INDIAN INSTITUTE OF TECHNOLOGY

BOMBAY
MA 204 - Mathematics IV
Lecture 1

Basic Properties of Complex Numbers

§1 Prerequisites
§1.1 Reals Numbers:
I The law of commutativity: a + b = b + a; ab = ba, for all a, b ∈ R.
II The law of associativity: (a + b) + c =
a + (b + c); (ab)c = a(bc), for all a, b, c ∈ R.
III The law of distributivity: (a + b)c = ac + bc, for all a, b, c ∈ R.
IV The law of identity: a + 0 = a; a1 = a, for all a ∈ R.
V The law of additive inverse: Given any a ∈ R, there exists a unique x ∈ R such that
a + x = 0.
VI The law of multiplicative inverse: Given a ∈ R, a 6= 0, there exists a unique x ∈ R
such that ax = 1.
Furthermore, there is a total ordering ‘<’ on R, compatible with the above arith-
metic operations, which makes R into an ordered field. Recall that < is a total ordering
means that:
VII given any two real numbers a, b, either a = b or a < b or b < a.
The ordering < is compatible with the arithmetic operations means the following:
VIII a < b =⇒ a + c < b + c and ad < bd for all a, b, c ∈ R and d > 0.

1 Algebra of Complex Numbers

We define the algebra of complex numbers C to be the set of formal symbols x +
ıy, x, y ∈ R together with the addition and multiplication defined as follows:

(x1 + ıy1 ) + (x2 + ıy2 ) = (x1 + x2 ) + ı(y1 + y2 );

(x1 + ıy1 )(x2 + ıy2 ) = (x1 x2 − y1 y2 ) + ı(x1 y2 + y1 x2 ).

ı2 + 1 = 0; i.e., ı2 = −1.

Theorem 1 The set C of all formal expressions a + ıb where a, b ∈ R forms the

smallest field containing R as a subfield and in which ı is a solution of the equation

X 2 + 1 = 0.

1
 Observe that a complex number is well-determined by the two real numbers, x, y
viz., z := x + ıy. These are respectively called the real part and imaginary part of z. We
write:

ℜz = x; ℑz = y. (1)

If ℜ(z) = 0, we say z is (purely) imaginary and similarly if ℑ(z) = 0, then we say z

is real. The only complex number which is both real and purely imaginary is 0. Observe
that, according to our definition, every real number is also a complex number.

equating the real and the imaginary parts of the two sides of an
equation

is indeed a part of the definition of complex numbers and will play a very important
role.

2 Conjugation and Absolute Value

Definition 1 Following common practice, for z = x + ıy we denote by z = x − ıy and
call it the (complex) conjugate of z. and call it
the conjugate of z.

z+z z−z
ℜ(z) = ; ℑ(z) = . (2)
2 2ı

z1 + z2 = z1 + z2 , z1 z2 = z1 z2 , z = z. (3)

Definition 2 Given z ∈ C, z = a + ıb, we define its absolute value (length ) |z| to be

the non-negative square root of a2 + b2 , i.e.,
√
|z| := (a2 + b2 ).

Remark 1 |z|2 = zz. Therefore

z ∈ C, |z| = 6 0 ⇐⇒ z 6= 0.
Also, for z 6= 0,

z −1 = z|z|−2 .

2
Basic Identities and Inequalities
(B1) |z| = |z|.
(B2) |z1 z2 | = |z1 ||z2 |.
(B3) |ℜ(z)| ≤ |z| ( resp. |ℑ(z)| ≤ |z|); equality holds iff ℑ(z) = 0 (resp. ℜ(z) = 0).
(B4) Cosine Rule:
|z1 + z2 |2 = |z1 |2 + |z2 |2 + 2ℜ(z1 z2 ).
(B5) Parallelogram Law :

|z1 + z2 |2 + |z1 − z2 |2 = 2(|z1 |2 + |z2 |2 ).

(B6) Triangle inequality : |z1 + z2 | ≤ |z1 | + |z2 | and equality holds iff one of the zj is a
non-negative multiple of the other.
(B7) Cauchy’s1 Inequality :
 2   
 n n n
X  X X
|zj |2   |wj |2  .


 zj wj  ≤
j=1  j=1 j=1

Polar form:
r(cos θ , sinθ )

E (ϕ) r

θ
ϕ
0

Fig. 0

Given (x, y) = z 6= 0, the angle θ, measured in counter-clockwise sense, made by

the line segment [0, z] with the positive real axis is called the argument or amplitude of
z:
θ = arg z.
1
Augustin Louis Cauchy (1789-1857) was a French mathematician, an engineer by training. He did
pioneering work in analysis and the theory of permutation groups, infinite series, differential equations,
determinants, probability and mathematical physics.

3
 x = r cos θ; y = r sin θ (4)

Let us temporarily set-up the notation

E(θ) := cos θ + ı sin θ. (5)

Then the complex number z = x + ıy takes the form

z = r(cos θ + ı sin θ) =: rE(θ).

Observe |z| = r. Now let z1 = r1 E(θ1 ), z2 = r2 E(θ2 ). Using additive identities for sine
and cosine viz.,

sin(θ1 + θ2 ) = sin θ1 cos θ2 + cos θ1 sin θ2 ,

(6)
cos(θ1 + θ2 ) = cos θ1 cos θ2 − sin θ1 sin θ2 ,

we obtain

z1 z2 = r1 r2 E(θ1 + θ2 ). (7)

If we further remind ourselves that the argument can take values (in radians)
between 0 and 2π, then the above identity tells us that arg(z1 z2 ) = arg z1 + arg z2
(mod 2π) provided z1 6= 0, z2 6= 0.
Put zj = rj E(θj ) for j = 1, 2, and let θ be the angle between the vectors represented
by these points. Then z1 z¯2 = r1 r2 E(θ1 − θ2 ) and hence ℜ(z1 z¯2 ) = r1 r2 cos θ. Thus,

ℜ(z1 z̄2 )
cos θ = . (8)
|z1 z2 |

Now, we can rewrite the cosine rule as:

|z1 + z2 |2 = r12 + r22 + 2r1 r2 cos θ. (9)

Note that by putting θ = π/2 in (9), we get Pythagoras theorem.

Remark 2 Observe that given z 6= 0, arg z is a multi-valued function. Indeed, if θ is

one such value then all other values are given by θ + 2πn, where n ∈ Z. Thus to be
precise, we have
arg z = {θ + 2πn : n ∈ Z}
This is the first natural example of a ‘ multi-valued function’. We shall come
across many multi-valued functions in complex analysis, all due to this nature of arg z.
However, while carrying out arithmetic operations we must ‘select’ a suitable value for
arg from this set. One of these values of arg z which satisfies −π < arg z ≤ π is singled
out and is called the principal value of arg z and is denoted by Arg z.

4
Example 1 The three cube roots of unity are
2π 2π 4π 4π
1, cos + ı sin , cos + ı sin
3 3 3 3
which we can simplify as: √ √
−1 + 3 −1 − 3
1, , .
2 2
Remark 3 deMoivre’s2 Law Now observe that, by putting r1 = r2 = 1 in (7) we
obtain:
E(θ1 + θ2 ) = E(θ1 )E(θ2 ).
Putting θ1 = θ2 = θ and applying the above result repeatedly, we obtain

E(nθ) = E(θ)n .

This is restated in the following:

deMoivre’s Law:

cos nθ + ı sin nθ = (cos θ + ı sin θ)n . (10)

Remark 4 Roots of complex numbers: Thanks to our geometric understanding, we can

now show that the equation

Xn = z (11)

has exactly n roots in C for every non zero z ∈ C. Suppose w is a complex number that
satisfies the equation (in place of X,) we merely write

z = rE(Arg z), w = sE(Arg w).

Then we have,
sn E(nArg w) = w n = z = rE(Arg z)
√ q
Therefore we must have s = n r = n |z| and arg w will contain the values

Arg z 2kπ
+ , k = 0, 1, . . . , n − 1.
n n
q
Thus we see that (11) has n distinct solutions. One of these values viz., n
|z|E( Arg
n
z
) is
√
called the principal value of the nth root function and is denoted by n z.

2
Abraham deMoivre(1667-1754) was a French mathematician. He also worked in Probability theory.

5
 INDIAN INSTITUTE OF TECHNOLOGY
BOMBAY
Spring 2007

MA 204 - Mathematics IV
Lecture 2
/5th Jan 07

1 Geometry of Complex Numbers

The picture below illustrates how to add two complex numbers geometrically.

y1+ y2 Q

P1
y1
y2 P2

O x1 x2 x1+ x2

Fig.1

The ‘parallelogram law’ (B5) now becomes:

The sum of the squares of the lengths of the diagonals of a

parallelogram is equal to the sum of the squares of the lengths
of the sides.

Remark 1 Geometric Multiplication of complex numbers:

1
 Z1 Z2

Z2
Z1

0 1
Fig. 2

In the picture above, the two triagles a similar. It tells you how to multiply two
complex numbers.
−→
Similarly, to get the inverse of z2 6= 0, we first re-scale the vector OP2 to have
length r2−1 and then change its direction to have an amplitude −θ2 . Also, it follows that
−→ −→
the amplitude of z1 z2−1 is the angle that OP1 makes with OP2 .
It follows that given 0 6= λ ∈ C, the assignment z 7→ λz defines a linear map
2 2
R −→ R , which is a composite of a rotation (through an angle2 θ = Arg 2
λ) and a
dilation or a scaling ( by a factor r = |λ|). Such linear maps R −→ R are called
similarities. For the converse part you will have to wait.

The dot product Think of two complex numbers as vectors in R2 . Then their dot
product is given by
(x1 + ıy1 ) · (x2 + ıy2 ) = x1 x2 + y1 y2
which we can rewrite in terms of multiplication of complex numbers as:

z1 · z2 = ℜ(z1 z¯2 ) = ℜ(z¯1 z2 .

Similarly, very that that cross product can be rewritten is the form

z1 × z2 = ℑ(z¯1 z2 )

Equation of a line: Let ax + by + c = 0 represent a line in cartesian coordinates.

WLOG we may assume that a2 + b= 1. Put w = a + ıb; z = x + ıy. Then ax + by =
Re(wz̄) = 1/2(wz̄ + w̄z). Thus, we see that the general equation of a line in the plane
can be given by complex number as:

wz̄ + w̄z = t, t ∈ R.

Equation of a circle:
(z − w)(z − w) = r 2 , r ∈ R.

Definition 1 By a rigid motion or an isometry of the plane, we mean a mapping

f : C → C which preserves distances, i.e., |f (z) − f (w)| = |z − w| for all z, w ∈ C.

2
Remark 2 Identity, translation, rotation about a point and reflexion in a line are ob-
vious examples of rigid motions. The first three preserve the sense of orientation of the
plane and the third one changes it. The first one keeps all points fixed. The second one
has no fixed point but it preserves orientation. The third one fixes exactly one point and
also preserves the oriention. The fourth one fixes exactly a line and (hence) changes the
orientation. Are there other rigid motions such as those which fix no points and change
the orientation? We shall investigate this right now.
It is clear that composite of two rigid motions is again a rigid motion. It is also
clear that a rigid motion is a one-one mapping but it is not easy to see that it also onto.
This will follow as an easy consequence of what we are going to do.

Reflection in a Line Let

wz̄ + w̄z = t
represent a line. If z ∗ denotes the image of z under the reflection in L, then z ∗ − z is
parallel to w and is bisected by L. Therefore, we obtain,

z ∗ − z = sw, s ∈ R; & w(z ∗ + z) + w̄(z ∗ + z) = 2t.

Upon simplification this yields
t − wz̄
z∗ = = wt − w 2 z̄.
w̄
Definition 2 By a glide-reflection we mean a RM which is a reflection in a line followed
by a translation by a non zero vector in the direction of L.

It is easy to see that a glide-reflection does not have any fixed point and does not
preserve the orientation. The converse follows from what we see below.

Theorem 1 Let f : C → C be a rigid motion such that f (0) = 0. Then there exist
unique a, b ∈ C with |a| = 1 such that

f (z) = az + b, ∀z ∈ C OR f (z) = az̄ + b, ∀z ∈ C.

Proof: Put b = −f (0) and define g(z) = f (z) + b. Then g is also a RM. and g(0) = 0.
Now |g(1)| = 1. So, put a = g(1)−1 and define h(z) = ag(z). Then h is RM and
h(0) = 0, h(1) = 1, and h(ı) = ±ı.
Case 1: Assume h(ı) = ı. Now consider any z = x + ı ∈ C and put h(z) = u + ıv. Then
it follows that
u2 + v 2 = x2 + y 2;
(u − 1)2 + v 2 = (x − 1)2 = y 2 ;
u2 + (v − 1)2 = x2 + (y − 1)2 . Solving these, yields, u = x, v = y. Thus h(z) = z, ∀z ∈ C.
This is the same as saying f (z) = az + b, ∀z ∈ C.
Case 2: Assume that h(ı) = −ı. Put h̄(z) = h(z). Then h̄ is a RM and h̄(0) = 0, h̄(1) =
1, h̄(ı) = ı. So, we are in case 1. ♠.

3
Theorem 2 Let f : C → C be a rigid motion.
(i) Suppose f fixes two distinct points. Then all points on the line passing through these
two points are also fixed by f.
(ii) Suppose f fixes three non collinear points. Then f = Id.
(iii) Suppose f fixes an entire line L. Then it is either Id or the reflexion in that line.
(iv) Suppose f fixes exactly one point. Then it is a rotation around that point.
(v) Suppose f fixes no points. Then either f is a translation or a glide reflexion.

Proof: (i) If z1 , z2 are such that azi + b = zi , i = 1, 2 then for any real numbers t1 , t2
such that t1 + t2 = 1, we have

a(t1 z1 + t2 z2 ) + b = t1 z1 + t2 z2 .

(ii) Argue as in (i).

(iii) We may assume that 0, 1 are fixed by f, so that the entire real line is fixed. Now if
f fixes ı also, then by (ii), f = Id. If f (ı) = −ı then we consider the RM f¯ which must
be Id.
(iv) Choose the fixed point f as the origin. Then f (0) = 0. This means b = 0. Also,
observe that if f (z) = eıθ z̄ then f (reıθ/2 ) = reıθ), shows that f fixes a whole line.
Therefore f (z) = eıθ)z for some θ.
(v) First consider the case f (z) = az + b. If a 6= 1 then we can solve for f (z) = z and
therefore a = 1. Now consider f (z) = az̄ + b. If a = 1, it follows that f is a reflection
in the line y = ℑb followed by translation by ℜ(b) 6= 0. Now say a 6= 1 Put λ2 = a. If
b = tλ for some t ∈ R again f is the reflection in the line joining 0 and λ. So, b 6= tλ.
Therefore there exists s ∈ R such that s(1 − a) = b − tλ. Let L be the line parallel to
the vector λ and passint through s. Then it follows that f is the reflection in the line L
followed by the translation by the vector tλ.
Story Time This story is due to George Gamow, the well-known physicist and
an ingenious story-teller. We quote from his book:
ONE TWO THREE · · · INFINITY (pp. 44-45).
There was a young and adventurous man who found among his great-grand fa-
ther’s papers a piece of parchment that revealed the location of a hidden treasure. The
instructions read:
“ Sail to · · · North latitude and · · · West longitude where thou wilt find a deserted
island. There lieth a large medow, not pent, on the north shore of the island where
standeth a lonely oak and a lonely pine. There thou wilt see also an old gallows on
which we once were wont to hang traitors. Start thou from the gallows and walk to the
oak counting thy steps. At the oak thou must turn right by a right angle and take the
same number of steps. Put here a spike in the ground. Now must thou return to the
gallows and walk to the pine counting thy steps. At the pine thou must turn left by a
right angle and see that thou takest the same number of steps, and put another spike
into the ground. Dig half-way between the spikes; the treasure is there.”

4
 Lecture 3
Complex Differentiabilty
Notation: Br (z) = {w ∈ C : |w − z| < r}
Open ball of radius r center z.

Definition 1 A ⊂ C is said to be an open set if it is the union of open balls.

Thus if A is open and z ∈ A then it follows that there is r > 0 such that Br (z) ⊂ A.

Definition 2 Let z0 ∈ A ⊂ C, such that there exists r > 0 with Br (z0 ) ⊂ A. Let f : A −→ C
be a map. Then f is said to be complex differentiable (written C−differentiable at z0 )
if the limit on the right hand side of the following formula exists, and in that case we call this
limit, the derivative of f at z0 :

df f (z0 + h) − f (z0 )
(z0 ) := lim . (1)
dz h→0 h

We also use the notation f 0 (z0 ) for this limit and call it Cauchy derivative of f at
z0 . For an open subset A of C, if f is complex differentiable at each z0 ∈ A then we say f is
holomorphic on A. If f is holomorphic on A, the map z 7→ f 0 (z) is called the derivative of f
on A and is denoted by f 0 .

Remark 1 We caution you that the words ‘holomorphic’ and ‘analytic’ have been used by
different authors to mean somewhat different things. Luckily, this is not a serious matter, since
we shall see that they ultimately they all mean the same thing.

Example 1 Let us work out the derivative of the function f (z) = z n , where n is an integer,
directly from the definition. Of course, for n = 0, the function is a constant and hence, it is
easily seen that it is differentiable everywhere and the derivative vanishes identically. Consider
the case when n is a positive integer. Then by binomial expansion, we have,
! ! !
n n−1 n
f (z + h) − f (z) = h z + hz n−2 + · · · + hn−1 .
1 2

Therefore, we have,
f (z + h) − f (z)
lim = nz n−1 .
h→0 h
This is valid for all values of z. Hence f is differentiable in the whole plane and its derivative
is given by f 0 (z) = nz n−1 . Next, consider the case when n is a negative integer. We see that the
function is not defined at the point z = 0. Hence we consider only points z 6= 0. Writing n = −m
z m − (z + h)m
and f (z + h) − f (z) = and applying binomial expansion for the numerator as
(z + h)m z m
above, we again see that
m
f 0 (z) = − m+1 = nz n−1 , z 6= 0.
z

1
Remark 2 As in the case of calculus of one real variable, the Cauchy derivative has all the
standard properties: sum of two C−differentiable function is C−differentiable and so
on.

(αf + βg)0 (z0 ) = αf 0 (z0 ) + βg 0 (z0 ), α, β ∈ C. (2)

Thus for each fixed z0 , the assignment f 7→ f 0 (z0 ) defines a linear functional on the space of all
complex differentiable functions defined in a neighborhood of z0 . Moreover, just like in the real
case, it is also a derivation, i.e.,

(f g)0 (z0 ) = f 0 (z0 )g(z0 ) + f (z0 )g 0 (z0 ). (3)

We even have the same formula for the derivative of a quotient:

!0
f f 0 (z0 )g(z0 ) − f (z0 )g 0 (z0 )
(z0 ) = ; if g(z0 ) 6= 0. (4)
g (g(z0 ))2

The proof of the following theorem is exactly the same as the proof of the corresponding result
for real valued function of a real variable.

2
Theorem 1 (The Increment Theorem:)
Let f : A → C, z0 ∈ A, r > 0 such that Br (z0 ) ⊂ A. Then f is complex differentiable at
z0 iff ∃ α ∈ C, and a set theoretic function φ : Bs (0) \ {0} → C, (0 < s < r) such that for all
h ∈ Bs (0) \ {0}

(f (z0 + h) − f (z0 ) = hα + hφ(h); limh→0 φ(h) = 0. (5)

Proof: For the given f, we simply take

f (z0 + h) − f (z0 )
ψ(h) := , h ∈ Br (0) \ {0}.
h
Then f is holomorphic at z0 iff limh→0 ψ(h) exists. In that case, we simply put α equal to
this limit and take φ(h) = ψ(h) − α and observe that φ(h) → 0 iff ψ(h) → α. On the other
hand, if there is such a function φ and a constant α then clearly, limh→0 ψ(h) = α, and so, f is
holomorphic at z0 and f 0 (z0 ) = α. ♠

Remark 3 We may assume that the error function φ in (5) is defined on the whole of Br (0),
its value at 0 being completely irrelevant for us. The increment theorem enables one to deal with
many tricky situations while dealing with differentiability. As an illustration we shall derive the
chain rule for differentiation.

3
Theorem 2 (Chain Rule :) Let f : A → C, g : B → C, f (A) ⊂ B and z0 ∈ A. Suppose
that f 0 (z0 ) and g 0 (f (z0 )) exist. Then (g ◦ f )0 (z0 ) exists and (g ◦ f )0 (z0 ) = g 0 (f (z0 ))f 0 (z0 ).

Proof: Let

f (z0 + h) − f (z0 ) = hf 0 (z0 ) + hη(h); 
(6)
g(f (z0 ) + k) − g(f (z0 )) = kg 0 (f (z0 )) + kζ(k) ; 

be as in the increment theorem, where

η(h) → 0 as h → 0; & ζ(k) → 0 as k → 0

Let us say η : Bs (0) → C and ζ : Br (0) → C. Now, if we take s to be sufficiently small, then
it follows (from the continuity of f at z0 ) that for h ∈ Bs (0), we have f (z0 + h) − f (z0 ) ∈ Br (0).
Hence we can put k = f (z0 + h) − f (z0 ), in (6) to obtain,

g(f (z0 + h)) − g(f (z0 )) = [hf 0 (z0 ) + hη(h)][g 0 (f (z0 )) + ζ(k)]
= hg 0 (f (z0 ))f 0 (z0 ) + hξ(h),

where, ξ(h) = η(h)g 0 (f (z0 )) + (η(h) + f 0 (z0 ))ζ(f (z0 + h) − f (z0 )). Observe that as h → 0, we
have, k = f (z0 + h) − f (z0 ) → 0 and ζ(k) → 0. Hence ξ(h) → 0, as h → 0. Thus by the
increment theorem again, (g ◦ f )0 (z0 ) exists and is equal to g 0 (f (z0 ))f 0 (z0 ) as desired.

Exercise: (i) Derive the 00 -form of L’Hospital rule for complex differentiable functions. (Hint:Use
increment theorem.) Apply this to! find the following limit:
1 1 1
lim n
− .
z→z0 z − z
0 (ξ − z) (ξ − z0 )n

4
 The following result, which is completely obvious from the definition of differentiability,
is extremely useful in extending the domain of definition of a holomorphic function.

Theorem 3 Patching-up Lemma: Let U1 , U2 be any two open sets in C and fi : U → C be

holomorphic functions. If f1 (z) = f2 (z) for all z ∈ U1 ∩ U2 then the function defined by

 f1 (z) z ∈ U1
g(z) =
 f2 (z) z ∈ U2

is holomorphic on U1 ∪ U2 .

Remark 4 So far, we have considered derivatives of functions at a point z ∈ A only when the
function is defined in a nbd of z. We can try to relax this condition as follows: Thus, if B ⊂ C
and f : B → C, we say f is holomorphic on B if f extends to a holomorphic map fˆ : A → C
where A is an open subset containing B. However, we can no longer attach a unique derivative
to f at points of B in general. With some more suitable geometric assumptions on B, this can
be made possible. For instance if B is a closed disc or a closed rectangle, and if f : B → C is
differentiable on B, then even at the boundary points of B, the derivative of f is unique.

Cauchy-Riemann Equations
Let U be an open subset of C, z0 = (x0 , y0 ) ∈ U and f : U → C be a given function.
Write z = x + ıy, z0 = x0 + ıy0 , h = h1 + ıh2 , f (x + ıy) = u(x, y) + ıv(x, y). Suppose
f (z0 ) exists and let f 0 (z0 ) = α + ıβ. By the increment theorem, we have
0

f (z0 + h) − f (z0 ) = hf 0 (z0 ) + hη(h); η(h) → 0 as h → 0. (7)

[Put L(h) = hf 0 (z0 ). Then L : C → C is a real linear map. Also

|f (z0 + h) − f (z0 ) − L(h)|

→ 0; as h → 0.
|h|

Therefore, it follows that f is real (total) differentiable at z0 and D(f )(z0 ) = L. In particular,
all directional derivatives of f exist at z0 .]
Put η(h) = η1 (h) + ıη2 (h), where η1 , η2 are real and imaginary parts of η. Then clearly,

ηj (h) → 0 as h → 0. (8)

Put h = h1 in (7), i.e., take h2 = 0, to obtain

f (x0 + h1 , y0 ) − f (x0 , y0 ) = h1 (α + ıβ) + h1 η(h1 ). (9)

Comparing the real and the imaginary parts on both the sides we get

u(x0 + h1 , y0 ) − u(x0 , y0 ) = h1 α + h1 η1 (h1 ); 
(10)
and v(x0 + h1 , y0 ) − v(x0 , y0 ) = h1 β + h1 η2 (h1 ). 

5
 Therefore, by the increment theorem for 1-variable functions, it follows that ux and vx
exist at (x0 , y0 ) and we have,

ux (x0 , y0 ) = α, vx (x0 , y0 ) = β. (11)

Now put h = ıh2 , i.e., take h1 = 0 in (7), to obtain

f (x0 , y0 + h2 ) − f (x0 , y0 ) = ıh2 (α + ıβ) + ıh2 η(ıh2 )

Again, comparing the real and imaginary parts and using increment theorem, we obtain

uy (x0 , y0 ) = −β; vy (x0 , y0 ) = α. (12)

Thus (11) and (12) together give

ux (x0 , y0 ) = vy (x0 , y0 ); uy (x0 , y0 ) = −vx (x0 , y0 ) (13)

These are called Cauchy–Riemann(CR)-equations.1 Observe that we also have.

f 0 (z0 ) = ux (x0 , y0 ) + ıvx (x0 , y0 ) (14)

= vy (x0 , y0 ) − ıuy (x0 , y0 ) (15)

and

|f 0 (z0 )|2 = u2x + vx2 = u2y + vy2 = u2x + u2y = vx2 + vy2 = ux vy − uy vx . (16)

The last expression above, which is the determinant of the matrix

 
u uy 
 x (17)
vx vy

is called the jacobian of the mapping f = (u, v), with respect to the variables (x, y) and is
denoted by

J(x,y) (u, v) := ux vy − uy vx . (18)

1
Bernhard Riemann(1826-1866) a German mathematician.

6
Remark 5 A simple minded application of CR-equations is that it helps us to detect easily when
a function is not C-differentiable. For example, <(z), =(z) etc are not complex differentiable
anywhere. The function z 7→ |z|2 is not complex differentiable for any point z 6= 0. However,
it satisfies the CR-equations at 0. That of course does not mean that it is C-differentiable at 0.
(See the exercise below.)
2
Example 2 If f (z) = zz , z 6= 0 and f (0) = 0, show that Cauchy-Riemann equations are
satisfied at z = 0, but f 0 (0) does not exist.

Sol:
3 2 2 3
For the first part, put f = u + ıv. Then u(x, y) = xx−3xy 2 +y 2 ; v(x, y) =
−x y+y
x2 +y 2
. Direct
computation shows that ux (0, 0) = 1; uy (0, 0) = 0; vx (0, 0) = 0, vy (0, 0) = 1. Hence CR equations
are satisfied.
However, for z = reiθ , we have

f (z) − f (0) z̄ 2
lim = lim = e−3iθ .
r→0 z r→0 z
2
which shows that f 0 (0) = limz→0 z̄z does not exist.
Cauchy-Riemann equations under orthogonal Transformations
“If f (z) = u + iv is differentiable at a point z0 = z0 + iy0 of a domain G, then

∂u ∂v ∂u ∂v
= , =− (∗)
∂s ∂n ∂n ∂s
∂ ∂
at (x0 , y0 ) where ∂s and ∂n denote directional differentiation in two orthogonal directions s and
n at (x0 , y0 ), such that n is obtained from s by making a counterclockwise rotation.”
Solution: Put s = cos θi + sin θj, n = − sin θi + cos θj. Then
∂u
us = = ∇u · s = ux cos θ + uy sin θ;
∂s
∂v
vs = = ∇v · s = vx cos θ + yy sin θ.
∂s
Similarly find un , vn . These quantities can be expressed in matrix form as:
      
u un   ux uy   cos θ − sin θ   ux uy 
 s = = R−θ
vs vn vx vy sin θ cos θ vx vy
R−θ is rotation through an angle −θ which is a multiplication by a non zero complex number.
Therefore the first matrix is a multiplication by a complex number iff the second one is. In
view of the above ex. this establishes (*).

7
CR equations in polar coordinates
Take a point other than the origin. (At the origin polar coordinate is singular.) Say,
z0 = (x0 , y0 ) 6= (0, 0) and let f = u + ıv. We claim that equations

rur = vθ ; rvr = −uθ . (19)

are equivalent to CR-equations and obtain the formula:

f 0 (z0 ) = e−iθ0 (ur + ivr ) = − zi0 (uθ + ivθ ) (20)

Sol: Since x = r cos θ; y = r sin θ, by chain rule:

xr = cos θ; xθ = −r sin θ; yr = sin θ; yθ = r cos θ.
Moreover, ur = ux cos θ + uy sin θ; vr = vx cos θ + vy sin θ;
uθ = −ux r sin θ + uy r cos θ; vθ = −vx r sin θ + vy r cos θ.
The latter can be expressed in the matrix form as:
    
cos θ sin θ   ux vy   rur vθ
r = .
− sin θ cos θ uy −vx uθ −rvr

Rewriting the above matrix equation in the form

    
u vy   cos θ − sin θ   ur
 x
vθ /r 
=
uy −vx sin θ cos θ uθ /r −vr

and substituting in f 0 (z) = ux + ıuy , gives

uθ uθ
 
0
f (z) = cos θur + sin θ + i − sin θur + cos θ
r r
= cos θur + sin θvr − i (sin θur + cos θvr )
= e−iθ (ur + ivr )
e−iθ −i
= (vθ − iuθ ) = (uθ + ivθ ).
r z

8
 Lecture 4

Theorem 1 If f is differentiable at a point then it is continuous at that point.

Theorem 2 Let U be an open subset of C and f : U → R be a function. Then f is real-

differentiable on U and the function f 0 : U → C is continuous iff both the partial derivatives of
f exist on U and are continuous on U.

We shall not give a proof of this theorem here. It suffices to remark that one has to use
the mean value theorem of 1-variable calculus. Such a function is said to be continuously
differentiable at x.
Observe that the assignment x 7→ (Df )x defines a map Df of U into the space of
all linear maps R2 into R viz., the dual vector space R2? which is isomorphic to R2 . So, one
can define f to be continuously differentiable if Df is defined and continuous. Since the two
coordinate functions of Df are nothing but the two partial derivatives, the continuity of Df is
equivalent to that of the continuity of the two partial derivatives of f.
Real-linear VS Complex Linear
Let us first consider a purely algebraic problem: Treating C as a 2-dimensional real vector
space, consider a real linear map T : C → C given by the matrix
 

a b 
c d

When is it a complex linear map? We see that, if T is complex linear, then T (ı) = ıT (1) and
hence, b + ıd = T (ı) = ıT (1) = ı(a + ıc). Therefore, b = −c and a = d.
Conversely, a R-linear map T : C → C given by
 

a −b 
b a

has the property that

T (z) = T (x + iy) = ax + ibx + (−by + iay) = (a + ib)(x + iy) = wz

which is merely multiplication by a complex number.

Theorem 3 Let f : U → C be a continuous function, f = u + ıv, z0 = x0 + ıy0 be a point of U.

Then f is C-differentiable at z0 iff considered as a vector valued function of two real variables,
f is real differentiable at z0 and its derivative (Df )z0 : C → C is a complex linear map. In
that case, we also have f 0 (z0 ) = (Df )z0 .

Proof: Suppose that f is C-differentiable at z0 . Then we have already seen that f is real
differentiable and its total real derivative D(f )(z0 ) is equal to f 0 (z0 ).
Conversely, assume that Df (z0 ) exists and is complex linear. This means that there exist
error functions ζ and γ say, such that

u(z0 + h) − u(z0 ) = (h1 ux + h2 uy ) + |h|ζ(h) 
(1)
v(z0 + h) − v(z0 ) = (h1 vx + h2 vy ) + |h|γ(h) 

1
with ζ(h) → 0 and γ(h) → 0 as h → 0. Complex linearity of Df means that ux = vy :=
α say, uy = −vx := β say. Then, multiply the second equation by ı and add it to the first
equation above to obtain

f (z0 + h) − f (z0 ) = h(α + ıβ) + |h|(ζ(h) + ıγ(h)),

where,
lim (ζ(h) + ıγ(h)) = 0.
h→0

Finally, in this case, the complex linear map D(f )z0 given by the matrix (??) is nothing but
multiplication by the complex number ux + ıvx = f 0 (z0 ). Hence, the proof of the theorem is
complete. ♠
As an immediate corollary, we have,

Theorem 4 Let f : U → C be a complex differentiable function in a convex domain U such

that f 0 (z) ≡ 0. Then f is a constant on U.

There are certain useful partial results that relate the two notions of differentiability. We
shall mention some of them here without proof. The most popular one is:

Theorem 5 Let f be a complex valued function of a complex variable defined on an open subset
U. Then f is complex differentiable in U iff both the first order partial derivatives of f exist,
are continuous and satisfy the Cauchy–Riemann equations in U.

Remark 1 In the ‘if ’ part of this theorem, the continuity of the partial derivatives ensures via,
MVT, the existence of real derivative. Now the Cauchy-Riemann equations are equivalent to
the complex linearity of the real derivative. Hence we can appeal to the previous theorem. In the
‘only if ’ part, the only thing that is not proved already is the continuity of the partial derivatives.
We shall not prove it here. It will follow once we show that any Cauchy differentiable function
has a continuous derivative. Indeed, we shall see later on that complex differentiability ensures
that the function has continuous derivatives of all orders!

(Looman-Menchoff)

Theorem 6 Let U be an open subset of C and f : U → C be a continuous function, f = u+ıv.

Suppose the partial derivatives of u, v exist and satisfy Cauchy-Riemann equations throughout
U. Then f is holomorphic in U.

Remark 2 We have seen a number of examples of functions that are complex differentiable at
a point but not so at any other points in a neighborhood. As far as the differentiation theory
is concerned such functions are not of much use to us. We would like to concentrate on those
functions which are continuously differentiable in some non empty open subsets. Such functions
will be called ‘holomorphic.’ Once again, we emphasize the fact that holomorphic functions are
those which have continuous first order partial derivatives satisfying C-R equations in a non
empty open subset of C. However, in practice, using C-R equations to see whether a function

2
is holomorphic or not, would be the last thing that we would like to do. We should have a large
class of holomorphic functions readily known to us and then often a new one could be expressed
in some nice way in terms of these known ones. To this end, the next two lectures will be
devoted to the task of getting familiar with the so called elementary functions.

Exercises:

1. Write down all possible expressions for the Cauchy derivative of a complex function f =
u + ıv in terms of partial derivatives of u and v.

2. Write down as many details as possible for the proof of ‘if’ part of theorem 5.

3. Verify that the functions <z, =z, |z| and z̄ do not satisfy CR equations.

4. Show that the function f (z) = z<z is complex differentiable only at z = 0 and find f 0 (0).

Some optional Material for those of you who can

The operators ∂z and ∂z̄ .
Consider

the

following
differential

operators:
∂ 1 ∂ ∂ ∂ 1 ∂ ∂
∂z
:= 2 ∂x − i ∂y ; ∂ z̄ := 2 ∂x + i ∂y ;
We claim that for f = u + iv the following three conditions are equivalent to each
other.
¯
(i)f satisfies CR equations. (ii) ∂f ∂ z̄
= 0. (iii) ∂∂zf = 0.
Sol: ∂f
∂ z̄
= 12 (ux + ıuy + ıvx − vy ) = 0 iff CR equations hold. This proves equivalence of
(i) and (ii). The proof of equivalence of (i) and (iii) is similar.
Next we establish chain rule in terms of these operators.
Now let f : Ω1 → Ω2 , g : Ω2 → Ω1 be any two functions with partial derivatives.
We take the variables in the domain Ω1 to be z = x + iy and those in Ω2 to be
w = u + iy. For brevity of exposition let us denote ∂f ∂z
by fz etc. Then the chain rule
can be stated as follows:

(g ◦ f )z = gw fz + gw̄ f¯z ; (g ◦ f )z̄ = gw fz̄ + gw̄ f¯z̄ (♥).

By symmetry, it is enough to establish the first one of (♥). So, we put f (z) = w =
u + iv. Then by definition, we have
2(g ◦ f )z = (g ◦ f )x − i(g ◦ f )y = (gu ux + gv vx ) − i(gu uy + gv vy ).
4gw fz = (gu − igv )(fx − ify ) = gu fx − gv fy − i(gu fy + gv fx ).
4gw̄ f¯z = (gu + igv )(f¯x − if¯y ) = gu f¯x + gv f¯y − i(gu f¯y − gv f¯x ).
Now adding the last two identities and observing that

fx + f¯x = 2ux ; fy + f¯y = 2uy ; fx − f¯x = 2ivx , fy − f¯y = 2ivy ;

we obtain the result.

3
Remark 3 The computations that we have carried out has the following interpre-
tation: Let f (T, S) be a polynomial (or a power series) in two variables S, T over
the complex numbers. Let g(z) = f (z, z̄) be the function obtained by substitution.
Then
∂f
gz = (z, z̄).
∂T
We shall illustrate the usefulness of this in the following example.