the first.

neighborhood of .x = cl

that the expression .T(x)

this equation state

T

defined,the constant function

In a

Therefore

ft 4

P(a)(X - ci)

Ai"

11M
"-) c1

P(x)
lim'
(x
x -4/c1

c2)(x

- c

(c

whence,
(c14

(8)

Al --(c

- c )(c
2
1
1

(c

c31)

- c

Thidast expression can be written tidily if we observe that since_ Q(c1 ) = 0

Q(x)
(x - c

lim
x -4 c

Q(x)

'" C(el)

lim

X-

CV(ol).
1

'

1:41

tblus A

Since

is simply a symbol for any one of the roots,

C
1.

,-6,FFIT.

i.

it does not matter which for the purpose of this discussion, We'have in general,

(9)

P(fic.)
I

Ak

ITTt
724

3'2 4

.f

*4

Exwmas A4-3b. 'wa obtain the partial fraction decomposition of

2
X + x -1
(x. + 1)x(x 1)

.denqminator

has simple zeros at

and

-1, 0,

P(-1)
Qt(-1)

1.

From

Pc0)

-1

' W(0)

-1 ' Q'(l)

-,

we have
P(x)
g(x)

1
1
+
x
2(x

f5747717

.1)

which is esipiy :verified to be correct.

There tire general techniques for the case of mill.tiple real. roqs*or

"imaginary roots, but in such cases it is often easier to determine ,the

decomposition by the method of equate coefficieii2b!-- if

Ebcimple A4-3c.

From

x3,

x(x

- 1

p x + q
1
1 +
x2 + 1

x'

+ 1)2

4, we obtain on multiplying both sides by

132x

q2

(X`0 .+

1)2

x(x2 +3.1)2
J

x3 . 1 =

r(x4 K2x2

= er + pi)x

+ 1.) + pi(x4'+'x2) + ql(x3 + x)

i
. 3

+ qix

+ (2r +

pl +

p2)x2
+ (q1

provided x / 0." Now the coefficients of like powers on the right and left
must be, equal (Exercises A4-3, N0.3) Thus we obtain the equations
r + p = -0-" '.
1

q1 =1
2r + p

+ p2= 0
q2

r = -1,

from which

r = -1,

pl = 1,

ql = 1,

q2 = -1,

= 1.

Also called the method of undeterMined coefficients.

:'

725

315

This yields

x - I

x + 1

x3 - f
, 2
,2
xkx + 1)

2
+ 11
(x

+ 1

which is easily verified t9 be correct.

Given the partial fraction- decomposition of a rational function we

complete the work of formal integration. by showing how td integrate the

For (1) the integrals are already 'found.

standard forms (1) and,(2).

If

n >1, we have
A

t-

(10a)

and if

n = 1,

+ C
(x

then

(10b)

,x - 1

dx = r.log Ix -'11 + C.

For (2) we intr6duce the substitution

V

'

(x - a) = b tan u

where we assume ,b > 0

<

(compare Example A4-1b).

b
2

dx -

Using

cos

VX
<2),

du '

we 'obtain

p *tan u + pa+ a' b

du
2u
2n
tan2u1n cos
b

px + q
[(x - a)2 +

b2111

--2-b

Of the last

2 n -1

cos211-3u sin u du +

922-1.

2n- 1

cos211-2U du.

o integrals, the first is immediatel/fqrmally'irrtegrable and

the second- is given by the recurrence relation of EXample.A4-2i.

We iepvq

as aaexelcise the problem of completing the integAtion and representing

fords,' integral in terms of

The resulting integral is_ssum of terms

of the following types,

Ax t B
(11a)

[(x - a,4. b21k

is a positive integer,

k <n,

. A log [(x - a)
'A arctan

2
+ b2],

x - a

1(
726

"

3 '416

Is

,.....

i''

A4-3

,\

Finally, we observe that if we, know the factorization of

P(1

the fort of the integral

eg

Cgx,

from (10) and (11).

we know

Therefore 1t is

Q(x)

'sufficient to differentiate this form and determine the constants by the 44'
method Hof equated coefficients..

Example, A4 -3d:

Consider

5,

x + 1
x?(x2 + 4)

v'

The integral must be of the form

a log x +

+-a log' (x2 + ijj + 13 actan E 4- C

. The derivative of this expression is

;
b

2ox

x2
x

(a.+ 2U)x3 + (213- b)x2 + 4ax - 4b

x
x2 + -4

x2 + 4

x2 (x2,

+ 4)

-Since the numerator of this expression should be

A

we have on equating

)6.+ 1

coefficients

,.

a._-1- aa = o,

4a = 1,

20. - b = 0,

-4b = 1,

whence
,

a =

, b

1'

g1

(L:=

It is easy to verify that this yields the correa integral.

Exercises A4-3
Integrate the following

1.
.

x + 2
2

(a)

a)(x

(x

c)

x3 + 1
(f)

+ 3x + 1
x3

(b)'

(c)

."te)

'

+ 3x - 10

x3
2

x -+ 2a)i. + b

(b > lel)

.6

x3, Q1

(g)

1
'x3 + a3
...

2
(d) -x + ax + p

.
,

(h)

(x - a),(x. -,b)

.__
SZ122

x(x -'1)2
(Consider the cases

a i b

and

a = b)

A4-3

--4---

(i)

(/) :t71

_2
(j)

- 1
1

(k)

"6

x
2.

+ x

Prove from Eq

a, < a

<

/OIL

ion (3) that if

Q x)

(x - si)(x

< a

then

n'

where

(x

a2)

has a 'decomposition into partial,

fractiond of the form

It

1
.x - al

72,77c

3.

r2

4-

x - a

-4-

...

-4-

rn

x - an

Prove if
a nxn + an-lxn-1

fox'. all0but finitely many numbers

bnxn + bn-lxn-1

x,, that the coefficients of like

,powers on the right and left are equal; i.e., pk = bk

4.

Verify that

px + q

.2,

[(x - a)

+ b

'dx

+ b0

for k = 0, l

can -be expressed as the sum' of terms

of the forms (11a, b, c).

.1

to rt

ear
8

.
.

..

Definite Integrals

4.4.

In Chapter 9. and earlier sections of this appendix we addressed ourselves

:Pribfiarii

-"ro the problem" of finding the indefinite integral of a,given function.

--,,.

In principle, this solves the.problem of evaluating any definite integral of the

function. In practice, it is of't'en desirable or necessary to ev:aluate_a definite
1
*
integral, not by formal iltegratioU,'but by some other method altogether.
.

It may be impossible iactotain'an explicit representation.of the indefinite

integral in termssof elementary functions, yet'some,special symmetry may
41eld the value. or a given definite integral.effortlesaly.

Even if the formal

.expression for.the indefinite #tegral is obtainable, the use of a symmetry
concetion may be(a worthwhile shortcut. Often the idea of integral remains
appropriate when the Riemann integral, as, strictly defined, does not exist
because the range or domain of the integrand may be unbounded.
In these
cafes, we b,av(toeittend the definit

411, of. integral in a meaningful sway.

these problems are treated in this s;Alon.

All

*.
(i)

Symmetry.

Watch for-symmetries; the observation that a symmetry

0
exists often provides a direct solution' to a problem,or an important simplification. We have already pointed out
one useful symmetry in Section 6-4.
.

If

is.san odd function and integrdble pn

C-a,a);

then.

(1)

,
.f(X)dx
= O.

-a.
0

Example

Consider

Ti

Avoi.

x e

I -=,..

A>

sin

x dx.

.-7(

It is hopeless; to find the indefinite integral, and it, i

I.= O.
:401.4

' t needed, since

..//
...---

If

is an 'integrable even function on

then

c2)

, a

f(x)dx -= 2

-a

f(x)dx.
0

b'

:729

320

IF

Consider

Ex, ple A4 -4b.

+ a

(a0 +a1 t+a2 t

t
2n

211\

ldt.

-x

.0

The odd powers contribute zero and for the even.powers'we obtain
2

(a0 + a2t

I =.2

2n

2n

dx

.10

3
a

(ox

= 2

2n+1
x
2n
2n + 1 )

Often an integral which exhibits no obvious symmetry can be'transformed

.

into a symmetric integral.

'finis is specific for each case and no general rule

_for discovering such symmetries can be given.

1,

Example A4-4c.

Consider

-;

ros

dx

N
Since the graph 'y =
u = x - 2

we set

has a center of symmetry at

317-72

and find
3

U du = O.

=
-3

Another-important syMmetry of a'function is periodicity.

0

If the function f is integrable and periodic with period

p, then the integrals of t offer intervals of length p are
p11%the same; i.e.,

b+p

a+p

f(x)dx

f(x)dx =

`(3)
a.

for all/ a

and Jn--

The graph y = f(x)

The statement is gepmetrically obvious.

interval of length ;la

represents the complete graph in the sense that the

picture,of'the function from

*

kp

to

a + (k

over any

1)p

where

to

is identical to the picture from

lc' is a

nteger.

The entire graph can be

thought of as a sequenCe of identical 'pictures of width

end (Figure A4 -4a).; If a frame of width

p,

laid

is laid over the graph (the

730,

330

la

ti

JP
.

a+(k-l)p

a+ky

b+p

a+(k +1)p

.
+(k+2)p

I-

...

Figuie

A4-4a

733,

331

,interval .

in the figure) then, the part of the total graph within:

(b,b + p)

d reassembled to -form the

a + kp

the frame may be cut along a line

original picture by interchanging the two piece formed by the cut.

geometrical discussion is

xs'tctly paraphrased by the analytical proof.

The proof is' left to EXer

ses A44, Number 12.

Consider

Example A4 -4d.

n4.144+

n st

cos 2kyx)dx.'

+ a

(a0 + a1 cos 27tx +

I =
./ 0

'

od

Since the intdgrand is periodic with pe

*1

1,

n1

This

av cos :2vIcx dx.

2vtx dx +

v.cos

as

v=0

0 v=0
v > 0,

For

1
At

cos 2vItx dx 0

sin 2vvx

2m1

= 0
11

and

1/4

sin (Ill)

'

cos 2vrtr dx

2vn

Xoneequentlyi
\

I = tn + .71 a

,
(ii)

.al

0 + 2y

-10n `.

Special reductions.

The general form of a recurrence relation,--

I.

for a definite integral is

. ,

fn(x)dx F g

n(x)1

i:

b'

+ c
a

(x)dx.
i. f
n-3.
a

-- _
,

Quite often speCific problems lead

to integrals for which the "boundar y" term

A.
,-.-)
-,
,

b
gnZx)

="gn(.b1 r OA)

is zero for. n > 0, say. -If so, we immediately have

.

732

332

A4.-4

i,

fn( 2;')

1,

.c

n-1

l'

f0(x)

,ThUs IrC-Example Ali. -2f, we could conclude at once from

Sxm(1 n x)m+i
x 1 (1 - x)m ."

ni;+ n + 1 ) n + m+ 1 i xm(1 - x)m-4' dx

Az

hat'',.

,
1

xm(1 - x)

dx -

n(fi - 1)

II

... 1

In + m+ 1)('n + n) ... (m + 2
n n - 1
n + m

."J

dx

. .. 1

Thus we obtain an important c nnection with the binomial coefficients:

-1'
it

xm(

x =

o.
EiaMple A4 -4e.

A case

{(n

m + 1)

(Ili

f special interest is

cos

x dx.

From the result of Exatpl

-2g, we haye
c s

v-1

y/2-1
x sin x
V

",

v V

v-2

For v > l,' this yields si ply

( 4) '4%

For V

Iv =
even, V = 2i,
-41W0

(5a)

For

%It

v .4odd,.v = 2n +

v .-- 1

v-2

we obtain
(2n -0.)(2n - 3) ... 1 y
.2n(2n - 2)
2
2

we obtain
2n(2n - 2)
2
n+1 -:(2n +/1)(2n - 1) ... 3

.733 '

333

t"

.From. (5a) and.(5b).there van be obtained a graceful represed-

"cation of

It

known as Wallis's Product.

2
4
3

all

'

and since '121.11

12n+1

on 1041

0 <cos x < 1

< I.

so thi.It' I

(2n- 1).(2n+ 1)

I2n

'

(2n)2

5 '5
.

Now, since

Observe that

2n *.1
2n

cosN*ix <cos9 x

we have

It follows that

I-

2n+1

< I

2n

< I

for

2n-1'

that' \

12n+1'

I2n

1 <y---=

2ntl
-

2n

lim 7---- -

Taking limits we obtain

2n

1,

lence

'2n4:1
.

,
2

1.3 35 57

where by this infinite product, we mean simply

e
lim

2
6.

n-4 [1 ' 3 .3. 5

(2r0

(2n - 1)(2n + 1)1

7
.

lim
n

'5

122n. (n02.1 2
1
j
(2n)!
2n + 1

The verification that the to expressions in tletese limits are equal

is left as an exercise.
it

'John Wallis (1616 - 1703)

nglish.

*.

734

3'34

Exercises A4-4
Evaluate the following defipite integrals: *
sin99

99

/2

)(

99

1.

dx
'a + b cos x

'x2 +'(99)2

-99

g.

6.

1x3

-3x

a >.b > 0

k.

smis

g/2

2
7,

dx

sin

cod

x dx

o
e

2.

-8:_

log3% x

dx
rr

a.

1 x + x-

iv

n/2

4.
0

dx, (12,a positive

integer)

ir/2

sin

5.

dx

- x

9.

m x qOpm x dx ,

sins e + 1

10.

2
a

(m, a positive integer)

+ b 2 cost e

s n

a > 0 , b > 0

Ir
I

-a
ii.

f(x)dx

Compare

f(x)dx -when

with

is even or ?aird to

-\

-a

derive the resplts (1) and (2) of the text by a method other than the
f

one you employed for-Exerq.ses 6=4, NUmber 4.

e
12.

Prove if

is integrable andperiodic of period

f'

then for all

p,

4-

a and b

a+p

f(x)dx.

f(x)dx =
A
13.,

Prove that if

n > 2

then
k
dt

< .524,

.500 <

1/F77
2x(1 + sin x)
14.

Prove, that

,71(

1 + doS2

dx =

-2

sr

X
I

15.

Show

22
1

62
42
3 5 .5 7

(2n)
(2n - 1)(2n..+ 1)

2n(no2

2n + 1
,

tad

7350

335

(2h):

!2

16.

Determine the value exact to two decimal places of

"

e36.1

1
17.

sin(v log x)
x

411..

Evaluate
v/4

j -v/4 2
(Hint:

cos 2t

dt.

Express the integrand as the sum of a s ,

integrable part.)

11111

etric part and an

Appendix 5
.

THE INTEGRAL FOR MONOTONE, FUNCTIONS

A5-1.

Introduction

Area, as we treated the idea in Chaptyr 7, was not defined analytically

We did not question the

but accepted as a geometrically understoodrconcept.

idea that a region with a curved boundary has a definite area but began with
the implicit assUmption that it does.

Our intuition did lead us to the

Fundamental Theorem of Calculus enabling us to calculate areas by finding

integrals.- In this appendix we shall take the concept of area arrived at

O

intuitively am express it in precise analytical terms.

Underlying our method for determining the area of a regioh, there are a
few elementary ideas.

These ideas are commonly accepted properties of area

which we postulate as the basis for the formal analytical definition of area.

The area function a

which associates with each region

real number, the area of

Property 1.

a(R) > 0

Property 2.

If

S. and

R,

(every point of
Property 1.

If

of the plane a

should satisfy the following properties.

are two regions and if

S

is contained in T

is alsoia point of

T) then a(S) < a(T).

is the union of two nonoverlapping_reilons

(every point of

lies in

Ri

or

their occomon boundary lie in both

R2
Ri

Ri

and

and only the points on

and

R2), then

0(R) = a(R1) T g(R2).

4

Property 4.

If

R, is a rectangle of height 'h

and width

then

a(R) = hw.

Property 2 is called the order property of area and Property 3 the

additive property.

Properties 2-4 are illustrated in Figure A5 -la.

737

337

"so

Property 3

Property ,4

Figure A5-la

Exercises A5-1"

1. Prove from Property 3 that if a, region

is the union

nonover-

lapping regions then

O

a(R) = a(Ri) + a(R2) +

2.

a(Rn).

Show that Property 2 is actually ,a consequence- of Property 3 ziiien that

.,area is nonnegative. lngorpebrate 'the, notion of, complementary legions.
Using ,the given f14-c3perties of area obtain the area. Of
AO,
by elementary geometrical arguments

lb the 'same, fora trapezoid.

4.

If Property 4 is replaced
Property 4, The, area of a
Property
COngruenp
show that tiNgrea, of a, squar
6

Using the previous exercise, show

h and width w :is )11w:

:.,..

it 4!qtr is Dap.
:5

have,

he same area,

aide
t,,,,,

ls of length

a'.

I.4

a?.

t tba area of a /..ectangle 9f,h.eigiq

,.
4 A, %
: ''''
,;: ,
k. .1 ' ! .

44. *.tif
4

tr

uic

A5-2-..Evaluation of anArea
This section describes, in general terms, the estimation prooedure of
Section 7-1.

Let

be a nonnegative bounded function defined oh

define the standard region

under the graph of

points bounded above by the graph of

the vertical line

a < x < b

R ..'((x;St)

(Figure A5-2a).

telow by t

f,

a 'and on the right by

on

and

To estimate the area of

[a,bl

x2

We

x-axis, on the left by

that is,

b;

< f(x))

we subdivided the standard region

into smaller standard regSons by subdividing the base interval

o =

[a,b].

as the. set of

Nt-1

n-1

parts, setting

x* = a, x

[a,b].

Figure

We subdividethe interval into

t
xl, x2, 74., xn_i

choosing points of subdivision

= b

and

such that

_
.

< x

....r

On each inter-f6I--tik:IlTx0,

region Rk

< x2 < .... < xn_l < xn.

where

It-= 1,

n,

we have a standard

where

and 0<y < f(x) ).

< xk

Rk = [(xfy)

)-

We then estimate.-the-area:Wea.eh, subregion

tangular approximations.
mk

and an, upper bound

Rk

from above and below by rec-

In each interval/ .[xk,_1,xk]_ we obtain .a _lower,bound

Mk 'for

f(x):

Thiszprocess is sometimes referred

o as establishing the net.

739-

389

ie.

mk < f(x) < Mk,

(xk_i 5 x

'
The region _Rk

is therefore contained in a rectangle of height Mk and,"in

turn containva rectangle of height

,

on the Ammon base

m,

Exk-i,xkl.

We

conclude from Property 2 and Property 4 (SeATOV.A5-,1), that

nt(xit

a(IV 5-Mk(xk

xk-1)

xk-1)

Using the additive property, Property 3, we then he4e

a(R) = a(R.04.,+ a(R2) +

It follows

+ a(Rn).

that

(xi - x0) +

a(R) >

- x1),+...

md(x2

mn(xn -Ixn-1)

and
<'M1(x.1

- xo)

xi) +

m2(k2

In abbreviated sum notatio'

).

ction A3-2) we have'

xi-1)

nik(xk
k=1

Nock

a(13)
k=1

x 4

Let us review this method for the function

Consider the region

region in Figure A5-2b(1)).

under the graph

Since

will be easy to-approximae a(R)

y = x

x2.

on

[0,1],

(the shaded

is an increasing function on

from above an

[0,1]

it

below in the manner of

Sgaion 7-1
4

We use a subdivision of
subdivision points

k-t

inteiwalof

the aubdi-loision,

f(xit_i) < f(x) < f(xk)

into

[0,1]

since" f

n
x

n-1

equal intervals by means of the

n
= --=

Ase.

we have

< x < xk,

is increasing.

'

VW.

74(16. 4 ;.)
.

On the

,qtn

A5r2

I
( I)

Figure A5 -2b
.

We conclude that the standard region

contains the rectangle

tangle:Tk

of height

Sk

of height

f(xk),

f(xk..1), and is contained in the rec-

both on. the same base.

The union of the non1

6.reriapping reiangies

dk- forms a region

and the union of the rectangles , T..

k
-1

(xk_vxk)

Rk based on the interval

_:

which is contained within

contains. R.

From the properties,

of area we may then obtain upper and lower estimates for the area a(R).

3"4 1

R,

CIP

We have a(S) <

where

a(T),
R

ti

ti

V,
c=4).(?ck

'plc -1)

- 1\2 1
n

4c-

k=1
n

(k2 - 2k + 1)

n
=

(2k

k2

n,

k=1

and
n

k=1,

n
k

=1.1L3 145.71

k=1

AO

We recognize the second sum in the braces within the:formula for

oad.,natural pttAers'',

as the sum of an arithmetic progression, the first' n

whose sum is

k2. of the first

The sum

n2.

a(S)

squares appears in both

l;

the formulit for d(S)

and.that'for

A general 'treatment of such'stOs

de(T).

For this particuldr sum we have (Example A3 -1g)

is given in Section A3-2.

,

k2

n33

2
n2

.=

k=1,

1,d1:11eqt.terY,

1 L1

a(S) =

43

(2n
/1.

a(T)

Since

3'3

is contained in 'R,

2'
2

= 1

and: R

is

2n

,,2

1
2

on_

on

cogtain4t-in

T,

Property 2 of
,

area states that

-742

340
sr

'

-: 7

a(s) < a(R) < a(T) )

t-

A'13 _ '2,11+

a.

7 < a(R) '5.:

6n

a.

+ 'f17-+ ,

on

,,

As we increase the number'of subdivisions

3'

steadily better approximations to the number

1".

FikR) =

both

n,

a(S)

become

and we conclude that

..

,\

a(T)

and

Fotmally, given any tolerance E >0 we choose n

to satisfy

the inequality
o,

'..4

,.,., -.--.,-- -,-,,:;),...A-kw4,1....-

I.

..._.

'''

,
then a(R)

differs from a(S),

a(T)

oi'

by at matt

Ei NO the estimate
i

.a(s)

from below and

a(T),.., from above differ from each other by at most

'2E.

.1Sp4cial sUmmation;techniques can be used to obtain the areas of standard

14.4,

regions ftor other functions.

used for the power Function

In Section A5-3 such summation techniques are.

x -)x

and the circular function

x -).cos x.-

Often it is not convenient, sometimes not possible, to represent the area as

c'El' limit of sums which may be easily evaluated.

The Fundamental Theorem of

CalcUlus offers simpler and more genet* techniques but these, too, may fail.
The idea of approximation is the fundamental one, and if alL else fails we
can always resort to obtaining approximations from above and below by the
Trapezoidal Rule or Simpson's Rule to find the area of a standard region.

713
1
)1111144110
L

'0,

0,fott,r,

f,:,'

-343
0r

'

1,

Aj 1

Exercises A5-2

, l., Use tre summation (method to find the area of the standard region. defined

(a),f:x-4e,0<x<b,c>0.
1

(b);

< x < b, c > 0.

(c)' f : x
,(d)

(e)

x2

+ 2x, 0 < x < b.

f : x sin (ax + b); 0 < x < c; a, b, c such that

ysin (ax. + b) > 0 on [0,c].

x -4.cos

x, 0 < x < c.

C.

2. ,Determine the area of the standard region for f

x -416-c

on

[0)1].

(The summation encountered will be similar ihtl! thel'one encountered in this

saction.)
3.

4.

5.

-0b5iin the result of Exercise 2 using only the fact that the area under
1
-.the graph of f : x x2 on (0,1] is
' together with the basic
'properties Of_area, without resort to summation technitfues.
Show how the upper estimating sums for c ate related term-by.-tera to
the lower estimating sums for x 2 . ,klicint: Sketch a graph of y = x2.
Use this graph and the y-axis to represent the standard region defined
by

4:-.)

If

S n -=

yn

shOw that

sft

34

3.

SuMmation Iechniques

InteEirstion

(i) integral of a polynomial.

)

In Section 7-5 we noted that integration is a linear operation, that the

integrallof,a linear cpmbinlItion of functions is the same linear combination

of theirlibtegAls:

----I
,--

(91f1(x) + c f (x) +
.

2 2

"

cnfn (x)]dx

b
=

fi(x)dx + c

c,

s.

-4

fn (x)dx.

+ cn

(x)dx +

In particular fOr a polynomial, we have

f 1:
a

x =

crxr

aria xrdx.

r=0

r=v

In.order to integrate a polynomial, then, it is sufficient to be fable bro

,1,ntegratepositive integral powers.

.0

ti

We here
,*

f(x)dx_s,f,x)dx

fa f(x)dx__
provided that

is 1nteg

lei over an interval containing the points

e,

bj c: (See' the discussionl)re-Ceedine,Example 7-5e.)

In partiCular,-for, a:

polynomial ye have
,

b :(x)

.f(x),cbc.,

_fa

f(x)./13c.

0
- -*4 "71..

We need therefore consider only integrals of the type :

f(x)dx.
.p
0 /c.

Considpr, in particular, the . integrtil of

--it /0 the-function
_

partition

,h

is incregsing on the interval.

which subdivides the simterval into

v(a) =

-oc':r

xr, over 9[0,a].

we ford the upper sum

over, a

Since

We:tall.e.a

/--

eq61,parts of length
using tht:mwsimum of

'n

in web subinterval; tines.

75

3'4 '5
-

Ki(xk - xkli),
k=1
n
(kh)rh

k=1
hr+1

-kr

k=1
.

AcordingAto Equation (4) of Section A3-2 (ii)'

1)r +1

411kr-41

have

+ P(k)'

r +

4.41,

where P

is a polynomial of degree

r -.1.

It follows that

=,

r+1
h
U = 1=1.

.(2)

1: r r+1

i+1,

- (k

Lk

1)

+ Q(h)

k=1.
-

where.

Q(h) = h

(3)

r+I

P(k)

k=1
and

is a polynomial of degree

r - 1.

We recognize the sum in (2) as'telescoping (Section`

1)) and obtain

=,1( Tth)r41.

hr+.1.,

;'

Since hh = a,

r + 1

--I. a_

we haye
r +1

r +1_ +Q(h).'

We can show that

:ar

Q(h)_ can be made closer to zero thaany given error

--i4eranc:e' using only-tfint.thedl ree

...

PO is at most
I

We, set

r-1

P(k) =

r - 1.

Iliki;

Since lk <'n

it follows;
',.,

1=1
.

...

r-1
,11)(k)1.5..
.

r-1

e-

r-1

IPilki < Z' IP 'Ini

1=1

J.1.
1,

'

740 4 6

.
A

. i

r-1 7r

<E ip i.ini-71<nr-i E- ipii,

1=1
.

1=1

*SI

In shi to we have found

34

(5)

JP(k), < Cnr-a

where the constant C is simply the sum of the abso3_ute:values of the

coeffiCients of P(x) Entering the result of H in (3), we have
n

''(6)

1Q(01 < hr+1 E ip(k)

k=1

n'

hr+1 E cnr-1
k=1

< hr+1

n(Co

r-1)

< Car h.,

where agarp we use the fact that nh = a

lim Q(h) = 0.

It follows at once that

h-O

Wft-uld also form the lower sumi__L_nyer

by taking the minimum value

of xr

as lower bound in each

r' x 1 In this way we could
obtain la result for L similar to (11) and so prove
a

xrdx

(1)

1.+1

r +, 1

' the details are left to the. reader.

._-

(ii) A cosine tftegral

seet us attempt to find the integral of cos x over [0,81, where we
suppose a < vikso that's x is degreasing on the interval. We take a subthe interval Into n equal parts cif length h = a . Setting
_

x = kh,
k

'f"

(k=r, ?

,
de,

we obtain a lower sum

over a

L =E -(cos

xk),(x.k- xk_i)

k=1

E
k=1

.4

cos kh

n),

ioirer. a

and'' an upper svM 'U

..

li

,i..

"

= -17

E c,:,,s(i, - 1)11
k=1

= L + 11[1 '- cos .a] ;

Frot Equation (7) of Section A3-2(ii), on setting

sin(n
+
2

n(a + 1) sin na

-cos

- sin

we obtain
n

Ecos k,2 = u(n)

(2)

u(0)

,
,
sin(n + -)z
1

12

2 sin f z

k=1

(2) permits us to evaluate

in Equ'ation (1):'
Equation

7*.i

lim L =

ih
2

the

,
1
sin(a + h)

h-O sin 7 h

Using the Pact that- lim sin u - 1

u-0

limit of the lower sum given

h
2

-we Alave

u
lim

L = sin -a.

h-(2

Since the diffrence

J-

between
--

and

U has the limit 0,, we conclude that

.)

cos ,x dx = spin a.

748

3.4 8

, Exercises A5-3
1.

hf subsection (i)jof this section we state that it follows "at. once" from

the inequality (6) that

lim Q(h) = 0.
11-0

Actually, what theorems on limits are beihg used?

1

2.

Show simply, without repeating the argument of the text, that the lower

sum L over

.E

xL1 (x- xk)

also has the limit (7).

k=1

3.

Employ Equation (8) of Section A3-2(ii) to obtain

dx

for

0 < a <
"

.40

7r
\or

41

749

.349

--

TheO6ncept of Integral.

Integrals of Monotone Functions

befinition of integral.,

In the compUtation of the area of the standard region under the graph of a
bounded function

on a closed interval we gave upper and lower estimates of

the area in terms of upper and lower bounds, fox

If the functiQn

division.

takes, on maximum and minimum values on each

subinterval, as it would if

on each interval of a sib-

were continuous or monotone, thenthese would

When

give the sharpest,.possible bounds.

is continuous it may be easier to

use slacker bounds than to attempt,..tip determine the extrema.

functions, however, the situation is especially simple:

For monotone,

The extreme values

on an interval are taken on -at tit endpoints:

We may allow

to take on negative, values so that the interpretation of

the upper and lower sums as upper and lower estimates of an area may not be
immediate.

Still these upper and lower sums may serve as upper and adwer

estimates for some unique number which lies below all upper estimates and above
all lower estimates; Lf such a unique number exists it is called the integral
]

of

oven the base interval.

The idea of integral has farms- reaching appli-

cations, and.its interpretation as area, although useful for visualizing the

concept of inttgral, is no

necessarily the most important realization of the

concept.

We consider a bounded function

into

A subdivision of [a,b]

(x

xo = a, xn

where

defined on a closed interval

(a,b), a < b.

interGals is defined by a set of points

x _

, x

n-1

and

x < x < x < ....< x

1. 2
rt

n-1

n.

We shalt call a set

of
_let

(a,b1.

of poind satisfying-Mese-requirements a partition

On the k-th salantervaf Exk..1,xk]

defined by the partition

mk be a lower bound, Mk_an,upper bound for

m' < f(x) < M

for all
bounds

mk

in the subinterval.

We define the lower sum over

as

759

a,

r(x), 'so that

a for the lower

N.

n
mk(xk

L =

and the upper sum over

k=1

a for the upper bounds

as

Mk

U =-14EN.Oxit.- xk.1).

If f

is a nonnegati4e

'unction then the lower and upper sums correspond

to lower and upper estimates, respectively, for the area under the graph of
on [a,b]. More generally, without restricting thelsignof

DEFINITION A5-4.
number

Let

f. be defined on

just one'number.

if it exists.

f,

[a,b).

is the,integral of f over *(a,b)

We Say that the-

if there exists

such that for each choice of partitions

a2 and all lower sums

Li

over

we use the..

f,

loker and upper-sums to define the integral of

al' and upper sums

U2

al,

over

a2, 'we have

L, < I <

We raise the question of existence of such a number

not immediately clear.
than any upper sum.

I because it is

It is possible to prove that no loWer sum is greater

Still, there

ay be a gap separating the values of the

upper sums from those 'LIT the lower sums.

,-If so, there is more than one number

between-the lower and upper sums and the integral is not,defined, On the other
hand, if for each e > 0 it is possible to find lower and upper sums which
differ by less than e, there'issuch a number I which these lower and
upper sums approximate within the error tolerance
able to define

e;

in, other words, we are

as the limit of upper and lower sums.

ciaple result here as a thebrehl which we shall use.

We state the prin-.

THEOREM' A5 -4a.

Let' f"be a:bounded fundtIon on

every positive

there exists apartition

lower and upper sums

than
f

e,

over

and U over

then there exists a number

[s,b].

Conversely, if

_then there exists a partition

and U

of of

ff-fo-r
fa,b)

which is the integral of

is integrable over

f.a,b)

a with lower and upper sums

such that U - L < e.

751

51.
4kt

and

a which differ by'less

If

over

f',,Inasan integral

over

we'say that

[a,b1

f ,is Integrable

(a,b).

A5,-14e requires a verification of the con

Definition A5-4.

ions of

FirA we Must have a derricitistratiOn that ho uppell sum.is lessq

than any lower .swn.

In that event, there.exists at least one number whidh is

A"
.p:,

both a lower bound for the set of upper sums and an upper bourid for thetet
It must then be shown that there isattilost

If lower sums (Separation Axiom).

one number I'between the upper and lower sums.. This follows from the exis=
tence df an upper and a lower sum which are diAer together than any prescribed
tolerance

E.

lower sums.

(ii)

Thus the4integral ts determined by a squeeze between upper and

For the details see Appendix

Integrability of monotone functions.

For monotone functions $'e' may choose

8.

at,the endpoints of

[x,

hick]

mk

and

as as function values

and it is particlalarly easy to obtain an

estimate of the difference between the upper and lower the error
We picture the situation in.terms

of approximations to the integral.

of the area of a standard region for a nonnegative increasing function

1(b)- f(o)

v(o)
.x.eb

xk

-.

Figure A5-4a
In Figure A5 -4a, -the. shaded.re.jtangle over the inlerval

Mk - mk,

where

Mk

f(xk)

and

[xk_12x0 -has height-

n'k= f(xk-l):

The total are;, of the shhded rectangles is.the difference between the

upper an lower sums for the given partition.

752

nr.-o

/.

Since the function

is monotone we ,cat imagine sliding these rectangles

.

...

`0"14..**parallel to the X.axisinto an arrangement with their right sides aligned-

In

.--,

this ch.Tangemen:t, the rectangles are contained Without overlapping in a single

)

.4

ree*igle.of

height' f(b) - f(a)

and base equal to the length of the largest

illt4rai of the aubdiVision., The, length of the largest interval,

,

w( a)

- :x..k.:1),

4 7 - m a x

is a measure of the,coarseness'of the subdivision and is.called the norm of the

--partiti9n,.., a.

We have'depieted a bound On,the difference between the upper

and lower sums:

U - L < [f(b) - f(a)]v(a)

Al*

Clealky, we can'make the 'difference between U and

tolerance

leset than any error

by making the xubdivision fine enough, namely, by choosing

so that

v(a) <
,Since the area I

and

f(b) - f(a)

must then lie in the interval of length at most

its value cannot differ from either by more than, e

between

andwe have

satisfied the condition Of Theorem. A5-4a.

AlthoUgh.we have obtainedphe last result by a geametrical argument we

Can obtain the same result analytically with ease.

We now prove:

a finite monotone funption on a Closed interval is integrebl.

THEOREM A5-4b.

If

is monotone on

(a,b),

then 'f' is integrable

overl4a,b] .

Proof:

partition

Washov'that for each positive

a

of

fa,13)

fort which the difference between -the upper and lower

sums on the partitiontalrbe.made less than

.

e:

U. - L < e.

*
For'thisose_we_let____Mk be the maximumand
[x.

x. ].

mk

the= minimum of

We shall prove that it is sufficient to use a subdivision

r. norm satisfying,

V(a)

when

it is possible to find a

f(a)l

If(b)

/ f(a).

753

353

with

*w.

Mk = mk

In this cage, we have

constant function.

is trivial since the function

f(b) = f(a)

The case

must then be a

and

- L = 0
a.

for all sul2divisions

We consider the case of a weakly increasing function

[xk_i,x0

The maximum and minimum on

decreasing case is similar).

(the weakly`
are

given by the endpoint values

Mk = f(xkr and mk = f(xk_1).

intervals of the subdivision we have

Summing over

Eu
k=1

=I:lk(xit

xk_i)

fock)(xk

xk_i)

k=1

L =Elnk(xk.
-

=E

xk-,1)

xk-1)

=Ef(xk-1)(xk
k=1

k=1

Consequently,

U - L

=E

,f(?ck)

,,

.,_imck

x,,,

k=1
:7
n

ft(xk) - t(xk_i)) (0
k=1

3-,

k=1

We observe that
-......

.,

1:71(xk)

,-

= f(xl)

f(x2) +

+ f(x

Ef(xk _1) = f(x0) + f(xl) +

+, f(xn..1).

k=1

354

A)-4

Subtracting the second of these sums from the first, we have

J.,

faxk) - f(xk..1)] = f(cn) - f(x0)

f(b)

f(a);

consequehtly,

L < v(o)[f(b) - f(a)]

zs

To make the difference less' than

above.

as indicated

We have satisfied the condition of Theorem A5-4a and it follows that,

is integrable over

( iii)

we need only choose 'qo)

Riemann sums.

(a,b),

Notation.

We have employed a method for defining area by approximation from above

and below an
integral.

extended our approach to define the more general concept of

This method has the great. advantage of logical sImpilcity in the

derivation of prqerti,es of the integral.

A more dirlesc,t method, but one which requires somewhat more complicated

argument, is to utilize' values of the function in the int.ervals of a subdivision,

instead of upper and lower bounds for approximating the area.

function ,f

defined on

La,b1 -and a partition

Thus, for a

a = (x0 , x1 , x2

,; xn)

Of .[a,b] 'we introduce sums of the form

-,

.(1)

=,01E fk)(xk

xk.:1)
,

k.1
.A

-where =E

is any value in: tbe'Ls4iiiterval

[-X.

These are called

-.X..].

x-1' x

Riemann sums

For a general Riemarin sum the 'rectangle over

[xk_i,x0

will

usually not. include all of the standard region under the graph and will usually
i

(,

include some region above the curve (Figure A5-'4b) %o that there will be a
partial cancellation of errors.

Sihce

mk < Otk) < 14k,

no matter how

_it' chosen, we see that the Riemann sums are sandwiched between the upper and
40;

lower sums
;., .< R <U.

1,44

/'

.
.

.
I..

DI
.

After Bernhard Riemann, a German mathematician of the early 19th century,

a Rionelir in the careful study 'of the concept of inktgral and. in other impOitent
N

areas,,:.-

.:..

-A.',
;...

.,

755

st,

vs. ,

a!
1

4
C

s
A

V.

If f

we can therefore approximate 'I, by Riemann sums.

him an integral ,I,

In fact, the approximation'to

by liemannpums can be kept within any pre-

avribed tolerance of error fOr every sufficiently' fine subdivision

and

'15p,

corresponding choice-of

We shall then have determined the intggral,as

tit.

,a new kind of limit, a limit of Riemann sums:

R.

lim

v(s)-4 0

:R

xk

'Ck

14-1

Figure A5-4h.
.

.,

-..

ser
.....

..

..._

.1

7'

It is natural to suppose that if this limit of HieMannSul6 exists, then

so does.the integral
the same.

of Definition*A5-4, and to suppose that the tw6 are

This is not an obvious proposition, but it is true.

These \remarks

aresummarized in the folibwing theorem.

THEOREM A5`-4i.

The value II

is the, integral of ,f, dyer %eta],

in the senselg Definition A5-4, if and `Only if it is the limit

of Riemann 'sums,

I =

lim

R.

v(a), -' 0

The proof will be found in Appendix 8.

ti

75

356.

0,

.11

The integval

71.-

tOtati, on of'Leibniz.

of

over

[a',11]

is pa9a14r written in the elegant

In Leibnizian notation, the Riemann sum

I.

(1)

is written

...

VIere .6De

represents the

integral Leibniz used

ifference
.0

xk - xk.a.. In representing the 4 1P

form,reminiscent of the Xemann sums,

.../

'

Aft

1)

'S C1C6( x) dx.

.---Afthough) as we have seen, the Ceibnizian notation for integral nicely

complements the Leibnizian notation for'derivative, itstems from conceptions

'iit.whtch are difficat to make prebise. In the thinking of Leibniz and most of
the early 'users of the calculus, the integral sign
.

which is an elongated

,.;

Roman
.

vS"- is a special sumOtion -symbol which rleaces the corresponding

b
Greek. symbol .;'1.' The integral
f(x)dx w
ought of as the suo?of
a

the areas of the-inginite set of "rectangles" having

"immeasurably smaIrotase
"s"

in'.."dx"

dx

an

replaces the Greek

height

f(x)

Of the

"A"

75.
k

.357

for
i

nfinitesi*" or
x

111 (tom Roman

e Riemann3sum). -

Exercises A5-4
1.

By,using upper and lower 'sum estimates evaluate the integral of each -

function

2.

over the indicated interval.

4f- 2

0 < x <1

(a)

f(x) = 2 - x

(b)

f(x) = x'

1 < x < 2.5

(c)

f(x) =

2.5 < x

(d)

f(x)

(a)

Find the

s.

<3

< x < 5

5 -

min-i and the maximum values of

2 + 2x - x2

f(x)

on

and use them to find tWo numbers respectively .

I

the intervit10,11,

bve the-velue of

below and

0"/

f(x).,,dx.

et.

0
(b)

3.

V.
.

Che-ck

yoUr result bir evaluating the integral.

ati loWer
the; graph. Oef : x
T-

Find upper

sums differing by less than. .1

it

on

(1,2

for the area under.

Evaluate earth of the following integrals, using upper and lower sum
estimates.
(a)

6.

k function

function

}
;.z

ye

...,

n.

[a,b]

[alb) ''-41Fifbr some Partition

on

interval,
2,

f deflOd on the interval

f(x)

Thus

a =

is said to be a step-

fx

... xns

-..

2,f the

O' 1
is -ao4ant on each open subinterval' (xi;_i;:ck,
sgn x

is a step function on

[-1,11.,

4'-++ is defined by
-1 ,
0 ,

sgn x
' e

X<p
= 0 .

.-" 1 , x >.0

758

358

)k ='1g.
sgn x

7.

Eyaluate each of . the'followineintegrals:

The gunct4ion

(x]

is defined

in Appendix 1.

(a)

[3x + 4]

dx

.
N

1r

dx

[i]

(.1))

(a)

[frxi

dx

a
8.A 'show that

f(x)dx = 0.
a

-1

).
r

O
A

"O.

A5-5I

.14.4

4
s

A5-.5i

Elementary

Proper es

of Integrals

In Section-7-4 a ember of elementary properties of area'were interpreted

.

in terms of the - integral notation. .These

red properties are, in fact, bimp?.e

.I.',

Our purpose in I:.=

, consequences of the four proPertiet stated in Section 'A5.-1.

this section is to show that indeed these properties hold for the integral as

defined by Definition A5+ We shall make considerable use-of Theorem

A5-114V

In this discuss,?on.1,,
f

Let

and

on

be nonnegatiVe,functIons with -f(x) < g(x)

[a,b].'
,

Since the standard region under the

graph of'

is contained in the

standard region undethe graph of

(Figure A5-5a), from Property 2 of

Section 6-1 the area of the former
must be no greater than the area of
the latter.

A stmilar inequality holds

foxPintegrals in general.',
x

Figure 5-5a

`THEOREM A5-5a.

on

[a,b]

If

and

f(x) < g(x)'

are integrable and

then

g(Adx.

f(x)dX <
a

Proof.

integral of

Let ,I
g.

f*

f, over

denote the integral of

[a,b],

We know (Theorem A5-4a) that for every positive

.

exist upper and lower sums

L < J,< U

and

had

is an upper sum for

I >J we could bake

follow that

for

such that

(Definition A5-4) we conclude that U -

upper sums as close as desired to

\J.

and

since
e = I -

J.

< e.

J
e

the
there

U - L < . Since
Thus we can find

At-the same time, every upper sum for

f(x) < g(x).

> 0

We have

I < J,

for if we

and from U - J < I - J* -ft would

U < I, *a contradiction, since

760

360

is an upper sum for

I.

2.A5 -5
ti

Considerthe:decomposition Of the Atanlard region over

into the

[e,c]

two'standard regions over

,A.:(1,O]

Y=49

where

A5-5b).

(alb]

and

a < b <c. (see,Figure

.The., additive property of area

A(Property 3 Of SeCtion A5-1),tstates,1

that the sum.of the areas of the two

-'-'subregions must be the.area of the

entire region.

This corresponds to

a general statement for integrals.

.1

If f is integrable over

THEOUNA5-5b.

ol

Proof.

The proof this will make use of the following result,Which

will be established in Appendix 8.

If
(2)

a < c << d < b

then

[a,b]

Let us assume that

Then (2) tells us that
e > 0,
[aTb]

and

and U ", L"

is integrable on

is-integrable on

is integrable on
is integrable on

(c,d].

[a,b]

[a,c]

and that

and on

a < c < b.

(c,bi,

so far

aabording to Theorem A5-4a, we can find subdivisions

ei

[b,c]

of

with corresponding upper and lower sums,

II'

Li < e

U',

for -the partition

U" - L" < e.

and L = LI +1",..ale upper and lower sums over

constructed by taking the two partitions

together as a partition Of

(a,ci.

U
For the i.ntegrals

and
.

Clearly, 'U = U' + U"

I,

al

(U' - L') + (U" - L") < 2e.

I"

overthe'intervals

[a,c],

(a,b],

(b,

- I < 2e,

111 -I' < e, -U" -I" <

el

'whence,.iar every positive

C.

a,.

7613 6 1

...""A

-'y -

[a,c]

and 47"

Furthermore,

respectively, wehave
a,

of

a'

such that

..

and

w.

77177:6777,7-7777:7-6--77"
I" - U ") - (I.-.U) I,

Ul) ;4-

= I(I!

In,-

< E.+ e +

..
as we sought to prove.

I' 4-.I".r I,

It follows that

f =0.

In ExerciseA5-4,,NuMber,8, we noted that

By defining, for

b < a,

f = -

we then seesthat if
f

a, b

and

b.

are any points of an interval over which

is integrable, then

414,

f =

f .

Ja

-liJigea2tity Of integration.

For positive constants

and

integratiOn is a linear'operation:-

ff

[ce(x)

U'

fox if

g(x)dx,

'f(x)dx + 3

13g(x)]cl-x =

and

L'

It is amediate that

are upper and' lower sums for

U = aU' + pu"

sums for the linear combination

on the signs of a and

Pg(x).

and'' L"

for

This result does not depend

0- as we now prove.

.
.THEOREM 45-5c.

If

linear coMbination

and' g' are integrable over

pg

is-integrable over

fa,b1

then -any

[a,b] and

rb
f(x)d:4. 13

Sb g(x)dx,
a
.

To simplify the considerations which depend on the signs of a and

p we divide the proof into two part's.

6'
r

762
r3

g,

are upper and lower'

L = aL' + pp"

and

0f(x)

Un

f,

P
Part LEI;

If

is integrable over

ja,b)

then for any constant_ -a-,-

.''. e

the function af

is integrablQ, and

b
,---....afT)Ox = a
,

Proof.

over

Let

be a partition of

"

f(x)dx.

a,

Pa,b)

and take upper and lower .sums

:1

=E Nock

).cit_1)

k=1.

=2>k(xk

xk-1)'

k=1

for which U - L < e.

If a > 0,

then

Cdilic(xk

and

xk..1)

k=1

`.-)

are upper and lower sums, respectively, for

It follOwS that

af..

aU'- aL < ae

..
and hence that the difference between upper and lower sumS for 4 .can -be
made 1.is than &ny desired tolerance. It follows that af is integrable,
Furthermore, for the" integral

e.vo,

of

U - I < e,

and

J "Of

af

over ..ta.-dsj

we

hive

-,
.

...
.

7,

aU - J < ae

-..

'from which it ittlows that

441k.

1J' - all =

- au)

alu -

< 13 -QUI
)

< acce.

Since this result holds for all positive

If

a < 0 then aU

a(U'- 1)1

vwe conclude that

e,

is a lower sum and

al,

= aI.

an upper slim for af

The proof is thus reduced iortihe preceding.!

If a = 0,

the lemma follows trivially.

44

763

363
To0o,-0.-0

5
)

If

Part Cii).
integrable over

and

(a,b)

and

g- are integrable over

f + g

is

then

(a/b],

f(x)dx +

(f(x) + g(x)]dx

/".

We make use o

Given any fixed

an auxiliary result (from Agg ndix 8):

tolerance, for any integrable function all suffic ently,fine partitions have
.
I

_upper and lower sums closer then that toleran

!

there .exists' some

such ,that any partition

6,

/^

,Thus for each positive

e,

will have an upper stun U

and a itn,,r sumT L

sat'

fu - LI <e
i

whenever
v(o) < 8,

Let
and

g,

for

f,

and take

respectivel

v(a) < 6.

with
L'

be the controls gorreSponding to the given

=6in(61,62).

Let

U"

and

L"

for

LII < e

IU1

a, ir

such that

g
and

111 < f.

IU"

Recall that

4-J

U' =

and

un

=E

mk"

L"

=21mk"(xx

- xk_i)

k=1

k=1

where
t
g(x)
<M,".

mkt 5 f(x) 5 Mkt. and

f()

+ mk"

it'followt that U = U' + u"

far f

over

a.

g(x)' <

is an upper sum and' L = Lt + L"

U - b = (y,

L,) + (u" - L") < 2e,

"764

3u
4

141i"

We conclude that

for

be any partition

There then exist upper and lower sum's, over

and

and

A5-17797,7,r7^-4.-'..-- r "

;.? ;

f + g

and it follows that

I"

It%

and

of

and f+ g,

- II .

II, + I),

f,

;$ inte/grable.

- U') + (I. - un) - (I -

< lit - IP!

..

respectively.1 we have the estimate,

/

1(I,

11[

?urtheIrmore,for the integrals:

4-

li;'

01
y

- Uni + II - Ili

<
- e + E + 2
:

< 4e

for each positive

i= I' + I".

It follows that

e.

The derivation of Theorem A5-50- from the Weeding is simple and Is

left as an exercise.

In one of the examples of Section 7=1 we used sums to find the Areaunder
2
x -ix .

the graph of

Employing Theorem A5-5c, we can .integrate any quadratic

function without resorting to estimates bg.upper and lower sums:

b

a(Ax2 +

+ C)dx

A I

x2 dx + B
a

x dx + C
a

dx
a

An immediate application of Theorem A5-5c dyes the area between

the
and g on [a,b],
where f(x) < g(x),, as the
integral of their difference. If f(x) >0 as in Figure A5-5a
then the area
between the two graphs is simply the area of the standard region under the

graphs of two functions f

graph of
thht is,,

less the area of the standard region under the graph of

f,

1b

b
S

i( x)dx

a-

f(x)dx

b [g(x) - f(x)]dX:'..

There is no reason...to restrict these considerations to nonnegative functions,

for if f(x) <'0 for some x in [a,b]
and m is a lower bound of f(x)

on

bY,b].,

we translate the x-axis vertic

Im(

units in the negative

direction so that

cx,y) -x,y* (m1).

NW*

765

365
4-

es

lathe neV coordinate system the region lies between the graphs of the non-

negativE-

ctions

->f(x)

A5-5e4).

Iml

and

'

g :,x -)g(x) + Imi.

(Figure

lak

Since i(x) -17(x)-= g(x) - f(x)

the definition of the area of the region

between the graphs of

as the integral of .t fl

and

clearly appropriate whenever

,g

f(x) < g(x)

nri

[a,b].

standard region under the graph of F : x -,g(x) = f(x)

function g - f

Thus, tHe area of the

on

&a,b1

A5-5d) is equal to the area of the_region between the graphs of

-

I.a,bJ. (Figure A5-5c)

is-'

(Figure

f end

on

pxam,,ple,A2.21.

Consider the area of the'region between the graphs of the

functions

2
f : x-4 eos xa and

g : x -, -sin2 x

on

(0,4).

(Figure A5-5e.)

1.
.

, We mIght'attempt to represent the area of the region as. -tiie limit of sums
4
,
of areas of rectangles. en the.other hand, we know that the area is given by

-N.

ti

f(x) - g(x))dx,

since' f(x) > g(x)

for all

(0,4).

in the interval

't 4

But

(f(x) - g(x)]dx =

: x -,,f(x) - g(x)

1'3(1=7

since

f(x) :-g(x) = cost x - (- -sing x) = 1

'for all

is shoWn in Figure A5-5f.)

x.

(The graph of

'

Id conclusion we note that

.

the area of the region shaded in Figure A5-5f

is equal to the area of the

region shaded in Figure A5 -5e.

y=-sintx

If
Figure A5=5e

}1.

Figure A5-5t
(
.;

st.

Exercises A5-5
1.

Exhibit the details of the proof of Part (i) of when

If the graph of

(a)

is odd.

is symmetric with respect to the origin, then

Prove that if

is odd and integrable on

[ -a,a],

.'.

then

a <0.

r
a

f(dx = 0
/,'.
(b)

If the graph of
is even.

is symmetric with respect to the y-axis, then

Prove for sn even funCtion

that

C:a,a]

ia

f(x)dx = 2

f(x)dx,

-a

Interpret this result geometOlcally.

-

which is integrable on

3. 'Prove Theorem A5-5c as a consequence of Part (f) and Part (4i).

Con-

y verseTy, derive theg_d-74as corollaries of Theorem A5-5e.

4.

Prove:

If

and

are integrable where

the;

f(x)dx1 5 f
a

5.

(a,b),

Ijrb

g X'-4 If(x)1 ,on

If(x)Idx.
a

Compute the values of the given integrals using Theorem A5-5c.

7-

' (a)

t
k3x

.' 5x + l)dx

2
(b)'

(x - 1)(x + 2)dx

(x + 2)(x

(c)

(a)

3)dx

Find the area of the region below the parabola

-the x-axis and between the lines

(b)

x = -3, x =

Find the area of the region between thelfraph of

2
-=f

: x

*- x - 6,

th

x-axis, and the lines'

First draw a rough sket h of

= -2, x = 3.

and indicate (by shading) the

region whose area isto be computed.

.

Find all values of

for which
a

(x + x )dx = 0
0

768

et

368

f(x)dx

8. ,Compute

where

'

f (X)

0 <x <1, %

0,

x2

=
5 -

1 <x <

II.x

fl

3 ..

".

,
9.

f : x --ox

Verify that the following property holds for

,

..

c-a

'

f(c

X)dx =

,.o

f(x)dx:

q=b

DoAggp think that

Explain the property geometrically in terms of areas.

the property holds for other functions that are integrable?

Justify

your answer.
.3,

10.

If a function

is periodic with period

5,

and integrable:for all

x,

show tha#
a+n).

a+T.

..i*x)dx = n

'f(x)dx,'

(n,integu).

/
Interpret geometrlically.i

11.

EValudte'(without using the,Fundamental Theor m Of Calculus)

1'

100/

.0
12.
s

PrOve that if

,x

S(1 + sin 2x)dx,

is integrable on

[a,b]

and if

f(e' > 0

for all

then

(a,b),

in

'

0N

f(x)dx > O.

c
.

13.

Prove that if-

and

are integrable over

- f(x).1dx

IS(g(x)

and

(a)

If the strong inequality

f(x)

[a,b],

< g(x),

and

f(u) < g(u)

e.> 0,

g(x)dx. S

for,some

on

b
f(x)d3c <,

If

f(x) < g(X)

proVe the strong inequality

sb

(b)

g be integrable-and suppose that

? on any subinterval of

then

Ig(x) Idx +

Let

[a,b),

are continuous at

x = u

in

[a,b]

and

prove that strong inequality holds as pbove.

769

36.9'

-,71

[a,b].

holds

.,

1.5.5if funOions

and

tf.

-,

<

sr f(x),Ox
s

6.,

....

'

7w
-

.4

1,bi, .dpigs it follow that

a .9. ''

g'are''.integrable,
and; fk) < h(x) < gkx,-,on
..a.

",

'i

h(x;dx< f

g(X)dx?

.-,

a,

Illustrate by an example.
/"..4.

4.

(a)

If

Prove the)-Mean Value Theoi-em of integrbl calculus:

is Ciill-

[a,b],, then there exists a value u

tinuous and integrable on

in

such that

(a,b)

/the-operti.nterval

/7.

f(x)dx = f(U)(b - a).

a
.e

Show that the value

(b)

f(u), in
f

f(U) = lim
h-O
h = L12-1211

where

Thus f(u)

n.

and

satisfies

(a)

+ f

+ f

11.

n + 1

= f(a +

k =.0, 1,'2,

for

khm)

can be interpreted as an <extension of the idea of

mean or arithmetic average to the values of a function'on an-interval.

a

17,

0
+
n + 1

If

a xn + a x
0
1

has atleast one root in

18,

n-1 .
+

n.

n-1

show that

= 0,

+ a

n-le

= 0
A

(0,1),

f(x), is integrable over

Prove that if

+ a

[alb],

then

If(x)1 "is
r's

integrable over
19.

Suppose

(a,b).

(The converse is not true.

See No. 19.)

if

-1

if

is rational

f(x) =
is irrational

.-

Show that-if

If

21,

and

,' min
)

and

thenpU > l

[0,1]
20.

(f,g)

Let

.g

L
and

are upper and lower sums for a partition of.,

L < -1.

are integrable over

If

f- and

integrable' on

(a,b),

are also integrable over

[0,1]?

then bothwax
'

(a,b].

be bounded and integrable over

Prove (a) The function

(b)

Is

(f';g)

gl.

'and

..,,,

[a,b].

f g is integrable over la,b];

is bqunded away from zero, then

sr

770
9' V

- "is integrable on

(a,b].

(a f(x) + /3 g(x)..) 2 dx

4aJ

22.

If f

and

are
botindid4trId integr able, then
to-

,#*-

exists and is > 0 for, all constant a and 0.

Show from this that
b

3f(x)2dx

.g(x)-dx >

f(x)

g(x)dx

Ja

with equality if and only if (for f

and g

continuous)

(x) = cg(x), a < x < b

23.

If f is integrable and its graph is convex on an interval [0,a),

show that *
a

f(x)dx >- af(24

2 .

Interpret geometrically.
24.

poll that

o.

4-

1)(b2
3

25.

4-

l) > C

j0

if(x2

a2..2
A

4- b )

Show Ilthat
1

(a) g

1f

3/2-

<

4.

Or

dx <.
,2
.15

>

(b)
x
0 )(---75

215-

This is known as 'the BAniakowsky-Schwartz, Inequality.

.

771

371

dx

.
26.

.
.

Find a continuously differentiable function F

in

[0,1]

(i.e.,

F'

is continuous)

which satishes the three conditions

F(0) = 0, F(1)

(a)

F(x) 2d x = a

(4))

and

.z.

F'(x)2dx dx

(.c)

is a minimum. .

tr
1

772

372_

Appendix 6
INEQUALITIES ANf) LIMITS

A6-1.

Absolute Value and Inequality,

The absolute value of a real number

.written

a,

is defined by

lel,
V

a,
-a

If

; a > 0

0, if

lal

if

a < O.

we think of the real numbers 4n, their representation on the number line,

then

lel

is the distance between 0

any real numbers

the diStance between

Ell and.

In general, for

a (Figure. 46-1).

and.

and.

is

-1b1

.1a1

C9

Figure A6-1
lb

al

bl:

.--:

then clearly
lxi < e.

If

x- lies within the span

-e .<

Conversely, if

where

x. is no farther from the origin 'than

e > 0,

and we must have

1x1 < e,

then

-.c <

< x

lxi.

It follmis potmetliately

that

-Ix'

(1)

R.

(See Exercises'A6-1, No/. 13a.)

Frbra the inequalities

..1a1.< a < lal

and ';.1b1 <,10

<1b1

e
we obtain
`f.

-(1a1 +

+ 1bl,

ithence
(/2)

",

la + b

tal

fi;

I-

773

tid
%

443 7

1.1

(This relation is known as the "triangle inequality.")

In words, the absolute

value of a sum of two terms is not greater than the sum Rf the absolute value

of the terms._ Sire any sum can be built up by successive additions, the
result holds in general, viz.,

la

+b

+el =I(a + b
< la + b) + 1cl
lal

We say that

is an wooer estimate foi x,

estimate for y

< y.

absolute value of the sum

0.

(y11

is a lower

Such an estimate can be

We then obtain

b = -y,1 in turn.

lx1

11x1 -

a = x + y and then setting

ix

Since

and that

"It is'often useful to have a lower estimate

obtained from (2) by the device of setting

and

Icl.""

In (2) we have found an upper estimate for, the

a + b.

which is better than the obvious estilhate

b = -x

1101

ly1

5. Ix

yi

IA.

is one or the other of the values

lx1

lit

or lyl

we have
4-

3).

IYII < Ix 4 YI.

Ilxi

Rs-

Specip/Simbols:
denotes the largest of the numbers

The symbol` ciax(ri, r2,...., rn)

r2,,

sm

rn;. similarly, .the symbol

lest'of the numbers.

min(ri, r2,

r )
n

""

.31

ample A6.la
7.max(2, 8, -3, -1) = 8

bin(2, 8, -3, -10) = -10

.

mak(-a,

7 lal.

774

274)

denotes the

Ix',

Exercises A6-1

1.

Find the absolute value of the following numbers:

Ak

(a)

-1.75

(b)

2.

3.

(a)

For what real numbers' x

(b)

For what real numbers

(c)

sin (j)

(d)

cos (2)

does

x noes

= -x?
?

410'

Solve the equation's:

(a)

x r = x -

11

13 - xi = 1

svo

(b)` I4x + 31 = 1

(c), lx + 21 =x
(d)
(; )

12x + 51 +

(f)

125c+ 31'1=-

(g)
4.

ix

3I

21 =\0
xl

15

213x + 41 + Ix

21 = i'+

For whlt values of

xl

(Express your

each of the following. true?

answer in terms of inequalitiessatisfied by. x')

ta)

1x1 < 0

60"

I)

40

Ix

11 + Ix

(m)

0 < I

IxI < 3

(n)

Ix - al < 8

(o)

0 <

(p)

Ix -

(4)

Ix - 1 <,2

fX1

21 =1

a21
.

(c)

<d)
(e)

(f)

61: <

Ix - 31 ? 2

.12x - 31 <

(g)

1x-al <a,

(h).

1k2.

; t\

- 31 < 1

lx

(s)

(i)` t(x -'2)(X

3) 1

>

-(t)

<S
11 < 2

YI = kel

an,4)

Ix t 11 <

and

I2x
for, all

IYI,

Isin xI = 0
!sin

>

<j)

Ix - 11 > IX

lx - 5) + 1 = ix 1;

st

(u)

11 'II <1

.(v)

>

y'

5. Sketcthe graphs of the following equations:

.(a)

Ix - 11 + 1y1 =1

IX -

(b)

ix +

(c)

y = Ix - 11 + IX - 31

yl = 2

(d) y= Ix - 1 1 + lx - 3I +21x - 41
,

11 + lx -31

(e) y=

6. -(8)

Show that if

> b > 0,

,ja

+ 31x - 51

+ 2k then

ab
<
a + b
(b)-,

ThUs, show that for positive numbers

6 < min(a,b)

7.

(a)

1..)

ShoW for positive

and

is satisfied by

a, b

b,

the conditon,

ab
+ b

that'

a 4, b
< max(a,b)
2

-a

a ,L b.

/.

(b) 'Prove for all

if

a, b

that

max(alb). = -(a + b + Ia - bl)

(c)

Prove for all i, b

that

min(a,b).=

8.

2a

+ b - Is - b1).

Shows that

max(a,b) + max(c,d) > max(a + c, b

9.

Sh w that if

ab > 0

then
ab VMIn(a2,102).
,
.

a = max(a,b,4,, then

81
b
1

82,

"b2 ' .. ' ban }

%r = 1,

7
a

min
r

2,

( r

by

...,

and similarly for max.

ur

('
,

prove that
... + qi
2 I

1.
.45.

min
'r

a1 + a

-a = min(-a,-b,-c).

a
n ,......

+ b

..

+ bn

:I

max
r

_Lb
r

If

ttj

77 6
1
I

1
n

;:

13.. (a)

4-

<1

or

+22 +12

+ (n - 1)2

Prove directly from the properties of order for

,-E- <

<E

lxl < E.

then

Ix

Conversely, if

Tor
14.

is an element of an ordered field and if

all positive values 4,t then x

Prove that

(a)

then6.

Prove that if

(b)

> 0 that if

<E

- <x < e.

n = 1 ,2 , 3 ,

(b) 'Prove that

,

lxl

<E

= 0.

f,

tab! = la

lb

It-1 =

, 14 /

O.

y I < lx(

Ix

15.

Prove that

16.

Under what conditions do the equality signs) hold for

Ira'

17.

If
x

0 < x < 1,
to obtain

x4 < x3
:

then
18.

la!

we can multiply both sides of the inequality

x2 < x

(and, similarly, we can show that

and so on).

ix

Use this result to show that if

x3 < x2 ,

o < 1,x1 <1,

+ axl < 31x1..

Prove the folleiwing inequalities:

1

(a)

x +

(b)

x+

(c)

Ix +11

Prove:

> 2 , x > O.
x< 0 .
> 2

x2 >xlxl

Show that if

Ix.-

;C/ 0.
for all real

al <

x.

then

<

'al.

< 1

Prove- for positive

and

b,

`for all

312a

here

a /

a / b,' that
ISA

- al

(a + b)

<a

+b.

r-a< lb - a,

847

777

Ot.

377.

by

. ;

Definition`of Limit of a Function

A6 -2..

We have used the concept of limit of a function in defining derivative.

formulation of the limit concept and derive

At this poin t we present a precis

the laws which govern operations with limits.

.

Although the concept of liMit'of (1 faction is more general than the

,-

idea Of derivative, our study of limits was initially motivated by the basic
.

r(x),

as the limit of the ratio

example of the derivative of a futction* 0

which- cep be written'
e

m =
?.

lim

r(x),

x -, a

as

m = lim r(x),
x-:1

where
r(x)

0(x)
x

0(a)
a

In order to be sure that the description,of the deriVativlias the limit.,of

r(x)

makes sense we must be sure that we have an ahequate set of approxima.

tions, that

r(x)

is defined for numbers

UsuallYmotte dorLin of
a

(excluding

itself)

arbitrarily close to

will contain an entire neighborhood,of

a.

but either for theoretical or practical reasons it

is often useful to analyze the behavior of

r(x)

a.

on only one-side of

For example7there is a natural, starting point in the motion of a rocket and

it is essential to know the initial directiop of the rocket in order tojeter\

mine -the rest of the-trajecto

,

In framing the general, efinitio/ of the limit of a function

a

poirit

at a

we theilrquire th t we hake an adetuate set of approximations.

Specifically, the definition marnot include the valuefka)

among the

approximations, even if it shouid bOiefined, but it must involve values

for

f(x)

close to

lneighborhood,pfl

a,

a,

For this purpose we introduce the deleted

x for welch

that is, the set of all

0 < Ix - al <

As the set of approximations to;-be used in aefinj.ng,the limit of

f(x)

we take the set of values

deleted neighborhood of

for all, X

from the domain

at 'a
in some

a.

In some texts this important case i

tions of "right - sided" "and "left-sided" 1

778

en care of by separate
(See'Exercises A67:4

its.

efini-'

No. 16.)

1'

378
.

A5 -2

1
With these ideas in mind we are now able to express the Urea of limit
completely in analytical terms.

If

then for any error tolerance

we keep

ing

has a limit

f(x)

x to be any number from the domain of

neighborhood of

within

Let

has the limit- L at

6,

of

L by.restric.C-

in a sufficiently' small

be a point'for whiCh every deleted

neighborhood contaille points of the domain of .f.

number

a,

e.

DEFINITION A6-21.

as .,x ..approaches

The function;

if (and only if) for each positive

there exists a positive number

such that

If(x) - LI < e

for every .x

In the domain of

We then write

which satisfies the inequality

aj < b.

lim f(x) = L.2

x-a

4. l`

It follows from the definition of limit, since the value

itself

f(a)

does not lie in the class of approximations considered, thatany function

which takes on the same values as
. would have the same limit at

a.

in some deletea,neighborbood

For example, the two functions

defined below have the same limit at every point

of
f

and

a of the real axis.

"t

f(x)' = 1

g(x)

0 ,

'or any integer

fpr non-integral

x,

.
,

x.

Although we do not rei,y upon pictures tor our precise understanding,q,,

't4 concept of /iMit, it is desirable to have a geometrical interpretation ol'

the idea.
I

ample A6-2a.

The graph of the functi n

f

: x-.2x - 4

6
The.definition of limit can be recapitulated in terms of neighborhoods:
the nusib4 L is said to be the limit of f at a if every deleted neigh: borhood of a contains points of the domain of f and if for esich
e-neighbothood of L there is at least one deleted 8-neighborha% wherein f
yaps the points of its domain.intio the e-neighborhood,

We shall now make use of this =tation, rai erthan-

lim

f(x) = L.

is shown in Figure A6-2a.

In order to show that

Iim .(2X - 14) = 2

x-3
we must show, for every e> 0,

iP
that there, is a

It is easy to see

0 < ix - 31 < S.

in the deleted neighborhood

from Figure A6-2a how

.0

4) - 21 < e

1(2x

for all

5 > 0 iso 'that

may be found.

E ==1
1
E

= 2

J-

-1-' .
4:3

ins

1
3

Figure A6-2a-.
$

Given a horizo

al band of width, 2e

find ayertical band of width

26

y =2,; we can

centered on-the line

about ,x

= 3 so

t the graph

lies entirely within the rectangle wliere the bands ov rlap.

v
';'".

we.,;,infer that for

1
,z

e =

,,5 =

/p.

es,--$

we may take

'6 = -32.:,

fon-

e =

t,

of

From the graph

5 =

There seems to be nR obstacle to..finding a

4- / and

far

S 'for axe'

no matter how. small, .but we cloexly cannot relk..,on pictures to do so.

,

Instead, we proces. analytically.

If we require

.0.'<

lx - 31

.<-'8,

then

4k,

1f(x) - 21 = I(2x - 4) - 21
1,2x - 61

= t2(X - 3)1'
= 2Ix - 31

<28.
Consequently, if..we take

e
,

the

If(x)'- 21 < e.

The preceding example was made especially simple to reveal the basic
picture.

We 11mi-explore the concept of limit in a variety of situations.

Example A6-2b.

gt4en by

Figure A6-2b preseuta

f (x: = sa (-1.)
1
k!
,7-11
x2 '

a graphs_ofthe three functions

x?:f 2(x)

f (x)'

41'

Y = f (x)

Y = f (x)

"x

.0

er

sgn(-L)

?c

tgn lc

41

;;,.

11

'

,y,= i'(x),
t

'0"

f33

4,

Figure A6.121:.:"":
*Cr,.
e",

'1;

sgn

if a > 0
if a = 0

if a <0

'

1'
1:!3:

1'-Ottserve that x = 0
,

1'

is a point of the domains of

but not of
3
For each of these functions,we. wish to consider the limit, if it exists
f

0.

as -; approaches

and

ince the threi functions 'coincide wherk x

clear 'that all -three functions, have the same limit.

obvious candidate .,for the limitl,

and that the graph of.' fi

btft

f2(0)

is

the

Verify that the conditions of Definition AC-'

Observe - hat -there is a gap in the graphs of

has a limit at

In each .case

the

it is

x = 0,

at ,x = 0.

L = .1

and the value

0,

limits does hot depend on how the functionq are defined at

are _satisfied by

and

is continuous, it has no gap.

bu-5 is

is not its limit,

f2

not .defined there,

f3

has a limit at

at

The function

= 0,
f*
1

is defined at sx = 0

x = 0 and the lirnit4iis

the function value.

Example
-

Figure A6-2d presents the graphs of the two functions

given by

= x.- .+ ..lsgn(x - a)

= x2 + sgn

17=-7.

x -4 x2 + sgnix -

782

'3 8-2
A

Thei function

is defined for all values of

sists only of those values of

the -same values as

number

for which

xt The domain of

eon2
x > ai and on this domain it has

g1. _It seems clear'froMthe graph that there is no single

L which is approximated by the values

g (x)- as

x.

On the contrary, in any neighborhood Of

approaches

a.

it is Possible to find valuesof

'

approximates .

for which,_ g (x)

.J-

- 1

2
a + 1.

other values whlch apprdximate

within any given error tolerance and

Verify; then, that the conditions of,

refinitionA6-2. cannot be satisfied, that

i.

..,

For the function

g2,

has no limit at x = a.

On the other hand, it appears that no matter what

the error tolerance, there isa deletpd neighborhood of

approximates

wherein 826),

+ 1

within the tolerance for all, x in'the domain_of the

41
function.; ThiS is easily verifi/id. in a 'deleted sb-neighborhood of
we have
e

,g

(x) = x

+ 1,

for

a < x < a + 5.

'We have forthe absolutt error of approximation

1g2(x)

(a2 + 1)1 = 1x2 -

a21

- al

Ix + al

<5(1x1 t
A

< 51(lal + 6) + 'all

'

< 6(21 al
-

6).

Tiais absolute error cah be kept within any given error tolerance
reatricting

to a small enough 5-neighborhood of 'a.

'by

For simplicity, we

first restrict ourselves to neighborhoods_of.radius no large'r than-

< 1

in the inequality above, we obt

in terms of the radius

Taking

1.

n a simpler bdund on the absolute error

6:
.

I82 (x) -

Nbw if we choose

(a2 + 1)1 < 6(21a1 + 1).1.

o that

- .

then we have ensuredRthat-

c.

ft

1g2(x)

+ 1)1- < e,
al

783,

"3 8
O

namely, that the error hb.s been kept within the tolerance

e.

a proscription for controlling the error uLthin any tolerance

accomPlashed our liurpos* and proved

r.

completely. i

e, we,hAve,,

VP'
2

lim g2(x) = a
,

Since this is

A.

the analytic terms oi.befiniiion A6-2.,

I
4

784

4.

Exercises A6-2
Show that if , 0.< lx - aj < 1,

2-14

0 < lx - al <1

2.

Sh Ow' that if

3.

Show thatAfki0 < lx - 21. < 1,

then

1,x + 2a1 < 1 + 31a1

then

1X3 - a31 <,( 31a21 - 31'81

then

< 1.

It!
1

Hint:

- 41

4.

Show that if

5.

Show . that if

then

1;

< 1.

lx

.1 , then

lx - al <

,0 < Ix - 11 < 1

then

14x + 11 < 9

and

1-4--1 < 1.

, x + 2

6. 'Show that

0 < lx .- 21 < 1,

then

lx + 11 <: 4

.. lx

+ 2x + 4

x2 + 1

-;

,t

7). Estimate how large

8.

< 1.

'

interval

'i

ana'

can become if

is restri,cte& to the'open

-3 < x < 1.

Use inequality properties to find a positive number M such that

O < 1,x - 11 < 3

9.

for all

1x2,+ 2x +

(b)

13x2 - 2x + 31 < M.

Sho

that if

and,

Ca)

(ak)

p\
0 < lx - 31 < 1

and --z0 <'fx

< E,

pair of inequalities

then

-=

Show that t

31 <

< I

and

(or

S .<...5

",o0".%
,6

10,

minfl $7
!))

satisfied by

Find a number M >.1, .such

li-cc141

< M for all' sx

sucithat

O < Ix T 21 < 1.

(See No.

For the given value of

11.

(a)

E = 0.1

(b)

E = 0.01

,rs yoniechoice of

lx2 - 91

f nd a number

E,

O < lx - 31-< 6,

)'

such that if

<

in (b)

a ceptale as an answer in (a)?

85

.385

Explain.

2 --:

,:

For the following functions, fibd the limit

12.

A
For each value, of

.
.

.4

whenever

Ix - al

< S.

1f(x) = 3x - 2,

a =

(1:)r

f(x) = roc + b,

(m

x ',approaches'. a.

as

such that,
.1

,(a)

CO. f(x) =,1 +"

r.

exhibit a number

e,

If(x)

-.LI <

'.

1
2

0).

a = 0.

tl

e,

. .

''

A6-3. ,Epsilonic Technique

It is conventional 4.1.1 discussions. of approximations to -a limit to use

.
.

the Greek letter epsilon for the error tolerance.

For this rea on the sub-

ject devoted to techniques for the control of error is colloquial /- called

ep-silonics.

We shall make immediate use of epsilonic technique in deriving

Ahe limit theorems which follow this section.

Eventually, in applications,

skill in epsilonic technique,,,will be extremely valuable for making estimates

when it is difficult to work with precise values.

.

Tp develop this skill it

fs helpful to set up a routine pattern in which to'present an epsilonic argument.

We shall first describe the pattern in general and then, for several

examples', carry out the proofs as indicated in the ;pattern.

Statement of the problem.

To'pro've that

lim f(x) = L:
x-a

For each tolerance

Show:

if

e > 0

obtain a control

0 < Ix -fal < 3,

then

8.

If(x -LI < e.

We have stated the problem in outline.

The proof is based on Definition'A6-2.

within the error tolerance e by
restri,ctirig the values of .x, to a sufficiently small deleted neighborhood
'of a. The p of '1.1.completed by verifying for a imitable radius
5
that
i
does give
e desired degree of control.
The crucial openquestion is,
how do we ch oseNa suitable'

We must coktrot the err If(x)-- LI

SimgffiLtion.

Step 1.

Find a g(6)

such that if

0- < lx -.al <'3,

therk,'If(x) - LI "< g(5).

.

The idea here is i'eip obtain an upper bound g(5)' for the absolute error where
g(5)
can be'held within the.tolerance e by taking sufficiently small values
`of 8.
If we have g(8) <E, then If(x) - LI < g(5) < e andoour objective
is achieved. In some of the following examples the work of simplification is.
divided into-bigee stages:
(a)
f(x)
is expressed in terms of x - a; 1
(b)
from the Alwquality.. 0 < lx - al < 5 there is derived a inequality of
dbe form. If(x) - LI < g(Of
(c)
a b
is chosen for each wE
in such a way
(that 'g(3) < t, In general, g is to be a simple function; one for whiqh it
is easy'to find a 5 such that g(5) < E. More typically,it will even be
.posaible to solve for 5. in tIcie equation 'g(E)) = e. For most of the cases
in this text it is possible to obtain g(8) =,c3
a pqsitive constant-of'
proportionality c. Mapipulations yielding a simple expression for
g(5) are.illustrated in the Examples below.'
4
Step 2.

-.Chooae

Choice of
5

5.

so that..g(8) < e.

This.ts the place' where the work.bf simplifi ation in Step 1 pays' off.
e
LI
,

7t57

3.87

In

we m a y chockse

the most typical c a i e where, g(8) . ct.

8 = 5- .
,

Steps'l and 2 show how the solution is found.u.. The next step is the
r

actual proof where we verify that the solution, hap been found.
.

Step 3.

Verification

Return 'to the ,statement of the problem From the given expkssion for
deduce the conclusion.
1st

Fir

we try out the methodrin a case where no complications arise, the

case of the general lineal function.

Example A6-3a.
tt

Statement of the problem.

(m

To prove that ,lim (mx +.b) = may+ b,.

'For each
Show:

e > 0

obtain' a

8 .

then

0 < Ix - al < 8,

if

0).

vx-a.

If(x)'- LI
.

step 1.

Simplification.
)

f(x) - L = (mX

(a)

b) - (ma + b)

m(x

(b)

a).

Ix- al < 8,.

If

If(x) - LF

a)1

- al

< Ind
(c)

Take

Step 2.

g(8) = 11118:

Obtain
4

To make

g(8) < e,

set

e.
ITT
( allowable, singe

Step 3.

Iml / 0

assumption) .

'Verificatio

Enter the result

in the statement of the probl

m.

The verifi.

cation follows the pattern of Step 1 with one additional step

;
788
:23 16: 8

"

If(x) - LI < Im16

''

< 1ml -4-4.

11m5

< e.
Since there is a strong inequa ity in this chekin,, we have
lf(x) - id <
1

In the following examples we shall omit repetitiou's material.

'EXample A6-3b.
Statement of the Problem.
- To prove that

lie
x-0

For each

e >

Show: if

- 1.
.1%

obtain 'a
-

S.

Q1 <

then

1 + ixl

< e.

Xf

Step 1.
4

(a)

1 +

l'x1

3.

+11,z1-

1,
1
+

Ixl

lx1
1 + Ix'

(b)

If

of < a,.

0 <

-1
11

11 +1 'xi

-.1x1

11

Ix'

ix'

1+ ix'
(since/ 1 + 1x1 > 1)
< S.
(c)

Take

Step`2.

g(6) = 8:

To Make

g(b) < c,

set

,Step 3.
ti

Set

8 = e. in the statement of t e probleM.

verificatioh folloWing Step 1 where we set, 6 = e

769`

We carry out the

at the last line:

The next example shows that it i-s--nOt_always sufficient to choose

proportional to ,e.:.
Example A6-1,3c.
V

Statement of the Problem.

Tb.prove that
For each
Show:

'( a ->,0)

lim 17.=
x-a

e > 0 obtain a

if

6.

0 < Ix - al < 6

IA

then-

15.1 <

where c is a poSitive constant, cannot

The choice 5) =
work when a = 0. fil taiat case we observe that if 0 < x'< 6 = ce,
..
then iTc < 1/E "T.
1.

We must then ma.ks

small.
.

c,' no matter how

for all

IC Irj.'< c

ft

_
;t-follows that we must find a positIVe number
otlee
e < e forailfpsi
iC. < VT or, equivalently,
number exists; hence, 5 = cE cannot work.

'

Step 1.

e.

satisfying
No such

From
< PT(

11/7 -

we obtainlon.multiplying by

(Section A6-1, Formula 3)

,h71

< I x - a},

.113-c- 4=71
,/

whence

Thus, if

0 < I

0,]V7

Step 2.
Step 3.

Ch oose

Take

<

b = e 2 .
)

6 = e

5.

in the statement of the problem.

is a mecapitulAtion of Stepl'llor this choice of

Step 1.

6.

by an ctuxiliary condition in

It is often expedient to restrict

The verification

The following example's are"tepical.

7

,"

4t.

Example A6-3d.
.

' Statement of the Problem.

Ar----77---

If

'z

To prove that' lim (Z' - 5x -,1) = -3.

1

w 4

:',
141

r.-2%

;"

!
4

.1

790
.
.

..

0. 7.

es,

*,

,
V

)
*11W

',,

3 (A
4

A6-3

e > 0 obtain a .5.

For each
Show:

0 s

if

.<

'then

(x3 - 5x

< E.

1) - ,1-3)

1.10/111

Step 1.

x3

(a)'

5x - 1

x3

5.x +2

[(x - 2) + 2f

x - 2) +
.

= (x - 2)3 + 6(x - 2)2 + 7(x - 2).

(b)

1x3

1 - (-3)1 4---,,1\(x

1(x

= Ix
<

2)2.

+ 7(x

2)

2)( (x - 2)2 + 6(x - 2) + 7)1.

1(x

- 21

52 + 6s +. 7)

Ix

(At the last line.we used,

For convenience' we restrict

2)2 + 6(x - 2)

(1x -

1-X - 21

<

'(c)

+ 6(x

2),3

71

212 + 61x - 21 + 7)

IS

.--

- 21 < 5. )l

by.lqqi-ring '5 .< 1.

Under

this condition
- 5x - 1,-

+ 65. + 7), ".*

<%5( 1 + 6 '.;--

1.

i,),,,.

< 145

In order to git tin upper bound in the simple form. c5, we put a constant
bound on the second factor in
particular value

in

S < 1

5(52 +.65 + 7)

1C

tricting

5.

(The

, e could have requiebd

is inessential.

5 <.K. where

by

is any positive constant.)

/

Step.2.. We now wish to obtain a value

/satisfying two . conditions

.
.

simultaneously:

5 <

-5-1

and

5.< 1.

One way of satisfying these conditions

,i; to set
-4-

'

where, we have chosen the denominator simply as a c.onvenient value which -is

greater than either 14 or c

Step 34 Set
tion

follows

( See Exercise, A6-1, No. 6a, b.)

S =

the statement of the problem. 'The verifica.

Step 1 through (b).'

In (c)"

we use 5< 1

791

391

and

A6-?3

t
6

s<

to Atain
= (-3)1 < E.

,1(X3 - 5x,-

Alternative Step 1.
(a)

.x3

- 5x - 1 - (-3) .= (* - 2)(X2+ 2x - 1) -

Ix3 - 5x + 21 = lx

2
-

21

+ 2x - 11

'

<'51x2 + 2)c - 11

3.45

6 < 1

Where, at the last line, imposing the condition

we

obtained, from.Ix - 21 <5 < 1.

1 < x < 3

utilize the result

-

'

Since we.do not use ths.foiMula for

..,,Alternative Step 2.

verification above tut only the conditions' S < 1

'

and

in tie

5 <S',.it is natural

(A6-1) to.' set

= minh- ; 1).
.
,

Set 5F min(

Alternative Step 3,

iri the statement of the prob-

,l)

The verification follows alternative Step.16aboye.

lem.

From"thespreceding example we see that we have great freedom in choosing

our control

b.

We can alwaysilise more stringent cOntrols,than necessary:

that is, given any deleted neighborhood of

< E

lf(x).-

for any

al < 5,

Ix

So chosen that

in the' neighborhood, then for all

x'

in any sub-

set of the neighborhood and, in pairticular,-for any smaller deleted neighbor-

In other terms, given any S

we satisfy the same inequality.

hood of. a,

which keepsthe error within the specified tolierance,'any smaller value of

follows that we may impose the condi-

will certainly have the same effect.

K where

tion' 5
found a

any larger

is any convenient positive constant.

for a particular
c.

E,

we know that the same

5,

Similarly., having

will suffice for

Hence Sae need concern ourselves onlyith these

satis-

where M is any convenient positive constant.

4
(
I
We.conclude the li st of examples by applying the techniques of the out-

fying,

E < M,

line to fina,some derivatives.

V

r(x),-

For a given
'f(.x)

f,

,; :ba)
,

x - a

we Let

x / a.

&ample A6-3e.

x:41 ,x/ 0

atertftt of the Problem

prove for

a / 0

that

1
a
a

lim r(x) = lim x

'4

X^11" X

,,

1
2

' For each

e > 0

9btain a 8.
#.4

ShOW:

if

0 < IX 7 al '< 6

(0b6ervethat

r(k)

then

Ir(x).- LI

iribt defined at

e,

x = 0 ,or

=fa:)

Step 1.
1

L =

f(x)

3'41

x - a

.....-

.1

= -

+ --

(x

2 '
a

ax

ax,

(Note that we used

Ix - al > 0

- a

in settIng

x / a.)

1 .for

.1

(b),

(1)

air problem now is to obtain a Constant upper bound for the factor
1

It is sufficient to bound the demonimator away'

21x1

a' 1(x - a) + al

from

0 or tit guarantee

''

,
- a). + al > C >0, ,

Ixl =

for some number

C.

We hale. (Appendix A6-1, Formula (3))

= 1(x

"a)

4'

al

793

1 al ,- Ix- al--

1-

393,

a)

in this relation, we obtain

al,< 5

To obtain a constant lower bound

IX

al > lal

Ix

IX'

we restrict

- 5

"> lad

S.

5*

l<

In that,case

>-11-'1"- >0

and Cn121

It follows from

- al

1-x

6,

>

lxk 1

that

lal

and

1
7T
<
lx

Exercises A6-1, No. 20).

2
.

(See

1111

Consquently, 'frOm (1), we have

'Ir(x)

28
a !XI/

LI

<s

all

<6
18-J3,

Step 2.

The value of

is restricted by two conditions:

and* -24 <-

5 <

To katisfy -both conditions we take

5 = min(
'

=1
'

rt

*
Of course,,
way so that It,";

in general, we could have restricted

lal

For definiteness we took

in any convenient

...-

,I

5 <12.1- 2:
2

r.
lo

719:1

4
---

Step 3.

Enter the above value of

in the- statement Of-the problem.

E)

At the last
line we use
.

:The verification follows the pattern of Step 1.

ILL!
E

Ir(x) -;LI 5'E.

Example A6-3f.

x > 0.

ITC,

Statement-oethe Problem.
a > 0

To.prove fo

Fdr ea h

- va
, 1
x - a

.--V7

lim r(x) = lim

x-a
x-a

thdt

c >,0 obtain a

Show:

if .0< ix - al < 6

(-Oise

e that 'r(x)

21e.

= L.

b.

then 1r(x)

is defined only for

LI <

x->

Step 1.

r(X) -

(a)

a -

(Note that

we guarantee

17

is not defined for negattVe values, and there ore

0 < x

by impodlig the restrictions

For thig purpope we require

S < a.)
-

795

Ix

al < a.

710 ,R,r71,11.1^. """"

Ir(x) - LI

(b)

-t

a- 4

24.(4. + VR)2
lx - al

a.-

(from 'lx - al < 5),

M( /71 + "IX)2
$

> 0)
a

24:(1.g.)2

r
2(4.)3

5 = min(26/a0e,

Step 2.'1- Take

Step 3.

defined

For the above value of

or all

requiies 1 x / a
last line we use

every expression used in Step 1 is

'5

in the deleted 8-lheighborhood 0< Ix -, al.< 5.

At the

The veriffcatiOn.follOws Step 1.

x > 0%)

and

8 < 2(4Ti.)3e, to obtain

I r(x) -

(This

Ll .< 6,

0')

In the preceding examples we have not always followedthe

outlthe to the letter but used it only as a serviceable guide.
Special difficulties are likely-to appear in Step land we cannot
The only absolutely general pattern
anticipate all contingencies.
the construction of a non-decreasing Chain of expressions.

where

460 < 01

00=, Ir(x) - LI,,

02 <

= g(5)

'

< con

and Op.

2'

...,'n-1

may

involve both x and 8...To construct such a sequence in a particular

case may require the greatest ingenuity.
In these examples we have verified that a given value
limit but have not shown how the limit

I8

actually the

L was obtained., In the next section

we shall develop general theorems which will enable us to discover the valUe.

of the limit and tdprOve that the value is correct.

,

Epsilonics will be

.neaessery only to'prove the theorems, not to apply them.

7963(,)(i.,

A6 -3

D.

Exercises'A6-3
.

Prove lim

1.

g x - 3) = -1:

(.A

''x44

error and find

', k

obtaip-sa_upper bound g(b) for the absdlute

.
---..----;6 rin,,
terms oI' c.
-------......
,

-..-----...............

A'

Gi7ve arguments the' prove:

2.
a

fa)

lim c ='c,

any constant.

'

x..a

lim x,. a.

(b)

4
-.-.

.,

x..a

'

,(c) ' lim kw = ka , k

.,.,

any constant.

'

x..a

;,,,

.4

Oise the reults of Example A6-3a for parts b and .c.)

3.

,-

Invoke the definition directly to prove. the existence of the limits in

.Nun6ir 2.

-,-

4.

In each of the following guess the limit, and then prove that your guess
it correct.

(a)

'

(e)

2
x..0 1 + x

lim

40.

lim

2tx
,

lim x
x -.2

- 8

=.3)

x- 3

(f)

lim.x

(g)

lim

3x
+ 2

- 3x - 1
x + 2,
.

797

3,97

Oh

777T
A6-4
s.
4'

.#

A6-4.

Limit Theorems
If the epsilonic definition of limit were required in every calculation

with limits, the development of thecalculus would be so disjointed and so

overburdened with elaborate detail that it could only be mastered by a few
devoted specialists.

We need and we shall derive theorems that broadly cover

In the end it will only be

most of the significant calculations with limits.

theexceptional cases for which epsilonic techniques are necessary.

The firgt general results apply to rational combinations of fpnctions,

that is, expressions formed from the functions of a given set by the rational
o

operations ofaddition, subtraction, multiplication, and divisi2n.

approaches 'a,' then the limit

function of the given set has a limit as

If each

of any rational combl:nation of these functions is the same rational combination

of the corresponding limits (with divisions by .zero excluded).

There are certain special rational combinations, called linear combina7

tioni; which rew often in different contexts.4. It is worth distinguishing

them as a class because of their importance. 'A ,linear combination is built
A

up by .addition of functions and multiplication of functions by constants.

Such g linear combination can be put in -the form

0:
where

c.

2?

'

+ ,cifIfn(x),

0(x) --..cifl(x) +.c2f2(x)

c
n

are constantt., In particular, a polynomial of

'can be written in the form'

degree less than oggtlual to

;,4(

O(x) = c

+ c x

+ cnxn'

and may therefore be thought of as a linear combination of powerslt, x,

2
x ,

The evaluation of the limit of a;linear combination is an instructive

instance o

the general method of evaluating the limits of rational combina-

tions.

Example A6-4a.
,

lim (61/3 + 5x +
x-4

7)
''

= lim 6V3..+_lim 5x + lim it

x-4

x-4

x-4

. tlim 6)(lim 16) + (lim 5)(lim x) + lim IT

x-4

x-4

x-4

x-4

et,,

= 6

lim 1/7<;+'5

x-4

x-4

.r,-.1....

4.4.

798

`,33

liM x + g.
x-4

:I
,

HA-4
sed three limit theorems without proof;

Note, that in the example we have

in essence,these are:

..

(111

The limit of the sum of two functions is the stim of the limits.

(2)

The

of ,the product of t o functions is the product of the

limits.,

of a constant is that constant.

(3)

Copsider'the statement

)'

liM c = c.

>

x-a
Vote that the interpretations of
are slightly different.

on the right and left of this equation

On the left,

stands for "f(x),

where

is the particular value assumed by the function for each

and on the right

Value,of

x. .With this in mind we have

For a constant fuhction

0011.124A6-4a.

f : x -4T,

"lim f(x) =

xa
proof.

We have

K.

= It

f (x )

cl = 0 < e,

e.,4ind every choice of . .(The constant function is a

for every positive

trivial case, of. Course, but we include it for completeness.)

If

TtE0ElitA6-4b.
.

then"for any Constant

lira f(x)*= L,

lim c f(x) = c lim f(x) = cL.

x -a
x-a

O.

;
Eroof.

c / 0,

We max assume

1%.

that of,Theorem A6-4a.

Given any

'

by restricting .x

c,

xa

lc

for if
e > 0,

c = 0,

to adeleted neighborhoo

SN

0 <

799

the problem is reduced to

we 'Wish to make

cLl

43 f3 9

From the'hypothesis we know that for any -d

we can find a

so that if

0 < lx - al <S ,
then
.If(x) - LI < E*,

lc f(x) - cri = Ici

Accordingly, we choose

T-T, obtain the

Lr <

ppropriate value

for this

*
E

and set

6 = S

In the following theorems we require that in some deleted

neighborhood of a the domains of the functions entering the
combination all coincide. This requ'irement eliminates nonsensical Combinations Oath as f(x) + g(x) when, f(x)
is defined
only for x > a
1 g(x)
is defined only for x < a. The
likelihood of eve , oAking such' s. mistake is extremely small and
therefore we do not mention this restriction on-the functions
.explicitly'in the statements or proofs of the theorems.

If lim f(x) = L and lim g(x) = M, "then

THEOREM A6-4c.

x-a

,
lim [f(x) + g(x)i = L + M.
x-a
,Proof.

We must show.that for any given

E > 0

there is some

such
/

that

if(x) +g(x).for all

fn the common domain o

Iq

M)1 < E

and g

Satisfying

0 < lx Dal < b.'

,-

From the hypothesis we know that for any positive

small, we can f

d 51

and

If(x)

such that

L11 < el' when

< e2

when

no matter how

2'

0 <>Ix - at < 81,

1`

Ig(x) k-

and

-1

0 < lx - aJ <

But
If(x) + g(x). - (L + M)l = If(x) r L + g(x) - MI

< If(x) - LI Ig00

1

,soo

400

141

-----

A6-4
,

1:

To leepiwithin the tolerance '-we can choose

tive quantities whose sum is',,.

to be

and

nt' posi-

conveniehge,-we fiX

Fo

El= 62
..-

:.
,.

Taking-Ithe apprOp;iatd values ,o

..

1! ;2

:-

.for these values

we set

l'

5 4 min(51,,52)

5, whenever

For this choice of

<5,

0 < lx

then
If(x) + g(x) - (L +14)1 <

<-e.

+
(

$ince a linear combination can be built lip by succes ive operations of

4

addition of two functions and multiplication by a const

,. we obtain,

-/,,,z

//
ctions is the same

The limit of a 2ihear combination- of f

Corollary,

if

linedr combinsftion of the limits of the functions; i.e.

= 1* 2 ,

lim,f (x)
a
then'

lim[clfi(x)' + c f *(x) +
2 2

+ cnfn (x)]-= c

x-a

lim f (x) + c
1

x -a

+ ... i. c

..

lim f (x)
2
x-a

lim f (x) =.c11.1 + c' L + ... + cn n.

n
2 2
x-a
._ _.
-.

The proof is left as an exercise,-

For general rational combinations we have the further operations of

multiplication dna division.
r

Example A6-4b.
.

lim
x-4

- 2x

r.

'

,fat]
-= lim
xL4

1.

- (lim2)(lim x2)(lim V)
x>

1im.x
x-4
=

-2

801

.r

2(lim x)(lim x)(11m VX)

X-4
x-4
x.4
4 : 4

3,4

4 01

For .0(x) = - - 2x
simple steps.

I
vx

let us see in detail, how

can be built up in

We set

1Tc,

(x)

f2(x) = xf (x

(multiplication)

1
.

f3(x) = xf(X):

(multipll.cation)

f4(x) = -2f3(x),

(multiplication)

I
,

...

(x) I :g 16

(division)

:"FT37c
2

.. t
...!.

w4ere

.
and

gl(4 7

k(x) 1= x

'

and then,'_

(additio

O(x) =f4( ) .4-'f5(x,

It is, of course, tedious and unnecessary to decompose axy ratio al .,combinetion into its elemeatary'building blo
it can be" done and t8 know\ how to do

s; but'it is important to realize othat

t.

(For example, it would be necessary

to do so in wAting computer programs ) In the process we have een,that to

04
prove the general theorem concerning imits of rational combinati nt we no
need to prove only the two special theorems for the limits of the product and,
quotient of two functions.
'THEOREM A6-4d.

If

lim f(x) = L
x-a

'and

lira g(x) = M,

then

x.a

/1 c
N44) lim if(X)-"; g(x))

x-a
Proof.

We with to estimate the difference

knowledge of the differences

iis.

f(x)g(x) - LM,, using the

f(x) - L and g(x) - M given in toe hypothe7

Nor

f(x)g(x).-, LM =(f(x) - L(x) + L(g(x)

=

(f(x),- LXg(x)

M)

M)+ M(f(lc) -,L)+L(g(x)- M);

s't

8402
kw,

hence,

(1)

If(x)ex)- LM1 <If(x)- LI

lex)

MI

If(x)-1,1 +114

IMl

Ig(x)

MI.

From the hypothesis we know that fOr any positive numbers

-Nand

are corresponding controls

and 62

61

there

2Y

1.

such that

If(x) - LI < el

0 <lx

for

al <
l'

Ig(x)

Thu's if we Choose

0 < lx - al < S

- MI <,c2, for

min(61,62),

0 < Ix

al <

it will follow from (1) that when

then'

If(x),g(x) - LMl < cic2 + IMIcl'+ 111E2.

(2)"

In order to keep from exceeding the tolerance

and

E2

we shall choose

c
1

so that

cle2

IUIC1

./

C;

11'1E2

....

this will then deter

ne our choice of

6
1

For convenience; we

itie are now ready

equire that

o choose

oose -ihe corresponding

a. +

are themselves products,

p(x)

and that

LE

and -6- and_let

0 < Ix

al < 6

f(4g(X) -:11q1 < V(1 +

Since a polynomial

2'

cl = E2 = V

1,

and in turn that

and verify (3Y.

V :r2 min

ollows from (2) and (4) when

and

Li +

I.
0

6,

Then

Let

I MI

6 = mint6

Then it

that

IMI ) <

6..

is a linear combination of powers, and powers

ro

. xk

(k

factors,

>1),

we can establish the foll&ing corollary.

'For any polynomial fundtion

Corollary.

lim'p(x) = p(a)
x-a
e

The proof of this corollary is left as an exercise lExercises,A6-4, No. 2).

of ,
f(x)

TO prove the limit theorem for a quotient

prove the limit theorem; for a zeciprocal T57.:

it is only necessary to

The rule for general quo-

tients then follolLfrom

,f(x)
iTR7-

f(x)r

-P
iTR"

First we prove a useful preliminary result.

If

Lemma

hood of

a 'where 'g(x).> 0 for

Proof, .Since
of

lim g(1
= M and' M > 0,
.

a Wherein

g(x)

then there exists a neighbor,

in the domain of

has the limit M at

a,

g.

there is a 6-neighborhood

M 'than to zero:

is closer to

5
.

In thid neighborho45d,

=324- >g(x) >

If the function

> 0.

has a. negative limit at

x = a

Lemma A6-4 to the function 4, we see at once that

deleted neighborhood of

a.

then', upon app criTig---

0(x)

is negative in some

As further consequences of Lemma A6-4 we have

the following two corollaries.

Corollary 1.

If

lim g(x) = M and

M /00 then there exists a neigh-

x-a

borhood of

a where

Corollary 2.

141 > I g(x) I

for

>

in the domain of g.

A limit of a function whose values are nonnurive.is non-

negative.

801.

404

A6-4

The proofs of these corollaries are left as exercises.

THEOREM A6 -4e.. If

Exercises A6-4, No% 3)

lim g(x) = M and M A 0,, then

x -a

*lim
x-a

M.
A

Proof,

We have

14

(2)'

?x(

*er

ILT.1-8011
provided

g(x) / 0.

&neighborhood of

However, from Corollary 1,to Lemma A6-4 there is a

a

wherein

Ig(x)I .>-'14

Furthermore, for any

.*
E

N.

the
,

neighborhood can be taken so small that also

\

>

, From (2), t

Ig(x) - MI < e*:

refore, we have
1

irRT

RI
*

imr

114

<

2E
e

M2

where in the last line we have taken

t

*
E

M2e
2

To complete the proof we choose the value of

Corollary 1.

If

appropriate to this

L and liT g(x) = M. ,where M / 0,

lim f(x)

x-a
I-

x-a

lim=
f(x)

x -a

ir(7)-

L
-14

805 A

r.;

then

Corollary p2.

c.

If

are,DolynomialA and if

and

106

4(a) /, 0

then

.42(a) 'e
...,P(x)
q(x) -.-Ca)
x-a

t-

In connection with these cOrollaries,,,we observe that if

.

= f(xY

iim,g(x) = 0,
x-a

Under these conditions,

the quotient r-r may still have

gkxj.

is a necessary but not sufficient condition for existence of

limf(x) = 0
x ..a

lim\f(x)
F7

The primary example -is the derivative of a function exp'resseS.as

x-a

*(1

the limit of a ratio for which the numerator and denominator both approach
zero.

It is pot possible to make any general statement'about the exidttence

and yei that'

of the limit for such cases; itmis possible that ,lim f(x) = 0
x-a

the limit of the quotient does not exist (for example, lim
x-O x?
Exercises A6-4f,Nos,,, 14 and 154
.

In estimating

lim f(x)

).

(See

we can often bound

A
cit

below and above by

x -a

tions

and

h which have limits as

expect that the limit of

approaches1

a.

In that case yet\

is bounded below and above by the limits

ofSts'

and

h.

This result is a direct consequence of the following theorem.

A6-4f.

a,

and

# Proof:

f(x) < g(xl

If

lim f(x) = L
x-a
Singe

g(*)

and

in some deleted neighborhood

lim g(x) = M,

then L < M,

f(x)

is nonnegative it *glows that

rim
x-a

f(x),1 = M = L > O.

(Thsdrem A6-4c and Corollary 2 to LemmalA6-.4.)

Corollary I.

[Sandwich Theorem.)

If

h(x) <f(x)' < g(x)

in some deleted neighborhood of

a,

.and if

and

lim h(x) = K
x-a

lim g(x) =

/ft

8.06

4 0

them, if

lim f(x) - exists,

x-a

< lim f(x) <_ M.

K _
x-a

,
yaw%
,...

Corollary 2.

[Squeeze Theorem.]

deleted neighborhood of

If

h(x) < f(x) <g(x).

and if

lim h(x)
''x-a

A lim g(x) = M,
x-a

then
.4

lim f(x) = M.
x-a

8d7
.

407.

in some w

Exercises A6-4

c
1.

Prove the corollary, to Theorem A6-4c .

2.

Prove the corollary to.Theorem A6 -4d.

3.

Prove the corollaries to Lemma A6-4.

4.

grove the cOrollaries, to Theorem A6-4e.

5.

Find the following limits, giving at each step the theorem on limits it,

-ro

which ,justifies .it.

lim (2 + x)

x- 3
(5x -,2)

(b)

lim

(c)

lim'(T74417

(d)

aim (x3 + ax 2

biT).:#dhere

"

2
+, a

where

x + a3),

and

b are constants.

is constant.

xa
6. Find the following limits, giving at each stepthe theorem which justifies
it.

x3 - 1
lim
2
x-1 x -

(a)
a

(p)

lim
;3 - 27
4

7.

lim
x-1

Find

- x

n -1

a positive integer.

Verify first that

tc.

- 1

x - l
8.

xn 2 1
, for
x - I

+ xn-2

'4.

(x

... + x + 1_,

Determine whether the following limits exist and, if they do 'exist

their values.

(a)

lim
,

l' +

1 - x

Fim Lx
xka

(9)

n%

- a );

is'a positive integer,

lim

x-. -1
#

(d)

lim (x-x

(e)

42)(c.

+ x - 2

lim
x -.l

8b8
.

fr

4 U 8

is /consta t,

find

9.

Using the algebra of limits show that' lim 1)

x-a
if

Assuthe

lim sin x.= 0

x-0

and

if and only

O.

lx - al

x-a

dm
10.

f(x) - f(a)- L(x - a)

lim'

f(a) - L
- a

lim cos x = 1.
x-0

Find each of the following

limits, if the limit exists) giving at each step the theorem on limits
which justifies it.
(a)

x -0

(b)

A-0

lim tan x
x-0

x-0

cos 2x
lim
cos x + cln x
x-0

(a)

Prove*Corollary 1 to TheoreniA6-4f.

(b)

Prove Corollary 2 to Theorem A6 -4f.

Prove

12.

(e).e lim

:(c)' lim sin 2x

x-0
11.

sin x
tan x

lim sin 3x

lim f(x)
x-a

exists.)
.

Foryhat integralvalues of

and

does

lim
x--a

+ a

n-

exist?

+ a-

Find the limitfor these cases.

'13.

Prove that'if

'A

liM f(x) = 0

and

g(x)

is bounded in a neighborhood of

x-a-

x = a,

then

lim f(x)

g(x) = 0.

x -a

.14.

(a)

Verify that if

lim T-,))
x-a g x

exists and if

lim g(x)4=.6,

'then

)Z-a

lim f(x) = 0.
x-a
(b)

Describe functions

and

At

lim g(x) = 0
x-a

for which lim f(x) = 0

and

x-a
_

..

yet the limit of their quotiene'does not exist._.

f

15.
.

Prove. that if

lim g(x) = 0
x -a

limit of t e quotient

and lim f(x)

----

&oes not exist, then the

.

...

f)
g(x)
(x

does not exist.

5

,
809

4 i) 9

16.

q a function is the Limit

The right-hand limit at a point

P(p,f(p))

of the function at the,point

for a right -hand domain

(p, p +

Similarly for the left-hand limit, the domain is restricted to

,

(P

We denote them, symbolically, by lim

6 .0.

f(x)

and

x-p
respectively,.. In particular,

lim
k-2+

[x]

= 2,

lim .[x] = 1.
x:2

lim
.
x-p

f(x)

Determine

the indicated limit's, if they exist, of the following:

2

x)
(

a)
x4:21'

Agir

Ex,

x-2-

x2 - 4

, 'a > 0

lim
x-0-

.1

(g)

lim
x-0

b 50

a > 0, 1 > 0

1/

47-7/i--

Airmeg.

4*
810

4
r-

A,

Appendi:X 7

CONTINUITY MORIN'S
A7-1.

Completeness of the Real Number. System.

TIE Separation Axiom

Simple algebraic and order properties do not'alone serve to define the

real number system; the rational,numbers satisfy the-same properties and so do
other systems. Although na physical measurement requires anything more than

the rational numbers, they are not adequate for either geometry or analysis.
Foroexample, the hypotenuse of a right trtangle with legs of unit length has

the irrational length v;

thus the Pythagorean Theorem would not be true

if length were measured by rational values alone.

In the rational field the

concept of infinite decimal would be limited to terminating and periodic

decidals; an infinite decimal like i5.1O1100111000...

with chains of ones and

zeros of increasing length is uninterpretable in the rational field.

The

system of rational numbers has theoretical gaps, but thi real number system
is complete in thatfreal numbers are adequa, te to represent all the points on a

line (lengths); and all infinite decimals.

At the same time, it is posse le

to represent any real number by a4point on a4line.or an infinite decimal; in

fact, we use the concepts of poini,on the number line or infin

e decimal as

synonymous with real number.

The completeness of the real number system, its lack of the

cti!cal gaps,
=

is a,consequence of a geometrically plausible axiom.

The Separation Axiom.
for which every number in

If

then there is a real number, s

x le A

and

y t13

we have

B' are non-empty sets ofreal-numbera

and

is less than o4 eqUal to each number,in

which separates

2nd

lies to the right Of lacy.,

B;

that is, for each

x <es < y. .

In geometrical terms, if no point of a set

.

B,

point of a set: ,B,

s,

then there is a point

should it happen to be a point of

points of

B (but s:

if

such that all points of A _(but

A ) lie to,the left of s, and aid

lie to the right of

c B )

(see Figure A7 -1a).

1,,

811

4.11

Figure A7-la

A simple example of two'sets satisfyinthe separation aom is given by

x < -1),

A =.

: y > 1),

( y

in the interval

Clearly, any number

[ -1,1]

serves tseparate these

sets.

If two sets are separated -by an entire interval, as in the preceding

example, then it is possible to find a rationalseparation nufter

.,ti

every interval on the nugber line contains rational points.

cases are those for which.tNre.are elements of the two sets

because

s,

The interesting
A wand

,closer together,thanany given positive distance. (Gaps in-the system of

,

rational numbers can be exhibited as failUfes4114,;the separation axiom for

such sets%. For example, let

satisfying--

a2

satisfying

< 2, ,and, et

It,is possible to findrational'Values

> 2.

be iheAtAgarositive rational numbers

--B. be the set of positive rational numbers
A

an& A

p
'

tolerance (see Exercises A7 -1, No. 18) but'a

closer together -than any
2
separation number s would have to satisfy s = 2 and no rational number
1

has that property (Exercise A7-1, No. 3c

'real nuibhr which separates

).

In fact, any .real number can be

B.

A "'and

We can define J ,as the unique

defined as a separationnuAber for suitable classes of rationals.

general'

More

it Will be convenient for somet purposes to determine a real.number

as the unique separation number for two ,ets by the criterion of. the following
lemma.

A and

Lemma A7-lr'l Consider tVo sets of real numbers

x < y for

each

exist a t A

number
p -

a<

s,

and

ating A and B

and each

x,c A
c B'

y i 'B.

such that

is unique.

p - a < e,, then the number

there exist a

100C
.

\
Proof.

Let

lie between a and p

I

there
separ-

and B.

Given

it follows from thefa,ct that

(Figure A7 -lb) that

d
812

And., p with
,

and t be separation points for A

a t A , and p eB such that p - a < e,

-.mad

such that

Conversely, if thgA is just one separation

then for every positive

If for every positive'

4t2

Is - tI < e:

e,
s

Since this

Figure A7-lb

is true for every positive

s = t

it follows that

Is - tI . 0

and hence that

(see Exercises A7-1, No. 13b).

is

For the proof of the converse, let

A and B.

For every positive

'e

denote the one number separating

there must exist points a cA and

p t B such that

s -

and 0 < s +

fbr should one of these inequalities fail, tlien we would have

s +

as a seppration number.

We conclude that

s - = or

p - a < c.

Next we derive an important consequence of the Separation Axiom.

The Least Upper Bound Principle.

Let A be a set of numbers which is

bounded above; i.e., there exists a value M such that

a < M 4,or all

,

t A.

In the set of al/ upper bounds of A

iS\Smallhan
'Proof.

any other, the least upper bound.

Let

satisfy the conditions of the Separation Axiom.

separation number.

for

B denote the set of upper bounds of

exists at least one separation number for A

there is one upper bound which

Since

A.

The sets

and

It follov that there

and

B. ;Let s,be such a

is a separation'nuMber i'tr,is.an''Upper bound of

and is by definition an element of

B it is the least element of B

B.

Since

s'

also,a lower bound

and therefore thedeast wiper bound

of' A.

The Least Upper Bound Prindiple is also a. way of ekpressing the

completeness of .the real numbers; it is equivalent to the Separation.

'

Axiom in t he.Sente that eithe4 may replace the axiom and that the

separation_p

ert

then follow.

*
This number is also called the supremum of
The abbreviatkon

tub A

and is denoted by

is also common.

813

4t3,

sup A.

In order to verify that the Separation Axiom and the Least Upper Bound.
Principle are equivalent formulations of the comPlekeness Of the neat number_ '441.,
system it is necessary to prove that in an ordered field the Least Upper Bound
The proof is left as an exercise.

"Principle implies the Separation Axiom.

each positive

A,

If M is the least upper bound of the

Corolley 1.

there exista an

t A

then for

such that a > M - c.

A set Of numbers which is bounded below has agreatest

Corollary 2.

loxer bound.
Or

The

roofs of these Corollaries are left as exercises.

There are, various methods for constructing t>e;realnumbers frolp the

rational numbers so that the usual algebraic and order properties and the
Separation Axiom will hold.

These will be discussed in ,subsequent courses.

Exercises A7-1

Prove Corollary, 1 4,9 the Least Upper-Bound Principle:

Prove Corollary 2 to the Least Upper Bound Principle.

(a)

Consider the sets

2

<2,

and

,="2.

Prove' if a F A -and

a satisfying

of positive rational numbers

of positive rational numbers

E B

that

satisfying

a < P.

Show that a separation number

satisfy

2
s

Prove that
4.

(a)

i.e.,

a,

B must

that there is an integer

(Principle of Archimedes).
c > 0

there is an integer

< 1,7 < c.

814

greater

a..

and

is irrational.

Prove that given any

0

for the sets A

s =

Prove for every real number

than

(b)

= 2k

4t

such that

We defizie infinite decimal

5.- \ (a)

c .c c c
0 1 2 3

where

is an integer, and

r.

the number
c

lq

are digits, by

where

c2,

c1,

2
0 +
10"

< r <c

10

Show that the pregedirlg

1
+
10

10

+ 1

10

Aoes, in fact, .define a unique

real number.

Given a real number

(b)

ve define its decimal representation

recursively-in terms of the integer part function

[x]

as follows:

co-= [r]
c2

1
10

cn = [10n.r - c0

102

ShoW that the inequality in part (a) is satisfied for this choice
( of,

,1

Show also thit.decimals consisting entirely, of

on are avoided.

(Thus, we obtain

91s

from some point

2 = 2.000 ... but not

2 = 1.999 ...).
#

6.

An' infinite decimal

fixed value

n, satisfying

p,

.c c

1 2 3

.-,is said to be periodic if for some

the period of the decimal, we have

n > n0,

wheie we require that

integer satisfying this condition.

c
for all
n+p
n
is the smallest positive
c

In words, from some place on, the

,decimal consists of the indefinite repetition of the same

Thus1
3

digits.

= 0.33333...

= 0:34090909...

are periodic decimals.

It is convenient to indicate a cycle of p digits

by,underlining, rather than repetition; e.g.,

22

= 3.142857

7-

815

I5

g)

Prove that every periodic decimal represents a rational number.

(Hint:

(b)

Consider the decimal as a geometric progression.)

Prove. that every rational number has a periodic decimal representation.

(A "terminating" decimal in'which each place beyond a certain

point is zero is considered hi-a special case of periodic decimals.)

If

represents a'rational nUxber given in lowest terms, find

r =

the largest possible period of the infinite decimal representation

of

From ,b

t.

in terms of the depominator

we conclude that a decimal which ls not periodic represents ap

irrational number, -And conversely,

(c)

Prove for every positive prime

other, than

and

exists an iri4ger, all of whose digits are ones, for which

factor; i.e.y-

(a)

is a

is a factor of some number of the form A

+ 10 + 1...

lOn + 10n-1 + 10n-2

7.

that there

Considera polynomial with integer coefftcierrts:

anxn + a

n-1

n-1

+ a

+ a

(a
0'

lowest terms, then p

is a factor of

and

/ 0)
n 4

is a rational root of this polynomial given in

Prove that if

+ x + 1

has no rational root.

(b)

Show that

(c)

Prove that if I is rational then it is integral.

(d)

Prove that

is irrational.

816

4
A

is a factor of

e Value and IXrmediate Value Theorems for Continuous

The Extr

A7-2.

Functions

In Section 8 = we stated two theorems, which we reiterate here in more

precise terms':

pEOREC8-2a.
Suppose

and that

The InterMediate Value Theorem.

is co tipuous at each point of the interval

f(a) A f b).

If

there is at least on'point

lies between
between

f(a)

and

a <,x < b
f(b)

then

and b such that

f(c) = d.

THEOREM 8 -2b.

Suppose

is continuous at each point of ;Ghe inter-

val

a <

.and

a < d < b

< b.

Then there are points

and

d,

with

a < c < b

such that

f(d) < ffx) < f(c). for

.a <x < b.

These two theorems will be proved in this section.

Our proof of Theorem

8-2a makes use of the Least Upper Bound Principle and the following simple
lemma:

Lemma A7-2a.

lim

If
x

number

= L

and

L > 0,

then there is a positive

such that
a

f(X) >

if x

is in the domain of,

and

0 < Ix - dl <

Proof.

gin

e > 0,

The definition of limit tells ,us that. for any

there is a

8 > 0

such that
< E'

1,f(X)

if.x

s.

is in the dom)in of

and

lx - al< b.
By assumption -L > 0,
12.1)

so that,

is also positive.

we can find a positive number

(1)

so that

If(x) -

Ll < 124

817

4t7
ti

Therefore, (taking

"_- -:.-

-0-

,- - '

`-'.

is in the domain of

f,41c

'and

_.

-......

0 < lx - al <8.
The inequality (1) can be rewritten as
,

L < 2'

< f(x)

N.

to both sides we have

Adding L

L -

cannot be less than

f(x)

so that, in particular,

+ L. < f (x) <

0.

'This.

completes the proof of the lemma.

-

This lemma.has been implicitly used before in the form of the assertion
f(x)-, approximates a positive number as x approaches a, then the

__Lthat if

.f(x)

Values

must be positive if

is close enough to

< d < f(b).

< b' and*

the

) 4.

Our tiurRose is to show that there is a number

(c) = d.- Such a 'number can be found as follows:

e set

y not empty (since

A.

c 'such that
4

be

Let 'A

f(x) < d,

a I A) and bounded above /by

inciple implies the existenbe of a number

e Least

b).

such thdt

(a,b]

in the interval

i(a) < f(b).

Suppose that

is analogous.

f(a) > f(b)

et of all n hers R
&

We give the proof for the case when

Proof of Theorem 8-2a.

e proof for the case

a.

such ghat
x < c

(2)

if

x CA

and

$(3)-

<a

(since (2) holds and

A).

,is an pper bound for

define

1s any upper bound for

a < c' < b

e shall show that

a < c

if

f(c) = d.' First, we note that

and that

c < b

a t A) and that

by

a <

g(x) .'d - f(x),

By a4s option

d > f(a),

so that

lim g(x) =
-4 a

g(a) > 0.

d -

13m
X -4a

< b.'

Furthermore,

lim

X -4

f(x)
a

f(a)

818

4 8

(since 43) hblds and

consirnhe function

a < c,

To show that

A.

,since

is continuous at

Therefore,

a.

lim

g(x) > 0
a

and we can apply Lemma

and

x > a

7-2a-to conclude that if

then

is'close enougb to

6
g(x),> O.

In particular, ttir

is an

in

[a,b1..such that

x > a

and

g(x) > 0,

that isn
f(x) < d.

.i.
...

Such a

must belong to

(from the definition- of ,4A4)

c > x,> a.

A similar argument (applied to

shows tat

c < b.

f(c) 4 d
...-

or

c(c)= d.

f(c) > d.

so'that (2) implies

h(x) = f(x).7- d,'instead of

This completes the proof that

Now we show that

Ai

g)

a < c < 'b.

SuppoSe this is false, so that

Consider the case

f(c) < d,

and again let

41,

'

g(x) = d - f(x),,,

.,

Since

is continuous at

c,

lim
x

we h

lim

d -

lim

X -4 C

f(x)
c

= d - f(c) > O.
,

'Again apply Lemma A7-2a to conclude that,

t-

g(x) > 0

if x' is sufficiently close'to

defined for

and

b ant g(x) > 0,

c <

so that

c,

(from (2)).

b > c,

that

argument (applied to
tag) > d

that indeed

g(x)

is defined.* Since

there niut be a'poi t

f(x) < d.

Such an

This contradicts the fact that

The assumption that 0c) < d

tion

and

g(x) 'is

c < x <1)

such that

x,

ust belong to' A

< x < b.

has led us to a contradictio

h(x) = f(x) - d,

A similar,.

instead of

g) shows that''t p assump-

must also lead to a dontradiction.

We are fOrced to conclude

f(cp = d.

This completes the proof of Theorem 8-2a.

Our frof of Theorem 8-2b will make use of the following lemma whose proof
_is a simple consequence of the definition of limit.

819

4 i.i

f(x) = f(a),

(Lemma

X -4

such that

then'there

a.number

S > 0

Ask
?"

-;4

If(x) I k 1 + If(a)-

such that ,*

in the domain of

for'all x

a - 8 < x < A

47.,e,.

116

?root.

lim f(x) = f(a), the definition of limit tells us that

x ,4a
c > 0 we can find a- 6. > 0 Au-a-that

Since

'

for-any given

If(x).- f(a)1_,< e

is in the domain of

and

1.

al < 6

0 < Ix

In particular, we-%ircan find a positive number

such that

'

If/x) - f(a)1

(5)

if x

is in the domain of

and

<_1Ix -

(6)

<

x.

If(x) - f(a)1 = If(a) - f(s) F 10)

(6) can be replaced by

< lx

(7)
.

a,t (forthen

The inequality (5) -certainly holds.-if

If x

in the domain of

S.

then

f(x) = f(x) - f(a) + f(a)

so that the triangle inequality gives
If(x)1 < lf(X1 - f(a)1 + If(a)1.
Thus, if

also satisfies (7) we- an apply (5)sro conclude that

If(x)1 < 1 4-1f(a)1.

This is our desired result for (7) and can be rewritten es

- S <x <a + 8.

82b

4 V3

<

Proof of Theorem 8-2b.

interval

x < b.

Suppose

if continuous at each point of the

We first show that

is bounded on the interval,

that is,

there ia.a number M such that

-(8)

for a< x. <b.

< M

1f(x) 1

'Let

A' be the set of numbers

a _
<.t _
< b

(9)

and

is bounded on the

interval
Certainly

is nit empty (for

such that

t,

a < x < t.

a s A) and bounded above by

b,

so it has a

least upper bound, say a.

We shall show that a 6 A and that a =b.

will establish that

and hence that (8) holds.

b c A

The number-a,

being the least upper bound of

A;

This

satisfies. the two

conditions

(10v

t < a

if

t c A

is an upper bouhd for

(a

A)

and

t-< 0

if

for all t c A

then a < 0,

is not

(a

(11)

larger than any other upper bound

Since

a-6 A,

upper bpund for

be continuous'at

it follows from (10) that

A,

a,

0).

a <a.

Also since

it follows from (11) that a < b.

is an

Therefore,

must

so that
lim f(x) = f(a).
x -)a
.

Apply A7-2b

o conclude,that there is a positive number

(12)

If(x)1 < 1

is in the domain of

if .x

such that

If(a)1

and

a - S< x < a t

(13)

This will be used to show that a e A

and that 'a = b.

2.0

To show that a 6 A,

we first observe that a - 5 <a so that

cannot be an upper bound for

one

t e A

(10)).

such thtt

(from (11)).

t >a - S.

(14)

Therefore, there is at least

Such a number

Furthermore, the values of

t 'cannot exceed , a

must be bounded in the interval

a < x < t "(fronv(9)), so there is a number M1

If(X)1 < M1

-,8.

such that

< x <.t

'--- 821

421.

(from

.Noti.pg that if , t<- x <a,. then

8

a - 6 < x < ct

(since d-8<t<x<a<a+ 8) we conclude from (12) that

Q

e the larger of

M2

Let

< 1 +.!f(a)1,

If(x)

(15)

.1f(x)1

and

NL

t: <

< a.

tl*n-(14) and "(15)-iell us

that
a < x < a,

Lf(x) 1 < M2,

(18)

so that

is bounded on the interval

,a1

and hence a must .6I-.1.9k2..k.7

411m,

we-first reca 1 that a <-b From (11)):" -If it

To show that a -,b,

then since a <

mere trub that a < b,

fa,b1

the interval

such that

a < tl

(127)
a < x < t

Therefore, if

ives

then (12)

Kx)I <1 + If( )1,

(18)
.

1 +

(of (16)) and

M2

larger of

so that

must belong to

(a)I.

be the.

Combining (16? and (18) we have

- 1

if, a <x
nd, hence,

A,

'

t1 <,a.

This contradicts (17)

a cannot be, less than

and we are forced to conclude tha

the proof of (8).

, <_

Let M3

'so that (13) holds):

If(x)1 <

if a <x < ti

(since a - 6 <'a < x < t1 <a + 6,

.

in

t1

we can find a number

This completes

b.

e
We now complete the proof
interval

[a,b]. under

f,

f Theorem 8-2b.

Leto B

the

be the image
II

th t is,

-1,-

is the' set

,The set 3 is non-empty, (sin e

all numbers

f(x),

a <x < b.

f(a) E B) and bounded aboVe (from' (8)) so it

has a le. ast upper bound, whi h we denote by 'a.

Thus

It

(20)

f(x) <a if a < x < b Ze

iS an upper bound for

-and

B)

/
fi

if

(21),

(a

f(x) < 0

for

a <x <b then a,< 0'

is not larger than any other upper-bound-for

822

A2

B).

It will be shown that there is a number

f(c)

= a.

(20)'

, From

<a,

a < x < b,

the function

(,a,b1

such that

a < x < b

for

is our desired maximum value of

f(c)

Suppose there is no
f(x)

in

we will then have

f(x) < f(c)

so that

A74

c
in

.g

a,b]

Lr

on the interval

such that

,jie must, therefore, have

f(c)

= a:

< a,

f(x)

(a,b].

Since,

a < x < b,

so that

defined,by
fi

g(x) - a - f(x)

is defined for each

interval).

in

[a,b]

(for the denominator is not zero in the

Furtheymore, for each

lim

g(x5 =
t

'x

in, [a,b]2,

(a

lim
x

we then have
1

a-

f(x))

lim f(xj
x

a - f(t)

g(t)

so that g --ds Continuous at each point of

A

is

that there

an

in

Apply

(8) to g to conclude_

such that

[a,b],

a <-x < b.

< M,

ig(x)

For each

[a,b].

we have:

ti

1
a - -f(x)

g(x)

so that

0 <
,

1
f(x)

M.

Taking reciprocals we have

4

a - f(x) >

C that is;

<a

f(x)

a'--

Hence,

a>> a

is no

- .

M
c- in

1
-

a < x < b.

for

is an upper bound 'for

B.

This contradicts (21) since

This contradidtion was a consequence of the assumption that there

that is, there

(a,b1

such that
number

f(c)
in

= a.

Hence, there must be such a

such thiit

[a,to]

f(x) < f(c),

a < x < b.

823
s

'

423

c,

A7-2:_
The proof that there is a

in

[a,b]

such that

f(d) <f(x),

-;

Of course,:now that .!!e know that,continuous functions

a < x < b

is analogous.

on

hive maximums, we can apply this to the function 41 A maximum for

[a,b]

will be a minimum for

so that continuous functions on closed intervals

must haye both maximum and minimum points.

v.

4
824

'1

A7-

Exercises A7-2,
1.

Let

0
1

x >0

sin T ,

ir(x)
0

Show that
[0,b1,

x = 0

satisfies the conclusion of Theorem 8-2a on any interval

but

is not continuous at

x = 0.
1

2.

is continuous and'haa an inverse on

,f(e) < f(b)

3.

then

Prove that if

Prove that if
la,b1

and

is strictly increasing.

is continuous on

is a closed interval.
4.

ra,bi

(Hint:

ra,b3

is continuilio ip

tlien there is an x

then the image of

[a,b1

Use Theorems 8-2a and b)--:

in

fa,b1_ and all values of f

Ca,b1

are in

for Which' f(x) = x.

.

j.

Suppose
x

'

f(x)
0 ,

x = 0

Does 'fa satisfy the hypothesis of Theorem 8-2b on the interval [0,1) ?*
COes (8) h8/d for

on

[0;1]?

on

[10-100;1]?

6.

Is the Continuity Of

7.

Cab a discontinuous function whose domain is a

8.

DolitiMbers 6 and 7 amount to the same question?

9.

Can a rionccibstant function whose domain is the set of real numbers bef:

essential to the hypothesis of,(8)?

lose

interval be bounde

04.

44$.0:

10.

bounded? ce

44k

Show that a function .f

which is increasing in a neighborhood at etch

.

point of an interval
(Hint:
e.

Let. A

increasing in

la,b7

is.an increasing foliation in la,b).

'be the set of alit, 4n

la,t1.

Show that if
.

N1,1)1 'such that

Cc ,= lub, then

'function has the Property that for.each poin

is defined, there is a neighbot.hOod in which

example as is Number'10 where

is

and a = b).

elk.

of an interval where it

e fActiod is bounded.

,Show
that the function is bounded over:the whole interval.
,
,

If-

(This is an

',' local pkoperty implies a global one.

isfeclear that the global property here implies the,:loCal one.)

I,

-,

1
1I

825

425

It

. A7-3-

A -3.

The Mean Value Theorem

oug

In Section 8-3 we discussed the Mean Value Theorem,- We amplify that

I

discussion here,
'1,

.*

The Mean Value Theorem.

Suppose

is continuous at each point of the interval

a < x < b.'.

.and differentiable at each point of

least one number

cy

such that

a < x_< b

Then there is at

and

a < c < b

'A

f(a)4:)f,(c).
b.- a

f(b)
'(1)

In this section we give a proof of this result, and show how it can be
used to obtaill error estimates

rn

ap proximation formulas.

Further applica-

tions will be discUssed in the next section.

In geometrical terms, the Mean Value Theorem states that on the arc
between any two points of the graph of a differentiable function there exists
a point where the curve has the same slope as the chord.* Thus, let (p,f(p))

and

(q,f(q))

with p < q,

be any two points on the grag} of a differentiable function

say (see Figure A7-3a)
y

I
P

Figure

X3a

The word "mean" here signifies "average". The slope pf the chord is
interpreted as average rise in function value per rise in value of x. The
Mean Value Theorem states that this average is equal to a value of the
derivative at some point of the interval.,
826

it

is 6

'

--Ace&-ding to the Mean Value Theorem there exists a point

between

and

q , where

ft(u)

f(q) - f(p)
10,

We can make the Mean Value Theorem plausible by an argument similar to

that by which we found that the slope of a graph at en interior extremum is
zdro.

Take a parallel to the chord at a point

(u,f(u)), which

}des on the arc at maximum distance from the chord.

,Since no point of the

arc lies at a greater distance from the chord, the arc cannot cross the
parallel.

The arc cannot meet the parallel at an angle for then it.would

cross; therefore the two must have the same direction at

(u,f(u)).

(See

Figure A7-3b.)

(q

f( q))

(u , f (11))

,..

/ 0
_/
__.

/
//

/
//
//

p, f(p))

- /e

Figure A7-3b
In order to derive the Mean Value Theorem we first prove it for, the
special case in whi,ch.the chord is horizontal.

827

427

Lemma A1:2 (Rollets Theorem).

interval

If

differentiable itthe open interval

(a,b),

then there is.at least one

f(a) = f(b)

is continuous in-the closed

ft(c) = O.

such that

(a,b)

in

and

(a,b)

If

Proof.

If. f

(a,b).

is constant then this is certainly true for any

is not constant, then there is a point a in

that f(a) / f(a).

Let us suppose

same arguments to

-f), so that if

(which

is a_maximum pg,i.nt for

c

must he in

(fqr f(a) = fN) and, hence, Theorem 8-2c implies that

ft(c) = 0.

Certainly

exists.by Theorem 8-2b)'then f(c) > f(a).

consequences of Rollets Theorem (Lemma A7-3).

Let

(a,b)

us examine some of the other

so{Before proving the Mean Value Theorem 1 1

Corollary 1.

such,

(otherwise'we can apply the

f(a) > f(a)

c

(a,b)

in

%.

Any zeros of

be differentiable on an interval.

within the interval are separated by zeros of the derivative.

Proof.

If xl < x2

and

f(xl) = f(x2) 4 0,

are satisfied and there exits a value

the conditions of Lemma A/-3

such that

<

< x

and

'

f.1(u) = 0.
4

As a Consequence of this result we observe further_thatj,in a'given inter.

val, a function may have at most one more zero than its derivative.

From this

fact there follows a familiar. result:,

Corollary 2.

ti

can have no more tha

A polynomial of degree

14

distinct real,zeros.

ercises A7-3, No. 1).

The proof is left as an exercise

Example A7-3.
(i)

know-thit
.

ft(x) = 3x

and

- 3

has zeros at

f(t) = -1.

and

f(x)

between

-1

and

It

has a zero for

1.

Clearly

negative for sufficiently large negative values

and positive for sufficiently large positive values.

f

= -1.

By the Intermediate Value Theorem we

conclude that there is a z7ro of

we can take

x = 1

may have as many as three zeros,, We observ4 that

follows that
f(-1) = 3

x < -1

It follows that

and another for x > 1: Specifically, we

828
eic

We

Let us'apply Corollary 1 to the zeros'of" f(x) = x3 - 3x +

AT-3
have
-2

f(-2) = -1
and

-1

f(2) =

and anothgr between

The function

(ii)

and

f(x) = x3 + 3x + 1

so that there is one zero between

1

and

2.

has the derivative

f'(x) = 3x2+3

which is alway's positive.

Sinc-e the derivative.is always positive f

can have at most one zero.) Observing that f(-1) = -3 and f(0) = 1

we see that a zero exists and lies between

Proof of Theorem 8-2g.

points

(a,f(a))

(2)

and

O.

is

in

a) f(01)

(x

y = 5.(z) = f(a)
x

and

The equation of.the straight line joining the

(b,f(b))

It follows for any point

x = -1

(a,b)

af(a)

that the height

h(x)

of

(x,f(x))

above the chord is given by

h(x)

(3)

f(x) - g(x) = f(x) -"f(a) - (x - a) f(b)

f(a)

G.

From this equation it follows straightforwardly that

conditions of Rolle's Theorem (Lemma A7-3) on

[a,b].

h(x)

satisfies the

First, as you May

verify directly,

h(a) = h(b).= O.

the sum of

and a linear function;_ since both terms of this sum are

f(x)

Nexl observe that

h(x) = f(x)

g(x)

is

differentiable on the open interval

(a,b)

interval

also is.differentiable on the open inter-

(a,b]

it follows that

val and continuous on the closed interval.

that for some value in

and continuous on the closed

From Eolle's Theorem, we conclude

(a,b)

h'(c) = f'(c) - g'(c) = 0,

or, from Equation (3) for

h(x)

.1"(c)

above,...)

f(b)
f(a)
- O.
b - a

vole

Linear Interpolation.'

r.,

Line-arintenpolation is a useful method of approximation to the values otX

a function in an interval when the endpoint values are known.

If bounds on the

range of the derivative can be qtained, the Mean Value Theorem gives a way of
estimating,,the error of approximation.
,

:#1

829

4 '49
At

Geometrically,, linear.interpolation consists of replacing the arc of the

Thus, on (a,b)
on (a,b)* by 'the chord joining the endpoints.
graph of
given in Equation (2). The
we approximate f(x) by tile linear function g(x)
is given by Equation (3). For
error of the approximatidniiig(x; f(x) = -h(x)
dUr purposes'it is convenient to repast Equation (3) in the form

f(x) - f(a)1.
x -

off(a)

f(b)b - a

(x -

.g(i) - f(x)

Now, by the Mean Value Theorem

g(x) - f(x) = (x - a)[f'(u2) - f'(u1))

(4)

where
say

iJ

a < u

a < ul < x < b,

ift(z)1 < M,

for

-lg(x)

If the derivative is bounded in

< b.

(a,b),

in

then from Equation 4)

(a,b),

f(x)l < lx - al(lft(u2)1 + 1f2(u1)1) -

whence

(5)

Ig(x)

- f(x)l < 214.11x - al.

e

ii5 by linear,interPolation for the

Example A7-3b. 'Let us estimate

function

Since

f : x

Equation e2) and obtain

we take

3.< .VD5 < 4

22

a = 9 and

as our estimate for

g(10) =

b = 16 .in

/O. On the interval

(9,16) , we have
1.

fs(x) =

1^

<

<g

2V

2 IX

Entering this bound in (5) we obtain

4
.

124 -

of <

We observe, however,. that

484

22 2

and we suspect thatlour estimate of error ip rather crude.

4

0.

850

'3 a' 4

'I

.1

q
<1.

(a,b)

If on the interVal

has a derivative

,ft

f",

the second drive-

tae may apply the Mean Value Theorem again to the difference

tive of f,

ft(u ) .- ft(u

in Equation (4) to obtain

g(x) - f(x) = (x - a)(u2 - ui)f".S.v)

where

is somewhere between

points of

(a,b)

and

Since

u1.

and

are both

we know that the distance between the two points is less

than the length of the interval:

u2 - u11 < b - a'.

Suppose, in addition, that.we have-a bound on the second derivative,

Jf "(x)l < M

on

(a,b)

Then we obtain an upper estimate for the error in

terms of thq'second derivative:

(6)

Ig(x)

ti

- f(x)

< (x - a)(b

Example A7-3c.

(6)

Now let us use Formula

error of approximation to

AT

to obtain an estimate_ for the

by the linear interpolation scheme of Example

A7 -3b., We' have

((-

hh,hrhhil
If.(x)1

3,2

<
for

x 'in (9;0)..

---7I32
,

4.9

108

Consequently, from
1272

<

(6),
_

101 <

It fcalows.that
3.07 < 1/1715 < 3.21.

We have obtained sharper estimates for

and now we can repeat

11.6

process to obtain still sharper estimates using

a = (3.07)2

mot

831

431

and b = (3121)2.

Exercises A7-3
1.

Prove

orollary 2 to Lemma A7-3.

2.

Sketc

the graphs of the functions ,in Example A7 -3a.

3.' Is th following converse of Rolle's Theorem true?

on the closdd interval

and differentiable on the open interval

(p,q)

and if there is at least one point

(p,q )I

f'(u) = 0,
such that

f .is continuous

If

then there are two points Jh

u
and

in the open interval where

n

where

.p <m<u<n< q

f(m) = f(n).
Jr-

Rolle's Theorem justify the conclusion that 21 = 0

dx

Ibe

4.

(y + 1)3 = x

f(x) = x(x - 1)(x - 2)(x - 3)(x -4).

n:

value

sf.,

< x < ]( for

in the interval

of lx

fOr

Determine how many solu-

fl(x) =,0

tions

outalculating
..15.

has and find intervals including each of these withf'(x).

Verify that Rolle's Theorem (Lemma A7-3) holds for the given function in

the given interval or give a reason why.it dos not.

(a)

f : x

(b)

>x3 + 4x2 - 7x

[ -1,21

2 .- x2

4-- ,
x

7.

10,

Prove that the equation

f(x) = xn + px + q = 0
cannot have more than two real solutions for an even integer
more than three real solutions for an odd

8.

A function
(a,b1.

(a,b)

g(x) = 0

Use-Rolle's Theorem.

has three different solutions in,the':

Show that the equation

one solution in 'the open interval

9.

nor

hds a continuous second deriyative\on the closed interval

The equation

open interval

n.

g "(x) = 0

has at-least

(a,b).

Show that the conclusion of the Mean Value Theorem does not follow for
f(x) = tan x' ifi the jnterval

1.5 < x < 1.6.

432

832

,
.

For each of the following functions show that, the Mean Value Theorem fails

10.

to hold on the interval

(a)

f : x

(b)

Explain why the theorem fails..

,a > 0.

if

[ -a,a]

Ix)

x
/

Show that the equation

- 11.

the open interval

x

+ x

- x - 2 = 9 has exactly one solution in

(1,2).

x sin x + cos x, for exactly two real- values of

12.

Show.that

13.

Find a number that can be chosen as the number

x.

in the Mean Value

Theorem for the given function and interval.

0 < x <

x -4cos x,

(a)

(b)

f :x )x3

(c)

(d)

x ) cos x + sin x,

-1 < x < 1

- 2x

< x < 0

+ -1,

0 < x < 2v

Derive each.Of the following inequalities by applying the Mean Value 'Jr

14%

Theorem.
(a)

Isin x - sin yl,< Ix - yl

< arctan x < x

(12)

if

1* x
A

15. %Ike the Mean Value TheOrem to approximate

-A.008.

16.

Use the Mean Value Theorem to approximate 'cos 61.

17.

'Show that .a.(1 +.

< 14717+

<

E )

n(an + e)
for

e > 0, a > 1, n > 1

!Tan

(n

rational).

18. *Using Number 17, obtain the following approximations.

(a)

3 +.116- < 3-k/5 < 3 +

(14)

3 +

(c)

Show that the approximation

< 5

< 3 +

5(244)

1 \

75, ,_o

5 r-rr

Y24

is correct to at least 5 decimal'places.

.
833

.13r

19.

thaw that a straight line can Intersect tbe graph of a polynomial

(a)

of n-th degNle at most

times.

(b)

Obtain the corresponding result for rational,functions.

(c)

Could

sin x

or

Justify you

be rational functions?

cos x

Prove the intermediate value property for derivatives; namely, if

20.

differentiable on the closed interval 4,0 then

value between

and

fl(p)

answer:
f

is

takes on every

f/(x)

in the open interval 1p,g)

fl(q)

(See Section 2- .)

Estimatefor Newton/ s Method.

21.

are positive, on

e [a,b]

and that

[a,b]

Suppose

where

f(r) = 0,

f'

and

r t [a,b].

Let

f"

and put

1,

f(xl)

x2 = xi

Suppose

< M

li'"(x) I

(a)

a <x < b.

If 1(x) I > m > 0,

and

Show that

ri <

lx2

02 !

f(xl) - f(r) e
(flint:

r = x

x2

between txi

Find

ft (x

and

such that

x2 - r = xl

fl-g)

f,x1)

(-x

fl(t)

fqx-1

1,

- r)
7

li)

f! (Xi)

(*X1

r)

.
Then -find

E.

.....

between

x,`
,,, - r =
,

(b)

If

b --,,a <

al

k,

xi id

vf':(t )
1

,'

.,

0 < k < 16 show

.that

'

-,,crfv-Section 2-

Ix

-'9
a' k7

-_,r1
4

.
.

,-- M

7-ib

.'.,,

4;

831

43

!O-

ar-

Applications of-The Mean Value Theorem

A7-4v;

This is an extension of some of the ideas of Section 8-4.

THEOREM:A7-4a.

fl(x) > 0

If

creasing on -(a,b);

f1(x) < 0

Suppose

a < x < b;

ff .f1(x) < 0

Only the increasing case

Proof.

the case

for

then

is in-

:t

then

is decreasing.

will be considered here,

f1(x) > 0

is similar (or can be obtained by considering

f'(x) > 0

on

For any two numbers

(a,t).

"gikskir

xl

and

-f).

X2

in they

0 o

interval with

the Mean,Value Theorem tells us that

< x2,

f(x2) - f(x1) = fi(c)(x2f- Xi)

4,
44.

for some

Since

(x1,x2).

in

f1(c) > 0we must have

-f(x2) - f(x1) > 0,

f(xl) < f(x2).

that is

This proves the theorem.

If we replace the weak ineqUalities

' inequalities
.

(>,

and

<,

and

(>,

<e)

by the stronger

respectively) the same proof yields

'

THEOREM A7-4b." If
increasing in

f' (x) > 0 for

(a,b);

if

a' < x < b

f'(x) < 0 then

is strictly

is strictly decreasing.

Theorem 7-3b-is a simple corollary to Theorem A7-4b, for if

Ft(x) =
then-

(F,- G)' = 0

on 'a < x < b

a < x < b

for

so that

1'

decreasing and, hence; must be constant on

F(x) = G(x) + c,

where
F

then

and

11,a constant.

- G

(a,b),

is both increasing apd

that is,

a < x < b,-

This also holds at the endpoints

must also be continuous at

and_b,

since,

a an b.

lo

835

435

fi

on

(a,b); if

Proof.

and 7:2

f"(x) > Os for

If

THEOM A7-4c.

f"(x) < 0 then

We discuss only the case

x

(a,b). and

are .in

< x2,

then

a < x < b,

is conve*,

is concave.

we wish to show that if

f"(x) > 0,

x1

then

f(x2) > f(x1) t fqx1)(x2 - x19

Aos

(See (3).of Section 8-2).' To prove this, we apply the Mean Value Theorem to
find a

in '('X

such that

f(x2) - f(xl) = f'(c)(x2 - x1). ;

f(xl) - fqx )(x2 - xl) = fqc)(x2,'.:-X1) - ft(x1)(x2 - xl)

= (f'(c) - f'(x1))(x2 - xl).

Apply the Mean Value TheoreM to

f'(c) - fqx1)."This gives a number, c

(xl,c) such.that

f'(c)

v(xl)

e(c1)(c

x1).

Hence

f(x2)

f(;(1),

f'(x1)(x2

which is non - negative, since

f
as (1),:

f"(c1)(c

f"(c,-4 ) > 0

f (xi)

and

xl)(x2

c > xl, x'2"4> xl.

(x).) (-x2

xl)
Therefore,

x 1.) > 0
4

A similar argument shows that*(1) holds if.

836

4 (3 b

in
.,

A7-4

----7''Exercises A7-4
1.

Let 'f

be-differentiable on a neighborhood of a point

f'(a) = 0.

f'(x) < 0

If

x <a and

when

for which

f'(x) > 0

when

.,

x > a,

2.

then

f(a)

when

x >.a

Let

is a minimum.

then

f(a)

If

when

f'(x) > 0

is a maximum.

on the open interval

(a,b).

and

f'(x) < 0

Give a proof.

be continuous on the closed interval

x < a

Suppose

and differentiable,

[a,b]

is the one pant in

(a,b)
.

where
of

ft(u) = 0.

Prove that if

then

is the` global extremum -of

f(1.Q

fi(x)

reverses sign_in a neighborhood

on

[a,b]- approp'iate

to the sense of reversal.

3.

Given a function

such that

f(1) = f(2) = 4,

andSuch that

exists and is positive thrciughtout t4 interval

.1 < x

e(x)-

3.

(,a.)

What can you conclude about

(b)

Prove your statement, stating whatever_ theorems you use in your

f'(2.5)?

proof.,
4.

Let ,f be a 'differentiable function on

ment that

(00).

Prove that the require-

f he increasing is equivalent to the condition that

everywhere but Vat every interval contains poi4ts

5., A function
interval

is such that

(p,q).

equations

g"

where

f'(x) _> 0

f'(x) > 0:

is continuous and positive in the

What is the maximum number bf roots of each of the

g(x) = 0

and

gi(x)

Prove your result and give so

in

(p,q)?

41ustrative
_

6.

Suppose that
f

(n)

(a)

(1)

(a) = f

2).
,

(a) =

Determine whether

0.

is, which kind.

(Hint:

= f

(n - 1)

(a) = 0

but that

.'

f(a)- is a local extremum, aid if it

Consider separately the cases

n'-even and

odd.)
7.

Prove that a necessary and sufficient condition that the graph of a

.differentiable function
each point

in

4to the fixed point

I,

be concave on an interval

is that for

the slope of the chord joining anoint

(a,f(a))

is aodepreasing function of

8 4a7

'

(x,f(x))

on

I.

;K.
4

8.

(a)

Let

be differentiable and its Wraph is concave'on an interval

I.

fi

Prove that the function

f(x) - f(a)
x - a

X 1 a

0(x) .
f'(a)

is decreasing, where the fixed point

(b)

x-= a

is any interior point of

I.

From the result of (a), prove that a necessary and sufficient condition that the graph of

is that

be concave on

f'

be

decreasing.
9.

(a)

Let

'function

(x,f(x))

'

f.

be two points on an interval

and

in the domain of a

Show that a point is on the chord joining the points

4c
on thgraph
of

(y,f(y))

and

if, and ad; if, its

coordinates are

(ex + (1 - 0)y,ef(x+(1 - e)f-(y))

(b)

such that

some

0 < e < 1.

Show thdt a'differentiable function, f

x

only if, for all

and

and all

I.

if, and

rs"*..1

such that '0.< e < 1,

ef(x) + (1 - e)f(y).

flex + (1 - 0)y)
(c)

in

is convex. on

Use (b) to show that the graphs of the following functions are
convex.

10.

(a)

ax + b

(i)

f.

(ii)

f : x 2+%x2

(iii)

Derive the following property of differentiable functions' If the

graph of

12) is concave on an interval

then for all points

I,

-a,

in

and any positive numbers

1)4.

>

Pf(a)

p, q
qf(b)

*/
.

In words, the function value of a weighted average is not'less than

the. weighted average of the function values.
(b)

Prove that this p.roperty is sufficient for concavity.

f,
C

-838
et

438:

"11%

Prove that if

is differentiable then,a, nec4sary and svfficient

condition for its graph to be'concave,is that

a + b

f(a) +f(b)

tf(---)
2
A

12.

The'graph of a differentiable function

for all

x.

'Show that.

is concave and.is positive

f' is a constant function.

1. Under what circumstances will the graph of a function

6t

both be concave?

and its inverse

one concave and the other conveic?

1

14.

If either of

D xF(x)

is of.one sign for

F(-)d

the other one has the same sign.

x > 0,

show that

Interpret geometrically and illustrate

by..several examples.
.

15.

If
that

16.

(a)

F(x)

'is concave and

F(x)

Let

F(a) = F(b) = F(c)

here

a < b < c,'

show

a < b < c,

and

is constant in "(a)c).
a, b,

and

be points in

suppose that the graph of

f(b) <

f(a)
c

such that

is-convex in

c - a

I.

Show that

f(c).

2
.

(Hint: Usthe result of HuMber 13.);

f(a) >

f(c),
c

f(c) > 177

(b)

If the graph of

hence,

7 f(b) - 7 : 7b f(a).
a

is convex in a closed interval, show that

:s!

is bounded in the interval.

(c)

Show by-a counter-example that the result in (b) is not valid for
an open_interval.

0A

..

839 '

139

Appendix 8
MORE ABO6T INTEGRALS

A8-1. 'Existence of the Integral,,

The purpose of this section,i4to establish necessary and sufficient

conditions for the existence of '.'integral of a function

over

[a,b]..

Recall that the integral is definpe the unique separation number between
the upper and lower sums.

first to establish that the upper and

We nee"

lower yams are in fact sdparatedthat every lower sum is le1;4 than or equal
to every upper sum.

If it is posible to find an ,upper :Aim and a lower sum

closertogether than any given fixed tolerance IL,

,exists a unique separation' number, a number

than by Lemma A1-5 there

hick is the integral of

.over ,[a,b].

Leihms A8 -la.

Let

Any fixed partition

be a function defined and bounded on

of

each upper sum U 'over

[a,b],

than or'equal to each lower sum L over

Proof.
(a,b]

lower sums, the points of

and

is greater

is simply a set or points of

To construct upper and

b.

are arranged in increasing order; ile!,

a = xo < xl < x2 <

An-upper sum U

<xn =.41).

is defined as

U =

mijxk _ x,_1)

k=1

where f(X) < Mk

on

[x.k..1,xk],

L =

a loser sum as,

.,(xk

x,_1)

k=1
Where f(x) .>

on

[x.k..1,xk].

Thus

mk < Mk

and term- for -.term

or
Ink(xk

xk-1)

For,'

0.

We recall that the partition

which includes the endpoints

[a,b].

Mk(xk

xk-1)

from which this lemma follows.

841

440

It is necessary to find a, means of comparing upper and lower sum's for

any two partitions

partition
.

a = al U a2

over
a

which consists of all points of the two partitions

Let U1

taken together.

For this purpose we introduce the joint

and

al

be any upper sum over

U1

shall show that

a2.

and that

al

L2

any lower sum

is an upper sum for the joint subditision

similarly is a lower sum for . a.

L2,

and

The resu t we seek will

then follow from the preceding lemma.

For any partitions

Lemma A8=1b.

and lower duns

U11.

and

al

a2

of [a,b]

and any upper

I

over' the respective subdivisions,

L2,

?-'.L

Proof,

Let

al,

(k = 1,

from

x.

...,;since
4

Mit

and mk

i = 0,

2,

up_f With

ul,

say,

Setting
137

uo =

p,

f(x)

is,a partition -of

[:.1c.

Further
Kj.

f(x) 'th all of

in each,of .the subinter'vals

(see Figure A8-1)

1p see

and ..\11.1) = xk,

are upper and lower b6unds for

AXk..1,Xk1 they are bounds for

,= 0, .1;

There may be points of the subdivision

(kie_xk),

up_i < xk,

)ck_i < u1 < u2 <

(ui

be a pair of cdneecutive points of subdivision

n).

.2,

in the open interval

..,,that the set

i-1

,u ],
i

If 'we form the-upper' sum

-

1;x0. using the upper bopd

over the partitigkof

)41

!'

Figure A8-1
,

441
842

AtS-1

=Z Mk(ui

Mk

- ui_

ui.1) 7telk(xk-xkI1)

dui

hus,the upper sum

i=1
,

E-uk

for the partition

is also an upper,sum for

Similarly

a.

k=1

is a lower sum for both

and

a.

L2 <1U
If form Lny partitions

Corpi!lary.

It follows from Lennie A8-la that

1.

an upper sum U1 over

a1

and

and a lower sum L

finC".r

of

over

[a,b)
a

there exist

satisfying

- L < E

1,*
A

then there exists a partition

which has upper and lower sums

and

satisfying

U - L

a = a U a
Since U and L
are upper and lower
1
2'
1
2.
sums for the'joint paltition, the result is immediate.
'4'

Take

Proof.

Let

THEOREM 6 -3a.

be a bounded functicin on
:(

positive

there -exists.a:partition

upper sums -L
co)

and

therp.exists a mumber-

oyez,

[a,b).

If for every

of

[a,

d lower and

a which differ by 471

less than

whiCh is the integral of

then

E,

over

ra,b].

"'Conversely, if

ip integrable over

[a,b),. then there exist a

-

partition
that

a .and lower nne-upperputs L

U 'over

and

Pr--_-oof.

L < E.
t

r'

upper sum.

such

From LommIA8-ib every lower sum is lest than or equal to each

If for every

E >0 ,there exist lower and upper sums

and U

satisfying 'U

L < E, then ty Lemma Al-1 the nutber Separating the set of

'lower sums from the set of upper sumb is up4ie. By definition this

-separation number is the integral of f over

Conversely, if

16' '

[a,bj.

f, is lntegrabl

,
t * if.the integral'If f over
.
.
'exists, then by definition the geparation'numbdc between lower and.
''upier sum! in unique. It follbws from the converse statement in Lemma
A7-1

'[a,b)

C.

1843

442

:.

that there exist lower and upper sums, not necesSarily, over the same partitiOn,
Say

over

otter

and..0

for which

- L

< c.

From the

corollary to Lemma A8 -lb, we conclude that there exists a single partition

haSing upper and lower sums

and

for which

L < e.

Next we proVe a useful corollary to Theorem 6-2a.-'

4,

Lemma A8-2c. If
over any subinterval

[a,b]

is integrable over

then

is integrable

[a,13]s.

'

Proof.

U an

her
and

There exists a partition

p.

of fa,b]

denotl upper and lower sums over

are points of

a,

With 'a

apariition

a'

and

13

U - L
all terms are nonnegative.

the sums

and

included in

and

U
a,

(see the proofof

it follows that

contains

E(1.1k - mk)(xk - xk_i)

If we let

and

U'

which are taken over a,

We may assume a

Now in the sum

[aA].

of

a.

for if they were not so originally they could be

introduced without affecting the values of

Lemma A8-2b).

for which U - L < es

U'

IA

denote those parts of

it follows that

L' <U - L < c.

According to Theorem 6-3a, the function

is integrable over

fa,P.1.

Iv

<

3
/

/
1 4,

a.

)7

Exercises A8-1
. ,Let

f be a function whic17 takes on a maximum an

closed interval

Let U (a)

and

minimum on every

(e.g., f could be a continuous

ction, or monotone).
*
L (a) be the upper and lower sums obtained by

using the maximum aneminimuft values of

f'x)

as the appropriate bounds

in each interval of the subdivision.

Let

and a
be any partitions of' [a,b].
1
2,
joint subdivision a = al U a2 that
a

u*(a ) >u*(a) >

1

Prove for the

"(a) >1.*(a

In other terms, by adding new points to a subdivision we may reduce

the

between the upper and lOVer sums, and we cannot

increase it;

2.

Consider the function

defined on
0 , c

by

[0,1]

irrational

f(x) =

11
Prove that the integral of

, x

rational

does not exist.

4
3.

Consider the function

defined on

[0,1]

by

(00 , x irrational
f(x) =
.

x, rational;

Prove that t(le integral of

4.

fl over

x =

in lowest terms.

[0,1]

exists and find its value.

Give an example of a nonintegrable function

fg

where

and

are,

each integrable.
v

44

The Integral of a Continuous Function

A8-2.

In this section it will be shown thatjf ...f, is continuous On the interexists, that is, there is 4a unique separa(a,h1 then the integral of f

val

tion number between.the upperanclaower sums.

If

and

sums

U 1 ; U
:

[a,x1

are any two partitions of

and that

[a,b]

is continuous on

Suppose

and lower sums

2

l'

is a point of

[a?b].

with corresponding upper

then we know that

2

Ul.

L2

(1)

In particular, if

possible partitions of

denotes the set of all possible upper sums for al]:

[a,x]

B denotes-the set of all possiblt lower _

and

is a lower bound for

number in B

then (1) tells_ns that each

(a,x1

sums for all possible', partitions'of

is an upper

and each, number in A

bound for B. . The symbol

(read "the upper integral of

x,f

to

x")

,a.

to

xlif ';'

from

will denote the greatest lower bound of

The symbol

A.

(read "the lower integfal of

.f

Since each Ti

will dote the least upper bound of B.

inn

A'

is an upper-

must -have

bound tor-":13'

fa f < U

SO that

is a lower bound for A

andlbence cannot exceed the greatest

,,

.4. T

lower bound -for

A,

that is

\...
4

...,.-

f.S',.,'=-.--.---.

rx f < i

(2),
-- -- .

_J-a_

Our purpose' is to show

f.
a

Mat
f

(3)

is,Ihere is a unique separation number for the upper and aower sums on
each subinterval

[a,x].

The method of proof is as follows:

8"
41

443

Let F

encl.; V

.40,-

be the functions- defined by these upper and lower integrals, that is

----eaves=e116.tie.",
ti

A <x <b

'f,

"i7 ( X )
Ai

F(x)

a <x < b.

f,
a

We shall show that F and F have the same'derivatives (namely

their difference is constant
the same (namely
Certainly

to shod that
proof of

(Theorem 7-3b).

are

0) they must_ Abe the s ame functions, which is statement (3).

F(a) .= F(a) = O.

F' = f =

F' = f

(See Exercises A5-4, lb. 8), so it is enough

Waishall

being quite similar.

THEOREM A8-2.

F(x) =

f) and hence

Since their values at

tahlish the fact that

f, a < x

the

In summary, we shall prove

f A.s continuous on

If

F' = f,

[a,b]

and

'then

1,(1) = f(x), a < x < b.

The proof of this theorem is quite analogous to the proof of the Area
Theorem (Theorem A7-3a), with some complications due to the fact that
not assumed to be increasing.
A
tt.
Lemma A8 -2a.

If

is continuous on

[a,b]

and

a <*c < b,

[c,b]. ,The union

be a partition of .(a,c]

a1
a

deVote Upper sums for

a1

sum for

[a,b].

is a partition
and

4:52

and

then

r:

l'f---faf-+Taf.
Proof. ,Let

f 'is, '

We first establish three lemmas.

a2

ta,b12

a partition of
If

U1

and U2

then U1 + U2. is certainly any upper

The number

,
lab f
is the greatest lower bound of the upper sums of partitions of
must have
b

f <

+ 1J2

Y
/

446

[a,b]

so we,

that is

j;f
rb

- U2 <U1 .

In other words
/IT

f - U2
a

doesn't exceed Ak upper.sunt \-U1

al

for any partition

of

[a,c]

and hence

for all such

cannot exceed the greatest lower bound of all such upper sums
[st,c],

partitions of

that is
c

f - U <

f.

This can be written as

17,

f < U

a
e

which fells

us, that

f -

partition

of

[c,b]

doesn't exceed any upper sum Apr,party

and hence cannot exceed'the greatest lower bound of

4

such sums, that.is.

rb

-J f<J

f.

.We -have, t4erefore, established the ineg414.ty,

a.

%(4)

f.

the reverse
To complete the proof df Lemma A8-2a we need to establish
inequality
1:;"

f> r f+

(5)

To,do this, suppose

sum V.

c-

It may be assumed that c s a,

"points of
Let

contained in

U1

and

for if not we can add c

(See the rofof Lemma A8-1b).

(a,c]

and

02

the points of

to

with,-

Let. al 1);the
a contained in

Vi denote the.)pper sums obtained from

U by including.
.

848

I
A

with a cor'respvonding upper

[i,b]

is a partition of

'out disturbing the sum U.

[c,b].

.447

'

-onlrthe

`tehntrof

'

II

which correspond to points of

and.. a

2,'

respectively.

Then.

U =

+ U2.
4

r"

Since

'

is an upper sum corresponding to a partition of

[a,c]

and

is the greatest lower bound of all upper sums of all partitions of.

we must tave

(a,c]

f < U

1
4

Similarly, we have

f < U
a
.

so that

(6)

gives

f +

) f'< U.

T
In other words

a1
f+

O doesni,t exceed-gny upper sun

C/

for any partition

of

[a,b],

so.it

cannot exceed the greatest lowbr bound of-such sums, that is

b

<

f <

f +
c

f.
a

------:

This is the desired inequality (5), which-combined with (4) completes the
proOtila Lemma A8-2a.
.

'Lemma A8-2b.

numbers such that

If

is continuous on

[a,b]

and if

.t

m < f(t) < M,

then

for

a < t < b

,
m(b

a) <

b
f <11(b - a ) .
a

J
849

448

m, and- M are

[a,bV

Consider the partition of

al = (a,b)

and the corresponding upper and lower sums

a);
= M(b - a).

= m(b - a).

The number

,,

is the greatest lower bound of all upper sums of all

-

A
a.

partitions

'of

[a,b]

so, since

rb

jaf<U1

is one such partition, we must have

pM(b-a).

The same argument also gives:- .,

b

f > m(b - a).

a

Recall that (see

b

_<
a
-al

f
a

'kr

so that
b

m(b

f <

a) <

f < M(b - a)
a

which gives the degired result.

_The.observant student will note that the, continuity of
*A.

particular role in these lemmas, except to insure that

that the upper and%lower sums can be defined.
an ameitrary boznded function

f.

played no

iS'bohAded so'

Hence, both lemmas hold for

In our third lemma, the Continuity of

is essential.

xLemma A$ -2c.

point of

,n6Aer ,5

(a,b1,.

Suppose 1'
If

is continuous on ja,b.le and that

is a

is agiven posVtive number thpn there ie'a positive

such that
p

where a and

are -khe-respective maximum and minimum viluq'sUNI.rgn the

clOed interval
(4,b1 n (x - 5,x + 5].

850

449

A0-2

r
(-:-

'Proof'.

The. That that

....,

ti.
,

(a,b) n fx - 8,x + 81

poaiive;length is easy to establish.

is a closed interval of

(See Exercises A8-2, No. 1.)

The lemma

**asserts that thedifferenCe between the Maximum and minimum values of

the interval -Ex --8,x + ,81 n

choosing csmall enough.

Theorem 8-

Since

on

can be made as small as we please, b'

Its prgpf makes use of.the definition of limit and

.

is continuous at

We know, that:

x,

v
lim 'f(t) =
t -)x,

Therefore, if

8
1

is any given poditive number, we can find a positive number

such that

(7)

If(t)

if

is in the domain of

and

(8)

- f(x)I < el

o < It - xl -<

The inequality (7) also holds if

x = t,

so.(8) can be replaced by

0 < It - xl <
If% e

is a given positive number, let

(9)

- f(x) I

if' t

is in the domain of

el =

It - xl < 8

on the interval

Let a and

ra,b1

Thus, if

is in the domain of

<
sp that (9)*olds.

81 > 0

so that

1'

..,

then

Choose

< el

8 be a positive number smaller than

x - 8 < t < x + 8

and

o<
*Let

a < t < b

and

and

- xl < a <
p

be the maximum and minimum values'of fo4

1-"8,x + 81

and choose points

and

d,

in this

*-AnterVtil such that

a = f(c);

{The existence of, c; d, a and g

p = f(d)

is guaranteed by Theorem 8-2b).

Therefore,

If(c)

< el 'and

If(d) - f(x)I < el

so that

851

450

"

a-0

-f(d)

= f(c)

f(x)).

f(x) - (f(d)

= f(c).

The triangle inequality ((2) of Section A6-1) gives

a - 0 '<

<

If(e) - f(Ix)I + If(d)_- f(x)

e.

'

el

El

This proves Lemma A8-2c.

is,defined by

The function F

-Proof of Theoregt A8-2.

x
7(x)

f for x

.1

..-,

[a,b].

in

..-

,-,..

PSX) = f(x),

Our purpose is to show that

lim 7(xt)
xt-4x

(10)

If

tf

1(x)

for

f(x);

that is,

'x

in

(ail)].

is a given positive number, use Lemma ;8-2c to find

where a and

8 >`0

are the respective maximum and minimum values of

so that

on the

interval

Suppose
(13)

n [x . 8,x + 8].

[a,b]

(12)
x'

ff

F and

is in the domain of

0< lx' -x! <8

.

so that

xl. is in ja,b]

is a point of (12).

(slnce

is only defined on

x* >x, then

In particular) if

(a;b1) and hence

is a*subinter:

(x,x'l

val of (12) (seeeExercises A8-2, No. 2) and so.

0 < f(t) <a if t,e (x,x1].

Lemma A8-2b, then gives

x'

f <

- x) <

a(X1 -

x).

Jx

--- 1

In this case Lemmi A8:2a gives

X

f =

+-

f"'

84,5.

x'

;t

so that

P(xt):- F'(x)

f -

f =

J-a

J a

and hence (14) gives

OW that is

F(x')

x) <

- F(x)

x' > x)

since we -are here assuming that

F(xl) -T(x)

<
-Subtract

f(X)

<a(x, - x)

x!

a.

throughout to obtain

<Pixt) - F(x)

<a

f(x)

x' - x

- f(x)

and now use the fact that

-f(x) < -a

-f(x),> -0

and
.

(siti'de

is in (12)) to obtain

- a <

F(x')

- F(x)

0.

1r0C, < (i(

"'

Oft

Using (11) we conclude that if

if

..xt > x

x'

is in the domain of

then
17(xi) - F(x)

A similar result holds if

Th1

and (13) holds and

x' < x

f(x)1

and we conclude that indeed (10) is true.

completes the proof of Theorem A8-2 and establishes that'the Integral of

b

continuous function on a'clOsed interval exists.

The Antegral

f ,is.then
a

17

defined to be the common value of

and

4.

853

'f.

452

47-

'

11,11

'"

,S

'Exercises A8-2

--4P0"
i;/

x st(a,b1

Show that if

1.

(Hintil.r Let

-slose'd interval,

[a,b]

(a1;b11

6,

and stow that

b.-vend. x + 6

a('

the -sitna.

is.a

+ 6)

the larger of a : and

hi,

8,x

[a,b]rt (x

t en

6'> 0

an

[x - '6,?cs+ 6]).-

0
v

and

x' > x

Show that if

2.

then

[a,b)

is a

(x,xl]

oft.

la,b1.

. subinterv,1 of

Show %hat

3.

Ja
4

'

Ja

F' = f

I. 'Deduce from Number 3 and Theorem A8-2 that

.

5:

.r

on

el

(a,b1.

is continuous

if

-.

I'''.

s 'continuous on

Show that if

then there is a number

ita,b44,

in

...

[a;b]

such that

4t
N
-

(Hint:

Choose

and, d

in

f(c

are the respective maximum'and mini

f(di)

f(c1)

ch that

[a,b

la ,b]

Show -that

Jf
a

* f(d ) <

b - a

'and'apply the Intermediate Value Theorem).

.

6.

choose

.7.

a < c < b?

so that

Show that if

f(x) > 0

for some

111

[a,b]

f > 0.

then
.1'

:` (Hint:

ShoW that there is. a

_..5,:> 0 _

and

wiihi a <b

and if

(a,b]

is continuous and nonnegative on

Can you then-

Use the Mean Value Theorem tclow that Number 5 ii true.

in >0 such that

f(x)

44

_---

on (a] Ilvfx - 8,x +

-,_

8.

fl(x)

Deduce from Numbef 7 that if

continuous on

[a,b]

then

fof

a <fr x <b and

is strictly increasing on

(a,b1:

s
C

"

854

f'

is

9.

SubpO'se

1 , 0 <X

a
f(X)r-,

2 , 1 < x < 2
(a)

Show directly

from

the defitlition and properties of upper integrals

that:

,0<x<1

F(x)

(b)

Does

1 <,x < 2-

2x - 1

.0

F have *a derivative at

x = 1?

Why doesn't this contradict

Theorem A8-2?

'10. Suppose

is bounded on

[a,b],

and

f.. shoT thEit

l'(x) =

a
is continuous on' [ a ,b] .

for

Make use. of Lemmas A8 -2a, b, which hold

(Hint:

bounded functions.

ti
4,

855

454

Appendix 9
'

.44

LUTIONS TOIFYERERTIAL EQUATIONS

LOGARITHM AND EXPONENTIAL FUNCTIONS AS

The Logarithm as Integral

. A9-1.

The ZI;garithm function

is the unique solution to the problem

loge

(1)N,
and can be expressed in the integral form
x
.

f(x) = log

(2)

x =
e

dt, x > O.

Our purpose ilathis section is to show how the properties of,thy.ogarithm

function can be obtained by using the fact that it is the unique solutiOn to

.(1) and that it is the area from

to

under the graph of

1
t

In order noto be prejudiced by the known properties of the logarithm

let us use the letter

to denote the function defined by

L(x) =

(3).

It will be shown that ,L

is reasonable to write

x > O.

dt

has all the properties of the logarithm and_that it

L(x) = loge x:

,N
L

Certain elementary properties of

are easy to obtain from (3).

First,

note that

L(i)

(4) .

o,

.
.,,1

since' L(1) =

dt =. O.
1

45

Second, the Area Theorem (Section 7-2) gives

From (5) and the fact that

(6)

1
:7 > 0

if

, x > O.
0, we conclude that

0,

L(1) = 0,

while.the values

the valUes

L(x)

L(x? for x > l'

That is,
,(7)

L'(x) > Caltand

L is a strictly increasing function.

In particular, since
less than

(x )

(5)

L(x) < 0,' if

0 < x < 1

857

/"
400

for

0 < x < 1

must be

must be greater than

O.

and

(8)

L(x )" > 6,

'if

x >
1

The secoW derivative of

(9)

is the derivative of' x -4

so that

.e4

, x >

L (x) =

o.

.-- ds negative, ik
The:expression 'Ir1

so that

x f 0,

L' is a concave func-

Already we have'enough information to know that the graph of

A
,
something like that shown in Figure A9-1a.'

tItn.

'looks

N./

Figure A9-la.

The graph of a strictly increasing, concave

through

and passing,

x > 0,

function, defined for

(1,0).
,

The basic logarithm property

L(ab) = L(a) + LW, a ;Op b > 0

(10)

can be obtained by using the

L =

that

i1 the unique solution to the

.problem
f(1) =.0.

f'(x) = X,

(11)

a > 0 -and

For suppose

g -.

is the function defined by

g(x) = L(ax) - L(a).

Certainly g(1) = 0.

Furthermore, since

L(e)

Dg(x) = D(L(ax)):

858

456

is a constant

A The chain rule (with "u = ax, ui = a),

DL(ax) =
In other.words,

then:gives
1

a=

ax

is also a solution to problem (11). Since

solution to this we must have

L = g,

is the only

that is

L(x) = L(ax) - L(a).

Adding L(a)

to both sides and replacing

by b

then gives the result

(10).

The formula "L(ab) = L(a) + L(b)

from

to b.

to

tells us that the area under

ab

is the __gum of the area from

(See Figure A9-1b.)

to

andthe area,from

Figure A9-lb

The area of the shaded regiOn is the area

from

to

plus the area from

to

b.

A
,

Frail (10) we have

L(a2) = L(aa) = L(a),+ L(a

= 2L(a)

.41*

2
2
L(a3) = L(a a) = L(a ) + L(a) = 2L(a) +1,(a) = 3L(a)
and in general
L(an) = nL(a)

"Iithermore, if

if

is any positive integer.

n -is aYositive,integer and

alin
ih
a

859;
A

457

44

971

thin bn

so that
n.

L(a

niAni

that is

Wn)
Suppose

Jan positive integer.

if

= 3;7; L(a)

r is a positive rationa*Mber so that

r =

where m and

awe poldtive integers.. Then

Walin)7)

L(ar) = L(amin) =

mL(alin)
n.

-111 L(a)

that, is,
.

L(ar) = rL(a)

...

if

a positive rational 'number.

This result will, in fact,. be true for ty rational'hudber

r.

If r = 0,

then
L(ar) = L(a0) = L(1) = 0 F OL(a) =

si

If

r < 0,

p . -r

then

is positive and
.crap =

so that

0 = L(1)= L(araP)
L(ar) + L(aP) = L(ar) + 'pL(a);
.

that is,
rt(a).

L(ar) = -pL(a)
In summary:

, 1'

L(ar) = r L(a),

(12)

a > 0

if

and

It will-now be shown that the range of L

and that L

(is)

is rational.

consists of all real numbers,.

In other words:

has(an inverse action.

If

is,any red6humber, there is a unique.

positive real number

suchthat 11(d) = c.

/860

458

A9.-1

c theree positive

. To prbve this we first show that for a given

numbers

and

d
2

such that

(14)

L(d1)

To do this, note that

integer n1

L(2) > 0

<c

from (8)).

and a positive integer

n

< L(d2).

Hence, there is a negative

such that

h2

L(2) < c < n

L(2)

We Can then choose

n
d
= 2
-Jr-

and

It follows that

L(d ) =,L(2

nl.
)

= n

L(2)

and,

L(d2) = L(2

so that

d1

d2

The function

are positive numbers which satisfy (14.

-13` is differentiable for each

no,
).= n2 L(2) >,c

x > 0;

theN.ore, it4is continuous for each

.x > 0 (SactiOn 8-1).. The Intermediate Value Theorem (Section.8-2)

implies that there is a pos/tive real number, d

between d1 ;and

d2

such

that
L(d) = c.

.84

Furthermore,

must be Vague since

is strictly increasing.
-

pletes the proof of (13).

861

,-

This com-

ercises,A9-1
Use upper and lower sums to show that

1.

2.

(a)

<

n > 1

Show. that for each integer

1

< 1 +

k2) <

1 +

Use upper and lower sums to estimate

(Hint:

10
(b)

100

Estimate

ln

n=1
(a)

Show that if

4P

a > 1
1

-.71 <,11,(a)

<a - 1.

u
(b)

Show that if

a > 1

L(2a) > L(a) +

(c)

Show that if' a > 1

1
2

then

L(a) <

L(a) = 2L(4).)

tHint:

lim

4,, Show that

x
5.

Find

(a).

fl(x)

f(x) =

L i11

).tint:

Use No. 3(c).)

03

for each of the followi

IJ()P)

(b), ,f(x) = L(x

..

f(x) = L(L(x))
.

Sketch the graph of

L(x),

using its deriyative.

dt.)

eaX.ponential Functions;

Denoting the inverse of

by 'E,

L'

it follows that

is defined for

c by

each real'humber

if

E(c) = d

L(d) = c,

front ihich ye have

E(1(d)) = d

(15)
.

for each

L(E(c)) =

The values

d > 0

o=-each

of the function E are positive (becquse iA is the

E(x)

inverse of a function whose domain consists of poSitive nuMbers).

Further-

E is strictly increasing and continuous because it is the inverse of

more,,

(See (3)ff, Section 8:11.)

a strictly increasing continuous function.

the function E

For'

the two results toy and (12) now take the form:
E(a + b) = E(a) E(t).

(16)

rEka

r.

= E(a)r

for any rational number

we note that

E(a + b) = E(a)E(b),

Fbr example, to show that

L(E(a + b)) = a b
+11

and that

1(E(a)E(0),= LW) + L(E(b))

= a +b'

so that L(E(a + b)) = L(E(a)E(b)).

have

Since

is strictly increasing we must"

E(a + b') = E(a)E(b).

It

If r

is a rational number then (15) tells us that

ar =*E(L(ar)Y.

Since

L(ar

= r L(a)

we therefore

ar = E(r. L(a)),

(17)

Let us now define

for

if

is rational and

a > 0

a > 0 .

and x arbitrary by

ax = E(x L(a)),

that is, by Wending (17) to all real numbers

x

-definition agrees with the definition of

864

x,' We shall show that this

used in Chapters 5 and

6.

a.
,

The laws of exponents hold for 'our new definition (18).

x+y

= E((x 1-.Y)L(a)) = E(xL(a)

,` a

For example,

yL(a))

so,tnat (.6) giv6s

.

ax+Y = E(xL(a))E(yL(a)) = axaY.

We prove that

(ax)y

axy

as follows.

From (18) we have

(ax)Y. E(Wax)).

We replace 3 by

E(xL(a))

to obtain

(ax)Y =

E(YLN(4a))).

Now use the fact that

L(E(xL(a)));.-- x L(a)

Zan application of the second formula of (15)) to obtain

= E(yx

L(a) = E(xy L(a)).

Now use the definition of powers (18) again to write

E(xy L(a))

We conclude that

(ax.)Y

Note that if
a > 1 -and

= a Y.

x < x2,

is strictly increasing.

so that

L(a) > 0

then

.x

Since

x -*a

the function

a > 1,

L(a) < x

L(a).

fr

is strictly increasing, we must have

a 1

A similar argument shows that

The function ,x --,ax

a )) < E 6c

= Ex1-

L(a)) = a x2

is strictly decrea\sing if

is continuous, for

lim
x

lim E(x L(a))

ax
a
x

x L(a))

= E( lira

(.....feef...

\.\-

864

,A

62

0 < a < 1.

A9-2

"

Since,

(since E is continuous).

lid x L(e) = b L(a)

and

E(b 1)(a))x.= a.

we indeed:have
lim

= a.13`\

x
_

In summary, if arbitrary powers are defined by (18) then the 10; of

\,increasing if

a' > 1,

exponents hold, the furiction

is continuous, and is strictly

It appears that,

0 < a < 1.

strictly decreasing it

indeed, the definition (18)) results in desirable properties for exponential

functions.

The results of Section 8-11 enable us to find the derivative of

x

hence, using the chain rule, the derivative of

1

of
4

L. is the function x -0-

x > 0

x -)a

E,

and

Since the derivative

which ha's only posi,t4ve values, we

know that
,

E'.(x)

The formula

(See (0 ), Section 8-11).

L': x -4'pthen gives

a..

L,"(E(x)).-

so that

El(x) - 1,4(x)) - E(x).

E

In summary, the function

f(x) = E(x)

is, it,s own derivative.

Therefore,

is a solution to the problem

f' = f; f(0) = 1.

(19)

In our preiious discussions (Chapters 5, 6) it was shown that if

lim

(20)

and if

.e

+."

exists and is .k

.0

is defined to be

21/1',

and

ex

to be

2(1/k)x,

then

x ,)ex must be a solution to (19):

We conclude that if (20) holds, then

865

and the function

x -' ex

must be the

same function, that is

(21)

E(x) =.

for all

In 'our setting, arbit;ary powers are defined by

(18)

(20) and (21) are true if we use

First we use the result

'

tive of* f : x

ax.

u(x) = x Lea),

(18),.

Let us show that indeed

to define pbwers.

and the chain rule to find the derive-

We have
ax =

Put

x.

g = E,

L(a)) .`

so that

f(x) = eu(x)).
and, hence, .

f'(x) = rex = g'(u(x))u'(x).

it

Since

g' = g = E and

1.0(x) = L(a) we have

De

= E(x L(a))

L(a)

)
= ax

L(a)

'that is

(22)

f'(x) = ax-L(a),

if

x -4 ax

In particular,

f' (0) = a L(a) = L(a).

Expressing thedefivat,ive as the limit ofa difference quotient, we have:

ft (0) =

lim
h=4 0

h
a

lim

f(0 + h) - f(0)

0+h

a0

lim
h--+ 0

We conclude that

lim a
h -'0

- 1

indeed exists and, in fact

h."

L(a),=

lim
h

464

Put

lim
h -40

- 1

= L(2). 'and

1 /k

that is, since we are using (18) to define powers,

e = E(1- L(2)) .
ir

Since

L(2) = k,

this means that.

0
5

so that' L(e) = 1.

E(1)',

'

tbus;*indeed we have
x
e

= E(x L(e)) = E(x)

Exercises A9-2
1-

1.

2.

'3.

(a)

f(x) = (1 - x).

(b)

f(x). = (L(x))x

(c)

f(x) = xl/x

Find the minimum value of

x -4

Show that if'yo = ay. where

where

fl(x)

Use the definition (18) to find

xx.,

is a coxstant then there is a constant

.41

K such that
.

y = KE(cx).
v.

(Hint:

PUt

i = E(-cx)y

and show that 2.4 = O.)

G

IP

867

4 65

oS2;

4.,

'L

Recall, that

if

s'
r

is au initis4.est/mateSto Vzero

f "then

of

Newto0A-meth od (Sectiony2-10), under, suitable conditiont gi.ges the,

;

better estimate
0

-x

f(x 1l)
TITTY
1

and the subsequent estimates

>

f(x )

xn

f
* - This can be
x

= 2

and

f(x) = L(x) - 1.

-Using

x3.

and

x2

find

the zero of

e,

.sed to estimate

L(2) 4 0.7

fqx n)

xn

5.

(at) .Show that

e =

lim (1 +
0

h
(Hint:

L(1 + h) -

L'(1) = 1 =

lim L(1 +

=
h

(b)

Show that

-) 0

h)1/h)

lim (1 +

e =

L(1)

311)n

n
(c)

Show that

ea =

lim (1 + a)/1
n

The Circular Functions

solutions
The sine and cosine functions can be constructed as 'inverses of

to

0
1

Yi
2

AS in our logaritM disThe followineexercises outline this construction.

they
cussion we shall,introduce new symbols for these functions, then show'that
desired functions.

We let

A. be the function deffnea for

A(x) =
J 0

868

dt.

1 - t

,.

460

< 1--

A9-2
6.

Rind v.0

and show that-+ is strictly increasing

and continuous, and,
a,

hence, has an inverse

7.

(a)

S.

What is sS(0)T
ce

(b)

Show that
St =

byuting the formula for the derivative of the inverse.

(c)

Whatois

(d)

Show that

Let

C = St

(a)

Show that

C" + C = 0.

(b)

Show that

C' =-S.

\:(c)

S1(0).2

S" + S = 0.

and use Number 7.

What is

[C(x)]2 + [S(x)]2 = 1.

Show that
9.

Show that if
Put

.(Hint:

C'(o)?

C(0)?

y(0) = 0

= 0,

y"

and

yt(0) = 1

then

y = S(x).

and use the fact that

z = y
;

0 = (z" + z)z' =
and
10.

1)((zt)2 + M2),

z(0) = zt(0) = 0 to show that

Use Number 9

z = 0.)

to show that

S(x + a = S(x)C(a) + S(a)C(x)

if

x +

X!, a, and

are 'in the domdin of

S,

Remark.

The above defines the functions

zero (that is, for

-1 < t < 1).

in the range of A

functions

and

to all

Cr only for

near

and 4 sbovId that

y =.0, y(0) = 1 W0) = 0 sO,we a e able

y"

S(x) = sin x for

C

and

where A `is defined on the interval

The intuitive dscltissiort of Chapters

y = sin "X 'ls'e solution to

to colplude t

near zero

is discussed in

Volume 2.

869

4 ti

"A Method for exte ding the

ppendix 8, SMSG Cal taus,