Академический Документы
Профессиональный Документы
Культура Документы
A complete list of books in the series appears on the inside back co~
COMPLEX NUMBERS
LIBRARY OF MATHEMATICS
edited by
WALTER LEDERMANN
D.Sc., Ph.D., F.R.S.Ed., Professor of
Mathematics, University of Sussex
WALTER LEDERMANN
t The Teaching of Algebra in Sixth Forms, Chapter 3. (G. Bell & Son"
Ltd., London, 1957.)
v
Contents
Preface page v
2. Geometrical Representations
3. Roots of Unity 33
Answers to Exercises 60
Index 62
vi
CHAPTER ONE
Algebraic Theory of Complex Numbers
1. NUMBE.R SYSTEMS
~1------~I---4I--+I----~I-+I---I~·--+------+I------+I--~~
-i -2 -i -I -t 0 -f 2 3
Figure I
Examples.
li(1+')'1=4.
l =l-i l =l, 1- 51=5. 11+il=v(lI+12)=v2.
Itan 8+il=v{(tan 8)2+12}=lsec 81'
We can now describe a simple solution to the prob em of
division. Let o:=a+ib and {3=c+id be given complex
numbers and suppose that O::F O. It is required to find a
number e=x+,y such that
""=/J (cx~O) (1.12)
8
THE ALGEBRAIC THEORY
Let us assume for a moment that there exists sut..h a num-
e.
ber Then on multiplying (1.12) by ex we find that exiif=
a.p, that is
(al +b2)(x+iy) =(a- ib)(e+id) =(ae+bd) +i(ad- be).
On comparing real and imaginary parts on both sides we
see that
ae+bd ad- be
X= a2+b2' y= a2 +b2'
Thus if there is a solution at all, it must be given by
ae+bd .ad- be 1
f= a2+b2 +, a2+b2 = Iex I2ciP· (1.13)
Conversely, it is easy to see that (1.13) does in fact satisfy
(1.12). Indeed, exl~liiiP=I:12P=P. The unique solution,
exhibited in (1.13), will be written P/ex or Pex- l , or ex- l p.
There is no need to memorize the explicit formula for P/ex.
The argument which shows that there is such a complex
number is equivalent to the familiar 'rationalization of
denominators' in working with surds. Indeed, it should be
borne in mind that, after all, i=v -lis a surd. To simplify
a complex fraction we multiply numerator and denomina-
tor by the conjugate complex of the denominator, dius
e+id (e+id) (a- ib) (ae+bd)+i(ad- be) (1.14)
a+ib (a+ib)(a- ib) a2 +b2
which is equivalent to (1.13). In particular, we note that if
ex rF 0,
(1.15)
Example 1.
_1_· +_1_ 1- i + 1+21
1+; 1- 2i (1+,)(1-,) (1- 2,)(1+2,)
1- i 1+2i 7 1.
=2+-5-=10-10'·
Example 3.
1- ," 1- (;')1
1+;+,"1+;'+;'+;I+;'+i'=--. 1-, = 1-,. o.
Let f(x)=a~+alx"-l+ ... +all_1x+a,. be a poly-
nominal with real coefficients a o, all az, ... , all. If we sub-
stitute for x a complex number ex, we obtain the number
f(ex) =aoexll +a1cxll - 1+aacx,.-I +... +all_1ex+a.. (1.16)
We now wish to findf(Ci). Bya repeated application of (1.6)
and (1.7) we may do this by placing a bar across each term
on the right-hand side of (1.16) and across all the factors
of this term. But since the coefficients are real Qo=ao,
10
THE ALGEBRAIC THEORY
a1=a1, • Hence f(a)=ao(Ci)"+al(ii),,-l+ . • . +a". Or,
briefly,
f(a) f(ii). (1.17)
Suppose now that f(a) =0. Then also f(a)=O and (1.17)
implies that f(a) =0. Thus we have proved the important
result that if f(x) =0 is an equation with real coefficients, its
roots are either real numbers or occur in pairs of conjugat,
complex numbers a, ii.
We have so far confined ourselves to the four rules of
arithmetic for complex numbers. A systematic study of
more complicated functions will be deferred until geo-
metrical and analytical ideas have been introduced into the
subject. However, the extraction of square roots is an ele-
mentary process which can be discussed already at this
stage.
Let a=a+ib be a given complex number. It is required
to find a number ~=x+iy such that ~=a, that is
(x+iy)2=a+ib.
On splitting this equation into real and imaginary com-
ponents, we have that
r- y2=a, 2xy=b. (1.18)
Suppose first that such a number ~ exists. By taking
moduli, we deduce that lal = I~I = Iell, whence on squar-
ing <I eI2)2=laI2, that is
(r+y2)2=a 2 +b2. (1.19)
Combining this with (1.18) we find that
r=!(a+v'(a2 +b2) ), y2=!( - a+v'(as+b2) ).
The expressions on the right-hand sides of these equations
are never negative, so that the extraction of the square roots
yield real values for x and y, as it should. But when choosing
the alternative signs for x and y it must be remembered
that xy=!b. Thus when b is positive, x and y have the
same sign, and when b is negative, they have the opposite
sign. There are therefore exactly two, and not four solu-
tions of the equation ~=a «X:;l: 0), as one would expect: if
11
ALGEBRAIC THEORY OF COMPLEX NUMBrlRS
'1 =X1 +iYl is one solution, then the other solution is
'.= - '1= - Xl- ,yl' In a numerical problem it is usually
best to find the value of one of the unknowns, say :Je, and
then to obtain the other from the equation XY=ib.
I
IXI I(XI •
S· Show that /J =TPi' Fmd the modulus of (l-a1) ' .
(I +2i)1.
13
CHAPTER TWO
Geometrical Representations
x-3+2i. 2i.
t x 2,+i.
-3 -2 -t 2.
x -t
-i- L
-·2L
Fiaure.z
14
GEOMETRICAL REPRESENTATIONS
such a plane) is often called the Argand plane or Argand
diagram or simply the complex plane. Any algebraical pro-
position between complex numbers can be translated into a
geometrical relation between the corresponding points of the
Argand plane, and conversely, any relation between points
of It plane can, in principle, be regarded as a relation be-
tween complex numbers. The reader should, however, keep
the logical situation clearly in mind: we have defined com-
plex numbers as algebraical objects obeying certain laws of
composition. The points of the Argand plane afford a
pictorial representation, which is helpful for many theoreti-
cal and practical purposes. But the representation must not
be confused with the object itself. By way of analogy, tem-
perature is often represented by the height of a column of
mercury, and this is a very useful device. But nobody would
suggest that temperature is a column of mercury. Besides
there are other ways of representing complex numbers (and
temperature).
From now on we shall frequently use the traditional nota-
tion for a complex number by means of a single Latin letter,
thus :r=x+iy or w=u+iv, where x, y, u and v are real.
For the sake of brevity, we often speak of 'the point :r',
when we should say 'the point representing :r'. Since the
x-axis consists of all the real points, it is called the real axis.
Similarly, the y-axis contains all the purely imaginary points.
It is, somewhat misleadingly, called the imaginary axis. (It
is, of course, no more imaginary than the real axis.) The
origin 0 corresponds to the number zero.
A different, but related interpretation of complex num-
bers is obtained, if we associate with the complex number
--+-
x+iy the vector OP, where P=(x, y), or more generally, by
• --+-
the duected segment P1Pz, where PI = (Xl> YI), P 2 =(XZ' YII)
and x a- Xl =x, yz- YI y. Again, if no misunderstanding
can arise, we speak of the vector :r instead of the vector
representing :r, and we shall even commit the convenient
--+-
abuse of language by writing z=OP.
15
GBOMBTRICAL REPRESENTATIONS
Let us translate the formula for the sum of z and w into
vector language, where z=x+iy and w=u+ifJ. We repre-
---+ -+ .
sent z and w by the vectors OP and OQ respectively (see
Fig. 3).
o x
Fiaure 3
o L---~w:-:---
Figure 4
PI, P,. P•• ••. are the points at the ends of In. %., :t•• •••
The vector that joins 0 to the end point of the vector .....
then corresponds to fOe In particular, the relation .8'1 +Zll+
••. +Z,,=O means that the broken line OP1P2 ••• p. is
a closed polygon (O=P,.). .
The difference between two complex numbers can also
be simply represented by the vector diagram. If and oP
oQ correspond to:t and ro respectively, then 'W- .8' is repre-
........
sented by PQ, that is by the vector that joins the end point
of z to the end point of ru.
If :t is represented by the vector oPt
where P=(x,y).
then the length of this vector is given by
l:tl =y'(~+r)·
17
GEOMETRICAL REPRESENTATIONS
--+-
More generally, if z is represented by P 1PI , then Izl =
--+-
P1P., the length of the segment P1P•. For let OP1=tor1 =
--+-
Xl +;Y1' OP.=X.+;Ya, Z=%I- %1' Then 1%11=(xa- XJ2+
(y.- Y1)I, that is 1%1 =P1P., the length of the segment.
If ,\ is real, the vector ,\z is parallel to the vector z and
its length is ,\ times that of %. If '\>0, the vectolS % and M
have the same direction, whilst if ,\ <0, their directions are
opposite to each other (see Fig. 5, where ,\>1). In particu-
lar, note that 1- zl=I%I. If %:;'0, the vector I!I% is a unit
vector (that is a vector of unit length) and has the same
direction as %.
Figure 5
Figure 6
Z=X+ly
o x
Figure 7
21
GEOMETRICAL REPRESENTATIONS
Note that
(2.8)
which immediately gives us a geometrical interpretation of
the modulus: 103'1 is the distance of the point 03' from the
origin. The angle 8 will always be measured in radians,
unless the contrary is stated. Its value is, however, not com-
pletely determined by the equations (2.7), since arbitrary
whole multiples of 211 can evidently be added or subtracted
from it. In order to obviate this ambiguity, we impose the
further condition that
(2.9)
For a given non-zero number 03' there exists one and only
one value of 8 which satisfies (2.7) and (2.9). This value is
called the argument of 03', and we write
8=arg o3'. (2.10)
We do not define arg O. The relation tan 8 ylx, which is an
immediate consequence of (2.7), does not determine 8
uniquely, if, as is customary the function tan-It is defined
in such a way that, for all t,
- t11 ~ tan-It ~ 111.
Generally, the value of tan-I(Ylx) is equal either to 8 or to
8+11 or to 8- 11, and it is only after examining the signs of
both x and y and observing the condition (2.9) that we can
decide which of these three values is the correct one.
ElI'amples.
(i) arg 3=0, (ii) arg (- 3)=11, (iii) arg (1+t)=1114,
(iv) arg (- 1- i)=- 11/4. (Note that in both (iii) and (iv)
tan- I (y/x)=11/4.), (v) arg 2i=t11, (vi) arg (- t)= - t11.
In virtue of (2.7) the complex number z=x+iy can now
be written in what we shall call the polar form, that is
o3'=r(cos 8+i sin 8) (2.11)
where r=lzl and 8=arg z. Conversely, if 03' has been ex-
pressed in the form (2.11), where rand 8 are real and satisfy
22
GEOMETRICAL REPRESENTATIONS
the inequalities r ~ 0, and - 7T<8 ~ 7T, then this is the polar
form of z, that is we may infer that r=lzl and 8=arg z.
Next, consider two complex numbers in polar form,
ZI =rt(cos 81+i sin 81), z2=r 2(cos 82 +i sin 8s).
We then have zlz2=r1rS(cos 81+isin 81)(cos 82 +isin 8s)
=r1r2{(cos 81 cos "2- sin "I sin 82)+i(sin 81 cos 82 +
cos 81 sin 82)} so that
zlzS=r1r2{cos (8 1 +82 )+i sin (81+82)). (2.12)
From this equation we immediately deduce the fact, already
known to us, that
IZ l z 21 =r1r 2 =l z lll z 21·
On the other hand, it would not be correct to say that
arg Z1Z2 is equal to 81 +8 2, because this number might fall
outside the range (2.9). All we can assert is that
(2.13)
where k=O or 1 or -I, and it is only on examining condition
(2.9) that we can decide which is the correct value of k.
If z has the polar form (2.11), then
z=r(cos 8- i sin 8)=r{cos (- 8)+i sin (- 8)}.
When z¥- 0, we can use (1.15), that is z-1=r- 2z. Hence
!=!(cos
% ,
8-isin 8). (2.14)
It follows that
_ 1
arg z=arg z= - arg %,
x
-z
Figure 8
24
GEOMETRICAL REPRESENTATIONS
More generally, if a is a fixed non-zero complex number
and s is a complex variahle, we may interpret the product
as (or sa) as an operation to be performed on s, that is the
vector s is carried into the vector@. If a=p (cos o:+i sin 0:)
the operation s-+@ amounts to multiplying the length of
the vector s by p and rotating its direction through an
angle 0:.
o a
Figure 9
that lal =Ibl:# 0 and a:# ±b (see Fig. 9). The diagonals are
then represented by a- band a+b respectively. Our geo-
metrical proposition amounts to the assertion that arg
(a- b)/(a+b) = ± 'IT/2, or, in other words, that q=
(a- b)/{a+b) is a purely imaginary number. The necessary
and sufficient condition for this is that q+q=O. Indeed we
readily find that _
a- b ii- b 2aa- 2M
+
q+q=a+b a+6={a+b){a+6) 0,
because lal =Ibl.
Figure 10
26
GEOMETRICAL REPRESENTATIONS
angle ~ at a point Z of the circumference. The comple-
mentary arc AZ' B is described by the equation
z'-b
arg - - = -
z'- a 1T+ ~
(see Fig. 10).
We now return to the multiplication formula (2.12).
which we generalize as follows. Let Zl. ZB • •.. , z" be a set
of complex numbers, each presented in polar form, say
zt=rt(cos Ot+i sin Ot) (k=1, 2, .... , n).
By a repeated application of (2.12) we find that the product
of these numbers is given by
ZlZ., .. Ztt=
T 1r•... rn{COS (0 1 +b.+ . • . +On) +isin (0 1 +0.+ .. . + Ott)}.
In particular, when Zl =z.= . . . =Zn=Z, we deduce that
ztt=rtt(cos nO+i sin nO). (2.15)
Specializing still further let us take a number Z for which
r=1, thus z=cos O+i sin O. Equation (2.15) then becomes
(cos O+i sin O)n=cos nO+i sin nO (2.16)
This remarkable formula is known as de Moivre's Theorem.
We have so far proved it only when n is a positive integer.
The formula is evidently true when n=O, if as is usual we
equate ,.0 to 1. Next, let n be a negative integer, say n= - q,
where q>O. Then
(cos 8+i sin 0)"= [(cos O+isin 0) q]-l=[COS qO+isinqO]-l
=cos( - qO)+i sine - qO) by (2.14)
=cos nO+i sin nO,
which establishes the validity of (2.16) for negative ".
Summarizing we can say that the "Ih power of the func-
tion cos O+i sin 0 may be obtained simply by multiplying
the argument 0 by n. The reader will know this phenomenon
from the general exponential function. For let !(:c)=a'Y.,
where a and 'Yare constants and a is positive. Then.
evidently, [f(x)]" f(nx). We shall now demonstrate that
cos 8+i sin 8 is in fact an exponential function of a special
kind, and it is only with this knowledge that we can gain •
27
GEOMETRICAL REPRESENTATIONS
real understanding of de Moivre's striking but seemingly
mysterious discovery.
At this point we have to anticipate some results that will
be systematically treated in Chapter Four (p. 43). When
11 is • complex number, the function e' (or exp z, as it is
more conveniently, but more clumsily printed) is defined
by the infinite series
z zll;r3 11"
e'=exp z=1+11+21+31+"'+ nl+···· (2.17)
When 1I=X is real, this series reduces to the well-known
formulat for es. We will not stop here to give an exact
explanation of what is meant by an infinite series with com-
plex terms, and we ask the reader to take it on trust that the
function defined by (2.17) has the properties expected of it.
In particular,
exp(z1 +z.+ . .. +zm) =exp 111 exp z•... exp 11m
whence exp(mz)=(exp z)m. Let us now put 1I=i8, where
8 is real. We then obtain that
,8 i 8 (i8)'iI (i8)S (i8)4 (i8)fI
e =l+n+2T+3T+4T+" '+111+' ..
(2.22)
We have derived these equations only for real (J. But we shall
later see that they are in fact valid for complex values of the
variable (p. 54).
As a further illustration we discuss the use of complex
numbers in the summation of certain trigonometric series.
C+ •. S =e ,ex,'~
.. ''''-1 ,ex+-'"+1-8 e,,,·/I- e-,,,'/I
e e"- 1 =e I e"/I- e-f'/I ,
whence by (2.22),
C . s- ,ex+"+l,O sin n8/2
+, -e I sin 8/2 •
Finally, on separating real and imaginary parts we find that
C n+10) sin nO/2
=cos ( ac+-2- sin 8/2
and
. ( n+10) sin n8/2
S =SID ac+-- 2 sin 0/2 •
32
CHAPTER THREE
Roots of Unity
Figure II
.,).
..
2k",. . 2k",
.....up (2k m .. -cos-+'8.n-.
..
34
ROOTS OF UNITY
'/) 2(n- 1)'17' •• 2(n- 1)'17'
En-1 =exp( 2(n- 1) ."., n =cos
n
+um n
.
are in fact distinct. For any equality between them, say
. 2'ITi(k-l)
Ek=E, (k>l) would lead to the equation exp 1
n
and on comparing real parts we should find that cos 2'IT(k-l)
n
=1. Moreover, every solution of (3.1) is included in this
list. For we have seen that any solution is of the form
exp (2'1Tik/n) where k is an integer which may clearly be
taken to lie in the range 0 ~ k ~ n-1. Notice that if E is an
n C" root of 1, so are all its powers
1, E, E2, ... , EII - 1. (3.3)
However, in general, these numbers are not distinct. An
nth root of unity for which the powers (3.3) are distinct, is
called a primitive nt" root of unity. It follows that if E is a
primitive n'" root of unity, the powers (3.03) constitute a
complete set of solutions. Whatever the value of n, the
number
(3.4)
is a primitive n C" root of unity, because its successive powers
are identical with the solutions listed in (3.2). In fact, Ek=
exp (2k'lTi/n)=(exp (2'1Ti/n»k=Et(k=O, 1, ... , n- 1). When
the nt" roots of unity are plotted in the complex plane, they
form the vertices of a regular n-gon inscribed in the unit
circle, one vertex being at the point 1 on the real axis. The
case n=6 is illustrated in Fig. 12.
sm 5= -'8 9 +. 7 I.
E.=COS
4'17' ••
T+' 4'17' 0 0 58 8 .
35
ROOTS OF UNITY
-I 1
Figure Ia
I
by multiplying out the brackets on the right, preferably
starting with the last two brackets. A neater way, which
uses simple properties of d~er~in;ntsl is as follows: let
A= y x z
z y x
By expanding the determinant in the usual way we find that
A=xII+y+z3- 3xyz. On the other hand, by adding the
second and the third rows to the first row, we see that
x+y+z is a factor of A. Next we modify the first row by
adding to it w times the second and w2 times the third row.
The first row now reads
x+wy+w2z, z+wx+w2y, y+wz+w2x
that is
x+wy+wsz, w(x+wy+w2z), w2(X+wy+w2z).
This makes it plain that x+wy+w2z is a factor of LI.
Similarly on repiacing w by w 2( =w) we find that x+
c.I;'+wz is a factor of A. Since A is of degree 3 in each inde-
37
ROOTS OF UNITY
terminate, there can be no further factors except a constant.
By examining the coefficient of xl we find that the constant
factor is in fact unity. This establishes the result.
(::r=(::r =(::r=1.
When interpreted geometrically, the conditions are equiva-
lent to the statement that the vectors £1' £2' ea form the sides
of an equilateral triangle. Therefore arg (e2/e1)= ±21T/3,
the sign depending on the orientation of the triangle. Hence
£2=CU£1 £2=W£1'
or and in both cases (e2/£1)3=1. The
other relations are proved by similar arguments.
Having now learned how to extract the nITa root of unity,
we shall have no difficulty in finding the n determinations
of al / n, where a is any given complex number. (We may
clearly assume that a #= 0.) Let a p exp;rx be the polar
form of a and assume that %=T exp ;0 is a solution of the
equation
z"=a. (3.8)
Then,.,. exp (iOn) p exp ;rx, whence T="P, by which we
mean the unique positive value of the root and On=rx+21Tk
(k=O, ±1, ±2, ...), so that 0=(rx+21Tk)/n. We may there-
fore assert that al/n=';;p exp «rx+21Tk)i/n), because the
nITa power of the number on the right is equal to p exp ;rx=a,
whatever the value of the integer k. Confining attention to
distinct determinations of alln , we may write
a1/,,="vp eXPC:)£ik=O, I, 2, ••. , n-I), (3.9)
n runs through all the ntTa roots of unity.
where ~k=exp ( 21Tki)
Note the following rule: if b is a particular root of (3.8),
the complete set of roots can be expressed in the form
b, be, be2, ••• , ben-I, where 4: is a primitive nITa root of
unity. If a p exp irx, we may put b='{;p exp {ia:/n}.
38
ROOTS OF UNITY
Example 3. Evaluate (2- 2111'•. In this case p=/a/=v'8
and ex=arg a= - 'I't/4. Hence we may take b=
\l8(cos ~-isin ~). The four values of (2- 211* are
therefore b, bi, - band - bi.
(3.13)
A similar procedure leads to the factorization of zft_ a,
where a is a given complex number. In fact,
Zft-a=(z-b)(Z-Eb)(Z-E 2b) ... (Z-Eft-1b),
where b is any number such that bft =a.
40
ROOTS OF UNITY
The algebraical identities which we have just established,
can be used to derive numerous trigonometrical relations.
( z2 -~
,,_ 8'TT+1)
cos- (3.14)
9
and deduce that
2 cos 38+1=8 (cos 8- cos29'TT)(cos 8- cos49'TT)
(cos 8- cos89'TT) (3.15)
Use (3.13) when n=9 and observe that the factor that cor-
responds to r=3, is z8- 2z cos 23'TT +1=z2+z+1, the re-
maining three quadratic factors being as on the right-hand
side of (3.14). This expression is therefore equal to (z'- 1)/
(z- 1) (z2+z+1) = (z'-1)/(z8- 1) = z'+z'+1, which
proves (3.14). Next, divide (3.14) throughout by z8 and
then put z=e'. With this value of z, z+z-1=2 cos 8,
z8+z-a=2 cos 38, and (3.15) is an immediate consequence.
(cos 0- coal;).
9. Resolve r+ I into linear and quadratic factors with real coefncienta
Deduce that 4 sin~ cos"'!: = I.
10 5
10. Show that the roots of (... - 1)8+(... + I)~=O are
±icot'!!:',
12
±icotS7T
U
, ±i.
42
CHAPTER FOUR
Elementary Functions of a Complex Variable
1. INTRODUCTION
In the preceding chapters we studied complex numbers
from an algebraical point of view, coupled with geometrical
interpretations, and this enabled us to arrive at a sensible
and consistent definition of powers :II', where r is an integer.
We also discussed fractional powers and their many-
valuedness. Thus we are in the position to handle expres-
sions like (:ri+2:11- Ii)!(3zl+.), in which several such powers
are combined.
However, there remains a vast class of functions, familiar
to the reader from the real field, which has not yet been
extended to the case of a complex variable. We have
already mentioned, without proof, the properties of exp :II,
when :II is complex, but we have not considered the other
'elementary' functions sin z, cos z, and log z.
It is clear that the geometrical definition of sin 8 as
(opposite side)/(hypotenuse) in a right-angled triangle loses
its meaning, when the 'angle' 8 is replaced by an arbitrary
complex number. We therefore abandon the geometrical
approach and base our definitions on the series
z zl zl
exp z=1+11+21+31+... (4.1)
• :z3 Zli :;;7
smz=z- 3'+51-71+..• (4.2)
zl :r' Zl
cos z=l- 2'+ 4'- 6'+· .. (4.3)
To be sure, this raises the question as to what is meant by
43
FUNCTIONS OF A COMPLEX VARIABLE
an infinite series with complex terms, and we are obliged at
this stage to embark upon a digression on the convergence
. of complex sequences and series. Although we shall en-
deavour to make the next two sections as nearly self-con-
tained as possible, the reader is expected to be acquainted
with the correspondingf work on real sequences and series.
2. SEQUENCES
The notion of convergence of a sequence of complex
numbers can be reducod to the idea of a null-sequence of
real numbers, that is of a sequence of real numbers, which
tend to zero.
.2, x
.
~ Z2
z·~ .. •
*~
o
Figure 13
Examples.
(1.) z. (3+i1l)1
n'" Separatmg
. .mto reaI an d Imaginary
. . parts
9-n" 6 i . 9-nl
we have that z.=-a-+-=x.+,y.,
n n where x" nI
= -1+ 92 " , -1
n
andy.=~n...O. Hencez.... -1+0;=-1.
. (2 3,,\· S' 12 3'JI
4 9 145
(11) z,,= 3+4")' mce 3+4" ='9+16=144>I,ltIO-
. r I
lows that Iz.1 =(145/144)t" ten s to infinity, and hence
lim z. does not exist.
(iii) z.=( - 1).+~II· Here z. does not tend to a limit, be-
cause the real parts, that is x" =( - 1)·, do not form a con-
vergent real sequence.
(iv) If Izi <1, then z"",O, as n ...O. For we know that the
real sequencef Izl" tends to zero. Since Izl"=lzal, Iz"I"'O
or lza- 01 ...0 which means that z ....O.
(v) z.=(cos~+;sin~r·+l. By de Moivre's theorem
I'.=cos (211+1)," +; sin (2,,+1),". Since 211+1 ...2• it fol-
n " n
lows from the continuity of the functions cos B and sin B
that lim z.=cos 2,"+; sin 2'7T=1.
t J. A. Green, Ioc. cit., Example 2, p. 6.
46
SBRIBS
The rules for calculating with convergent sequences of
complex numbers are analogous to those for sequences of
real numbers. t
3. SERIES
The definition of convergence of an infinite series
W1+W Il + . . ,+WA+' • • (4.6)
with complex terms WA =XA +iYA is the same as for series
with real terms.t In both cases the concept of convergence
of a series is reduced to the more fundamental notion of
convergence of a sequence.
because the real and the imaginary parts each form a con-
vergent series. In fact, the series converges absolutely, since
(~-_l~)"_+~t_'c_os_n_ol ~ 1+1 cosnO\ ~ -2
1 n2
2 [Th'IS step uses (2 .1 )]
n-
0)
n'
(ii) 2:G~~D" diverges, because ~he general term does
not tend to zero, for
IG~~J"I=I~~~~I"=G!:)"/2=1
for all n (see p. 48).
(iii) 2: (1 ~ z)" converges absolutely provided that
Iz/(1- z)1 <1. This condition means that z is closer to 0
than to 1, that is, 9lz<t. (Alternatively, we can obtain this
result by calculation, thus IzI2<11- zI2 means that x2+y2<
(1- X)2+yll and so x<t.)
4. POWER SERIES
A power series is a series of the form
co+C1z+cllz2+Caz3+ . • •+cnz"+ • •• , (4.10)
where z=x+ry is a complex variable and where the co-
efficients COl CII Clll ••• are given numbers, in general com-
t J. A. Green,loc. cit., p. 34-
49
FUNCTIONS OF A COMPLEX VARIABLE
plex. It may happen that the series converges for all z,
obviously a very desirable situation, or again, at the other
extreme, for no value of z except zero, evidently a rather
useless case. In general, there exists a positive number R
such that (4.10) converges absolutely if Izl <R and diverges
if 1.:::1 >R. We omit the proof of this fact, but we note its'
geometrical interpretation, namely that each point of the
divergence
Figure 14
Examples.
(i) l+z+zZ+ ... +zn+ . •.. In order to examine for
absolute convergence we consider the limit, as n-+cx>, of
the ratio Iznlllzn-11 =Izi. Since the ratio happens to be
independent of n, it follows that lim Iznlllzn-11=lzl. The
ratio test now tells us that the series converges absolutely
when Izi <1 and that it diverges when Izi >1. Thus in
this case R=l. In fact, when Izi <1, it is shown as in the
real case, that 1+z+z2+ ... +zn+ . .. =(1- Z>-1.
(ii) The series
zzSz3 zn
E{z)=1+ 1l + 2!+31+"'+ n!+'" (4.12)
.I
converges absolutely for a 1 z. The proof is the same as in
zn+1
the real case·t Indeed, It (n+l)! III z1ln!1 =hm
• n+1
Izl =OJ
irrespective of the value of z.
(iii) The series C(z)=l-;;+~-z:!+ . .. and
z3 Z5 Z7
S(z)=z- 31+5!- 71+" also converge for all z, as the
reader can readily prove.
As long as the variable i restricted to lie inside the circle
of convergence, the manip lation of power series is in many
ways analogous to that of finite sums (polynomials). Thus
two power series may b added and subtracted in the
obvious way and if
1(z)=aO+a1z+a2z2+• ••
and
g(w)=bo+b1w+b sw2+ • ••
both converge absolutely for certain values of :l and fD,
fez) g(w)=aobo+(a1bOz+aObl W )+{a2boZl+alb1zw+aobaw2) •••
f J. A. Green, loco cie.,p. 48
51
FUNCTIONS OF A COMPLEX VARIABLE
Thus absolutely convergent power series may be multiplied
term by term, and the terms may be arranged in any man-
ner,t for example according to increasing degrees in the
variables III and aI.
Example.
(1- IIltl=(1- IIltl(1- IIltl
=(1 + III +1Il2 + ...)(1 +1Il+1Il2 + . •• )
=1+211l+311l2 +4z3+ • ..
Complex series are sometimes a useful tool for obtaining
results about real series. The following example will serve
as an illustration. In the formula (1- IIlt 1 =1+1Il+z3+ . ..
(11Il1 <1) put lIl=r e'8 (r<1) and compare real parts. Thus
1+r cos 8+rcos 28+ •• •=&1(1- re'8t1
1-rcos8
1- 2r cos 8+rI'
Assume now that 0<8<!1T. It follows that o<cos 8<1,
so that the last equation is valid for r=cos 8. Hence
l+(cos 8)'+(cos 8)2 cos 28+(cos 8)8 cos 38+ ..•
1-cosl! 8 1
1-cos· 8
and therefore
cos 8+cos 8 cos 28+(cos 8)1 cos 38
+(cos 8)3 cos 48+ ••• =0.
6. THE LOGARITHM
All logarithms in this book refer to the base e. We recall
that if x is positive, log x is that number u for which
x=exp u. It is known that there is one and only one real
number with this property, so that log x is unambiguously
defined as the real solution, for u, of the equation
x=exp u,
where x is a given positive number.
By analogy, let us try to define log z as the solution,
for w, of the equation
z=exp w, (4.23)
where z is a given complex number. However, we now meet
with the difficulty that, if w is a solution of (4.23), so is
w+211'ik, where k is an arbitrary integer (see (2.21». It
follows that if w=log z exists at all, it has infinitely many
56
THE LOGARITHM
values differing from one another by integral multiples of
2m. Let us now examine these values more closely.
Since exp fO is never zero, no meaning can be attached to
log z, when z=O. Accordingly we shall henceforth assume
that z#;O. Write z=rexp (i8) and w=u+iv. Then (4.23)
becomes
(4.24)
Comparing moduli, we firtd that r=exp u and therefore
u=log r,
where the logarithm has the usual unique real value which
exists because r>O. Equation (4.24) now becomes exp i8
=exp iv, from which we can only deduce that
v=8+21Tk,
where k is an integer. Remembering that "=Izl and
8=arg z, we can express the final result in the form
log z=log Izl +i(arg Z+21Tk), (4.25)
where log Izl on the right-hand side denotes the real
logarithm.
The value that corresponds to k=O, is called the principal
value of the logarithm, and it is sometimes distinguished by
a different notation, such as
Log z=log Izl +i arg z.
The principal value reduces to the usual meaning of
log x( =Log x), when x is positive, because in that case
argx=O.
Examples.
(i) log (- 2)=Logl- 21 +i arg (- 2)=Log 2+1Ti.
(ii) Log i=Log lir +arg i=Log 1+i1T/2=i1T/2.
(iii) Solve 4cos z=3+i. The equation is equivalent to
2 exp (iz)+2 exp (- iz)=3+i. Hence
2 exp (2iz)- (3+,) exp (iz)+2=0.
Regarding this as a quadratic for exp (i.r) we find, after
lome calculations. that
57
FUNCTIONS OF A COMPLEX VARIABLB
exp (iz)=l +i or t- ti, and hence either
iz=log (l+')=tLog 2+i(~+2'ITn),
8n+l i
z=-4-'lT- 2Log 2 (n=O, ±1, ±2, ••. )
or
iz=log (t- tt)= - tLog 2+i( -~+2'ITm),
8m-l i
z=-4-'IT+'2Log 2 (m=O, ±1, ±2, ...).
(iv) The equation tan z=i has no solution, even when !Z
is complex. For if Z=Zo were a solution, it would follow
that sin zo=i cos zo, sinl zo= -cosB zo, sins zo+cos2 zo=O,
in contradiction to (4.20).
59
Answers to Exercises
Clraptnl
1. (I.)5+101,
-
. (II
..) -3 - -a. ( ...)
' , 111 - 7 + a4l,. (.)
IV -
7 + -6"
a . ()II a.
V - - ~,
13 13 130 5 5 5
(vi) - i.
a. (i) 1- i, (il) 3- ai.
3. n(n- I)+i"'.
4. (i) S, (il) la-ill=(v's)'=12S, (ii) IS+l2il- 1-{v'(aS+I44)}-1
-1/13, (iv) I.
s'lil-lex~I-'exd il =lexll;1
I+ai U (v's)J
11- ail' -(v's)' =(v'S)'-Sv'S.
6. (i)± (Ja + Ja,) (ii) ±{(va+ I)i+i(v'a- I)t}, (iii) ±(a+.).
7. a- i, I+ai.
8. ~- 4t+ 5=0, roots a+i, a-i.
10. ±!(v'6±v'a.) (four combinations of ligna).
Chapter 11
I. If X is represented by :t, the condition is that (a- x),'(b - x) should be
purely imaginary: X=3.
3. Interior and circumference of ellipse with foci at I and i and major
ais 4.
4. The conditions are equivalent to (.11'1- .11'.)/(.11'1- .11'.) =,\ (see Example
2, p. as).
6. By the well-known property of cyclic quadrilaterals it suffices to show
that the Bum of opposite angles, say at .11'1 and .11'., is equal to 'IT. Thul
arg (.11'1- .1)/(.11'.- .ll'1)+arg (.11',- .II',)f(.II'I- .11',)= ±'IT or 0
arg (.11'.- .1)(.'-.11'./(.11',- .11'1)(.11'1- .11'.)= ±'IT or 0,
that is {.II'I' .11',; a., a,} is real.
7. u- a==q(b- a), fI- b=u- a.
8. When A-I, that is when 1.11'- al-I.II'- bl, the equation represents all
points equidistant from a and b, the perpendicular bisector of the
IelPDeDt .liB.
60
ANSWERS TO EXERCISES
4m+I)tI(1+,)/v'a.
10. (- I)m ( se~7T
Chapter 111
I. 1. t+liv'3. - t+liv'3. - I, - t- liv'3. t- liv'3.
a. b, bit - b, -bi, where b==cos (7T18)+i sin (71'/8).
3. v'a(cos..!.+isin 71'). -1+i, -v'a(cosS7T+isinS!\.
Ia Ia la 12/
6. The sides of the triangle are represented by the vectors fl. 11- fl.
- o. Let JII=( - o)/fI, w=(o- fI)/fI. It is required to show that arg JII-
±a7T/3. arg w= ±a7T/3. This follows from the fact that .r+JII+ 1 -
wI+W+l=O.
,. Put fI=.I- JIll. 0-=.1- JIIa (moving the origin to JIIa)'
8. In the product formula for a8+ I divide each side by %4 and put
JIIs=e ill•
9. In the product formula for a-6+ I put .=;.
Chapter IV
I. (I) - i, (ii) no limit, (iii) 0, (iv) - I. (v) i.
a. (i) conv.• (ii) div., because resl parts div., (iii) aba. conv., (iv) conv.,
I
(v) div., because
Sin
~ - zi~
I
~I_I e-tl- e" e"- I
>~CX).
3. (i) I, (ii) a/3, (iii) 1/4, (iv) 11,.
S. le'l =eS. The greatest and least values of" on the circle are 3 and - I
respectively..
6. (i) (Zk+I)7Ti. (ii)log seccx+i(-cx+ak7T),(iii) C08\Og a+i.in lOll a
(iv) up (- tw<4k+ I) ).
7. (4k+1)7T/a- ilog (a±v'3).
(.) sinh a,,+; sin a" (.') 4k+ I .
9· 1 com ax+cos a,,' 11 - .. -71'1.
61
Index
radius of convergence, So
distributive law, 2 real axis, IS
real part, 6
Euler's formula, 29 root of unity, 33
63
LIBRARY OF MATHEMATICS
CALCULUS OF VARIATIONS A. M. Arthurs
COMPLEX NUMBERS W. Ledermann
DIFFERENTIAL CALCULUS P. J. Hilton
ELECTRICAL AND MECHANICAL OSCILLATIONS D. S. Jones
ELEMENTARY DIFFERENTIAL EQUATIONS AND OPERATORS
G. E. H. Reuter
FOURIER AND LAPLACE TRANSFORMS P. D. Robinson
FOURIER SERIES I. N. Sneddon
*FUNCTIONS OF A COMPLEX VARIABLE, I AND II (2 vols.)
D.O. Tall
INTRODUCTION TO ABSTRACT ALGEBRA C. R. J. Clapham
INTRODUCTION TO MATHEMATICAL ANALYSIS C. R. J. Clapham
LINEAR EQUATIONS P. M. Cohn
*LlNEAR PROGRAMMING K. Trustrum
MULTIPLE INTEGRALS W. Ledermann
NUMBER THEORY T. H. Jackson
NUMERICAL APPROXIMATION B. R. Morton
PARTIAL DERIVATIVES P. J. Hilton
PRINCIPLES OF DYNAMICS M. B. Glauert
PROBABILITY THEORY A. M. Arthurs
SEQUENCES AND SERIES J. A. Green
SETS AND GROUPS J. A. Green
SETS AND NUMBERS S. Swierczkowski
SOLID GEOMETRY P. M. Cohn
SOLUTIONS OF LAPLACE'S EQUATION D. R. Bland
VIBRATING STRINGS D. R. Bland
VIBRATING SYSTEMS R. F. Chisnell
*Also available in cloth edition