David Calvis)
BaldwinWallaceCollege)
PEARSON Prentice
flaIl)
Library
of Congress CataloginginPublication
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)
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, 017602098
Tel:5086477000
Fax:
5086477101
of Waterloo Inc.
Inc.)
Maple,
is a registered trademark
of Wolfram
Research,
10 9 8 7 6 5 4 3 2
of America
1)
ISBN
0132397307)
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
C
T
f
\"
,..._\"z\037;\037,.
..
,\037/
0N
J
_:'
,
i
\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)
FirstOrder 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 LinearFirstOrderEquations 46 .6 Substitution Methodsand ExactEquations 59
1 1 1 1 1 1 1
1)
.7
1.8
CHAPTER)
85)
2
.
'\"\". > /\037,
,z
\037
\037
.. ,
\"riP'
\037,)
General olutions f Linear Equations 113 S o C Equationswith Constant oefficients 124 Homogeneous
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
\"'. .::::::
.......,)
J{
\037\037:::'
Series 194 Introductionand Review of Power Series Solutions NearOrdinary Points 207 RegularSingularPoints 218
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
and of Derivatives, ntegrals, Products Transforms 297 I Periodic Piecewise and ContinuousInput Functions 304 a Impulsesnd DeltaFunctions 316)
5)
5.1 FirstOrderSystems and Applications 326 5.2 The Methodof Elimination 338 and 5.3 Matrices LinearSystems 347 5.4 The EigenvalueMethodfor Homogeneous Systems 366 5.5 SecondOrder A Systems and Mechanical pplications 381 5.6 Multiple EigenvalueSolutions 393
5.7
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 RungeKuttaMethod 453 6.4
6)
CHAPTER)
464)
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
References FurtherStudy 555 for Existence Uniqueness and ofSolutions559 Appendix: Answersto Selected Problems573
Index
11)))
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 problemsolving 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.Iv.,/Nv'''''vN.lv.,/N.N'NhV>....
N.lN.lN.I'XVN.N.N\"'....><N
.v\"/,/\"\"\",/\"\"/N'.IN.I/hnv,,..MN'I<NN.I\"NNv\"\"'
: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 solidclasstested 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 firstorder 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 massspringsystems(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 (page149150); 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 (pages183184). 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 reductionoforderformula with a simplenonseries roblem p
3:
(page239).
istenceof Laplacetransforms (page273);new Remark discussingthe mechanics of partialfraction decomposition (page 279);new muchexpanded t discussionof the proof of the Laplacetransform existenceheorem and its
extensionto include the jump discontinuities that play an important rolein many practical applications (page286287). 5: for Chapter New Problems2023 student exploration of threerailwayn 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 secondorder 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 nonrotating coordinate systems moonearth orbit problems 473). (page Chapter7: New remarks on phase plane portraits, autonomous systems, and critical points (page488490); introduction of linearized systems new Lorenz and new 3dimensional illustrating (page502); Figures6.5.18 6.5.20 and Rosslertrajectories (page552553).)
figures show students Throughout the text, almost 550computergenerated 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
. .
The final two sectionsof Chapter 1 (on populations and elementary mechanics)offer an early introduction to mathematical modeling with significant ap
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 firstorder systems and models. he chapter continues with a selfcontained 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 perioddoublingin 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(0136006132), 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
ques
. . . . . .
ulations? (Section 1.7) How do a unicycle and a twoaxle 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)
. . .
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 nonlinearity lead to chaosin biological nd mechanical a Why 7.6)) systems?(Section
The answer sectionhas beenexpanded to its considerably increase value as a learning aid. It now includesthe answersto most oddnumbered roblemsplus a good p M The 60Spagenstructor's olutions anual (013I S many evennumbered ones. 6006140) accompanying this book provides workedout 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 (013600615 problemsin the book,and the 345page 9) contains solutions for most of the oddnumbered 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 (0136006795) erably in the 320pageApplications 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
Irfan
SemionGutman,
University University
MiklosBona,
It
of Oklahoma of Florida
UIHaq,
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
.
'.
(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')))
FirstOrder 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)
dt
function
dx =
x2 + t 2
derivative
x/(t) = dxjdt.The
dy +3+7y=O dx2 dx
d2y
y
first two
.)
The study of differential equations has three principal goals: differential equation that describes a specifiedphysical situation. 2. To findeither xactly or approximatelytheppropriate solution of that e a equation. To interpret the solution that is found.)
1)))
3.
Chapter FirstOrderDifferential
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
to
find
all
y(x) = Ce
x2
,)
(1))
=C
\037\037
(2xe )
x2
= (2x)
in
x2 (ce ) = 2xy.
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).
.)
Example3)
Temperature
A)
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
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 thenwith the aid of this formulawean predictthe future temperature of the c
bod\037
...... ...... ....
.)
Example4)
(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))))
where k is a constant. If the tank is a cylinder with vertical sidesand crosssectional 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)
(7))
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
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:
that
1000,
= 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
= pet) = 1000e(ln2)t
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
FirstOrderDifferential 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 twodimensional plane,and no two intersect.Moreover, the selection e amounts to a determination of P(O).Because xanyone point Po on the Paxis 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.)
in
of a realworld 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 realworld situationfor example,answering the question originally posed.)
1.The formulation
3.
FIGURE
modeling.)
In the population example,the realworld 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 realworld 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 realworld situationtheactual bacteria populationwe 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
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 realworld 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 realworld 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 realworld 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 realworld experience. But in Example6 we simply ignoredany complicating factors that
dx
if
xiC.
= (C  x)2
y(x) =
1
2
Y)
Thus)
Cx)
(8))
on any interval of real numbers not containing the point x = C.Actually, Eq. (8) defines a oneparameter 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)
Ix) 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,\"'
FirstOrderDifferential Equations)
\"\"\"''''''I'. \"\"',
M,,\",..,
UW
u,.,.
Example8)
Verify
that
the
function
+ y = 0)
(10))
Solution
y/(x) =
Then substitution
_!x/ 2 1nx
1
and
y\"(x)
= ix3/2Inx 4x3/2.)
into
Eq. (10)yields)
3/ + y = 4x2 (ix3/2Inx!x2) + 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 (realvalued) 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 (realvalued) 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 firstorder equations, and)
y(4)
is a fourthorder equation. The most general form of an nthorder differential with independent variable x and unknown v function or dependent ariable equation
y=y(x)is)
(13))
where F is a specific realvalued 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
interval.
a differentiablefunction on continuous there. This is why only a continuous necessarily as a (differentiable) solution of a differential equation on an
.)))
Example7)
Continued
showsthe two \"connected\" ranchesof the graph y = 1/(1 x). The b Figure lefthand 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 righthand 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)
(14))
two
successive differentiationsyield)
x=l)
;;:..
0)
for all x. Consequently, Eq. (14)defines what it is natural to call a twoparameter differential equation) family of solutionsof the secondorder 5
5)
o
x)
5)
y\"
+ 9y = 0)
(15))
FIGURE
y/
of
solutions.
5)
the differential equations in (11) (12)are exceptionso the gent and Although eral rule, we will seethat an nthorder differential equation ordinarily has an nparameter family of solutionsone 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 socalled (n) (nI) , G (16) y
In both
/\" x,y,y,y,...,y)
5
3)
o
x)
3)
FIGURE
solutions
Y2(X) Y3(X)
YI
3 cos3x,
where G is a realvalued function of n + 1 variables. In addition, we will always seek realvaluedsolutions 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
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 concentrateonfirstorder differentialequationsof the form
dy
dx
= f(x,y).)
(17))))
Chapter FirstOrderDifferential
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)
dy
Yo)
Example10)
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
 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 lefthand 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 righthand branch passesthrough therefore the graph on (0) of the solution of the different initial value problem
\037)
1.1.7
5
y/
(\037,
.)
5)
o
x)
5)
FIGURE
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 simpletechniquesseparation 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 firstorder equations having impressive applications.)
\037\302\243\037\037\037
\037\037\037
In
Problems
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)
that
y/
eX
equation.
with
derivatives
y\" y\"
2x
YI
Y2
2x
YI
eX
Y2
eX sin
= In x;YI = X
In
x,Y2 =
1 x)))
In
12. 2y\" x
In
xy' +
=
YI
2y = 0; YI
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
= erx
36.
2y
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
diseaseand
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 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
38. ' = y y
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
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 xaxisat 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, writein the manner of Eqs.(3) through (6) of this sectiona differential equation that is a mathematical model of the situation described.
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 oneparameter 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)
= kv 2.Theinitial
coasting in water
= = 10. =
= Ij(
1.1.8
=
 x).
b?)))
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)
FirstOrderDifferential Equations)
100
80 60
C=2C=1c=oC=1C=2 C=3
1)
C=4)
\037
40
\037
C=4
1) 2)
20 0
2 3 4 5
y
FIGURE
equation dy/dx
of the
FIGURE
= Cx4 for
various values
that
of differentiable
y(x)
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 whyin contrast to the situaexist infinitely many tion in part (c) of Problem 47there 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_.______________________)
= 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
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)
(3))
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
1.2.1
problem) dy
\037)
dx
= I(x), y(xo)=
Yo.)))
11)
6)
C=1
C=2)
\037
4)
2)
0)
0)
1 2 3 4 4321 0
x)
\037
2)
4) 1 2 3 4
6 6
C=4
4 2
0
x)
We will see that this is the typical pattern for solutionsof firstorderdifferential 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 firstorder differential equation is simply a oneparameter 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.
y(x} =
2
\037
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)
FIGURE
1.2.3. Solution
curves
equation in
C=
Example 1.)
12
Chapter
1
FirstOrderDifferential Equations)
Secondorder equations.The observation that the special firstorder equation dyjdx = f(x) is readily solvable (providedthat an anti derivativeof f can be found)
extendsto secondorder differential equations of the specialform) d2y
dx2
in
= g(x),)
(4))
which the function g on the righthand side involvesneither the dependent ariv able y nor its derivativedyjdx. We simply integrate onceto obtain)
=
\037\037
C\\,)
y(x) =
C\\x
+ C2,)
where C2 is a secondarbitrary constant. In effect, the secondorder ifferential d in (4) is one that can be solved by solving successively firstorder 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 xaxis) described its is by positionfunction)
x = f(t))
giving its
(5))
\037)
= f' (t);)
that is,)
v=
.
dx
dt)
(6))
 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)
i
to)
v(s) ds,
theorem of calculus
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))
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
(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
FirstOrderDifferential 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))
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
metersthats, 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 unitssuch distancein milesor as use kilometers time in hoursin and situations (suchas in a probleminvolving special an auto trip). However, the footpoundsecondfps) and meterkilogramsecond ( (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
'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:
... .. .... ','
..
....
pound slug
(lb)
newton (N)
foot (ft)
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.
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
In.)
follows that)
1 mi
ft)
Thus a postedU.S. international speedlimit of 50mi/h means thatin legalspeedlimit is about 50x 1.609 80.45 lan/h.)
\037
termsthe
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
so its massis)
= 444.8N2
9.8 m/s
\037
45.4 kg.
To discussvertical motion it is natural to choose yaxisas 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)
(16)) (17))
yet)
(t
16
Chapter
1
FirstOrderDifferential Equations)
Example3)
its maximum height when its velocity (Eq. (16)) zero,) is vet) and
thus
straight
reaches
0) with
when
is)
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
yaxis)
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 yaxisits 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
(a,
0))
(a,0)
xaxis)
= Vo
(1 =:)
.)
(18))
VR)
Vs)
FIGURE
1.2.5.swimmer's A
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,
a=
.
VR
Vs)
using
(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)))
17)
y(x) = 3x
Then)
y
 4x +
3
1.)
.)
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)
19.
VX2
+ 9; y(4) = 0
dy
dx
x 2 + 1 y(O) 1 ; y(O) = 0 I x V
2)
10 ;
=0
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.
16. (t) = a
17.a(t)
==
1 vt
1,
8)
+4
1
, Vo
= 1,xo 1
==
\037)
6)
4)
(5, 5):)
'
8)
2)
In Problems through 22, a particle starts at the origin and travels along the xaxis 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.
21.)))
18
22.
10 8 6
;:::..
Chapter
1
FirstOrderDifferential 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 200fttall 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
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.
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
36.Supposea woman
= ,J2g . h
h above which it
the on the moonwhere surface gravitational acceleration is (approximately) 5.3 ft/s 2howhigh 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
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
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,
of the
= 3 mi/h
is given
10
VR
= Vo
dv
dt)
(1
\037:
))
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.
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 fastin km/hwasthe 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
19)
fires a prosimultaneously jectile(straight ahead in the samedirection) that travels at 8 onetenth of the speed = 3 x 10m1s of light. How long c will it take the spacecrafto catch up with the projectile,) 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
dx
= I(x,y))
(1))
y)
x)
where the righthand 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 simplelooking 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
. . .
A 1.3.1.solution
curve
slopeml
= Xl
slopem2 = X2
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 xyplane, 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 passesthat y'(x) = I(x, curve of the differential equation y' = (x,y)thegraph of a solution of the equationis a curve in the xyplane 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
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
FirstOrderDifferential 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 43
\\ \\ \\ \\ \\ \\ \\ \\
2 3 4
x)
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.)
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,
given equation. The numerical at the intersection of the horizontal xrow and the very appears tical ycolumn of the table. If you inspectthe pattern of upperleft to lowerright diagonalsin this table, you can seethat it was easily and quickly constructed.(Of)))
=x
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
2 3
y for
1
  // / /
\\
'\" '\"
/)
;::....
\\
'\"
5 4
3
2
1
;::....
\"')
\\
\\
'\"
\\
'\"
0)
\\
5
/ /
'\"
\\
 / /
'\"
/
/)
5)
o
x)
2 3 4 5
1
5)
o
x)
5)
FIGURE
curve through
(4,4).)
y) on the righthand 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
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.)
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
FirstOrderDifferential Equations)
the
appropriate solution curve in this figure, infer that y(3) initial value problemy' = x y, y(4)= 4?)
\037
dv
dt
=g
kv
.)
(3))
= 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
throw
100 / / / / / / / / / / / / / / / / / / / / / / / / / / / / / 5 10 15 20
/) t)
dv
dt)
= 32 0.16v.
(4))
25)
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 impliesthathowever as you throw itthe 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 = 200x 60
88 ftls
mi 136.36
h)
.
milh 100
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 ftlswhateverinitial velocity it starts with. You might like to verify .))) o absence f air resistancethis would hit the ground at over mi/h. ball
200,
300
200 thatin
23)
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 longterm basis(in terms of the maximum availablefood, for instance).)
that
Example4)
If we take k
dP
dt)
the
form)
(6))
50
t)
75)
100)
FIGURE
typical solution
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 impliesthatwhatever the initial populationtheopulation p t a pet) approacheshe limiting population P = 150s t 00.
1.3.8
\037
P'= O.06P
 O.0004p
2
curves for
.)
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
.)
Example5)
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 yaxisto plunge downward so that none can passthrough the point (0,
1.3.9,
0).)))
24
>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
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 xaxis 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 xaxis, and its partial derivative jay are continuous provided that both the function a near the point (a,b) in the xyplane.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 xyplanethat 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 xinterval of Theorem and the solution curve y y (x) b).) through the point
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.
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 xyplane,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 solutionnecessarily a continuous functionof 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 someopenxinterval containing a
o)
y(x) = 1 x)
that
4)
2
4)
2)
o
x)
2)
FIGURE
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 rectanglewherein solutionsof the differential equation are guaranteed to existbefore 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 manyeven infinitely
manysolutions.)
.....,..... \037\"A\"'\" ........h
\"......
\037
..
\037
....
.....
.......A.......
........
,......
..,,,
......
Example6)
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 xyplane where x i= O. that) Indeed,we seeimmediately by substitution in (11)
y(x) = Cx
2)
(12))))
26
4)
Chapter FirstOrderDifferential
1
Equations)
(0,b)
(0,0))
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 xaxis
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 yaxis. 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.)
2 if parabolasy = Cx . Hence, a
point off the yaxisthere 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
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 yaxisfor Then for any value of the constant C the function defined by)
1.3.14.
42) FIGURE
1)
o
x)
1)
2)
1.3.14. Thereare
(1,1).) 
curves
y(x) = Cx)2 if x I
x2
if x
< 0, > 0)
(16))
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 x00< x < axis,but this solution is unique only on the negativexaxis
(1,
1.3.14.
O.)))
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)
\\
2)
\\e\\
.'
\\ \\
\\ \\ \\ \\ \\ \\\\ \\\\
\\ \\
\\\\\\\\\",\\\\\\\",..../
\\\\\\ \\\\\\
\\ \\ \\ \\
0)
,, ,, ,)
\\ \\ \\
\\
\\
'Y
\\e\\ \\ \\ \\
\\.
\\
\\ \\ \\
\\
1) 2)
\\ \\
\\ \\
\\
\\
\\ \\ \\
\\ \\ \\
\\ \\
\\
\\ \\
\\ \\
1)
\\ \\
\\
\\
\\
\\
\\
\\
0)
I I I / I I
2 1) o FIGURE1.3.18.)
X)
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)
1)
0)
,,,
\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)
\\\\\\\\\\\\\\\\\\ \\\\\\\\\\\\\\\\\\)
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)
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
FirstOrderDifferential Equations)
. = SIn x + SIn y
........................................
\037
2 1
\037
.............,) ,..../)
1)
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
,/ /1
12.dy = x In y; dx
dy 14.dx = \037;
dy
y(1)=
1
y(1)= 1
1 2 3 3
/ / /
y(O) = 0
2
FIGURE1.3.22.
y(2) = 2 y(2) = 1
dy 9. =X2 y2 dx
3 2 1
\037
y(O) = 0 y(O) = 1
0
I / / / / / /
/ /
/
1 / 2 /// // / / / 3 2 / 1 3
/ /
It
/ /
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,
y(4) = 0; y(4)=?)))
29)
21
29.Verify
that
if c is a
constant,
defined
by) piecewise
23.y/ =
24.
 1,
y(x) satisfiesthe
y2 ,
= 0;
{(x
for
c))
for
x x
< c, >
c)
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 xyplane 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.)
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
limiting
deerpopulation?)
30.Verify
that
if c is a
constant,
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.
= f(x,
y(x)
= cos(x
+1
11)
that
if c is a constant,
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/ 2JY, y (0) 0 has infinitely many different solutions. (b) For what values of b doesthe initial value problem y/ 2JY, y(O) b have (i) no solution, (ii) a unique solution that is defined for all
= c).
equation
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
= y,
satisfies the differential equation y/ = 4xJY 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
FirstOrderDifferential 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 constantvalued 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. =
direction
values at x
ential equation
field
1 of the
initial
values
(b)
3 of the
tial equation
y( 3)= 2.99.)
3)= 3.01
initial
..........\"\"
\\
;::..,.
I \\
1\\
\\
1\\
..........\"
\\
1\\
\\\"\" \"\"
..........
\\
\\ \"
Thelessonof this problem is that small changesin conditions can make big differences results.) in
35.(a) Usethe
direction
==.
values at x
ential equation
..........\"\"
\\ \\
\"..........)
\\ \\ \\
+ 1 with
initial
values
\\ \\
\\
\\
(b)
3 of the
tial equation
x)
3)= +0.5.
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).)
with
5< x,
initial y
and solution
curves for
the differential
. = sln(x
y)
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):
31)
{x, Plot.Vect.orField[{l, Sin[xy]}, 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
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 userfriendly MATLABbased
dfield
1999]
1.3.29),
dfield
for y'
= sin(x
dfield setup
to construct slopefield
y).)
Usea graphing
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
= ax + b in
y(x)=x2tan(
1
X2C . xc) )
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
FirstOrderDifferential 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
(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
_\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)
dy
dx)
= H(x,y)
(1))
dy
dx
= g(x)h(y)= g(x) ,
fey))
where h (y) = 1/ (y). In this case variables x and y can be separatedisolated the on oppositesidesof an equationbywriting informally the equation)
fey) dy = g(x)dx,)
which we understand to
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;)))
33)
(3))
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,
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)
dx
Solution
Informally,
= 6xy,
y(O) = 7.)
y
sideby dx to get)
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
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/
y(x) = 7e
3x2
.)
Example 1.)
6xyin
in
Fig.1.4.1.)
.)))
34
Chapter
1
FirstOrderDifferential 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.)
y(x) =
Fig.1.4.1.)
.)
,,,,nn\"'\"''
In\"
\"\"\"
nn\"n,,,',
Example2)
2x  dx
dy
3y2
5.)
(5))
Solution
6)
When
, .. . \"\",.\".\"'\"')
. : :
,
.)
)
(1, 3) ,
.........\"\"\"
f (3l5)dy = f (4 22x)dx; _
y3
5y
= 4x x + C.)
(6))
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
function)
Example 2.)
= (4
 2x)/(3  5)
y
_
2
in
H(x,y) = x2 4x + y3
5
y)
 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 1.4.2the 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,
.)))
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
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)
 
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 ysolutions of Eq. (8)for xvalues 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
.)
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
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:
(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
(9))))
36
Chapter
1
FirstOrderDifferential Equations)
having the
factor (y
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)
15105
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 linearfirstorder differential equation is contained in its general solution. By contrast, it is common for a nonlinear firstorder 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 constantvalued 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
1.5
1.4.5,
.................
Example4)
dy
dx)
= 6x(y 1)2/3.
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.
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
FirstOrderDifferential 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
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
.)
Because is a constant, so is A C
== eC It
kt x(t) == xoe
(16))
in
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
39)
x = Xo e (k > 0))
kt
Xo)
Xo)
t)
growth.)
Natural FIGURE1.4.8.
decay.)
Example5)
the According to data listed at www.censlls.gov, world'stotal population reached 6 billion personsin mid1999, 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 increaseenfoldtherebyreach(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
natural
growth
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
60== 6eO.0129t;)
and
thus in
that
is, when
t ==
0.0129) 178;
\037
In
10
.)))
40
Chapter
1
FirstOrderDifferential Equations)
Note: Actually, the rate of growth of the world population is expectedto slow somewhat during the next halfcentury, 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 halflife of the isotope,because parameter is more this convenient. The halflife 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))
Example6)
Solution
turns out to contain 63% as much 14C specimenof charcoal found at Stonehenge as a sampleof presentday 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
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 bedates 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 firstorderdifferential equation with constant coefficients:)
\037)
dx =
ax + b.
(20))
It
Example7)
4lb 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.
150\302\260F
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
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.
= 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 plausibleand under ideal conditionstrue,
the velocity of water exiting
through
= 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 crosssectional of the tank at height y. Then, appliedto a thin horizontal))) area
42
Chapter
1
FirstOrderDifferential Equations)
A (y )
and
1
Y
A (y ) d y
\302\267)
hencethat)
A(y)
and)
dV
dt
From
dV
dy
. dy
dt
= A(y) dy
dt
\302\267)
(23))
obtain)
(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 yvalues)
With
2 dy
)= Jr
dt
5/2
2
V
/2.32y;
\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)
= 0,thus
= 72 .
\037
41\0378
that
to
.)))
drain.
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
dy
x 2) dy dx
= 2y
(x
17.y' =
18. 2y' = 1  x 2 + y2 _ x 2 x
Find explicit in Problems
side. )
 1)y5 
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.
x)
righthand
19 through
X
particular
solutions
28.
of the
initial
value problems
dy 19.dx = ye ,
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.
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 onesixth as much 14C carbon extracted from presentday bone.How old is the skull?
from a purported relic 10 atoms of 14C contained 4.6 x Carbon extracted from a presentday 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
y(l) = 3 y(l) = 1
I Y\"2]\"[
10 5.0 10
dx +
1 = 2y,
= y,
I =\"2]\"[
37. (Continuously
x dx
dx
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,
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.
(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
also that sodium pentobarbitol is elimweight. Suppose inated exponentially from the dog'sbloodstream, with a halflife of 5 h. What single dose should be administered in order to anesthetize a 50kgdog for 1 h?
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 halflife 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?
cataclysmperhapsthe
itself
4.51 10
10 pu
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 halflife x 10 is about 1.28 9 years) and that oneof every nine potas
(c) If it
reaches100pu,
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 longtotal number of months or fraction thereofwill 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).
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
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)))
45)
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
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 yaxis. 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 yaxis. 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
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?)
it
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)
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
of
where k
dx 1 k=dt
t)
be empty?)
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)
:30
(0.6)nr
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 12hwater clockis to be designedwith the dimensions shown in Fig. 1.4.10,
shaped like the surface obtained by revolving the curve (x) around the yaxis. 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 6month 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 solutionthe curve of minimal descentime is
1697.
an)))
46
Chapter
1
FirstOrderDifferential 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 xaxis (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 cycloidtothe 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 yaxis pointing downward. Then a mechanical
analogue
of Snell's law
a denotesthe angle of deflection (from the vertiof the tangent line to the curvesocota = y' (x) cal)
where
in
opticsimplies
constant,
that)
sIn
a=
(i))
v)
descended distance y a
mgy
= 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 firstorder equation)
= j
p.
a FIGURE
(a)
First
dv
dx
2 = J1+v.
wirethe brachistochroneproblem.)
derive from
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)
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)
In
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
.=2x; dx
1 dy
y
that
).
(2))))
a Because sideof the equation in (2)is recognizable s a derivative (with respect each
to the independent variable x), all
that
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 firstorder 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)
Because)
Dx
[IP(x)dx] = P(x),)
alent to)
e Dx [y(x).f P(X)dX]
Integration of both
f = Q(x)e P(x)dx.
1(Q(x)e
f
P(X)dX)
dx + C.)
Finally,
in
(3):)
[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 OFFIRSTORDERQUATIONS E factor p (x) = ef P(x)dx. 2. Then multiply both sidesof the differential equation by p (x). the Next, recognize lefthand side of the resulting equation as the derivative of a product:) Dx [p(x)y(x)]= p(x)Q(x).)))
3.
48
Chapter
1
FirstOrderDifferential Equations)
4. Finally,
p(x)y(x)=
then tion.)
d f p(x)Q(x) x + C,)
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)
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
ex / 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
,)
(7))
11
e4x/3.
Hence integration
with
eXy =
and multiplication by
eX
8 e4x/3dx
= _ e4x/3+ C,
\037\037
y(x) = Ce
Substitution solution is)
j\037
ex/3.)
1
(8))
of x
0 and
1now gives C =
1
j\037
32
y(x) = 32eX
1
.)))
1.5 LinearFirstOrderEquations
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
.)
initial condition y(O) = Yo. The two types of behavior are separated by the particular solution y(x) = ex / 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 yaxis 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
.)
(x2 + Solution
After
+ 1)dy
dx)
3xy
= 6x.
(9))
+x + y= x dx
dy
3x
6x
1)
2+
Multiplication by)
p(x) = exp
yields)
(I
x23
:
2
1).
1 dX)
(x2 + 1)3/2dx)
7
dy
and
thus)
6 5 4
3
;::....
2
1
(x2 + 1)3/2y=
1 2 3 54321 0
x)
4 5
(10)
FIGURE
.)
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
FirstOrderDifferential 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
1.5.2.
.)
The preceding derivation of the solution in Eq. (6)of the linear firstorder 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 isas unique value of C such that y(xo) = Yo. previously noteda
on
theorem.) we Consequently, have proved the following existenceuniqueness
(Q(x)e P(X)dX) existon I. Our derivationof Eq. (6)showsthat if y = y (x) is a solution of Eq. (3)
I containing
(11))
dx + P(x)y
dy
= Q(x), y(xo) =
Yo)
in
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
Theorem 1 tellsus that every solution ofEq. 3)is included in ( Eq. (6). Thus a linearfirstorderdifferential 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
(12))))
dt
XQ)
LinearFirstOrderEquations 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)
( 1)
= Yo.
(13))
dy 1
= dx + y x X
sinx
2)
with
IS)
Xo
1 the
integrating
factor in (12)
p(x) = exp
Yo
sin t dt + [X
11
\302\267)
(14))
rule, given x) needto be approximatednumericallyusing 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
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
sin t
FIGURE
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 exceptionrather the rulewhena 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
it
52
Chapter
1
FirstOrderDifferential Equations)
MixtureProblems)
As a first application of linear firstorder equations, we consider tank containing a a mixture of solute and solventsuch 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 tankkept a mixedby stirringflows 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
solutiona
\037x
Input:
Yi
LIs,
Ci
g/L)
\037t
\037t
liters
co(t) =
\037)
ri
second
(\037t
seconds)
Output:
Yo
Lis,
g/L)
Co
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
We now
divide by
\037t:)
\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
Co denotesthe
co(t)=
of solute in the tank at time the differential equation)
x(t)
Vet))
 roc
o,
dt Vet) = Va + (ri
satisfies
dx =
rici
x(t)
 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 equationexamination the behavior of the system of
(18).
cubickilometers.
.)
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 LakeEriecaused past by five industrial pollution that has now beenorderedto ceaseis 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?)
350(kIn 3/yr),
concentrationof LakeHuron), and
separable equation)
dt
this
notation,
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) = ept
[xo
it qe
+
.)
Pt
dt]
= ept
[xo
;
= ert / V
x(t) == cV
[SCV
(e /
\037\037
rt v
 1)]
In
;
4
(ePt
1)]
(21))
+ 4cV ert/V
for
t ==
;
4 ==
480
350
In
\037
1.901 (years).) .)
Example5)
in 120gallon 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 wellstirred 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)\037t3
(90+ )
t)))
\037t,
54
Chapter
1
dx
dt
3 90 +
t)
x = 8.
An integrating
factor is)
p(x) = exp
which gives) Dt
(I
3 90 +
t)
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.)
.)
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.
3y
27. (x + ye
of dy/dx 2
(X
)=
dy dx)

of the
function) error
= 1 + 2xy in terms
erf(x) =
.j7r10
initial
et2 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
55)
31
that
31.a) Show (
that)
that)
dt y(t),
5x
100
200'
5y
find
ever in
Yp(x)
= ef 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 wellmixed 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 mixturekept 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
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 inflowoutflow 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 400galtank 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 wellmixed 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
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) = nnenjnL (d) Conclude from Stirling's n approximation n! nne J2nn that Mn (2nn)1/2.
\037 \037 :>.
woman accepts an engineering 30yearold 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 secondlawaccording 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
+ y.)
= vto
d
dt
(mv)
= mg, =
v(O) = 0,
where m is the variable mass of the hailstone, v dyjdt is its velocity, and the positive yaxis points downward. Then show that dvjdt gj4. Thus the hailstone falls as though it were under onefourth 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 wellmixed 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'slongterm pollutant content. How long doesit
:>.
2 4 6 8 10
5) o
x) 5)
10 8 6 4 2 0
in the
reservoir to reach
x=20)
46
10 20 30 40 50 60
curves for
 y.
and solution
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.)
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
concentration
c(t)
10
57)
oscillatory
that the graph of x (t) does,indeed,resemble the curve shown in Fig. How long does
1.5.9.
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)
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 daytoday 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 firstorderdifferential equation)
5,
5,J3
(2))
du dt
that is,)
= k(u
A(t));)
du
dt)
(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
FirstOrderDifferential 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
= 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
.
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 longterm \"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

nt
(6))
10
30
40
Consequently, the longterm 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 Observethatwhatever the to hours initial temperaturethe 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
'\"
\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
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 wintersummerdifference 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
59)
dy
dx)
= f(x,y),
(1))
with
combination)
v=a(x,y))
of x and
equation)
y that
(2))
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
x(4))
yields)
dy
dx
')
where the partial derivatives af3jax = f3x(x, v) and af3jav = f3v(x, v) are known functions of x and v. If we substitute the righthand 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)
dy
dx)))
= (x + y + 3)2.
60
Chapter
1
FirstOrderDifferential Equations)
As indicated earlier,
v Then)
Solution
let's try
the
substitution) that
= x + y + 3;)
dy
is,
=v x

3.)
 dx
dv
dv
dx)
1,
dx)
= 1 + v 2.
in
obtaining its
solution)
x=
So v =
tan(x
dv
 C).Because = x +
v
l+v) 2
y
= tan
+ C.
 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
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
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 firstorder equations for which there are standard substitutions that are known to
succeed.)
Homogeneous Equations
A homogeneous firstorder 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))))
61)
x = F(v) dv
dx)
v.
Thus every homogeneousirstorderdifferential 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 resultbecause yn Ix +n = (yIx)n,and xm m
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.)
Example2)
the form)
dy
dx)
4x2 +3y2
2xy
=2
()
y
Y + 2 (x) ) .
The substitutions
y
in
= 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
FirstOrderDifferential 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
Note that the lefthand 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 positivevalued and negativevalued 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)
x dy = y + x2 dx
where Xo
J x+
Y
y2,
y(xo) = 0,)
> O.)
x and find dx
dy =
that)
Solution
We
)1( ) 2,
Y
x)
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 == sin0 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 
63)
\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)
= yln)
(10))
transforms
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)
........
.....
....
the
form)
y=, dx
dy
3 2x
2x
y
we
n
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 =
2xvI/2.
 dv
dx
x)
= 4x
with integrating
factor p
3 4 Dx(x v) = 2\"; x)
x3 v = x3y2
y2
+ C; = + C;
4
x)
x)
= 4x2 Cx3 +
.)
.)))
64
Chapter
1
FirstOrderDifferential Equations)
The equation)
Example5)
x dy + 6y == 3Xy 4/3 dx
n ==
,)
 1.he T
n ==
it
substitutions)
v == yI/3,)
== v
3
and)
transform
it into)
  dy == dy dv == dx dv dx
3v_4dv dx)
Division by
2 dv v==1 dx
x) with integrating
2 Dx(x v)
and finally,)
==
2\"; x2v ==  + 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
x
and hence)
f 3x dx==x +c,)
y(x) == 41nIx
+ cxl.)
.)))
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
that Suppose an airplanedeparts from the point (a,0) locateddue eastof its intended
1.6.5
(a,0)
x)
dy. dt dt
Vo Sin
VOX
x2
'
y2
voY
x2
Yx2
+y2
/'/'
of the
/' /'/' 8
/'/'
x)
/'/'/\037
Iy I
I I I
\037
:,
y)
dx/dt
VOX)
+ w.
y2)
equation)
(voY
w
J
w
x2
+ y2 ) .
(12))
If we set)
k
=,
Vo)
\037
(13))
FIGURE
velocity airplane.)
neousform)
=
\037\037 \037
_k[1+()2]'/2.)
to)
(14))
The substitution
f ,Jl +
By trigonometric substitution, we find that)
In
v2
  f x.)
k
dx
(15))
integral
on
the left,
(v + J1 + v ) = kInx + C,
= y(a)/a = 0 yields)
C = kIn a.)
(16))
and the
initial
condition v(a)
(17))
=
\037
[G
X
 G ) l)
k
(18))
Because = XV, y
we finally
obtain)
y(x) _ 2
  [()
a a
lk
X
a)
l+k
(19))
66
Chapter
1
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))
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
in
Eq. (20)yields)
5
dx
==
\037
2 5
) [ (200
(200 ] )
400
\037
')
Ymax
==
100
< 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 firstorder 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
vided that Eq. (21)implicitly defines y as a differentiable function the original differential equation in the form)
respectto x. Pro
of x, this gives
==
0;
(22))
that is,)
== Fy(x,y).)))
67)
form)
(23))
form. The general firstorder 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
implicitly
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 mixedsecondorder partial derivatives Fxy and Fyx are continuous on an open set in the xyplane,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)
.
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)
dx + 3xy 2 dy = 0)
(25))
= y3
and
Fy
F(x,y) = xy3 has the function = 3xy 2. Thus a general solutionof Eq. (25)is)
xy3
= c;)
.)))
kxI / 3
.)
68
Chapter
1
FirstOrderDifferential Equations)
But
8 by y2 to obtain)
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)
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
they
is exact.)
THEOREM 1
tinuous
Supposethat the functions M(x,y) and N (x,y) are continuous and have confirstorder 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)
 a f M(x,y)dx.
ay)
(28))))
69)
To seethat there is such a function of y, it sufficesto show that the righthand sidein Eq. (28)is a function of y alone.We can then find g(y) by integrating with respect to y. Because righthand 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
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) =
(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)
exact ecause) b
== 6xy
y3
and
 .
N(x,y)
==
 3xy2) 4y + 3x  3
2 2
y
dy == O.)
(30))
aM
ay
== 6x
3
2 == aN
ax
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
aF =
ay)))
2 2 3x _ 3xi+ g'(y) = 4y + 3x
 3xi,
70
Chapter
1
FirstOrderDifferential Equations)
and
it
thus)
F(x,y)
equation)
2 == 3x y
xy3
+ 2y 2 + C
1 .)
by
the
2 3x y
xy3
+ 2y 2 == C)
(31))
1 2 3
4
into
.)
5
54321 0
x)
2 3 4 5
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 figureat(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.
.)
function
== O.)
(32))
If either the dependentvariable y or the independent variable x is missingfrom a secondorder 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))
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 secondorder in of differential equation).)))
and ExactEquations 1.6 Substitution Methods Example10) Solvethe equation xy\" + 2y' == 6x in
Solution The substitution
which the
71)
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
10
x)
5
54321012345
y
+ C + C2 x)
I
:f:100.)
= 0,
+ x
\0373,
of the secondorder 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.)
.)
(35))
Then the
\037)
p==y)
dy
\"
dx')
 dx dp
\037\037
  dp dy dy dx p dp
dy)
(36))
(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 secondorder ifferential equation.))) d
72
Chapter
1
FirstOrderDifferential 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 firstorder equation) dp yp = P2.
dy)
f d:= f d;, =
In
y p In y + C (because = C1y) p
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)
 C2) =
1
= (y')2is)
Ae Bx
,)
where A
1
that
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 xaxis, nd with negative values of B they are reflected in the yaxis. 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
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 
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)))
73)
31
60.Usethe
equation)
method
in
Problem
59 to solvethe
differential
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
(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?
curves of the
differential
equation)
37. (cosx+lny)dx+
( + +eY)dY=O x
\037
dx
))
=0
y) dx
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 secondorderifferd ential equation in Problems4354. 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
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
In
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))
parameter
is called a Clairaut equation. Show that by) family of straight lines described y(x)
one(38))
57. Show
ential
= 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
 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
i
y
i(y')2)
that the line)
= Cx
+y +3
 iC2)
=
iC2).
= +
hand so that
dv
du
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 oneparameterfamily of straight line solutions are illustrated in Fig. 1.6.10.))) is tangent
74
Chapter
1
FirstOrderDifferential 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
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 secondorder
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.)
only
plane
curve
l/r.)
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 incrementsthat by one birth or death at a time. o Supposethat the population changesonly by the occurrencef births and deathstheres 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)
births:
length
\037t
fJ
(t) . P(t) .
in
\037
,)
\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
+ 0,sotaking
(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)Pand thus increasess the population doesthen (1)gives the initial value problem) a Eq. dP = 2 , (0.0005)P P(O) = 100 dt)
(with
t in
= dt; f ;2 dP f (0.0005)
1
Substitution
 P = (0.0005)tC. +
we readily solve for)
of 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)
1.7.1.
solution
1.)))
curves for
the equation
76
Chapter
1
FirstOrderDifferential Equations)
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 ==aPbP dt
(2))
the form)
dP
\037)
dt
== kP(M
 P),)
(3))
Example2)
In
by
dt)
==
(4))
grate.We
get)
J P(1  P) = J 0.0004dt, _ J ( + 150 P ) dP = J 0.0004dt [partial fractions], + IPIIn1150PI 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
P(t) ==
(5))))
1.7PopulationModels 77)
p)
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)
100)
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
(7))
populations are positive valued. If Po == M, then (7)reduces to the unchanging (constantvalued) \"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
Po + {pos. number})
<
P(t) ==
p)
MPo
Po + (M Po)e
MPo
kMt)
MPo
Po)
==
M.
MPo
Po)
== M.
lim
t\037+oo
P(t) ==
MPo
Po + 0)
== M.
(8))
M/2)
P = M/2)
curve
P == M/2 has an
on this
below the
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 longterm 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
FirstOrderDifferential 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),
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
.)
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 19thcentury 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
,)
so r
==
growth
P (t) == (5.308)e(O.026643)t)
(with t in
(9))
years and
lation
growth
and between1800
.)
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)
(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 18001990 censuspopulafigures with those predictedby the exponential growth model in (9) and the Both agreewell with the 19thcenturyfigures. 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 populationpredicting a billion in the year 2000whereas 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
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
92.228
129.849 169.492
221.237
288.780 376.943 492.023 642.236 838.308 1094.240)
76.213
90.834
0.001
1.394
0.545
248.710
281.422)
2000)
Exponential)
812.818) 177.038
growth and
589.598
40%) 20%)
U.S.ensus c
40%)
FIGURE1.7.5. Percentage
errors in the exponential and logistic population models for
18001950.)
The two modelsare comparedin Fig. 1.7.5, where plots of their respective shown for the 18001950 actual populationare We seethat the logisticmodel tracks the actual populationreasonablywell period. this ISOyear 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 18001900 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.
errorsaspercentage f the a o
80
Chapter
1
FirstOrderDifferential 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.)
describeome situations s
that
which
the
1.Limited environment
 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 constantsize 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
. pet) == 10 90elOOkt +
Then substitution of t ==
1000
(12))
1000 10 90e+
100k)))
1.7PopulationModels 81)
that
e100k ==
With
\037,
so
k == 160In
\037
\037
0.008109.
which we solve for e100kt== 316 It follows that 80% of the population has heard the rumor when)
t==
thus
In36
lOOk
9 4)
\037
4.42,
.)
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 righthand sidein Eq. (13)is the negative of the righthand 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,
==
C,)
P150) P
==
Ce ::i:e0.06t== Be0.06t
(15))))
82
Chapter
1
FirstOrderDifferential Equations)
(a) Substitution
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 situationaeal 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)
P'
variables and use partial fractionsto solvethe initial Separate value problems in Problems 18. either the exactsolution Use ora computergeneratedslope 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)
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(5x),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?
IjJP.
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))
11.upposethat when S
JP.
a certain birth and death rates f3 and tional to (a) Show that)
= 4x(7 x),x(O)= 11
17)
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?)))
dozenalligators
1988,
P.
in the
1998. 22.
swamp?
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
now happens
to the
rabbit population
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
= 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
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
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)
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)
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.))
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
FirstOrderDifferential Equations)
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
doomsday
occur?
initial
value problem
P' =
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
Throughout this period,P (t) remained tion of the initial value problem
pet) (t
PII(t) = 2k
PcP 4M)(p
M).)
closeto
123.2 million.
the
solu
dP
dt (a)
What
dict?
(h) What
population
doesit predict?
continued
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)
17901840 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 Sshaped 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
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
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.
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.)
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,
r(t) =
the
sameinitial
equation populations
P'= k P (sameconstant
after compare the
value
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 (downwarddirected) 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) ==
yet) is given
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
2 (m).) (4.9)(5+ (49)(5)== 122.5 ) to the ground when y == (4.9)t(t 10) == 0,and thus
== y(5) ==
Now we want
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
FirstOrderDifferential 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
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 ydirection 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,)
Thus)
\037)
dv
dt)
= pv g,
(4))
where p
should verify for yourself that if the positive yaxiswere directeddownward, then Eq.(4) would take the form dv/dt = pv + g. firstorderdifferential equation, and its solution is) Equation (4) is a separable
v(t)
= k jm > O. You
(v
+
\037
) e
pt
\037
.)
(5))
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))))
occasional survival of peoplewho fall planes. We now rewrite Eq. (5)in the form)
dy
dt)
(8))
Integration gives)
We substitute
we find
that
C = Yo
0 for t
+ (va
yet)
1 P = (vo v,)et + v,t + C. p and let Yo = y(O) denote the initial height
yet)
v, )Ip,and so)
1 = Yo + V,t + (va
p)
 v,)(le ).
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 someparachutejump 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
v,
 
.)
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
122.5
= 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
FirstOrderDifferential 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
.)
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 ydirection 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
> O. We must
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
13
substitution
:
v
(12))
apply the
= v Jpig and
to derive the
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)
(14))))
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
to make the 1
substitution
= v v'pjg and
+c
f!
tanh
(C2
I
 t..[jig)
I
with)
C2 = tanh 1
(vo
ID
.)
(16))
p
1
In
(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
in
the
caseof downward
lv,
I
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
FirstOrderDifferential 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.)
functions
in
the Comparisonof the last two lines of data here indicateslittle differencefor motion of our crossbow boltbetween 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 situationssuch for instance, the atmospheric reentry and descentof a space as,
vehicle.
.)
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 freefalling 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).)))
variable t, we substitute)
dr =, r   dr dr dt)
.vd2 dt 2
dv dv dv
dt
dt
dr)
(as in Eq. (36)of Section and obtain the firstorder equation) 1.6)
dv v = T  GM r dr
2)
FIGURE
with
r. Integration with
equation)
1 GM v2 =Tr++C 2
r)
(21))
that
= 4).
Beforegnition:Substitution i
1 v2 = GM + C r) 2
 GM/ro
with)
Vo
1h
3600 s)
m 14770
km
36
s)
and ro
velocityposition
into
1  2 = 4r + GM + C2 2
v
r)
(21)
(21b))
the values v
r = R at touchdown.)
At
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 righthand 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) kilometersjust .))) the velocity v = mls at the instant of ignition. gives
(21b). 1.78187 10 26
(21b). 41,870
450
92
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
r2 respectto r yields)
GM
1 v2 2
=+c.
GM
r)
Now v
v2
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
j2\037M
(24))
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 onefifth of the escape velocity from the earth'ssurface, a fact that greatly facilitates the return trip (\"From .))) the Moonto the Earth\.")
\037
Remark: Equation
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)
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.
= 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)
ekt ).
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
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
to a newspaperaccount, a paratrooper
s.
100
ft.)
+00.
x(t)
wastesindrums 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
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.
= 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 upwardmotion position initial condition y (0) Yo.
(13)
16.
of its
so.)
limiting
substitute u v pig the downwardmotion velocity function given in Eq. with initial condition v(O) Vo. to obtain the Integrate the velocity function in Eq. with downwardmotion position function given in Eq. initial condition y(O)
in
a Eq.(15) nd
= ,J
(17)
(16)
(16)
Yo.)))
94
Chapter
1
FirstOrderDifferential 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
= j
00.
Vo
r(t)
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)
= lOs, and
also find
.
28.(a) Supposethat
=
Why
the limiting
is shot
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
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
(
1
In
1+
PV5 with
is
VWM
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,
Substitute
= v(dvjdy)
v
holethe = 10
= v5
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
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
failedon an
are strandedyour 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
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 Owhenits rocketsare turned off or burned out, for instancethen (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)
mo
mentum during
tlP
\037
(m tlv m tlv
+ v tlm + tlm tlv) + (tlm)(v c) + c tlm + tlm tlv. take the limit as tlt 0, so tlm 0,assuming
\037 \037
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
FirstOrderDifferential Equations)
PROBLEM
equation)
(m
 fJt) +
dv
dt)
kv
= fJe
  fJt)g.
(mo
(5))
= voMk/{J +
\037c
(1
Mk/{J)
(1 ::ok
fJt
Mk/{J),)
(6))
where Vo
= v(O) and)
M
  met)
mo mo
t.)
mo)
No Resistance)
PROBLEM2
to
obtain)
For the
= Vo
gt
+ e In
mo mo
.
fJt)
(7))
Because mo
IS)
fJtl
VI
= V(tl) = Vo
gtl
mo + eln.
ml)
(8))
PROBLEM3
Start
with
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)
em  gt  In. 2
1 2
1
mo
ml)
(10))
fJ
The V2 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 V2 at N burnout under the assumption that it is launched vertically upward from rest on the PROBLEM4
initial
an
ground.)))
=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 constantthrust 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 initialtotinal 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 firstorderdifferential 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 firstorderdifferential 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
(1))
the form)
y dx [G(x, )]
= o.)
(2))))
98
Chapter
1
FirstOrderDifferential 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 specificfirstorderdifferential 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 firstorder 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)
tox.)
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
1.x3 + 3y  xy' = 0
17.eX + ye
xeYX)y'
=0
y3
26.9X
27.
/2y4/3
 12x
I y)
3 3y + x y4
+ 3xy' = 0
/5y3/2
+ (8X 3/2yI/3
 15x/ yI/2)y' = 0
65
31
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'Yy
tan
x)))
n Chapter 1
(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 secondorder quation e takes(orcan be written in) the form)
\037)
A(x)y\"
(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\"
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)
I,then
(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 firstorder differential equation (as in Section1.6). f course,t is not unusualeithern i i O mathematicsor in the Englishlanguage more generallyfor same word to have the
In
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
springdashpot
> 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)
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 homogeneousecondorder equationgovernsthe freevibrations linear s of the mass;we will return to this problemin detail in Section 2.4.)))
102
If,
in
addition to
then
FsandFR,
the
be added to the
equation is)
\037)
c
F(t).
(6))
vibrations of This nonhomogeneous linear differential equation governs the forced the massunder the influenceof the external force F (t).)
(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
the form)
(8))
We will
\037)
(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
(10))
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)
\"
\"
.)))
Example
1)
We
x (x) = cos
and
Y2
(x) = sin x)
+y= 0
.)
Theorem 1 tellsus
that any
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
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
/)
0)
31:
X)
231:)
331:)
2.1.3. Solutions
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 secondorder for equation.)))
2.1.1.
104
3 2 1
;;..\"
(8))
1)
2 1
satisfiesthe
initial
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 afirstorderdifferential 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 secondorder 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 econdorder 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....,......''''',..\",.....\"......,.....
\"'...
\",.....
._
\"'......
......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 homogeneoussecondorder 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
(12))
to solve the)
ClYI
ClY\037
(13))
C2.)))
105)
the
functions) x YI(x)==e
and
Y2(x)==xe
==
x)
 2y' +
initial x
0,
and
then find
Solution The verificationis routine; we omit it. We imposethe given initial conditionson the
general solution) 10
8 6 4 2 0
for which)
\037
== 3,
2 4 6 8 10 2)
y'(O)== CI + C2 == 1.)
initial
2)
FIGURE
y\" initial
solutions y (x)
2.1.8.x Different =
Figure
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
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 constantvaluedfunction 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
x+l
x
xex . , x2. ,
Ix
I.)))
e2x . ,
cos x;
106
2 Chapter
I(x)= sin2x
and
are linearly dependent n any interval o metric identity sin 2x = 2 sin xcos x.
.)
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
in
Section 2.3.
show, finally,
that
and
Y2
of the homogeneousquation) e
y\"(x)
(9))
(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
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)
=. 
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.)))
107)
kg'
g'
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 econdorder equation, then the strong conversestated in part (b) of Theorem 3 holds.)
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
(]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 secondorder 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
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
G(x) = CIYl(X)
+ C2Y2(X))
of Eq. (9);then)
and)
G'(a)==
CIY\037
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.
Example4)
....
.........
.)
Y2
......
......
Therefore, Yl and
 4y =
d
O.)
(15))
But
Y3(X)
= cosh2xand Y4(X) = sinh 2xare also solutionsof Eq. (15), because) = \037(cosh2x) dx
d2
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) = e2x Of course,his is no surprise,because) t
sinh
2x== 4e2x
 4e2x)
.)
in
Remark: Because2x , e2x 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
+ C2e2x)
the form)
2.1Introduction: SecondOrder Equations 109) Linear Linear SecondOrder with Coefficients Equations Constant
As an
illustration
this
section,we discussthe
(16))
+ by' + cy == 0)
with
and
c.We first
and
(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)
== 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 ==
== o.)
will
satisfy the
differential
ar2 + br + c == o.)
(18))
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,
\037)
(19))
110
2 Chapter
Example5)
 7y' + 3y
== o.)
Solution
We
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)
3 2
1
\037
with
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
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 + C2e2xf 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.)
this
\"
'  2rly+r1y==.) 0
2
that y ==
that)
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)
Example7)
To solve the
initial
value problem)
y\"
+ 2y' + y = 0;
+ 2r + 1 = (r + 1)2= 0)
the general solution provided by Theorem6
x
= r2 =
1. ence H
X
10)
Differentiation yields)
y
,)
5)
so the
;;::....)
initial
o)
5 321 0 4
x)
2 3 4 5
x
which imply
initial
that
C} = 5 and C2
of the
FIGURE
y(x)
y\"
==
+ 2y/ + y
Cl
e + 2xe of
X
2.1.10. Solutions
==
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
Problems)
1through 16, homogeneoussecondorderlina 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
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,
5)
(0) == 1)))
112
2 Chapter
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, = x3; 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
\" \"
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
(b) Solvethis
mula)
firstorder
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 ,
where
(c) Why does Abel's formula W (Yl , Y2) is either zero everyor nonzero everywhere (as stated in Theorem 3)?)
Determine whether
the
independent 2 cos x 2
Apply Theorems5 and 6 to find generalsolutions of the differential equations given in Problems 33 through 42. Primes
denotederivatives
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.)
37. 2y\"
39.4y\" 41.6y\"
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
Each of Problems 43 through 48 gives a general solution equation y (x) of a homogeneous secondorder'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/
x are two
(0). 3
= Cl + C2elOx 44. y(x) = cle + C2elOx = cle + C2xelOx = cle + C2elOOx 47. y(x) = CI + C2X xh = eX (c1e + C2exh))
lOx lOx lOx
30.
).
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 thirdquadrant point of intersection of the solution curves shown in Fig. 2.1.7.)))
ax
y//
+ bxy' + cy = 0
(22)
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 constantcoefficientlinear equation d2 a y2 + (b a) dy + cy 0 (23) dv dv
(for x
general
Problems
52
56.)
with
independent
variable
the
characteristicequation
of the form)
show that our discussionin Section of secondorderinear equations 2.1 l differential equalinear natural way to the general nthorder Po(x)y(n)+ Pl(x)y(nl)
+...Pnl(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) i0 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)
(2))
yen)
+... +
(3))
Just
tion
as in the secondorder a homogeneousnthorder 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 verificationas of Theorem 1 of Section essentially samea
THEOREM 1 of for Principle SuperpositionHomogeneous Equations
n
..., 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)
Example
1)
It
is easyto verify
YI
that
114
2 Chapter
2Y3(X)
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)
y(x) = cle
3x
.)
saw in Section that a particular solution of a secondorder differential 2.1 linear equation is determined by two initial conditions.Similarly, a particular solution of an nthorder 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
...,
,)
Pn, and
... 
+ PI (x)y(nl)
+ +...
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)
...
,)
y(nI)(a)= bn 
1 .)
10
Equation (2) and the conditions in (5) constitute an nthorder 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)
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
Remark:Because general solutioninvolvesthe three arbitrary constants its CI, C2, and C3, the thirdorder 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 solutionsfor of which, two of thesethree initial valuesare zero. each .)
Alternatively,
5 4 3 2
1
\037
3)
2)
/)
y'(O)
=3
1)
/)
y\"(0)=3
1
2 3
4
\037
0)
1) 2)
o)
1)
'\"
y'(O)
= 3)
5 1)
2
x)
5)
3 1) 2
y\"(O)
= 3)
1
x)
4)
11)
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).)
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))
that
conditions)
yea) = y'(a)

=... =
and
Example2)
..
It
is easy to verify
that)
YI
(x) = x 2
x 2y\"
Y2(X)
=x
3)
 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
yen)
== 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)
and 11 12 impliesthat
(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
...
...
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,)
...
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, ...
Example3)
= sin 2x,
and
f3 (x) = eX
.)
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 homogeneousthorderlinear 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(nI) finI)
We write
f\037nl))
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 (17781853
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(nI)(x)
I
+...+)
+...+
cnfn(x)= 0,
cnf\037(x)
= 0,)
(9))
f\037nl)
(x) =
0,)))
118
2 Chapter
I.
(11, ...
...,
... .
\037.
I.
11, ..
Example4)
.....
Y2(X)
cos2x, and
Y3(X))
sin 2x (of)
e)3x
cos2x)
sin 2x)
3e3x 3x ge
2sin2x) 2 cos2x)
4cos2x 4sin
2x) 2x)
2x)
=e)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
.)
Example5)
Showfirst
that
Yl(X)=X, Y2(x)=xlnx,
of the thirdorder equation
x 3 y(3)
and
Y3(X)=X
2)
x 2y\"
+ 2xy'
 2y =
0)
are linearly independent on the openinterval x of Eq. (10)that satisfies the initial conditions)
> 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.
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)
+
y(I) = Cl
Cl
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
...
conditions of the form in (5). Theorem 3 provides the of W in the case linearly independent solutions.) of necessary nonvanishing
that
initial
Supposethat
equation)
YI, Y2,
...,
Yn
y(n)
on an openinterval
..
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.)
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
(a) + (a) +)
C2Y2 (a)
c2y\037(a)
+ ... + + +)
...
CnYn(a)
cny\037(a)
= 0, = 0,)
(12))
clyi
in
cny\037nl)
(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
(13))
trivial initial
in
= y(nl)(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 thorder 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)
+ Pnl(X)Y'+ Pn(x)y = 0) + Pl(X)y(nl) (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))
+...
..., ...
is a linear combination)
\037)
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 saysthatgivenbo, b I , and b2therexists 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
\"
\"\"
.
that)
y(x) == Cle
to find
+ 3CI
CI the (See application for this
C2
== b o, == b I , == b 2
.)
section.))
.)
equation)
Nonhomogeneous Equations
We now
\037)
(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
linearity
of differ
entiation))
Yc
(n)
/ (nI) + PlYc + . + PnIYc+ PnYc ... == [(y(n) + PI yenI) + + PnIy/ + Pn y] / (n) (nI) [(yp + PIYp + . + PnIYp + PnYp == I(x) I(x)== o.)
..
..
Thus Yc == Y
\037)
Yp
and
\037)
it
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
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
+ 4y == 12x,)
(17))
and
that
y(O) == Cl
== 5,
.)
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)
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)= x2, hex) = x2 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 thirdorder 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
= e , Y2 = e
X
, Y3 =
e)))
30.Verify
that YI
= x and
x 2y\"
Y2
= x 2 are linearly
2y
independent equation)
 2xy' +
= 0,)
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 nthorder linear differential
(a) Given
y
\",
the equation)
,)
+ py + qy = 0
y\"(I)
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
has a solution
satisfying
y(O) =
if and only if
1,
y'(O)= 0,
C = 5.
homogeneous linear
an nthorder
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=
iI.))
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
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)
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)
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 lnI is nonzero if
r nI
2)
r nI
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)
independent
on the whole
real line.)
124
35.According
W (YI
2 Chapter
to Problem 32 of Section the Wronskian , Y2) of two solutions of the secondorder equation
2.1,
is given
by
W(x)
for
= K exp
someconstant K. It can be shown that the Wronskian of n solutions YI, Y2, , Yn of the nthorder 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).
y(n)
+ PI (X)y(nI)
+...PnI +
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 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)
36.
and then show that W/ PI W. Integration now gives Abel's formula. of the homogeneous Supposethat one solution YI secondorder 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,)
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
the
(18)
(18),
(18).
44. First
YI v\"
(2y\037
(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.)
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 nthorder linear equation if it has constant coefficients. The)))
in
the form)
(1))
any(n)
an
f=. O.)
_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 secondorder 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 + anlrn1e +
...
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
(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 nthdegree polynomialsuchs 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 seconddegree 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
126
2 Chapter
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
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)
..., r
rnx)
(4))
Example
1)
y(O) = 7,
+ 3y\"
 10y' 0; =
y/
(0) = 0,
y\"
(0) = 70.)
+ 3r2
 lOr =
O.)
solve by factoring:)
and r
0,r = 5,
= Cl + C2e5x + C3e2x
.)
initial
y(O) = Cl +
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
.)
.)))
1.
1,
dx)
(5)
on operates the ntimes differentiablefunction y(x) to producethe linear combination) + Ly = anyCn) + an_1yCnl) ... + 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),)
+... +
(6))
useful to think of the righthand sidein Eq. (6)as a (formal) nthdegree polynomial in the \"variable\" D;it is a polynomialdifferential operator. A firstdegree 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.)
any
a)y)
in
(7))
(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  expandsby collecting coefficientsin 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
R Repeated ealRoots
Let us now considerthe possibility the characteristic equation) that
anr
n
1
..
(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.
 
rl)k == o.)
as) rl)k,)
in
(8))
 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
(7).
1.
Ly == (D
 ro)[(D
rl)k y ] == O.)
(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))
+ u(rle
rIX
 rl(ue
rlX)
rIX
r1x ) == (Du)e
(12)
(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
l
,)
y(x) == uer1x
== (CI
...
+ CkxkI)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.)))
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
.)
(14))
We may observethat, accordingto Problem 29 of Section2.2,he k funct rx involved in tions erx , xerx , x2erx , , and xk1e (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)
doubleroot r
==
x x + C2XeO. + c3x2eO. == Cl + C2X + C3x2 to the solution, while the doubleroot r == 1contributes C4ex/3 C5xex/3. encea H +
contributes
== 0
CleO.
3
.)
.)
H.
t4
et
If we substitute so on, we get)
==
L00
tn n =0
n!
== 1
t2
t == i () in this
i 2 ==
i 1, i,i 4
3 ==
==
1,and
ei(J =
n=O
>;
(io)n
n!)
== 1
==
02
i ()5 5!)
+0 _ 4!
+i
0 3 05 0_ +
5!)))
130
2 Chapter
Because two real seriesn the last line are the Taylor seriesor cos and sin (), the i f () respectively,this impliesthat)
>)
(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)
+ i sin y).)
(16))
Thus it appearsthat complexroots of the characteristic equation will lead to s function complexvaluedolutions of the differential equation. A complexvalued F of the real variable x associateswith eachreal number x (in its domain of definition) the
complexnumber)
(17))
The realvalued functions and g are calledthe realand imaginary parts, respectively, of F.If they are differentiable, we define the derivative F'of F by)
(18))
Thus we simply differentiate the real and imaginary parts of F separately. We say that the complexvaluedfunction F(x)satisfies the homogeneous linear differential equation L[F(x)]== 0 provided that its real and imaginary parts in == + (17)separately satisfy this equationso L[F(x)]== L[f(x)] iL[g(x)] O. The particular complexvalued functions of interest here are of the form F(x) == erx , where r == a We note from Euler's formula that)
::i: bi.
+ i sin bx))
(19a))
i sin bx).)
19b
))
(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
.)))
== Cle(a+bi)x
== Cleax(cosbxi
ax
 C2)e
ax
sinbx,)
constants CI and C2 can be complex. instance, the choice For arbitrary CI == C2 == 4 gives the realvalued solution YI (x) == eax cos while the choice bx, CI == i , C2 == 4i gives the independent realvaluedsolution 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)
(21))
Example3)
+ 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
.)
Example4)
 4y' + 5
==
== 0)
== ::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
so the
initial
.)
(22)) upon
writing)
+ iy == re
ie)
(x, y))
of the complexnumber
z=r
x)
( +i
\037
= r(cose i sine) = re +
i6)
x + iy.)
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/ 2 ::f::,Jre
,)
(23))
where ,Jr denotes(as usual for a positive real number) the positive square root of the modulus of z.)
Example5)
+ 4y == O.)
4
== (r2)2
 (2i)2
iJf
== (r2
,J .
::
)
i sin
i
== ei3Jr / 2, we find
that)
= 1+
i)
,J
2i
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,
+ A 2x +
...
Lx e
p=o)
kl
...
+ Bkxkl)eCabi)x
(24)
P ax
bx (Ci cos
+ di sin bx).
It can be shown
that
the 2k
functions)
,)
Example6)
Solution
O.)
has as its roots the conjugate pair gives the general solution)
3::f::2i of multiplicity
k ==
2. HenceEq. (24)
.)
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)
co'''
C\"'
C\"'\"
..
C\"'
\037.
_N..
..
,_u.'u
+ y'
 10y
== 0)
is the cubicequation)
r3
+r10==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)
y(x) == Cle
2x
134
2 Chapter
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.
is)
3x + C2X + C3x2 + C4x3 + cse + C6esx + C7xesx + e2x (cscos3x + Cg sin 3x)+ xe2x (CIOcos3x + CII sin 3x).
.)
IfI1Problems
Find the
u)
lems
 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)
18,
_
==
12, y\"(O)
that y(4)(x)
==
==
value problem)
5y\"
+ 100y' 500y == 0;
y' (0)
==
y(O)
given that differential In
YI
0,
==
==
250)
8y(3) 3y(3)
8y\" y\"
16y\"
==
(x)
solution
of the
3y\"
==
9y(3)
(3)
12y\" lly\"
==
==
6y(4)
==
equation.)
==
+ 2y' + y
0 (Suggestion: Expand
in
through 42, find a linear homogeneous equation with the given generalsolution. constantcoefficient 2 2x y(x) == (A + Bx + Cx == Ae 2x + B cos 2x + C sin 2x 40. y(x)
Problems39
39.
)e
26.
Problems through 21
41.y(x)
42. y(x)
== ==
A cos2x
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
  
quadratic
formula
solve the
following
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)
y(3)
y;
y(O) ==
1,
y' (0) ==
O.)
3x
2y\"
8y
==
==
2x
3
y
==
5y(3)
lly\"
25y\"
==
cos2x
x)
given initial
conditions
on the
==
==
sin
Ae x
x ax + Be + Ce/3
,)))
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 secondorder Euler equation ax2y\" + bxy' + cy = 0 to a constantcoefficient homogeneous linear
substitution equation. Show similarly that this same transforms the thirdorder Euler equation)
50.Thedifferential
equation)
y\"
+ (sgnx)y= 0)
coefficient function)
(25))
'
has the
discontinuous
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
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
I I
Equilibrium position)
FIGURE2.4.1. massA
springdashpot 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
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 dashpota 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))
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
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)
my\"
+ ey' + ky
== F(t))
(5))
gravity).)))
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)
( )
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(Icos()).)
is a constant C therefore gives)
2
mL2
2
We
( ) )( ) (
d()
dt)
+mgL(Icose)==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).
()\"
(7))))
138
2 Chapter
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))
2 + wax == 0
(8'))
(10))
a so that)
this
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.
(11),
negative. Thus)
A)
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),)
B.
==
. .
.)
With
(12))
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)
2n T= (Va)
1
(Va
(13))
2n)
(14))
t) I I I I I I I I I I
II I
C
1I 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
50x= 0;that
is,)
\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
with
x(t)
Hence amplitude of motion its
 .
1
2)
SIn
10
t.
is)
x(t) =
0 (0cos  0 2
2 lOt
1
sin lOt
0cos(10t 2 1
ex),)
cosex=
0
1
>0
and)
sin a ==
o)
1
< o.
angle)
= 2:rr + tan
( 2/0)
is)
\037
5.8195,)
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.)
T)
I)
0.5)
c)
.5)
0.5) 1)
x(t) = C cos(wot a) in
C
\037 \037
Example
1,
\037
FreeDamped Motion
With
takesthe form
damping
but
mx\"
,)
(15))
> o.
1/2)
(16))
(17))
that
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
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)
o)
t)
(19))
FIGURE2.4.7. verdamped O
motion:
rl
< 0 and
x(t)
r2
are graphed
position
Xo
with the
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 wouldbe 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
(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 +
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:
+ B sin wIt),)
2
(21))
x(t)
Solution curves are graphed with the same initial position Xo and different initial velocities.)
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) = Cet
COswlt+
\037 \037
sin
Wit)
,)
so)
\037)
a))
(23))
where)
\037)
C=
A2
=A + B2,) cosa C)
,
and)
. Slna =
.
B
C)
a)
,WI
\"
I)
/',.I)
X X
t)
.......( = +Cept
= Cept 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 = Cet 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 Cet its timevarying 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.
exponentially damps the oscillations,in accord with the timevarying amplitude. 2. It slowsthe motion; that is, the dashpot decreases frequency of the motion.))) the
1.It
lagas
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) == et (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.
(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
It follows that)
x(O)== A
whence we find
the body is
== 1
and
that A ==
1 and
x(t) = et
its Hence timevarying
(cas,J99
 \037sin,J99 . t)
is)
amplitude of motion
CIet
==)
.)))
144
2 Chapter
therefore write)
x(t) =
4  vTIS 115
sin
J99t
and
sinal =
4 vTIS<
O.)
+ al = 2][ tanl
4/vTIS = 2][ tanl 4 IN> (v99)) 5.9009, 99/ 115 (\037 vTIS)
\037
is)
81
= al
WI)
1
in
t
S \0370.5931
\037
< 0.5820 8
Example1).With the
timevarying
x(t)
\037
amof the
(25))
and its graph is the dampedexponential that is shown with the undamped oscillations Example1).) of
x)
\\
.......)
t x(t) = C]ecos(w]t a]) of Example 2 (damped the oscillations), position function x(t)= C cos(wot a) of
position function
x(t) = ::l:C1e.)))
the
curves envelope
 aI)
==
0,
wIt
that
 al
,
3n
2)
.
+
..
,)
is, when)
t == 8 1
We
seesimilarly
t == 80
that
when)
,  , , ,  ,  , , ,
 3n
2WI)
81
81
81
2WI)
2w})
3n
2WI)
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)
(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) = Clecos( wit
a\037)
FIGURE2.4.11. Graphs
illustrating the additional
on the interval
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 4kg 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
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
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 andwhenlocatedat the respective istances I and R 2 from the center R d of the earthhave periodsPI and P2. Show that)
PI P2
A 11.
RIJL;
R2
slugs in ftIbs (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;.)
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.)
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),
x(t)
tial equation
=x
Xo in this
differential
equation.))
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
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 massspringdashpot 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
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 12lbweight (mass m = 0.375 slugs
Xo Vo Xo Vo Xo Vo
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.
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, timevarying 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
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,
of the
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 timevarying amplitude be 1% of its initial value?)
20 10
\037
\\
\"
Problems 24 through 34deal with a massspringdashpot 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)
problem.
24. (Critically
10
t)
damped) Show
in this
casethat)
.)
15)
20) function
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)
only if
Vo
and
Vo
casethat)
(vo
r21
x(t)
= e
al).
where rl , r2
r2xO)erjl
wB and y
(rl
O.)))
148
2 Chapter
Xo
3.125 slugs
in
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 poundseconds foot) and spring constant per pounds per foot).)
(in
x(t)= e pI xOcosw)t+
Vo
+ pXo.
WI
smw)t.) )
that)
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
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.
(26))
initial
condiofan ideal
32.(Underdamped)
of)
localmaxima
X(t)
p = Cet 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 massspringdashpot 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 102n.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 et 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)
and Figs.2.4.7
the graphs
of Xl (t)
and
X2
(t)
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.)
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 nthorder linear equation with constant coefficientshas the form)
any(n)
\037)
(1))))
= Yc +
Yp)
(2))
mogeneous equation)
an y (n)
(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 thenby equating coefficientsof like powersof x on the two sidesof the resulting equationattempt to determine the coefficients Ao, AI, ..., Am so that y p will, indeed,be a particular solution of Eq. (1). Similarly, supposethat)
that
.)
Example
1)
+ B.)
Then
y\037
=A
and
y\037
= 0,so yp
will
2)))
150
2 Chapter
yp(x) ==
\"'''',h _,'_'\"
\037x
/6
.)
.)
..,.
,..\"....\"
.....,.,......\037.
....
...
...\"
,..
\037
...
\"
Example2)
Solution
4
\",..
\037'.....'nn'.n.vu.....\"\"..
3x == 2e
.)
yp(x) == Ae
Then
y\037
3x
.)
== 9Ae 3x ,
 4(Ae 2
y
3x == )
3x 2e
;)
that
is, 5A
==
2,so that
A ==
\037.
\037
e3x .
.)
Example3)
Solution
+ y'
== 2 cos x.)
signalsthat
the presencef y' on the lefthand side o we probably needa term involving sin x as well. Sowe try)
yp(x) == Acosx+
y\037(x)
y\037
==
(x) ==
cos x
B 
sin x.)
Then substitution 3(
that
5A + B == 2,
with
A
 5B
== 0)
13
13
.)
that
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)
4
2x == 2e
.. .......
.... 0....... .)))
(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
(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
2. An
1.A polynomial in x;
exponential function coskx or sin kx
.)
3.
Any
erx ;
(4))
such functionforexample,
RULE
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
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)
+ 4y == 3x3
.)
(5))
+ 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
+ 4C == 0,)
4B == 0 , 2B + D == 0)
and
with
solution A ==
\037,
Eq. (5)is)
B == 0,C ==
\037,
== 3 3 yp (X ) 4x
9
gX.)
.)
Example6)
Solvethe
initial
value problem)
y\"
 3y' + 2y
1,
==
x 3e 10cos3x;
y(O) ==
(6))
 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)
==
y\037
Ae  9B cos3x  9C 
sin 3x.)
y;
We
3y\037
equate the coefficientsof the terms involving those involving sin 3x.The result is the system)
7B  9C 9B  7C
with
6A == == ==
10,
0)
9.
which, however,does not have the requiredinitial values in (6). To satisfy thoseinitial conditions, e begin with the generalolution) w s
2x}4
9.
i\037
 e  ij sin3x+
x
cos3x.)
(0) == c}+ C2 + 4 + ?3 == 1,
== 2)
i\037
.)
.....
Example7)
.)
(7))))
154
2 Chapter
0,r
==
3i,
and r ==
3i.So
Yc(x) == Cl
the
Yp(x) == A
.)
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
.)
differential equation)
.)
(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
== 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)
rx
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))))
= II(x) + Iz(x),)
Yz
(12))
(x) of the
two
equa(13))
= II(x)
and
Ly
= Iz(x),)
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
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
If the function
I(x)is of either
form
in
solution)
rX
+... + ...
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
in
8 Examples through
10.)))
156
I{x}
,)
YP)
Pm
+ 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
+...
...
+... +...
of undetermined coefficients.)
Nn
\"\037..\037.,,\037.,,,..,...'...'\037nN''''.''\"''A'....N..,._..,_.
\037
,\037
,...
....
Example8)
+ y\"
x == 3e
+ 4x2.)
== r2 == 0 and r3 ==
(16))
s 1,o
the
+ C2X + C3ex
2
.)
As a first
sum)
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
6C+
has the solution A == solution is)
\037,
3,
2B + 6C== 0,
== 12D 4)
24D == 0,)
B == 4, C ==
Yp(x) ==
\037,
and
\037eX
+ 4x2
\037x3
+ 1x4
.)
.)))
Example9)
.)
equation)
Example10)
3i,
sum)
be multiplied by x , and the secondpartcorresponding sin 3xmust e b to x we multiplied by x. Hence would take) yp(x) =
(Ax
3
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))
expreceding
independent
2 sec x,
x,
158
(that
+ Pn_l(X)y(nl)
+... +
(18))
== ClYl
+ czyz + ... +
CnYn)
(19))
+ Pn_l(X)y(nl)
+... +
(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)
+ . + 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 secondorder onhomogen n neousequation)
\037)
L[y] == y\"
complementary function)
(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
y\037
and
y\037.
yp
== (UlYl 1
.)
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))))
, == , UIYI
, + U2Y2')
(26))
Yp
== (UIYI \"
.)
(27))
But both
YI
and
Y2
satisfy the
homogeneousquation) e
y\"
+ Py' + Qy == 0)
\"
==
P '  Q
Yi
I
Y\037
Yi)
(28))
for i ==
that)
(u'lY\037
u;Y\037)
U2Y\037)
Q.
(u IYI
+ U2Y2).)
In view of Eqs.(24)and
Yp
== (uIYI
\"
 Q
+u2
\"
YP')
hence)
L[YP ] ==uIYI
The requirement that
Y2.)
(29))
Yp
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
(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))
160
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 )
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,)
We
,.
==
 SInx
tan
x ==
U2
, == COS tan x
. x == SIn x
x  cosx)
sin 2
.)
== cosx
 secx,
Hence take) we
Ul
f (eosx
tanxl)
and) U2
f sinxdx =  eosx.
U2 )Y2 tan
+ (x)Isec(x x+ (sinx
In
(x)
Yp(x) ==
.)))
Problems
given equation.
1through
x.
(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)
==
10.
9.
+ 16y == e3x
 6 + +  4y y/ y/ y
==
==
2 sin 3x
2
==
sin
sinh
Use trigonometric
equations
in
identities
generalsolutions of the
==
17

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
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
+ C2X1 is a
second
==
3x sin sin
y(4) y(4)
5y\" 9y\"
2y\"
==
==
sin
+ xy/
==
72x5
.)
==
2x
2x
+y
==
x 2 cos
X)
Solvethe
initial
==
value problems in
== == ==
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 secondorder 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
(0) ==
y/
(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)
==
+ (cosx i sinx)3
63.Carry
l)y\"
Yc
==
CIX
derive the
derive the
==
\037
identities
3 cos x sin 3 x
cosx i cos3x, +
sin
out the solution processindicated in the text to variation of parameters formula in (33) from and
==
\037
i
(x)
==
==
sin 3x.)
2 sin
formula in
(33)to
find
_
I I
162
+ ex'+ kx = F(t))
(1))
motion of a mass m that is attached to a spring governs the onedimensional constant k) and a dashpot (with constant c) and is also acted on by an external force F(t).Machineswith rotating components commonly involve massspring systems(or their equivalents) in which the external force is simpleharmonic:)
that (with
Equilibrium position
(2))
l)
FIGURE2.6.1. cartThe
withflywheel
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
wt)
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;)
.)
+ 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 frontloading 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).)
mx\"
whosecomplementary function
\037)
(4))
== Fo cos wt,)
Fo
mw 2) Fo/m
Fo/m
2 w o)
w2
'
(5))
and
thus)
(6))
+ xp is given
o)
by)
(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(wota)+w2 coswt, 2 w
o)
x'(0).
(8))
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 ==
\037
...
\037
x\"
Find
O.)
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.
 5 cos
5t,)
x'(t)
==
5t.)))
164
15 10
5)
2 Chapter
Period = 2n
II I I I
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
If we imposethe
that)
initial
= 5 cos3t
 cos
5
5t
in
Example 1.)
m(w6
Fo
(
2)
and
C2 ==
0,
x(t) ==
A == \037(wo
Fo 2 m(w o
w 2)
(coswt
+ w)t and B ==
x(t) == m
\037(wo
(9))
with
2Fo
2 (wo
w2
+ w)t.
Iwo
(10))
))
Supposenow that
+ w)t
\037
Wo,
so that
Wo
is a rapidly
+ w),)
 is
wi.
x(t) == A(t)
but with
sin (wo
\037
+ w)t,)
(t) == m
2Fo
2 (wo
1 sIn 2 (wo w2
))
w) t
Example2)
1.5 1.0 0.5 I
\037
With m ==
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(wow) == 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) 254soundwithahears a beatan of the combined of) 0.0 frequency
0.0
.)
0.5
\\
\\
\\
\\
\\
\\)
I)
\\../
\\.
t)
(wo
 w)/2
258
 254
2)))
2n)
== 2 (Hz).
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)
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))
1.0) 0.5)
\037
0.0)
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
.......)
(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
10rad/s; that
little
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, thenjust as in our o imately equal will I simpleexampleof a masson a springresonance occur. ndeed,the resulting)))
166
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
= 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
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
disk.
Because righthand 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
167)
divide eachterm by
\037mx'
to obtain
x + 3m
\"
x
2k
==
o.
natural frequency of horizontal back andforth 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=)
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);)
+ kx ==
in
ky)
(16))
== 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
.)
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
a))
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)
the
phaseangle a lies in
a ==
B
A
== k
cw
< a < n,
(22))))
169)
1
k
ot=)
1T
cw
1
mw 2
if k if k
> mw 2 < mw
2)
,)
+ tan kmw) 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 valueand practical resonance ccursat 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
\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 massandwith 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,
motion
x(t) =
Xtr(t)
value
x\"
(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
we substitute
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
point we
C1 == 1 and C2 ==
imposethe
x(O) (0) 3.Therefore, the transient motion and the steady periodic
conditions
==
6,x'
Xtr(t) ==
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)
x\"
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)
10,
initial
value problem in
(24)
20.)))
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)
41  (676 + (
occurswhen)
(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 massandspring'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).
14,
2.6.2)
Each of Problems 15through 18 gives the parameters for a + system with equation mx\" +cx' forcedmassspringdashpot 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)
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
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
conditions.
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
g.)))
172
2 Chapter
26.
w,
exceptwith
mx \"
 a)
expectthe
of C a).
and
termsderive the
periodicsolution)
xsp(t) =
where
J(k 
+ F.2 0 a 2 + m( )2 (ew)2)cos(wt
\302\2432
 
fJ),
a is defined
in
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,
(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
ex'
J(k
m(
ecr
Fo
2
)2 + (ew)2)
vi 4km
ecr/\037' where
= Wm =
J m  \037 < 2m
\037
Wo
= fk.
V
;;)
FIGURE
A 23. building
2.6.11. massspringThe
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 cartwithflywheel 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\"
(with
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)
2.25 =
occursat
wm
the frequency
given
by)
secondfloor?
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 \"
Thus the resonance frequency in this caseis larger (in contrast with the result of Problem 27) than the natural frevlk/m (Suggestion: Maximize the square quency Wo
of C
.))))
the function
nance frequency
thesegraphically basedconclusions by analyzing C(w ). In particular, find the practical resoand the corresponding amplitude.)
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:
00
20
40
60
80
100
Velocity (mi/h))
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)
w resistor ith
inductorwith
R)
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
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)
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 secondorder 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
TheMechanicalElectrical Analogy
It
\037)
is striking
that
+ ex'+ kx = F(t))
(5))
of a massspringdashpotystem with external force F(t).The table in Fig.2.7.3 s detailsthis important mechanicalelectrical 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 analogeomputerselectrical 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)
LI\"+ RI'+
As
I= wEocoswt.
1
C)
(6))
c periodicurrentIsp ; thus)
in a massspringdashpot 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)
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)
< ) a <
z=
R2
(WL
W\037
(ohms),)
(10))
Eo
Z)
cos(wt a)
(11))
Z')
Eo
(12))
176
2 Chapter
(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
Eo
Z
coswt+
FIGURE2.7.4.Reactance and
delay angle.)
=
Therefore,)
Eo
Z)
(cos8 sin 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).)
When
we want to
find
given the
initial t == 0 in
values
Eq. (2)
(16))
Example
1)
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.)))
values of R, L, C,and
w == 377 in
that the
59.58 (ohms))
110 (volts)
==
1.846 amperes(A).
With
the
0.575 ==
377)
is)
w)
0.0015 s,
== 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
these initial
their solution is Cl ==
The observation that after onefifth 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
Solution
have E(t) =
(0)
0.1)
(A/s),
Example1:)
We
= 1(0) Cl + C2= 0, = 1'(0) 44cl 456c2= 1100) = for Cl = C2 2.670. Therefore,) 44t I (t) = (2.670)(e e456t).) 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 loudestand simultaneously tune out stations at other frequencieswe therefore want to choose to maximize C broadcasting
10.)))
glanceno
wL
that
 wC
1
==
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
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 firstorderequation) Problems
6 deal with
seriescircuit
the inductor
LI'+ RI = E(t).)
3. Supposethat the battery in Problem 2 is replacedwith an alternatingcurrent 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 steadystate current Isp in Write the solution in the form Isp(t)
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 firstorder 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
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 10F; sin E(t) = 110 601(tV 4 16.R = 25 Q, L = 0.2 C = 5 X 10F; H,
E(t) = 120cos377t
V)
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
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). steadystatecurrent? 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) = 100eV; 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
= tan](wRC).)
In Problems
with input
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 LCcircuitthat is, an RLCcircuit with R = Owith 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
= 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 solutionitapproaches 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
\037
.)
+ 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
such as)
+ p(x)y' + q(x)y == 0;
(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)
+ 3y == 0;
(3))
2
1)
1
2 3
4)
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 sinnJ3
\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
+ 4y == 0;
y
(4))
2
1)
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
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 Aiscalledan 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\"
characteristic values).
+ Ay = 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 eigenfunctionindeed, 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 Apositive, ero, nd negative. z a = 0,then the equation is simply y\" = 0 and its general solution is) y(x) = Ax
B.)))
a
\"
2 0 ay==,)
y(x) == cle
ax
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,)
+ eax )j2and
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,)
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,)
..
,)
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
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 halfperiod.
.)))
184
2 Chapter
< 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)
(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)
+ AY == 0;
(11))
Solution
Virtually
the
sameargument as that usedin Example3 showsthat the only possible > 0 (a > 0) to be specific.Then the
+a
== 0,)
with
general solution)
0 provided that
aL ==
y)
2')
n
3n 9n2
2') ...
n=2)
n=I)
that
is, if)
n A==4L2'
4L2' ...,
2
multiple
of
4L2
are (11)
Thus the
given
1)
by)
nth
FIGURE2.8.5. The
eigenfunctions
Yn
 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 halfperiod.
4L2)
and
Yn
(x) == sin
 l)n
2L)
(12))
== A Y 1 (x,A)
+ BY2 (x,A).
We write
which)
Yi
(x,A) because YI
and
Y2 will
YI(X) ==
ax == sinxJI.)
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
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
(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 xaxis. 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 xaxis, achpoint moves in a e circle centeredat that point's equilibrium position on the xaxis. 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 ydirection is) w
\037x
We plan to
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)
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 ydirection. ecause == p \037x, substitution of B m this and (15)in F == ma yields
y)
\037x,)
\037x)
\037x
\037
pw
y.
We now
string:)
\037x
/
/)
+ pw 2y
== 0
.)
(16))
x + L\\x
x)
If we write)
pw A==T)
(17))
string
and
are fixed, we
finally
get the
(7))
eigenvalue problem)
+ Ay
==
0;
y(O) == 0, y(L) == 0)
that in
(7)are)
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
rium
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
increase
further,
the
string
pops
%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 phenomenonthehapeof 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 yaxisdirected 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 fourthorderdifferential
equation)
Positive
yvalues)
Ely(4) == F(x),)
where)
(19))
. . .
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
Ely(4) == W)
where E,
(20))
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
Eq. (20)is a fourthorderdifferential equation, its solution involves successive simpleintegrations. Ely(3) == wx
+C
;)
EIy Ely'==
\"
==
2 wX
. + CIX + C2,)
+ C2X + C3 ; + C3x + C4
\037wx3
\037CIx2
Ii
<,:
;.;.;:)
x=L
or hinged)
Simply
supported
11) where C
x=L
Ely ==
wx4
2\037
\037CIx3
\037C2x2
,)
x=o t)
Built
in)
a beam.)
t)
where A,
(21))
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 cantilevera 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
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
189)
EndpointCondition)
Freeend)
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
(22))
FIGURE
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
= 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
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
Ely(x) =
Now y(O) = 0 impliesthat
wx4
2\037
/2 wLx3
0= wL42\037
l\037
wL4+CL.)
It follows that
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)
. radIus: a =
Its
1.  In. 2
)(
length)
\037
is)
dyn 38484.6 . cm
==
so)
/=
Therefore Eq. (25)yields)
\037
1 c n(1.27)42.04m4. 4)
\037
Ymax
,)
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 onesixteenth as muchonlyabout 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.
y)
= y(x))
p) x)
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)
< <
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
+ AY == 0;
y(O) == y(L) == 0)
that
(7))
that
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
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)
and
I = JT [<0.5in.)(2.54 ) 4
In.)
r
\037
\037
e 2.04m4.
the critical force for this rod
Uponsubstituting
IS)
that
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
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
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
\037
 I)n/2 = 
anx}f, = z.
+ 'Ay = 0;
y/(O)
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.)
has no
ically
z)
11.
12.
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
7. Considerhe t
y\"
+ 'Ay = 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)
= 1.
16.a) Supposethat (
x
= L.Show
that its
= 0 and
13.
nx.
y(x)
(x
 2Lx + L
3
2X
).
y\"
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
onefifth that
=
4
=
3
(25).
curve
IS)
by)
w
24\302\2431
(x  4Lx + 6L
4 3 at
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
that its maximum deflection occurswhere x of the maximum deand is about flection that would occur if the beam were simply supported at eachend.)))
Power Series
Methods)
(:)\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)
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
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
 a)n + ..
..
(1)
If a
\037)
= 0,this is a powerseriesn i
Lcnx =
n
n=O)
00
Co
(2))
194)))
We will
xa.
00
n=O
Lcnx
existsfor all x in
lim
N\037oo
L
n=O)
cnx
(3))
I. In this
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
. 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)
Ix Lx
=
n=O)
00
= 1 + x + x2 + x3 + ... ;
(11))
and)
ex
 
..
(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) (xa)n==f(a)+f'(a)(xa)+ (xa)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
==
(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,)
. . .
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
(16))))
addition and the series (16)is the result of formal multiplicationmultiplying 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
(x
=x
...
\037
\037
\037
4 = x x3 + 16 5 6 120)
x ...
)
3
=
for all x.
1 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
x +
1
x3
3
 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
15
x7 15 + 7 17x
315)
FIGURE
3.1.1. Obtaining
the
for series
tan
by division
of series.)
f(x) =
Lcnx =
n
n=O)
00
Co
(19))
of the function
and)
00
\037)
+...
(20))
at
eachpoint of I.)
For example, ifferentiationof the geometric series) d
( 11))))
(1
1
X)
00
The
that it
will satisfy a given differential equation dependsalso on Theorem This theoremalso 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
series = y
L cnxn
2.
so
prooftells
THEOREM2
If)
IdentityPrinciple
L
n=O
00
anx
L
n=O)
00
bnx
x in someopeninterval
an
I,then
an
Theorem 2 that
.....
In particular, if
.....
someopen interval,
follows from
Example
1)
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
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=O)
00
cnx
= 0;
that is,)
L[(n+ I)C +l + 2c ]x = O.
n n n
n=O)))
00
200
some interval,
then it
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 firstorderdifferential 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
it
c= (I)
n
2nco
, n.)
23co 3
23co
n
3!.)
'
> 1.)
00
n ( 2X )
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
.)
ShiftofIndexof Summation)
In the solution of Example1 we wrote)
L
n= I
00
ncnx
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
point by
1,from n =
1 to
= 0,thereby
series
(24))))
.)
+2
\037
00 \037 n=3
\037anx
nl = \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 validityin 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
. 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.
. 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
Example2)
Solution
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
first
n == 0 with
L
n=O
00
ncnx
 3 L(n + I)C
00
n=O n
n +lX
+2
Lcnx
n=O)
00
==
0;
that is,)
== O.
 3(n + I)C +l + 2c
n
== 0,)
relation) recurrence
C n+l
 n+21) +
3(n
n)
for
> O.)
and
find that)
We
with n ==
0,n
==
a 1, nd n == 2,in turn,
')
co, 3)
C2 ==
3 Cl 3 .2
==
3 co 32
and
C3 ==
pattern
evident; it is not
to show by
C == n
n+l
3n
Co)
if
> 1.)
OOn x. L 3+ 1n
n
n=O)
(26))
n)
hm
n+oo
3n
00)
Cn+l)))
+ 3 == 3. n + 2)
3
1
3x)
  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
.)
==yxl.)
substitutions
y
==
n
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
righthand 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
lefthand side contains neither a constant term nor a term containing n power, the identity principlenow yieldsCo == 1,C1 == 1,and C == 1)Cnl for n > 2.It follows that)
C2 == 1 Cl ==
I!,
C3 == 2 C2 ==
C == (n n
 I)! for
00
n=2)))
> 2.)
y(x)
== 1
n . +x + L(n l)!x
204
series is)
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)
+ y = O.)
00
= LCn , xn
n=O)
we find
\037)
that)
y' =
L
n=})
00
} ncnx
n y\" in
and)
\"
= \037 n (n
'\"\"'\"
00
1) Cnx n2
n=2)
Substitution
for y and
L  1)cx  + Lcnx = O.
00
00
n=2
n=O)
We shift the
and
n with
in
the
first sum by
n +2(replace = 2 with
=0
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
this
formula
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 secondorder equation. When we apply the recurrence relation in (27)with we get)
C2 =
n
Taking
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
for k
>
1,)
\037)
k 1)co
(2k)
4
and)
C 2k+l
!)
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) =
(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
(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, andrecognizing imthe of the differential equation y\" + y == Owe 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
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
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)
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.
by
of earlierchapters!))
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
2.y/ = 4y
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
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)
(as in
Example
15.y/ x
+y =0 17.x 2y/ + y = 0)
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
a (12) s the
1.
(12)
initial
value problem)
y\"
= y/ + y,
y(O) = 0,
y(l) =
1)))
207)
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
.. =
Fn2
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,
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
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
1,
y\" (0)
= 1.)
tan
1 = x + x3 + 3
2 17 xs + _x 15 315
(b) Solvethis
initial
that)
(d)
in
it!)
Would you
+....
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.
3.1
(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\"
(3))))
208
is the equation)
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\"'\"
...
series.
.)
Example2)
The point
x = 0 is not an ordinary
y\"
+ 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
becausethe coefficient functions A(x), B(x), and C(x) are polynomials with
A(O)
iO.
.)
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
xa.
209)
(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))
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)
(x2 + 9)y\"
in
+ xy' + x2y = 0)
(6))
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
210
Example5)
powersof x of)
(x2
4)y\"
+ 3xy' + y = o.)
(7))
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)
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)
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
L[(n +
2
n=O)
00
2n
+ l)c n
+ 1)2c n
= 0,)
cn+2
for n
> O. With
c2
 4.2')
Co
in this
 n
(n
(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
=)
1 3 5
n
find that)
With
(2n +
\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)
\037)
(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
Alternatively,)
y(x) == Co 1 + x 8
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
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)
...
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 righthand side in Eg.(11), that which the general solution in the form Y == CoYo + CIYI. .))) expresses
212
Translated eriesSolutions) S
If
in
and
initial
values yea)
the form)
(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)
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)
 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
,)
(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
< 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.)
Example7)
Solution
 x
xy
I
=0
.)
(15))
We
substitution
of the powerseries y
n n
the
equation)
= 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)
 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
nCn
C n
xn 2] = O.
n=2)
recurrence relation)
\037)
0,that
C3 =
C +2 = n
+ Cn 2 (n + 2)(n+ 1))
nCn
214
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
(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
(1 x2)y\"
where the real number the inequality has extensive applications, anging from r integration formulas (such Gaussianquadrature) to the problemof determining the steadystate 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)))
==
L cmx
in
Cm+2
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
CI,)
Co,
S!)
> 0,)
C2m ==
(1)maCa
 2)(a  4)...(a
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
a.
nonterminating
series.)))
216
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(2n2k)!n2k X
 k)!  2k)!
(n
')
(26))
where N
part of
nj2.The first
Po(x) = 1,)
2 P2(x) = (3x
PI (x) = x,)
P5(x) = \037(63x5
8)
 70x + 15x),
3
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\"
1.(x2 2
l)y\"
+ 4xy' + 2y = 0
 
7xy'
+ 16y = 0
 
Usepowerseriesto solvethe
initial
value problems in
Prob
(b) Differentiate
to succession
eachsideof the
(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
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\"
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.)
(c) Show
is)
 2xy' + 2ay
== O.)
(a) Derivethe
YI
two
power seriessolutions)
 2) ...(a (2m)!)
2m + 2)
x2m
In Problems 23 through 26, find a threeterm 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\"
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 nthdegree 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
Ho(x)
 1,
2
coefficients.
28. y\" + eXy = 0 29. (cos y\" + y = 0 x) 30. xy\" + (sin x)y' + xy = 0 31. erivethe recurrencerelation D
12x,)
general formula
Hn(x)
in
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
nthdegree 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 nthdegree 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
2nv
O.)
=xy)))
218
servesto
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
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)
31/6r(\037)
define the standard Airy functions that appear in mathematical tables and computer algebra systems. Their exhibit trigonometriclike 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).) highprecision 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
of convergence a (as
IIIJ RegularSingularPoints)
We now
\037)
+ 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,)
2 )y
= 0,
(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
 , xI
_
dy
dt
dy
dt)
dt dx
1/
d2
and 1
x = 2 1
)]
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.)
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)
will
it
(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
\037)
(3))
where)
\037)
p(x) = xP(x)
and
(4))))
220
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 xn = 0,
x)2
y
noting
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
\"
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
l+x
\037
),+ _ . y
2\"Y x)
Example
1)
x2(1+ x)y\" + x(4 x2)y' + (2 + 3x)y = O. In the standard form y\" + Py' + Qy = 0 it is
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):)
\"
and
Q(x) = x2 + (1 x)
2 + 3x
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
x+o)
x2Q(x).
(6))
If Po = 0
x2y\"
. .
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.)
the form)
II
(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)
\"
we first write
it in
the form
II
(3):)
x)/ + (sinx x, + (1 y
 cosx)/ x
x)2
2
y
sIn x
= O.
lim qo = x+o
 cosx=
X2
lim
x+o
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
q(X)
[  (    +...] 2! )  2! 6! +...
=
1
 cosx =
X2
1 1 X2
X2
+ 4!
6!)
x2
x4
4!
.)
= q(O) =
and
\037,
.)
TheMethod ofFrobenius)
We now
l approach the task of actually finding solutionsof a secondorderinear differential equation near the regular singular point x = O.The simplest uch equation s is the constantcoefficientequidimensional 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)
(8))
stants, it
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
(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.
(18481917),
series.
..
;)
it
is not a series integral powersof x. in To investigate the possibleexistencef Frobenius seriesolutions, we begin o s
the equation)
with
\037)
(10))))
obtained by
point,
then
multiplying
the equation in at x =
.. ..
,
.)
(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)
(13))
and)
\037)
y\"
L
n=O)
00
cn(n
(14))
Substitution
of the
+ [r(r l)coxr+ (r + l)rclxr+1 ...] + [pox + PIX2 + ...] . [rcoxrl (r + I)CIXr + ...] + r + [qO+ q X + ...] . [cox+ C X r+ + ...] = O.
I I I
(15))
then
initial terms of the two products on the lefthand side here and coefficientsof xr , we seethat the lowestdegree 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 higherdegree 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)
(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
y\"
+ Po + PIX + P2x2 + X
...
y' +
qo
==
o.
(17)
Example3)
Find the
 
== O.)
the form)
y\"
==
0,
and
thus
seethat r(r
Po == and qo ==
\037
\037.
 1)+  1
\037r
\037
== r 2
with then
T 1.he
Yl
(x) == X 1/2
L
n=O)
00
anx
and
Y2(X) ==
xI
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))
Let p
L
n=O)
00
and
q(x) = Lqnxn.
n=O)
00
Por + qo = O.Then (a) For x > 0,there existsa solution of Eq. (10)of the form)
>)
rt Yl (x) = x
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
>)
= xr2
L
n=O)
00
bnx
(bo i= 0))
(19))
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.
Example4)
Solution
(20))
in
3 + 1. + y' X
 x
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
\037)
0,)
= xl/2
L
n=O)
00
anx
and
Y2 = xI
L
n=O)
00
bnx
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
than
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
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)
\037r
\037)
Co =
O.)))
3.3RegularSingularPoints
The terms corresponding n to
227)
1 yield)
o.)
\037
or r
1,
it
(24))
in
in
Eq. (23)is)
C n
C n
= o.)
We
\037)
2 2(n + r)2 + (n + r)
C n
for n
1)
> 2.)
r
(25))
CASE1: rl =
=
\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.
   616 ao
14')
a4
a2
ao
44
')
and
a6 =
a4
90 55,440
.)
ao
CASE2: r2 = We now write b n in place of C and n relation) (25).This gives the recurrence Eq.
\037)
1.
4
3n)
substitute
1
in
bn
 2b2
b2
bn 2
for n
> 2.)
n
(27))
Again,
 , _ 20_
bo
2)
b4
bo
40')
an d
b6 _
 _ b4
bo
54
2160)
= box1
.)))
228
Example5)
Solution
\"
,....,,'..'..n.'\"
\"\"'\"
...........
\"'\"
........
,,\"\"\"''''''
'\"
........
\"......\037
\037\"......
\".......
.....................
.....\037.......\"
(28))
+ 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
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
shift
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))
Cn
C2
= 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
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.)
3.3RegularSingularPoints
229)
ri.
.....
Example6)
+ 2y / + xy = 0
2
.)
(31))
Solution
+ y/ + y = 0, X x
2)
x2
with
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.
in
00
00
n=O
n=O
n=O)
We
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
1,
couldexist.)))
230
that
all C n
2 =n(nC n
,)
for n
> 2.)
1
1))
(33))
N
1 1
is) C2n
co
C3= 3 .2CI,
Co 1
')
Cs
C7
CI
')
co (l)n
for n
(2n)!'
C2n+I
Lcnx
n=O)
00
=
y)
Co
x
Co
\037
1_
x2
2!+ 4!
n X
2n
x4
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
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))
Oa
.)
Summary
When
\037)
+ 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
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)
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
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 secondorderinear 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)
O. (b)
= x r L cnx n
can satisfy
n = L !x.
= L cnx n
232
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
Show
with
IS)
most
one
Po = qo
= 1.(b)
C n
recursion formula
= =
Cnl
n2 +
2rn
= L
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 realvaluedsolutions of the form)
YI
of
(x) = A(x)
order 4,)
to derive its
Y2(X)
cos(lnx) B(x)sin(lnx),
x 2y\"
41.
solu
= Lanx
= Cocosx+ CISlnx .
graphs
  l)y =
0)
Figure tions.)
of the
two indicated
appearedin
in the
program
ematical Monthly.
singular follows from
YI)
(a)
Show that
point with
exponents rl
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
(x)
that
= cosxand
Problem
38.
y\"
,JX
Y2(X)
Slnx
in
C +2 =) n (n
2
,JX
 n)cnl +  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
2.
has exponents
series correspondingto
rl
= 1 and
x
= 1at x = 0, and rl = 1 is
')
= (I)n+1
X
1 ::::(!).
(b)
(x) =
= J1(x) 2
00
(1+ X)2)
to
(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)
= XI LCnxn.
n=O)
t 40. Considerhe
equation
x2
y\"
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. wideranging 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
= xi Lqnxn
n=O)
y(x)
=X
L
n=O
00
cnx
L
n=O)))
00
cnx
senes
Co i=
IS)
O. Show
C n+1
that the
recurrence relation
n)(fJ n) C n n) Co
==
for this
+  (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, ) Ix (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).
==
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)
cp(r) = r(r
(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 secondorderinear differential equation with variable l
coefficients.)
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)
(4))
LFn(r)x
n=O)
00
n +r
=0
(5))))
234
r. It turns
out
that
the coefficientof xr
is)
(6))
>
1,
(7))
CI,..., CnI).)
HereLn is a certain linear combination of Co, CI, CnI. Although formula is not necessary our purposes,t happensthat) for i
...,
the exact
nI
(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
(9))
This is a recurrence relation for Cn in terms of Co,CI, , CnI. 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, , CnI 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
(10))
it
stageCo,
...,
happensthat)
CNI)= 0,)
arbitrarily
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))))
5;
= 5.We substitute
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
00
+ (n + r)cnl = 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.
(14))
If n i= 5,we can solve this equation for Cn to obtain the recurrence relation)
\037)
C n
Cnl
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
00
5 1 5
X C5
in
Frobeniusseriesolutions) s
arbitrary
two
YI(X)
= x5
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)
 xy' + (x  8)y =
2 r2
0)
(18))
Solution Here = Po
is) 0) substitution
= r(r 1)
with
   8 =  2r  8 = = 6. On = 2
r differing by N
of
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)
sum and
00
n r
Cn_2Xn+r
= o.
n=O
n=2)
The coefficientof xr gives the indicial equation, and the coefficientof xr+l gives)
 8]
CI =
o.)))
 8]
C n
Cn
+ Cn 2= 0,)
the
(19))
which
ficient of Cn is (n
+ r).
this
coef
 6)c +
n
2= 0)
> 2, n i= 6).)
(20))
for n
\037)
> 2.For n
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
is)
\037=2 .4 ...
. (2n). 8 . 10..(2n + 6)
+6
00
co (I)n
=.
Co 10)
(I)n6co
2 n!(n
2n
+ 3)!)
\037
( _ 1) X 22nn!(n 3)!)
n
2n
)
.)))
of Eq. (18).)
238
tion to
find
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))
then
be given by)
Y2(X) == V(X)Yl
(x).)
(25))
the
the and
== vY\037'
We get)
[vy\037'
+ 2v ' + vI!YI] + P
Y\037
[vy\037
==
0,
.)
I!
== 0
But the
bracketedexpression in
(2y\037
(26))
method is that Eq. (26)is linearin v'. Thus the has reducedthe secondorder 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:
p=exp(1 (2
thus)
\037:
+P(X\302\273)dX)
P(X)dX);) =exP(21nIYll+
p(X) =
Y\037
exp
(IP(x)dx)
.)))
uy;exp
(IP(X)dX)= C,
Y2 Yl
so
Vi
=U=
\037
exp
(IP(X)dX).)
=V=C
f
=
f exp( P(x)dx)dx + K.
Y;)
1 and K
\037)
Y2
= Yl
exp
(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 reductionoforderformula 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
 
+ Poxy' +
Beforewe can apply the reductionoforderformula 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 reductionoforderformula 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 reductionoforderformula can be found in Problems3744of Section we 2.2where introduced the method of reductionof order
in
Problem 36 (though
without
itself).)))
240
equation)
/!
\037)
(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
> 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 ==
(,\037
 'IN)
== 0,)
so we seethat)
Po
that is,)
 2'1;)
N.)
in
(30))
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  ... == xpo exp (PIX 1P2x2 ...)
X
...
)) ,)
In X
so that) exp
( f
(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
..
= Yl = Yl
f f
...
)
dx
(32))
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)
is always presentwhen
'1= '2.)
+ N).
\302\267 \302\267
With
> 0,
= Yl = Yl
f(
  ..
1
\302\267
\302\267
\302\267
) dx)
CN
X
Cl + xN+l + x + .
1
N)
xN =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)
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
to solutions for x
> o.)
THEOREM 1
Supposethat
(4))
Let p
p(x) = LPnxn
q(x) = Lqnxn.
n=O)
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)
\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.
of orderzero,)
= case'l r2 by
....
deriving the
the
first solution)
= Jo(x)=
00 \037
(_ 1)n X 2n
22n (n!)2)
(38))
= Yl lnx + Lbnxn .
n=1)
00
(39))
'\"' nx++ b
Yl X
00
\037n
n=1)
nX
nl
Y2
nX
n2
We substitute
to
obtain)
2!/+
Y2
= (x2y\037'
00
+ x2Y2 + + X 2Yl) In x +
XY2
XY\037
2xy\037)
00
L
n=l)
00
bn xn +2,
and
it
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 =
1
4)
(41))
For n
 
(n!)2)
(42))))
244
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 lefthand 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 righthand
 _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
22n2[(n
2n2 (ltc_
(2)(2)n(2n)
22n (n !)2)
1)!]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) ==
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
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)
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
(49))
With
in
(28)yields)
Y2
\037
xYI)
dx
1
3 5
7
== Yl
dx Yl f x3 (Ix/8+x4/192x/9216+...)2) 2 6
== 4YI
== 4Yl
 f (
3
))
==4Yl
x3
7 4
by lOng division)
4x
) (++ +....
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
n=l
2n
(51)
the terms
.) 1)\"
Eq. (50)agree
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
00
n= 1
2n
(n
.)
4 Examples and 5 illustrate two methodsof finding the solution in the logarithmic casesdirect 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
1.
xy\"
+ (3  x)y'  
=0 =0
 +  4 64
x2 x4
solution) 2
x6 + 2304)
. .
..

 
the method
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
four
nonzero terms
16.ind F
seriessolutions
11. x
y\"
+ (x
 
17.(a) Verify
that
YI
(x)
3x)y'+ 4y = 0)
x 2y\"
of
O.)))
= 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))
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
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
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)
a secondsolution
oo
= YI (x) In x + 1
the differential
 L n!  I)!
n=I
C2n(r) =
With
(n
+ HnI x n .
(r
(1)n
(61)
19. upposethat S
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);
Substitution
of this
result in
is)
O.)
21. erivethe D
tion ing
in
(45).
that is,)
Ln(r;
Co,CI,
<p(r
+ n))
..., .
CnI)
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))
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
C=
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
C2
= 1 = HI + Ho and
C2n
deducethat)
22n (n
 1)!n!)
)n C2n
= Hn + HnI.)
> 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 (17841846),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
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
3.5Bessel's Equation
The general pattern
is)
249)
a2m ==
, 1)(p+2)...(p+m))
solution)
ao l)m
== p
YI(X) == ao
\" f;;o
00
2m
(4))
If p == 0 this is the only Frobeniusseriesolution; with ao == 1 as well, it is the s function Jo(x)we have seenbefore.)
in
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 integerthat 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))
In comparing (6)with
p.Thus
22m
in this
(4),
\037)
Y2(X) == bo
\" f;;o
00
(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
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 ett
dt.)
reI)==
1
o
X
00
==
1.
x [bxett dt
I
(9))
0)
u == t X
and dv ==
lim
b\037oo
et dt:)
b
r(x + 1) =
lim
b\037oo
[bett dt = 10
([
_ettx] + 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
in
general that)
r(n + 1) == n!)
00
for n
> 0 an integer.)
is) 00
(11))
An important
r G) =
et t / 2 dt =
1
21
2)
e
u2
du = ...(ii,)
(12))
u 2 for t in 00
the
first
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 righthand 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
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)
function:)
= 1p(x)
00
.)
(13))
P ho Similarly, if p > 0 is not an integer, we choose == Ij[2f( 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) C2Jp(X))
(I)m
2mp (14))
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
kind
of integral order.Thus)
x2 2
12
=X
+ 22.4 22.4.6 2 2 2
1 x
3
x4
x6
+...

(17))
and)
1 (x) =
1
00
\037
2 I)m2m+
+ 2!.3! 2 ) (
....
they
(18))
if you examine the seriesn (17),you can seepart of the reason why lo(x) and i cosx might be similaronlyinor changesin the denominatorsin (17)are needed m to producethe Taylor seriesor cos As suggested Fig.3.5.2, zerosof f x. the by
the)))
in
re
252
'
''' <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
10.2102 13.3518
16.4934)
3.9270 7.0686
16.4706)
functions
FIGURE
functions Jo(x) and J1 (x) are interlacedbetween 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))
( +
2
Jr
In
(n
\037 Jr m=O)
n 2m
with
\037
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.
\037)
(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.)
3.5.5
y)
y)
0.5)
0.5)
x)
12 (x))
x)
J2(x) and
Y2(x).)
Jp(x) =
in
J;
=
(13))
the
dx
p [x
p (x) ] ==
d
00
00
'\"\"\"\"
x2m+2p (I)m
and
thus
J;
XP
00
'\"\"\"\"
f:=o2
2m
+plm !(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
(23))
If we carry out the differentiationsin Eqs.(22)and (23)and then divide the resulting identities by x P and xP , respectively,we obtain (Problem 8) the identities)
P
x)
Jp(x)
(24))))
254
= 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
(26))
which can be used to express esselunctions of B f functions of lower orders. the form) In
high
(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.
==
.)
Example2)
Usingfirst p
== 2 and then
== 1 in
J3(X) == J2(X)
X
so that) J3(X)=
With
similar manipulations every Bessel function in expressed terms of Jo(x) and Jl (x).
.)
Example3)
by
Eq. (23)with p
==
dx)))
3.5Bessel's Equation
and integrate by parts with) 2 u == X
,)
255)
dv ==
and)
V
xIJ2(X) dx,
du == 2x dx,)
This gives
== _X1 J1 (x).)
secondresult of Example1.)
.)
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))
Eq. (28)is)
\037)
(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
..
= (ak)2 =
,)
(33))
(Y2\037)2
and
that
its
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
Yn(X))
x)
Besselunction f
I (x).)
n
Yn2, Yn3,
..
of the
_.J>_r,obleltl\037_
.)
1.Differentiate
that
J\037(x)
sine functions).
termwise 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...(3ml) r (3).
3m
part
x+o)
y(x)
= x2
(a) and
t Eq.(13)hat
[J /2(X) + J /2(X) ]
I I
JS/2(X) + JS/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
terms
of Besselfunctions
of
the form
f x I (x) dx
m n
can be
and
the
f;
integral
11.1
13
21.
5. ExpressJ4(x) in
Construct tions.
terms
recursion
of o formula
J (x)and JI (x).
in
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).
17.
that)
= ax
(29).
Jo(x) = Jr
e by showing that the righthand sidesatisfiesBessel'squaO. tion of orderzero and has the value o (0) when x this constitutes a proof.))) Explain why
1
1
7r
257)
 1 cos(en
0)
x sin
e)de
Z\"
(1
p2X\037
by showing that the righthand 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
 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))
> 2, show that the righthand 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)
ex
2n
In(x)
2
nx)
(36))
first that)
10
then
{2n eixsinO dB
= {n (eiXSinO + eiXsinO)
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
X 1/2z
Bessel's
ratio
+
Yo(x)
\037
equation
of orderp,)
x 2y\"
j j:x
nx)
cas(x
sin
 in)  in)
in that the
(x
approximations
approachesnity u
the form)
dz)
(Z2
==
0,
(1))
and
substitute)
ex w == X y
,)
z == k x
t3
.)
(2))
Then a routine
but
x 2y\"
+ (1 2a)xy'+ (a2
f32 p2
f3
2k 2x 2t3
)y ==
0;)))
258
x 2y\"
(3))
 2a,
a==
fJ2 p2,
C == fJ 2k 2,
== q
and
q == 2fJ.
(4)
It
IA
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)
(6))
where
Jp is to be replacedwith
\"..\"
Example
1)
Solvethe
equation)
4x2y\"
+ 8xy' + (x 4
(\037
 3)y
==
o.
(7))
Eqo
x 2y\"
seethat
a ==
A ==
\037,
fJ
==
2,k ==
Eqo
2,B ==
\037,
+ 2xy' +
C
==
\037.
==
\037,
\037,
and
p
I
y(x) == x
/2
[CIJI/2(\037x2)
Ifwe recallrom f
1
J / 2(Z) =
Jnz
sinz
and
y(x) = x /
32
( 4 + B 4)
Acos
sin
x2
.)))
259)
(8))
Solution
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,
.)
.)
3.6.1.
(9))
FIGURE
column.)
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. physicalreasonsno For bending at the free top of the column and no deflection at its imbedded bottomthe boundary conditions are)
(10))
to solve it
form. With)
  EI')
Y
gp
(11))
+ y 2x() = 0;
(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.
(13))))
260
initial
00
( m!r(p+m+l)2 ) Jp(X)=\037
and
find
(l)m
2m+p
')
after somesimplificationsthat)
B(x)=
 y2x4 + 504 ( 12
1_ y 2x3
4 7 y x
+
From
this it
c231/ 3
y 1/3
r (D (
+ y180
4x6
... ) ... )
.)
Cl
= 0, so)
(14))
.)
J / 3
1
(\037y
L 3/ 2)
= o.) =
\037y
>.)
Thus the column will buckleonly if z JI /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)
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
(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 crosssectional we finally get) area,
;;
1/3
Y/2r/
Ll
\037
EI (1.986)8A) g
( )
(17))))
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.,, __ _ ____, '_' _',__.'_,) .... ...
'
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
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\"
J /4 ( 1.X2) . 2
(\037X2)
   xy
\"
dy dx)
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
+ 2y ' + xy
14. erify that the substitutions in V (Eq.(1)) yield Eq.(3). 15.a) Show that the substitution) (
y
is y(x)
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
3)
= x 2 + y2.
262
circular
crosssection,subject to an axial forceP of compression. As in Section2.8, deflection curve y == y (x) satisfies its
Note that
if
==
b, this
result
+ 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)
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
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
to horn;
.. ...
==
x 4y\"
where
A
==
+ AY == 0, yea) == y(b)
==
0,
the
J12
Pb4
jE 10. (a)
general solution of
Apply
theo
y(x)
=x (ACOS + Bsin
\037
\037
).
A FIGURE3.6.6. variablelength
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)
Bessel functions.)
dx
dy
dx)
= x2 + l;)
(1))
_ = x2 _ y2.
(2))))
263)
(3) (4) (5) (6)
dx
= y2
2 ;
dx
dy
dy
= x + y2; =x
dx
y2;
dy =y2 x.
dx)
this
y(x) = x
 cJ3/4
(\037x2)
1
CJ1 / 4 ( 2X2 )
+ Jl/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
(8))
Kp(x)= 2i
that
T(
P [Jp(ix)+ Yp(ix)]
p.For instance, the general solution of Eq. (5)is given for x > 0
_ X 1/2I2/3 ( 3x3/2) y(x) 2
11/3x3/ 2) (3
 I2/3  C/l/3
C
by)
2 3/2 ( 3x ) 2 / , ( 3x3 )
2)
(9))))
264
c = y(o)r /3
1
3
(\037)
(10))
(\037))
4)
2)
\037o)
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
...
y2.
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 secondorder equation)
y\"
3.5. (18)in Section The secondorder differential equations of the form y\" = f (x,y) with the same righthand 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 secondorder 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
2 0
\037
\"
,.,.\"y2
=x
2
4
0 2 4 6
x)
10
y\"
12
==
0, y'(O)==
y2
 x,
265)
) (12))
(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 righthand side with x , where s is an odd integer. (With parities other than these, more exoticpecialfunctions are involved.)))) s
(12).
 1,
LaplaceTransforll1
Methods)
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 massspringdashpot 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 discontinuitiesfor 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)
of
f is the
(1))
F(s)=
\302\243{f(t)}
1 e
00
st
f(t) dt)
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
in
(1)gives)
\037
st roo e dt Jo
OO
_\037est
]0
lim
b\037oo
_\037ebS
S)
],
(3))
and therefore) 1
\302\243{1}=s)
for
s>O.
in (3), it's goodpracticeto specify the domain of the Laplacetransformin 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
...
}=
1
o
00
este dt =
at
1
0
00
e d e(sa)t t = _
(sa)t 00
a ]
t =0)))
268
\037
0 as t
at
\037
\037)
\302\243{e
} =)
sa)
i{3t
for
\302\243
> a.)
(5))
= a + if3,)
\037
in
a.
as 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 ett  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) = 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
a
\302\243{t
}=
1 estt
00
a dt.)
this
If we substitute
integral, we get)
sa+l)
(10))))
}=
f(a + 1)
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.)
(12))
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
r
5
3 3
()
\037
= 0T)
(13))
()
2222224'))) () ()
=
\037r
\037
0T
270
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)
\"..,
..,
'\"
.....n....;...'....\"N\"
'\"
................\"
VN,';,'N.V
\037\"\"'....;....;..'N..'........
....nnn
\"\"
\037_....\037
.V',','\",.\"
,....\".,....\".,....\".
'\"
,\037'\"
\".
........
...
\037....\"
\".
\037,nnn.\"
\037\037,....\".
....NU.N....\",....\".,....\".
\037
\"',....\".
__\037........
1
_\302\243{e
kt
+2
kt
_\302\243{e
}=
1 1 1  sk+ 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))
(5)(with a = ik)
kt {cos } =
1 1  s 1 ik + s +1 ik =  . S22s 2( ) 2 (ik)2
')
and
thus)
\302\243{coskt}
s s 2 +k)2
fors > O.
(16))
\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
we get)
\302\243{3e
2 + 2sin3t} =
+ 1 cos6t}
1
O.)))
36))
.)
f(t) = \302\243I{F(s)}.)
(18))
Example7)
I
{s3 }
 =t 2')
I 1 2
\302\243
I
{s+2) }
= e2t ,
2
\302\2431
{ +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
eat)
sa)
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).
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)
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.)
step function.)
= 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)}
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
1
o
estua(t)dt= =
consequentl y,)
\302\243{Ua(t)}
eas
S)
1
a
est dt =
lim
b\037oo
st =S
t=a)
(S > 0, a >
0).
(22)
.)
l
c piecewiseontinuous for t
b
g(t) dt)
the
dt)
J is
integral)
< +00.But
in
order for F (s
l e !(t))the
as
as)))
J(t)
\037
4.1Laplaceransforms and InverseTransforms 273) T T if order +00. he function is saidto be of exponential as t +00 there
\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 boundedfunctionsuch cosktor as sin ktis 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
+00as t
(t)
= et2 is not
of
e Laplacetransform. The followingtheoremguarantees that piecewise functions of exponential orderdo have Laplace transforms.)
the function
10
00
estet2 dt
e st because  et2
\037
as +00 t
w +00, e seethat the improperintet2 L{e } doesnot exist(for any s), and therefore
\037
t if transform +00,hen its is piecewiseontinuous and satisfiesthe condition in (23),hen F(s) existsfor c t all s > c.)
\037
(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 integralsthe 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
existsfor s
00
integral)
lest(t)1 f
dt)))
274
B +00. ut
l
if
lest(t)1dt < f
M lest ect dt = M
I
<M
s
1
o
00
00
M e(sc)t t = d
e(sc)tt) d
sc)
.)
(24))
IF(s)1<
if s
> c.When
1
o
sc
we take limits as s
\037
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 polynomialscan (and is, as we erally, a rational functiona 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
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.)))
integrals in his work on probability theory. The socalled operational methodsfor differential equations, which are basedon Laplacetransforms, were not exsolving I ploited by Laplace.ndeed,they werediscoveredand popularized by practicing O the (18501925 engineer liver Heaviside engineersnotably 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(17491827),
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.)))
15.(t) = 1 + cosh5t J
0 
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
Apply the
geometric seriesto
\302\243{f(t)}
23.F(s)= 4 s
1 25.F(s)= S
1e) . s(l
S
5/2
s+5
3s + 1 s2 +4
function 00
f can be written
n=O)
f is shown
in
Fig.
4.1.10.
in the form)
27. F(s)=
s4
29.F(s)= 5 +3s S2 9
show
that)
s(l+ eS)
.)
33.Derivethe transform
in the text to
derive the
of f (t)
used used
(16).
34. Derivethe
transform
in the text to
derive the
of f(t)
IL
1 Problem40.)
eo eo
2
(14).
. .
0
4 3 5
......
0
6
function
t)
35.Usethe tabulated
integral)
f
S 36. how
of
to obtain
directly
of
the
41.The graph
transform. Laplace
that the function
\037
order as t
t
of the squarewave function get) is shown in in terms of the function f ofProblem 40 and hencededucethat)
\302\243{g(t)}
is not.)
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 eS)
es
= 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
f,
f,
345
t t
6)
t)
show that
\302\243{f(t)}
bs as = sl(e e ).)
of
T 39. heunit
f(t) = n
1<
to
< n,
1,2,3,...
.)
seewhy
u(t
its name
is appropri
if n
f(t) =
L 00
n=O)
{b
if n
1< 1<
n)
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
+ 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
for t
> T.)
(3))
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)
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
t=O
2 of Section
preceding expression.
given in Eq. (4). We will defer the has isolateddiscontinuities to the end of this section.)))
est f(t)
f (0)
to the evaluation of
by
> c) as t
which
+
Theorem
in case
existswhen
f'(t)
278
1,
\302\243{f\"(t)}
= = = S\302\243{f'(t)} = s [s\302\243{f(t)}
\302\243{g'(t)}
and
\037)
thus)
\302\243{f\"
 f' (0).)
(5))
S\302\243{f\"(t)}
(6))
After finitely
many
COROLLARY
Transforms f HigherDerivatives o
Supposethat the functions f, f', f\", .. fCnI) 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)
.,
...  ...
C
C
  f nl)(o)  f nl)(o).
C
(7))
Example
1)
....
xxx=;)x(O)= 2,
II I
x'(O)= 1.)
Solution
With
the given
initial
s\302\243{x(t)}
and)
\302\243{X\"(t)}
= s2\302\243{x(t)}
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
(s3)(s+2))))
4.2Transformationof Initial
By the method of partial fractions (of integral
Value
Problems 279)
B such that)
2s
3
(s3)(s+2)
and multiplication of both
s3+ s+2,
identity)
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\243l{l/(s
 a)} = eat,
7 + s+2. s3 5
it
follows that)
\037e3t
x(t) =
\037e2t)
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 corollarythe given initial conditions.
Remark:In Example1 we found the values of the partialfractioncoefficientsA 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))
\"surefire\"
resultedfrom clearing fractions. In lieu of any such shortcut, the method is to collect coefficientsof powersof s on the righthand 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
samevalues A =
B=
\037.)
.)
Example2)
Such a problemarises in the motion of a massandspringystem with external s force, as shown in Fig.4.2.2.)))
280
Solution Becauseoth b
J(t) = sin 3t
initial
transformed equation)
\302\243{x\"
Therefore,)
2.The
X(s)
(s2 + 4)(s2+ 9)
.)
+ B + Cs+ D s2 + 9 s2 + 4
.)
The surefire approach would be to clearractions by multiplying both sidesby the f common denominator, and then collect coefficientsof powersof s on the righthand 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 odddegree 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
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)
\302\243{sin
to
sin 2t
\037
sin 3t
2)
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))
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 constantcoefficientlinear 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.)
..
Example3)
Solvethe
system)
2x\" ==
y\"
6x+ 2y,
== 2x
 2y + 40
sin 3t,)
(8))
(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. massandspring 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
Because
equations)
\302\243{sin3t}
+ 6X(s) 2Y(s),
120
(s2+3)X(s))
 Y(s) =
X
0,)
120
(10))
 2 = (S2+ 1)(s2+ 
Cramer's for instancetheystem in (10)for) rule, solveusing s 3 120 8 = 5 + s2+9 X(s) = (lla)) s2+1 s2+4 (s2+1)(s2+4)(s2+9)
and)
Y(s) =
(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
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
yet)
in
two
masses.
..)))
4.2Transformationof Initial
The TransformPerspective
J(t))
Value
Problems 283)
transformed
FIGURE4.2.7. massA
springdashpot system
external
with
kX
(s) = F(s).
(13))
forceJ(t).)
Note that Eq. (13)is an algebraic a linear equationin \"unthe equationindeed, 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
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 righthand sidein Eq. (14). Much of the remainder of this chapter is devoted to finding Laplacetransforms and inverse transforms. In particular, we seekthose thatunlike methods that are sufficiently powerful to enableus to solve problems thosein Examples1 and 2cannot 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
+ ateat } =
at
\302\243{e
\302\243{f'(t)}
= s\302\243{f(t)} =
that)
}.)
linearity
of the transform
}
at
a\302\243{te
}=
at
s\302\243{te
}.)))
284
{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}.)
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}.
2ks (s2
+ k2 2
))
(16))
the
\302\243{t
sin kt}
00
.)
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\2431
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
g(t) =
it f('c)dr:)
and
t
s piecewisemooth for t
> O.Furthermore,)
< Ig(t)1
this gives)
i
o
IfCr)1di < M
i
0
eCTdi = _(ect
C
M  1) < _e
C)
ct
\302\243{g'(t)}
s\302\243{g(t)}
 g(O).)
\302\243{f(t)}
Now
\302\243
{itf(r:)dr:
\302\243{g(t)}
})
s)
which
\037)
Example6)
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\2431
= t {s(s a) } 10
1
\302\2431
We now
\302\2431
{s2(s a) }
\302\2431
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
We
10)
{best (t) dt
I'
existsand alsoneed to find its value. With b fixed, let tl, t2, ..., tkI 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 (tnI,t n ) where f' is continuous.
This yields)
st (b e (t) dt =
10 o
l'
I: j
k k
tn
n=1 tnl
=
Now the
k first summation)
t
\037[est
+
\037
tn
tnl
est I(t)dt.)
(19))
(20))
sb down (19)telescopes to f(to) + estk 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 est (t)dt = 1(0) esb I(b)+ S est I(t)dt. + in
I'
But from
(with
== 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 thatwithin c eachinterval [tnI,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+
'>..
'nl
f(t)
and
f(t;)== lim
t \037;;) t
f(t)
not agreewith the actual values f(tnI) 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
st ,
tnl
(t )
st
\\
st2
== where)
n)
+ esb f(b),
(20/)
(21))
that
\302\243{f'(t)}
denotesthe
f(t)
at
t == tn.
Assuming
00
exists,we
==
n)
(22))
of \302\243{f'(t)}
==
in
(19)as b
==
\037
+00.)
Example7)
6
5
J(t)
4
3
2
1)
f(O) ==
Eq. (22)yields)
.... 1,2,3,
in
Fig.4.2.8.
Hence
0==sF(s)  Le1
n=l)
is)
123456
x == es <
F(s)
==
1  Les
00
n=O
ns ==
s(1 e
S
))
In the last step we usedthe formula for the sum of a geometric series,)
00
Lx
n=O)
1)
Ix),
with
1.)
.)
3.x\"  X
lIB Problems)
UseLaplacetransforms
Problems
1through 16.
to
solvethe
initial
value problems in
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
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 + et ; 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' = 2xz;x(O)= 1, (O) = 0, y
is shown
i (l)[t]s the
in
Fig.
then) 4.2.9,
squarewave function
whose
\302\243{f(t)}
1 =  tanh s
.
2)
.)
(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
17.F(s)=
s(s 3)
1
s(s2+ 4)
1 1 S2(S2
in
\302\243{f(t)}
s(l+ eS)
.) .)
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
5)
26. pply A
for
\302\243{cos
\302\243{sinh
27. (a)
Apply
Theorem1 to
at
\302\243{tne
show that)
=)
}
sa)
\302\243{t
 eat }.
1
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
get))
28. 29.
\302\243
456
\302\243{t
30.
\302\243{t
coshkt}
31.pply A
that)
Fig.
4.2.12,
then)
sawtooth function
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)}= Sle
(Suggestion:Note that
J(t))
1)
f'(t) = 1 where
it
is defined.))
=e
as ebs
s)
456
37.)))
t)
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 oCI{R(s)} based on the same method of partial fractions that we use in ele
pes)is lessthan
that
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)
(2))
where A I,A 2,
RULE
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
An s
+ Bn
(3))
where AI,
A 2,
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
onthesAxis 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,
>)
Thus the translation s + s a in the transform correspondso multiplication of t the original function of t by eat.)))
290
a
00
in
the definition of
F(s)=
at
\302\243{e
\302\243{f(t)},
F(s a) =
1
it
00
e(sa)t(t) dt = f
1 e
st
dt [eatf(t)]
f(t)}.)
.)
If we apply the translation theorem to the formulasfor the Laplacetransforms coskt, and sin kt that we already knowmultiplyingeachof thesefunctions eat and replacing with s a in the transformswe the following additions s by get to the table in Fig. .1.2.) 4
eat t n
eat
coskt
eat sin kt
(s > a)
k2 k2
(6)
(7)
(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\037d=\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
+ 6x'+ 34x = 0;
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)
.)
x(t)in
Example 1.)
.)))
Example2)
of)
 
S2+ 1
8s)
.
s(s+ 2)(s 4). Hence)
4)
 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)
= 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)
+ 4y' + 4y = (2;
= + + yes) = s3 (s + 2)2 s3 s2 S
To find A,
ABC   +
D (s + 2)2
+.
E s + 2)
(9)
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))
Eq. (12)yieldsC =
To find D and
4,and
substitution
of s
= 0 in
= D + E(s+ 2) + (s + 2)2G(s),
then
(13))
Substitution of s
Thus)
2
in
\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
\037te2t
\037e2t.)
.)
4, Examples 5,and 6 illustrate techniques for dealing with quadratic factors in partial fraction decompositions.)
Example4)
Considerthe massspringdashpotystem 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.)
60
X(s) =
60
As
+B +
Cs + D
we multiply
both
Eq. (15);
equate real parts and imaginary parts on each sideof this equation to obtain
== 24A + 30B 60
and
which are readily solved for A == To find C and D, we substitute (s 3)2 25 in Eq. (15)and get)
\037\037
we equate real parts and imaginary parts; this yieldsthe two linear equations)
== 186C 12D 60
and
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
10s 10 + (s + 3)2 + 25 )
\0371
0.5)
Transient)
After
x(t) == 19(2cos2t
2
sin 5t).)
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
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)
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)
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).)
(18)
and
\037
envelope
and
then
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. .)))
Example6)
Solvethe
.......................
initial
y(4)
+ 2y\" + y = 4tet ;
Solution
First
weobservethat)
oC{y\"(t)}=
t
\302\243{te}
(s _ 1)2
')
4
1))
yes)
 (s  1)2(s2+ 1)2
4
=
If we multiply
A(S2
B  1)2+ (s
A
.)
(19))
the
equation)
+ 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. + +
(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
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))
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
obtain)
yes) =
2 (s  1)2
+ 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
III) Problems)
Apply
the translation
in
Jrt
of the functions
Problems
1.f(t) = t 4e
t)
Laplacetransforms
UseLaplacetransforms
to
solvethe
initial
value problems in
))
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 = et ; x(O)= x'(O)= 0
31.(3) + x
s2 + 4s + 4 +5 9. F (s) = s2 3s6s + 25
2s
3
Problems
11 through
Laplacetransforms of
22.)
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'
illustrate
2 6))
s massspringdashpotystem
conditions
with
x (0) = x'(0) = o.
16)
20.F(s)= S4
 8s + 16
2
a figure
t Eq.(16)o
21. (s)= F
s2 + 3 (s2+ 2s + 2)2)
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
2
))
A(t)
tetj5 is A(5)
Fig.
4.3.5)
= 5je.Thus
of the
26.
23.
24.
\302\2431
4a4 {S4:3 }
= coshat cosat =
2)
\302\2431
L4:
s2
4a4
}
}
2:
2a
2 sinh
at sin at
/)
2
oscillation in Problem
sinh
x = + te t15
t
IOn
25.
\302\2431
4 { s4 +4a
\037 x = te t15)
cosat)
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\"
(initially
+ x == cost;
=
\302\267
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
\302\243{cos
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
I and g is defined
(3))
(f
We will
* g)(t) =
1f(i)g(t i)di.)
1
also write
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
=
g(u)f(t
u) du
= g(t) * f(t).)
I.)))
* g == g *
298
Example
1)
(cost) * (sint) =
We
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\243l{F(s)
(5))
that
Thus we can find the inverse transform of the product F(s) G(s),provided
integral)
\037)
. aC1{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 {ter sin2T dT = et eT (sin2T  2 COS 2T) 5
,.,
2
\302\2431
{ (s 1)(s+4)}
2
t e10
t t
sin2TdT
= et
so)
\302\243
10
]0')
.)))
1
{(sI)(s2+4) }
==
2 e  5 5
1.  SIn
2t
5)
cos 2t.
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
If J(t)is piecewiseontinuousfor t c
for s
\037)
> c.Equivalently,)
== L1{F'(s)}. J(t) == L1{F(s)}
t)
(7))
(8))
for n ==
1, ,3, ... 2
.)
Example3)
.)
(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
.)
s}
{ds
s)
= =
roC { 1 + (l/S)2) }
1
t
I/s2
1}
r
oC
 I
{S2+
r(
sint).)))
300
\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)
\"\"
n\037.....,_,\",
'\"
'\"
NY''''
\"\"
,...\"\",.\037....,....,,...
.\"
tx\"
+ x' + tx
== 0,)
such that
denotedby Jo(t).Because)
\302\243{x'(t)}
and
\302\243{X\"(t)}
== s2X(s) s,)
and
because and x
d
transformed equation)
ds
[s2X(s)
==
o.
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))
.)))
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+
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
t\302\243I
{[(X)F(a)da}.)
(13))
Example6)
Solution
condition in
sinh t
t
= lim
t+O
2t
t+O
2)
with
t, yields)
  1 (0'10'+12a+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
Fi\037d\037:/:'
f{
i;i(\037i\037
\037\037{)Ii:)))
302
Solution
We
it
{ (s2
 1)2} =
t\302\2431
{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
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
eS(tT) get
 T)dt
(u
=t
T),)
estg(t T)dt,
= F(s)G(s) G(s) = =
00
eSTJ(T) dT =
00
1 1 1 (1 e eSTJ(T)
00
(eST
est get  T)
T)dt)
00
eSTf(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: McGrawHill, 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
= F(s)G(s)
limit
of the inner
F(s)=
d F(s)=
=
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\2431
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)
t.
1 e f(t)dt.)
at
00
[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 =
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\2431
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)
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)=
s(s 3)
1
8. F (s) 10.F ( )
9. F(s)= 13.F(s)=
In
(s2 + 9)2 s
 S(S2+ 4) s +
1
(s  3)(S2+
1))
1 S2(S2 k 2) 1
37. x\" + 2x'+ x = f(t); x(t) = rer f(t 1/ = 38.x\" + 4x' + 13x f(t); x(t) = 3
l
In
r) dr
i
0)
f(t
 T)e
2Tsin
3TdT
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)
21.(t) = e f
3t
Laplacetransform of f(t).
1
F(s)=
where 0 < k < 1 and s > c. Then) the 00
\037
LS
n=O)
00
an n+k+l
of the
functions
in
Problems23
S2+ 1
seriesconvergesbsolutely for a
23.F(s)= In s+2
25. F(s)= In
s2
(1
f(t)=
Apply this result in
S2+ 1
' 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)
s (S2+ 1)3)
 .)
In Problems 29 through 34, transform the given differential equation to find a nontrivial solution such that x (0) o.
Expand with the aid of the binomial seriesand then compute the inverse transformation term by term to obtain)
31.
35.Apply
\302\2431
Jo(t)
Finally, that)
=C
00
\037
2n (_I)nt
22n(n!)2
that
theorem to show
40. Expand
e F(s) = SI/2l/s in
C = 1.
powers of S1
to show that)
 I1)viss } = {(s
2\037 V JT 10
[./I e
u2
du
= e/erf.Jt.
\302\2431
(Suggestion: Substitute u
= ,Jt.)) = x' =
{vis
\037el/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\2431
{\037eI/s
= Jo
(2.Jt)
.)
.\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 simpleonoff function is the unit step function that we introduced
in)))
Functions 305)
the
unit
.)
x = Ua(t))
a)
FIGURE
unit
step function
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
U a (t)
\037)
oC{u(t
 a)}= es)
\037
.
as as
(2))
Because oC{u(t)}= lis, (2)impliesthat multiplication of the transform of u(t) Eq. t t a in the originalindependentvariable. byeas correspondso the translation t
this
Laplacetransformation.)
(3a))
(3b))
for s
> c + a.)
u(t
Note that)
 a)f(t a) =
f
1f(t
a))
(4))
Thus Theorem 1 impliesthat oC1 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(ta)
a.
= + a then
00
= eSTJCC)di
00
eS(T+a) J(i)di.)
yields)
easF(s)=
1 e J(t 00 st
a)
dt.)))
306
Eq.(4) we seethat
this
is the sameas)
\302\243(u(t
easF(s)=
becauseu(t
Theorem 1.)
 a)f(t a)},) eS1u(t a)f(t a)dt= 100  a)f(t  a) == 0 for t < a. This completesthe proof of
\037)
Example
1)
With
eas == u(t { s3 }
........ \037'n.'n.'n_\"\"\"\",,,,,,,,, \"'.,;,.....
_
if)
 a)(t a)2 2
1
\037\"'\" ,.......,,,.. \"..ou,\", .... ..... ........\" ..._..,............. ... nn..... \"\"',.............
==
14(ta)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
== \302\243{u(t
 3)J(t 3)}
2t)
e3sF(s)== e3s
3\" s
6 9 + 2\" +  . s) s
.)
Example3)
Find
\302\243{f(t)}
if
f(t) =
\037os
if 0 if t
(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)
Functions 307)
note first
that)
f(t)
==
[1
  2n)]cos2t
u(t
== cos 2t
2JT))
2n s
oC{cost} == 2
s(l e2ns ) .
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
Solution
We
initial
value problem)
(s2+4)X(s)==F(s)== s2 +4)
so)
X (s) =
 e(s2 + 4)2
s
S
2ns
Because)
oC
1
{ + 4)2 }
(s2
u (t
== 1 t Sin 2t 4
by
n)
sin 2t [t
nl2)
/ /)
x = n 12)
==
\037
(t
sin 2 (t
.)
2JT)
u (t
(t
sin 2t
that
be written
If we separatethe
in
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.)
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
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
......
Solution
L)
We
\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
(6))
switch.
= 90[1
  1)],orresponding c
u(t
closingof the
apply
=dq
dt)
(7))
to obtain the
secondorder equation)
2i 2
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))
i i(r)dr=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.)))
Functions 309)
presentexample, q. (9)is) E
 110i
di dt
  1)].
u(t
(10))
Because)
\302\243.
{10
[ti(r)dr
} =;)
/ (s)
by Theorem 2 of Section
4.2 transforms on
90
We
I(s)=
But)
90 s2 + 11s + 1000) O
so we have)
s + 10) s + 100')
.)
 e
100t thus ;
is)
i(t) == e10t
After
_ e100(tI)].)  e   1)[e10(tl)
100t
u(t
> 1, e find w
that
the current
in
the circuit is
given
i (t) ==
The portion e10t e100t of the solution would describethe current if the switch .))) wereleft closedor all t rather than beingopen for t > f
I e
el
Ot
10t
e_
lOOt
if t if t
< >
1,
1.)
1.
310
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
J
f(t + p) == f(t))
I)
(11))
holdsis called for which Eq. (11)
for all t
the
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 ePS
 1 e f(t) dt.
st
0)
gives)
1
o
00
est f(t) dt == L
n=O
00
i + np
l
np)
cn
+l)p
in
the
nth
cn
+l)p
est f(t) dt ==
np
1
0
escr+np)
1
0)
eSTf(i)di
F(s)=
\037
(e
nps
= (1+ ePs +
Consequently,)
l e I(r:)dr:) e ...) l
P 2ps
eST: d I(r:)r:.)
F(s)=
We
 e
PS 10)
Ix) ==1+x+x+x
with
23+...,
i seriesn
the
the
final
.)
transform of a periodicfunction
the
improperintegral.)))
Functions
311)
D
Example6)
/(t)
1
4.5.9 shows the graph of the squarewave 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.)
F(s)=
 e (2ae f(t) dt r +  e (Jo eS1dt1 (l)e )  e ([;eSIJ: [;eslIa))  eas (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))
==
1 s
tanh
as
2)
(13b)
.)
shows Figure 4.5.10 the graph of a triangularwavefunction 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) triangularwave
t)
a 2a 3a 4a 5a 6a
G(s) == s
as F(s)== 1 tanh. 2 2) s
(14)
.)
Example8)
Considera massspringdashpotystem 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
312
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
so that)
20
4.5.11.
== n
is)
20 s) (1 20 s) (1
ns
(1 +
ns
2ns
_ e3ns
+ ...)
ns
2ns
3ns
+ ...) ,
20 40 F(s)== s + S
00
enns.
(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))
in
L
')
g(t) == L
+ {s[(s 2) + 16] }
2\302\2602
10)
r 5e,sin4r di.
2
g(t) == 1 e 2t (cos4t +
sin 4t) == 1
(t )
,)
(18))
where)
.)
(19))
term
Now we apply Theorem 1 to find the inverse transform of the righthand The Eq. (17). result is)
u x(t) == g(t) + 2 L(I)n (t
n=1)
00
 nn)g(t  nn),
in
(20))
g(t nn) == 1
== 1
 nn)])
4 sin
Functions
313)
(21))
if Hence 0 < t
< n,then)
X(t) == 1
h(t).)
If n
 )]  2 [1 e
h
(t
:rr:
(t )] ==
 + (t) 1
h
:rr:
2h (t)
[1 e2
(t ) ]
:rr:]
.)
If 2n
   
4:rr:
:rr:]
.)
(t) + 1)  2h
(
n
 +...(_1)ne +
:rr:
2n
:rr:])
(t )
(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
e2
:rr:
 2 .e(+ e
1) n 2
:rr:
2:rr:
e2(tn:rr:) (cos4t +
in
!
2)
sin 4t )
solution)
(23))
for nn
Xtr(t)
term
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 e2, (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
\\ \\ \\
\\
2t (2.23)e
\\ \\)
\037/ )
\"'1)
/
/
1
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
.)
1through 10.hen sketch the graph of f.) T 3s S_ 3s e1.F(s)= \037 2. F(s)=) e S2es S 2 eS e)2s =) 3. = e4.
Problems
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 squarewavefunction of Fig. 4.5.13
23.Apply
{
F(s)
 e
271:
1)
oC{f(t)}=
+ s(1 eas))
10. (s)= F
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(e71: S _ e S2+ 4)
given in
rI rI
I I I I
rI)
271:
s)
2a
3a 4a 5a
6a)
t)
Problems
26.Apply
Theorem2 to show
that the
Laplacetransform of
is)
))
J(t))
a 2a
3a
4a 5a
6a)
t)
Functions
315)
as) s(1e
')
and construct
31.
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)
Problems 36 through 40, the values of the elements RLCcircuit are given. Solvethe initial value problem)
In
ojan
29.) Suppose that (t) is the halfwave 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 _
.)
e1Csfk)
7r
k)
27r
k)
37r
k)
t)
impressed voltage e(t). 36.L = 0, R = 100,C = 103; = 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 = 104; = 100sin lOt if e(t) o < t < n; e(t) = 0 if t > n 39.L = 1,R = 150,C = 2 X 104; = lOOt if 0 t < 1; e(t) e(t) = Oift > 1 R 40. L = 1, = 100,C = 4 X 104; (t) = 50t if 0 t < 1; e e(t) = 0 if t >
::::
::::
\037
1)
rectification
of
30. et g(t) = L
function
of Problem 29 and k > O. Note that h(t) = f(t) + get) is the fullwave 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 massspringdashpot 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
1])
7r
k)
37r
k)
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 squarewave function that has amplitude A and period2n. also (See the definition of a squarewavefunction 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 squarewave function with k amplitude 4 and period 2n. 42. m = 1, = 10, = 2; f(t) is a squarewave function k c with amplitude 10 and period
2n.)))
316
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))
p=
lbd
dt)
[mv(t)] dt
= mv(b) mv(a).
(2))
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
I1
\342\202\254
if a
da,E
(t) =)
E)
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)
p=
(t).)
l
a
da,E(t)dt= a
a+E
1 dt=l.
E E may
00
da,E(t)dt =
1.)
(4))))
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 (19021984), the early who in in 1930s introduced a \"function\" allegedly enjoying the properties Eqs.(6)and (7).)
at except a singlepoint,
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
+\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
(8))
get)
,)
1
We
00
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).)
whichwhenappliedto 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
1
We
00
(10)) to be)
(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\"
(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))
But
what
x(t) =
where X E (t) is a solution of)
Ax\"
lim
E\037O)
XE
(t),
(16))
(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 . (
eas
S S as 1 )
SE
SE)))
\037
e)SE =
SE)
by
,)
(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
, ,
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)
We
 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 taxis (Theorem 1 of Section 4.5),we seethat the inverse transform is)
 2Jr)
sin 2(t
 2Jr)
1
= 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
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
x=3)
\037
0)