You are on page 1of 645

ELEMENTARY DIFFERENTIAL EQUATIONS)))

ELEMENTARY DIFFERENTIAL EQUATIONS)


SixthEdition)

C.Henry Edwards David E. Penney)


The University ofGeorgia)
with the assistance of

David Calvis)

BaldwinWallaceCollege)

PEARSON ---------Prentice
flaIl)

Saddle River, NJ 07458))) Upper

Library

of Congress Cataloging-in-Publication

Data on file.)

Editorial Director, Computer Science, Engineering, and Advanced Mathematics: Senior Editor: Holly Stark Editorial Assistant: Jennifer Lonschein

Marcia J.Horton

Senior Managing Editor: Scott Disanno Production Editor: Irwin Zucker Art Director and Cover Designer: Kenny Beck Art Editor: Thomas BenfaUi Manufacturing Manager: Alan Fischer Manufacturing Buyer: LisaMcDowell Senior Marketing Manager: Tim Galligan)

@ --------- 2008,2004,2000,1996 Inc.


PEARSON

Prentice

Hall)

by Pearson Education, Pearson Education, Upper Saddle River, New Jersey 07458)

Inc.

All rights reserved. No part of this book may be reproduced, in any form or by any means, without permission in writing from the publisher.)

TRADEMARK INFORMATION) MATLAB isa registered trademark For MATLAB product information, The MathWorks, Inc. 3 Apple Hill Drive USA Natick, MA, 01760-2098

ofTheMath Works, Inc. pleasecontact:

Tel:508-647-7000
Fax:

508-647-7101
of Waterloo Inc.
Inc.)

E-mail: info@mathworks.com Web: www.mathworks.coml) Maple is a registered trademark


Mathematica

Maple,

is a registered trademark

of Wolfram

Research,

Printed in the United States

10 9 8 7 6 5 4 3 2

of America
1)

ISBN

0-13-239730-7)
London LTD.,
Australia Pty, Limited, Sydney Singapore, Pte.Ltd North Asia Ltd., Hong Kong Canada, Ltd., Toronto

Pearson Education Pearson Education Pearson Education Pearson Education Pearson Education Pearson Education Pearson Education Pearson Education

de Mexico, S.A.de C.V.


Japan, Tokyo Malaysia, Pte.
Ltd.)))

C
T

f
\"

,..._\"z\037;\037,.

..
,\037/

0N
J
_:'

,
i

\037 ..

::'.

... .\037 :;:

\037:

\037

:f./

\037\037

_'N_\"\"\037:\037

,.;:-

I
-..

i'. :; T E N:. T
t/

..

\037\"

';

_N_\037\"\"_\" :;

.,:>

:\037

';

\037

,:

::

\037 ,_

\037;

:\037

,j

...;.

-,.N'_

,:.

:] .,.

S
_._._\037\"..)

:,.,
)

Preface

vii)

CHAPTER)

First-Order ifferentialEquations 1 D
.1 Differential Equationsand Mathematical odels M .2 Integrals s General nd Particular a a Solutions 0 .3 Slope and SolutionCurves 19 Fields .4 Separablequationsand Applications 32 E 1.5 LinearFirst-OrderEquations 46 .6 Substitution Methodsand ExactEquations 59
1 1 1 1 1 1 1

1)

.7

1.8
CHAPTER)

Models 74 Population Models Acceleration-Velocity

85)

2
.
'\"\". > /\037,

LinearEquationsof HigherOrder 100


2.1 2.2 2.3 2.4
2.5
2.7

,z

\037

\037

.. ,
\"riP'
\037,)

Mechanical ibrations 135 V

General olutions f Linear Equations 113 S o C Equationswith Constant oefficients 124 Homogeneous

Second-Order Equations 100 Linear Introduction:

C Nonhomogeneous Equationsand Undetermined oefficients148 2.6 Forced scillations Resonance162 O and Electrical Circuits 173 2.8 Endpoint Problems nd Eigenvalues a
180)

CHAPTER)

3
\"'. .::::::

Power SeriesMethods) 194)


3.1 3.2 3.3 3.4
3.5
3.6)

......-.,)
J{
\037\037:::'

Series 194 Introductionand Review of Power Series Solutions NearOrdinary Points 207 RegularSingularPoints 218

The Cases233 Methodof Frobenius: Exceptional

Bessel's Equation 248

F Applicationsof Bessel unctions 257)

v)))

VI

Contents)

CHAPTER)

4)
CHAPTER)

T 4.1 Laplaceransforms and InverseTransforms 266 4.2 Transformation of Initial ValueProblems 277 4.3 Translationand Partial Fractions 289

Transform Methods 266 Laplace

4.4 4.5 4.6

and of Derivatives, ntegrals, Products Transforms 297 I Periodic Piecewise and ContinuousInput Functions 304 a Impulsesnd DeltaFunctions 316)

5)

5.1 First-OrderSystems and Applications 326 5.2 The Methodof Elimination 338 and 5.3 Matrices LinearSystems 347 5.4 The EigenvalueMethodfor Homogeneous Systems 366 5.5 Second-Order A Systems and Mechanical pplications 381 5.6 Multiple EigenvalueSolutions 393
5.7

Linear Systemsof DifferentialEquations 326)

5.8 CHAPTER)

Matrix Exponentials Linear Systems 407 and Linear Nonhomogeneous Systems 420)

NumericalMethods) 430)
Euler'sMethod 430 6.1 Numerical Approximation: Look 6.2 A Closer at the Euler Method 442 6.3 The Runge-KuttaMethod 453 6.4

6)
CHAPTER)

Numerical Methods Systems for

464)

NonlinearSystemsand Phenomena 480


7.1
7.2 7.3
7.4

7)

7.5 7.6

480 Stabilityand the PhasePlane 488 Linear and Almost Linear Systems 500 513 Models: Predators Competitors and Ecological 526 NonlinearMechanical ystems S i Chaosn Dynamical Systems 542)
Equilibrium

Solutions Stability and

References FurtherStudy 555 for Existence Uniqueness and ofSolutions559 Appendix: Answersto Selected Problems573
Index
1-1)))

P'REFACE)
he evolution of the presenttext in successive editions is basedon experience differential equations coursewith an emphasis on teaching the introductory conceptual ideas and the use of applications and projectsto involve students in active problem-solving experiences. various points our approach reflects the At use widespread of technical computing environments likeMaple,Mathematica,and MATLAB for the graphical, numerical, or symbolicsolution of differential equations.Nevertheless, continue to believe that the traditional elementaryanalytical we methods of solution are important for students to learn and use.One reason is that effective and reliableuse of computer methods often requirespreliminary analysis using standard symbolictechniques;the construction of a realistic computational modeloften is basedon the study of a simpleranalytical model.)

.....N\037\"

Princinal eaturesofThisRevision F
.../N;....//lfll\"'.M...'NN'..N'N\"/h'//..\037N;I'In' ...I::/\"...N/\"/N//N//..#NI'NJIVI'N\"\"\"\"......./H'N Nd.lh./hNh./\037\"'\" /I'Nk>' nN>Vfl \037N'..fl_ nv..;.;\302\253/.........x'..y,\037....'/Y.'h;;I.N'N/.9\037//JP...I#//N/N..\"\"'*NhYn.N'N'.IN\"'N\037 N.lN.N'.\302\253.t'N'/N.lN'.I/NIN'.IN..Nh>vVN.I-v.,/Nv'''''-vN.l-v.,/N.N'NhV>....

N.lN.lN.I'XVN.N.N\"'....-><N

.v\"/,/\"\"\",/\"\"/N'.IN.I/hnv,,..MN'I<NN.I\"NN-v\"\"-'

:Jo.'.dN\"\"'N\037\"'\037N'-'..\037N'-'..I<N\037..

I'MJV.I

.....,......

\037

I.JVN.lMv><NY\"\"1'

N..;\\V\".....\"\"\"'..,..........................

....,...

<'.............\037\"'.....v.v...............

W -.<v.....

.....,

..............'\"'

,..........,.,

........

...............\".....,..,_...\037

\"\"

\037,\037

,\302\253'i'.\037.'''''

\037)

successful eatures of preceding editions have beenretained, the exposif tion has beensignificantly enhanced in everychapter and in most individual sections of the text. Both new graphicsand new text have been inserted where neededfor H improved student understanding of key concepts. owever, the solidclass-tested chapter and sectionstructure of the bookis unchanged, so class notes and syllabi will not require revision for use of this new edition.The following examples f this o revision illustrate the way the local tructure of the text has been augmented and s polishedfor this edition.)
While the

additionaldetailsinserted of homogeneity for first-order equations (page61); in the derivation of the rocketpropulsion equation (page95),and new Problem 5 insertedto investigate the liftoff pause of a rocket on the launch pad sometimes observed before blastoff (page97).

and Chapter New Figures showing directionfields that illuso trate failure of existencend uniqueness f solutions (page24);new Problems a 34 and 35 showing that small changesin initial conditionscan make big differencesin results,but bigchangesin initial conditions may sometimesmake only small differencesin results(page30);new Remarks 1 and 2 clarifying the conceptof implicit solutions (page35);new Remark clarifying the meaning

1:

1.3.9 1.3.10

Chapter2: New explanation of signs and directions of internal forces in new mass-springsystems(page 101); introduction of differential operators n and clarificationof the algebra of polynomial operators (page 127); ew introduction and illustration of polar exponential forms of complexnumbers fuller (page132); explanation of method of undeterminedcoefficientsin Exnew amples1 and 3 (page149-150); Remarks 1 and 2 introducing \"shooting\" and why someendpoint) illustrating terminology, and new Figures2.8.1 2.8.2
VII)))

..

VIII

...

Preface)
value problems have infinitely many solutions, while others have no solutions different types of at all (page181); Figures2.8.4 2.8.5 and new illustrating eigenfunctions (pages183-184). Chapter New Problem 35 on determination of radii of convergence of new solutions of differential equations (page218); Example3 power series t before the subsection logarithmic cases the method of Frobenius,o on in just illustrate first the reduction-of-orderformula with a simplenon-series roblem p

3:

(page239).

istenceof Laplacetransforms (page273);new Remark discussingthe mechanics of partial-fraction decomposition (page 279);new much-expanded t discussionof the proof of the Laplace-transform existenceheorem and its

4 Chapter :New discussion clarifying functions

of exponential orderand ex-

extensionto include the jump discontinuities that play an important rolein many practical applications (page286-287). 5: for Chapter New Problems20-23 student exploration of three-railwayn cars systemswith different initial velocity conditions (page392); ew Remark the relation betweenmatrix exponential methods and the generillustrating new alized eigenvalue methods discussedpreviously (page416); exposition insertedat end of sectionto explain the connection between matrix variation of parameters here and (scalar)ariation of parameters for second-order v equations discussed in Chapter 3 (page427). previously and 6: clarifying Chapter New discussionwith new Figures6.3.11 6.3.12 in the difference betweenrotating and non-rotating coordinate systems moonearth orbit problems 473). (page Chapter7: New remarks on phase plane portraits, autonomous systems, and critical points (page488-490); introduction of linearized systems new Lorenz and new 3-dimensional illustrating (page502); Figures6.5.18 6.5.20 and Rosslertrajectories (page552-553).)
figures show students Throughout the text, almost 550computer-generated of direction fields, solution curves, and phaseplane portraits that bring pictures symbolicsolutions of differential equations to life. About 15application modulesfollow key sectionsthroughout the text. Their students is more extenis to add concrete ppliedemphasisand to engage purpose a sive investigationsthan afforded by typical exercises problems. and A solid numerical emphasis provided where appropriate (as in Chapter 6 on Numerical Methods) the inclusion of generic numerical algorithms and a limited by number of illustrative graphing calculator, BASIC,and MATLAB routines.)
vivid

and Organization C?Iltent)


f The traditional organization of this text still accommodatesresh new material and combinationsof topics. instance:) For

. .

The final two sectionsof Chapter 1 (on populations and elementary mechanics)offer an early introduction to mathematical modeling with significant ap-

plications. The final sectionof Chapter 2 offers unusually

early exposureto

endpoint)))

Preface)

IX)

. . . . .
in

problemsand eigenvalues, with interesting applications to whirling strings and buckledbeams. a Chapter 3 combines completeand solidtreatment of infinite seriesmethods functions in its final section. with interesting applications of Bessel Chapter 4 combinesa completeand solid treatment of Laplacetransform
methods with brief coverage of delta functions and their applications in
its final

section.

Chapter 5 provides an unusually flexible treatment of linear systems. Sections and 5.2 offer an early, intuitive introduction to first-order systems and models. he chapter continues with a self-contained treatment of the T necessarylinear algebra, and then presentsthe eigenvalueapproach to linear systems. It includesan unusual number of applications (ranging from brine tanks to railway cars)of all the various cases f the eigenvaluemethod. The o from earlier edicoverage of exponential matrices in Section5.7is expanded

5.1

ti

ons.

with the elementary Chapter 6 on numerical methods beginsin Section6.1 with the RungeEuler method for singleequations and ends in Section6.4 Kutta method for systems and applications to orbits of cometsand satellites. and Chapter 7 on nonlinear systems phenomenaranges from phaseplane analto ecological mechanical systems an innovative concludingsection and to ysis an on chaos and bifurcation in dynamical systems.Section7.6presents eleintroduction to such contemporary topicsas period-doublingin biomentary a t logicalnd mechanical systems,he pitchfork diagram, and the Lorenz strange attractor (all illustrated with vivid computer graphics).)

This bookincludesadequate material for different introductory courses varying length from a singleterm to two quarters. The longer version, Elementary contains Differential Equations with Boundary Value Problems(0-13-600613-2), additional chapters on Fourier series ethods and partial differential equations(inm cluding separation of variables and boundary value problems).)

:,\\p\037li\037

\037\037__.______)
To samplethe range of applications in
tion s:)
this

text, take a look at the following

ques-

. . . . . .

ulations? (Section 1.7) How do a unicycle and a two-axle car react differently to road bumps? (Sections 2.6 5.5) and are flagpoles hollow instead of solid?(Section .6) 3 Why If a mass on a spring is periodicallystruck with a hammer, how does the behavior of the mass dependon the frequency of the hammer blows? (Section

explainsthe commonly observedlag time between indoor and outdoor 1.5) (Section daily temperature oscillations? and extinction in alligatorpopWhat makesthe differencebetween doomsday
What

4.6)

If a moving train hits the rear end of a train of railway cars sitting at rest, how can it happen that just a singlecar is \"popped\" off the front end of the second
train?

(Section.5)))) 5

Preface)

. . .

comet? (Section 6.4)


What

Howcan you predictthe time of next perihelion passageof a newly observed

determineswhether two species ill live harmoniously together, or w whether competition will result in the extinction of one of them and the survival of the other? (Section .4) 7 and when does non-linearity lead to chaosin biological nd mechanical a Why 7.6)) systems?(Section

and M Applications Solutions anuals


(\"h'/\037'\302\273'I\037\",Z\037'>W//\"//hW\"\"NV,:V/hV/.\302\253>'//h':.'..w\"'h\"..:-:?:>'h/\"'/hhlh\"''''Yh////hlh/'''''''h/\"<'1'('v/Y/lhl.-wI\"hW//n'I\302\273\"\037h/lY/h//I'MYl'h//n'...:q\037.\302\253\302\253(o\"/.o(.Q'\037\037'>mY/..mm'MX-V.\302\253\302\253(o\"\037..vh;\302\253,..wI\302\273\",c.y'-\302\273\302\273y\037\"M\"\"\037..w\037o\302\273\302\273\302\273(\037.....v...\037\037..:.w..w\037....WIY/.../_..w\037\037..w'(O\037W/..wh\037Y....../..:vhV/..:.w..w.\302\253r\302\273\302\273'\037)

The answer sectionhas beenexpanded to its considerably increase value as a learning aid. It now includesthe answersto most odd-numbered roblemsplus a good p M The 60S-pagenstructor's olutions anual (0-13I S many even-numbered ones. 600614-0) accompanying this book provides worked-out solutions for most of the

t\\\037o\037\037m\037

\037____)
of different computational systems.)
In preparing this revision we profited greatly from the advice and following very capableand perceptive reviewers:

M StudentSolutions anual (0-13-600615 problemsin the book,and the 345-page 9) contains solutions for most of the odd-numbered roblems. p The approximately 15application modules the text contain additional probin lem and projectmaterial designedlargely to engage students in the exploration and considapplication of computationaltechnology. Theseinvestigationsare expanded the that accompanies Manual (0-13-600679-5) erably in the 320-pageApplications text and supplements with about 30 additional applications modules. ach section E it in this manual has parallel subsections \"Using Maple,\"\"Using Mathematica,\" and methods and techniques of each sys\"Using MATLAB\" that detail the applicable to comparethe merits and styles tem, and will afford student users an opportunity

of assistance the

Raymond A. Claspadle, University of Memphis

Irfan

SemionGutman,
University University

MiklosBona,
It

of Oklahoma of Florida

of Wisconsin-Platteville Carl Lutzer, Rochesternstitute of Technology I SigalGittlieb, Dartmouth University of Massachusetts,


University

UI-Haq,

is a pleasure to (onceagain) credit Dennis Kletzing and his extraordinary TEXpertisefor the attractive presentation of both the text and the art in this book. a Finally, but far from least, I am especiallyhappy to acknowledge new contributor to this effort, David Calvis,who assistedin every aspectof this revision and
contributed tangibly
to the improvement of every chapter in the book.)

C.H.E.
h.edwards@mindspring.com)))

A i E E L E M.. N T,. R y
.: \" .,..,' '...........' \\ ...

,.'.d

d.......;

. .. .

,......

;:
...Z ;\037\037\037 ,.,: ,',' .x'''... ..

. '
v v

\037VM;l ..

\".>..

'.

!
\037

:\"'

r\"

\"'W.\037...:

. . ,.;
N{

.,\037'

c:::,\302\273,\"

\037

D
._,......,.........,;.;x. ...,,,,,,,,.,

E
.'\302\253.0\"/'
..,';....,..........,.,.,,\037

a
.
'.

:,..d (;.: I' F F E R :'.'.'


:.\"'\"::, ,....,.\037\":.-.v... i
r..-. \302\267

(09'

...>:...;.:.\037.\037<::

EN T A L S U AT I Q N::.
i
\"

\037

r\037

'\"'.''''
j

!!

,::\037

.v

'-

',:/{

,.._..y.;I.\037..

:<-)

it

\037.: (o,:;A.

: '

. (',w...,
'.

'.

I ' ',.
i
\037\" .\037\"

,.

(.

...:

'..N'

\037

\037;

\037i/':

\037

,,:;

......x\037?'.<;.::\037y

\037:.

\"?\037;-:\037')))

First-Order DifferentialEquations)

11II

Diff'\037\037\037\037ti\0371.\0379\037.\037.tion_s_

\037.\037<!__\037.\037th.:\037_\"1.;}t.i<:al

\037o<!els.)

he laws of the universe are written in the language of mathematics. Algebra is sufficient to solve many static problems,but the most interesting natural phenomena involve change and are describedby equations that relate changing quantities. is the rate at which Because derivative dxjdt = f'(t)of the function the the quantity x = f(t) is changing with respectto the independent variable t, it is natural that equations involving derivatives are frequently used to describethe changing universe. An equation relating an unknown function and one or more of its derivatives is called differential equation.) a

......

Example

1)

The differential equation)

dt
function

dx =

x2 + t 2
derivative

involves both the unknown differential equation)

x(t) and its first

x/(t) = dxjdt.The

dy -+3-+7y=O dx2 dx

d2y
y

involves the unknown function derivativesy/ and y\" of y.

of the independent variable x and the

first two

.)

The study of differential equations has three principal goals: differential equation that describes a specifiedphysical situation. 2. To find-either xactly or approximately-theppropriate solution of that e a equation. To interpret the solution that is found.)
1)))

1.To discover the

3.

Chapter First-OrderDifferential
1

Equations)

In algebra,we typically seek unknown numbers that satisfy an equation the suchas x3 + 7x2 llx+41= O. By contrast, in solving a differential equation, we are challenged find the unknown functions y = y(x) for which an identity such to as y'(x) = 2xy(x)-that s, the differential equation) i

dy

dy = dx 2xy
will want

-holds some interval on


Example2)
If C is a constant and)
then)

of real numbers. Ordinarily, we solutions of the differential equation, if possible.)

to

find

all

y(x) = Ce

x2
,)

(1))

=C
\037\037

(2xe )

x2

= (2x)
in

x2 (ce ) = 2xy.

Thus every function y(x) of the form of-the differential equation)

thus Eq. (1) satisfies-and is a solution

dx

= 2xy

(2))

for all x. In particular, Eq. (1) defines an infinite family of different solutions of constant C. By the this differential equation, one for eachchoicef the arbitrary o method of separation of variables (Section it can be shown that every solution 1.4) of the differential equation in (2)is of the form in Eq. (1).

.)

Models Differential a EquationsndMathematical


The following three examples of illustrate the process translating scientificlaws and principlesinto differential equations. In eachof these examplesthe independent in variable is time t, but we will seenumerous examples which somequantity other than time is the independent variable.)

Example3)
Temperature
A)

Newton'slaw of cooling may be stated in

this way: The time rate of change (the rate of change with respect time t) of the temperature T(t) of a body is proporto tional to the difference betweenT and the temperature A of the surrounding medium That is,) (Fig.

1.1.1).

dT
dt)

= -k(T

A),

(3))

Temperature

T)

FIGURE

the cooling,Eq.(3),describes coolingof a hot rock in water.)

1.1.1. Newton's

law

of

where k is a positive constant. Observethat if T > A, then dTjdt < 0,so the B function of t and the body is cooling. ut if T < A, temperature is a decreasing then dTjdt > 0,so that T is increasing. Thus the physicallaw is translated into a differentialequation. If we are given the values of k and A, we should be able to find an explicitformula for T(t), and then-with the aid of this formula-wean predictthe future temperature of the c
bod\037
...... ...... ....

.)

Example4)

Torricelli's impliesthat law


draining tank in the tank:)

(Fig.1.1.2)

the time rate of change of the volume V of water in a is proportional to the squareroot of the depth y of water

dV
dt)

= -k\037,

(4))))

Models 3) Differential Equationsand Mathematical


\302\267

where k is a constant. If the tank is a cylinder with vertical sidesand cross-sectional area A, then V = Ay, so dV/dt = A (dy/dt). In this case (4) takes the form Eq.

dy =
where h

= k/ A is a constant.)

dt)

-h..jY,

(5))

.)

The time rate of change of a population pet) with constant birth and death rates is, in many simplecases, proportional to the sizeof the population. That is,

dP = kP' dt
that

(6)

where k is the constant of proportionality.

Let us discussExample5 further. Note first


kt pet) = Ce

each function of the form

(7))

is a solution of the differential equation


FIGURE Torricelli'saw l of draining, Eq.(4),describes the draining of a water tank.)

1.1.2.

dP -=kP dt
in

(6).We

verify this

assertionas follows:
kt kt P'(t)= Cke = k (Ce ) = kP(t)

of eachfunction of the form given in all such functions are solutionsof Eq. (6). Thus, even if the value of the constant k is known, the differential equation kt dP/dt = kP has infinitely many differentsolutionsof the form pet) = Ce , one for eachchoicef the \"arbitrary\" constant This is typical of differential equations. o It is also fortunate, because may allow us to use additional information to select it from among all thesesolutions a particular one that fits the situation under study.) for all real numbers

t. Because substitution

(7)into Eq. (6)producesan identity,

C.

Example6)

the population at time t = 0 (hours, h) was and that the population doubled after 1 h. This additional information about pet) yieldsthe followingequations:

kt Supposethat pet) = Ce is the population of a colony of bacteria at time t,

that

1000,

It follows that of k the differential equation in

= 1000 P(O) = Ceo = C, = 2000 P(l)= Cek . With C = 1000 that ek = 2,so k = In 2 0.693147. and
\037

this

value

(6)is

Substitution

dP = dt (In2)P (0.693147)P. in of k = In 2 and C = 1000 Eq. (7)yieldsthe particular solution


\037

= pet) = 1000e(ln2)t

1n2 .2 1000(e = 1000 t )t

e (because1n2 = 2)) = 1.5)

that satisfies the given conditions. We can use this particular solution to predict number of future populations of the bacteria colony. For instance, the predicted bacteria in the population after one and a half hours (when t is)

.2 = P(I.5) 1000 3/ 2

\037

2828.)

.)))

4
8 6 4 2 0

Chapter
1

First-OrderDifferential Equations)
3)

C= 12 C= 6 C=

Q...

-2
-4 -6 -8 -2 C= -12)
t)

The condition P(O) = 1000 Example6 is calledan initial conditionbein causewe frequently write differential equations for which t = 0 is the \"starting time.\" Figure showsseveral different graphs of the form P (t) = Cekt with = In 2.The graphs of all the infinitely many solutions of dPjdt = kP in fact fill k of the entire two-dimensional plane,and no two intersect.Moreover, the selection e amounts to a determination of P(O).Because xanyone point Po on the P-axis that actly one solution passesthrough eachsuch point, we seein this case an initial condition P(O) = Po determines unique solution agreeing with the given data.) a

1.1.3

Mathematical Models)

FIGURE

P(t) = Ce

1.1.3.= Graphs
kt

of

with

In 2.)

which of process mathematical modeling (Fig.1.1.4), involves the following:

Our brief discussion population growth of

in

5 Examples and 6 illustrates the crucial

of a real-world problemin mathematical terms; that is, the constructionof a mathematical model. 2. The analysis or solution of the resulting mathematicalproblem. The interpretation of the mathematical resultsin the context of the original real-world situation-for example,answering the question originally posed.)

1.The formulation

3.

FIGURE

The o 1.1.4. process f mathematical

modeling.)

In the population example,the real-world problemis that of determining the m c population at somefuture time. A mathematical odel onsistsof a list of variables(P and t) that describe given situation, together with one or more equations the to relating thesevariables (dPjdt = kP, P(O) = Po) that are known or are assumed

f hold. The mathematical analysis consistsof solving theseequations (here, or P as a function of t). Finally, we apply thesemathematical resultsto attempt to answer the original real-world question. As an exampleof this process, think of first formulating the mathematical the modelconsistingof the equations dPjdt = kP, P(O) = 1000, describing bacof teria population of Example6. Then our mathematical analysis there consisted = 2t for the solution function P(t) = 1000e(1n2)t 1000. as our mathematsolving ical result. For an interpretation in terms of our real-world situation-theactual bacteria population-we to substituted t = 1.5 obtain the predictedpopulation of P (1.5) 2828acteria after 1.5ours. If, for instance, the bacteria population is b h growing under ideal conditions of unlimited spaceand food supply, our prediction we may be quite accurate, in which case concludethat the mathematicalmodel is
\037

quite adequate for studying this particular population. differential On the other hand, it may turn out that no solution of the selected accurately fits the actual population we're studying. For instance, for no equation choicef the constants C and k doesthe solution P(t) = Cekt in Eq. (7)accurately))) o

M 1.1 Differential Equationsand Mathematical odels 5)


concludethat the differential equation dPjdt = kP is inadequate for modeling the world population-,-which in recent decadeshas \"leveled off\" as compared with the steeplyclimbing graphs in the upper half (P > 0) of

describe actual growth the


centuries.We
must

of the human population of the world over the past few

With Fig.1.1.3.

sufficient insight, we might formulate a new mathematical model differential equation, one that that takes into including a perhapsmore complicated account such factors as a limited food supply and the effect of increased population on birth and death rates. With the formulation of this new mathematical model, we in a counterclockwise may attempt to traverse onceagain the diagram of Fig. manner. If we can solve the new differential equation, we get new solution functions to comparewith the real-world population. Indeed,a successful opulation p analysis may require refining the mathematicalmodel still further as it is repeatedly

1.1.4

might affect our bacteria population. This made the mathematical analysis quite simple, t A perhapsunrealistically so. satisfactory mathematicalmodel is subjecto two cont I tradictory requirements:t must be sufficiently detailed to representhe real-world situation with relative accuracy,yet it must be sufficiently simpleto make the maththe ematical analysis practical. If the model is so detailed that it fully represents physicalsituation, then the mathematical analysis may be too difficult to carry out. If the modelis too simple,the resultsmay be so inaccurate as to be useless. hus T there is an inevitabletradeoff betweenwhat is physically realistic and what is mathThe ematically possible. construction of a model that adequately bridgesthis gap betweenrealism and feasibility is therefore the most crucial and delicate step in the process. ays must be found to simplify the model mathematically without W sacrificing essentialfeatures of the real-world situation. Mathematical modelsare discussedthroughout this book. The remainderof this introductory sectionis devoted to simpleexamples and to standardterminology usedin discussing differential equations and their solutions.)

measuredagainst real-world experience. But in Example6 we simply ignoredany complicating factors that

and Examples Terminology)


Example7)
If C is a constant and y(x) =

1j(C x), then


dy

dx
if

xi-C.

= (C - x)2
y(x) =
1

2
Y)

Thus)

C-x)

(8))

defines a solution of the differential equation)


dy -=y2 dx
(9))

on any interval of real numbers not containing the point x = C.Actually, Eq. (8) defines a one-parameter family of solutions of dyjdx = y2, one for each value of the arbitrary constant or \"parameter\" C.With C = 1 we get the particular solution)

I-x) that satisfies the initial condition y(O) = 1. indicated in Fig.1.1.5, solution this As is continuouson the interval (-00, )but has a vertical asymptote at x = 1. .))) 1

y(x) =

Chapter
1
NUl Nu,\"'

First-OrderDifferential Equations)
\"\"\"''''''I'. \"\"',

M,,\",..,

UW

u,.,.

Example8)

Verify

that

the

function

y(x) = 2xl/2 x1 / 2 lnx satisfies the differential equation) 4x2y\"

+ y = 0)

(10))

for all x > o.)

Solution

First we compute the derivatives)

y/(x) =
Then substitution

_!x-/ 2 1nx
1

and

y\"(x)

= ix-3/2Inx 4x-3/2.)

into

Eq. (10)yields)

3/ + y = 4x2 (ix-3/2Inx!x-2) + 2xl/2 x / 2 lnx = 0) if x is positive, so the differential equation is satisfiedfor all x > o.)

4x2y\"

.)
(11))

that it

The fact that we can write a differential equation is not enough to guarantee has a solution. For example,t is clearhat the differential equation) i t
(y/)2

+ y2 = -1)

has no (real-valued) solution, because sum of nonnegative numbers cannot be the negative. For a variation on this theme, note that the equation)
(y/)2

+ y2 = 0)

(12))

obviously has only the (real-valued) solution y(x) = O. In our previous examples any differential equation having at leastone solution indeedhad infinitely many. The orderof a differential equation is the orderof the highest derivative that appearsin it. The differential equation of Example8 is of secondorder, those in 2 Examples through 7 are first-order equations, and)
y(4)

+ x2y (3) + x5y = sin x)

is a fourth-order equation. The most general form of an nth-order differential with independent variable x and unknown v function or dependent ariable equation

y=y(x)is)

/\" F (x,y,y,y,...,y 0 ) =,)


Cn\302\273

(13))

where F is a specific real-valued function of n + 2 variables. Our use of the word solution has been until now somewhat informal. To be we precise, say that the continuous function u = u (x)is a solutionof the differential u\", , u ) exist equation in (13)on the interval I provided that the derivativesu', on I and) F (X,U,U,u ) =) 0

...

Cn

for all x in For the sake of brevity, we may say that differential equation in (13)on

I.

/\",...,u
Cn\302\273

I.

= u (x) satisfiesthe

an open interval is function can qualify

Remark:Recall from elementary calculusthat

interval.

a differentiablefunction on continuous there. This is why only a continuous necessarily as a (differentiable) solution of a differential equation on an
.)))

1.1 Differential Equationsand Mathematical Models 7)


\"\" \"\" \037 A'''NA'n\037N;,'Nn''''nn,..,

Example7)
Continued

showsthe two \"connected\" ranchesof the graph y = 1/(1 x). The b Figure left-hand branch is the graph of a (continuous) solution of the differential equation branch is the graph y/ = y2 that is definedon the interval (-00, The right-hand of a different solution of the differential equation that is defined (and continuous) on the different interval (1, 0).So the singleformula y (x) = 1/( x) actually 1 ( defines two different solutions (with different domains of definition) of the same differential equation y/ = y2.

,____

'\"\" \"\"\"'''''''''''''...0''....

1.1.5
'\" .......\037....... \"\"'.., '-'

'\"

\"\"\"'\",\"\"\",,

.........,.,.......

\"\"'.......\"

....

\"\"',..........\"'\"

\"\"'''''.......

\"\"

_,

...........\"

'\"

\"'\"

...

\"\"

...

..,..

_...

\"A\"

...

.\037.\037.

\037

...

oJ

_\"\"._.\"..\037_

...\".......

\037\037 _\037

\037'n'\037\"\"\"\"'\"

1).

.)

\037

.\".......

.....

..,.........

.....

.....

Example9)
5)

If A and B are constants and)

y(x) = A cos3x+ B sin 3x,)


then

(14))

two

successive differentiationsyield)

x=l)
;;:.-.

0)

3x, (x) = -3Asin 3x + 3B cos = -9Acos3x 9Bsin 3x = -9y(x)) y\" (x)


y/

for all x. Consequently, Eq. (14)defines what it is natural to call a two-parameter differential equation) family of solutionsof the second-order -5
-5)

o
x)

5)

y\"

+ 9y = 0)

(15))

FIGURE
y/

= y2 defined by y(x) = 1/(1 x).)

1.1.5. solution The

of

on the whole real number line. Figure

solutions.

1.1.6 the graphs of several such shows

5)

the differential equations in (11) (12)are exceptionso the gent and Although eral rule, we will seethat an nth-order differential equation ordinarily has an nparameter family of solutions-one involving n different arbitrary constantsor pa-

rameters.
tion

;;:.-.0)

as an implicitly definedfuncordinarily assumethat any differential equation under study can be solvedexplicitly for the highest derivative that normalform appears;that is, that the equation can be written in the so-called (n) (n-I) , G (16) y
In both

causes F complications. or this reason,we will

and Eqs. (11)

the of (12), appearance y/

/\" x,y,y,y,...,y)

-5

-3)

o
x)

3)

FIGURE
solutions
Y2(X) Y3(X)

= 2 sin 3x, and = -3cos3x + 2 sin 3x of the differential equation y\" + 9y =


O.)

YI

1.1.6. three The =


(x)

3 cos3x,

where G is a real-valued function of n + 1 variables. In addition, we will always seek real-valuedsolutions unlesswe warn the readerotherwise. only All the differential equations we have mentioned so far are ordinary differential equations, meaning that the unknown function (dependent ariable)depends v on only a single independent variable. If the dependentvariable is a function of two or more independent variables, then partial derivativesare likely to be involved; if they are, the equation is called partial a differential equation. For example,the temperature u = u (x,t) of a long thin uniform rod at the point x at time t satisfies t (under appropriate simpleconditions)he partial differential equation

- ' --kau at a2u

ax2

where k is a constant (called thermal diffusivity the of the rod). In Chapters 1 7 we will be concernednly with ordinary differential equations and will through o refer to them simply as differential equations. In this chapter we concentrateonfirst-order differentialequationsof the form
dy

dx

= f(x,y).)

(17))))

Chapter First-OrderDifferential
1

Equations)

also will samplethe wide range of applications of such equations. A typical mathematical modelof an appliedsituation will be an initial value problem, consistingof a differential equation of the form in (17)together with an initial condition y(xo) = Yo. Note that we call y(xo) = Yo an initial condition whether or not Xo = O.To solvethe initial value problem
We
\037)

means to find Eq. (18)on someinterval containing Xo.)

= f(x,y), y(xo) = a differentiable function y = y (x) that dx = I/(C x) of the


dy

dy

Yo)

(18)) satisfies both conditions in


.....

Example10)

Given the solution y(x)

discussedin Example7, solve the initial value problem

differential equation dyldx

y2

dx
Solution
We

2 = y, y(l) = 2.)

initial

find a value of C so that the solution y(x) = I/(C x) satisfies the condition y(l) = 2.Substitution of the values x = 1 and y = 2 in the given solution yields)

need only

5)
I

so 2C 2
y = 2/(3

2 = y(l)

C_
this

'

1,and henceC =

\037.

With

value of C we obtain the desired)


2x)

- 2x)
(1,
2))

I I I I I

solution)

y(x) = 3
x = 3/2)
2)

-2 --x 3
1

\\
:>-..0)

showsthe two branchesof the graph y = 2/(3 2x). The left-hand Figure branch is the graph on (-00, of the solution of the given initial value problem the point (2, 2) and is y/ = y2, Y (1) = 2.The right-hand branch passesthrough therefore the graph on (0) of the solution of the different initial value problem
\037)

1.1.7

-5

y/

= y2, y(2) = -2.

(\037,

.)

-5)

o
x)

5)

= y2 defined by y(x) = 2/(3 2x).)


y/

FIGURE

1.1.7. solutions The

of

The central question of greatest immediateinterest to us is this: If we are given a differential equation known to have a solution satisfying a given initial condition, how do we actually find or compute that solution? And, oncefound, what can we do with it? We will seethat a relatively few simpletechniques-separation of variables (Section solution of linear equations (Section 1.4), 1.5), elementary substitution methods (Section 1.6)-are to enableus to solve a variety of enough first-order equations having impressive applications.)

-\037\302\243\037\037\037

\037\037\037

In

Problems

given function is a solution Throughout theseproblems, spectto

1 through

x. 1.y/ = 3x2; Y = x 3 + 7 2x 2.y/ + 2y = 0; y = 3e3.y\" + 4y = 0; YI = COS2x, Y2 = sin 2x 4. y\" = 9y; YI = e3x , Y2 = e3x)

--------,-------------------------------------------------------) --------------each 5. = + 2e-; = - e12, of 6. + 4y' + 4y = 0; = e- , = xereprimes denote 7. - 2y' + 2y = 0; = cosx, = x


verify by substitution

that

y/

eX

the given differential

equation.
with

derivatives

y\" y\"

2x

YI

Y2

2x

2x 8.y\" +y = 3 cos2x, = cosx-cos = sinx-cos 2x, 1 9. y/ + 2xy 2 = 0; y = 1 +X2


YI Y2

YI

eX

Y2

eX sin

10. 2y\" + xy/ x

= In x;YI = X

In

x,Y2 =

1 -x)))

In

M 1.1 Differential Equationsand Mathematical odels 9)

11x 2'\"+ 5xy + 4y = 0;


\302\267

12. 2y\" x
In

-xy' +
=

YI

2y = 0; YI

In x = 1 Y2 = \037 X x = X cos(lnx), = x sin(lnx)) Y2 y

2'

35.In a city

into the given differential equation to determine all values of the constant r for which y erx is a solution of the equation.)

Problems

13 substitute through 16,


= 0)

= erx

36.

13. y' = 2y 3 15.\" + y' y


In

2y

14.y\" = y 4 16.y\" + 3y' 3


verify

t having a fixed population of P persons,he time who rate of change of the number N of those persons have heard a certain rumor is proportional to the number of those who have not yet heard the rumor. t In a city with a fixed population of P persons,he time rate infected with of change of the number N of those persons i a certain contagious diseases proportional to the product

4y

= 0)
In

of the number do not.)

who have the

diseaseand

the number who

satisfiesthe given differential equation. Then determine a value of the constant C so that y(x) satisfies given initial condition. Usea the computer or graphing calculator(if desired)to sketch several typical solutions of the given differential equation, and highthe light the one that satisfies given initial condition.) x 17.y' + y = 0; y(x) = Ce-, y(O) = 2 2x 18. ' = 2y; y(x) = Ce , y(O) = 3 y
that y (x)

Problems 17through 26,first

Problems 37 through 42, determine by inspection at least one solution of the given differential equation. That is, use your knowledge of derivatives to make an intelligent guess.
Then

test your

hypothesis.

37. y\" = 0

39.xy' + y = 3x2 41.y' + y = eX


solution

38. ' = y y

40. (y')2 + y2 = 1 42. y\" + y = 0


show that a general (one-parameter) differential equation)

19. ' = y + 1; (x) = 1, (O) = 5 y y y 20.y' = x y; y(x) = Ce- +x 1,y(O) = 10

x3 21. ' + y = 0; y(x) = Ce- , y(O) = 7 y = 1; (x) = In(x + C),y(O) = 0 22.eYy' y dy 3 23. dx + 3y = 2x5; y(x) = + Cx-, y(2) = 1 24. xy' 3y = x 3; y(x) = x 3(C + In x),y(l) = 17 25.y' = 3x2(y2 + 1);y(x) = tan(x 3 + C),y(O) = 1 26.y' + y tan x = cosx; (x) = (x + C)cosx, (n) = 0) y y In Problems through 31,function y = g(x)is described 27 a

Cex

- \037x5

43. (a) If k is a constant,


of the

3x2

dx dt is given
constant. by

= kx 2
kt), where

x-

x(t) = Ij(C
inspection

C is an arbitrary
initial value

(b) Determineby

44. (a)
(b)

problem

x' = kx 2, X (0) = O.

a solution of the

31.
In

somegeometric property of its graph. Write a differential the form dy jdx = f (x,y) having the function g equation of as its solution (or as one of its solutions). 27.The slopeof the graph of g at the point (x,y) is the sum of x and y. 28. heline tangent to the graph of g at the point (x,y) interT sectsthe x-axisat the point (xj2,0). 29.Every straight line normal to the graph of g passeshrough t the point (0, 1). an you guess C what the graph of such a function g might look like? 30.The graph of g is normal to every curve of the form 2 y = x + k (k is a constant) where they meet. Theline tangent to the graph of g at (x,y) passeshrough t
by

Continuing Problem 43, assume that k is positive, and kx 2 with sevthen sketch graphs of solutions of eral typical positive values of x(O). How would these solutions differ if the constant k

x' =

45. Supposea population P of rodents satisfiesthe differenthere are P(O)= 2 tial equation dPjdt = k p2. Initially,
rodents,

were negative?

and their number is increasing at the rate of rodPjdt per month when there are P dents. How long will it take for this population to grow to a hundred rodents? To a thousand? What's happening

= 1 rodent

= 10

here?

46. Supposethe
satisfies the

velocity
differential

of a motorboat
equation dvjdt

the point

(-y,x).)

Problems 32 through 36, write-in the manner of Eqs.(3) through (6) of this section-a differential equation that is a mathematical model of the situation described.

32. hetime rate of change of a population T to the square root of P.

P is proportional

33. The time rate of change of the velocity v of a coasting motorboat is proportional 'to the square of v. 34. Theacceleration dvjdt of a Lamborghini is proportional to the difference between 250kmJh and the velocity of the
car.)

at (m1s), and v is decreasing the rate of 1 m/S2 when v 5 m1s. How long doesit take for the velocity of the boat to to decrease 1 m1s? To /o mls? When doesthe boat come to a stop? defines a 47. In Example 7 we saw that y(x) of solutions of the differential equafamily one-parameter tion dyjdx y2. (a) Determine a value of C so that (b)Is there a value of C such that y(O) O? y(10) Can you nevertheless find by inspection a solution of O? (c) Figure shows y2 such that y(O) dyjdx C typical graphs of solutions of the form y(x) Does appearthat thesesolution curves fill the entire xyit

of speed the

motorboat is v(O)

= 10 meters per second = = Ij(C x)

= kv 2.Theinitial

coasting in water

= = 10. =

= Ij(

1.1.8
=

- x).
b?)))

plane? Can you concludethat,


the

given any point (a, b) in the differential equation dyjdx y2 has explane, actly one solution y(x) satisfying the condition y(a)

10

Chapter
1
3) 2)

First-OrderDifferential Equations)
100
80 60

C=-2C=-1c=oC=1C=2 C=3

1)

C=4)
\037

40

\037

C=-4
-1) -2)

-20 -40 -60 -80 -100 -2-1 0 -5-4-3


x)

20 0

2 3 4 5
y

FIGURE

equation dy/dx

1.1.8. of solutions =Graphs


y2.)

of the

FIGURE

1.1.9. graph The


of C.)

= Cx4 for

various values

48. (a) Show


ily

that

of differentiable

y(x)

= Cx4 defines a one-parameter famsolutions of the differential (b)Show that equation

xy'

= 4y (Fig. 1.1.9).
y(x)

-x4
{

if x x 4 if x

< 0, > 0)

definesa differentiable solution of x y' = 4y for all x,but 4 is not of the form y(x) = Cx (c) Given any two real numbers a and b, explain why-in contrast to the situaexist infinitely many tion in part (c) of Problem 47-there solutions of x y' = 4y that all satisfy the differentiable

condition

y(a) = b.)

1IIJ._.!!l
!\037g\037\037!S_

\037S._\037\037\037\037\302\243\037\037...\037J?:\037}>..\037\037!!\037\037\037.\037\037__\037\037\037_!\037\037_\037_.______________________)

The first-order equation dy/dx


f right-hand-sideunction
\037)

= I(x,y) takes an especiallysimpleform if the v doesnot actually involve the dependent ariable y, so)

dy

dx

= I(x).

(1))

In
\037)

this

weneedonly integrate both sidesof Eq. (1)to obtain) specialcase y(x) =

f f(x)dx+ c.)

(2))

This is a generalolutionof Eq. (1), s meaning that it involves an arbitrary constant I o C,and for every choicef C it is a solution of the differential equation in (1).f G(x) is a particular anti derivativeof -thatis, if G'(x) = (x)-then)

y(x) = G(x)+ C.) G(x)+C2 on the sameinterval


The graphs of
any

(3))

two such solutions

and 1.2.2. There we seethat the constant C is geometrically the vertical distance betweenthe two curves y(x) = G(x)and y(x) = G(x)+ C. To satisfy an initial condition y(xo) = Yo, we needonly substitute x = Xo and = Yo into Eq. (3)to obtain Yo = G(xo)+ C,so that C = Yo G(xo).With this y choicef C,we obtain the particularolutionofEq.(1)satisfying the initial value o s

I are \"parallel\"

Yl

(x)
the

in

G(x) + Cl illustrated sense

and Y2(X) by Figs.

1.2.1

problem) dy
\037)

dx

= I(x), y(xo)=

Yo.)))

1.2 a Solutions Integrals s Generaland Particular


4 3 2
1) \037

11)

6)

C=-1
C=-2)
\037

4)

2)

0)

0)

-1 -2 -3 -4 -4-3-2-1 0
x)
\037

-2)

-4) 1 2 3 4

-6 -6

C=-4
-4 -2
0
x)

FIGURE 1.2.1. of Graphs 2 y = x + C for various values of C.)

FIGURE1.2.2. Graphs of y = sin x + C for various values of C.)

We will see that this is the typical pattern for solutionsof first-orderdifferential equations. Ordinarily, we will first find a generalsolution involving an arbitrary

constant C.We can then attempt to obtain, by appropriate choice C,a particular of solution satisfying a given initial condition y (xo) = Yo.)

Remark:As the term is usedin the previousparagraph, a generalolution s of a first-order differential equation is simply a one-parameter family of solutions. A natural question is whether a given general solution contains every particular solution of the differential equation. When this is known to be true, we call it the general solution of the differential equation. For example,becauseany two antiderivativesof the samefunction (x) can differ only by a constant, it follows that every solution of Eq. (1)is of the form in (2).Thus Eq. (2)serves to definethe general solution of (1).

.)

Example

1)

Solvethe

initial

value problem)
dy

dx)

= 2x+3, y(l) = 2.

Solution Integration of both sidesof the differentialequation as in Eq. (2)immediately yields


4 2 0 the general solution)

y(x} =

-2
\037

f (2x+ 3}dx= x + 3x+ C.)

-4

-6 -8 -10 -6

showsthe graph y = x2 + 3x + C for various values of C. The Figure to particular solution we seek corresponds the curve that passesthrough the point the initial condition) (1, ), thereby satisfying 2 -4

1.2.3

-2
x)

4)

y(l) = (1)2+ 3 . (1)+ C = 2.)


It follows that

FIGURE

1.2.3. Solution

curves

for the differential

equation in

C=

-2,so the desiredparticular solution is)


2 y(x)=x +3x-2.)
.)))

Example 1.)

12

Chapter
1

First-OrderDifferential Equations)

Second-order equations.The observation that the special first-order equation dyjdx = f(x) is readily solvable (providedthat an anti derivativeof f can be found)
extendsto second-order differential equations of the specialform) d2y

dx2
in

= g(x),)

(4))

which the function g on the right-hand side involvesneither the dependent ariv able y nor its derivativedyjdx. We simply integrate onceto obtain)

=
\037\037

f y\"(x)dx= f g(x)dx= G(x)+


arbitrary

C\\,)

where G is an antiderivative of g and C1 is an


integration yields)

constant. Then another

y(x) =

f y'(x)dx= f [G(x)+ Cddx = f G(x)dx+


= dx g(x)
dv
and)

C\\x

+ C2,)

where C2 is a secondarbitrary constant. In effect, the second-order ifferential d in (4) is one that can be solved by solving successively first-order the equation
equations)

dy
dx)

= vex).

a Velocity ndAcceleration
Directintegration is sufficient to allow us to solve a number of important problems concerning the motion of a particle (or mass point) in terms of the forcesacting on it. The motion of a particle along a straight line (the x-axis) described its is by positionfunction)

x = f(t))
giving its

(5))

x-coordinate at time t. The velocity of the particle is defined to be)


vet)

\037)

= f' (t);)

that is,)

v=

-.
dx
dt)

(6))

Its acceleration is aCt) = v'(t) = x\"(t);n Leibniznotation,) i aCt)


\037)

- dV dt

d2x dt 2

.)

(7))

vet)

Equation (6)is sometimes appliedeither in the indefinite integral form x(t) = dt or in the definite integral form)

x(t) = x(to) + because xjdt= (precisely d

i
to)

v(s) ds,
theorem of calculus

which you should recognizes a statement of the fundamental a


v).)))

1.2 Solutions a Integrals s Generaland Particular


and

13)

Newton'ssecond of motion says that if a force F(t) acts on the particle law is directedalong its line of motion, then)

ma(t) = F(t);)

that is,)

F=ma
,)

(8))

where m is the mass of the particle. If the force F is known, then the equation x\"(t) = F(t)lm can be integrated twice to find the position function x(t) in terms of two constants of integration. These arbitrary constants are frequently detertwo mined by the initial position Xa = x(O)and the initial velocity Va = v(O) of the

particle. Constantacceleration. instance, supposethat the force F, and therefore the For accelerationa = F1m, are constant. Then we beginwith the equation)

dv
dt)

=a

(a is a constant)

(9))

and integrate both

sidesto obtain)
v(t)

f adt=at+C!.)
0,
of this information
into

when t = and substitution the fact that C1 = Va. So) preceding equation yields
We know that v
Va

the

vet)
A

= dx = at + Va. dt)

(10))

secondintegration

gives)

x(t) =
and the
substitution t

f v(t) dt = f (at +
= 0,x = Xa
X(t)

vo) dt

= 4at2 + vot + C2,)

gives C2 = Xa. Therefore,)

= 4at2 + vat + Xa.)

(11))

the position,of Thus, with Eq. (10)we can find the velocity,and with Eq. the particle at any time t in terms of its constant acceleration a, its initial velocity Va, and its initial position Xa.) . . . .

(11)

Example2)

lander is falling freely toward the surface of the moon at a speedof 450 meters per second(m/s). Its retrorockets, hen fired, provide a constant decelw 2 eration of 2.5 meters per secondper second(m/s ) (the gravitational acceleration At to produced the moon is assumed be included in the given deceleration). what by above the lunar surface should the retrorockets e activatedto ensurea \"soft b height touchdown\" (v = 0 at impact)?)))
A lunar

14

Chapter
1

First-OrderDifferential Equations) denote by x(t) the height of the lunar lander above the surface, as indicated We Fig.1.2.4. let t = 0 denotethe time at which the retrorocketsshould be fired. Then Vo = -450 and the (mjs,negative because height x(t) is decreasing),
We
in

Solution

the

= +2.5, because upward thrust increases velocity v (although an the Then Eqs. (10)and (11) become) Ivl). speed
vet)

it

decreases
(12))

= 2.5t

- 450)
xo,)

and)
Lunar surface)

x(t) =
lander

2 1.25t

- 450t +

(13))

FIGURE1.2.4. lunar The of Example 2.)

where Xo is the height of the lander above the lunar surface at the time t = 0 when the retrorockets should be activated. = From Eq. (12)we see v = 0 (softtouchdown)occurs hen t = 450j2.5 that w 180 (that is, 3 minutes); then substitution of t = 180, = 0 into Eq. (13)yields) s x
Xo

= 0 (1.25)(180)2 = 40,500) + 450(180)


\037 25\037

meters-thats, Xo = 40.5 km i

should be actimiles. hus the retrorockets T vated when the lunar lander is 40.5 kilometers above the surfaceof the moon, and it will touch down softly on the lunar surface after 3 minutes of decelerating descent.

.)

Units) Physical
Numerical work requiresunits for the measurement of physicalquantities such as distanceand time. We sometimes ad hoc units-such distancein milesor as use kilometers time in hours-in and situations (suchas in a probleminvolving special an auto trip). However, the foot-pound-secondfps) and meter-kilogram-second ( (mks)unit systemsare usedmore generally in scientific and engineering problems. In fact, fps units are commonly used only in the United States (and a few other w countries), hile mks units constitute the standard international system of scientific
uni ts.)
.\037:,:;::\037':';:':;) \037,:..:\037) \037:':\\/:T/e/(>:,::; \"\037:_::; .--. -.-;:\037,<

>'.;:;i U
'>;\037'.?
\037

'])/>1(1

?\037:r:\037\037\037: ::\\.>.:. -'::\037:\037{>}h:/:.:::;\037) }>;r<r\037:\037fHSJ;::}{<;-;un:i;\\ :'\" ;',-';'-';\":\"':,; -.\037' '-;. -.,)

'lY\"iiil\037
:;

\037>r

'<>::::;',\302\253\\:<\037-';:

.. '',' . .... a\",.. . ... s ...:,.< .'.;.,...'..'.''.......:'......'m..'...1;,;...'''''.....u...n.'........8..., .... . ... . . . . . . .' . . . '
... .. .. .. .. ;. .; .. .. \" ': ';';;\"i\037i,T;1';! ..

,1

....

...,

.. .. ..

......

..

... ... ..

.......

..

..

...

...

..

..

...

...

..

,.

.. ..

:(\037,>t/y>c

..

..

\"\"..

,.

..

..

:\037/i:t'\037p;:,:::?r-

::.:_::{;;)l

';)

.
40
:L:
... .. .... ','

..

....

Force Mass Distance


Time g

pound slug

(lb)

newton (N)

foot (ft)

kilogram (kg) meter (m) mls 9.8

second(s) 32 ft/s 2

second(8)
2)

The last line of this table gives values for the gravitational acceleration g at the surface of the earth. Although these approximate values will suffice for most 2 v ftj s (at examplesand problems,more precisealues are 9.7805 js2 and m sealevel at the equator). Both systemsare compatible with Newton'ssecondlaw F = ma.Thus 1 N is (by definition) the force requiredto impart an accelerationof 1 mjs2 to a massof 1 an kg. Similarly, 1 slug is (by definition) the massthat experiencesaccelerationof 1 ftj s2 under a force of will use mks units in all problems (We requiring mass units and thus will rarely need slugsto measure mass.))))

32.088

lIb.

1.2 a Solutions Integrals s Generaland Particular

15)

and Inches centimeters (as well as milesand kilometers) lso are commonly a used in describing distances.For conversionsbetween fps and mks units it helps to remember that) 1 in. =

2.54m (exactly) c

and

lIb

\037

4.448N.)

For instance,)
1 ft and
it

= 12 x 2.54 = 30.48 in. cm,


\037m

In.)

follows that)
1 mi

cm = 5280 x 30.48 = 160934.4 1.609 ft cm km.


\037

ft)

Thus a postedU.S. international speedlimit of 50mi/h means that-in legalspeedlimit is about 50x 1.609 80.45 lan/h.)
\037

terms-the

Vertical Motionwith Gravitational Acceleration)


The weight W of a body is the force exertedon the body by of a = g and F = W in Newton'ssecondlaw F = ma gives)
gravity. Substitution

W=mg)
\037

(14))
\037

for the weight W of the massm at the surfaceof the earth (where g 32 ft/s 2 9.8 m/s2). For instance, a massof m = 20 has a weight of W = (20kg)(9.8 mjs2)= kg N 196 . Similarly, a massm weighing 100oundshas mks weight) p
W

= (100 Ib)(4.448 N/lb) = 444.8N,)


m

so its massis)

= 444.8N2

9.8 m/s

\037

45.4 kg.

To discussvertical motion it is natural to choose y-axisas the coordinate the for position,frequently with y = 0 corresponding \"ground level.\" If we to system choose upward direction as the positive direction, then the effect of gravity on a the its its vertically moving bodyis to decrease height and alsoto decrease velocityv = if we ignore air resistance,hen the accelerationa = dv/dt of t dy/dt. Consequently, the body is given by)
\037)

---g
dv

dt

.)

(15))

This acceleration equation provides a starting point in many problemsinvolving vertical motion. Successive integrations (as in Eqs.(10)and (11)) the velocity yield
and height formulas)
and)

vet)

= -gt+ va) = _\037gt2 + vat + Yo.)


velocity.)))

(16)) (17))

yet)

Here, o denotesthe initial Y

(t

= 0) height of the body and va its initial

16

Chapter
1

First-OrderDifferential Equations)

Example3)

(a) Supposethat a ball is thrown


initial

its maximum height when its velocity (Eq. (16)) zero,) is vet) and
thus

2 velocity Vo = 96(ftjs, so we use g = 32 s in ftj

straight

upward from the ground (Yo = fps units). Then it

reaches

0) with

= -32t+ 96= 0,)


height
that

when

= 3 s.Hence maximum the


y(3) =

the ball attains

is)

3 + -4.32.2 + 96.3 0 = 144(ft)


initial

the aid of Eq. (17)). (b) If an arrow is shot straight upward from the ground with (mjs,so we use g = 9.8 in mks units), then it returns mjs2
(with

velocity Vo
0,)

= 49

to the ground

when)

yet)
and A
y-axis)

+ -4.(9.8)t +49t= (4.9)t(-t 10)=


2
air.)

thus

after 10 in the s

.)
2a.

Swimmer's roblem P

Figure

The linesx = :f:a river of width w = that the velocity the banksof the river and the y-axisits center. uppose represent S v R at which the water flows increasess oneapproacheshe center of the river, and a t indeedis given in terms of distancex from the center by)
VR

1.2.5 a northward-flowing shows

(-a,

0))

(a,0)

x-axis)

= Vo

(1 =:)

.)

(18))

VR)

Vs)

FIGURE

1.2.5.swimmer's A

problem (Example 4).)

can use Eq. (18)to verify that the water does flow the fastest at the center, where VR = vo, and that VR = 0 at eachriverbank. that 0) on the westbank and swims Suppose a swimmer starts at the point his dueeast (relative to the water) with constant speedVs. As indicated in Fig. velocity vector (relative to the riverbed) has horizontal component Vs and vertical the direction angle a is given by) component VR. Hence swimmer's
You

(-a,

1.2.5,

tan tan Because a = dyjdx, substitution

a=

-.
VR
Vs)

using

(18)gives the differential equation)

for the swimmer'strajectory y


\037.....,...........,..\"\"\"\"'\" N \"\"'\" \037\037\037\"\"\"\"\"'__,_\"\"\"\"\"'\037_\"'....'_\037\037........ \"\"'''''''V\037\037\037\037'''''''''''''''\037\037N ....,\037\"\"'_\"'\"

X2 dy = Vo 1 dx Vs ( a ) = y(x) as he crosses river.) the


2)
''''\037.....,..''''''''''''.... N'\037\"\",,,,,,,,,,,, \"\",\037\037r-v \037\037\037\"\",,,,,,.... .....,.._......\037\037\"\"'.....,......

(19))

4) Example

Supposethat the river is 1 mile wideand that its midstream velocity is Vo = 9 mijh. If the swimmer'svelocity is Vs = 3 mijh,then Eq. (19)takesthe form
v.....,..\037\037'_....'N

_....._,_,.\",\037

....,_...\037\037\037...

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

dx)

dy = _ 3(1 4x2).

Integration yields)

y(x) =

f (3 - 12x)dx = 3x- 4x +
2 3

C)))

1.2 a Solutions Integrals s Generaland Particular


for the swimmer's trajectory. The initial condition y

17)

y(x) = 3x
Then)
y

- 4x +
3

(-4)= 0 yieldsC = 1,so)

1.)

(4) = 3 (4) 4 (4)3 + 1 == 2,

so the swimmer drifts 2 milesdownstream while he swims1 mile acrossthe river.

.)

P 11IIIroblems)
In through ing the given differential condition.

Problems

10, find

a function
and
the

equation

prescribed

= f (x)satisfyinitial

19. 10
8 6
\037

2. dx = (x - 2)2;y(2) = 1
dy 3. dx = -IX;y(4) = 0

dy 1.-=2x+l;y(0)=3 dx

4.

dy 5. dx =

dy 6. dx = X

7.

9. dx =
dy In

- -;y(l) - dy

dy

(5,5))

4 2 .... ..
..

dx

1 x2

=5

0 0

4)
t)

6)

8)

10)

-Vx

+2

; y(2) =

-1
20.)
dy 8. dx = cos2x; = 1 y(O)

FIGURE 1.2.6. Graph of the velocity function v(t) of Problem


10)
8)

19.

-VX2

+ 9; y(-4) = 0

dy

dx

x 2 + 1 y(O) 1 ; y(O) = 0 I x -V
2)

10 ;

=0

10. = xe-x ; y(O) = 1


dy dx)

6)
\037)

(5,5).)
4)

Problems through 18, the positionfunction x (t) of a find a moving particle with the given acceleration (t), initial position Xo = x (0), and initial velocity Vo = v (0). a(t) = 50, Vo = 10, = 20 Xo 12. (t) = -20,o = Xo = 5 a V 13. (t) = 3t, Vo = 5, Xo = 0 a = 14. (t) = 2t + 1,Vo = -7,xo 4 a = 4(t + 3)2,Vo = Xo = 1 15. (t) a

11

2)

2)

4
t)

8)

10)

11.

-15,

21.

FIGURE 1.2.7. Graph of the velocity function v(t) of Problem 20.


10)

16. (t) = a
17.a(t)
==

1 -vt

-1,

8)

+4
1

, Vo

= -1,xo 1
==
\037)

6)
4)

(5, 5):)

= 0, Xo = 0 18. (t) = 50sin 5t, Vo = -10, a Xo


3 (t + 1)
Vo
==

'

8)

2)

In Problems through 22, a particle starts at the origin and travels along the x-axis with the velocity function v(t) whose Sketch the graph graph is shown in Figs. through ofthe resulting positionfunction for 0 < t < 10.)

19

2)

4)
t)

6)

8)

10)

1.2.6

x(t)

1.2.9.

FIGURE 1.2.8. Graph of the velocity function v(t) of Problem

21.)))

18
22.
10 8 6
;:::..

Chapter
1

First-OrderDifferential Equations)

32.Supposethat
. (3,5)

(7,5))

4)

2)

4)
t)

6)

8)

10)

dropped from a height of 20 ft hits the ground in 2 If a ball is dropped from the top of a 200-ft-tall uilding on Gzyx, how long will it take to hit b will it hit? the ground? With what speed 34. A personcan throw a ball straight upward from the surface of the earth to a maximum height of 144 ft. How high could this personthrow the ball on the planet Gzyx

100 33.On the planet Gzyx, a ball

a car skids 15 m if it is moving at 50 km/h when the brakes are applied. Assuming that the car has how far will it skid if it is the sameconstant deceleration, at km/h when the brakes are applied? moving

s.

FIGURE1.2.9. Graph of the velocity function vet) of Problem 22.

35.A
of part

of Problem 29? stone is droppedfrom rest at an initial height with the surface of the earth. Show that the speed
strikes the ground is v

23. hat W

is the maximum

the top of a building 400 ft high. does it take to reach the ground? With what speeddoesthe ball strike the ground? 25.Thebrakes of a car are applied when it is moving at 100 o meters per km/h and provide a constant deceleration f 10 2 secondper second(m/s ). How far doesthe car travel before coming to a stop? 26.A projectileis fired straight upward with an initial velocity of 100 l s from the top of a building 20 m high and m falls to the ground at the baseof the building. Find (a) its maximum height above ground; (b) when it passes the the top of the building; (c) its total time in the air. 27. A ball is thrown straight downward from the top of a tall of building. The initial speed the ball is 10 s.It strikes ml the ground with a speed 60 ml s.How tall is the buildof How long

(b) of Example 3? 24. A ball is dropped from

height attained by the arrow

36.Supposea woman

= ,J2g . h

h above which it

the on the moon-where surface gravitational acceleration is (approximately) 5.3 ft/s 2-howhigh abovethe surface
will

(on earth) from the ground to a height of 2.25 feet. If she jumps straight upward with the same initial velocity

has enough

\"spring\"

in her

legsto jump

sherise?

37.

38.At noon
constant at

at At noon a car starts from rest at point A and proceeds a straight road toward point constant acceleration along P.M. ith a velocity of w If the car reachesB at 60 mi/h, what is the distance from A to B?

B.

12:50

w from rest at point A and proceeds ith accelerationlong a straight road toward point C, a car 35 miles away. If the constantly accelerated arrives at C with a velocity of 60 mi/h, at what time doesit arrive

a car starts

39.If a = 0.5mi and

C?

Vo

28.

29.

straight downward with an initial top of the Washington Monulong doesit take to reach the doesthe baseballstrike the ground, and with what speed ground? A dieselcar gradually speeds p so that for the first s u its acceleration given by) is

ing? A baseballis thrown

speedof 40 ft/s from the ment (555 ft high). How

40. Supposethat a = 0.5 mi,


as in Example 4, but
by the fourth-degree

must the swimmer's speed be in order that Vs the only 1 mile downstream as he crosses river?
Vo

= 9 mi/h

as in Example 4, what
he
drifts and v s river

= 9 mi/h,

that the velocity


function)

of the

= 3 mi/h
is given

10

VR

= Vo

dv
dt)

(1

\037:

))

2 = (0.12)t+ (0.6)t (ft/s 2).

If the car starts from rest (xo = 0, Vo = 0), find the distance it has traveled at the end of the first s and its

10

41.

30.A car traveling


brakes

velocity at that time. at 60 mi/h

31.

ft (88 ft/s) skids 176 after its are suddenly applied. Under the assumption that the braking system provides constant deceleration, hat w is that deceleration?or how long doesthe skid continue? F Theskid marks made by an automobile indicated that its brakes were fully applied for a distanceof 75 m beforeit cameto a stop.Thecar in question is known to have a cono stant decelerationf20 m/s2 under theseconditions. How fast-in km/h-wasthe car traveling when the brakes

Now find rather than the quadratic function in Eq. the how far downstream the swimmer drifts as he crosses flver. A bomb is dropped from a helicopter overing at an altih tude of 800feet above the ground. From the ground diis rectly beneath the helicopter, a projectile fired straight after the bomb upward toward the bomb, exactly 2 seconds W is released. ith what initial velocity should the projectile be fired, in order to hit the bomb at an altitude of exactly

(18).

400feet?
i 42. A spacecrafts in freefall of at a speed 1000mph
2 At o fired, provide a constant deceleration f 20,000mi/h astronauts what height abovethe lunar surface should the fire the retrorockets to insure a soft touchdown? (As in 2, ignore the moon'sgravitational field.)))) Example toward the surface of the moon (mi/h). Its retrorockets, when

were first applied?)

1.3 FieldsandSolutionCurves Slope


43. Arthur Clarke's Wind from The
the Sun

19)

fires a prosimultaneously jectile(straight ahead in the samedirection) that travels at 8 one-tenth of the speed = 3 x 10m1s of light. How long c will it take the spacecrafto catch up with the projectile,) t

= m1s 0.001g 0.0098 2.


from rest at time t

Diana, a spacecraft ropelled the solar wind. Its alup by minized sail provides it with a constant acceleration of

describes (1963)

44. A driver

= 0 and

t Supposehis spacecraftstarts

and how far will it have traveled by then? c involved in an accident laimshe was going only 25 mph. When policetested his car, they found that when its brakes were applied at 25 mph, the car skidded only 45 feet before coming to a stop. But the driver's skid feet. Assummarks at the accidentscenemeasured determine the speed ing the same(constant) deceleration, he was actually traveling just prior to the accident.)

210

and Solution Curves) Sl\037peFields


Considera differential equation of the form)
dy
\037)

dx

= I(x,y))

(1))

y)

x)

where the right-hand function I(x,y) involvesboth the independentvariablex and dependentvariable y. We might think of integrating both sidesin (1)with redx + C.However,this approach spectto x, and hencewrite y(x) = J I(x, doesnot leadto a solutionof the differentialequation, because indicatedintegral the involves the unknown function y(x) itself, and thereforecannot be evaluatedexplicitly. Actually, there existsno straightforward by procedure which a generaldifferential equation can be solved explicitly. ndeed,the solutionsof such a simple-looking I in differential equation as y' = x2 + y2 cannot be expressed terms of the ordinary functions studiedin calculustextbooks. evertheless,the graphicaland N elementary numerical methods of this and later sectionscan be used to construct approximate solutions of differential equations that sufficefor many practicalpurposes.)
the
y(x\302\273

Fields Graphical and Solutions Slope


FIGURE
for the differential equation y' = x y together with tangent lines having

. . .

A 1.3.1.solution

curve

slopeml

= Xl

point (Xl, YI);

slopem2 = X2

point (X2, Y2); and

slopem3 = X3

YI

at the at the

Y2

There is a simplegeometric way to think about solutions of a given differential y equation y' = I(x,y). At each point (x,y) of the xy-plane, the value of I(x, ) determines slopem = (x,y). A solution of the differentialequation is simply a a differentiable function whosegraph y = y(x) has this \"correctslope\"at each point Thus a solution is, (x, through which it passes-that y'(x) = I(x, curve of the differential equation y' = (x,y)-thegraph of a solution of the equation-is a curve in the xy-plane whosetangent line at each point (x,y) simply has slope m = (x,y). For instance, Fig. showsa solution curve of the differential equation y' = x y together with its tangent lines at three typical

y(x\302\273

y(x\302\273.

1.3.1

Y3 at the

point (X3, Y3).)

This geometric viewpoint suggestsa graphical ethod for constructing apm solutions of the differential equation y' = I(x,y). Through each of a proximate o representative collectionf points (x,y) in the plane we draw a short line segment the properslopem = (x,y). All these line segments constitute a slope having field (ora direction for the equation y' = (x,y).) field)

points.

I
........

\302\245.....

..

Example

1)

(a)-( Figures1.3.2 d) show slopefields and solutioncurves for the differential equa........... ..... ....... ......... .............. ......... .........-....u.............. ...... ..... .............. \302\245-..........

..

.-.

......

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

.........
\037..... ...........

tion)

dx)
with

dy = ky

(2))

= 2, 0.5, and of the parameter k in Eq. (2).Note that each field yields important qualitative information about the set of all solutions))) slope
the values k

-1, -3

20

Chapter
1

First-OrderDifferential Equations)
4 3 2
1
\037

4 3 2
1
\037O

-1
-2) o
x)

-1
-2
1

234
= 2y.)

-3 -4 -2 -1 0 -4-3
\\ \\ \\ \\ \\ \\ \\ \\

2 3 4

x)

FIGURE 1.3.2(a) field Slope


and solution

curves for y'

field FIGURE 1.3.2(b) Slope and solution

y' 4 3 2
1
\037O \037O

= (O.5)y.)

curves for

4 3 2
1

-1 -2
o
x)

-1
-2)
1

234
=

o
x)

234
= -3 y.)

FIGURE 1.3.2(c) field Slope


and solution

curves for y'

field FIGURE 1.3.2( Slope d) and solution

y.)

curves for y'

of the differential equation. For instance, Figs.1.3.2(a) (b) suggestthat each and solution y(x) approaches and :l::oox +00if k > 0,whereasFigs.1.3.2(c) as that y(x) 0 as x +00if k < O. Moreover, although the sign (d) suggest of k determines direction of increase r decrease y(x), its absolute value Ikl the o of to determine the rate of change of y(x). All this is apparent from slope appears fields like those in Fig. even without knowing that the general solution of kx is given explicitly by y(x) = Ce Eq. (2) A slope field suggestsvisually the general shapesof solution curves of the differential equation. Through eachpoint a solution curve should proceed such in a direction that its tangent line is nearly parallel to the nearby line segments the of slopefield. Starting at any initial point (a,b), we can attempt to sketchfreehand an approximate solution curve that threads its way through the slopefield, following the visible line segments closelyas possible.) as
\037 \037 \037

1.3.2,

Example2)

Construct a slopefield for the differential equation y/ = X Y and use it to sketch an approximate solution curve that passesthrough the point ( 4).)

-4,

Solution Solution Fig.1.3.3 a table of slopesfor the shows


slopem

given equation. The numerical at the intersection of the horizontal x-row and the very appears tical y-column of the table. If you inspectthe pattern of upper-left to lower-right diagonalsin this table, you can seethat it was easily and quickly constructed.(Of)))

=x

1.3 Fieldsand SolutionCurves Slope


x\\y

21)

-4 -3 -2
0
1

-4 -3 -2 -1 0 1 0 -1 -2 -3 -4 -5 1 0 -1 -2 -3 -4 1 2 0 -1 -2 -3 -1 3 2 1 0 -1 -2
4
5 6
7

4
5 6
7

2 3

0
1

-6 -5 -4 -3 -1 -2
0
1

2 3

4
5 6

2 3

-6 -5 -4 -3 -1 -2
0
1

4
5

2 3

-6 -5 -4 -3 -1 -2
0
1
0)

-7 -8 -7

FIGURE 1.3.3. of the slopey' = x Values


5)
\\

2 3
y for

-1

-4 < x, y < 4.)

- - // / /
\\
'\" '\"
/)
;::....

\\

-'\"

5 4
3

2
1
;::....

-\"'-)
\\

\\

'\"

\\

'\"

0)
\\

-5

-/ /
'\"

\\

-- / /
'\"

/
/)

-5)

o
x)

-2 -3 -4 -5

-1
-5)

o
x)

5)

FIGURE 1.3.4. field for Slope


y' = x y correspondingto the table of slopesn Fig. 1.3.3.) i

FIGURE

curve through

The 1.3.5. solution

(-4,4).)

y) on the right-hand sideof the differential equation would necessitate more complicated shows calculations.) igure F the corresponding showsan approximate solution curve field, and Fig. slope sketchedthrough the point 4) so as to follow as this slopefield as closelyas At each in it appears proceed the directionindicatedby the nearby to possible. point line segments the slopefield. of

a function course, more complicated

f(x,

(-4,

1.3.5

1.3.4

.)

4
3

2
1
\037

-1
-2 -3
-4)

-4 -3 -2 FIGURE
typical

-1 0
x)

4)

y'

=x

solution
y.)

1.3.6. field and Slope


curves for

a spreadsheet rogram (for instance) readily constructs a table of p it can be quite tedious to plot by hand a sufficient number Fig. of slopesegmentsas in Fig. However, most computer algebra systems includecommandsfor quickand ready constructionof slopefields with as many line segmentsas desired;such commands are illustrated in the application material for this section. he more line segments T are constructed, the more accurately solution curves can be visualized and sketched.Figure showsa \"finer\" slopefield for the differential equation y/ = X of Example2,together with typical solution Y curves treading through this slopefield. If you look closelyat Fig. you may spot a solution curve that appears to be a straight line!Indeed,you can verify that the linear function y = x 1 is a solution of the equation y/ = X y, and it appearslikely that the other solution curves approach this straight line as an asymptote as x +00.This inference illustrates the fact that a slopefieldcan suggesttangible information about solutions that is not at all evident from the differential equation itself. Can you, by tracing
Although slopesas in

1.3.3,

1.3.4.

1.3.6

1.3.6,

\037

the)))

22

Chapter
1

First-OrderDifferential Equations)
the

appropriate solution curve in this figure, infer that y(3) initial value problemy' = x y, y(-4)= 4?)

\037

2 for the solution y(x) of

of fields Applications Slope


The next two examples illustrate the use of slopefields to glean useful information in physicalsituations that are modeled differential equations. xample isbased 3 E by on the fact that a baseballmoving through the air at a moderate speedv (lessthan about 300 encounters air resistancethat is approximatelyproportional to v. If ftls) the baseballis thrown straight downward from the top of a tall building or from a t both hovering helicopter,hen it experiences the downwardaccelerationof gravity and an upward acceleration of air resistance.f the y-axis is directeddownward, I 2 then the ball's velocity v = dy/dt and its gravitational acceleration g = 32 ftls are both positive, while its accelerationdue to air resistance negative. Hence total is its acceleration of the form) is

dv

dt

=g

kv

.)

(3))

typical value of the air resistance proportionality constant might be k

= 0.16.)

Example3)
400
300
\037

.....\"\"'=\037

200)

\"\"\"\"\"\"\" ','
\037 \037 \037 \037 \037 \037 \037 \037 \037 \037 \037

,,,,,,,,,,,,, ..................................................)
\037\"\"\"\"\"\"\"\"\"\"\"\"\"\"\"\"\"\"
\037 \037 \037 \037 \037 \037 \037 \037 \037 \037 \037

a baseballstraight downward from a helicopter hovering at an wonder whether someonetanding on the ground below s couldconceivably catch it. In orderto estimate the speedwith which the ball will land, you can use your laptop'scomputer algebra system to construct a slopefield for the differential equation)

Supposeyou
altitude of

3000 You feet.

throw

100 / / / / / / / / / / / / / / / / / / / / / / / / / / / / / 5 10 15 20
/) t)

dv
dt)

= 32 0.16v.

(4))

25)

FIGURE1.3.7. field Slope


typical

and

v'

= 32

solution curves

- O.16v.)

for

The result is shown in Fig. 1.3.7, together with a number of solution curves to corresponding different values of the initial velocity v(O) with which you might throw the baseball downward. Note that all thesesolutioncurves appear to approach the horizontal line v = 200 an asymptote. This impliesthat-however as you throw it-the ftls baseballshould approach the limiting velocity v = 200 instead of accelerating indefinitely (as it would in the absenceof any air resistance). The
handy

fact that

60milh = 88 ftls
v

yields)
\037

ft milh = 200-x 60

88 ftls

mi 136.36
h)

-.

to Perhapsa catcher accustomed this speeding ball. fielding v' (0) = 32

milh 100

fastballs would have some chance of

then Comment If the ball'sinitial velocity is v(O) = 200, Eq. (4) gives = 0,so the ball experiences initial acceleration. Its no (0.16) (200) = 200 a constant \"equilibis therefore remains unchanged, and hence v(t) velocity rium solution\" of the differential equation. If the initial velocity is greater than 200,

then the initial acceleration given by Eq. (4) is negative, so the ball slowsdown as it falls. But if the initial velocity is less then the initial acceleration given by than is positive, so the ball speedsup as it falls. It therefore seemsquite reasonable (4) of the that, because air resistance, baseballwill approach a limiting velocity of the ftls-whateverinitial velocity it starts with. You might like to verify .))) o absence f air resistance-this would hit the ground at over mi/h. ball

200,

300

200 that-in

1.3 Fieldsand SolutionCurves Slope


In

23)

Section we will discussin detail the logisticdifferential 1.7

dP
dt)

equation)

= kP(M P)

(5))

often is used to model a population P (t) that inhabits an environment with carrying capacity M. This means that M is the maximum population that this environment can sustain on a long-term basis(in terms of the maximum availablefood, for instance).)
that

Example4)

If we take k

= 0.0004 M = 150, then and

dP
dt)

the

logisticequation in (5)takes the

form)

2. = 0.0004P(150 = 0.06P 0.0004P P)

(6))

300 250 200 Q.. 150 100 50 o


o) 25)

50
t)

75)

100)

FIGURE

typical solution

1.3.8. field and Slope

The positive term 0.06P the right in (6)correspondso natural growth at a 6% on t 2 annual rate (with time t measured years). The negative term -0.0004p in representsthe inhibition of growth due to limited resources the environment. in showsa slopefieldfor Eq. (6),together with a number of solution Figure curves corresponding possibledifferent values of the initial population P(0). to Note that all these solution curves appear to approach the horizontal line P == 150 as an asymptote. This impliesthat-whatever the initial population-theopulation p t a pet) approacheshe limiting population P = 150s t 00.

1.3.8

\037

P'= O.06P

- O.0004p
2

curves for
.)

Comment If the initial

then 150, 150) P'(O)= 0.0004(150)(150 = 0,)

population is P(O) =

Eq. (6)gives)

so the population experiencesinitial (instantaneous)change.t thereforeremains I no and henceP(t) = 150 a constant \"equilibrium solution\" of the difis unchanged, ferential equation. If the initial population is greater than 150, the initial rate of then is negative, so the population immediately beginsto decrease. change given by (6) But if the initial population is less then the initial rate of change given by than 150, is positive, so the population immediately beginsto increase.therefore seems It (6) reasonable concludethat the population will approach a limiting value of to quite

ISO-whatever (positive) initial population. the

.)

Existencend Uniqueness Solutions a of


Beforeone spendsmuch time attempting to solve a given differential equation, it is wise to know that solutions actually exist. We may also want to know whether there is only one solution of the equation satisfying a given initial condition-that is, whether its solutions are unique.)

Example5)

T (a) [Failure of existence] he initial value problem)


1 , y=x)

y(O) = 0)
I

(7))

has no solution, because solution y (x) = J (l/x) dx = In Ix + C of the differno ential equation is defined at x = O. We seethis graphically in Fig. which showsa direction field and sometypical solution curves for the equation y' = l/x. It is apparent that the indicated direction field \"forces\"all solution curvesnear the y-axisto plunge downward so that none can passthrough the point (0,

1.3.9,
0).)))

24

Chapter First-OrderDifferential Equations


1

>-0

;::....

//////) //\037//////
o)

/ / / / / /: / /////// /
Y2(X)

/ / / / /

/ / / / /

/ / / / /

/ / / /

/ / / /

/Yl(X)

= X2

/ / / /

/ / / /

= 0)

o
x)

(0,0)
o) x)

1)

Direction field and typical solution curves for the equation y' l/x.)

FIGURE

1.3.9.
=

FIGURE

1.3.10. field and two Direction


= 2-JY, y(O) =
curves for the
O.)

different solution

initial

value

problem y'

(b) [Failure of uniqueness]On the other hand, you can readily verify value problem) y/ == 2\037, y(O) == 0)
has the

that

the

initial

(8))

Problem 27).Figure different solutions Yl (x) == x2 and Y2(X) = 0 (see showsa direction field and thesetwo different solution curves for the initial value problemin (8).We seethat the curve Yl (x) == x2 threads its way through the indicated direction field, whereasthe differential equation y/ == 2,JY specifies slope y/ = 0 along the x-axis 2(X) == o. Y Example5 illustrates the fact that, before we can speakof \"the\" solution of an initial value problem,we need to know that it has one and only one solution. o and Questionsof existence uniquenessof solutions also bear on the processf mathematical modeling.Supposethat we are studying a physical system whosebeb havior is completely determined by certain initial conditions, ut that our proposed mathematical model involves a differential equation not having a unique solution satisfying thoseconditions. This raises an immediate question as to whether the the mathematical modeladequately represents physicalsystem. The theorem stated belowimpliesthat the initial value problem y/ == I(x,y), yea) = b has one and only one solution definednear the point x = a on the x-axis, and its partial derivative jay are continuous provided that both the function a near the point (a,b) in the xy-plane.Methodsof proving existencend uniqueness theorems are discussed the Appendix.) in

1.3.10

two

al

Y)

THEOREM 1

b)

continuous on some rectangle R in the xy-planethat contains the point in its interior. Then, for someopen interval I containing the point a, the value problem)
dy

Supposethat

of Existence UniquenessSolutions and both the function I(x,y) and its partial derivative Dyl(x,y) are

(a,b)

initial

FIGURE

R and x-interval of Theorem and the solution curve y y (x) b).) through the point

1.3.11. rectangle The


I
=
(a,

x)

\037)

dx

= I(x,y), yea) =

b)

(9))

1,

has one and only one solution that is definedon the interval (As illustrated in the solution interval I may not be as \"wide\" as the originalrectangle Fig. R of continuity; seeRemark 3 below.))))

1.3.11,

I.

1.3 Fieldsand SolutionCurves Slope


both pIe 1 and Fig.1.3.2(c), the function

25)

Remark

1:

In the

a jay = are continuous everywhere, so Theorem 1 impliesthe existence a of unique solution for any initial data (a,b). Although the theorem ensuresexistence x only on someopeninterval containing x = a, each solution y(x) = actually

-1

of case the differential equation dy/dx = -y of ExamI(x,y) = -y and the partial derivative

is definedfor all x.

Ce-

In the case the differential equation dyjdx = of of is continuouswherever Example5(b) and Eq. (8),the function y) = but the partial derivative aljay = Ij,JY is discontinuous when y = 0, and y henceat the point (0, This is why it is possiblefor there to existtwo different solutions Y1 (x) = x2 and Y2 (x) = eachof which satisfies the initial condition y(O) = o.)

Remark 2:

> 0,

I(x,

-2,JY

-2,JY

0).

0,

Remark In Example of Section we examined the especially 7 simdifferential equation dyjdx = y2. Here have I(x,y) = y2 and aljay = 2y. we ple Both of these functions are continuous everywhere in the xy-plane,and in particular on the rectangle < x < 2,0 < y < 2. Becausehe point (0,1) lies in t the interior of this rectangle, Theorem 1 guarantees a unique solution-necessarily a continuous function-of initial value problem) the

3:

1.1

-2

6)

dy
\037)

4)

y = 1/(1

dx

2 = y,

y(O) =

1)

(10))

x))

;::....

2)

--,
IR
I I
I)

on someopenx-interval containing a

= O. Indeedthis is the solution)


1

o)

y(x) = 1 -x)
that
4)

-2

-4)

-2)

o
x)

2)

FIGURE

point (0, leavesthe rectangle R beforeit reachesthe right sideof

curve through the

1.3.12. solution The


initial

1)

R.)

we discussedin Example7. But y(x) = Ij(1 x) is discontinuousat x = 1, so our unique continuous solution doesnot existon the entire interval < x < 2. Thus the solution interval I of Theorem 1 may not be as wide as the rectangle R where and aljay are continuous. Geometrically,the reasonis that the solution curve provided by the theorem may leave the rectangle-wherein solutionsof the differential equation are guaranteed to exist-before the it reaches one or both ends of the interval (see Fig.

-2

1.3.12).

.)

The following exampleshowsthat, if the function (x,y) and/or its partial derivative aljay fail to satisfy the continuity hypothesis of Theorem 1,then the initial value problemin (9) may have either no solution or many-even infinitely

many-solutions.)
.....,..... \037\"A\"'\" ........h

\"......

\037

..

\037

....

.....

.......A.......

........

,......

..,,,

......

Example6)

Considerthe first-orderdifferential equation)

x dy = 2y.) dx
Applying

(11))

Theorem 1 with I(x,y) = 2yjx and aljay 2jx,we conclude that must Eq. (11) have a unique solution near any point in the xy-plane where x i= O. that) Indeed,we seeimmediately by substitution in (11)

y(x) = Cx

2)

(12))))

26
4)

Chapter First-OrderDifferential
1

Equations)

(0,b)

(0,0))

x. In particular, the initial

satisfies Eq. (11) any value of the constant C and for all values of the variable for value problem)

2)

x dy = 2y, dx = 0.)

y(O) = 0)

(13))

;::.,.

0)

-2)

has infinitely many different solutions, hosesolutioncurves are the parabolas y = w Cx2 illustrated in Fig. C (In case = 0 the \"parabola\" is actually the x-axis

1.3.13.

-4

-2)

-1)

o
x)

2)

Observethat all these parabolaspass through the origin (0,0), but none of passesthrough any other point on the y-axis. It follows that the initial value problemin (13)has infinitely many solutions, but the initial value problem)
them
X
\037\037

FIGURE
infinitely

solution curves but no through the point (0, solution curves through the point
many

1.3.13. Thereare
0),

=2Y, y(O)=b)
any

(14))

(0,b) if b 1=o.)

has no solution if b i= O. Finally, note that through

2 if parabolasy = Cx . Hence, a

point off the y-axisthere passesonly one of the i= then the initial value problem)

0,

x dy = 2y, yea) = dx
4)

b)

(15))

x = 0 In summary,

has a unique solution on any interval that contains the point the initial value problemin (15)has)

x = a but

not the origin

2)

;::.,.

0)

. . .

a unique solution near (a,b) if a i= 0; no solution if a = 0 but b i= 0; infinitely many solutions if a = b = O.)

.)
(-1,

-2)

Still more can be said about the initial value problemin (15). onsidera C instance the point 1) indicated in Fig. typical initial point off the y-axis-for Then for any value of the constant C the function defined by)

1.3.14.

-4-2) FIGURE

-1)

o
x)

1)

2)

1.3.14. Thereare
(1,1).) -

infinitely many solution through the point

curves

y(x) = Cx)2 if x I

x2

if x

< 0, > 0)

(16))

is continuous and satisfies the initial value problem)

x- = 2y,
dy
dx)

y(-I)= 1.

(17))

For a particular value of C,the solution curve defined by (16)consistsof the left half of the parabola y = x2 and the right half of the parabola y = Cx2 Thus the 1) branchesat the origin into the infinitely many unique solution curve near solution curves illustrated in Fig. We therefore seethat Theorem 1 (if its hypotheses are satisfied) guarantees of uniqueness the solution near the initial point (a,b), but a solution curve through is (a,b) may eventually branch elsewhereso that uniqueness lost. Thus a solution may existon a larger interval than one on which the solution is unique. For instance, the solution y(x) = x2 of the initial value problemin (17)existson the whole x-00< x < axis,but this solution is unique only on the negativex-axis

(-1,

1.3.14.

O.)))

1.3 Fieldsand SolutionCurves Slope

27)

lIB)ProbleD1s)
In Problems through we have provided the slopefieldof the indicated differential equation, together with oneor more solution curves.Sketch likely solution curves through the additional points marked in eachslopefield.

10,

4.

dy dx)
\037

=x
I I I I le\\ \\ \\ \\ \\ \\ \\

3)
\\ \\ \\ \\ \\ \\ \\ \\
\\)

1.

dy dx)
3)

-y \\ \\ \\ \\ \\ \\

2)

sIn

x)
1)

\\\037

\037

-3 -2 -1 0 1 -3 FIGURE 1.3.15.)
X)

I ///////// ///////// -2 I111111111 /e/ I/ I / // I 1/11/1111/I /I /I /I I


I I I I I I I I / / / I I I I / / / / / I I / I I I I e/ / I
\037

,, ... \\.. \\\"\"\"'\\ \"\"\"'\ ,, ,, \"\"\"\ //P/////// / ////// -1 / /11111// I/


\\ \\ \\ \\ \\ \\ \\ \\ \\\\ \\\\

\\

2)

\\e\\

.'
\\ \\

\\ \\ \\ \\ \\ \\\\ \\\\

\\ \\

,,.. . .., \",--..-/) \\\",--..-/


\\ \\ \\ \\ \\ \\ \\ \\ \\ \\e\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\

\\\\\\\\\",-\\\\\\\",--..--..-/

\\\\\\ \\\\\\
\\ \\ \\ \\

0)

,, ,, ,)
\\ \\ \\

\\

\\

'Y

\\e\\ \\ \\ \\

\\.
\\

\\ \\ \\

\\

-1) -2)

\\ \\

\\ \\

\\

\\

\\ \\ \\

\\ \\ \\

// II //I/I / /II / I /II // I


I I
\037

\\ \\

\\

\\ \\

\\ \\

1)

\\ \\

\\

\\

\\

\\

\\

\\

0)

I I I / I I

-2 -1) o FIGURE1.3.18.)
X)

1111111 1111111 111111/ 111////)


I I leI
1)

I I I I

2)

3)

\037

I
I)

2)

3)

dy 2. dx) =x+y
3)

\037

5.
\037\037

=y
3)
/ /

+
/
/

1)

2)

11111/ /1111/ 1/1111 I


\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ e\\ \\ \\ \\ \\ \\ I \\ \\ \\ \\

1)

0)

-1 \\.\\\\.' -2 \\\\\\\\\\\\ \\\\\\\\\\\\ \\\\\\\\\\\\ -3 -2)\\\\\\\\\\\\) o -3 -1) FIGURE 1.3.16.)


\\
\037

,,,
\037

//II / / I II I / // / / I I
I)

/ / / / / / /el I I I / Ie / / / / I I I I I I / / /II/IIIII
I I I I I I I I I)

2//
/ 1 /,e, / I / I
\037

/ / / /1

I I

,
......--\",\\\\\\ --\",\\\\\\\\ - ,,

0)

-1) -2)

,, ,, ,, ,
\\ \\ \\ \\ \\

\\ \\

\\ \\

\\e\\ \\ \\ \\ \\
\\\037

\\

\\

\\

\\ \\ \\ \\

\\ \\ \\

\\ \\

\\ \\ \\

\\

\\

\\ \\

\\

\\ \\

\\

\\ \\

\\

\\

\\

\\

\\

1)

2)

3)

-3 -2 -1) o 1 -3 FIGURE 1.3.19.)


X)

\\\\\\\\\\\\\\\\\\ \\\\\\\\\\\\\\\\\\)
2)

\\e\\

3)

X)

dy 3. dx) = y - sIn x
3)

\037

dy 6. dx) = x - y + 1
I I I I I I /e I I I I fI
/) 1)

\037

2)

I I I I I I I I
I)

111111 I /I // // / I 111111 / / / / / I lieII


lei

3)

///////// ////////\037

1\\\\\\\\\\\\\\\\ \\\\\\\\\\\\\\\\\\
\\e\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\

2)

1)

0)

0)

\\e...\\\\\\\",\\\\\\\\\",-\",--......)
\\ \\ \\

.,,,,,, .
\\ \\ \\

\",\",',' -2 \\\"\"\"'\\
-1)
\\\\,e\",\037\\\\
\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\

,................-..-..-.................,)

-1) -2)
\037/

I /

I I I I

I I I

/ /
/

/ /
/ /
/)

-3 -2 -3

/ / / /)

leI

-1)

o
x)

1)

2)

3)

-2) -1)

o
x)

1)

2)

3)

FIGURE 1.3.17.)

FIGURE 1.3.20.)))

28
7.

dy. dx)
3)

Chapter
1

First-OrderDifferential Equations)

. = SIn x + SIn y
---........................................--

. 1 . dx) = -x2 + SIn y O dy


3) 2)

\037

2 1
\037

..-.-...--.......----,) ,.-------...-/)

/...,....-----//) /...-...----,.-/) /------...-/)


\"\"\"'-.;:\037 ,, ,,
\\e\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\
\037\\

1)

........--.--.....---....... ,.......-------........ ,.....----........, ,.......-------............... -2 , ........---------......... -3 \"'\\\\\\'\"0 -------///,..--3 -2 -1 x 2 3 FIGURE1.3.21.)


\\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\ \\

-1 '
0)

0)
\\

,-\\

\\

\\

\\

, , , , ,,

'-----.-1

-1 ,, -2 , -3 -2 -1 -3
\\

\\

\\

\\\037
\\

\\

\\

\\

0
x)

FIGURE 1.3.24.)
In

dy 8. dx = x 2 - y

\037

Problems through 20,determine whether Theorem 1does or does guarantee existence fa solution ofthe given initial not o is value problem. If existence guaranteed, determine whether Theorem 1doesor doesnot guarantee uniqueness of that solution. dy 11. = 2x2y2; dx

11

3 2 1 0

/ / / / / / / / / / / / / I
/ / / I

/1//\"\\\\ / -,
\\ \\ \\ \\ \\ \\ \\
\037

,e, / /... /., /1//--'


/ I /
/ / I / / /

/ - ,, /,.-....., 1//-\", / / -.....\"

,/ /--1

12.dy = x In y; dx
dy 14.dx = \037;

dy

y(1)=

-1

y(1)= 1

-1 -2 -3 -3
/ / /

dy 13. =\037; y(O) = 1 dx

y(O) = 0

-2

FIGURE1.3.22.

dy 15. = y'x - y; dx dy 16.dx = y'x - y; dy 17. dx = x 1;

y(2) = 2 y(2) = 1

dy 9. -=X2 -y-2 dx

3 2 1
\037

y- - y(O) = 1 18. dx y-=x-l; y(l) = 0


dy 19.dx = In(1 + y2);

y(O) = 0 y(O) = 1

0
I / / / / / /

/ /
/

-1 / -2 /// // / / / -3 -2 / -1 -3
/ /

It
/ /

/-,, ////-\", /1/--\" ////-\" /1//--' / ..-,/// /-/ // ///


/
I
\\

dy 20._ = x2 _ y2; dx
/ / / / / /

,e,
/ / / /

In Problems and 22, first use the method of Example 2 to construct a slopefieldfor the given differential equation. Then sketch the solution curve corresponding to the given initial condition. Finally, use this solution curve to estimate the desired value of the solution y

21

FIGURE1.3.23.

21. ' = x + y, y

22.y' = y - x,

(x). y(O) = 0; y(-4) =?

y(4) = 0; y(-4)=?)))

1.3 Fieldsand SolutionCurves Slope


Problems 23 and 24 are like Problems and 22, but now use a computer algebra system to plot and print out a slope fieldfor the given differential equation. If you wish (and know how), you cancheck your manually sketchedsolution curve by
plotting

29)

21

29.Verify

that

if c is a

constant,

then the function

defined

by) piecewise

23.y/ =
24.

it with the computer. x 2 + y2 y (0)

- 1,

y(x) satisfiesthe

y/=x+\"2 25.You bailout of the


ripcord of your
your downward

y2 ,

y (2) = ? y(-2) =0; y(2)=?

= 0;

{(x

for

c))

for

x x

< c, >
c)

of helicopter Example 3 and

Now k velocity satisfies the initial parachute.

= 1.6 Eq. (3), so in


value problem)

pull the

dv

dt
In

= 32 1.6v,

v(O) = O.)

3y2/3 for all Can differential equation y/ (x c)3in you alsouse the \"left half\" of the cubic y a solution curve of the differential equapiecing together Sketch a variety of such solution tion? Fig. curves. Is there a point (a, b) of the xy-plane such that the initial value problem y/ 3y2/3, yea) b has either no solution or a unique solution that is defined for all x? Reconcile your answer with Theorem 1.)

-x.

(See

1.3.25.)

order to investigate a slopefield for this

appropriate ity be? Will a strategically located haystack do any good? How long will it take you to reach95%of your limiting

your chancesof survival, construct differential equation and sketch the solution curve. What will your limiting veloc-

y)

y = (x

c)3)

velocity? 26.) Suppose deerpopulation P(t)in a small forest satisfies the the logistic equation)

x)
c)

dP
dt

2 = 0.0225P0.0003p
.)

Construct a slopefield and appropriate solution curve to answer the following questions: If there are 25 deer at time t 0 and t is measured in months, how long will it take the number of deer to double? What will be the

A suggestion FIGURE 1.3.25.

for Problem 29.) defined

limiting

deerpopulation?)

30.Verify

that

if c is a

constant,

then the function

by) piecewise

the fact that, if the hypothesesof Theorem are not satisfied, then the initial value problem y/ b may have either no solutions, y), yea) finitely many solutions, or infinitely many solutions.

The next sevenproblems illustrate

= f(x,

y(x)

= cos(x

+1

1-1)

1 X _ c, f < c) if c < x < c + T{, if x > c +


T{)

27. (a) Verify

that

if c is a constant,

then the function defined

piecewise by)
y(x) satisfies the

=
{\037x

C)2)

for x for x
y/

< c, >
c)

differential

28. erify V
many

x?

Construct a figure illustrating the cluding the point x fact that the initial value problem y/ 2-JY, y (0) 0 has infinitely many different solutions. (b) For what values of b doesthe initial value problem y/ 2-JY, y(O) b have (i) no solution, (ii) a unique solution that is defined for all

= c).

equation

= 2-JY for all

31.
x (in-

Doesit suffice

32.

cos(x c) with sin(x c) simply to replace in piecing together a solution that is defined for all x? Verify that if c > 0, then the function defined piecewise
y(x)

I y2 for all x. satisfiesthe differential equation y/ = a preliminary sketch with c = 0 will be helpful.) (Perhaps Sketch a variety of such solution curves. Then determine (in terms ofa and b) how m any diffe rent solutions the iniI y2, yea) = b has. tial value problem y/ = Carry out an investigation similar to that in Problem 30, exceptwith the differential equation y/ = + I y2.

-J -

-J -

J --

by)

that if k is a constant, then the function y (x) = kx satisfies the differential equation xy/ = y for all x. Construct a slope field and severalof these straight line solution curves.Then determine (in terms of a and b) how

none, or infinitely yea) = b has-one,

different solutions the initial value problem xy/


many.)

= y,

satisfies the differential equation y/ = 4x-JY for all x. Sketch a variety of such solution curves for different values of c.Then determine (in terms ofa and b) how many different solutions the initial value problem y/ = 4x JY, yea) = b has.)))

{(x - c)2
2

if x 2 < C, if x 2 > c

30
33.If c
c.

Chapter
1

First-OrderDifferential Equations)

xj(cx 1)(with

that the function defined by y (x) satisfies graph illustrated in Fig. the differential equation x 2y/ + y2 0 if x i= Sketch a variety of such solution curves for different values of 0 that Also, note the constant-valued function y(x) doesnot result from any choiceof the constant Finally, determine (in terms of a and b) how many different solutions the initial value problem x 2y/ + y2 0, yea) b has.)

i=

- 0,

verify

1.3.26) 1jc.
= c. =

34. (a) Usethe

direction

values at x

ential equation

x + 1 with y/ = Y = -1.2 y(-l)= -0.8.) and y(-l)

field

1 of the

of Problem 5 to estimate the two solutions of the differ-

initial

values

(b)

Usea computer algebra system


ues at x

3 of the

tial equation

y( -3)= -2.99.)

to estimate the valtwo solutions of this differenwith initial values y( and

-3)= -3.01
initial

..........\"\"
\\
;::..,.

\" \\ \\ \" \" \\

I \\
1\\

\\

1\\

--..... -=--) :::---..........\" -....................\"


\\ I
\037

--..........\"
\\

1\\

\\\"\" \"\"
..........

\\

\\ \"

Thelessonof this problem is that small changesin conditions can make big differences results.) in

35.(a) Usethe

direction

-==.

values at x

field of Problem 6 to estimate the 2 of the two solutions of the differy/

ential equation

..........\"\"
\\ \\

\"..........--)
\\ \\ \\

and y( -3)= -0.2 y( -3)= +0.2.) at

+ 1 with

initial

values

\" \" \\ \" \\ \\

\\ \\

\\

\\

\" \" \\ \" \" \\ \\


..........

(b)

Usea computer algebra system


ues y( x

3 of the

tial equation

x)

-3)= +0.5.

to estimate the valtwo solutions of this differenwith initial values y( and

-3)= -0.5

FIGURE1.3.26. field for x 2y/ + y2 Slope and graph of a solution y(x) = xj(cx 1).)

=0

The lessonof this problem is that big changes in initial conditions may make only small differences results.) in

1.3Application)

\037hN_\037\037'\037PN\037\037\037,\037

:-Q

\037N\037\037\037\037t\037N\037__\037\037!\037p\037m,\037\037elds

and SolutioI?Curv\037,s)

Widely

available computer algebrasystemsand technical computing environments includefacilities to automate the constructionof slopefields and solution curves, as do somegraphing calculators (see Fig.1.3.27).)

FIGURE 1.3.27. field Slope


equation) dy

with

-5< x,

initial y

dx points (0, b), b = < 5 on a TI-89graphing

and solution

curves for

the differential

. = sln(x

y)

and -3,-1, 2,0,2,4 window calculator.)

Maple\342\204\242,

The applications manual accompanying this textbook includesdiscussion of and f resourcesor the investigation of difMathematica\342\204\242, ferential equations.For instance, the Maplecommand
MATLAB\342\204\242

wit.h(DEt.ools):

y=-S..S);))) DEplot.(diff(y(x),x)=sin(x-y(x\302\273, x=-S..S, y(x),

1.3 Fieldsand SolutionCurves Slope


and the Mathematica command
\302\253

31)

{x, Plot.Vect.orField[{l, Sin[x-y]}, -5, 5}, {y, -5, 5}] itFigure 1.3.28 produceslope fields similar to the one shown in Fig. 1.3.28.

Graphics\\Plot.Field.m

o 1 2
x)

345
equation

FIGURE1.3.28. Computergenerated slopefield and solution


curvesfor the differential y' = sin (x y).)

self was generated with the MATLAB program [John Polking and David Arnold, Ordinary DifferentialEquations UsingMATLAB, 2nd edition, UpperSaddle River, NJ: Prentice Hall, that is freely available for educational use W (math.rice.edul\037dfield). hen a differential equation is entered in the p you can (with mousebutton clicks) lot both a slopefield setup menu (Fig. and the solution curve (or curves) through any desiredpoint (or points). Another ODE packagewith impressive freely available and user-friendly MATLAB-based

dfield

1999]

1.3.29),

dfield

i (www.math.uiuc.eduliode).) graphical capabilitiess lode

MATLAB FIGURE 1.3.29.

for y'

= sin(x

dfield setup

to construct slopefield

and solution curves

y).)

Usea graphing

You might warm up Problems1 through

calculator or computer system in the followinginvestigations. by generating the slope fields and some solution curves for lOinthis section.)

INVESTIGATION Plot a slopefield and typical solutioncurvesfor the differenA: sin(x y), but with a larger window than that of Fig.1.3.28. < x < 10, < Y < 10,or instance, a number of apparent straight f With line solution curves should be visible.
tial equation dy/dx =

-10

-10

(a) Substitute

cientsa and b must be in orderto get a solution.

= ax + b in

the differential equation to determine what the coeffi-

(b) A computer algebra system gives the general solution)

y(x)=x-2tan-(
1

X-2-C . x-c) )
1.3.28.

Plot this solution with selected valuesof the constant C to comparethe resulting that Can you see no value of solutioncurves with thoseindicated in Fig. = x n/2 correspondingo the initial condition t C yieldsthe linear solution y y(n/2) = O? Are there any values of C for which the correspondingsolution curves lieclose this straight line solution curve?))) to

32

Chapter
1

First-OrderDifferential Equations)
digit

INVESTIGATION B: For your own personalinvestigation, let n be the smallest in your student ID number that is greater than and considerthe differential

equation)

-dy

1,

dx

= 1cos(x ny). n
of possibility straight line

(a) First investigate (as in


solutions.

part (a) of InvestigationA) the

(b) Then generate a slopefield for this differential equation, with the viewing window chosen that you can picture someof thesestraight lines,plusa sufficient so

y(x) behaves as x --* +00. (c) A computer algebrasystem gives the general solution) y(x) =
\037

number of nonlinear solution curves that you can formulate a conjecture about what happensto y(x) as x --* State your inference as plainly as you can. Given the initial value y(O) = Yo, try to predict(perhapsin terms of Yo) how

+00.

[x

+ 2 tan- x c ( )J.)
1
\037

....
_ _ ___\" __ ____

Can you make a connection between this symbolicsolution and your graphisolution curves (straight linesor otherwise)?) cally generated

Separable Equationsand Applications -.JIIIIiI,\" , ____


\037
_\037_ \037 ___

_\037\"___

___'_'Nm'_N______\037

_N_

\",\"

______O\"\"N'N_N

__o\037

,__

_\037,__

\"0'

\"Nm_____w_,,_,_

\037m,_

_mN____

,_

_m

___\"'

NN

__

___

____

_\"

_,.w

__

__\"N'\"

___

___

__N_

N_O\037\"y\"_'_.'.N\037..'NY'

N)

The first-orderdifferential equation)

dy
dx)

= H(x,y)

(1))

is calledeparablerovided that H(x,y) can be written as the product of a function s p


of x and a function of y:)
\037)

dy

dx

= g(x)h(y)= g(x) ,
fey))

where h (y) = 1/ (y). In this case variables x and y can be separated-isolated the on oppositesidesof an equation-bywriting informally the equation)

fey) dy = g(x)dx,)
which we understand to

be conciseotation for the differential equation) n


dy f(y)- = g(x). dx)
(2))
integrating

It is easy to solve this specialtype of differential equation simply by both sideswith respectto x:)

f(y(x\302\273
\037\037

dx =

f g(x)dx +

c;)))

1.4 Equationsand Applications Separable


equivalently,)

33)

f fey) dy f g(x)dx + c.)


All that

(3))

is requiredis that the antiderivatives)

F(y) =

f f(y)dy
,
(y(x\302\273y

and G(x) =

f g(x)dx)

can be found. To seethat Eqs. (2)and (3)are equivalent, note the following consequenceof the chain rule:)
Dx[F(y(x\302\273J

= F,

(x) = f(y)- = g(x) = Dx[G(x)J, dx)

dy

which

in turn

is equivalent to)
F(y(x\302\273

= G(x)+ C,)
only

(4))

because two

functions have the same derivative on an interval if and differ by a constant on that interval.)

if

they

Example

1)

Solvethe initial value problem)


dy

dx
Solution
Informally,

= -6xy,

y(O) = 7.)
y

sideby dx to get)

we divide both sidesof the differential equation by

dy
y)

and

multiply

each

= -6xdx.

Hence)

f d;
8 6 4 I 2 I 0 I
I I I I \\

d f (-6x) x;
that

In Iyl

= -3x2 C.) +
y(x) is positive near x = 0, so we

We may

seefrom

the

initial

deletethe absolute value symbols:)


lny

condition y(O) = 7

;;:....

/
\"-

-2 -4 -6 -8

= -3x2 C,) +

\\ \\ I I I I

and hence)

.-2

where A

3x2 c 3x2 C 3x2 y(x) = e- + = e- e = Ae- , = eC. The condition y (0) = 7 yieldsA = 7, so the desiredsolution is)

(0,-4))

FIGURE
solution

curves for y/

1.4.1. = field and Slope

y(x) = 7e-

3x2
.)

Example 1.)

-6xyin

This is the upperemphasized solution curve shown

in

Fig.1.4.1.)

.)))

34

Chapter
1

First-OrderDifferential Equations)

Remark: Suppose,instead,that the initial condition in Example1 had been Then it would follow that y(x) is negative near x = O. We should y(O) = therefore replace with -y in the integrated equation In Iyl = -3x2 C to + Iyl obtain

-4.

In(

+ -y) = -3x2 C.
3x2 _4e.)

The initial condition then yieldsC = In4, so In(-y) =

-3x2 In4,and hence) +

y(x) =

This is the lower emphasized solution curve in


\037 ,,\"\"

Fig.1.4.1.)

.)

,,,,nn\"'\"''

In\"

\"\"\"

nn\"n,,,',

Example2)

Solvethe differential equation)

2x - -dx
dy

3y2

5.)

(5))

Solution
6)

When

we separatethe variables and integrate both sides,we get)

.,.'., .. .....' //,.,.......----- ,.


....-

----,-- .. -. \"\",.\".-\"'-\"'---------------)
. ------: :---

,
.)

--)

(1, 3) ,

.-----........\"\"\"

f (3l-5)dy = f (4 -22x)dx; _
y3

5y

= 4x x + C.)

(6))

This equation is not readily solved for y as an explicitfunction


;;:....

of x.)

.)

0)

-2) -4)

-6 -6 -4 -2 0

. . . -----------------) 2
x)

6
and

Example2 illustrates, it mayor may not be possibleor practical to solve Eq. (4) explicitly for y in terms of x. If not, then we call (4) an implicit solution of the differential equation in (2). Thus Eq. (6) gives an implicit solution of the differential equation in (5).Although it is not convenient to solve Eq. (6)explicitly in terms of x, we seethat each solution curve y = y(x) lies a contour (or level) on
As

curve where the

function)

FIGURE1.4.2. field Slope


solution curves for
y/

Example 2.)

= (4

- 2x)/(3 - 5)
y

_
2
in

H(x,y) = x2 4x + y3

-5

y)

is constant. Figure 1.4.2 showsseveralof thesecontour curves.)


\037-------_.

._---------,-\037----,------,---,--,,------_..,--,\037--,..,----,------------'--'--\037--------\037-'--To solve the initial value problem)

-- -dy

----------.-(7))

2x
5)

dx

3y2

y(l) = 3,

we substitute x = 1 and y = 3 in Eq. (6)and get C = 9.Thus the desiredparticular solution y(x) is defined implicitly by the equation)
y3

_ 5y = 4x x2 + 9.)

(8))

The corresponding solution curve y = y(x) lieson the upper contour curve in 1.4.2-the passingthrough (1, ). Because graph of a differentiable one the 3 Fig. solution cannot have a vertical tangent line anywhere, it appearsfrom the figure that this particular solution is defined on the interval 5) but not on the interval
(

-3,7).

(-1,

.)))

1.4 Equationsand Applications Separable


20 15 10 5 0

35)

Remark
attempt to

in Eq. (8),we can When a specificvalue of x is substituted solve numerically for y. For instance, x = 4 yieldsthe equation)

1:
\037

-\037

fey) = y3

- 5y - 9 =

o.)

-5

-10 -15 -20 -4 -2 -6

showsthe graph of With a graphing calculatorwe can solve for the Figure 1.4.3 of This singlereal root y 2.8552. yieldsthe value y(4) 2.8552the particular solution in Example3.

f.

\037

0
y)

FIGURE1.4.3. Graph of fey) = y3 5y 9.)

- -

0 particular solution through (1, ) is defined on the interval (0,4) but not on the we get the interval On the other hand, with the initial condition y(l) = lower contour curve in Fig.1.4.2. particular solution is defined for all x. Thus This the initial condition can determine whether a particular solution is defined on the whole real line or only on someboundedinterval. With a computeralgebra system one can readily calculate a table of values of the y-solutions of Eq. (8)for x-values at desiredincrements from x = to x = 5 (for instance).Sucha table of values serveseffectively as a \"numerical solution\" of the initial value problem in (7).

Remark 2: If the initial condition in (7) is replacedwith the condition = 0,then the resulting particular solution of the differential equation in (5) y(l) lieson the lower \"half\" of the oval contour curve in Fig.1.4.2.appears that this It

(-1,5).

-2

-1

.)

and S General, Singular olutions Implicit,


3
2)

1)

x2 + y2 = 4 yields)

The equation K (x,y) = 0 is commonly calledan implicitsolutionof a differential equation if it is satisfied (on some interval) by some solution y = y (x) of the differential equation. But note that a particular solution y = y (x) of K (x,y) = 0 mayor may not satisfy a given initial condition. For example,differentiation of

\037

0)

-1) -2) o
x)

x+

dy y- = 0, dx

so x2 + y2 = 4 is an implicit solution of the differential equation x only the first of the two explicitsolutions)
1

+ yy' = O.But

FIGURE1.4.4. field Slope


solution

curves for

y/

y(x) = + 4
satisfies the
initial

and xjy.)

J -x

2 and

y(x) =

-J4 -

2)

should not assumethat every possiblealgebraic solution solution satisfies the same differential equation. For iny implicit stance,if we multiply the implicit solution x2 + y2 4 = 0 by the factor (y 2x), then we get the new implicit solution)
You

Remark = y (x) of an

1:

condition y (0) = 2 (Fig.1.4.4).

(y
that

- 2x)(x + - 4) =
2
y2

0)

yields (or \"contain s\") not only the previously noted explicitsolutions y 4 x2 and y = ,y'4 x 2 of the differential equation x + yy' = 0,but also the +-J additional function y = 2xthat doesnot satisfy this differentialequation. Remark 2: Similarly, solutions of a given differential equation can be either gained or lost when it is multiplied or divided by an algebraic factor. For instance, considerthe differential equation)

- -

(y

- 2x)y-= -x(y- 2x)


dy
dx)

(9))))

36

Chapter
1

First-OrderDifferential Equations)
having the

factor (y

-obvious solution 2x), we get


then

y = 2x. But if we divide both sidesby the common differential equation) the previously discussed

dy
y

dx

= -x,)

or

x+

dy y- = 0,
dx)

(10))

= 2xis not a solution. Thus we \"lose\" the solution y = 2xof Eq. (9) upon its division by the factor (y 2x);alternatively, we \"gain\" this new solution when we multiply Eq. (10)by (y 2x). Suchelementary algebraicoperations to
of which
y

75)

50)
;>..,)

25)

o)

-15-10-5

0
x)

10 15

FIGURE1.4.5. general The solution curves y = (x C)2and the singular solution curve y = 0 of the differential equation

(y/)2 = 4y.)

general solution of (2),but the general solution of (2). In Section we shall seethat every particular solution of a linearfirst-order differential equation is contained in its general solution. By contrast, it is common for a nonlinear first-order differential equation to have both a general solution involving an arbitrary constant C and one or several particular solutions that s cannot be obtained by selecting value for C. Theseexceptional olutions are a singularsolutions.In Problem 30 we ask you to show that the frequently called general solution of the differential equation (y')2 = 4y yieldsthe family of parabolas y = (x C)2 illustrated in Fig. and to observethat the constant-valued function y(x) = 0 is a singular solution that cannot be obtained from the general solution by any choicef the arbitrary constant C.) o

of practice,but the possibility loss or gain of such \"extraneoussolutions\" should be kept in mind. A solution of a differential equation that contains an \"arbitrary constant\" (like the constant C in the solution of Examples1 and 2) is commonly calleda general o solution of the differential equation; any particular choicef a specificvalue for C yieldsa singleparticular solution of the equation. The argument preceding Example1 actually sufficesto show that everyparticular solution of the differential equation f (y) y' = g (x) in (2)satisfies the equation F(y(x)) = G(x)+ C in (4).Consequently,it is appropriate to call (4) not merely a

simplify

a given differential equation before attempting to solve it are common in

1.5

1.4.5,

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

Example4)

. Find all solutions of the differential equation)


......

dy
dx)

= 6x(y 1)2/3.

Solution) Separation of variables gives)


4)

2)

! !
1 2/ 3 dY (y 2

= 3(y - 1) - 1)1/3= x + C;

2XdX;

;>..,

0)

2 y(x) = 1 + (x + C)3.)

-2) -4)

-2

-1

0) x)

2)

FIGURE1.4.6. Generaland
singular solution
y/
==

6x(y

curves for

1)2/3.)

Positive values of the arbitrary constant C give the solution curves in Fig. 1.4.6 that lie above the line y = whereasnegative values yield those that dip below it. The value C = 0 gives the solution y(x) = 1 + x6, but no value of C gives the singular solution y(x) = 1 that was lost when the variables were separated. Note that the two different solutions y(x) = 1 and y(x) = 1 + (x2 1)3both Indeed,the whole singular solution curve satisfy the initial condition y(l) = = 1 consistsof points where the solution is not unique and where the function y .))) (x,y) = 6x(y 1)2/3is not differentiable.

1,

1.

1.4 Equationsand Applications Separable


Natural GrowthandDecay
The differential equation)
\037)

37)

dx =
dt)

kx

(k a constant)

(11))

servesas a mathematicalmodelfor a remarkably widerange of natural phenomenaany involving a quantity whosetime rate of change is proportionalto its current size. Here someexamples.) are
POPULATION GROWTH: Supposethat pet) is the number of individuals in a population (of humans, or insects,or bacteria) having constant birth and death rates {3 and 8 (in births or deaths per individual per unit of time). Then, during a short time interval \037t, approximately {3P(t) births and 8P(t) deaths occur, so the change in pet) is given approximately by)
\037t \037t

\037P

\037

({3

- 8)P(t)

and therefore)

dP
dt)

lim
\037t--+O

where k

= {3

\037P
\037t)

\037t,)

= k P,

(12))

8.)

COMPOUND INTEREST: Let A(t) be the number of dollarsin a savingsaccount supposethat the interest is compounded continuously at r. (N ote that 10%annual interest means that r = 0.10.) Continuous compounding means that during a short time interval \037t, the amount of interest addedto the account is approximately \037A = r A(t) \037t, so that)
at time t (in years), and an annual interest rate

- dA

dt

\037A
\037t)

lim

= r A.

(13))

\037t--+O

RADIOACTIVE DECAY: Consider sampleof material that contains N(t) atoms a of a certain radioactiveisotopeat time t. It has beenobserved a constantfraction that of those radioactiveatoms will spontaneouslydecay(into atoms of another element or into another isotopeof the sameelement) during each unit of time. Consequently, the samplebehaves exactly likea population with a constantdeathrate and no births. To write a modelfor N(t), we use Eq. (12)with N in place of P, with k > 0 in of place 8, and with {3 = O. We thus get the differential equation)

dN
dt)

= -kN.

(14))

The value of k dependson the particular radioactive isotope. The key to the method of radiocarbondating is that a constant proportion of the carbon atoms in any living creature is made up of the radioactive isotope 14Cof carbon.This proportion remains constant becausehe fraction of 14Cin the t remains almost constant, and living matter is continuously taking up atmosphere carbon from the air or is consuming other living matter containingthe sameconstant ratio of 14Catoms to ordinary 12Catoms. This sameratio permeatesall life, because seem organic processes to make no distinction betweenthe two isotopes.)))

38

Chapter
1

First-OrderDifferential Equations)
The ratio of 14Cto normal carbon remainsconstant in the atmospherebecause, 14Cis radioactive and slowly decays,the amount is continuously replenalthough ishedthrough the conversionof 14N(ordinary nitrogen) to 14Cby cosmic bomrays barding the upperatmosphere.Over the long history of the planet, this decay and has into replenishment process come nearly steady state. Of course, hen a living organism dies,it ceasests metabolism of carbon w i and the process radioactive decaybeginsto depleteits 14Ccontent. There is no 'of a replenishment of this 14C, nd consequently the ratio of 14Cto normal carbon begins to drop. By measuring this ratio, the amount of time elapsedsince death of the the to organism can be estimated.For such purposesit is necessary measure the decay constantk. For 14C,t is known that k 0.0001216 measured in years. i if t is (Matters are not as simpleas we have made them appear.In applying the techmust be taken to avoid contaminating the nique of radiocarbon dating, extreme care samplewith organic matter or even with ordinary fresh air. In addition, the cosmic ray levels apparently have not been constant, so the ratio of 14Cin the atmosphere has varied over the past centuries. y using independent methods of dating samB in r ples, esearchers this area have compiledtablesof correction factors to enhance the accuracy of this process.))
\037

m o bloodstream, easured the excessver the natural level of the drug, by


at a rate proportional to the current

DRUGELIMINATION: In many

cases the

amount A(t)

of a certain drug
will

in the

dA
dt) In

decline

excessmount. a
== -AA,

That is,)
(15))

where A > O.The parameter A is called elimination constant the drug.) of the

The NaturalGrowthEquation
The prototype differential equation dxjdt == kx with x(t) > 0 and k a constant (either negative or positive) is readily solved by separating the variables and integrating:)

f
Then we solve for x:)

dx =
\037

f k dt;
== Ae kt

x == k t + C.)

e1nx

== ekt +C;

x == x(t) == eCekt

.)

with particular solution of Eq. (11) the


\037)

Because is a constant, so is A C

== eC It

is also clearhat A == x(0) == Xo, so the t initial condition x(O)== Xo is simply)


.)

kt x(t) == xoe

(16))
in

Because the presencef the of o differential equation)


\037)

natural

dt)

exponential function
kx

its solution,

the

dx ==

(17))

is often called exponential naturalgrowth equation.Figure 1.4.7 a the or shows of x(t) in the case > 0;the case < 0 is illustrated in Fig.1.4.8.))) k k typical graph

1.4Separable Equationsand Applications


x) x)

39)

x = Xo e (k > 0))
kt

Xo)

Xo)

t)

FIGURE 1.4.7. Natural

growth.)

Natural FIGURE1.4.8.

decay.)

Example5)

the According to data listed at www.censlls.gov, world'stotal population reached 6 billion personsin mid-1999, was then increasing at the rate of about 212 and thousand personseachday. Assuming that natural population growth at this rate continues, we want to answer thesequestions: (a) What is the annual growth rate k? st (b) What will be the world population at the middle of the 21 century? How long will it take the world population to increaseenfold-therebyreach(c) t b billion that somedemographerselieve to be the maximum for which the ing the 60 planet can provide adequate food supplies?)
in years. so take t == 0 to correspond (mid) 1999, Po == 6.The fact that P is increasing to or 0.000212 personsper day at time t == 0 means that) billion, by 212,000,

Solution (a) We measure the world population pet) in billionsand measure time
We

0.07743) p'(0) == (0.000212)(365.25)


\037

billion per year. From the


obtain)

natural

growth

equation P' == k P with


\037

t == 0

we now

k ==

P'(O) 0.07743
\037

P(O)

0.0129.
in

Thus the world population was growing at the rate of about 1.29% annually This value of k gives the world population function)

1999.

P(t)
(b) With
t ==

== 6eO.0129t.)

51we obtain

the prediction)
\037

P(51)== 6e(0.0129)(51)11.58 (billion))


for the world population in mid-2050(sothe population will almost have doubled in the just over a half-century since1999). (c) The world population should reach 60 billion when)

60== 6eO.0129t;)
and
thus in

that

is, when

t ==

0.0129) 178;
\037

In

10

the year 2177.)

.)))

40

Chapter
1

First-OrderDifferential Equations)

Note: Actually, the rate of growth of the world population is expectedto slow somewhat during the next half-century, and the best current prediction for the 2050population is \"only\" 9.1 billion. A simple mathematical model cannot be to expected mirror preciselythe complexity of the real world. The decayconstant of a radioactive isotopeis often specified terms of anin other empirical constant, the half-life of the isotope,because parameter is more this convenient. The half-life 1: f a radioactive isotopeis the time required for half of o it to decay. To find the relationship betweenk and a we set t = 1: nd N = 4 No kt kT When we solve for in the equation N (t) = Noe , so that we find 4 No = Noe
that)

= 1: In2
k)
\037

-.
(In

1:, .

1:,

(18))

For example,the half-life of 14Cis 1: years.)

2)/(0.0001216), approximately 5700

Example6)
Solution

turns out to contain 63% as much 14C specimenof charcoal found at Stonehenge as a sampleof present-day charcoal of equal mass.What is the ageof the sample?)

take t = 0 as the time of the death of the tree from which the Stonehenge charcoal was made and No as the number of 14Catoms that the Stonehenge sample contained then. We are given that N = (0.63)Noow, so we solve the equation n kt = Thus (0.63)No Noe- with the value k = 0.0001216.we find that)
We
t

In(0.63) 3800 -0.0001216) (years).


\037

Thus the sampleis about 3800 old.f it has any connection with the buildersof I years our computations suggesthat this observatory,monument, or templet Stonehenge, whichever it may be-dates 1800 .C. earlier. from B or

.)

a Coolingnd Heating
the time rate of According to Newton'slaw of cooling (Eq. (3) of Section in change of the temperature T(t) of a body immersed a medium of constant temperature A is proportional to the difference A T.That is,)

dT
dt) dt)

= k(A

- T),

1.1),

(19))

where k is a positive constant. This is an instanceof the linear first-orderdifferential equation with constant coefficients:)
\037)

dx =

ax + b.

(20))

It

includesthe exponential equation as a specialcase(b solve by separation of variables.)


50\302\260F, 375\302\260F

= 0) and is also easy to


P.M. 5:00 After
75

Example7)

4-lb roast, initially at is placed in a oven at it is found that the temperature T(t) of the roast is roast be (medium rare)?)))
A

minutes

125\302\260F.

When will the

150\302\260F

1.4Separable Equationsand Applications


Solution
We

41)

P to take time t in minutes, with t = 0 corresponding 5:00.M. We alsoassume (somewhat unrealistically) that at any instant the temperature T (t) of the roast is uniform throughout. We have T(t) < A = 375,T(O) = 50,and T(75)= 125. Hence)

dT
dt
\037\037\037

= k(375 T);
k

' - T dT = f dt. -In(375 T) = + 375 - T = Be375


kt

C;)
.)

kt

Now T(O) =
that

kt . 50impliesthat B = 325,so T(t) = 375 325e- We alsoknow T = 125 when t = 75.Substitution of thesevalues in the preceding equation

yields) k

1 75 In (

\037

0.0035.

Hence finally solve the equation) we

= 150 375
for
t
\037

- 325e(-o.0035)t)

= -[In(225j325)]j(0.0035) 105(min), the total cookingtime required. Because roast was placedin the oven at 5:00.M.,t should be removedat about the P i P.M. .) 6:45

Torricelli's Law)
that

Supposethat a water tank has a hole with area a at its bottom, from which water is leaking.Denoteby yet) the depth of water in the tank at time t, and by Vet) the volume of water in the tank then. It is plausible-and under ideal conditionstrue,
the velocity of water exiting
through

the hole is)


,)

= J2gy

(21))

which is the velocity a dropof water would acquire in falling freely from the surface Onecan derivethis formula of the water to the hole (see Problem 35 of Section beginning with the assumptionthat the sum of the kinetic and potentialenergyof the systemremains constant. Underreal conditions,aking into account the constriction t of a water jetfrom an orifice, v = c,J2gy, where c is an empirical constantbetween o and 1 (usually about for a small continuous stream of water). For simplicity we take c = 1 in the following discussion. As a consequence we have) ofEq.

1.2).

0.6

\037)

dV
dt)

(21),

= -av = -aJ2gy ;

(22a))

equivalently,)

\037)

dV
dt)

= -k,JY

where k

= aJ2i.

(22b))

This is a statementof Torricelli's for a draining tank. Let A(y) denote the horilaw zontal cross-sectional of the tank at height y. Then, appliedto a thin horizontal))) area

42

Chapter
1

First-OrderDifferential Equations)

sliceof water at height y with area method of cross sectionsgives)


v (y) The fundamental

A (y )

and

thicknessd y , the integral calculus

1
Y

A (y ) d y

\302\267)

hencethat)

theorem of calculustherefore impliesthat dVjdy

A(y)

and)

dV

dt
From

dV

dy

. dy

dt

= A(y) dy

dt

\302\267)

(23))

Eqs. (22)and (23)we finally

obtain)

dy A(y)- = -ay\037= -k\037, 2gy dt) an alternative form of Torricelli's law.)


\037

(24))

.) ..
....

Example8)

b hemispherical owl has top radius 4 ft and at time t = 0 is full of water. At that moment a circular hole with diameter 1 in. is openedin the bottom of the tank. How
A

long will

it

take for all the water to drain from the we triangle in Fig.1.4.9, seethat)
A(y)

tank?)

Solution

From the

right

= Jrr 2 = Jr [16 (4

- - y)2] = Jr(8y _
( 24)
A
1

y2).)

Positive y-values)

With

= 32 ftj s2, Eq.(24)becomes)


Jr(8y

2 dy

)-= -Jr
dt
5/2

2
V

/2.32y;

d f (8y I/2 -l/2)y = - f


16 3/2 _ y
3 t
I)

\037

5y

- -1.. +
72 t
\037

dt;
C
.)

Now y(O) =

4, so) C

FIGURE1.4.9. Draining a
hemispherical
tank.)

16 43/2 _ 3
when)

.45/2 _ 448 15.)


2150 (s);)
tank

The tank is empty when

= 0,thus

= 72 .

\037
41\0378

that

is, about 35 min 50 s. So it takes slightly lessthan 36 min for the

to
.)))

drain.

1.4Separable Equationsand Applications

43)

__ Problems)
generalsolutions (implicit if necessary, xplicit if convee 1 nient) of the differential equations in Problems through 18. Primes denotederivatives with respectto x.)
Find dy 1.dx + 2xy = 0 . 3. dy = SIn x

30.Solvethe differential
=

dx dx

dy 2. dx + 2xy 2 = 0

dy

4. (1+ x)dy = 4y
dx

5. 2..jX = Jl _ y2 dx dy = j 7. (64xy)l 3

9. (1

11.' = xy3 y
dy

13.3 dx = (y4 + 1)cosx y 15.dx


dy

dy

x 2) dy dx

= 2y

(x

17.y' =

18. 2y' = 1 - x 2 + y2 _ x 2 x
Find explicit in Problems

side. )

x 2(2y3 y) 1+x+ y +xy

- 1)y5 -

6. dx dy 8. dx = 2x secy dy 10. 1+ X)2 dx = (1+ y)2 ( 12.y' = X(y2 + 1) y dy 14.dx = 1 + ..jX 1+


= 3vfXY
,JY)

t 31. iscusshe differencebetween the D

4y to verify the equation (dy/dx)2 general solution curves and singular solution curve that Then determine the points are illustrated in Fig. (a, b) in the plane for which the initial value problem 4y, yea) b has (a) no solution, (b) infinitely (y')2 (c) on some many solutions that are defined for all of the point x a, only finitely many soluneighborhood ti 0n

1.4.5.
=

x,

s.

(dy/dx)2 = 4y and dy/dx = 2,JY. Do they have the same solution curves? Why or why not? Determine the points (a, b) in the plane for which the initial value problem y' = 2,JY, yea) = b has (a) no solution, (b) a unique
solution,

differential

equations

32.Find

(c) infinitely

many

solutions.

16.x2 + l)(tan y)y' = (


(Suggestion:Factor the
y2)

x)

right-hand

19 through
X

particular

solutions

28.

of the

initial

value problems

dy 19.dx = ye ,

y(O) = 2e y(O) = 1 y(5)

a general solution and any singular solution s of the Deterdymyslashdx = y y2 (a,b) in the plane for which the initial value problem y' = y y2 yea) = b has (a) no solution, (b) a unique solution, (c) infinitely many solutions. (Population growth) A certain city had a population of Assume 25000in 1960and a population of30000in 1970. that its population will continue to grow exponentially at a constant rate. What population can its city planners expect in the year 2000? 34. (Population growth) In a certain culture of bacteria, the sixfold in 10 h. How long number of bacteriaincreased
differential equation mine the points

.J

- 1,

.J

- 1.

33.

did it take for the population

to

double?

20.

dy

dx

= 3x2(y2 + 1),
x
-VX2

35.(Radiocarbon dating)

dy 21. y dx = 2

22. 23.

dx
dy dy dy

dy

= 4x3y

, - 16
y,

=2

36.(Radiocarbondating)

Carbon extracted from an ancient as skull contained only one-sixth as much 14C carbon extracted from present-day bone.How old is the skull?
from a purported relic 10 atoms of 14C contained 4.6 x Carbon extracted from a present-day specimen per gram. x 10 atoms of 14C of the samesubstance contained per gram. Compute the approximate ageof the relic.What is your opinion as to its authenticity?

Carbon taken

of the time of Christ

y(l) = -3 y(l) = 1
I Y\"2]\"[

10 5.0 10

dx +

1 = 2y,

24. (tan x)dy dx 25. 26. 27.


\037:

= y,

I =\"2]\"[

37. (Continuously

x- dx
dx

= 2x2y, y(l) = 1 = 2xy2 + 3x2y2 , Y ( 1)= -1


y

compounded interest) Upon the birth of their first child, a coupledeposited $5000in an account that pays 8% interest compounded continuously. The interest payments are allowed to accumulate. How much will the account contain on the child's eighteenth birth-

= 6e2x- ,

dy 28. 2..jX

dx

y(O) = 0

38.(Continuously
discoverin

day?

2 = cos y,

29. (a) Find

differential equation Find a singular solution that is not in(b) dy/dx cluded in the general solution. (c)Inspecta sketch of typicalsolution curves to determine the points b) for which the initial value problem y' b has a unique y2, y(a) solution.)

= y2.

y(4) = ]\"[/4 a general solution of the

(a,

39.(Drug elimination)
used to

compounded interest) Supposethat you your attic an overdue library book on which your grandfather owed a fine of30 cents 100years ago.If an overdue fine grows exponentially at a 5% annual rate compounded continuously, how much would you have to pay if you returned the book today?

Supposethat sodium pentobarbital is anesthetize a dog. The dog is anesthetized when its bloodstream contains at least45 milligrams (mg) of sodium pentobarbitol per kilogram of the dog's body)))

44

Chapter
1

First-OrderDifferential Equations) 48. According


amounts to one cosmological theory, there were equal of the two uranium isotopes 235U and 238U at the creation of the universe in the \"big bang.\" At present there are 137.7toms of 238U for eachatom of 235U. Using the a x 8 x 9 half-lives 4.51 10years for 238U and 7.10 10 years for 235U, calculate age of the universe. the F and left to cool 49. A cakeis removed from an oven at which is F. After 30 min the at room temperature, When will it be temperature of the cakeis 50.The amount A (t) of atmospheric pollutants in a certain mountain valley grows naturally and is tripling every 7.5 years.
210\302\260 70\302\260

also that sodium pentobarbitol is elimweight. Suppose inated exponentially from the dog'sbloodstream, with a half-life of 5 h. What single dose should be administered in order to anesthetize a 50-kgdog for 1 h?

40. Thehalf-life of radioactive cobaltis 5.27 years.Suppose


that

a nuclear accidenthas left the level of cobaltradiation in a certain region at 100times the level acceptable for human habitation. How long will it be until the region is again habitable? (Ignore the probablepresence other of

140\302\260F.

100\302\260F?

radioactive

41.Suppose that

isotopes.)
a
mineral

body formed in an ancient formation of the earth originally contained the uranium isotope238U (which has a half-life of x 9 years) but no lead,the end product of the radioactive decayof 238U. If today the ratio of 238U atoms to lead atoms in the mineral body is when did the cataclysm occur?

cataclysm-perhapsthe

itself-

4.51 10

(a) If the initial amount is a formula for A (t) giving


after t

10 pu

(pollutant units), write the amount (in pu) present present

0.9,

years.
to stay in the valley when the how long will has left the surround-

(b)

What will be the amount (in pu) of pollutants in the valley atmosphere after 5 years?
will be dangerous amount of pollutants this take?

42. A certain

sium atom disintegrations yields an argon atom. What is the age of the rock, measured from the time it contained only potassium?

of potassium and argon atoms. Assume that all the argon is the result of radioactive decayof potassium (its half-life x 10 is about 1.28 9 years) and that oneof every nine potas-

moon rock was found to contain equal numbers

(c) If it

reaches100pu,

51.n accidentat a nuclear A


ing

43. A
0\302\260

pitcher of buttermilk initially at 25 C is to be cooled is by setting it on the front porch, where the temperature C. Supposethat the temperature of the buttermilk has dropped to 15 C after 20 min. When will it be at 5 C?
\302\260 \302\260 \302\260

natareapolluted with radioactive material that decays urally. The initial amount of radioactive material present suo is 15su (safeunits), and 5 months later it is still 10 (a) Write a formula giving the amount A(t) of radioactive

power plant

(b)

material (in su) remaining after t months. What amount of radioactive material will remain after

44. When

8 months?
How long-total number of months or fraction thereof-will it be until A = 1 su, so it is safe for peopleto return to the area? 52.Therearenow about 3300different human \"language families\" in the whole world. Assume that all these are de-

in sugar is dissolved water, the amount A that remains undissolved after t minutes satisfies'the differential -kA (k > If 25%of the sugar disequation dAjdt solvesafter 1 min, how long does it take for half of the sugar to dissolve?

(c)

0).

45. The intensity


the

surface? 46. The barometric pressurep (in inches of mercury) at an altitude x miles above sea level satisfies the initial value (a) problem dpjdx= (-0.2)p(O)= 29.92. Calculate p, the barometric pressureat 10,000 and again at 30,000 ft ft. (b) Without prior conditioning, few peoplecan survive when the pressuredrops to lessthan 15 in. of mercury. How high

of light at a depth of x meters below surface of a lake satisfies the differential equation d jdx = (a) At what depth is the intensity half the intensity 10at the surface (where x = O)? (b) What is the intensity at a depth of m (as a fraction of lo)? (c) At what depth will the intensity be 1% of that at the

1 I. (-1.4)

10

into language families every 6 thoudevelops 1.5 sand years. About how long ago was the single original human language spoken? 53.Thousands of years agoancestors f the Native Americans o
family

rived from

a single original

language,

and that

a language

is that?

the crossed Bering Strait from Asia and entered the western hemisphere. Sincethen, they have fanned out across North and South America. The single language that the original Native Americans spokehas sincesplit into many Assume (as in Problem 52) Indian \"language families.\" that the number of theselanguage families has beenmulevery 6000years.Thereare now 150Native tiplied by American language families in the western hemisphere. About when did the ancestorsof today's Native Ameri-

1.5

47. A certain pieceof dubious

information about phenylethylamine in the drinking water began to spreadone day in a Within a week, 10,000 city with a population of peoplehad heard this rumor. Assume that the rate of increaseof the number who have heard the rumor is proportional to the number who have not yet heard it. How long will it be until half the population of the city has heard the rumor?)

cansarrive?

54. A

100,000.

tank is shaped like a vertical cylinder; it initially contains water to a depth of9 ft, and a bottom plug is removed 0 (hours). After 1 h the depth of the water has at time t dropped to 4 ft. How long doesit take for all the water to drain from the tank?

55.Supposethat

the tank of Problem 48 has a radius of 3 ft and that its bottom hole is circular with radius 1 in. How)))

1.4Separable Equationsand Applications


long will it take the water (initially pletely?)

45)

9 ft deep)to drain com-

y)

56.At time t

bottom plug (at the vertex) of a full conis removed. After 1 h the water in the tank is 9 ft deep. hen will the tank be empty?) W
==

0 the

ical water

tank

16 high ft

57. Suppose that


lons of water

tank initially containing Vo galdrains (through a bottom hole) in T minutes. UseTorricelli's to show that the volume of water in law the tank after t < T minutes is V == Vo [1 (t/T)]2.)

a cylindrical

4ft)

58.) A water

tank has the shapeobtained by revolving the curve == X 4/ 3 around the y-axis. A plug at the bottom is rey moved at noon, when the depth of water in the tank is ft. At 1 P.M. the depth of the water is 6 ft. When will the tank be empty?)

or x = g(y))
x)

12

12

t
I)

Water flow

59. A

the shape obtained by revolving the by around the y-axis. The water depth is 4 ft at noon, when a circular plug in the bottom of the tank is removed.At 1 P.M. the depth of the water is 1 ft. (a) Find the depth yet) of water remaining after t hours. (b) When will the tank be empty? (c) If the initial radius of the top surface of the water is 2 ft, what is the radius of the circular holein the bottom?)

parabola x

water tank 2 ==

has

The FIGURE 1.4.10. clepsydra.)

12

65.Just

tim ture

apparent homicide vicat a constant temperaAt noon the temperature of the body is of Assume that the temperature and at 1 P.M.it is F and that it has of the body at the time of death was cooledin accordwith Newton's law. What was the time the body
that

beforemidday
in
70\302\260F.

of an

is found

a room

12

is kept

80\302\260F

75\302\260F.

98.6
\302\260

60. A

cylindrical tank with length 5 ft and radius 3 ft is situated with its axis horizontal. If a circular bottom hole with a radius of 1 in. is openedand the tank is initially half full of xylene, how long will it take for the liquid to drain completely?)

of death? 66.Early onemorning


7 A.M. a
snowplow

it

61. spherical tank A


circular bottom
will

hole with

be required

of radius 4 ft is full of gasolinewhen a radius 1 in. is opened.How long for all the gasoline drain from the tank?) to
full

had traveled 2 miles, but it took two more hours (until lOA.M.) for the snowplow to go an additional 2 miles. (a) Let t == 0 when it began to snow and let x denote the distance traveled by the snowplow at time t. Assuming that the snowplow clearssnow from the road at a constant rate (in cubicfeetper hour, say), show that)

set off to cleara road. By 8 A.M.

began to

snow at

a constant rate. At
it

62. Supposethat
radius

1 m has its flat sideas its bottom. It has a bottom hole of radius 1 cm. If this bottom hole is opened 1P.M., at
when will the tank

an initially

hemispherical water tank

of
where k

dx 1 k-=dt
t)

be empty?)

63. Considerhe initially t ample 8, exceptthat

full hemispherical water tank ofExthe radius r of its circular bottom hole is now unknown. At 1 P.M. the bottom hole is openedand at 1 P.M.the depth of water in the tank is 2 ft. (a) Use 2 Torricelli'slaw in the form dV/dt == ,J2gy (taking constriction into account) to determine when the tank will be empty. (b) What is the radius of the bottom

(Answer: 6 A.M.) 67. A snowplow setsoff at 7 A.M. as in Problem 66. Suppose now that by 8 A. M. it had traveled 4 miles and that by 9 A.M. it had moved an additional 3 miles. What time did

is a constant. (b)

What time did it start snowing?

:30

-(0.6)nr

shows 68.Figure 1.4.11

This is a more difficult snowplow problem because now a transcendental equation must be solved A.M.) numerically to find the value of k. (Answer: 4:27
it start snowing?

hole?)

64. (Theclepsydra,or water clock)A 12-hwater clockis to be designedwith the dimensions shown in Fig. 1.4.10,
shaped like the surface obtained by revolving the curve (x) around the y-axis. What should be this curve, y == and what should be the radius of the circular bottom hole, in orderthat the water level will fall at the constant rate of 4 inches per hour (in./h)?)

a bead sliding down a frictionless wire from point P to point Q. Thebrachistochrone problem asks what shape the wire should be in order to minimize the bead'stime of descentfrom P to Q. In June this of 1696,John Bernoulli proposed problem as a public challenge, ith a 6-month deadline (later extended to w Easter 1697at GeorgeLeibniz'srequest). IsaacNewton, then retired from academiclife and serving as Warden Bernoulli's challengeon of the Mint in London, received The very next day he communicated January 29, t his own solution-the curve of minimal descentime is

1697.

an)))

46

Chapter
1

First-OrderDifferential Equations)
of the cycloidthat is generated by a point on the rim of a circular wheel of radius a as it rolls along the xaxis. [See Example 5 in Section9.4of Edwards and Early Transcendentals,7th edition Penney, Calculus: (Upper SaddleRiver, NJ: Prentice Hall, 2008).] between 69.) Supposea uniform flexible cable is suspended two points (:l::L, at equal heights locatedsymmetriH) Principles cally on either sideof the x-axis (Fig. 1.4.12). of physics can be usedto show that the shapey = y (x) of the hanging cablesatisfies the differential equation)

arc of an inverted cycloid-tothe Royal Societyof London. For a modem derivation of this result, supposethe beadstarts from rest at the origin P and let y = y (x) be the equation of the desiredcurve in a coordinate system with the y-axis pointing downward. Then a mechanical
analogue

of Snell's law

a denotesthe angle of deflection (from the vertiof the tangent line to the curve-socota = y' (x) cal)
where

in

opticsimplies
constant,

that)

sIn

a=

(i))

v)

descended distance y a
mgy

v i (why?)-and = ...j2gys the bead'svelocity

= -PE).)

vertically

(from KE

when it

!mv

has

a-=
d 2y dx
2)

1+

dy 2 dx )

')

p)

T p is the ratio of the cable's where the constant a 0 (where y' (0) 0 tension T at its lowest point x If we substitute ) and its (constant) linear density d 2yjdx 2 in this secondv dymyslashdx, dvjdx orderdifferential equation, we get the first-order equation)

= j

p.

a FIGURE
(a)
First

dv

dx

2 = J1+v.

wire-the brachistochroneproblem.)
derive from

1.4.11. sliding A bead


Eq.(i) the dY =
dx

down

Solve this differential equation for y' (x) = v(x) sinh(xja).Then integrate to get the shapefunction

differential
Y

equation

y(x)=acosh )+C C
of the hanging the Latin word
(ii))
Y)

cable.This curve is calleda catenary, from


for chain.)
(L, H))

J2a
=

y)

where

(b)

appropriate positive constant. 2a sin 2 t, dy Substitute y 4a sin t cost to derive the solution)

a is an

(-L,H)) dt
in

(ii)
Yo)

= a(2t sin 2t), y = a(l cos2t)) (iii)) for which t = Y = 0 when x = O. Finally, the substitution of e = 2a in (iii) yields the standard parametric equations x = ace sin e), y = a(l cose))
x

x)

The FIGURE 1.4.12. catenary.)

E lID) LinearFirst-Order quations


........... .....)

In

Section we saw how 1.4


multiplying

after

both

differential equation by integrating to solve a separable sides by an appropriate factor. For instance, to solve the

equation)

dy

dx

= 2xy

(y

> 0),

(1))

we multiply

both

sidesby the factor l/y to get

-.-=2x; dx
1 dy
y

that

2 is, Dx(lny) = Dx(x

).

(2))))

a Because sideof the equation in (2)is recognizable s a derivative (with respect each
to the independent variable x), all
that

remains are two simpleintegrations, which)

LinearFirst-Order 1.5 Equations


yield In
y

47)

ential equation by p (x, y) yieldsan equation in which each sideis recognizableas a derivative. With the aid of the appropriate integrating factor, there is a standard technique for solving the linear first-order equation)
\037)

= x2 + C.For this reason,the function p(y) = l/y is calledan integratA for the original equation in (1). n integrating factor for a differential ing factor equation is a function p (x,y) such that the multiplication of each sideof the differ-

dy
dx)

+ P(x)y = Q(x)

(3))

on an interval on which the coefficientfunctions P(x)and factor) multiply eachside in Eq. (3)by the integrating
\037)

Q(x)are continuous.We
(4))

p(x) = ef P(x)dx.)
f f ef P(x)dx dy + P(x)e P(x)dx y = Q(x)e P(x)dx.
dx)

The result is)


(5))

Because)
Dx

[IP(x)dx] = P(x),)

the left-hand sideis the derivativeof the producty(x)

alent to)

.ef P(x)dx, so Eq.(5)is equiv-

e Dx [y(x).f P(X)dX]
Integration of both

f = Q(x)e P(x)dx.

sidesof this equation gives)


f y(x)e P(x)dx =

1(Q(x)e
f

P(X)dX)

dx + C.)

Finally,
in

solving for y, we obtain the general solutionof the linearfirst-orderequation

(3):)

y(x) = e-f P(x)dx

[I(Q(x)ef

P(X)dX)

dx +

c].)

(6))

I This formula should not be memorized. n a specific problem it generallyis the simplerto use the method by which we developed formula. That is, in order to solve an equation that can be written in the form in Eq. (3)with the coefficient functions P (x) and Q (x) displayed explicitly, you should attempt to carry out the

following steps.) METHOD: OLUTION S OFFIRST-ORDERQUATIONS E factor p (x) = ef P(x)dx. 2. Then multiply both sidesof the differential equation by p (x). the Next, recognize left-hand side of the resulting equation as the derivative of a product:) Dx [p(x)y(x)]= p(x)Q(x).)))

1.Beginby calculating the integrating

3.

48

Chapter
1

First-OrderDifferential Equations)

4. Finally,

integrate this equation,)

p(x)y(x)=
then tion.)

d f p(x)Q(x) x + C,)

solve for y to obtain the general solutionof the original differentialequa-

Remark Given an initial condition y(xo) = Yo, you can (as usual) substitute x = Xo and y = Yo into the general solution and solve for the value of C yielding the particular solution that satisfies this initial condition. Remark 2:
you
find You

1:

the

integrating

neednot supply explicitly a constant of integration when factor p (x). For if we replace

f
in

P(x)dx

with

P(x)dx + K)

Eq. (4),the result is)


P(x)dx. p(x) = eK+!P(x)dx = eKe!

But the constant factor eK does not affect materially the result of multiplying both sides of the differential equation in (3)by p(x),so we might as well take K = O. You may therefore chooseor f P(x)dx any convenient anti derivative of P(x), without bothering to add a constant of integration.

.)

.-.

.....-.

\"\"

'\"

\037

.....

Example

1)

Solvethe

initial

value problem
11 -y=ge- /3, y(O)=-l.) dx

dy

Solution Here have P(x) = we

e-x / 3, so the integrating = x p(x) = e!(-I)dxe- . x Multiplication of both sidesof the given equation by e- yields)
1 18

-1and Q(x)=

factor is

11 e-xdy e-xy = ge-4x/3 dx which we recognize as)

,)

(7))

d _ (e-Xy)= dx respectto x gives


11

11

e-4x/3.

Hence integration

with

e-Xy =
and multiplication by
eX

8 e-4x/3dx

= _ e-4x/3+ C,
\037\037

gives the general solution

y(x) = Ce
Substitution solution is)

j\037

e-x/3.)
1

(8))

of x

0 and

-1now gives C =
1
j\037

32

' so the desiredparticular


.)

y(x) = 32eX
1

x e-x / 3 = 32 (eX - 33e-/ 3)

.)))

1.5 LinearFirst-OrderEquations
y

49)

= _ 33exp(-x/3)
32)

o)

;:....

-1)

-2 -3
-4)

-1 0
FIGURE
solution

2
x)

y'

x = y + e- /3
1
18

curves for
.)

1.5.1. field and Slope


Example2)

initial condition y(O) = Yo. The two types of behavior are separated by the particular solution y(x) = e-x / 3 for which C = 0 in Eq. (8),so Yo = for the solution curve that is dashedin Fig. If Yo > , then C > 0 in Eq. (8),so the term eX eventually dominates the behavior of y(x), and hence as y(x) +00 x +00.But if Yo < , then C < 0,so both terms in y(x) are negative and therefore y(x) -00as x +00. hus the initial condition Yo = is critical T in the sense that solutions that start above on the y-axis grow in the positive direction, while solutions that start lower than grow in the negativedirectionas x +00. he interpretation of a mathematicalmodel often hinges on finding such T a critical condition that separates kind of behaviorof a solution from a different one kind of behavior.

Remark:Figure 1.5.1 a slopefield and typical solution curves for shows Note that somesoluEq. (7),including the one passingthrough the point (0, tions grow rapidly in the positive direction as x increases, while others grow rapidly in the negative direction.The behavior of a given solution curve is determinedby

-1).

its

\037

\037\037

1.5.1.

\037\037

\037\037

\037

\037\037

\037

\037

\037\037

\037\037

\037\037

\037

.)

Find a general solution of)

(x2 + Solution
After

+ 1)dy
dx)

3xy

= 6x.

(9))

division of both sidesof the equation by x2 + 1, recognize result) we the

-+x + y= x dx
dy

3x

6x
1)

2+

2 2 as a first-orderlinear equation with P(x) = 3x/(x + 1)and Q(x) = 6x/(x +

Multiplication by)

p(x) = exp
yields)

(I

x23

:
2

1).

1 dX)

= exp (\037ln(x2 + 1)) = (x2 + 1)3/2)

(x2 + 1)3/2dx)
7

dy

2 2 + 3x(x + 1)1/2y= 6x(x + 1)1/2, 2 + 1)3/2 ] = 6x(x + 1)1/2.) y

and

thus)

6 5 4
3
;::....

Dx [(x Integration then yields)

2
1

(x2 + 1)3/2y=

-1 -2 -3 -5-4-3-2-1 0
x)

2 d 6x(x + 1)1/2 x = 2(x2 + 1)3/2+ c.)

2 -3/2 gives the general solution) Multiplication of both sidesby (x + 1) 2


3

4 5

2 y(x) = 2 + C(x + 1)-3/2.)

(10)

FIGURE

solution curves for the differential equation in Eq.(9).)

1.5.2. field and Slope


Eq. (9). Note that, as x
\037

.)

Remark:Figure 1.5.2 a slopefield and typical solution curves for shows a +00, ll other solution curves approach the constant solution curve y(x) = 2 that corresponds C = 0 in Eq. (10). his constant))) to T

50

Chapter
1

First-OrderDifferential Equations)
solution can be described an equilibrium solution of the differential equation, beas all x (and thus the value of the solution remains forever where it starts). Moregenerally, the word \"equilibrium\" connotes so by an equilibrium solution of a differential equation is meant a \"unchanging,\" constant solution y(x) = c, for which it follows that y'(x) = O. Note that substitution of y' = 0 in the differential equation (9)yields3xy = 6x,so it follows that we seethat y (x) = 2 is the only equilibrium solutionof this y = 2 if x i= O. Hence differential equation, as seemsisually obvious in Fig. v

causey(O) = 2 impliesthat y(x) = 2 for

1.5.2.

.)

Closer ookat the Method) L

The preceding derivation of the solution in Eq. (6)of the linear first-order equation e + Py = Q bearscloserxamination. Supposethat the coefficientfunctions P(x) y' Then the and Q (x) are continuous on the (possiblyunbounded) open interval antiderivatives f dx P(x)dx and

I.

I, then y(x) is given by the formula in Eq. (6)for somechoiceof the constant t C.Conversely,you may verify by direct substitution (Problem 31)hat the function in Eq. (6) satisfies Eq. (3). Finally, given a point Xo of I and any numy(x) given ber Yo, there is-as unique value of C such that y(xo) = Yo. previously noted-a
on
theorem.) we Consequently, have proved the following existence-uniqueness

(Q(x)e P(X)dX) existon I. Our derivationof Eq. (6)showsthat if y = y (x) is a solution of Eq. (3)

1 THEOREM TheLinearFirst-Order Equation


If the functions P (x) and Q (x) are continuous on the openinterval the point Xo, then the initial value problem)
\037)

I containing
(11))

dx + P(x)y

dy

= Q(x), y(xo) =

Yo)

has a unique solution y(x) on priate value of c.)

I, given by the formula

in

Eq. (6)with an appro-

Remark Theorem 1 gives a solutionon the entire interval I for a linear differential equation, in contrast with Theorem 1 of Section which guarantees only a solution on a possiblysmaller interval.

1:

1.3,

Remark 2: Remark 3:
to solve the
wri ting) initial

the general solution given in has no singular solutions.

Theorem 1 tellsus that every solution ofEq. 3)is included in ( Eq. (6). Thus a linearfirst-orderdifferential equation

n The appropriate value of the constant C in Eq. (6)-aseeded be selected value problemin Eq. (1 I)-can \"automatically\" by

p(x) = exp
y(x) =
1

P(t) (1: dt) 1 p(t)Q(t) (x) [


,
x
Yo

(12))))

dt

XQ)

LinearFirst-OrderEquations 1.5

51)

The indicated limits Xo and x effect a choicef indefinite integrals in Eq. (6) that o guarantees in advance that p(xo)= 1 and that y(xo) = Yo (asyou can verify directly
by
substituting

x = Xo in Eqs. (12)).

.)

Example3)

Solvethe initial value problem)

. x2dy + xY = sIn x, dx)

( 1)

= Yo.

(13))

Solution Division by x2 gives the linear first-order equation)

dy 1

= dx + -y x X

sinx
2)

with
IS)

2 P(x) = Ijx and Q(x) = (sinx)jx . With

Xo

1 the

integrating

factor in (12)

p(x) = exp

= = x, (IX dt) exp(lnx)


\037

so the desiredparticular solution is given by) y(x) = x


In

Yo

sin t dt + [X

11

t accordwith Theorem 1,his solution is definedon the whole positivex-axis..)

\302\267)

(14))

rule, given x) needto be approximatednumerically-using Simpson's for instanceto find the value y (x) of the solution at x. In this case, owever, we have the sine h integral function) x sin t

Comment:n general,n integral such as the one in Eq. (14)would (for I a


Si(x)=

3
2)

appearswith sufficient frequency in applications that its values have been tabulated. A good set of tables of specialfunctions is Abramowitz and Stegun, Then Handbook Mathematical Functions (New York: Dover, 1965). the particuof lar solution in Eq. (14)reducesto)
which

l
o

dt,

;::....

-1 -2
-3
0 (1,-3) 5 10
x)

y(x)=- [Yo+ x
15 20

. . l -dt-1 -dt] =-[Yo+SI(x)-SI(I)].


x

sin t

FIGURE

1.5.3. solution Typical


Eq.(15).)

curves defined by

and The sineintegral function is available in most scientificcomputing systems can showsa be used to plot typical solution curves defined by Eq. (15). 1.5.3 Figure selectionof solution curves with initial values y (1) = Yo ranging from Yo = -3to a 0 as x +00,nd this is Yo = 3.It appearsthat on each solution curve, y(x) in fact true because sine integral function is bounded. the
\037 \037

tot
1

sin t

x)

(15))

than is the exception-rather the rule-whena in solutionof a differentialequation can be expressedterms of elementaryfunctions. We will study various devicesfor obtaining good approximations to the valuesof numerical the nonelementary functions we encounter. In Chapter 6 we will discuss integration of differential equations in somedetail.)))

In the

sequelwe will seethat

it

52

Chapter
1

First-OrderDifferential Equations)

MixtureProblems)
As a first application of linear first-order equations, we consider tank containing a a mixture of solute and solvent-such salt dissolved water. There is as in both inflow and outflow, and we want to compute the amount x(t) of solute in the tank at time t, given the amount x(0) = Xa at time t = O. Supposethat solution with a concentration of Ci grams of solute per liter of solution flows into the tank at the constant rate of ri liters per second, nd that the solution in the tank-kept a mixedby stirring-flows ut at the constant rate of r 0 liters per second. o thoroughly To set up a differential equation for x(t), we estimate the change in x \037t]. The amount of solute that flows into the during the brief time interval [t, t tank during secondsis riCi grams. To checkthis, note how the cancellation of dimensions checks computations:) our

solution-a

\037x

Input:

Yi

LIs,

Ci

g/L)

\037t

\037t

Amount x(t) Volume V(t)


Concentration

liters
co(t) =
\037)

ri

second

grams (Ci lIter) ) ) .

(\037t

seconds)

Output:
Yo

Lis,
g/L)

Co

FIGURE1.5.4. single-tank The


mixture problem.)

The amount of solute that flows out of the tank during the sametime interval dependson the concentration co(t)of solute in the solution at time t. But as noted in co(t) Fig.1.5.4, = x(t)/V(t),where Vet) denotesthe volume (not constant unless
ri

yieldsa quantity measuredin grams.

= r 0) of solution in the tank at time t. Then) = {gramsinput} {gramsoutput}


\037x

We now

divide by

\037t:)

\037x \037t \037

\037

rici

\037t

- roc

o \037t.

\037

rici

- roc .
o

Finally,

we take the limit as if all the functions involved are continuous and x(t) is differentiable, then the error in this approximationalso approaches ero, z and we obtain the differential equation
\037t

0;

\037)

dt)

dx =

rici

in

which

ri, Ci,and r 0 are constants, but

Co denotesthe

co(t)=
of solute in the tank at time the differential equation)

t. Thus the amount x(t) of solute in

x(t)
Vet))

- roc

o,

(16)) variable concentration)


(17)) the
tank

If Va = V(O), then differential equation for the amount

dt Vet) = Va + (ri

satisfies

dx =

rici

x(t)

- ro)t, so Eq. (18)is a linear first-order


of solute in the
tank

- -x.
ro
V

(18))))

at time

t.

over a short time interval [t, t + \037t]-that you should strive to understand, because it is a very useful tool for obtaining all sortsof differential equations.
in deriving But any other consistent ystem of units can be used to measure amounts Eq. s of solute and volumes of solution. In the following examplewe measure both in

cesswe used to

Important: Equation (18)neednot be committed to memory. It is the proobtain that equation-examination the behavior of the system of

(18).

Remark: It was convenientfor us to use g/L mass/volumeunits

cubickilometers.

.)

1.5 LinearFirst-Order Equations


Example4)

53)

Assumethat LakeErie has a volume of 480 kIn 3 and that its rate of inflow (from kIn that LakeHuron) and outflow (to LakeOntario) are both 350 3 per year. Suppose at the time t == 0 (years),the pollutant concentrationof LakeErie-caused past by five industrial pollution that has now beenorderedto cease-is times that of Lake Huron.If the outflow henceforth is perfectly mixed lakewater, how long will it take to reducethe pollution concentration in LakeErieto twice that of LakeHuron?)

Solution Here have) we


v == 480 (km 3),
ri == r 0 == r ==

350(kIn 3/yr),
concentrationof LakeHuron), and

Ci == C (the pollutant V,) xo == x(O)== 5c

and the question is this: When

is x(t) == 2cV? With

separable equation)

which we rewrite in the linear first-order form)

dt

this

notation,

Eq. (18)is the


(19))

dx ==

rc

- -x
r
V

')

dx
dt)

+ px == q

(20))

constant coefficients p == r /V, q == rc, and integrating factor p == ePt You The can either solve this equation directly or apply the formula in (12). latter gives)
with

x(t) = e-pt

[xo

it qe
+
.)

Pt

dt]

= e-pt

[xo
;

= e-rt / V
x(t) == cV

[SCV

(e /
\037\037

rt v

- 1)]
In

;
4

(ePt

-1)]
(21))

+ 4cV e-rt/V

To find when x(t) == 2cV, e therefore needonly solve the equation) w


rt cV + 4cVe- / V == 2cV)

for

t ==

-;

4 ==

480

350

In

\037

1.901 (years).) .)

Example5)

in 120-gallon tank initially contains 90 lb of salt dissolved 90 gal of water. (gal) Brinecontaining 2 lb / gal of salt flows into the tank at the rate of 4 gal/min, and the well-stirred mixture flows out of the tank at the rate of 3 gal/min.How much salt doesthe tank contain when it is full?)

Solution The interesting feature of this exampleis that, due to the differing rates of inflow and outflow, the volume of brine in the tank increasesteadily with V (t) == 90 + t s in the amount x of salt in the tank from time t to time gallons.The change
\037x

\037t

(minutes)

is given by)

\037X\037(4)(2)\037t-3

(90+ )
t)))

\037t,

54

Chapter
1

First-OrderDifferential Equations) so our differential equation is)

dx
dt

3 90 +
t)

x = 8.

An integrating

factor is)

p(x) = exp
which gives) Dt

(I

3 90 +
t)

dt = e31n(90+t) = (90+ t)3,

+ + [(90 t)3x]= 8(90 t)3; = 2(90 t)4 + c.) + (90+ t)3x

Substitution of x(0) = time t is)

The tank is full

90gives C = (90)4,so the amount of salt in 4 90 . + x(t) = 2(90 t) (90+ t)3 w after 30 min, and when t = 30, e have) 4 90 202 x(30)= 2(90 30) + (lb) 3 120)

the

tank

at

-_

\037

_
Find

of salt in the

tank.)

.)

Problems) 21. y' = 3y + x 4 2 y2(2n) = 0 cosx, x


22.y' = 2xy + 3x exp(x ), y(O) = 5 23.xy' + (2x - 3)y = 4x4
24. (x2 + 4)y'+ 3xy = x,y(O) = 1
Solvethe
regarding

3x = 2xe= eX2 2xy 5. xy' + 2y = 3x, y ( 1)= 5 6.xy' + 5y = 7x2, y(2) = 5 7. 2xy' + y = 10,JX 8.3xy' + y = 12x 9.xy' y = x,y(l) = 7 10. xy' 3y = 9x3 2 11. + y = 3xy,5 y(l) = 0 xy' 12. y' + 3y = 2x , y(2) = 1 x 13. ' + y = eX, y(O) = 1 y 14. y' 3y = x 3, y(l) = 10 x 15. ' + 2xy = x,y(O) = -2 y 16. ' = (1- )cosx,y(n) = 2 y y 17.(1 +x)y' + y = cosx, (O) = 1 y 18. y' = 2y + x 3 cosx x 19. ' + y cotx = cosx y 20. ' = 1 + x + y + xy, y(O) = 0) y

generalsolutions of the differential equations in Prob25. If an initial condition is given, find the corresponding particular solution. Throughout, primes denote derivatives with respect x.) to y' + y = 2, y (0) = 0 2.y' 2y = 3e2x , y(O) = 0
lems 1 through

3.y' + 4. y' -

1.

dy 25.(x2+1)-+3x3y=6xexp(-\037x2),y(0)=1 dx) differential equations y as the independent


in Problems 26 through 28 by variable rather than x.)

3y

dy 26.(1- 4xy2)dx = y3 dy 28.(1+ 2xy) dx = 1 + y2 29.Expressthe general solution

27. (x + ye
of dy/dx 2
(X

)-=
dy dx)

--

of the

function) error

= 1 + 2xy in terms

erf(x) =

.j7r10
initial

e-t2 dt.)
value problem)

30.Expressthe solution

of the
y

2x-=
dy dx)

+ 2x cosx, y(l) = 0
section.)))

as an

integral

as in Example 3 of this

LinearFirst-Order 1.5 Equations


Problems and 32 illustrate-for the special aseoffirstc order linear equations-techniquesthat will be important when we study higher-orderinear equations in Chapter 3.) l

55)

31

time t. Show first

that

dy and then part

31.a) Show (
that)

that)

J yc(x) = Ce- P(x)dx

solvefor (a). (c) Finally,


tank 2.)

dt y(t),

--5x

100

200'

5y

find

using the function x(t) found in the maximum amount of salt

is a general solution of dy/dx

+ P(x)y = O. (b) Show


f

ever in

Yp(x)

= e-f P(x)dx

Tank 1

[I(Q(x)e

P(X)dX)

dx

Volume

VI
x)

Amount

32.

33. 100

long will it be until only 10 of salt remains in the tank? kg 34. Consider reservoir with a volume of 8 billion cubicfeet a There (ft3) and an initial pollutant concentration of 0.25%. is a daily inflow of 500 million ft 3 of water with a pollutant concentration of 0.05%and an equal daily outflow of the well-mixed water in the reservoir. How long will it take to reducethe pollutant concentration in the reservoir to

S dy/dx + P(x)y = Q(x). how that y(x) = yc(x) + yp(x) is a general solution of dy/dx + P(x)y = Q(x). (a) Find constants A and B such that y p (x) = A sin x + B cosx a solution of dy/dx + y = 2 sinx. (b) Usethe is result of part (a) and the method of Problem to find the general solution of dy/dx + y = 2 sinx. (c) Solvethe initial value problem dy/dx + y = 2 sinx,y(O) = A tank contains 1000liters (L) of a solution consisting of in kg of salt dissolved water. Pure water is pumped into the tank at the rate of 5 L/s, and the mixture-kept uniform by stirring- pumped out at the samerate. How is

particular solution of dy/dx + P(x)y = Q(x). that (c) Suppose yc(x) is any general solution of dy/dx + P (x) y = 0 and that y p (x) is any particular solution of

is a

Tank

2
V2
y)

Volume

Amount

31

f)

1.

FIGURE

A 1.5.5.cascadeof two tanks.

tank 1 in the cascadeshown in Fig. contains of pure ethanol and tank 2 iniinitially gal tially contains 100 gal of pure water. Pure water flows into tank 1 at gal/min, and the other two flow rates arealso gal/min. (a) Find the amounts x(t) and y(t) of ethanol in the two tanks at time t > O. (b) Find the maximum amount of ethanol ever in tank 2. 40. A multiple cascadeis shown in Fig.

39.Supposethat

10

100 10

1.5.5,

1.5.6.)

0.10%?

35. Rework Example 4 for the caseof Lake Ontario, which River and receivesinflow empties into the St. Lawrence from LakeErie (via the Niagara River). The only differencesare that this lake has a volume of 1640Ian 3 and an inflow-outflow rate of 410 3/year. Ian 36. A tank initially contains 60 gal of pure water. Brine 1 lb of salt per gallon enters the tank at containing 2 gal/min, and the (perfectly mixed) solution leavesthe

tank at 3 gal/min; thus the tank is empty after exactly 1 h. (a) Find the amount of salt in the tank after t minutes. (b) What is the maximum amount of salt ever in the tank? 37. A 400-galtank initially contains gal of brine containing 50 lb of salt. Brine containing 1 lb of salt per gallon enters the tank at the rate of 5 galls,and the well-mixed brine in the tank flows out at the rate of 3 gal/so How much salt will the tank contain when it is full of brine? 38. Considerthe cascade two tanks shown in Fig. of with VI and V2 200 (gal) the volumes of (gal) brine in the two tanks. Each tank also initially contains 50 lb of salt. The three flow rates indicated in the figure areeach5 gal/min, with pure water flowing into tank of salt in tank 1 at time t. (a) Find the amount that (b) Suppose y(t) is the amount of salt in tank 2 at)

100

= 100

1.5.5,

\037)

1.

x(t)

A multiple cascade. FIGURE = 0, tank 0 contains 1 gal of ethanol and 1 gal At time t of water; all the remaining tanks contain 2 gal of pure water each. Pure water is pumped into tank 0 at 1 gal/min,)))

1.5.6.

56

Chapter
1

First-OrderDifferential Equations) 44. Figure


and typical solution (a) Show that every 1 t solution curve approacheshe straight line y as x \037 For eachof the five values YI (b) determine the initial value Yo (accurateto 0, 5, and five decimal places)such that y(5) YI for the solution Yo.) satisfying the initial condition y(

and the varying mixture in each tank is pumped into the one below it at the samerate. Assume, as usual, that the mixtures are kept perfectly uniform by stirring. Let Xn (t) denote the amount of ethanol in tank n at time t. t /2 (a) Show that xo(t) (b) Show by induction on n

1.5.8 a shows
the equation

curves for

slope field

y'

= x + y.

= etn

-5,

-00.

10,
10 8 6 4 2 0

= -x = -10,

that)

xn(t)

t/2 = en! 2
n

for n

> O.

= -5)=

A 41.

(c) Show that the maximum value of Xn (t) for n > 0 is Mn = xn (2n) = nne-njnL (d) Conclude from Stirling's n approximation n! nne- -J2nn that Mn (2nn)-1/2.
\037 \037 :>-.

woman accepts an engineering 30-year-old position with a starting salary of $30,000per year. Her salary S(t) increasesexponentially, with S(t) = 30et /20 thousand dollars after t years. Meanwhile, 12%of her salary is deposited continuously in a retirement account, which accumulates interest at a continuous annual rate of 6%. in terms of to derive the differential (a) Estimate
\037A \037t

-2 -4 -6 -8 -10
-5) o
x) 5)

equation satisfied by the amount A (t) in her retirement account after t years. (b) Compute A (40), the amount available for her retirement at age 42. Suppose a falling hailstone with density 8 1 starts that from rest with negligible radius r O. Thereafter its radius is r kt (k is a constant) as it grows by accretion to during its fall. UseNewton's secondlaw-according which the net force F acting on a possibly variable mass m equals the time rate of change dpjdt of its momentum set up and solvethe initial value problem p m

70.

FIGURE

curves for y'

1.5.8. field and solution = Slope


x

+ y.)

= v-to

d
dt

(mv)

= mg, =

v(O) = 0,

where m is the variable mass of the hailstone, v dyjdt is its velocity, and the positive y-axis points downward. Then show that dvjdt gj4. Thus the hailstone falls as though it were under one-fourth the influence of gravity. 43. Figure shows a slope field and typical solution curves for the equation y' x y.)

1.5.7

shallow reservoirthat has a one square kilometer water surfaceand an averagewater depth of 2 meters. Initially it is filled with fresh water, but at time t = 0 water contaminated with a liquid pollutant begins flowing into the reservoirat the rate of 200 thousand cubic meters per month. The well-mixed water in the reservoirlows f out at the samerate. Your first task is to find the amount x (t) of after t months.) pollutant (in millions of liters) in the reservoir 45.) The incoming water has a pollutant concentration of liters per cubic meter (L/m 3). Verify that c(t) = 10 the steadily rising curve in the graph of x (t) resembles which a Fig. 1.5.9, approachessymptotically the graph of to the equilibrium solution x (t) = 20 that corresponds the reservoir'slong-term pollutant content. How long doesit

Problems 45 and 46 deal with a

:>-.

-2 -4 -6 -8 -10
-5) o
x) 5)

10 8 6 4 2 0

take the pollutant concentration 5 L/m 3?)


25 20)
15) 10)
5) x)

in the

reservoir to reach

x=20)

\037 Problem \037Problem 45)

46

10 20 30 40 50 60

curves for

FIGURE1.5.7. field Slope


y' = x

- y.

and solution

FIGURE Graphs of solutions Problems 45 and 46.)

1.5.9.

in

(a) Show that every solution curve approaches the x 1 as x \037 (b) For each straight line y of the five values Yl and 4.000, to 4.002,determine the initial value Yo (accurate four decimal places) such that y(5) Yl for the solution satisfying the initial condition y( Yo.)

+00. = 3.998, 3.999, 4.001,

46. The incoming 1O(1 + cost)


tion

-5)=

of slightly over 6* months. Doesit seempredictable oscillate that the lake's polutant content should ultimately about an average level of 20 million liters?))) periodically

L/m 3 that varies between 0 and 20, with an concentration of L/m 3 and a period of oscillaaverage

water has pollutant

concentration

c(t)

10

1.5 LinearFirst-Order Equations


Verify

57)

oscillatory

that the graph of x (t) does,indeed,resemble the curve shown in Fig. How long does

1.5.9.

1.5 Application) Indoo\037

For an interesting appliedproblemthat involves the solution of a linear differential equation, considerindoor temperature oscillations that are driven by outdoor o oscillations f the form) temperature
A(t)

it)

take the pollutant concentration 5 L/m 3?)

in the

reservoir to reach

\037\037l:\037\037rat\037re

_9scill \037otions)

= ao + al coswt + b i sinwt.)

(1))

If w = 7T /12, theseoscillations then have a periodof 24 hours (so that the cycle of outdoor temperatures repeatsitself daily) and Eq. (1)provides a realisticmodel for the temperature outsidea houseon a day when no change in the overall day-to-day weather pattern is occurring. or instance, for a typical July day in Athens, GA F with a minimum temperature of when t = 4 (4 A.M.) and a maximum of when t = 16 P.M.), e would take) w (4
70\302\260F 90\302\260F

A(t)
We

= 80 10cosw(t 4) = 80

sin sin,B to derived Eq. (2)by using the identity (3) f3 = 80, l = and b i = in Eq. (1). a get ao If we write Newton'slaw of cooling (Eq. (3) of Section for the correindoor temperature u(t) at time t, but with the outside temperature A(t) sponding given by Eq. (1) instead of a constant ambient temperature A, we get the linear first-orderdifferential equation)

-5,

-5,J3

- 5 coswt - 5,J3sinwt. a cos(a- = cos cos + a


1.1)

(2))

du dt
that is,)

= -k(u

A(t));)

du
dt)

. + ku = k(ao + al coswt + b I slnwt)

(3))

coefficient functions P(t) = k and Q(t) = kA(t). Typical values of the to proportionality constant k range from 0.2 0.5 (although k might be greater than 0.5 a poorly insulated building with openwindows, or less than 0.2 a wellfor for insulated building with tightly sealedwindows).)
with

therefore want to investigate the resulting indoor temperatures that we must endure for the next several days. Beginyour investigation by solving Eq. (3)with the initial condition u(O) == Uo (the indoor temperature at the time of the failure of the air conditioner). You a o may want to use the integral formulas in 49 and 50of the endpapers, r possibly computer algebra system.You should get the solution)
kt wt u(t) = ao + coe- + CI cos + dl sin wt,)
(4))))

SCENARIO: Supposethat our air conditioner fails at time to = 0 one midnight, and we cannot afford to have it repaireduntil payday at the end of the month. We

58

Chapter
1

First-OrderDifferential Equations)
where)

Co = CI =)

Uo

- - k2al + k2al d _
aO
k2
W 2)

kwh l

kwh l
2)

k2

+W

kwat k2

+ k2bt +W
2)

with w With

t (for instance),his solution reducesapproximately)to) (

= n/12. and ao = 80, l = -5, I = -5,J3 in Eq. (2)),w = n/12, k = 0.2 a h (as

100 95 = 95 Uo 90
I I
1

u(t) = 80 + e-t / 5 (uo

- 82.3351)(2.3351) nt - (5.6036)nt cos + 12


sin

-.
12)

(5)

I I

,-..85
OI)
\037

I I I

'-\"
\037

80
75

zero Observefirst that the \"damped\" exponential term in Eq. (5)approaches as t --* +00,eaving the long-term \"steady periodic\"solution) l
I I I I I I

I I I

70

65 60 0

= 65 Uo

: = It 12 20
t (h))

It

= 36

usp(t)

= 80 + (2.3351) nt cos

--

sin-. 12 (5.6036) 12)

nt

(6))

10

30

40

1.5.10. = curves Solution Uo 65,68, given Eq.(5) 71,...,92,95.)


FIGURE
by with

Consequently, the long-term indoor temperatures oscillateevery 24 hours around the sameaverage temperature as the average outdoor temperature. to showsa number of solution curves corresponding possible Figure initial temperatures Uo ranging from Observethat-whatever the to hours initial temperature-the indoor temperature \"settlesdown\" within about 18 to a periodicdaily oscillation. ut the amplitude of temperature variation is less B indoorsthan outdoors.Indeed,using the trigonometric identity mentioned earlier, Eq. (6)can be rewritten (verify this!)as)

1.5.10

80\302\260F

65\302\260F

95\302\260F.

100 95 90
85)
,,-.....

u(t) = 80

OI)
\037

- (6.0707) ( - 1.9656) cos = 80 - (6.0707) -(t- 7.5082). cos


7\037

'-\"
\037

80
75 70
I ,\\1

12)

(7))

Indoor
temperature
I

Do you seethat
of about
74\302\260

65 60
0)

It

10

20

t (h)

30

= 36 40

FIGURE

1.5.11. Comparison of

indoor and outdoor temperature oscillations.)

this impliesthat the indoor temperature varies between a minimum F and a maximum of about 86 F? of Finally, comparison Eqs. (2) and (7) indicates that the indoor temperature 4 hours, as illustrated lagsbehind the outdoor temperature by about in Fig. Thus the temperature insidethe housecontinues to rise until about 7:30 each evening, so the hottest part of the day insideis early evening rather P.M. than late afternoon (as outside). For a personalproblemto investigate, carry out a similar analysis using ava erageJuly daily maximum/minimum figures for your own localend a value of k appropriate to your own home. You might also considera winter day instead of a summer day. (What is the winter-summerdifference for the indoor temperature the problem?)You may wish to explore use of available technology both to solve the differential equation and to graph its solutionfor the indoor temperaturein comparisonwith the outdoor temperature.)))
\302\260

1.5.11.

7.5082

\037

3.5

and ExactEquations 1.6 Substitution Methods

59)

S Methodsand ExactEq\037ations) 11IIIubstitution


The first-orderdifferential equations we have solved in the previous sections have all beeneither separable linear. But many applications involve differential equaor tions that are neither separable linear. In this sectionwe illustrate (mainly with nor substitution methods that sometimes be used to transform a given can examples) differential equation into one that we already know how to solve. For instance, the differential equation)

dy
dx)

= f(x,y),

(1))

with

v dependent ariable y and independent variable x, may contain a conspicuous

combination)

v=a(x,y))
of x and
equation)
y that

(2))

suggestsitself as a new independentvariable v. Thus the differential

dx

dy = + + (x y 3)2
in

practically demandsthe substitution v = x + y + 3 of the form If the substitution relation in Eq. (2)can be solved for)
y

Eq. (2).
(3))
function

= f3 (x,v),)
of

then

v application of the chain rule-regarding as an (unknown)

x(4))

yields)

dy

dx

dv dx af3 dv = + av dx f3x + f3v dx axdx


af3

')

where the partial derivatives af3jax = f3x(x, v) and af3jav = f3v(x, v) are known functions of x and v. If we substitute the right-hand sidein (4) for dyjdx in Eq. (1) and then solve for dvjdx, the result is a new differential equation of the form)

dv
dx)

= g(x,v)

(5))

or new dependent variable v. If this new equation is either separable linear, sectionsto solve it. we can apply the methods of preceding If v = v (x) is a solution of Eq. (5),then y = f3 (x,v (x))will be a solution of such that the transformed the original Eq. (1). he trick; is to select substitution a T is one we can solve. Even when possible,this is not always easy; it may Eq. (5) require a fair amount of ingenuity or trial and error.)
with then

Example

1)

Solvethe differential equation)

dy
dx)))

= (x + y + 3)2.

60

Chapter
1

First-OrderDifferential Equations)
As indicated earlier,
v Then)

Solution

let's try

the

substitution) that

= x + y + 3;)
dy

is,

=v x

--

3.)

so the transformed equation is)

- -dx
dv

dv

dx)

1,

dx)

= 1 + v 2.
in

This is a separable equation, and we have no difficulty


10 8 6 4 2 0

obtaining its

solution)

x=
So v =
tan(x

dv

equation dy/dx = (x + y + 3)2 is x + y + 3 = tan(x

- C).Because = x +
v

l+v) 2
y

= tan-

+ C.

+ 3,the general solution of the original

- C);

that is,)

\037

-2 -4 -6 -8 -10
-5) o
x) 5)

y(x) = tan(x

- C) - x seethat,

3.)

.)
the function

FIGURE

solution curves

1.6.1. field and Slope


for

y' = (x + y

+ 3)2.)

(x+y +3)2 is continuously differentiablefor all x and y, each solution is continuous the only on a boundedinterval. In particular, because tangent function is continuous on the open interval (-T[2,T[ 2), the particular solution with arbitrary constant / / value C is continuouson the interval where -T[2 < x C < T[ 2; is, C T[ 2 < / / that / x < C + T[/2.This situation is fairly typical of nonlinear differential equations, in contrast with linear differential equations, whosesolutionsare continuouswherever

the differential equation of Example We

Remark: Figure 1.6.1 a slopefield and typical solution curves for shows

1.

although

f (x,y) =

the coefficientfunctions in the equation are continuous. Example1 illustrates the fact that any differential equation of the form)

dy
dx)

= F(ax+by+c)

(6))

can be transformed into a separableequation by use of the substitution v = ax + o by + c (seeProblem 55). The paragraphs that follow deal with other classesf first-order equations for which there are standard substitutions that are known to

succeed.)

Homogeneous Equations
A homogeneous first-order differential equation is one that form)

can be written

in

the

dy
\037)

dx
substitutions)

=F

(x) ) .
Y

(7))

If we make the

-,
y

x)

= vx,

-=v+x-, dx
dy dv
dx)

(8))))

1.6 Substitution Methods and ExactEquations


then

61)

Eq. (7)is transformed into the separable equation

x- = F(v) dv
dx)

v.

Thus every homogeneousirst-orderdifferential equation can bereduced an inteto f means of the substitutions in (8). gration problemby

Remark:A dictionary definition of \"homogeneous\"s \"of a similar i or nature.\" Considera differential equation of the form)
dy Axmyn_ = BxPyq+ Cxrys dx)

kind

(*))

in whosepolynomial coefficient functions are \"homogeneous\" the sensethat each of their terms has the same total degree, + n = p + q = r + s = K. If we m divide each side of (*) by xK, then the result-because yn Ix +n = (yIx)n,and xm m

so forth-is equation) the

- dx _ - ( ) (x) --B(x) +Cy n

dy

x)

(by another division) in the form of Eq. (7).More a differential equation of the form P (x,y) y/ = Q (x,y) with polynomial generally, coefficients P and Q is homogeneousf the terms in thesepolynomialsall have the i sametotal degree The differential equation in the followingexample is of this K. form with K = 2.)

which evidently can be written

Example2)

Solvethe differential equation)


dy 2xy- = 4x2 + 3 2. dx)
y

Solution This equation is neither separablenor linear, but we recognize as a homogeneous it


equation by
writing it in

the form)

dy
dx)

4x2 +3y2
2xy

=2

()
y

Y + 2 (x) ) .

The substitutions
y

in

(8)then take the form)

= vx,)

-=v+x-, = -, dx x
dy dv
dx) v
y

and

1
v)

x)
y)

These yield)
and hence)

2 3 dv v+x-=-+-v, dx
v
2)

dx
dv dv

+2 v

v2

+4.
')

2v

+4 f x) 2 In(v + 4) = In Ixl + In

2v 2 v

1 - dx;

C.)))

62
6 4 2
\037

Chapter
1

First-OrderDifferential Equations)
We

apply the exponential function to both sidesof the last equation to obtain)
V

+ 4 == Clxl;

y2 -+4==Clxl; 2
X

-2
-4) o
x)

y2

+ 4x2 == kx3.

6
and

FIGURE1.6.2. field Slope


solution curves for 4x2 + 3y 2.) 2xyy'

Note that the left-hand side of this equation is necessarily nonnegative. It follows that k > 0 in the case solutions that are defined for x > 0,while k < 0 for of solutions where x < O. Indeed,the family of solutioncurves illustrated in Fig.1.6.2 exhibitssymmetry about both coordinate axes.Actuall y, there are positive-valued and negative-valued solutions of the forms y(x) == .vk x3 4x2 that are defined for x > 4/k if the constant k is positive, and for x < 4/ k if k is negative.
-::1::.

.)

Example3)

Solvethe initial value problem)

x dy = y + x2 dx
where Xo

J x+
Y

y2,

y(xo) = 0,)

> O.)
x and find dx
dy =
that)

Solution

We

divide both sidesby

)1-( ) 2,
Y

x)

so we make the substitutions

in

(8);we get)
v

+x
v2

\037:

=v+

J1

v2 ;

f
50 40 30 20 10 0
I

.vI

dv ==
1

\037

x)

dx;

sin- v == In x + C.) We need not write In Ix because > 0 near x == Xo > O. Now note that x == 0,so C == sin-0 -lnxo -lnxo. ence) == H y(xo)/xo
1

v(xo)

y v == == sin (In x

In

xo) == sin

( ),
\037

In

Xo)

\037

-10
-20 -30 -40 -50

and therefore)

y(x) = x sin

is the desiredparticular solution. Figure 1.6.3 showssometypical solution curves. 10 20 30 40 50 Because the radical in the differential of equation, thesesolutioncurvesare confined to the indicated triangular region x > Iyl. You can checkthat the boundary lines FIGURE1.6.3. on curves Soluti of y == x and y == -x (for x > 0) are singular solutioncurves that consist points of for xy' = y + x 2 y2.) .))) with the solution curves found earlier. tangency
o) x)

(In

:J

J -

1.6 Substitution Methods and ExactEquations


Bernoulli Equations
A

63)

first-orderdifferential equation of the form)

\037)

dy
dx)

+ P(x)y =

Q(x)yn

(9))

is called Bernoulli quation.If either n = 0 or n a 1,then Eg.(9)is linear. e as we ask you to show in Problem 56,the substitution) Otherwise,
\037)

= yl-n)

(10))

transforms

Eq. (9)into the linear equation)

dv
dx)

+ (1 n)P(x)v = (1 n)Q(x).

Rather than memorizing the form of this transformed equation, it is more efficient to make the substitution in Eq. (10)explicitly, as in the followingexamples.)
...............

Example4)

If we rewrite the homogeneous quation 2xyy' = 4x2 + 3y2 of Example 2 in e

........

.....

....

the

form)

---y=-, dx
dy

3 2x

2x
y

we
n

seethat it is also a Bernoulli equation with P(x) = and 1 n = 2.Hence substitute) we

-1, v

Q(x)) -3/(2x),
1 dv

2x,)

= y2,)

= vI /2
1

,)

and)

-1/2 -=--=-v dx dx
dy dy dv dv

dx)

This gives)

_v2

/ 2 dv

3 _ _ _v dx 2x

/2 =

2xv-I/2.

Then multiplication by 2v 1 / 2 produces linear equation) the

-- dv

dx

x)

= 4x

with integrating

factor p

= e!(-3/x)dx x-3 . Sowe obtain) =

-3 4 Dx(x v) = 2\"; x)

x-3 v = x-3y2
y2

--+ C; = --+ C;
4
x)

x)

= -4x2 Cx3 +

.)

.)))

64

Chapter
1

First-OrderDifferential Equations)
The equation)

Example5)

x dy + 6y == 3Xy 4/3 dx
n ==

,)

is neither separablenor linear nor homogeneous, but


\037,

- -1.he T
n ==

it

substitutions)

v == y-I/3,)

== v

-3

and)

transform

it into)

4 dv 4 -3xv- dx + 6v-3 == 3xv- .

- -- dy == dy dv == dx dv dx

is a Bernoulli equation with

-3v_4dv dx)

Division by

4 -3xv- yieldsthe linear equation)

2 dv ---v==-1 dx
x) with integrating

factor p == eI(-2/x)dx == x-2 This gives)

-2 Dx(x v)
and finally,)

==

-2\"; x-2v == - + C; x) x)
1 1

v == x

+ Cx2

;)

y(x) =

1 2 (x + Cx )3
')

.)

Example6)

The equation)

2y 2xe dy

dx)
,)

== 3x4

+ e2y

(11
))

is neither separable, or linear, nor homogeneous, is it a Bernoulli equation. But n nor 2y we observethat y appearsonly in the combinations e2y and Dx (e2y ) == 2e y,. This
prompts the
substitution)

v == e2y
that

- dv

dx

2y == 2e dy dx)

transforms Eq. (11) the linear equation xv/ex)== 3x4 + vex); that into

-- dv 1

is,)

dx

x)

v == 3x3

.
that)

After multiplying

by the v==

integrating

factor p == l/x,we find


3

x
and hence)

f 3x dx==x +c,)
y(x) == 41nIx

4 so e2Y ==v==x +cx,)

+ cxl.)

.)))

and ExactEquations 1.6 Substitution Methods

65)

Flight rajectories T
y)

y =I(x))

/')

/' /'

/'/'

/
Va

The (0,0). plane travels with constant relative to the wind, which is blowing due north with constant speedw. speed Vo As indicated in Fig. 1.6.4, assumethat the plane'spilot maintains its heading we toward the origin. directly Figure helps us derive the plane'svelocity components relative to the are) ground. They

destination-an airport locatedat the origin

that Suppose an airplanedeparts from the point (a,0) locateddue eastof its intended

1.6.5

(a,0)

x)

FIGURE1.6.4. airplane The headedfor the origin.)

dx --= -vocose - + = J =e+ = J

dy. dt dt
Vo Sin

VOX

x2

'
y2
voY

x2

Yx2

+y2

/'/'
of the

/' /'/' 8

/'/'
x)

/'/'/\037
Iy I
I I I

\037

:,
y)

Hence trajectory y = f (x) of the plane satisfies the differential the


dy -- = dy/dt = dx 1

dx/dt

VOX)

+ w.
y2)

equation)

(voY
w

-J
w

x2

+ y2 ) .

(12))

If we set)
k

=-,
Vo)
\037

(13))

FIGURE

The 1.6.5. components


vector of the

the ratio of the

velocity airplane.)

neousform)

to windspeed the plane'sairspeed,then Eq. (12)takes the homoge-

=
\037\037 \037

_k[1+()2]'/2.)
to)

(14))

The substitution

= XV, y' = V + xv' then leadsroutinely


dv

f ,Jl +
By trigonometric substitution, we find that)
In

v2

- - f x.)
k

dx

(15))
integral

or by consulting a table for the


2

on

the left,

(v + J1 + v ) = -kInx + C,
= y(a)/a = 0 yields)
C = kIn a.)

(16))

and the

initial

condition v(a)

(17))

As we ask you to show in Problem then solving for V is)


v

t 68,he result of substituting (17)in Eq.(16)and

=
\037

[G
X

- G ) l)
k

(18))

Because = XV, y

we finally

obtain)

y(x) _ 2

- - [(-)
a a

l-k

-X
a)

l+k

(19))

for the equation of the plane's trajectory.)))

66

Chapter
1

First-OrderDifferential Equations) pass through


that k ==

1), then Eq. (19)takes the form y(x) == \037a(l x2ja2), so the plane's T trajectory approachesthe point (0,aj2) rather than (0,0). he situation is even worseif w > Vo (sok > I)-in case follows from Eq. (19)that y it as this +00 x O.The three cases re illustrated in Fig.1.6.6.) a
\037 \037

Note that only in the case < 1 (that is, w < vo) does the curve in Eq. (19) k the origin, so that the plane reaches destination. If w == Vo (so its

Example7)

If a == 200 mi,
will

Vo == 500 mijh,and w == 100 ijh,then k == wjvo == so the plane m succeed reaching the airport at (0,0). ith thesevalues, Eq. (19)yields) in W
\037,

y(x) == 100

4/5

) [(200

- (200 )
X

6/5

.)

(20))

o < x < 200?)


y)

Now supposethat we want to find the maximum amount by which the plane is blown off course during its trip. That is, what is the maximum value of y(x) for

Solution Differentiationof the function


dy

in

Eq. (20)yields)
5

dx

==

\037

2 5

) [ (200

4 \037 -1/5 6 \037 1/5 _

(200 ] )
400
\037

')

and we readily solve the equation y/


(a,0) (0,0)
W
x)

== (x) == 0to obtain (xj200)2/5 Hence)


\037.

Ymax

==

100

FIGURE1.6.6. three cases The


wind

< Va (plane velocity exceeds velocity), W = Va (equal and velocities), w > Va (wind is
greater).)

[(

2 2 3

- (2 ) ) 3

==

27

14.81.)

Thus the plane is blown almost 15 north at one point during its westwardjoumey. mi The (The graph of the function in Eq. (20)is the one used to construct Fig. 1.6.4. vertical scale there is exaggerated a factor of 4.) by

.)

ExactDifferential Equations
We

have seenthat a general solution y(x) of a first-order differential equation is often defined implicitly by an equation of the form)

F(x,y(x)) == C,)
where C is a constant. On the other hand, given the identity in the original differential equation by differentiating each side with

(21)) we (21), can recover

vided that Eq. (21)implicitly defines y as a differentiable function the original differential equation in the form)

respectto x. Pro-

- -aF aF dy ax + ay dx) N(x,y)

of x, this gives

==

0;
(22))

that is,)

M(x,y) + N(x,y)- == 0, dx)


dy

where M(x,y) == Fx(x, ) and y

== Fy(x,y).)))

and ExactEquations 1.6 Substitution Methods


It
\037)

67)

is sometimes convenient to rewrite Eq. (22)in the more symmetric

form)

M(x,y) dx + N(x,y) dy = 0,)

(23))

called differential its f (x,y) can be written

form. The general first-order differential equation y' == The prein this form with M = (x,y) and N = discussion showsthat, if there existsa function F(x,y) such that) ceding

ax)

-1.

aF =

M and

aF -=N
ay

,)

then

the equation)

\037)

F(x,y) = C)
defines a general solution of Eq. (23).In
this

exact ifferential equation-the differential) d

implicitly

case, g.(23)is calledan E

dF = Fxdx+ Fydy)
of F(x,y) is exactly M dx + N dy. Natural questions are these:How can we determine whether the differential h equation in (23)is exact?And if it is exact,ow can we find the function F such that Fx = M and Fy = N? To answer the first question, let us recall that if the mixedsecond-order partial derivatives Fxy and Fyx are continuous on an open set in the xy-plane,then they are equal: Fxy = Fyx. If Eq. (23)is exact and M and N have continuous partial derivatives,it then follows that)

aM
ay
y3

= Fxy = Fyx = aN
ax)

-.

Thus the equation)

aM
\037)

aN
ax) (24))

ay

is a necessarycondition that the differential equation M dx + N dy = 0 be exact. That is, if My =1= Nx, then the differential equation in question is not exact, so we need not attempt to find a function F (x,y) such that Fx = M and Fy = N -there is no such function.)

Example8)

The differential equation)

dx + 3xy 2 dy = 0)

(25))

is exact ecause can immediately seethat the b we


property that Fx

= y3

and

Fy

F(x,y) = xy3 has the function = 3xy 2. Thus a general solutionof Eq. (25)is)
xy3

= c;)
.)))

if you prefer, y(x) =

kx-I / 3

.)

68

Chapter
1

First-OrderDifferential Equations)
But

8 by y2 to obtain)

supposethat we divide each term of the differential equation in Example

dx + 3x dy = o.) This equation is not exact ecause, with M = y and N = 3x,we have) b

- -.
aM
ay

(26))

=1

=1=

3=

aN
ax)

Hence necessarycondition in Eq.(24) is not satisfied. the


We

are confronted with a curious situation

have exactly the samesolutions, a yet one is exact nd the other is not. In brief, whether a given differential equation is exact r not is related to the precise M dx N dy = 0 in which it is written. o form Theorem 1 tellsus that (subjectto differentiability conditions usually satisfied in practice)the necessary condition in (24) is also a sufficient condition for exactI ness. n other words,if My = Nx , then the differential equation M dx N dy = 0

(25)and (26)are essentiallyequivalent, and

here. he differential equations in T

they

is exact.)

THEOREM 1
tinuous

Criterion Exactness) for

c < y < d.Then the differential equation)


is exactn R if and only i

Supposethat the functions M(x,y) and N (x,y) are continuous and have confirst-order partial derivatives in the open rectangle R: a < x < b,

M(x,y)dx + N(x,y)dy = 0)
if)

(23))

aM
ay at

aN
ax)

(24)) R
with

eachpoint of R. That is,there exists a function F (x,y) defined on aFjax = M and aFjay = N if and only if Eq.(24)holdson R.)

Proof:We have seenalready that it is necessaryfor Eq. (24) to hold if T w (23)is to be exact. o prove the converse, e must show that ifEq.(24)holds, Eq.
then we can construct a function F(x,y) such that a Fjax = M and aFjay Note first that, for any function g(y ), the function)

= N.
(27))

F(x,y) =

f M(x,y)dx + g(y)
f
dX

satisfies the condition a Fjax = M. (In Eq. (27), he notation M(x,y) dx denotes t an antiderivative of M(x,y) with respectto x.) We plan to choose(y) so that) g
N

= aF =
ay

( f M(x,y) ) + g'(y)
\037 ay)

as well;that is, so that)


, g (y) = N

- a f M(x,y)dx.
ay)

(28))))

1.6 and ExactEquations Substitution Methods

69)

To seethat there is such a function of y, it sufficesto show that the right-hand sidein Eq. (28)is a function of y alone.We can then find g(y) by integrating with respect to y. Because right-hand sidein Eq. (28)is defined on a rectangle, and hence on the an interval as a function of x, it suffices to show that its derivative with respect o x t is identically zero. ut) B
\037

- J M(x, ) ax (
N
\037

ay

Y)dX

==

==

- -ax
\037\037

- ax J M(x,y)dx aN - a a ax ayax J M(x,y)dx


aN
ay
ay)

aN aM ==---==0 ax
by

hypothesis. So we can, indeed,find the desiredfunction Eq. (28).We substitute this result in Eq. (27)to obtain)

g(y) by

integrating

F(x,y) =

J M(x,y)dx+ (N(X,y)- J M(X,y)dX)d J


aay

(29))

as the desiredfunction

with

Fx == M and

Fy

== N.

\037)

Insteadof memorizing Eq. (29),it is usually better to solve an exact equation M dx+ N dy == 0 by carrying out the process indicated by Eqs.(27)and (28). irst F we integrate M(x,y) with respectto x and write)

F(x,y) =
thinking

J M(x,y)dx +g(y),)

of the function g (y) as an \"arbitrary constant of integration\" as far as the variable x is concerned. Then we determine g (y) by imposing the condition that a Fjay == N (x,y). This yieldsa general solution in the implicit form F(x,y) == C.)

Example9)

Sol e the differential equation) v


(6xy y3) dx + (4y + 3x

Solution Let M(x,y)

exact ecause) b

== 6xy

y3

and

- -.
N(x,y)
==

- 3xy2) 4y + 3x - 3
2 2
y

dy == O.)

(30))

xy 2. The given equation is

aM
ay

== 6x

-3

2 == aN

ax

Integrating a Fjax == M(x,y)

with

respectto x, we get)
2 y

F(x,y) =
Then we differentiate with

J -i)dx = 3x - xl+
(6xy
to respect y and

g(y).)

set a Fjay

== N(x,y). This yields)

aF =
ay)))

2 2 3x _ 3xi+ g'(y) = 4y + 3x

- 3xi,

70

Chapter
1

First-OrderDifferential Equations)
and
it

follows that g'(y) == 4y. Hence g(y) == 2y 2 + C1 , and

thus)

F(x,y)
equation)

2 == 3x y

xy3

+ 2y 2 + C

1 .)

Therefore, a general solution of the differential equation is defined implicitly


5 4 3 2
1
\037

by

the

2 3x y

xy3

+ 2y 2 == C)

(31))

-1 -2 -3
-4

(we have absorbedthe constant C1

into

the constant C).)

.)

-5

-5-4-3-2-1 0
x)

2 3 4 5

FIGURE1.6.7. field Slope and solution curves for the exact


equation in Example 9.)

Remark: Figure 1.6.7hows a rather complicatedstructure of solution s curves for the differential equation of Example9. The solution satisfying a given initial condition y(xo) == Yo is defined implicitly by Eq. (31), with C determined by x == Xo and y == Yo in the equation. For instance, the particular solution substituting 2 2 satisfying y(O) == 1 is defined implicitly by the equation 3x y xy3 + 2y == 2. The other two specialpoints in the figure-at(0,0) and near (0.75, 12)-are 2. ones where both coefficientfunctions in Eq. (30)vanish, so the theorem of Section1.3 doesnot guarantee a unique solution.

.)

Reducible econd-Order S Equations


A

second-order differentialequation involves the secondderivativeof the unknown


y(x), and thus has the general form)

function

F(x,y, y', y\")

== O.)

(32))

If either the dependentvariable y or the independent variable x is missingfrom a second-order equation, then it is easilyreducedby a simplesubstitution to a firstorderequation that may be solvable by the methods of this chapter. variable y missing. If y is missing,then Eq. (32)takesthe form) Dependent
\037)

F(x,y', y\")
substitution)

== O.)

(33))

Then the
\037)

== y I == dy

\"
')

dp
dx)

dx

y)

(34))

resultsin the first-orderdifferential equation)

F(x,p,p') == O.)
If we can solve this equation for a general solution p(x,C1 ) involving an constant C1,hen we needonly write) t
arbitrary

y(x) =

J y/(x)dx = J p(x,Cddx +

C2)

to get a solution of Eq. (33)that involves two arbitrary constants CI and C2 (as is to be expected the case a second-order in of differential equation).)))

and ExactEquations 1.6 Substitution Methods Example10) Solvethe equation xy\" + 2y' == 6x in
Solution The substitution
which the

71)

v dependent ariable y is missing.)

defined in (34)gives the first-order equation)

x dp + 2p== 6x;) dx

that

is,

-dp
2
1

dx + x) p

==

6.
by

Observing that the equation on the right here is linear, we multiply 21nx 2 factor p == exp (f (2Ix) dx) == e == x and get)
25

its integrating

2 2 Dx(xp) == 6x ,

20
15 10 5
;>, 0)

3 x2P == 2x + C ,

I dy P == dx == 2x+ x
2)

- -.
C

-15 -20 -25

-10
x)

-5

A final integration with

respectto x yieldsthe general solution) y(x) == x


2

-5-4-3-2-1012345
y

+ C + C2 x)
I

FIGURE1.6.8. Solution curves C] for 2 of the form == x C


1

:f:100.)

= 0,

+ x

\0373,

\03710, \03720, \03735, \03760,

of the second-order equation xy// + 2y' == 6x. Solution curves with CI == 0 but C2 i= 0 are simply vertical translates of the parabola y == x2 (for which CI == C2 == shows 0). Figure 1.6.8 this parabola and sometypical solutioncurves with C2 == 0 but C1 i= O. Solution curves with C1 and C2 both nonzero are verticaltranslatesof those(other than the parabola) shown in Fig.

1.6.8.
== O.)

.)

variable x missing. If x is missing,then Eq. (32)takes the form) Independent


\037)

F(y, y', y\")


substitution)

(35))

Then the
\037)

p==y)

dy

\"

dx')

- dx dp
\037\037

- --- dp dy dy dx p dp
dy)

(36))

resultsin the first-orderdifferential equation)

(y,

p,p

= 0)

for p as a function of y. If we can solve this equation for a general solution p(y, C1) involving an arbitrary constant CI,then (assuming that y' i= 0) we needonly write)

x(y)==

-dy== dy

dx

J dyldx

dy==

Jp

-dy==

J p(y, C

dy

I ))

+C2.
solution

If the final integral P == f (11 dy can be evaluated,the result is an implicit p) x(y) == P(y, C1)+ C2 of our second-order ifferential equation.))) d

72

Chapter
1

First-OrderDifferential Equations) Solvethe equation yy\" = (y')2 in which the independent variable x is missing.)
We

Example11)
Solution

,.....\037.

.....

assumetemporarily that y and y' are both nonnegative,and then point out at the end that this restriction is unnecessary.The substitution defined in (36)gives the first-order equation) dp yp- = P2.
dy)

Then separation of variables gives)

f d:= f d;, =
In

y p In y + C (because = C1y) p

> 0 and p = y' > 0),

where C1
5

= eC. Hence) dx
1 1 dy

4
3

C1 y'
1 .)

2
1
;>,

= C1x

f d; = loy + C

-1
-2 -3 -4 -5 -5 -4 -3 -2 -1a
x)

The resulting general solution of the second-order equation yy\"


1x y(x) = exp(C

- C2) =
-1

= (y')2is)

Ae Bx

,)

where A
1

that

= C2 and B the constants A and

e-

The FIGURE1.6.9. solution curves y = Ae Bx with B = 0 and A = 0, :f: are the horizontal lines 1 y = 0, :f: Theexponential curves are with B > 0 and A = :f:1 in color,those with B < 0 and A = :f: are black.) 1

1.

values of A, we get all horizontal lines in the plane as solution curves. The upper half of Fig.1.6.9 showsthe solution curves obtained with A = 1 (for instance) and these solution curves are reflected different positive values of B. With A = a in the x-axis, nd with negative values of B they are reflected in the y-axis. In particular, we seethat we get solutions of yy\" = (y')2, allowing both positive and f negative possibilitiesor both y and y'.

satisfiesyy\"

= C1.Despiteour temporary assumptions, which imply B are both positive, we readily verify that y(x) = Ae Bx = (y')2 for all real values of A and B. With B = 0 and different

.)

IIIIIJ Problems)
Find

lems

1through

generalsolutions of the differential equations in Prob30. Primes denotederivatives with respectto x


Y

throughout.)

5. x(x+ y)y' = y(x y) 7. xy2y' = x 3 + y3 9.x2y' = xy + y2 11.x2 y2)y' = 2xy ( 12.yy' = y2 + xJ4x2 + y2 x 13. y' = y + Jx 2 + y2 x 14.y' + x = Jx 2 + y2) y

3.xy' = y + 2,JXY

1.(x + y)y' = x -

2.2xyy' = x 2 + 2y 2 4. (x - y)y' = x + y 6.(x + 2y) y' = Y 8.x 2y' = xy + x 2e /x 10. yy' = x 2 + x


Y

15. (x+ y)y' + y(3x + y) = 0 x 17.y' = (4x + y)2 16. ' = ,Jx + y + 1 y =1 x 19. 2y' + 2xy = 5y 3 18.x + y)y' ( 3 = 6x 21. ' = y + y3 20.y2y' + 2xy y 23. xy' + 6y = 3xy 4/3 22.x 2y' + 2xy = 5y 4 24. 2xy' + y 3e- = 2xy
2x

3y2)

25.y2(xy' + y)(l + X4)1/2= x x 26.3y2y' + y3 = e= 3x4 + y3 27. 3xy2y' 28.xeYy' = 2(e + x 3e2x ) 29.(2xsin y cosy) y' = 4x2 + sin 2 y 30.(x + eY)y' = xeY Y

1)))

and ExactEquations 1.6 Substitution Methods


In equation

73)

Problems through 42, verify is exact;then solveit.)

31

that the given differential

60.Usethe
equation)

method

in

Problem

59 to solvethe

differential

31.(2x + 3y) dx + (3x + 2y) dy = 0

32. (4x y) dx + (6y x)dy = 0 33. (3x2 + 2y2) dx + (4xy + 6y2) dy = 0 34. (2xy 2 + 3x2) dx + (2x2y + 4y3) dy = 0 35.
\037

dy

dx M 61. ake an appropriate

xy 36. (1+ yexy ) dx + (2y + xe ) dy

(x3+ )dx+(y2+1nx)dy=O =0
1

62.Show

substitution to find solution of the this general solution sin(x dyldx equation x ]f12that is readily contain the linear solution y (x) verified by substitution in the differential equation?

-x - 3y + 718 4x - y). Does- a


2y

that the solution dy

curves of the

differential

equation)

37. (cosx+lny)dx+

=0 38. (x + tan- y) dx + 1 + y2 dy 39.(3x2y3 + y4) dx + (3x3y2 + y4 + 4xy3) dy


y

( + +eY)dY=O x
\037

dx

3 y(2x _ y3) x(2 3 x 3 y

))

=0

40. (eX sin y + tan

y) dx

2 + (eX cosy + x sec y) dy = 0 2Y

41. 2X

42.

2X 5/2

-3

3y2 x4

dx+
dx +

x3

X2

areof the form x 3 + y3 = 3Cxy. 63.Theequation dyldx = A(X)y2 + B(x)y+ C(x)is called that a Riccatiequation.Suppose oneparticular solution Yl (x) of this equation is known. Show that the substitution)
y

y5/3

3y5/3

y2

+\037 dY=O
\037 dy

2X 5/ 2

2X 5/ 2y2/3

3X 3/ 2y5/3

=0
transforms the

= Yl +

1 v)

Riccati equation
dv

into the

linear equation)

Find a generalsolution of eachreducible second-orderifferd ential equation in Problems43-54. Assume x, y and/or y/ positive where helpful (as in Example

11).

dx

+ (B + 2AYl)V = -A.)

43. xy\" = y/ 45. y\" + 4y = 0 47. y\" = (y/)2 49. yy\" + (y/)2 = yy/ 51.\" = 2Y (y/)3 y 53. y\" = 2yy' 55. Show that the substitution
the differential equation

50.y\" = (x + y/)2 52. y3 y\" = 1 54. yy\" = 3(y')2 v = ax + by + c transforms dyldx = F(ax+ by + c) into a

44. yy\" + (y/)2 = 0 46. X y\" + y/ = 4x 48. x 2y\" + 3xy' = 2

Usethe method of Problem 63 to solve the equations in Problems 64 and 65, given that Yl (x) = x is a solution of each.

64. dy + y2 = 1 + x 2
dx
dy 65. dx + 2xy = 1 + x 2 + y2

separable equation. 56. Supposethat n 1= 0


dy Idx + P (x) y = stitution
v

I-n

and n 1= Show that the subtransforms the Bernoulli equation Q(x) yn into the linear equation

1.

66.An equation

of the

form)
y

= xy/ + g(y/))
the

(37))

= + (1- n)P(x)v(x) (1- n)Q(x). dx


dv

parameter

is called a Clairaut equation. Show that by) family of straight lines described y(x)

one(38))

57. Show
ential

that the substitution

equation linear equation 58.Usethe ideain

= In y transforms the differIn dyldx + P(x)y = Q(x)(y y) into the dvldx + P(x)= Q(x)v(x).
v

= Cx + g(C))

Problem 57 to solvethe

equation

t 67. Considerhe

is a general solution of Eq.(37).


Clairaut equation)
y

- 4x dx
dy dy

+ 2y In y = o.

= xy/

59.Solve the

differential

equation

dx
k by finding k transform v y

x x

- -1
y

for which g(y/)

-i
y

i(y')2)
that the line)

(y/)2 in Eq. (37).Show

= Cx

+y +3

- iC2)
=
iC2).

= +

hand so that
dv
du

the substitutions x u + h, it into the homogeneousequation u u

v
v)

x 2 at the point (\037C, to the parabola y x 2 is a singular soluExplain why this implies that y tion of the given Clairaut equation. This singular solution and the one-parameterfamily of straight line solutions are illustrated in Fig. 1.6.10.))) is tangent

74

Chapter
1

First-OrderDifferential Equations)

71.A river
A

72.

north at w feet per second. 4 ft/s, always at and swims at Va heading toward a tree at (0,0) on the west bank directly 2 ft/s, acrossfrom the dog'sstarting point. (a) If w 4 ft/s, show that the dog reachesthe tree. (b) If w show that the dog reachesinstead the point on the west bank 50 ft north of the tree. (c) If w 6 ft/ s, show that the west bank. the dog never reaches In the calculus of plane curves, one learns that the curvature K of the curve y y) is given y (x) at the point

100ft wide is flowing

dog starts

(100,0)

= =

(x,

by)

FIGURE Solutions of the Clairaut equation of Problem 67.The \"typical\" straight line with equation y = Cx C2 is tangent to the parabola at the point (4 C2). 68. eriveEq.(18)n this sectionfrom Eqs.(16)nd (17). D i a I 69.n the situation of Example 7, supposethat a = 100mi, Va = 400 mi/h, and w = 40 mi/h. Now how far northward doesthe wind blow the airplane? 70. As in the text discussion, upposethat an airplane mains tains a heading toward an airport at the origin. If Va = 500 mi/h and w = 50 mi/h (with the wind blowing due north), and the plane begins at the point (200, 150), show that its ed trajectory is describ by

1.6.10.

c,

\037

\037

[1 + Y/(X)2]3/2' r and that the curvature of a circleof radius r is K = 1/ of [See Example 3 in Section11.6Edwards and Penney, Calculus:Early Transcendentals,7th edition (Upper Saddle River, NJ: PrenticeHall, 2008).] Conversely, substitute p = y/ to derivea general solution of the second-order

K=

ly\"(x)1

differential

equation)
ry\"

= [1 (y/)2]3/2) +
(y
2
.)

(with

constant)

in the form)

(x
Thus

- a)2 + - b)2 = r
curvature

9 + x 2 + y2 = 2(200x ) 1/10.)

a circleof radius r (or a part thereof) is the


with constant

only

plane

curve

l/r.)

lIB P()pulationModels) 1.4 I Sectionoccurs = P(t)


n with

we introduced the exponential differential equation dPjdt = kP, solution Poekt , as a mathematical model for natural population we as a result of constant birth and death rates. Here present growth that a more general population modelthat accommodatesirth and death rates that are b not necessarily constant. As before,however, our population function P(t) will be a continuous approximation to the actual population, which of course changesonly is, by integral increments-that by one birth or death at a time. o Supposethat the population changesonly by the occurrencef births and deaths-theres no immigration or emigration from outsidethe country or envii ronment under consideration.It is customary to track the growth or decline a of population in terms of its birth rate and death rate functions defined as follows:)

. .

(t) is the number of births per unit of population per unit of time at time t; 8 (t) is the number of deaths per unit of population per unit of time at time t.)
fJ

[t, t

Then the numbers of births and deaths that occurduring \037t] is given (approximately)by)

the time interval


t

births:
length
\037t

fJ

(t) . P(t) .
in

\037

,)

deaths: 8 (t) . P(t) .

\037

.)

Hence change \037P the


is
\037P

the population during the time interval [t, t


\037

\037t]

of

= {births} {deaths}

fJ(t)

. P(t) .

\037t

- 8(t) . P(t) .

\037t,)))

so)

\037P
\037t

1.7PopulationModels 75)
[,8(t) 8(t)]P(t).
\037t

\037

The error in
the
\037)

this

limit-weget the differential equation)


dP
dt

approximation should approach zero as

-+ 0,so-taking
(1))

= (,8

- 8)P,)

in which we write ,8 = ,8(t),8 = 8(t), and P = P(t) for brevity. Equation (1)is the general opulation equation.If ,8 and 8 are constant, Eq. (1)reducesto the p But it alsoincludesthe possibilitythat f3 natural growth equation with k = ,8 and 8 are variable functions of The birth and death rates need not be known in advance; they may well dependon the unknown function P(t).)

t.

- 8.

Example

1)

initially, and that its death rate is Supposethat an alligator population numbers 100 = 0 (sonone of the alligators is dying).If the birth rate is ,8 = (0.0005)P-and thus increasess the population does-then (1)gives the initial value problem) a Eq. dP = 2 , (0.0005)P P(O) = 100 dt)

(with

t in

years).Then upon separating the variableswe get)

= dt; f ;2 dP f (0.0005)
1
Substitution

- P = (0.0005)tC. +
we readily solve for)

of t

= 0,P = 100 C = -1/100, then and gives


P(t) =

w in Fig. indicate that a population explosion always occurs, hatever the size of the (positive) initial population P (0) = Po. In particular, it appearsthat the unbounded in afinite periodof time. population always becomes
500) 400)

2000 20 . = = so For instance, P(10) 2000/10200, after 10 years the alligatorpopus lation has doubled. But we seethat P -+ +00 t -+ 20, o a real \"population as occurs 20 in explosion\" years.Indeed,the direction field and solution curves shown

t)

1.7.1

.)

300
Q..

200 (0,100)

100)

10 20 30 40 50
t)

field FIGURE Slope and p dP/dt = (0.0005) 2 in Example

1.7.1.

solution
1.)))

curves for

the equation

76

Chapter
1

First-OrderDifferential Equations)

Bounded opulations the Logistic quation and P E


In situations as diverse as the human population of a nation and a fruit fly population a closed as container, it is often observedthat the birth rate decreases the population itself increases. reasonsmay range from increasedscientificor cultural The sophistication to a limited food supply. Suppose,for example,that the birth rate function of the population size so that fJ == fJo fJ 1 fJ is a lineardecreasing where fJo and fJI are positive constants.If the death rate 8 == 80 remains constant, then Eq. (1)takes the form)
in

P,

P,

dP
dt
that
\037)

== (fJo

- P - 80)
fJI
')

P;)

is,

where a == fJo 80 and b == fJI. If the coefficientsa and b are both positive, then Eq. (2)is calledthe logistic equation. For the purposeof relating the behavior of the population P(t) to the values of the parameters in the equation, it is useful to rewrite the logisticequation
in

dP 2 -==aP-bP dt

(2))

the form)

dP
\037)

dt

== kP(M

- P),)

(3))

where k == band M == a/b are constants.)

Example2)

In
by

1.3 Example4 of Section we exploredgraphically a population that is modeled the logisticequation) dP 2


\"....\" \"

dt)

==

. 0.0004P(150 == 0.06P 0.0004p P)

(4))

To solve this differential equation symbolically,we separatethe variables and inte-

grate.We

get)

J P(1 - P) = J 0.0004dt, _ J ( + 150 P ) dP = J 0.0004dt [partial fractions], + IPI-In1150PI 0.06t C, P c c - :f:eeO.0 BeO.0 [where B :f:e]. 150 If we - 0 P Po 150 last equation, we B P /(150 Po).Hence) P PoeO.06t - 150 150
5\037

1\0370

\037

In

== ==

6t ==

6t

==

P)

substitute

t ==

and

==

i=

into

this

find

that

== o

P)

Po)

Finally, this

equation is easyto solve for the population)

P(t) ==

150P o Po + (150 Po)e-0.06t)

(5))))

1.7PopulationModels 77)
p)

300 240 180 150 120

showsa number population Po == P(0).Figure 1.7.2 of solutioncurves corresponding different valuesof the initial population ranging to from Po == 20 Po == 300. ote that all these solution curves appear to approach to N the horizontal line P == 150 an asymptote. Indeed,you should be able to see as whatever the initial value Po > O. directly from Eq. (5)that limt\037oo P(t) == 150,
at time
t in

terms of the

initial

.)

60
20)
25)

and Limiting opulations CarryingCapacity P


50)
75)

100)

FIGURE1.7.2. Typical solution


curves for the logistic equation

P'

==

O.06P O.0004p2.)

The finite limiting population noted in Example2 is characteristicof logisticpopulations. In Problem 32 we ask you to use the method of solution of Example 2 to show that the solution of the logisticinitial value problem)

dP
dt
IS)

== kP(M

- P), -

P(O) == Po)

(6))

P(t) ==
Actual animal

MP0 kMt) Po + (M Po)e-

(7))

populations are positive valued. If Po == M, then (7)reduces to the unchanging (constant-valued) \"equilibrium population\" P(t) = M. Otherwise, the behavior of a logisticpopulation dependson whether 0 < Po < M or Po > M. If 0 < Po < M, then we seefrom (6)and (7)that P' > 0 and)

P (t)

==

MPo kMt) Po + (M Po)e-

MPo
Po + {pos. number})

<

However, if Po > M, then we seefrom (6)and (7)that

P' < 0 and)


>

P(t) ==
p)

MPo
Po + (M Po)e-

MPo
kMt)

Po + {neg. umber} n or \"negative


number\" in

MPo
Po)

==

M.

MPo
Po)

== M.

In either case,he \"positive number\" t


\037

the denominator has

absolutevalue lessthan Po and-becausef the exponential factor-approaches o 0 as t I +00.t follows that)


M)

lim
t\037+oo

P(t) ==

MPo
Po + 0)

== M.

(8))

M/2)

P = M/2)

curves for the logistic equation


==

FIGURE1.7.3. Typical solution

curve

P' kP(M P). Each solution


that starts

P == M/2 has an
on this

below the

line. (See Problem 34.))

line inflection point

Thus a population that satisfies the logisticequation does not grow without bound like a naturally growing population modeledby the exponential equation P' == k P. Instead,it approachesthe finite limiting population M as t --+ +00. As illustrated by the typical logisticsolution curves in Fig. 1.7.3, population the M a P(t) steadily increasesnd approaches from belowif 0 < Po < M, but steadily decreases approachesM from above if Po > M. Sometimes is called the and M it carryingcapacityof the environment, considering to be the maximum population that the environment can support on a long-term basis.)
in the population of a certain country was 50 million and was at the rate of 750,000eopleper year at that time. Supposealso that in growing p its population was million and was then growing at the rate of 1 million per year. Assume that this population satisfiesthe logisticequation. Determineboth the limiting population M and the predicted population for the year 2000.)))

Example3)

Supposethat

1885

1940

100

78

Chapter
1

First-OrderDifferential Equations)
We substitute the two given

Solution

== 1.00 100k(M 100).) We solve simultaneously for M == 200 k == 0.0001. the limiting Thus and
0.75== 50k(M 50),

pairsof data in Eq. (3)and find

that)

tion with

populaof the country in question is 200 million. With these values of M and k, and we t == 0 corresponding the year 1940 which Po == 100), find thatto (in will to Eq. (7)-the in the year 2000 be) according population
= P(60)= 100 (200 +

- 100)e-(O.OOOl)(200)(60)
,)

. 100200

about 153.7 million people.)

.)

Historical Note)
The logisticequation was introduced (around 1840) y the Belgianmathematician b and demographer . F. Verhulst as a possiblemodel for human population growth. P In the next two examples comparenatural growth and logisticmodel fits to the we for 19th-century U.S.opulation census data, then compareprojections the 20th p
century.)

Example4)

The U.S. w m population in 1800as 5.308illion and in we take Po == 5.308 t == 0 in 1800)the natural in (with and substitute t == 100, == 76.212, find that) P we
== 76.2125.308e
IOOr
,)

million. If w 1900as 76.212 rt


growth

model P(t) == Poe

so r

==

-In76.212 0.026643. 100 5.308)


1
\037

Thus our natural


IS)

growth

modelfor the U.S. population during the 19th century)

P (t) == (5.308)e(O.026643)t)
(with t in

(9))

years and

lation

growth

and between1800

the P in millions). Because eO.026643 1.02700,averagepopu\037

w 1900as about 2.7%per year.

.)

Example5)

and The U.S. million. If we take Po == 5.308 subw population in 1850as 23.192 P and stitute the data pairs t == 50,P == 23.192 1850) t == 100, == 76.212 (for (for in 1900)the logisticmodelformula in Eq. (7),we get the two equations)

(5.308)M 5.308 (M - 5.308)e-IOOkM 76 212 +


==
.)

(5.308)M == ' 50kM 5.308 (M 5.308)e- 23.192 +

(10))

k and M. Nonlinear systemslike this ordinarily are solved an appropriate computer system. But with the right algebraic numerically using trick (Problem36 in this section)the equations in (10)can be solved manually for k == of thesevalues in Eq. (7)yieldsthe M == Substitution model) logistic
in

the two

unknowns

0.000167716,188.121.
P(t) == 5.308 +

998.546

(182.813)e-(O.031551)t)

( 11))))

1.7PopulationModels 79) U.S. The table in Fig.1.7.4 comparesthe actual 1800-1990 censuspopulafigures with those predictedby the exponential growth model in (9) and the Both agreewell with the 19th-centuryfigures. But the exlogisticmodel in ponential model diverges appreciably from the census data in the early decades of the 20th century, whereas the logisticmodel remains accurate until 1940. By the end of the 20th century the exponential model vastly overestimatesthe actual the U.S. over population-predicting a billion in the year 2000-whereas logistic modelsomewhatunderestimates it.)
tion

(11).

.:.......';.......:...);\037c\037\037Q\037:......:..........

>......:..]\037\037DQJ.1\037D.tial

\302\245\037at

;;\\:.\\:i.;\037\037\037\037:illftR\037>,:;)

:,;;;;.cii:Mfjelel
i)

Exponential Error)

Logistic Model
5.308 7.202 9.735

Logistic

Error

1800

1810

5.308

7.240

5.308 6.929

0.000 0.311
0.594
3.079

0.000 0.038

1820 1830 1840 1850 1860 1870 1880 1890 1900

12.861 23.192
31.443 17.064

9.638

11.805 20.113
26.253 34.268

9.044

1.056
1.655

-0.097
-0.437
0.000 1.038

15.409

5.190
4.290
5.459 4.593

23.192
30.405 39.326 50.034 62.435

13.095 17.501

-0.234
-0.768
0.155

50.189
62.980

38.558

44.730
58.387

1910
1920 1930 1940 1950 1960 1970 1980 1990
Error)

76.212

76.212
99.479

106.022 123.203 132.165 151.326


179.323
203.302 226.542

92.228

129.849 169.492

221.237
288.780 376.943 492.023 642.236 838.308 1094.240)

0.000 -7.251 -23.827 -46.289 -89.072 -137.454 -197.620 -288.721 -415.694

76.213
90.834

-0.001
1.394

0.545

105.612 0.410 119.834 3.369 132.886 -0.721


144.354 154.052
6.972 25.271

248.710
281.422)

2000)
Exponential)

-812.818) 177.038
growth and

-589.598

161.990 41.312 168.316 58.226


173.252
76.458 104.384)

40%) 20%)

FIGURE 1.7.4. Comparison of exponential


populations (in millions).)

logistic models with

U.S.ensus c

-40%)

FIGURE1.7.5. Percentage
errors in the exponential and logistic population models for

1800-1950.)

The two modelsare comparedin Fig. 1.7.5, where plots of their respective shown for the 1800-1950 actual population-are We seethat the logisticmodel tracks the actual populationreasonablywell period. this ISO-year eriod. However, the exponential error is considerably throughout p larger during the 19th century and literally goesoff the chart during the first half of the 20th century. In orderto measure the extent to which a given model fits actual data, it iscuserror(in the model)as the squareroot of the average tomary to define the average the squaresof the individual errors (the latter appearing in the fourth and sixth of columns of the table in Fig. 1.7.4). data, Usingonly the 1800-1900 this definition 3.162 the average error in the exponential model,while the averageerror for gives in the logisticmodel is only 0.452. w Consequently,even in 1900e might well have that the logisticmodel would predictthe U.S. population growth during anticipated .))) the 20th century more accurately than the exponential model.

errors-aspercentage f the a o

80

Chapter
1

First-OrderDifferential Equations)
The moral of Examples and 5 is simply that one should not expect much 4 too of modelsthat are basedon severely limited information (suchas just a pair of data o points). Much of the sciencef statisticsis devoted to the analysis of large \"data sets\" to formulate useful (and perhapsreliable)mathematicalmodels.)

MoreApplications the Logistic quation of E


We next

describeome situations s

that

in illustrate the varied circumstances

which

the

logisticequation is a satisfactory mathematicalmodel.)


situation. A certain environment
tion of at most M individuals. It is then reasonable expect growth rate to the 8 (the combined birth and death rates)to be proportional to M befJ causewe may think of M P as the potential for further expansion.Then 8 == k(M so that fJ

1.Limited environment

can support a popula-

- P),

- P,

dP dt

== (fJ

- 8)P

== kP(M

- P).)

The classicxample a limited environment situation is a fruit fly population of e in a closedontainer. c 2. Competition situation. If the birth rate fJ is constant but the death rate 8 is proportional to P,so that 8 == a P,then)

dP
dt)

== (fJ

- aP)P

== kP(M

- P).

3.

This might be a reasonableworking hypothesis in a study of a cannibalistic population, in which all deaths result from chance encounters between individuals. Of course, ompetition between individuals is not usually so deadly, c nor its effects so immediate and decisive. Joint proportionsituation. Let pet) denote the number of individuals in a constant-size susceptible population M who are infected with a certain conand incurable disease. he disease spreadby chance encounters. T is tagious Then P'(t) should be proportional to the product of the number P of individuals having the diseasend the number M P of those not having it, and a therefore dPjdt == kP(M P). Again we discoverthat the mathematical modelis the logisticequation. The mathematical description the spreadof of a rumor in a population of M individuals is identical.)

Example6)

population M == 100 thousand people have heard a certain rumor. After 1 week the number pet) of those who have heard it has increased P(l)== 20 to thousand. Assuming that pet) satisfies a logisticequation, when will 80% of the city's population have heard the
..........,.. ..............

.....

Supposethat at time

t ==

thousand 0,10

....

peoplein

a city

with

rumor?)

Solution

Substituting

and Po == 10 M == 100 (thousand) in Eq. (7),we get

. pet) == 10 90e-lOOkt +
Then substitution of t ==

1000

(12))

1,P == 20gives the equation)


20==

1000 10 90e+

100k)))

1.7PopulationModels 81)
that

is readily solved for

e-100k ==
With

\037,

so

k == 160In

\037
\037

0.008109.

pet) == 80, q. (12)takes the form) E


80 ==

1000 ' 10 90e-100kt +


In36
In

which we solve for e-100kt== 316 It follows that 80% of the population has heard the rumor when)

t==
thus

In36

lOOk

9 4)

\037

4.42,

after about 4 weeksand 3 days.)

.)

v Extinction Doomsdayersus
Considera population pet) of unsophisticatedanimals in which females rely solely on chance encounters to meet males for reproductive purposes.It is reasonableto such encounters to occurat a rate that is proportional to the product of the expect number P/2 of males and the number P/2 of females, hence a rate proportional at to p2 We therefore assumethat births occurat the rate kP2 (per unit time, with k constant). The birth rate (births/time/population) then given by f3 == kP. If is the death rate 8 is constant, then the general population equation in (1)yields the differential equation)

\037)

dP
dt)

== kP2

- 8P

== kP(P

- M)

(13))

(where M == 8/k > 0) as a mathematicalmodel of the population. Note that the right-hand sidein Eq. (13)is the negative of the right-hand side in the logisticequation in (3). We will seethat the constant M is now a threshold w population, ith the way the population behaves in the future dependingcritically on whether the initial population Po is less than or greater than M.)

Example7)

Consideran animal population pet) that is modeled the equation by dP 2 == 150)== 0.0004P 0.06P. dt 0.0004P(P We want to find pet) if (a) P(O) == 200; P(O) == 100.) (b)

(14))

Solution To solve the equation in (14),we separate the variables and integrate. We get
dP

f
-\037
\037

PcP
1

- 150))

==

f 0.0004dt,

150 ( P P - 150) dP f 0.0004dt[partial fractions], f ) + IPI -InIP - 1501 -0.06t


==
In

==

C,)

-P150) P

==

Ce ::i:e-0.06t== Be-0.06t

[where B == ::i:e ].)

(15))))

82

Chapter
1

First-OrderDifferential Equations)

(a) Substitution

of t == 0 and we solve Eq. (15)for)

P == 200 (15)gives B == 4. With into


pet) == 4e-.06t
O

this

value of B (16))

.06t 600e- .
O

1)

Note that, as t increasesnd approachesT == In(4)/0.06 23.105, positive a the denominatoron the right in (16)decreases approaches . Consequently pet) and O as r +00 t --+ This is a doomsday situation-aeal population explosion. With this value of Substitution of t == 0 and P == 100 (15)gives B == into (b) B we solve Eq. (15)for)
\037

T-.

\037

-2.

Pet )
Note
that,

300e- 2e- + O

.06t

.06t

300 2 + e .06t
O

.)

(17))

as t increases ithout bound, the positive denominator on the right in w T +00. onsequently, pet) --+ 0 as t --+ +00. his is an (eventual) (16)approaches C extinctionsituation.

.)

p)

M)

t)

curves for the explosion/extinction = kP(P M).) equation

FIGURE1.7.6. Typical solution

P'

variables and use partial fractionsto solvethe initial Separate value problems in Problems 1-8. either the exactsolution Use ora computer-generatedslope fieldto sketch the graphs ofseveral solutions of the given differential equation, and highlight
the indicated

_
5. dx dt dx dt dx
7.
dt

Thus the population in Example7 either explodes is an endangered or species threatened with extinction, depending whether or not its initial size exceeds the on threshold population M == 150. n approximation to this phenomenon is someA times observedwith animal populations, such as the alligator population in certain areas of the southern United States. showstypical solution curves that illustrate the two possibilities Figure 1.7.6 for a population pet) satisfying Eq. (13). Po == M (exactly!), the populaIf then tion remains constant. However, this equilibrium situation is very unstable. If Po M exceeds (even slightly), then pet) rapidly increases ithout bound, whereas if w the initial (positive)population is less than M (however slightly), then it decreases S Problem 33.) (more gradually) toward zero as t --+ +00. ee

Problems)
the population numbers 100 rabbits and is increasing at the rate of 20 rabbits per month. How many rabbits will there beone year later?

particular
X (0)

solution.)

3. dx = 1 - x 2, x(O)= 3 dt

1.dx = x - x 2, dt

= 2 2. dx = lOx -x2,x(0)= 1 dt
4.
dx
dt)

10. upposethat the fish population S


by

pet) in a lake is attacked disease time t = 0, with the result that the fish at cease reproduce that the birth rate is = 0) and the to (so
a
f3

=9

- 4x , x(O)= 0
2

-=3x(5-x),x(0)=8

death rate 8 (deaths per week per fish) is thereafter proporIf there were initially 900 fish in the lake tional to and 441were left after 6 weeks,how long did it take all the fish in the lake to die?

Ij-JP.

6. = 3x(x - 5),x(O)= 2

9.The time rate of change of a rabbit population P is proportional to the square root of P.At time t = 0 (months))

dx 8. dt = 7x(x - 13), (O)= x

11.upposethat when S
-JP.

a certain birth and death rates f3 and tional to (a) Show that)

with fish, the lake is stocked 8 are both inversely propor-

= 4x(7 x),x(O)= 11
17)

pet) = Okt + \037r


')

100and after 6 months where k is a constant. (b) If Po there are 169fish in the lake, how many will there be after
1 year?)))

1.7PopulationModels 83) 12. he time rate T


of change of an alligator population P in a swamp is proportional to the square of The swamp contained a dozenalligators in two dozenin
When will there be four What happens thereafter?
f3

dozenalligators

1988,

P.

in the

1998. 22.

swamp?

13. onsidera prolific C


rates,
tion and

breedof rabbits whose birth and death 8, areeachproportional to the rabbit populaP = pet),with > 8. (a) Show that)
f3

P (t )
Note that
\037

- - Pot'
Po
k

k constant.

P (t) +00 as t 1/(k Po). This is dooms(b) Supposethat Po = 6 and that there are nine day. rabbits after ten months. When does doomsday occur? 14. epeatpart (a) of Problem 13 in the case < 8. What R
\037 f3

15. onsidera population C


= aP =

now happens

to the

rabbit population

in the long run?

P (t) satisfying the logistic equation dP/dt is the time rate bP2, where B at which births occurand D b p2 is the rate at which deaths occur. If the initial population is P (0) Po, and Bo births per month and Do deaths per month are occur0, show that the limiting population is ring at time t

= aP

16. onsidera rabbit C

= BoPo/Do.

17.Considera rabbit

population P (t) satisfying the logistic If the initial population is equation as in Problem rabbits and there are 8 births per month and 6 deaths per month occurring at time t 0, how many months doesit take for P (t) to reach95%of the limiting population M?

15. =

120

in KN03 dissolves methanol, the number x(t) satisfies of grams of the salt in a solution after t seconds 2 the differential equation dx/dt = 0.8x 0.004x (a) What is the maximum amount of the salt that will ever in dissolve the methanol? (b) If x = 50 when t = 0, how long will it take for an additional 50 g of salt to dissolve? 24. Supposethat a community contains 15,000peoplewho t are susceptibleo Michaud's syndrome, a contagious disease.At time t = 0 the number N (t) of peoplewho have Michaud's syndrome is 5000and is increasing developed at the rate of 500 per day. Assume that N (t) is proportional to the product of the numbers of those who have and of those who have not. How long caught the disease to will it take for another 5000people developMichaud's

P' 23.As the salt

growing at the rate of 1 million per year. Predict this country's population for the year 2000. 0, half of a \"logistic\" populaSupposethat at time t tion of personshave heard a certain rumor, and that the number of those who have heard it is then increasing at the rate of 1000persons per day. How long will it take for this rumor to spreadto 80%of the population? (Suggestion: Find the value of k by substituting P(O)and (0) in the logistic equation, Eq.

100,000

(3).)

'

syndrome?

25.The data

in the table in

18. onsidera C
=

population pet) satisfying logistic If the initial population is 240 equation as in Problem rabbits and there are9 births per month and deaths per month occurring at time t 0, how many months doesit take for P (t) to reach105%of the limiting population M?

15.
=

the

population (a) What is the limiting the approximation)

P (t)

that

are Fig. 1.7.7 given for a certain satisfies the logistic equation in (3).
population

M? (Suggestion: Use
h)

12

P I (t)
h = 1 to 25.00 when and
with

\037

pet + h)

- P(t 2h)

estimate the values

P (t) satisfying the extinctionpopulation ap2 where B ap2 explosion equation dP/dt is the time rate at which births occur and D b P is the rate at which deaths occur. If the initial population is P (0) Po and Bo births per month and Do deaths per month are occurring at time t 0, show that the threshold

- bP,
P (t)

of P'(t) when P = 47.54.Then substitute these values

in the logistic equation and solvefor k and M.) (b) Use the values of k and M found in part (a) to determine when 0 to correspond the to P (Suggestion: Take t

= 75.

19. onsider an C

population

is M

= DoPolBo.

year 1925.))

Year 1924 1925 1926 1974 1975


1976)

P (millions)
24.63 25.00 25.38) 47.04 47.54 48.04)
data for Problem

the alligator population satisfying as in Problem If the iniextinction/explosionequation tial population is 100alligators and there are births per month and 9 deaths per months occurring at time t 0, how many months does take for P (t) to reach times it the threshold population M?

18. 10 = 10

20. Consider an alligator population P (t) satisfying the If extinction/ explosion equation as in Problem 18. the initial population is 110 births per alligators and there are 11 month and 12 deaths per month occurring at time t = 0, how many months doesit take for P (t) to reach10%of the threshold population M? 21. upposethat the population P (t) of a country satisfiesthe S differential equation dP/dt = k P (200 P) with k constant. Its population in 1940 was 100million and was then)

FIGURE1.7.7.opulation P

25.
f3

26.A

population P (t) of small rodents has birth rate (births per month per rodent) and constant death (O.OOI)P rate 8. If P (0) = 100 and P'(0) = 8, how long
(in)))

84
27.

Chapter
1

First-OrderDifferential Equations)

dents? (Suggestion: First

months) will it take this population to double to 200rofind the value of Consider animal population P (t) with constant death an rate 8 (deaths per animal per month) and with birth rate f3 proportional to that Suppose P (0) 200 and P'(O) (a) When is P (h) When does

8.)

33.(a)

Derivethe

solution)

= 0.01 = 2.

pet) =
of the

P.

= 1000?

Po

MPo kMt) (M Po)e

28. uppose the number x(t)(with that S


2 0.0001x O.Olx. tors in

doomsday

occur?

kP(P- M), P(O)= Po.


extinction-explosion

initial

value problem

P' =

a swamp satisfies the

t in months) differential equation

of alligadx/dt

(a) If initially

there are25 alligators in the swamp, solve this differential equation to determine what happens to the alligator population in the long run. the
in

doesthe behavior of P (t) as t increases depend 0 < Po < M or Po > M? 34. If P (t) satisfies the logistic equation in (3), use the chain
(h) How
on whether rule to show
that)

29. uring D

(h) Repeat part

Throughout this period,P (t) remained tion of the initial value problem

pet) (t

(a),exceptwith 150 alligators initially. periodfrom 1790to 1930,the U.S. population


million 3.9 to

PII(t) = 2k

PcP 4M)(p

M).)

years) grew from

closeto

123.2 million.
the

solu-

dP
dt (a)

2 = 0.03135P 0.0001489p = 3.9. , P(O)

What

dict?
(h) What

1930population doesthis logistic equation prelimiting

(c) Hasthis logistic equation


in

population

doesit predict?
continued

curately model the U.S. population? [Thisproblem is basedon a computation

since1930to acby Verhulst,

1845usedthe

30.

dict accurately (long after his own death, of course).] A tumor may be regarded as a population of multiplying cells. It is found empirically that the \"birth rate\" of the cellsin a tumor decreases exponentially with time, so that at f3(t) f3oe- (where and f30 are positive constants), and hence)

1790-1840 population data to preU.S. the U.S. population through the year 1930

who

Concludethat P\" > 0 if 0 < P < 4M; P\" = 0 if P = 4M; P\" < 0 if 4M < P < M; and P\" > 0 if P > M. In particular, it follows that any solution curve that crosses line P = 4M has an inflection point the where it crosses that line, and therefore resembles neof o the lower S-shaped curves in Fig. 1.7.3. 35.Considertwo population functions PI (t) and P2(t), both of which satisfy the logistic equation with the same limiting population M but with different values k I and k 2 of the constant k in Eq.(3).Assume that k I < k 2 Which population approaches the most rapidly? You can reason M if geometrically by examining slopefields (especially ap-

successive alues of t. v 36.Tosolvethe two equations


ing

propriate software is available), symbolically the solution given in Eq. (7), or numerically
in

by analyzing by substitut-

f (10)or

the values

of k

and

M, begin by solving the first equation for the x = 50kM and the secondequation for x 2 =

e-

quantity

dP
dt

at = f3oe- P,

P(O)= Po.

Solvethis

initial

value problem for

pet) = Poexp

(1

\037o

-e- )) .
at

for x in Upon equating the two resulting expressions terms of M, you get an equation that is readily solvedfor M. With M now known, either of the original equations is readily solvedfor k. This technique can be used to \"fit\" the logistic equation to any three population values Po, PI, and P2 corresponding to equally spaced times to 0, tI, and t2 2tI

e- 2 .
IOOkM

37. Usethe
to the

31.

Observethat P (t) approaches finite limiting populathe tion Poexp(f3o/a) as t +00. For the tumor of Problem 30, supposethat at time t = 0 6 there are Po = 10cellsand that P (t) is then increasing 5 at the rate of 3 x 10cells per month. After 6 months the tumor has doubled (in sizeand in number of cells). Solve numerically for a, and then find the limiting population of
\037

for U.S. population data (Fig. 1.7.4) the years Solvethe resulting logistic equa1850,1900,and 1950. actual for

method

of Problem 36 to

fit the

logistic equation

38.Fit 39.

tion and compare predicted the and actual populations the years 1990and 2000.

32. erivethe solution D


pet) =
Po +

the tumor.

the logistic equation to the actual U.S.opulation data p Solvethe for (Fig. 1.7.4) the years 1900,1930,and the and resulting logistic equation, then compare predicted actual populations for the years 1980,1990, nd 2000. a Birth and death rates of animal populations typically are

1960.

of the logistic initial value problem P' = kP(M P), P (0) = Po.Make it clearhow your derivation depends on whether 0 < Po < M or Po > M.)

MPo kMt (M Po)e-

instead, they vary periodically with the paspopulation P satisfies the differential equation) not constant; F sage of seasons. ind P (t) if the

dP
dt)))

= (k + bcos2Jrt)P,

Models 85) 1.8 Acceleration-Velocity


where t is in years and k and b arepositive constants. Thus the growth-rate function k + b cos21ft varies periodically about its mean value k. Construct a graph that contrasts the growth of this population with one that has)

r(t) =

the

sameinitial

equation populations

P'= k P (sameconstant
after compare the

value

Po but satisfies the natural k). How would

growth the two

of passage many

years?)

Accel\037ration\037Yelocitr...
\037\037cl.\037l\037)

In Section we discussed vertical motion of a massm near the surfaceof the earth If under the influenceof constant gravitational acceleration. we neglect any effects of air resistance, Newton'ssecondlaw (F == ma)impliesthat the velocityv of then the massm satisfies the equation) dv mdt
==
\037

1.2

G ')

(1))

where FG == -mgis the (downward-directed) of gravity, where the gravitaforce tional acceleration is g 9.8 2 (in mks units; g 32 ft/s 2 in fps units).) m/s
\037 \037
v
_ \037 .. \037 \"\"_ \037 ........\" ..... ...\037 _................ \"',,\"n\"

Example

1)

bolt Supposethat a crossbow is shot straight upward from the ground initial velocity Vo == 49 (m/s).Then Eq. (1)with g == 9.8 gives)

..

'A\"

'n\".',

\037,.\"....\"

...\"

'n_\"

\"\"

\"\"......

......

'A''''''

............
,.\".............. ...... ................

\037

.......\037...\"\"

'\"\"

...........

(Yo == 0) with

dv
dt)

==

-9.8, so
function

vet) ==

+ + -(9.8)t Vo == -(9.8)t 49.


by)

Hence bolt's the height


y(t) =

yet) is given

+ f [-(9.8)t 49]dt = -(4.9)t+ 49t +

Yo

2 = -(4.9)t 49t.) +

The bolt reaches maximum height when v == -(9.8)t 49 == 0,hence when its + t == 5 (s).Thus its maximum height is)
Ymax

The bolt returns secondsaloft. The force

2 (m).) -(4.9)(5+ (49)(5)== 122.5 ) to the ground when y == -(4.9)t(t 10) == 0,and thus
== y(5) ==

FR exertedby air resistanceon the so Eq. (1), now)


dv mdt)

Now we want

.) in to take account of air resistance a problem like Example 1.


moving

after

10

mass m

must

be addedin

== FG

+ FR.

(2))

Newton showedin his PrincipiaMathematica that certain simplephysicalassumptions imply that FR is proportional to the square of the velocity: FR == kv 2 But on of empirical investigationsindicate that the actual dependence air resistance veit can be quite complicated. or many purposes sufficesto assumethat) F locity

FR

== kv P

,)

where 1 < p < 2 and the value of k dependson the size and shape of the body, as well as the density and viscosity of the air. Generally speaking, == 1 for relatively))) p

86

1 Chapter

First-OrderDifferential Equations)
low speedsand p
But how
that

determine the value of the coefficientk. Thus air resistanceis a complicated physical phenomenon. But the simplifyw ing assumption that FR is exactly of the form given here, ith either p = 1 or p = 2, yields a tractable mathematical modelthat exhibits the most important qualitative features of motion with resistance.)

= 2 for high speeds, whereas1 < p < 2 for intermediate speeds. slow\"low speed\"nd how fast \"high speed\"re dependon the samefactors a a

Resistance roportional Velocity to P


Y
I

Let us
R)

m)

(Note: FR acts upward when the body is falling.) INet force F = FR + FG)

first considerthe vertical motion of a body with mass m near the surface of the earth, subjectto two forces: downward gravitational force FG and a force a FR of air resistancethat is proportional to velocity (sothat p = 1) and of course directedoppositethe direction of motion of the body. If we set up a coordinate systemwith the positive y-direction upward and with y = 0 at ground level, then

FG
\037)

= -mgand)

FG)

FR = -kv,)

(3))

Ground level)

FIGURE
with air

resistance.)

1.8.1. Vertical

motion

= dyjdt is the velocity of the body.Note that the minus signin Eq. (3)makesFR positive (an upward force) if the body is falling (v is negative) and makesFR negative (a downward force) if the body is rising (v is the net force acting on the body is then) positive).As indicated in Fig.
where k is a positive constant and v

1.8.1,

F = FR + FG = -kv
and

mg,)

Newton'slaw of motion F = m(dv/dt) yieldsthe equation)


dv m-= -kv -mg.
dt)

Thus)

\037)

dv
dt)

= -pv g,

(4))

where p

should verify for yourself that if the positive y-axiswere directeddownward, then Eq.(4) would take the form dv/dt = -pv + g. first-orderdifferential equation, and its solution is) Equation (4) is a separable
v(t)

= k jm > O. You

(v

+
\037

) e-

pt

\037

.)

(5))

Here, a = v (0)is the initial V

velocity of the body.Note that)


v,

t\037oo

lim vet)

= g.
P)

(6))

Thus the speedof a body falling with air resistancedoes not increasendefinitely; i a instead,it approaches finite limiting speed,or terminalspeed,)
Iv

,-- - -.
g
mg
k)
I

(7))))

Models 87) 1.8 Acceleration-Velocity


This fact is
the
what

occasional survival of peoplewho fall planes. We now rewrite Eq. (5)in the form)

makes a parachute a practical invention;


without

it even helps explain parachutes from high-flying air-

-dy
dt)

_ (va v,)e-pt + V,.

(8))

Integration gives)

We substitute

we find

that

C = Yo

0 for t

+ (va
yet)

1 P = --(vo v,)e-t + v,t + C. p and let Yo = y(O) denote the initial height

yet)

of the body. Thus

v, )Ip,and so)

1 = Yo + V,t + -(va
p)

- v,)(l-e- ).
pt

(9))

Equations (8)and (9)give the velocity v and height y of a body moving verT tically under the influence of gravity and air resistance. he formulas dependon the initial height Yo of the body, its initial velocity Va, and the drag coefficientp, the constant such that the accelerationdue to air resistance aR = -pV. The two is alsoinvolve the terminal velocity v, defined in Eq. (6). equations For a persondescending with the aid of a parachute, a typical value of p is which correspondso a terminal speedof lv, 21.3 or about 14.5 i/h. t ftls, m With an unbuttoned overcoat flapping in the wind in placeof a parachute,an unlucky which gives a terminal speed skydiver might increase to perhapsas much as 0.5, p of v, 65 ftl s, about 44 mi/h. See for Problems10 11 someparachute-jump and computations.)

1.5,
I I

\037

\037

Example2)

again considera bolt shot straight upward with initial velocity Va = 49 m/s from a crossbow ground level. But now we take air resistance at into account, with in p = 0.04 Eq. (4). We ask how the resulting maximum height and time aloft comparewith the values found in Example1.)
We We substitute
obtain) t / 25 = 294e- 245, t / 25 yet) = 7350 245t 7350eYo

...........

..........

......

...........

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

........

.....

...........

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

Solution

= 0,Va = 49, and


vet)

v,

-g/p= -245in Eqs. (5)and (9),and

- -

.)

To find the time requiredfor the bolt to reach its maximum height (when we solve the equation t / 25 245 = 0 vet) = 294efor t m meters (as opposedto meters without the bolt strikesthe ground, we must solve the equation) yet)

= 0),
\037

108.280

= 25 In (294/245) 4.558 Its maximum (s).


\037

122.5

height is then Ymax = v(t m ) air resistance). find when To


t / 25

= 7350

- 245t - 7350e- =

O.)

and UsingNewton'smethod, we can begin with the initial guessto = 10 carry out the iteration t n +l = t n to generate successive y(tn)ly' (tn ) approximationsto the root.Or we can simply use the Solve command on a calculator or computer. We)))

88

Chapter
1

First-OrderDifferential Equations)
the bolt is in the air for tf 9.411 seconds(as opposedto 10 seconds air resistance). hits the ground with a reducedspeedof IV(tf)1 It 43.227 m/s (asopposedto its initial velocity of 49 m/s). Thus the effect of air resistance to decrease bolt'saximum height, the is m the total time spent aloft, and its final impact speed. Note also that the bolt now spends more time in descent(tf t m 4.853 than in ascent(tm 4.558 s) s).
find that without
\037 \037

\037

\037

.)

Resistance roportional Square Velocity to of P


Now we assumethat the force of air resistanceis proportional to the squareof the velocity:)
\037)

2 FR = ::I::kv

,)

(10))

with k O. The choicef signs here dependson the direction of motion, which o the force of resistancealways Taking the positive y-direction as upward, FR 0 for upward motion (when v 0) while FR 0 for downward motion (when v 0). Thus the sign of FR is always oppositethat of v, so we can rewrite

>

<

<

opposes. >

>

Eq. (10)as)

FR = -kvlvl.)
Then Newton'ssecondlaw gives
dv m-= FG + F
dt) that is,)
R

( 10'))

= -mg

kvlvl;

dv

where p = k/m separately.)


UPWARD

> O. We must

(11)) dt -g pvlvl,) o discussthe cases f upward and downward motion

the initial position Yo with the differential equation)

MOTION: Supposethat a projectileis launched straight upward from initial velocity Vo > O. Then Eq. (11) with v > 0 gives

= g
\037
\037

- - P = -g (1+ ) .
v

In Problem we familiar integral)

13

ask you to make the


1

substitution

:
v

(12))
apply the

= v -Jpig and

to derive the

projectile's velocity function

V(f)=

Because tan u du = f

!!
In
I

f l+u2 du = tan-

+c

tan(C,-f,Jlii)with
I

Cl = tan- 1

(v0

/r) .

(13)

cosu + C, a second integration (seeProblem 14)


1

yieldsthe position function)

cos(C t-JfJi) 1 yet) = Yo + -In cosC p


1)

(14))))

Models 89) 1.8 Acceleration-Velocity


DOWNWARD

Eq. (11)

straight

downward from the initial position Yo with with v 0 gives the differential equation)

MOTION: Suppose that a projectile is launched (or dropped) initial velocity Vo < O. Then

<

dv

dt
In Problem
integral)

P 2 =-g+pv =-g(1- 2) g
V

.)

(15))
apply
the

15we ask you

to make the 1

substitution

= v v'pjg and

= f l-u2 du tanhto derive the v(t)

+c

projectile's velocity function) =

f!

tanh

(C2
I

- t..[jig)
I

with)

C2 = tanh- 1

(vo

ID
.)

(16))

Becausetanh u du = In coshu + C,another integration f


position function) yet) = Yo

(Problem 16) yields the


(17))

-p
1

In

cosh(C2 t y'Pg) coshC


2)

(Note the analogy betweenEqs. (16)and (17)and Eqs. (13)and (14)for motion.) If Vo = 0,then C2 = 0,so vet) = y' gj tanh (t y'Pg). Because) p

upward

lim tanh
x\037oo

lim x = x\037oo

sinh x

coshx

lim
x\037oo

! - ee2 (eX
X

)
))

!(e +

= 1,

it

follows that speed)

in

the

caseof downward
lv,
I

t motion the body approacheshe terminal

f!)
with

(18))

with Vi = gjp in the case downward motion of (ascompared described Eq. (4)).) by
I I
\"
OA ....... .. \037 '\" .\037

linearresistance

Example3)

consideroncemore a bolt shot straight upward with initial velocity Vo = 49 mls from a crossbow ground level, as in Example But now we assume resistance 2. air at In in Eqs.(12) and (15). to the square of the velocity,with p = 0.0011 proportional Problems17and 18 ask you to verify the entries in the last line of the following we
We

table.)
,
,., j!i\037\";\\[\".!.\037.,.'.\037i\037;.'I;!f ..'.:.;.:.:.::: ',--,,-'--------

.,

];.;.'.!J:,]trll!_\037lgl!.!.!;;\037;;.!;.iti.,;.i:\"i[.;i.;:j'Iil\037.<';.;.;.,:

..c;::;,jA;seel1t
):']1\037i\037\037(sJ

Descent
Til11\037

Impact
Speed\037f\037s)

11\037*iillfJiilj:iiilj!\037\037!lillllli!'il'ii!'il\037iiiiii\\II;I\0371I\037\\il\\i

(s)

0.0
(0.04)v
2 (0.0011)v

122.5

108.28

10 9.41
9.41)

4.56
4.61)

108.47

4.85 4.80)

49 43.23
43.49)))

90

Chapter
1

First-OrderDifferential Equations)
120 100
80 60 40
20)
y)

Without
resistance)

1 2 3 4 5 6 7 8 9 10

t)

and resistance), Example 2 (with linear air resistance), Example 3 (with air resistance proportional to the square of the velocity) are all plotted. The graphs of the latter two are visually indistinguishable.)

FIGURE 1.8.2. height The

functions

in

Example 1 (without air

the Comparisonof the last two lines of data here indicateslittle difference-for motion of our crossbow bolt-between air resistanceand air resistanceprolinear where portional to the squareof the velocity. And in Fig.1.8.2, the corresponding height functions are graphed,the difference is hardly visible. However, the differencebetweenlinear and nonlinear resistancecan be significant in more complex situations-such for instance, the atmospheric reentry and descentof a space as,

vehicle.

.)

Variable Gravitational Acceleration)


Unlessa projectilevertical motion remains in the immediate vicinity of the earth's in surface, the gravitational acceleration acting on it is not constant. According to Newton'slaw of gravitation, the gravitational force of attraction betweentwo point massesM and m locatedat a distancer apart is given by)
\037)

F= GMm r
2) \037

(19))

where G is a certain empirical constant (G 6.6726 10- N.(m/kg)2 mks x 11 in units).The formula is alsovalid if either or both of the two massesare homogeneous in t betweenthe centersof the spheres. spheres; this case,he distancer is measured The following exampleis similar to Example2 in Section but now we take account of lunar gravity.)

1.2,

lander is free-falling toward the moon, and at an altitude of 53 kilometers above the lunar surface its downward velocity is measured 1477 kmIh. Its retroat when fired in free space, of a deceleration T = 4 mls 2 At what rockets, provide b height above the lunar surface should the retrorockets e activated to ensurea \"soft
A lunar

touchdown\"

(v

= 0at impact)?)

Solution Let r(t) denote the lander'sdistance from the center of the moon at time t When (Fig.1.8.3). we combinethe (constant) thrust acceleration T and the (negwe ative) lunar acceleration/m = GM/r2 of Eq. (19), get the (acceleration) F
differential equation)

- - -,
d2r =T dt 2 GM

2)

(20))

where M
R

22 = 7.35X 10 (kg) is the mass of the moon, which has a radius of 6 meters = 1.74 10 x kIn, (or 1740 a little over a quarter of the earth's radius).)))

Models 91) 1.8 Acceleration-Velocity


Noting
that this

variable t, we substitute)

second-order differential equation doesnot involve the independent

dr =-, r - - dr dr dt)

-----.--vd2 dt 2
dv dv dv

dt

dt

dr)

(as in Eq. (36)of Section and obtain the first-order equation) 1.6)
dv v- = T - GM r dr
2)

FIGURE

lander descendingto the surfaceof the moon.)

1.8.3. lunar The

with

the new independent variable

r. Integration with

respectto r now yields the

equation)

1 GM -v2 =Tr+-+C 2
r)

(21))

that

we can apply both before ignition (T = 0 ) and after ignition (T of T

= 4).

Beforegnition:Substitution i

= 0 in (21)gives the equation)


(21a))

1 -v2 = GM + C r) 2

where the constant is given by Cl = v5/2

- GM/ro

with)

Vo

m kIn = -1477 x 1000 x

1h
3600 s)

m 14770

km

36

s)

and ro

= (1.74 106) 53,000 1.79310m (from the initial = x x 6 +


of T

velocity-position
into

measurement).) After ignition:Substitution


gIves)

= 4 and v = 0,r = R (at touchdown)

1 - 2 = 4r + GM + C2 2
v
r)

(21)

(21b))
the values v

r = R at touchdown.)
At

where the constant C2 =

-4R

-GM/ is obtained
R

by

substituting

= 0,

the instant of ignition the lunar lander'sposition and velocity satisfy both and Therefore we can find its desiredheight h above the lunar surface (21a) This gives r = at ignition by equating the right-hand sides in (21a)and x 6 and finally h = r R = C2) = meters (that is, 41.87 \037(Cl over miles).Moreover, substitution of this value of r in (21a) kilometers-just .))) the velocity v = mls at the instant of ignition. gives

(21b). 1.78187 10 26

(21b). 41,870

-450

92

Chapter First-OrderDifferential Equations)


1

Escape Velocity
Ir

ml\037) r(t))

In his novel From the Earth to the Moon Jules Verne raised the question of the initial velocity necesaryfor a projectilefired from the surface of the earth to reach the moon.Similarly, we can ask what initial velocity Va is necessary for the projectile escaperom the earth altogether. This will be so if its velocity to f v = drldt remains positive for all t so it continues foreverto move away from the earth. With r(t) denoting the projectile's distancefrom the earth'scenter at time t (Fig. we have the equation)

(1865),

> 0,

1.8.4),

dv

dt
A FIGURE1.8.4.mass m at a great

2 = d r = GM 2

dt

--;:2')

(22))

distance from

the earth.)

24 X similar to Eq. (20),but with T = 0 (no thrust) and with M = 5.975 10 (kg) 6 = 6.378 10 X denoting the massof the earth, which has an equatorial radius of R = v(dvldr) as in Example 4 Substitution of the chain rule expression (m). dvldt
gIves)

dv v- dr

Then integration of both sideswith

r2 respectto r yields)

-GM

1 -v2 2

=-+c.
GM
r)

Now v
v2

= Va and r = R when t = 0,so C = !v5 GMIR,and hencesolution for


2 v 2 = va

1 R

gives)

+ 2GM r (

- )
')

.)

(23))

This implicit solution of Eq. (22)determineshe projectile's t velocity v as a function of its distancer from the earth'scenter.n particular,) I
v2

2 > va

- 2GM
R

so v will remain positiveprovided that v5 from the earth is given by)


Va

> 2GMI . Therefore, the escapeelocity R v


.)

j2\037M

(24))

In Problem 27 we ask you to show that, if the

earth.
Va
\037

-J2GMIR,
With

11,180

initial projectile's velocity exceeds from the 00 as t so it does, ndeed, \"escape\" 00, i the given values of G and the earth'smassM and radius R, this gives mi/h).) (m/s)(about 36,680 about 6.95 ils,about 25,000 ft/s, m then

r (t)

\037

\037

(24)gives the escapeelocity for any other (spherical) v when we use its mass and radius. For instance, when we use the planetary body massM and radius R for the moon given in Example we find that escapeelocity 4, v from the lunar surface is Va 2375m/s. This is just over one-fifth of the escape velocity from the earth'ssurface, a fact that greatly facilitates the return trip (\"From .))) the Moonto the Earth\.")
\037

Remark: Equation

Models 93) 1.8 Acceleration-Velocity

Problems) 11I1
ferencebetween 250 kmjh and the velocity of this sports car. If this machine can accelerate from rest to 100 kmjh in lOs,how long will it take for the car to accelerate from rest to 200kmjh? 2.Supposethat a body moves through a resisting medium with resistanceproportional to its velocity v, so that dvjdt = -kv. (a) Show that its velocity and position at time t are given by
v(t)
and)

1.Theacceleration a Maserati is proportional of

to the dif-

8.Rework
thrust

9.A motorboat
Then)

both parts of Problem 7, with the soledifference 2 n d that the decelerationue to air resistanceow is (0.001)v S2 when the car'svelocity is v feet per second. ftj

weighs 32,000Ib and its motor provides a i of 5000lb. Assume that the water resistances 100 of pounds for eachfoot per second the speedv of the boat.

dv = 1000- 5000- 100v. dt

= voe-

kt)

3.

4.

that (b) Conclude the body travels only a finite distance, and find that distance. Supposethat a motorboat is moving at 40 ftj s when its motor suddenly quits, and that s later the boat has slowedto 20 ftj Assume, as in Problem 2, that the resistanceit encounters while coasting is proportional to its velocity. How far will the boat coastin all? Consider a body that moves horizontally through a medium whoseresistance proportional to the squareof is 2 Show that the velocity v, sothat dvjdt

xU) = xo + ( ) (1

;Vo vokt)

e-kt ).

10.

s.

10

If the boat starts from rest, what is the maximum velocity that it can attain? A woman bails out of an airplane at an altitude of 10,000 ft, falls freely for 20 s, then opensher parachute. How long will it take her to reach the ground? Assume linwithout the ear air resistance ftj s2, taking p = pv with the parachute. (Suggestion: parachute and p = First determine her height abovethe ground and velocity

1.5

0.15

11.ccording A

when the parachute

opens.)

= -kv .

v(t)
and that)

1+ 1

survived when his parachute failed to by openbut provided someresistance flapping unopened in the wind. Allegedly he hit the ground at 100milh after of Testthe accuracy this account. (Suggesfalling for 8 tion: Find p in Eq.(4) by assuming a terminal velocity of the mijh. Then calculate time required to fall 1200

a training jump from 1200ft

to a newspaperaccount, a paratrooper

s.

100

x(t) = Xo + k In(l + vokt).


--* in contrast with the result of Problem 2, Which offers lessresistance when the +00 as t --* body is moving fairly slowly-themedium in this problem or the one in Problem 2? Doesyour answer seem consistent with the observed behaviors of as t --* oo? Assuming resistance proportional to the square of the velocity (as in Problem 4), how far doesthe motorboat of Problem 3 coastin the first minute after its motor quits? Assume that a body moving with velocity v encounters resistance the form dvjdt of -kV 3/2 Show that Note that,

o 12.t is proposedto disposef nuclear I

ft.)

+00.

x(t)

wastes-indrums with 640 Ib and volume 8 ft 3 dropping them weight W Theforceequation for a drum into the ocean(vo falling through water is)

= 0).

-by

5.
6.

x(t)

dv m-=-W+B+F , dt)
R

v(t)
and that

4vo

(kt0JO + 2))

xU) = xo + \0370JO 1 _ k

) kt0JO+2

13.Separatevariables in Eq.(12)and 14.ntegrate I


obtain the upward-motion Eq. with initial condition

where the buoyant forceB is equal to the weight (at 62.5 3 Ibjft ) of the volume of water displacedby the drum (Archimedes' principle) and FR is the forceof water resistance,found empirically to be 1 Ib for each foot per secondof the velocity of a drum. If the drums are likely to burst upon an impact of more than 75 ftjs, what is the maximum depth to which they can be droppedin the ocean without likelihood of bursting?

to

(13)

v(O) = vo.

substitute u velocity function

= v ,Jpig
given in

Conclude under a -powerresistancea body coasts that beforecoming to a stop. only a finite distance 7. Suppose a car starts from rest, its engine providing an that acceleration f 10 S2,while air resistance o ftj provides 0.1 for ft/ S2 of deceleration each foot per secondof the car's velocity. (a) Find the car'smaximum possible (limiting) velocity. (b) Find how long it takes the car to attain 90%
\037

15.eparatevariables S
to obtain

to obtain the the velocity function in Eq. function given in Eq. (14)with upward-motion position initial condition y (0) Yo.

(13)

16.

of its
so.)

limiting

velocity, and how far it travels while doing

substitute u v pig the downward-motion velocity function given in Eq. with initial condition v(O) Vo. to obtain the Integrate the velocity function in Eq. with downward-motion position function given in Eq. initial condition y(O)
in

a Eq.(15) nd

= ,J
(17)

(16)

(16)

Yo.)))

94

Chapter
1

First-OrderDifferential Equations)

17.Considerthe

18. 19.

m. dropped (vo = 0) from a height of Yo = 108.47 Then useEqs.(16)nd (17)o show that it hits the ground about a t 4.80s later with an impact speed about 43.49 of mjs. A motorboat starts from rest (initial velocity v (0) = Vo = of 0). Its motor provides a constant acceleration 4 ftj s2, but water resistance causesa deceleration v 2 400ftj S2 of
Find v when t
\037

crossbowbolt of Example 3, shot straight 0) at time t 0 with initial upward from the ground (y 49 Takeg velocity Vo mjs2 and p in Eq. Then use Eqs. and to show that the bolt reaches maximum height of about 108.47 in its m about Continuing Problem supposethat the bolt is now

(12).

= mjs.

= = 9.8
(13)

= 0.0011

on earth

(14)

27.

4.61. s

17,

in your legsto jump 4 feet straight up while wearing your spacesuit, can you blast off from this asteroid using leg power alone? (a) Supposea projectileis launched vertically from the su rfacer R of the earth with initial velocity Vo R y'2GMj so v5 k2 R where k2 2GM.Then solve the differential equation drjdt kjy'r (from Eq.(23) \037 00as in this section) that explicitly to deduce t \037

have enough spring

= j

00.
Vo

r(t)

i (b) If the projectle is lau nched vertically


locity

with

initial

ve-

> y'2GMjR,deducethat dr =
dt

as t
initial

of 160ftj s.It experiencesoth the decelb of gravity and deceleration 2j800 due to air resisv tance.How high in the air doesit 21.f a ball is projectedupward fromgo? ground with initial I the 2 velocity Vo and resistance proportional to v , deducefrom
velocity
eration

20. n arrow A

boat)

velocity +00 (that is, the maximum possiblepeed the of s

= lOs, and

also find

.
28.(a) Supposethat
=
Why

the limiting

/k2 +a > \037. r y'r


r(t)

is shot

straight upward from the ground

with an

\037 00as t \037 oo? again follow that a body is dropped(vo 0) from a distance ro > R from the earth's center, so its acceleration is dvjdt s -GMjr2 Ignoring air resistance,how that it reachesthe height r < ro at time

doesit

Eq.(14)that the maximum height it attains


Ymax

22. upposethat S
in

23.

25.

parachute open. If he jumps from an altitude of ft and opens his parachute immediately, what will be his terminal speed? How long will it take him to reachthe ground? that Suppose the paratrooperof Problem 22 falls freely for 30s with p 0.00075 beforeopening his parachute. How long will it now take him to reachthe ground? 24. Themass of the sun is 329,320times that of the earth and its radius is 109times the radius of the earth. (a) Towhat radius (in meters) would the earth have to be compressed in order for it to become black a escapeelocity v from its surface equal to the velocity c 3 x 8 mj s of (b) Repeatpart (a) with the sun in place of the light? earth. (a) Show that if a projectileis launched straight upward from the surface of the earth w ith initial velocity Vo less than escape velocity y'2GMjR,then the maximum distance from the centerof the earth attained by the projectile

f Eq.(15)or a paratrooper falling

2p g p = 0.075(in fps units, with g

-(
1
In

1+

PV5 with

is

VWM

\037 Jrro -r 2 + ro cos-rr . ( )


I

v\037

. )

= 32 ftjs2)

10,000

29.

to evaluate ro (Suggestion: Substitute r from y'rj(ro r)dr.) (b) If a body is dropped a height of 1000km abovethe earth's surface and air resistance is neglected, ow long does it take to fall and with what h will it strike the earth's surface? speed that is Suppose a projectile fired straight upward from the surfaceof the earth with initial velocity Vo < y'2GMj . R Then its height y (t) abovethe surface satisfies the initial

2 cos ()

value problem)

d 2y dt
2)

GM (y + R)2
dvjdt
2

')

y(O) = 0,

y' (0) = Vo.) to obtain

Substitute

= v(dvjdy)
v

hole-the = 10

= v5

and then integrate

2GMy R(R + y)

30.

IS)

r max
where

- 2GM_

2GMR
2 Rv o

for the velocity v of the projectile height y. What maxiat mum altitude does reachif its initial velocity is 1 kmjs? it In Jules Verne's original problem, the projectile launched from the surface of the earth is attracted by both the earth and the moon, so its distance r (t) from the centerof the earth satisfiesthe initial value problem

'
of the
earth,

and R

are the mass and


what initial

radius

re-

26.Supposethat
equal to that

velocity Vo must such a be projectile launched to yield a maximum altitude of 100 kilometers abovethe surfaceof the earth? (c) Find the maximum distance from the centerof the earth, expressed in terms of earth radii, attained by a projectile launched from the surface of the earth with 90%of escape velocity. spectively.

(b) With

failed-on an

are stranded-your rocket engine has asteroid of diameter 3 miles, with density of the earth with radius 3960miles. If you)
you

GMm GMe , + , r (0) = R, r (0) = Vo r2 (S r)2 where Me and Mm denote the massesof the earth and the moon, respectively; R is the radius of the earth and S = 384,400km is the distancebetween the centersof the earth and the moon. Toreachthe moon, the projectile must only just passthe point between the moon and earth where its net accelerationanishes.Thereafter it is \"under v the control\" of the moon, and falls from there to the lunar

-- -d 2r dt 2

for the

surface.Find the minimal launch to projectile make it \"From

velocity Vo that suffices the Earth to the Moon.\)

Models 95) 1.8 Acceleration-Velocity

1.8Application) RocketPropulsion)
y)

tv)

F
f)

blastsoff straight upward from the surfaceof Supposethat the rocketof Fig. the earth at time t = O. We want to calculate its height y and velocity v = dyjdt at time The rocketis propelledby exhaust gasesthat exit (rearward) with constant of speed c (relative to the rocket).Because the combustion of its fuel, the mass m = m (t) of the rocketis variable. To derive the equation of motion of the rocket,we useNewton'ssecondlaw in the form)

1.8.5

t.

III
\037

dP -=F
dt)

(1))

c)

FIGURE
rocket.)

1.8.5. ascending An

where P is momentum (the product of massand velocity) and F denotes external net force (gravity, air resistance, tc.).the massm of the rocketis constant so m' (t) = If e O-whenits rocketsare turned off or burned out, for instance-then (1)gives Eq.

F= d(mv) dt
which
(with

dv dm dv =m-+-v=m-' dt dt dt

second law. dvjdt = a) is the more familiar form F = ma of Newton's here m is not constant. Supposem changesto m + tlm and v to v + L\\ v during the short time interval from t to t + tlt.Then the change in the momentum of the rocket itself is
But

tlP
But the
with

\037

(m

+ tlm)(v +

\037v)

system also includesthe mass -\037m and approximate velocity


the time interval tlt
is)

= m tlv + v tlm + tlm exhaust gasesexpelled during this


mv

tlv.) time interval,


in

v-c. Hencethe total change

mo-

mentum during

tlP

\037

(m tlv m tlv

Now we divide by tlt and of met). The substitution continuity

+ v tlm + tlm tlv) + (-tlm)(v c) + c tlm + tlm tlv. take the limit as tlt 0, so tlm 0,assuming
\037 \037

yieldsthe rocket ropulsion p equation

of the resulting expression dPjdt in (1) for (2)

dv dm m-+c-= F. dt dt

If F = FG + FR , where FG = -mgis a constant force of gravity and FR = -kv is a force of air resistance proportional to velocity, then Eq. (2)finally gives
dv dm m-+ c-= -mg- kv. dt
dt)

(3)

Constant Thrust)
Now supposethat the the time interval to mI.Thus)

rocketfuel is consumed the constant \"bum rate\" fJ during at which time the massof the rocket decreases mo from [0, l],uring t d
m(O) = mo, met)

= mo

m(tl) = ml,
dm
dt) fJt,)
tl.)))

= -fJ

for t

< tl,

(4))

with burnout

occurring at time

96

Chapter
1

First-OrderDifferential Equations)
PROBLEM
equation)

Substitute the expressions (4)into Eq. (3)to obtain the differential in

(m

- fJt)- +
dv
dt)

kv

= fJe

- - fJt)g.
(mo

(5))

Solvethis linear equation for)


v(t)

= voMk/{J +
\037c

(1

Mk/{J)

(1 ::ok
fJt

Mk/{J),)

(6))

where Vo

= v(O) and)
M

--- - met)
mo mo
t.)

mo)

denotesthe rocket'sractional ass time f m at

No Resistance)
PROBLEM2
to
obtain)

For the

caseof no air resistance, k = 0 in Eq. (5)and integrate set


vet)

= Vo

gt

+ e In

mo mo

.
fJt)

(7))

Because mo
IS)

fJtl

= ml,it follows that

the velocity of the

rocketat burnout (t = tl)

VI

= V(tl) = Vo

gtl

mo + eln-.
ml)

(8))

PROBLEM3

Start

with

Eq. (7)and integrate to obtain)

y(t) =
It follows that the

(vo

+ c)t

e - -gt - -(mo2
1 2
fJ

,Bt) In

mo

mo-,B

(9))

t)

rocket'sltitude a
YI

at burnout

is)

= y(tl) = (vo + e)tl

em - -gt - -In-. 2
1 2
1

mo
ml)

(10))

fJ

The V-2 rocketthat was usedto attack London in World War IIhad massof 12,850 of which 68.5% fuel. This fuel burned uniformly was kg, for 70seconds ith an exhaust velocity of 2 kmIs. Assume it encounters air resisw tance of 1.45 per mls of velocity. Then find the velocity and altitude of the V-2 at N burnout under the assumption that it is launched vertically upward from rest on the PROBLEM4
initial

an

ground.)))

Models 97) 1.8 Acceleration-Velocity


PROBLEM5 Actually, our basicdifferentialequation in (3)applieswithout qualification only when the rocketis already in motion. However, when a rocket is sitting on its launch pad stand and its enginesare turned on initially, it is observed that a certain time interval passesbefore the rocketactually \"blasts off\" and begins to ascend. he reasonis that if v = 0 in (3),then the resulting initial acceleration) T

-=---g
dv

e dm

dt

m dt)

of the rocket may be negative. But the rocketdoes not descendinto the ground; m this just \"sits there\" while (because is decreasing) calculated accelerationincreasesntil it reaches and (thereafter)positive values so the rocket can begin to u 0 ascend.With the notation introduced to describedthe constant-thrust case, how s that the rocketinitially just \"sits there\" if the exhaust velocity e is less than mog/f3, and that the time tB which then elapsesbefore actual blastoff is given by)
it tB

= mog

- f3e.

fJg)

FreeSpace
gravity

where there is neither Supposefinally that the rocketis accelerating in free space, nor resistance, o g = k = O. With g = 0 in Eq. (8)we seethat, as the mass s of the rocketdecreases mo to m I,its increase velocity is) from in
\037

= v}

Vo

= e In

-.
mo mi)

(11))

Note that dependsonly on the exhaust gasspeede and the initial-to-tinal mass ratio molm}, but does dependon the bum rate fJ. For example,if the rocket not = t blastsoff from rest (vo = 0) and e = 5 km/s and molm} 20,hen its velocityat burnout is VI = 5ln 20 15 km/s.Thus if a rocketinitially consistspredominantly of fuel, then it can attain velocities significantly greater than the (relative)velocity of its exhaust gases.)
\037v \037

\037

U\037 \037\037!Ymm___mmm'\037m'\037

\037m'\037'm\037m'mm'

\"m_'\037'm\037_'_\037'\037m_\"__\"

'\"\037\037m\"\"_

____

\"_m'\"m_'_\037\"\037____

_m____\037__._________\037_\037_____\037_)

chapter we have discussedapplications of and solution methodsfor several types of first-orderdifferential equations, including those that are separable (Section linear (Section or exact (Section In Section1.6 also we 1.6). 1.4), 1.5), discussedsubstitution techniques that can sometimes used to transform a given be first-orderdifferential equation into one that is either separable,inear, or exact. l Lest it appearthat thesemethods constitute a \"grab bag\" of specialand unrelated techniques,t is important to note that they are all versions of a singleidea. i Given a differential equation)
In
this

important

f(x,y, y') = 0,)


we attempt to write
it in

(1))

the form)

y dx [G(x, )]

= o.)

(2))))

98

Chapter
1

First-OrderDifferential Equations)
It is precisely obtain the form to
in Eq. (2)that we multiply the terms in Eq. (1)by an appropriate integrating factor (even if all we are doing is separating the variables). But once have found a function G(x,y) such that Eqs.(1)and (2)are equivalent, we a general solution is defined implicitly by means of the equation)

G(x,y) = C)
that

(3))

one obtains by integrating Eq. (2). Given a specificfirst-orderdifferential equation to be solved,we can attack by means of the following steps:)

it

. .

If Isit separable? so, eparatethe variables and integrate (Section 1.4). s Isit linear?That is, can it be written in the form)

dy
dx)

+ P(x)y = Q(x)?

. .

If so, ultiply by the integrating factor p = exp (f P dx) of Section m Isit exact?That is, when the equation is written in the form M dx+N dy

1.5.

is aM/ay = aN/ax(Section 1.6)? If the equation as it stands is not separable,inear, or exact, there a plausible is l substitution that will make it so? For instance, is it homogeneous Section (

= 0,

1.6)
?)

to Many first-order differential equations succumb the line of attack outlined more do not. Because the wide availability of comof puters, numerical techniquesare commonly used to approximate the solutions of differential equations that cannot be solved readily or explicitly by the methods of this chapter. Indeed,most of the solution curves shown in figures in this chapter were plotted using numerical approximations rather than exact solutions.Several numerical methods for the appropriate solution of differential equations will bediscussed Chapter 6.) in

here.Nevertheless, any m

Chapter
Find

1)

Review ProbleIns) 30.Primesdenote derivatives with


respect

tox.)

s general olutionsof the differential equationsin Problems1through


x2

4. 2xy 3 + eX + (3x2y2 + sin y)y' = 0 5. 3y + x4y' = 2xy 6.2xy 2 + x 2y' = y2 7. 2x2y + x 3y' = 1 8.2xy + x 2y' = y2 9.xy' + 2y = 6x2,JY 10. ' = 1 +X2+ y2 +X2y2 y 11.2y' = xy + 3y2 x 12. xy 3 + 2y 4 + (9x2y2+ 8xy3)y' = 0) 6

2.xy2 + 3y2 - y' = 0 3.xy + y2 - x 2y' = 0

1.x3 + 3y - xy' = 0

13. xy 2 + y' = 5x4y2 4 14. 3y' = x 2y _ y3 x 15.' + 3y = 3x2ey 16.y' = x 2 - 2xy + y2


3x

17.eX + ye

3 19. x5y2 + x 3y' = 2y 2 20.xy' + 3y = 3X-3/2

+ (e + 18. x2y - x y' = 2


xy
Y

xeYX)y'

=0

y3

21.x2 - l)y' + (x - l)y = 1 (


Y Y

+ y' = y2 25.2y + (x + l)y' = 3x + 24. 9x2y2


32 X /

22.xy' = 6y + 12x4y2/3 23.e + ycosx + (xe + sinx)y'= 0


3)))

Models 99) 1.8 Acceleration-Velocity

26.9X
27.

/2y4/3

- 12x
I y)

28.y + xy' = 2e2x 29.(2x + l)y' + y = (2x + 1)3/2 30.y' = -Jx +


Each of the differential equations in Problems through 36 is of two different types considered this chapter-separable, in d linear, homogeneous, Bernoulli, exact,etc. Hence, erive gen-)

3 3y + x y4

+ 3xy' = 0

/5y3/2

+ (8X 3/2yI/3

- 15x/ yI/2)y' = 0
65

eral solutions for eachofthese equations then reconcile your results.)


dy 2 31. =3(y+7)x dx dy 33.dx dy 35. dx

31

32. 34. 36.

dy

in two different

ways;

3x2 + 2y 2
4xy 2xy + 2x x2 + 1

-= dy

dx

= xy3

xy

dx

x + 3y 3x y

dY dx

y'Y-y
tan
x)))

Linear Equations of Higher Order)

Second-Order IfIIm I!ltroduction: \037inearEquations)


we investigated first-order differential equations. We now turn to of higher ordern > 2,beginning in this chapter with equations that are equations linear. The general theory of linear differential equations parallelsthe second-order case(n = 2),which we outline in this initial section. Recall that a second-order differential equation in the (unknown) function is one of the form) y(x)

n Chapter 1

G(x,y, y', y\") = O.)

(1))

This differential equation is saidto be linear provided that G is linear in the dependent variable y and its derivatives y' and y\". Thus a linear second-order quation e takes(orcan be written in) the form)
\037)

A(x)y\"

+ B(x)y'+ C(x)y= F(x).)

(2))

Unlessotherwisenoted, we will always assumethat the (known) coefficientfunctions A(x),B(x),C(x), nd F(x)are continuouson someopeninterval I (perhaps a unbounded) on which we wish to solve this differential equation, but we do not require that they be linear functions of x.Thus the differential equation)
eXy\"

+ (cosx)y' (1+ -JX)y = tan- x) +


1

is linear because dependent ariable y and its derivatives y' and the v
early. By contrast, the equations)
y\"

y\"

appear lin-

= yy'

and

y\"

+ 3(y')2+ 4y 3 = 0)

are not linear because roductsand powersof y or its derivativesappear.) p


100)))

2.1Introduction: Second-Order Equations 101) Linear


If the function F(x)on the right-hand side of Eq. (2)vanishes identically on we call Eq. (2)a homogeneous linear equation; otherwise, it is nonhomoF geneous. or example,the second-order equation)

I,then

x2y\" + 2xy' + 3y = cos x)


is nonhomogeneous; associated its homogeneousquation is) e

x2y\" + 2xy' + 3y = O.)


In
\037)

the with l general, homogeneousinear equation associated Eq. (2)is)


A(x)y\"

+ B(x)y'+ C(x)y= O.)

(3))

casethe differential equation in (2)modelsa physical system, the nonhomogeneousterm F(x)frequently correspondso someexternal influenceon the system. t Remark:Note that the meaning of the term \"homogeneous\" for a secondorderlinear differential equation is quite different from its meaning for a first-order differential equation (as in Section1.6). f course,t is not unusual-eithern i i O mathematicsor in the Englishlanguage more generally-for same word to have the
In

different meanings in differentcontexts.)


A
Spring
Mass)

Typical pplication A

Dashpot)

m l
..

f:\

x=o x>o
Equilibrium position)

x(t)

Linear differential equations frequently appearas mathematical modelsof mechancircuits.For example,supposethat a massm is attached '..'. ical systemsand electrical : both to a spring that exertson it a force Fs and to a dashpot (shockabsorber)hat t exertsa force FR on the mass (Fig. Assume that the restoring force Fs of the spring is proportional to the displacement of the mass from its equilibrium x Then) positionand acts oppositeto the direction of displacement.

2.1.1).

FIGURE

2.1.1.massA
system.) F

spring-dashpot

> 0)) so Fs < 0 if x > 0 (springstretched) while Fs > 0 if x < 0 (spring compressed).
Fs = -kx
(with k

m)

We

the

Fs)

II()

assumethat the dashpot force FR is proportional to the velocity v = dxjdt of massand acts oppositeto the direction of motion. Then)

R)

x,

> 0)

FIGURE Directions f the o forces acting on m.)

2.1.2.

FR = -cv =
so FR

dx -cdt)

(with

c > 0)

< 0 if v > 0 (motion to the right) while FR > 0 if v < 0 (motion to the left). If FRand Fs are the only forcesacting on the massm and its resulting acceleration is a = dvjdt, then Newton'slaw F = ma gives)
mx\" that is,)

= Fs + FR

;)

(4))

\037)

dx dx m-+c-+kx = O. dt
2
dt)

(5))

Thus we have a differential equation satisfied by the position function x(t) of the massm.This homogeneousecond-order equationgovernsthe freevibrations linear s of the mass;we will return to this problemin detail in Section 2.4.)))

102

2 Chapter Linear Equationsof HigherOrder)


must F(t)-which

If,

in

addition to
then

FsandFR,

the

be added to the

mass m is acted on by an external force side in Eq. (4)-the resulting right-hand


==

equation is)
\037)

d2x dx m\037 + dt) + kx dt

c-

F(t).

(6))

vibrations of This nonhomogeneous linear differential equation governs the forced the massunder the influenceof the external force F (t).)

Second-Order Linear Homogeneous Equations


Considerthe general second-order equation) linear
A(x)y\"

+ B(x)y'+ C(x)y== F(x),)


it in

(7))

eachterm
\037)

where the coefficientfunctions A, B,C,and F are continuous on the openinterval Here assumein addition that A (x) i= 0 at eachpoint of I, so we can divide we

I.

in

Eq. (7)by A(x) and write


y\"

the form)
(8))

+ p(x)y' + q(x)y == f(x).) + p(x)y' + q(x)y == O.)

We will
\037)

discussfirst the associated homogeneousquation) e


y\"

(9))

useful property of this homogeneouslinear equation is the fact that the sum of any two solutions of Eq. (9)is again a solution, as is any constant multiple of a solution. This is the central idea of the following theorem.)
A particularly

1 THEOREM
Let Yl and
interval
Y2

for o PrinciplefSuperpositionHomogeneous Equations


the linear combination)
C2Y2)

I. If Cl and C2are constants, then


y

be two solutions of the homogeneousinear equation in (9)on the l


== CIYl

(10))

is also a solution of Eq. (9)on I.)

operation of differentiation, which gives)


Y

Proof:The conclusionfollows almost immediately from the linearity


, == , CIYl , + C2Y2 an d
\"

of the

== CIYl\"

+ C2Y2.)
\"

Then)

+ C2Y2)\" + P(CIYl+ C2Y2)'+ q(CIYl + C2Y2) == + + q(CIYl + C2Y2) + C2Y;)+ , , == Cl(Yl + PYI + qYl ) + C2 Y2 + PY2 + qY2 ) ( == Cl .0 + C2. 0 == 0) because and Y2 are solutions.Thus Y == ClYl + C2Y2 is also a solution. Yl
y\"

+ PY' + qy

== (CIYl
(CIY\037'

p(CIY\037

C2Y\037)

\"

\"

.)))

2.1Introduction: Second-Order Equations 103) Linear


:N \"\" ............__\037 \037 \037...

Example

1)

We

can seeby inspection that)


Yl

x (x) = cos

and

Y2

(x) = sin x)

are two solutions of the equation)


\"

+y= 0

.)

Theorem 1 tellsus

that any

linear combinationof thesesolutions, such as)

y(x) = 3Yl(X)

2Y2(X)

= 3cosx 2sinx,)

is also a solution. We will seelater that, conversely,every solution of y\" + y = 0 is a linear combination of thesetwo particular solutions Yl and Y2. Thus a general solution of y\" + y = 0 is given by) y(x) = Cl cosx+ C2sinx.)
this singleformula for the general solution encoma \"twofold infinity\" of particular solutions, because two coefficientsCl the passes and C2can be selected illustrate someof independently.Figures through the possibilities, ith either Cl or C2 set equal to zero, r with both nonzero. w o

It is important to understand that

2.1.3

2.1.5
y

.)

\037

8 6 4 2 0

\037

-2
-4

8 6 4 2 0

\037

-2
-4
0
x) 31:

-6 -8 -31:
FIGURE
y(x)

231:

-6 -8 -31:

0
x)

31:

231:

-2 -4 -6 -8 -10 -31:

10 8 6 4 2 0

= 6 cos 2 sin x x = 3 cos + 4 sin x x

/)

0)

31:
X)

231:)

331:)

= Cl cosx y\" + y = o.) of

2.1.3. Solutions

FIGURE2.1.4. Solutions y(x) = C2sin x of y\" + y = o.)

FIGURE Solutions of y\" + y = 0 with Cl and C2 both


nonzero.)

2.1.5.

Earlier in this sectionwe gave the linear equation mx\" + CX' + kx = F(t) as a mathematical modelof the motion of the massshown in Fig. Physicalconsiderations suggestthat the motion of the mass should be determined by its initial valuesof x(0)and x'(0), positionand initial velocity. Hence, given any preassigned Eq. (6) ought to have a unique solution satisfying these initial conditions. More generally, in order to be a \"good\" mathematicalmodel of a deterministicphysical situation, a differential equation must have unique solutions satisfying any approa t priate initial conditions.The following existencend uniquenessheorem (proved in the Appendix)gives us this assurance the general second-order for equation.)))

2.1.1.

104
3 2 1
;;..\"

2 Chapter Linear Equationsof HigherOrder)


THEOREM2

Existence Uniqueness LinearEquations and for

containing the point

Supposethat the functions p,q, and


y\"

a.Then, given any

I are continuous on the open interval I


two numbers bo and bl , the
equation)

+ p(x)y' + q(x)y = I(x))

(8))

-1)

has a unique (that

-2 -1

satisfiesthe

initial

is,one and only one)solution on the entire interval conditions)


yea) = bo, y' (a) = bl
.)

I that
(11))

2
x)

FIGURE2.1.6. Solutions of y\" + 3y' + 2y = 0 with the same initial value y (0) = 1 but different initial slopes.)
5 4 3 2 1 0
y(O)

=3 = 0)

y(O)

;;..\"

-2 -3
-4
-\0372

-1
-1) o 1
x)

234 2.1.6
\037,...
... '\" \037 .... \037...\"

FIGURE2.1.7. Solutions of y\" + 3y' + 2y = 0 with the same initial slopey' (0) = 1 but different
initial

values.)

tions in (8) are continuous. Recall from Section that a nonlinear differential equation generally has a unique solution on only a smaller interval. Remark Whereas afirst-orderdifferential equation dy/dx = F(x,y) admits only a single solution curve y = y(x) passingthrough a given generally initial point (a,b), Theorem 2 impliesthat the second-order equation in (8) has solution curves passingthrough the point (a, bo)-namely, ne for o infinitely many each(real number) value of the initial slopey' (a) = b l That is, instead of there being only one line through (a,bo) tangent to a solution curve, every nonvertical (a,bo) is tangent to some solution curve of Eq. (8). Figure straight line through showsa number of solution curves of the equation y\" + 3y' + 2y = 0 all while Fig.2.1.7 showsa number of solution having the sameinitial value y (0) = curves all having the same initial slopey' (0) = The application at the end of this sectionsuggestshow to construct such families of solution curves for a given s linear homogeneous econd-order differential equation.

Remark constitute a secondEquation (8) and the conditions in (11) order linear initial value problem. heorem 2 tells us that any such initial value T problemhas a unique solution on the whole interval I where the coefficientfunc-

1:

1.3

2:

1,

1.
\"'.........\"\" \"'...

.)

Example
Continued)

1)

saw in the first part of Example1 that y (x) = 3 cos 2 sin x is a solution (on x the entire real line) of y\" + y = O. It has the initial values y (0) = 3, y' (0) = Theorem 2 tells us that this is the only solution with these initial values. More generally, the solution
We

\037\037\"\"\037\"\"\037\"\",.. \037\"\"\037...\"\"\"\"\037...

\"\"\037\"\".\037 \"'\" \"'\"....\"\"\037\"\"..........,,\037\"\"\037....\037

\037\"\"\"'-\037\"\"\037\"'\037\"\"\037\"\"\"\"'\037\"\"\"'-\"'''

\037\"\"\" ...\"\037\",,,.....,............

....

....\"\"\",

,\"\",\"\",\"\"\"\"......\"'....\"

\",...\"\",,,,,,,,,,\037,,,,,

...\"\"\",.....,,,\"\"\"'...

''''',''''\037....,......''''',..\",.....\"......,.....

\"'...

\",.....

._

\"'......

......u

\037\"\"

'\"

\037....-u

....\"

...

...,.....

....

....,.........'.,....v\"',...\"\",\"\"......-...c<,...

...'\"

\"\"''''''''\"'''\037''''.....-..

'\"

...

-2.

y(x) = bo cosx+ b l sin x satisfiesthe arbitrary initial conditions y(O) = bo, y' (0) = b l ; this illustrates the existencef such a solution, alsoas guaranteed by Theorem 2. o

linear Example1 suggestshow, given a homogeneoussecond-order equation, we might actually find the solution y(x) whoseexistence assuredby Theorem 2. is we First, we find two \"essentiallydifferent\" solutions YI and Y2; second, attempt to imposeon the general solution)

the

initial conditions yea) simultaneous equations)

= ClYI + C2Y2) bo, y'(a) = b l . That is, we attempt


y

(12))
to solve the)

ClYI
ClY\037

(a) + c2Y2(a)= bo, (a) + c2y\037(a) = b


i)

(13))

for the coefficientsCI and

C2.)))

2.1Introduction: Second-Order Equations Linear


Example2)
Verify that

105)

the

functions) x YI(x)==e

and

Y2(x)==xe
==

x)

are solutionsof the differential equation


y\"

- 2y' +
initial x

0,

and

then find

a solution satisfying the

conditions y(O) == 3, y' (0)== 1.)

Solution The verificationis routine; we omit it. We imposethe given initial conditionson the
general solution) 10
8 6 4 2 0

y(x) == cle + C2xex,)

for which)

y'(x) == (CI +c2)e+C2xex,


X

to obtain the simultaneous equations)

\037

y(O) == CI The resulting solution is Cl == 3, C2 == value problemis


-1) o
x)
1)

== 3,

-2 -4 -6 -8 -10 -2)

y'(O)== CI + C2 == 1.)

-2.Hencethe solution of the original


x

initial

2)

FIGURE
y\" initial

solutions y (x)

3e + C2 eX of X 2y' + y = 0 with the same


value y(O) = 3.)

2.1.8.x Different =

Figure

sameinitial value y(O) ==


In
must

2.1.8 several additional solutions of shows


3.

y(x) == 3e

- 2xe .
x

y\"

- 2y' +

==

0,all having

the

.)

orderfor the procedureof Example2 to succeed, two solutionsYI and the have the elusive property that the equations in (13)can always be solvedfor CI and C2, no matter what the initial conditions bo and b l might be.The following definition tells precisely how different the two functions Yl and Y2 must be.)
Y2

DEFINITION

LinearIndependence Two Functions of

Twofunctions definedon an openinterval I are saidto be linearlyindependent on I provided that neither is a constant multiple of the other.) Twofunctions are saidto be linearly dependent n an openinterval provided o are not linearly independent there; that is, one of them is a constant multiple of the other. We can always determine whether two given functions and g are o whether either of the two linearly dependent n an interval I by noting at a glance or gj is a constant-valuedfunction on I.) quotients f jg
that they

Example3)

Thus it is clearhat the following pairsof functions are linearly independenton the t entire real line:)
sIn x and
eX eX

and and and and

x+l
x

xex . , x2. ,
Ix
I.)))

e-2x . ,

cos x;

106

2 Chapter

Linear Equationsof HigherOrder)


That is, neither sin xIcos = tan x nor cos I sin x = cot x is a constant-valued x x 2x 2x function; neither eX le- = e3x nor e- lex is a constant-valued function; and so forth. But the identically zero function I(x) = 0 and any other function g are 0 linearly dependenton every interval, because g(x) = 0 = I(x). Also, the functions

I(x)= sin2x

and

are linearly dependent n any interval o metric identity sin 2x = 2 sin xcos x.

g(x) = sin x cosx because (x) = 2g(x)is the familiar trigonoI

.)

General olutions S
does the homogeneous quation y\" + PY' + qy = 0 always have two linearly e s Theorem 2 saysyes! We needonly choosel and Y2 so that) independent olutions? Y
But
Yl

(a) =

1,

Y\037

(a) = 0

and

Y2(a) = 0, y\037(a)

= 1.)

is then impossiblethat either Yl = kY2 or Y2 = kYl because . 0 i= 1 for any k constant k. Theorem 2 tells us that two such linearly independent solutions exist; actually finding them is a crucial matter that we will discussbriefly at the end of
It
this

and section, in greater detail beginning


We want to

in

Section 2.3.

show, finally,

that

and

Y2

of the homogeneousquation) e
y\"(x)

given any two linearly independent solutions Yl

+ p(x)y'(x)+ q(x)y(x)= 0,)


= CIYl + C2Y2)

(9))

as every solution y of Eq. (9)can be expressed a linear combination)


y

(12))

s is ( Y2. This means that the function in (12) a general olution ofEq. 9)-it providesall possiblesolutions of the differential equation. As suggested the equations in (13), determination of the constants Cl the by and C2 in (12)dependson a certain 2 x 2 determinant of values of Yl, Y2, and their derivatives. Given two functions and g, the Wronskian of and g is the of Yl and
determinant)

\037)

w= I g = Ig'
I')
g')

- I'g.
\302\267

We write

two functions or the point

either W(/, g) or W(x), depending whether we wish to emphasize the on x at which the Wronskian is to be evaluated.For example,)

W(cosx,sIn x)
and)

=. -

cos SIn x = cosx + sIn2 x = 1 x 2 sIn x cosx)


X

X W(eX,xe ) = ex e)

xe = e2x eX + xe
X
x)

.)

Theseare examplesof linearly independent pairsof solutions of differential equations (see t Examples1 and 2). Note that in both caseshe Wronskian is everywhere
nonzero.)))

2.1Introduction: Linear Second-Order Equations


On the other hand, if the functions then) kg (for example), W(f, g) =
kg

107)

f and g are linearly dependent,with f =

kg'

g'

g = kgg' kg'g = O.)

Thus the Wronskian of two linearly dependentfunctions is identically zero.In Section we will prove that, if the two functions Yl and Y2 are solutions of a 2.2 s linear homogeneous econd-order equation, then the strong conversestated in part (b) of Theorem 3 holds.)

Solutions) THEOREM3Wronskiansof ( equation(]3q.9)))


of s linear SupposetliatYlandY2aretwo.solutionsthe homogeneousecond-order

y\"+ p(x)y' + q(x)y = 0

on allopen.Interval on which p and q are continuous. d pendent, then W (Yl, Y2) = 0on (a)i.I\037;)Il\0374\0372areli\037early 7 a t 0 (b)'lfYlandYi{re linearly independent,hen W (YI, Y2) 1:- at each point of I.)

I.

Wronskian

Thus, given two solutions of Eq. (9), there are just two possibilities:The W is identically zero if the solutions are linearly dependent;he Wront skian is never zero if the solutions are linearly independent. he latter fact is what T we needto show that Y = ClYl + C2Y2 is the general solution of Eq. (9)if Yl and Y2 are linearly independent solutions.) THEOREAA4. of Solutions Homogeneous General Equations

solutionsof the homogeneousequation LetYlandY2betwolinearlyindependent

(]3q.(9)))

y\"+p(x)y'+ q(x)y = 0)

the withjJan.dqcontinuouson open interval I. If Y is any solutionwhatsoever of Eq. 9) 1 ( on ,thenthere existnumbers Cl and C2such that)

Y(x) = CIYl(X)
for.....allix..........inil...)

+ C2Y2(X))

that when we have found two linearly independent solutions of the second-order e homogeneousquation in (9),then we have found allof its solutions. e therefore call the linear combination f = ClYl + C2Y2 W a general olution of the differential equation. s

In

Theorem 4 tells us essence,

a ProofofTheorem 4: Choose point


equations)

CIY\037

I, and considerthe simultaneous CIYl (a) + c2Y2(a)= f (a), (14)) = f'ea).) (a) +
a of
c2y\037(a)

The determinant of the coefficients in this system of linear equations in the unknowns Cl and C2 is simply the Wronskian W (Yl, Y2) evaluated at x =

a.

By)))

108

2 Chapter Linear Equationsof HigherOrder)


Theorem 3,this determinant is nonzero, so by elementaryalgebra it follows that the equations in (14)can be solved for Cl and C2. With these values of Cl and C2, we define the solution)

G(x) = CIYl(X)

+ C2Y2(X))

of Eq. (9);then)
and)

G(a) == CIYl (a) + c2Y2(a)== yea))

G'(a)==

CIY\037

(a) + c2y\037(a) = Y'(a).

Thus the two solutions Y and G have the sameinitial values at likewise, o do Y' s and G'.By the uniqueness a solution determined by such initial values (Theorem of 2),it follows that Y and G agreeon Thus we seethat)

a;

I.

Y(x) = G(x) = CIYl (x) + C2Y2(X),) as desired.)

Example4)

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

.)
Y2

If Yl (x) = e2x and


Y\037'

......

......

(x) = e-2x , then)


and Y;
2x = (-2)( 2)e- = 4e- = 4Y2.) _ 2x

2x = (2)(2)e = 4e2x = 4Yl


Y2

Therefore, Yl and

are linearly independent solutions of)


y\"

- 4y =
d

O.)

(15))

But

Y3(X)

= cosh2xand Y4(X) = sinh 2xare also solutionsof Eq. (15), because) = \037(cosh2x) dx
d2

= -(2sinh2x) 4 cosh2x dx)

and, similarly, (sinh 2x)\" = 4 sinh 2x.It therefore follows from Theorem 4 that the functions cosh and sinh 2xcan be expressed linear combinationsof Yl (x) == 2x as e2x and Y2(X) = e-2x Of course,his is no surprise,because) t

cosh2x = 4e2x+ 4e-2x and

sinh

2x== 4e2x

- 4e-2x)
.)
in

and by the definitionsof the hyperbolic cosine hyperbolic sine.)

Remark: Because2x , e-2x and coshx, sinh x are two different pairs of e Theorem 4 linearly independent solutions of the equation y\" 4y == 0 in (15),

impliesthat every particular solution Y (x) of this equation can be written both
the form)

Y(x) == cle
and
in

2x

+ C2e-2x)

the form)

x Y(x) == a cosh + b sinhx.


Thus these two different linear combinations (with arbitrary constant coefficients) o provide two different descriptions f the set of all solutions of the samedifferential 4Y == O. Hence eachof these two linear combinations is a general equation y\" solution of the equation. Indeed,this is why it is accurate to refer to a specificsuch linear combination as \"a general solution\" rather than as \"the general solution.\" .)))

2.1Introduction: Second-Order Equations 109) Linear Linear Second-Order with Coefficients Equations Constant
As an
illustration

s linear homogeneous econd-order differential equation)


\037)

of the general theory introduced in


ay\"

this

section,we discussthe
(16))

+ by' + cy == 0)

with

and

constant coefficientsa, b, and beginwith the observation that)

c.We first
and

lookfor a singlesolution of Eq. (16)


== r 2erx
,)

(erx ), == rerx

(erx )\"

(17))

so any derivative of erx is a constant multiple of erx . Hence,if we substituted rx then each term would be a constant multiple of erx , with the y == e in Eq. (16), constant coefficientsdependent n r and the coefficientsa, b, and c.This suggests o that we try to find a value of r so that thesemultiples of erx will have sum zero.f I we succeed, y == erx will be a solution ofEq.(16). then For example, f we substitute y == erx in the equation) i

we obtain)
Thus)

- 5y' + 6y e - 5re + 6e 0; (r - 2) (r - 3)e (r - 5r + 6)e


y\"

== 0,)

r2

rx

rx

rx == o.)

rx ==

rx ==

o.

be a or 3. So,in searchingfor a singlesolution, we actually have found two solutions: I (x) == e2x and Y2 (x) == e3x . Y To carry out this procedurein the general case,we substitute y == erx in With Eq. (16). the aid of the equations in (17),we find the result to be)

Hence y

== erx will

solution if either r == 2

r ==

ar2erx + brerx + cerx

== o.)
will

erx Because is never zero, e concludethat y(x) == erx w


\037)

equation in (16)preciselywhen r is a root of the algebraicquation) e

satisfy the

differential

ar2 + br + c == o.)

(18))

This quadratic equation is calledthe linear differential equation)


\037)

characteristic equation of the homogeneous


(16))

ay\"

+ by' + cy == o.)

If Eq. (18)has two distinct (unequal) roots rl and r2, then the correspondingolus tions YI (x) == er1x and Y2(X) == er2x of (16)are linearly independent. (Why?) This gives the following result.) THEOREM5
then) r1x y(x) == cle +C2er2x)

DistinctReal Roots)
distinct,

If the roots rl and r2 of the characteristic equation in (18)are real and

\037)

(19))

is a general solution of Eq. (16).)))

110

2 Chapter

Linear Equationsof HigherOrder)


Find the general solution of)
2y\"

Example5)

- 7y' + 3y

== o.)

Solution

We

can solve the characteristic equation)

by

factoring:)

(2r - 1)(r 3)
x

2r2 7r + 3 == 0)
== O.

The roots rl == 1 and r2 == 3 are real and distinct, so Theorem 5 yieldsthe general
solution)

y(x) == Cle /
\037

+ C2e3x

.)

.)

'\"

\"

\037\037

\037

.u\"\"

Example6)

The differential equation y\"

+ 2y' == 0 has characteristic equation)


r2

3 2
1
\037

with

distinct real roots rl

+ 2r == r(r + 2) == 0) == 0 and r2 == 2. Because x = 1, get the general eO. we

solution)

y(x) == Cl + C2e

2x
.)

-1
-2 -0.5
0

showsseveral different solution curves with CI == 1,all appearing to Figure +00. approach the solution curve y(x) = 1 (with C2 == 0) as x
\037

2.1.9

Remark: Note that


0.5
x)

1.5 2

equation into
only the

Theorem 5 changesa problem involving a differential one involving only the solution of an algebraic equation.

.
.)

FIGURE2.1.9. Solutions o y(x) = 1 + C2e-2xf y\" + 2y' = 0 with different values of C2.)

If the characteristic equation in (18)has equal roots rl == r2, we get (at first) The single solution Yl (x) == ertx of Eq. (16). problem in this caseis to the \"missing\" secondsolution of the differential equation. produce A doubleroot r == r} will occur preciselywhen the characteristic equation is a constant multiple of the equation)

(r rl)2 == r 2 2rlr + rf
Any

== O.)

differential equation with

this
\"

' - 2rly+r1y==.) 0
2
that y ==
that)

characteristic equation is equivalent to)


(20))

is easy to verify by direct substitution ( Eq. (20).It is clearbut you should verify)
But
it
Yl

xertx is a secondsolution

of

(x) == er1x

and

Y2

(x) == xertx)

are linearly independent functions, so the general solution of the differential equation
in

(20)is)

r1x y(x) == cle + c2xerlx.)))

Linear Second-Order Equations 111) 2.1Introduction:


THEOREM 6
then)
\037)

Roots Repeated = r2,


(21))

If the characteristic equation in (18)has equal (necessarily real) roots r}

y(x) = (CI + c2x )er1X)

is a general solution of Eq. (16).)

Example7)

To solve the

initial

value problem)
y\"

y(O) = 5, y' (0) = -3,) we note first


that

+ 2y' + y = 0;

the characteristic equation) r2

+ 2r + 1 = (r + 1)2= 0)
the general solution provided by Theorem6
x

has equal roots r}


IS)

= r2 =

-1. ence H
X

y(x) = cle- + C2xe-x.)

10)

Differentiation yields)
y

'( ) =-Cle + C2e-x -C2 -x -x xe


= C} 5, = -Cl+ C2 = -3,) y'(O)
y(O)

,)

5)

so the
;;::....)

initial

conditions yield the equations)

o)

-5 -3-2-1 0 -4
x)

2 3 4 5
x

which imply
initial

that

C} = 5 and C2

= 2. Thus the desiredparticular solution


x x

of the

FIGURE
y(x)
y\"
==

+ 2y/ + y

Cl

e- + 2xe- of
X

2.1.10. Solutions
==

value problem is)

y(x) = 5e- + 2xe-.

0 with

different

values

of Cl.)

This particular solution, together with several others of the form y(x) x 2xe-, is illustrated in Fig.

2.1.10.

x = cle- +

_
In

2.3.) case complexroots will be discussedin Section of

roots.The The characteristic equation in (18)may have either real or complex

Problems)
1through 16, homogeneoussecond-orderlina 2.y\" -9y 3.y\" + 4y
y/(O) y/(O)
== == == ==

Problems

ear differential

equation, two functions Yl and Y2, and a pair of initial conditions are given. First verify that Yl and Y2 are Then find a particular solutions of the differential equation. solution of the form Y == ClYl + C2Y2 that satisfies the given derivatives with respectto initial conditions. Primes denote

3x 0; YI == e3x , Y2 == e- ; y(O) == y/(O) == 15 0; Yl == cos2x, == sin2x;y(O) == 3, Y2


==

-1,

4. y\" + 25y

1.

y\"

x.

5. y\"
y/

==

0; Yl

==

eX , Y2

==

e- ; Y (0)
x

==

0, y/ (0) ==

6.y\" + y/

-3y'+2y -

-10

0; Yl
==

==

COS Y2 5x,

==

sin

5x; y(O)

==

10,

0; Yl == eX, Y2 == e2x ; y(O) == 1, /(O) == 0 y 3x 0; YI == e2x , Y2 == e- ; y(O) == 7, 6y


==

5)

(0) == -1)))

112

2 Chapter

Linear Equationsof HigherOrder)


Y2 3x

8.y\" Yl y/(O) = 9.y\" y Yl = y/(O) 10.\" y solutions? y/(O) = 13 11. -2y'+2y = 0; Yl = eX cosx, = eX sinx;y(O) = 0, 32.Let Yl and be two solutions of A (x)y\" + B(x)y/ + y\" y/(O) = 5 C(x)y = 0 on an open interval I where A, B, and C 12.\" + 6y' + 13y = 0; Yl = e- cos2x, = e- sin 2x; y are continuous and A (x) is never zero. (a) Let W = y(O) = 2, y/(O) = 0 W (Yl , Y2). Show 13. 2y\" - 2xy' + 2y = 0; Yl = X, = x 2; y(l) = 3, x y/ ( 1)= 1 dW 14. 2y\" + 2xy' - 6y = 0; Yl = x 2, = x-3; y(2) = 10, x A(x)-= (Yl)(AY2) - (Y2)(AYl). dx) y'(2)= 15 2 15. y\" - xy/ + y = 0; Yl = X, = x lnx; y(l) = 7, x from the original differand Then substitute for y'(I)= 2 ential equation to show 16. 2y\" + xy/ + y = 0; Yl = cos(lnx), = sin(lnx); x
x Y2 x YI 5x Y2 5x Y2 Y2 3x Y2 3x that) Y2
\" \"

x 7. y\" + y/ = 0; Yl = 1, 2 = e- ; y(O) = -2, /(O) = 8 Y y

- 3y' = 0; = 1, = e ; y(O) = 4, = -2 + 2y' + = 0; = e- , = xe- ; y(O) = 2, - = -1 25y = 0; = e , = xe ; y(O) = 3, 10y'+

S 31.how

2 are inthat Yl sinx 2 and Y2 linearly dependent functions, but that their Wronskian vanishes at x O. Why does this imply that there is no differential equation of the form y\" + p(x)y/+ q(x)y 0, with both p and q continuous everywhere, having both Yl and Y2 as

= cosx

Y2

Y2

Ay\037

Ay\037'

that)

y(l) = 2, y'(I) =

Y2

3)

Thefollowing three problems illustrate the fact that the not generally hold for nonlinear position principle does tions.)

superequa-

dW A(x)-= -B(x)W(x). dx)

(b) Solvethis
mula)

first-order

equation

to

deduceAbel'sfor-

17.Show

= 1/ is a solution of y/ + y2 = 0, but that if x y = c/xis not a solution. Show 18. 2 that y = x 3 is3 a solution of yy\" = 6x4, but that if c i= 1,hen y = cx is not a solution. t S 19. how that Yl = 1 and Y2 = -JX are solutions of yy\" + (y/)2 = 0, but that their sum y = Yl + Y2 is not a solution.)
c i= 0 and c i= 1,hen t
that y

W(x)
where
imply

= K exp

( f

A(x)

B(x)dX ,

K is a constant. that the Wronskian

where

(c) Why does Abel's formula W (Yl , Y2) is either zero everyor nonzero everywhere (as stated in Theorem 3)?)

Determine whether

the

through 26 are linearly the real line.

independent 2 cos x 2

pairs of functions in Problems 20 or linearly dependent on

Apply Theorems5 and 6 to find generalsolutions of the differential equations given in Problems 33 through 42. Primes

denotederivatives

20. f(x) = JT, g(x)=


x

x,g(x)= 1 cos2x 25.f(x) = eX sinx,g(x)= eX cosx 26.f(x) = 2cosx 3 sinx,g(x)= 3 cosx 2sinx + 27. Let y p be a particular solution of the nonhomogeneous equation y\" + py/ + qy = f(x) and let Yc be a solu-

21.f(x) =x3,g(x)=x lxl 22. f(x) = 1 +x,g(x)= 1 + Ixl 23.f(x) = xe2, g(x)= Ixle
x

+ sin x

33.y\" - 3y' + 2y = 0
35.y\" + 5y/ = 0
y/ Y

with

respectto

x.)

24. f(x) = sin

37. 2y\"

39.4y\" 41.6y\"

- - =0 + 4y' + = 0 - 7y' - 20y =


y

34. y\" + 2y' 15y= 0 36.y\" + 3y' = 0 2 38. 4y\" + 8y' + 3y = 0 40. 9y\" 12y'+ 4y = 0
0)

42. 35y\"

- - 12y =
y/

0)

tion
y

= Yc + y p is a solution = 1 and
YI

of its associated homogeneous equation. Show that of the given nonhomogeneous


x C2sin Cl COS Yc 27,find a solution of y\" y

Each of Problems 43 through 48 gives a general solution equation y (x) of a homogeneous second-order'differential ay\" + by' + cy = 0 with constant coefficients. Find such an
equation.

28.

equation. With y p
initial

of Problem
that

conditions

29. Show

tions of x 2y\" 4xy' + 6y conditions y(O) 0 y/ Explain why thesefacts do not contradict Theorem2 (with respectto the guaranteed

= x 2 and

(0) =

= = = (0). =

Y2

-1= =

+ = 1 satisfying

in the notation

the

y/

different solu0, both satisfying the initial

x are two

(0). 3

43. y(x) 45. y(x) 46. y(x) 48. y(x)

= Cl + C2e-lOx 44. y(x) = cle + C2e-lOx = cle- + C2xe-lOx = cle + C2elOOx 47. y(x) = CI + C2X xh = eX (c1e + C2e-xh))
lOx lOx lOx

30.

uniqueness x 3 and (a) Show that Yl solutions on the independent 3 x2 3x O.


y\"

).

Y2

y/

tically

zero. Why

real line of the equation (b) Verify that W (Yl , Y2) is idendo these facts not contradict Theorem

Ix

3
1

are linearly

3?)

Problems 49 and 50deal with the solution curves of y\" + 3y/ + and 2y = 0 shown in Figs. 49. Find the highest point on the solution curve with y (0) = 1 and y/ (0) = 6 in Fig. 50.Find the third-quadrant point of intersection of the solution curves shown in Fig. 2.1.7.)))

2.1.6 2.1.7. 2.1.6.

o 2.2General olutions f Linear Equations 113) S


A 51. second-order equationis oneof the form Euler 2

ax

y//

+ bxy' + cy = 0

(22)

concludethat a general solution of the (22)is y(x) = CIX rl + C2Xr2.)


Make the substitution
solutions
v

Euler equation in

where a, b, c are constants. (a) Show that if x > 0, then the substitution v In x transforms Eq. (22) into the constant-coefficientlinear equation d2 a y2 + (b a) dy + cy 0 (23) dv dv

(for x

> 0) of the Euler equations

= In x of Problem 51to find


in

general

Problems

52-

56.)

with

independent

variable

the

characteristicequation

v. (b) If the roots rl and r2 of of Eq.(23)arereal and distinct,)

=0 2 54. 4x y\" + 8xy' 3y = 0 56.x 2y// 3xy' + 4y = 0) 52.x 2y\" + xy/


Y

53. x 2y\" + 2xy' 12y = 0 55. x 2y\" + xy/ = 0)

IIfIJ GeneralSoluti()ns)f Lineargq\037\037!\037_()!!\037) (


We now
tion in generalizes a very

of the form)

show that our discussionin Section of second-orderinear equations 2.1 l differential equalinear natural way to the general nth-order Po(x)y(n)+ Pl(x)y(n-l)

+...Pn-l(x)y'+ +

Pn(X)Y ==

F(x).

(1))

Unlessotherwisenoted, we will always assumethat the coefficientfunctions Pi (X) and F (x) are continuous on someopen interval I (perhapsunbounded) where we wish to solve the equation. Underthe additional assumption that Po(x) i-0 at each point of I, we can divide each term in Eq. (1)by Po(x) to obtain an equation with
leading coefficient of the form)
\037)

1,

yen)

+ PI (x)y(n-l) ... + Pn-l(x)y'+ Pn (x)y == j(x).) + + PI(X)y(n-l)

(2))

The homogeneous linear equation associated Eq. (2)is) with


\037)

yen)

+... +

Pn-I(X)y'+ Pn(x)y == O.)

(3))

Just
tion

as in the second-order a homogeneousnth-order linear differential equacase, has the valuable property that any superposition, linearcombination,of soor lutions of the equation is again a solution. The proof of the following theorem is 2.1.) that the routine verification-as of Theorem 1 of Section essentially same-a
THEOREM 1 of for Principle SuperpositionHomogeneous Equations
n

Let Yl, yz,


the interval
\037)

..., be n solutions of the homogeneousinear equation in (3)on l I. If CI,Cz, ..., C are constants, then the linear combination) == CIYl + czyz + ... + (4))
Yn y
CnYn)

is also a solution of Eq. (3)on I.)


\".v, \037 ........ .................... '\" .-.< .-.' '\" '\"

Example

1)

It

is easyto verify
YI

that

the three functions) and


Y3(X) == sin 2x)))

2x, (x) == e-3x , yz(x) == cos

114

2 Chapter

Linear Equationsof HigherOrder)


are all solutions of the homogeneoushird-order equation t
y(3)

+ 3y\" + 4y' + 12y = 0)


that any

on the entire real line. Theorem 1 tells us solutions,such as)

linear combination of these

y(x) = -3Yl(X) + 3yz(x)

2Y3(X)

3x = -3e- + 3cos2x 2sin2x,)

is also a solution on the entire real line. We will seethat, conversely,every solution of the differential equation of this exampleis a linear combinationof the three particular solutions Yl, yz, and Y3. Thus a general solution is given by)

+czcos2x +c3sin2x.) ExistencendUniqueness Solutions a of


We

y(x) = cle-

3x

.)

saw in Section that a particular solution of a second-order differential 2.1 linear equation is determined by two initial conditions.Similarly, a particular solution of an nth-order linear differential equation is determined by n initial conditions.The following theorem, proved in the Appendix, is the natural generalizationof Theorem 2 of Section 2.1.)

2 THEOREM
interval

and for Existence Uniqueness LinearEquations

Supposethat the functions PI,pz,

nth-order linear equation (Eq.(2)))


6 4
2) \037O)

I containing the point a. Then, given n


yen)

...,
,)

Pn, and

are continuous on the open numbers bo, bI, , bn I, the

... -

+ PI (x)y(n-l)

+ +...

Pn-l(X)y' + Pn(X)Y = I(x))

has a unique (that is,one and only one)solution on the entire interval satisfies the n initial conditions)

I that
(5))

-2
-4
-6) o 2 4
x)

yea) = bo, y' (a) = b I

...

,)

y(n-I)(a)= bn -

1 .)

10

FIGURE2.2.1. particular The


solution y (x)
3x

-3e- + 3 cos2x - 2 sin 2x.)

Equation (2) and the conditions in (5) constitute an nth-order initial value T problem. heorem 2 tells us that any such initial value problemhas a unique solution on the whole interval I where the coefficient functions in (2)are continuous. It tells us nothing, however, about how to find this solution. In Section we will 2.3 seehow to construct explicitsolutions of initial value problemsin the consta\037tcoefficient case occurs that often in applications.)
We

Example
Continued)

1)

saw earlier that)

y(x) = is a solution of)


y(3)

3x _3e- + 3 cos2x - 2 sin 2x)

+ 3y\" + 4y' + 12y = 0

on the entire real line.This particular solutionhas initial values y (0) = 0,y' (0) = 5, and y\" (0) = 39, nd Theorem 2 impliesthat there is no other solution with these a same initial values. Note that its graph (in Fig.2.2.1) periodicon the right. looks 2x 2 sin 2x because the negative exponent, we seethat y (x) 3 cos of Indeed, .))) for large positive x.

\037

o 2.2General olutions f Linear Equations 115) S

Remark:Because general solutioninvolvesthe three arbitrary constants its CI, C2, and C3, the third-order equation in Example1 has a \"threefold infinity\" of solutions,ncluding three families of especially i simplesolutions:)

. . .

3x y(x) = cIe- (obtained from the general solution with 2x y(x) = C2cos (with Cl = C3 = 0),and = C3sin2x(with Cl = C2 = 0).) y(x)

C2 = C3 =

0),

Theorem 2 suggestsa threefold infinity of particular solutionscorreof spondingto independent choices the three initial values y (0) = bo, y' (0) = b l , and y\" (0) = b2. Figures2.2.2 illustrate three corresponding families through 2.2.4 of solutions-for of which, two of thesethree initial valuesare zero. each .)
Alternatively,

5 4 3 2
1
\037

3)

2)

/)

y'(O)

=3

1)

0.8 0.6 0.4 0.2


\037

/)

y\"(0)=3

-1
-2 -3
-4

\037

0)

-1) -2)
o)
1)

'\"
y'(O)

= -3)

-5 -1)

2
x)

5)

-3 -1) -2

-0.2 -0.4 -0.6 -0.8

y\"(O)

= -3)

1
x)

4)

-1-1)

1
x)

345

FIGURE2.2.2. Solutions of y(3) + 3y\" + 4y' + 12y = 0 with y/ (0) = y\" (0) = 0 but with
different values

for y(O).)

FIGURE2.2.3. Solutions of y(3) + 3y\" + 4y' + 12y = 0 with y (0) = y\" (0) = 0 but with
different values for y/ (0).)

different values for y\" (0).)

FIGURE2.2.4. Solutions of y(3) + 3y\" + 4y' + 12y = 0 with y(O) = y/ (0) = 0 but with

Note that Theorem 2 impliesthat the trivial solution y(x) solution of the homogeneousequation)
yen)

= 0 is the only
(3))

+ PI (x)y(n-I) ...+ Pn-l(x)y'+ Pn(x)y = 0) +


initial

that

satisfies the trivial

conditions)

yea) = y'(a)
-

=... =
and

y(n-l) (a) = o.)

Example2)

..

It

is easy to verify

that)

YI

(x) = x 2
x 2y\"

Y2(X)

=x

3)

are two different solutions of)

- 4xy' + 6y =

0,)

that both satisfy the initial conditions y (0) = y' (0) = O. Why does this not contradict the uniqueness art of Theorem 2? It is because leading coefficientin the p this differential equation vanishes at x = 0,so this equation cannot be written in the form ofEq. 3)with coefficientfunctions continuouson an openinterval containing ( the point x = O. .)))

and

116

2 Chapter

Linear Equationsof HigherOrder)

Solutions Linearly Independent


On the basisof our knowledge general solutionsof second-orderinear equations, of l we anticipate that a general solution of the homogeneousnth-order linear equation)
\037)

yen)

+ PI (x)y(n-1) ... + Pn-1 + Pn(X)Y + (x)y'


Y

== 0)

(3))

will
\037)

be a linear combination)
== C1 Y1

+ C2Y2 + ...+ Cn Yn

,)

(4))

where Y1, Y2, , Yn are particular solutions of Eq. (3).But thesen particular solutions must be \"sufficiently independent\" that we can always choose coefficients the C1,C2, , Cn in (4) to satisfy arbitrary initial conditions of the form in (5). The of question is this: What should be meant by independence three or more functions? Recall that two functions 11 12 linearly dependentf one is a constant and i are == 2 == multiple of the other; that is, if either 11 kl or 12 kl1 for someconstant k. If we write theseequations as)

...

...

+ (1)/1 (-k)/2== 0
we seethat the linear dependence of C1and C2 not both zero such that)

== or (k)/1+ (-1)/2 0,)

and 11 12 impliesthat

there existtwo constants

+ c1/1 c2/2== o.)

(6))

and Conversely,if C1and C2are not both zero, then Eq. (6)certainly impliesthat 11 are 12 linearly dependent. In analogy with Eq. (6),we say that n functions 12, In are linearly that somenontrivial linear combination) dependent rovided p

+ c1/1 c2/2+ + cnl


n)

...
...

11, ...,

of them vanishes identically; nontrivial means that not allof the coefficientsCI, C2, , C are zero (although someof them may be zero).) n

...

DEFINITION

LinearDependence ofFunctions

The n functions o 12, , In are saidto be linearly dependent n the interval I provided that there existconstants Cl, C2, , Cn not all zero such that)
C1

11, ...

+ 11 c2/2+ ... + Cn In == 0)

(7))

on

I;that
in

is,)

(x) cl/1 + c2/2(X)+ + cnln(x)== 0)

...

for all x

I.)

If not all the coefficients in Eq. (7)are zero,hen clearly we can solve for at least t one of the functions as a linear combinationof the others, and conversely.Thus the functions 12, , In are linearly dependentif and only if at least one of them is a linear combinationof the others.)))

11, ...

o 2.2General olutions f Linear Equations 117) S

Example3)

The functions) fl (X)

= sin 2x,

x, f2 (x) = sin x cos

and

f3 (x) = eX

are linearly dependent n the real line because) o

+ (l)fI+ (-2)f2 (0)f3= 0)


(by the familiar trigonometric identity sin 2x

= 2 sin x cos x).)

.)

on The n functions fI, f2, , fn are calledlinearly independent the interval I provided that they are not linearly dependenthere. Equivalently, they are linearly t independent on I provided that the identity) C1 fl

...

+ C2f2 + ...+ Cn fn = 0

(7))

holdson I only
that

in

the

trivial

case)
CI = C2 =

... = C

= 0;)

is, no nontrivial linear combination of these functions vanishes on Put yet another way, the functions fI, f2, , fn are linearly independent if no one of them is a linear combinationof the others.(Why?) Sometimesone can show that n given functions are linearly dependent by values of the coefficients so that Eq. (7)holds. 3, finding, as in Example nontrivial But in orderto show that n given functions are linearly independent,we must prove that nontrivial values of the coefficientscannot be found, and this is seldom easy to do in any direct or obvious manner. n of Fortunately, in the case n solutionsof a homogeneousth-orderlinear equao tion, there is a tool that makesthe determinationof their linear dependence r indeThis pendencea routine matter in many examples. tool is the Wronskian determin nant, which we introduced (for the case = 2) in Section Supposethat the n 1 times differentiable.Then their Wronskian functions fI, f2, , fn are each n

...

I.

is the n x n

...

2.1.

determinant)

fl
\037

f2
f\037

fn
f\037

W=

f{

(8)

fI(n-I) fin-I)
We write

f\037n-l))

on , fn) or W(x), depending whether we wish to emphaW(fI, size the functions or the point x at which their Wronskian is to be evaluated. The Wronskian is named after the Polishmathematician J. M. H.Wronski (1778-1853
We saw in Section that the Wronskian of two linearly dependentfunctions 2.1 vanishes identically. Moregenerally, the Wronskian of n linearly dependentfunctions T , fn is identically zero. o prove this, assumethat Eq. (7)holds on f2, W the interval I for somechoicef the constants CI, C2, , C not all zero. e then o n differentiate this equation n 1 times in succession, obtaining the n equations)

f2,...

fl, ...

...

( clf!x) + c2f2(X)

clf{(x)+ c2f\037(x))
CIfI(n-I)(x)
I

+...+)

+...+

cnfn(x)= 0,
cnf\037(x)

= 0,)
(9))

n + c2fi- ) (x) + ...+ Cn

f\037n-l)

(x) =

0,)))

118

2 Chapter

Linear Equationsof HigherOrder)


in f We recallrom linear algebra that a system of n linear homogeneous quations in n unknowns has a nontrivial solution if and only e if the determinant of coefficients vanishes. In Eq. (9) the unknowns are the constants Cl, C2, Cn and the determinant of coefficients is simply the Wronskian W w 12, , In) evaluated at the typical point x of Because e know that the i Ci are not all zero,t follows that W (x) = 0,as we wanted to prove. Therefore, to show that the functions 12,. , In are linearly independent on the interval I, it sufficesto show that their Wronskian is nonzero at just one point of I.)

which hold for all x

I.

(11, ...

...,
... .
\037.

I.

11, ..

Example4)

3x Show that the functions Yl (x) = e- , Example1)are linearly independent.) is)

.....

Y2(X)

cos2x, and

Y3(X))

sin 2x (of)

Solution Their Wronskian W=

e)-3x

cos2x)

sin 2x)

-3e3x 3x ge-

-2sin2x) 2 cos2x)

-4cos2x -4sin
2x) 2x)

2x)

=e)-3x

-2sin 2 cos 2x cos sin + 3e-4cos2x -4sin -4cos2x -4sin


3x)

2x)

2x) 2x)

+ geW Because

3x)

cos2x)

sin 2x

-2sin 2x

2 cos 2x)

3x = 26e- i= O.)

i= 0 everywhere, it follows that Yl, Y2, and on any openinterval (including the entire real line).
......

Y3

are linearly independent

.)

Example5)

Showfirst

that

the three solutions)

Yl(X)=X, Y2(x)=xlnx,
of the third-order equation
x 3 y(3)

and

Y3(X)=X

2)

x 2y\"

+ 2xy'

- 2y =

0)

(10)) a particular solution


( 11))

are linearly independent on the openinterval x of Eq. (10)that satisfies the initial conditions)

> O.Then find


1.)

y(l) = 3, y'(I) = 2, y\"(I) = Solution Note that for x


neouslinear equation of the standard form of the three given solutions, e find that) w
x
in

> 0,we coulddivide each term in (10)by x 3 to obtain a homoge(3).When we compute the Wronskian
x2 x In x 1

W=

+ In x
1
x)))

2x 2

= x.

2.2General olutions f Linear Equations 119) S o


x > O. To find
(
o 11)n)

Thus W (x) i= 0 for x > 0,so Yl, Y2, and Y3 are linearly independenton the interval the desiredparticular solution, we imposethe initial conditionsin

y(x) = ClX

y'(x) = Cl
y\"(x)

= 0)

+ C2X In x) + C3X 2 + c2(1 In x) + 2C3X,) + C2 + 2C3.) x)


,)

+-

This yieldsthe simultaneous equations)

+ C3 = 3, y'(I) = Cl + C2 + 2C3 = 2,) C2 + 2C3 = 1;) y\" (1) =)


we solve to question is)
find

y(I) = Cl

Cl

1,C2 = -3,and C3 = 2. Thus the particular


y(x) = x

solution

in)

- 3x

In

x + 2x2

.)

.)

Provided that W (Yl, Y2, , Yn) i= 0,it turns out (Theorem 4) that we can always find values of the coefficientsin the linear combination)
y

...

= ClYl + C2Y2 + ... + CnYn)

conditions of the form in (5). Theorem 3 provides the of W in the case linearly independent solutions.) of necessary nonvanishing
that

satisfy any given

initial

THEOREM3 Wronskians of Solutions)

Supposethat
equation)

YI, Y2,

...,

Yn

are n solutionsof the homogeneousth-orderlinear n


(3))

y(n)

+ PI (x)y(n-l) ... + Pn-l(X)y' + Pn(x)y = 0 +

on an openinterval

I,where eachPi is continuous. Let) w = W (YI, Y2, . ,

..

Yn).)

are linearly dependent,hen W = 0 on t Yn are linearly independent, then W i= 0at each point of Thus there are just two possibilities: ither W = 0 everywhereon I, or W E everywhere on I.)

(a) If YI, (b) If YI,

..., Y2, ...,


Y2,

Yn

I.

I.
f.=

to

assumethat
Y2,

Proof:We

tions YI,

...

W (a)

have already proven part (a). To prove part (b), it is sufficient = 0 at somepoint of and show this implies that the soluB , Yn are linearly dependent. ut W (a) is simply the determinant

I,

of)))

120

2 Chapter

Linear Equationsof HigherOrder)


coefficientsof the systemof n homogeneousinear equations) l
CIYl
CIY\037

(a) + (a) +)

C2Y2 (a)
c2y\037(a)

+ ... + + +)

...

CnYn(a)
cny\037(a)

= 0, = 0,)
(12))

clyi
in

- ) (a) + c2yi -l)(a) + ... +


1

cny\037n-l)

(a) = 0

the n unknowns Cl, C2, ..., C . Because (a) = 0,the basic fact from linear W n algebraquoted just after (9) impliesthat the equations in (12)have a nontrivial solution.That is, the numbers Cl, C2, ..., Cn are not all zero. We now usethesevalues to define the particular solution)
Y

(x) = C Yl (X) + C2Y2 (X) + ... + Cn Yn (X))


1

(13))
trivial initial

of Eq. (3). The equations

in

= y(n-l)(a)= O. yea) = Y/(a) = Theorem 2 (uniqueness)herefore impliesthat Y(x) = 0 on I. In view of (13) t and the fact that Cl, C2, ..., C are not all zero,his is the desiredconclusion that t n the solutions Yl, Y2, the proof of Yn are linearly dependent. This completes
Theorem 3.

conditions

(12) then

...

imply that

satisfies the

...,

.)

General olutions) S
can now show that, given any fixed set of n linearly independent solutions of a homogeneous th-order equation, every (other) solution of the equation can be n as a linear combination of those n particular solutions. Usingthe fact expressed from Theorem 3 that the Wronskian of n linearly independent solutions is nonzero, the proof of the following theorem is essentially sameas the proof of Theorem the 4 of Section (the case = 2).) 2.1 n
We

THEOREM4

Let Yl,
tion)

Y2,

...,
Yn yen)

of General Solutions Homogeneous Equations

be n linearly independent solutionsof the homogeneousquae

+ Pn-l(X)Y'+ Pn(x)y = 0) + Pl(X)y(n-l) (3)) on an openinterval I where the Pi are continuous. If Y is any solutionwhatsoever (3), ofEq. then there existnumbers Cl, C2, Cn such that) Y(x) = CIYl (x) + C2Y2(X) + + cnYn(x))

+...

..., ...

for all x in I.)

is a linear combination)
\037)

Thus every solution of a homogeneous th-order linear differential equation n


Y

= CIYl + C2Y2 + ...+ CnYn)

of any n given linearly independent solutions.On this basis we call such a linear combination a generalolutionof the differential equation.))) s

Example6)

3x 2x, According to Example4, the particular solutions YI (x) == e- , Y2 (x) == cos == sin 2xof the linear differentialequation y(3) +3y\"+4y' + 12y== 0 are and Y3(X) e N linearly independent. ow Theorem 2 saysthat-givenbo, b I , and b2-therexists a particular solution y(x) satisfying the initial conditions y(O) == bo, y/ (0) == b I , and y\" (0) == b2 Hence Theorem 4 impliesthat this particular solution is a linear combination of YI, Y2, and Y3. That is, there exist coefficients CI, C2, and C3 such
\"

\"\"
.

o 2.2GeneralSolutions f Linear Equations 121)

that)

y(x) == Cle-

differentiation Uponsuccessive thesecoefficients,we needonly solve the three linear equations)

+ C2cos2x+ C3sin 2x. and substitution of x == 0,we discover that


3x

to find

+ -3CI
CI the (See application for this

C2

== b o, == b I , == b 2
.)

- 4C2+ 2C3 9CI

section.))

.)
equation)

Nonhomogeneous Equations
We now
\037)

considerthe nonhomogeneousnth-order linear differential


yen)

+ PI (x)y(n-I) ... + Pn-I(x)y/ + Pn(x)y == I(x)) + + PI (x)y(n-I) ... + Pn-I(x)y/ + Pn(x)y == +


in

(2))

with

associated homogeneousquation) e
yen)
O.)

\037)

(3))

Supposethat a single fixed particular solution yp of the nonhomogeneous equation in (2)is known, and that Y is any other solutionof Eq. (2).If Yc == Y Yp,
then subsitution

of Yc

the differential equation gives (using the

linearity

of differ-

entiation))
Yc

(n)

/ (n-I) + PlYc + . + Pn-IYc+ PnYc ... == [(y(n) + PI yen-I) + + Pn-Iy/ + Pn y] / (n) (n-I) [(yp + PIYp + . + Pn-IYp + PnYp == I(x) I(x)== o.)

..

..

Thus Yc == Y
\037)

Yp

is a solutionof the associated homogeneousquation in (3).Then) e y==Yc+YP')


(14))

and
\037)

it

follows from Theorem 4 that)

where YI, Y2, homoYn are linearly independent solutions of the associated We call Yc a complementaryfunction of the nonhomogeneous geneousquation. e s equation and have thus proved that a general olutionof the nonhomogeneousequation in (2)is the sum of its complementary function Yc and a singleparticular solution Yp of Eq. (2).)))

...,

Yc

== CIYI

+ C2Y2 + ... +

CnYn,)

(15))

122

2 Chapter

Linear Equationsof HigherOrder)


THEOREM 5

of SolutionsNonhomogeneous Equations

Let Yp be a particular solution of the nonhomogeneous equation in (2) on an open interval I where the functions Pi and f are continuous. Let Yl, Y2, ..., Yn be linearly independent olutionsof the associated s homogeneousquation in (3). e If Y is any solution whatsoever of Eq. (2)on I, then there exist numbers CI, C2, ..., Cn suchthat) Y(x) == CIYl (x) + C2Y2(X) +
for all x in I.)

...

+ cnYn(x) + Yp(x)

(16))

Example7)

It

is evident that

Yp

== 3xis a particular solution of the equation)


y\"

+ 4y == 12x,)

(17))

and

that

of Eq. (17)that satisfiesthe

Yc(x) == Cl cas2x+C2sin 2xis its complementary solution. Find a solution


initial

conditions y(O) == 5,y' (0) == 7.)

Solution The general solution of Eq. (17)is)

2x y(x) == CI COS + C2sin 2x+ 3x.)


Now)

Hence initial the

s y' (x) == -2CIin 2x+ 2C2cas2x+ 3. conditions give)

y(O) == Cl

== 5,

y'(O)== 2C2+ 3 == 7.)


We find that Cl == 5 and C2 ==

2.Thus the desiredsolution is)

+ + y(x) == 5 cos2x 2sin2x 3x.)

.)

lem IIf!JProb s)
In

6, show directly that the givenfunctions are linearly dependent on the real line. That is, find a nontrivial linear combination of the given functions that vanishes
identically.

Problems

1through

6.f(x) = eX, g(x)= coshx,hex) = sinh In Problems7 through 12, the Wronskian use
given functions
terval.)

4. f(x) = 17, (x)= 2 sin x,hex) = 3 cos X g 2 5. f(x) = 17, (x)= cos x,hex) = cos2x g
x)

2.f(x) = 5, g(x)= 2 - 3x2, hex) = 10 15x + 2 = 0, g(x)= sin x,hex) = eX 3.f(x) 2 2


to prove that the the indicated in-

1.f(x) = 2x, g(x)= 3x2, hex) = 5x - 8x2

7. f(x) = 1, (x)= x,hex) = x 2; the real line g 8. f(x) = eX, g(x)= e2x , hex) = e3x ; the real line 9. f(x) = eX, g(x)= cosx,ex) = sin x;the real line h 10.(x) = eX, g(x)= x-2, hex) = x-2 1n x;x > 0 f 11.(x) = x,g(x)= xe , hex) = x2ex ; the real line f x 12.(x) = x,g(x)= cos(1nx), = sin(1nx); > 0) hex) f
X

Problems through 20, a third-order homogeneous linear equation and three linearly independent solutions are given. Find a particular solution satisfying the given initial conditions.)
In

13

are linearly

independent

on

13.y(3) + 2y\" - y/ - 2y = 0; y(O) = 1, /(O) = 2, y\"(O) = 0; y -2x -x


Yl

= e , Y2 = e
X

, Y3 =

e)))

2.2GeneralSolutions f Linear Equations 123) o


14.(3) - 6y\" + IIy'- 6y = 0; y(O) = 0, y'(O) = 0, y = eX , = e , Y3 = e3x y\" (0) = 3; 15.(3) - 3y\" + 3y' - y = 0; y(O) = 2, y' (0) = 0, y\" (0) = 0; y = XeX , Y3 = X 2eX Y = eX , 16.(3) -5y\"+8y'-4y = 0; y(O) = 1, '(O)= 4, y\"(O) = 0; y y = eX , = e , Y3 = X e 17.y(3) + 9y' = 0; y(O) = 3, y'(O)= -1,\"(O) = 2; = 1, y = cos3x, Y3 = sin 3x 18.(3) -3y\"+4y'-2y= 0; y(O) = 1, '(O)= 0, y\"(O) = 0; y y = eX, = eX cosx, 3 = eX sinx. Y 19. 3y(3) - 3x2y\" + 6xy' - 6y = 0; y(l) = 6, y'(I) = 14, x 2 3 y\"(I) = 22; = x, = x , Y3 = x 3 2 - 4y =2 0; y(l) = 1,y'(I) = 5, 20. y(3) + 6x y\" + 4xy' x 2
YI Y2 2x I Y2 YI Y2 2x 2x YI Y2 YI Y2 YI Y2

30.Verify

that YI

= x and
x 2y\"

Y2

= x 2 are linearly
2y

solutions on the entire

real line of the

independent equation)

- 2xy' +

= 0,)

31. his problem indicates T


conditions equation.

but that W (x, x 2) vanishes at x O.Why do these observations not contradict part (b) of Theorem 3?
why we can impose only n initial on a solution of an nth-order linear differential

(a) Given
y

\",

the equation)
,)

+ py + qy = 0

y\"(I)

= -11; = x,Y2 = x- , Y3 = xYI

lnx)

In

21through 24, a nonhomogeneous differential a complementary solution Yo and a particular soequation, lution y p are given. Find a solution satisfying the given initial
Problems
Yc

explain why the value of y\"(a) is determined by the values of y (a) and y' (b) Prove that the equation)

(a).

y\"

- 2y' -

5y

= 0) = C)
differential

conditions.)

2y' + 2y = 2x; (O) = 4, y' (0) = 8; y = CIeX cosx + C2exsin x;y p = x + 1 25.LetLy = y\" + py' + qy. Supposethat YI and

24. y\"
Yc

22.y\" - 4y = 12;- = 0,YP (0) = 10; y(O) y' e ; YP = -3 Yc = cle + C2 23.y\" - 2y' - 3y = 6; y(O) = 3, y'(O)= 11; e3x Yc = cle-+ C2 ; YP = -2 2x 2x x

21.\" + y = 3x; y(O) = 2, y'(O)= -2; y


= CI cosx C2sin x; = 3x +

has a solution

satisfying

the conditions) and


y\"(O)

y(O) =
if and only if

1,

y'(O)= 0,

C = 5.
homogeneous linear

32.) Prove that

an nth-order

Y2

are two

equation satisfying the hypotheses of Theorem2 has n lin, Yn. (Suggestion: early independent solutions YI, Y2, Let Yi be the unique solution such that)

...

functions

such that)

(i
Yi

-I)(a-) 1 )

and

Yi(k)

(a) = 0

if k

i=

i-I.))

26.(a) Find

LYI = I(x) and LY2 = g(x).) Show that their sum y = YI + Y2 satisfiesthe nonhomogeneous equation Ly = I(x)+ g(x).

33.Supposethat

by inspection particular homogeneousequations

solutions

of the

two non-

the three numbers rl, r2, and r3 are distinct. Show that the three functions exp(rlx), xp(r2x), e and exp(r3x)are linearly independent by showing that their Wronskian)

y\"+2y=4
27.
lution

and

y\"+2y=6x.
y\"

1
W = exp[(rl + r2 + r3)x]. rl

1
r2 2 r2

1)

(b) Usethe method of Problem 25 to


differential equation Prove directly that the functions

of the

a particular so+ 2y = 6x + 4.
find

r3)

r2
I

r2
3)

II(x)= 1,
are linearly

12(X) = x,

and

13(X) = x

2)

is nonzero for

all

x.
that the Vandermonde determinant)

34. Assume as known

28.

independent on the whole real line. (Suggestion: Assume that CI + C2X + C3X2 O.Differentiate this from the equations you get equation twice, and conclude that CI C2 C3 Generalizehe method of Problem27 to prove directly that t the functions)

1
r2 2 r2

1 rn
2 rn

= = = 0.)

v=

rl r2
I

10(x) = 1, II(x)= x, 12(x) = X 2, In (X) = X are linearly independent on the real line. 29.Usethe result of Problem 28 and the definition of linear to independence prove directly that, for any constant r, the
n)

...,

r ln-I is nonzero if

r n-I
2)

r n-I
n)

functions)

the numbers rl , r2, , rn aredistinct. Prove the method of Problem 33 that the functions) by

...

lo(x)= erx
are linearly

,)

II(x) = xe

rx
,)

...

,)

In

(x)

= xne

rx)

-) fi(x) = exp(rix ),) 1 < i < n are linearly


independent.)))

independent

on the whole

real line.)

124
35.According
W (YI

2 Chapter

Linear Equationsof HigherOrder)

to Problem 32 of Section the Wronskian , Y2) of two solutions of the second-order equation

2.1,

37. Before applying Eq. (19)with


second-order linear

is given

by

+ PI (x)y/ + P2(X)Y = 0 Abel's' formula s


y\"

W(x)
for

= K exp

someconstant K. It can be shown that the Wronskian of n solutions YI, Y2, , Yn of the nth-order equation

(f

PI (x)dx

...

a given homogeneous differential equation and a known solution YI the equation must first be written in the form of with leading coefficient 1 in order to correctly determine the coefficient function P Frequently it is v (x) YI (x) in more convenient to simply substitute Y the given differential equation and then proceeddirectly to find v Thus, starting with the readily verified solution YI (x) x 3 of the equation)

(x), (18) (x).

(x).

y(n)

+ PI (X)y(n-I)

+...Pn-I +
of the
YI) Y2)

satisfies the sameidentity. as follows: (a) The derivative of a determinant of functions is the sum of the determinants obtained by separately

(X)y/ + Pn(X)Y = 0 Prove this for the casen = 3

x 2y\"

- 5xy' + 9y = 0
x, =

(x

> 0),)

substitute Y vx 3 and deducethat x v\" + v/ O.Thence solvefor vex) C In and thereby obtain (with C the secondsolution Y2 (x) x 3 In x.)

= 1)

cludethat)

differentiating

the rows

original determinant.
Y3)

Con-

In

w/

=
Y\037 Y\037 Y\037

eachof Problems 38 through 42, a differential equation and onesolution YI are given. Usethe method of reduction oforder as in Problem 37 to find a secondlinearly independent solution
Y2.

Y1

(3)

(3)
Y2

Y3

(3)

(b) Substitute

for y(3)

Y?), yi ), and

yj3)

from the equation

+ PI y\" + P2Y/ + P3Y = 0,

36.

and then show that W/ PI W. Integration now gives Abel's formula. of the homogeneous Supposethat one solution YI second-order linear differential equation)

(x)

(x > 0); YI (x) = x 3 xI2 YI (x) = e 40. x 2y\" x(x+ 2)y'+ (x + 2)y = 0 (x > 0); YI (x) = x (x + l)y\" (x + 2)y' + Y = 0 (x > YI(X) = eX 2 =0 < x < 1); YI (x) = x 42. (1 x )y\" + 2xy' 2y 43. First note that YI (x) = x is one solution of Legendre's equation of order 1,)

38.x 2y\" + xy/ - 9y = 0 39.4y\" - 4y' + Y = 0;

41.

(-1

-1);
=

y\"

+ p(x)y/+ q(x)y = 0)

(18))

(1
secondsolution)
Y2(X)

- x - 2xy' +
2
)y\"

2y

O.)

is known (on

an interval

The method of reduction of order consists functions). of substituting Y2(X) and attempting V(X)YI (x) in to determine the function v (x) so that Y2 (x) is a second After substituting linearly independent solution of use the fact that YI (x) is a Y V(X)YI (x) in Eq. solution to deduce that)

I where P and

q are continuous

Then use the method of reduction of order to derive

the

(18)

(18),

(18).

-In +x 1-2 I-x)


x 1

(for-l< x < 1).


=
is one
\037,)

44. First

YI v\"

(2y\037

+ PYI )v/ = o.)

(19))

X -1/2 cosx verify by substitution that YI (x) solution (for x > 0) of Bessel'squation of order e

If YI (x) is known, then (19) a separable is equation that is readily solvedfor the derivative v/ (x) of v(x). Integration of v/ (x) then gives the desired(nonconstant) function
v (x).)

x 2y\"

+ xy/ + (x2

\037)y

O.)

Then derive by reduction I/2 sinx.) Y2(X) = X-

of order the secondsolution

I-I0mogeneous Eq1Jatios with C()nstantCoeffic\037ents) ll


In

equation is a linear combination of n linearly independentparticular solutions, but we said little about how actually to find even a singlesolution. The solution of a linear differential equation with variable coefficientsordinarily requiresnumerical methods (Chapter 6) or infinite series ethods (Chapter 3). But we can now show m how to find, explicitly and in a rather straightforward way, n linearly independent solutionsof a given nth-order linear equation if it has constant coefficients. The)))

2.2 l Section we saw that a general solution of an nth-order homogeneousinear

2.3Homogeneous C Equationswith Constant oefficients 125)


general such equation may be written
\037)

in

the form)
(1))

any(n)

+ an_Iy(n-l) + ... + a2Y// + alY/ + aoy == 0,)

where the coefficientsao, aI,a2,

..., an are real constants with

an

f=. O.)

The Characteristic Equation


We first

lookfor a singlesolution of Eq. (1), begin with the observationthat) and

_e dx
dk
k(

rx

rx ) == rke

')

(2))

so any derivative of erx is a constant multiple of erx . Hence,if we substituted rx each term would be a constant multiple of erx , with the constant y == e in Eq. (1), coefficientsdepending r and the coefficientsak. This suggests we try to find on that r so that all thesemultiples of erx will have sum zero,n which case == erx will be i y a solution of Eq. (1). For example,n Section we substituted y == erx in the second-order i 2.1 equa-)
tion)

ay
to derive the characteristic equation)

//

+ by / + cy

==)

ar2 + br + c == 0)
r must satisfy. To carry out this technique in the general case, e substitute w and with the aid of Eq. (2)we find the result to be)
that
y

== erx in

Eq.(1),

a n rne rx
that is,)

rx + an-lrn-1e +

...

+ a 2r 2erx + a rerx + a0erx == o.


I
,)

n n 1 + + a2r2 + aIr + ao) == O. (anr + an_Ir e Becauserx is never zero, e seethat y == erx will be a solutionof Eq. (1)precisely w when r is a root of the equation)

erx

- ...

\037)

anr

+ an_Ir n - + ...+ a2r2 + aIr + ao == o.)


1

(3))

This equation is called characteristic the equationor auxiliary equation of the differential equation in (1). ur problem,then, is 'reducedto the solution of this O purely algebraic equation. According to the fundamental theorem of algebra, every nth-degree polynomial-suchs the one in Eq. (3)-has zeros,hough not necessarily a n distinct t and not necessarily real.Finding the exact values of thesezerosmay be difficult or even impossible;he quadratic formula is sufficient for second-degree t equations, but for equations of higher degree we may need either to spot a fortuitous factorization or to apply a numerical technique such as Newton'smethod (or use a

solve command).))) calculator/computer

126

2 Chapter

Linear Equationsof HigherOrder)

Distinct ealRoots) R
Whatever the method we use,let us supposethat we have solvedthe characteristic equation. Then we can always write a general solution of the differential equation. The situation is slightly more complicated the case repeatedroots or complex in of which the characteristic roots of Eq. (3),so let us first examine the simplest equation has n distinct (no two equal) realroots r2, , r n. Then the functions)

ert x , er2x

e ...,)
rnx

case-in rl, ...

are all solutions of Eq. (1), (by Problem 34 of Section thesen solutionsare and 2.2) linearly independent on the entire real line.In summary, we have proved Theorem 1.) THEOREM 1
then)
\037)

DistinctReal Roots)

If the roots rl,r2,

..., r

of the characteristic equation in (3)are real and distinct,

rtx y(x) = Cle

+ C2er2x + ... + cne

rnx)

(4))

is a general solution of Eq. (1).)


...........

Example

1)

Solvethe initial value problem


y(3)

y(O) = 7,

+ 3y\"

- 10y' 0; =
y/

(0) = 0,

y\"

(0) = 70.)

Solution The characteristic equation of the given differential equation is


r3
We

+ 3r2

- lOr =

O.)

solve by factoring:)

r(r2 + 3r 10)= r(r + 5)(r 2) = 0,)


and

and r

so the characteristic equation has the three distinct real roots r =

= 2.Because = 1, heorem 1 gives the general solution eO T


Y (X )

0,r = -5,

= Cl + C2e-5x + C3e2x

.)

Then the given

initial

conditions yield the linear equations)

(0) = y\" (0) =)


y/

y(O) = Cl +

+ C3 = 7, 5C2+ 2C3= 0, 25c2 + 4C3= 70)


C2

the coefficients Cl, C2, and C3. The last two equations give y\" (0) 2y/ (0) 35c2 = 70,so C2 = Then the secondequation gives C3 = 5, and finally the first equation gives Cl = O.Thus the desiredparticular solution is)
in

2.

y(x) = 2e-

5x

+ 5e2x

.)

.)))

2.3Homogeneous Coefficients 127) Equationswith Constant Differential Polynomial Operators


If the roots of the characteristic equation in (3)are not distinct-there repeated are solutions of Eq. (1)by the linearly independent method of Theorem For example,if the roots are 2,2,and 2,we obtain only the two functions eX and e2x The problem,then, is to producethe missing linearly i F independent solutions. or this purpose,t is convenientto adopt \"operator notation\" and write Eq. (1)in the form Ly = 0,where the operator)

roots-then cannot producen we

1.

1,

d2 d L=an-+an-l - +...+a2-+aI-+aO dx dx dx2


dn dn
1

dx)

(5)

on operates the n-times differentiablefunction y(x) to producethe linear combination) + Ly = anyCn) + an_1yCn-l) ... + a2y(2)+ aIY/ + aoy of y and its first n derivatives. We also denote by D = d/dx the operation of differentiation with respectto x, so that) Dy
and

= y/,

D2y = y\",

D3y = y(3),)

so on. In terms of D,the operator L in (5)may be written)


L
n = anD n + an_1D -I

+... +

a2D2 + aID+ ao,)

(6))

useful to think of the right-hand sidein Eq. (6)as a (formal) nthdegree polynomial in the \"variable\" D;it is a polynomialdifferential operator. A first-degree polynomial operator with leading coefficient 1 has the form D a, where a is a real number. It operateson a function y = y(x) to produce)
and we will
find it

(D a)y = Dy

- ay = y/

ay.)

The important fact about such operators is that

any

two of them commute:)

(D a)(D b)y = (D b)(D


for any twice differentiable function the following computation:)
y

a)y)
in

(7))

= y(x). The proof of the formula


by)

(7) is

- - - aCy' = - (b + a)y/ + aby = - (a + b)y' + bay = D(y' - ay) - b(y' - ay) = (D - b)(y' - ay) = (D - b)(D - a)y.) seehere also (D - a)(D - b) = D - (a + b)D + ab. can be of factors of on operator product (D - al)(D- a2) ... (D - expands-by collecting coefficients-in same as does product (x - aI)(x - a2) ... (D a)(D b)y = (D a)(y/ = D(y' by)
y\"

by)

y\"

We

that

Similarly, it

shown by induction the number that an the form out and an) multiplying the an ordinary way (x an) of linear factors, with x denoting a real variable. Consequently,the algebra of polynomial differential operators closelyresembles algebra of ordinary real the
polynomials.)))

128

2 Chapter

Linear Equationsof HigherOrder)

R Repeated ealRoots
Let us now considerthe possibility the characteristic equation) that
anr
n
1

+ an_lr n - + .+ a2r2 + aIr + ao == 0)

..

(3))

has repeatedroots. or example, upposethat Eq. (3)has only two distinct roots, ro F s of multiplicity 1 and rl of multiplicity k == n 1 > Then (after dividing by an) Eq. (3)can be rewritten in the form)

1.

(r rl)k(r ro) == (r ro)(r


Similarly, the

- -

rl)k == o.)
as) rl)k,)
in

(8))

o corresponding perator L in (6)can be written L == (D

- rl)k(D- ro) -

== (D

- ro)(D-

(9))

the

Two solutions of the differential equation Ly == 0 are certainly Yo == erox and rt Yl == e x. This is, however, not sufficient; we need k + 1 linearly independent the solutions in order to construct a general solution, because equation is of order k + To find the missingk 1 solutions, e note that) w

o orderof the factors making no difference because f the formula

(7).

1.

Ly == (D

- ro)[(D-

rl)k y ] == O.)

Consequently,every solution of the kth-order equation)

(D

rl)ky == 0)

(10))

will also be a solution of the original equation Ly == O. Hence our problem is reducedto that of finding a general solution of the differential equation in (10). The fact that Yl == erl x is one solution of Eq. (10)suggeststhat we try the
substitution)

y(x) == U(X)YI(X)

rIX == u(x)e

,)

(11))

where u (x) is a function yet to be determined. bservethat O


rIX (D rl) [uerIX ] == (Du)e

+ u(rle

rIX

- rl(ue
rlX)

rIX

r1x ) == (Du)e

(12)

Uponk applications of this fact, it follows that)

(D

rI)k
Ck

[uerIX ]

==

(Dk u )e

(13))

for any sufficiently differentiablefunction u (x). Hence == uerl x will be a solution y of Eq. (10)if and only if Dku == u ) == O. But this is so if and only if

u(x) == Cl + C2X + C3x2 + ... + Ckxk


a polynomial of degree most k at

l
,)

- 1.Henceour desiredsolution of Eq. (10)is


+ C2X + C3x2 +

y(x) == uer1x

== (CI

...

+ Ckxk-I)er\\x.)

original differential equation Ly == O. The precedinganalysis can be carriedout with the operator D rl replaced with an arbitrary polynomialoperator. When this is done,the result is a proof of the following theorem.)))

rlX In particular, we seehere the additional solutions xe , x2erlX ,

..., xk-Ier1x of the

2.3Homogeneous C Equationswith Constant oefficients 129)


THEOREM2

Roots Repeated

If the characteristic equation in (3)has a repeatedroot r of multiplicity k, then the part of a general solution of the differential equation in (1)correspondingto r is of the form)
>-)

(CI

+ C2X + C3x2 + ... + Ckxk -l )erx

.)

(14))

We may observethat, accordingto Problem 29 of Section2.2,he k funct rx involved in tions erx , xerx , x2erx , , and xk-1e (14)are linearly independenton the real line. Thus a root of multiplicity k corresponds k linearly independent to solutions of the differential equation.)

...

.\037..

Example2)

Find a general solution of the fifth-orderdifferential equation

9y (5) _ 6y (4) + y(3) == o.)

Solution The characteristic equation is

9r5 6r4 + r 3 == r 3 (9r2 6r + 1) == r 3 (3r 1)2== o.


It has the triple root r == 0 and the

doubleroot r

==

x x + C2XeO. + c3x2eO. == Cl + C2X + C3x2 to the solution, while the doubleroot r == 1contributes C4ex/3 C5xex/3. encea H +

contributes

1. The triple root r

== 0

CleO.

general solution of the given differential equation is)

/ + y(x) == Cl + C2X + C3x2 + C4ex/3 C5xex

3
.)

.)

Functionsnd Euler's a Formula Complex-Valued


the coefficientsof the differential equation and its characteristic equation are real, any complex(nonreal) roots will occur complex in bi This raisesthe question conjugate pairsa ::f:: where a and b are real and i == of what might be meant by an exponential such as eCa+bi)x. To answer this question, we recallrom elementary calculus the Taylor(or f f MacLaurin) seriesor the exponential function)

Because e have assumedthat w

H.
t4

et
If we substitute so on, we get)

==

L00
tn n =0

n!

== 1

+ t + 2!+ 3!+ 4!)+ ... .


t3

--t2

t == i () in this

seriesnd recall that a

i 2 ==

i -1, -i,i 4
3 ==

==

1,and

ei(J =

n=O

>;

(io)n
n!)

== 1

+ i() 2! 3!+ 4!+


1_

==

02

- - - --... -... ... 2! ) ( 3! ).


- ()2 4 i ()3
()4

i ()5 5!)

+0 _ 4!

+i

0 3 05 0_ +
5!)))

130

2 Chapter

Linear Equationsof HigherOrder)

Because two real seriesn the last line are the Taylor seriesor cos and sin (), the i f () respectively,this impliesthat)
>-)

eie == cos + i sin ().) ()

(15))

function

This result is known as Euler's formula.Because it, we define the exponential of eZ , for z == x + iy an arbitrary complexnumber, to be)

== ex +iy == eX eiy == eX (cos y

+ i sin y).)

(16))

Thus it appearsthat complexroots of the characteristic equation will lead to s function complex-valuedolutions of the differential equation. A complex-valued F of the real variable x associateswith eachreal number x (in its domain of definition) the

complexnumber)

F(x) == f(x) + ig(x).)

(17))

The real-valued functions and g are calledthe realand imaginary parts, respectively, of F.If they are differentiable, we define the derivative F'of F by)

F'(x) == f' (x) + i g'(x).)

(18))

Thus we simply differentiate the real and imaginary parts of F separately. We say that the complex-valuedfunction F(x)satisfies the homogeneous linear differential equation L[F(x)]== 0 provided that its real and imaginary parts in == + (17)separately satisfy this equation-so L[F(x)]== L[f(x)] iL[g(x)] O. The particular complex-valued functions of interest here are of the form F(x) == erx , where r == a We note from Euler's formula that)

::i: bi.

e(a+bi)x == eax eibx == eax (cos bx


and)
ibx e(a-bi)x== eax e- == eax (cos bx

+ i sin bx))

(19a))

i sin bx).)

19b
))

The most important property of erx is that)


rx rx Dx(e ) == re
,)

(20))

if r is a complexnumber. The proof of this assertionis a straightforward tion basedon the definitions and formulas given earlier:)

computa-

rx ax + Dx(e ) == Dx(e cosbx) iDx(eaxsinbx) ax ax == bx beax sin bx) + i (ae sin bx + beax cos ) bx (ae cos rx ax ax == (a + bi)(e cos + i e sin bx) == re bx

.)))

2.3Homogeneous Coefficients 131) Equationswith Constant


R Complex oots
rx It follows from Eq. (20)that when r is complex Gust as when r is real), e will be a solutionof the differentialequation in (1)if and only if r is a root of its characteristic equation. If the complexconjugate pair of roots rl == a bi and r2 == a bi are then simple(nonrepeated), the correspondingart of a general solution of Eq.(1)is) p

r1x y(x) == Cle + C2er2x

== Cle(a+bi)x

== Cleax(cosbxi
ax

+ C2e(a-bi)x C2eax(cosbx i sinbx) sinbx)+

+ I y(x) == (CI+ C2)e cosbx i(C


where the

- C2)e

ax

sinbx,)

constants CI and C2 can be complex. instance, the choice For arbitrary CI == C2 == 4 gives the real-valued solution YI (x) == eax cos while the choice bx, CI == i , C2 == 4i gives the independent real-valuedsolution Y2(X) == eax sinbx. This yieldsthe following result.)

-4

THEOREM3

R Complexoots)

If the characteristic equation in (3)has an unrepeatedpair of complex conjugate roots a ::f:: i (with b =f. 0), then the correspondingart of a general solution of b p Eq. (1)has the fonn)
\037)

eax (CI cos + C2sin bx).) bx

(21))

Example3)

The characteristic equation of)


y\"

+ b 2y

== 0

(b > 0))
a == 0) gives the general

is r 2 + b 2 == 0,with roots r
solution)

==

::f::bi.Theorem 3 (with So

bx y(x) == CI COS + C2 sin bx.)

.)

Example4)

Find the particular solution of)


y\"

- 4y' + 5
==

== 0)

for which y(O) == 1 and y' (0) == 5.)

Solution Completion of the square in


r2

the characteristic equation yields)

a == 2 and b == 1 gives the general solution)

== ::f::i. we obtain the complex so r 2 == ::f:: i (which Thus conjugate roots 2::f:: couldalsobe found directly using the quadratic formula). Hence Theorem3 with
2x y(x) == e (CI cosx+ C2 sin x).)))

- 4r + 5 (r - 2)2 +

1 == 0,)

132

2 Chapter

Linear Equationsof HigherOrder)


Then)
2x y'(x) == 2e (CI COSX + C2 sin x) + e2x (-clsin x + C2cosx),)

so the

initial

conditions give) y(O) == CI == 1 and

y' (0) == 2CI+ C2 == 5.)

It follows that C2 == 3, and

so the desiredparticular solution is)

+ y(x) == e x(cosx 3 sin x).)


In
y)

.)
(22)) upon
writing)

Example5 belowwe employ the polarform)


z == x

+ iy == re

ie)

(x, y))

of the complexnumber

z. This form follows from Euler'sformula

z=r
x)

( +i

\037

= r(cose i sine) = re +

i6)

x + iy.)

FIGURE Modulus and number argument of the complex

2.3.1.

terms of the modulus r == x2 + y2 > 0 of the number z and its argument() indicated in Fig. For instance, the imaginary number i has modulus 1 and iJr == e3Jr /2.Another consequencethe fact is argument Jr/2,so i == e /2. Similarly, ie has the two that the nonzero complex roots) number z == re square
in

2.3.1.

-i

ie v'Z == ::f::(re == )I/2

ie/ 2 ::f::,Jre
,)

(23))

where ,Jr denotes(as usual for a positive real number) the positive square root of the modulus of z.)

Example5)

Find a general solution of y(4)

+ 4y == O.)

Solution The characteristic equation is)


r4

-4

== (r2)2

- (2i)2
iJf

== (r2

+ 2i)(r2 2i) == 0,)

and its four roots are ::f:: ::f::2iSince == eiJr / 2 and i


iJf

,J .

= .../2i (2e / 2)1/2= he / 4 = h (cos + i sin


and)

::
)
i sin

-i

== ei3Jr / 2, we find

that)

= 1+

i)

,J

2i

i3Jf = (2e / 2f/ 2 =

he

i3Jf

/4 =

h (cos3:+ 3: = -1+ )
-1
x

i.)

Thus the four (distinct) roots of the characteristicequation are r == ::f::(::f:: These 1+i). two pairsof complex ::f:: give a general solution) i roots, 1 ::f:: and conjugate

i,

y(x) == eX (CI cosx+ C2sinx) + e- (C3cosx+ C4sinx))


of the differential equation y(4) + 4y == O.)
.)))

2.3Homogeneous C Equationswith Constant oefficients 133)


R Repeated omplex oots C
I bi Theorem 2 holdsfor repeatedcomplex roots.f the conjugatepair a ::f:: has multhen the correspondingart of the general solutionhas the form) tiplicity k, p
(AI

+ A 2x +

...

+ Akxk-l)eCa+bi)x + (Bl + B2X +


==

Lx e
p=o)

k-l

...

+ Bkxk-l)eCa-bi)x
(24)

P ax

bx (Ci cos

+ di sin bx).

It can be shown

that

the 2k

functions)

ax xP eax cos bx,) xP e sin bx that

,)

< p < k-1)

appearin Eq.(24)are linearly independent.)

Example6)
Solution

Find a general solution of By completing the

(D2 + 6D+ 13)2y==

O.)

square,we seethat the characteristic equation)


== (r2 + 6r + 13)2 [(r + 3)2 + 4]2 == 0)

has as its roots the conjugate pair gives the general solution)

-3::f::2i of multiplicity

k ==

2. HenceEq. (24)

3x 3x 2x 2x y(x) == e- (Cl cos + dl sin 2x) + xe- (C2cos + d2 sin 2x).

.)

In applications we are seldompresentedin advance with a factorizationas convenient as the one in Example Often the most difficult part of solving a homo6. l geneousinear equation is finding the roots of its characteristic equation. Example 7 illustrates an approach that may succeed hen a root of the characteristicequation w can be found by inspection.)
\037 \"

Example7)

The characteristic equation of the differential equation)


y(3)

co'''
C\"'

C\"'\"

.-.

C\"'

\037.

_N..

..

,_u.'u

+ y'

- 10y

== 0)

is the cubicequation)

r3

+r-10==0.
2.

By a standard theorem of elementary algebra, the only possiblerational roots are the factors ::f:: ::f::2, and ::f:: of the constant term ::f::5, By trial and error (if not we discoverthe root The factor theorem of elementaryalgebra by inspection) and divisionof the former into the latter impliesthat r 2 is a factor of r 3 r producesas quotient the quadratic polynomial)

1,

10

10.

- 10,

r2

+ 2r + 5 == (r + 1)2+ 4.)

The roots of this quotient are the complexconjugates have found now yield the general solution)

-1::f::2i.The three roots we


.)))

y(x) == Cle

2x

+ e-x (C2cos2x + C3 sin 2x).)

134

2 Chapter

Linear Equationsof HigherOrder)


\037 \037 _ A

Example8)

The roots of the characteristic equation of a certain differential equation are 3, 5, 0,0,0,0, 2 ::f:: i,and 2 ::f:: Write a general solution of this homogeneous 3 differential equation.)

-5,

3i.

Solution The solution can be read directly from


y(X) == CI

I the list of roots.t

is)

3x + C2X + C3x2 + C4x3 + cse + C6e-sx + C7xe-sx + e2x (cscos3x + Cg sin 3x)+ xe2x (CIOcos3x + CII sin 3x).

.)

IfI1Problems
Find the

u)

lems

20. (4) + 2y(3) + 3y\" y 2 (r +r+l)2.))


Solvethe
initial

- 4y -0 3. + 3y' 10y 0 5. + 6y' + 9y 0 7. - 12y'+ 9y 0 9. + 8y' + 25y 0 11. - + - 0 12. + y' 13. + + 4y' 0 15. - + 16y 0 17. + + 4y 0 19. + - y' - 0
1.
y\" y\" y\" y\"
== == ==

1through

generalsolutions of the
20.)

differential

equations
==

in

Prob-

37. Find
and

4y\"

==

==

2. 2y\" - 3y' 0 4. 2y\" - 7y' + 3y 0 6. y\" + 5y' + 5y 0 8. y\" - 6y' + 13y 0 10.5y(4) + 3y(3) 0)
== == == ==

38.Solvethe initial
y(3)

a function y(x) such y'(O)== y(O) ==

18,
_
==

12, y\"(O)

that y(4)(x)
==

==

and 13, y(3)(0) 7.


==

y(3)(X) for all

value problem)
5y\"

+ 100y' 500y == 0;
y' (0)
==

y(O)
given that differential In
YI

0,
==

10, y\" (0)

==

250)

y(4) y(4) y(4)


y

8y(3) 3y(3)
8y\" y\"

16y\"

==

(x)

e5x is one particular

solution

of the

3y\"

==

9y(3)
(3)

12y\" lly\"

==

==

6y(4)

==

14.y(4) + 3y\" - 4y 0 16.y(4) + 18y\" + 81y 0 18.y(4) 16y


== == == ==

equation.)

==

+ 2y' + y

0 (Suggestion: Expand
in

through 42, find a linear homogeneous equation with the given generalsolution. constant-coefficient 2 2x y(x) == (A + Bx + Cx == Ae 2x + B cos 2x + C sin 2x 40. y(x)

Problems39

39.

)e

26.

value problems given


==

Problems through 21

41.y(x)
42. y(x)

== ==

A cos2x

2 2 )sin2x) (A+Bx+Cx )cos2x+(D+Ex+Fx

+ B sin 2x + C cosh2x + D sinh 2x

0; y(O) == 7, y'(O)== 11 22.y\" + 6y' + 4y == 0; y(O) == 3, y'(O)== 4 9 23.\" 6y' + 25y == 0; y(O) == 3, y'(O)== 1 y 24. 2y(3) 3y\" 2y' == 0; y(O) == 1, '(O)== y y\"(O) == 3 == 0, == 1 25.3y(3) + 2y\" == 0; y(O) == y'(O) y\"(O) 26.y(3) + 10y\" + 25y' == 0; y(O) == 3, y' (0) == 4, y\" (0) ==

21.\" - 4y' + 3y y

-- -

-1,

-1,

5)

Find general solutions of the equations in Problems 7 through 2 32. First find a small integral root of the characteristic equation by inspection; then factor by division. 27. y(3) + 3y\" 4y == 0 28. y(3) y\" 5y' 2y == 0 2 29. (3) + 27y == 0 y 30.(4) y(3) + y\" 3y' 6y == 0 y y(3) + 3y\" + 4y' 8y == 0 32.(4) + y(3) 3y\" 5y' 2y == 0) y

Problems 43 through 47 pertain to the solution of differential equations with complexcoefficients. num43. (a) UseEuler'sformula to show that every complex ber can be written in the form rei(), where r > 0 and -Jr < e < Jr. (b) Expressthe numbers 4, 3i, 1 + and in the form rei(). (c) Thetwo square + Find roots of rei() are :!::.-Jiei()/2. the square roots of the and numbers 2 + 2i

i, -1 i,J3

-2,

- 2i,J3

-2 ,J3.
to the

44. Usethe

- - -

tions. Note in eachcase that

quadratic

formula

solve the

following

roots are not complex

equa-

conjugates.

31.

-- - - ==

(b) x 2 2ix + 3 == 0 (a) x 2 + ix + 2 == 0 45. Find a general solution of y\" 2iy' + 3y == O. 46. Find a general solution of y\" iy' + 6y == 0 47. Find a general solution of y\" == + y.

48. Solvethe

(-2 2i,J3)
y\" (0) ==

initial
==

value problem)

In

equation

Problems 33 through 36, one solution of the is given. Find the generalsolution.

differential

33.y(3) + 3y\"
34. 3y(3)

35.6y(4)

36.9y(3)

0; - -+54y - e0; e / 12y' + + + 20y' + 4y 0; e+ + 4y' - 14y 0;


y
==

y(3)

y;

y(O) ==

1,

y' (0) ==

O.)

3x

2y\"

8y

==

==

2x

3
y
==

5y(3)
lly\"

25y\"

==

cos2x
x)

(Suggestion: Imposethe general solution) y(x)


==

given initial

conditions

on the

==

==

sin

Ae x

x ax + Be + Ce/3

,)))

2.4Mechanical ibrations 135) V


where and fJ arethe t 0, to discoverhat)

3 complex conjugate roots of r

-1=

Y YI
(x)) 4)

y(x) is a solution.)

1 =\"3

eX

x/ + 2e- 2 cos

2))
x\037

-4)
Y2(x))

49. Solvethe

initial

value problem)

= y(3) + y\" + y' + 2y; y(O) = y' (0) = y\" (0) = 0, 2y(3)(0) = 30.)
y(4)

FIGURE2.3.2. Graphs of YI (x) and Y2 (x) in Problem 50. 51.)According to Problem 51in Section2.1, substitution the v = In x (x > 0) transforms the second-order Euler equation ax2y\" + bxy' + cy = 0 to a constant-coefficient homogeneous linear
substitution equation. Show similarly that this same transforms the third-order Euler equation)

50.Thedifferential

equation)

y\"

+ (sgnx)y= 0)
coefficient function)

(25))

ax311I bx 211 cxy + dy =) 0 y + y +


(where a, b, c, d are coefficientequation)
constants) into the constant-

'

has the

discontinuous

+1 if x > 0, sgnx = -1) if x <


{

ad 3y
dv 3

+ (b

d - 3a)-+ (c -b + 2a)-+
2
y

dy

dv 2

dv)

dy

= O.
general

o.)

Make the

substitution

Show that Eq.(25)neverthelesshas two linearly dent solutions YI (x) and Y2 defined for all x

Each satisfies Eq.(25)at eachpoint x i= 0; Each has a continuous derivative at x = 0; = 1 and Y2(0) = (0) = O. YI (0) = Each Yi (x) will be defined by one formula (Suggestion: for x < 0 and by another for x > 0.)The graphs of these two solutions are shown in Fig. 2.3.2.)
y\037(O)
y\037

. . .

(x)

indepensuch that

solutions throug h

58.

(for x

> 0) of the

= In x of Problem 51to find


Euler equations
in

Problems

52

52.x2y\" + xy' + 9y = 0 53.x2y\" + 7xy' + 25y = 0 54. X 3y + 6x2y\" + 4xy' = 0 55.X 3y lll x2y\" + xy' = 0 56.x 3y'\" + 3x2y\" + xy' = 0 57. X 3y 3x2y\" + xy' = 0 58.X 3y lll + 6x2y\" + 7xy' + y = 0)
lll lll

Vibrations) Ell) Mechanical


m
c)

I I

Equilibrium position)

FIGURE2.4.1. massA
spring-dashpot system.)

The motion of a mass attached to a spring servesas a relatively simple example of the vibrations that occurin more complexmechanical systems.For many such systems,the analysis of thesevibrations is a problemin the solutionof linear differential equations with constant coefficients. We considera body of massm attached to one end of an ordinary spring that resistscompression well as stretching; the other end of the spring is attachedto as a fixed wall, as shown in Fig. Assume that the body rests on a frictionless horizontal plane, so that it can move only back and forth as the spring compresses and stretches. enoteby x the distance the body from its equilibriumpositionD of its position when the spring is unstretched. We take x > 0 when the spring is a stretched, nd thus x < 0 when it is compressed. law, According to Hooke's the restorative force Fs that the spring exerts on the massis proportional to the distance that the spring has beenstretchedor comx Because is the sameas the displacement of the mass m from this x pressed.

2.4.1.

its)))

136

2 Chapter

Linear Equationsof HigherOrder)


equilibrium

position,it follows that)

Fs == -kx.)

(1))

The positive constant of proportionality k is calledthe springconstant.Note that Fs and x have oppositesigns:Fs < 0 when x > 0,Fs > 0 when x <o. showsthe mass attached to a dashpot-a device,like a shock Figure 2.4.1 absorber,that providesa force directedoppositeto the instantaneous direction of motion of the mass m. We assumethe dashpot is so designedthat this force FR is proportional to the velocity v == dxjdtof the mass;that is,)

FR

==

-ev == dx dt)

-e-.

(2))

The positive constant e is the damping constantof the dashpot. Moregenerally, we may regard Eq. (2) as specifying frictional forcesin our system (including air resistanceto the motion of m). to If, in addition to the forcesFs and FR , the massis subjected a given external force == F(t),hen the total force acting on the massis F == Fs + FR + FE. t FE

UsingNewton'slaw)

F == ma == m

d2x dt 2

== mx \"

'
(3))

we obtain the second-order linear differential equation)


\037)

mx\"

+ ex'+ kx == F(t))

governs the motion of the mass. If there is no dashpot (and we ignore all frictional forces),hen we set e == 0 t in Eq. (3)and call the motion undamped;t is damped motion if e > O.If there is i no external force, we replace (t) with 0 in Eq. (3).We refer to the motion as free F in this case and forced the case in F(t) =f. O.Thus the homogeneousquation) e
that
\037)

mx \"

+ ex, + k x

==)

(4))

$j

i)

U nstretched
spnng)

Static
equilibrium)
y)

System in motion)

FIGURE2.4.2. mass A

suspendedvertically from a spring.)

of a masson a spring with dashpot but with no external forces defer discussion forced motion until Section of 2.6. For an alternative example,we might attach the mass to the lower end of a In spring that is suspended vertically from a fixed support, as in Fig.2.4.2.this case the weight W == mg of the mass would stretch the spring a distanceSo determined by Eq. (1)with Fs == Wand x == So.That is, mg == ks o, so that So == mgjk. This o gives the static equilibrium position of the mass. If y denotesthe displacement f the mass in motion, measureddownward from its static equilibrium position, then we ask you to show in Problem 9 that y satisfies Eq. (3);specifically,that)

describes motion free


applied.We
will

my\"

+ ey' + ky

== F(t))

(5))

if we includedamping and external forces(meaning thoseother than

gravity).)))

2.4Mechanical ibrations 137) V

TheSimple Pendulum
The importance of the differential equation that appearsin Eqs. (3)and (5) stems from the fact that it describes motion of many other simplemechanicalsystems. the For example, simple endulum consistsof a massm swinging backand forth on a p the end of a string (or better, a masslessrod) of length L, as shown in Fig.2.4.3. We may specify the position of the mass at time t by giving the counterclockwise angle () == e(t) that the string or rod makeswith the vertical at time t. To analyze the motion of the massm, we will apply the law of the conservation of mechanical to energy, according which the sum of the kinetic energy and the potential energy of m remains constant. The distancealong the circular arc from 0 to m is s == L(),so the velocityof the massis v == dsjdt == L(d()jdt), nd therefore its kinetic energy is) a T == \037mv2 == \037m 2 2
We next

o)

FIGURE2.4.3. simple The


pendulum.)

( )
0,

dS 2 = \037mL2 2 dt

( )2.
dt)

de

the lowestpoint 0 reachedby the mass (see Then its potential energy V is the product of its weight mg and its Fig. 2.4.3). vertical height h == L (1 cos above so) ())

chooses referenceoint a p

==mgL(I-cos()).)
is a constant C therefore gives)
2

The fact that the sum of T and

-mL2
2
We

differentiate both sidesof this mL 2


d() dt

( ) -)( ) (
d()
dt)

+mgL(I-cose)==c.
with

identity

respectto t to obtain)
d()

d2()

so)

- -g.
d 2() dt 2

\037 dt

. == + mgL(sIn())- 0, dt)
SIn () == 0

L)

(6))

after removal of the common factor mL 2(d()ldt). This differential equation can be derived in a seemingly more elementary manner using the familiar secondlaw F == ma of Newton (appliedto tangential components of the accelerationof the mass and the force acting on it). However, derivations of differential equations based on conservation of energy are often seenin more complexsituations where Newton'slaw is not so directly applicable, nd it may be instructive to see energy a the method in a simplerapplication like the pendulum. Now recall if () is small, then sin () that () (this approximationobtainedby f retaining just the first term in the Taylor seriesor sin e).In fact, sin () and e agree to two decimal places hen I()I is at most ]T/12(that is, In a typical pendulum w for example,e would never exceed 15 It therefore seemsreasonableto clock, simplify our mathematicalmodel of the simplependulum by replacing sin e with e in Eq. (6). If we also insert a term C()'to account for the frictional resistance the of medium, the result is an equation in the form of Eq. (4):) surrounding
\037
\302\260

15\302\260).

()\"

+ C()'+ k() == 0,)

(7))))

138

2 Chapter

Linear Equationsof HigherOrder)


where k == g/L. Note that this equation is independent of the mass m on the end between () and of the rod.We might, however,expect effects of the discrepancy the sin () to accumulate over a periodof time, so that Eq. (7)will probably not describe accurately the actual motion of the pendulum over a long periodof time. In the remainder of this section,we first analyze free undamped motion and then free dampedmotion.)

FreeUndamped Motion
Eq. (3)takesthe simplerform)
If we have only a mass on a spring,with neither damping nor external force, then
mx\"

+ kx == O.)
=

(8))

It

is convenient to define)
Wo

[!;)
.)

(9))

and rewrite Eq. (8)as)


X
\"

2 + wax == 0

(8'))

The general solution of Eq. (8')is)

x(t) == A coswot+ B sinwot.)

(10))

a so that)

To analyze the motion described by

this

c solution, we chooseonstants C and


sIn a ==

C == vi A2

A + B2,) cosa == C)

-,

and)

-,
B
C)

(11))

as indicated in Fig. Note that, although tan a == BIA, the angle a is not given by the principal branch of the inverse tangent function (which gives values only in the interval x n/2). Instead,a is the angle between 0 and 2n whose cosineand sine have the signsgiven in where either A or B or both may be

2.4.4.

-n/2 < <

(11),

negative. Thus)
A)

FIGURE2.4.4. angle The

Ci.)

tan- l (B/A) a==) n + tan- l (B/A) 2n + tan- l (BIA)) where tan- l (B/A) is the angle in In any event, from (10)and
A

if if if

A A A

> 0,B > 0 (first quadrant), < 0 (second third quadrant), or > 0,B < 0 (fourth quadrant),)

(-n/2, /2) given by a calculator or computer. n we (11) get)

w x(t) == C C cos ot + C slnwot

B.

==

c C(cosaosw ot + sIn a slnwot )


that)

. .

.)

With

the aid of the

cosine addition formula, we find

x(t) == C cos(wot a).)


Thus the massoscillates and from about its equilibrium position with))) to

(12))

V 2.4Mechanical ibrations 139)

3.Phaseangle)

2. Circularrequency f

1.Amplitude

C,
(Va,

and

a.)

Suchmotion is calledsimple armonic h motion. If time t is measured seconds, circular frequency (Va has dimensionsof in the of radians per second(rad/s). The period the motion is the time required for the to complete one full oscillation, o is given by) s system
x)

x(t) = Ccos(root- a))

its seconds; frequencyis)


v
in

2n T=- (Va)

---1
(Va

(13))

2n)

(14))

t) I I I I I I I I I I

I-I I

-C

--1----I I

(Hz), which measuresthe number of completecyclesper second. Note in whereascircular frequencyhas the frequency is measured cyclesper second, dimensions radians per second. of A typical graph of a simpleharmonic position function)
hertz
that t x(t) = C cos(wa

I.
harmonic motion.)

-I)

ex) =

C cos

FIGURE2.4.5. Simple

= ( (t :a)) C cos(Wo(t
Wa

8\302\273)

is shown in Fig.2.4.5, where the geometric significance of the amplitude C, the periodT, and the time lag) a

8=(Va)

are indicated. If the initial position x(0) = Xa and initial velocity x'(0) = Va of the mass t are given, we first determine the values of the coefficientsA and B in Eq. (10),hen find the amplitude C and phaseangle a by carrying out the transformation of x(t) to the form in Eq. (12), indicated previously.) as

Example

1)

body with mass m = 4 kilogram (kg) is attached to the end of a spring that is stretched 2 meters (m) by a force of 100ewtons (N).It is set in motion with initial n (rn/s). (Note that these initial position Xa = 1 (m) and initial velocity Va = conditions indicate that the body is displaced the right and is moving to the left to at time t = 0.) Find the position function of the body as well as the amplitude, a frequency,periodof oscillation, nd time lag of its motion.)
A

.....

......

.....

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

.....

....,.

......

......

..\"..

-5

Solution The spring constant is k =

50x= 0;that

is,)

(100 )/(2 m) = 50(N/m),so Eq. (8)yields N = x\" + 100x O.

\037X\"

Consequently,the circular frequency of the resulting simpleharmonic motion of the = (rad/s). ence will oscillatewith period it H body will be (Va = v'100 10 T
and
with frequency)

-- - = 2n = 2n
(Va

\037

10)

0.6283 s

(Va

2n

= 10 1.5915 Hz.
\037

2n)))

140

2 Chapter

Linear Equationsof HigherOrder)


We now
function)

imposethe initial conditions x(0) == 1 and x'(0) ==

-5on the position

x(t) == A coslOt + B sin lOt


It follows readily that
A == 1 and

with

X'(t) == -lOA sin lOt + lOBcoslOt.)

B == 4, so the position function of the body is)


== coslOt

x(t)
Hence amplitude of motion its

-- .
1
2)

SIn

10

t.

is)

c = J(1)2+ (-i)2= i.J5 m.)


To find the time lag, we write)

x(t) =

0 (0cos - 0 2
2 lOt
1

sin lOt

0cos(10t 2 1

ex),)

where the phaseangle a satisfies)

cosex=

0
1

>0

and)

sin a ==

-o)
1

< o.

a Hence is the fourth-quadrant


ex

angle)

= 2:rr + tan-

( 2/0)
is)

-1/0= - tan- (i)


2:rr

\037

5.8195,)

and the time lag of the motion

8 ==

a
(Va)

\037

s. 0.5820

With

the amplitude and approximatephaseangle shown explicitly,the position function of the body takesthe form)

x(t)
and its graph

\037

- 5.8195),) !.J5cos(10t
.)))

is shown

in

Fig.2.4.6.)

V 2.4Mechanical ibrations 141)


x)
8) I-)

T)

I)

0.5)

c)

.5)

-0.5) -1)

x(t) = C cos(wot a) in
C
\037 \037

FIGURE2.4.6. Graph of the

1.118, T 0.628, period

Example

position function with amplitude and time lag 8 0.582.)

1,

\037

FreeDamped Motion
With

takesthe form

damping

but

no external force, the differentialequation we have beenstudying

mx\"

+ cx'+ kx == 0;alternatively,) ' 2 X + 2px + wax == 0


II

,)

(15))

where Wo == ,Jk/mis the corresponding ndamped circular frequency and) u P == 2m


C

> o.
1/2)

(16))

The characteristic equation r 2 + 2pr + w6 == 0 ofEq.(15)has roots

rl, r2 == -p ::i: - (6) (p2


c ==--- 2
k

(17))

that

dependon the signof) p


2 2 -wa

C2

4m 2

m)

4km 4m
2)

(18)) three

(0, Xo))
\037)

The critical amping Cer is given by Cer == ,J 4km , and we distinguish d as C > Cer C == Cer or C < according

'

'

cases,

Cere)

o)

OVERDAMPED CASE:C

w case, e are dealing with

a strong resistancein comparison with a relatively weak or a small mass. Then (17)gives distinct real roots rl and r2, both of which spring are negative. The position function has the form)

> Ccr(C2 > 4km). Because is relatively large in this C

o)
t)

+C2 x(t) == CIerI! er2!.)


It is easyto see that
without

(19))

FIGURE2.4.7. verdamped O
motion:
rl

< 0 and

r1t x(t) = cIe + C2er2t with

x(t)

r2

< O.Solution curves position


sameinitial

are graphed
position
Xo

with the

and different initial

velocities.)

w Xo graphs of the position function for the overdampedcase; e chose a fixedpositive number and illustrated the effects of changing the initial velocity va. In every case the would-be oscillations re dampedout.))) a

any

0 as t and +00 that the body settlesto its equilibrium shows sometypical oscillations(Problem 29). Figure 2.4.7
\037 \037

142

2 Chapter

Linear Equationsof HigherOrder)


2 CRITICALLYDAMPEDCASE: C = Ccr(c = 4km). In this case,(17)gives equal roots rl = r2 = p of the characteristic equation, so the general solution is)

x(t) = e-pt (CI + C2t ).)

(20))

(0,XO))
\037)

o)

0 and CI C2t has at most one positive zero,the body passes its equilibrium positionat most once,and it is clear that x(t) --+ 0 as Somegraphs of the motion in the critically dampedcaseappear in In the and they resemblethose of the overdamped case(Fig. Fig. t critically dampedcase,he resistanceof the dashpot is just large enough to damp out any oscillations, ut even a slight reduction in resistancewill bring us to the b t remaining case,he one that showsthe most dramatic behavior.)
through t --+

Because -pt > e

2.4.8,

+00.

2.4.7).

o)
t)

2 UNDERDAMPED CASE: C 4km). The characteristic equation now C (c cr has two complex roots p i wB p2, and the general solution is) conjugate
\037)

<

< - ::i: J

FIGURE2.4.8. Critically
damped motion:

wIt x(t) = e-pt (A cos

+ B sin wIt),)
2

(21))

x(t)

Solution curves are graphed with the same initial position Xo and different initial velocities.)

= (c] + c2t)e-ptwith p > o.

where)

WI

_ - J - - .J - c _
Wa

p2

4km

2m)

(22))

we addition formula as in the derivation of Eq. (12), may rewrite Usingthe cosine as) Eq. (20)
p x(t) = Ce-t

COswlt+
\037 \037

sin
Wit)

,)

so)
\037)

p x(t) = Ce-t COS(WIt

a))

(23))

where)

\037)

C=

A2

=A + B2,) cosa C)

-,

and)

. Slna =

-.
B
C)

a)

,WI
\"

I)

/',--.I)
X X
t)

.......( = +Ce-pt

= Ce-pt cos(w It a)

\037

0)

o)

o The solution in (22)represents exponentially dampedoscillations f the body around its equilibrium position. The graph of x(t) liesbetweenthe \"amplitude pt and x = Ce-t and touches them when wIt p a is envelope\"curves x = an integral multiple of The motion is not actually periodic,but it is nevertheless useful to call WI its circularrequency(more properly, its pseudofrequency), l = T f p 2nIWI its pseudoperiod oscillation,and Ce-t its time-varying amplitude. of Most of these quantities are shown in the typical graph of underdamped motion in Fig.2.4.9. ote from Eq. (21)that in this case N WI is lessthan the undamped circular frequency Wa, so TI is larger than the period T of oscillation of the same mass without damping on the same spring. Thus the action of the dashpot has at leasttwo effects:)

-Cen.

FIGURE2.4.9. Underdamped oscillations: pt x(t) = Ce- cos(w]t a).)

exponentially damps the oscillations,in accord with the time-varying amplitude. 2. It slowsthe motion; that is, the dashpot decreases frequency of the motion.))) the

1.It

V 2.4Mechanical ibrations 143)


As the following example illustrates, damping typically alsodelays the motion further-that is, increases time the comparedwith undamped motion with the sameinitial conditions.) The mass and spring of Example1 are now attached also to a dashpot that provides 1 N of resistancefor each meter per secondof velocity. The massis set in motion with the sameinitial position x(O) == 1 and initial velocity x'(0) == as in Example Now find the position function of the mass,its new frequencyand of pseudoperiod motion, its new time lag, and the times of its first four passages the initial position x == O.) through

lag-as

Example2)

1.

-5

Solution Rather than memorizing the variousformulas given in Hence (4) is 1X\" Eq.

+ x' + 50x== 0;that is, == x\" + 2X' + 100x O.) == The characteristic equation r 2 + 2r + 100 (r + 1)2+ 99 == -1 ,J99 so the general solution is) ::i: i, r2 t + x(t) == e-t (A cos,J99 B sin,J99 t).)

better practicein a particular case set up the differential equation and then solve to it directly. Recall that m == 1 and k == 50;we are now given c == 1 in mks units.

it the preceding discussion, is

o has roots rI,)

(24))
\037

is (as Consequently,the new circular (pseudo)frequency WI == ,J99 9.9499 comin and with Wa == 10 Example1). he new (pseudo)period frequencyare) T pared
TI ==
and)

2n
WI

==

2n
\037

V 99)

fAi\\

VI

==

-1

TI

==

WI

2n

==

\037

0.6315 S

,J99 1.5836 Hz 2n
\037

< > with T 0.6283TI and V 1.5915VI in Example 1). (as compared We now impose initial conditionsx(0) == 1 and x'(0) == 5 on the position the
\037

function

in

(23)and the resulting velocity function)

t x'(t) == -e-(A cos,J99 B sin,J99+,J99 t (-Asin,J99 B cos t).) t + et + J99 t)

It follows that)

x(O)== A
whence we find
the body is

== 1

and

= x'(O)== -A + B,J99= -5,)

that A ==

1 and

B == -4/,J99. the new position function of Thus


t

x(t) = e-t
its Hence time-varying

(cas,J99

- \037sin,J99 . t)
is)

amplitude of motion

CIe-t

==)

4 e- = {115 (1)2+ 99e( ,J99 )


2
t t V

.)))

144

2 Chapter

Linear Equationsof HigherOrder)


We

therefore write)

x(t) =

vTIS e-t \037 cosJ99t 99 115 \037 ( vTIS


V

4 - vTIS 115

sin

J99t

(ill- cos( 99 - al),) e J


9g
t

where the phaseangle al satisfies)

= \037 cosal vTIS>0 115)


Hence is the fourth-quadrant al
angle)

and

sinal =

4 -vTIS<

O.)

+ al = 2][ tan-l

-4/vTIS = 2][ tan-l 4 IN> (v99)) 5.9009, 99/ 115 (\037 vTIS)
\037

and the time lag of the motion

is)

81

= al
WI)
1

in
t

S \0370.5931

(as comparedwith 8 masstakesthe form)

\037

< 0.5820 8

plitude and approximate phaseangle shown explicitly, the position function

Example1).With the

time-varying

x(t)

\037

f{le- cos(J99 - 5.9009),


15
99)
t in x = Ctet)

amof the

(25))

and its graph is the dampedexponential that is shown with the undamped oscillations Example1).) of
x)

(in Fig.2.4.10 comparison

\\
.......)

t x(t) = C]ecos(w]t a]) of Example 2 (damped the oscillations), position function x(t)= C cos(wot a) of

FIGURE 2.4.10. of the Graphs

position function

x(t) = ::l:C1e.)))

and Example 1 (undamped oscillations), t

the

curves envelope

V 2.4Mechanical ibrations 145)


From (24) we seethat the masspassesthrough == and thus when) when COS(WIt

- aI)
==

0,

its equilibrium position x == 0

wIt
that

- al

--,
3n
2)

n n 3n 2') 2') 2') n

.
+

..
,)

is, when)

t == 8 1

We

seesimilarly
t == 80

that

when)

-, - -, -, -, - -, - -, -, -,
- 3n
2WI)

81

81

81

2WI)

2w})

3n

2WI)

the undamped massof Example1 passesthrough 80

equilibrium

3n

2wa

2wa

80

2wa

80

3n

2wa
t4

The following table comparesthe first four values tI, t2, t3, undamped and dampedcases, respectively.)
n tn tn

we calculate for

the

(undamped)

0.11070.42490.7390 1.0532 0.43520.7509 1.0667) 0.1195

(damped)

(where only the first three equilibrium passagesare Accordingly, in Fig. 2.4.11 shown) we seethe dampedoscillationsagging slightly behind the undamped ones. l

.)

x)

x = Clet)

x(t) = Ccos(Wo t

a))

- -)

---)

- --)- --)
-1)
t x(t) = Cle-cos( wit

a\037)

FIGURE2.4.11. Graphs
illustrating the additional

on the interval

with delay associated

0 < t < 0.8


damping.)

1IIl\037.!'!\037 In
\037 \037_!:

\037..\037.\037

_)

1.Determine
2.

the

monic motion
spring constant Determine the monic motion

period and frequency of the simple of a 4-kg mass on the end of a spring

harwith

stretched
initial

period and frequency of the simple harof a body of mass 0.75 on the end of kg a spring with spring constant 48 N Im. 3.A mass of 3 kg is attached to the end of a spring that is)

16 Im. N

position Xo Find the amplitude, motion.)

20 cm by a forceof 15N. It is set in motion with = 0 and initial velocity Vo = m/s. period,and frequency of the resulting

-10

4. A body
that

with mass 250g is attached to the end of a spring is stretched 25 cm by a forceof 9 N. At time t =

0)))

146

2 Chapter

Linear Equationsof HigherOrder)

the body is pulled 1 m to the right, stretching the spring, and set in motion with an initial velocity of 5 mjs to the left. (a) Find in the form C cos(wot (b) Find the amplitude and periodof motion of the body.)

x(t)

- a).

Problems5 through 8, assumethat the differential equation of a simple pendulum of length L is LO\" + gO = 0, where 2 at g = GM R is the gravitational acceleration the location ofthe pendulum (at distanceR from the centerof the earth; M denotes mass of the earth). the
In

5. Two pendulums

areof lengths LI and L2 and-whenlocatedat the respective istances I and R 2 from the center R d of the earth-have periodsPI and P2. Show that)
PI P2

A 11.

RI-JL;
R2

slugs in ft-Ib-s (fps) units) floats in water with its W axis vertical (as in Problem 10). hen depressed slightly and

3.125

The FIGURE2.4.12. buoy of Problem 10. cylindrical buoy weighing 100 Ib (thus of mass m released,it oscillatesup
and down four times every the radius of

-JL;.)

10 Assume that friction s.

is negligible. Find

6.A certain
radius 2 min result bulge

pendulum keeps perfecttime in Paris, where the of the earth is R = 3956(mi). But this clockloses 40 s per day at a location on the equator. Usethe of Problem 5 to find the amount of the equatorial of the earth.)

12. ssume that the earth A


with

is a solid sphereof uniform density, mass M and radius R = 3960(mi). For a particle of mass m within the earth at distance r from the center of

the buoy.

7. A

in., pendulum of length 100.10 locatedat a point at sea level where the radius of the earth is R = 3960(mi), has the same periodas doesa pendulum of length 100.00 in. atop a nearby mountain. Usethe result of Problem5 to find the height of the mountain.)

8.Most grandfather

clockshave pendulums with adjustable One such clockloses10 per day when the min lengths. length of its pendulum is 30 in. With what length pendulum will this clockkeepperfecttime?) of a mass attached to suspended spring. (Suggestion: displacement of the mass below of the spring; set up the differeny

the earth, the gravitational force attracting m toward the -GMr mjr 2, where Mr is the mass of the centeris Fr of the earth within a sphereof radius (a) Show that part 3 that a small hole is Now suppose Fr -GMmrjR (b) drilled straight through the centerof the earth, thus connecting two antipodal points on its surface.Let a particle of mass m be dropped at time t 0 into this hole with inibe its distance from the center tial speed zero,and let from NewConclude of the earth at time t (Fig. 2 where ton's secondlaw and part (a) that r\" (t) k 2 GMjR 3 gjR.)

r.

r(t)

2.4.13).

= -k r(t),

9.DeriveEq.(5)describing the motion


the bottom of a vertically First denote by the the unstretched position

x(t)

tial equation

for x. Then substitute

=x

Xo in this

differential

equation.))

10. onsidera floating C


gj

cylindrical buoy with radius height h, and uniform density p < (recallthat the density of water is 1 cm3 The buoy is initially suspended at rest with its bottom at the top surface of the water and is releasedat time t O. Thereafterit is acted on by two forces: a downward gravitational force equal to its n r 2hg and (by Archimedes' principle of weight mg 2 buoyancy) an upward forceequal to the weight xg of w water displaced, here x is the depth of the bottom of the buoy beneath the surface at time t (Fig. Conclude the buoy undergoes simple harmonic mothat tion around its equilibrium position Xe ph with period P 2n phjg Compute P and the amplitude of the motion if p gjcm3, h 200cm, and g 980cmjs2.)

0.5

r,

).

= x(t)

nr

2.4.12).

= ,J . = 0.5

(Problem 12). from part (b) that (c) Take g = ftjs2, and conclude the particle undergoes simple harmonic motion back and forth between the endsof the hole,with a period of about 84 min. (d) Lookup (or derive) the period of a satellite with the that just skims the surface of the earth; compare

A FIGURE2.4.13.mass m falling a hole through the centerof the earth

down

32.2

result in part

(c). How do you explain the coincidence? Or is it a coincidence? (e) With what speed(in miles)))

V 2.4Mechanical ibrations 147) per hour) does the particle passthrough the centerof the earth? (f) Look up (or derive) the orbital velocity of a satellite that just skims the surface of the earth; compare with the result in part (e). How do you explain the coincidence?Or is ita coincidence? Supposethat the mass in a mass-spring-dashpot system with m = 10, = 9, and k = 2 is set in motion with c x(O) = 0 and x'(0) = 5. (a) Find the position function x(t) and show that its graph looksas indicated in
Fig.
right function u(t) on the spring that would result if the mass motion with the same initial position and velocity, but with the dashpot disconnectedso c ( 0). Finally, construct a figure that illustrates the effect ofdamping and u(t). by comparing the graphs of

= Cocos(wot
were set in

- ao)
x(t)
Xo

13.

2.4.14. (b)
5 4 3 2
\037

beforestarting backtoward

Find how far the mass the origin.)

moves to

the

m 19. = 4, c = 20,k = 169; = 4, = 16 k 20.m = 2, c = 16, = 40; = 5, = 4 m k 21. = 1, = 10, = 125; = 6, = 50 c 22.A 12-lbweight (mass m = 0.375 slugs
Xo Vo Xo Vo Xo Vo

16. = 3, c = 30,k = 63; = 2, = 2 m 17.m = 1, = 8, k = 16; = 5, = -10 c m k 18. = 2, c = 12, = 50; = 0, = -8


\037,

m 15. =

c = 3, k = 4; Xo

= 2, = 0
Vo

Va

Xo

Vo

Xo

Vo

in fps units)

-1
-20 14. upposethat S
with with

10
t)

15

20

23.

x(t)of Problem 13.


x(O) = 20 and
m

FIGURE2.4.14. position The

function

is attached both to a vertically suspended spring that it stretches 6 in. and to a dashpot that provides 3 lb of resistancefor every foot per secondof velocity. (a) If the weight is pulled down 1 ft below its static equilibrium pofrom rest at time t = 0, find its position and then released sition function x(t). (b) Find the frequency, time-varying amplitude, and phaseangle of the motion. This problem dealswith a highly simplified model of a car of weight 3200lb (mass m = 100slugs in fps units). Assume that the suspension system acts like a single spring like a single dashpot, so that its and its shock absorbers

mass in a mass-spring-dashpot system and k = 226 is set in motion 25, c =


the

function

Fig.

2.4.15.

x (t) and show that its graph looksas indicated in o (b) Find the pseudoperiod f the oscillations

x'(0) = 41.(a)

10,

Find the position

and the equations in the figure.)

of the

\"envelope curves\" that

aredashed

vertical vibrations satisfy Eq.(4) with appropriate values of the coefficients. (a) Find the stiffness coefficient k of the spring if the car undergoesfreevibrations at 80cyw clesper minute (cycles/min) hen its shockabsorbersare disconnected.(b) With the shock absorbers onnected, c the car is set into vibration by driving it over a bump, and the resulting damped vibrations have a frequency of 78 cycles/min.After how long will the time-varying amplitude be 1% of its initial value?)

20 10
\037

\\

\"

Problems 24 through 34deal with a mass-spring-dashpot system having position function x (t) satisfying Eq. (4). We write Xo = x (0) and Vo = x'(0) and recall that p = c/(2m), p2. The system is critically w6 = k / m, and wi = w6

-10 / -20/ /
0
5)

in damped, overdamped, or un derdamped, as specified each

problem.

24. (Critically
10
t)

damped) Show

in this

casethat)
.)

15)

20) function

pt x(t)= (xo + vot + pxot)e-

FIGURE2.4.15. position The x(t)of Problem 14.)

25.(Critically

The remaining problems in this sectiondeal with free damped motion. In Problems a mass m is attached through to both a spring (with given spring constant k) and a dashThe mass is set in pot (with given damping constant motion with initial position Xo and initial velocity Vo. Find the position function and determine whether the motion is overdamped, critically damped, or underdamped. If it is underdamped, write the position function in the form Also, find the undamped position) x(t) C1 pt COS(WI t

15

21,
c).

Deducefrom Problem 24 that the = 0 at someinstant t > 0 if and if Xo and Va + pXo have opposite signs. only 26.(Critically damped) Deducefrom Problem 24that x(t) has a localmaximum or minimum at someinstant t > 0 if and
damped)

mass passes through x

27. (Overdamped)Show x(t) = \037


2y)

only if

Vo

and

Vo

+ pXo have the same sign.


in this

casethat)
(vo
r21

x(t)

= e-

-al).

where rl , r2

- -r,xo)e ] = - r2)/2 > = -p ::l:p2 J


[(vo

r2xO)erjl

wB and y

(rl

O.)))

148

2 Chapter
Xo

Linear Equationsof HigherOrder) 34. (Underdamped)


A body weighing 100 Ib (mass m = fps units) is oscillating attached to a spring and a dashpot. Its first two maximum displacements of in. in. are observed occurat times 0.34 6.73 and to sand s, respectively. Compute the damping con-

28.(Overdamped)If = 0, deducefrom Problem 27 that) X(t) = e- sinh yt.


Vo y) pt

3.125 slugs

in

29. (Overdamped) Prove that

30.

in this casethe mass can pass 0 at most once. through its equilibrium position x (Underdamped) Show that in this case)

1.46 1.17

stant (in pound-seconds foot) and spring constant per pounds per foot).)

(in

x(t)= e pI xOcosw)t+

Vo

+ pXo.
WI

smw)t.) )
that)

31.Underdamped) If the damping (


WI
\037

Differential Equations and Determinism Given a mass m, a dashpot constant c, and a spring 2.1 that the equation) k, Theorem 2 of Section implies
mx//

constant

constant parison with v'8mk , apply the binomial

c is small in com-

serieso show t

WO

- \037) .
8mk)

has a unique
tions X (0) and minima differential

= Xo,

solution

x/ (0) = vo.

+ cx/ + kx = 0) for t > 0 satisfying given

(26))
initial

Thus the future motion

condiofan ideal

32.(Underdamped)
of)

Show that the

localmaxima

X(t)

p = Ce-t COS(Wlt

occurwhere)

Conclude t2 tl = 2njWI if two consecutive axima that m occurat times tl and t2. 33.(Underdamped) Let Xl and X2 be two consecutivelocal maximum values of x(t).Deduce from the result of Problem 32 that) Xl = 2np
In
X2
\037

tan(wlt

- a) =

--.
p
WI

a))

system is completely determined by the mass-spring-dashpot equation and the initial conditions. Of coursein a real physical system it is impossible to measure the parameters m, c, and k precisely.Problems 35 through 38 explore the resulting uncertainty in predicting the future behavior of a

physical system.

= 1, = 2, and k = 1 in Eq.(26). how S c = 0 and x/ (0) = 1 t Xl (t) = te- . 36.Supposethat m = 1 and c = 2 but k = 1 10-2n.Show that the solution of Eq.(26)with x(O)= 0 and x/ (0) = 1 35.Supposethat
m
that the solution with X (0)
is)

WI

= 2npjWI is called the logarithmic The constant decrementof the oscillation. ote alsothat c = mWI \037jn N becausep = cj(2m).)
Note: The result of Problem 33 providesan accuratemethod for measuring the viscosity of a fluid, which is an important parameter in fluid dynamics but is not easyto measure directly. According to Stokes'srag law, a sphericalbody of radius a d moving at a (relatively slow) speed through a fluid of viscosity a experiences resistive force FR = 6n f.La v. Thus if a spherical mass on a spring is immersed in the fluid and set in motion, this drag resistance with damping condamps its oscillations stant c = 6naf.L. Thefrequency WI and logarithmic decrement can of the oscillations be measuredby direct observation. Thefinal formula in Problem 33 then gives c and hence the viscosity of the fluid.)
f.L
\037

n n = 10 t sinh 10t. e2n 37. Supposethat m = 1 and c = 2 but that k = 1 + 10-. Show that the solution of Eq. (26) with x(O) = 0 and x/CO)= 1 n n e-t X3(t) = 10 sin 10-

IS

X2(t)

is)

t.)

38.Whereas
shown in
with

in Fig. but N very long pseudoperiod. evertheless, show that for eachfixed t > 0 it is true that)

damped oscillations like those illustrated

and Figs.2.4.7

the graphs

of Xl (t)

the 2.4.8, graph

and

X2

(t)

resemblethose of X3(t) exhibits

2.4.9,

n\037oo

lim X2(t)

n\037oo)

lim X3(t)

= Xl (t).

tions

Conclude on a given finite time interval the three soluthat are in \"practical\" agreement if n is sufficiently large.)

and Nonhomogeneous Eqt:iations Undetermined C()efficients)


We

learned in Section how to solve homogeneousinear equations with constant 2.3 l but we saw in Section that an external force in a simplemechanical 2.4 coefficients, systemcontributesa nonhomogeneousterm to its differential equation. The general nonhomogeneous nth-order linear equation with constant coefficientshas the form)
any(n)

\037)

+ an_Iy(n-l) + ...+ alY' + aoy = f(x).)

(1))))

2.5Nonhomogeneous C Equationsand Undetermined oefficients 149)


By Theorem 5 of Section
\037)

a 2.2,general solution of Eq. (1)has the form)


y

= Yc +

Yp)

(2))

where the complementary function

h yc(x) is a general solutionof the associatedo(n

mogeneous equation)
an y (n)

+ an -1 -1)+ ... + a y' + aOY = 0,) y


1

(3))

Thus (x) is a particular solution of Eq. (1). our remaining task is to find y p. The method of undetermined coefficients is a straightforward way of doing this when the given function (x) in Eq. (1)is sufficiently simplethat we can make an intelligent guess as to the general form of y p. For example,suppose that f (x) is a polynomial of degree Then, because derivativesof a polynomial are the m. themselves polynomials of lower degree, is reasonableto suspecta particular it solution
and
yp

yp(x) = Amxm

+ Am_1x m -

+... +

A1x

+ Ao

is alsoa polynomial of degree but with as yet undeterminedcoefficients.We m, for may, therefore, substitute this expression y pinto Eq. (1),and then-by equating coefficientsof like powersof x on the two sidesof the resulting equation-attempt to determine the coefficients Ao, AI, ..., Am so that y p will, indeed,be a particular solution of Eq. (1). Similarly, supposethat)
that

f (x) = a coskx + b sin kx

.)

Then it is reasonable expect particular solution of the same form:) to a

+ yp(x) = A coskx B sinkx,)


a linear combination with undetermined coefficients A and B. The reason is that kx any derivativeof such a linear combinationof cos and sin kx has the same form. We may therefore substitute this form of yp in Eq. (1), then-by equating coand kx efficients of cos and sin kx on both sides of the resulting equation-attemptto determine the coefficients A and B so that y p will, indeed,be a particular solution. It turns out that this approach does succeed whenever all the derivatives of the (x) have the sameform as (x) itself. Beforedescribing method in full generality, we illustrate it with several preliminary examples.)

Example

1)

Find a particular solution of y\"

+ 3y' + 4y = 3x+ 2.)


yp(x) = Ax

Solution Here (x) = 3x+ 2 is a polynomial of degree so our guessis that) 1, f

+ B.)

Then

y\037

=A

and

y\037

= 0,so yp

will

satisfy the differential equationprovidedthat)

(0)+ 3(A) + 4(Ax + B) = 3x + 2,)


that is,)

(4A)x+ (3A + 4B) = 3x+

2)))

150

2 Chapter

Linear Equationsof HigherOrder)


for all x. This will be true if the x-termsand constant terms on the two sidesof this I equation agree.t therefore sufficesfor A and B to satisfy the two linear equations 4A == 3 and 3A + 4B == 2 that we readily solve for A == and B == /6 Thus we have found the particular solution)
\037

yp(x) ==
\"'''',h _,'_'\"

\037x

/6

.)

.)

..,.

,..\"....\"

.....,.,......\037.

....

\037 \037 ,..\"

...

...\"

,..

\037

...

\"

Example2)
Solution

Find a particular solution of y/l


Any

-4

\",..

\037'.....'-nn'.n.vu.....-\"\"..

3x == 2e

.)

derivativeof e3x is a constant multiple of e3x , so it is reasonable try) to

yp(x) == Ae
Then
y\037

3x
.)

== 9Ae 3x ,

so the given differential equation will be satisfied provided that)


9Ae3x

- 4(Ae -2
y

3x == )

3x 2e

;)

that

is, 5A

==

2,so that

A ==
\037.

Thus our particular solution is y p (x) ==

\037

e3x .

.)

Example3)
Solution

Find a particular solution of 3y\"


A first

+ y'

== 2 cos x.)

x, guessmight be yp(x) == A cos but

signalsthat

the presencef y' on the left-hand side o we probably needa term involving sin x as well. Sowe try)

yp(x) == Acosx+
y\037(x)
y\037

==

(x) ==

Bsinx; -A sin x + B cosx,


A

cos x

-B -

sin x.)

Then substitution 3(
that

cos B sin x) + (-A sin x + B cos x x)

of y p and its derivativesinto the given differential equation yields)


2(A

x cos + B sin x) == 2 cos x,)

is (collecting coefficientson the left),)

(-5A + B)cosx+ (-A 5B)sin x == 2cosx.)


This will be true for all x provided that the cosine sine terms on the two sides and of this equation agree.It therefore suffices for A and B to satisfy the two linear
equations)

-5A + B == 2,

with

readily found solution A ==


Y

' B == /3 . Hencea particular solution is) . == - 5 cos + SIn x.) x p (x)


t3

-A

- 5B

== 0)

13

13

.)

that

have made it appear.)


.,..\".\"......,..\".\"..'.. .n.nnn'\",,,', \"

The following example,which superficially resemblesExample indicates the method of undetermined coefficients is not always quite so simpleas we
n
\"... \" .0-' ....... \037 _ .,...,..\",,_ ...., \037

2,

Example4)

Find a particular solution of y\"

-4

2x == 2e

.. .......
.... 0....... .)))

2.5Nonhomogeneous C Equationsand Undetermined oefficients151)


Solution If we try
yp

(x) = Ae 2x , we find
y;

that)

4yp

= 4Ae2x

- 4Ae

2x

=0

=1=

2x 2e

.)

reasonable guessis)
for which)

A Thus, no matter how A is chosen, e 2x cannot satisfy the given nonhomogeneous 2x equation. In fact, the precedingcomputation showsthat Ae satisfies instead the associated homogeneousequation. Therefore, we should begin with a trial function 2x else yp(x) whosederivativeinvolves both e and something2x that can cancel upon substitution into the differential equation to leave the e term that we need. A

yp(x) = Axe

2x
,)

and y\037(x)=Ae2x+2Axe2x
Substitution into

2x y;(x)=4Ae +4Axe .
2X

the original differential equation yields

(4Ae

2x

+ 4Axe2x)

- 4(Axe
\037xe2x.)

2x

2x = 2e

.)

2x The terms involving xe2x obligingly cancel, leaving only 4Ae = Consequently,a particular solution is) A
\037.

2x 2e , so that)

yp(x) =

.)
2x

The General pproach A


Example4 resultedfrom the fact that f (x) = 2e satisfies the associated homogeneousquation. Rule 1, iven shortly, tells what to do when e g we do not have this difficulty, and Rule 2 tells what to do when we do have it. The method of undeterminedcoefficientsapplieswheneverthe function f (x) in Eq. (1) is a linear combination of (finite) products of functions of the following three types:
Our initial
difficulty in

2. An

1.A polynomial in x;
exponential function coskx or sin kx
.)

3.
Any

erx ;

(4))

such function-forexample,

f(x) = (3 4x2)e5x 4x3 coslOx,)


has the crucial property that only finitely many linearly independent functions appear as terms (summands) in f(x) and its derivativesof all orders.In Rules 1 and 2 we assumethat Ly = f(x) is a nonhomogeneous linear equation with constant coefficientsand that (x) is a function of this kind.)

RULE

1 Method Undetermined of Coefficients

the

yp a linear combination of all linearly independent such terms and their derivatives. Then determine the coefficients by substitution of this trial solution into the nonhomogeneousequation Ly = f(x).)))

Supposethat no term appearing either in f (x)or in any of its derivativessatisfies associated homogeneousquation Ly = O. Then take as a trial solution for e

152

2 Chapter

Linear Equationsof HigherOrder)


Note that this rule is not a theorem requiring proof; it is merely a procedure to be followed in searchingfor a particular solution y p. If we succeed finding y p, in then nothing more needbe said. (It can be proved, however,that this procedure will succeednder the conditions specified u here.) always In practicewe check supposition the made in Rule 1 by first using the charac-

teristic equation to find the complementary function Yc, and then write a list of all the terms appearing in (x) and its successive derivatives. If none of the terms in this list duplicates term in Yc, then we proceed a with Rule 1.)

\037

Example5)

Find a particular solution of)


y\"

+ 4y == 3x3

.)

(5))

Solution The (familiar) complementary solution of Eq. (5)is)


Yc(x) == Cl cos2x

+ C2sin 2x.)
try)

The function

f (x) == 3x3 and its derivativesare constant multiples of the linearly f B none independentunctions x3 , x2, x, and 1. ecause of theseappearsin Yc, we
+ Bx2+ Cx+ D, == 3Ax 2 + 2Bx+ C, y; == 6Ax + 2B.)
yp

== Ax 3

y\037

Substitution

in

Eq. (5)gives)

y;

+ 4yp == (6Ax + 2B)+ 4(Ax3 + Bx2+ Cx+ D) 2 == 4Ax 3 + 4Bx + (6A + 4C)x+ (2B+ D) == 3x3
4A == 3,
6A

.)

We

equate coefficientsof like powersof x in the last equation to get)

+ 4C == 0,)

4B == 0 , 2B + D == 0)
and

with

solution A ==

\037,

Eq. (5)is)

B == 0,C ==

\037,

D == O. Hence particular solution of a

== 3 3 yp (X ) 4x

9
gX.)

.)

Example6)

Solvethe

initial

value problem)
y\"

- 3y' + 2y
1,

==

x 3e- 10cos3x;

y(O) ==

y' (0) == 2.)


== 0

(6))

Solution The characteristic equation r 2


complementary function
is)

- 3r + 2

has roots r

1 and r ==

2,so the)

Yc(x) == Ctex+c2e2x.)))

C 2.5Nonhomogeneous Equationsand Undetermined oefficients 153) x x The terms involved in f (x) == 3e- - lOcosx and its derivativesare e- , cos3x, 3
and sin 3x.Because of theseappearsin none yp==
y\037

Yc,

we try)

+ Bcos3x+Csin3x, x == -Ae3Bsin3x 3Ccos3x, +


Ae-x
x

==
y\037

Ae- - 9B cos3x - 9C -

sin 3x.)

we substitute theseexpressionsnto the differentialequation in (6)and collect i we get) coefficients,


After

y;
We

3y\037

+ 2yp == 6Ae-x + (-7B 9C)cos3x + (9B-7C)sin3x x == 3e10cos3x.)

equate the coefficientsof the terms involving those involving sin 3x.The result is the system)

e-x , those involving cos3x,and


3,

-7B - 9C 9B - 7C
with

6A == == ==

-10,
0)

solution A == 4, B == }73 , and C == i3 This gives the particular solution)

-x 7 3 3 ( ypx) ==\"2 e +TIcosx+TIslnx,)


}

9.

which, however,does not have the requiredinitial values in (6). To satisfy thoseinitial conditions, e begin with the generalolution) w s

y(x) == Yc(x) + yp(x))

x 7 3 3 ==c}e+C2e +\"2e x +TIcosx+TIslnx,) with derivative) x y'(x) == c}e + 2c2e2x

2x}4

9.
i\037

- e- - ij sin3x+
x

cos3x.)

The initial conditions in (6)leadto the equations)


y

(0) == c}+ C2 + 4 + ?3 == 1,

y' (0) == c}+ 2C2 4 +


with

== 2)
i\037

solution c}== 4, C2 == }6 The desiredparticular solution is therefore) 3


y
} x (x) == -\"2e

7 9 . 6 + TIe2x + e-x + TI cos3x + TI SIn 3x.)


}
\"2

.)

.....

Example7)

Find the general form of a particular solution of)


y(3)

+ 9y' == x sin x + x2e2x

.)

(7))))

154

2 Chapter

Linear Equationsof HigherOrder)


==

Solution The characteristic equation r 3 + 9r == 0 has roots r


the complementary function
is)

0,r

==

-3i,

and r ==

3i.So

Yc(x) == Cl

+ Cz cos3x + C3sin 3x.)

The derivativesof the right-hand side in Eq. (7)involve the terms)

x cos , SIn x, xcos , x SIn x, x Zx zx e , xe , and xZezx


.)

Because is no duplication with there


trial

solution takesthe form)

the terms of the complementary function,

the

Yp(x) == A

zx cosx+ B sin x + Cxcosx+ Dxsin x + Eezx + Fxezx + Gxze

.)

Yp in Eq. (7)and equating coefficientsof like terms, we get seven equations determining the seven coefficients A, B,C,D,E, F,and G.

Uponsubstituting

.)

The Caseof Duplication


Now we turn our attention to the situation in which Rule 1 doesnot apply:Some of the terms involved in f(x) and its derivatives satisfy the associated homogeneous equation. For instance, supposethat we want to find a particular solution of the

differential equation)

(D r)3y == (2x 3)erx


a o Proceedings in Rule 1, ur first guesswould yp(x) == Ae Eq. (8)
is) rx

.)

(8))

be)
.)

rx + Bxe

(9))

This form of yp(x) will not be adequate because complementary function the
Yc(x) ==
rx rx cle + czxe + c3xzerx,)

of (10))
than

so

(2x 3)erx .

substitution

of (9) in the left-hand side of (8) would yield zero rather

To seehow to amend our first guess, e observethat) w


rx (D r)z[(2x 3)erx ] == [Dz (2x 3)]e

== 0)

by

2.3.If y(x) is any solution of Eq. (8) and we apply the Eq. (13)of Section to both sides,we seethat y (x) is also a solution of the equation operator (D r)z (D r)5 y == O.The general solution of this homogeneousequation can be written

as)

y(x) == Ct erx + czxe


\\.

rx

4 + c3xzerx+ Ax 3erx + Bx erx .


\037 \\..

Yc

yp

Thus every solution of our original equation in (8)is the sum of a complementary function and a particular solution of the form)
3 rx yp(x) == Ax e

4 + Bx erx

.)

(11))))

2.5Nonhomogeneous C Equationsand Undetermined oefficients 155)


Note that the right-hand side in Eq. (11) be obtained by multiplying each term can of our first guess in (9) by the least positive integral powerof x (in this case, 3) x that sufficesto eliminate duplication between the terms of the resulting trial solution This succeeds yp(x) and the complementaryfunction Yc(x) given in (10). procedure in the general case. To simplify the general statement of Rule 2,we observe to find a particular that solution of the nonhomogeneouslinear differential equation) Ly
it suffices to find tions)

= II(x) + Iz(x),)
Yz

(12))

separatelyparticular solutions Y I (x) and


Ly

(x) of the

two

equa(13))

= II(x)

and

Ly

= Iz(x),)

For respectively. linearity L[YI

then gives)

and therefore yp = Y I \"superposition principle\" for nonhomogeneouslinear equations.) Now our problemis to find a particular solution of the equation Ly = f(x), where (x) is a linear combinationof products of the elementaryfunctions listed in can be written as a sum of terms each of the form) (4).Thus

+ Yz] = LY + LYz = II(x) + Iz(x),) + Yz is a particular solution of Eq. (12).This is a type of (


1

I I(x)

Pm

(x)e coskx or
rX

Pm

(x)e

rX

sinkx,)

(14))

where Pm (x) is a polynomial in x of degree Note that any derivative of such m. a term is of the sameform but with both sinesand cosines ppearing. The procea dure by which we arrived earlier at the particular solution in (11) Eq. (8)can be of to is generalized show that the following procedure always successful.)
RULE

2 Method ofUndetermined Coefficients

If the function

I(x)is of either

form

in

(14),take as the trial

solution)
rX

o yp(x) = xS[(A + Alx

+ Azx z + (Bo + B1x Bzxz + +

+... + ...

Amxm)e

+ Bmxm)e

rX

coskx sinkx],

(15))

where s is the smallestnonnegativeinteger such that such that no term in yp duplicatesa term in the complementaryfunction Yc. Then determinethe coefficients in Eq. (15)by substituting yp into the nonhomogeneousequation.)

(x) exhibiting the full practicewe seldomneed to deal with a function in (14). he table in Fig.2.5.1 the form of y p in various common T lists generality t m cases, correspondingo the possibilities = 0,r = 0,and k = O. On the other hand, it is common to have)
In

I(x)= II(x) + Iz(x),)


where II(x) and Iz(x) are different functions of the sort listed in the table in In Fig.2.5.1. this event we take as yp the sum of the trial solutions for II(x) and Iz(x),choosings separatelyfor each part to eliminate duplicationwith the complementary

function. This procedure illustrated is

in

8 Examples through

10.)))

156

2 Chapter Linear Equationsof HigherOrder)

I{x}
,)

YP)

Pm

= bo + b1x + b2x 2 + ...+ bmx m


a coskx + b sin kx rx e (a coskx + b sin kx)
Pm Pm
rx (x)e

+ Amxm) + A 2x 2 S coskx X (A + B sin kx) rx X Se (A cos kx + B sin kx) 2 S X (Ao + A1x + A 2x + + Amxm)e rx
xS(A o + A1x xS[(A o

+...
...

cos (x)(a kx + b sin kx))


in the method

+ A1x + Amxm) coskx + Bmxm) sinkx]) + (Bo + B1x

+... +...

FIGURE 2.5.1. Substitutions


\037 'N' ..:-: ....N 'u.,.... '''\"N N..N..U..,......,.... N N \037.-N\037.... .

of undetermined coefficients.)

Nn

\"\037..\037.,,\037.,,,..,...'...'\037nN''''.''\"''A-'....N..,._..,_.

\037

,\037

,...

....

Example8)

Find a particular solution of)


y(3)

+ y\"

x == 3e

+ 4x2.)
== r2 == 0 and r3 ==

(16))

Solution The characteristic equation r 3 + r 2 == 0 has roots rl


complementary function
is) Yc(X) == CI

s -1,o

the

+ C2X + C3e-x
2

.)

As a first

step toward our particular solution, we form the


(Ae ) + (B + Cx+ Dx ).)
X

sum)

The part Ae X corresponding to

H we duplication. ence take)


Yp

tary

function,

but

the part

x 3e does not duplicate any part of the complemenB + Cx + Dx2 must be multiplied by x2 to eliminate

y\0373)

+ Bx2 + Cx3 + Dx4, 2 3 == Ae x + 2Bx+ 3Cx + 4Dx , 2 == Ae x + 2B + 6Cx+ 12Dx , Y; == Ae x + 6C+ 24Dx.)
== Ae x
y\037

and

Substitution

of thesederivatives in Eq. (16)yields) 2Aex


x 2 + (2B+ 6C)+ (6C+ 24D)x+ 12Dx == 3e + 4x2
.)

The systemof equations)


2A ==

6C+
has the solution A == solution is)
\037,

3,

2B + 6C== 0,
== 12D 4)

24D == 0,)

B == 4, C ==
Yp(x) ==

\037,

and

D == 1.Hence desiredparticular the

\037eX

+ 4x2

\037x3

+ 1x4

.)

.)))

2.5Nonhomogeneous C Equationsand Undetermined oefficients 157)


........

Example9)

Determine the appropriate form for a particular solution of)


y\"

+ 6y' + 13y = e-3x cos2x.)


3x

= Solution The characteristic equation r 2 + 6r + 13 0 has roots -3::f:: so the complemen2i,


yc(x) = e- (CI cos2x+ C2 sin 2x). 3x This is the same form as a first attempt e- (A cos2x + B sin 2x) at a particular we solution, so we must multiply by x to eliminate duplication. Hence would take)
tary function is)

+ yp(x) = e- X(Axcos2x Bxsin2x).)


3
N,\"NN.\",'N,'NN.'n\037\"'N,\"\",... ,-';\"\" '\" \"'''..........nN,.''...... ....;\".\"...... .........\"....n.....,..\".... .......\"\"....\"\" \" \",..\" \"''''''' ...........'A...........\" \"\"\"\"\"'......... ;\"\".,..... ... \037 .... '\" ,..... ........ \037 .... \037._.\", .....,_\"....... _., ....... \037v

.)
equation)

Example10)

Determine the appropriate form for a particular solutionof the fifth-order


2 (D 2)3(D + 9)y = x2e2x + x sin3x.)

Solution The characteristic equation (r 2)3(r2 + 9) = 0 has roots r = 2,2,2, 3i,and


so the complementary function
is)

-3i,

3x + yc(x) = Cte2x + C2xe2x c3x2e2x+ C4cos + C5sin 3x.)


As a first

steptoward the form of a particular solution, we examine the


[(A

sum)

2x + Bx+ Cx2)e ] + [(D+ Ex)cos3x+ (F + Gx)sin 3x].)

Toeliminate duplication with terms of Yc (x),the first part-correspondingx 2e2x to


must
3

be multiplied by x , and the secondpart-corresponding sin 3x-must e b to x we multiplied by x. Hence would take) yp(x) =
(Ax
3

4 + Bx + Cx5)e2x + (Dx+ Ex2)cos3x+ (Fx+ Gx2) sin 3x. .)

Variation Parameters) of
let us point out the kind of situation in which the method of undetermined coefficientscannot be used.Consider, or example,he equation) f t
Finally,
II

+y=tanx,)
the
many linearly

(17))

which at first amples.Not


derivatives)

t so;he function f (x) = tan x has infinitely 2 2 sec x tan

glancemay appearsimilar to thoseconsideredin

expreceding
independent

2 sec x,

x,

4 2 4 sec x tan 2 x + 2 sec x,)


trial

Therefore, we do not have available a finite linear combination to use as a


solution.)))

158

2 Chapter Linear Equationsof HigherOrder)


We discusshere the method of variation of parameters, which-inprinciple is, if the integrals that appearcan be evaluated)-can always be usedto find a

(that

particular solution of the nonhomogeneouslinear differential equation)


yen)

+ Pn_l(X)y(n-l)

+... +

Pl(X)y' + po(x)y == f(x),

(18))

provided that we already know the general solution)


Yc

== ClYl

+ czyz + ... +

CnYn)

(19))

of the associated homogeneousquation) e


yen)

+ Pn_l(X)y(n-l)

+... +

Pl(X)y' + Po(x)y == o.)

(20))

i Here,n brief, is the basicideaof the method of variation of parameters. Supwe replace constants, or parameters,Cl, Cz, ..., Cn in the complementhe pose function in Eq. (19)with variables: functions U 1,Uz, Un of x. We ask tary whether it is possibleto choose thesefunctions in such a way that the combination) that

...,
(X)

yp(x) == Ul (X)Yl (x) + UZ(X)yZ(X)


this

+ . + Un (X)Yn

..

(21))

is a particular solution of the nonhomogeneousequation in (18). turns out that It is always possible. The method is essentially samefor all ordersn > 2,but we will describe the it in detail only for the case == 2.Sowe beginwith the second-order onhomogen n neousequation)
\037)

L[y] == y\"
complementary function)

+ P(X)y' + Q(x)y == f(x)) + czyz(x))


P and
Q are continuous. We

(22))

with

Yc(x) == ClYl(X)

(23))
want to

on someopen interval I where the functions find functions uland Uz such that)
\037)

yp(x) == Ul(X)Yl(X)

+ uz(x)yz(x))

(24))

is a particular solution of Eq. (22). B One condition on the two functions Ul and Uz is that L[yp] == f(x). ecause two conditions are requiredto determine two functions, we are free to imposean additional condition of our choice. will do so in a way that simplifies the comWe as much as possible. first, to imposethe condition L[yp] == f(x),we But putations
must

compute the derivatives


1

y\037

and

y\037.

The product rule gives)


1

yp

== (UlYl 1

+ uzyz ) + (U1Yl + uzyz )


I I

.)

Toavoid the appearance the secondderivativesu { and of the additionalcondition that we now imposeis that the secondsum here must vanish:)
u\037,

'

U1Yl

+ uzYz
1

== 0

.)

(25))))

2.5Nonhomogeneous C Equationsand Undetermined oefficients 159)


Then)
Yp

, == , UIYI

, + U2Y2')

(26))

and the product rule gives)


\"

Yp

== (UIYI \"

+ U2Y2 ) + (UIYI + U 2Y2 )


\"
\" \"

.)

(27))

But both

YI

and

Y2

satisfy the

homogeneousquation) e
y\"

+ Py' + Qy == 0)
\"

with the nonhomogeneousequation in (22), o) associated s


Yi

==

-P ' - Q
Yi
I
Y\037

Yi)

(28))

for i ==

2 1, .It therefore follows from Eq. (27) - P . (u + == +


Y\037

that)

(u'lY\037

u;Y\037)

U2Y\037)

-Q.

(u IYI

+ U2Y2).)

In view of Eqs.(24)and

(26),this means that)


\"

Yp

== (uIYI
\"

' + u ,2Y2') PYp


\"

- -Q
+u2
\"

YP')

hence)

L[YP ] ==uIYI
The requirement that

Y2.)

(29))

impliesthat) L[yp] == f(x)-therefore


\"

Yp

satisfy the nonhomogeneousequation in

is, (22)-that that


(30))

uIYI

+ u 2Y2 == f(x.) )
\"

Finally, Eqs. (25)and (30)determine the these Collecting equations, we obtain a system)

functions

UI

and

U2 that

we need.

,
UIYI
\037)
\"

uIYI

+ U ,2Y2 == 0, + u 2Y2 == f(x))


\"

(31))

of two linear equations in the two derivatives u'l and Note that the determinant of coefficients in (31)is simply the Wronskian W (YI, Y2). Once we have solved the equations in (31)for the derivatives u'l and u;,we integrate each to obtain the functions uland U2 such that)
\037)

u;.

Yp

== UIYI

U2Y2)

(32))

is the desiredparticular solution of Eq. (22). In Problem 63 we ask you to carry


out
this theorem.)))

process explicitly and thereby verify

the formula for Y p (x) in the following

160

2 Chapter Linear Equationsof HigherOrder)

1 THEOREM Variationof
by)

Parameters
+ P(x)y'+ Q(x)y == f(x) has comple+ C2Y2(X), then a particular solution is given
Y2 (X )

If the nonhomogeneous equation y\"


mentary function Yc(x) == CIYl (x)

Yp

(x)

-...\"

Yl (X )

Y2(x)f(x) X + d
W(x)

Yl(x)f(x)dx,
W(x))

(33))

where W == W(Yl, Y2) is the Wronskian of the two independent solutions Yl and Y2 of the associated homogeneousquation.) e
. Find a particular solution of the equation y\"
..\037... ...

Example11)

+ Y == tan x.)

..,.

\037

,......

.....

\037

Solution The complementary function is Yc(x) == Cl cosx+ C2sinx,and we couldsimply substitute directly in Eq. (33).But it is more instructive to set up the equations in (31)and solve for u; and u;,so we beginwith)
Yl

, == Yl

== cos x,

-.

SIn x,)

== sIn x, , == COS x.) Y2


Y2

Hence equations in (31)are) the

We

easilysolve theseequations for)


u
1

,.
==

+ (u;)(sinx)== 0, == (u;)( sinx) + (u;)(cosx) tanx.)


x) (U'l)(cos

- SInx

tan

x ==

U2

, == COS tan x

. x == SIn x

x - cosx)
sin 2
.)

== cosx

- secx,

Hence take) we
Ul

f (eosx

- seex)dx = sinx -In seex +


I

tanxl)

and) U2

t (Do you seewhy we choosehe constants of integration to be zero?) Thus our

f sinxdx = - eosx.
U2 )Y2 tan

particular solution is)


Yp(x) == U 1 (X)Yl
==
that is,)

+ -(x)Isec(x x+ (sinx
In

(x)

xI) cos + (-cos x x)(sinx);)

Yp(x) ==

In + -(cosx) Isecx tanxl.)

.)))

2.5Nonhomogeneous C Equationsand Undetermined oefficients 161) _Pro\037lem\037)


In

Problems

given equation.

In all tives with respectto

1through

x.

20,find a particular solution y p of the theseproblems, primes denotederiva-

(b)

Usethe

result

of part (a) to
y\"

find

a general

solution

of)

3.
5.
7.

1.

sinx + x cosx 18. (4) 5y\" + 4y == eX 2xe2x y 19. (5) + 2y(3) + 2y\" == 3x 1 y 20.y(3) Y == eX + 7)
==

11. 13. 15.(5) + 5y(4) y


17.y\" + y

10.

9.

+2y' 3y 1 +xex + 9y == 2 cos3x + 3 sin 3x 1 y(3) + 4y' == 3x y\" + 2y' + 5y == eX sin x


==

y\" y\" y\" y\" y// y\"

+ 16y == e3x

- -6 + + - 4y y/ y/ y
==

==

2 sin 3x
2

==

sin

sinh

2. y\" - y' - 2y 3x + 4 4. 4y\" + 4y/ + y 3xeX 6. 2y\"-+ 4y/ + 7y x 2 8. y\" 4y cosh2x


== == == == == == X

3 + 4y == cos x.) to find

Use trigonometric
equations
in

identities

Problems 44 through 46.

generalsolutions of the

==

17

--

12.y(3) + y/ 2 - sinx 14.y(4) - 2y\" + y xe 16.y\" + 9y 2x2e + 5


==

44. y\" + y' + y == sinxsin 3x 45. y\" + 9y == sin 4 x 3 46. y\" + y == x cos
X)

3x

In

Problems 47 through 56, use the method of variation ofparameters to find a particular solution of the given differential

equation.

Problems through 30, set up the appropriate form of a particular solution YP' but do not determine the values of the
In

21

21.\" --2y' + 2y y
22.y(5) y(3) 23.y'/ + 4y
y(3)
y// y\" y\"
== ==

coefficients.

30.y(4)

x - 24. 25. + 3y' + 2y x(e- - e- ) 26. - 6y' + 13y xe 2x 27. + + 4y x + cos2x 28. + (x + 1) 3x 29.(D -1)3(D- 4)y xe + e + e==

eX sin

+ 2x2 5 3x cos2x 3x 12y'== X 2xe2x


eX
==
X

47. y\" + 3y' + 2y == 4ex 49. y\" 4y' + 4y == 2e2x 51.// + 4y == cos3x y 53.y\" + 9y == 2 sec3x 55.y\" + 4y == sin 2 x

2x 48. y\" 2y' 8y == 3e== sinh 2x 50. y\" 4y 52. y\" + 9y == sin 3x 2 54. y\" + y == csc x 56. y\" 4y == xe

- ==

57. You

can verify

by substitution

that Yc

CIX

+ C2X-1 is a
second-

complementary function for the nonhomogeneous

==

3x sin sin

order equation) x 2y\"


But

y(4) y(4)

5y\" 9y\"
2y\"

==

==

sin

+ xy/

==

72x5

.)

==

2x

2x

+y

==

x 2 cos

X)

Solvethe

initial
==

value problems in
== == ==

31 Problems through 40.


== == == ==

the method of variation of paramemust first divide this equation by its leading coefficient x 2 to rewrite it in the standard form)

beforeapplying

ters, you

31.

2 2x;y(O) 1, y\" + 4y 32.y\" + 3y' + 2y eX; y(O)y'(O)y'(O) 3 0, 33.y\" + 9y sin 2x;y(O) 1, /(O) 0 y 34. y\" + y cosx; (O) 1, /(O) -1 y y 35.y\" - 2y' + 2y 2 x + 1; (O) 3, y/(O) 0 y 36.y(4) -4y\" x ; y(O) y'(O) 1,\"(O) y(3) (0) -1 y x 37. y(3) - + y/ 1 + xe ; y(O) y/(O) 0, y\"(O) 1 38.\" + 2y' + 2y sin 3x; y(O) 2, y/(O) 0 y 39. (3) + y\" x + e- ; y(O) 1, ' (0) 0, y\" (0) 1 y - Y 5; y(O) x y' (0) y\"y(0) y(3) (0) 0 40. y(4) 41. ind a particular solution of the equation F
== == == == == == == == == == == ==

\"

+ xy

1, - - x1
2)

==

72x3

Thus
Yp
==

f(x)
3x5
.)

==

equations

in

a (31)nd

N to 72x3 in Eq. (22). ow proceed solve the thereby derive the particular solution

2y\"

==

==

==

==

==

==

==

==

==

==

==

==

==

==

==

==

y(4)

y(3)

y\"

_ y' _ 2y == 8x5

Problems 58 through 62, a nonhomogeneous second-order linear equation and a complementary function Yc are given. Apply the method of Problem 57 to find a particular solution
In

42. Find the solution of the initial value problem of the differential equation of Problem 41 and
conditions)
y

consisting the initial

(0) ==

y/

(0) ==

y\" (0) == y(3) (0) ==

O.)

43. (a)

Write

- 6y + x3 - 4xy' + 4y x ;; x + c2 ) 3xy' + 60.4x - 4xy' + 3y 8x / ; + x3/ 61. + + x + x; cos(1nx) 62.(x - - 2xy' + 2y x - 1; + c2(1+ x )
58. y\" 59.x 2y/'
2 2
y\"
==

of the

equation. 2 x

Yc Yc

==

CIX

C2

==

==

(Cl

lnx

y\"

==

43

Yc

==

CIX

C2

xy/

==

In

Yc
==

==

CI

C2sin(lnx)

cos3x + i sin 3x == e3ix


by
inary

==

+ (cosx i sinx)3

63.Carry

l)y\"

Yc

==

CIX

Euler'sformula, expand, and equate real and imagparts to

derive the

derive the
==
\037

identities

3 cos x sin 3 x

cosx i cos3x, +
sin

Eqs.(31) (32). 64. Apply the variation of parameters


the particular solution y p geneous equation y\" + y

out the solution processindicated in the text to variation of parameters formula in (33) from and

==
\037

-i

(x)
==

==

sin 3x.)

2 sin

-xcosx of the nonhomox.)))

formula in

(33)to

find

_
I I

162

2 Chapter Linear Equationsof HigherOrder)

ForcedOscillations Resonance) and


In
\037)

Section we derived the differential equation) 2.4


mx\"

+ ex'+ kx = F(t))

(1))

motion of a mass m that is attached to a spring governs the one-dimensional constant k) and a dashpot (with constant c) and is also acted on by an external force F(t).Machineswith rotating components commonly involve mass-spring systems(or their equivalents) in which the external force is simpleharmonic:)
that (with
Equilibrium position

F(t) = Focoswt or F(t) = Fosinwt,)

(2))

l)

FIGURE2.6.1. cartThe
with-flywheel

system.)

where the constant Fo is the amplitude of the periodicforce and w is its circular frequency. For an exampleof how a rotating machine component can provide a simharmonic force, considerthe cart with a rotating vertical flywheel shown in ple The cart has massm mo, not including the flywheel of massmo. The Fig. centroid of the flywheel is off center at a distancea from its center, and its angular T speed is w radians per second. he cart is attached to a spring (with constant k) as shown. Assume that the centroid of the cart itself is directly beneath the center of the flywheel, and denoteby x(t) its displacementrom its equilibrium position f the spring is unstretched). igure 2.6.1 us to seethat the displacement (where F helps x of the centroid of the combinedcart plus flywheel is given by)

2.6.1.

_= x

(m

- mo)x + mo(x + a cos


m
mx\"

wt)

= x + moa cos wt.


m)

the Let us ignore friction and apply Newton'ssecondlaw m x\" = -kx,because force exertedby the spring is -kx. substitute for x in the last equation to obtain) We

- moaw cos
2

wt

= -kx;)
.)

that is,) mx\"

+ kx = moaw2 coswt

(3))

Thus the cart with its rotating flywheel acts like a mass on a spring under the influence of a simpleharmonic external force with amplitude Fo = moaw2 Sucha model of a front-loading washing machine with the clothes system is a reasonable T beingwashedloadedoff center. his illustrates the practical importance of analyzing solutions of Eq. (1)with external forcesas in (2).)

Forced Oscillations Undamped


To study undamped oscillations under the influence of the external force F(t) coswt, we set c = 0 in Eq. (1), thereby beginwith the equation) and Fo
\037)

mx\"

whosecomplementary function
\037)

+ kx = Fo coswt) is Xc = Cl cos + C2sin wot. Here) wot


wo= (f)))

(4))

and 2.6Forced scillations Resonance 163) O


(as in Eq. (9)of Section is the (circular) natural frequencyof the mass-spring 2.4) in radians reminds system.The fact that the angle wot is measured (dimensionless) in in us that if t is measured seconds(s),then Wo is measured radians per secondthat dif that is, in inverse seconds(S-I). lso recallrom Eq. (14)in Section2.4 A vision of a circular frequency w by the number 2n of radians in a cyclegives the correspondingordinary) frequency v == w/2n in Hz (hertz == cyclesper second). ( Let us assumeinitially that the external and natural frequencies are unequal: w -1= WOo We substitute xp == A cos in Eq. (4) to find a particular solution. (No wt sineterm is neededin xp becausehere is no term involving x' on the left-hand side t in Eq. (4).)This gives)

-mw2A cos + kA cos wt wt


so) A== k

== Fo cos wt,)

Fo
mw 2) Fo/m

Fo/m

2 w o)

w2

'

(5))

and

thus)

wt. xp(t) == W 2 -w)2 cos o

(6))

Therefore, the general solution x == Xc


wot x(t) == Cl cos

+ xp is given
o)

by)

Fo/m + C2sin wot + w 2 -w2 coswt,

(7))

Equivalently,

where the constants Cl and C2 are determined by the initial values x (0) and as in Eq. (12)of Section we can rewrite Eq. (7)as) 2.4,
Fo/m x(t)==Ccos(wot-a)+-w2 coswt, 2 w
o)

x'(0).
(8))

of one so we seethat the resulting motion is a superposition two oscillations, with


natural
\"\"N,'N'.'nn_'n\037n\037n\037,-\" '\" '\" '\" ..

circular frequency Wo, the other with the frequency w of the externalforce.)
\" '\" '\" N \"\"\"\"\" \"'\037\"\" A.' \037 .....,\037_ \037..... \037 \"\"'\"'

,,,
_n..

\037

Example

1)

Supposethat
IS)

m ==

k 1, == 9, Fo == 80, and w == 5,so the differential equationin (4)


'\" \037\"N\"\"\"\" \037\"\"\" N '\" '\" \"\"\"\", A.' \037\037N' \037\037 N' \"\"\"\", ..,\"\"\",......\".,,,'\" '\" \"\"'\"' \"'N\037 '\" '\" ... '\" \". \"'.. .....

\037

...

\037

x\"

5t.) + 9x== 80cos

Find

x(t) if x(O)== x'(0) ==


natural

O.)

Solution Here the

frequency Wo == 3 and the frequency w == 5 of the externalforce 5t are unequal, as in the precedingdiscussion.First we substitute x p == A cos in Thus a the differential equation and find that -25A+ 9A == 80, o that A == s particular solution is

-5.

xp (t)

==

-5cos5t.

The complementaryfunction is Xc == Cl cos 3t the given nonhomogeneousequation is)

+ C2sin 3t, so the general solution of

x(t) == Cl cos3t + C2 sin 3t


with derivative)

- 5 cos

5t,)

x'(t)

==

s -3Clin 3t + 3C2cos3t + 25 sin

5t.)))

164
15 10
5)

2 Chapter
Period = 2n
II I I I

Linear Equationsof HigherOrder)


The initial conditions x(0) == 0 and desiredparticular solution is)

x'(0) == 0 now yield Cl == 5 and C2 == 0, so the

-I
I I I
I)

x(t) == 5 cos3t

- 5 cos

5t.)

\037

0)

-5

As indicated in Fig. the periodof x(t) is the least common multiple terms. periods2n13 and 2n15of the two cosine

2.6.2,

2n of the

.)

-10 -15
o
n 2n 3n
t)

Beats)
4n 5n 6n

FIGURE2.6.2. response The


x(t)

If we imposethe
that)

initial

conditions x(0) == x'(0) == 0 on the solution in (7),we find


C1 ==

= 5 cos3t

- cos
5

5t

in

Example 1.)

m(w6

Fo

-(

2)

and

C2 ==

0,

so the particular solution is)

x(t) ==
A == \037(wo

Fo 2 m(w o

w 2)

(coswt

The trigonometric identity 2 sin A sin B ==

+ w)t and B ==
x(t) == m

\037(wo

w)t, enablesus to rewrite Eq. (9)in the form)


1 sin 2(wO

- coswot). cos(A- B) - cos(A+ B),applied 1 w)t sin 2(wO

(9))

with

2Fo
2 (wo

w2

+ w)t.
Iwo

(10))

))

Then sin \037(wo


slowly
with varying

Supposenow that

+ w)t

\037

Wo,

so that

Wo

circular frequency \037(wo

function, function. We may therefore interpret


varying

is a rapidly

+w is very large in comparison with

+ w),)

a whereassin \037(wo w)t Eq. (10)as a rapid oscillation

- -is

wi.

x(t) == A(t)
but with

sin (wo
\037

+ w)t,)

a slowly varying amplitude


A

(t) == m

2Fo
2 (wo

1 sIn 2 (wo w2
))

w) t

Example2)
1.5 1.0 0.5 I
\037

With m ==

0.1, == 50,Wo == 55,and w == 45, Eq. (10)gives Fo


x(t) == sin 5t sin 50t.
\037

x = sin 5t
I
\\ I)

\\

I)

showsthe corresponding scillationof frequency (wo + w) == 50 that o Figure 2.6.3 is \"modulated\" by the amplitude function A(t) == sin 5t of frequency \037(wo-w) == 5.

A rapid oscillation with a (comparatively)slowly varying periodicamplitude exhibitsthe phenomenon of beats. For example,if two horns not exactly attuned to one another simultaneously play their middleC, one at wo/(2n) == 258 Hz and -1.0x I= sin 5t sin 50t) == the other audible variation in the Hz, then one -1.5 0.5 1.0 1.5 2.0 2.5 3.0 amplitudeat w/(2n) 254sound-withahears a beat-an of the combined of) 0.0 frequency

0.0

.)

-0.5

\\

\\

\\

\\

\\

\\)

I)

\\../

\\.

t)

FIGURE2.6.3. The phenomenon of beats.)

(wo

- w)/2

258

- 254
2)))

2n)

== 2 (Hz).

2.6ForcedOscillations Resonance 165) and Resonance


Lookingat Eq. (6),we seethat the amplitude A of xp is large when the natural and external frequencies and wareapproximately equal. It is sometimes useful to Wo rewrite Eq. (5)in the form)
A== k

Fo
mw
2)

Folk (wlwO)2

==

p Fo ::i: k

')

(11))

where Folk is the static displacementa spring with constant k due to a constant of force Fo, and the amplification factor p is defined to be)

p- 11It is clear that p the increase without


\037 \037
----+ ----+

1
(wlwo)21.)

(12))

+00as w wo. This is the phenomenon of resonancebound (as w wo) in the amplitude of oscillationsof an undamped system with natural frequency Wo in responseto an external force with frequency w wo. We have been assuming that w i= WOo What sort of catastrophe should one if expect wand Wo are,precisely equal?Then Eq.(4),upon divisionof each term by m, becomes)
2 x/I + wox == Fo coswot.
m)

(13))

Because wot is a term cos


1.5)
....)

of the complementary function, the method of undetermined coefficientscallsfor us to try)


....) ....)

1.0) 0.5)
\037

Xp(t) == teA We substitute this in

cos + B sin wot).) wot


that A == 0

0.0)

Hence particular solution is) the

Eq. (13)and thereby find


Fo

and

Fo/(2mw o).)

-0.5) -1.0)

0.000.250.500.75 1.001.251.50 The graph of xp(t) in Fig.2.6.4 which m == 1,Fo == 100, Wo == 50)shows and (in how the amplitude of the oscillation theoretically would increasewithout vividly FIGURE2.6.4. The bound in this case pure resonance, == wo. We may interpret this phenomenon of w phenomenon of resonance.) as reinforcement of the natural vibrations of the system by externally impressed
t)

-1.5

.......)

xp(t) == 2mwo) t SIn wot.


.......

(14))

vibrations at the

samefrequency.)

Example3)

Then the

Supposethat

natural

Hz.We

would therefore expect oscillations very large amplitude to occur if of revolves at about 95 revolutionsper minute (rpm). flywheel

m == 5 kg and k == 500 N/m in the is Wo == -Jklm == frequency

10rad/s; that
little

cart with the flywheel of Fig.

is, 10/(2n)
\037

2.6.1.
1.59
the

(1.59)(60)
\037

resonanceibrations. A spectacular xamplecan occurwhen a column of soldiers v e marches in stepover a bridge. ny complicated structure such as a bridgehas many A natural frequencies f vibration. If the frequency of the soldiers' o cadence approxis to one of the natural frequencies f the structure, then-just as in our o imately equal will I simpleexampleof a masson a spring-resonance occur. ndeed,the resulting)))

In practice,a mechanical system with very

damping can be destroyedby

166

2 Chapter Linear Equationsof HigherOrder)


resonanceibrations can be of such large amplitude that the bridgewill collapse. v This has actually happened-for o example,the collapsef Broughton Bridgenear it is Manchester, ngland, in 1831-and the reasonfor the now-standard practice E of breakingcadence when crossing bridge. Resonance a may have been involved in the 1981 ansasCity disasterin which a hotel balcony (called skywalk) colK a of lapsedwith dancerson it. The collapse a building in an earthquake is sometimes due to resonanceibrations causedby the ground oscillating at one of the natural v frequenciesof the structure; this happenedto many buildings in the MexicoCity a o 1 On earthquake of September19, 985. occasionn airplane has crashedbecause f resonant wing oscillations causedby vibrations of the engines.It is reported that for some of the first commercial jet aircraft, the natural frequency of the vertical vibrations of the airplane during turbulence was almost exactly that of the masshead R springsystemconsistingof the pilot's (mass)and spine (spring). esonance occurred, causingpilotsto have difficulty in reading the instruments. Largemodem commercial jetshave different natural frequencies, o that this resonance s problem no longer occurs.)

M Modeling echanical Systems)


The avoidance of destructive resonanceibrations is an ever-present consideration v in the design of mechanical structures and systemsof all types. Often the most important step in determining the natural frequency of vibration of a system is the formulation of its differential equation. In addition to Newton'slaw F == ma, the useful for this purpose(as in the principleof conservation of energy is sometimes derivation of the pendulum equation in Section 2.4).The following kinetic and formulas are often useful. potential energy

1.Kinetic energy:T == 1mv2 for translation of a massm with velocity v; 2. Kinetic energy: T == 1/w 2 for rotation of a body of a moment of inertia / with angular velocity w; 3.Potential energy: V == 1kx2 a spring with constant k stretched or comfor
presseda distancex; 4. Potential energy:V = mgh for the gravitational potential energy of a massm
constant.) g may be regardedas essentially at height
h

above the reference level (the level at which

= 0),provided that

Example4)

Find the

natural frequency of a mass m on a spring with constant k if, without friction, it is a uniform disk of radius a that rolls without sliding as shown in Fig

instead of

2.6.5.)

slipping,

Solution
Equilibrium position
I I
I)

With

the

precedingnotation, the principleof conservationof energy gives 1mv2 + 1/w2+ 4kx2 == E

where E is a constant (the total mechanical energy of the system). We note that v == aw and recallhat / == ma 2/2 for a uniform circular disk. Then we may t the last equation to) simplify
\037mv2

\037kx2

==

E.
with

x =0

FI GURE

2.6.5. rolling The

disk.

Because right-hand sideof this equation is constant, differentiation the to t (with v == x' and v' = x\") now gives III I 3 2mx x + k xx == 0
.)))

respect

and 2.6Forced scillations Resonance O


We

167)

divide eachterm by

\037mx'

to obtain

x + 3m
\"

-x
2k

==

o.

natural frequency of horizontal back -and-forth oscillation of our rolling disk is y'2k/3m, which is y'2/3 0.8165 the familiar natural frequency times y'k/m of a masson a spring that is slidingwithout friction rather than rolling without sliding.It is interesting(and perhapssurprising) that this natural frequencydoesnot dependon the radius of the disk. It couldbe either a dime or a large disk with a .) radius of one meter (but of the samemass).

Thus the

\037

Example5)

Supposethat a car oscillatesvertically as if it werea massm == 800 kg on a single 4 spring (with constant k == 7 X 10N/m),attached to a singledashpot (with constant == 3000 os/m). Supposethat this car with the dashpot disconnected driven is c N
along a washboard road surface with an amplitude of 5 cm and a wavelength of L == 10 (Fig.2.6.6).what car speedwill resonanceibrations occur?) v m At
m)

m)
\037}\037)

c)

Surface)

y=acosT)
s=o)
Equilibrium position)

2ns

_=Iy=-)

road surface of Example 5.)

FIGURE2.6.6. washboard The

T FIGURE2.6.7. he\"unicycle model\" of a car.)

Solution

of the car as a unicycle, as pictured in Fig.2.6.7. x(t) denote the Let o displacement f the mass m from its equilibrium position; we ignore the force of gravity, because merely displaces equilibrium position as in Problem it the 2.4. 9 of Section We write the equation of the road surface as
We think

upward

== a

cos

27TS
L)

m m). (a == 0.05 , L == 10

(15))

When the

car is in motion, the spring is stretched by the amount secondlaw, F == ma, gives)
mx\" ==

y,

so Newton's

-k(x

y);)

that is,) mx\"

+ kx ==
in

ky)

(16))

If the velocity of the car is v, then s == vt


, mx' + kx

so Eq. (15), Eq. (16)takes the form)


27TVt
L)

== ka cos

( 16'))))

168

2 Chapter Linear Equationsof HigherOrder) This is the differential equation that governs the vertical oscillations the car. In of with circular comparing it with Eq. (4), we seethat we have forced oscillations will occur when w == Wo == ,Jklm We use our frequency w == 2nvjL. Resonance numerical data to find the speedof the car at resonance:)

==

-If- -\037
L

2n

==

10

104

2n

\037

800)

14.89 (m/s);

that

is, about 33.3 ijh (using the conversionfactor of 2.237 ijh per m/s). m m

.)

Forced Oscillations Damped


In real
nothing
\037)

physicalsystemsthere is always some damping, from frictional effects if

else. complementary function The


mx 1/
I

Xc

of the equation)
17

Thus Xc is a transientsolution of Eq. (17)-one dies out with the passageof that time, leaving only the particular solution X p. The method of undeterminedcoefficientsindicates we should substitute) that
wt x(t) == A cos in

+ cx + k X == rocoswt) (17)) is given by Eq. (19), or (21)of Section depending whether c > Cer = on 2.4, (20), ,J4km , C == Cer' or C < Cere The specificform is not important here. What is t 0 as t +00. important is that, in any case, hese formulas show that xc(t)
\037 \037

+ B sin wt)

W wt terms, and equate coefficientsof cos and Eq. (17). hen we do this, collect sin wt, we obtain the two equations) (k

- m(
(k

)A

+ cwB == Fo,)
for)

-cwA + (k

- m( )B
2 2

== 0)

(18))

that

we solve without
A==

difficulty

- m( )2 + (cw)2
(k

- m( )Fo
2 2

')

B==
(k

cwFo - m( )2 +

(cw)2)

(19))

If we write)
A

coswt + B sin wt

== C (cos wt

cos + sin wt a

sin a) == C cos (wt

a))

as usual, we seethat the resulting steady periodicoscillation)

xp(t) == C cos(wt
has amplitude)

a))

(20))

c = .JA2 + B2 =
Now (19)impliesthat sin a == BjC the first or secondquadrant. Thus)
tan

.J(k

- m( )2 +
with

Fa 2

(21))
(cw)2)

> 0,so it follows that


mw 2)

the

phaseangle a lies in

a ==

B
A

== k

-cw

< a < n,

(22))))

and 2.6Forced scillations Resonance O


so)
tan

169)

-1
k

ot=)
1T

-cw
1

mw 2

if k if k

> mw 2 < mw
2)

,)

+ tan- k-mw) 2

cw

as c > 0,then the \"forcedamplitude\"-defined a function C(w) in of remains finite, in contrast with the case resonance the unby (21)-always casewhen the forcing frequency w equals the critical frequency Wo = damped -Jklm But the forced amplitude may attain a maximum for somevalue of w, in which casewe speak of practicalesonance.To seeif and when practical resr onance occurs,we need only graph C as a function of wand look for a global function maximum. It can be shown (Problem 27)that C is a steadily decreasing of w if c > y'2km But if c < y'2km, then the amplitude of C attains a maximum value-and practical resonance ccurs-at o somevalue of w less than wo, so and then approaches zero as w +00.It follows that an underdamped system w typically will undergo forced oscillations hoseamplitude is) Note that if

if ot (whereas = 1T12 k = m(2).

\037

. . .

to Largeif w is close the critical resonance frequency; Closeo Folk if w is very small; t Very small if w is very large.)
.... ........... ............

Example6)

of Find the transient motion and steady periodicoscillations a damped mass-andwith m = c = and k = 26 under the influenceof an external spring system force F(t) = 82 cos with x(O)= 6 and x'(0) = O.Also investigatethe possibility 4t of practical resonance this system.) for

1,

2,

Solution The resulting


problem)

motion

x(t) =

Xtr(t)

+ xsp(t)of the mass satisfies the initial


x(O)= 6, x'(0) = O.

value

x\"

+ 2X' + 26x= 82 cos4t;

(23))

i Insteadof applying the general formulasderivedearlier in this section,t is better in a concrete roblemto work it directly. The roots of the characteristicequation) p

r2

+ 2r + 26 = (r + 1)2+ 25 = 0)
function is)

are r

-1::l::so the complementary 5i,


the
trial solution)

5t xc(t)= e-t (Cl cos + C2 sin 5t).)


When

we substitute

x(t) = A cos4t+ B sin4t)


4t the given equation, collect terms, and equate coefficientsof cos and sin 4t, like we get the equations)
in

+ 8B = 82, -8A+ lOB =


lOA
0)))

170

2 Chapter Linear Equationsof HigherOrder) with solution A == 5, B == 4.Hence the general solution of the equation in (23)is)
X(t) ==
At this

e-t (C1cos5t + C2sin 5t) + 5 cos4t + 4 sin 4t.)


initial

point we

C1 == 1 and C2 ==

imposethe

oscillation the massare given by) of

x(O) (0) -3.Therefore, the transient motion and the steady periodic

conditions

==

6,x'

== 0 and find that

Xtr(t) ==

e-t (cos5t - 3 sin 5t))

and)

4t xsp(t)= 5 cos + 4 sm 4t
==

,J41 cos(4t
\037

. =.J41
a))

(.J4T

cos + 4t

.J4T

sm 4t

where a == tan- 1 0.6747. 2.6.8 graphs of the solution x(t) == Xtr(t) shows Figure value problem)
(\037)

+ xsp(t)of the initial


(24))

x\"

+ 2X' + 26x== 82cos4t,x(O)== Xo, x'(O)== 0

for the different values Xo == 20, a 0,10,nd 20of the initial position.Here we seeclearly what it means for the transient solution Xtr(t) to \"die out with the passageof time,\" leaving only the steady periodicmotion xsp(t).Indeed,because Xtr(t) --+ 0 exponentially, within a very few cyclesthe full solution x(t) and the steady periodicsolution xsp(t)are virtually indistinguishable(whatever the initial

-10,

positionxo).)
x)

10)

t)

-10)
-20)

FIGURE2.6.8. Solutions of the with Xo = -20, and 0, 10,

-10,

initial

value problem in

(24)

20.)))

and 2.6Forced scillations Resonance O


10 9 8
7
Practical resonance)

171)

To investigate the possibilityof practical resonancein the given system, we c = and k = 26 in (21)and find that the forced substitute the values m = at frequency w is) amplitude

1,

2,

6 u 5 4 3 2
1

C(w) =
The graph of C(w) is shown
5)

,J676- 48w +
2

82

w 4)

.
amplitude

in 3

00)

10
w)

15

20)

CI (w)

Plot FIGURE2.6.9. of amplitude C versus external


frequency
w.)

-41 - (676- -+ (

The Fig.2.6.9. maximum

occurswhen)

(4w 96w) 4)3/2) 48w2

(676 48w

2 -164w(w 24) = o. 2 4

+ ( )3/2)

Thus practical resonanceccurs when the external frequency is w = J24(a bit less o than the mass-and-spring's undamped critical frequency of Wo = ,Jk/m = ,J26).

.)

IfD Problems)
In Problems through 6, express the solution of the given initial value problem as a sum of two oscillations in Eq. as Throughout, primes denotederivatives with respectto time t. In Problems graph the solution function x (t) in such a its peway that you can identify and label (as in Fig. riod.)

(8).

1-4,

2.6.2)

Each of Problems 15through 18 gives the parameters for a + system with equation mx\" +cx' forcedmass-spring-dashpot kx = Fo coswt. Investigate the possibility ofpracticalresonance of this system. In particular, find the amplitude C(w) w forcedoscillations ithfrequency w. Sketch of steady periodic the graph of C(w) and find the practicalresonance frequency
w

cos + 9x = 10 2t; x(O)= x'(0) = 0 2.x\" + 4x = 5 sin 3t; x(O)= x'(O)= 0 3.x\" + 100x = 225cos5t + 300sin 5t; x(O)
x\"

1.

(if any).

375, x'(O)= 0 4. x\" + 25x = 90cos4t; x (0) = 0, x'(0) = 90 5. mx\" + kx = Fo coswt with w i= Wo; x (0) = Xo, x'(0) = 0 6.mx\" +kx = Fo coswt with w = Wo; x(O)= 0,x'(O)=
Vo)

m 15. = 1,c = 2, k = 2, Fo = 2) 16. =1,c=4,k=5,F m o =10)

17.m=1,c=6,k=45,F o =50) m k 18. = 1, = 10, = 650,Fo = 100) c


A 19. mass weighing units) 1 in. by mass.At what frequency (in hertz) will tions occur?Neglect amping.) d

In each of Problems 7 through lution xsp(t) C cos(wt

cx'

find the steady periodicsothe given equation mx\" + with periodic + kx forcing function offrequency w. Then graph xsp(t) together with (for comparison) the adjustedforcing function F] (t) F(t)jmw.

= F(t)

- a)10, of

is

F(t)

100lb (mass m = slugs in fps to the end of a spring that is stretched a forceof 100lb. A force Fo coswt acts on the
attached

3.125

o resonance scilla-

8.x\" + 3x'+ 5x = -4cos5t 9.2X\" + 2x'+ x = 3 sin lOt x 10.\" + 3x'+ 3x = 8 coslOt + 6 sin lOt)
In steady

7. x\" + 4x' + 4x = lOcos 3t

20.A front -loading washing machine is mounted on a thick rubber pad that acts like a spring; the weight W = mg the (with g = 9.8 mjs2) of the machine depresses pad exactly

each of Problems through 14,find and plot both the periodicsolution xsp(t) = C cos(wt a) of the given differential equation and the transient solution Xtr(t) that sat-

11

its rotor spins at w radians per second, newtons on the machine. At what speed revolutions per minute) will (in vibrations occur?Neglect friction.) resonance the rotor

cm. 0.5 When

exertsa vertical force Fo coswt

isfies the given initial


x\"

conditions.

21.igure 2.6.10 shows F

11. + 4x' + 5x = lOcos3t; x(O)= x'(O)= 0

12.\" + 6x'+ 13x= 10 5t; x(O)= x'(0) = 0 x sin 13.\" + 2x'+ 26x = 600cos x lOt;x(O)= 10, = 0 x'(O) x 14.\" + 8x'+ 25x = 200cost + 520sin t; x(O) = -30, x'(O)= -10)

on the end of a pendulum horizontal spring (with constant k). Assume small oscillations of m so that the spring remains essentially horizontal and neglect damping. Find the natural circular frequency Wo of motion of conthe mass in terms of L, k, m, and the gravitational stant

a mass m

(of length L) also attached to a

g.)))

172

2 Chapter

Linear Equationsof HigherOrder)


In particular, show that it is what one would sameas the formula in (20)with the same values

26.

in placeof cos(wt sin(wt Given the differential equation)

w,

exceptwith
mx \"

- a)

expect-the
of C -a).
and

D' + ex' + kx = E0 coswt + 1'0SIn wt)


sine
forcing

-with both cosine and


steady

terms-derive the

periodicsolution)

FIGURE2.6.10. pendulumThe and-spring system of Problem 21.

xsp(t) =
where

J(k -

+ F.2 0 a 2 + m( )2 (ew)2)cos(wt
\302\2432

- -

fJ),

22. mass m hangs A

a is defined

in

Eq. (22) and

fJ

tan- 1 (Fo/Eo).

on the end of a cord around a pulley of radius a and moment of inertia as shown in Fig. Therim of the pulley is attached to a spring (with constant so k). Assume small oscillations that the spring remains essentially horizontal and neglectfriction. Find the natural circular frequency of the system in terms of m, a, k, and g.)

I,

2.6.11.
I,

25).) 27. According

(Suggestion: Add the steady periodicsolutions separately corresponding to Eo coswt and Fo sin wt (seeProblem to

periodicoscillations for
Fo coswt is given
by)

Eq. (21),the

amplitude of forced steady the system mx\" + + kx

ex'

C(w) = (a) If e >


nance) when)
W

J(k

-m(
ecr

Fo
2

)2 + (ew)2)
-vi 4km

ecr/\037' where

a steadily decreasess w increases. (b) show that C attains a maximum value

, show that C If e < ecr/\037, (practical reso-

= Wm =

J m - \037 < 2m
\037

Wo

= fk.
V

;;)

FIGURE
A 23. building

2.6.11. mass-springThe
of Problem 22.

28.) As

pulley system

consistsof two floors. The first floor is attached rigidly to the ground, and the secondfloor is of mass m = 1000 slugs (fps units) and weighs tons as (32,000 Theelasticframe of the building behaves a lb). spring that resistshorizontal displacements of the second the floor; it requires a horizontal forceof 5 tons to displace secondfloor a distance of 1 ft. Assume that in an earth-

indicated by the cart-with-flywheel example discussed section,an unbalanced rotating machine part typically results in a forcehaving amplitude proportional to the square of the frequency w. (a) Show that the amplitude of the steady periodicsolution of the differential
in this

16

equation)
mx\"

+ ex'+ kx = mAw 2 coswt)


to that in

(with

a forcing term similar

i Eq.(17))s given

by)

quake the ground oscillateshorizontally with amplitude Ao and circular frequency w, resulting in an external horizontal force mAow2 sin wt on the secondfloor. (a) What is the natural frequency (in hertz) ofoscillations of the secondfloor? (b) If the ground undergoes one oscillation every s with an amplitude of 3 in., what is the amplitude of the resulting forcedoscillations f the o

F(t) =

C(w) =
(b) Supposethat e2
amplitude

J(k -m( )2
<
k m

mAw 2 2 +

(ew)2)

2mk. Show that the maximum


Wm

2.25 =

occursat
wm

the frequency

given

by)

secondfloor?

24. A mass on a spring

25.

damping is acted on by the 3 external forceF (t) Fo cos wt. Show that there are two values of w for which resonance occurs,and find both. Derive the steady periodic solution of) without

=)

2mk 2mk e2 )

.)

mx \"

17. + ex' + kx = 1'0SIn wt.)

Thus the resonance frequency in this caseis larger (in contrast with the result of Problem 27) than the natural fre-vlk/m (Suggestion: Maximize the square quency Wo

of C

.))))

2.7 Electrical Circuits 173)


Automobile Vibrations Problems 29 and 30 deal further with the car of Example 5. Its upward displacement function satisfies the equation mx\" + cx'+ kx = cy' + ky when the shock absorberis connected(sothat c > 0).With y = a sin wt for the roadsurface, this differential equation becomes
where v amplitude slightly over 5 cm. Maximum resonanceibrations with amplitude about 14 cm occuraround 32 mi/h, but then subside to more tolerable levels at high speeds.
Verify

the function

nance frequency

thesegraphically basedconclusions by analyzing C(w ). In particular, find the practical resoand the corresponding amplitude.)

+ cx'+ kx = Eo coswt + Fo sin wt) Eo = cwa and Fo = ka.


mx\"
\037 '-'\"

15 12

29. Apply the result of Problem 26 to show that the amplitude C of the resulting steady periodicoscillationfor the car is
given
by)

e u
::::s

(1) \"'0

9 6 3

c=
Because w

30.

2.6.12 the graph of the amplitude function shows C(w) using the numerical data given in Example 5 (including c = 3000 N.s/m). It indicates that, as the car accelerates radually from rest, it initially oscillates with) g
m\",

v, this Figure

gives C as a function of v.

= 2nv/L when

J - m( )2 + (cw)2 car is
(k
2 the

aJP + (cw)2

.\037

\037

<r:

moving with velocity

00

20

40

60

80

100

Velocity (mi/h))

FIGURE2.6.12. Amplitude of vibrations of the car on a washboard surface.)

\037Electrical Circuits \037\"..


UN< _\".\" _,_\"_\",\"_\037,, __\037, ,,__..,_m_)

o
Switch)

Here examine the RLC circuit that is a basicbuilding blockin more complicated we electrical circuits and networks.As shown in Fig.2.7.1, it consists of)
L)

a resistance R ohms, of an inductanceof L henries,and A capacitor with a capacitance of C farads)


An

w resistor ith

inductorwith

R)

FIGURE2.7.1. series The RLC


circuit.)

a source of electromotive force (such as a battery or a generator) suppliesa voltage of E(t) volts at time t. If the switch shown in the circuit of Fig.2.7.1 closed, resultsin a current of I (t) amperesin the circuit and is this a charge of Q(t) coulombson the capacitor at time t. The relation between the
in that

series with

functions Q and
<;ir\037ui\037

I is)

,..YQJ\037ftgl,>

ElemeDt)
Inductor) Resistor)

::,lJi!Op..<

dl Ldt RI 1

\037;

= I(t).)

(1))

Capacitor)

C Q)

units, in which time is measured in seconds. always use mks electric to elementary principlesof electricity, the voltage drops across According the three circuit elements thoseshown in the table in Fig.2.7.2. can analyze are We the behavior of the seriesircuit of Fig.2.7.1ith the aid of this table and one of c w Kirchhoff's laws:

We will

FIGURE2.7.2. Tableof voltage


drops.)

The (algebraic) of the voltage dropsacrossthe elements in a sum of an electrical circuit is equal to the appliedvoltage. simpleloop
As a consequence, current and charge in the the satisfy the basiccircuit equation)
\037)

simpleRLC circuit of Fig.2.7.1

1 dI L-+RI+-Q=E(t). dt C)

(2))))

174

2 Chapter Linear Equationsof HigherOrder) If we substitute (1)in Eq. (2),we get the second-order linear differential equation)

LQ\"+ RQ' + C)

1 -Q= E(t)

(3))

for the charge Q(t),under the assumption that the voltage E(t) is known. In most practical problemsit is the current I rather than the charge Q that is of primary interest, so we differentiate both sidesof Eq. (3)and substitute I for Q'
to
\037)

obtain)

LI\"+ RI'+
We

-I= E'(t).
1
C)

(4))

do not assumehere a prior familiarity with electrical circuits.It sufficesto circuit as \"black boxes\" resistor,inductor, and capacitor in an electrical that are calibrated by the constants R, L, and C.A battery or generator is described by the voltage E (t) that it supplies.When the switch is open,no current flows in the there is a current I (t) in the circuit and a charge circuit; when the switch is closed, on the capacitor.All we need to know about theseconstants and functions is Q(t) that they satisfy Eqs. (1)through (4), our mathematicalmodel for the RLC circuit. We can then learn a gooddealabout electricity by studying this mathematicalmodel.)
regard the

TheMechanical-Electrical Analogy
It
\037)

is striking

that

Eqs. (3)and (4) have preciselythe sameform as the equation)


mx\"

+ ex'+ kx = F(t))

(5))

of a mass-spring-dashpotystem with external force F(t).The table in Fig.2.7.3 s detailsthis important mechanical-electrical most analogy. As a consequence, of the resultsderived in Section for mechanical systemscan be appliedat onceto 2.6 electrical circuits.The fact that the samedifferentialequation servesas a mathematical modelfor such different physical systemsis a powerful illustration of the unifyof ing role mathematics in the investigation of natural phenomena. Moreconcretely, the correspondencesFig.2.7.3 be used to construct an electrical in model of a can mechanical system,using inexpensive and readily availablecircuit elements. given The performance of the mechanical system can then be predictedby means of accurate but simplemeasurements in the electrical T model. his is especiallyuseful when the actual mechanical system would be expensive construct or when meato surements of displacements velocities would be inconvenient, inaccurate, or and even dangerous.This ideais the basis of analogeomputers--electrical of models mechanical systems.Analog computers modeledthe first nuclear reactors for commercial powerand submarine propulsion before the reactors themselves werebuilt.)

Me\037II.a.rdc\037....$\037$\037gdt..!....i?!.!..........;.......)

':'!\"li_{II_\037{\"\":tt.)
c
Inductance L R Resistance 1/ Reciprocal apacitance C c Charge Q (using (3) (or current I using (4))) Electromotive forceE (or its derivative E'))

Massm
Damping constant Spring constant k
Position

ForceF)

FIGURE 2.7.3. Mechanical-electrical analogies.)))

2.7 Electrical Circuits 175)


takesthe form)
In the typical case an of
alternating

current voltage E(t) = Eo sinwt, Eq.(4)

LI\"+ RI'+
As

-I= wEocoswt.
1
C)

(6))

c periodicurrentIsp ; thus)

in a mass-spring-dashpot ystem with a simpleharmonic external force, the s solution of Eq. (6) is the sum of a transientcurrent Itr that approaches zero as t +00(under the assumption that the coefficients in Eq. (6) are all positive, so the roots of the characteristic equation have negative real parts), and a steady
\037

I = Itr + Isp.)

(7))

soluRecall from Section (Eqs.(19)through (22)there) that the steady periodic 2.6 tion of Eq. (5)with F(t) = Fo cos is wt

xsp(t)=
where)

Jcw

Fo cos(wt m(2)2 (k

- a)

+ (cw)2)

a = tan -1 2 k-mw) '


If we make the substitutions the steady periodiccurrent)

0 <a

< Jr.

L for m, R for c, IIC for k, and wEo for Fo, we get Eo cos(wt a)

Isp(t)=

J
with

R2

(WL

(8))

W\037

r)
Jr.
(9))

the

phaseangle)

-1 wRC 2 a=tan 1 - LCw) '


Reactancend Impedance a
The quantity
in

< -) -a <

the denominator in (8),)

z=

R2

(WL

W\037

(ohms),)

(10))

is calledthe impedance the circuit. Then the steady periodiccurrent) of Isp(t)=


has amplitude) 10 = reminiscent of Ohm'slaw, I =
EIR.)))

Eo
Z)

cos(wt a)

(11))

Z')

Eo

(12))

176

2 Chapter

Linear Equationsof HigherOrder)


function, whereas Equation gives the steady periodiccurrent as a cosine the input voltage E(t) == Eo sin wt was a sine function. To convert Isp to a sine function, we first introduce the reactance)
R)

(11)

S == wL
-5)

-.
1
wC)
Z)

(13)) delay

Then Z == ,JR2 + S2,and we seefrom Eq. (9)that now angle 8 == a Equation (11) yields)

- !n.

with a is as in Fig.2.7.4,

5)

lsp(t) =
==

R)

D=a-!!
2)

SR. -(-- - )
Eo
Z)

a wt (cos cos
Z

+ sin a sin wt)


Slnwt

Eo
Z

coswt+

FIGURE2.7.4.Reactance and
delay angle.)

=
Therefore,)

Eo
Z)

(cos8 sin wt

sin 8 cos wt).

Isp(t)==

Eo
Z)

sln(wt

- 8), 1 wRC)

(14))

-1
I I I
I)

where)

r-\037
I I I

8 == tan- 1

S ==

tan- 1

LCw2

(15))
Isp

o This finally gives the time lag 8/w (in seconds) f the steady periodiccurrent behind the input voltage (Fig.2.7.5).)

Initial alue Problems) V


current behind

FIGURE2.7.5.Time lag of imposed voltage.)

When

to obtain the equation)

I (0)and Q(O).SOwe must

we want to

find

the transient current, we are usually


first find

I'(0).To do so,we substitute

given the

initial t == 0 in

values

Eq. (2)

1 LI'(0)+ RI (0)+ C) Q(O)== E(O) to determine

(16))

I'(0)in terms of the initial


==

values of current, charge, and voltage.)

Example

1)

Consideran RLC circuit with R


5
4 x 10-farad (F).At

50 ohms (Q), L == henry (H),and C when both (0) and Q (0)are zero, the circuit is connectedto a Hz alternating current generator. Find the current in the V, circuit and the time lag of the steady periodiccurrent behind the voltage.)

110- 60-

time

t ==

0,

0.1

Solution

S Hz frequency of 60 means that w == (2n)(60)rad/s,approximately377 rad/s. o we take E(t) == 110 377t and useequality in place f the symbol for approximate o sin The equality in this discussion. differential equation in (6)takes the form)
A == cos + SOl' (0.1)1\" + 20001 (377)(110) 3771.)))

2.7 Electrical Circuits 177)


We substitute the given is impedance Z ==

59.58, so the steady periodicamplitude is) Q


10==

values of R, L, C,and

w == 377 in

Eq. (10)to find

that the

59.58 (ohms))

110 (volts)

==

1.846 amperes(A).

With

the

samedata, Eq. (15)gives the sinephaseangle:)


== 8 == tan- 1 (0.648) 0.575.)

Thus the time lag of current behind voltage is)


8)

0.575 ==
377)
is)

w)

0.0015 s,

and the steady periodiccurrent Isp ==

(1.846) sin(377t 0.575).)


\037

== The characteristic equation (0.12 + 50r+ 2000 0 has the two roots'l )r -44and r2 -456. ith theseapproximations,the general solution is) W
\037

44t I(t)== cle- + c2e-456t+ (1.846) sin(377t 0.575),) with derivative)

I'(t) == -44cl -44t - 456c2e-456t+ 696 e cos(377t 0.575).)


Because (0) Q(O) == 0,Eq. (16)gives I'(0) == 0 as well. With 1 ==
values substituted, we obtain the equations)
= == 1(0)= Cl + C2 1.004 0, I'(0) == -44cl 456c2+ 584 == 0;)

these initial

their solution is Cl ==

Thus -0.307, == 1.311. the transient solution is) C2


Itr(t) ==
44t 456t + (-0.307)e- (1.311)e.)

The observation that after one-fifth of a secondwe have Itr (0.2) < 0.000047 A t (comparableo the current in a singlehuman nerve fiber) indicates that the transient solution dies out very rapidly, indeed.
I I

.)

Example2)

still Supposethat the RLC circuit of Example1, with I (0) == Q (0) == 0, is connectedat time t == 0 to a battery supplying a constant 110. Now find the current V in

the

circuit.)))

178

2 Chapter

Linear Equationsof HigherOrder)


We now

Solution

have E(t) =

so 110, Eq. (16)gives) = I, = E (0) = 110 1100

(0)

0.1)

(A/s),

and the differential equation is)

= + SOl' (0.1)/\" + 20001 E'(t) = O.)


Its general solution is the complementary function we found
44t I (t) = cle- + c2e-456t.) in

Example1:)

We

solve the equations)

= 1(0) Cl + C2= 0, = 1'(0) -44cl 456c2= 1100) = for Cl = -C2 2.670. Therefore,) 44t I (t) = (2.670)(e- e-456t).) Note that I (t) --+ 0 as t --+ +00 even though the voltage is constant.)

.)

Electrical Resonance)
Consideragain the current differential equation in (6)corresponding a sinusoidal to E (t) = Eo sin wt. We have seenthat the amplitude of its steady input voltage periodiccurrent is)
10)

10=

EO

R 2 + (w L -Y

10 =
w1c

2)

Eo

Eo

R2

(wL

w\037

(17)) Y)

Wm

W)

T FIGURE2.7.6. heeffectof
frequency on 10.)

For typical valuesof the constants R, L, C,and Eo,the graph of 10as a function of (J) value at Wm = 1jv'LC and then approaches zero as w --+ +00;he critical frequency Wm is the resonance t frequency of the circuit. In Section we emphasized importance of avoiding resonance most in 2.6 the mechanical systems(the cello an exampleof a mechanical system in which resois nance is sought).By contrast, many common electrical devicescouldnot function The radio is a without taking advantage of the phenomenon of resonance. properly familiar example. highly simplified model of its tuning circuit is the RLC circuit A we have discussed.ts inductance L and resistanceR are constant, but its capaciI tance C is varied as one operatesthe tuning dial. Supposethat we wanted to pick up a particular radio station that is broadcasting at frequency w, and thereby (in effect) provides an input voltage E(t) = Eo sin wt to the tuning circuit of the radio.The resulting steady periodiccurrent Isp in the tuning circuit drives its amplifier, and in turn its loudspeaker, with the volume of sound we hear roughly proportional to the amplitude 10of Isp.To hear our preferred station (of frequency w) the loudest-and simultaneously tune out stations at other frequencies-we therefore want to choose to maximize C broadcasting

resembles one shown in Fig.2.7.6.reaches maximum the a It

10.)))

2.7 Electrical Circuits 179)


But examine Eq. (17),hinking of w as a constant with t at a calculusrequired-that is maximal 10

glance-no

wL
that

- wC
1

C the only variable. We see


when)

==

o.
')

is, when) C ==

-.
Lw
2)

(18))

set the capacitance this value. to old crystal radiosworked,but modem AM radios have a more sophisticated A are design. pairof variable capacitors used.The first controls the frequency selected described as the secondcontrols the frequencyof a earlier; to signal that the radio itself generates,kept close 455 kilohertz (kHz)above the desiredfrequency. The resulting beat frequency of 455 kHz,known as the intermediate frequency, is then amplifiedin several stages. his techniquehas the advantage T that the severalRLC circuits usedin the amplificationstageseasily can be designed to resonateat 455kHz and reject ther frequencies, o resulting in far more selectivity of the receiver as well as better amplificationof the desiredsignal.)
the dial to
that

Sowe merely turn

This is the way

Proble ms
1through
an

\"_)

RL circuit of Fig. 2.7.7, a with an inductance of L containing henries, a resistorwith a resistance R ohms, and a sourceof of electromotive force (emf), but no capacitor. n this caseEq. (2) I reducesto the linear first-orderequation) Problems

6 deal with

seriescircuit

the inductor

LI'+ RI = E(t).)

3. Supposethat the battery in Problem 2 is replacedwith an alternating-current generator that supplies a voltage of E(t) = 100 60t volts. With everything elsethe same, cos now find I(t). with 4. In the circuit of Fig. 2.7.7, the switch in position 1,
lOt

L)

supposethat L = 2, R = 40, E (t) = 1OOe- , and 1(0)= O. Find the maximum current in the circuit for t > O. with 5. In the circuit of Fig. 2.7.7, the switch in position 1, lOt supposethat E(t) = 1OOe- cos60t, R = 20, L = 2, = and 1(0) O.Find I(t). with 6.In the circuit of Fig. 2.7.7, the switch in position 1, and take L = 1, = 10, E(t) = 30cos60t 40sin60t. R + (a) Substitute Isp (t) = A cos60t + B sin 60t and then
determine A and B to
the circuit.

R)

FIGURE2.7.7.The circuit
Problems 1 through 6.)

for

C cos(wt a).)

(b)

find the steady-state current Isp in Write the solution in the form Isp(t)

1.In the circuit

of Fig. that suppose L 5 H, R 25 Q, and the sourceE of emf is a battery supplying 100V to the circuit. Supposelsothat the switch has beenin poa sition 1 for a long time, so that a steady current of 4 A is 0, the switch is thrown flowing in the circuit. At time t to position 2, so that 4 and E 0 for t > O.Find samecircuit as in switch is initially in position at time t 0, so that Find (t) and show that (t)

2.7.7,

2.Given the I

I(t).

1(0)=

10 with the RC circuit in Fig. 2.7.8, deal a resistor (R ohms), a capacitor(C farads), a containing switch, a sourceof emf, but no inductor. Substitution of L = 0 in Eq. (3) gives the linear first-order differential equation)
Problems 7 through

Problem 1, upposethat the s 2, but is thrown to position 1 1(0)= 0 and E = 100for t > O. I 4 as t +00.)
\037 \037

dQ 1 R-+-Q=E(t) dt C)
for the chargeQ = Q(t) on the capacitorat time t. Note

I(t) = Q/(t).)))

that

180

2 Chapter Linear Equationsof HigherOrder)


SWitch

1
for

C)

R)

FIGURE2.7.8. he circuit T
Problems7 through 10.)

12.R = 200 Q, L = 5 H, C = 0.001 F; E(t) = 100sin lOt V 13.R = 20 Q, L = 10 C = 0.01 H, F; V E(t) = 200cos5t 14.R = 50 Q, L = 5 H, C = 0.005F; E(t) = 300coslOOt + 400sin lOOt V 6 15.R = 100Q, L = 2 H, C = 5 X 10-F; sin E(t) = 110 601(tV 4 16.R = 25 Q, L = 0.2 C = 5 X 10-F; H,
E(t) = 120cos377t
V)

11. = 30 Q, L = 10 C = 0.02; E(t) = 50sin2tV R H, F

7. (a) Find
if

current (t) in the RCcircuit voltage supplied by a battery) and the switch is closedat time t = 0, so that Q(0) = O. (b)

E(t) = Eo (a constant
lim
t\037+oo

the

charge Q(t) and

In

Show that)

Problems 17through 22, an RLCcircuit with input voltF age E(t) is described. ind the current I(t) using the given initial current (in amperes) and chargeon the capacitor(in coulombs ).)

Q(t) = EoC

and that

lim
t\037+oo)

I(t) = O.
5t

8.Supposethat in the circuit of Fig. 2.7.8, have R = 10, we C = 0.02, = 0, and E(t) = 100eQ(O) (volts). (a) Find Q(t) and I(t). (b) What is the maximum charge on the capacitoror t > 0 and when doesit occur? f 9.Supposethat in 4 the circuit of Fig. 2.7.8, = 200, R C = 2.5X 10- Q(O) = 0, and E(t) = 100cos 120t. In each of Problems 20 through 22,plot both the steady peri, (a) Find Q(t) and I(t). (b) What is the amplitude of the odiccurrent Isp(t) and the total current I(t) = Isp(t) + Itr(t). steady-statecurrent? with A 10. n emf of voltage E(t) = Eo coswt is applied to the RC 20.Thecircuit and input voltage of Problem 11 1(0)= 0
lOt
1)

17.R = 16 L = 2 H, C = 0.02 F; Q, V E(t) = 100; 1(0)= 0, Q(O)= 5 18.R = 60 Q, L = 2 H, C = 0.0025F; t V; E(t) = 100e- 1(0)= 0, Q(O)= 0 19.R = 60 Q, L = 2 H, C = 0.0025F; E(t) = 100e-V; 1(0)= 0, Q(O)=

circuit of Fig. at time t 0 (with the switch closed), and Q(O) O.Substitute Qsp(t) A coswt + B sin wt in the differential equation to show that the steady periodic i charge on the capacitors

2.7.8

Qsp(t) =
where
fJ

EoC cos(wt fJ) -J1 + w 2R2C2

= tan-](wRC).)

In Problems
with input

11 the through 16, parametersof an RLCcircuit


E(t) are given.
Substitute Isp (t)

and Q(O)= 0 = Thecircuit and input voltage of Problem 13with 1(0) 0 21. and Q(O)= 3 22.Thecircuit and input voltage of Problem 15with 1(0)= 0 and Q(O)= 0 23.Consideran LCcircuit-that is, an RLCcircuit with R = O-with input voltage E(t) = Eo sin wt. Show that unbounded oscillations current occurfor a certain resoof this nance frequency; express frequency in terms of Land

C.

voltage

24. It was stated

= A coswt + B sin wt =

Eq. (4), using the appropriate value of w, to find the periodiccurrent in the form Isp(t) 10sin(wt 8).)
in

steady

R, L, and C are positive, of LI\" + RI'+ 1/ = 0 is a transient C Prove solution-itapproaches eroas t +00. this. z 25.Prove that the amplitude 10of the steady periodicsolution of Eq.(6) is maximal at frequency w = 1/-JLC
in the text that, if

then any solution

\037

.)

Problemsand Eig\037nvalues) Endp()int


are now familiar with the fact that a solutionof a second-order differential linear is uniquely determined by two initial conditions.In particular, the only equation solution of the initial value problem)
You
y\"

+ p(x)y' + q(x)y == 0;

yea) == 0, y'(a) == 0)

(1))

is the
by

rectly, on the

solution y (x) = O. Mostof Chapter 2 has beenbased,directly or indio (as uniqueness f solutionsof linear initial value problems guaranteed Theorem 2 of Section 2.2).)))
trivial

a 2.8Endpoint Problemsnd Eigenvalues 181)


In this

sectionwe will seethat the situation is radically different for a problem


y\"

such as)

+ p(x)y' + q(x)y == 0;

yea) == 0, y(b) == O.)

(2))

The difference between the problemsin Eqs. (1)and (2)is that in (2)the two conditions are imposedat two different points a and b with (say)a < b.In (2)we are to find a solution of the differential equation on the interval (a, b) that satisfiesthe conditions yea) == 0 and y(b) == 0 at the endpoints of the interval. Sucha problem is called endpoint or boundary value problem. an Examples1 and 2 illustrate the sortsof complications that can arisein endpoint problems.)

Example
4
3
y

1)

Considerthe endpoint problem)


y\"

+ 3y == 0;

y(O) == 0, yen) == o.)

(3))

2
1)

The general solution of the differential equation is)

-1
-2 -3
-4)

+ y(x) == A cosx-J3 B sinx-J3.


so the condition y(O) == 0 impliesthat A == O. Therefore the only B possiblesolutions are of the form y (x) == B sin x,J3. ut then)
Now y(O) == A,

FIGURE

2.8.1. Various
=

B sin solutions y (x) of the endpoint value problem in For no B i= 0 doesthe Example solution hit the target value y 0 for x T(

x.J3

possible

yen) == B sinn-J3
\037

-0.7458B,)

1.
.)

so the other condition yen) == 0 requiresthat B == 0 also.Graphically,Fig.2.8.1 illustrates the fact that no possiblesolution y (x) == B sin x,J3 with B i= 0 hits the desiredtarget value y == 0 when x == n. Thus the only solution of the endpoint value problemin (3)is the trivial solution y(x) = 0 (which probably is no surprise).

.)

..\037..

d..\037

Example2)
4
3
y

Considerthe endpoint problem)


y\"

+ 4y == 0;
y

y(O) == 0, yen) == O.)

(4))

The general solution of the differential equation is)

2
1)

(x) == A cos2x+ B sin 2x.

-1
-2 -3

y(O) == A, so the condition y(O) == 0 impliesthat A == O.Thereforethe only possiblesolutions are of the form y(x) == B sin 2x.But now yen) == B sin 2n = 0 no matter what the value of the coefficient B is.Hence, s illustrated graphically a -4) in Fig.2.8.2, B=-4) solution y (x) == B sin 2xhits automatically the desired every possible target value y == 0 when x == n (whatever the value of B). Thus the endpoint value FIGURE2.8.2. Various possible problemin (4) has infinitely many different nontrivial solutions.Perhaps this does solutions y (x) = B sin 2x of the seem bit surprising. a endoint value problem in Example 2. No matter what the coefficient 1 Remark Note that the big difference in the results of Examples and 2 B is, the solution automatically stemsfrom the seemingly small difference betweenthe differential equationsin (3) hits the target value y = 0 for X=T(.) and (4),with the coefficient3 in one replacedby the coefficient4 in the other. In mathematics as elsewhere, sometimes doorsturn on small hinges.\" \"big
Again,

1:

Remark The \"shooting\" terminology usedin Examples1 and 2 is often useful in discussing solution endpoint value problems.We considera possible which starts at the left endpoint value and ask whether it hits the \"target\" specified by the right endpoint value.)))

2:

182

2 Chapter Linear Equationsof HigherOrder)

P Eigenvalue roblems
Rather than being the exceptionalcases, xamples1 and 2 illustrate the typical E situation for an endpoint problemas in (2): It may have no nontrivial solutions, or in it may have infinitely many nontrivial solutions.Note that the problems (3)and both can be written in the form) (4)
\037)

the situation in an endpoint problemcontaining a numerical value can (and generally will) dependstrongly on the specific parameter of the parameter. An endpoint value problemsuch as the problemin (5)-one contains an that The A-iscalledan eigenvalue problem. question we askin unspecified parameter an eigenvalue problemis this: For what values of the parameter A does there exist a nontrivial (i.e., nonzero) solution of the endpoint value problem? Such a value of A is called eigenvaluef the problem.One might think of such a value as a an o value of A for which there existproper(nonzero) solutionsof the problem. \"proper\" is Indeed,the prefix eigen a German word that (in somecontexts) be translated may v as the Englishword proper,o eigenvalues are sometimes calledproper alues (or s

+ p(x)y' + Aq(X)y = 0; yea) = 0, y(b) = 0,) (5)) with p(x) = 0,q(x) = 1, = 0,and b = n. The number A is a parameter in a If the problem(nothing to do with the parameters that were varied in Section 2.5). we take A = 3,we get the equations in (3);with A = 4, we obtain the equations in
y\"

(4). Examples1 and 2 show that

characteristic values).

Thus we saw in Example that 2


y\"

= 4 is an eigenvalueof the endpoint problem)


(6))

+ Ay = 0,

y(O) = 0, yen) = 0,)

whereasExample1 showsthat A = 3 is not an eigenvalue of this problem. Supposethat A* is an eigenvalue of the problemin (5) and that y*(x) is a nontrivial solution of the endpoint problemthat resultswhen the parameter A in (5) is replacedby the specificnumerical value A*, so)

y: +

p(x)y\037

+ A*q(X)y* = 0

and

y*(a) = 0, y*(b)

= O.)

Then we call y* an eigenfunctionssociated a with the eigenvalue A*. Thus we saw with the eigenvalue in Example that y* (x) = sin 2xis an eigenfunction associated 2 A* = 4, as is any constant multiple of sin 2x. that Moregenerally, note that the problemin (5)is homogeneousin the sense one any constant multiple of an eigenfunction is again an eigenfunction-indeed, associated with the sameeigenvalue. That is, if y = y* (x) satisfies the problem in I (5)with A = A*, then so doesany constant multiple cy*(x).t can be proved (under mild restrictions on the coefficient functions p and q) that any two eigenfunctions associated with the sameeigenvalue must be linearly dependent.)

Example3)

Determine the eigenvalues and associated eigenfunctions for the endpoint problem)
y\"

+ Ay = 0;

y(O) = 0, y(L) = 0

(L > 0).)

(7))

Solution

We must

If A

considerall possible(real) values of A-positive, ero, nd negative. z a = 0,then the equation is simply y\" = 0 and its general solution is) y(x) = Ax

B.)))

a 2.8Endpoint Problems nd Eigenvalues 183)


Then the endpoint conditions y(O) == 0 == y(L) immediatelyimply that A == B == 0, so the only solution in this case the trivial function y (x) = O.Therefore,A == 0 is is not an eigenvalue of the problemin (7). 2 If A < 0,let us then write A == (with a > 0) to be specific.Then the differential equation takesthe form)

-a

\"

2 0 -ay==,)

and its general solution is)

y(x) == cle

ax

+ C2e-ax== A coshax + B sinh

where A == CI + C2 and B == CI C2.(Recall that coshax == (eax ax eax that sinh ax == (e )j2.)The condition y(O) == 0 then gives)

ax,)

+ e-ax )j2and

y(O) == A cosh0 + B sinh 0 == A == 0,)

x)

so that y(x) == B sinh ax.But now the secondendpoint condition, y(L) == 0, gives a y (L) == B sinh a L == O. This impliesthat B == 0,because i= 0,and sinh x == 0 for x == 0 (examine the graphs of y == sinh x and y == coshx in Fig.2.8.3.) only A Thus the only solution of the problemin (7)in the case < 0 is the trivial solution = 0,and we may therefore conclude the problemhas no negativeeigenvalues. that y the The only remaining possibility that A == a 2 > 0 with a > O. In this case is
differential equation is)
with
y\"

+ a2y

== 0,)

FIGURE2.8.3. hyperbolic The sine and cosinegraphs.)

y(x) == Acosax+Bsinax. The condition y (0) == 0 impliesthat A == 0,so y (x) == B sin ax.The condition y(L) == 0 then gives) y(L) == B sinaL== O. Can this occurif B i= O? Yes,but only provided that a L is a (positive) integral
multiple of n:)

general solution)

aL == n,)
== a 2 ==

that

is, if)
A

n2
An

2n,) 2

3n,)

...
2

,)

nn,)

..
,)

' 4n 9n ..., L2 L2' L2'


n 2n

n 2n 2

L2'

y)

of Thus we have discovered the problem in (7)has an infinite sequence positive that eigenvalues)
n=1)
n=3)

==

L
x)

L2

'

n ==

1,2,3,

....)
is)

(8))

With

B == 1,he eigenfunction associated t with the eigenvalueAn


Yn

-I)

FIGURE2.8.4. The
. . eigenfunctions forn =
Yn

1,2,3,4.)

. nn x (x) = SIn

(x) == sin

-,
nnx
L)

n ==

1, ,3,... . 2

(9))

showsgraphs of the first several of theseeigenfunctions. We seevisFigure 2.8.4 just ibly how the endpoint conditions y (0) == y (L) == 0 serve to select those sine at functions that start a periodat x == 0 and wind up at x == L precisely the end of a half-period.
.)))

184

2 Chapter

Linear Equationsof HigherOrder)


s 9.1, precisetatement we will defer until Section under the assumption that q (x) > o on the interval [a,b], any eigenvalue of the form in (5)has a divergentincreasing

Example3 illustrates the general situation. According to a theorem whose

< A2 < A3 < ... < An < ... ---+ +00 of eigenvalues, eachwith an associated eigenfunction. This is also true of the folmore general type of eigenvalue problem,in which the endpoint conditions lowing involve values of the derivative y' as well as values of y:)
AI

sequence)

+ p(x)y' + Aq(X)Y == 0; alyea) + a2Y'(a) == 0, b i y(b) + b2Y'(b) == 0,)


y\"

(10))

where aI,a2, b I , and b2 are given constants. ith al == 1 == b2 and a2 == 0 == b I , we W the problemof Example4 (in which p(x) = 0 and q (x) = 1, in the previous as get example).)

Example4)

Determine the eigenvalues and eigenfunctions of the problem)


y\"

+ AY == 0;

y(O) == 0, y' (L) == O.)

(11))

Solution

Virtually

2 eigenvalues are positive, so we take A == a differential equation is) 2


y\"

the

sameargument as that usedin Example3 showsthat the only possible > 0 (a > 0) to be specific.Then the

+a

== 0,)

with

general solution)

+ y(x) == A cosax B sin ax.

The condition y(O) == 0 immediatelygives A == 0,so)

y(x) == B sin ax and y'(x) == Bacosax.)


The secondendpoint condition y' (L) == 0 now gives)

aL y' (L) == Bacos == O.)


This will hold with B
n /2:)
=1=

0 provided that

a L is an odd positive integral


,)

aL ==
y)

2')
n

3n 9n2

2') ...

n=2)

n=I)

that

is, if)

n A==4L2'

4L2' ...,
2

- l)n ... (2n - 1)2n


(2n
2)
,)

multiple

of

4L2
are (11)

Thus the
given
-1)
by)

nth

eigenvalue An and associated eigenfunction of the problem in (2n (2n

FIGURE2.8.5. The
eigenfunctions
Yn

. (2n (x) = SIn 2L forn =

- l)n

1,2,3,4.)

for n == showsgraphs of the first several of theseeigen3, Figure functions. We how the endpoint conditions y (0) == y' (L) == 0 serve to visibly select those sine functions that start a periodat x == 0 but wind up at x == L just .))) preciselyin the middleof a half-period.

- - 1)2n 2.8.5 1,2, .... see


An

4L2)

and

Yn

(x) == sin

- l)n
2L)

(12))

2.8Endpoint Problems nd Eigenvalues 185) a


A general procedure for determining the eigenvalues of the problem in (10) can be outlined as follows. We first write the general solution of the differential equation in the form)
Y

== A Y 1 (x,A)

+ BY2 (x,A).

We write
which)

Yi

(x,A) because YI

and

Y2 will

3 dependon A, as in Examples and 4, in


Y2(X) == sin

YI(X) ==

== cosax cosx-JI and

ax == sinx-JI.)

Then we imposethe two endpoint conditions, oting that each is linear in Y and y', n and hence also linear in A and B.When we collect coefficientsof A and B in the of equations, we therefore get a system of the form) resulting pair

al (A)A + fJI
a2(A)A

(A)B == fJ2(A)B == O.)

0,

(13))

eigenvalueif and only if the system in (13)has a nontrivial solution (one B not both zero).But such a homogeneousystem of linear equations s has a nontrivial solution if and only if the determinant of its coefficientsvanishes. We therefore concludethat the eigenvalues of the problem in (10)are the (real) solutions of the equation)
Now
with

A is an A and

D(A)

== a I (A) fJ2 (A)

-a

2(A) fJ I (A) == O.)

(14))

To illustrate Eq. (14)in a concrete roblem,let'srevisit the eigenvalueprobp lem of Example3. If A > 0, then the differential equation y\" + Ay == 0 has the general solution y(x) == A cos(,J):x) sin(,J):x). endpoint conditions The +B == 0 and y(L) == 0 then yield the equations) y(O)
== 0, + B .0 == +B y(L) == A cos(,J):L) sin(,J):L) 0)
Y

(0) == A

.1

terminant

(as we saw in Example3). For more general problems,he solutionof the equation D(A) == 0 in (14) t may presentformidabledifficultiesand require a numerical approximationmethod (such as Newton'smethod) or recourse a computer algebra system. to Mostof the interest in eigenvalueproblemsis due to their very diversephysical applications. he remainderof this sectionis devoted to three such applications. T
Numerous additional applications are included in Chapters 8 and 9 (on partial differential equations and boundary value problems).)

and B) which correspond the equations in to The de== 0 corresponding (14)is then simply the equation to equation D(A) == sin(,J):L)== which impliesthat,J):L nn, or A == n 2n 2jL2 for n == (in the
unknowns A

(13).

0,

1,2,3,

...

TheWhirling String
of us has not wondered about the shapeof a quickly spinning jump rope?Let considerthe shapeassumedby a tightly stretchedflexible string of length Land constant linear density p (massper unit length) if it is rotatedor whirled (like ajump a rope)with constant angular speed()) (in radians per second)round its equilibrium W position along the x-axis. e assumethat the portion of the string to one sideof any point exertsa constant tension force T on the portion of the string to the other)))
Who

us

186
y)

2 Chapter

Linear Equationsof HigherOrder) side of the point, with the direction of T tangential to the string at that point. We assumethat, as the string whirls around the x-axis, achpoint moves in a e circle centeredat that point's equilibrium position on the x-axis. hus the string is T so elastic, that as it whirls it also stretchesto assumea curved shape. Denote by o y (x) the displacement f the string from the point x on the axis of rotation. Finally, we assumethat the deflectionof the string is so slight that sin e tan e == y' (x) in
further
\037

T x=O
Equilibrium
(a))

T)
x)

x=L
position

c). Fig.2.8.6(
law
x)

F == ma to the pieceof string of mass p corresponding the interval to are [x,x + \037x]. The only forcesacting on this piece the tension forcesat its two ends.From Fig.2.8.7e seethat the net vertical force in the positive y-direction is) w
\037x

We plan to

derive a differential equation for y(x) by application of Newton's

x=O)
Whirling string
(b))

x=L)

F == T sinCe+ \037e)
so that)

-T

sin e

\037

T tan(e

\037e)

-T

tan

e,)

F
__
Vertical component: I (J __J) Tsin (J
I

\037

Ty'(x + \037x)

- Ty'(x).)

(15))

String)

(c)

FIGURE2.8.6. whirling The


string.)

Next we recallrom elementary calculusor physicsthe formula a == rw 2 for the f (inward) centripetal accelerationof a body in uniform circular motion (r is the radius of the circle w is the angular velocity of the body). Here have r == y, so and we the vertical accelerationof our piece string is a == -w2y, the minus sign because of the inward direction is the negative y-direction. ecause == p \037x, substitution of B m this and (15)in F == ma yields

Ty'(x + \037x) so that)


T)

y)

- Ty'(x) -pw + - y'(x) T . y'(x


\037

\037x,)

\037x)
\037x

\037

pw

y.

We now
string:)

take the limit as

\037x

--+ 0 to get the differential equation of motion of the TY \"

/
/)

+ pw 2y

== 0

.)

(16))

x + L\\x

x)

FIGURE2.8.7. orceson a short F segment of the whirling string.)

If we write)
pw A==T)

(17))
string

and

imposethe condition that the ends of the


y\"

are fixed, we

finally

get the
(7))

eigenvalue problem)

+ Ay

==

0;

y(O) == 0, y(L) == 0)

that in

(7)are)

we considered Example We found there that the eigenvaluesof the problem in 3.


n 2n

2 ,
n ==

An

==

L2)

1, ,3, ..., 2

(8))

the eigenfunction Yn(x) == sin(nnxjL) associated with An. But what does all this mean in terms of the whirling string? It means that unlessA in (17)is one of the eigenvalues in (8),then the only solution of the problem)))
with

a 2.8Endpoint Problems nd Eigenvalues 187)


in

rium

value W n corresponding An,) to

the (7)is the trivial solution y(x) = O. In this case string remains in its equilibposition with zero deflection. But if we equate (17)and (8)and solve for the

Wn

=
A\037T

n; \037)

(18))

for n == 2,3, , we get a sequencef critical peeds f angular rotation. Only o o s at thesecritical angular speedscan the string whirl up out of its equilibrium position. At angular speedW it assumesa shapeof the form Yn == C sin(nnx/L) illustrated n in Fig. 2.8.4 O (where Cn = 1). ur mathematicalmodel is not sufficiently complete much smallerdeflections to (or realistic) determine the coefficientCn , but it assumes than thoseobserved Fig. .8.4,the numericalvalue of C would necessarily e b in 2 so n smaller than significantly Supposethat we start the string rotating at speed)

1,

...

1.

< WI =
\037

,)

\037 So

when W is increasedstill back into its undeflectedposition along the axisof rotation!)

then gradually its speedof rotation. long as W < WI, the string remains in its undeflected position y O. But when W == WI, the string popsinto a whirling

position y

= == Cl sin(n x/L). And

increase

further,

the

string

pops

The Deflectionf a Uniform o Beam)


\"70,\">'''''''''''''''''.'''''' /jlil!!f!!l ;;.r;.i.:.)
\"

%Ci;;;ry\037.t\,!:")

FIGURE2.8.8. Distortion of a
horizontal

beam.)

L)

include now an example the use of a relatively simpleendpoint value problem of to explain a complicated s physical phenomenon-thehapeof a horizontal beam on which a vertical force is acting. uniform both in crosssecConsiderthe horizontal beam shown in Fig. tion and in material. If it is supported at its ends, then the force of its own only axis of symmetry into the curve shown as a dashed weight distorts its longitudinal line in the figure. We want to investigate the shape y == y (x) of this curve, the deflectioncurve of the beam. We will use the coordinate system indicated in with the positive y-axisdirected downward. Fig. A consequence the theory of elasticity is that for relatively small deflections of
We

2.8.8,

2.8.9,

\037\037\037-____---\037

x)

of such a beam (so small that [y'(x)]2is negligible comparison with unity), an in adequate mathematicalmodel of the deflectioncurve is the fourth-orderdifferential
equation)

Positive
y-values)

Ely(4) == F(x),)
where)

(19))

FIGURE2.8.9. deflection The


curve.)

. . .

E is a constant known as the Young's modulus of the material of the beam, I denotesthe moment of inertia of the crosssectionof the beam around a horizontal line through the centroid of the cross ection,and s denotesthe density of downward force acting vertically on the beam at F(x)
the point
x.)))

188

2 Chapter

Linear Equationsof HigherOrder)


Density of force?Yes;this means that the force acting downward on a very segment[x,x + \037x] of the beam is approximately F(x)\037x. The units of F (x) are thoseof force per unit length, such as poundsper foot. We will consider here the case which the only force distributed along the beam is its own weight, in w poundsper foot, so that F(x) = w. Then Eq. (19) takesthe form)
short

Ely(4) == W)
where E,

(20))

I,and ware all constant.

Note: We assumeno previous familiarity with the theory of elasticity or with I Eq. (19)or (20)here.t is important to be ableto begin with a differential equation that arises in a specificapplieddisciplineand then analyze its implications; thus we develop an understanding of the equation by examining its solutions.Observe that, in essence, (20)impliesthat the fourth derivative y(4) is proportional to Eq. the weight density w. This proportionality involves, however, two constants:E, which dependsonly on the material of the beam,and I,which dependsonly on the s shapeof the cross ectionof the beam.Values of the Young'smodulus E of various materials can be found in handbooks physical constants; I == \037Jl'a4 for a circular of cross sectionof radius a.)
Although

only the solution of simplefirst-order equations by One integration of Eq. (20)yields)

Eq. (20)is a fourth-orderdifferential equation, its solution involves successive simpleintegrations. Ely(3) == wx

+C

;)

a secondyields) another yields)

EIy Ely'==

\"

==

2 wX

. + CIX + C2,)
+ C2X + C3 ; + C3x + C4

\037wx3

\037CIx2

a final integration gives) x=o


. .: .
...

Ii
<,:
;.;.;-:)

x=L
or hinged)

Simply

supported

11) where C
x=L

Ely ==

wx4
2\037

\037CIx3

\037C2x2

,)

1, C 2, C3 , and C4 are arbitrary Eq. (20)of the form)

constants. Thus we obtain a solution of

x=o -t)
Built
in)

a beam.)

t)
where A,

2 3 2 y(x)== 24EI)x +Ax +Bx +Cx+D,

(21))

Two FIGURE2.8.10. ways of


supporting

from the four integrations. in which the beam is supat its ends,where x == 0 and x == L. Figure showstwo common types ported of support.A beam might alsobe supported way at one end but another way at one the other end.For instance, Fig. beam firmly fastened showsa cantilever-a at x == 0 butfree (no support whatsoever) at x == L. The following table showsthe c to boundary or endpoint onditions orresponding the three most common cases. c

These four constants are determined by the way last

B,C,and D are constants resulting

2.8.10

2.8.11

seethat theseconditions are appliedreadily in beam problems,although a discussion here of their origin would take us too far afield.)))
We will

a 2.8Endpoint Problemsnd Eigenvalues


(;:\",\";\037;Sr.PP\037f.ta:)

189)

EndpointCondition)

Simply supported Built -in or fixed end

Freeend)

= y\" = 0 y = y' = 0 y\" = y(3) = 0)


y

For example,the deflection curve of the cantilever in Fig. would be with the coefficientsA, B,C,and D determinedby the conditions) given by Eq. (21), y(O) = y'(O)= 0 and
Cantilever)

2.8.11

y\"(L)= y(3)(L)= 0,)

(22))

FIGURE

2.8.11. cantilever.) The


Example5)
Solution

t correspondingo the fixed end at x = 0 and the free end at x = L. The conditions in (22)together with the differential equation in (21)constitute an endpointvalue problem.)
....-. ...........

Determine the shapeof the deflectioncurve of a uniform horizontalbeam of length L and weight w per unit length and simply supported at each end.)
We

have the endpoint conditions) y(O) = y\"(O)

= 0 = y(L) = y\"(L).)

Rather than imposing then directly on Eq. (21), us begin with the differential let y(4) = wand determine the constants as we proceedwith the four equation E 1 successive integrations. The first two integrations yield)

Ely(3) = wx

Hence y\"(O) = 0 impliesthat

+ A; Ely\" = !wx2 + Ax + B.) B = 0,and then y\"(L)= 0 gives

0= !wL2 +AL.)
It follows that
A

= -wLj2and thus
\"

that)
1

EIy = 2x2
Then two more integrations give)

- 2wL
1

x.)

Ely'= iWx3
and
finally,)

\037wLx2

+ C,) + Cx+ D.)


(23))

Ely(x) =
Now y(O) = 0 impliesthat

wx4
2\037

/2 wLx3

D = 0;then, because (L) = 0,) y

0= wL42\037

l\037

wL4+CL.)

It follows that

from Eq. (23)we obtain) H C = wL3j24. ence

y(x) = 24EI)(x

- 2Lx + L x)
3 3

(24))))

190

2 Chapter

Linear Equationsof HigherOrder) as the shapeof the simply supported beam. It is apparent from symmetry (see also Problem 17)that the maximum deflection Ymax of the beam occursat its midpoint x = L/2,and thus has the value)
L
Ymax=Y

( )
\"2

L 24E/( 16

- 2L + 2L ) 8
4 1 4

;)

that is,)

Ymax

4 = 5wL

384E/.)

(25)

.)

\037

\037

\037

Example6)

For instance, supposethat we want to calculate the maximum deflectionof a simply steelrod 20 ft long with a circular cross ection1 in. in diameter. From a s supported 3 handbook we find that typical steelhas density 8 = 7.75g/cm and that its Young's 2 12 = 2 X 10 g/cm.s, so it will be more convenient to work in cgs modulus is E units. Thus our rod has) length: L
and)

= (20ft)

em) (30.48 ;) = 609.60 e

. radIus: a =
Its

1. - In. 2

)(

cm 2.54=--) = 1.27m. c In.)

linearmassdensity (that is, its massper unit

length)
\037

is)

2 \037, p = na 8 = n(1.27)2(7.75)39.27 cm)


so)
w \037 = pg = 39.27 980 ( cm ) ( s )
e\037 \037

dyn 38484.6 . cm
==

The area moment of inertia of a circular disk of radius a around a diameter is /


\037na4,

so)

/=
Therefore Eq. (25)yields)
\037

1 c -n(1.27)42.04m4. 4)
\037

Ymax

(5)(38484.6)(609.60)4 16.96 em 12 ( (384)2 x 10 ) (2.04)


\037

,)

to note that

about 6.68 as the maximum deflectionof the rod at its midpoint. It is interesting in., 4 Ymax is proportional to L , so if the rod were only 10ftong, its maxil mum deflection would be only one-sixteenth as much-onlyabout 0.42 Because in. / = \037na4, we seefrom Eq. (25)that the samereduction in maximum deflection .))) couldbe achieved by doubling the radius a of the rod.

a 2.8Endpoint Problemsnd Eigenvalues 191)

y)

= y(x))
p) x)

TheBuckled od) R showsa uniform Figure 2.8.12


\"buckled\" by this buckling
I

p)

x=O)
rod.)

Ix=L)

rod of length L, hinged at each end, that has been an axial force of compressionP appliedat one end. We assume to be so slight that the deflection curve y == y (x) of the rod may be regardedas defined on the interval 0 x L. In the theory of elasticity the linear endpoint boundary value problem)

< <

FIGURE2.8.12. buckled The

Ely\" + Py

==

0,

y(O) == y(L) == 0)

(26))

of is usedto model the actual (nonlinear) behavior of the rod. As in our discussion the deflection of a uniform beam, E denotesthe Young's modulus of the material of the beam and denotesthe moment of inertia of each crosssectionof the beam around a horizontal line through its centroid.

If we write)
A==

EI')

(27))

then

the problem in

the (26)becomes eigenvalueproblem)


y\"

+ AY == 0;

y(O) == y(L) == 0)
that

(7))

that

we considered Example3.We found in


n 2JT2 An==

its eigenvalues {An} are given

by)

L2

' n==I,2,3,...)

(8))

with

and

P == AE I. The forces)
Pn

with An. (Thus whirling strings the eigenfunction Yn == sin(nJTxjL) associated buckledrodsleadto the sameeigenvaluesand eigenfunctions.) To interpret this result in terms of the buckledrod, recall from Eg.(27)that

== An

E1== n

2n 2

EI

L2)

n==I,2,3,...)

(28))

are the criticaluckling forces the rod. Only when the compressiveorce P is f of b one of thesecritical forcesshould the rod \"buckle\" out of its straight (undeflected) f shape.The smallestcompressiveorce for which this occursis)

==

JT2EI
L2)

(29))

for This smallestcritical force PI is calledthe Eulerbuckling force the rod; it is the f upperbound for thosecompressiveorcesto which the rod can safely be subjected without buckling. (In practicea rod may fail at a significantly smallerforcedue to a contribution of factors not taken into account by the mathematicalmodel discussed

here.
))))

192

2 Chapter

Linear Equationsof HigherOrder)


...... \037 ,.... ...............

Example7)

For instance, supposethat we want to compute the Euler buckling force for a steel rod 10ftong having a circular cross ection1 in. in diameter. In cgsunits we have) s l

E=2X
L

2 12 10 gjcm.s,

= (10 ft)

em, (30.48 ;) = 304.8 e


\037m

and

I = JT [<0.5in.)(2.54 ) 4
In.)

r
\037

\037

e 2.04m4.
the critical force for this rod

Uponsubstituting
IS)

thesevalues in Eq. (29)we find


\037

that

X PI 4.34 10 dyn 5 using the conversionfactor 4.448 x 10 dynjlb.)


8

976lb,

.)

m
1.

Problems)
=

The eigenvalues in Problems through 5 are all nonnegative. First determine whether 'A 0 is an eigenvalue; then find the positive eigenvalues

and associated eigenfunctions.)

y\" + 'Ay = 0; y/(O) = 0, y(l) = 0 2.y\" + 'Ay = 0; y/(O) = 0, y/(n) = 0 3.y\" + 'Ay = 0; y( -n)= 0, y(n) = 0 4. y\" + 'Ay = 0; y/( -n)= 0, y/(n) = 0 5. y\" + 'Ay = 0; y( 2) = 0, y/ (2) = 0

t 6.Considerhe eigenvalue problem)


y\"

8.
O.)

(a) Show that 'A 0 that the eigenfunctions where an is the nth positive are the functions {sin root of the equation tan z (c) Draw a sketch indicating the roots {an}f as the points of intersection of the tan z and y curves y z. Deducefrom this sketch when n is large. that an (2n t Considerhe eigenvalue problem)
all its

is not

eigenvaluesare nonnegative. an eigenvalue. (b) Show

\037

- I)n/2 = -

anx}f, = -z.

y\" all its

+ 'Ay = 0; y(O) = 0, y(l) = y/ (1);)

+ 'Ay = 0;

y/(O)

= 0, y(I) + y'(I) = = ax ax = = I/z.

2 All the eigenvalues are nonnegative, so write 'A where a > 0 is not an eigen(a) Show that 'A A cos value. (b) Show that y + B sin satisfies the endpoint conditions if and only if B 0 and is a positive root of the equation tan z Theseroots of {an} are the abscissas the points of intersection of the curves y tan z and y z, as indicated in Fig. Thus the eigenvalues and eigenfunctions of this problem are the numbers {a\037} and the functions {cosnx } rea

0.

=a a

eigenvaluesare nonnegative. (a) Show that 'A = 0 is an eigenvalue with associated eigenfunction Yo (x) = x. (b) Show that the remaining eigenfunctions are given by Yn (x) = sin fJnx, where fJn is the nth positive root of the equation tan z = z. Draw a sketch showing theseroots. Deducefrom this sketch that fJn (2n + I)n/2 when n is
\037

r =

= 1/

2.8.13.

r,

9.Prove that the eigenvalue problem negative eigenvalues. 10. rove that the eigenvalue problem) P
y\"

large.

of Example 4 has

no

spectively.)

+ 'Ay = 0; y(O) = 0, y(l) + y/(I) = 0)

has no
ically
z)

11.
12.

z = 0.) Usea method similar

negative eigenvalues. (Suggestion: Show graphis that the only root of the equation tanh z

-z

The FIGURE2.8.13. eigenvaluesare determined by the intersections of the graphs of y = tan z and y = I/z
(Problem 6).)

in to that suggested Problem show that the eigenvalue problem in Problem 6 has no ative eigenvalues. t Considerhe eigenvalue problem) y\"

to 10

neg-

+ 'Ay = 0; y(-n) = y(n), y/(-n) = y/(n),)

7. Considerhe t
y\"

eigenvalue problem) y(O)

+ 'Ay = 0;

= 0, y(l) + y/ (I) = 0;)

becausethe two endpoint which is not of the type in conditions are not \"separated\" between the two endpoints. 0 is an eigenvalue with associated))) (a) Show that 'Ao

(10)

a 2.8Endpoint Problemsnd Eigenvalues 193)


(b) Show that there areno negeigenfunction Yo (x) ative eigenvalues.(c) Show that the nth positive eigenvalue is n 2 and that it has two linearly independent associated eigenfunctions, cosnx and sin Considerhe eigenvalue problem) t

= 1.

16.a) Supposethat (
x

= L.Show

that its

a beam is fixed at its ends x shapeis given by)


w
24\302\2431

= 0 and

13.

nx.

y(x)

(x

- 2Lx + L
3

2X

).

y\"

+ 2y' + AY = 0; y(O) = y(l) = O.)

14.

1 is not an eigenvalue. (b) Show (a) Show that A that there is no eigenvalue A such that A < (c) Show that the nth positive eigenvalue is An n 2n 2 + with X associated sin nn eigenfunction Yn (x) Considerhe eigenvalue problem) t

= e-

1.
x.

(b) Show that the roots of y'(x) = 0 are x = 0, x = L, and x = L/2, so it follows (why?) that the maximum deflectionof the beam is)
Ymax

1,

=Y

2\"

WL4
384\302\2431')

y\"+2y'+AY=0;
Show that the

y(O)

=0, y'(I)=O.)

17.For the

15.

eigenvalues are all positive and that the nth + 1 with associated positive eigenvalue is An = eigenfunction Yn (x)= sin an x , where an is the nth positive root of tan z = z. (a) A uniform cantilever beam is fixed at x = 0 and free at its other end, wherex = L. Show that its shapeis given

eX

a\037

18.

of a beam with simply supported ends. simply supported beam whose deflection curve is given by Eg.(24),show that the only root of y'(x) 0 in [0,L] is x L/2, so it follows (why?) that the maximum deflectionis indeedthat given in Eg. (a) A beam is fixed at its left end x 0 but is simply supL. Show that its deflection ported at the other end x
one-fifth that

=
4

=
3

(25).

curve

IS)

by)

y(x) (b) Show that

w
24\302\2431

(x - 4Lx + 6L
4 3 at

y(x) (b) Show

w
48\302\2431)

(2x

- 5Lx + 3L
41.6%

2X

).

2 2X

).

x = 0, and thus that it follows (why?) that the maximum deflectionof the cantilever is Ymax = y(L) = wL I).)
4/(8\302\243

y' (x)= 0 only

that its maximum deflection occurswhere x of the maximum deand is about flection that would occur if the beam were simply supported at eachend.)))

- L (15 -J33) /16

Pow-er Series

Methods)

Introdu 11II_ ctionand Review

(:)\037

_\037_?w\037...__\037e

ries)

constant coefficientscan be reducedto the algebraic problem of finding the roots of its characteristic equation. There is no similar procedure solving linear for differential equations with variable coefficients,at leastnot routinely and in finitely of many steps.With the exception specialtypes,such as the occasional equation that can be solved by inspection,linear equations with variable coefficients generally of require the powerseriesechniques this chapter. t These sufficefor many of the nonelementarydifferentialequations techniques that appearmost frequently in applications. erhaps the most important (becausef o P its applications in such areasas acoustics, eat flow, and electromagneticradiation) h is Bessel's equationof ordern:)
with
\037)

Section we saw that solving a homogeneous 2.3 linear differential equation

x2y\" + xy' + (x2

2 )y

= O.)

\037)

It Legendre's equationof ordern is important in many applications. has the form) (1 x2)y\" 2xy' -t- n(n + l)y = O.) In this sectionwe introduce the powerseries ethod in its simplest orm and, f m

basicfacts about powerseries. Recall first is an infinite series f the form) o

along the
00

way,

state (without

proof) severaltheorems that constitute a reviewof the in that a power series (powers of) x a

L (x - a)n =
C n
n=O)

Co

+ Cl(x a) + C2(x a)2+ ... + Cn (x


x:)

- a)n + ..

..

(1)

If a
\037)

= 0,this is a powerseriesn i

Lcnx =
n
n=O)

00

Co

+ CIX + C2X2 + ...+ cnxn + ... .

(2))

194)))

3.1Introductionand Review of PowerSeries 195)


in confine our review mainly to powerseries x, but every general property of powerseries x can be converted to a general property of power seriesin x a in by replacement of x with The powerseries (2)convergesn the interval I provided that the limit) in o

We will

x-a.
00
n=O

Lcnx
existsfor all x in

lim
N\037oo

L
n=O)

cnx

(3))

I. In this

case sum) the f(x) =

Lcnx
n=O)

00

(4))

o is defined on I,and we call the series cnx n a power series representationf the L function f on I.The following powerseries representationsof elementaryfunctions should be familiar to you from introductory calculus:)
eX

x '\"\"' cos = \037 (_I)n = 1 X


2n
n=O

. sIn x =
cosh = x
sinh x =

+ 10(1 x) =
1

L, l+x+,+,+.\"; 2. - - -- ....' '. '. ' L I)! - - -- ... . 3! 5! - - ... . L 2! - - .... L 5! - 2 3 - ... L
00

xn =
n.

x2

x3

n=O

3.)

(5))

00

(2n)
.)

x2 x4 +4 2

(6))

00

(_ 1) n X 2n + = x
1

x3

n=O

(2n +

')

(7))

00

n=O

= 1 + x + x4 + 4! (2n)!
x2n

')

(8))

00

x2n+l

n=O

(2n + I)!

= x + x + x5 + 3!
3

')

(9))

00

(_I)n+lxn = x
n
n

x2

x3

(10))

n=l)

I-x Lx
=
n=O)

00

= 1 + x + x2 + x3 + ... ;

(11))

and)
ex

3 l)x2 (1+ x) = 1 + ax + a(a2! + a(a 1)(a 2)x + . . 3!

- -

..

(12)

In compact summation notation, we observethe usual conventionsthat O!= 1 and in that xO = 1 for all x, including x = O. The series (5) through (9)convergeto in the indicated functions for all x. In contrast, the series (10)and convergeif in 1 but diverge if Ix if Ix = I?) The series is the geometric Ix (What The series in with a an arbitrary real number, is the binomial series. If a is a nonnegativeinteger n, then the series (12)terminates and the binomial in serieseduces a polynomial of degree which converges for all x. Otherwise, r to n the seriess actually infinite and it converges if Ix 1 and diverges if Ix > its i behavior for Ix = 1 dependson the value of

series.

<

> 1. (12),
I

(11) (11)

<

1;

a.)))

196

3 Methods) Chapter PowerSeries Remark: Power series as thoselisted in (5) through (12)are often such derived as Taylor series. he Taylor series T with center == a of the function f is x
the

powerseries)

\037

L
n=O

00

fen) (a)

n!

f\" (a) (x-a)n==f(a)+f'(a)(x-a)+ (x-a)2+... (13) 2!) under the hypothesis that

in

(sothat the coefficients in Eq. (13)are all defined).If a is the Maclaurin series)
\037)

powersof x

- a,
!
(0)

f is infinitely ==

differentiable at x

==

0,then

the

in series (13)

L =0
n

00

en)

xn

==

(3) 0 (O f (0)+ f' (O)x + f 2!) x2 + f 3 ( ) x3 + ... .


\"
!)

(13'))

For example, supposethat (x) == eX. Then fen) (x) == eX, and hencefen) (0) == 1 > O. In this caseEq. (13') for all n in series (5). reduceso the exponential t

.)

PowerSeries Operations
If the Taylor series the function of converges to interval containing a, then we say that the function example,)

f(x) for all x in some open f is analytic at x == a. For

. . .

every polynomial function is analytic everywhere; every rational function is analytic wherever its denominator is nonzero; more generally, if the two functions and g are both analytic at x == a, then so are their sum + g and their product g, as is their quotient /g if

g(a)

f.

=1= O.)

For instance, the function h(x) == tan x == (sinx)/(cosx) analytic at x == 0 is == 1 =1= 0 and the sine and cosine functions are analytic (by virtue of their convergent power seriesepresentations in Eqs. (6) and (7)). It is rather r awkward to compute the Taylor series f the tangent function using Eq. (13)beo causeof the way in which its successive derivatives grow in complexity (try it!). series ay be manipulated algebraically in much the sameway Fortunately, power m as polynomials. or example,if) F

because 0 cos

00

f(x) == Lanxn
n=O)

and

g(x) ==

L
n=O)

00

bnx

(14))

then)

f(x) +
and)

00

g(x) ==

L(a + bn)x
n
n=O)

(15))

f(x)g(x)==

Lcnx
n=O)

00

== aobo

+ (aobl + a1bo)x (aob2 + a1b + a2bo)x2+ ... , +


1

(16))))

addition and the series (16)is the result of formal multiplication-multiplying in eachterm of the first series y each term of the secondand then collecting coefb ficients of like powersof x. (Thus the processes resembleaddition and strongly i multiplication of ordinary polynomials.)The seriesn (15)and (16)convergeto f(x) + g(x) and f(x)g(x),respectively, on any open interval on which both the

where Cn

= aobn

3.1Introductionand Review of PowerSeries 197) i +a1b- +...+anbo. The seriesn (15)is the result of termwise
n
1

seriesn (14)converge. For example,) i


sIn xcos = x

(x

=x

14 15 - 613+ 120 )( - 212+ 24 -...) x x x ) + -6 - 2 x3 + 24 + 12+ 120) x5 + ... ) ( ) (


1 x
\037 \037

-...
\037

\037

\037

4 = x -x3 + 16 5 6 120)

-x -...
)
3

=
for all x.

1 2

1. (2x (2x (2x)- 3! + 5! - ... = - 2x [ ]


)
5 SIn 2)

c Similarly, the quotient of two powerseriesan be computed by long division, as illustrated by the computation shown in Fig.3.1.1. division of the Taylor This seriesor cos into that for sin x yieldsthe first few terms of the series) f x
tan

x = x + -x3 +
3

2 17 -x + -x +.... 15 315)
5

(17))

Division of power series more treacherous than multiplication; the seriesthus is obtained for /g may fail to converge at somepoints where the seriesfor f and g both converge. For example,the sine and cosine seriesonverge for all x, but the c series (17)convergesonly if Ix < JT/2.) in tangent

ThePowerSeries Method)
The power series ethod for solving a differentialequation consists substituting m of the powerseries)
\037)

= LCn xn
n=O)

00

(18))

in the differential equation and then attempting to determine what the coefficients must be in order that the power series ill satisfy the differential w Co, Cl, C2, This is reminiscent of the method of undetermined coefficients,but now equation. we have infinitely many coefficients somehow to determine. This method is not always successful,but when it is we obtain an infinite seriesepresentation of a r form\" solutions that our previous methods have solution, in contrast to the \"closed

...

yielded. Beforewe can substitute the powerseries (18)in a differential equation,we in must first know what to substitute for the derivativesy', y\", .... The following theorem (stated without proof) tells us that the derivativey' of y = L cnx n is obtained of by the simpleprocedure writing the sum of the derivativesof the individual terms
in

the

seriesor f

y.)))

198

3 Methods) Chapter PowerSeries

x +
1

x3
3

2x 17x + 15 + 315 + ...


5

- 2 + 24 - 720+ ...
x2 x4 x6

x
X

x3
6

+ 120 +
x5
24

x5

x7
5040

+ ... + ... + ...

x3
2

x7
720

x3
3

x5
30

x7
840

x3
3

x5
6 2x5

+ 72
4x7

x7

...

15
2x5

... 315 + ... ...

15

x7 15 + 7 17x
315)

FIGURE

3.1.1. Obtaining

the

for series

tan

by division

of series.)

1 of THEOREM Termwise Differentiation PowerSeries)


If the powerseriesepresentation) r
\037)

f(x) =

Lcnx =
n
n=O)

00

Co

+ ClX + C2x2 + C3x3 + ...

(19))

of the function
and)

f convergeson the openinterval I,then f is differentiableon I,


= f'(x)= LnCnxn-l Cl + 2C2X + 3C3X2
n=l)

00

\037)

+...

(20))

at

eachpoint of I.)
For example, ifferentiationof the geometric series) d

I-x =\037x =1+x+x+x +...


1
\037 n
n=O)

( 11))))

3.1Introductionand Review of PowerSeries 199)


gives)

(1

1
X)

= \"nXn-l 1 + 2x + 3x2 + 4x3 + ... . = 2 \037


n=l)

00

The
that it

will satisfy a given differential equation dependsalso on Theorem This theorem-also without stated us that if two powerseriesrepresent the samefunction, then they are the sameseries. particular, the Taylor seriesin (13) In is the only powerseriesin powersof x a) that represents function f.) the (

o processf determining

the coefficients in the

series = y

L cnxn
2.

so

proof-tells

THEOREM2
If)

IdentityPrinciple

L
n=O

00

anx

L
n=O)

00

bnx

for every point

x in someopeninterval
an

I,then

an

= bn for all n > o.)


it

Theorem 2 that
.....

In particular, if
.....

0 L anx n => for all x in = 0 for all n


o.)

someopen interval,

follows from

Example

1)

Solvethe equation y' + 2y = O.)


We substitute the

Solution

series)
y

\037)

= LcnXn
n=O)

00

and

y' =

L
n=l) n

00

ncnx

-,
1

and

obtain)

L
n=l

00

- +2 ncnx
n
1

L
n=O)

00

cnx

= O.

(21))

To comparecoefficientshere,we needthe general term in each sum to be the term n this, we shift the index of summation in the first sum. containing x To accomplish To seehow to do this, note that)

L
n=l

00

ncnx

= Cl + 2C2X + 3C3X2 + ... =

L
n=O)

00

(n

+ I)C +lX n . n

Thus we can replace with n + 1 if, at the sametime, we start counting one step n lower;that is, at n = 0 rather than at n = This is a shift of +1 in the index of summation. The result of making this shift in Eq. (21)is the identity)

1.

00

L(n + I)Cn +lX n + 2


n=O

L
n=O)

00

cnx

= 0;

that is,)

L[(n+ I)C +l + 2c ]x = O.
n n n
n=O)))

00

200

3 Methods) Chapter PowerSeries If this equation holds on that (n n n I)C +I

+ 2c = 0 for all n > 0;consequently,)


Cn+I

some interval,

then it

follows from the

identity

principle

\037)

2c -n+l)
n

(22))

for all n > O. Equation (22)is a recurrence relationfrom which we can succesin terms of co;he latter will turn out to be the arbitrary t sivelycompute CI,C2,C3, constant that we expect find in a general solutionof a first-orderdifferential equato

..

tion.) With n

= 0,Eq. (22)gives)
CI =

--.
2co
1)

With n

= 1, q. (22)gives) E
C2 =

--=+-=-.
2CI 2 22co 1 2

22co
2!)

With n

= 2,Eq. (22)gives)
2C2 3

- - 1.2 C3----- . By now


\037)

it

should be clearhat after n such steps,we will have) t

c= (-I)
n

2nco
, n.)

23co 3

--23co
n

3!.)

'

> 1.)
00
n ( 2X )

(This is easyto prove by induction on n.) Consequently,our solutiontakes the form)

y(x) =
In the
final

L
n=O

00

Cnx

L(-1)-x = .
00
n

2n Co

n=O

n'

Co

L n' n=O.)

-2x = coe .

in stepwe have usedthe familiar exponential series Eq. (5)to identify our 2x seriesolution as the same solution y(x) = coe- we couldhave obtained s power

by the method of separation of variables.

.)

ShiftofIndexof Summation)
In the solution of Example1 we wrote)

L
n= I

00

ncnx

= L(n + I)C +IX n n


n=O)

00

(23))

by
n

the index of summation by +1 in the series the left. That is, we on shifting the n simultaneously increased index of summation by 1 (replacing with n + 1,
\037

o e is obtaining the seriesn the right. This procedure valid because achinfinite in (23)is simply a compact notation for the singleseries)
CI

the + 1)and decreased starting

point by

1,from n =

1 to

= 0,thereby

series
(24))))

+ 2C2X + 3C3X2 + 4C4X3 + ...

.)

3.1Introductionand Review of PowerSeries 201)


Moregenerally, we can shift the index of summation by k in an infinite series the n+k) and decreasing by simultaneouslyincreasing summation index by k (n the starting point by k. For instance, a shift by n 2) yields) (n
\037

+2

\037

00 \037 n=3

\037anx

n-l = \037
n=l)

00

\037an+2x

n+l

. -k = Ikl. Thus a

If k is negative, we interpret a \"decreasey k\" as an increasey b b shift by n (n 2) in the index of summation yields)

-2

\037

L
n= 1

00

ncnx

= L(n
n=3)

00

- 2)C

n _2Xn

-;
3

we have decreased index of summation by 2 but increased starting point by the the 2,from n = 1 to n = 3.You should checkthat the summation on the right is merely another representation of the series (24). in We know that the power series obtained in Example1 converges for all x because is an exponential series.More commonly, a power series solution is it not recognizable terms of the familiar elementary functions. When we get an in unfamiliar power series solution, we need a way of finding where it converges. After all, y = L cnxn is merely an assumedform of the solution. The procedure illustrated in Example1 for finding the coefficients {cn}is merely a formal process and mayormay not be valid. Its validity-in applying Theorem 1 to compute relation for the coefficientsy' and applying Theorem 2 to obtain a recurrence dependson the convergence of the initially unknown series = L cnxn Hence y this formal process justified only if in the end we can show that the power series is we obtain converges on some open interval. If so,t then represents solution of i a the differential equation on that interval. The following theorem (which we state without proof) may be usedfor this purpose.)

THEOREM3

Radiusof Convergence Given the powerseries cnx n , supposethat the limit) L


\037)

. hm p= n-+oo

C n
Cn+l)
I

(25))

exists(p is finite) or is infinite (in this case will write p = (0). hen we T

(a) If p = 0,then the seriesiverges for all x i= O. d t (b) If 0 < p < 00,hen L cnxn converges if Ix < p and divergesif Ixl t (c) If p = 00,hen the seriesonvergesfor all x.) c

> p.

o L cnxn . For instance, for the powerseriesbtained in Example1,we have)

The number p in (25)is calledthe radiusof convergence the power series of

. hm p= n-+oo

+ (_1)n+12n+lco/(nI)!)

n (-1)2nco/n!

n-+oo

lim

n+l = 00,
2)

and consequently the series e obtained in Example1 converges for all x. Even w if the limit in (25)fails to exist,there always will be a number p such that exactly one of the three alternatives in Theorem 3 holds. This number may be difficult to will be quite w in find, but for the powerseries e will consider this chapter, Eq. (25) sufficient for computing the radius of convergence.)))

202

3 Methods) Chapter PowerSeries

Example2)
Solution

Solvethe equation (x 3)y' + 2y == O.)


As

before,we substitute)
00
y

\037)

==

LcnX n=O)
00

and

y' ==
00

L
n=l)

00

ncnx

to

obtain)

(x - 3) L
n= 1

ncnx
00

- +2
1

Lcnx
n=O)

==

0,

so that)

Lncnx - 3 Lncnx - + 2 L
00
00
n n
1

n cnx == O.

n=l

n=l

n=O)

In the

T secondsum we shift the index of summation by +1. his yields)

first

n sum we can replace == 1 with

n == 0 with

no effect on the sum. In the

L
n=O

00

ncnx

- 3 L(n + I)C
00
n=O n

n +lX

+2

Lcnx
n=O)

00

==

0;

that is,)

L[nc - 3(n + I)C +l + 2c ] x


00
n n
n=O)

== O.

The identity principlethen gives)


nCn from which we obtain the
\037)

- 3(n + I)C +l + 2c
n

== 0,)

relation) recurrence

C n+l

- n+21) +
3(n

n)

for

> O.)
and
find that)

We

apply this formula


Cl ==

with n ==

0,n

==

a 1, nd n == 2,in turn,
')

-co, 3)

C2 ==

3 -Cl 3 .2

==

3 -co 32

and

C3 ==

4 4 -C2 3cO. 3.3


==
3) difficult

This is almost enough to make the induction on n that)


\037)

pattern

evident; it is not

to show by

C == n

n+l
3n
Co)

if

> 1.)

Hence proposedpowerseriesolution is) our s y(x)==co


Its radius of convergenceis) p == nlim
---+

OOn x. L 3+ 1n
n
n=O)

(26))

n)

hm
n---+oo

3n

00)

Cn+l)))

+ 3 == 3. n + 2)

3.1Introductionand Review of PowerSeries 203)


< x < 3 and diverges if Ixl > 3. In this Thus the series (26)converges if in particular examplewe can explain why. An elementary solution (obtainedby separation of variables) of our differential equation is y == 1/(3 x)2.If we differentiate termwise the geometric series)

-3
1

3-x)

- - 3 L3 '
3

x=
3)

00

xn
n

n=O

we get a constant multiple of the series (26).Thus this in constant Co appropriately chosen)epresents solution) the r

serieswith (

the

arbitrary

y(x) =

(3 _ X)2)

< x < 3, and the singularity at x == 3 is the reason why the on the interval radius of convergenceof the powerseriesolution turned out to be p == 3. s
Example3)
Solution
Solvethe equation)
We

-3

.)

==y-x-l.)
substitutions
y

make the usual

==
n

L cnxn and y' == L ncnx n -l , which


00
n
n=O)

yield)

x2

L
n=l

00

ncnx

- == -1- x + cnx L
1

LnCnxn+1 -1 x + LCnxn. Because the presencef the two terms -1 -xon the of and o side,we need to split off the two terms, Co+ Clx, of the seriesn the o for comparison.If 2 we also the on the left by -1 (replace 1 - index of
00 00

so that)

==

n=l

n=O)

first

and

shift n with n

summation

right-hand right n ==

with n

==

1),we get)

L(n - I)C
00

n _lX

==

-1- x +

Co

n=2

xn + CIX + LCn .
n=2)

00

Because the x to the first


(n

left-hand side contains neither a constant term nor a term containing n power, the identity principlenow yieldsCo == 1,C1 == 1,and C == 1)Cn-l for n > 2.It follows that)
C2 == 1 Cl ==

I!,

C3 == 2 C2 ==

2!, C4== 3 .C3 == 3!,)


n

that) and, in general,


\037)

C == (n n

- I)! for
00
n=2)))

> 2.)

Thus we obtain the powerseries)

y(x)

== 1

n . +x + L(n -l)!x

204

3 Methods) Chapter PowerSeries


But the radius of convergenceof this

series is)

. (n I)! lim 1 hm p = n-+oo n! = n-+oo n) = 0,

so the series converges only for x = O. What does this mean? Simply that the differential equation doesnot have a (convergent)powerseriesolutionof the s given assumedform y = L cnxn . This exampleservesas a warning that the simpleact of n writing y = L cnx involves an assumption that may be false.

.)

Example4)

Solvethe equation y\"

+ y = O.)
00

Solution If we assumea solution of the form)


\037)

= LCn , xn
n=O)

we find
\037)

that)

y' =

L
n=})

00

-} ncnx
n y\" in

and)

\"

= \037 n (n
'\"\"'\"

00

1) Cnx n-2

n=2)

Substitution

for y and

the differential equation then yields) n(n


n n

L - 1)cx - + Lcnx = O.
00

00

n=2

n=O)

We shift the

and

n with

index of summation n + 2). This gives)


00

in

the

first sum by

n +2(replace = 2 with

=0

n L(n + 2)(n+ l)c+2xn + n=O

L
n=O)

00

cnx

= O.
identity

The identity (n + 2)(n+ l)c+2 + Cn = 0 now follows from the n and thus we obtain the recurrence relation)
\037)

principle,
(27))
with

C n+2

-- + 1)(n+
C n (n
will

2))

for n

> O. It is evident that

this

formula

determine the coefficientsCn

even

subscriptin terms of Co and those of odd subscriptin terms of C};Co and C}are not and to predetermined thus will be the two arbitrary constants we expect find in a
general solution of a second-order equation. When we apply the recurrence relation in (27)with we get)
C2 =
n

Taking

and 1,3, 5 in turn,


C3 =

-2!'
Co C}
')

C4 =

Co

4!'

and)

C6 =

we find
5

that)

- 3! C-- 5!')
C}

and)

C7

--. = --.
Co 6!) C}
7!)))

= 0,2,and 4 in

turn,

3.
1

Introductionand Review of Power eries 205) S


that

Again, the pattern

is clear; e leave it for you to show (by induction) w


C2k ==)

for k

>

1,)

\037)

k -1)co

(2k)
4

and)

C 2k+l

!)

c1 -1)k + I)!.) (2k


( 5
X
7

Thus we get the powerseriesolution) s


6

y(x) == Co 1
that

. - x2 + x - x + ) + (x - x3 + x - 7! +... 3! 5! ( 2! 4! 6! )
Cl
')

- - - ... - - (-1 )n

is, y(x) == Cocosx+ Cl sinx.Note that we have no problem with the radius of f functions convergefor convergence here;the Taylor seriesor the sine and cosine
\037lx.

.)

The solution of Example4 can bear further comment. Supposethat we had never heard of the sineand cosine We functions, let alone their Taylor series. would then have discovered the two powerseriesolutions) s

C(x) =
and)

L
n=O)

00

2n

(2n)!

=1

- 2!+ x 4!
X

-...
.

(28))

S(x) =

x2n+l (-I)n = x - x3 + x5 - ...) 3! Sf \037 (2n + I)!


00

(29))

of the differential equation y\" + y == O. Both of these powerseriesonvergefor c all x. For instance, the ratio test in Theorem 3 impliesconvergencefor all z of the series (-1)n /(2n)!obtained from (28)by writing z == x2 Hence followsthat it zn L in (28)itself convergesfor all x, as does (by a similar ploy) the series (29). It is clearhat C(0) == 1 and S(0) == 0,and termwise differentiation of the two t

series (28)and (29)yields) in

C'(x)== -Sex) and S'(x)== C(x).)

(30))

Thus with the aid of the power series ( ( Consequently, C' 0) == 0 and S' 0) == method (all the while knowing nothing about the sine and cosine functions), we have discovered y == C(x) is the unique solution of) that
y\"

1.

+y

== 0)

conditions y(O) == 1 and y' (0) == 0,and that y == Sex) initial conditions y (0) == 0 and y' (0) == It follows that C(x) and Sex)are linearly independent, and-recognizing imthe of the differential equation y\" + y == O-we agree call C the cosine can to portance function and S the sine function. Indeed,all the usual properties thesetwo funcof tions can be established, sing only their initial values (at x == 0) and the derivatives u in (30);there is no need to refer to triangles or even to angles.(Can you use the == seriesn (28)and (29)to show that [C(x)]2 [S(x)]2 1 for all x?) This demoni + strates that)
that

satisfies the

initial

is the unique solution that satisfies the

1.

The cosine nd sinefunctions are fully determined the differena by tial equationy\" + y = 0 of which they arethe two natural linearly s independentolutions.)))

206

3 Methods) Chapter PowerSeries

x and show Figures3.1.2 3.1.3 how the geometric characterof the graphs of cos and sin x is revealed by the graphs of the Taylorpolynomial approximationsthat we get
by
truncating

the

infinite

This is by no means an uncommon situation. Many important specialfunctions of mathematics occurin the first instance as powerseriesolutions of differs ential equations and thus are in practicedefined by means of these powerseries. In the remaining sectionsof this chapter we will seenumerous examples such of
functions.)
y) y)

series (28)and (29). in

n=8)

= 16)

n=24)

n=5)

13)

n = 21)

2)

2)

-1
-2) n=6)
n = 14)

x)

-1
-2)

n=22)
approximations to

n=7)

= 15)

= 23)
to

FIGURE cosx.)

3.1.2. Taylor

polynomial

FIGURE
SIn x
.)

3.1.3. Taylor

polynomial

approximations

P 11II roblems)
Problems through solution of the find a powerseries Determine the radius of convergiven differential equation. genceof the resulting series,and use the seriesin Eqs.(5) to identify the series solution in terms offamilthrough
In

10,

iar elementary

(12)

functions.

from checking your work


methods

by

of earlierchapters!))

(Ofcourse,no onecan prevent you also solving the equations by the

relation Problems through 22,first derive a recurrence n giving C for n > 2 in terms of Co or Ct (or both). Then apply the given initial conditions to find the values of Coand Ct. Next determine C (in terms of n, as in the text) and, finally, n identify the particular solution in terms offamiliar elementary In functions.)

19

3.2y' + 3y = 0
5. y' = x 2Y
7. (2x l)y' + 2y = 0
0)

1.

y/

=Y

9. (x - l)y' + 2y =
11

4. y/ + 2xy = 0 6.(x 2)y'+ y = 0 8.2(x + l)y' = Y 10. (x l)y' = 3y) 2

2.y/ = 4y

19.\" + 4y = 0; y(O) = 0, y/(O) = 3 y 20.y\" - 4y = 0; y(O) = 2, y/(O) = 0

21.\" - 2y' + y = 0; y(O) = 0, y/(O) = 1 y


x 2y\"

22.y\" + y/ - 2y = 0; y(O) = 1, /(O) = -2 y 23.Show that the equation)


+ x 2y' + y = 0)
y

In Problems through 14,usethe method ofExample 4 to find two linearly independent power seriessolutions of the given differential equation. Determine the radius of convergence f o eachseries, nd identify the generalsolution in terms offamila

has no power seriessolution of the form

24. Establish
lowing
initial

= L cnx n .
fol-

iar elementary

functions.)

13.\" + 9y = 0) y
Show

11. =
y\"

12.\" = 4y y 14.\" + y = y
3) that
the

x)

solution yield a powerseries differential equations in Problems

(as in

Example

powerseriesmethod fails to the form y = L cnx n for the of

15.y/ x

+y =0 17.x 2y/ + y = 0)

15through 16. xy' = y 2 18. 3y/ = 2y) x

18.)

25.For the

binomial seriesin solution of the initial value problem in part (a), if and that this seriesconverges Ix < (c) Explain why follows the validity of the binomial seriesgiven in from parts (a) and (b). that the

steps. (a) Show that y = (l + x)Q' satisfies the value problem (1+x)y' = ay, y(O) = 1.b) Show (
power seriesmethod gives the
I

the binomial

seriesin (12)y means of the b

a (12) s the

1.

(12)

initial

value problem)

y\"

= y/ + y,

y(O) = 0,

y(l) =

1)))

N 3.2Series olutions earOrdinary Points S


derive the power series solution y(x)
where

207)

27. This sectionintroduces


F.
\037xn

L
n=l

00

the useof infinite seriesto solve equations. Conversely, differential equations F can sometimes be used to sum infinite series. or examdifferential the ple,consider
infinite

n!)

series)

26.(a) Show

Fn

.. =

{Fn}\037o is the sequence0, 1,1,2, 3, 5, 8, 13, of Fibonaccinumbers defined by Fo = 0, Fl = 1,

Fn-2

+ Fn-l for n >

that the solution

of the

1.

initial

value problem

Y (0) = 0 is y(x) = tanx. (b) Becausey(x) = tan x is an odd function with y' (0) = 1,ts Taylor series of the form i is 7 s X3 y = x + C3 + csx + C7X + ... . Substitute this seriesin y' = 1+ y2 and equate like powers

y' = 1 + y2,

note the + + the terms evaluate this

on

. 1+---+-+---+... 5! I! 2! 3! - - ... e.
1 1 1 1 1

4!

')

++ pattern of signs superimposed We could of the seriesfor the number seriesif we could obtain a formula for the
1 13 14 1 - -x + -x + -x - -xs +... 2! 3! 4! 5!
2
')

function)

f(x) = 1 + x

of x to derive the
3C3 =

following

relations:)

1,

5cs = 2C3,
9cg

7C7 = 2cs + (C3)2, = l1cII 2cg + 2C3C7+ (CS)2.


that) (c) Conclude

2C7+ 2C3CS,)

because sum of the numerical seriesin question is simthe to ply f(I). (a) It's possible show that the power series hereconvergesor all x and that termwise differenf given tiation is valid. Given thesefacts, show that f (x) satisfies
the initial y(3)

value problem)

= y;
1

y(O) = y' (0) =

1,

y\" (0)

= -1.)

tan

1 = x + -x3 + 3

2 17 -xs + _x 15 315

(b) Solvethis

initial

value problem to show

that)

(d)
in
it!)

t (13)o derive the

Would you

62 9 1382 II x + x + 2835 155925 preferto use the Maclaurin seriesformula


tangent
in series

+....

v'3 .v'3 f(x) = -ex+ _e-x/2 cos 2 + v'3 sIn 2) 3 3 ( ) 2

-x

-x

part

Think (c)?

about

For a suggestion, seeProblem 48ofSection (c) Evaluate ( 1)to find the sum of the numerical seriesgiven
here.)

2.3.

Se:ries Sollltions ar Ordina\037Po\037nts) \037e


The powerseries ethod introduced in Section can be appliedto linear equam tions of any order(as well as to certain nonlinear equations), but its most important l s applications are to homogeneous econd-orderinear differential equations of the
form) A(x)y\"

3.1

+ B(x)y'+ C(x)y== 0,)

(1))

where the coefficients A, B, and C are analytic functions of x. Indeed,in most applications thesecoefficientfunctions are simplepolynomials. We saw in Example3 of Section that the series ethod does not always m a seriesolution. To discoverwhen it does succeed, rewrite Eq. (1)in the we s yield

3.1

form)
\037)

y\"

+ P(x)y' + Q(x)y == 0)

(2))

with

considerthe equation)

leading coefficient 1,and with P == B/ A and Q == C/ A. Note that P(x) and Q(x)will generally fail to be analytic at points where A(x) vanishes. For instance,

xy\"

+ y' +xy == O.)

(3))))

208

3 Methods) Chapter PowerSeries The coefficientfunctions in (3)are continuous everywhere. But


in

is the equation)

the form of (2)it

y\"

1 + x) y' + y = 0

(4))

P (x) = 1/ not analytic at x = O. x The point x = a is calledan ordinary point of Eq. (2)-andf the equivalent o that the functions P(x) and Q(x) are both analytic at x = a. Eq. (I)-provided x Otherwise, = a is a singularpoint.Thus the only singular point of Eqs. (3)and is x = O. Recall that a quotient of analytic functions is analytic wherever the (4)
with

denominator is nonzero.t follows that, if A(a) 0 in Eq. (1)with analytic coefI then x = a is an ordinary point. If A(x),B(x),and C(x) are polynomials ficients, with no common factors, then x = a is an ordinary point if and only if A(a) O.)

i-

i-

n..,.....u..u..........\".u...................

.....

...\037v

.....

.....,..\",..,....

Example

1)

The point

x = 0 is an ordinary
xy\"

....\"'...,...\"'.\"_....\037

\037\"

,,,'..,........_,.....'....,

\",..........\037\"'\"

point of the equation)

+ (sinx)y'+ x2y = 0,)

despitethe fact that

(x) = x vanishes at x = O.The reasonis that)


x3 x2 x - 3!+ x - ) = - 3!+ Sf - ...) Sf (
5 4 1

sin x 1 P(x) = \037= x x

...

is neverthelessanalytic at x = 0 because division by x yieldsa convergentpower the

series.

.)

Example2)

The point

x = 0 is not an ordinary
y\"

point of the equation)


1

+ x2y' + x / 2y = O.)

For while P(x) = x2 is analytic at the origin, Q(x) = x1 / 2 is not. The reasonis that Q (x) is not differentiable at x = 0 and hence not analytic there. (Theorem 1 is of Section impliesthat an analytic function must be differentiable.)

3.1
..............

.)

Example3)

The point

x = 0 is an ordinary

point of the equation)

4 (1 x3)y\" + (7x2 + 3x5)y' + (5x I3x )y = 0)

becausethe coefficient functions A(x), B(x), and C(x) are polynomials with
A(O)

i-O.

.)

Theorem 2 of Section impliesthat Eq. (2) has two linearly independent 2.1 solutions on any open interval where the coefficient functions P(x) and Q(x)are continuous. The basicfact for our presentpurposeis that near an ordinary point a, these solutions will be powerseries powersof A proof of the following in theorem can be found in Chapter 3 of Coddington, An Introduction to Ordinary P Cliffs, N.J.: rentice Hall, 1961).))) DifferentialEquations (Englewood

x-a.

N 3.2Series olutions earOrdinary Points S


THEOREM 1

209)

Near Solutions an OrdinaryPoint


A(x)y\"

Supposethat a is an ordinary point of the equation)

+ B(x)y'+ C(x)y= 0;)

(1))

is, the functions P = BIA and Q = CIA are analytic at x = e has two linearly independent solutions, achof the form)
that
\037)

a.Then Eq.(1)
(5))

y(x) = LCn(X a)n.


n=O)

00

The radius of convergence of any such seriesolution is at leastas large as the s T distancefrom a to the nearest (real or complex) singular point of Eq. (1). he in coefficients in the series (5)can be determined by its substitution in Eq. (1).)
..

\037

\037

'\"

Example4)

Determine the radius of convergenceguaranteed by Theorem' 1 of a seriessolution


of)

(x2 + 9)y\"
in

+ xy' + x2y = 0)

(6))

in powersof x. Repeat for a series powersof x 4.)

Solution This exampleillustrates the fact that we must take into account complexsingular
B points as well as real ones. ecause)

P(x) =

x x2 + 9)

and)

2 = 2x Q (x) x +9

')

Fig.3.2.1).

The distance(in the complex the only singular points of Eq. (6) are plane) of each of these from 0 is 3, so a series solution of the form cnxn has radius of convergence at least The distanceof each singular point from 4 is 5, so a seriesolution of the form Cn (x 4)n has radius of convergenceat least 5 (see s

::l::3i.

3.

.)

y)

x)

FIGURE

3.2.1. Radius

of convergences distance to nearest singularity.))) a

210

3 Methods) Chapter PowerSeries

Example5)

Find the general solution in

powersof x of)

(x2

4)y\"

+ 3xy' + y = o.)

(7))

Then find the particular solution with y(O) = 4, y/ (0) = 1.)

Solution The only

w so the singular points of Eq. (7) are ::l::2, series e get will have radius of convergence at least 2. (See Problem 35 for the exact radius of convergence.)
Substitution

of)
00
/ Y

\037)

= LcnXn ,
n=O)

= \037 ncnxn - ,
00
'\"\"\"\"

and

y\"

n=l)

L - 1)cx 00

n(n

n=2)

in

Eq. (7)yields)

L - l)cx - 4 L - 1)cx - + 3 Lncnx + Lcnx = O.


00 00

n(n

n(n

00

00

n=2

n=2

n=l

n=O)

We

can beginthe first and third summations at n = 0 as well,because o nonzero n terms are thereby introduced.We shift the index of summation in the secondsum by +2, replacingn with n + 2 and using the initial value n = O.This gives)

L - l)cx - 4 L(n + 2)(n+ 1)c


00 00

n(n

n +2xn

+3

n=O

n=O

L
n=O n

00

ncnx

Lcnx = o.
n
n=O)

00

After

c collectingoefficientsof Cn

and C +2, we obtain) n

L[(n +
2
n=O)

00

2n

+ l)c n

- 4(n + 2)(n+ 1)c+2]x = O.


n n +2

The identity principleyields)


(n which
\037)

+ 1)2c n

- 4(n + 2)(n+ 1)c

= 0,)

leadsto the recurrence relation)

cn+2

for n

> O. With
c2

- 4.2')
Co
in this

-- n

(n

+ l)c n 4(n + 2))


C6 =

(8))

= 0,2,and 4 in
C4 =

turn,

we get)
')

4 .4

3C2 =

3co 42 2 4

. .

and)

.
n
n!)))

5C4 = 3 5co 4 6 43 2 4

. . .6).

Continuing

fashion, we evidently would


C2n

=)

1 3 5
n

. . ...(2n - 1) 4 .2 .4 ... (2n)) Co.

find that)

With

the common notation)

(2n +

= I)!! 1 .3 .5 ...(2n + 1) = (2n + I)! 2 .

N 3.2Series olutions earOrdinary Points 211) S

\037)

.4 .6 ...(2n)== 2 .n!, e w obtain) (2n - I)!! C2n == Co. 23 .n!) (We alsoused the fact that 4 .2 == 2 .) With n == 1,3, 5 in Eq. (8),we get) and 2 .4c 6c == 2 .4 .6c 2CI 4C3 5 C3 . and) C7 == Cs == . == 2 . . 4.3 4 5 4 3 5 4 . 7 43 .3 .5 .
and the observation that 2
n finally n

(9))

-- n

3n

')

')

n!

7)

It

is apparent that the pattern


C2n+I == n 4

is)

\037)

. . ...(2n) n! CI. . . . ...(2n + 1)CI == 2 . (2n + I)!!)


2 4 6 1 3 5
n

(10))

formula
solution)

The formula in (9)gives the coefficientsof even subscriptn terms of Co;the i in (10)gives the coefficients of odd subscriptin terms of CI. After we the we o separately collect terms of the seriesf even and odd degree, get the general

y(x) == Co 1 +

f
n

=I 2
(2\037

- 1)!! x
.n!

2n

+ CI

f
n

=I

2n

x2n+I . (2n + I)!!) ) . (11)

Alternatively,)

y(x) == Co 1 + -x 8

123 + 56+... + -x 128 )


4

1024)

13 + CI x + -x + -x + 140) 6 30
(
Because(0) == Co and y/ (0) == CI, the given initial Y
and CI == the Using these values in Eq. solution satisfying y (0) == 4 and y' (0) == 1 are)

15 -x +... 17 .
)
first

11 '))

1.

(11/),
3

conditions imply that Co == 4 few terms of the particular

y(x) ==4+x+ x 2

1 2

13 + 6 x + 32x4 + 30x

15 +....

(12)

.)

of y == L cnxn in a linear secondorderequation with x == 0 an ordinary point typically leadsto a recurrencerelation that can be used to expressach of the successive e coefficients C2, C3, C4, in terms of the first two, Coand CI.In this event two linearly independentsolutionsare obtained as follows.Let Yo(x) be the solution obtained with Co== 1 and CI == 0,and let YI (x) be the solution obtained with Co == 0 and CI == Then)

Remark:As in Example5, substitution

...

1.

Yo(O) ==

1,

yb(O) == 0

and

YI(O) ==

0,

y\037(O)

== 1,)

so it is clearhat Yo and YI are linearly independent.n Example5, Yo(x) and YI (x) t I are defined by the two series appear on the right-hand side in Eg.(11), that which the general solution in the form Y == CoYo + CIYI. .))) expresses

212

3 Methods) Chapter PowerSeries

Translated eriesSolutions) S
If
in

and

y/(a),we would have neededthe general solution in


00

Example5 we had sought a particular solution with given

initial

values yea)

the form)

_a)n; y(x) == LCn(x


n=O)

(13))
with

that

is,in

the form

in

powersof x a rather than in powersof x. For only (13)is it true that the initial conditions) yea) == Co and y/(a) == Cl)

a solution of

determine the arbitrary constants Co and Cl in terms of the initial values of y and y/. Consequently, to solve an initial value problem,we need a seriesxpansion of the e general solution centeredat the point where the initial conditions are specified.)

Example6)

Solvethe initial value problem)


(t 2

d - 2t - 3)-+ 3(t - 1)+ dt


2
y

dy
dt)

==

0; y(l) == 4, y'(I) ==

-1. (14)
this

Solution

We

seriesn (14)to determine the coefficients,it simplifies the computations if we first i make the substitution x == t 1, o that we wind up looking for a seriesf the form o s L cnxn after all.To transform Eq. (14)into one with the new independent variable x, we note that)

needa general solution of the form

L Cn (t - l)n. But instead of substituting

- 2t - 3

== (x

- -- dy == dy

+ 1)2 2(x+ 1)
dx dt dx == dy
== dx)

-3

== x 2

- 4,

dt

/ y,

and)

d2y dt 2
into)

_ - [dx (dx)]dx _ dx - _
\037

dY

\037

dt

(y )

II

,)

where primesdenotedifferentiation with respectto x. Hence transform Eq. (14) we

(x2

4)y\"

+ 3xy' + y

== 0

T conditions y == 4 and y/ == 1 at x == 0 (corresponding t == 1). his is to value problemwe solved in Example5, so the particular solution in (12) is available. We substitute t 1 for x in Eq. (12)and thereby obtain the desired particular solution)
with initial the initial

yet) == 4 + (t

- 1)+ -(t - 1)2+ -(t - 1)3 2 6 + -(t- 1) + -(t- 1) +.... 32


1 1 3 4 1 5
30)))

Near 3.2Series olutions Ordinary Points 213) S

< t < 3.(Why?) A seriesuch as this can be usedto s This seriesonverges if c estimate numerical values of the solution. For instance,) + + y(0.8)= 4 + (-0.2) 2 (-0.2)2 6 (-0.2)3
5 4 + -(-0.2) -(-0.2) + 30 32

-1

1 -

1 -

+...
')

so that y(0.8)

\037

3.8188.)

.)

The last computation in Example6 illustrates the fact that seriessolutionsof differential equations are useful not only for establishing general properties of a of but also for numerical computations when an expression the solution in solution, terms of elementary functions is unavailable.)

R Typesof Recurrence elations


The formula in Eq. (8)is an example a two-termrecurrence of relation; it expresses coefficients.A manyeach coefficientin the series terms of one of the preceding in in termrecurrence each relation expresses coefficient in the series terms of two or more precedingcoefficients. In the case a many-term recurrence relation, it is of to generally inconvenient or even impossible find a formula that gives the typical coefficient Cn in terms of n. The next exampleshowswhat we sometimescan do with a three-term recurrence relation.)

Example7)
Solution

Find two linearly independent solutions of)


/I

- -x
xy
I

=0

.)

(15))

We

make the usual


00

substitution

of the powerseries y
n n

= L cnxn . This results in


nX n +2

the

equation)

Ln(n - l)cx - LncnX - Lc


2
00 00
n

= O.

n=2

n=l

n=O)

We

can start the secondsum at n = 0 without changing anything else.To make each term include xn in its general term, we shift the index of summation in the first n in the third sum (replace sum by +2(replace with n + 2),and we shift it by n
with n

- 2).These
00
n=O

-2

shifts

yield)

n L(n + 2)(n+ l)c+2xn

- LncnX - LC -2
00 00
n n n=O
n=2)

Xn

= O.

The common range of these three summations is n > 2, so we must separate the terms corresponding n = 0 and n = 1 in the first two sums before collecting to coefficientsof xn This gives)

2C2+ 6C3X

CIX

L[(n + 2)(n+ l)c - 00


n +2

nCn

C n

xn -2] = O.

n=2)

The identity principlenow impliesthat 2C2=

recurrence relation)
\037)

0,that

C3 =

iCl,and the three-term


(16))))

C +2 = n

+ Cn -2 (n + 2)(n+ 1))
nCn

214

3 Methods) Chapter PowerSeries for n > 2.In particular,)


C4 =)
C7 =)

2C2+ Co
12)

Cs =)

3C3 Cl 20)

+ +

C6 =

4C4+ C2
30) (17))

5cs+ C3
42)

6C6 C4 Cg =) 56)

Thus all values of Cn for n > 4 are given in terms of the arbitrary constants Co and Cl because = 0 and C3 = C2 To get our first solution Yl of Eq. (15), chooseo = 1 and Cl = 0,so that we C C2 = C3 = O.Then the formulas in (17)yield)
\037Cl.

C 4
thus)

-- 12')
1
Yl

Cs = 0,)

C6

-1

90')

C7 = 0,)

Cg

. - 1120')
3
(18))

(x) =

x x x + 12 4 + 90 6 + 1120
g

+....)

Because = C3 = 0,it is clearrom Eq. (16)that this seriesontains only terms of f c Cl even degree. To obtain a second ( we linearly independent solution Y2 ofEq. 15), take Co = 0 = 1, o that C2 = 0 and C3 = Then the formulas in (17)yield) and Cl s
C4 = 0,) Cs

\037.

40')

C6 = 0 ,

C7 =

13
1008')

so that)

Y2(X)=X+-X3 6

3 13 +-xs + 1008)x +.... 40


7

(19))

Because = C2 = 0,it is clearrom Eq. (16)that this seriesontains only terms c f Co of odd degree.The solutions Yl (x) and Y2 (x) are linearly independent because A (0) = 0,whereasY2 (0) = 0 and (0) = 1. general solution of Yl (0) = 1 and in (15)is a linear combination of the powerseries (18)and (19). Eq. Equation (15) has no singular points,so the powerseriesepresenting Yl (x) and Y2 (x) converge r .) for all x.
Y\037 Y\037

The Legendre quation E


The Legendrequation ordera is the second-order of linear differential equation) e
\037)

(1 x2)y\"

where the real number the inequality has extensive applications, anging from r integration formulas (such Gaussianquadrature) to the problemof determining the steady-state temperature within a solid sphericalball when the temperatures at points of its boundary are known. The only singular points of the Legendreequation are at +1 and so it has two linearly independent solutions that can be expressed powerseries as

- 2xy' + a(a+ l)y = a satisfies a > -1. differential equation This as numerical
0,) (20))

-1,

in)))

N 3.2Series olutions earOrdinary Points 215) S


T powersof x with radius of convergenceat least 1. he substitution Problem 31)o the recurrence t relation) Eq. (20)leads(see
\037)

==

L cmx

in

Cm+2

+ - - (a - m)(a+ +2)1) + 1)(m


m (m

m)

(21))

for m > O. We are using m as the index of summation because e have another role w for n to play. In terms of the arbitrary constants Coand CI,Eq. (21)yields)
C2 == C3 ==

C4 == Cs =)
We

+ -a(a2! 1) - (a-l)(a+2) 3! a(a- 2)(a + 1)(a + 3) 4! - 1)(a - 3)(a + 2)(a + 4)CI. (a


Co,)

CI,)

Co,

S!)

can show without much trouble that for m

> 0,)

C2m ==

(-1)maCa

- 2)(a - 4)...(a-

2m

+ 2)(a + 1)(a 3)..(a+ 2m 1)Co + .


(2m)

(22))
and)

C2m+! =

- - (a - 2m + 1)(a+ (a (-1) (a 1)(a 3)... (2m + 1)! 2)(a + 4)...+ 2m) C!.
m

(23))
Alternatively,) C2m

= (-I)ma2mcO

and

C2m+1 =

(-I)ma2m+lCl,)

where a2m and a2m+1 denote the fractions in Eqs. (22)and (23),respectively.With this notation, we get two linearly independent powerseries solutions)
YI(X)

= CoL(-I)m 2m a2mx
m=O)

00

and

Y2(X)

= CI L(-I)ma2m+IX2m+l
m=O)

00

(24))

a = n, a nonnegative integer. If a == n is even, we see = 0 when 2m > n. In this case, I (x) is a polynomialof from Eq. (22) Y I nand Y2 is a (nonterminating) infinite series. f a == n is an odd positive degree integer, we seefrom Eq. (23)that a2m+1 == 0 when 2m + 1 > n. In this case, Thus nand YI is a (nonterminating) infinite series. Y2 (x) is a polynomial of degree in either case, ne of the two solutions in (24) is a polynomial and the other is a o
Now

of Legendre's equation of order

a.

supposethat that a2m

nonterminating

series.)))

216

Methods) 3 Chapter PowerSeries


With an appropriate choice (madeseparately for each n) of the arbitrary constants Co(n even) or CI (n odd), the nth-degree polynomial solution of Legendre's

equation of ordern,)
\037)

(1 x2)y\"

- 2xy' +
j

n(n

+ l)y = 0,)

(25))

is denotedby
tomary
that

n. the Pn (x) and is called Legendreolynomialof degree It is cusp the (for a reasonindicated in Problem 32)to choose arbitrary constant so n It the coefficientof xn in Pn (x) is (2n)! [2 (n !)2].then turns out that)

\037)

pn(X )

-L
N k=O

2n k!(n

(-1)k(2n-2k)!n-2k X

- k)! - 2k)!
(n

')

(26))

where N

= [nj2], integral the


1 1

part of

nj2.The first

six Legendre polynomials are)

Po(x) = 1,)
2 P2(x) = -(3x

PI (x) = x,)

- 1), 2 P4(X) = (35x - 30x + 8


4 2
in

P3 (x) = (5x3 3x), 2


3),)

P5(x) = \037(63x5
8)

- 70x + 15x),
3

and their graphs are shown

Fig.3.2.2.)
y)

x)

FIGURE3.2.2. polynomials Graphs y = Pn (x) of the Legendre n = 1, , 3, 4, and 5.The graphs are distinguished by the fact that 2 < x < 1.) zerosof Pn (x) lie in the interval

for
all n

-1

Problems)
1 15. 8. (2 x 2)y\" xy' + 16y= 0 9. (x2 l)y\" + 8xy' + 12y= 0 10.y\" +xy'- 4y = 0 3 11.y\" - 2xy' + 10y= 0 5 12.\" - x 2y' 3xy = 0 y 13.\" + x 2y' + 2xy = 0 y 14.y\" + xy = 0 (an Airy equation) 15.\" + x 2Y = y
0)))

Find generalsolutions in powersof x of the differential equarelation State the recurrence tions in Problems through in and the guaranteed radius ofconvergence eachcase.

7. (x2 + 3)y\"

2. (x + 2)y\" + 4xy' + 2y = 0 3.y\"2+xy' + y = 0

1.(x2 2

l)y\"

+ 4xy' + 2y = 0

4. (x + l)y\" + 6xy' + 4y = 0 5. (x2 3)y\" + 2xy' = 0 6. (x2 l)y\" 6xy' + 12y == 0)

-- -

7xy'

+ 16y = 0

- -

N 3.2Series olutions earOrdinary Points 217) S


lems

16 17. and 16. 1+ X 2)y\" + 2xy' (


17.y\"
initial

Usepowerseriesto solvethe

initial

value problems in

Prob-

(b) Differentiate
to succession

eachsideof the

last equation n times in

2y - 2y = 0; y(O)= 0; y(O) ==0, y'(O)= 1 = 1, '(O) + xy' y


0)

(1
22. find a
Thus u tion of in u is formula.

obtain)

x 2)v(n+2)

2xv(n+1)

Solve the

First make a a, then n solution cnt of the transformed differential equation. State the interval of values of x for which Theorem of this section

value problems in Problems substitution of the form t = x

18through

= v(n) = Dn (x2
order n.

(2n)!n!; /

guarantees convergence.)

6x + 10)y\" 6y = 0; y(3) = 2, y'(3) 0 21.4x2 + 16x+ 17)y\" = 8y; y( -2)= 1, '(-2)= 0 ( y 22. (x2+6x)y\"+(3x+9)y' = 0; y( -3)= 0, y'(-3)= 2) -3y

20. (x

18.\" + (x l)y' + y = 0; y(l) = 2, y'(I)= 0 y 19.2x - x 2)y\" - 6(x l)y' - 4y = 0; y(l) = 0, y'(I)= 1 (
2

-=

- - 4(x 3)y' +

33.TheHermiteequationof order a
y\"

(Note 2 (2n) / [2n (n !) ] )

the coefficient of xn then state why this proves Rodrigues' that the coefficient of x n in Pn (x) is

+ n(n + l)v(n) ==
that

O.)

l)n satisfies Legendre's quae

(c) Show

is)

- 2xy' + 2ay

== O.)

(a) Derivethe
YI

two

power seriessolutions)

2m a(a = 1 + \037 \037(_I)m


m=l

- 2) ...(a (2m)!)

2m + 2)

x2m

In Problems 23 through 26, find a three-term recurrenceren lation for solutions of the form y Then find the cnx three nonzero terms in each of two linearly independent first solutions.)

=L

and) Y2

=x)
\037 + \037(_I)m m=I

23. y\" + (1+ x)y = 0 24. (x2 l)y\" + 2xy' + 2xy = 0 25. y\" + x 2y' + x 2y = 0 26. (1+ x 3)y\" + x 4y = 0 27. Solvethe initial value problem)

2m (a

- 1)(a - 3)
(2m

+ I)!)

...(a

2m

+ I ) x2m+I.

y\"

+ xy' + (2x2 + l)y = 0; y(O) = 1, '(O)= -1.) y


sufficiently many

Determine

curate to four In

decimal places.)

terms to compute

(1/2)ac-

if a is an even integer, YI is a polynomial whereas Y2 is a polynomial if a is an odd integer. (b) The Hermite polynomial of degreen is denoted by Hn (x).It is the nth-degree polynomial solution of Hermite's equation, multiplied by a suitable constant so that the coefficient of x n is 2n Show that the first six Hermite polynoShow that

mials are)

the first three nonzero terms in each of two linearly independent solutions of the form n == Substitute known Taylor seriesor the analytic cnx Y f functions and retain enough terms to compute the necessary

Problems 28 through 30,find

Ho(x)

- 1,
2

coefficients.

28. y\" + e-Xy = 0 29. (cos y\" + y = 0 x) 30. xy\" + (sin x)y' + xy = 0 31. erivethe recurrencerelation D

H2(x) = 4x 2, 4 H4 (x) = 16x 48x2 + 12, 3 Hs(x)= 32xs 160x + 120x.)


A

(x) = 2x, H3(x) == 8x3


HI

12x,)

general formula
Hn(x)

for the Hermite polynomials x = (_I)ne2 d e-2) . dx (


X
n)

in

(21)for the Legendre


RoVerify

equation. 32. Follow the stepsoutlined in this problem to establish formula) drigues's

is)

Pn(x)
for the
v

--(x 1 dn n! 2n dx
2
n)

I)n
that

= (x2

nth-degree Legendre polynomial. (a) Show l)n satisfies the differential equation)

(1
Differentiate

- x )v' +
2

2nxv

O.)

are \"interlaced\" -that is, the n zerosof Hn lie in the n open intervals whose endpoints are successive n pairs of zerosof H +1. 34. The discussion 3.1 following Example 4 in Section sugthat the differential equation y\" + y = 0 could be gests used to introduce and define the familiar sine and cosine
bounded functions. In

that this formula does in fact give an nth-degree polynomial. It is interesting to use a computer algebra system to investigate the conjecture that (for each n) the zeros of the Hermite polynomials Hn and Hn+I

(1

eachsideof this
x 2)v\"

+ 2(n

- l)xv'+

equation to obtain)

a similar

fashion, the Airy equation)


\"

2nv

O.)

=xy)))

218
servesto

3 Methods) Chapter PowerSeries


introduce two new

ranging from radio waves to molecular vibrations. Derivethe first three or four terms of two different power series solutions of the Airy equation. Then verify that your results agreewith the formulas)
in applications

specialfunctions

that

appear

y)

y,

(x)= 1 + \037

1 4

. .....(3k - 2)
(3k)!

t:r
2 5
k=l

3k)

x)

and)

Y2(X)

==

+ \037 \037

. .....(3k - 1)
(3k

+ I)!

X3k+1

The FIGURE3.2.3. Airy


Y
==

function graphs

Ai(x) and

==

Bi(x).)

for the solutions that satisfy the initial conditions YI (0) == The (0) == 0 and Y2 (0) == 0, (0) == respectively. c special ombinations)

1,

Y\037

Y\037

1,

35.(a) To determine
solution
in

A I(X)
and)

o the radius of convergence f the series write the seriesof terms of even Example 5, in the form) degreein Eq.

(11)
==

YI

(x)

32/3r(\037))

(x) 3113 (
Y2

00
))

00
n=1)

\037

Yo(x)

1 + LC2nX2n == 1 + Lanzn
n=l

Bi(x)_

YI

(x)

Y2(X)

31/6r(\037)

3-1/6r(\037)

define the standard Airy functions that appear in mathematical tables and computer algebra systems. Their exhibit trigonometric-like osgraphs shown in Fig. cillatory behavior for x < 0, whereas Ai (x) decreases and Bi(x) increases exponentially as exponentially It is interesting to use a computer algebra x -+ system to investigate how many terms must be retained in the YI- and Y2 -series boveto produce figure that is visua a from Fig. (which is basedon ally indistinguishable to the Airy functions).) high-precision approximations

3.2.3

where an == C2n and z == x 2 Then apply the recurrence to show 3 relation in Eq.(8) and Theorem in Section in of that the radius of convergence the series z is 4. Hence in o the radius of convergencef the series x is How does T 1 this corroborate heorem in this section? (b) Write the series terms of odd degreein Eq. in the form) of

3.1 2.

(11)

+00.

YI(X)

=x

(1

+
\037C2n+,x2n)
that

=x
its radius

(1

+
\037bnzn))

3.2.3

to show similarly

in power series x) is also2.)

of convergence a (as

IIIJ RegularSingularPoints)
We now
\037)

l s investigate the solution of the homogeneousecond-orderinear equation)


A(x)y\"

+ B(x)y'+ C(x)y= 0)

(1))

near a singular point. Recall that if the functions A, B, and C are polynomials having no common factors, then the singular points of Eq. (1) are simply those points where A (x) = O.For instance, x = 0 is the only singular point of the Bessel equation of ordern,)

x2y\" + xy' + (x2

2 )y

= 0,

whereasthe Legendre equation of ordern,)

(1 x2)y\"

- 2xy' +

n(n

+ l)y = 0,)

has the two singular points x = and x = It turns out that someof the features of the solutionsof such equations of the most importancefor applicationsare largely determined by their behavior near their singular points.)))

-1

1.

3.3RegularSingularPoints 219)
We will restrict our attention to the case which x = 0 is a singular point in of Eq. (1). differential equation having x = a as a singular point is easily transA formed by the substitution t = x a into one having a correspondingingular point s at O.For example, et us substitute t = l into the Legendre equation of order n.

Because)

_ _ - dx dy

- ---, x-I
_
dy

dt

dy
dt)

dt dx

1/

d2

and 1

-x = 2 1

- we get d + -t(t + 2)\037- 2(t + 1)dt


(t

= d dY dx2 [dt (dx


y

)]

2 dt =dy dx dt 2
equation)

')

+ 1)2= -2t
2
y

t 2,

the

dy
dt)

n(n

+ l)y = O.

This new equation has the singular point t = 0 corresponding x = 1 in the to it has also the singular point t = to original equation; corresponding x =

-2

-1.)

P Typesof Singular oints


A

seriesolutions of the form y(x) = L cnxn , so the straightforward method of Secs tion 3.2 fails in this case. o investigatethe form that a solutionof such an equation T
might
in
\037)

differential equation having a singular point at 0 ordinarily

will

not have power

the standard form)

take, we assumethat Eq. (1)has analytic coefficient functions and rewrite


y\"

it

+ P(x)y' + Q(x)y = 0,)

(2))

where P = BIA and Q = CIA. Recall that x = 0 is an ordinary point (as opposed to a singular point) of Eq. (2) if the functions P(x) and Q(x) are analytic at x == that is, if P (x) and Q (x) have convergent power series xpansions powers in e of x on some open interval containing x = O. Now it can be proved that each of the functions P (x) and Q (x) either is analytic or approaches ::f::oos x --+ O. a x = 0 is a singular point of Eq. (2) provided that either P(x) or Consequently, ::f::oo s x -+ O. For instance, if we rewrite the Bessel a Q (x) (orboth) approaches of ordern in the form) equation

0;

1/

+ -y + X

1, ( - -) x
1
n2
2)

= 0,

x P(x) = 11and Q(x)= 1 (nlx)2both approach infinity as x --+O. seepresently that the powerseries ethodcan be generalizedto apply m near the singular point x = 0 of Eq. (2),provided that P (x) approachesnfinity no i more rapidly than 1Ix, and Q (x) no more rapidly than 1Ix2, as x -+ O. This is a way of saying that P (x) and Q (x) have only \"weak\" singularitiesat x = O.Tostate
we seethat
We will
this

more precisely, e rewrite Eq. (2)in the form) w


II

\037)

q(x) + p(x)y , + \037y _ 0, x x)


2 q(x) = x Q(x).)

(3))

where)
\037)

p(x) = xP(x)

and

(4))))

220

3 Methods) Chapter PowerSeries

RegularSingularPoint The singular point x = 0 of Eq. (3)is a regular singularpoint if the functions and q(x) are both analytic at x = O. Otherwise is an irregularingular it p(x) s
DEFINITION
point.) In particular, the singular point x = 0 is a regular singular point if p (x) and are both polynomials.For instance, we seethat x = 0 is a regular singular q(x) point of Bessel's equation of ordern by writing that equation in the form
/I

+ -y, + x

22 x-n = 0,
x)2
y

noting

1 and q (x) = x2 n 2 are both polynomials in By contrast, considerthe equation


that

p(x) =

x.

which
get)

3 2x y\" + (1+ x)y' + 3xy = 0,) has the singular point x = O.If we write this equation in
\"

the form of (3),we

+ (1+ x)j (2x x

Because) as x

0 (although singular point. We will not discussthe solution of differential equations near irregular singular points;this is a considerably more advanced topic than the solutionof differential equations near regular singular points.)
\037
\037

p(x) = 2x2 = 2x2 + 2x 00 q(x) = is a polynomial), we seethat x = 0 is an irregular


\037

l+x

-\037

),+ _ . y

2\"Y x)

Example

1)

Considerthe differential equation

x2(1+ x)y\" + x(4 x2)y' + (2 + 3x)y = O. In the standard form y\" + Py' + Qy = 0 it is
y
\"

+ x2 + y = O.) x(1+x) (1+x)


4
2 3x
y

-x
2

Because)

P(x) =
\037

-x

both approach 00 as x point. To determine the nature of this singular point we write the differential equation in the form of Eq. (3):)
\"

x(1+ x) 0,we seethat x = 0 is a singular

and

Q(x) = x2 + (1 x)

2 + 3x

+ + + (4 x2)j(1 x),+ (2+ 3x)j(1 x) y= 0. y 2 X


X)

Thus)

p\037)=

a Becausequotient of polynomials is analytic we seethat p(x) and q(x) are both analytic

wherever the denominatoris nonzero, x at x = O. Hence = 0 is a regular .))) of the given differential equation. singular point

l+x

-x

2
\037d

q\037)=

2 + 3x

l+x .

3.3RegularSingularPoints
form
in

221)

It may happen that when we beginwith a differential equation in the general Eq. (1)and rewrite it in the form in (3),the functions p(x) and q(x)as given in (4) are indeterminateforms at x = O. In this case the situation is determinedby the limits)
lim Po = p(O) = x---+o p(x) = lim
x---+o)

xP(x)

(5))

and)

qo =

lim q(O)= x---+o q(x) =

lim
x---+o)

x2Q(x).

(6))

If Po = 0
x2y\"

. .

= qo, then x = 0 may be an ordinary point + xp(x)y'+ q(x)y = 0 in (3).Otherwise:)

of the differential equation

If both the limits in (5) and (6) exist and are finite, then x = 0 is a regular singular point. If either limit fails to exist or is infinite, then x = 0 is an irregular singular
point.)

Remark:The most common case applications, the differential equain for


tion written in

the form)
II

q(x) + p(x)y , + -z-y _ 0, x x)

(3))

is that the functions p(x) and q(x) are polynomials. In this casePo = p(O)and qo = q(O)are simply the constant terms of these polynomials, so there is no need .) to evaluate the limits in Eqs. (5)and (6).

Example2)

To investigate the nature of the point


X y

x = 0 for the differential equation)

\"

+ (x2 sinx)y'+ (1 cosx)y= 0,)


in

we first write

it in

the form
II

(3):)

x)/ + (sinx x, + (1 y

- cosx)/ x
x)2

2
y

Then I'H6pital's gives the values) rule


lim Po = x---+o
and)

sIn x

= O.

= lim cosx =1)


x---+o)

lim qo = x---+o

- cosx=
X2

lim
x---+o

for the limits


Alternatively,

in (5)and (6). Since that x = 0 is not they are not both zero, we see an ordinary point. But both limits are finite, so the singular point x = 0 is regular.

1 sin x

2x

1 2

we couldwrite)
sin x 1

p(x) = \037= x x

x3 x x2 x - 3!+ Sf - ) = - 3!+ Sf (
5

...

...)))

222

3 Methods) Chapter PowerSeries


and)

q(X)

-[ - ( - - -- -+...] 2! ) - 2! 6! ----+--...
=
1

- cosx =
X2

1 1 X2

X2

+ 4!

6!)

x2

x4

4!

.)

moreover that Po = p(O) = 1 and qo x = 0 is a regular singular point.

These s (convergent)powerserieshow explicitly that p(x) and q(x) are analytic

= q(O) =

and

\037,

thereby verifying directly that

.)

TheMethod ofFrobenius)
We now

l approach the task of actually finding solutionsof a second-orderinear differential equation near the regular singular point x = O.The simplest uch equation s is the constant-coefficientequidimensional equation)
X y

+ poxy, + qoY

=) 0

(7))

to which

I Eq. (3)reduceswhen p(x) = Po and q(x) = qo are constants. n this case we can verify by direct substitution that the simplepowerfunction y(x) = xr is a solution of Eq. (7)if and only if r is a root of the quadratic equation)

r(r 1)+ por + qo = O.)


In the general case,n which i

(8))

stants, it

tion of the form)

r p(x) and q(x) are powerseriesather than conis a reasonable that our differential equation might have a soluconjecture 00 00

\037)

y(x) = x

L
n=O

cnx

Lcnx
n=O)

n +r

= coxr + CIXr+1 + C2Xr +2 + ...

(9))

and a powerseries. his turns out to be a very fruitful conT to Theorem 1 (soonto be stated formally), every equation of the jecture; according form in (1)having x = 0 as a regular singular point does,ndeed,have at leastone i such solution. This fact is the basis for the method of Frobenius, amed for the n German mathematician GeorgFrobenius who discovered method the in the An infinite series f the form in (9) is called a Frobenius o Note that a Frobeniusseriess generally not a powerseries. or instance, with r = -\037 the i F i seriesn (9)takesthe form)

-the product of xr

1870s.

(1848-1917),

series.

= cox-/ 2 + cIX 1/2 + C2x3/ 2 + C3x5/2 . +


1

..
;)

it

is not a series integral powersof x. in To investigate the possibleexistencef Frobenius seriesolutions, we begin o s
the equation)

with

\037)

x2yH + xp(x)y'+ q(x)y = 0)

(10))))

obtained by
point,
then

multiplying

p(x) and q(x) are analytic

the equation in at x =

3.3RegularSingularPoints 223) (3) by x2. If x = 0 is a regular singular


0,so)

p(x) = Po + PIX + P2x2 + P3x3 + . q(x) = qo + qlX + q2x2+ q3x3 + .


s Supposethat Eq. (10)has the Frobenius seriesolution)

.. ..
,
.)

(11))

\037)

= LcnXn +r .
n=O)

00

(12))

We may

m the (and always do) assumethat Co i= 0 because series ust have a first nonzero term. Termwise differentiation in Eq. (12)leadsto)

\037)

+r)xn +r -l y' = Lcn(n


00
n=O)

(13))

and)

\037)

y\"

L
n=O)

00

cn(n

+ r)(n + r l)xn +r -2.

(14))

Substitution

of the

+ [r(r l)coxr+ (r + l)rclxr+1 ...] + [pox + PIX2 + ...] . [rcoxr-l (r + I)CIXr + ...] + r + [qO+ q X + ...] . [cox+ C X r+ + ...] = O.
I I I

in series Eqs. (11) now yields) through (14)in Eq. (10)

(15))

then

initial terms of the two products on the left-hand side here and coefficientsof xr , we seethat the lowest-degree term in Eq. (15) is collecting r If is to be satisfiedidentically,then the coefficient co[r(r +por +qo]x Eq. (15) of this term (as well as those of the higher-degree terms) must vanish. But we are that Co i= 0,so it follows that r must satisfy the quadratic equation) assuming

Upon multiplying

-1)

\037)

r(r 1)+ por + qo = 0)

(16))

of precisely sameform as that obtained with the equidimensionalequationin (7). the of a (16)is calledthe indicial quation the differentialequation in (10), nd Equation e its two roots (possibly are the exponents f the differential equation (at the o equal) regular singular point x = 0). Our derivationof Eq. (16)showsthat if the Frobenius series = xr L cnxn is y to be a solution of the differential equation in (10), the exponent r must be one then of the roots rl and r2 of the indicial equation in (16). rl i= r2, it follows that there If are two possibleFrobenius series solutions, whereas if rl = r2 there is only one s possibleFrobenius seriesolution; the secondsolutioncannot be a Frobeniusseries. The exponents and r2 in the possibleFrobenius seriesolutions are determined s rl the indicial equation) by the values Po = p(O) and qo = q(O) that we have (using In discussed. practice,particularly when the coefficientsin the differential equation)))

224

3 Methods) Chapter PowerSeries


the original form in (1)are polynomials, the simplestway of finding Po and qo is often to write the equation in the form)
in

y\"

+ Po + PIX + P2x2 + X

...

y' +

qo

+ qlX + q2x2 + ...


x
2)

==

o.

(17)

Then inspection the of and qo.) Po

series appearin that

the two numeratorsreveals the constants

Example3)

Find the

in s exponents the possibleFrobenius seriesolutions of the equation)

2x2(1+ x)y\" + 3x(1+ x)3y' (1 x2)y


Solution
We

- -

== O.)

divide each term by

2 2x (1+ x) to recastthe differential equation in

the form)

y\"

2 _1(1 x) y 2 + 2x + x ) 2 + l(l X y' + x


2)

==

0,

and

thus

seethat r(r

Po == and qo ==
\037

\037.

Hence indicial equation is) the

- 1)+ - 1
\037r
\037

== r 2

+ 1r 1 == (r + l)(r 1)== 0,)


two

with then

roots rl == and r2 == of the forms)

T -1.he

s possibleFrobenius seriesolutions are

Yl

(x) == X 1/2

L
n=O)

00

anx

and

Y2(X) ==

x-I

L
n=O)

00

bnx

.)

Frobenius eriesSolutions) S
s and r2 are known, the coefficients in a Frobenius seriesoin are determined by substitution of the series Eqs. (12)through (14)in the differential equation, essentiallythe same method as was used to determine coefficients in power series 3.2.If the exponentsrl and r2 are solutions in Section then there always existtwo linearly independent Frobenius secomplexconjugates, ries solutions.We will restrict our attention here to the case which rl and r2 are in both real.We also will seek solutions only for x > O. Once such a solution has been found, we need only replacer1 with Ix Ir 1 to obtain a solution for x < o. x A The following theorem is proved in Chapter 4 of Coddington's n Introduction to Ordinary Differential Equations.)))

Oncethe exponentsrl

lution

3.3RegularSingularPoints
THEOREM 1

225)

Frobenius Seriesolutions) S
x2y\" + xp(x)y'+ q(x)y = o.)
(10))

Supposethat x = 0 is a regular singular point of the equation)


>-)

Let p

> 0 denote the minimum


p(x) =

of the radii of convergenceof the power series)


Pn xn

L
n=O)

00

and

q(x) = Lqnxn.
n=O)

00

Let rl and r2 be the (real) roots,with rl

Por + qo = O.Then (a) For x > 0,there existsa solution of Eq. (10)of the form)
>-)
rt Yl (x) = x

> r2, of the indicial equation r(r 1)+

L
n=O)

00

anx

(aO i= 0))

(18))

to corresponding the larger root rl . (b) If rl r2 is neither zero nor a positive integer, then there existsa second

linearly independent solution for x > 0 of the form)


Y2(X)

>-)

= xr2

L
n=O)

00

bnx

(bo i= 0))

(19))

t correspondingo the smaller root r2.

The radii of convergence of the powerseriesn Eqs. (18)and (19)are each at i least The coefficients in these series be determined by substituting the can i seriesn the differential equation)

p.

x2y\" + xp(x)Y'+ q(x)y = O.)


have already seenthat if rl = r2, then there can existonly one Frobenius seriesolution. It turns out that, if rl r2 is a positive integer, there mayor may s not exista secondFrobenius seriesolutionof the form in Eq. (19) s correspondingto the smallerroot r2.These in c a exceptional ases re discussed Section3.4. xamples E 4 through 6 illustrate the process determining the coefficientsin those Frobenius of seriesolutions that are guaranteed by Theorem 1.) s
We

Example4)
Solution

Find the Frobenius seriesolutions of) s


2x2y\"

+ 3xy' (x2 + l)y = O.)


put

(20))
in

First we divide each term by 2x2 to


y\"

3 + 1. + y' X

--- -x
1

the equation in the form 2 2


1

(17):)
(21))
\037

2
Y

2)

= O.

seethat x = 0 is a regular singular point, and that Po = and qo = Because = and q (x) = x2 are polynomials, the Frobeniusseries p(x)
We now
\037

-!

\037.

we)))

226

3 Methods) Chapter PowerSeries obtain will converge for all x > O.The indicial equation is)

r(r

- 1)+ \037r
\037

so the exponentsare rl and r2 not by an Theorem 1 guarantees the two linearly independent solutions.Rather than separately substituting)
YI

- (r + 1)= = = -1.hey do differ T integer, so existencef o Frobenius series


\037

= (r

\037)

0,)

= xl/2

L
n=O)

00

anx

and

Y2 = x-I

L
n=O)

00

bnx

recurrence relation for the seriesor Y2. f


When

Eq. (20),it is more efficient to beginby substituting Y = xr L cnxn . We will then get a recurrence relation that dependson r. With the value rl = it becomes a recurrenceelation for the series YI, whereaswith r2 = r it becomes a for
in

-1

\037

we substitute)
Y n = LcnXn +r , y' = L(n + r)cnx +r -l ,

00

00

\037)

n=O)

n=O)

and)

\037)

y\"

= L(n + r)(n + r
n=O)

00

- 1)cx
n

n +r

-2
get) Eq. (21)-we

in

Eq. (20)-the original differential equation, rather


00

than

n L(n + r)(n + r l)cxn +r + 3 L(n + r)cnxn +r n=O


n=O)

00

- Lc x - Lcnx
00
n n +r + 2

00

n +r

= O.

(22)

n=O

n=O)

A stagethere are several ways to proceed. goodstandard practiceis to shift indicesso that each exponent will be the sameas the smallestone present.In this example,we shift the index of summation in the third sum by 2 to reduceits
At this

exponent from
00

+ r + 2 to n + r. This gives)
00

L(n + r)(n +
n=O

- l)cx + + 3 L(n + r)cnx


n n r
n=O)

n +r

-L
00

C _2Xn +r n

n=2

- Lcnx
00
n=O)

n +r

= O.

(23)

The common range of summation is n > 2,so we must treat n = 0 and n = 1 t separately. ollowing our standard practice,the terms correspondingo n = 0 will F always give the indicial equation)

[2r(r 1)+ 3r l]co= 2 (r2 +

\037r

\037)

Co =

O.)))

3.3RegularSingularPoints
The terms corresponding n to

227)

1 yield)

[2(r+ l)r+ 3(r+ 1)


follows that)

- I]Cl = (2r + 5r + 2)CI =


2 Cl =
0)

o.)

Because coefficient2r2 + 5r + 2 of Cl is nonzero whether r = the

\037

or r

-1,
it

(24))

in

either case. The coefficientof xn+r 2(n + r)(n + r

in

- l)c + 3(n + r)c - -2n n

Eq. (23)is)

C n

C n

= o.)

We

solve for Cn and simplify to obtain the recurrence relation)


Cn

\037)

-2 2(n + r)2 + (n + r)
C n

for n
1)

> 2.)
r

(25))

We now write an in This gives the recurrence relation)


\037.

CASE1: rl =

of place Cn and substitute

=
\037

in

Eq.(25).

\037)

an

2n 2

a n -2

+ 3n)

for n

> 2.) 2,

(26))

With in

this formula we can determine the coefficients in the first Frobenius solution In view of Eq. (24)we seethat an = 0 whenever n is odd. With n = 4, and 6 Yl.

Eq. (26),we get)


a2

-- - - 616 ao

14')

a4

a2

ao

44

')

and

a6 =

a4

90 55,440
.)

ao

Hence first Frobenius solution is) the


Yl

(x) = aox 1/2

CASE2: r2 = We now write b n in place of C and n relation) (25).This gives the recurrence Eq.
\037)

-1.

x2 x x + 14 + 616 55,440)+ ... . +

-4
3n)

substitute

-1

in

bn

- 2b2
b2

bn 2

for n

> 2.)
n

(27))

Again,

Eq.(24)impliesthat bn = 0 for n odd. With


b2 =

= 2,4, and 6 in (27),we get)

- -, _ 20_
bo
2)

b4

bo

40')

an d

b6 _

- _ b4

bo

54

2160)

Hence secondFrobenius solution is) the


Y2(X)

= box1

x2 x 1+-+-+x +.... ( 2 40 2160) )


4 6

.)))

228

3 Methods) Chapter PowerSeries


\037 '-'

Example5)
Solution

Find a Frobenius solution of


\037)

Bessel's equation of orderzero,) 2 x2y\" + xy' + x y = O.)


'-\" '\" '\" \"...\037....'\" \",\037\037_..,\037 '\"\"\" _ \"\"No\"\" \"'..\",_ '\" '\" ....,',.,........'-'\037........

\"

,....,,'..'..n.'\"

\"\"'\"

...........

\"'\"

........

,,\"-\"-\"'''-'''

'\"

........

\"......\037

\037\"......

\037.... .....\037 .............

\".......

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

.....\037.......\"

(28))

In the form of (17), q. (28)becomes) E


II

+ -y + -y = O. X x
2)

1, x
with

1 and q(x) = x2, so our series = 0,the indicial equation is will 1 and qo r(r 1)+ r = r 2 = O.) Thus we obtain only the singleexponent r = 0,and so there is only one Frobenius seriesolution) s 00 n cnx y(x) = x

Hence = 0 is a regular singular point x

converge for all x > O. Because = Po

p(x) =

L
n=O)

of Eq. (28);it is in fact a powerseries. Thus we substitute y = L cnxn in (28);the result is)

L - l)cx + L
00 00

n(n

ncnx

n=O

n=O

Lc x + = O.
n n

00

n=O)

We

combinethe

first

two sums and

shift

the index of summation

in

the

third

by

-2

to

obtain)

L
n=O
\037)

00

2c xn
n

L
n=2)

00

C _2Xn n

= O.

The term corresponding xO gives 0 = 0:no information. The term corresponding to to x1 gives CI = 0,and the term for xn yieldsthe recurrence relation)
C n

-C n

-2

n 2)

for n

> 2.)
n

(29))

Because = 0,we seethat CI


6 in Eq. (29),we get)
C2 = Evidently,
\037)

Cn
C2

= 0 whenever n is odd. Substituting


Co

= 2,4, and
2
.)

-2

Co 2

')

C 4

-- 42
\037n=

22

.42

')

and

the pattern

is)

22 42

. ...(2n)2

co (-l)n

The choice o = 1 gives us one of the most important specialfunctions in mathC f ematics,the Besselunction of orderzeroof the first kind, denotedby Jo(x).
Thus)

=.
co (-l)n 22n (n !)2)
4
X

C6 =

-6 = -2 .4 .6
C4 2 2 Co 2

\037)

lo(x)=

00 \037

(_ 1)n X 2n = 22n (n!)2

x - 4 + 64 - 2304 + .
X

..
\302\267)

(30))

examplewe have not beenableto find a secondlinearly independent solution it of Bessel's W equation of orderzero. e will derive that solution in Section3.4; will .))) not be a Frobenius series.
In this

When r1
Recall

2 if rl then Theorem 1 guarantees only the that, r2 existencef the Frobeniusseriesolution correspondingo the larger exponent o t s 6 illustrates the fortunate casein which the series method nevertheless Example a secondFrobenius seriesolution. The case which the secondsolution is in s yields not a Frobenius series ill be discussed Section w in 3.4.)

- r Is an Integer - is a positive integer,

3.3RegularSingularPoints

229)

ri.

.....

Example6)

Find the Frobenius seriesolutions of) s xy


1/

+ 2y / + xy = 0
2

.)

(31))

Solution

In standard form the equation becomes)


// y

+ -y/ + -y = 0, X x
2)

x2

so we seethat x indicial equation)

o is a regular singular point

with

r(r 1)+ 2r = r(r + 1) = 0

Po)

2 and qo

O. The)

has roots rl = 0 and r2 = when rl which differ by an integer. In this case r2 is an integer, it is better to depart from the standard procedureof Example 4 and t beginour work with the smallerexponent.As you will see,he recurrencerelation will then tell us whether or not a secondFrobenius seriesolution exists. it does If s exist, our computations will simultaneouslyyield both Frobenius seriessolutions. If the secondsolution doesnot exist,we begin anew with the largerexponentr = rl to obtain the one Frobenius seriesolution guaranteed by Theorem s Hence begin by substituting) we

-1,

1.

= x-ILCn = LCn xn xn n=O)


00 00
n=O

in

This Eq. (31). gives)

L(n - 1)(n- 2)c x + 2 L(n - l)cx + LcnX = O.


00
n n

00

00

n=O

n=O

n=O)

We

combinethe first two sumsand shift the index by

L - l)cx - + L
00

n(n

00

Cn _2Xn

-2 the -2 = O.
in

third

to obtain)

(32))

n=O

n=2)

The cases n

= 0 and n = 1 reduceto)
o Co = 0 and

O.Cl = O.)

arbitrary constants Co and Cl and therefore can expectto find a solution incorporating two linearly independent Frobenius seriessolutions. general If, for n = we had obtained an equation such as 0 Cl = 3, which can be satisfied for no choicef Cl, this would have told us that no secondFrobenius seriessolution o

Hence have two we

1,

couldexist.)))

230

3 Methods) Chapter PowerSeries Now knowing


\037)

that

all C n

relation) is well,from (32)we read the recurrence

-2 =-n(nC n
,)

for n

> 2.)
1

1))

(33))

The first few values of n give


C2 = C4 = C6

-N
1 1
is) C2n

co

C3= 3 .2CI,
Co 1
')

--C2 4! 4.3) = --C4 =--' 6 .5 6!


Co

Cs
C7

= = --C3 5! 5 .4 = --C6 --. .6 = 7!


CI
')

CI

')

evidently the pattern


\037)

co (-l)n

for n

> 1. herefore, a general solution of Eq. (31)is) T


y(x) = x-I

(2n)!'

C2n+I

CI (-l)n (2n + I)!

Lcnx
n=O)

00

=
y)

Co

x
Co
\037

1_

x2

2!+ 4!
n X
2n

x4

_...+ x ( _ x3!+ x _... ) )


3 5
\037

5!)

=
Thus)

00
\037

(-1)

(2n)!

+
1

C1 00
\037

(-1)

2n

+1
.)

\037

(2n +

1)!

1)

FIGURE
Yl(X) in

= cosxand
x

- x)

We

have thus

4n)

Frobeniusseriesolutions) s
YI

cosx+ CI sIn x). y(x) = -(co x found a general solution expressed a linear combinationof the two as
(x) = x

3.3.1. solutions The


Y2(X)

cosx

and

Y2(X)

Slnx
x)

Example 6.)

but s one of theseFrobenius seriesolutions isbounded the Fig. in other is unboundednear the regular singular point x = common occurrence the case exponents of an integer. differing by

As indicated in

3.3.1,

-.
sIn x
x)

(34))

O-a

.)

Summary
When
\037)

confronted with a linear second-order differential equation)


A (x)y//

+ B(x)y'+ C(x)y= 0)

with

seriesolutions we first s
\037)

o analytic coefficient functions, in orderto investigate the possibleexistencef write the equation in the standard form)

If

P(x) and the equation has two linearly independent powerseries solutions.)))

+ P(x)y/+ Q(x)y = O.) Q(x) are both analytic at x = 0,then x = 0 is an ordinary point, and
y//

3.3RegularSingularPoints
x Otherwise, = 0 is a singular point, and we next write the differential
tion
\037)

231)
equa-

in

the form)
II

q(x) + p(x)y / + \037y _ 0. x x)

If p(x) and q(x) are both analytic at x = 0,then x = 0 is a regular singular point. In this case find the two exponents and r2 (assumed we real, and with rl > r2) by rl solving the indicial equation)
\037)

r(r 1)+ por + qo = 0,) = xrl

where Po = p(O) and qo


y

associatedwith the larger exponent rl, and if rl r2 is not an n o s integer, the existencef a secondFrobenius seriesolution Y2 = xr2 L bnx is also

L anx

= q(O). There always existsa Frobenius seriessolution

guaranteed.)

IEI1Proble\"ms
_)

In

Problems

nary point, a regular singular point, or an irregular singular point. If it is a regular singular point, find the exponents of the o. differential equation at x 3 x )y' + (sinx)y 0 xy\" + (x + x 2y' + (eX l)y 0 xy\" x 2y\" + + xy 0 4. 3x3 x2 0

1through

8, determine

whether

= 0 is an ordi-

2. 3.

1.

+ 2x2y'+ (1 )y = 5. x(1+ x)y\" + 2y' + 3xy = 0 2 6.x 2(1 x 2)y\" + 2xy' 2y = 0 7. x y\" + (6 sinx)y'+ 6y = 0 8. (6x2 + 2x3)y\" + 2lxy' + 9(x2 l)y = 0)
y\"

(cosx)y'

= =

+ 2y' + 2y = 0 21.x2y\" + xy' (1+ 2x2)y = 0 2 22.2x2y\" + xy' (3 2x2)y = 0 23.6x2y\" + 7xy' (x2 + 2)y = 0 24. 3x2y\" + 2xy' + x 2y = 0 25.2xy\" + (1+ x)y' + y = 0 26.2xy\" + (1 2X2)y' 4xy = 0)

19.xy\" 2 20.3xy\"

- y' -

=0

- -

l point of a second-orderinear difa transforms ferential equation, then the substitution t = x it into a differential equation having t = 0 as a singular point. We then attribute to the original equation at x = a the behavior of the new equation at t = O. Classify (as regular or irregular) the singular points of the differential equations in Problems 9 through 16. If x

= a i= 0 is a singular

9.(1- x)y\" + xy' + x 2y = 0 10. 1- X)2y\" + (2x - 2)y'+ y = 0 ( 11.1- x 2)y\" - 2xy' + 12y = 03 ( 12.x2- 2)3y\" + 3(x - 2)2y'+ x y = 0 ( 13.x2 - 4)y\" + (x - 2)y'+ (x + 2)y = 0 ( 14.x - 9)2y\" + (x2 + 9)y'+ (x2 + 4)y = 0 ( 15.x - 2)2y\" - (x2 4)y'+ (x + 2)y = 0 ( 16. 3(1- x)y\" + (3x + 2)y'+ xy = 0) x

> O. 27. xy\" + 2y' + 9xy = 0 28.xy\" + 2y' 4xy = 0 29.4xy\" + 8y' + xy = 0 30.xy\" y' + 4x3y = 0 31.x2y\" 4xy' + (3 4x2)y = 0) 4
graphs for x

Usethe method of Example 6 to find two linearly independent solutions ofthe differential equations in ProbFrobenius series Then construct a graph showing their lems 27 through

31.

Problems 32 through 34,find the first three nonzero terms solutions. of eachof two linearly independent Frobenius series 32.2x2y\" + x(x+ l)y' (2x + l)y = 0 33.(2x2 + 5x3)y\" + (3x x 2)y' (1+ x)y = 0 34. 2x2y\" + (sinx)y' (cosx)y= 0 35.Note that x = 0 is an irregular point of the equation)
In

solutions (for independent Frobenius series x > 0) of each of the differential equations in Problems through 26.
Find two linearly

x 2y\"

+ (3x

- l)y' +
y

= o.)

17

(a)

Show that y

17.4xy\" + 2y' + y = 0

18.xy\" 2

+ 3y' -

= 0)

only if r the \"formal\"

Substitute solution y n of this series?))) convergence

O. (b)

= x r L cnx n

can satisfy

n = L !x.

this equation to derive What is the radius of

= L cnx n

232

3 Methods) Chapter PowerSeries


A and

36. (a) Suppose that


that the equation

siblevalues of r.) 37. (a) Usethe method of Frobenius to derive the solution 3 YI = x of the equation x y\" xy' + y = O. (b) Verify l substitution the secondsolution Y2 = xe- /x Does Y2 by

solution of the form y x r cnx n (b) Repeatpart (a) 3 with the equation x y\" + Axy' + By O. (c) Show that the equation x 3y\" + Ax2y' + By 0 has no Frobenius seriessolution. (Suggestion: In each casesubstitute x r cnx n in the given equation to determine the posy

x 2y\"

= L

+ Ay' + By = 0 has at

B are nonzero constants.

Show

with
IS)

most

one

Po = qo

= 1.(b)
C n

Show that the

recursion formula

= =

Cn-l
n2 +

2rn

= L

Apply this formula with r to obtain qn

r = i to obtain Pn = Cn , then with = Cn. Concludethat Pn and qn are complexconjugates:Pn = an + ibn and qn = an ibn ,

-i

38.

have a Frobenius series representation? Apply the method of Frobenius to Bessel's equation

where the numbers {an} and {b } are real. (c) Deduce n from part (b) that the differential equation given in this problem has real-valuedsolutions of the form)
YI

of

(x) = A(x)

order 4,)
to derive its

Y2(X)

= A(x) sin(lnx) + R(x) cos(lnx),) = Lbnxn .

cos(lnx) B(x)sin(lnx),

x 2y\"

+xy'+ (X2 -\037) Y = 0,


> 0,)
,JX

general solution for x y(x) ,JX

41.
solu-

n and B(x) where A(x) t Considerhe differential equation)

= Lanx

= Cocosx+ CISlnx .
graphs

+ + x(x 1)(x 1)2y\" + 2x(x 3)(x l)y' 2(x


that

- - l)y =

0)

Figure tions.)

3.3.2 the shows


Y)

of the

two indicated

appearedin
in the

program

an advertisement for a symbolic algebra March 1984issueof the American Math-

ematical Monthly.
singular follows from
YI)

(a)

Show that

Theorem1 that this differential a power seriessolution of the form)


YI

point with

exponents rl

x = 0 is a regular = 1 and r2 = O. (b) It


equation has

1)

x)

= x + C2X2+ C3x3 + . in Substitute this series(with CI = 1) the differential tion to show that C2 = -2, 3 = 3, and) C
(x)
.)

..

equa-

FIGURE3.3.2. solutions The


Yl

(x)
that

= cosxand

Problem

38.
y\"

,JX

Y2(X)

Slnx
in

C +2 =) n (n
2

,JX

- n)cn-l + - 5n - 2)c (n
2
n

(n

+ 7n + 4)C+1 n
in part

(n
2

+ 1) + 2)) (n

39.a) Show (
Frobenius

Bessel'squation of order 1, e
x
2

+ xy' + (x - l)y =
r2

0,) that the

for n > to prove by

2.

relation (c) Usethe recurrence

has exponents

series correspondingto

rl

= 1 and
x

= -1at x = 0, and rl = 1 is
')

n for n that C n that) Hencededuce(using the geometric series) induction


YI

= (-I)n+1
X

1 ::::(!).

(b)

(x) =

= J1(x) 2

x (_1)n2n 2n n!(n + 1)!2 \037

00

(1+ X)2)
to

for 0 < x < 1. 42. This problem is a brief introduction

(b) Show that there is no Frobenius solution correspondthat is, show that it ing to the smaller exponent r2 is impossibleto determine the coefficients in)

= -1;
00

metric equation)

Gauss's ypergeoh

Y2(X)

= X-I LCnxn.
n=O)

t 40. Considerhe

Show that its exponents are solutions) Frobenius series


y+

equation

x2

y\"

so :l:i, it has complex-valued


y_

+ xy' + (1 x)y = O. (a)


00

x(1 x)y\" + [y (a + fJ + l)x]y' afJy = 0, (35)) where a, fJ, and yare constants.This famous equation has in mathematics and physics. wide-ranging applications (a) Show that x = 0 is a regular singular point ofEq. (35),
exponents 0 and 1 y. (b) If y is not zeroor a negative integer, it follows (why?) that Eq.(35)has a power seriessolution)
with

=x i\037 n \037Pnx
n=O)

00

and

= x-i Lqnxn
n=O)

y(x)

=X

L
n=O

00

cnx

L
n=O)))

00

cnx

3.4Methodof Frobenius: Exceptional The Cases 233)


with

senes

Co i=
IS)

O. Show
C n+1

that the

recurrence relation
n)(fJ n) C n n) Co
==

for this

for n > O. (c) Conclude that part (b) is) y(x)


==

+ - (a + n)(1+ +
(y
with

for n where an == + + 2) . (a + n in and fJn and Yn are defined similarly. (d) Theseries (36) is known as the hypergeometric seriesand is commonly denoted by Show that fJ, y,

a(a 1)(a

..

- 1)

\037

1,

1 the

in series

1+

L n!
n=O

00

a fJ \037xn
Yn)

(36))

1 F(I,1,1, ) I-x (the geometric series); x 1 -x)2 In(1 +x); (ii) xF(I, ,2, (iii) x F (4, 1, -x ) tan x; 1 1(iv) F(-k, , , x) (l+x)k (the binomial series).) (i)
== == ==
I
\037,

F(a,

x).

==

Methodof Frobenius: he Exceptional T Cases)


We

continue our discussion the of


// y

equation)

\037)

+ p(x)y , + q(x) = 0 X x
2)

-y

(1))

where p(x) and q (x) are analytic at x = the roots rl and r2 of the indicial equation)
\037)

0,and x = 0 is a regular singular point. If


0)

cp(r) = r(r

+ -1) por +qo=

(2))

do not differ by an integer, then Theorem 1 of Section3.3 guarantees that Eq. (1) has two linearly independent Frobenius seriesolutions. e consider now the more s W complexsituation in which rl r2 is an integer. If rl = r2, then there is only one s exponent available, and thus there can be only one Frobenius seriesolution. But we saw in Example6 of Section that if rl = r2 + N, with N a positive integer, 3.3 then it is possiblethat a secondFrobenius seriesolution exists. e will also see W s that it is possiblethat such a solution does not exist.In fact, the secondsolution involves In x when it is not a Frobenius series. you will seein Examples and 4, As 3 theseexceptional ases ccur the solution of Bessel's c o in equation. For applications, this is the most important second-orderinear differential equation with variable l

coefficients.)

The Nonlogarithmic with Case


In

rl = r2 + N
the power series
00

Section we derived the indicial equation by substituting 3.3 (x) = L Pn xn and q (x) = L qnxn and the Frobenius series) p
\037)

r y(x) = x

Lcnx Lcnx
n

00

n +r

(co i= 0))

(3))

n=O

n=O)

in
\037)

the differential equation in the form)

x2y// + xp(x)y'+ q(x)y = O.)


substitution,

(4))

The result of this

x, is an equation of the form)

after collectionf the coefficientsof like powers of o

LFn(r)x
n=O)

00

n +r

=0

(5))))

234

3 Methods) Chapter PowerSeries


in

which the coefficientsdependon

Fo(r)= [r(r 1)+ por + qo]co= cp(r)co,)


which gives the indicial equation because i= 0 by assumption; also, for Co the coefficientof xn+r has the form)
n

r. It turns

out

that

the coefficientof xr

is)

(6))

>

1,
(7))

Fn(r)= cp(r + n)cn + Ln(r;Co,

CI,..., Cn-I).)

HereLn is a certain linear combination of Co, CI, Cn-I. Although formula is not necessary our purposes,t happensthat) for i

...,

the exact

+ Ln = L[(r k)Pn-k+ qn-k]Ck.


k=O)

n-I

(8))

Because the coefficientsin (5)must vanish for the Frobenius serieso be a all t solution of Eq. (4),it follows that the exponent r and the coefficientsCo,CI, , C n must satisfy the equation)

...

cp(r

+ n)cn + Ln(r;Co,CI,..., = O.) Cn-I)

(9))

This is a recurrence relation for Cn in terms of Co,CI, , Cn-I. Now supposethat rl = r2 + N with N a positive integer. If we use the larger exponent rl in Eq. (9),then the coefficient (ri + n) of Cn will be nonzero for every n > 1 because (r) = 0 only when r = rl and when r = r2 < ri. Once Co, CI, , Cn-I have been determined, we therefore can solve Eq. (9)for C and continue n to compute successive to coefficientsin the Frobenius seriesolution corresponding s the exponent rl But when we use the smallerexponent r2, there is a potential difficulty in computing CN. For in this case (r2 + N) = 0,so Eq. (9)becomes)
cp

...

...

cp

cp

O.CN + LN(r2;Co,CI,..., = o.) CN-I)


At this

(10))
it

stageCo,

CI,..., have already beendetermined.If CN-I


LN(r2;Co,CI,

...,

happensthat)

CN-I)= 0,)

then we can chooseN C

cientsin a secondFrobeniusseriesolution. But if it happensthat) s


L N (r2; Co,CI,
then

arbitrarily

and continue to determine the remaining coeffi-

..., CN-I)i= 0,)

there cannot exista o Eq. (10)is not satisfied with any choicef CN; in this case secondFrobenius seriesolution correspondingo the smaller exponent r2. Exams t ples 1 and 2 illustrate thesetwo possibilities.)
....

Example

1)

Considerthe equation)

x2y// + (6x+ x2)y' + xy = O.) Here = 6 and qo = 0,so the indicial equation is) Po
cp(r) = r(r

(11))

- 1)+ 6r =

r2

+ 5r = 0)

(12))))

The Cases 235) 3.4Methodof Frobenius: Exceptional


with roots rl = 0 and r2 = the roots differ by the integer N r the Frobenius series = cnxn + and get) y

-5;

= 5.We substitute

r n L(n + r)(n + r l)cxn +r + 6 L(n + r)cnxn +


n=O
n=O)

00

00

n + L(n + r)cnx +r + +
1

00

n=O

Lcnx + + = o.
n r
1

00

n=O)

When

we combine first two and alsothe last two sums,and the index by the result is)

-1,
00
n=O

in

the latter

shift the

r n L[(n+ r)2 + 5(n+ r)]cnx+r + L(n + r)cn_lXn + = o.


n=l)

00

The terms corresponding to n > 1 we get the equation)

0 give the indicial equation in (12),whereas for

n [en + r)2 + 5(n+ r)]c

+ (n + r)cn-l = 0,)

(13))

which in this examplecorresponds the general solution in (9). Note that the to coefficientof Cn is (n + r). We now follow the recommendation in Section for the case = r2 + N: 3.3 rl We begin with the smaller root r2 = With r2 = Eq. (13)reducesto)
c/J
\037)

-5, n(n - 5)c + (n - 5)C -l = o.)


n n

-5.

(14))

If n i= 5,we can solve this equation for Cn to obtain the recurrence relation)
\037)

C n

-Cn-l
n)
')

for n i= 5.)

(15))

This yields)
Cl =
C3 =

-co,)

C2 =

-3 = -6
C2

Co

and)

C4

-2 = 2') - -- - 24.) 4
Cl Co
C3

Co

(16))

case = r2 + N, it is always the coefficientCN that requiresspecialconsidrl eration. Here = 5, and for n = 5 Eq. (14)takesthe form 0 Cs + 0 = O. Hence N is a secondarbitrary constant, and we can compute additional coefficients,still Cs using the recursion formula in (15):)
In the

C6 = and

-6')

Cs

C7 =

-7= 6 .

C6

Cs 7

')

Cg

-8 = -6 .
C7

Cs 7 8

')

(17))

so on.)))

236

3 Methods) Chapter PowerSeries


When

we combinethe resultsin (16)and (17),we get)


5 Y = x-

Lcnx = cQx- - x + x2 - x + x 2: 6 24) ( x6 x - -+ 6.7- 6.7.8) ) + +... ( 6


n
n=O)

00

5 1 5

X C5

in

Frobeniusseriesolutions) s

terms of the two

arbitrary

constants Co and C5. Thus we have found the

two

YI(X)

= x-5

x - x + x2 - x6 +-) 2 (
3

--

24)

and)
Y2

(x) =

L 6. ...
n=l

00

x (-I)nn
(n

+ 5)

00 xn = 1 + 120 (-I)n

n=l (n

+ 5)!)

of Eq. (11).)
.,.

.)
x2y//

Example2)

Determine whether or not the equation)

- xy' + (x - 8)y =
2 r2

0)

(18))

has two linearly independent Frobenius seriesolutions.) s

Solution Here = Po

-1and qo = -8,so the indicial equation


cjJ(r)

is) 0) substitution

= r(r 1)

with

roots rl = 4 and r2 we L cnxn +r in Eq. (18), get)


00

- - - 8 = - 2r - 8 = = 6. On = -2
r differing by N

of

n L(n + r)(n + r l)cxn +r L(n + r)cnxn +r n=O


n=O)

00

+
If we shift the index by
get) 00 in

L
n=O n

00

cn xn +r +2

- 8 Lcnx + = o.
00
n r
n=O)

-2 combine L[(n+ r)2 - 2(n+ r) - 8]cx + + L


the
third

sum and

the other three sums,we

00

n r

Cn_2Xn+r

= o.

n=O

n=2)

The coefficientof xr gives the indicial equation, and the coefficientof xr+l gives)

[(r + 1)2 2(r + 1)

- 8]

CI =

o.)))

3.4Methodof Frobenius: Exceptional The Cases 237)


Because coefficientof CI is nonzero both for r = 4 and for r = -2,it follows the that CI = 0 in each case. or n > 2 we get the equation) F
\037)

[(n + r)2 2(n+ r)


i correspondsn
c/J

- 8]
C n

Cn

+ Cn -2= 0,)
the

(19))

which

ficient of Cn is (n

We work first with the


\037)

+ r).

this

exampleto the general equation in (9);note that


smaller root r
n(n
n

coef-

- 6)c +
n

= r2 = -2. Then Eg.(19)becomes)

-2= 0)
> 2, n i= 6).)

(20))

for n
\037)

> 2.For n

i= 6 we can solve for the Cn

recurrence relation)
(n (21))

-2 =-n(nC -6))

Because = 0,this CI

formula gives)
C2 =

C Now

- C2 Co 4- 8 - 64
.

8'

Co

C3 = 0,)
and)

,)

Cs = o.)

Eq. (20)with

= 6 reducesto)
o C6+

- o.
Co = 64) (n

there is no way to choose so that this C6 by assumption, and hence solution corresponding the to equation holds. Thus there is no Frobenius series smallerroot r2 = 2. To find the singleFrobenius seriesolution correspondingo the larger root s t relation) rl = 4, we substitute r = 4 in Eq. (19)to obtain the recurrence
But Co i= 0

\037)

C n

-2 =-n(nC +
n

6))

> 2).)

(22))

This gives)
C2 =

-2 . 8 '
Co

C4 =

-4 . 10= 2 .4.8 .
C2

The general pattern

is)

\037=2 .4 ...

. (2n). 8 . 10..(2n + 6)
+6
00

co (-I)n

This yieldsthe Frobenius seriesolution) s


4 Yl (x) = x 1

=.
Co 10)

(-I)n6co

2 n!(n
2n

+ 3)!)

\037

( _ 1) X 22nn!(n 3)!)
n
2n

)
.)))

of Eq. (18).)

238

3 Methods) Chapter PowerSeries Reductionf Order) o


When only a singleFrobeniusseries solution exists, we need an additional technique. We discusshere the method of reductionof order,which enablesus to l use one known solution YI of a second-order homogeneousinear differential equa-

tion to

find

a secondlinearly independent solution Y2. Considerthe second-order


yl!

equation)
\037)

+ P(x)y' + Q(x)y == 0)

(23))

that on an openinterval I on which P and Q are continuous. Suppose we know one solution YI of Eq. (23).By Theorem 2 of Section there existsa secondlinearly independent solution Y2; our problemis to find Y2. Equivalently, we would like to

2.1,

find

the

quotient)

veX)

Y2(X)
YI

(x))

(24))

Oncewe know vex), Y2 will

then

be given by)

Y2(X) == V(X)Yl

(x).)

(25))
the

We begin by substituting derivatives) == vY\037 v' Yl


Y\037

the and

expressionin (25) in Eq. (23), using


Y\037

== vY\037'

+ 2v ' + v// Yl.)


Y\037

We get)
[vy\037'

+ 2v ' + vI!YI] + P
Y\037

[vy\037

+ v' YI] + QVYI


\"

==

0,
.)

and rearrangement gives)


V [YI

I!

+ PYI' + QYI ] + v // YI + 2v YI + Pv 'YI


this

== 0

But the

of Eq. (23).This leaves the equation)


vI!YI

bracketedexpression in

last equation vanishes becauseYI is a solution

(2y\037

+ PYI)V ' == O.)

(26))

The key to the


substitution in

method is that Eq. (26)is linearin v'. Thus the has reducedthe second-order linear equation in (23)to the first(25) order(in v') linear equation in (26).If we write u == v' and assumethat YI (x) never vanishes on then Eq. (26)yields)

successf this o

I,

u' +
An integrating

(2

+
P(X\302\273)

= O.)

(27))

\037:

factor for Eq. (27)is)

p=exp(1 (2
thus)

\037:

+P(X\302\273)dX)

P(X)dX);) =exP(21nIYll+

p(X) =

Y\037

exp

(IP(x)dx)

.)))

3.4Methodof Frobenius: Exceptional The Cases 239)


We now

integrate the equation in (27)to obtain)

uy;exp

(IP(X)dX)= C,
Y2 Yl

so

Vi

=U=
\037

exp

(-IP(X)dX).)

Another integration now gives)

=V=C

f
=

f exp(- P(x)dx)dx + K.
Y;)

With the particular choices C formula)

1 and K

0 we get the reduction-of-order

\037)

Y2

= Yl

exp

(-f p(x)dx) dx. 2


Yl)

(28))

This formula provides a secondsolution Y2(X) of Eq. (23)on any interval where N a never vanishes,Y2 (x) Yl (x) is never zero. ote that becausen exponentialfunction is a nonconstant multiple of YI (x),so Yl and Y2 are linearly independentsolutions.)

Example3)

For an elementaryapplication of the reduction-of-orderformula considerhe differt ential equation x2y\" 9xy' + 25y = O.

= 0 has the powerfunction y(x) = xr as a solution if and only if r is a root of the quadratic equation r 2 + (Po l)r + qo = O. Here = -9and qo = 25, so our Po 2 lOr +25 = (r 5)2 = 0 and has the single(repeated)root quadraticequation is r r = 5. This gives the singlepowerfunction solution YI (x) = x5 of our differential
qoY

In

Section we mentioned that 3.3

- -

the equidimensionalequation x2y\"

+ Poxy' +

Beforewe can apply the reduction-of-orderformula to find a secondsolution, we must first divide the equation x2y\" 9xy' + 25y = 0 by its leading coefficient x2 to get the standard form) 9 25 II y -yI + = 0

equation.

-y x
2)

dent solution)

in Eq. (23)with leading coefficient 1.Thus we have P(x) = -9/xand Q(x) 2 25/x , so the reduction-of-orderformula in (28)yieldsthe secondlinearly indepen==

Y2

(x) = x5

f 5 x=x f

exp

(X\037)2

(-f - dx) dx
\037

IO

d exp (91nx) x = x

x- x9 dx = x5 1nx)
lO

for x > O. Thus our particular equidimensionalequation has the solutions Yl (x) = x5 and Y2 (x) = x5 In x for x > O.

two

independent

.)

Similar applications of the reduction-of-orderformula can be found in Problems37-44of Section we 2.2-where introduced the method of reductionof order
in

Problem 36 (though

without

deriving there the reduction-of-orderformula

itself).)))

240

3 Methods) Chapter PowerSeries

The Logarithmic Cases


We now

investigate the general form of the secondsolution of the


y

equation)

/!

\037)

q(x) + p(x)y , + -z-Y_ 0, x) x

(1))

under the assumption that its exponents and '2 == N differ by the integer N > O. We assumethat we have already found the Frobenius series solution)
\037)

'1

'10))

Yl

(x) == xrl

L
n=O)

00

anx

(aO

\037

(29))

for x
and

2 Q(x)for q(x)/x . Thus we can rewrite Eq. (1)in of Eq. (23).

> 0 corresponding the larger exponent '1.Let us write P(x) for p(x)/x to the form y\" + Py' + Qy 0
==

Because indicial equation has roots '1and '2== ,} N, it can be factored the r2 + (Po

easily:)

- l)r+

qo ==

(, - '1)(' '1+ N) == ,2+ (N - 2'1)' + 1 == N

(,\037

- 'IN)

== 0,)

so we seethat)
Po
that is,)

- 2'1;)
N.)
in

-Po - 2'1 -1==

(30))

In preparation for use of the reduction of orderformula

P(x) == Po
Then)

+ PIX + P2X 2
X

+...- +
== Po
X)

(28),we write)

PI

+ P2X + . .

..

exp

= ( f P(x)dx) exp (-f [:0+ PI + P2 + ]dx)) == exp Po (- - PIX - 1P2x2 - ... == x-po exp (-PIX -1P2x2 ...)
X

...
)) ,)

In X

so that) exp

( f

P(x)dx = x-po(1+ Alx + A 2x2 + ...) .

(31))

In the last step we have used the fact that a composition of analytic functions is analytic and therefore has a power seriesepresentation; the initial coefficientof r that series (31)is 1 because O== in e We now substitute (29)and (31)in (28);with the choiceo == 1 in (29),this a

1.

yields)
Y2

== Yl

x-po(1+ A x + A 2x2 + . ) 2 dx. X 2rl (1 a X + a2x2 + ... +


1 1
))))

..

3.4Methodof Frobenius: Exceptional The Cases 241)


We

expandthe denominator and simplify:


Y2

= Yl = Yl

f f

x-Po-2rl + A1x + A2 X2 + (1 1 + B1x B2x2 + ... +


1

...
)

dx
(32))

x- -N (1+ C1x C2X 2 + ...) dx +

of

the quotient seriess i N We now considerseparately the cases = 0 and N O. We want to ascertain the general form of Y2 without keepingtrack of specificcoefficients.)

we (Here have substituted (30)and indicated the result of carrying out long division series illustrated in Fig.3.1.1, in particular that the constant term of as noting

1.)

>

CASE

1: QUALEXPONENTS(rl = r2). E
Y2

With

= 0,Eq. (32)gives

= Yl

f(

+ C + C2x2 + ... dx
1

\037

= y1lnx+ Yl (C1x 4 C2X2 + = y1lnx+ xrl (1+ alX + (C1x 4C2X2 = Yl In x + xrt (box + b x2 + b2x3 + ...) . t of Consequently,in the case equal exponents,he general form of Y2 is)
1
\037)

+...) +...)
00

+...)
(33))

Y2(X) it

n = Yl(x)lnx+xl+rlL:bnx.
n=O)

Note the logarithmic term;

is always presentwhen

'1= '2.)
+ N).
\302\267 \302\267

CASE2: POSITIVE INTEGRALDIFFERENCE(rl = r2 Eq. (32)gives


Y2

With

> 0,

= Yl = Yl

x- -N (1+ C1x C2x2 + + CNX N + . + f

f(

- - ..
1
\302\267

\302\267

\302\267

) dx)

CN
X

Cl + xN+l + x + .
1
N)

x-N =CNy1lnx+Yl -N (
r =CN y lnx+x2+N
1

-+ -N +
OO

) dx
1)

CIX -N+l

+... )
--+-C1x +... + ) (
1 N N 1
')

(L: )
n=O

ax xn n

so that)
\037)

Y2(X)

= CNYI (x)lnx + xr2 L:bnx n ,


n=O)

00

(34))

of i=- O.This gives the general form of Y2 in the case exponents a positive integer. Note the coefficient CN that appearsin (34)but not in (33). happensthat CN = 0,then there is no logarithmic term; if so, Eq. (1) has a secondFrobenius seriesolution (as in Example s
differing
by If it
1).)))

where bo =

N -aol

242

3 Methods) Chapter PowerSeries


In our derivationof Eqs. (33)and (34)-whichexhibit the general form of the have secondsolution in the cases = '2and '1 '2 = N > 0,respectively-we '1 that said nothing about the radii of convergenceof the various powerseries appear. Theorem 1 (next) is a summation of the precedingdiscussion also tells where and the series (33)and (34)converge. As in Theorem 1 of Section we restrict our in 3.3,
attention

to solutions for x

> o.)

THEOREM 1

Supposethat

The Exceptional Cases x = 0 is a regular singular point of the equation)


x2y\"

+ xp(x)y'+ q(x)y = O.)


00

(4))

Let p

> 0 denotethe minimum


00
n=O)

of the radii of convergenceof the powerseries)


and

p(x) = LPnxn

q(x) = Lqnxn.
n=O)

Let'land r2 be the roots,with rl > r2, of the indicial equation)


r(r 1)+ Por + qo = O.)
(a) If rl = r2, then Eq. (4) has two solutions Yl
\037)

and

Y2

of the forms)
(35a))

r Yl (x) = x

L
n=O)

00

anx

(aO 0))
\037

and)

\037)

Y2(X)

= Yl (x)In x + xq+l

L
n=O)

00

bnx

(35b))

(b) If rl

r2 = the forms)

N, a positive integer, then Eq.(4) has two solutions Yl and Y2 of

\037)

Yl

(x) = xr

L
n=O)

00

anx

(aO 0))
\037

(36a))

and)
n = CY1(X)In x +xr2 Lbnx .
n=O)

00

\037)

Y2(X)

(36b))

In Eq. (36b), o b 0 but C may be either zero or nonzero, so the logarithmic term mayor may not actually bepresentin this case. he radii of convergenceof T the powerseriesf this theorem are all at least The coefficientsin theseseries o (and the constant C in Eq. (36b)) ay be determined by direct substitution of the m i seriesn the differential equation in (4).)))
\037

p.

3.4Methodof Frobenius: Exceptional The Cases 243)


Example4)
.
We will illustrate the
tion

of orderzero,)

= case'l r2 by

....

deriving the

secondsolutionof Bessel's equa(37))

x2y !I + xy' + x2y = 0,)


for which

3.3 '1= '2= O. In Example5 of Section we found


Yl(X)

the

first solution)

= Jo(x)=

00 \037

(_ 1)n X 2n
22n (n!)2)

(38))

the According to Eq. (35b) secondsolution will have the form)


Y2

= Yl lnx + Lbnxn .
n=1)

00

(39))

The first two derivativesof Y2 are)


, Y2=Yl I
and)

'\"' nx+-+ b
Yl X

00

\037n
n=1)

nX

n-l

Y2

00 !! !!nx + 2 ' - + '\"' (n - l)b = I 2\" \037n


Yl Yl Yl X X
n=2)

nX

n-2

We substitute

thesein Eq. (37)and use the fact that Jo(x)alsosatisfiesthis equation


O= X

to

obtain)

2!/+
Y2

= (x2y\037'
00

+ x2Y2 + + X 2Yl) In x +
XY2
XY\037

2xy\037)

+ Ln(n l)bn xn + Lnbnxn +


n=2 n=1

00

L
n=l)

00

bn xn +2,

and

it

follows that) o = 2 \037


n=1

'\"'(-l) 2nx22n + b x + 22b2x2 + '\"' 2b + b -2)x . \037(n 22n (n' .)


00
n

00

(40))

n=3)

The only term involving x in Eq. (40) is b 1 x, so b 1 = O. But n 2bn + bn 2 = 0 if n is odd, and it follows that all the coefficientsof odd subscriptn Y2 vanish. i Now we examine the coefficients with even subscripts Eq. (40).First we in

seethat)

b2 =

-2. 22(-1)(2) .22 . (1 !)2)

1
4)

(41))

For n

> 2,we read the recurrence relation)


2 (2n) b2n

n + b2n-2 = (2)( 1) (2n) 2


2n

- -

(n!)2)

(42))))

244

3 Methods) Chapter PowerSeries


from

cients)on the

are typical of the exceptional ases the method of Frobenius, nd their solution c of a often requiresa bit of ingenuity. The usual strategy dependson detecting the most of o conspicuous ependence b 2n on n. We note the presencef 22n (n !)2on the rightd hand side in (42);in conjunction with the coefficient (2n)2 on the left-hand side, we are inducedto think of b2n as something divided by 22n (n !)2. oting also the N alternation of sign,we make the substitution)
b 2n
in

t (40). Note the \"nonhomogeneous\"erm (not involving the unknown coeffiside in (42).Such nonhomogeneous recurrence relations right-hand

- _I
(

22n (n

) n+lC2n !)2

')

(43))

the expectation that the recurrence relation for C2n will be simplerthan the one for b 2n We chose(_I)n+l rather than (-I)n b because 2 == > 0;with n == 1 in == Substitution of (43)in (42)gives) (43),we get C2

1.

\037

(2n) which

2 (_1)n+lC2n

22n (n!)2

22n-2[(n

2n-2 (-ltc_

(-2)(-2)n(2n)
22n (n !)2)

1)!]2)

boilsdown to the extremely simplerecurrence relation)


C2n == C2n-2

+
1

-.
1
n)

Thus)

C4 == C2 C6 ==c 4
Cg == C6

+ - == 1 + -, 2 2
1 1 1 1 +-==1+-+3 2 3'

+4

and

so on. Evidently,)
C2n

==1+-+-+... + -==H 2 3
n

111
== 1

+2 +3 +4

111
')

n,)

(44))

where we denoteby

Hn the nth partial sum of the harmonic a Finally, keepingin mind that the coefficientsof odd subscript re all substitute (43)and (44)in to obtain the secondsolution)

series (I/n). L

zero, we

(39)

Y2(X) ==

'\"'(_I)n+lHnx2n lo(x)In x + \037 22n n! 2 ( ) n=l)


x+

00

== lo(x)In

-- x2
4
4 3x

+ 128 13824)

6 l1x

...

(45))

of Bessel's in T equation of orderzero. he powerseries (45)convergesfor all x. The most commonly usedlinearly independent [of lo(x)] secondsolution is)
Yo

+ (x) == -(y -ln2)Yl -Y2; J[


J[)))

3.4Methodof Frobenius: Exceptional The Cases 245)


that is,)

Yo (x) ==

H - ( +In-) lo(x)+\"(-I)n+1
2 J[
X

00

nx

2n

\037

22n(n!))2

'

(46))

where y denotesEuler'sonstant:) c
y == lim (Hn
n\037oo)

- Inn)

\037

0.57722.

(47))

This particular combination Yo (x) is chosenbecause its nice behavior as x --+ of i kind. function of orderzeroof the second +00;t is calledthe Bessel

.)

Example5)

As an alternative to the method of substitution, by employing the technique of reduction of

Bessel's equation of order 1,)


the

we illustrate the case rl r2 == N order to derive a secondsolution of


== 0;)

x2y\"

+ xy' + (x2

- l)y
X

(48))

associatedindicial equation has roots rl == 1 and r2 Problem 39 of Section one solution of Eq. (48)is) 3.3,
Yl(X)

-1. According
X
7

to)

= J (X) = x
1

00 \037

22nn! (n

-1)
n

\037n

+ I)!

=X

- 16+ 384 - 18432 +.... 2


3
X

(49))

With

P(x) == l/xfrom (48),the reduction of orderformula


== YI

in

(28)yields)

Y2

\037
xYI)

dx
1
3 5
7

== Yl

dx -Yl f x3 (I-x/8+x4/192-x/9216+...)2) 2 6
== 4YI

dx + f x(x/2- x /16+ x /384 x /18432...)2) 4

+ f x (1- x /4 + 5x /192 7x /4608 ... dx

== 4Yl

- -f (
3

))

==4Yl

+...x )d 7x 19x -+-+-+ +...dx f ( x 192 )


1

x3

x 19x + x2 + 192 4608) + 4


1
3 3

7 4

by lOng division)

4x

==Yl lnx + 4Yl


Thus)

) (--+-+ +....
1

4608)

7x2

4 19x

2x2

384

18432)

Y2(X)==Yl(x)lnx--+-+-x 8 32 4608)

x3

5 l1x

+....

(50))))

246

3 Methods) Chapter PowerSeries Note that the technique of reduction of orderreadily yieldsthe first several terms of the series, does not provide a recurrence but relation that can be used to detennine the general term of the series. With a computation similar to that shown in Example (but more complicated 4 Problem 21), method of substitution can be usedto derive the solution) the
Y3(X)

-see

= Yl(X) lox

-- + 2:(-l)n(H+ Hn_1)x2n-1 2 n. _ x '(


1
00
n

n=l

2n

(51)
the terms

.) 1)\"

where Hn is defined in (44)for n


shown
in

Eq. (50)agree

> 1; o == O.The readercan verify H


3 -J (x) + Y3(X). 4)
1

that

with)

Y2(X) ==

(52))

The most commonly usedlinearly independent [of J1]olutionof Bessel's s equation of order 1 is the combination) 2 2 f (x) == -(y -ln2)Y1(x)-Y3(X) + J[) J[
1

==

2 J[

[(

In

-) J1(x) -- + 2:(-1)n(Hn+Hn_1)x2n-l (53) . 2 n! - I)! 2 x


X

00

n= 1

2n

(n

.)

4 Examples and 5 illustrate two methodsof finding the solution in the logarithmic cases-direct substitution and reduction of order.A third alternative is outlined in Problem 19.)

roblems BJ..r .
In

..)

or find one such solution and show (as in Example 2) that a secondsuch solution doesnot exist.)

1to find

Problems

1through

8, either apply
independent

the method

two linearly

Frobenius

seriessolutions,

of Example

12. 2y\" + x 2y'2 -2y = 0 x 2


y\" y\" with)

2.xy\" + (5 x)y' y 3.xy\"+(5+3x)y'+3y=0


4. 5xy\" + (30+ 3x)y'+ 3y = 0 5. xy\" (4 + x)y' + 3y = 0 6.2xy\" (6 + 2x)y'+ y = 0 7. x 2y\" + (2x + 3x2)y' 2y = 0 8.x (1 x)y\" 3y' + 2y = 0)

1.

xy\"

+ (3 - x)y' - -

=0 =0

13.2 + (2x - 3x)y'+ 3y = 0 x 14. + x(1+ x)y' -4y = 0 x 15. egin B


Jo(x) = 1
Using
linearly

- -+ - 4 64
x2 x4
solution) 2

x6 + 2304)

. .

..
-

-- -

the method

of reduction of order, derive the second

independent

Y2(X)

= Jo(x)In x + x

-- 3x4

+ 128 13284)

6 l1x

...

In
in

a Frobenius seriessolution of the given differential equation. Then usethe reduction of ordertechnique (asin Example 4) tofind the logarithmic term and thefirst three nonzero terms in a second linearly independent solution.

Problems9 through

14,irst f

find the first

four

nonzero terms

16.ind F

of Bessel'squation of orderzero. e of Bessel'squation of order e


x 2y\"
two linearly independent Frobenius
\037,)

seriessolutions

9.xy\" + y' xy = 0 10. 2y\" xy' + (x2 + l)y = 0 x2 2

11. x

y\"

+ (x

- --

+xy' + (X2 -\037) Y = O.)

17.(a) Verify

that

YI

(x)

3x)y'+ 4y = 0)

x 2y\"

= x eX is one solution x(1+ x)y' + y =

of

O.)))

3.4Methodof Frobenius: Exceptional The Cases 247)


(b) Note that
rl

= r2 = 1.ubstitute) S
Y2

Hencededucethat)
YI

= yIlnx + Lbnxn +I
n=I)

00

= y(x, rI) =

L
n=O)

00

C (rdxn +r n

\\

(58))

in the differential that)

equation

nb n

bn

in relation and (c) Substitute bn = cn/n! this recurrence conclude from the result that C = Hn. Thus the second n

- = n!
I

-1

to deducethat b I for n

-1

and)

> 2.

is one solution of Eq. (54). (b) with respect r to show that) to L[Yr(x, rl)) =
\037

Differentiate

Eq. (57)

8r [xr (r

- rl)2]
r=r\\)

= O.

solution

is)

Y2(X)

x = xe In x

18. onsiderthe equation C


series solution)
rl

1 and

r2

= 0 at
Yl

x y\" Y = 0, which has exponents x O. (a) Derivethe Frobenius ' L n.'(x_ 1)1
n=I

=
00

-L H
00
n=I
n

n xn+I

Deduce that Y2 = Yr (x,rI) is a second solution of Eq. (54). (c) Differentiate Eq. (58)with respectto r to
show that)

n!)
Y2

= ylin

+ xr

\\

L
n=l)

00

c\037

(rI)x

(59)

20.Usethe
tions in

(x)=

.)

method of Problem 19 to derive both the solu(38) and (45) of Bessel'squation of order zero. e The following stepsoutline this computation. (a) Take s in Co = 1;how that Eq. (55)reduces this caseto)

(b)

Substitute)

(r
Y2

+ 1)2CI(r) = 0

and)

= CyIlnx+

in the equation

xy\"

L
n=O)

00

bnx

cn(r) (b)

C -2(r) n (n

+ r)2

for n

> 2.)

(60))

o
bn

to

derive the recurrence)


2n

relation)

n(n

+ l)bn +I

- =-

+1

(n

+ I)!
Hn

C. n!)
is)

= c;(0) = 0, and then deduce (60) that Cn (0) = (0) = 0 for n odd. Henceyou needto compute Cn (0) and (0) only for n even. (c) Deducefrom (60)that)
Next show that CI(0)
c\037

from

c\037

Conclude from
Y2(X)

this result that

a secondsolution
oo

= YI (x) In x + 1
the differential

- L n! - I)!
n=I

C2n(r) =
With

(n

+ Hn-I x n .

(r

+ 2) (r + 4)2 ...(r + 2n)2


2

(-1)n

(61)

19. upposethat S

r = rl = 0 in (58),this gives ate (61)o show that) t

Jo(x).(d)
.)

Differenti-

equation

2 L[y] = x y\" + xp(x)y'+ q(x)y = 0) (54)) has equal exponents rl = r2 at the regular singular point x = 0, so that its indicial equation <p(r) = (r rI)2 = Let Co = 1 and define Cn (r) for n > 1 by using Eq.(9);

H , c2n (0) = (_I)n+I n 22n (n !)2

Substitution

of this

result in

is)

solution (59)gives the second

O.)

21. erivethe D
tion ing

in

(45).

that is,)

cn(r) Then define the

-Ln(r;

Co,CI,
<p(r

+ n))

..., .
Cn-I)

o logarithmic solution in (51) f Bessel'squae of order 1 by the method of substitution. Thefollowstepsoutline this computation. (a) Substitute)

(55)
in

Y2=CJ(x)lnx+x1 1

(1+

[\037

bnxn))

function Y

(x,r) of x and

r to be)

Bessel'squation e
00
I n=2)

to obtain)
n

y(x,r) =
(a) Deducefrom
the

L
n=O)

00

cn(r)xn +r

(56))

2 -b + x + L[(n - l)b + + bn_tJx


n
1

discussion preceding (9) that) Eq.


xr

L[y(x,r)] = (r

+C x+

rI)2.)

(57))

1)x \"(-I)n(2n++I)! + 2
00
2n 2n

\037

(n

n!)

= O.

(62))))

248
for n

3 Methods) Chapter PowerSeries

(b) Deducefrom Eq. (62)that

C=

odd. (c) Next deducethe recurrence relation

-1

and that

bn

=0

in

Eq.(63)to

obtain)

[(2n + 1)2
for n > determined

- 1]
n

b2n +2 + b2n

1. ote that N
for all

2 n(n + I)! ! (63) n if b2 is chosenarbitrarily, then b2n is > (d) Takeb2 = and substitute

= (\037l)n(2n + 1)

C2n+2

C2n

n+l + -.
1 1
n)

b2n

1.

-1

(e) Note that

C2

= 1 = HI + Ho and
C2n

deducethat)

22n (n

- 1)!n!)
)n C2n

= Hn + Hn-I.)

lID Bessel's Equation)


We
\037)

have already seenseveral cases f Bessel's o equation of orderp


x2y\"

> 0,)
(1))

+ xy' + (x2

- p2)y

== O.)

Its solutions are now called Besselunctions of order p. Such functions first apf in the 1730s the work of Daniel Bernoulli and Euler on the oscillations in peared of a vertically suspendedchain. The equation itself appearsin a 1764article by Euleron the vibrations of a circular drumhead, and Fourier usedBessel functions in his classical But treatise on heat (1822). their general propertieswere first studied

m systematically in an 1824 emoir by the German astronomer and mathematician Friedrich W. Bessel who (1784-1846),was investigating the motion of planets.The standard source information on Bessel of functions is G.N. Watson's A Treatiseon the Theory of Bessel Functions, 2nd ed.(Cambridge: ambridge University Press, C which cover only the period up to 1922, 1944).Its 36 pages of references, give someidea of the vast literature of this subject. 2 Bessel's p2 == 0,with roots r == p. equation in (1)has indicial equation r m +r If we substitute y == L cmx in Eq. (1), find in the usual manner that Cl == 0 we

::l::
(2))

and

that)

\037)

[em + r)2

- p2]

C m

+ Cm -2== 0)

for m

> 2.The verificationof Eq. (2)is left to the reader(Problem 6).)


p and write
am in

The Caser=p > 0)


If we use r
formula)
\037)

of place Cm, then Eq. (2) yields the a -m(2p-2 +


m

recursion)

am ==

m))

(3))

coefficients are) a

Because == 0,it al

follows that am == 0 for all odd values of m. The first few even

- ao - ao 2--2(2p+2)--2(p+1)') a2 - 2 .2(p+aol)(p + 2)') a -4(2p + 4)


2 4 4

ao . a6==- a4 ==-26 .2.3(p+1)(p+2)(p+3)))) 6(2p+6)

3.5Bessel's Equation
The general pattern
is)

249)

a2m ==

22m m!(p+ we get the

, 1)(p+2)...(p+m))
solution)

ao -l)m

so with the larger root r


\037)

== p

YI(X) == ao

\" f;;o
00

2m

x2m+p (-l)m . m!(p + l)(p 2) ...(p + m)) +

(4))

If p == 0 this is the only Frobeniusseriesolution; with ao == 1 as well, it is the s function Jo(x)we have seenbefore.)

The Caser = -p < 0)


If we use r == p and write bm
\037)

in

placeof Cm, Eq. (2)takes the form)

m(m

- 2p)b +
m

bm 2 == 0)

(5))

for m > 2,whereasb I == O. We seethat there is a potential difficulty if it happens 2p is a positive integer-that if p is either a positiveinteger or an odd positive is, b integral multiple of For then when m == 2p, q. (5)is simply O. m + bm 2 = o. E Thus if bm 2 i= 0,then no value of bm can satisfy this equation. But if p is an odd positive integral multiple of we can circumvent this For supposethat p == k/2 where k is an odd positive integer. Then we difficulty. needonly choosem == 0 for all odd values of m.The crucial stepis the kth step,) b
that

1.

1,

k(k and
this

- k)b +
k

bk 2 == 0;)

equation will hold becausebk == bk 2 == O. Hence p is not a positive integer, we take bm == 0 for m odd and definethe if coefficientsof even subscriptn terms of bo by means of the recursion formula) i
\037)

bm ==

- m(m

bm 2 2p))

> 2.)
in

(6))

with except p replacedwith

In comparing (6)with

(3),we seethat (6)will leadto the same result as that

-p.Thus
22m

in this

we case obtain the secondsolution)


2m

(4),

\037)

Y2(X) == bo

\" f;;o
00

x (-l)m -p . m!(-p+ l)(-p 2)... +m)) + (-p

(7))

The series (4) and (7)converge for all x > 0 becausex == 0 is the only singular in point of Bessel's equation. If p > 0,then the leading term in YI is aoxP, whereas the leading term in Y2 is box p. Hence (0) == 0,but Y2 (x) ::1::00 0, so as x YI it is clearhat YI and Y2 are linearly independent solutions of Bessel's t equation of

\037

\037

orderp

>

O.)))

250

3 Methods) Chapter PowerSeries

TheGammaFunction
The formulas in (4) and (7)can be simplified by use of the gamma function which is definedfor x > 0 by)
\037)

rex),
(8))

r(x) =

is not difficult to show that this improper integral converges for each x > O.The w gamma function is a generalization for x > 0 of the factorial function n!, hich is defined only if n is a nonnegative integer. To seethe way in which r (x) is a w generalization of n!, e note first that)
It

00

x e-tt-

dt.)

reI)==

1
o
X

00

e-t dt == lim [_e-t ]


b\037oo

==

1.
x [bxe-tt- dt
I

(9))

0)

Then we integrate by parts with

u == t X

and dv ==
lim
b\037oo

e-t dt:)
b

r(x + 1) =

lim
b\037oo

[be-tt dt = 10

([

_e-ttx] + 0

10)

== x
that is,)
\037)

( lim
b\037oo

[b
10)

e\037ttx\037l

dt ;

rex + 1) == xr(x).)

(10))

This is the most important property of the gamma function. If we combineEqs. (9)and (10), seethat) we

r(2)== 1 . reI)== I!, r(3)== 2.r(2)== 2!, r(4) == 3 . r(3)== 3!,)


and
\037)

in

general that)

r(n + 1) == n!)
00

for n

> 0 an integer.)
is) 00

(11))

An important

specialvalue of the gamma function

r G) =

e-t t- / 2 dt =
1

21
2)

e-

u2

du = ...(ii,)

(12))

where we have substituted

u 2 for t in 00

the

first

integral; the fact that)

1e
o

-u2 d

\037 u==-

is known, but is far from obvious.[See,or instance, Example5 in Section of 13.4 f Edwardsand Penney, Calculus: arly Transcendentals, 7th edition (UpperSaddle E River, NJ: Prentice Hall,2008).] r (x) is defined in (8) only for x > 0,we can use the recursion Although formula in (10)to define r (x) whenever x is neither zero nor a negativeinteger. If

-1< x < 0,

then)

rex) ==

rex + 1);
x)))

3.5Bessel's Equation 251) the right-hand sideis defined because0 < x + 1 < 1. he same formula may then T be used to extend the definition of f (x) to the open interval (- -1), to the 2, then a openinterval (-3,-2), nd so on.The graph of the gamma function thus extended is shown in Fig.3.5.1. student who would like to pursue this fascinatingtopic The
further

Winston,

should consult Artin's The Gamma Function (New York: Holt, Rinehart and i In 1964). only 39 pages,this is one of the finest expositionsn the entire literature of mathematics.)

FIGURE

extended gamma

3.5.1. graph The


function.)

Bessel unctionsf theFirstKind F o


of the
P If we chooseo == Ij[2 f(p + 1)]in (4), where p a

f(p + m + 1) == (p + m)(p+ m 1)...(p + 2)(p + I)f(p+ 1)) we function of the first by repeatedapplication of Eq. (10), can write the Bessel
kind of orderp very conciselywith the aid of the gamma
\037)

> 0, and note that)

function:)

= 1p(x)

X 2m+p (-I)m (2 ) + \037 m!r(p m + 1)

00

.)

(13))

P ho Similarly, if p > 0 is not an integer, we choose == Ij[2-f( p + 1)]in (7) to obtain the linearly independent secondsolution)

00
\037)

( m Lp(X)=\037 !r(_p+m+l)2))
+ y(x) == C1Jp(X) C2J-p(X))

(-I)m

2m-p (14))

of Bessel's equation of order

p.If p is not an integer, we have the general solution)


(15))

for x > 0;x P must be replacedwith Ix IP in Eqs. (13)through (15)to get the correct solutions for x < o. If p == n, a nonnegativeinteger, then Eq. (13)gives)
\037)

In(x) _

(-l)m -m=om!(m+n)!-) (
\037 \037
X

2m+n

2)

(16))

for the

Bessel functions of the first lo(x)=


00

kind

of integral order.Thus)
x2 2

x (_I)m 2m = 22m (m!)2 \037

1-2
=X

+ 22.4 22.4.6 2 2 2
1 x
3

x4

x6

+...
-

(17))

and)

1 (x) =
1

00

\037

1)! 22m+lm!(m+ 2 2!( 2 )


(

2 -I)m2m+

+ 2!.3! 2 ) (

....
they

(18))

The graphs of Jo(x)and

if you examine the seriesn (17),you can seepart of the reason why lo(x) and i cosx might be similar-onlyinor changesin the denominatorsin (17)are needed m to producethe Taylor seriesor cos As suggested Fig.3.5.2, zerosof f x. the by
the)))

sembledampedcosine sineoscillations,espectively (see Problem 27).Indeed, and r

J1(x) are shown

in

In Fig.3.5.2. a general way

re-

252

3 Methods) Chapter PowerSeries


y)
.>,......,....';<.i

'

''' <u:Df'l\"x .. . .
\"//\037O\037 \\ J)

n tll.<Zeoo.'.i.
>

(nl}Jr)
2.3562 5.4978 8.6394

nth

Zero of JI(x)

(n

1r)
\037)

2.4048

2
x

5.5201
8.6537

4
5

11.7915 11.7810 13.3237


14.9309
14.9226)
1

3.8317 7.0156 10.1735

10.2102 13.3518
16.4934)

3.9270 7.0686

16.4706)

functions

FIGURE3.5.2. graphs of the Bessel The Jo(x)and J1 (x).)

FIGURE

3.5.3. Jo(x)and J (x).) Zerosof

functions Jo(x) and J1 (x) are interlaced-between two consecutive zerosof any Jo(x) there is preciselyone zero of J1 (x) (Problem 26)and vice versa. The first four zeros of Jo(x)are approximately 2.4048, For 5.5201, and 8.6537, n large,the nth zero of Jo(x)is approximately (n the nth zero of J1 (x) Jr; is approximately (n + Jr. Thus the interval between consecutive zerosof either x Jo(x) or J1 (x) is approximately Jr -another similarity with cos and sin x. You can seethe way the accuracy of these approximations increases ith increasing n w to two decimal places. by rounding the entries in the table in Fig. It turns out that Jp (x) is an elementary function if the order p is half an odd in Eqs. (13)and (14), integer. For instance, on substitution of p = and p = the resultscan be recognized as) (Problem 2) respectively,

11.7915.

\037)

\037)

3.5.3
\037

\037

J1j2(X) =

2
JrX)

sin x

and)

Llj2(X)=

2
JrX)

cosx.

(19))

Bessel unctions f theSecond F Kind) o


The methods of Section must be used to find linearly independent secondso3.4 lutions of integral order. A very complicated generalization of Example3 in that sectiongives the formula)
Yn(x)

-( +
2
Jr

In

1 n-l -) In(x)- - \"2 - m!x-- - I)! 2 x


n 2m

(n

\037 Jr m=O)

n 2m

with

\037

m (-l)m(H+ Hm+n) Jr m=O m!(m + n)!

cr+
2
is)

2m
')

(20)

usedthere.If n = 0 then the first sum in (20)is taken to be zero. is called Besselunction of the second the kind of integralorder f Here, (x)
the notation
Yn

> o.

The general solution of Bessel's equation of integral ordern

\037)

y(x) = CIJn(X) + C2Yn(X).)

(21))

is important to note that Yn (x) --+ -00as x --+ 0 (Fig.3.5.4). C2 = 0 in Hence Eq. (21)if y(x) is continuous at x = O. Thus if y(x) is a continuous solution of Bessel's equation of ordern, it follows that)
It

y(x) = cJn(x))))

3.5Bessel's Equation
c == y (0).In Section we will seethat 9.4
for some constant c. Because Jo(O) ==

253)
0, then

1,we seein
this

addition

that

if

==

has numerous physical applications. illustrates the fact that for n > 1 the graphs of n (x) and Yn (x) Figure lookgenerally likethose of J1 (x)and Y 1 (x). In particular, n (0)== 0 while Yn (x) --* -00as x --+ 0+, and both functions undergo dampedoscillationas x --+ +00.)

functions singlefact regarding Bessel

3.5.5

y)

y)

0.5)

0.5)
x)

12 (x))
x)

-0.5) FIGURE3.5.4. graphs of the Besselunctions The f


Yo(x) and Y 1 (x).)

-0.5) The FIGURE3.5.5. graphs of the Bessel functions

J2(x) and

Y2(x).)

Bessel unction F Identities)


Bessel functions are analogous to trigonometric functions
number of standard identities of frequent utility, 00

functions. Differentiationof) gralsinvolving Bessel

in that they satisfy a large in especially the evaluation of inte-

Jp(x) =
in

J;
=

X 2m+p (-I)m ( 2)) m!r(p+ m + 1)

(13))

the

case p is a nonnegativeinteger gives) that

dx

p [x

p (x) ] ==

d
00

00

dx f:=o22m +Pm!(p + m)!)


x2m+2p-l (-I)m 22m +p-1 m!(p + m 1)!)

'\"\"\"\"

x2m+2p (-I)m

and
thus

J;
XP

00
'\"\"\"\"

f:=o2

2m

+p-lm !(p + m

x (-I)m

2m

+p

- I)!')
(22))

we have shown

Similarly,)

that)

dx)

[x Jp(x) ]
p

_p J -l(x). X

d dx

p p [x- Jp(x)]== -x- Jp+l(x).

(23))

If we carry out the differentiationsin Eqs.(22)and (23)and then divide the resulting identities by x P and x-P , respectively,we obtain (Problem 8) the identities)

= J;(x) Jp- (x) 1

P
x)

Jp(x)

(24))))

254

3 Methods) Chapter PowerSeries


and)

= J;(x)

P
x)

Jp(x) Jp+I(X).

(25))

functions Thus we may express derivativesof Bessel the functions in terms of Bessel themselves. ubtraction of Eq. (25)from Eq. (24)gives the recursion formula) S

2p Jp + l (x) == -Jp(x) Jp -l (x),


X)

(26))

which can be used to express esselunctions of B f functions of lower orders. the form) In

high

order in terms of Bessel

== 2p Jp-l(x) -Jp(x) Jp+l(x),


x)

(27))

it can be used to express Bessel functions of large negative orderin terms of Bessel functions of numerically smaller negative orders. The identities in Eqs. (22)through (27)hold wherever they are meaningfulthat is, wheneverno Bessel functions of negative integral orderappear.In particular, hold for all nonintegral values of p.) they

Example

1)

With

==

0,Eq. (22)gives)

f xJo(x)dx
Similarly,
with

= xJI(X)+C.

==

0,Eq. (23)gives) + f J (x)dx = -Jo(x) C.)


I

.)

Example2)

Usingfirst p

== 2 and then

== 1 in

Eq. (26),we get)


==
X
X)

J3(X) == -J2(X)
X

4 2 - Jl(x) - [-Jl(x)- Jo(x)] - Jl(x),

so that) J3(X)=
With

similar manipulations every Bessel function in expressed terms of Jo(x) and Jl (x).

-;Jo(x)+ (:2- 1)J (x).


I

of positive integral order can be

.)

Example3)

To antidifferentiatex J2(x),we first note that)

= f x-If2(x)dx -x- J (x) + C


I I

by

Eq. (23)with p

==

1.We therefore write)


f f xf2(x)dx f x [x-I2(x)]

dx)))

3.5Bessel's Equation
and integrate by parts with) 2 u == X
,)

255)

dv ==
and)
V

x-IJ2(X) dx,

du == 2x dx,)
This gives

== _X-1 J1 (x).)

= + f xh(x)dx -xJ1(x) 2 f J1(x)dx= -xJ1(x) 2Jo(x)+ C,


with

the aid of the

secondresult of Example1.)

.)

The Parametric essel quation B E


The parametric essel B equation of ordern is
\037

x 2y\"

+ xy' + (a2x 2

2 == 0,) )y

(28))

where a is a positiveparameter. As we will see Chapter this equationappearsin in the solution of Laplace's I equation in polar coordinates.t is easyto see(Problem9) that the substitution t == ax transforms Eq. (28) into the (standard) Bessel equation

9,

t with

2d

_ + t- +
2
y

dy
dt)

dt 2

(t

2 == 0 )y

(29))

the general solution yet) == C1Jn(t)+ C2Yn(t). Hence general solution of

Eq. (28)is)
\037)

y(x) == C1In(ax) + C2Yn(ax).)


Now

(30))

considerthe eigenvalueproblem
x 2y\"

+ xy' + (AX 2

2 == )

0,

y(L)

== 0)

(31))

on the interval [0, ]. We seek positive values of A for which there existsa the L nontrivial solution of (31)that is continuous on [0, ].If we write A = a2, then L the differential equation in (31)is that in Eq. (28),so its general solution is given in Y Eq. (30).Because n(x) --+ -00as x --+ 0 but In(O) is finite, the continuity of y(x) The requiresthat C2 == O. Thus y(x) == C1Jn(ax). endpoint condition y(L) = 0 now impliesthat z == a L must be a (positive) root of the equation)

I (z) == O.)
n

(32))
infinite

For n

o It follows that the kth (see 3.5.6). sequencef positive zerosYn1, Yn2, Yn3, Fig. of the problem in (31)is positive eigenvalue
Ak

> 1,In(x)oscillatesrather like J1(x) in Fig.3.5.2 hence has an and

..

= (ak)2 =

,)

(33))

(Y2\037)2

and

that

its

associated eigenfunction is)


Yk(X) == n

I (LX)
Ynk

.)

(34))

of Eq. (32)for n < 8 and k < 20are tabulated in Table 9.5of M. Abramowitz and I.A. Stegun, Handbook MathematicalFunctions (New York: of Dover, 1965).))) The roots
Ynk

256

3 Methods) Chapter PowerSeries


y)

Yn(X))

x)

The FIGURE3.5.6. positive zerosYnl,

Besselunction f

I (x).)
n

Yn2, Yn3,

..

of the

_.J>_r,obleltl\037_
.)

1.Differentiate
that
J\037(x)

2.(a) Deducefrom Eqs.(10)nd (12)hat a t

sine functions).

= -JI (x) (another

termwise the

seriesor Jo(x)to show f


analogy with the

cosineand

directly

11. se the U

r (n + 2) =
(b)

II.
and
I

3 5 (2n

. .

- 1)vn.
the formulas in construct a figure

relation rex + 1) = xr(x) to deducefrom t a Eqs.(13)nd (14)hat if p is not a negative integer, then Jp(x)=

(xj2)P

00

r(p+ 1) [
because only

1+
\037

m! (p +

(-I)m(xj2)2m 1)(p+ 2) (p + m)

.. .

.)

Usethe

for Eq.(19)

3.

showing the graphs of thesefunctions. that (a) Suppose m is a positive integer. Show that 2 2

JI /2(X)

result

of part (a) to

J- /2(X), and

verify

12.

This form is more convenient for the computation of Jp(x) the single value (p+ 1)of the gamma function is required. Usethe seriesof Problem to find y (0) = lim y (x) if

11

r (m + 3) = 2.5.8...(3m-l) r (3).
3m
part

x-+o)

y(x)

= x2

from (b) Conclude

(a) and

t Eq.(13)hat

[J /2(X) + J- /2(X) ]
I I

JS/2(X) + J-S/2(X)

.
evaluated

J- /3(X) =
1

Usea computer algebra system to graph y (x) for x near O. Does graph corroborateour value of y (O)? the y

(_I)m3mx2m (xj2)-1/31 + 00 r (D ( \037 22m m! .2.5 ...(3m 4. Apply Eqs.(19),26),and (27)to show that (

- 1)) .

Any in

integral

J3/2(X) = J
and)

2
JrX)

3 (sin x

- x cosx)
of thesetwo
func-

integral indefinite cannot be simplified further, but the function Jo(t) dt is tabulated in Table of Abramowitz and Stegun. Use the identities in Eqs.(22) and (23)to evaluate the integrals in Problems through

f Jo(x)dx. The latter

terms

of Besselfunctions

of

the form

f x I (x) dx
m n

can be

and

the

f;

integral

11.1

13

21.

L3/2(X) = -J JrX (cosx+xsinx). 23


a figure
showing the graphs

5. ExpressJ4(x) in

Construct tions.

13. x 2Jo(x)dx f 15. x4Jo(x)dx f

6.Derive the 8. 9. 10.


7.

terms

recursion

of o formula

J (x)and JI (x).
in

Eq. (2) for Bessel's

equation. Verify the identity in (23)by termwise differentiation. Deducethe identities in Eqs.(24) and (25) from those in Eqs.(22)and t transforms the paraVerify that the substitution metric Besselquation in (28)into the equation in e Show that

(23).

f x J (x)dx 19.f x 4J (x)dx 21. i)(x)dx f 22.Prove


2
1 1

17.

14. x 3Jo(x)dx f 16. xJ1(x)dx f 18. x 3J (x)dx f 20. ]z(x) x d


1

that)

= ax

(29).

Jo(x) = Jr

p 4J;(x)= Jp-2(x) - 2J (x)+ Jp+2(x).)

e by showing that the right-hand sidesatisfiesBessel'squaO. tion of orderzero and has the value o (0) when x this constitutes a proof.))) Explain why

-1
1
7r

sin cos(x e)de

3.6Applicationsof Bessel unctions F


23.Prove that)
I JI (x) = n
yields)

257)

- 1 cos(en
0)

x sin

e)de

Z\"

(1

p2X\037

by showing that the right-hand sidesatisfiesBessel'squae tion of order 1 and that its derivative has the value (0) when x O.Explain why this constitutes a proof. 24. It can be shown that)

J{

the latter equation reduces z\" + z to suggests (without proving it) that if Bessel'squation, then) e

(b) If x is so large that (p2

- i))/x
\037

z = O.
2

O.Explain

is negligible,

then

why this

y(x) is a solution of

I (x)= 1
n

I n

cos(ne- x

sin

e)de.
with

y(x)

\037

X- 1/2(A
1/2 CXand

cosx B sin x) +

=
C and
ex

0)

cos(x
x

ex))

(35))

assertion. prove the preceding 25.Deducefrom Problem 22 that) Jo(x) = 2n


(Suggestion:Show 1

> 2, show that the right-hand side satisfies Bessel'squation of ordern and also agreeswith the vale ues In (0) and J (0).Explain why this doesnot suffice to
With n
\037

constants,

large.)

Asymptotic Approximations It is known that the choices C = -J2/n and = (2n + l)n/4 in (35)yield the best approximation to

-1
0)

I (x) for x large:)


n
\037

ex

2n

eix sine de.


Similarly,)

In(x)

2
nx)

cas[x i(2n+ l)n].

(36))

first that)

10
then

{2n eixsinO dB

= {n (eiXSinO + e-iXsinO)
10

dB;) In particular,)

Yn

(x)

\037

2
nx)

sin

[x

-i

(2n

+ l)n] .

(37))

useEuler'sformula.) 26.UseEqs.(22)and (23)and Rolle'stheorem to prove that between any two consecutive erosof In (x) there is prez ciselyonezeroof In +1 (x).Usea computer algebra system
to construct a figure
instance)

Jo(x)
and)

\037

27. (a)

illustrating

this
y

fact with

n
in

= 10 (for

Show that the substitution

X- 1/2z

Bessel's
ratio
----+

Yo(x)

\037

equation

of orderp,)
x 2y\"

+ xy' + (x2 p2)y = 0,)

if x is large. Theseare asymptotic

j j:x
nx)

cas(x
sin

- in) - in)
in that the

(x

of the two sides eachapproximation in as x +00.)

approximations

approachesnity u

of lID Applications Bess\037lFlInctions)


The importance of Bessel functions stemsnot only from the frequent appearanceof Bessel's b equation in applications,ut also from the fact that the solutionsof many other second-order linear differential equations can be expressedterms of Bessel in functions. To seehow this comesbout, we begin with Bessel's a equation of order p
in

the form)

d2 dw Z2-+ z-+ - p2)w 2 dz


w

dz)

(Z2

==

0,

(1))

and

substitute)

-ex w == X y

,)

z == k x

t3
.)

(2))

Then a routine

but

somewhat tedioustransformation (Problem 14)of Eq. (1)yields)

x 2y\"

+ (1 2a)xy'+ (a2

f32 p2

f3

2k 2x 2t3

)y ==

0;)))

258

3 Methods) Chapter PowerSeries


that is,)
\037)

x 2y\"

+ Axy' + (B + Cxq)y == 0,)

(3))

where the constants A,


A == 1

- 2a,
a==

B,C,and q are given by


B == a 2

fJ2 p2,

C == fJ 2k 2,
== q

and

q == 2fJ.

(4)

It

is a simplematter to solve the equations in (4) for)

I-A
2
and)
fJ q)

\037)

2./C k = ,

J(1- - 4B p==
A)2
q)

'
(5))

Underthe assumption that the squareroots in (5)are real, it follows that the general
solution of Eqo (3)is)
where)

y(x) == x<Xw(z)

== x<Xw(kx

fJ

),)

w(z) == CIJp(z)+ C2Y -p(z) that (assuming p is not an integer) is the general solution of the Bessel equation in This (1). establishesthe following result.)
in Solutions Bessel Functions If C > 0,q i= 0,and (1 A)2 > 4B,then the general solution (for x > 0) of

1 THEOREM
Eq. (3)is)
\037)

+ y(x) == x [cIJp(kx,B) c2J_p(kx,B)],)

(6))

where

J-p is to be replacedwith
\"..\"

a,fJ, k, and p are given by the equations in


Y p .)

(5). If p is an integer, then

Example

1)

Solvethe

equation)

4x2y\"

+ 8xy' + (x 4
(-\037

- 3)y

==

o.

(7))

Solution To compareEqo (7)with


and

Eqo

(3),we rewrite the former as

x 2y\"

seethat

a ==

A ==

\037,

fJ

==

2,k ==
Eqo

2,B ==
\037,

+ 2xy' +
C
==
\037.

==
\037,

\037,

+ ix4) Y == 0) and q == 4. Then the equations in (5)give + C2J-I/2 (\037x2)] .


1

and

p
I

Thus the general solution in (6)of Eq. (7)is

y(x) == x-

/2

[CIJI/2(\037x2)

Ifwe recallrom f
1

3.5 (19)of Section that

J / 2(Z) =

Jnz

sinz

and

2 L /2(Z) = Jnz cosz,


in

we seethat a general solution of Eq. (7)can be written

y(x) = x- /

32

( 4 + B 4)
Acos
sin

x2

the elementary form 2 X


\302\267)

.)))

3.6Applicationsof Bessel unctions F


Example2)
Solvethe
Airy equation)
yll

259)

+ 9xy = O.) + 9x3y = O.)


\037,
\037.

(8))

Solution

First we rewrite the given equation in the form) x2yll

This is the specialcase Eq. (3)with A = B = 0,C = 9,and q = 3.It follows of = k = 2,and p = Thus the general from the equations in (5)that a = fJ solution of Eq. (8)is)

i,

3 1 y(x) = X 1/2 [Cl1/3 (2x / ) + C21 (2x / )] -1/3 3

.)

.)

o Column Bucklingf a Vertical


For a practical application, we now considerthe problemof determining when a uniform vertical column will buckle under its own weight (after, perhaps,being We nudged laterally just a bit by a passingbreeze). take x = 0 at the free top end of the column and x = L > 0 at its bottom; we assumethat the bottom is rigidly imbeddedin the ground, perhapsin concrete; Fig. see Denotethe angular deflection of the column at the point x by () (x). From the theory of elasticity it follows that)

3.6.1.

d 2() E I \037 + gpx() = 0,


x=L) dx)

(9))

FIGURE
column.)

3.6.1. buckling The

where E is the Young'smodulus of the material of the column, I is its crosssectionalmoment of inertia, p is the linear density of the column, and g is gravitational acceleration. physicalreasons-no For bending at the free top of the column and no deflection at its imbedded bottom-the boundary conditions are)

()'(0) = 0, ()(L)= O.)


We will

(10))

to solve it

accept(9)and (10)as an appropriate statementof the problem and attempt


in this

form. With)

- - EI')
Y

gp

(11))

we have the eigenvalue problem)


()II

+ y 2x() = 0;

()'(O)= 0, ()(L)= O.)

(12))

The column can buckleonly if there is a nontrivial solution of (12); otherwise the column will remain in its undeflectedvertical position. The differential equation in (12)is an Airy equation similar to the one in Example 2. It has the form of Eq. (3) with A = B = 0,C = y2, and q = 3. The equations in (5)give a = fJ = k = \037y, and p = Sothe general solution is)

i,

\037,

\037.

()(x)= x / 2 [Cl1 (\037yx3/2) + C21 (\037yx3/2)].) l/3 -1/3


1

(13))))

260

3 Methods) Chapter PowerSeries


In

orderto apply the

initial

w in) conditions, e substitute p = :i::


\037

00

( m!r(p+m+l)2 ) Jp(X)=\037
and
find

(-l)m

2m+p
')

after somesimplificationsthat)

B(x)=

c1yl/3 x 3 /3r (1)


1

- y2x4 + 504 ( 12
1_ y 2x3

4 7 y x

+
From
this it

c231/ 3
y 1/3

r (D (

+ y180

4x6

-... ) -... )
.)

is clearhat the endpoint condition ()'(0) = 0 impliesthat t

Cl

= 0, so)
(14))

x1 ()(x)= C2 / 2J- /3 (\037yx3/2)


1

.)

The endpoint condition ()(L)= 0 now gives)


2)

J- / 3
1

(\037y

L 3/ 2)

= o.) =
\037y

(15)) L3/ 2 is a root of the equation

>-.)

Thus the column will buckleonly if z J-I /3(Z) = O.The graph of)
o)

(Z/2)-1/3 1 + Ll/3(Z)=

00 \037

-1
o)

r n)

22m

m!.2.5 . (3m

z2m (-1)m3
m

- )
1))

(16))

5)
z)

10)

15)

FIGURE3.6.2. graph of The

J- /3(Z).)
1

Problem 3 of Section is shown in Fig.3.6.2, we seethat the smallest where (see 3.5) zero z is a bit less than 2.Mosttechnical computing systems can find roots positive like this one. or instance, eachof the computer system commands) F
1

x, 1..2) fsolve(BesselJ(-1/3,x)=O, FindRoot[BesselJ[-1/3,x]==O, {x,2}]

(Maple)
(Mathematica)
(MATLAB))

fzero(lbesselj(-1/3,x)l, 2))

(rounded accurate to five decimal places). yield the value ZI = 1.86635 The shortest length L 1 for which the column will buckleunder its own
IS)

weight)

Ll =
\037

\037\037

Y/3

[ (
3\0371

If we substitute ZI and 1.86635 p = 8A, where 8 is the volumetric density of the material of the column and A is its cross-sectional we finally get) area,

;;
1/3

Y/2r/

Ll

\037

EI (1.986)8A) g

( )

(17))))

F 3.6Applicationsof Bessel unctions

261)

for the critical buckling length. For example,with a steelcolumn or rod for which 3 7 2 and X E = 2.8 10 Ibjin. g8 = 0.28 , Ibjin.the formula in (17)gives the results shown in the table in Fig.3.6.3.)
i' R Cros ...s..............Secti()u.....ot . .... i>d Circular with r = 0.5 in. Circular with r = 1.5 in. Annular with nnner = 1.25 and router = 1.5 in. in.)
H ..,..,...-'__'_\",'_ '._.._. ,,-,._. ,___ \"_.H_.H__H.,-, __ _ ____, '_' _',__.'_,) .... ...

'

ShortestBuckling Length Ll)


30 ft 6 in. 63 ft 5 in. 75ft 7 in.)

FIGURE3.6.3.) have used the familiar formulas A = Jr r 2 and I disk. The data in the table show why flagpoles are hollow.)
We

iJr r 4 for a circular)

ED Problems)
Problems through expressthe generalsolution given differential equation in terms of Bessel functions. 2 2 x x 0
In

12,

of the

16.a) Substitute (

y\" y =0 2 6. 16x y\" + 24xy' + (1+ 144x3)y = 0 7. x 2y\" + 3xy' + (1+ x 2)y = 0 8.4x2y\" 12xy'+ (15+ 16x)y= 0 2 9. 16xy\" (5 144x3)y = 0 10.x2y\" 3xy' 2(14 x 5)y = 0 2 11.\" + x4y = 0 y 12.\" + 4x3y = 0 y

- xy' + (1+ )y = 2. + 3y' + = 0 3. - y' + 36x = 0 4. x - 5xy' + (8 + x)y = 0 5. 36x + 60xy' + (9x - 5)
1.
y\" xy\" xy\"

of Section in 3.5 the seriesof Problem the result of Problem 15here to show that the solution of the initial value problem
dy

11

xy

dx
IS)

= x 2 + y2, = x J3/4
I

y(O)

= 0)

y\"

y(x) (b) Deducesimilarly


problem)

J- /4 ( 1.X2) . 2
(\037X2)

- - - xy
\"

dy dx)

that the solution

of the

initial

value

= x 2 + y2,
(\037X2)

y(O)

=1
(\037X2)

IS)

13. pply A

y(x)
of)

Theorem1 to show

= x 2r(*)J3/4 3
I

that the

general solution
=) 0

J-3/4 +r 2r( 4) J- /4 ( 2X 2) - r ( 4) J /4 ( 2X2) .


(\037)

+ 2y ' + xy

14. erify that the substitutions in V (Eq.(1)) yield Eq.(3). 15.a) Show that the substitution) (
y

is y(x)

= X-I (A cosx B sin x). +


(2) in Bessel's equation

Somesolution curves of the equation dyjdx = x 2 + y2 are The shown in Fig. 3.6.4. location of the asymptotes where +00 can be found by using Newton's method to y(x) in the formulas for the find the zerosof the denominators solutions as listed here.)
\037 3) 2)

--1 du
u dx)
\037

1)

dyjdx = x 2 + y2 is)

transforms u\" + x 2u

= O.

the

Riccati equation dyjdx = x 2 + y2 into (b) Show that the general solution of

0)

-1) -2)

J3/4 y(x) = X
Section3.5.))
(Suggestion: Apply

(\037X2)

CJI /4 2X2 +
the identities

(')(')'
- CJ- /
I

J- /4

3 4 (\037X2)

2x2)

-3 -2 -1 -3

0) x)

in

Eqs.(22)and (23)of

dy Solution curves of FIGURE3.6.4.


dx)))

3)

= x 2 + y2.

262

3 Methods) Chapter PowerSeries


tapered rod
with

17.Figure 3.6.5 shows a linearly


the endpoint value problem) Ely\"
y)

circular

crosssection,subject to an axial forceP of compression. As in Section2.8, deflection curve y == y (x) satisfies its

2.8. 18. onsider a variable-length C Assume that its Fig. 3.6.6.


tion
LO\"

Note that

if

==

b, this

result

reducesto Eq.(28)of Sec-

+ Py

==

0;

yea) == y(b)

==

O.

(18))

as indicated in pendulum length is increasing linearly with time, L(t) == a + bt. It can be shown that the oscillations of this pendulum satisfy the differential equation)

+ 2L'O' gO == +

0)

under the usual condition that 0 is so small that sin 0 is very well approximated by sin O. Substitute 0 L == a + bt to derive the general solution)

0:

\037

x)

B(t) x=a) x=b)

.Jr [Al'(\037JgL)+ BY, (\037JgL)]


.)

FIGURE3.6.5. taperedrod of Problem 17. The Here, however, the moment of inertia 1 == 1 of the (x) crosssectionat x is given by)

of this solution to a discussion the of (\"its nether extremity was steadily descending pendulum formed of a crescentof glittering steel, about a foot in
For the
application

lex)=
where

\037n(kx)4

= 10'

Gf '
b. Substitution of

the value of at x the differential equation in value problem)

1(x) in

10 ==

I(b),

edge as keen as that of a razor. and the whole hissed as it swung through the air down and still down it m classic\"The Pit and came\") of Edgar Allan Poe's acabre
the Pendulum,\" seethe article by Borrelli, Coleman, and Hobsonin the March 1985issueof Mathematics Maga-

length from horn

to horn;

the horns upward, and the under

.. ...

==

(18) yields the eigen==

zine (Vol. 58, pp.78-83).)

x 4y\"
where
A
==

+ AY == 0, yea) == y(b)
==

0,
the

of this sectionto x 4y\" + J12y == 0 is)


rem

J12

Pb4

jE 10. (a)

show that the

general solution of

Apply

theo-

y(x)

=x (ACOS + Bsin
\037

\037

).
A FIGURE3.6.6. variable-length

that the nth eigenvalue is given by J1n (b) Conclude nnabjL, where L == b a is the length of the rod, and hencethat the nth buckling forceis)

Pn

==

n 2n 2

----v-( b )

E 10.)

pendulum.)

3.6Application)
......

\037ic\037ati_

Eq.\037ati\037_\037,\037,,_\037n\037...,

\037,.\037.\037i\037ie\037

B\037ss\037__\"

\037unc:.\037i\037\037s)

Riccati equation is one of the form)


dy
dx)

= A(x)y2 + B(x)y+ C(x).


oneslistednext can be solved explicitly in terms of
dy

Many Riccati equations likethe

Bessel functions.)

dx
dy
dx)

= x2 + l;)

(1))

_ = x2 _ y2.

(2))))

3.6Applicationsof Bessel unctions F


dy

263)
(3) (4) (5) (6)

dx

= y2

2 ;

dx
dy

dy

= x + y2; =x

dx

y2;

dy -=y2 -x.
dx)

For example, roblem 15in P


given
by)

this

sectionsaysthat the general solution of Eq. (1)is


J3/4 (\037x2)
1

y(x) = x

- cJ-3/4

(\037x2)
1

CJ1 / 4 ( 2X2 )

+ J-l/42 (

(7))

X2)

Seewhether the symbolicDE solver command in your computeralgebra system, such as the Maplecommand
= dsolve(diff(y(x),x) x
or the Mathematica command)
A A

2 + y(x)A2,

y(x\302\273

DSolve[y'[x] == x 2 + y[x]A2, y[x],x ] functions other than those appearing in Eq. (7)are agreeswith Eq. (7). If Bessel involved, you may need to apply the identities in (26)and (27)of Section3.5to transform the computer's\"answer\" to (7). Then seewhether your system can take the limit as x 0 in (7)to show that the arbitrary constant c is given in terms of
\037

the

initial

value y (0)by)

to plot typical solution curves likethoseshown in Fig.3.6.4. Next, investigate similarly one of the other equations in (2)through (6).Each has a general solution of the samegeneral form in (7)-a quotient of linear combinations of Bessel functions. In addition to Jp(x) and Yp(x),thesesolutionsmay involve the modifiedBessel functions)

Now you

c = y(O)f (i) 2f (D should be able to use built-in Besselunctions f


.)

(8))

Ip(x) = i-P Jp(ix))


and)

Kp(x)= 2i
that

T(

-P [Jp(ix)+ Yp(ix)]

satisfy the modifiedBessel equation

x2y\" + xy' (x2 + p2)y = 0)


of order

p.For instance, the general solution of Eq. (5)is given for x > 0
_ X 1/2I2/3 ( 3x3/2) y(x) 2

1-1/3x3/ 2) (3

- I-2/3 - C/l/3
C

by)

2 3/2 ( 3x ) 2 / , ( 3x3 )
2)

(9))))

264

3 Methods) Chapter PowerSeries


where)

c = y(o)r /3
1

-3

(\037)

(10))

(\037))

4)

2)

\037o)

-2) -4) -5)


o) x) 5)

showssometypical solution curves, together with the parabola y2 = x Figure 3.6.7 that appearsto bear an interesting relation to Eq. (6)-we seea funnel near y = = -,JX. +,JXand a spout near y The Bessel functions with imaginary argument that appear in the definitions of Ip (x) and Kp (x) may look exotic, the powerseries f the modified function but o without the alternating In (x) is simply that of the unmodified function n (x) except minus signs. or instance,) F

10)

Io(x)=
and)
X

+ 4 + 64 + 2304)+
x3 x5
X

-x2 x4

x6

...

FIGURE3.6.7. olution curves S of dy dx)


==

y2.

+ II(x) = 2 + 16+ 384 + 18432 ... .


Checkthesepowerseriesxpansions sing your computer algebra system-look at e u Besselln either Mapleor Mathematica-and i them with Eqs. (17)and compare

in however, cannot be expressed terms of elementary functions and/or \"known\" functions. Nevertheless, specialfunctions such as Bessel they can be investigated showssoluusing an ODEplotter. For instance, the interesting pattern in Fig.3.6.8 tion curves of the second-order equation)
y\"

3.5. (18)in Section The second-order differential equations of the form y\" = f (x,y) with the same right-hand sidesas in Eqs. (1) through (6) have interesting solutions which,

= y2 _

X)

(11))

sameinitial value y(O) = 0 but different slopesy/(O) = -3.3, -3.1, 0.7. quation (11) a form ofthejirstPainleve transcendant, an equation that arose is E d historically in the classificationof nonlinear second-order ifferential equations in
with

the

...,

w Equations,New York: Dover Publications, 1956). Figure 3.6.8as suggested by an article by Anne Noonburgcontaining a similar figure in the Spring1993 of issue the C.ODE. Newsletter.) E

terms of their critical points (seeChapter

14of E.L. Ince,Ordinary Differential

2 0
\037

\"

----------,.,.\"-y2

=x

-2
-4
0 2 4 6
x)

10
y\"

12
==

FIG URE 3.6.8. first Painleve transcendant The


y(O)
==

0, y'(O)==

-3.3, ...,0.7.))) -3.1,

y2

- x,

F 3.6Applicationsof Bessel unctions


The c package. Mapledsolveommand yieldsthe general solution)
Finally,

265)

here's related examplethat a

was inspiredby a Mapledemonstration

+ y(x) = x-I(CIJIO(X) C2Y IO(X))


2 806400x 9600x 100x x 4 6 + x-II + + + + 51609600x + (1857945600
8
IO

) (12))

of the nonhomogeneous second-orderquation) e

x2y\" + 3xy' + (x2 99)y = x.)

(13))

Showthat Theorem 1 in this sectionexplainsthe \"Bessel solupart\" of the alleged tion in Eq. Can you explain where the rational function part comes from, or at least verify it? For further examples this sort, you can replacethe coefficient of 99 in Eq. (13)with r 2 where r is an even integer, and/or replacethe x on the S right-hand side with x , where s is an odd integer. (With parities other than these, more exoticpecialfunctions are involved.)))) s

(12).

- 1,

LaplaceTransforll1
Methods)

TransforD1snd InverseTransforD1s) a Laplace 11II)

f(t),)

I
D{fU)} = f '(t))

n Chapter 2 we saw that linear differential equations with constant coefficients have numerous applications and can be solved systematically.There are common situations, however, in which the alternative methods of this chapter are preferable. For example, recallhe differential equations) t

mx\"

+ ex'+ kx = F(t))

and

LI\"+ RI'+

-I= E'(t)
1
C)

f(t)

,)
\037{f(t)}

= F(s)
with

to R corresponding a mass-spring-dashpot ystem and a series LC circuit, respecs It often happens in practice that the forcing term, F(t) or E'(t),has tively. circuit is discontinuities-for example,when the voltage suppliedto an electrical turned off and on periodically. n this case I the methods of Chapter 2 can be quite awkward, and the Laplacetransform method is more convenient. The differentiationoperator D can be viewed as a transformation which, when = T appliedto the function f(t),yieldsthe new function D{f(t)} f'(t). he Laplace transformation involves the operation of integration and yieldsthe new function = F(s)of a new independent variable s.The situation is diagrammed in 4.1.1. learning in this sectionhow to compute the Laplacetransform F(s) After Fig. of a function f(t),we will seein Section that the Laplacetransform converts 4.2
\302\243

\302\243{f(t)}

FIGURE
D.)

of a function:

4.1.1. Transformation
\302\243

in analogy

a differential equation in the unknown function f(t) into an algebraic equation in F(s). Becauselgebraicequations are generally easiero solve than differential t a equations, this is one method that simplifies the problemof finding the solution

(t).)))

266)

4.1Laplaceransforms and InverseTransforms 267) T


DEFINITION
function
>-)

The Laplace ransform T

Given a function f(t) defined for all t F defined as follows:)

> 0,the Laplacetransform

of

f is the
(1))

F(s)=

\302\243{f(t)}

1 e-

00

st

f(t) dt)

for all values of s for which the improper integral converges.)

Recall that an improperntegralover an infinite interval is definedas a limit i of integrals over boundedintervals; that is,)

1
a

00

g(t) dt =

lim
b\037oo

l
a)

g(t) dt.

(2))

then otherwise, (2)exists, we say that the improper integral converges; or fails to exist. ote that the integrand of the improper integral in (1) N diverges contains the parameter s in addition to the variableof integration t. Therefore,when the integral in (1)converges,t converges not merely to a number, but to afunction i F of As in the following examples,it is typical for the improper integral in the definition of {f(t)}to converge for somevalues of s and diverge for others.)

If the limit
it

in

s.

\302\243

Example

1)

With

f (t) = 1 for t > 0,the definition of the Laplacetransform


\302\243{l}

in

(1)gives)
\037

st roo e- dt Jo

OO

_\037e-st

]0

lim
b\037oo

_\037e-bS

S)

],
(3))

and therefore) 1
\302\243{1}=s)

for

s>O.

in (3), it's goodpracticeto specify the domain of the Laplacetransform-in Also, problemsas well as in examples. in this computation we have usedthe com-

As

mon abbreviation)

[g(t)]a =

OO

lim
b\037oo

[g(t)]b.
a)

(4)

.)
\302\243{1}

Remark:The limit we computed in Example1 would not existif s < 0, bS for then (l/s)e- would becomenbounded as b H u +00. ence is defined > O. This is typical of Laplacetransforms; the domain of a transform is only for s normally of the form s > a for somenumber
\037

a.

.)

Example2)

With

f(t) = eat for t


at
\302\243{e

> 0,we obtain)


....\037

...

}=

1
o

00

e-ste dt =
at

1
0

00

e d e-(s-a)t t = _

-(s-a)t 00

-a ]

t =0)))

268

4 T Chapter Laplaceransform Methods) If s a > 0,then e-(s-a)t

\037

0 as t
at

\037

s +00,o it follows that)


1)

\037)
\302\243{e

} =)

s-a)
i{3t

for
\302\243

> a.)

(5))

Note here that the improper integral giving


noting

= a + if3,)
\037

also that the formula

in

It is worth {eat}diverges if s < (5)holds if a is a complexnumber. For then, with

a.

as t

+00, rovided that p

e-(s-a)t=e e-(s-a)t 0 s > a = Re[a];ecall that ei = cosf3t + i sinf3t. r


\037) {3t
\302\243{t

a of a The Laplacetransform } powerfunction is most conveniently expressedin terms of the gammafunction rex), which is defined for x > 0 by the
formula)
\037)

r(x) =

00

x e-tt - dt.)
1

(6))
function in

For an elementary discussion r (x), seethe subsection the gamma of on Section where it is shown that) 3.5,

r(l) =
and
that)

1)

(7))

rex + 1) = xr(x))
for x > O. It then follows that if n is a positive integer, then)

(8))

r(n + 1) = nr(n) = n . (n l)r(n 1) = n . (n 1). (n 2)r(n = n(n = n(n


thus)
\037)

- - - - 1)(n- 2)...2. r(2) - 1)(n- 2) ...2. . r(I);)


2))

r(n + 1) = n!)

(9))

if n is a positive integer. Therefore, the function rex + 1), hich is defined and w continuous for all x > with the factorial function for x = n, a positive agrees

-1,
......

integer.)
\037

Example3)

Supposethat

Then) f (t) = t a where a is real and a > -1.


..n
.....

a
\302\243{t

}=

1 e-stt

00

a dt.)
this

If we substitute

= st,t = ujs, and dt = dujs in


a
\302\243{t

integral, we get)
sa+l)
(10))))

}=

\037 [00 a du = e-uu


sa+l Jo

f(a + 1)

for all s > 0 (sothat integer, we seethat)


\037)

4.1Laplaceransforms and InverseTransforms 269) T u = st > 0). Because r(n + 1) = n! if n is a nonnegative

Lt n } {
L{t}=

=n!
sn+l) s)

for s

> o.)
3 = L{t} 4. s)

(11))

For instance,)

2'
s)

2 L{t } =

2 3'

and

As

in

nition,

Problems1 and 2,theseformulas can be derived immediately without the use of the gamma function.

from the

defi-

.)

ofTransforms Linearity
It is not necessaryfor us to proceed uch further in the computation of Laplace m transforms directly from the definition. Once we know the Laplacetransforms of several functions, we can combine them to obtain transforms of otherfunctions.The reasonis that the Laplacetransformation is a linearoperation.)

THEOREM 1 LinearityoftheLaplace ransform T If a and b are constants, then)


\037)

+ L{af(t)+ bg(t)}= aL{f(t)} bL{g(t)})

(12))

for all s such that the Laplacetransforms of the functions

f and g both exist.)


linearity

The proof of Theorem 1 follows immediately from the tions of taking limits and of integration:)
\302\243{af(t)

of the opera-

+ bg(t)}=
=

1 ec-+oo 10)
lirn lirn

00

st

[af(t) + bg(t)]dt
st

=a

t et ( 10ec-+oo

[af(t) + bg(t)] dt
st

f(t) dt

+b

c-+oo 10)

lirn

t e-

st

get)dt

= aL{f(t)} bL{g(t)}.) + Example4)


n The computation of L{t / 2} is basedon the known specialvalue)

r
5
3 3

()
\037

= 0T)

(13))

of the gamma function. For instance, it follows that)

()

2222224'))) () ()
=

\037r

\037

0T

270

4 T Chapter Laplaceransform Methods)


using the formula rex 1) = xr(x) in (9),first Now the formulas in (10)through (12)yield)

with

x=

and
\037

then with

x = 4.

2
\302\243{3t

2' 4r ( 2 6 + 4t 3/ 2} = 3 . s3 + s5/2) = s3 + 3
\037

\037

- Hs -.
s5)
\".,,\037 N\037\"\"\"\" '\" ,..,,,'\037\037\"\" ,.., \037....'\"

.)
...............,..............,...\"....;\037...............\037\037......,'-\"\"-\"\"'\"

Example5)

Recall that coshkt = (ekt together give)


\302\243{coshkt}

\"..,

..,

'\"

.....n....;...'....\"N\"

'\"

................\"

VN,';,'N.V

\037\"\"'....;....;..'N..'........

+ e-kt )/2.If k > 0,then


,.,,\037 ...

....nnn

\"\"

\037_....\037

.V',','\",.\"

,....\".,....\".,....\".

'\"

,\037'\"

\".

........

...

\037....\"

\".

\037,nnn.\"

\037\037,....\".

....NU.N....\",....\".,....\".

\037

\"',....\".

__\037........

Theorem 1 and Example2

1
_\302\243{e

kt

+2

kt

_\302\243{e-

}=

1 1 1 - s-k+ s+k) ; 2( )

that is,)

\302\243{coshkt}

s s2
2 -k)

for s

> k > O.

(14))

Similarly,)

.
\302\243{slnhkt}

s2

2 -k)

for s
in

> k > O.
yields)

(15))

Because kt = (eikt + e-ikt )/2,the formula cos


\302\243

(5)(with a = ik)

kt {cos } =

1 1 - s -1 ik + s +1 ik = - . S2-2s 2( ) 2 (ik)2

')

and

thus)

\302\243{coskt}

s s 2 +k)2

fors > O.

(16))

(The domain follows from s

> Re[ik]= 0.) Similarly,)


=
k

\302\243{sinkt}

s2 +k)2

for s

> O.

(17)

.)
....

Example6)

..
n....

...,\037,.'N.\037\"

\"'..d.........,...

,,\037

\"\"\"......,

..\"\"

..\"'.'N.\"\"...,....\".,....\".\"\"

.n..........\037__,....\".

....

......\037__.\"...N..\"\"..n,....\".,....\".\"\"\"...,...

...,....\".'......'..........,.........\"......,....\".,....\".

......

..

....

\"\"'\037,....\".,........,

\"\"...,.,...,...,...\"...,.,,\037...,.,......

....,..................................\"

_,....\".\037

..........\",.......,,\"\"....

\"'\"

\"\"....\037....

\"....................

Applying

linearity,

the formula
2t

in

and (16), a familiar trigonometricidentity,

we get)

\302\243{3e

2 + 2sin3t} =

3 = - + - s s 2 s s2 + 36 3s + 144s- 72 for s > s(s- 2)(s2+


2t
\302\243{3e

+ 1 cos6t}
1

O.)))

36))

.)

4.1Laplaceransforms and InverseTransforms 271) T Inverse Transforms)


According to Theorem 3 of this section,no two different functions that are both continuous for all t > 0 can have the sameLaplacetransform. Thus if F(s) is the transform of somecontinuous function f(t), then f(t) is uniquely determined.This observationallows us to make the following definition:If F(s) = \302\243{f(t)}, then we call f(t) the inverseLaplace transform of F(s) and write)
\037)

f(t) = \302\243-I{F(s)}.)

(18))

Example7)

2, Usingthe Laplacetransforms derived in Examples 3, and 5 we seethat)


\302\243

-I

{s3 }

- =-t 2')
I 1 2

\302\243

-I

{s+2) }

= e-2t ,

2
\302\243-1

{ +9 }
s2

2
3)

sin 3t

and

so on.)

.)

E{s))
1) 1)

s)
1) t)

S2) t n (n
a

> 0))

n!
sn+l)

(s > 0))

NOTATION: UNCTIONS AND THEIR TRANSFORMS. Throughout this chapter F we denotefunctions of t by lowercase letters. The transform of a function will alT is transform ways bedenotedby that sameletter capitalized. hus F(s) the Laplace (s > 0)) of (t) and x(t) is the inverse Laplacetransform of X (s). A table of Laplacetransforms servesa purposesimilar to that of a table of lists integrals.The table in Fig.4.1.2 the transforms derived in this section;many (s > 0)) additional transforms can be derived from these few, using various general properties of the Laplacetransformation (which we will discussin subsequent ections).) s

rea+ 1) (s > 0)) (a > -1)) sa+l)


1)

Piecewise ontinuous C Functions)


As we remarked at the beginning of this

eat)

s-a)
S
t)

section,we needto beableto handlecertain of discontinuousfunctions. The function c (t) is saidto be piecewiseontintypes > 0)) uous on the boundedinterval a < t < b provided that [a,b] can be subdividedinto (s finitely many abutting subintervals in such a way that)

cosk
sin

S2+ k
k

2)

(s > 0)) (s > 0)) (s > Ikl)) (s > Ikl)) (s > 0))

kt)

1.f is continuous in the interior of eachof thesesubintervals;and 2. f (t) has a finite limit as t approaches endpoint of each subinterval each
its
interior.)

from

coshk
sinh k

t)

t)

S2k

S2+ k S S2 k
k

2)

2)

2)

u (t

a))

es)

as)

is piecewiseontinuous for t > 0 if it is piecewise c continuous on say that c + T every boundedsubinterval of [0, (0). hus a piecewiseontinuous function has only simplediscontinuities (if any) and only at isolatedpoints. At such points the value of the function experiences a finite jump, as indicated in Fig. The jump in (t) at the point e is defined to be (e+) (e-), where)
We

-f

4.1.3.

f(e+)=
unit

lim
E-+O+

f(e+ E)

and

f(e-) =

lim
E-+O+)

f(e- E).

FIGURE4.1.2. short table of A transforms.) Laplace

c Perhaps the simplest iecewiseontinuous (but discontinuous)function is the p It is defined as follows:) function, whosegraph appearsin Fig. step

4.1.4.

u(t) =

0 fort 1 for t

> o.)

<0,

(19))))

272
y)

4 T Chapter Laplaceransform Methods)

v)
.........--)

.)

(0,
b
x)

u(t))
1))

(a, 1).)
t

Ua(t)

= u(t

- a)

a)

= a)

FIGURE4.1.3. graph of a The c piecewiseontinuous function; the solid dots indicate values of the
function at discontinuities.)

FIGURE4.1.4. graph of the The


unit

The FIGURE4.1.5. unit step


function
Ua

step function.)

(t) has a jump at t

= a.)

Because(t) = u

1 for

values of a function

> 0 and because Laplacetransform the for t > 0,we seeimmediately that)
t
\302\243{u(t)}

involves only the

1 s)

(s > 0).

(20))
in

The graph of the unit stepfunction ua(t) = u(t occurs t = a rather than at t = 0;equivalently,) at ua(t)

- a) appears
< a, > a.)

Its Fig.4.1.5.jump

= u(t

- a) =

0 for t 1 fort

(21))

Example8)
Solution

Find
We

\302\243{u

a (t)}if a

> O.)
We obtain)
00 00

beginwith the definition of the Laplacetransform.


\302\243{Ua(t)}

1
o

e-stua(t)dt= =

consequentl y,)
\302\243{Ua(t)}

e-as
S)

1
a

e-st dt =

lim
b\037oo

-st -=-S

t=a)

(S > 0, a >

0).

(22)

.)

General roperties fTransforms o P


It

is a familiar fact from calculusthat the integral)

l
c piecewiseontinuous for t
b

g(t) dt)
the
dt)

if exists if g is piecewiseontinuous on the boundedinterval [a,b]. Hence c

> 0,it follows that


st

J is

integral)

exists for all b


b
\037

+oo-to we needsomecondition to limit exist,

< +00.But

in

order for F (s

l e- !(t))-the

limit of this last integral the rate of growth of

as
as)))

J(t)

\037

existnonnegativeconstants M, c, and T such that)

4.1Laplaceransforms and InverseTransforms 273) T T if order +00. he function is saidto be of exponential as t +00 there

\037

< I/(t)! Mect


Thus a function
t
\037

for t

> T.)

(23))

is of exponential orderprovided that it grows no more rapidly (as a constant multiple of some exponential function with a linear exponent. The particular values of M, c, and T are not so important. What is important is that somesuch values existso that the condition in (23)is satisfied. ct The condition in (23)merely says that I(t)/e liesbetween -M and M and is therefore boundedin value for t sufficiently large.In particular, this is true (with c = 0) if I(t) itself is bounded.Thus every boundedfunction-such cosktor as sin kt-is exponential order. of t If pet) is a polynomial, then the familiar fact that p(t)e- 0 as t --+ +00 = c = 1.Thus every impliesthat (23)holds (for T sufficiently large) with M polynomial function is of exponential order. For an exampleof an elementary function that is continuous and therefore boundedon every (finite) interval, but neverthelessis not of exponentialorder, considerthe function I(t) = et2 = exp(t2). Whatever the value of c, we seethat)

+(0)than

\037

because2 t

- ct

t-+oo
\037

hm

I e
(t)
ct

=
\037

et 2 = . t 2_ct = hm e +00 t-+oo ect t-+oo)


hm

+00as t

+00.Hencethe condition in (23)cannot hold


function

for any (finite) value M, so we concludethat the exponential order.


Similarly, gral
that

(t)

= et2 is not

of

e Laplacetransform. The followingtheoremguarantees that piecewise functions of exponential orderdo have Laplace transforms.)
the function

10

00

e-stet2 dt

e st because - et2

\037

as +00 t

that would define t2 doesnot havea

w +00, e seethat the improperintet2 L{e } doesnot exist(for any s), and therefore
\037

THEOREM2 If the function


t

t if transform +00,hen its is piecewiseontinuous and satisfiesthe condition in (23),hen F(s) existsfor c t all s > c.)

\037

> 0 is of exponentialorder as IisLaplace continuousfor= L{/(t)}xists.Moreprecisely, piecewise e F(s)


t

Existence Laplace ransforms of T


and

(23). piecewise I(t)! is bounded< M[0,T].IncreasingM (23)> necessary,we can therefore assume e for > 0, 0 < < T. Because I < Me for > o. follows I a standardtheorem convergenceof improper integrals-the fact abthat continuity,

Proof:First
!
that
I

we note
I

we can take T
if
t

on

in in

For by
t

if

that
I

(t)
ct

ct

it then

(t)

all

By

on

that

solute convergenceimpliesconvergence-it sufficesfor us to prove that the integral)

existsfor s

> c.To do this, it sufficesin turn

00

le-stf (t) dt)


I

to show that the value of the

integral)

le-st(t)1 f

dt)))

274

4 T Chapter Laplaceransform Methods)


remains boundedas b
that)
\037

B +00. ut

the fact that


b

J (t) < Mect for all t > 0 implies


I

l
if

le-st(t)1dt < f

M le-st ect dt = M
I

<M
s

> c.This proves Theorem 2.)


We

1
o
00

00

M e-(s-c)t t = d

e-(s-c)tt) d

s-c)

.)
(24))

have shown, moreover, that)

IF(s)1<
if s

> c.When

1
o

M le-st (t)1 < J dt

s-c

we take limits as s

\037

w +00, e get the following result.)

COROLLARY F(s) for s Large


If J (t) satisfies the hypotheses Theorem 2,then) of
lim
s\037oo)

F(s)= O.

(25))

The condition in (25)severely limits the functions that can be Laplacetransforms. For instance,the function G(s) = s / (s + 1)cannot be the Laplacetransform of any \"reasonable\"unction because limit as s f its +00is not O. Moregenbe of two polynomials-can (and is, as we erally, a rational function-a quotient than that of of shall see)a Laplacetransform only if the degree its numerator is less its denominator. On the other hand, the hypotheses Theorem 2 are sufficient, but not necesof of sary, conditions for existence the Laplacetransform of J(t).For example,the function J(t) = c fails to be piecewiseontinuous (at t = 0), but nevertheless > its Laplacetransform (Example3 with a =
\037

1,

1/,J!

-! -1)
\302\243{t-

1/2} =

r (1)= (if
s 1/2
Y

--;

existsand violates the condition in (24),which would imply that s F (s)remains boundedas s +00. The remainder of this chapter is devoted largely to techniques for solving a differenti\037l equation by first finding the Laplacetransform of its solution. It is then
both
\037

determines the solution of the differential we have found has only one inverse Laplace transform that couldbe the desiredsolution. The following theorem is proved in Chapter 6 of Churchill'sOperationalMathematics, 3rd ed.(New York: McGrawfor us to know that equation; that is, that the
vital
this uniquely function of s

Hill, 1972).)
THEOREM 3

T of Laplace ransforms UniquenessInverse

of Supposethat the functions J(t) and get) satisfy the hypotheses Theorem 2, so that their Laplacetransforms F(s)and G(s) both exist.If F(s)= G(s) for all s > c (for somec), then J(t) = get) whereveron [0, (0) + both J and g are
continuous.)))

T 4.1Laplaceransforms and InverseTransforms 275)

integrals in his work on probability theory. The so-called operational methodsfor differential equations, which are basedon Laplacetransforms, were not exsolving I ploited by Laplace.ndeed,they werediscoveredand popularized by practicing O the (1850-1925 engineer liver Heaviside engineers-notably Englishelectrical Thesetechniques were successfully widely appliedbefore they had been rigand orously justified, and around the beginning of the twentieth century their validity was the subjectof considerable controversy. One reason is that Heavisideblithely assumedthe existencef functions whoseLaplacetransforms contradict the condio tion that F (s) 0 ass 0,therebyraising questionsas to the meaning and nature of functions in mathematics. (This is reminiscent of the way Leibniz two centuries earlier obtained correct had resultsin calculususing \"infinitely small\" real numbers, of as to the nature and role numbers in mathematics.)) thereby raising questions
\037 \037

Laplacetransform can differ only at their isolatedpoints of discontinuity. This is of no importance in most practical applications, o we may regard inverse Laplace s transforms as beingessentially unique. In particular, two solutionsof a differential must be the same solution if they have equation must both be continuous, and hence the sameLaplacetransform. Historical emark: Laplacetransforms have an interesting history. R The integral in the definition of the Laplacetransform probably appeared first in the work of Euler. It is customary in mathematics to name a technique or theorem for the next personafter Euler to discoverit (else there would be several hundred In this case, he next personwas the t different examplesof \"Euler's theorem\.") who employedsuch French mathematician Pierre Simonde Laplace(1749-1827),

Thus two piecewiseontinuous functions of exponential order with the same c

11II\" \"Problems)
to find directly the LaplacetransApply the definition in the Junctions described Jormula or graph) in ProbJorms oJ (by lems through 10.) t t2

(1)

9.) (1,1))
t

1.J(t) =

2.J(t) =

3. J(t) = e3t+
5. J(t) = sinht
7.)

4. J(t) = cost

FIGURE4.1.8.

6.J(t) = sin 2 t
(1,1))

10.
(0, 1))

(1,0)

t)

FIGURE4.1.9.)
t)

FIGURE4.1.6.)
8.)

ojthe Junctions

Usethe

(1,1) (2, 1) .)
o) c)
t)

Problems through 22. A preliminary tegration by parts may be necessary.) 5 2 4t 3 (t) = 3t / J(t) = + 3t 3/2 3t (t) = t J(t) = t 2e
in

transJorms

in

Fig.

4.1.2find to

11

the

LaplacetransJorms
in-

11.
13.

FIGURE4.1.7.)))

2 17.J(t) = cos 2t 19.J(t) = (1+ t)3 21. (t) = t cos2t) J

15.(t) = 1 + cosh5t J

0 -

22. (t) = sinh 2 J

20. J(t) = tet

12. J - e14. J 16. (t) = sin 2t + cos2t J 18. (t) = sin 3t cos3t J
lOt

3t)

276
Usethe

4 T Chapter Laplaceransform Methods) to transforms in Fig. 4.1.2 find the inverse Laplace
of the functions
3
in

transforms

23 Problems through 32.)

Apply the

geometric seriesto
\302\243{f(t)}

obtain the result)

23.F(s)= 4 s
1 25.F(s)= S

24. F(s)= S-3/2) 2


S)

-1e-) . s(l
S

5/2

26. F(s)= 28. F(s)=

s+5
3s + 1 s2 +4

40. (a) The graph of the


Show that

function 00

f can be written
n=O)

f is shown

in

Fig.

4.1.10.

in the form)

27. F(s)=

s-4

29.F(s)= 5 +3s S2 9

u f(t) = L(-I)n (t n).

10s- 3 31. (s)= 25 -s) F 2

9+s 30.F(s)= 4-s) 2


32.F(s)= 2s-e1
3s)

(b) Usethe method of Problem 39 to


\302\243{f(t)}

show

that)

s(l+ e-S)
.)

33.Derivethe transform
in the text to

derive the

of f (t)

sin kt by the method formula in

used used

(16).

34. Derivethe

transform

in the text to

derive the

of f(t)

sinh kt by the method formula in

IL
1 Problem40.)

e---o e---o
2

(14).

. .
0
4 3 5

...-...
0
6
function
t)

35.Usethe tabulated

integral)

f
S 36. how

eaxcosbxdx=eax 2 (acosbx+bsinbx)+C a 2 +b)


\302\243

The FIGURE4.1.10. graph of the

of

to obtain

{cost} k +00 but

directly

from the definition

of

the

41.The graph

transform. Laplace
that the function
\037

Fig. 4.1.11. g Express

order as t
t

f(t) = sin(et2 ) is of exponential


that its derivative

of the square-wave function get) is shown in in terms of the function f ofProblem 40 and hencededucethat)
\302\243{g(t)}

is not.)

37.Given a > 0, let f(t) =


> a. First,
clearits
G 38. iven

1 if 0

<

sketch the graph of the function making Then express in terms of unit value at t as step functions to show that \302\243{f(t)}

< a, f(t) = 0 if

+ s(1 e-S)

-e-s

= 1 tanh s
\037

2.)

= a.

0 < a < b, let f (t) b, either t < a or t > b. First, sketch the graph of the function making clear its values at t = a and t = b. Then express in terms of unit step functions to f

f(t) = 0 if

that

= s-l(1 e- ). = 1 if a < t <

f,

IL) e---o) e---o) ...-...)


-1 e---o e---o e---o)
FIGURE Problem 41.) The 4.1.11. graph of the
function
1)

f,

345
t t

6)

t)

show that

\302\243{f(t)}

bs as = s-l(e- e- ).)

of

T 39. heunit

staircase function is defined as follows:)


if

f(t) = n

-1<
to

< n,

1,2,3,...
.)

42. Given constants a and b, define h (t) for t > 0 by)


h(t)

(a) Sketch the graph of f ate.(b) Show that)

seewhy
u(t

its name

is appropri-

if n

f(t) =

L 00
n=O)

{b

if n

-1< -1<

< nand n is odd; . < nand n IS even.)


oneof the preceding prob-

n)

Sketch the graph of h lems to show that)

and apply

for all t > (c) Assume that the Laplacetransform of the infinite series part (b) can be taken termwise (it can).) in

o.

\302\243{h(t)}

s = a + beS

+ s(1 e-

))))

4.2Transformationof Initial

Value

Problems 277)

of III!JTransformation InitialValueProblems)
We now

equation with constant coefficients,such as) axil (t)


with

discussthe application of Laplacetransforms to solve a linear differential

+ bx'(t) + cx(t)

== (t),)

(1))

given initial conditions x(0) == Xo and x'(0) == xb. By the linearity of the Laplacetransformation, we can transform Eq. (1)by separately taking the Laplace transform of each term in the equation. The transformed equation is)
a\302\243{x\"

(t)}+

b\302\243{x

'(t)}+

c\302\243{x(t)}

== \302\243{f(t)};)

(2))

it involves the transforms of the derivativesx' and x\" of the unknown function x(t). The key to the method is Theorem which tells us how to expressthe transform of the derivative of a function in terms of the transform of the function itself.)

1,

y)

THEOREM 1
Continuous function)

Transforms f Derivatives) o

smoothfor t > 0 and Supposethat the function f (t) is continuous and piecewise is pf exponential orderas t --+ +00,o that there existnonnegativeconstantsM, s c, and T suchthat)
b
I I I I I I I I

a
I I I I

I I I I I I I I I

If(t)1 < Mect


Then
\302\243

for t

> T.)

(3))

( {fl t)}existsfor s > c, and)


== L{f'(t)}
s\302\243{f(t)}

y'

\037
I I I I

...)
x

- f(O)

==

sF(s) f(O).)

(4))

: a

I I

\037 .
I I I I I I

I I

\"-J)
Piecewise continuous derivative)

FIGURE4.2.1. The discontinuities of to correspond \"comers\" on the graph of f.)

The function f is calledpiecewisemoothon the bounded interval [a,b] if it s is piecewiseontinuous on [a,b] and differentiable exceptat finitely many points, c with f' (t) being piecewiseontinuous on [a,b]. We may assignarbitrary values c to f (t) at the isolatedpoints at which f is not differentiable. We say that f is s s piecewisemooth for t > 0 if it is piecewisemooth on every bounded subinterval of [0, (0). igure 4.2.1 indicates how \"comers\"on the graph of f correspondo t + F discontinuities in its derivative The main ideaof the proof of Theorem 1 is exhibited bestby the case which in (t) is continuous (not merely piecewiseontinuous) for t > O. Then, beginning c with the definition of (t)}and integrating by parts, we get)

f'

f'.

f'

\302\243{f'

\302\243

[00 ( [00 ( {f't)}= 10 e-sf I' t) dt = [e-s f (t )] + s10) e-sf I(t) dt .


t

00

t=O

Because f (3),the integrated term o


the

a +00,nd its value at the lower limit

2 of Section

preceding expression.

given in Eq. (4). We will defer the has isolateddiscontinuities to the end of this section.)))

> c, and its value is that

the 4.1, integral converges when s > c. Then

t == 0 contributes The integral that remains is simply

e-st f(t)

z approaches ero (when s


\302\243{f(t)}; \302\243{f'(t)}

-f (0)

to the evaluation of
by

> c) as t
which

--+

Theorem

in case

existswhen

f'(t)

278

4 T Chapter Laplaceransform Methods) Solution Initial alue Problems) of V


In

well. If we assumethat get) =


that

order to transform Eq. (1), e need the w


theorem impliesthat)

transform of the secondderivative as of satisfies the hypotheses Theorem then f'(t)

1,

\302\243{f\"(t)}

= = = S\302\243{f'(t)} = s [s\302\243{f(t)}
\302\243{g'(t)}

- f'CO) g(O) - f(O)]- f'CO),)


s\302\243{g(t)}

and
\037)

thus)

\302\243{f\"

(t)}= s2F(s) sf(O)

- f' (0).)

(5))

repetition of this calculation gives)


\302\243{f\"l(t)}

S\302\243{f\"(t)}

- f\"(O) = s3F(s)- s2f(O) - Sf'(O)- f\"(O).

(6))

After finitely

many

such stepswe obtain the following extensionof Theorem 1.)

COROLLARY

Transforms f HigherDerivatives o

Supposethat the functions f, f', f\", .. fCn-I) are continuous and piecewise smooth for t > 0,and that eachof thesefunctions satisfies the conditions in (3) with the samevalues of M and c.Then \302\243{f(n)(t)} existswhen s > c, and)
\302\243{fCn)(t)}
\037)

.,

= sn sn-I(O) sn-2f'CO) f = sn F(s) sn-I(O) sf n-2)(0) f


\302\243{f(t)}

... - ...
C
C

- - f n-l)(o) - f n-l)(o).
C

(7))

Example

1)

Solvethe initial value problem)

....

x-x-x=;)x(O)= 2,
II I

x'(O)= -1.)

Solution

With

the given

initial

values, Eqs.(4) and (5)yield)


\302\243{X'(t)}

s\302\243{x(t)}

and)
\302\243{X\"(t)}

= s2\302\243{x(t)}

- x(O)= sX(s) - sx(O)- x'(O)= s2X(s)- 2s + 1,


2)

where (according our conventionabout notation) X (s) denotesthe Laplacetransto form of the (unknown) function x(t). Hence transformed equation is) the

[S2X(S) 2s + 1] [sX(s) 2]
which we quickly simplify
to)

- - 6 [Xes)]=
3
O.)

0,)

Thus)

- - 6)X(s) - 2s + = 2s - 3 2s Xes) =
(S2 S

s2

-s-)6

(s-3)(s+2))))

4.2Transformationof Initial
By the method of partial fractions (of integral

Value

Problems 279)

B such that)

2s

-3

there calculus), existconstantsA and

(s-3)(s+2)
and multiplication of both

2s - 3 = A(s + 2) + B(s that A

s-3+ s+2,
identity)

sidesof this equation by (s 3) (s +2) yieldsthe


3).)

If we substitute s B = Hence)
\037.

3, we find

= =

\037;

substitution 3 5

of s)

-2 shows

that)

X(s) =

\302\243{x(t)}

Because \302\243-l{l/(s

- a)} = eat,

7 + s+2. s-3 5

it

follows that)
\037e3t

x(t) =

\037e-2t)

is the solution of the original initial value problem.Note that we did not first find the general solution of the differential equation. The Laplacetransform method directly yieldsthe desiredparticular solution, automatically taking into account.) via Theorem 1 and its corollary-the given initial conditions.

Remark:In Example1 we found the values of the partial-fractioncoefficientsA and B by the \"trick\" of separately substituting the roots s = 3 and s = -2 of the original denominator s2 s 6 = (s 3)(s+ 2) into the equation)

-- 2s - 3 = A(s + 2) + B(s -3 =
(A

3))
\"sure-fire\"

resultedfrom clearing fractions. In lieu of any such shortcut, the method is to collect coefficientsof powersof s on the right-hand side,)
that

2s

+ B)s + (2A

3).)

Then upon equating coefficientsof terms of likedegree, e get the linear equations) w
A

2A

B 2, - 3B = -3) =
and
\037

which are readily solved for the


h
\".'\037N\037'h... _\"'''\"'''''''\",,, \"'...,.... ..-,,,....'''' ...\" ..., '\" '.......'N..\"\"\"... ....,..._\"'.... ... ....... \"'... '\"

samevalues A =

B=

\037.)

.)

Example2)

Solvethe initial value problem)

. x + 4x = SIn 3t;) x(O)= x'(0) = O.)


1/

Such a problemarises in the motion of a mass-and-springystem with external s force, as shown in Fig.4.2.2.)))

280

4 T Chapter Laplaceransform Methods)

Solution Becauseoth b
J(t) = sin 3t

initial

the transform of sin 3t from the table in

values are zero, q. (5) yields E 3

transformed equation)

4 and (Section .1) thereby get the Fig.4.1.2

\302\243{x\"

(t)} = s2X (s). We read

s2X(s)+ 4X(s) = s2 +9) .


A FIGURE4.2.2. mass-andspring system satisfying value problem in Example massis initially at rest in its equilibrium position.) the initial

Therefore,)

2.The

X(s)

(s2 + 4)(s2+ 9)

.)

The method of partial fractions callsfor) 3


As

(s2 + 4)(s2+ 9))

+ B + Cs+ D s2 + 9 s2 + 4

.)

The sure-fire approach would be to clearractions by multiplying both sidesby the f common denominator, and then collect coefficientsof powersof s on the right-hand side. quating coefficientsof likepowerson the two sidesof the resulting equation E would then yield four linear equations that we couldsolve for A, B,C,and D. n However, here we can anticipate that A = C = 0,because either the numerator nor the denominatoron the left involves any odd powersof s,whereasnonzero values for A or C would leadto odd-degree terms on the right. Sowe replaceA and C with zero before clearing fractions. The result is the identity) 3 = B(s2+ 9) + D(s2+ 4) = (B + D)s2+ (9B+ 4D).)
When

we equate coefficientsof like powersof s we get the linear equations)

B + D = 0,

9B+4D=3,)
which are readily solved for B = and
\037

D=

-\037.

Hence)

(s) =

\302\243{x(t)}

3
\302\267

10 s2 + 4

- 5 . S23+ 9
1
.)

.)

x)

Because

\302\243{sin

1
2)

+ 2t} = 2/(s2 4) and


x(t) =

\302\243{sin

+ 3t} = 3/(s2 9), it follows that)

to

sin 2t

\037

sin 3t

-2)

The FIGURE4.2.3. position


function

x (t)

in

Example 2.)

showsthe graph of this period2n position function of the mass. Note Figure 4.2.3 that the Laplacetransform method again gives the solution directly, without the of necessity first finding the complementaryfunction and a particular solutionof the original nonhomogeneous differential equation. Thus nonhomogeneousequations are solved in exactly the samemanner as are homogeneousquations. e

.)

4 that Examples1 and 2 illustrate the solution procedure is outlined in Fig. .2.4.)))

4.2Transformationof Initial
Differential

Value

Problems 281)

equation in x(t))

Solution x(t) of differential


equation)

Algebraic equation

Solution Xes)

of algebraic
equation)

in Xes))

FIGURE4.2.4. Using
value problem.)

the

Laplacetransform

to

solvean

initial

Linear Systems
of two or more constant-coefficientlinear differential equations involving two or more unknown functions x(t), yet), of the independent variable t. When initial conditions are specified,he Laplacetransform reducessuch a linear system of dift ferential equations to a linear system of algebraic equations in which the unknowns are the transforms of the solutionfunctions. As Example illustrates,the technique 3 for a system is essentially sameas for a singlelinear differential equation with the constant coefficients.)

Laplacetransforms are frequently used in engineering problemsto solve a system

..

Example3)

Solvethe

system)
2x\" ==
y\"

-6x+ 2y,

== 2x

- 2y + 40

sin 3t,)

(8))

subjectto the initial conditions)

x(0) == x'(0) == y (0) == y' (0) == O.)

(9))

This initial value problemdetermines the indicated displacement functions x(t) and of the two massesshown in Fig.4.2.5, that the force J(t) == 40 sin 3t yet) assuming is suddenly appliedto the secondmass at the time t == 0 when both massesare at rest in their equilibrium positions.)
J(t) = 40sin 3t)

value problem
equilibrium

A FIGURE4.2.5. mass-and-spring system satisfying the initial in Example 3. Both massesare initially at rest in their positions.)

Solution

We write imply
that)

Xes) ==

\302\243{x(t)}

and

yes) ==

\302\243{y(t)}.

Then the

initial

conditionsin (9)
.)))

\302\243

// 2 {X (t )} == S X

(s) and

\302\243

// 2 {y (t )} == S Y (s)

282

4 T Chapter Laplaceransform Methods)

Because
equations)

\302\243{sin3t}

= 3/(s2+ 9), the


2s2X(s) =

transforms of the equations in

(8) are the

+ -6X(s) 2Y(s),

s2y(s) = 2X(s) 2Y(s) + s2 . +9)


Thus the transformed systemis)

120

(s2+3)X(s))

- Y(s) =
X

0,)

+ -2X(s) (s2+ 2)Y(s)= s2 +9) .


The determinant of this pair of linear equations in

120

(10))

(s) and Y (s) is)


4),)

s2_i s2-12= (s2+ 3)(s2+ 2)


and we readily

- 2 = (S2+ 1)(s2+ -

Cramer's for instance-theystem in (10)for) rule, solve-using s 3 120 8 = 5 + s2+9 X(s) = (lla)) s2+1 s2+4 (s2+1)(s2+4)(s2+9)

and)

Y(s) =

18 - s2+9. s2+1 s2+4 (s2+1)(s2+4)(s2+9) + 120(s23)


= 10 +
8

(lIb))

The partial fraction decompositions Eqs. (1Ia)nd (lIb) readily found using in a are the method of Example2. For instance, noting that the denominator factors are linear in s2, we can write)
, s2+1 (s2+1)(s2+4)(s2+9) + s2+4+ s2+9) and
it

120

ABC
-4

follows that) (12))

= 120 A(s2 + 4)(s2+ 9) + B(s2+ 1)(s2+ 9) + C(s2+ 1)(s2+ 4).


y(t))
10)

-10) FIGURE4.2.6. position The


functions

is, s = i,a zero of the factor s2 + 1)in Eq. (12)gives in Eq. (12)yields substitution of s2 = = 3. Thus we obtain the partial B = -8,and substitution of s2 = yields C fraction decompositionhown in Eq. (1Ia). s At any rate, the inverse Laplacetransforms of the expressions Eqs. (11 in a) and ( 11 give the solution) b)
Substitution A

= 120 . 3 . 8, so A = 5. Similarly,

of s2 =

-1

(that

-9

x(t) = 5 sin t
y

-4

sin 2t

(t)

Example 3.)

x(t)and

= 10 t + 4 sin 2t sin

-6

sin 3t, sin 3t


.)

yet)

in

showsthe graphs of thesetwo period2n position functions of the Figure 4.2.6

two

masses.

..)))

4.2Transformationof Initial
The TransformPerspective
J(t))

Value

Problems 283)

Let us regard the general constant-coefficientsecond-orderquation as the equation e


of motion of the familiar equation is)

+ ex'+ kx = f(t) Then the s mass-spring-dashpot ystem (Fig.4.2.7).


mx\"

transformed

FIGURE4.2.7. massA
spring-dashpot system
external
with

[s2X (s) sx(O)

- x'(0)]+ c [s (s) - x(O)]+


X

kX

(s) = F(s).

(13))

forceJ(t).)

Note that Eq. (13)is an algebraic a linear equation-in \"unthe equation-indeed, known\" X (s). This is the source the powerof the Laplace of transform method:) Linear differential equations are transformed into readily solved algebraic equations.) If we solve Eq. (13)for X(s),we get)

X(s) =
where)

F(s)+ I (s)
Z(s)

Z(s)

')

(14))

Z(s) = ms2 + cs + k
\302\243{x(t)}

and

I (s) = mx(O)s+ mx'(0)+ cx(O).)

Note that Z (s) dependsonly on the physical system itself. Thus Eq. (14)presents X (s) = as the sum of a term depending only on the externalforce and one dependingonly on the initial conditions.In the caseof an underdamped system, thesetwo terms are the transforms)
\302\243{xsp(t)}

= F(s)
Z(s))

and

\302\243{Xtr(t)}

= I (s)
Z(s))

of the steady periodicsolution and the transient solution, respectively.The only potential difficulty in finding thesesolutions is in finding the inverseLaplace transform of the right-hand sidein Eq. (14). Much of the remainder of this chapter is devoted to finding Laplacetransforms and inverse transforms. In particular, we seekthose that-unlike methods that are sufficiently powerful to enableus to solve problems thosein Examples1 and 2-cannot solved readily by the methodsof Chapter2.) be

Additional TransformTechniques)
.....

Example4)

Showthat)
\302\243

a {t e t}

(s

a)2)

.
at
s\302\243{te

Solution If f (t) = teat, then


\302\243{e

Theorem 1 gives) f (0) = 0 and f' (t) = eat + at eat. Hence


at

+ ateat } =
at
\302\243{e

\302\243{f'(t)}

= s\302\243{f(t)} =
that)

}.)

It follows from the

linearity

of the transform
}

at
a\302\243{te

}=

at
s\302\243{te
}.)))

284

4 T Chapter Laplaceransform Methods)


Hence)
\302\243

{teat}==

\302\243

because

at

\302\243{e

} == 1/(s a).)

s -a

{ea } ==

(s -

1 a)2) (15))

.)

Example5)

Find

\302\243{t

sin kt}.)

Solution Let f (t) == t sin kt. Then f (0) == 0 and)

f' (t) == sin kt + kt coskt.)


The derivative involves the new function differentiate again. The result is)
t

cos so we note that kt,

f' (0) == 0 and

kt f\" (t) == 2k cos

k 2t sin kt.)

== s2\302\243{f(t)} by the formula in (5) for the transform derivative, and \302\243{coskt} == s/(s2 k 2), so we have)

But

\302\243{f\"(t)}

of the second

2ks s2 +k)2
Finally,

\302\243

k2

\302\243{t

sinkt} ==

s2\302\243{t

sinkt}.

we solve this equation for)


{t sin k t} ==

2ks (s2

+ k2 2
))

(16))
the

This procedureis considerably more pleasant than the alternative of evaluating


integral)

\302\243{t

sin kt}

00

51 te- sin kt dt.)

.)

Examples4 and 5 exploitthe fact that if f (0) == 0,then differentiation of f to the corresponds multiplication of its transform by s. It is reasonableto expect inverse operation of integration (antidifferentiation) to correspond divisionof the to
transform by s.)

THEOREM2

Transforms f Integrals o
I

If f(t) is a piecewiseontinuous function for t > 0 and satisfies the conditionof c ct exponential order If(t) < Me for t > T, then)
\302\243

{10

{tf('e) = d7:
}
\302\243-1

1
\037\302\243{f(t)}

= F(s)
S)

(17))

for s

> c.Equivalently,)
{ F;S) } =

1
1

f(7:)d7:.)

(18))))

4.2Transformationof Initial
culus impliesthat)

Value

Problems 285)
theoremof cal-

continuous, the fundamental Proof:Because is piecewise f

g(t) =

it f('c)dr:)
and
t

s piecewisemooth for t

is continuous and that g'(t) = f(t) where f is continuous; thus g is continuous

> O.Furthermore,)

< Ig(t)1
this gives)

i
o

IfCr)1di < M

i
0

eCTdi = _(ect
C

M - 1) < _e
C)

ct

so get) is of exponential orderas t --+ +00. ence can apply Theorem 1 to g; H we


\302\243{f(t)}

\302\243{g'(t)}

s\302\243{g(t)}

- g(O).)
\302\243{f(t)}

Now

g(O)= 0,so division by s yields)

\302\243

{itf(r:)dr:

\302\243{g(t)}

})

s)

which

the completes proof.)

\037)

Example6)

Find the inverse Laplacetransform of) 1

G(s))

s2(s a)

.)

Solution

In effect, Eq. (18)means that we can deletea factor of s from the denominator, find the inverse transform of the resulting simplerexpression, nd finally integrate from a o to t (to \"correct\"for the missingfactor Thus)

s).

\302\243-1

= t {s(s- a) } 10
1

\302\243-1

dr:= teaTdr:= !(e - 1). 10 {s - a }


1
at
a)

We now

repeat the technique to obtain)

\302\243-1

{s2(s- a) }

i {s(s- a) } i _(eaT - 1)dT - = = e )] a (e at [a (a


=
t

\302\243-1

dT =
1 2

a)

aT

r:

at

1).)

This technique is often a more convenient way than the method of partial fractions .))) for finding an inverse transform of a fraction of the form P(s)j[sn Q(s)].

286

4 T Chapter Laplaceransform Methods)


in the general that the limit)

Proof fTheorem 1: o casein which

c f' is merely piecewiseontinuous. We needto prove


lim
b\037oo

We

concludethis sectionwith the proof of Theorem 1

10)

{be-st (t) dt

I'

existsand alsoneed to find its value. With b fixed, let tl, t2, ..., tk-I be the points interior to the interval [0, ] at which f' is discontinuous. et to == 0 and tk == b. L b Then we can integrate by parts on eachinterval (tn-I,t n ) where f' is continuous.
This yields)
st (b e- (t) dt =

10 o

l'

I: j
k k

tn

n=1 tn-l

e-st I' t) dt ( I(t)t-l


tn
k

=
Now the
k first summation)
t

\037[e-st

+
\037

tn

tn-l

e-st I(t)dt.)

(19))

+ [-e-stlI(tl)+e-stzI(t2)] I:[e-stI(t)r= [-I(to)+e-st1I(tl)] n=1


tn-l

+ ... + [_estk- (tk-2)+ e-stk-l (tk-l)] + [_estk-lI(tk-l) e-stk I(tk)] +


Z

(20))

sb down (19)telescopes to f(to) + e-stk f(tk) == f(O) + e- f(b),and the second summation addsup to s times the integral from to == 0 to tk == b.Therefore (19) reduceso) b t b e-st (t)dt = 1(0) e-sb I(b)+ S e-st I(t)dt. + in

I'

But from

Eq. (3)we get)


sb le- f(b)1 < e-sb . Mecb == Me-b(s-c)-+ 0)

> c.Therefore, finally taking limits equation, we get the desiredresult)


if s
\302\243{f'(t)}

(with

in s fixed) as b --+ +00 the preceding

== s\302\243{f(t)}

- f(O).)

\037)

Extensionf Theorem o
1)

c supposethat the function f is only piecewiseontinuous (instead of continuand let tl , t2, t3, . be the points (for t > 0) where either f or f' is discontinous), uous.The fact that f is piecewiseontinuous includesthe assumption that-within c eachinterval [tn-I,t n ] betweensuccessive with a f points of discontinuity- agrees function that is continuous on the whole closed interval and has \"endpoint values\
Now

..
I

f(t:_) == t lim t+
---'>..

--'n-l

f(t)

and

f(t;)== lim
t \037;;) t

f(t)

not agreewith the actual values f(tn-I) and f(tn ). The value of an inon an interval is not affected by changing the values of the integrand at tegral
that may
the)))

4.2Transformationof Initial
endpoints.However, if the

Value

Problems 287)

fundamental theorem of calculusis applied to find the value of the integral, then the anti derivative function must be continuous on the closed interval. We therefore use the \"continuous from within the interval\" endThe values above in evaluating (by parts) the integrals on the right in (19). point result is)

eL[ st 1(t)
n=l

r = [-1(tt) + e- 1 1 ] + [_e- I(t{)+ e- l(t;)]


t

st ,

tn-l

(t )

st

\\

st2

+ e-stk-Il(tk-I)] + ... + [_estk-2 + + [_estk-Il(tL1) e-stk l(tk)]


1\302\253(L2)

== where)

k-l -f(O+) - Ljj(t


n=l)

n)

+ e-sb f(b-),

(20/)

jj(tn ) == f(t:)- f(t;))


(finite) jump in

(21))
that
\302\243{f'(t)}

denotesthe

therefore get the generalization)


\302\243{f'(t)}

f(t)

at

t == tn.

Assuming
00

exists,we

==

sF(s) f(O+) - Le- jj(t


stn n=l)

n)

(22))

of \302\243{f'(t)}

==

sF(s) f(O) when we now take the limit


unit

in

(19)as b
==

\037

+00.)

Example7)
6
5

Let f (t) == 1 + [t]be the


Then

J(t)

4
3

2
1)

.....-0 .....-0 .....-0

... .... .....-0


t)

f(O) ==

Eq. (22)yields)

1,f'(t)= 0,and jj(n)== 1 for each integer n

staircasefunction; its graph is shown


00 ns

.... 1,2,3,
in

Fig.4.2.8.
Hence

0==sF(s) - Le1
n=l)

so the Laplacetransform of f(t)

is)

123456
x == e-s <

F(s)

==

1 - Les
00
n=O

ns ==

s(1 e-

S
))

The FIGURE4.2.8. graph of the unit staircase function of Example


7.)

In the last step we usedthe formula for the sum of a geometric series,)
00

Lx
n=O)

1)

I-x),

with

1.)

.)
3.x\" - X

lIB Problems)
UseLaplacetransforms
Problems

1through 16.

to

solvethe

initial

value problems in

2.x\" + 9x = 0; x(O)= 3,x'(O)= 4)

1.

x\"

+ 4x = 0;x(O)= 5,x'(O)= 0

4. x\" + 8X' + 15x= 0;x(O)= 2,x'(O)= -3 5. x\" + x = sin 2t; x(O)= 0 = x'(O) 6.x\" + 4x = cost; x(O)= 0 = x'(O) 7. x\" + x = cos3t; x(O)= 1, '(O)= x
0)))

2x = 0; x (0) = 0, x'(0) = 2

288

4 T Chapter Laplaceransform Methods) 34. If f(t) =


graph

x 8.x\" + 9x = 1; (0) = 0 = x'(0) 9.x\" + 4X' + 3x = 1; (O)= 0 = x'(O) x x 10.\" + 3X' + 2x = t; x(O)= 0, x'(O)= 2 x' 11. = 2x + y, y' = 6x + 3y; x(O)= 1,y(O) = -2 x 12. ' = x +2y, y' = X + e-t ; x(O)= y(O) = 0 x 13. '+ 2y' + x = 0, x' y' + y = 0;x(O)= 0, y(O) = 1 x 14.\" + 2x + 4y = 0, y\" + x + 2y = 0; x(O)= y(O) = 0, x'(0) = y' (0) = -1 15.\" + x' + y' + 2x y = 0, y\" + x' + y' + 4x - 2y = 0; x x(O)= y(O) = 1, '(O)= y'(O)= 0 x x 16. ' = x+z,y' = x+y, z' = -2x-z;x(O)= 1, (O) = 0, y

is shown

i (-l)[t]s the
in

Fig.

then) 4.2.9,

square-wave function

whose

\302\243{f(t)}

1 = - tanh s

-.
2)
.)

(Suggestion:Usethe geometric series.))


J(t)
1

(0) = 0)

t -1 ........0........0........0)

123456

.. ........0........0....

Apply Theorem 2 to find the inverse Laplacetransforms through 24. functions in Problems

17

of the

The FIGURE4.2.9. graph of the w square- ave function of Problem 34.)

17.F(s)=

s(s- 3)
1

19. (s)= F 23.F(s)=


25.Apply

s(s2+ 4)
1 1 S2(S2

F 21. (s) = s2(s2 + 1)

18. (s)= 3 F s(s+ 5) 20.F(s)= 2s + 1 + S(S2 9) 1 22.F (s) = s 2 (s


9)
1

35.If f (t) is the unit


Fig.
then) 4.2.10,

on-offunction whose graph is shown f

in

\302\243{f(t)}

s(l+ e-S)
.) .)

Theorem to derive 1
kt}. kt}.

- 1)

24. F(s)=
\302\243

J(t)
1 1

s(s+ l)(s + 2)
the formula

.. ........0........0....
2
3

{sin kt} from the formula for

5)

26. pply A
for

\302\243{cos

Theorem1 to derive {cosh from kt}


\302\243

\302\243{sinh

The FIGURE4.2.10. graph of the on-offfunction of Problem 35.

27. (a)

Apply

Theorem1 to
at
\302\243{tne

show that)

=)
}

s-a)

\302\243{t

- eat }.
1

36.If get) is the


shown in

then) Fig. 4.2.11,

triangular

wave function

whose graph is

(b) Deducethat

at
\302\243{tne

3,
Apply

...

= n!j(s

a)n+1 for n

= 1,2,

\302\243{g(t)}

1 = 2 tanh s

-.
2)
t)

.)

Theorem
in

forms

Problems 28 through 30.

1as in Example 5 to derive the Laplacetrans=


S2 k 2 (s2 + k 2)2 = 2ks 2 (s2_ k )2 + 2 = S2_ k2 (S2 k )2

get))

28. 29.

\302\243

{t cos kt} sinh kt}

456

\302\243{t

The FIGURE4.2.11. graph of the triangular wave function of Problem 36.)

30.

\302\243{t

coshkt}

31.pply A
that)

37. If f(t) is the


5 and Problem 28 to 1
show

the results in Example

Fig.

4.2.12,
then)

sawtooth function

whose graph is shown s

in

\302\243

-1

I
2 + k2 2 ) }

{ (s

. = _ (sIn kt - kt coskt).
2k)3

\302\243{f(t)}

1 =2

- e- . s s (1- -S)
e)

Apply

the extension

Laplacetransforms
\302\243{u(t

for a > O. 32. 33.If f(t) = 1 on the interval [a, b] (where 0 < a < b) and f (t) = 0 otherwise, then)
as
\302\243{f(t)}

- a)}= S-le-

of Theorem 1 in Eq. (22) to derive the 32 given in Problems through 37.

(Suggestion:Note that
J(t))
1)

f'(t) = 1 where

it

is defined.))

=e

-as e-bs
s)

456
37.)))

t)

The FIGURE4.2.12. graph of the sawtooth function of Problem

4.3Translation and Partial Fractions 289)

11I1,

TraJ?-sl\037ltio\037

and

\037ar!\037\037I}:'\037\037\037\037.!\037ol1\037.w

..... ._
.

..)

the 4.2, solutionof a linear differential equation with constant coefficientscan often be reducedto the matter of finding the inverse Laplacetransform of a rational function of the form As illustrated by Examples and 2 of Section 1

R(s) =
where the degreeof

P(s)
Q(s)
of

(1)
finding

mentary calculusto integrate rational functions. The following two rules describe the partial fraction decomposition of R(s),in terms of the factorizationof the denominator Q(s) into linear factors and irreducible quadratic factors corresponding to the real and complex zeros, respectively,of Q (s).)
RULE

is oC-I{R(s)} based on the same method of partial fractions that we use in ele-

pes)is lessthan

that

Q(s).The technique for

1 LinearFactor PartialFractions

The portion of the partial fraction decomposition R(s) corresponding the to of linear factor s a of multiplicity n is a sum of n partial fractions, having the
form)

+...+ , s-a+ (s-a)2 (s-a)n)


Al
A2 An

(2))

where A I,A 2,
RULE

..., and An are constants.)

2 Quadratic actor artialFractions F P

The portion of the partial fraction decompositionorresponding the irreducible to c factor (s a)2 + b 2 of multiplicity n is a sum of n partial fractions, quadratic
having the form)
A 2s + B2 Als + BI + + b2 [(s a)2+ b2]2 (s a)2+

\302\267

\302\267

\302\267

' [(s _ a)2+ b2]n


find

An s

+ Bn

(3))

where AI,

A 2,

..., An, BI,B2, ..., and Bn are constants.)


-I{R(s)}involves two steps.First we must
the partial fraction

Finding oC

o decompositionf R(s),and then we must find the inverse Laplacetransform of each of the individual partial fractions of the types that appearin (2)and (3). The latter stepis basedon the followingelementaryproperty of Laplacetransforms.) THEOREM 1

onthes-Axis Translation at If F(s)= oC{J(t)}xistsfor s > c, then oC{e J(t)} e existsfor s > a + c, and)
>-)

= oC{e J(t)}
at

F(s

- a).)

(4)) (5))

Equivalently,

>)

oC-I{F(s = eatJ(t).) a)}

Thus the translation s -+ s a in the transform correspondso multiplication of t the original function of t by eat.)))

290

4 T Chapter Laplaceransform Methods)


we obtain)

s Proof:If we simply replacewith

-a
00

in

the definition of

F(s)=
at
\302\243{e

\302\243{f(t)},

F(s a) =

1
it

00

e-(s-a)t(t) dt = f

1 e-

st

dt [eatf(t)]

f(t)}.)

This is Eq. (4),and of tn,

is clearhat Eq. (5)is the same.) t

.)

If we apply the translation theorem to the formulasfor the Laplacetransforms coskt, and sin kt that we already know-multiplyingeachof thesefunctions eat and replacing with s a in the transforms-we the following additions s by get to the table in Fig. .1.2.) 4

eat t n

eat

coskt

eat sin kt

-n! (s -s-a (s a)2 + (s -a)2 +


a)n+l k

(s > a)
k2 k2

(6)
(7)

(s > a) (s > a))

(8)

For ready reference, ll the Laplacetransforms derived in a listedin the table of transforms that appearsin the endpapers.)
\"_.\"

this

chapter are

Example

1) --\"-Co\037\037id\037;-;'m\037_\037\037\037=\037\037-d=\037p;i\037g--\037y-\037t\037\037'O;ith\"';;;
'''\037''\037'''k

o As (Fig.4.3.1). usual, let x(t) denote the displacement f the mass m from its f equilibrium position.If the massis set in motion with x(0) = 3 and x/ (0) = 1,ind x(t) for the resulting dampedfree oscillations.) value problem)

= 17,and c = 3 in

mks units

Solution The differential equation is !x\"+ 3x'+ 17x=


k = 17)
x\"

0,so we need to solve the initial

+ 6x'+ 34x = 0;

x(O)= 3, x/CO)= 1.)

The FIGURE4.3.1. massspring-dashpot system of Example


1.)
x) 3)

We

take the Laplacetransform of eachterm of the differential equation. Because = 0,we get the equation) (obviously)
\302\243{O}

[S2X(S) 3s
which we solve for)

- - 1]+6[sX(s)3] + 34X(s)=

0,)

2)

X(s) =
Applying
t)

1)

5 s +3 =3.(S+3)2+2S 2 . (S+3)2+2S + s2+6s+34 the formulas (7)and (8) a = -3and k = 5,we now see
.)

3s + 19
in

with

that)

1t
\"2)

FIGURE4.3.2. position The


function

x(t) = e-3t (3 cos5t + 2 sin 5t)

.)

x(t)in

Example 1.)

showsthe graph of this rapidly decaying dampedoscillation.) Figure 4.3.2

.)))

4.3Translation and Partial Fractions 291)


Example2 illustrates a useful technique for finding the partial fraction coefficientsin the case nonrepeated linear factors.) of

Example2)

Find the inverse Laplacetransform

of)

R(s) = s3 2s2 Solution Note that

- -

S2+ 1

8s)

.
s(s+ 2)(s- 4). Hence)
4)

the denominatorof R(s)factors as Q(s) = 1

ABC -S2+ - 8s =-+ s + 2 + s - . s3 2s2 s


+ 2)(s
and

Multiplication of eachterm of this equation by Q(s)yields) S2+ 1 = A(s


When

- 4) + Bs(s- 4) + Cs(s+
24C = 17.)

2).)

we successively substitute the three zeross = denominator Q(s)in this equation, we get the results)

0,s = -2,and s = 4 of the

= -8A= 1, 12B 5,
Thus A

-!, =
B

5
1

2 and S2

'

C=
1

\037\037

, so)

+ = --+ + s s +2 s s3 - 2s2 - 8s
\0372 \037\037

4)

and therefore)
\302\243

-1

5 + - 2s2- 8s } = -- + -e + -e . 8 12 {s3
S2 1 1 _ 2t 17
4t

24)

.)
fraction

Example3 illustratesa differentiation technique for finding the partial coefficientsin the case repeatedlinear factors.) of
................

Example3)

Solvethe initial value problem)


y\"

+ 4y' + 4y = (2;

y(O) = y/(O) = O.)

Solution The transformed equation is) s2y(s)+ 4sY(s) + 4Y(s) = 3\". s)


Thus)

= + + yes) = s3 (s + 2)2 s3 s2 S
To find A,

ABC - - -+

D (s + 2)2

+.
E s + 2)

(9)

B,and C,we multiply


2

both

sidesby s3 to obtain)
(10))))

(s + 2))2

=A+Bs+Cs2+s3F(s),

292

4 T Chapter Laplaceransform Methods) is + F(s)= D(s+ 2)-2 E(s + 2)-1 the sum of the two partial fractions to corresponding (s + 2)2.Substitution of s = 0 in Eq. (10)yieldsA = 4. To find BandC,we differentiate Eq. (10)twice to obtain)
where

-4
(s + 2))
and) 3

2 = B+2Cs+3s F(s)+s3F'(s)

(11))

12 = 2 2C+6sF(s)+6s F'(s)+s3F\"(s). (s + 2))4


Now substitution of s
\037.

(12))

Eq. (12)yieldsC =
To find D and

= 0 in Eq. (11) B = yields

-4,and

substitution

of s

= 0 in

E, we multiply eachsidein Eq. (9)by (s + 2)2 to get)


2
3\" s)

= D + E(s+ 2) + (s + 2)2G(s),
then

(13))

1 2 where G(s) = As-3 + Bs- + Cs-, and

differentiate to obtain) (14))


and

6 -4 = E + 2(s + 2)G(s)+ (s + 2)2G'(s).


s)

Substitution of s
Thus)

-2

in

Eqs. (13)and (14)now yields D =

-\037

E=

-\037.

4 _ + yes) = -1...-1...S _ s3 s2 (s + 2)2 so the solution of the given initial value problemis)
\037

3 -

3
8

+ 2')

yet)

\037t2

4t

\037

\037te-2t

\037e-2t.)

.)

4, Examples 5,and 6 illustrate techniques for dealing with quadratic factors in partial fraction decompositions.)

Example4)

Considerthe mass-spring-dashpotystem as in Example1,but with initial condis tions x(O) = x'(O)= 0 and with the imposedexternal force F(t) = 15 Find sin2t. the resulting transient motion and steady periodicmotion of the mass.)

Solution The initial value problemwe needto solve is)


x\"

+ 6x'+ 34x = 30 sin 2t;

x(O)= x'(0) = O.)

The transformed equation is)

s2X(s)+ 6sX(s)+ 34X(s)= s2 +4) .


Hence)

60

X(s) =

60

s2+4 (S+3)2+2S (s2+4)[(s+3)2+2S]


\)

As

+B +

Cs + D

4.3Translation and Partial Fractions 293)


When

we multiply

both

sidesby the common denominator,we get)


(15))

60== (As + B)[(s+ 3)2 + 25]+ (Cs + D)(s2+ 4).


To find
in

Eq. (15);

A and B,we substitute the result is)

the zero s == 2i of the quadratic factor s2 + 4

60== (2iA + B)[(2i+ 3)2 + 25],)


which we simplify
to)

60== (-24A + 30B)+ (60A + 12B)i.)


We now

the two linear equations)

equate real parts and imaginary parts on each sideof this equation to obtain
== -24A + 30B 60

and

== 60A + 12B 0,)

which are readily solved for A == To find C and D, we substitute (s 3)2 25 in Eq. (15)and get)

\037\037

and B == the zero s ==


\037\037

-3+ 5i of the quadratic factor

60== [C(-3 5i) + D][(-35i)2+ 4],) + +


which we simplify
to)

60== (186C 12D) (30C 30D)i.) +


Again

we equate real parts and imaginary parts; this yieldsthe two linear equations)
== 186C 12D 60

and

30C 30D== 0,)


\037\037

and we readily find their solution to be C == D == With these values of the coefficients A, B,C, and o decompositionf X (s) is) X (s) =
x
1

D, our

partial fractions

29(

-1 + 50+ Os
S2+ 4

Periodic

0.5)

\\)

== \037

-IOS 25 .2 lO(s+ 3) 4 .5 . + + + + 25) 29 ( s2 + 4 (s 3)2 ) + 5 sin 2t) + {g e-3t5 cos5t (

10s 10 + (s + 3)2 + 25 )

\0371
-0.5)
Transient)

After

we compute the inverse Laplacetransforms, we get the position function)

x(t) == 19(-2cos2t

-2

sin 5t).)

FIGURE4.3.3. periodic The forcedoscillationxsp(t), damped


transient motion xtr(t), and solution xsp(t) + xtr(t) in

Example 4.)

x(t) =

The terms of circular frequency 2 constitute the steady periodic forcedoscillationof the mass,whereas exponentiallydamped the terms of circular frequency5 constitute its transient motion, which disappears Note very rapidly (see Fig.4.3.3). that the .))) transient motion is nonzero even though both initial conditions are zero.

294

4 T Chapter Laplaceransform Methods)

Resonance Repeated uadratic actors) and F Q


The following two inverse Laplacetransforms are useful in inverting tions that correspond the case repeatedquadratic factors: to of
\302\243

partial frac-

-1
{

(s2

+
I

k 2) 2 }

-t .
1
2k)

sIn k t,

(16))

== (sinkt -ktcoskt). (17)) ) { + k 22} 2k 3 These follow from Example and Problem 31 Section respectively.Because 5 of 4.2,
\302\243

-1

(s2

_
1

of the presence Eqs. (16)and (17)of the terms t sin kt and t cos a repeated in kt, factor ordinarily signalsthe phenomenon of resonancein an undamped quadratic mechanical or electrical system.)

Example5)

UseLaplacetransforms to solve the initial II 2 D. X + WaX = ra SInwt;)


that

value problem)
X(O) = 0

= x/CO))

determines undamped forced oscillations f a masson a spring.) the o we transform the differential equation, we get the equation
w w

Solution

When

If w

Fa 2 s2Xes)+ woX(s)= 2Fa 2 ' so Xes) = 2 2 . 2 2 S +w (s +w)(s +wa) =1= Wa, we find without difficulty that X (s) =
w2

so it follows that)

x(t) =
But if
w == Wa,

Faw
w2

- (s2 + - s2 + ) -- ). - (
Faw

2 wa

2 wa

w 2)

2 wa

- 1.
1 sinwat
SInwt
Wa

we have)
X (s) =
Fawo

(s2

2 + w a) 2 '

4)

so Eq. (17)yieldsthe resonanceolution s

x(t) =
t)

Fa \037 (sInwat 2w
a)

wat

coswat).

(18)

.)

Remark:The solution curve defined in Eq. (18)bouncesback and forth betweenthe \"envelopecurves\" X = ::I::C(t) are obtained by that (seeFig.4.3.4)
FIGURE4.3.4. Theresonance
writing

(18)in

the form)

= 1, curves x = ::f:C(t).)

solution in with (Va Fa together with its

(18)

and
\037

envelope
and
then

x(t) == A(t) cos wat


defining the usual
\"amplitude\"

+ B(t)sin w at C = ,JA2 + B2.n this case find I we

that

C(t) =

Fa 2 2wa

J{Jit + 1.
a 2

This technique for constructingenvelope curvesof resonance solutions is illustrated further in the application material for this section. .)))

4.3Translation and Partial Fractions 295)


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

Example6)

Solvethe

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

initial

value problem) y(O) = y/ (0) = y\" (0) = y(3) (0) = O.)

y(4)

+ 2y\" + y = 4tet ;

Solution

First

weobservethat)
oC{y\"(t)}=

= s2y(s),) oC{y(4)(t)} s4y(s),) and

t
\302\243{te}

(s _ 1)2
')

Hence transformed equation is) the


(s4+ 2s2+ l)y(s) = (s

4
1))

Thus our problemis to find the inverse transform of)

yes)

- (s - 1)2(s2+ 1)2
4

=
If we multiply
A(S2

B - 1)2+ (s
A

+D +F + Cs+ + Es+ 1 1 (s2 1)2 s2

.)

(19))
the
equation)

by the common denominator (s

+ 1)2+ B(s 1)(s2+ 1)2+ Cs(s 1)2 + D(s 1)2+ Es(s 1)2(s2 1)+ F(s 1)2(s2 1) = 4. + +

- 1)2(s2+ 1)2,we get -

(20))

Uponsubstituting

holdsfor all values of s.To find the valuesof the remaining coefficients, we substitute in succession values s = 0,s = the s = 2,s = -2,and s = 3 in Eq. (20).This yieldsthe system)
that

s = 1 we find Equation (20)is an identity

that A

1.

-1,

D F 3, - 4C + 4D - 8E + 8F= 0, -8B + + 25B+ 2C+ D + 10E 5F = -21, - 18C 9D- 90E+ 45F= -21, -75B + +

-B

(21))

200B 12C 4D + 120E 40F= -96) + + +

of five linear equations in B,C, D, E, and F. With the aid of a calculator proC = 2,D = 0,E = 2,and grammed to solve linear systems,we find that B =

F=

1.

-2,

We now substitute in

Eq. (18)the coefficientswe have found, and thus


1 S

obtain)

yes) =

2 -(s - 1)2

+ 1

2s 2s+1 . + 1)2+ s2 + (s2


1)

the t Recalling Eq. (16), translation property, and the familiar transforms of cos and sin t, we seefinally that the solution of the given initial value problem is)
y

(t) = (t

- 2)e +
t

(t

+ 1)sin t + 2 cost

.)

.)))

296

4 T Chapter Laplaceransform Methods)

III) Problems)
Apply

the translation
in
Jrt

theorem to find the

of the functions

Problems

1.f(t) = t 4e

3.f (t) = e- sin 3n


2t

t)

1through 4. 2. f(t) = t 3/2e4. f (t) = e-t12cos2 (t - k n


4t)

Laplacetransforms

Problems 27 through 38.

UseLaplacetransforms

to

solvethe

initial

value problems in

))

Apply the translation

theorem to find the inverse


in

forms of the functions

Problems 5 through

10.

Laplacetrans-

5. F(s)= 2s 3
7. F(s)=

-4
1

6.F(s)=

s 1 3 (s + 1) +

28.xl/ - 6X' + 8x = 2;x(O)= x'(O)= 0 29.xl/ - 4x = 3t; x(O)= x'(0) = 0 30.xl/ + 4X' + 8x = e-t ; x(O)= x'(O)= 0

27. xl/ + 6X' + 25x = 0;x(O)= 2, x'(0) = 3

31.(3) + x

s2 + 4s + 4 +5 9. F (s) = s2 3s6s + 25

8.F(s)= s2 s+2 5 4s 10. (s) = 9s2 - 12s+ 20) F


the inverse

2s

-3

Usepartial fractions to find


the functions
in

Problems

11 through

Laplacetransforms of

22.)

5 2s -4 13.F(s)= s+7s+10 14. (s)= s3 5ss2 2s) F 2

1 11. (s)= s2 -4 F

5s 6 12. (s)= s2 _ 3s F

15.F (s) =

S3

17.F(s)= S4

1
5s)2

- 16. (s)= + s F (s
1
2

= 0;x(O)= 0,x'(O)= xl/(O) = 1 32.X(4) x = 0;x(O)= 1, '(O)= xl/(O) = X(3) (0) = 0 x 33.X(4) + x = 0; x (0) = x'(0) = xl/ (0) = 0, x(3)(0) = 1 34. X(4) + 13xl/+ 36x = 0; x (0) = xl/ (0) = 0, x'(0) = 2, = x(3)(0) -13 35.X(4) + 8x\" + 16x = 0; x(O) = x'(O) = xl/(O) = 0, x(3)(0) = 1 36.X(4) + 2xl/ + x = e2t ; x (0) = x'(0) = xl/ (0) = x(3)(0) = 0 t 37. xl/ + 4X' + 13x = te-; x(O)= 0,x'(O)= 2 x 38.xl/ + 6X' + 18x= cos2t; x(O)= 1, '(0) = -1)

xl/

6X'

Problems 39 and 40 and with


the initial

illustrate

2 6))

s mass-spring-dashpotystem
conditions

with

x (0) = x'(0) = o.

two types of resonancein a given external force F(t)

16)

19. (s)= S4 S25s22s 4 F + +

S3 18. (s)= _ F (s 4)4

39.Supposethat m = 1,k = 9, c = 0, and F(t) = 6 cos3t.


Usethe
inverse transform given in solution x (t) t sin 3t. Construct that occurs.) the resonance

20.F(s)= S4

- 8s + 16
2

a figure

t Eq.(16)o

derive the that illustrates

21. (s)= F

s2 + 3 (s2+ 2s + 2)2)

2 3 22.F(s)= 22s 4sS + 5)2) (4s

40. Supposethat m = 1,k = 9.04,c = 0.4, and F(t) t IS 6e- cos3t. Derivethe solution)

Usethe factorization)
S4+ 4a4 = (s2
to through

x(t)= te-tl5 sin 3t.)


Show that the maximum value the oscillations of the mass increasein amplitude during the first 5 s beforebeing damped out as t ---+ +00.)

-2as+ 2a )(s2+ 2as+ 2a


2

2
))

A(t)

te-tj5 is A(5)

derive the inverse Laplacetransforms listed in Problems 23

Fig.

4.3.5)

= 5je.Thus

of the

amplitude function (as indicated in

26.

23.
24.

\302\243-1

4a4 {S4:3 }

= coshat cosat =

2)

\302\243-1

L4:
s2

4a4

}
}

2:
2a

2 sinh

at sin at

/)
-2
oscillation in Problem
sinh

x = + te -t15

t
IOn

25.

\302\243-1

4 { s4 +4a

= \037 (coshat sin at + sinh at cosat)


at

\037 x = te -t15)

1 26. -1 s4 +4a4 = \037 (cosha t sin at 3 { } 4a)


\302\243

cosat)

The FIGURE4.3.5. graph of the damped


40.)))

4.4Derivatives, Integrals, Products Transforms 297) of and

11II

De\037i\0373:ti\037.c::\037.!

I'.1\037c::.gE\037I.\037!

a.\037d

Pr_\037\037uct\037__\037f

Transforms)
unknown)

is sometimes as recognizable the product of the transforms of two known functions. For example, hen we transform the initial value problem w
x\"

The Laplacetransform of the

(initially

solution of a differential equation

+ x == cost;
=
\302\267

x(O)== x'(0) == 0,)

1 s . = s2 + 1 s2 + 1 \302\243{cost}. \302\243{smt}. (S2+ 1)2 This strongly suggeststhat there ought to be a way of combining the two functions sint and costto obtain a function x(t) whose transform is the product of their transforms. But obviously x(t) is not simply the product of cos and sin t, because t

we get)

Xes)=

\302\243

t {cossin t} =

\302\243

H sin 2t}=

1 S2

S
=1=

Thus

sin t} =1= Theorem 1 of this


\302\243{cos

\302\243{cos

sectionwill tell us that the function

t}

+4

(s2

+ 1)2

\302\243{sin

t}.

h(t) =
has the desiredproperty that)
\302\243{h(t)}

1f('r)g(t- i)
1

di)

(1))

== R(s) == F(s)

. G(s).)

(2))

The new function of t defined as the integral in (1)dependsonly on and g and is called convolutionof and g.It is denotedby * g, the ideabeing that it is a the new type of product of and g, so tailored that its transform is the product of the transforms of and g.)

II

DEFINITION

TheConvolution Two Functions) of

The convolution * g of the piecewiseontinuous functions c for t > 0 as follows:)


\037)

I and g is defined
(3))

(f
We will

* g)(t) =

1f(i)g(t- i)di.)
1

also write

product, Theorem 1 of this sectionsaysthat)


\037) \302\243{I

I(t)* get) when convenient. In terms of the convolution


* g}==
u == t

If we make the

substitution
1

\302\243{/}

\302\243{g}.)

T in the integral

in

(3),we seethat)

f(t)

* get) =

Thus the

convolution is commutative: I
=

1f(i)g(t- i)di= fO f(t - u)g(u)(-du)


1

g(u)f(t

u) du

= g(t) * f(t).)
I.)))

* g == g *

298

4 T Chapter Laplaceransform Methods)

Example

1)

The convolutionof cos and sin t is t

(cost) * (sint) =
We

it COST sin(t - T)dT.

apply the trigonometric identity

cosA sin B == 4 [sin(A + B)


to
obtain)

sin(A

B)])

(cos * (sint) = t)
==
that is,)

it
1[

\037

[sin t

- sin(2T - t)] dT
t

T sint

_ + 1COS(2T t)

;
T=O)

And

we recallrom f

is indeeds/(s2+ 1)2. Theorem 1 is proved at the end of this section.) THEOREM TheConvolution 1 Property

(cost) * (sint) == 4t sin t. 4.2 Example5 of Section that the Laplacetransform of 1tsin t

c Supposethat f(t) and get) are piecewiseontinuous for t > 0 and that If(t)! are boundedby Mect as t -+ +00. hen the Laplacetransform of the T !g(t)! convolution f(t) * get)existsfor s > c;moreover,)
and
\037)

\302\243{f(t)

* get)}== \302\243{f(t)}.

\302\243{g(t)})

(4))

and)
\037)

\302\243-l{F(s)

. G(s)}== f(t) * get).) .

(5))

that

we can evaluate the

Thus we can find the inverse transform of the product F(s) G(s),provided
integral)

\037)

. aC-1{F(s)

G(s\302\273)

it f(T)g(t - T)dT.)

(5'))

Example2 illustrates the fact that convolution often provides a convenient alternative to the use of partial fractions for finding inverse transforms.)
cO

Example2)

With

f(t)

== sin 2t and

get) == et , convolutionyields
t = = (sin2t)*e t {te-r sin2T dT = et e-T (-sin2T - 2 COS 2T) 5

,.,

2
\302\243-1

{ (s -1)(s+4)}
2

t e10
t t

sin2TdT

= et
so)
\302\243

10

]0')
.)))

-1

{(s-I)(s2+4) }

==

2 -e - 5 5

1. - SIn

2t

5)

cos 2t.

4.4Derivatives, Integrals, Products Transforms 299) and of

Differentiation ofTransforms
of f(t) 4.2, According to Theorem 1 of Section if J(O) == 0 then differentiation to multiplication of its transform by Theorem 2, proved at the end corresponds of this section,tells us that differentiation of the transform F (s) correspondso t of the original function f (t) by t.) multiplication

s.

THEOREM2

of Differentiation Transforms < > 0and IJ(t)1 Mect as t -+ +00, then)


(6)) == L{-tJ(t)} F'(s))

If J(t)is piecewiseontinuousfor t c

for s
\037)

> c.Equivalently,)
== --L-1{F'(s)}. J(t) == L-1{F(s)}
t)

(7))

Repeatedapplication of Eq. (6)gives)


oC{t
n

J (t)}== (_I)np(n) (s))

(8))

for n ==

1, ,3, ... 2

.)

Example3)

Find oC{t2 sin kt}.)

Solution Equation (8)gives)


2 oC{t sinkt}=
==

(-1)2 C2:2 ) :S22 k

d -2ks == 6ks2 2k3 ds [ (s2+ k 2)2] (s2+ k 2)3

.)

(9))

.)

The form of the differentiation property in Eq. (7)is often helpful in finding an inverse transform when the derivativeof the transform is easier work with than to the transform itself.)

Example4)

Find oC

-1{tan -1(1 } Is)

.)

Solution The derivativeof tan-

(IIs) is a simplerational function, I -1 tan _II == -oC-1 d tan -1- 1 oC


1

s}

{ds

so we apply Eq. (7):)

s)

= =

--roC { 1 + (l/S)2) }

-1
t

-I/s2
1}

--r

oC

- -I
{S2+

--r(-

sint).)))

300

4 T Chapter Laplaceransform Methods)


Therefore,)
1 sin t -1 tan -I - ==---. s

\302\243

.)

t)

Equation (8)can be appliedto transform a linear differential equation having polynomial, rather than constant, coefficients.The result will be a differential equation involving the transform; whether this procedure leads to success epends,of d o course, n whether we can solve the new equation more readily than the old one.)
,.\" ...,,'\" \"'\"\"

Nunn,..

Example5)

Let x(t) be the solution of Bessel's equation of orderzero,)


\"'\".... .-.v .\"...., ...\".n...'..o,...,.....\" ...... '\" ...., ,.\"

\"\"

n\037.....,_,\",

'\"

'\"

NY''''

\"\"

,...\"\",.\037....,....,,...

.\"

tx\"

+ x' + tx

== 0,)

such that

denotedby Jo(t).Because)
\302\243{x'(t)}

x(0) == 1 and x'(0) == O.This solution of Bessel's equation is customarily


== sX(s)
x\"

and

\302\243{X\"(t)}

== s2X(s) s,)

and

because and x
d

transformed equation)

are each multiplied by t, application of Eq. (6) yieldsthe

-ds

[s2X(s)

- s]+ [sX(s)- 1]- d [Xes)]


ds)

==

o.

The result of differentiation and simplificationis the differential equation)


= (S2+ I)X'(s) +sX(s) = o.)

This equation is separable-)

X'es)
Xes)
its general solution is) X (s) == In Problem 39 we outline the argument

s2+1')

v'

c . s2 1
+
C
==

that

follows that)
==

1.Because es) == X .
1)

\302\243{Jo(t)},

it

\302\243{Jo(t)}

v's2 +

(10))
.)))

and of 4.4Derivatives, Integrals, Products Transforms 301)

o IntegrationfTransforms
Differentiation of F (s) corresponds multiplication of f (t) by t (together with to a change of sign). It is therefore natural to expect integration of F(s) will that to correspond division of J(t) by t. Theorem 3, proved at the end of this section, confirms this, provided that the resulting quotient (t )/t remains well behavedas t 0 from the right; that is, provided that)

\037

t--+O+

. J (t) Im- exists


t)

and

is finite.)

(11))

o Integrationf Transforms c Supposethat J(t) is piecewiseontinuous for t > 0,that f(t) < ditionin (11), that IJ(i)1 Mect as t T and +00. hen)
\037
\302\243

THEOREM 3

satisfiesthe con-

{J;t)}

[00F(a)

da)

(12))

for s

> c.Equivalently,) J(t) =


\302\243-I{F(s)}

t\302\243-I

{[(X)F(a)da}.)

(13))

Example6)
Solution

... ... . . . Find {(sinh )/t}.) t


\037... \302\243,

We first verify that the lim


t--+O

condition in

sinh t
t

= lim
t--+O

holds:) (11) et - e-t = et + e-t = lim 1,

2t

t--+O

2)

with

the aid of l'Hopital's Then Eq. rule. sinh t


\302\243,

with (12), J(t) = sinh

t, yields)

- - 1 (0'-10'+1-2a+l - ) - - [ 0'-1 -2 ]
l
00

[
s

00
\302\243'{sinht}

dO' =
1

[
s 1
In

<X>

a 2_ 1
dO'
1
00
.)

dO'

In

Therefore,)
sinh t
\302\243,

_ -- + }
1

s s

1
1)

becausen 1 = I

O.)

.)

The form of the integration property in Eq. (13)is often helpful in finding an inverse transform when the indefinite integral of the transform is easier handle to
,

Example7)

..
.

than

the transform itself.)

Fi\037d\037:/:'

-f{

i;i(\037i\037

\037\037{)Ii:)))

302

4 T Chapter Laplaceransform Methods)

Solution

We

coulduse partial fractions, but


2S
\302\243-1

it

{ (s2

- 1)2} =

This is much simplerto apply Eq. (13). gives)

t\302\243-1

{s

1 (a2
cr 2 _ 1 s

00

2a
1)2)
oo

da
t\302\243

t\302\243

-1

{[

-1

-1

{S2-

1}

')

and therefore)
1
\302\243-

2s { (s2 1)2)}

sinh t

.)

.)

Proofs Theorems of

ProofofTheorem Theorem 2 of Section


gIves)

1:The transforms F(s) and G (s) exist when s > c by 4.1.or any i > 0 the definition of the Laplacetransform F
SU

G(s) =
and therefore)

00

e-

g(u)du =

G(s) = eST
00

because may define J(t) and get)to be zero for t < O. Then) we

1 1
00

00

e-S(t-T) get

- T)dt

(u

=t

T),)

e-stg(t T)dt,

= F(s)G(s) G(s) = =
00

e-STJ(T) dT =
00

1 1 1 (1 e- e-STJ(T)
00

(eST

e-st get - T)
T)dt)

00

e-STf(T)G(S)dT dt)
dT

00

st

f(T)g(t

dT.)

Now our hypotheses on and g imply that the orderof integration may bereversed. of (The proof of this requiresa discussion uniform convergence of improper inteM grals, and can be found in Chapter 2 of Churchill'sOperational athematics, 3rd ed.(New York: McGraw-Hill, 972).) ence H 1

= F(s)G(s) = =
and therefore,)
We

1 (1 e1 e- (it
00 st

00

00

st

- ) f(T)g(t - T) dT )
* get)}.
with t

J(T)g(t T) dT dt
dt

00

e-st [f(t)* get)]dt,)


\302\243{f(t)

= F(s)G(s)
limit

integral i >replace upper the proof of Theorem 1. t. This completes the

of the inner

because et-i)= 0 whenever g


\037)))

and of 4.4Derivatives, Integrals, Products Transforms 303) Proof fTheorem 2:Because) o


00

F(s)=
d F(s)=

e-st f(t) dt,)

differentiation under the integral sign yields)

=
thus)

10)

-1 e00

st

ds

f(t) dt
dt

d o ds

00

[e- f(t)]

st

1 e- [-tf(t)]
st
0)
\302\243-1

00

dt;

F'(s) = -tf(t)},
\302\243{

The and then dividing by of differentiation under the integral sign dependson uniform convergence of the resulting integral; this is discussed Chapter 2 of the bookby Churchill just in mentioned.
which
validity
\037)

is Eq. (6).We obtain Eq. (7)by applying

-t.

Proof fTheorem 3:By definition,) o


F(eT) =

1 e- f(t)dt.)
at

00

Sointegration of F(a)from s to +00 gives)

[00

F(eT)deT

[00
00

(1 e- f(t) dt)
at

00

deT.)

Underthe hypotheses the theorem, the orderof integration of Churchill'sbookonceagain); it follows that)

may

be reversed(see

[00

F(eT)deT =

1 ([00e- f(t) ) e-at = 1 [ ] f(t)dt


at
deT 00 00

dt

a =s)

1 e- f;t)
st

00

dt = 4-

{f;t) }

.)

This verifies Eq. (12), Eq. (13)follows upon first applying and
plying

by

t.

\302\243-1

and

then multi\037)

Problems) 11II
Find the convolution

f(t) * g(t)in Problems1through 6.) 1.f(t) = t, g(t)= 1 2. f(t) = t, g(t)= eat 3.f(t) = g(t)= sin t 4. f(t) = t 2, g(t)= cos 5. f(t) = g(t)= eat)

6. f(t) = eat, g(t) = e

bt

(a i= b)) Laplace

t)

Apply the convolution theorem to find the inverse transforms of the functions in Problems 7 through

14.)))

304
7. F(s)=

4 T Chapter Laplaceransform Methods)


1

s(s 3)
1

8. F (s) 10.F ( )

9. F(s)= 13.F(s)=
In

S2 11. (s) - (s2+4)2 F

(s2 + 9)2 s

- S(S2+ 4) s +
1

l 36.x\" + 4x = f(t); x(t)= - f(t - T) sin 2Td-r 2 0

(s - 3)(S2+

1))

12.F(s)= + 4s + S(S2 5) s 14.F (s) S4+ 5S2+ 4)

1 S2(S2 k 2) 1

37. x\" + 2x'+ x = f(t); x(t) = -re-r f(t 1/ = 38.x\" + 4x' + 13x f(t); x(t) = 3
l
In

-r) dr

i
0)

f(t

- T)e-

2Tsin

3TdT

rem 3 to find the


2t

Problems 15through 22, apply either Theorem 2 or Theo-

TermwiseInverseTransformation of Series
Chapter2 of Churchill's Operational Mathematics, the folthat f (t) is continuous for lowing theorem is proved. Suppose t > 0, that f(t) is of exponential orderas t --+ +00,and that)

15.f(t) = t sin 3t 17.f (t) = te cos3t 19. (t) = sin t f t

21.(t) = e f

3t

Laplacetransform of f(t).

-1

16.f(t) = 18.f (t) = te-t sin 2 t 1 - cos2t 20. f (t ) = t t et - e22.f(t) = t


t 2 cos2t

F(s)=
where 0 < k < 1 and s > c. Then) the 00
\037

LS
n=O)

00

an n+k+l

Find the inverse transforms through 28.)

of the

functions

in

Problems23
S2+ 1

seriesconvergesbsolutely for a

23.F(s)= In s+2
25. F(s)= In

s-2
(1

f(t)=
Apply this result in

S2+ 1

24. F(s)= In 2 s +4 26. F(s)= tan- 1


3

' rcn+k+l)
ant

n k

Problems 39 through
that)

41.

27. F (s)= In

(s + 2)(s 3) +
s12 ))

28.F (s)

s+2)

39.In Example 5 it was shown


\302\243{Jo(t)}

s (S2+ 1)3)

1 -1/2 C =C 1+ y'S2+ 1 s ( S2))

- .)

In Problems 29 through 34, transform the given differential equation to find a nontrivial solution such that x (0) o.

29.x\" + (t - 2)x'+ x = 0 t 32. 33.


34.

Expand with the aid of the binomial seriesand then compute the inverse transformation term by term to obtain)

31.

30. tx\" + (3t


tx\" tx\" t x\" tx\"

35.Apply
\302\243-1

+ (4t 2)x'+ (I3t 4)x = 0


the convolution

- (4t + I)x'+ 3x =+0 I)x= 0 I)x'+ 2(2t + 2(t - I)x'- 2x = 0 - 2x'- x = 0 +


t

Jo(t)
Finally, that)

=C

00
\037

2n (_I)nt

22n(n!)2
that

note that Jo(O)= 1 implies


the function

theorem to show

40. Expand

e F(s) = S-I/2-l/s in

C = 1.

powers of S-1

to show that)

- I1)viss } = {(s

2\037 V JT 10

[./I e

u2

du

= e/erf.Jt.
\302\243-1

(Suggestion: Substitute u

= ,Jt.)) = x' =

{vis

\037e-l/S

= \037cos2vt.

5i)

41.Show

that)

In Problems 36 through 38, apply the convolution theorem to derive the indicated solution x (t) of the given differential (0) o.) equation with initial conditions x (0)

\302\243-1

{\037e-I/s

= Jo

(2.Jt)
.)

and lIED Per\037odic

.\037ie\037\037\037_\037\037._\037.\037\037t.\037\037.':!\037

\"\037

\037.!.\037p!:l\037.E\037\037

\037t

._____________ \037\037_....
.___.)

Mathematical modelsof mechanical or electrical systemsoften involve functions with discontinuities corresponding external forcesthat are turned abruptly on or to off. One such simpleon-off function is the unit step function that we introduced
in)))

4.5Periodic Piecewise ontinuousInput and C


Section Recall that 4.1.
x)

Functions 305)

the

unit

stepfunction at t = a is defined by)


(1))

.)

x = Ua(t))

< . _ a, ua(t)=u(t-a)=0 if t > a.) 11ft

a)

FIGURE
unit

step function

4.5.1. = The graph


at t

of the

a.)

indicatessuccinctly where the unit upward step in value takes useful idea of a \"time a) connotesthe sometimes delay\" a before the stepis made. In Example8 of Section we saw that if a > 0,then) 4.1
The notation

whereasu(t place 4.5.1), (Fig.

U a (t)

\037)

oC{u(t

- a)}= es)
\037

-.
as as

(2))

Theorem 1 tells us that


the

Because oC{u(t)}= lis, (2)impliesthat multiplication of the transform of u(t) Eq. t t -a in the originalindependentvariable. bye-as correspondso the translation t
this

fact, when properly interpreted, is a general property of

Laplacetransformation.)

on Translation the'-Axis If oC{f )}existsfor s > c, then) (t


THEOREM 1
oC{u(t
and)

- a)f(t- a)}= e- F(s)) - a)f(tif 1 f


a))

(3a))

as = oC-1 {e- F(s)} u(t

(3b))

for s

> c + a.)
u(t

Note that)

- a)f(t- a) =
f

1f(t

a))

< a, . tt > a.) _

(4))

Thus Theorem 1 impliesthat oC-1 asF (s)}is the function whosegraph for t > a {eis the translation by a units to the right of the graph of f (t) for t > O.Note that the part (if any) of the graph of (t) to the left of t = 0 is \"cut off\" and is not translated In (Fig.4.5.2). some applications the function f(t) describesan incoming signal that starts arriving at time t = O.Then u(t denotesa signal of the same \"shape\" but with a time delay of a, so it doesnot start arriving until time t =

-a)f(t-a)

a.

Proof fTheorem 1:From o


a)

we the definition of oC{f(t)}, get)

e-as F(s)= e-as


The substitution
t

FIGURE4.5.2. Translation of J(t) a units to the right.)

= + a then

00

= e-STJC-C)di

00

e-S(T+a) J(i)di.)

yields)

e-asF(s)=

1 e- J(t 00 st

a)

dt.)))

306

4 T Chapter Laplaceransform Methods)


From

Eq.(4) we seethat

this

is the sameas)
\302\243(u(t

e-asF(s)=
becauseu(t
Theorem 1.)

- a)f(t- a)},) e-S1u(t a)f(t- a)dt= 100 - a)f(t - a) == 0 for t < a. This completesthe proof of
\037)

Example

1)

With

J(t) == 4t2, Theorem 1 gives)


\302\243-1

e-as == u(t { s3 }
........ \037'n.'n.'n_\"\"\"\",,,,,,,,, \"'.,;,.....

_
if)

- a)-(t- a)2 2
1
\037\"'\" ,...-....,,,.. \"..ou,\", .... ..... ........\" ..._..,............. ... nn..... \"\"',.............

==

14(t-a)2)
\"\"'..... \"\"',......... .. --..._... ...........'\"\"\"',...,. ......,,................. N

if if

t t

< a, >
a)
\"'\" \"',...... ... \"\"\"'--............ __.. ,___ ........... .........................,......

4 (Fig. .5.3).

.)
\037 '\"'

\037\037

\"_'\037'_'

\037....

;,\";,'\"

\037,......

..........\"........,.......,.

Example2)

Find

...............
get) =
1\0372)

.....

\",Mo.

\"'.....,,\"\"'...

'\"

....\"

........

\037...............,

..,

\037

...

...........

\302\243{g(t)}

if if

t t

> 3)

< 3,

(Fig.4.5.4).)

Solution Beforeapplying Theorem 1, e must first write get) in the form u(t 3)f(t 3). w The function f(t) whose translation 3 units to the right agrees(for t > 3) with get) == t 2 is f(t) == (t + 3)2 because J(t 3) = t 2. But then

F(s)==
so now Theorem 1 yields)
\302\243{g(t)}

2
\302\243{t

+ 6t + 9}== 3\" + 2\" + -, s s s)


==

== \302\243{u(t

- 3)J(t- 3)}
2t)

e-3sF(s)== e-3s

3\" s

6 9 + 2\" + - . s) s

.)

Example3)

Find

\302\243{f(t)}

if

f(t) =

\037os

if 0 if t

< t < 2n, > 2n)


x)

(Fig.4.5.5).)

x)

\037
x)

20)
15)
Tt)

x =f(t))

2Tt)

3Tt

t)

10)

1 x = '2ua(t)(t
a)

2
a))

5)

-----;
1)

x=t2//
2)

\",/

/)
\\
3)

x=g(t)
4)
t)

The FIGURE4.5.3. graph of the inverse transform of


Example 1.)

FIGURE4.5.4. graph of the The function g(t) of Example 2.)

FIGURE4.5.5.The function J(t) of Examples 3 and


4)))

4.5Periodic Piecewise ontinuousInput and C


Solution
We

Functions 307)

note first

that)

f(t)

==

[1

- - 2n)]cos2t
u(t

== cos 2t

- - 2n) cos2(t u(t

2JT))

because f the periodicity of the cosine o function. Hence Theorem 1 gives)


oC{ (t )} == oC{cos 2t}

- e,.., ..... \037\"\" ,-..0,..,,___

2n s

oC{cost} == 2

s(l- e-2ns ) .
s2 +4)

.)

Example4)

.. .

.....\037

\"\"\"

.....\"\"\"',___

,___,___

'-'

\"\"

-..

,___

,.-<'\"

'\"'.....

'\"'

'-'

\037 '-'

'\"

mass that weighs32 lb (massm == 1 slug)is attached to the free end of a long, The massis initially light spring that is stretched 1 ft by a force of 4 lb (k == 4 lb 1ft). at rest in its equilibrium position. Beginning at time t == 0 (seconds), external an force f(t) == cos is appliedto the mass,but at time t == 2JT this force is turned off 2t a (abruptly discontinued)nd the massis allowed to continue its motion unimpeded.
A

Find the resulting position function

x(t) of the mass.)

Solution

We

needto solve the

initial

value problem)

xl!+ 4x == f(t); x(O)==x/CO)== 0,)


where (t) is the function of Example3.The transformed equation is)

(s2+4)X(s)==F(s)== s2 +4)
so)

s(l- e-2ns ) '


s (s2+ 4)2
.)

X (s) =

- e(s2 + 4)2
s
S

2ns

Because)

oC

-1

{ + 4)2 }
(s2
u (t

== 1 t Sin 2t 4

by
n)

4.3, Eq. (16)of Section it follows from Theorem 1 that)


x(t) ==
t
\037

sin 2t [t

nl2)

/ /)

x = n 12)

==
\037

- - 2n) . - 2n) - - 2n) . - 2n)]


\037

(t

sin 2 (t
.)

2JT)

u (t

(t

sin 2t
that

be written

If we separatethe
in

t cases < 2n and t > 2n,we find

the position function

may

the form)
t

-n

12)

'-o)

x = -n12 4n
t)

x(t) ==
6n

\037

sin 2t

if if

t t

< 2n,)

-n

2n

I !nsin 2t)

> 2n.)

FIGURE4.5.6. graph of the The function x(t)of Example 4.)

As indicated by the graph of x(t) shown in Fig. the massoscillates with circular frequency w == 2 and with linearly increasing amplitude until the force is removed at time t == Thereafter, the mass continues to oscillatewith the samefrequency but with constant amplitude n The force F(t) == cos would 2t if producepure resonance continued indefinitely, but we seethat its effect ceases .))) immediately at the moment it is turned off.

4.5.6,

2n.

12.

308

4 T Chapter Laplaceransform Methods)


If we were to attack Example4 with the methods of Chapter 2, we would need to solve one problemfor the interval 0 < t < 2n and then solve a new I problemwith different initial conditions for the interval t > 2n. n such a situation the Laplacetransform method enjoys the distinct advantage of not requiring the solution of different problemson different intervals.)

Example5)

Consider RLCcircuit shown in Fig. .5.7, R = 110 , L = 1 H, C = 0.001 the Q 4 with V. F, and a battery supplying Eo = 90 Initially there is no current in the circuit and no charge on the capacitor.At time t = 0 the switch is closed left closedfor and 1 second. t time t = 1 it is openedand left open thereafter. Find the resulting A
......

current in the circuit.)

Solution
L)

We

recallrom Section the basicseriesircuit equation) f 2.7 c


1 di L-+ Ri + -q= e(t); dt
C) (5))

\037)
EO) C)

we use lowercase letletters for current, charge, and voltage and reserve uppercase ters for their transforms. With the given circuit elements, q. (5)is) E

T)
circuit

'VV'v
R)

T
where e(t)

di
dt

= + 110i 1000q e(t), +


to the opening and

(6))

T FIGURE4.5.7. he series RLC of Example 5.)

In Section our strategy was to differentiateboth 2.7 the relation)


1

switch.

= 90[1

- - 1)],orresponding c
u(t

closingof the
apply

sidesof Eq. (5),then

=dq
dt)

(7))

to obtain the

second-order equation)
2i 2

d di 1. L-+ R- + -1= e (t). dt


, dt
C)

Here do not use that method, because = 0 except t = 1,whereas the we at e'(t) to when t < 1 to e(t) = 0 when t > 1 would seem require jump from e(t) = 90 that e' 1) = 00. hus e' t) appears have an infinite discontinuity at t = 1. his T T to ( (

4.6. phenomenon will be discussedin Section For now, we will simply note that it is an odd situation and circumvent it rather than attempt to deal with it here. To avoid the possibleproblemat t = 1,we observethat the initial value = 0 and Eq. (7)yield,upon integration, q (0)
q(t) =
We substitute

1i(i)di.)

(8))

Eq. (8)in Eq. (5)to obtain) di 1 L-+Ri+dt C

i i(-r)d-r=e(t).
t
0)

(9))

This is the integrodifferentialequationof a series LC circuit; it involves both R the integral and the derivativeof the unknown function i (t). The Laplacetransform method workswell with such an equation.)))

4.5Periodic Piecewise ontinuousInput and C


In the

Functions 309)

presentexample, q. (9)is) E

- 110i
di dt

+ 1000 i(-r)d-r== 90[1


0)

- - 1)].
u(t

(10))

Because)
\302\243.

{10

[ti(-r)d-r

} =---;-)

/ (s)

by Theorem 2 of Section

4.2 transforms on

of integrals, the transformedequation is)

(s) == + s/(s)+ 110/(s)1000--(1 e-S). s s) /

90

We

solve this equation for / (s) to obtain)

I(s)=
But)

90(1 e-S) s2 + lIOs 1000 +


.)

90 s2 + 11s + 1000) O
so we have)

s + 10) s + 100')

- ) -e I(s)= s+lO- s+l00 -s (s+10 s+100


1 1 1 1
We now
10t apply Theorem 1 with f(t) == e-

.)

- e-

100t thus ;

the inverse transform

is)

i(t) == e-10t
After

_ e-100(t-I)].) - e- - - 1)[e-10(t-l)
100t

u(t

we separatethe cases < 1 and t


by)

> 1, e find w

that

the current

in

the circuit is

given

i (t) ==

The portion e-10t e-100t of the solution would describethe current if the switch .))) wereleft closedor all t rather than beingopen for t > f

I e-

e-l

Ot

10t

-e_

lOOt

_ ee-10(t-l) 100t + e-100(t-l))

if t if t

< >

1,
1.)

1.

310

4 T Chapter Laplaceransform Methods) Transforms f Periodic unctions) o F

Periodic forcing functions in


I I I I

systemsoften are practical mechanical or electrical more complicated pure sinesor cosines. nonconstantfunction (t) defined than The for t > 0is said to be periodic there is a number p >0such that) if

\037 P

FIGURE4.5.8. graph of a The function with periodp.)

-J

f(t + p) == f(t))
I)

(11))
holdsis called for which Eq. (11)

for all t
the

computation of the Laplacetransform of a periodicfunction.)

periodof f. Sucha function is shown

> O.The leastpositive value of p (if any)


in

Theorem 2 simplifiesthe Fig.4.5.8.

2 THEOREM Transforms f Periodic o Functions) Let f(t) beperiodicwith periodp and piecewiseontinuousfor t c == transform F(s) oC{f(t)}xistsfor s >0and is given by) e F(s)== 1 e-PS

> O.Then the


(12))

- 1 e- f(t) dt.
st
0)

Proof:The definition of the Laplacetransform F(s)==


The substitution
t ==

gives)

1
o

00

e-st f(t) dt == L
n=O

00

i + np

l
np)

cn

+l)p

e-st f(t) dt.


p

in

the

nth

integral following the summation sign yields)

cn

+l)p

e-st f(t) dt ==

np

1
0

e-sc-r+np)

f(i + np) di == e-nps


sr

1
0)

e-STf(i)di

because + np) == f (i) by periodicity. hus) T f(i

F(s)=
\037

(e-

nps

= (1+ e-Ps +
Consequently,)

l e- I(r:)dr:) e- ...) l
P 2ps

e-ST: d I(r:)r:.)

F(s)=
We

- e-

PS 10)

(Pe-ST: dr:. I(r:)

use the geometriceries) s


1

I-x) ==1+x+x+x
with

23+...,
i seriesn

derived Eq. (12).

x == e-Ps < 1 (for s > 0) to sum

the

the

final

step. Thus we have

.)

transform of a periodicfunction

The principal advantage of Theorem 2 is that


without

the

enablesus to find the Laplace of necessity an explicitevaluationof an


it

improperintegral.)))

4.5Periodic Piecewise ontinuousInput and C


..,.. ,..,,,

Functions

311)
D

Example6)
/(t)
1

4.5.9 shows the graph of the square-wave function f(t) == (_I)[t/a of == 2a;[x]denotesthe greatest integer not exceeding. By Theorem x periodp
Figure 2 the Laplacetransform of (t) is)

-1
6.)

--.0 --.0 .-...) a 2a 3a 4a 5a 6a --.0 --.0 --.0)


t)

F(s)=

FIGURE4.5.9. graph of the The square-wavefunction of Example

- e- (2ae- f(t) dt r + - e- (Jo e-S1dt1 (-l)e- ) - e- ([-;e-SIJ:- [-;e-slIa)) - e-as (1- e+ s(1 es(1 e1 2as Jo) sl 1 2a
S1

2as

dt

a)

1)

2as)

as )2 2as
))

as

))

Therefore,)

F(s)

-e - s(1+ eI

-as
as
))

(13a))
==

eas/ 2 e-as/ 2 s(eas/ 2 + e-as/ 2) Example7)


g(t))
a)

1 s

tanh

as
2)

(13b)

.)

shows Figure 4.5.10 the graph of a triangular-wavefunction g(t) of periodp == 2a. Because derivative g/ (t) is the squarewave function of Example 6,it follows the from the formula in (13b) and Theorem 2 of Section that the transform of this 4.2 function is) triangular-wave
t)

a 2a 3a 4a 5a 6a

FIGURE4.5.10. graph of The the triangular-wave function of


Example 7.)
\"

G(s) == s

as F(s)== 1 tanh-. 2 2) s

(14)

.)

Example8)

Considera mass-spring-dashpotystem with m == 1, == 4, and k == 20 in apc s propriate units. Supposethat the system is initially at rest at equilibrium (x (0) == x/ (0) == 0) and that the massis acted on beginning at time t == 0 by the external force f (t) whosegraph is shown in Fig.4.5.11:square wave with amplitude 20 the and period2n. ind the position function f (t).) F
'.on.... \"'.... \037 \037,....\" \037 \037 __ ... \"\"'\037\037 \" ,... ,..,. \"\"\037.'N\"\" \" ,... \" '\" V\" \" \"'.... \037 '\" \"\" n.';\037\" \037n\037 \"\"\037

Solution The initial value problemis)


x\"

+ 4x'+ 20x== f(t);

x(O)== x/(O) == O.)

The transformed equation is) S2 X(s) + 4sX(s)+ 20X(s)== F(s).)


(15))))

312

4 T Chapter Laplaceransform Methods)


From

Example6 with a

F(s)==
f(t)
20)

.---.0.---.0\037
I I I I I I I I I I I I I I
I)

==

I I I I
I)

I I I I I

I I I I
I)

==
t

1t
I I I I

21t 31t 41t 51t 61t


I I I I I I I I I I I I I I I I I I I I

so that)

-20

\037 \037 \037)

FIGURE Thegraph of the external-force function of


Example
8.)

4.5.11.

-. - ee- - e- - e- e- - 2e- 2e- 2e- -L(-I)n


20
s
1 1
ns

== n

we seethat the transform of f(t)


ns)

is)

20 s) (1 20 s) (1

ns

(1 +

ns

2ns

_ e-3ns

+ ...)

ns

2ns

3ns

+ ...) ,

20 40 F(s)== s + S

00

e-nns.

(16))

n=1)

Substitution

(16) ofEq. in Eq. (15)yields) F(s) X (s) s2 + 4s + 20) 00 -nn 20 20e s = 2 + + s[(s 2)2 + 16] 8(-l)n + 2)2 + 16]\" s[(s

(17))

From the transform

in

4.3, Eq. (8)of Section we get

20 2t == 5e- sin 4t { (s + 2)2 + 16 } so by Theorem 2 of Section we have) 4.2

L-

')

g(t) == L-

+ {s[(s 2) + 16] }
2\302\2602

10)

r 5e-,sin4-r di.
2

Usinga tabulated formula for f eat sin bt dt, we get)

g(t) == 1 e 2t (cos4t +

sin 4t) == 1

(t )

,)

(18))

where)

h(t) == e-2t (cos4t + 4 sin 4t)


in

.)

(19))
term

Now we apply Theorem 1 to find the inverse transform of the right-hand The Eq. (17). result is)
u x(t) == g(t) + 2 L(-I)n (t
n=1)

00

- nn)g(t - nn),
in

(20))

and we note that for any fixedvalue of t the sum

g(t nn) == 1
== 1

- e- [cos4(t nn) + - e e- (cos4t + 4t)


2(t-nn)
2t
2nn

Eq. (20)is finite. Moreover,


4 sin4(t
.)))

- nn)])

4 sin

and C 4.5Periodic Piecewise ontinuousInput


Therefore,)
2n g(t nn) == 1 e :rr:h(t).)

Functions

313)

(21))

if Hence 0 < t

< n,then)
X(t) == 1

h(t).)

If n

< t < 2n,then)


x(t) == [1

- )] - 2 [1- e
h

(t

:rr:

(t )] ==

- + (t) 1
h
:rr:

2h (t)

[1- e2
(t ) ]

:rr:]

.)

If 2n

< t < 3n,then)


x(t)

[1 h (t )] 2 [1 e2 h (t)] + 2 [1 e == 1 + h(t) 2h(t) [1 e2 + e4


==
:rr:

- - - -

4:rr:

:rr:]

.)

The general expression nn for

< t < (n + l)n is)

x(t) == h(t) + (-l)n 2h(t) [1 e2


== h

(t) + -1) - 2h
(
n

- +...(_1)ne +
:rr:

2n
:rr:])

(t )

e + (-l)n2(n+ ' 2 l+e:rr:)


l):rr:

(22))

Eq. (19),)

which we obtained with the aid of the familiar formula for the sum of a finite geometric progression. rearrangement of Eq. (22)finally gives, with the aid of A

e4t 2 x(t) == e:rr:+12t (cos + _ sin 4t) + (_l)n 2


1

e2

:rr:

- 2 .e(-+ e
1) n 2
:rr:

2:rr:

e-2(t-n:rr:) (cos4t +
in

!
2)

sin 4t )
solution)

(23))

for nn
Xtr(t)

< t < (n + l)n.The first


\037

term

Eq. (23)is the transient


\037

2t 2t cos(4t 0.4636). (24)) (O.9963)e- + 4 sin4t) (1.1139)e(cos4t

The last two terms in Eq. (23)give the steady periodicsolution xsp.To investigate == t Then) nn for t in the interval nn < t < (n + it, we write

i -

l)n.

Xsp(t) ==
\037

2e2 4r (-l)n[1 - e2 + 1 e-2, (cos + 4 sin 4r) ] 2T COS(4i (-l)n[1- (2.2319)e- - 0.4636)]
:n:
:rr:

(25)

.)

showsthe graph of xsp(t).Its most interesting feature is the appearFigure 4.5.12 anceof periodicallydampedoscillationswith a frequencyfour times that of the w imposedforce f(t). In Chapter 8 (Fourier Series) e will seewhy a periodicexternal force sometimes at excitesscillations a higher frequency than the imposed o .))) frequency.

314

4 T Chapter Laplaceransform Methods)


X
sp)

\\ \\ \\

\\

-2t (2.23)e

\\ \\)

\037/ -)
\"'1)

/
/
1

1/\\ - (2.23)e II)


-2t)
\\ \\ \\ \\

-2(t -1 (2.23)e +

n)

\"

-1)

-2(t

1 I
I)

v\\ -1 (2.23)e

- n)

The FIGURE4.5.12. graph of the steady periodicsolution Example 8; note the \"periodically damped\" oscillations with force.) frequency four times that of the imposed

for

P lID. roblems
.)

Find the inverse in

F(s) s+2) ns e5. F(s)= S2 1 + 7. F(s)=


1
S2 +

1through 10.hen sketch the graph of f.) T 3s S_ 3s e1.F(s)= \037 2. F(s)=) e- S2es S 2 e-S -e)-2s =) 3. = e4.
Problems

Laplacetransform f(t) ofeachfunction

given

s 2 Theorem with p = 1 to verify that oC I}= 1/ . = S/(S2+ k 2). 2 24. Apply Theorem to verify that oC{cos kt} 25.) Apply Theorem2 to show that the Laplacetransform of is) the square-wavefunction of Fig. 4.5.13
23.Apply
{

F(s)

- e-

271:

-3s) 9. F(s)= s(1+ en) 2 S2 +


Find the

6.F (s) = S2se- 2 +n - 2s 8.F(s)= s(1 e- )


S

1)

oC{f(t)}=

+ s(1 e-as))

10. (s)= F

Laplacetransforms of the functions 11 through 22.

if 0 < t < 3; f (t) = 0 if t > 3 12. (t)= 1 if 1 < t < 4; f(t) = 0 if t < 1 or if t > 4 ! = 13. (t) sin t if 0 < t < 2n;f (t) = 0 if t > 2n 14. (t) = cosntif 0 < t < 2; f(t) = Oift > 2 f 15. (t)= sint if 0 < t < 3n;f(t) = 0 ift > 3n ! 16. (t) = sin 2t if n < t < 2n; f (t) = 0 if t < n or if f t > 2n 17.f (t) = sin n t if 2 < t < 3; f (t) = 0 if t < 2 or if t > 3 18. (t) = cos if3 < t < 5; f(t) = Oift < 3 orift > 5 f 19. (t) = 0 if 4nt 1;f (t) = t if t > 1 t < f 20.f (t) = t if t < 1;f (t) = 1 if t > 1 21.(t) = t if t < 1;f(t) = 2 - t if 1 < t < 2; f(t) = 0 if f t >2 22.f(t) = t 3 if 1 < t < 2; f(t) = 0 if t < 1 or if t > 2)

11. (t) = 2

! !

S2 + n 2 2s(e-71: S _ e S2+ 4)
given in

r-I r-I
I I I I

rI)

271:

s)

2a

3a 4a 5a

6a)

t)

Problems

The FIGURE4.5.13. graph of the square-wavefunction of Problem 25.)

26.Apply

Theorem2 to show

the sawtooth function

f (t) of Fig. 4.5.14 as 1 eF(s)= as2 - s(l- e- . as

that the

Laplacetransform of
is)

))

J(t))

a 2a

3a

4a 5a

6a)

t)

The FIGURE4.5.14. graph of the sawtooth function of Problem 26.)))

4.5Periodic Piecewise ontinuousInput and C


27. Letget) be the staircase function of Fig. 4.5.15. that Show get) = (tja) f(t), where f is the sawtooth function of
Fig.
and 4.5.14, hencededucethat) as = e\302\243{g(t)}

Functions

315)

s mass-spring-dashpotystem with Solvethe initial value problem)


mx//

external forcing Junction.

+ cx/ + kx = f(t), x(O)= x/CO)= 0)


the graph

as) s(1-e-

')

and construct

31.

of the position function x (t). m = 1, = 4, c = 0; f(t) = 1 if 0 < t < n, f(t) = 0 if k


t

get))
4)

,...)
a)

3)

2)

32.m = 1, = 4, c = 5; f(t) = 1 if 0 < t < 2, J(t) = 0 if k t >2 33.m = 1,k = 9, c = 0; f(t) = sin t if 0 t < 2n, f (t) = 0 if t > 2n k 34. m = 1, = 1, = 0; f (t) = t if 0 < t < 1, (t) = 0 if c J t > 1 35.m = 1, = 4, c = 4; f(t) = t if 0 < t < 2, f(t) = 0 if k
\037

>n

t
2a) 3a)

> 2)

4a

t)

FIGURE4.5.15. graph of the The staircase function of Problem 27.)


28.) Suppose that

Problems 36 through 40, the values of the elements RLCcircuit are given. Solvethe initial value problem)
In

ojan

29.) Suppose that (t) is the half-wave shown in Fig. Show that)
\302\243{f

f(t) = t if oC{f(t)}.

function of period2a with (t) is a periodic 0 < t < a and f(t) = 0 if a < t < 2a.Find

di L-+ dt
with the given

Ri

+C

-i
1
0

i(-r)

di = e(t);

i (0) = 0)

f 4.5.16.
(t)} =
J(t))

rectification

of sin kt,

(S2+ k2)(1 _

.)

e-1Csfk)

7r
k)

27r
k)

37r
k)

t)

impressed voltage e(t). 36.L = 0, R = 100,C = 10-3; = 100if 0 t < 1; e(t) e(t) = 0 if t > 1 4 37. L = 1,R = 0, C = 10- e(t) = 100if 0 < t < 2n; ; e(t) = 0 if t > 2n 38.L = 1,R = 0, C = 10-4; = 100sin lOt if e(t) o < t < n; e(t) = 0 if t > n 39.L = 1,R = 150,C = 2 X 10-4; = lOOt if 0 t < 1; e(t) e(t) = Oift > 1 R 40. L = 1, = 100,C = 4 X 10-4; (t) = 50t if 0 t < 1; e e(t) = 0 if t >

::::

::::

\037

1)

FIGURE4.5.16. half-wave The


sin kt
.)

rectification

of

30. et g(t) = L
function

of Problem 29 and k > O. Note that h(t) = f(t) + get) is the full-wave rectification of sinkt shown in Fig. 4.5.17. Hencededucefrom Problem 29 that)
oC{h(t)}=
h(t))

u (t

-n

jk)

f (t

- n jk), where f
2 coth 2k)

(t) is the

s2 +k

-.
ns

In Problems 41 and 42, a mass-spring-dashpot system with external force is described. nder the assumption that U x (0) x/ (0) 0, use the method of Example 8 to find the transient and steady periodicmotions of the mass. Then construct the graph of the position function Ifyou would like to checkyour graph using a numerical DE solver, it may be useful to note that the function)

f(t)

x(t).

f(t) = A[2u((t
(t

- n)(t - 2n)(t - 3n) . - 4n)(t - 5n)(t - 6n))-

1])

7r
k)

37r
k)

FIGURE4.5.17. full-wave The


sin k t
In
.)

rectification

of

has the value +A if 0 < t < n, the value -A ifn < t < 2n, and soforth, and henceagrees on the interval [0,6n] with the square-wave function that has amplitude A and period2n. also (See the definition of a square-wavefunction in terms sawtooth and triangular- wave functions in the application material for this section.)

oj

stant k,

Problems through 35, the values of mass m, spring condashpot resistancec, and force (t) are given for a)

31

41.m = 1, = 4, c = 0; f (t) is a square-wave function with k amplitude 4 and period 2n. 42. m = 1, = 10, = 2; f(t) is a square-wave function k c with amplitude 10 and period
2n.)))

316

4 T Chapter Laplaceransform Methods)

and lID !\037pulses DeltaFunctions)


Considera force f(t) that acts only during a very short time interval a < t < b, with f(t) = 0 outsidethis interval. A typical examplewould be the impulsive forceof a bat striking a ball-the impact is almost instantaneous. A quick surge of voltage (resulting from a lightning bolt, for instance) is an analogous electrical phenomenon.In such a situation it often happensthat the principal effect of the force dependsonly on the value of the integral) p=

and does not dependotherwise precisely on how f(t) varies with time t. The number p in Eq. (1)is calledthe impulse the force f(t) over the interval [a,b]. of In the case a force f(t) that acts on a particle of massm in linear motion, of integration of Newton'slaw)

f(t)dt)

(1))

d f(t) = mv/(t) = dt) [mv(t)]


yields)

p=

lbd
dt)

[mv(t)] dt

= mv(b) mv(a).

(2))

replace (t) with f


x)

Thus the impulseof the force is equal to the change in momentum of the particle. Soif change in momentum is the only effect with which we are concerned, need we know only the impulseof the force; eneedknow neither the precise w function (t) nor even the precise time interval during which it acts.This is fortunate, because in a situation such as that of a batted ball, we are unlikely to have such detailed information about the impulsive force that acts on the ball. Our strategy for handling such a situation is to set up a reasonable mathematical modelin which the unknown force f(t) is replacedwith a simpleand explicit force that has the sameimpulse.Suppose simplicity that (t) has impulse 1 and for acts during some brief time interval beginning at time t = a > O. Then we can select fixed number E > 0 that approximates the length of this time interval and a

the

specificfunction)
1)

r
I I I I I I I I I

I-----1
\342\202\254

if a

da,E

(t) =)

E)

< t < a + E,)


(3))

Area

=1

IT
I I I I I I
I) \342\202\254

I!
t)

otherwise.)

This is a function of t, with a and E beingparameters that specify the time interval that [a,a + E].If b > a + E, then we see(Fig.4.6.1) the impulseof da,E over [a,b]
IS)

a
impulse function

a+\342\202\254)

The FIGURE4.6.1. graph of the


da,E

p=

(t).)

Thus da,E has a unit impulse,whatever the number computation gives)

l
a

da,E(t)dt= a

a+E

1 -dt=l.
E E may

the be.Essentially same

00

da,E(t)dt =

1.)

(4))))

a 4.6Impulsesnd DeltaFunctions 317)


the time interval during which the force acts seemsunimporat tant, it is tempting to think of an instantaneous impulse that occursprecisely the instant t We might try to model such an instantaneous unit impulse by taking the limit as E \037 thereby defining)

Because precise = a.

0,

8a(t)

= limda,E(t),
E\037O)

(5))
in

where a > O.If we couldalso take the limit under the integral sign it would follow that)

Eq.(4),then

1
But the limit
in

00

8a(t)dt =

1.)

(6))

Eq. (5)gives)
8a(t) =

+00 10)

if if

t t

= a,
i= a.)

(7))

(7)-ifa function is zero is not 1 but zero. evertheless,the symbol N delta function at a 8a(t) is very useful. However interpreted, it is calledthe Dirac after the British theoretical physicistP. A. M. Dirac (1902-1984), the early who in in 1930s introduced a \"function\" allegedly enjoying the properties Eqs.(6)and (7).)
at except a singlepoint,

Obviously, no actual function can satisfy both (6) and


then

its integral

DeltaFunctionss Operators a
The following computation motivates the meaning that we will attach here to the symbol 8a(t).If g(t) is continuous function, then the mean value theorem for inte-

gralsimpliesthat)

for somepoint
lim
E\037O

t in [a,a + E].It follows that)


00

+\342\202\254

g(t) dt = Eg (1)

1
0

g(t)da,E(t) dt

= lim
E\037O

If 8a(t) were a function in the strict by continuity of g at t = definition, and if we could interchange the limit and the integral in therefore couldconclude that)
Function

a.

l
a

a+E

1 g(t) . E dt = lim g (t) = g(a)


E\037O)

(8))

senseof the Eq. (8),we


(9))

get)

,)

1
We

00

a g(t)8 (t) dt = g(a).)

the delta function,

take Eq. (9)as the definition (!) of the symbol 8a(t).Although we call it the operation) it is not a genuine function; instead, it specifies

1 ...
00

8a(t) dt)

FIGURE4.6.2. diagram A

how the delta function illustrating \"sifts out\" the value g (a).)

which-whenappliedto a continuous function g(t)-sifts or selectsthe value out of this function at the point a > O. This idea is shown schematically in g(a) Note Fig.4.6.2. that we will use the symbol 8a(t) only in the contextof integrals such as that in Eq. (9),or when it will appear subsequently in such an integral.)))

318

4 T Chapter Laplaceransform Methods)


st For instance, if we take g(t) = ein

Eq. (9),the result is)


.)

1
We

00

e-S18a (t)dt = e-as


as

(10)) to be)

therefore define the Laplacetransform of the delta function


a oC{8(t)}= e-

(a

> 0).)
8a(t),)

(11))

If we write)

8(t) = 8o(t)
then with (11) a = 0 gives)

and

8(t

- a) =

(12))

= oC{8(t)} 1.)

(13))

Note that if 8 (t) werean actual function satisfying the usual conditionsfor existence of its Laplacetransform, then Eq. (13)would contradict the corollary to Theorem2 of Section But there is no problemhere;8(t) is not a function, and Eq. (13)is our definition of oC{8(t)}.

4.1.

DeltaFunction Inputs
Now, finally, supposethat we are given a mechanical system whoseresponse x(t) to the external force f(t) is determined by the differential equation)
Ax\"

+ Bx'+ Cx= f(t).)


t

(14))

To investigate the response f this system to a unit impulseat the instant o seemseasonableto replace(t) with 8a(t) and begin with the equation) r f
Ax\"

= a, it
(15))

+ Bx'+ Cx= 8a(t).)

But

what

of Eq. (15)provided that)

is meant by the solution of such an equation? We will call x(t) a solution

x(t) =
where X E (t) is a solution of)
Ax\"

lim
E\037O)

XE

(t),

(16))

+ Bx'+ Cx= da,E(t).)


1 - [ua(t)- ua+E(t)]
E)

(17))

Because) da,E(t) =
(18))

is an ordinary function, Eq. (17)makes sense.For simplicity supposethe initial conditions to be x(O) = x/(0) = O. When we transform Eq. (17),writing X E = E oC{x},we get the equation)
(As 2

+ Bs+ C)XE(s)= E
1

- e-(a+E)S) = (e- - e- . (
e-as
S S as 1 )
SE
SE)))

4.6Impulsesnd DeltaFunctions 319) a


If we take the limit
in

the last equation as E


lim
E\037O

\037

0,and note that)


1)

-e)-SE =
SE)

by

rule, I'Hopital's we get the equation) 2 as (As + Bs+ C)X(s)= e-

,)

(19))

if)

Xes) =
Note that
this

lim
E\037O)

XE(x).

= as Eq. (15)directly, using the fact that oC{oa(t)} e- . On this basis it is reasonable solve a differential equation involving a delta to function by employing the Laplacetransform method exactly as if oa(t) were an ordinary function. It is important to verify that the solution so obtained agrees with the one defined in Eq. (16), this depends but on a highly technical analysis of the limiting procedures of we considerit beyond the scope the present involved; discussion. formal method is valid in all the examplesof this section and will The resultsin the subsequent roblemset.) producecorrect p
o

is preciselythe sameresult that we would obtain if we transformed

, ,
1)

Example

massm = 1 is attached to a spring with constant k = 4; there is no dashpot. The massis released rest with x(O) = 3.At the instant t = 2Jr the massis struck from with a hammer, providing an impulsep = 8.Determine the motion of the mass.)
A

Solution According to Problem 15, e needto solve the initial value problem) w
x\"

+ 4x = 802n (t);
S2X (s)

x(O)= 3, x'(0) = O.)

We

apply the Laplacetransform to get)

- 3s + 4X(s) = 8e3s

2ns
,)

so)

X (s) = 2 s

2ns 8e+ s2 +4 . +4

as Recalling the transforms of sine and cosine, well as the theorem on translations on the t-axis (Theorem 1 of Section 4.5),we seethat the inverse transform is)

x(t) = 3 cos2t + 4u(t

- 2Jr)

sin 2(t

- 2Jr)
1

= 3 cos2t + 4U2n (t) sin 2t.)

= Because 3cos2t 4sin2t 5cos(2ta) with a = tan- (4/3) 0.9273, + separation of the cases < 2Jr and t > 2Jr gives) t
\037

x(t )

\037

if t 3 cos 2t 15 cos(2t 0.9273)) if t

< 2Jr, > 2Jr.)

The resulting motion is shown in Fig. .6.3. that the impulse at t = 2n results 4 Note in a visible discontinuity in the velocity at t = 2Jr, as it instantaneously increases the amplitude of the oscillations f the massfrom 3 to o .)))

5.

320

4 T Chapter Laplaceransform Methods)


t= 2n)
6)

x=3)
\037

0)