Burkill A First Course in Mathematical Analysis PDF
Burkill A First Course in Mathematical Analysis PDF
Burkill A First Course in Mathematical Analysis PDF
r s B N0 - 5 e 1 , - a t q t s - 1
ffifillrf
||ilXilillili
A FIRST COURSE IN
MATHEMATICAL
ANALYSIS
BY
J. C. BURKILL
Sc.D.,F.R.S.
@ CamblidgsUniversityhess 1962
Printedin GreatBritair! by
Redra,ood Pres6Limited
Melksham,Wiltshire
1. NUMBERS
Ll, The branchesof puic mathcmatici, p. 1. 1.2. The scopeof mathcmatical
analysis,p. 2, 1.3.Numbers, p.3. 1.4. lrrational numbe$, p,7. 1.5.Cuts
of the rationals, p. 8. 1.6. The field ol real numbers,p. 10. 1.7.Bounded sets
ofnumbers, p. 13, 1.8.'Ihe leastupperbouod (supremum),p, 15. l.9 Complcx
numbers,p. 18. 1.10.Modulus and phase,p, 19,
2. SreusNcss
2.1, scquencqs,p. 23. 2.2- Null sequencs,p, 23, 2,3. Sequencetcnding to a
limit, p. 25, 2.4. Sequencestending to infinity, p. 26. 2.5, Sum and product of
sequ6nces, p.28. 2.6, Increasingsequences, p. 31. 2.7. An important sequence
an,p.32, 2.8, Rccurrnce relations,p. J4, 2.9. Infiniteseries,p.33. 2.10,The
gcometric series:r.i, p. 39. 2.11, The serics I'n-k, p.40. 2.12. Ploporties of
infinite scries,p. 43.
3 . C o N r r N u o u sF u N c r r o N s
3.f, Functions, p. 47. 3,2. Behaviour of /(x) for large valucs of x, p. 49.
3.3.Sketchingofcurves,p.49.3.4.Continuousfunctions,p.5l.3.5.Exarnples
of continuous and discontinuousfunctions. D. 53. 3,6, Tbe intcrmediate-valu
property, p. J6. 3.7. Ilounds of a continuous function, p. 57. 3.8. Uniform
continuity, p. 60. 3.9. Inversofunctions, p. 62.
5 . I N F r N r r BS E R rE s
5.1. Serieso[ positive ternrs, p. 88. 5.2. Seriesof positive and negativeterms,
p. 90. 5.3. Condjtional convergencc,p. 92. 5.4. Seriesof complex terms, p. 94.
5.5. Power series,p. 96. 5.6. The circle of convergenceof a power series,p. 97.
5.7. Mutliplicationof serjes,p. 99. 5.8. Taylor'sserics,p. l0l.
6 . T H E S p E c r A LF u N c r r o N s o F A N A L y s r s
6.1.Thespccialfunctionsofanalysis,p.104,6.2.Theexponentialfunction,p.105.
6.3.Repeatcdlimits,p, 105. 6.4.Rateof increase ofcxp.r,p. 106. 6.J.exprasa
power, p. 107. 6,6, The logarithmic function, p. ll0, 6.7. Trigonometric func-
tions,p. | | 2. 6.8. Exponentialand trigonometricfunctious,Jr.I | 3. 6.9, The
inversc trigoflometric fuoctions, p. 116. 6.10, The hyperbolic functions and
thcir invcrses, p. I17.
vi CONTENTS
8. FuNcrroNs oF SEVERAL
VAnrAsr,ss
8.1,Functionsof .r and/, p. l Sl. S.2.Limitsandcontinuity,p. 152. 8.3.Partial
p. l5l. 8.4.Ditrerentiability,
differentiation, p. 156. 8.5,Composilefunctions,
p. 158. 8.6.Changps of variable.Homogeneous p. 159. 8.7.Taylor's
functions,
theorem,p.163, 8.8.MaximaaDdminima,p.164. 8,9.ImplicitfuoctioN,p. 165.
8^ - | -2-4'
1.2.
'WeThe scopeof mathematicalanalysls
cannow describemathematicalanalysisasincluding those
topics which dependon the notion of a limit. Thus it includes
the differential and integral calculus,and you may ask whether
a new title is necessary; doesnot /fte calculusadqrately specify
the subjectmatter? In a senseit does,and it is mainly by usage
and tradition that analysishas come to denote a rather more
formal (or more 'advanced') presentation, witl greater atten-
tion to the foundations and more insistence on logical de-
duction. The use of the word analysis has the advantageof
clearly including the summation of infinite series (which the
schoolboy would reasonably regard as algebra rather than
calculus).
Operations which are complete in a finite number of steps,
such as the evaluation of a determinant, belong to algebra,not
to analysis. The binomial theorem is a theorem of algebra
if the index is a positive integer; othefwise it belongs to
analysis.
Geometryis a subject separatefrom analysis,developedfrom
its own axioms. Its only impact on analysisis that we shall often
find it suggestiveand helpful to use geometricallanguageand
illustrations.
In the light of what we have said the subjecttrigonometry is
seento fall into two parts. The solution of triangles,'height and
r.21 NUMBERS J
1.3, Numbers
We haveseenthat the logicalorderof development
of mathe-
matical analysisis
Number
Funltion
Lilnit
1,2,3,...,
I (c)
Exerclses
Notes on theseexercisesare gioen on p. 17O.
The method of induction may b used for 1-6,
2 . 1 r + 3 r + 5 t . . .+ ( 2 4 - 1 ) r = ?
3. 1.|+3.2+5-2'+...+(2r+l\2 = A+(B+Cn)2, wh.e,{,,and C
are coostants (not dependiog on z) to be found.
.exr-'ll =
a'
lr 1- ** 1-r-6;'
5.2>rfifn>9.
6. 5r'-62+8 is divisible by 9.
7. lf b, d, q ate positive integers and
oDc
E' a' a,
prove that positive integsrs m, n can be found such that
2 _ m4+ nc
q mb+nd'
Construct a numerical example and solve it.
8, The density property of the rationals amounts to sayiDgthat there is
no rational which is r"rt to another. Obssrve the following plan of
arranging the positive rationals (not io order of magnitude) which does
assign a deonite place to each
*; ?,l; l, i,l; t,*,1,* t, ....
Prcre that plq occupiesthe {t(p+q-l) (p+q- 2) +4}th place. (Each
rational occursin-finitelyoften; e.g. I appearsas +, i,3, ,..,)
r.4l NUMBERS
I (r)
Exercises
Noies on theseexercisesarc gloen on p. 170
1. Adapt the argument of tlrcorm 1,4 to shon, that no rational number
has its cube equal to 16.
2. Extend the result ol I to show that a rational number r/4 in its lowest
terms can be the cube of a rational number only if? and 4 are cubes of
integcrs.
3. Provethe morc generaltheorcm(Gauss,177?-1855)that,ifA,p*.,.,p,
are integers, the only possible ralional roots of the equation
x"+ prxn-r+ prx^-t* ...4 p" = Q
are integers which divid p..
4, Solve the e4uations:
r{-xl-l6x'+4x+48 = O,
4d-8:f -3.r+9 = 0.
Ererclses 1 (c)
Notes on lhese exercites are gioen on p. 770.
1. Prove that, if mln is an approximation to J2 from below, then
(m+2n)l(n+n') is a closer approximation from above. Hence write do$,n
approximations to./2, obtaining two which differ by lecs than l/10,000.
2. Find similarly approximations to J3,
3, Prove that, if a,4 c, dare rational and
a+^lb = c+^/d,
then either (i) a = c, b = /, or (ii) b and d zre both squares of rational
numbrs.
4, Prove that, if a,4 c are rational and
c.l3= O,
s+ b^12+
then a=b=c=O.
5. lf a, b, c are rational and
= O,
a+b42+c44
what conclusioncan you draw?
6. lf a, b, c, d are rational and.x is irrationat, in what circumstances is
dx+ b
cx+d
rational ?
ExerclsesI (d)
Notes on theseexercisesate given on pp. l'l0-1.
Note, The relationsof inequality in exercises5-8 are repeatedlyuseful in
analysisand it paysto be familiar with them.They may alreadybe known
to tho readeras resultsin algebra. Proofs are given on pp, 170-1.
1. If a and b are numbers.and
a<b+:
1.9. Complexnumbers
The system of real numbers is comprehensiveenough to
carry very many theoremsof mathematicalanalysis, It may be
asked what we should miss by not admitting at any stage
numbers other than real numbers, In answeringthis question
it is customary to point out that some quadratic equations,of
which the simplest
- is
xu*l:O
have no roots in the systemof real numbers.This is a depri-
vation rather than a disaster. It is indeed satisfying that the
introduction of complex numbers enables us to prove the
theorern tiat every algebraicequation has a root. But an even
more cognt casefor admitting complex numbersrestson tleir
bringing to light close connectionsbetweensome of the most
common functions of analysis,the exponentialfunction on the
one hand and the trigonometric-sine and cosine-on the other.
If the variables are real thesefunctions are for ever unrelated.
The case for including complex numbers in ordinary analysis
is finally won by the beauty and generality of someof the later
theorems which are based on them (beyond the scopeof this
course),
We now sketch a method of introducing complex numbers
into analysis. We can extend the field of real numbers by
adjoining one or more new elementswhich are combined with
the original membersofthe field by the operations + and x in
accordancewith the axioms A. The new element which must
be adjoined to yield complex numbersis r, wherei by definition
satisfie$i!+ I : 0. The numbersc+Dr, wherea and b arc rcal,
which are elementsof the extendedfield are added and multi-
plied in accordance with the algebraicaxioms A l-1 1. The
number a+0i behaves in every respect like the real
number a.
No axioms of the type O (order) hold in the extendedfield.
It is not possibleto arrangecomplexnumbersin an order of
magnitude in the way that real numberscan be so arranged.
1.9t NUMBERS 19
Exercisesl (e)
Noteson lheseexercisesare giaenon p. l'11,
l. Provefrcm tlre axiomsthat, if a, r, c, d are real and.a * bi = c + di, then
a= candb -- d.
2. The inverse(A10) of a+bi is (a- bi)lb2+ b2) provided that a and ,
are not both 0.
3, Provethat, if the product of two complexnumbersis zero, at leastone
of them is zero.
4, From the addition formulae for the cosineand sine deducethat
(cosd+ i sina)(cosd+isin/) = cos(d+d) + t sin(A+d).
5. By induction or otherwiseprove de Moivre's theorem
(cosr+l sinr)' = cosn0+ i sinn0.
6. Extend de Moivre's theorem taking the index to be (i) a negative
integer,(ii) a rational plq.
ExerclsesI (/)
Notet on lhese exercisesare gh.tenon p, l7L
l. Taking a simplevalue of z (suchas 2+ i), mark in a diagramth poiltts
representingz+ l, 22,iz. llz, i, z+ z, zr.
2
SEQUENCES
2.1. Sequences
In chapters2-4 we shall operatewith real numbersonly,
Complexnumberswill be neededin chapter5. A sequence is a
in
set of numbersoccurring order, that is to say,thereis a first
number, a secondnumber and so on. If the sequenceis un-
ending,or, in other words, if, whateverpositiveintegern be
assigned, there is a correspondingrth nunber, we have an
infnite sequence.In simplecasesa sequenceis definedby an
explicitformulagivingthe nth numberin termsof n. The nth
numberof a sequence is convenientlydenotedby s" (or t,. or
tl., etc.),
Illustrations.
(l) r" = 1/a. The sequeoceis 1, 1, J, ....
(2)s" = (-r)l^!n. (3) s" = zr.
meaning lim so : 3.
(4) There is a further symbolismwhich savesmuch writing
and which you may adopt rvhen (but not before) you have
masteredits meaning.The above definition may be written
r"+rif
e>0; 3,ly'.ls,-.rl <e for alln> N.
In this notation,whateveris givenis written beforethe semi-
colon. Here'given6 greaterthan0'.
The symbol 3 (reversedE) taken togetherwith the next
followingstop means'there is (or thereexists)...suchthat'.
Here 'there existsly' suchthat ls,,- sl < e for all n > N.'
In this book symbolismof this kind will be usedfrom time to
time, but not on everypossibleoccasion.It is the experience
of many studentsof mathematicsthat argumentsare easierto
follow if brevity is not made the prirne considerationand if
symbolicstatementsare relievedby verbalsentences.
SEQUBNCES 12.3
Exerciscs 2 (a)
Notes on lhese erercises are giuen on p, 172.
l. Define a sequenceJ" satisfying the following requiremeots:
(a) 0 < s" < I for all n,
(6) there is no z for which s" = {,
(c) sn -) l as a -+ co.
For each of the sequencesdefined in ?-6, state whether or not it tends
to a limit. If a sequencehas a limit, make an (e, N) tablc as in $2.2,takiog
= l0-3 and aoy other values that you like.
^3n
t, Si= -.
t?t\s
3. o" = l-nl .
\,.F J,'
2 (r)
Exercises
Notes on theseexercites are gioen on p. 172,
For each sequences" defined in l-6, state whether it tends to a limit
(finite or infinite) or oscillates.
l. 100/r-r+(- I)", l0O+(-l)"x-r, 100+(-l)"r.
2. a+6(-l)", where a and, are consbnts.
3. r'{l + (- 1)"},n' + (- l)" n, sn2+ b(- l)" n.
4. The rmainderwhen r is divided bv 3.
5 .* + + + i + . +
. .G r .
6 . ( l + 2 + 3 + . . . + r ) l n L ,{ l - 2 + 3 - 4 + . . . - ( - l ) n l l n .
7. Give a value of N such that, if n > N, tf -4n > lU.
Establish the truth or falsity of the statemots in each of &10. This
meansthat ifa statementis true you hayeto prove it; ifit is false,construct
a counter-exemple,i.e, an example satisfying the hypolhesis but not the
conclusion,
8. If Jo+r- rn oscillates finitely, then Jn oscillates.
9, If .r"*1-.r, oscillatesinfinitely, then sn oscillatesinfinitely,
10. If, given .( (however large), we can find 1Vfor rvhich rd > K, tben
is a null sequence.
We can choose.ly'suchthat, for all n > N,
lr I > +ltl
I 2
and so I; ;I 'Flu'
Then t - t,, is a null sequence
and 1/t" t is a boundedsequence
and theorem2.52showsthat (t-tn)lt,Lt is a null sequence. I
Exercises2 (c)
Notes on theseexercisesare gioen on p. 172.
l Discussthe behaviouras z + co of the sequences rvhoserth terms are
1z-3\t /3x+l\3 n(n-l)
\3r+ l/ ' \z-3l' (n-2)(n-3)Qt-4)'
2. Discussthe behaviouras z + co of the seneralrational function of n
anne+afle-r+...+ae
R(tt) =
bono+ brno-r+ .,.+ be
where p and q are positive integers.
2.6. Increasingsequences
Defnition. If sn*, 2 s,for all ualuesof n we call s, increasing.
11is usefulto regardincreasein the wide sense,allowingthe
possibilityof equalityat any of the stepsfrom n to n+1. If
Ja+r ) f,n for all n, we call s,, strictly increasing.
If s,,*1( s" for all valuesof n, we call s, decreasing.A word
which usefullycoverseither increasingor decreasingis mono-
touc.
Examples.Whichof the followingsequences
areincreasingor decreasing
?
, '2A.|
(1),,,+1, @:j-, ( 3 ) l , ( 4 )n + ( - 1 r " , ( 5 )2 r + ( - l ) " .
t2 SEQUBNCB3 t2.6
We shall prove that a monotonic sequencehas the very
important property that it must tend to a limit or to +oo or
-co. In other words, a monotonicsequence cannotoscillate.
Theorem2,6. An increasingsequencetends either to a limi! or
to +@.
Proof. Let s* be the sequence.There are two possibilities.
Either
(l) a numberA aanbe found suchthat s* ( A for all n; or
(2) whatever number I is taken, there is a value of N for
which s1y> l.
Let us deal first with the possibility (2). Sinces" is increasing,
so > .l not merely for z : lV but for n >- N, Then, straight
from the definition,s" +oo.
Take now the case(1). The number ,{ is an upper bound of
the.ro. By theorem1.8,thereis a numbers : suprn with the
properties
s- < s for all n
and Jn > J- for someparticularvalueofn.
Sinces,,is increasing,the secondinequality will hold also for all
valuesof z beyondthat particular value,The two inequalities
show that so+ s. The theorem is proved.
The most usefulcaseof the theoremmay be summedup in
the statement
A boundedincreasingsequence,ends,o a limit.
The correspondingtheorem that a decreasingsequencetends
to a limit or to -@ can be proved, either by an analogous
argument or by using the fact that if so decreasesthen -s-
increases.
2.7. An important sequencea'
Let s* : &, whete a is a constant. The behaviour of the
sequenceas ,, + co dependson the value of a.
(1) If 4: l,.so = 1 for all z and lims,,: 1. If c=0,
lims": L
(2) Suppose a > l. Let a: l+k, wherek > 0.
Then s'" : (l +&)" > l+nk
(by taking only the frst two terms of the binomial expansion).
2.71 sEeuENcEs 33
Asn+oo, l+nk-->a andtherefore J,->co.
(3) Suppose
0 < a < l. Leta-t: l+/, where/ > 0.
rhen ocs":d-rr.-1,.1r,
As z + co, 1l(l + nl) -., O and therefores,.+ 0.
(4) Suppose a to be negative.If -l < c < 0and a: -b,
so that 0 < b < l, it follows from (3) that ,4 + 0 and hence
s,: (-b)a+ 0.
lf a:-1, r, takesthe values-l and 1 alternatelyand
oscillatesfinitely.
Il a < -1 and a:-b, b > l, then from (2) we have
D" +co. So s" : (-b) takesvalues,alternatelynegativeand
positive,numericallygreaterthan any assignednumber.That
is to say,ra oscillatesinfinitely. Summingup, we have
a" ->a (a > 1),
a"->l (a: l),
a" "->0 (-l < a < 1),
a" oscillatesfinitely (a : -l),
a" oscillatesinfinitely (a < -l).
Exercises2 (d)
Notes on theseexercisesarc gioei on p. 172.
1. If 11> 0 and .ro*1> -l(rn, where f > l, for all values of '|, thcn
Jn -) + CO.
2. If, for alt valuesof n, l.r"*11< Kls,l, where0 < ,( < l, thenrn -+0.
The conclusionremainstrue if the hypothesisis satisfiedonly for n > N.
3. If t**= t, -r<t<r,
provo that Jn -+ 0.
4, Discuss the behayiour, as n -+ @, of the sequencea"lnb, where k is
a positive integer.
5, Prove that, if a > 0, then {a -+ 1 as z -+ co.
6, Prove that {l + (l lrr')}"increasesas z increasesand that it tends to a
limit. Clhis limit is the very important number e-see chapter 6.)
34 sEeu8NcBtt I2'1
as ,t irsrsss,and dcducothat it ton& to thc
7. Provcthat {h decreares
limit l.
8. Give examplesof soqu,ncesr" for which
tm,n+t _ 1
Ji
""@)=#
tcndsto a limit as ,| -+ oo. Showin a diagramthe graph of
y = ltun4"(x).
11. Provothat,if -l <t< I,
a" - m(m-r).:!m-n+l)* = H *
tendsto zcro asn -+ .o.
bevclimits 8s ,t -t ooond'
12. Investigatewhetherthe following sequnces
if so, give the limits,
(i) {3"+2'), and cxtendto numbrsothcr than 3,2;
,n _ ,ri
(ii)
ffi ,wherea > 0 and6 > 0;
4nn
\ . . . , , r f+ l ' n r + z ' ' . . ' nt+I,
(iv) Y(n!);
(v) n-^l@+o,(n+ bJt
(vD 2f ln!.
2.8. Recurrencerelatlons
Hitherto we have supposeds,, to be given by an explicit
formula in terms of z, In practice a sequenceis often deter-
mined by a relation connectingtwo or more su@essive members
'solve' this for s".
of it, and it may or may not be possibleto
We illustrate by examplesof useful types.
2.81 sEeuENcEs 35
Example| . Thelinearrecurrence
relation(ordffirenceequation)
with conslantcoefficients.Supposethat we are giventhat
sn: sn-a+sn-z (n > 2)'
to: l, sl : 1.
2 (c)
Exerclses
Notes on these exercif,esarc gbm on p. 173.
Invstigatc the behaviour as z + @ of r" if .r, is givon by the recurrenco
relatioN stated in 1-6.
1. 3.r" = 2J"-t-1so-r (16= 7,3. = 3;.
2. f,o+r = 2/(I +r") (rr = 0),
2,81 sEeuENcBs 37
3, .r"*t = sl.17s,where0 < /r < 1 and rr lies betwenthe roots of th
equation xr-x+& = 0.
2(t"11)
4. ,-,, = 1",= 3y.
<" -1
5. s,+r = L* (sr = 2).
f
Consideralsoothervaluesof s1(e.g.s1= +).Whathappens
if rr = +?
6. s" = (sl-. * s"-, + 2)1/3 (s, = l,sr = *),
Fis. I
7. Prove that, if ^
ttt = 3 and zn*, = (4"+5)1/r,
then ll, tends to a limit /, and givc / to two decimal places.
ProYethat
0 < a1n*1-/ < 30-"(a.-/).
Note. In numericalapplicationsofanalysis,it is vital to know with what
rapidity we can rely on a sequenceapproaching its limit. This information
is giYenby the last inequality.
8. Discussthe possiblelimits of the ssquercedefinedby
t", = sl+
H. 1l '
ProYethat, if ro > 3, then
(i) 3 < s"*1 < t", (ii) f,n+r-3 ( r? (""-3).
9. Prove that. if
ct) bt) 0 and a,*1 = t{a"+b"), b"a = za"bJ@"+b"),
then a. > a,+t ) bn+r > ,a. Prove that, 39 7 -+ co,cn and ," both tetrd
to the limit .,/(a!rJ.
SEQUBNCES 12,9
t+*+...+*,
and nr-r + oo.
"(*)
Numerical illustration. Find a value of r large enough to
make the sum of z terms of
greaterthan 20. I +++++...
The reader should not disdain numerical work, which keeps
him in touch with reality. We will work this examplefor him,
The proof showedthat the sum of the fust 2- terms is greater
than I + +/rr. So 238 terms will certainly be enough. Now
238> 1011and a twelve-figure number in this context is sur-
prisingly large, We naturally sacrificed accuracy in approxi-
mating to the sums of blocks of terms. But more refined
methods would show that the number of terms which must be
taken to grve a sum greater than 20 exceedslS. Thesecalcu-
lations show, in homely language, that the diuergenceof this
seriesis slow.
2 (/)
Exercises
Not.t on theseexercisetare givenon p. 773.
l. For the seriestl/zr estimate a value of ff which will ensurethat the sum
of all terms after the Nth is less than 10r,
Mak the same calculation for the geometric series:(0.99)i. What do
you conclude about the relative 'slowness-or quickncss-of convergenc'
of the two series?
2. Find for what values of r the following seriesconverge, and sum them
rt
' '
l+r'(l+f)' (l+fF "''
rl rl .,
'r l - r' - -J- '
l + rr (l +rt)r (l + rr)t "''
1+2r+3rt+4rt+,.,.
2,lll sEQUENcrs 43
thesum
3. Estimate
,!Jr"
4. Provethat, if a> 0and, > 0, the serieswhosenth t.'fmis ll(a+nb,
diverges.
5. Sum to m tertnsand to infinity the serieswhosezth term is
I
n<*Tl-<*zl'
Deducethe convelgenceof Xr-3. Extend this to a denominatorwhich is
a product of & insteadof three consecutiveintegers.
there existsK suchlhat tu* < Kfor all N; and then lu" < K.
lt
This is a restatementof theorem 2.6.
(6) If, for eoeryn,
Exerciss2 (g)
Notet on theseexercisesare giuen on pp, 1734.
l. (N,8.) Explain necessarycondition, t llcient condition. Which of the
following are (a) necessary,(r) sufficient, for the real numbrs p,q to be
equal ?
(i) p'+q' = zpq, (ii) pz = q!, (iii) p+!=q+!-.
'q'p
l - l ++-'++-l+*-*+...,
,r .+1t ,! 1+ t f f l + .l ., . .
then the condition,4 > I is necessary for the convrgence of the scquence
rn; show that it is also sufrcient in the case of ro > l, by verifying that
f. > I for every n, and, for a suitable c > l,
c" lr"-cl < lro-cl.
6 sBeuENcBs 12.12
U. Define(if possiblQan oscillatingre4uenccsatisfying
&t! -.1'1.
J,
also tends to r.
15. Find the sum of the first 4+ I tmrs of the sries
I +2x+ 3.d+...+(z+ l).xi+....
Prove that, if l.rl < l, the seriesconvergesto the sum l/(l -.t)t.
3
CONTINUOUS FUNCTIONS
3.1. Functions
of one real numbery
You are familiar with the dependence
on another real number x which is commonlydeterminedby
formulaesuchas
.y = 1/("r'+l) or y:,,!{(2-x)(x-t)}.
In the former y is definedfor all valuesof x and in the latter
only for 1 < x < 2. In eithercase,if a value of x is assigned,
we cancalculatethe valueofy whichcorresponds to it and there
is only one such value of y (if it is understoodthat the ./ is
positive). Here r may be describedas the independent uariable
and ), as the dependentuariable.
Thesetwo examplesare simpleillustrationsof functions.We
go on to expressin generaltermsthe meaningof function.
Let X be a set of numbersx and Y a set of numbersy. If
rules are given by which, to eachx in X, a corresponding7
in I is assigned,theserules determinea function definedfor
xin X,
Noler. (1) Obseryethat, to a givenx, there is just onc corresponding),
The same/ may correspondto more than one x, Fo! instance,in thg
example/ = 1/(r'+l), y = ! corresponds to both.r = I and x = -1.
In other words, the correspbndence betweenX and yis in generalmany-
one and is only exceptionallyone-one.
In the ordinary graphicalrepresentation,no line parallel to O/ cuts the
cuNe in more than one point.
(2) You may come acrossin mathematicalliterature other words and
phmses equivalent to those used here, A function defined in X with
valuesin I is often called a transformation,or a mapping, of X into y.
According to the definition, y may include valueswhich are not taken by
/ for any value of x, If every value of I' in y is taken for some ,. in X,
we can say that the mapping of X is onto Y.
(3) The usual notatjon for a function is the letterl If more lettersare
needed,g, C, Fare commonly used,The value ), which a function/takes
for a particular x in X is written/(r), so that / = /(r).
,t8 coNTrNuous ruNcrroNs t3.1
(4) Strictly speaking, the function /is the set of all pairs of numbers
(r, y) which are related by the rules defining the conespondence. /(x) is
the 'function-value' for the number -r. It is convenint, and does not lead
to error, to spak sirnply of the function f(*).
A few further illustratiom will bring out the full meaning of the
definition of function. In each of thesg satisfy yourself that the rules are
adequate to determine a uuique value of/ for each value ofr concemed.
(D Y=r for o<.r<1,
./=0 for r<0 andfor r> 1.
In physical applications the independent variable is commonly the timc;
this function could represent,for example, a force of unit magnitude which
&cts for unit time and then ceases.
(2) Suppose that rt = t this equation assigns values to / for t > 0,
and, if .x > 0, there are two values of/ equal in magnitude and of opposite
signs. We said in tbe definition that a function is to be sizgle-aalzed. So
y' = x docs not define in our snse),as a function of .r. We may, however,
di$ect it into two functioDs f = +^lx and y = -^!x each defined for
x>o.
(3) y = 7 +- .
r^ t) (.r-2)'
This function is defined for all values of x except .x = I and x = 2.
(4) r = the large'stprime factor oft
This statement has meaning only when .r is ao integer.
(5, s" = lln, where n is a positive integer. Sequencesare functions of
a special t'?e, in which the indepndent variable is restricted to be a
positivc integer. The same defining formula with r replaced by r, namely,
y = !lx, may defrnea function for values of r otber than positive intgers
(here, for all values of x except r = 0).
(O .Y = 0 when x is irrational,
Y= 1 whe! x fu ntional.
An 'utrnatural' function like this may seemto have little interest. Such
functions are, in fact, of use in analysis in deciding for how wide a class
of functions some proposd theorem is true. We sball come across iuus-
tratious of this later in the book.
(7) / is fhe tmperature in degresat time r at a givNrplace.
This type of function is very common in scisnce and in everyday Uf.
It differs from the exampls(lXO in that there is no analytical formula by
wbich it can be represcnted.In practice the function-valuec may be given by
a graph (drawn, say, by a pen attachd to a recording thermometer). These
values are then knowu withiq thc limits of accuracy which can b9 attained
io obsrvatioD.
3.21 CONTINUOUS FUNCTIONS 49
\
\ \/!
Fis,,2
(x- 1)(x-3)
Examptet. ), =
Gc_ilix_6,
(1) y is defncd for all r xcpt2 and 4.
(2) As.r+o or r+ -@, y+1. Draw the horizontalasymptotc
=
.y |. Also it is clear that, if x is large and positivg .r- I > x-2 and
x-3 > x-4, so the curve approschesthe asymptotefrom above as
x-+@.
(3) x = 4 and .r = 2 aro vcrtical arymptotcs. From the signs of the
factors in the numcrator and denominator,
8a t-t4+, /++@ a n da s x - - 4 - , y - + - q i
as x-+2+,.r++@ a n da s x * 2 - , y-+ -q,
(4) y=0 gvcs .r= I or 3,)
Mark these Poin$'
x = o givs .y = *. l
It is often worth while to find where the cuwe cuts its horizontal
ssymptotc, Hcre y = I gives x = t (and, what is more importalt, no
othcr point).
We have by now a good idea of the shape of the cuwo (69. 2a).
3,31 coNTrNuousFUNcrroNs 51
Example 2, y = x "' +. 7t ,
(l) Sincey is unchangedwhen -x is put io place of r, the curve is
symmetricalabout Oy. y is definedand is positivefor all x except0.
(2) As .r -+ + co or Jc+ - @, y -+ c.. Moreover, when x is large,/ is
a little greatcr than x2, so we sketch the parabola/ = xr as a guide for
largex.
( 3 ) A s . r - +0 + o r x ' + 0 - , J . , + c o .
(4) The curve does not cut either axis. When r= tl, /= 2, (see
frs.2b).
Exercises3 (o)
Sketchthe generalshapesof the curvesgiven by the following equatioos.
l.Y=x*. 2. y = 76-to. 3. J, = xro+x-ro.
4. y =:. l. -vx =
x'
-----
ylf I
o Y = (i-1) (x
4,
8. yr = -i .. 9. y1 = -i-= .
x+ I
3.4. Continuousfunctions
The readerwill haveacquiredfrom examplesthe impression
that the commonfunctionscan reasonablybe calledcontinuous,
thoughsomeof them may presentdiscontinuitiesfor particular
valuesof x, For instance,he would regardthe function whose
graph is sketchedin exampleI of $3.3 as being continuous
exceptat x : 2 and x -- 4. He would think of a function as
continuousso long as its graph can be drawn without taking
the pencil off the paper.
We must now refine theserough ideasinto analyticalcon-
cepts.The reader,after reflection,will (we hope)agreethat for
a function/to be continuous at a valuex - c he requiresthat
(l) /(c) is defined,(2) as x approachesc, the value of /(x)
approaches /(c). Thus the assertionof continuity is nothing
nore or lessthan a statementabout limits, namelythat
lim /(x) : lim f(x) : f(c).
2-+')+ ri+C-
52 coNTrtrdous FUNcrroNs ?,41
We neednot keepthe c+ and c- separateand can now setup
two basic definitions,the former dealingwith the limit of a
function and the latter with continuity.
Definition, f(x) -->I as x -t c tf, gioene, there exists I suchthat
l , / ( x ) - r l. e f o r a l l x r z 0 < l . x - c l < d .
The d dependson the e, and, in general,the smallerthevalue
of e, the smallermust be the value of d. The dependence may
be emphasised if we wish, by writing d(e).
Note that the value .r : c is excludedfrom the set of .r for
which the e-inequalityis required to hold in the above de-
finition. The statementconcernsbehaviouras x getsnearto c,
not behaviourat c.
Defnition. Thefunctionfis continuousat c y'/(r) -->f(c)asx -> c.
Combiningthe last two definitionswe seethat the following
is an equivalentdefinitionof continuity,
f is continuousat c if, gh.tene, there is 3 suchthat
lf(x)-.f(c)l<e for lx-cl <6.
The valuesofx such that l.r-cl < d, in other words, the
open interval (c - 4 c + d) may be aptly alled a neighbourhood
of c. More generallywe canapply this word to an openinterval
c-8, < x < c+d2, wheredr, d, may be different. It is worth
while to statethis as a definition.
Defirtdoa. An open interaal is called a neighbourhoodof any
one of its points.
The definition of continuity of f rt c then statesthat, given
any ueighbourhoodN, of/(c), there is a neighbourhoodN, of c
such that, if x is in 1Vr,then/(x) is in /V,.
We have defned continuity of f at c, that is to say, con-
tinuity at a point. We go on to definecontinuityin an interval.
Supposefirst that tbe interval is closed.
Defirtion. f is said to be continuous in the closed interual
(a, b) if
(1) for eachc in a < c < b,f is continuous
at c;
(2) lim /(.r) - f(a), tim f(,) :.t(b).
.+a+ .+b-
3.41 coNTrNuouspuNcTroNs sir
The purposeof the specialtreatmentof the end-points4 anc
D is clear; we wish to avoid any mentionof valuesofx outside
(a, b). lf we are defining continuity in an open interval say
a < x < b or theinfiniteinterval.r > a. thereareno end-ooints
and no condition (2).
Definition. f is said to be continuousin an open interual if it is
continuousat eachpoint of the interual.
Exercises3 (6)
Nolet oratheseexercisesare gioen on pp, 171-5,
1. Stateany valuesof r for which the following functions are discootinuous
^lKx-a)l(b-x)1, l/J(*r+ 1),
tanr, secr, U(l +tanr),
2. Sketch the graph of the 'saw-tooth function'
x- [r] -].
3. Sketch tbe graphs of the functions
lxl, ty'xl, dtxt,
pointing out for what values of x there are discontinuities.
4. Proye that the function defined by
,f(x) = x sin (t/*) (x * 0),
/(0) = o
is continuow for all values ofr, Sketch its graph.
5. Provo that. if -
xr-6x+5
J(x) =:t;F:';T-i8,
and /< * I, there are two values ofx for which,t"(x) = &, Illustrate by a
graph.
6. Prove that
x,-2x+ |
xry2x+2
is bounded for all values of;r, and find its supremum and infimum.
7. ArBwer the same question for
4x2+3
x'+l'
8. Inwstigate the limit as { + I of
*-3x+2
p 11
7z= z,z t32'
9. Investigate the limits as x -+ 0 of
M -J \X)
which contradictssup/(x) : ff I
Secondproof (by the bisectionmethod), Bisectthe interval
(a, D).The key to the proof is that thereis at leastone of the
two halves in which sup/(.r) = ]l1. Selectthis half (or, if
sup/(x) = M in both halves,select(say) the left-handhalQ.
We thushavean interval,which we letter(ar,br),whereeither
ar = a ot Dr - D, such that
br- a, = \(b - a).
3.9. Inversefunc{iore
You will havemet situations in which, uncritically, a function
has been constructed as the inoerse of an already known
function, For example,the equation x : sin / is usedto define
/ as a function arc sin .x of .r, This illustration shows that the
procedure requires care, because,for a given .r, there are in-
finitely many valuesof1 and if we are to keep to single-valued
funotions we must inpose some restriction on the permitted
valuesof y,
We shall now give al existencetheoremwhich assuresus lha!
if certain simple conditions are fulfilled, we can obtain a new
function inverse to a kf,own function.
Defnirton. /rs increasingfor a < x < b if f(x) < f(x) for
all x1, x2 such that 4 ( rr < xs < b. If f(x) < f(x), we say
that f is s|aic/.lyincreasing.
Tlrcorcm3.9. Let f be continuous@td strictly increasingfor
4 < ;r < b. I'et f(a) : c, f(b) : d. Tlpn there is a ftotction g,
continuous and strictly hcreasing for c 4 y < d, su& that
fie$l = y (so that gO) is thefunction inoerseto f(x)).
Remarks. The situation is easily appreciated graphically.
Moreover, we can seethat the hypothesesare natural. If/were
not strictly increasing (or strictly decreasing)tlere could be
valuesofy correspondingto more ttran one value ofx, and then
there would not be a single-valuedinverse. If / were discon-
tinuous, we might have, for some,f(+'l : tt > I = f(-\
and an interval (,\, p) ofy in which tlere is no inversefunction.
Proof. l*t k be any number such tbat c < k < d.
By theorern 3.6, there is a value fr such that
.f(h) _ k,
and, since;f strictly increases,there is only one such lr corre-
sponding to a given &.
The inversefunction g is then definedby } : g(/<).
It is easy to see that g is strictly increasing. For a formal
proof of this, let !, < yn and y, : /(.r), yz : f@), From tle
last paragraph,x, and x, are uniquely defined. Ifx, g xr, then,
3.91 coNTrNuousFUNcrroNs 63
since/ is increasing,f(x") < f(rr), i.e. yz 3 yy which contra-
dicts the assumption/r < ./2. So x, < x, ar.dg is strictly in-
creasing.
It remainsto prove that g is continuous.
Withe > 0, let
.f(h-u) : k1 and f(h+e) = kr-
Then,since/is increasing,
kl <k<k2
and h-u . g(y) < h+e if kt < ! < kz.
Since e is arbitrary, g is continuous at y = k.
Here k is any number in the open interval (c, {. A similar
argument establishesone-sided continuity at the end-points c
and d.
Exrcises3 (c)
Noles on lhese exetcisesare gil)en on p. 175,
1. Discussthe continuity of the following functions:
(i) f(x) = llq wheo x is a rational numberplq in its lowest terms,
=
,f(x) 0 when x is irrational;
(ii) /(x) = x log sin'zx(x + o), /(0) = 0;
tl
l ll l(-xl = - cosec -_
x-a x-a
2, Constructa function which takeseveryvalue/in 0 < / < I onceand
onceonly for valuesofr in 0 <x < I and which is discontinuousfor some
valuesofxin0<r< l.
3. /is bounded for a ( x < b, and,Mk) is the suprcmumof the values
oflio the closedinterval (a, x). Prove that, iflis continuousfor a value
xo in a < xo < ,, and ifl(xo) < M(ro), then thereis an intervalcontaining
xo in which M(x) is constant,
4. Can a function be continuousfor one value ofx and discontinuousfor
all other values?
5. In theorem 3.82, take f(.x) = xtt3, a = -1, D = l,e = rt and give a
value of d.
6, Which of the following setsof data (i)-(iii) are sumcientto determine
the value/(0) ?
(i) /is continuousat x = 0 and in any neighbourhoodof x = 0 it
takes both positive and negativevalues;
A coNTrNuout PUNcrtoNs 13.9
Gi) girtn c, thcrctu tsuch that l/(r)l < efor0< l.rl < f;
fui) as i -t Q /(i) -+ a aad
flh)+f(-hr-2f(o, - ,
7. /is boundedfor a < x < t atr4 for cvery pair of vsluesxt, .r with
a < x1 < x1< b'
/{}(x1**J} < }{,f(rJ+,f(xr)}.
Provthat /b continuousfor a < r < 6.
8. /is deEacdfor all r and
lf(x)-f(x)l I' alx- x'l
for all r and.r/, wberethe constint c is lcsstbaDl. TbGscqucn@.rr,.a, ...
fu d0ncdby a givenro aod
x, = f(x"-) (z = 1,2,...).
Provothat x" -t g a34 + co,whero
=flil.
ProYalsothat,ifl(o) > q thcn
f(o\ -,- fiol
i;;<Ei';'
65
4
THE DIFFERENTIAL CALCULUS
{f,:-D:f@
h
tendsto a finite limit as lr tends to 0, this limit, denotedby
/'(x), is calledthe deriuatiueor diferentialcofficientof f fot lhe
value of x. The function;f is said to be dffirentiable for that
valueof x.
To statethe geometricalcounterpartof this definition,sup-
posethat P is a givenpoint on the curvey = f(x) and Q is a
variablepoint on the curve. If, as Q approaches P along the
curve,the gradientof the line PQ tendsto a limit, the curvehas
a tangentat P whosegradientis/'(x),
The definitioncallsfor a numberof remarks,
(l) An expressivenotation is to write ft, the changein x,
as d;r (d':ris a singlesymbol; it is not d'muitiplied by x). The
corresponding changein y, namely,/(x + 0x)-/(x) is calleddy.
The derivativeis then the limit, as d'xtendsto 0, of the ratio
d'v
ox
The derivativernaybe written
ay
dx'
wheredldx is, for the present,to be understoodas a symbol
specifyingan operationto be performedon whateverfollows
the d in the upper line. (The dy and dx ^rc not the numerator
and denominatorof a fraction.)
(2) A function, definedfor a set of valuesof r, may be
differentiablefor all thosevaluesor for none or for someand
not others,
6 TEB DITTBRBNTIAL
caLcuLUE 4.tl
Illusrrations.(i) /(r) = rt
f'(t) = 2:c for all valucsofr.
(ii) .f(r) = .t (t ntional),
0 (t irational)'
/(.r) is not differcntiablefor any valueof x,
(iiD l(x) = lxl.
f'(x) = | it x > 0 and/'(.r) = -1 if.r < 0. ,f ig not difreaentiablc
for
r=0.
(3) A necessary(but not sufficient) condition for / to be
differentiable for a given value of .r is that /is continuous for
that value.
For 8y/0x can tend to a finite limit as 8x -+ 0 only if dy -+ 0,
i.e. if/is continuour.
The function lxl is continuousbut not differeotiableat x = 0.
(4) The following trots are, in part, suggested by the
examplesin (2).
In the definition of the derivative we coul4 if wo wished,
consider separatelyl + 0* and i -+ 0-, defining what might
be called the right-hand and left-hand derivatives. If they are
equal there is a derivative in the ordinary sense/'(x). The
function l/ of (iii) above has at x - 0 right-hand derivative I
and left-hend derivative - l.
Considernext another exampleofa power of r. lff(x): xl,
then,forx:q
f@if$: n_+
and this tends to +oo as fr + 0. Geometrically, the curve
./3 : x has a vertical tangent at (O 0). In defining the deriva-
tive we specifiedthat the limit should be fnite, and in accord-
aucewith this definition we say that .xt is not differentiablefor
x:0.
The decisionwhether to admit or excludeinfinite derivatives
is based on convenience.The exceptional casesthat have to
be stated in theorems if inftdte derivatives are allowed sway
the balancein favour of exclusion, For instance,if, for a certain
value of x,/'(r) were +@ and g'(.r) were -o, it can be shonm
by simple examplesthat/+g might have a derivative with any
4.11 THr DTFFERENTTAL
cALcuLUs 67
Exerclses4 (a)
Notes on theseexercisesarc giuen or p. 175.
1. Find the equation of the tangentto the curve / = r!-4 at esch of its
points of intersectioowith (i) the x-axis, (ii) the /-axis.
2. Find the equation of the normal to the curve 2/ = (xa2)r at eachof
the points on it where, = 2.
3. Find the equation of the tangent at the origin to the curye
Y = x'(x-l)2+3x.
Prove that the tangent touches the curve again at a sccond point.
4. Provethat the equation of the tangentto the parabola/r = 44x at the
poirLt(at,,2At) t"
,, =
"*orr.
5. Give an exampleof a function which is continuousfor all valuesof r
and is differentiablefor all valuesof x excaptI and - l.
6. State for what yalues of x the following functions fail to be (a) con-
tinuous, (r) differentiable.
(i) [.t], (definedon p. 54).
(x-l ( x < l ) .'
(ii) /(x) = I
{xr-l (x > 1).
7. Find the anglesof intersectionof the pair of curves
(i) 2Y = f' Y = x2(2- xr,
Also of the pair
' (ri) ZY = x!, 2x = Yt.
5.2
68 rEB DTPFERTNTIALcALcuLUs a4.2
Considerthe fust term in this last sum. Since/ is differ-
entiableat x it is continuousat x andso/(x + ft) +/(x) asI + 0.
The other factot {g(x+h)-e@)}lh has limit g'(x). Sincethe
limit of a productis equalto the productof the limits,
_>f(x) g,(x) as I + 0.
s1*a 1;18(X!_Q:EG)
h
Similarlv'
s@tyl+3\x)_>g(x)f,(x) as ft+0.
Finalln we appeal to the theorem that the limit of the sum of
thesetwo functions of I is the sum of the limits. I
This formula for differentiating a product is typical of those
which are usedover and over again in the differential calculus.
The proof has beenset out fully to stressits dependenceon tle
repeatedapplication of the theoremsabout limits.
(3) tf Q@) : lls@) and g(x) + o, thm
: - "49)-'
q'1*1
{g(x)}''
This fo[ows from
h)
i@+h)-d@): e@)-s(x+
h hs@)
s(x+h)'
(4) rf$@) : f(x)ls(x),then
. .!pl@
6,61:t@s@_t.
{g(x)}n
combine (2) and (3).
(5) Functionof a fwrction.
1*1 u - g(x) and y : f(u), so that
y -.f{slx)} = 0@),say.
rhen Q'@): f'{s(x)} s'(x),
o\ in other notation,
dy dy du
7;: TITY'
Proof. Let 8x be a changein r, 0z the correspondingohange
in z by the functional relation u : S@). Let 8/ be the change
inyfromT:/(u),Then
a4 : {g(x) + e} dr,
1.21 T H B D T F F E R E N T T A Lc A L c u L U s 69
Exercises4 (6)
Noles on theseexercisesare giten on p. 176.
t. Differentiate
. . . . . J'.-( 2 . r + 3 ) - J ( x + 6:,)
(Ii) :___::'_:_: as x r 3.
{(x+ rr-z
3, Proveby induction that the derivativeofx" is rx"-l, r being a positive
lnregr.
4. lf yr,yr,,..,/. are functionsof r, provethat, if y = yry",.-y",lhen
Itly_ldy,,
-, yI d"yE" '
,Ai- iE-"
wherex has a value for which y is not zero.
5. Prove that, if a polynomial p(x) is divisible by (x-a)t, then p'(x) is
divisibleby r-a.
6. Prove that, if p'(.r) is divisible by (x-a;'-t and p(x) is divisible by
.r-a, thenp(x) is divisibleby (x-a)".
7. Show how 6 is of use in searchingfor multiple roots of equations.
Illustrate by solving the equatiol
xo-3x'+2 = 0.
8. Prove that the equation
llt^
x-a x-o x-c
can have a prir of equal roots only if a = b = c,
9. Prove that, if p(x) is a polynomial, then betiveenany two roots of
/(x) = 0lies a root ofp'(x) = 0.
(This is Rolle's theorem for polynomials. It will be proved for mor
generalfunctions in S4.5,)
fxz sin (l/x) r + 0,
10. If f(x\ = I
t 0 x=0,
prove that /'(x) cxistsfor all v;rluesof x and give the valuesof /'(x) for
,i + 0 aod ofl'(o). Prove that/'(x) is discontinuousat n = 0.
72 THB DTFTIRBNTIAL cALcuLUs 14.3
11. Tbe elemcnts of a determioant of order z are functions of x. Prove
that iB derivative is the sum of the r determinants formed by differentiating
the elementsof one row only leaving the other rows unaltered.
12, Prove that, ifr, 4, r are polynomials in .r of degreenot greater than 3,
then
p.lf
y'q'r'
p'{/
is a po[ynomial of dcgree not geater than
4.4. Repeateildifferentiation
Let y - f(x) be differentiable.The derivative/'(.x) may itself
be differentiable. Notations for this secondderivative of/(x)
arc fryldf or f'(x). Notice that the existencof the second
derivative implies the continuity of the first. The zth derivative
may be writtenla)(x). No new principlesare involved.
There is a theorem of interest on ttre zth derivative of a
product.
Lcibniz's ,heorcm. If f and S are functioru of x haoing n-th
deriuatiues,then
: f6rg+ iqf@-l,9' + ...+ nctfb'-ng:f,+ ... +j!'gtfui.
U:g)(,t,
Thc proof, by induction, is left to the reader. The algebraic
lernma
*1L, : nC,+ nC,-,
rs requrreo.
4 (c)
Exerclss
Notes on these cxercisesare gioen on p, 176,
In calculating an rth derivative it is necessaryto adopt the Eethod which
will give the result in the most compact form.
l, Find the nth derivativs of
x-l x+l
vi=4' G='f.
2, Show that the value of
W) ir=-l
for r = 0 is 0 if r is cven and -nlif z is odd and grcatcr than L
3. Prove that the nth dcrivntive of sin x is sin (.r+ lnz) and invcstig&t
corrcsponding results for cos.r, sin &r and ". sh b.r.
4.41 TrlE DIFTERENTTALcALcuLUs 73
4. Find the rth derivative of sir 3r sin 5x. (Do not use Leibniz's
theorem!)
5. Find the dh derivative of al(ar - xt).
6. Find th nth derivativeof a/(a!+ x!).
Ercrdr.4 (d)
Noteson thescexerclsctarc glucnonp, 176.
l. Find a valuefor c in Rollds thcoremwhon/(r) is
(i) (.r-a)" (.r- 6). (rn and n positiveintegers);
(ii) sin(rlx) in (llm,llmr).
2. Iffi.r) is a polynomial,provc that thcrc is a root of
P'(J)+ l@(x>= o
bctwccnany two rcal roots ofr(r) = 0.
3. If a and D arc successivo roo& of ,(x) = O then thc numbcf of roots
b e t w c c n a a n d--6 o f
P'(x)+k^x) = O
(eachcountcdaccordingto its multiplicity) is odd.
4. Provc that thc quation
t d\, , .
[6J (f-r)'= o
has z rcal roots, all diferent snd lying bctwen - I and l.
5. What information docs a kaowlcdgc of the rcots of ,.(, = 0 givo
about thc roots ofp(r) = 0?
4.51 T H E D T F F E R E N T T AcLA L c u L U s 75
Prove that the equation
Yl Yl
t-x+i-j+...+(-r)'; =0
has one real root if z is odd and no real root if a is even.
6. Prove that the equation
x"+kx+l = O
hasat most two real roots when ', is evenand at most threwhen n is odd.
c(b): Q@\
This gives -f(b)-f(a) : k(b-a).
Rolle'stheoremgivesc such thal $'(c) : 0, i.e.f,(c) - k. 1
It is ofteDconvenientto write the result,with b : a+h,
f(a + h) = f(a) + hJ,@+ 0h';,
whered is somenumbersuchthat 0 < d < l.
The meanvalue theoremhas important consequences
under
further assumptionsabout/'(:r).
Corollary 1. If f'(x) = 0for all x in a < x < b, thenf(x) is
constantfor a < x < D.
Proof. If x, and x, are any two points with
c<xr<xr<0,
then .f(x)--f(x,): (x"-x)f,(xs),
where xt < xB < xz,
and this is 0. l
Corollary 2. fff'(x) > Ifor a < x < b, thenf(x) is a strictly
increasingfunction in the interual a < x < D.
76 THE DITTBRENTIAL cll,cul.us t4.6
Proof. Wehave to prove that, if xt and x, are any points with
d ( Xr ( x, < 6, then/(.xJ < /(xJ.
f(x)-.f(*r): (x,-x)/(x.) (x1 < rs < nJ
>0. I
It is often useful to be able to apply corollary 2 under slightly
more general hypotheses.We can allow one value c at which
the only requirementis the continuity off eventhe existence
of/'(c) is not assumed.
To prove that /(x) is still strictly increasingin (a, D), apply
corollary 2 to the two intervals (a, c) and (c, b't. If .q is in (a, c)
and x, in (c, b), then/(x) < f(c) <.f(x). I
The extendcd statementholds for any finite number of ex-
ceptionalpoints c.
Corollary2 shouldbe comparedcarefullywith theorem4.51.
It assumesmore and provesmore. If, as in theorem4.51,we
know only that / is increasingat a point, there need be no
interval throughout which / is an increasing function (see
exercise4(fl, l3).
The following extension of the mean value theorem to two
functions will be useful later.
Theorem4.62. (Caucl1y'smean ualue theorenr). Supposethat
both thefunctionsf andg are continuousin the closedinterualand
diferentiable in the open interoal (a, b). Supposethat g'(x) ir
rliferent from 0 for all x in a < x < b. Then,for somec with
a<c<b,
f(b) -f(o) f'(c\
s(b)-s@) c'G)'
Proof. (Obsewe first that it is not sufficient to apply the
mean value theorem toif and g separately,becausewe get two
different c's.)
Defrne {(x) : f(x) .kg(x)
andchoosetheconstant& to make/(6) : C@),i.e.
f(t) -f(a) : k{g(b)- s@)}.
By Rolle'stheorern,thereis c suchthat
Q'k) -f'(c)-kg(c) :0.
4.61 T r I E T T T T T E R E N T I A cL A L c u L U s 77
Exercises4 (e)
Notes on theseexercisesare gioen on pp. 176-7.
l. In each of (i)-(iii) find, if possible,a number c satisfyingthe mean
yalue theorem, In any examplein which no c can be found, which of the
conditionsof the theorcm is not satisfied?
(i) /(x) = x(x-2) (x-4) (a = r, b = 3);
(ii) f(x) = llxz (a = -t, b = 3);
(iii) f(x) = xtt. (a = -1, b = l).
2. Taking /(r) to be (i) xt, (ii) x", prove that, as i -+ 0, the number d of
the mean valuo theoremtends to a limit.
3. lf f '(x) -+ I as x + cu,prcve that f(x)lx + L
4. Discussthe following argument.
Let f'(x) exist for a < x < b. l.f,t a < c < ,, Then, if q < c+ h < b,
the mean value theorgm gives
tgl_!\!(a =f,(c+oh).
h
let h->O. The left-hand side tends to /'(c). So the limit of f'(c+qh),
as t -+ 0, exists and is equal to /'(c). That is to say,/'(r) is continuous
5. Take in theorem4.62
f(t) = t", eft)= 4ts-3tt (a<t<6)
and find whethera c existswhen (i) a = O,b = l; (ii) a = -1, b = 2;
(iii) a = -1, 6 = 3. Illustrat geometdcally on the curve r = ,r,
y = 413-3t1.
where7 is a Positiveconstant.
Then C(0),d'(0), ..., /tn-tt(0) are all zero,and
00tr(o) _ lt.
: is
By thorem4.51, 6t^-1'(h)is increasingat h O and so
positivein an intervalto the dght of0' -. .
tnrs
By corollary 2 to theorem 4'61, applied to 0(n-')(")'
--nip"", is positivein an intervalto the right of 0'
function
tire argument.We find finally that, for sufficiently
smallpositiveft, O(h) > 0, that is to say'
I 6-t'(a)+'il.{.1\"'
(a)-'t l'
Ln-l hn
> nf' + ... +
f (a+ h) l(a)+ @) ffi
SimilarlY
/.-" 1o1
+\ gi'"'{o)+n\'
f(a+h)<f(a)+ry'@)+..'+ffi
ll'
A corresponding argument applies to negative values of
Collectingour results,we haveprovedthe theorem'
of
Apptication to maxima and minima' To illustrate the use
tfreoiem+.gf the readershould establishthe following extension
of the conditionsgivenin theorem4'73'
Supposethat
f'(c):f'(c) - ... -lr-l)(c):0'
la)(c) + 0.
Then (l) a necessarycondition for c to give a maximum or
minimum ofl(x) is that z is even; (2) supposingthat n is even'
4.81 THE DIFFERENTTALcALcuLUs 8l
if fQL)(c)< 0, then c givesa maximum ofl(x), and iflar(c) > O,
thencgivesaminimum.
The next theorem is the direct extensionto generalz of
theorem4.61.
4.9. IndetermiDatefotms
We shall extendtheorems4,81 and 4.82 to two functions.
(Compare tleorem 4.62, when n : 1,)
Thcorcm4.91. Supposethatf@r(a), Ci;t(a) existanilstd (a) + 0.
Then,as h -> O,
-I@r-
-F;G)'
fd(a)
and + #1 -l
8'^',(xl
a s x - > a\ o r a +o ra - ) ,
. fG\
then rl ot x->a (ora+ or a-).
ffi
We now indicatea field of applicationof thesetheorems.
Supposethat / and g are continuousin the closedinterval
(a, b) and thal f(a) : c@) :0. The limit as r + d+ of the
quotient .ftx)lg@) cannot be ascertaineddirectly by putting
r : a. Suchan expressionis traditionallycalledan indeterm-
inateform (0/0). On suitableassumptionsabout the deriva-
84 TEB DIFFERENTTALcALcuLUs 14.9
tives of / and g, the limit of f(x)lg(x) can often be found by
theorems4.91 and 4.92.
It is to be observedthat the hypotheses of theorems4.91and
4.92 are different, and neither tleorem includesthe other.
Illwtratior. Inoestigate lhe llmit as x -+ O of
tan kx- k lan x
I]inx-sinE'
.Sotttrbr. Writing tbe fraction as/(x)/g(x), we have
f'(x)-kser.'kx-kstrf:x,
g'(x) = kcosx-kcnskx.
If we us theorcm 4.91 it is necessaryto differentiate twice more, since
g"' is the derivative of lowest order which does not vanish at x = 0.
Theorem 4.92 (corollary) is, however, applicable to give
f'(x) cosx+coskr,
=;6Fr.*;a;'; ^,
id) t* + 't'
and this tendsto 2 asr to 0.
Exerclses4 (/)
Noleson lheseexercises aregiuenonp. 177.
1. Find the limits, asx tendsto +, of
,^ (l -x)'-x"
v/ (1=F=; '
wherem and ', are pogitiveintggrs;
-.., I +cos 2z.r
ou (k_lF.
2, Find theequations ofthe tangentandnormalto theellipsex - acos0,
y = 6 sin d at the point d.
A tangeutto an ellipsemeetsthe axsat P and O. Find the leastvalue
ot PQ.
3. By writing y = tr obtain parametricquationsfor the curv
xt+)" = 34xY'
Obtain the equationof the taogentafthe point with parametert, and
fnd th parameterof the point wherethis tanSentcuts the curve again.
What happcnsas , -+ - I ?
4, Stetch the locus(the cycloid)givenby
.x = a(r-sin l), / = 4(l -cos ,)
for valuesof , btwen0 and 2a
4.9I T I I E D I T F E R E N T T AC
LALCULUS 85
Prove thai thc normals to the curve are tangentsto the curye
x = a(t + sin t), ./ = -a(l -cos t),
and sketchthis secondcurve in your diagram.
5. Investigatemaxima aod minima of
oH, sin x
"'' sin (x -a)'
and sketchthe correspondinggraphs.
6, Prove that, if 9b > a > 0, the function
a sin x+ b sin 3x
has a maximum for somc value of r betwen0 and ]2. Taking a = 6,
b = l, find the greatestand leastvaluesof the function for 0 < x < jz.
7, P is a point on the circle whoseequation is
(x-h)'9+b/-h)2 = q2,
wherelr > d, artdPM, PN are the perpendicularson the coordinateaxes.
Find the positions of ,l' for which the area of the triangle PMN is a
maximum, and show that there ar two maximum positions or one
accordingas , is lessor greaterthan aJ2.
8. If y=ax'+bx2+cx+d,
what condition must be satisfiedby the coemcicntsa, b, c, d, if, cor-
respondingto any value ofy, there is only one value of x?
9. ll p(x) and q(x) are quadratic,and the roots of 4(x) = Q x16qqmp1g^,
provethat the maximum and minimum valuesofp(,r)/4(r) are the valuesof
k for which p(x) - k4(x) is a perfect squarc.
10. An open bowl is in the form of a segmentof a sphereol metal of
negligiblethickness. Find the shapeof the bowl if its volume is greatest
for a given area of metal,
11. The centresof two spheresof radii a and b are at a distance. apart,
where c > a+r. Wbere must a point sourceof light be placedon thc line
of centresbetweenthe two spheresso as to illuminate the greatesttotaL
surface?
12. A flat pieceof cardboardhas the form of an equilateral:JiangleABC
of height 3r. Points P, O, R are marked on the medianslG, BG, CG or,
oppositesides of G from A, B, C, with GP = GQ = cR. The iriangles
BCP, CAQ, ABR $e then cut away,and the rcmainingpiecaof cardboard
is folded about the edgesof the trianglePO-Rso as to form the surfaceof
a tetrahedron. Proye that the volume of this tetrahedroncannot xceed
(318)rh6.
13. Construct a function / for which /'(0) > 0, but there is no interval
(- h, h) in which / is an increasingfunction. (Try x, sin (l/;r) + &x, rvhere
k is a suitablconstaDt,)
t6 THE DrrDEnBNTraLcALcuLUs 14.9
14. Discus ,lverror'smethodof approxlmatioz,that, if .rr is ncar to a root
a of the equation/(x) = q then
l(x)
"'
'-7G6
"=
is likcly to b a better approximation to a,
Show in particular that if, in (a- h, a+ h), lf'(x)l > k > 0 and f(x)
has constant sign, then xl lies between x, and o if/(.r) andf(a) have the
same sign.
Approximale to the root near z of thc c4uation sin x = ?.r, wherc the
constant ? is smsll.
13. The rule of prcpotional pa s. If the values of a function /(.r) ar
tabulated at inlcrvals ,. and o+k lies bctween two successiveentries
4, a+r, thc rule is that approxiEatcly,
INFINIT; sERIEs
Excrcises5 (a)
Noles on lhesa exercisesare gi\en on p, 17'1.
l. Investigate tbe convergenccor divergenceof the serieswhose rth
termsare
jz-"'r' 7' 2-t" x"'
( n l ) " - 1. 3 . 5 . . .( 2 2- l ) -
' I AJ -(in-Dx '
i2n)t'
whrex > 0.
2. Prove that lhe sericswhose nth term is
( a + 1 , ( 2 a+ l \ . , . ( n a+ l )
(b + t) (2b+ t)-.-i;b + t')
convergcsifO < a < b and divergesif a > D > 0.
3. Prove that, if ay" g < I for all n > N, then >|l, convrges.
"
Show that this statementprovides a test which is more generalthan
theorem5,1l.
Discussthe series
a+b+a.+b.+ar+b3+....
where0<a<b<l-
4. Statea test for convergencawhich is more generalthan d'Alembert's
test in the sameway that exercise3 is more generalthan Cauchy'stest.
5. Provethat d'Alembert'stest doe8Dot determinethe convergence
of thc
seriesin exercise3.
6, Discussthe following statemen!.
The tcsts of Cauchy and d'Alembcrt, being derived from comparison
with a geometricprogression,cannot detcrminethe convergenceof the
seriestrt-r, which convergesmore slowly than any geometricprogresrion
($2.1I, exercis).
90 rNFrNrrB sERrBg t5.1
7. Prove that if, for all n, an*6/ao is less than &, whero /< < I, then lz.
convefges. Generalise.
L Establish the truth or falsity of thesc st tments:
(i) lf (u"lv") -+ | as n -+ @, then la" and Eo" both convergp or
both diverge.
(ii) If n" - u" + 4 theo Ez" and Eun both convergBor both diverge.
(iii) If (r"*Ja") > & > 1 for infinitely many r, then lao diverges.
lunl : uo+wo.
If now llz,l converges,then lo,n and )w* both converge.
Tlrereforeso doesI(o,- w"), i.e.2u*. I
5.21 sBRrEs
rNFrNrrE 9l
Defnition. If l,lu) is convergent,lhe sefies Zuo is called
absolutelyconvergent.
Exerclse. Give three examplesof serieswhich converge,but not abso-
lutely. (An exampleis I - I +,-*+ l- $+.... Theorcm5.22will suggest
many others.)
If we haveto determinewhethera seriesXu,"converges, the
firststepis to look at > lu"l. If I lu"l converges,
the matteris
settled.If X la"l divergeswe haveto try other methods,
Themostcommondistributionof signsin seriesis + and -
alternately.The following is a very useful theorem on such
'alternatingseries'.
Szn-Szo-z = -azn-t*azo 4 O.
Exerclses5 (6)
Notcs on thesc exerciset qre gioen on pp. 177-8,
1. &" ir a convergent seriesof positivs terms. Prove tltat
(i) if lr.l < a", then 16" is absolutely convergent;
(ii) :a"x' is absolutely convergent for -l < r < l;
(iii) Ea" cos ad and &o sia zd are absolutely convergent,
2. For what valusof & is t(-1)"/(22+ l)e absolutclyconvergent?For
what values of & is it conYergent?
3. For what values of x (if ary) are the following sertsconvergeni but
not absolutely convergent;
(i) tr", (ii) >x"/z?
4. Discuss tbe convergencof the series:
(i) 1-*+*-*+r%...;
(ii)l-*+l-3+i?-...;
( i i Dl - l + * - 3 + * - " ' '
lhe rspective deoominators beiDg of the forrn 8n,2, a'id nt.
5. Discuss the convergetrceof the sries
a b,a b, a b
. . . -,f
7-r'" 3- 4'f h_r-2i,-...,
where a and , are positivc constants.
6. Find two numbers diffedng by not more than I betwen which thc
sum of the series
r-1+3-1*l-l*...
must lie.
?. Establisb the tmth or falsity of the following statmeDts:
(i) U a" > O and an tends to 0 as r "+ @, then the series
ao-ar+a.-a.+...
convergs.
(ii) If &r" is absolutely convergent, and there is a coostant ,{ such that
- Alu"l < z" < I lu"l for all z,
then t,r" is absolutely convergent.
53. Conditionaloonvergence
Definition. If l,un convergesand Elu"l diverges,then 2u^ is
said to be conditionally oonvergent.
5.3I INTINITESERIES 93
Illustrction- The series
1-1+1-*+...
convergesconditionally.
The meaningof the word conditionalhereis that the sum to
which the seriesconverges is conditionalon the order in which
the termsare taken. If the termsare rearrangedthe sum.isin
generalaltered. As an illustration,rearrangethe last seriesso
that two negativetermsalwaysfollow one positiveterm
r-+-+++-*-*+....
Supposethat thesurnof ihe originalseriesis s (in fact,s : log"2
but we do not needto know this), We shall prove that the re-
arrangedseriesconvergesto sum+r. Let s, and /, be the sums
of the first n termsof the two series.Then
- 11 I I I
,l -s ^: : |l -_ 2 -- )- .l
4.r..._;1_ 42q_4n.
In eachblock of three terms (two negativeand one positive)
subtract the first of the negativeterms from the preceding
positiveterm, and we have
tlll 1 I
2468" 4n-2 4n
: tt'"'
Therefore,as n -> co, t"n-> \s; lr,,*, and ts,+2tend to the same
limit. So the rearrangedseriesconvergesto sum b.
This changeol sum by rearrangement is not paradoxical,as
sn and t^ are differentfunctionsof n. The more striking fact is
that, howeverwe alter the order of the termsof an absolutely
convergentseries,the sum is unchanged.
Theorent5,3. If 2u" is absolutelycont)ergent,
eueryseriescon-
sistingof thesametermsin any orderhaslhe sanrcsum.
Proof. We provethe theoremfirst in the specialcasein which
all the termsare positive(u" >. O).
Let Xai consistof the same terms as Xz, with the order
rearrangedin any way.
Iaf
If z : l, the seriesdiverges.
lf z i l, s,, : (l - z")l(l - z) andwe haveseenthat this tends
to a lirnit if and only if lzl < l.
So the values of e for which the sedesconvergesare the
points insidea circle in the complexplane.We shall seein the
next seclionthat this 'circle of convergence'existsfor a wide
classof series.
Exercises 5 (c)
Notes or theseexercisesare given oa p. 178.
1. Find the sum to n termsof the series
1 + 2 2+ 3 z t+ . . .+ ( n + l ) z n+ . . . .
Prove that, if lzl < 1, the seriesconvergcsand find its sum.
!16 tNrrNrrE gEnrEs 15.4
2. Provethat the sries
z , l z \ r -(r-,)
/ ' \!
1-r-"*(ri
1
*" '
convergesif snd only if re z < !. What is its sum?
3. Decide tlle cotrvergenceor divergenceof the serieswhose ath tcrms are
...i" ....(l+D. . . . . .F
(rr;, (ll, -;-, {ur) ;.
^lSl.,'
i.c.if Vl . !t.
And, if lzl > l/1, ta or+rl
liml:'4r-l\ 't '
I ttnz* 1-
so that the zth term a,nz' doesnot tend to zero and convergence
is impossible. I
. Wchavesupposcd
,lVole. rfinitoandnotzcro.Thereadrshouldconsider
thoxccpted cass.
Illsstrqtions. Esch of tbo scris
2*, 2(f ln), z(z"ln\
bas radiusof convergenccl. Thc first scricsconverg* at no point on tho
circlc lzl = l; the third is absolutclyconvergpntat all points of thc circl.
Thc scconddivrgpsfor z = 1,andit canbeshownto convrgpat all other
pointc oo the cidc lzl = I tty thc dwice suggcstodin exercisc5 (d),8.
The systematicteatmnt of th quGtion of convergencc of &.zi on its
circlc of convergencc is outsidethc scopoof this book-
Ercrcbes5 (d)
Nolecon thesecxerctses arc gloenonp. 178,
of the powrsriesof which thc general
1. Find tho radii of convergence
lcnns are
(i) nzi (D rfrtnt Au,0+)',
(iv)z!2", 0)q#l, $)#,
(vii) ztzn (viii) t3+(-1)").t
2. Provc tlEt, if la"lt," .' l/r as I -.o, thcNrthe rcdcs &"2" ha! ndius
of convrgpnccr,
DirsulE thc convcrgence of the srics
fuf zr
whcro 4 is o constant.
3. If thc soriei Ia" z" has radius of convergcnco.R,in what region of tho
z-plane doca the sri$ >a"(z-zo)i convcrge? Alswr the sams qucstion
for thc srica >a.z-".
5.61 rNFrNrrE sERlEs 99
4. Prove that the series
l+z+z'+.,.+2"+...
convergesfor all valuesof z for which the series
2 + 2(22- 1)+ 2(22- t)' + .,. + 2(22- rr" + ...
converges,and has the samesum.
5, lt for all n, la"l < &, what can you sayabout the radiusofconvergence
of 2anz"'t If, further, la"l > I > 0, what thcn follows?
6. If, for all r, la"l < l, then the equation
| = orz+arz.+,--
cannoi have a root with modulus lessthan +. If it is satisfiedby
z = +(cosd+ i sin 0),
theo a- = cos n0 - i sin n0.
7. If thc radiusof convergenceof lcn zn is / and of >b, zn is J' what can
you say about the radius of convergcnceof
z(a"+ b")2"?
($u"r)
(fr,"r),
and so the series2aoao convergesabsolutely.
By theorern 5.3 (which remains true for complex terms), its
sum is the samewhatever the order of the terms. But ttrere is
one particular order, namely, sumrnationby squares,in which
the sum is evident. For the sum of all terms with sumxesnot
exceedingn is
Qt!+ua+... + u") (oo+ o,+... + o")
atrd the limit of this is rr. I
5,71 r N r r N r r Bs a R r 8 s 101
Corollary. If 2a"2" and Ebnz" haue radii of conuergenceR
and S, then their product is
and a, :l!yn1o+0n1.
In the last line B is a numberbetween0 and I which depends
on a, lt and n.
If, as n-+ co, R,.>0, then S" -',f(a+ h). So we have the
followins result.
102 rNnNrrB SBRTEE l5.E
fheorcm 5.8. (Taylof s series).If, in thzorem4.82, R^ (the term
in l{) tendsto 0 as n tendsto @, then
'
f(a + h) : f(a)+ hf (a)+ ...+ X j\r(a) + ....
In accordancewith the remark on p.79, the expansionfor
a : 0 could be calledMaclaurin'sseries.
The binomialseries. As an exampleof the useof theorem 5.8,
we shall prove the binomial tleorem for/(.x) = (l + x)n, where
rz is a rational number, positive or negative,
f(i'(x) - mQn-l) '..(m-n+l)(l+xY-
and we shall prove that, if - 1 < x < l,
(.l + x ) - = l + llm\ /m\
','lx+... +1"'lx'+....
\t,f \nl
If m is a positive integer, fn+tr(x) = 0 and we have a poly-
nomial of degreem. In the generalcase
vn
R":"lfr't(ox) : lm\ Xtr
,
\i) .6an
where d dependson n.
If now 0 < x < 1, (l +02a1"'" > l for n > m, and,asstated
in exercise2(d),ll,
(!\o-,o as z-'oo,
\nl
so that R,| -+ 0,
If .r is negative,this argumentbreaksdown sinceI +dr is not
greaterthan I . We haverecourseto Cauchy'sfonn of rernainder
(theorem 4.83) which applies for the full range -l < .x < l.
Thisgives
- : m@-l)...(m-n+l)(1-01-t*
^" -IllllG.-lJ-
(lTatt-",-m'
Now (l - d)/(l + dx) is lessthan I and so
ln";' r-l/n-r\lr-r"
r \ z - t 7 1r " t
'
where K- dependson m (and x) but not on n. Again from
exercise2(d), ll, R"-+Q as z+ao. I
Other well-known Taylor's serieswill be found in the next
chapter.
s,El INTTNITE SERIES 103
Exerclses 5 (?)
Noles on theseexercisesare gioen on p. 178,
1. Discuss the convergenceof the serieswhosc ,rth terms arc (with a, ,, c
positive)
(i) 1- cos (z/z), (ii) a"J(b" + cn),
I
*"'(art+bn+c)*' (iu)
Gdi.,#TT,
(v) (- l)'avi (a > 0), (vi) (- l)"{r/(rf + 1) -d,
2, Provc that, if a > l, the series
1 . 2 .' 4 .-8-.
'd+l ' '
a+l' a,+l ar+l "'
convergesto the sum l/(a- 1).
3. If u" > n"*, and lrn convergps,prove that lim rtr, = 0.
4. From the serics l(l/z), every term which contains a spcified di8it
(say 7) is removed. Prove that the series formed by the remaining terms
converges.
5, The sries >anzi converges to sum /(z) if lzl < l. Prove that, if
sn=ao+ar+.,.+ai ahd lzl < 1, the seriesrrnz'convergs to sum
. f(z\|a.r-z).
Hence give a proof of the binomial theorem when the exponent is a
negalive integer.
6
THE SPECIAL FUNCTIONS
OF ANALYSIS
i.e.lg{dP=}g{m"o*?-9,
*' **{,ts*94} =Jg{***4@}
So we have got down to the root of the matter. The truth ofthe
theorerndepeudson interchangingthe order of the two limiting
operstions lim and lim
n-+0 /,.-+6
applied to the particular function of n and i (and also of x,
which remainsfixed while n alLdh vary),
All we can sayis that the interchangeof order of two limiting
operationsin generelgivesdifferent results, It is only when the
function to which they are applied satisfiesrestrictiveconditions
that the interchangeis valid. A simple illustration in which the
order of thc two limits sffects the result is the following:
: -r'
g'{:sii#} : }:r(-r)
.. 1.. | -nhr =
tim llim; .*l
I ' : l.
tim (1)
r-+- h-+o *nftl ,.-+.'
General tlmreme on interchange of limits are beyond the
scope of this book. From time to time we shall have to deal
with repeatedlimits of simple functions, and we shall give the
most straigbtforward argumentavailablein eachparticular case.
6.4. Rate of imeese of exp x
We prove the theorem which led us to discuss repeated
Iimits.
Theorcm6.11. d(ery x)ldx = exp r.
Proof, Using theorem 6.2, we have
exp(x+l)-exp x : exoh-l
exPx ---7- '
h
6.4| IHE SPICIAL FUNCTIONS OF ANALYSIS ro7
Now exoh-l -t - L - -hL - - ] -h z
-'2l'3l'"'
h
: 1+t!(h), say.
We wish to prove that /(l) + 0 as h -> 0. We have
t+hl
| -t+hl (it lhl < 2)
+0 as l+0. I
Corollary, expx is a continuousfunction,
This follows from $4.1(3). Alternativelyit could be proved
directlyby an argumentsimilar to that usedin the theorem.
Theorem6.42, expx is a strictly increasingfunction and, if
y : expx, y takeseueryoaluegreaterthanofor oneoalueof x.
Proof. If x > 0, then, at oncefrom the series,expx > L
If;r < 0, expx=l/exp(-.n)>0.
Sinceexp;r has a derivativewhich is positivefor all values
of x, it increasesstrictly.
As x + co,plainlyexpx-+ co.
As r -+ -ao, expx : l/exp(-r) + 0+. I
Theorem6.43. (TlE order of magnitudeof expx.)
For anyfxed k (howeoerlarge)
expJr->@
as x->@.
=F-
Proof. Let z be the integer next greaterthan *.
If x > 0, expx > x"ln!, since this is just one term of the
seriesdefiningexpx. l
You should acquirea vivid appreciationof this important
facl. For large x, thefunction expx is larger than anypower of x
(seefig. 3a, p. 110).
l+x+;t+...+:, = ke".
,-crpt
a
Fis. 3
!*0""o-f,.
log (ab) : loga+logb,
log.r+co as x->co,
log.r-+ -oo as x+0+.
If& > 0,
$r..o as ;->@.
This last property (correlative to that at the end of 96.4) is
important. In words, logx tends to infinity as x tends to in-
finity, but more slowly than x raised to any positive power
(howeversmall). The generalshapeof the graph of y : 169I ;t
shownin fig. 36.
There is a sirnple representation of log (l +t) as a powcr
series.
6.61 TrrE spEcIAL ruNcrroNs oF ANALYSIS 111
The general power a'. At the end of $6.5 we had defined a'
when ;r is rational and also for irrational x in the special case
a = e. We have still to attach a meaning to a' when x is
irrational and a + e. We adopt the following definition,
Exercises6 (6)
Notes on lheseexercisesare giaen on p, 179.
l. Prove that, as r -+ co,
4(xr/"- l) -' log x.
2. Prove that, if r > 0,
-r-,x' < log(l +n) < x- j,r'+ !.f.
Give an extensionof theseinequalities,
3. A differentiablef'rnction.f (not identically 0) satisfiesthe functional
equation
f(xy) = f\x) +fty).
Prove that/'(r) = ,4/.r.
ll2 TsE spEcral. FUNcrroNs oF ANALylrg t6.6
4. Prove that, ifl > 0,
=z(fi+t (i#)'.I(il*,)'*).
rcs'
Use this sriesto calculate log 2 to thre placs of decimals.
5. Find the limits, ag .r -+ 0 and as I -+ co, of
6.7. Trigonome&icfunctions
As predictedin $ 1.2we baseour accountof the trigonometric
functions on the definitions
"o'":r-{+fi-...,
,in": - t3' ! + It -
"- ""
Exercises6 (c)
Notes on theseexercisesqre giuen on pp. 179-80.
l. Refinethe argumentof theotem 6.81 to giye closerboundsfor {2, for
example,1.5 < lzr < 1.6. (The seriesfor cos x and sin.x do not provide
a practical way of obtaioing accurateapproximationsto z. For better
methodsse$7.9.)
2. Prove the statementfollowing theorem 6.82 that 2I is the smallest
period of cos x and sin x,
3. If a ard , are positiveconstantsand x is rcal, provo that
.f(-r) = c cot r+6 cosecx
takesall real valuesif a > 6, and all real valuesexceptfor a certainrange
if a < b.
Sketch graphs of y = f(x) for a > b, a = b, a < b.
4. Prove that, as ,l -+ @,
lqss :) _) a_1...
\nl
116 TEB SPICTAL FUNCTIONS O! ANALYSIS 16.9
a=_ r
e- -IT=t.
Iheorcm 6.92. Theequationx = tny delnes on imserse
ftmction
y = arc tanx dertnedfor all x. y is an increasing
function and
j:-r: -tr, =
lyyr tn.
Also dY : I
dx I +xr.
Proof. dxldy - ssczt = l*.x1
tany is a continuous increasingfunction for _\n < y <
+7t
and, as y -+ + tv, x -->1oo, respectivd. The exiitence of tie
inverse function and the value of its deriyative follow from
theorem3.9 and ga.2(6). I
6.91 THE spEcrAL FUNcrroNs oF ANALYSIS ll7
Nole. As in theorem 6,91, we are defininga principal value. If, for a
given value rr of r, the value / = ),r satisfies.t = tan /, so does yr+ nt,
whcrc ,r is any positive or negativeinteger.
Ercrc&s6 (O
Noleson theseerercisetqe gioenonp. IEU
l. Find the sum I cosh(c+ 2rO.
r-o
2. Prove that the equafion
.r = 2+logx
has two positive root!, 8ay a and t
Thc sequncer, is defircd by
;"+r = 2+lo8*r @= 1,2,3,...),
where a <.xl < 0. Provo that:" +6.
3. Provothat, if -1 <r< l,
1+ r coso+ ...+ r,rcosz0+.., = @,_"**r,
' =
r sina+...+2. sinrr0+ tsinO
"' l=r. @;6+,''
resultsliko thoseof 3 for hypcrbolic
4. Investigate function$
5. Proveth.t, if/ = e-'l'r, then
l"+1+rl"+rr+ (r,+l)'rd = 0.
Tho functions./l arc d6ncdby
t (x)= (-t.dd #,Qin.
Provc that
(i) l:+r = (n+l)f,b
(D fa = xf,-fi,
(iii) 4+r-xl,+r+(n+ l)l. = 0,
(iv) Ji is a polynomialof dcgreez.
6. Obtain orpansionsin powcr-series of the foUowingfunctions. Find thc
geDralterm if you can, otherwisothe 6rst threenon-vanishingterms.
(i) cosrr, (ii) tan r,
(iD (arcsinrf (iv) sin (m arc sin x),
(v) e. co62x, (vi) coslog (1+r).
tlg
Write S: >M,t,
,-*,
and t : 2_r*,6n
then s(o(s.
Our ultimate aim is to prove that, if/is a function of one of
the commonly occurring types (including continuous functions
and monotonic functions), the sum r tends to a limit as 0*
tends to 0. We shall define this fimit to be the integral of the
function / over (a, b).
You will observetlat this limiting procssis of a lesssimple
kind than those which we have so far encountered, If the
function/is given, the number a dependson the x, and the $;
in fact, a is a function whosedomain is a set of dissections.In
the passageto the limit, we suppose8+ to take a sequenceof
values tending to 0, and the permitted dissectionsare pro-
gressivelyrestricted by the requirement that their norms are to
be lessthan 8*.
On account of the complexity of the limit-operation, we fust
take an easierapproach to the integral though the botmdsof
the sums ,S,s for all dissections,
THB TNTEORAL CALCULUS
Theotem7.22, J > j.
Proof. l*t 9'and 0rbe any two dissectionsof (a, D),
l*t 9s& the dissectionwhosepoints of division are all the
points of division of eithler9, ot 9". So It is a refinementboth
of 91rnd of 9* The corollary to the last theoren give$
S(e) < s(e) afi 4e) > s(e).
But s(e) > 4e),
beingupper and lower sumsfor the samedissection. Combining
all theseinequalities we have
s(s) > 4s).
Sincethis is true for all dissections9t,
J : inf s(e) > 4e).
Sincetie last line is true for all dissections9e
J >- sttPdQ) : i' 1
If we assumeabout / oaly what we have assumedalreadn
that it is a bounded function, then it is possible for J to be
greater than j or equ4l to j.
Illustrction An o(ample in which J > J. De6no
'rt
f(x) = | ; is rational
/(.t) = 0 if x is inational,
and tsko th intcrval (a, D) to be (O l).
Thcn, whatever disscctiotr I is takcn, every M, = I and S(9) = l
E v c r ym , = 0 a n d . r ( 9 ) = 0 . S o J = l a n d j - 0 .
Herc y = /(.r), far from reprscDtinga curvc in a Ntraightforward scnso
capablc of bounding an era, is discontinuous for wry value of r.
Etercrse. Givo tho simplest examplc you can think of in whicb J = ,/.
= f,r<*lt,
" r-\
I-e<\,f(N)8,<I+e,
t-l
< I+k.
n
Similarly lrn"8, > I-2e.
,-l
Mr-^, a J-
D-A
Exercises7 (a)
Noter on theseexercisesare given on p, l8O,
Calculation of simple integralsfrom the definition,
IL
(0, l) into n equal parts.
1. Calculate I xdx by disse-cting
JO
fb
2. Calculatc I x.dx, wherek > 0, by dividing (a, ,) into r parts in geo-
*"tri" p.ogr"'rion the poiots 44, 44r, ,.., aq'-r, whercaqn = b.
"t
lc
3. Calculate I sin rd-:r by dissecting(0, a) into equal parts.
lo
4. By the method of2, prove that
12dx I
Jrr'= r'
Deduce that
I r I I )
,. =I
,llf itr*rl,"tr*a"+ "+ pn1,l 2'
A theoremof Darboux.
5, With the notation of $?.1, S '+ J as 8* + 0.
This is an important theorem, showing that the number "I which was
definedas a bound is in fact a limit. As the proofis a little mole difncult
than those of any resultsin the text, we have developedprope ies of the
integral independentlyof it.
I s(5,)d, : f -2/(f,)
r-l l
d,+X s(6,)6.
r-1
Ii'rl" |"'trt.
lJd I Ja
Ul")'"
ffi)ff:.)
Proof. Thts may be deduced from Cauchy's inequality
(exerciseI (d), 6) or from the fact that
fb fb fb tb
= A2)"f+2M)"-fs+p"J"{
l"Of+lrdz
is greater than or equal to 0 for all values of the constants
A, p.
Theorcm7.61. F is a continuous
function.
Proof. rx+tl
F(x+h)- F\x) : tOl at
),
With the notation of$7,5 (5), the absolutevalue of the right-
hand sidedoesnot exceedmax (lmil, lM&l). So
F(x+h)- F\x) -">0 aslt +0. I
If we make the additional assumptionof continuity ofJ we
cao prove a sbarper resulL
Thcorem7.62. If f is integrableh (a, b), then,for ay oalueof x
for whichf is continuous,
F'(x) : f(x).
1.61 r H E T N T E o R A Lc A L c u L U s 129
Prool Supposethat ft > 0. Let now
M: supf(t), m:
inff(t)
for x < , < x+h. M and ln then dependon :c and l. By
$7.5(5), k+h
^h fQ) dt < Mh.
" J,
^ -F(x + /r) - F(x)
\'-:-l---:--:-l-'
Theretbre ,n < < M.
n
Let ft -+ 0+. Since/is continuous at x, both m andM tend
to /(r). A similarargumentholds whenh -->O- . I
Theorem7.63. Let f be continuous in (a, b). Supposethat ( is a
function hauing the property
Q'@):f(x) for a4x{b.
fx
rhen a<x<b.
)"f<tlat:Q@)-C@) for
Proof. From theorem7.62,the function F-l hasderivative
0fora < x < b.
By theorem4.61(corollary1),,F-l is a constantand,since
F(a) : 0. we have
F(x): Qg)-Q@). I
The existenceof a function / having a given continuous
derivative/ is establishedby theorem7.62, Such a function
which, by theorem7.63. is determinedexceptfor an additive
constantis calledan indefniteintegralof f, written
If@)a''
An indefinite integral may or may not be readily expressiblein
terms of known functions; if it is, theorem 7.63 provides the
normal method of calculating the defnite integral
fb
I f(x) dx.
Ja
Ifr>1. l-!2.:G-a)t-]
J \x-a)" t-n
d, - (x > a),
Ifn : t.' I {log(x-a)
J x-a [log(c-x) (x < a.1,
or, conveniently,in one formula,
fdx
toglx_al.
);;=
A fraction whosedenominatoris a quadraticwith complex
roots is integratedas follows.
9-2
tl2 rHB INTBCRAL CALCULU8 T7.8
Example.
t 4x-l
I3xe-4x+5"'
The derivativeof tfte denominatoris 6x-4. writc the numerator4x- I
as l(6x-4)+f. The first termgives
2 t 6 x - 4 y'r =
; I 3#5 | log (3xr-4x+ 5)'
The secoudgives
5l dx 5l dx
9J.ri-tx+* 9/ (x-3)'+*
= 53 x-| = 5 3x-2
'
45arctanffi ffi4clan ^
To integrate a partial fraction of the form
>
T!!Lx"+2a*T*@ D'
write the numeratorrc p(x+a)+(q-ap) and the problern is
reducedto integrating 1l(x2+2ax+ b)^. This is achievedby
successive'reduction' of the index n; seethe paragraphbelow
on reduction formulae.
Trigonometricfunctions. (a)To i\lqgrate a product of cosines
and sines,turn the products into sumsand differencesby using
formulaelike
2cosacosD: cos(a+ D)+ cos(a- D).
Examples. (i) J cosr.rcos4x dr.
(ii) J sid .r costx dx.
In (D, costx cos4x = t(l+cos2x)cos4.x
= +cos4x+* cos6r+1 cos2r.
Hencc tbc intesral ig
I sin 4r++ sin 6J+t sin 2x.
In (ii), the odd power of cos x suggststhe substitution !l = sin r, giviDg
-ar)d/ and so * sin8.r-+ sin! x.
Jzr(l
(b) The integral of any rational function of cosr and sin.x
can be transformedinto the integral of a rational algebraic
function by the substitutionIanlx : r for which
7-t2 2t dx = 2
cosx - slnx:
l+-ts' l+rr' 7t 6;r.
7.81 THE INTEORAL CALCULUS T33
Examgle. , )_ r,r)r
: = I arctan(+tanlx)'
J s*o l"-t t I ;i
formulae. Supposethat we require
Reduction
I sin^xdx'
This comesunder(a) above,but it is moreeasilyfound by step-
by-stepreductionof the index ,r. A reductionformulais nearly
alwaysderivedby the appropriateintegrationby parts. Here
I sin"x dx : f isin"-r-x;sinxdx
: - sin"-rxcosx+Jtz-I) sin"-axcosx cosxdx
: -sin"-rrccosx+Qr-l) -
Jsin"-'zx(l sin'x)dx,
and so
n [sin"xdx = -sin'-rrcosr+(n-\Jsin"-txdx
and we have connectedthe integral of sin".r with that of
sin"-'x. This integral is particularly simple if the range OI
integrationis (0, jz), for then
fb rlt
nl- sin^xdx: (n-1)l-
' J o sin"-2
xdx.
Jo
By repeatedapplicationof this reductionformula we find that
l- sin"xd:ris
JO
(n-l\...4.2 /z-l\...3.1n
n . . .5 . 3 n ...4.2 2
accordingasn is odd or even.
Irrationalfunclions.We take only the simplestcases,
ExamDle.
tdx
I t-+q JG+6'
Thc function under the sign is linear, and the substitution x+.b = u.
"/
givesthe inte$al of a rational function of ll.
Exerclss7 (r)
aregiueain p. 181.
Noteson theseexercises
1. Integrate
I
(, - 2F-(r'+ r)' (;r;TtIFTFi' F +aqtai'
11
F + 1 , ) , 0 +1 _ .
2. Intecrate -(xrJ: .
+ cr)r '
(i) by substitutinglr = xr+cr, (ii) by substituting.r = a ta! A.Verify that
ihe two resultsagree.
3. Wallls'sproductfor r.
tl'l
Writing /" = I sin" x &, prove that IrJIi.+t lies between I and
to
r+Ql2n).
7.8I TnB INTEGIAL CALCULUS
Deduccthat
r .. 224466 2m 2m
t=;*Ilrssl '..6=n;i'
Establish thc altemativ formula
"t"=#.wffi
4, Integratcco$.3r sin 2r, tanrsecr, cosectx,
f* ttr
5. EvsluatcI sinT*sinz.rdx, I sin7x sinrrxdx.
JO JO
6. Showhow to integrato
a+D cosx+c sin.x I
,+qcosl+rsil' ;cofiTZ'cd.rsil+ciilF;'
7. Obtain reductionformulaefor thc integlalsof
l(f * 1)', {log.t)i .r"r/(d-xrL
tan;.r, secn.r, l/(a+t cost)'.
8. Find a reductionformula for tbe integral
[ ,"d"
J ^l@x'+2.hx+c)
and useit to evaluatc
tfdx
I l6TE+4'
9, Prove that, according as n is an cven or odd p$itive integer,
["'ii4 a = o or '.
Jo srna
If a is a positiveintger,cvaluate
[" sin"noN'
''
Io
"i*d
10. If the polynomialP"(x) is defned by
= ;a1 (;J
P"(,>
/,t\n
(-d-rr,
Provethat
(i) if O(x) is a polynomialof degreelessthan n,
IL
I P"(x) Q@)dx = 0,
J -l
tu : t-l+[-....
This givesa meansof calculatingr, but the seriesconvergestoo
slorvlyto be useful for numericalwork. The following simple
relations,whichthe readercanverifyfrom theadditionformulae
for the tangent,lead to serieswhich convergemore rapidly
rx'
+: r.
Geometrically,the areabetweenthe curvey : l/x2, its asymp-
tote the r-axis and the line.r = I is finite.
13E rEE TNTEGRALcALcuLUs
t6
!" .f{i a** r,
that its value is t
we say that I f(x) dx exists,or contserges,and
.,4
fx
If | /(x) dx existsfor all valuesof X geater than d, but does
Ja
P@
| 'f(x) dx'
.'-o
f.o la
If I f(x) dx : /r and JI - a f(x) dx = h,
Ja
wewrite : ,,*,".
l' -f4l*
It is easyto seethat the value of the last integral is independent
of the particular value of a.
f@ )-
Theorcm7.10. if andonlyifk > l.
conuerges
Lfi
Proof. lf k + l, _l
F dx = Xa_k
-1:E-'
J,l
The limit of the right-hand side is finite if and only if & > 1.
nk:r,
li!:rc'*'
which tendsto infrnity as X+co. I
Nole, Any Dumber greater than 0 would serveinstead of 1 as the lowr
limit of integration.
Integrals of unboundedfunctions. If I > q the function
l/y'x is continuousin (0, l) and
f'1-z-z,to.
Jr{x
7.101 T H E T N T E G R A Lc A L c u L U s 139
Exerclses7 (c)
Notes on theseexercisesare giuen on p. l8l.
1. Evaluatethe integrals
xtb dx
. l oF + x t + x + I ' J&@')1G'+r)'
2. Prove that, if O < d, < i,
dxe -
J o x 2 + 2 x c o s d + l s i nd '
l@
3. P(ove that x" e-a dx = nt
lo
f@
4. Evaluate e-d"cosbxdx (a > O),
Jo
[' sr,,nax ax.
J -6
tb )-
5. Prove that I -,,.- "1,,--* b > a\
J a lt\x- a) \o - x)t
exists. Calculateits value by two differentsubstitutions,
(i) x = acosrd+Dsin,d, (ii) (b-x)lG-a) = u'.
6, Evaluate
l-rcosd
t= l
) -E r-zrcos0+f'
w h e n ( i ) 0< r < l , ( i i ) r > l .
Prove that, d being fixed, f tends lo one limit as r + I through values
lessthan l, and to a difTerentlimit as r + I through valuesgreaterthan l.
Show, also, that neither liBit is equal to th value of l when / = l
TIIB INTECRAL CALCULUS 17.11
Fig. 4
I@-r)>f(x)>f(n).
t42 TIIB INTEORAL CALCULUS 17.11
Inte$ation from n* I to z then gves
1n-9>f_,1x1dx>f(n). (A)
Exercls37 (O
Noteson theseexercttesare gloenonp. lgl.
l. Provethat
.toEi--l-;=
rrlot
(n= 2'3'"'l
Jo6-,ne
Dooting either of thesc cxprecsionsby r., prove that
t 5. ProYethat,as&-)0+, -c
t x ; ; ; I. + 1 .
II 6. Pro\e that
LN
I srr
I i' ;oottr
tI corvcrgesif,c > I and divergesif& q l.
7. Prow that
t l''Ll'|
J rW
re < I togr <
I rlogx.k+logn
I
i
Invcstigrtetho limit asz -+ o, of (al)r,"/r.
t, Verify the statmentin tho numericalillustrstion at thecnd of thcorcrtr
Zll thst the sumof 10 rrrnsof lo/n) is lssthan 20,
i 9. Discussthe conlcrgenceof
ll
(i) r
n;1o;o;r ' ci) t nliogtroen)"'
t14 TEB INTEORAL CALCULUS 17.12
r-'-<FJaoxF + 1
r t )t-l1r
= ! loc(r+ 1)-* los(f -x+ l)+t5 arctan:fujtr
|
= +toqz+4 < o.tre.
J4t
the indcfinite integral being obtained by the method of $7.8.
7.121 T H E T N T E G R A Lc A L c u L U s 145
Thisgivesl<0.915.
To 6nd an approximation from below, observe that x'> :l for
0<.x<l,alrdso
-11
tr ,1r T
l' = l)}tr= log(l+J2)
/" 71"trr llog{x+./(x'+
> 0.896.
Methods of obtaining closerapproximationswill be Sivcn in $7.13.
Example2. Approximate to
ftn
t= ^lbin x) dx.
)o
(The indefinite integral is not cxpressibleby the speial functions
included in chaptr6.)
(i) J(sinx) > sio x gives1> l.
(ii) We have2.r/r < sinx < x(0 < x < t7). Integratiogthe square
r o o t s 'w e 6 n d
r'047< 1< l3r.
(iii) Schwarz'sinequality(as in exanrpll) yields
I<,J(+nl<1.254.
6@1:
fiet-n1le)at
: htf@)+f(o)\-zfrc>*
:
and so
l!<"t * rh{f(o)
fT
+R,
+f(h)}
where lRl < lMlJ O t(h-t)ilt = lrMhg. I
To approximate to the integral of/(x) over (4 b), we may
divide the interval into z equal parts, where 6-o : rt, and
apply the trapezium rule to eachpart. The approximate value
it'h"o n{tI@)* + h)+ + 2h)+ ...+ (a+ nt
^o f(a tf )J,
the first and last terms having the coefrcient |. If, moreover,
l"f'(x)l < ,ll4,then the enor is at most
lvMnhsor fzMA#.
Simpson'srule. TIis method, basedon the idea of approxi-
mating to the curvey - /(r) by a parabola drawn ttrrough three
of its points (instead of by a straigbt line through two points)
is likely to give a much closer estimateof the integral.
The Simpsonapproximation to
ftt
r:
I "f(x)dx
is :
,s s@-c){J@)+qf$G+d)l+f(o}.
To obtain an upper bound for the error we shall assumethat
/(x) has a bounded fourth derivative.
Lewru. lf y = p(x) : Ixz+ mx+n,is the parabola(with axis
parallel to Oy) dtawn through the points ofy : /(x) for which
x = -h,O, h, then
trh
of the 'error',
Definenow, as a measure
fh
E(h) = | f(x) dx- !hlf(- h)+ 4f(0)+f(h)}.
J-h
f tt-}.::-+l d,,
J o J(l -1';
--
(where II denotes the products of the factors given by r = 1,2" ..., n-1),
prove that (a beiug real)
ln
I log(l - 2acosx + ar)dx
JO
8
FUNCTIONS OF SEVERAL YARIABLES
&ffi)""^*.(fr)
In the suffix notation, the former of theseis (i)y or, by sup-
pressingthe brackets,fi.
It will be found that in all straightforward examplesthe order
in which partial differentiationsare carried out doesnot matter.
We shall usually have, in fact
*ffi
&(v):
and then we can write eachof them as
, Azf
(ot f-).
6ia
Our first theorem givesconditions for this interchangeability
of the order of partial differentiations.
Theotem8.3. If f"v andfu. are continuousfunctions of (x, y) at
a point (a, b), they are equal there.
Proof, Sincef,, andfr" are continuous at (a, b), there is a
square N, centre (d, b), in which they exist. The functions f
andf, from which they are derivedmust existin.lV, as mustl,
For tlre rest of the proof, h and k are assumedto be small
enough for (a + h, b + k) to be in ,lV.
The proof dependson taking increments in the function /
correspondingto incrementsin the variablesx and y in turn,
If
^
i@,y) : f(x,y+k)-f(x,y)
and 2t : Q@+h,y)-Q@,y),
8.31 F U N c r r o N s o F S E V E R A LV A R T A B L E S 155
Exercises8 (a)
Notes on theseexercisesare gi\en on p, 182,
l. Which of the following functions (with th definition suitably com.
pleted for t = 0,,y = 0) are continuousat (0,0)?
.., (x+y)'1 ..-, xy2 ,,,,, r'+y'
\t) -x4i, (r)
x+-r,
(iiD=;.r.
2. \f x = r cos0, y = r sin d, give the valuesof the partial derivativesof
r and d with lespectto r and /.
Is it true that
nr\ ni =
\u,)\;) t?
lllustrateby a diagram.
FUNCTIONI OA SEVERAL VAIIABLES 18.4
8.4. Differentiability
Let us consider what is the most natural extension of the
idea of diferentiability when the number of independent
variablesexceedsone. If we have a function;f of one variable,
differentiability at a point means the existenceof a tangent
fine-a linear approximation to the curve y - f(x). On a
surface z : .f(x, y) an approximation linear in the variables
suggestsa tangent plane. This leads us to the following de-
finition.
Defaition. Thefunction f(x, y) ir differentiable at (a, b) if, for
(o + h, b + k) in a neighbourhootlof (a, b),
f(a+h,b+k)-f(a, b) : Ah+Bk+e(lil+l&l),
whereA and B do tmt dependon h or k, and e ten& to O as
(lt, /c) + (0, 0).
This meansthat the plane
z -f(a, b) : A(x - a) + B(y - b)
is at a distancefrom the surfacez : /(x, y) which is small com-
paredwith the displacementof the p oint (a+ h, b + k) from (4 D)
and is the tangent plane to the surface.
The 'error-term' in the definitionof differentiabilitywhichwe
have written as
6(rrl+ Iel)
can be put in severaldifferent forms, which are easily seento
be equivalent. We could, for instance,say instead
e,,l(h2+k\,
Again, we could rewrite the condition as
f(a + h, b + k) -f(a, b) : d4h,k)h + p(h, k) k,
where a(i, k) -> A and p(h, k) -+ A as (i, &) + (0, 0).
If (in the original definition) we keep /< - O and let l tend
to 0. we have
A = fJa, b,),
and similarlv n = f"(a, b).
It can be seenthat differentiability, in the senseof this de-
finition, assertsmore than the existenceof partial derivatives
8.41 F U N c r r o N s o F s E V E R A Lv A R r a B L E s 151
with respectto .r and y. Geometrically,the existenceof the
partial derivativef(a, D) impliesthat thereis a tangentline to
the curvewhich is the sectionof the surfacez = f(x, y) by the
plane ).,: 6. The existenceof a tangent plane to a surface
requiresmore than the existenceof tangentlinesto two curves
whichare sectionsby perpendicular planes.The next theorem
shows that if we assumethe continuity, and not merely the
existence,of f" andfu, then the differentiabilityof/ is a con-
sequence.
Theorcm8.4. If f,andfrare continuous at a point (x, y) thenf
is dffirentiableat thatpoinr.
Proof. Let h and k be small enoughfor (x+i,y+/c) to lie
within a circular neighbourhoodof (.r,y) in whichl, and l,
exist.Then we have
f(x + h, y + k) -f(x, y)
: {f(x + h, y + k) - f(x, y + k)l + {1f(x,y + k) - f(x, y)\.
In the first bracket,only x is changed;in the second,only y.
Apply the rneanvaluetheoremto eachand we obtain
hfi(x + 0rh, y + k) + kfu@,y + ?rk).
By continr.rityof f" ^ndIt this is equalto
h{f"(x, y) + e,}+ kffi(x, y) + en},
wherethe e's tend to 0 as (1, ft) + (0, 0), and the definitionof
differentiabilityis satisfied.
Exerclses8 (6)
Notes on theseexerciset are gioen on p. 182,
l. Investigatewhetherthe following functionsare diFereotiableat (0, O):
(i) lrr-2t1, (ii.1lxyl*.
, t.
1xyl ,,/Q24- yr) when (x, /) + (0, 0),
't(x' /) =
io when (x, /) = (0, o),
jnvestigatefor (0, 0), (i) continuity ofli (ii) exisrence
of/,,i1(iii) differenti-
ability ofl
3, Gjve alr exatnpleof f(x,y) which is differentiableat (0,0) and dis-
continuousat all other points,
f58 ruNcrroNs oF sBvBnAL VARTABLE8 IE.5
8.5. Compositefunctions
We now extend to functions of two variablesthe formula
dy dy du
dx du dx
provedin $4.2(0.
Theorem85. If x - x(t) andy : y(t) arediferentiablefimctions
of t for a giuen t, ond z : z(x, y) is a diferentiablefunction of
(x, y) f* the corresponding(x, y), then
z = z{x(t),y(t)}
is a diferentiablefunction of t and
dz 0z dx 0z dv
a: ua+Tyii'
Proof. Let t be changedto t + 6t; let 8.r and 0y be the corre-
spondingchangesin x and y.
Then ^ = ldx \^
dx dt,
\A+et)
liv \
6' y : l \ + e 2'1l t t ,
\dt
wheree, and 2tend to 0 as d, + 0. Writing
8z : z(x+ 8x,y * 8y)- z(x,y),
we have,sincez is differentiable,
" :
8z --1*8x+
0 z^.02,
fi6t +e(llxl+layl),
where e -+ 0 as (dx, dy) * (0, 0).
If dx and dy are both 0 (which may happen for arbitrarily
small 6t if dxldt : dyldt - 0), e fails to be determined by the
last equationand we definee to be 0.
Substituting for dx and dy in terms of 81,we obtain
a : (%*+a34\
at oy dt
u+rtt.
\ox I
whereit is easilyseenthat ? + 0 as 8, + 0, I
8.51 r u N c r r o N s o F S E V E R A LV A R T A B L E S 159
oa{-s'\9# -+ T;
-Y l*"u# - sin
reader
There is a similar expressionfor trVl|yz, which the
should write down' Adding, he will find that
to be homogeneousof degreeI
Definition. /(x, y) is homogeneousol degreeft if' for every
oositivet and all x, y for which/is defined'
firc,o:.u{*r'{=i(t'{r.r't),
and, by hypothesis,the last expressionis hf(, ri)/t. So, if
u : f(E,D, ldo h
DAI T
8 (c)
Exercises
Noles on these exercises are giuen on p. 182.
l, (The lacobian.) Lei ll,I, be differentiable functions of r, /. The
deteminant
l' "l
loa 4l
a@,r)
writterl shortly as
aG,y)
is called the Jacobianof the transformationfrom (x, /) to (a, ,). Its role
correspondsto that of the derivativedultlx for a function a(x) of a single
variable. Prove the following propertieswhich illustrate this role.
(i) If a relation C(a,u) = 3 1to16r1or all (x, y) in a domain D, then
:0
=o#
,W,
reduces to
and hence solvc the equation.
ffi=o,
6. lf/(x, t) is transformed into f,(2, u) by thc transformatioD
x = ur_tf, y=2uo,
prove that
#.#,=d$ffi.^H.
Find the most general function/(x, y) which satisfie
a_?f
a:,'att=o
and is a function of r+,r,(x!+/) only,
7. Prove that, if r and d are polar coordinates and l =
log /, the cquation
a.v arv
af+aY'=o
bcomes t! -,!v = o.
6F+8F
8. For all positive vatues of \ the function g
satisfiesthe condition that
s'(,\'x, ,\?) = i?(x,,),
whcre ,,, J, |' are positive intcgers, prove that
^ff+vf;r=,g.
8.61 F U N c r r o N s o F S E V E R A Lv A n r A B L E s 163
9, The function F(4 r) becomes/(x, /) when the substitution
u = x3-3xy2,
u = 3x.y_ y.
is made. Prove that
I la,f atf\ = a2F a2F
g!r+rrj" \bx,+ay,l ar,* u".
rot=(nft+r{)n"o,
/ 22 A2 A2\
F"Q\ = l hz-+zhk
oa"
--l -fb, b't,
=, + kzoo'l
oaoD
\
and so on. Finally, F"(d) hasthevaluestatedin the theorem. I
The possibleextensionof this result,on the lines of theorem
5.8, to provide an expansionof f(x, y) in an infinite seriesof
powers of x and y is lessimportant than the expansionof/(x)
ts Zanx^,
v : +.lQ-xz),
y hasa derivative
givenby
.Y
''' , , _-_
J(t-;r')'
It is often inconvenientor impossibleto solve an equation
F(x,l) : 0 explicitly for y. The iollorving theorem givescon-
ditions under which we can nevertheless assertthat the equation
definesa functiony : /(x) and, further, that the functioncan be
differentiated.
F-+ - d x : o.
' F..4!
0t& and QrRn such that F > 0 for all points on QrQ2 and
F < 0 for all points on RrRr.
l.et now pR be any vertical line between QtRt and QsR2,
By (iii), F increases(strictly) from a negativevalue at R to a
positivevalue at p. SinceF is continuous,there is a unique
point P at which .F : 0. The ordinate of P definesy : /(x), and
(iv) is established.
Fis. 5
Exercises8 (d)
Notes on theseexercisesare giDenon p. 783.
l. Prove that, if A, B, C arc the anglesof a triangle
cosI +21(cos,+cos C) < 2.
2. Investigatemaxima and minina of
(i) (x- y)t (a2-xr-yr),
(ii) (x - ),) (x -2y,).
1. It f(x, y') = x3+2x,y-xy2-8y?, show that there is only one point
(xs,y.) where necessaryconditions for a maximum or minimum arc
satisfid. By consideringthe valuesoflon the line y = 1, prove that it
has no maximum or minimum.
4. If u = xt+yt, o = xs+y',
prove that, if x is consideredas a function of a, u,
A x v A- x l
= =
at) 2xG:i' at iG-r'
,/- i,\
Prove that ":;:2 = - I
a@.u) 6xy(x - y)'
5. lf f(x, y, z) = 0, where/ is a differentiablefunction of x, /, z, prove
that
laz\ lAx\ /Ay\ -t'
\ail"\a'1"\axl'=
where (0zl0y)" denotes the derivative of z with respect to / when r is
constant.
6. The equations f(x, y, z, w) = Q,
g(x' Y' 2' w) = O'
where/and g are differentiable,can be solvedto give z and w as functions
of x, /. ProYcthat
a3=_{r,s>1a1tr9 az__a(f.s) l4!.s,
ax a@,v)l 0(2,w)' at aU), a\2, w)l t!)'
Calq'rlate azlax, Az/a/ as furctions ot x, y, z, w if
xY+zte = O,
x2+y2-22-v2= l.
170
Ansryers or hints for solution are given to those exercissin which they
appear to be most useful. If an exerciseembodis an important result or
an instructive method, the solution is set out in more detail.
1 (o)
1. Write I(a), r(n) for the expresions on left and right sides. Assume, for
some z, I(z) = r(n). Then
,(z+ 1) = (r)+(r+ 1)r = r(n)+(n+ 1)l
= +(n+1) (n+2) (bt+ 3) = r(r+ 1).
But r(l) = I = r(l). Hence (z) = r(r) for all n.
2. Right-hand side can be found from 1, since
1(r)
3. Supposethat a/, is a root, where a and 6 arreintegerswithout a @mmon
factor, and 6 > 0. Write a/D for x and multiply by 6'-1. This givcs i = 1.
4. 12, -3,4i -r, t, z.
1 (c)
r. 99170,2391169.
3. (a- c+^lb)' = dgivesa = c or Jt rational.
6. ad = bc.
r(a
3. If /4 and B are sets of positive lumbrs, then sup D = sup,{ sup I
and inf D = inf A inf B. If I or B may contain nogative numbers, there
is no such simple result.
5. If the a's are not all equal, there are two a,, a. for which a, < A < a,.
Replace a,, a. by 6,,4 where ,, = A, b, = qr+ q.- A. T\eo.
b,b,- +a, = (A-a) (a,- A) > O.
So the replacement haskept the A.M.the sameand increasdthe c.M. After
at most n - I repetitions of the argument all the numbers are replaced
by ,4, The c,M., now equal to ,4, having been increasedby replacements,
was at first less thao l.
N O T E SO N T I I E E X E R C I S E S I7I
6. Use the identity
(',q,b,), = zolzbl-2(a,b,- a,b,),,
wherer, J take the values,1,2, ..., r,.
7. Prove that, for the positive integerr,
a
- i"T+ T
\_l
> a"_l
n'
Dividing by the positive factor a- I and then multiplying up, we have
proved this inequality if
n a " > a " - r + a " - 2 +. . . + l ,
which is true.
It follows that.' if z > n.
a^-l a"-l
mn
To extend to rational indices, let r = mlp and s = nlp (with m,n,p
positive integers), Pat art' = b.
8. Similar to 7.
| (e,
l. a-c = i(d-D) gives (a- c'12 = -(d-[)2. From the order properties
of real numbers,left-handside > O and right-hand side < 0.
3. (d+ bi) (c+ di, = O givesac - bd = ad+ bc = O- Hence
(a' + bl) (c2+ d,) = (ac- bd)r 4 (ad + be), = O.
Therefor a'+b2 = O or cr+d2 = O.
r(n
2, Circle,line, circle, hyperbola foci 1 l, conkadicts theorm l.lo (2).
3. (iii) Combining the conjugate2rr+zt = 3 with the original equation
wehaye z, = Z3and then zs = l.
4. Put z+p = Z.
5. If lzl < ], then l4rz+,...+a.z"l < 2{J+...+(})"} < l.
6. Circle.
8. If there is a root with lzl = l, then 2 = llz, and so crz+bt+a = O.
ConjugategivesZz,+62 + a = O. Combine with original equation to give
z(ab-bi)+aa-cZ = O.
For converse, write w: (6c-Ab\l@n-ca), so that lwl = t. aw+b
reducesto -cw. B\t w = |lw.
9. Roots of quadralic are t)+ 1)2
+ t)aand its conjugateu.+ l' + r!. Form
sum and product.
12. Proye first for z P(z).
13. Use 12.
t72 NOTES ON THE EXIRCISIS
2 (a)
r. t+ (- +)..
5. s" = 17111o* 11*Ur,* O,
8. If r +.r', 1uLe6=]ls-s'1. lN. ls"-sl < e andl,ro-.r'l< eforall
,r > lV. Contr&diction.
2 (b)
&10. All false. Counterexamplefor 9, .r, = 2141- 1;'n.
2 (c)
2. Limit 0, aolbo,*a, -co accordingas p <q,p=q,p> qanda6lbg
positive, p > q and aolbor',egfa,tiye.
2 (d)
6. By the binomial theorem
rr.., (t-;)"'('--,')
increasesas n increases. Moreover. the number of terms increaseswith r.
Therefore,s. increases,
From the above expansion
s "< r + r + r 1 + { + . . . + 1
< r + t + l r +j + . . . + ) < t .
By theorem 2.6, r" tends to a limit e, where 2 < e < 3.
t. a+r"(z+ 1) <fnif (n+1) <2"+', i.. if (l+7-9" < z which is true
ifz > 3 (proved in last exercise).
Write {h = I+r.Thenz = (l+x)" > 1++z(n- 1)xr,andso xt <2ln.
, H e n c ex + 0 a s n + o .
.. lt, u, = 0if x = 0 orm a positiveinteger. Otherwise
lu.*rlu^l -+ lxl.
12. (i) 3. (ii) 1,0, -1 accordingas a >, -, < b.
nn
(Iu, ,t > > ,
d+-i "" n, + n.
(iv) tt! > an for rr > /V; d(a!) -+ co.
(v) Method of 2 (a), 5. Limit is - t(q+ br.
(vi) Use 2 (4, r,
NOTES ON THE EXERCISES t73
2 (.,
By methodsof $8,limitsare
l, 4. 2. l. 3, Thesmallerroot.
4, 2. if,r. = f.
5. 3, but s. is undefrned
6.2.
7. I = r(l + l2l), if,*,-l'= u"-1,
u"*,-l = (u"-l)l@"t+ l) < (u"-l)121, 412> 30.
a, t,2,3.
9. a" ) an+t ) bn+r > 6n and an+r- bn+t < ,@"- b").
Also analbor, = snb..
2Ul
l. l/zt < l/(z- l)a givcs
Hl I
t N '
'1'7
Hcnce N = 10',
For t (0.99)",we have from (0 99)x < l0-.(l -0.99), N > 1376.
The serond seriesconvergesmore rapidly than the first.
2. First is c.p. convergentfor r > 0 to sum l*r;alsoforr = 0 to sum 0.
For thir4 if ,,
so =) zr.r,
2 (s)
l. General. lf A, B are statements,the notation A + B me ns'lf A,
thcn B'. We say also that (i),4 implies 4 or (ii) I is a sufficientcondition
for 4 or (iii) I is a necessary condition for,l.
The double-headedanow Ao B means that ,4 and , are logically
equivalentto each other, in other words: (i) lf and only if ,4, then 8;
(ii) ,4 is a necessaryand sufficientcondition for 8.
In the text we shall nearly always write words insteadof thc symbols
(r-;). (r-fr)
/ l\n ,/ | \tr+r
t a k ea 1 = 4 r = . . . = 4e = l-0/r),4n+r = I anduse
c.M. of (ar, .,., a,+J < l.v.
,n-rn = {l -(l -z-")"}/(1-tt")".
To prove numerator -+ 0, use (l - rr)n > 1-nli.
7. .r" = (- l)",. r" = (- l)" or (- l)"+1.
E. 6(rr- l)/9(n- 1)r (z+ l).
10. If r" -+ l, , satisnes,+(l/r) = 2,4. If roots of this are real, A > l. lf
c is the larger root, r"-c : (r"a-c)!r"-rc, and induction.
11. A sequencosillating more slowly as n gets larger, like
s" = 2+sir- (, n.
If rn*t/s" -+ -|, then .r" + 0.
12, Last part, u7 < @+t u7 < (l +Dai.
13. (i) True. (ii) False; we can say only r > t. Illustration of r = t
given by -rn = s"=lln. (iii) Falsity shown by r"=nr (r evcn),
s" = (r- l)r (z odd); then s" '+ co.
14. Write r" = (rr+s!+..,+d/2.
Given e, s- < sn < J+ for r > m,
Sum from m+l to n
(z-ra) (.r-e) < ,ran-mt^ < (r-m) (r+er.
Divide by n and rearrange
I.m\.,mI.m\..m --,
rd < r' < ($+er+; ri.
\r-;/(r-r+; \t
Kceping n fixed, we can choose aosuch that the first expression is greater
than J-2 and the last lssthan J+2.
so s-;l' < t. < J+2 for r > ,ro,
15. Sum of, terms is
I -x" nxn
1t-xy-1-"'
3 (r)
4. It is assumed(aDticipatingchapter O that the sine is continuous. lf
x + 0, sin (l/x) is continuous by theorem 3.5, and /(.r) is the product of
two continuous fuDctions. Continuity at x = 0 follows from the definition ;
l/(x)l < e if lxl <e.
N O T E SO N T H E E X E R C I S E S I75
5, If ,t - l, there is only one value of r.
lf k + I, f(x) = k eives
(l - k)x'9- (6-9k)x + 5 - l8k = 0.
This givesreal r if (6-9k)r > 4(l-&) (5-18&), i.e. if 9k -t6k+16 > 0
which is true for all &.
8. (i) x- I is a common factor.
9. (i) Multiply numerator and deDominarorby J(l +r)+y'(l -.r).
(iii) Case p < c. lf q-p is even, limit + co or - co according as
aolbo> or < 0. If 4-? is odd, considerx + 0+ and r + 0- separately,
10.(i) Wewant 1t\
t o r\t
\ P + " r + V ) ( 1++F: ,
to be as close as possibleto llx for large r. Equate coemcientsof l/xr
for& = 0,1,2.p = 9,q = I,r -_ _1.
(ii) As in (i), squaringto get rid of J. Or, anticipatilg 95.8,expand
{l + (4/x)F}.
3 (c)
1. (i) Continuousifx is irrational.
(ii) Discontinuousfor x = nn (n 4 O).
(iii) Discontinuousfor ll@- a) = ntr.
2, f(x) = I -tfor0 < x < l,/(0) = 0,/(l) = I org(x) = .r(rrational),
c(r) = 1-x (x irrational).
4. g(r) of 2 is continuousfor .r = 1 only,
5. l0-r. 6. (r),r(0) = 0, (iii) /(0) = c.
7. I6+d)-/(r) < tr{/(.r+ 2d)-/(x)}...
I
< + 2 6) -f(x)}.
2 {.^x
where4<x cb,a< x + 2 8 < b . L , e dt + 0 a n dz - +c o .
8, Iro-xr-rl { a"-11x.- xnl.
i(*"-",-J is (absolutely) i.e.x" -+g.
oonvergent,
1
= l"/(0-.r(0)l< al(l
16-,r(o)l
gives -eE < 6-l(0) < dE,etc.
4 (a)
r. (i)y:4(x-2),y = -4(x+2). (ii)r = -+.
2. y-2= -Ix,y-2 = '(x+4).
3. y = 3x. s. y = lr- ll+ lx+ 11.
6. (i) (a) and (6) x = integer, (ii) (6)r = l.
7. (i) Touchat (0,0),arctan l8 at (t,3).
(ii) li at both points.
176 NOTBS ON THA EXARCISIS
4 (b,
9. Writep(r) = (x - a)i (x - b\t' q(r), where a and 6 are conscutiveroots
of p(.r) = 0. Prove that, if p'(.r) = (x - d^a G - b)n-r r(x), then r(a) and
r(6) have opposite signs.
10. If r+ o,f'(x\ = 2x sin (Ux)-cos (l/x), by the rules of$2.
Ifr = 0, !r(r) -/(0))/ h = hsin(llh) -+O, siving/'(0) = 0.
Examples like this arc repestedly usd to settle some of the Iss asy
questions which occur in differentiation. This one gives the answer yer to
the questioq (in geometrical language!-can a curve have a taogent at
every point and yet the direction of the tangentnot vary continuously?
12. DiFcrentiate four times, using 1l and rejecting vanishing determinants.
4 (c)
1, 2,5. Put into partial fractions,
4. sin 3x sin 5.r = Xcos 2r-cos 8x), and use 3,
6. The only way of doing this systematically is to use complex partial
fractions
a = rl I I \
.
oza *t li \j=-;a- lllal
The zth derivative is
(--Zt
-l)'a! I I I I
k;=Fri-G-+ia)+i'
If r = ,v/(t'+ ar) and cos A = xh, sin 0 = 4//, the! de Moivro's
theorem (l (e), 5) giYes
(x+ ia)'*' = /"+r {cos(n+ 1) d+i sin (a+ l)d},
etc. The result finally takes the real form
(- 1)' n ! /{"+D sin (a+ l)d,
whero r, A have the assignedvalus.
4 (d)
2. Method of 4 (r), 9.
5. At most one real root ofp(r) = 0 lies betwentwo consecutiveroots of
P'(x) = o.
ln the example, if z is odd, p(x) = -(l +xi)/(l +r) < 0 for all .r.
p(x) decreasesas x increasesfrom large - to large + values. p(x) = 0
for one x.
If z is even, p'(r) < 0 if * < I, p'(x) > 0 if x > I andll) > 0.
6. Us first part ofs.
4 (c)
2. Limit *.
3. We may take / = 0 (and apply the special casto /(r)-rr).
Given e, choose X to make lf'(x'11 < e for x > X.
f(x)-f(x) = (x- x) !'(c).
NOTES ON TIiA EXERCISBS I71
5 (4)
l. C,C,Cforx < l, Cforx <4,Cforx<2,
5. b"lan -> @, a"+1|b"-+O,
6. lf u" = ,r-h, then 4u^ -+ I and u^nlun -+ 1,
8, (i) Z; (ii) F; (iii) F Ge3).
5 (r)
2. k>1,k>o.
3. (i) None;(ii) -1.
a, O fu"| + +, thereforeD; (ii) abs.C; (iii) C by rheorem5.22.
l7a NorEs oN THB ExERrsEg
5. a=b,C. a+b,D.
6, Sumsof 5 and 6 termsare f$ and f$,
7, (i) Thehypothesisin theorem5,22that 4odcreass
is omitted.To prov
the statemeDtfalse,coDsider
q. = llttt+(-1)ln'
(ii) llr"l < ,{ le"l. Hencct lr,"l convrgs.
3 (c)
r. tu2uDz.
2. lzl < ll - zl.
3. (i) Theorem 5.22to re and im parts. (ii) lr"l = 21. r-'-+ co. (iii) |l" is
lln or iln accordiag as z is even or odd.
5(A
r. (i) 1,(iD all z, (iii) 1, (iv) Q (v) 2, (vi) 1,(vii) 1,(viii) *.
2. a<l,Callz.
5. X> 1,R=1.
6, Uso theoreorl.l0 (sum).
7. lf r> 4 radiusof convergenco
is r. Ur=r, any trumber> r Givc
illustrations).
E.(r-dLt
- = ,,":.".,-z^*',
I ,t "-^2'
1 (n+l) r|t
For lzl = I, this tendsto a finitc limit,
9. X = *, If lzl = j!, modulusof nth term > 1, and so seriesdivergsson
thc circle.
5 (c)
l. (i) C; Qi) C if c < max c)i (iii) C if & > *; (ivJC ll k > 0; (v) D;
(r,
(vi) C.
^'' l l 2 =
a-l- o*l ar=i'
3 . ( n - m ) u ,< i *.3 u . . + O a s m - + @ .P o tm = * n ( i f r r e v e no) r
,n+l ''+l
t(r + l) (if ,t odd).
4. Of tarms v,/ith m-digit dnominators, the number remaining is 8 x 9.-1.
Sum is lese than
A g 9t {F_r \
= 80.
8(i+m+lo,+...+fu:i+...,
5. Thcorem5.7.
N O T E SO N T H E E X E R C T S I S t79
6 (a)
l. II e = |fu, 111s11
^(t -i.!\ =,"ri 1.l.
\m o nll mit nl m'
and so an integeris lessthan 1/m. Contradiction.
2. Included in 3.
3. The argumentof 2(9),6, showsthat, if r > x > e
(r*1".""n"<(r-:)-".
(The inequalitiesfor x < 0 will then follow by putting r = -/.)
Now we Drovcthat
I r --;/
1 1 - " +-/" ' '+ u.
\'I \'I
| ^n lt vh-n
= (r+1J
L.H,s. - rj
t(t-;)
//
< e x r x { ( r,rr\-r
_ 1_J.ll = f.'rf icxDx
(by use of easyinequality
ll - (b/,|)l-" < (l-61-t when 0<6<
l).
4, fi6) from l- k to - k.
< 0 for x > 0, and so/n decreases
-r+l--r
h+,G)= I"U)+ttrfr. gives/"*(x")> 0.
5. 9120. 6, Descending
orderasgiven.
7. exp (r/x).
6 (D)
t ' A si ' - r o '
i-t exD(rlosx)-l
n=:--__i-*toe,.
2. The derivativessatisfy'
1-x<(l+x)-r<1-x+x2.
3. yf'(xy) = /'(.r) and xf'(xy) = /'(y).
t. tt x = (y-DlQ+ l), L.H.s.= los (t +x)-loc (l -x).
5. (i) a,0; (ii) log,(alb)llog (cld). For x.> co, several cases, e.g. if
max (a, b, c, d) is a,limit is co.
6 (c)
l. Take four terms of seriesfor cos .r when x = * and three terms when
x=t.
2. From theorem6.81, sin.n > 0 for 0 < r < 1z and from 6.82(l),
t2-?
180 Norts oN TIIE ExERclsas
s i n r > O f o r t n < x < z . F r o m 6 . 8 2( 2 ) , s i n x < 0 f o r r < x < 2 n .
So sin (x+c) = sin .r cannot be true for all .r if c < 2z'
4. (logcosay)/l-+ -rd" (e.C.by $4.9). Puty= r/a
6 (d)
1. cosh (a + r/) sinh (r + 1) P cosech/.
2. Method of $2.8.
3, Real and imaginary pa(s of geometric serieslr" exp (rnd).
4. Analogous formulae valid for lrl < e-0 if 0 > O.
5. Induction.
6. (i) Use cos 3r. (ii) r+*ri+*rl.
(iii) (l-x')y'-xy' = 2. Leibniz gives
(l - xt) r{n+tt- (2n + l) xln+r't - n2lnt = O.
Putting r = 0, we obtain Maclaurin coefficients,
(iv) Method of (iii). Generalterm is
(-l)" m(m'-12) (m'-3', ..,ln'-(2t-l)1x!i+r/(22+ 1)!
(v) exp (l +2,)x - >(l +2i) x"lnl
D e f i n e d b y c o s d= l l " l 5 , s i n a = 2 1 1 5 .
Ral part is ,5li cos n?(x"ln!).
(vi) 1-lx'+}.ll.
7 (s)
2. Take E to be left-hand end-point of 6,,
1*.- *-,,
2{rrr, =,t @q*,1*
= 4.+r(4- 1) tq{'-Dt&+l)
= ar+\q - l) (qntt+rt- lj l(fr - 1)
= (b.+r - 4!+1) (4 - l)l(f, - lr.
{s a + oo,4 -+ I and (qr+r- 1)/(4-1) + &+1.
5. Civen , tbere is, by definition of .I, a dissection 90 for which
J<.t(90)<J+e.
I-et 9o li^avep points of division inside (a, ,). tJt I & a\y dissction,
with norm 6*, and 4r the dissction formed by all points of division of
9o arld 9.
'1.21)
Then (theorem S(9) <,s(gJ. Also, since 91 is formed from 9
by, extra points of division,
s(e)-S(eJ < p(M-m)8'.
Hence S(9) < J+e+p(M-m)8'.
$3. < el(M-m)p, then J < S(9) < J+ 26.
NOTES ON TIIB EXARCISBS 181
7 (b)
Most of the exercisesin this and the next set depcnd directly on the
mcthodsof $S7,7aod 7,6.
3, In (0, +lr), sio2a-r x > sinr' r > sintd+r r,
giving I,^-t ) Ir^ > -L.+r (in fact, >).
Divide by I1.*1 and us (2rr+ l) ,rnrr = 2m lz^-r.
7 (c)
6. 2^rc tan {(l +r) tan lrl(l -r)}+, (r<l),
-2arc tan {(r+ l) tan ltl(r - l)} +t (r> l).
L i m i t sr + r ( r < l ) , - z * d ( r > l). 1= Swhenr= l,
7 (d)
2, First sum lies between
[2n dx r2n )e
I --:: and | :
Jn x ttu+r x
sqd se + log 2. Secondsum -+ 0 by theorem5.22.
H and E follow from theorem7.11.
7, The idea of theorem 7.ll used aod applied to an increa$ngf.
9. (i) By $6.6,ifd > 0, (log z)a/r/ + 0. Ift < l, choosed = tr(l -?), say.
E(l/n,+,) diverges.Thereforeso docs
t4o
_
np+, (logn)q.
Similarly, convergence ifp > 1. Ifp = l, we have 6.
(ii) (log log z)t < log n if z > no. Divergent fot aU ?.
7 <e,
6. (a) Since y' < 0, tbe curve / = log t is concrve to O-r.
(i) First inequalityxpresses that the area uDderthe curve / = log r
betweenx = r-*, x = r+l is less than area of trapezium formed by
y - O,x = r-1, x = r*l and tangentat x = t
7U')
1. (i) d2, from theorem7.62. (ii) l.
lu Il
2. sU\ = (x-l) l Io U(A dE+x IIt G-Dl@dE.
Using theorem7.62
f, - ^.^ tr
e'Q)= I u f@de+J,G-rt I@ d,.
8(0)=8(l)=0.
182 NOTBS ON TITB BXERCISBS
s.tDf;s;
(ii)rogl(a)= -
I {t* (t-)....+r"e (r*)} ltorcu{t+x)dx.
6. Mcthodof 5.
7. Integrateby parts. lnduction, For last part,
E(o)
1. No, !o, ys,
8(r)
1. (i)Ys. (D lfe > l.
2. Yes,yes,no.
3. (Cf. 3 (c), 4.) If /(-r) = A for .r irrational and xi for .r rational,
f'(x) = Oat x = o and/is discontinuousfor all other x. To constructan
analogousfunction g(r, /) of two variables, de6ne g(.r,f) =/(r) if
*r+yt = 4 1'.ur1u* of rwolution),
t (c)
1. (i) 4"u"+0.a. = o,
6"u'+i.tt' = O
havesolutioB othcr rhan (O 0) for Cr, C.
(ii) Multiplication of determiosnts.
2. u = f(x)+fbt), o = (x+y)l(r-xy).
3. A = |ar sin8sin C/sin,4.
,tA ,r^a
;I t a = :=+cot +@t C 6c-ctr,tA tA
"'
and tA+EB+tC = O.
4. Method of 3, AAis gratrstwhcn
8e> Q,8b> g Ec> 0 (acute-angled),
or 8a < O,8b > Q dc > 0 (obtuseat ,{).
5. / = /(.r)+j'(u), wher, g ar arbitary (difrrntiablc)functions.
6 (also7, 9). TransfomEtionas in S8.6.
rl = x+^l(f+tD, s = Au+8.
E. As in $E.6.
NOTBS ON THE EXERCISES 183
8 (d)
l Put -cos (8+C) for cos l. Necesary condition for turning value is
.8= C. Maximum of -cos 2r+ 23ltcos I given by I = +2.
2. (i) Maxima at (t '4, + +ll).
(ii) f" = f" = 0 at (0, 0) This is not a maximum or minimum
because,near (O,O), f < 0 between the parabolas y2 = x, 2f = x ard,
> 0 elsewhere.
3. (0,0).
5. Mcthod of 6.
6. dx, 8y, dz, 6l? satisfy (approximaiely)
f" 6x+f' 3r+f,6z+f" 3w= O
and similarly for g.
Puf 6/ = 0, solve for dz/r.r and let 0r + 0.
zN-w"
185
INDEX
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: