Complex Differentiation 1

Chapter One

COMPLEX DIFFERENTIATION

1. Revision
A complex number is simply an ordered pair of real numbers. Therefore, any vector in the xy
plane corresponds to a complex number, and vice-versa. As a rule we shall write x + iy for the
pair (x, y); the special symbol i represents the pair (0, 1), which in vector analysis is usually
written ̂. Indeed, many engineering books (and the programming language python) use the
notation x + jy, which is probably more intuitive. Moreover, x and y are called the “real” and
“imaginary” part of z, respectively, despite the fact that they are both real numbers. All this
can be confusing at the beginning, but you’ll get used to it.
Addition of complex numbers corresponds to addition of vectors by the “parallelogram
law”. Multiplication of a complex number by a real number s has the effect of scaling the
corresponding vector; if s is negative, then the direction is also changed.
Multiplication between complex numbers is defined by the formula

def
(x + iy) · (a + ib) = (xa − yb) + i (xb + ya).

It follows immediately from this definition that

i2 = −1.

It’s also easy to show that in this way all standard properties of “usual” multiplication (com-
mutative, associative and distributive) are preserved: convince yourself of this.
In case you wonder, it’s impossible to carry this idea into 3-dimensional vector spaces: it
has been shown that no algebra may be constructed over triplets of real numbers, preserving all
the “usual” properties of addition and multiplication.

◮Definition: If z = x + iy is a complex number, then the quantity z = x − iy is called the

complex conjugate of z. ◭

Multiplying a complex number by its own conjugate, one always gets a real quantity:

(x + iy) · (x − iy) = x2 + y 2 + i(xy − xy) = x2 + y 2 .

The right-hand side of this expression is never negative; clearly, it’s zero only if x = y = 0 (i.e., if
z = 0). Except for this special case, the product of a complex number by its complex conjugate
is a positive real number.
2 Complex Differentiation

Division between complex numbers is done by the simple algorithm

x + iy x + iy a − ib (x + iy)(a − ib)
= · = .
a + ib a + ib a − ib a 2 + b2

◮Example 1 Divide 3 + i4 by 1 + i7.

Solution:
3 + i4 (3 + i4)(1 − i7) (3 + 28) + i(−21 + 4)
= = = 0.62 − i 0.34 ◭
1 + i7 12 + 72 50
◮Definition: For a complex number z = x + iy, the expression
√ p
|z| = zz = (x + iy) · (x − iy)

is called the modulus of z. Note that the modulus of a complex number is equal to the length
of the corresponding vector. ◭

Using this notation, the formula for division may be written

z1 z1 · z 2
= .
z2 |z2 |2

THE TRIANGLE INEQUALITY

In high-school geometry, you saw that in any tri-
angle the length of each side is always shorter than
the sum of the other two, and greater than the dif- y
ference.
Hence, if a, b and c are complex numbers such
that a + b = c, and we interpret complex numbers
as vectors, then the corresponding vectors form a
triangle with sides having length |a|, |b| and |c|, b
respectively, as shown in the picture. It follows c
that
|a| − |b| ≤ |c| ≤ |a| + |b|,
and finally, substituting c with a + b, that a
|a| − |b| ≤ |a + b| ≤ |a| + |b|. x

This important result is generally known as the triangle inequality.

POLAR FORM OF COMPLEX NUMBERS
Note the identity  
p
2 2
x iy
x + iy = x + y · p +p ;
x2 + y 2 x2 + y 2
since  2  2
x y
p + p ≡ 1,
x2 + y 2 x + y2
2
Complex Differentiation 3

then clearly the expressions in brackets on the left-hand side are the cosine and sine, respectively,
of some
p angle φ which is called the argument of x + iy. Recall that, if z = x + iy, then
|z| = x2 + y 2 . Therefore, we may write

z = |z| (cos φ + i sin φ),

where φ and |z| are real numbers; φ is the argument and |z| is the modulus of z.
Note however that the argument is not uniquely defined; adding 2π radians to φ does not
affect cos φ and sin φ.
Finally, applying Euler’s formula, we get:

z = |z| eiφ .

This is the polar form of the complex number z.

The polar form helps to visualize the multiplication and division of complex numbers: it’s
not difficult to show that

|z1 · z2 | = |z1 | · |z2 |, and arg(z1 · z2 ) = arg z1 + arg z2 .

In other words, when two complex numbers are multiplied, the modulus of the product equals
the product of the moduli, and the argument of the product equals the sum of the arguments.
Another important application of the polar form comes in the calculation of powers: if

z = |z| eiφ ,

where φ = arg z, then for any integer n we find that

z n = |z|n einφ = |z|n (cos nφ + i sin nφ),

which is called de Moivre’s formula.

√
◮Example 2 Calculate (1 − i 3)11 by de Moivre’s formula.
√
Solution: One could, of course, multiply z = 1 − i 3 by itself 11√times, but doing that would be
missing the point. By de Moivre’s formula,
√ we note that |z| = 1 + 3 = 2, and hence (writing
φ for arg z), cos φ = 1/2, sin φ = − 3/2. Because cos φ is positive and sin φ is negative, φ
must lie in the fourth quadrant: hence, φ = 5π/3 is correct, but φ = −π/3 is equally good. In
general, one should write φ = 5π/3 + 2kπ, where k may be any integer. So,

z = 2 ei5π/3+i2kπ .

Raising this expression to the 11th power, we get:

z 11 = 211 ei55π/3+i22kπ .

We may ignore the factor ei22kπ because it’s equal to 1. Moreover 55π/3 = π/3 + 18π; hence
√
cos(55π/3) = cos π/3 = 1/2, sin(55π/3) = sin π/3 = 3/2.

So, we finally find that

√
z 11 = 2048 [cos(π/3) + i sin(π/3)] = 1024 (1 + i 3). ◭
4 Complex Differentiation

De Moivre’s formula may be applied in reverse to find roots of complex numbers, but then one
must be careful: look at the following example.
√
◮Example 3 Find z 1/5 , if z = 1 − i 3.
Solution: Using the results of the preceding example, we note that arg z may be written in any
of the forms
arg z = 5π/3, = 11π/3, = 17π/3, = 23π/3, = 29π/3, etc.
All these values correspond to the same z because they differ by multiples of 2π. However, when
we take the 5th root, we must divide the argument by 5, and this does make a difference. We
get results like
√ √ √ √ √ √
2 eiπ/3 , 2 ei11π/15 , 2 ei17π/15 , 2 ei23π/15 , 2 ei29π/15 , 2 ei35π/15 ,
5 5 5 5 5 5

and so on. Further analysis (e.g., calculating cos π/3, cos 11π/15 etc.) shows that the first five
complex numbers in the list are different, but the sixth one is identical to the first. This had to
be expected, because 35π/15 and π/3 differ exactly by 2π. In the same way, one may discover
that the seventh value
√ coincides with the second, the third value with the eighth, and so on.
The expression (1 − i 3)1/5 has exactly five distinct values in the complex field. ◭

This example may be adapted to roots of any integer order: one finds that for any complex z
and integer n, the expression z 1/n has always exactly n distinct complex values.
2. Elementary Complex Functions
A complex function is a law that associates a complex variable z with one or more complex
variable(s) w1 , w2 , w3 , etc. If only one variable w is determined by z, the function is called
single-valued; if more than one, the function is called multi-valued.
For example f (z) = z 3 and f (z) = (z + 3)/(5z + i4) are single-valued; f (z) = z 1/3 is
multi-valued (three-valued, if you wish to be precise).
A single-valued complex function may be seen as an ordered pair of real functions of the
variables x and y; however, in many applications one doesn’t have to get into such details.

◮Example 4 Express f (z) = z 4 as a pair of real functions.

Solution: We find that
z 4 = (x − iy)4 .
Expanding the right-hand side, we get
(x − iy)4 = x4 − 4x3 (iy) + 6x2 (iy)2 − 4x(iy)3 + (iy)4 =
= x4 − 6x2 y 2 + y 4 − i 4(x3 y − x3 ).
As required, we have separated the real and imaginary part of f (z), which are, respectively,
u(x, y) = x4 − 6x2 y 2 + y 4 v(x, y) = −4(x3 y − x3 ).
This example illustrates another important point: the dependence on z may occur via expressions
like z, |z|, arg z, etc. ◭

The real and imaginary part of a complex number are themselves real functions of z: as a matter
of fact, it’s easy to see that
z+z z−z
x= , y= ;
2 i2
verify this. However, many applications of complex functions are based on functions that do not
depend on anything but the variable z alone. We’ll come back to this point.
Complex Differentiation 5

POLYNOMIALS AND RATIONAL FUNCTIONS

Polynomials and rational functions are perhaps the easiest examples of complex functions: for
instance,
√
f (z) = 7 + (3 − i2) z + (5 + i4) z 2 + (−3 − i) z 3 + (−i5) z 4 + ( 2 + i) z 5

is a fifth-degree polynomial in z alone.

Technically, the function f (z) = constant is regarded as a polynomial of degree zero. The
fundamental theorem of algebra says that for every polynomial in z of degree ≥ 1 one may find
at least a point λ where the polynomial vanishes (that is, both the real and imaginary part are
zero).
An immediate consequence of the fundamental theorem of algebra is that every polynomial
pn (z) of degree n may always be factorized as the product of n linear terms:

pn (z) = an · (z − λ1 ) · (z − λ2 ) · (z − λ3 ) · · · (z − λn ),

where an is the coefficient of the leading term. Some of the numbers λ1 , λ2 , . . . may coincide.
They are called the roots or also the zeroes of the polynomial; when some of them coincide, we
say the polynomial has a multiple root (or roots).
Rational functions are, by definition, fractions where both the numerator and the denom-
inator are polynomials, and the denominator is not constant. Obviously, if numerator and
denominator have common factors, they may be cancelled. If f (z) is a rational fraction without
common factors, then (by the fundamental theorem of algebra) there is at least one point λ
where f (λ) = 0, and at least one point µ where 1/f (µ) = 0, i.e., |f | becomes infinite.
THE EXPONENTIAL AND RELATED FUNCTIONS
The exponential function is defined by the series
∞
X zk
exp z = .
k!
k=0

It may be shown that this series converges absolutely at every point z of the complex plane. It
is often written ez , and we shall use this notation too, although we’ll see later that this is not
entirely correct.
Since this series converges absolutely, one may sum the even powers and the odd powers
separately:
X zk X zk
ez = + . (1)
k! k!
k=even k=odd

Bringing in the definitions

X zk X zk
cosh z = , sinh z = ,
k! k!
k=even k=odd

one gets immediately:

ez = cosh z + sinh z.
Reading these formulas backwards, one gets
ez + e−z ez − e−z
cosh z = , sinh z = . (2)
2 2
6 Complex Differentiation

The exponential function has two important properties, viz.

b
a
ea+b = ea · eb , eab = ea = eb ,

which hold for all complex a and b. Furthermore, one may show that

cosh2 z − sinh2 z = 1

for every complex z, and also that

cosh(a ± b) = cosh a cosh b ± sinh a sinh b, sinh(a ± b) = sinh a cosh b ± cosh a sinh b

for every complex a and b.

Complex sine and cosine are likewise defined by power series. Their definitions differ from
the ones for hyperbolic sine/cosine only by an alternating sign factor (−1)k :
∞ ∞
X z 2k X z 2k+1
cos z = (−1)k · , sin z = (−1)k · .
(2k)! (2k + 1)!
k=0 k=0

Recall that equation (1) is valid for any complex z. Hence, replacing z with iz in (1), we get:

X (iz)k X (iz)k
eiz = + .
k! k!
k=even k=odd

Now, note that i2 = −1, hence i4 = +1, i6 = −1 and so on. Raising i to an even power gives
always ±1, with alternating sign. From here, it’s easy to deduce that i3 = −i, i5 = +i, etc. As
an exercise, verify that the following expansions are valid:
∞ ∞
X z 2ℓ X z 2ℓ+1
eiz = (−1)ℓ · +i (−1)ℓ · =
(2ℓ)! (2ℓ + 1)!
ℓ=0 ℓ=0
= cos z + i sin z.

The last line is precisely Euler’s formula, which you saw in first year in the context of real
variables. Therefore, Euler’s formula may be extended to complex arguments.
Inverting Euler’s formula, one gets immediately:

eiz + e−iz eiz − e−iz

cos z = , sin z = ; (3)
2 i2
note the i in the denominator on the right-hand side, but not on the left-hand side. Note that
equations (2) and (3) are very similar. As an exercise, show that

cos iz = cosh z, cosh iz = cos z, sin iz = i sinh z, sinh iz = i sin z; (4)

these formulas are very useful. Also, from (3) one may immediately deduce that

cos2 z + sin2 z = 1
Complex Differentiation 7

for every complex z, and that

cos(a ± b) = cos a cos b ∓ sin a sin b, sin(a ± b) = sin a cos b ± cos a sin b

for every complex a and b; note the sign inversion on the left-hand side.
THE LOGARITHM
The complex logarithm is defined by requiring that the identity

elog z = z,

be valid for all complex z 6= 0. To understand this definition, it’s convenient to start from the
polar form of complex numbers,
z = |z| ei arg z , (5)
where one may always substitute |z| with eln |z| , because |z| is real (recall that we assume z 6= 0),
and ln |z| represents its natural logarithm, which you studied in first-year calculus. Having done
this substitution, we find that (5) becomes

z = eln |z|+i arg z .

Finally, by comparison, we get that

log z = ln |z| + i arg z.

This is the definition of the complex logarithm; for example, see R.P. Boas, Invitation to Complex
Analysis, Random House (New York), 1987. Note immediately that, unlike the real logarithm,
which is single-valued, the complex logarithm is multi-valued because arg z may be increased or
decreased by any integer multiple of 2π. The complex logarithm is an infinitely-valued
function. For this reason, among others, we keep the different symbols, “log” and “ln”, for the
complex, and the real, logarithm respectively.
The most important property of the logarithmic function, i.e.,

log(z1 · z2 ) = log z1 + log z2

is preserved, because

ln |z1 · z2 | = ln |z1 | + ln |z2 |, arg(z1 · z2 ) = arg z1 + arg z2 ,

and so
log(z1 · z2 ) = ln |z1 | + ln |z2 | + i arg z1 + i arg z2 = log z1 + log z2 .
Unlike its real counterpart, the complex logarithm is defined for all complex z (except z = 0, of
course). Negative numbers are not a problem, because eiπ = cos π + i sin π = −1.

◮Example 5 Observe that −3 = e+iπ · eln 3 · ei2kπ , where k may be any integer. Therefore,

log(−3) = ln 3 + iπ + i2kπ [k = 0, ±1, ±2, . . .]

Everything comes at a price. In order to give a meaning to the logarithm of a negative number,
we must give up the one-valued-ness of the function. ◭
8 Complex Differentiation

◮Example 6 Find log(1 + i).

√
Solution: First of all, we note that |1 + i| = 2, hence

√ √
 
1 1

1+i= 2 √ +i√ = 2 cos π/4 + i sin π/4 =
2 2
√
= 2 eiπ/4 .
√
So, |i + i| = 2 and arg(i + i) = 41 π + 2kπ, where k may be any integer. Hence,
√
1+i= 2 eiπ/4+i2kπ ;

note that ei2kπ = 1. Therefore,

1
log(1 + i) = 2 ln 2 + iπ/4 + i2kπ,

where k is an arbitrary integer. ◭

◮Example 7 Find log(−5 + i12).

Solution: First of all, observe that if z = −5 + i12, then

|z| = 13,

and z is a point in the second quadrant (x is negative, y is positive). Hence arg z is an angle
between π/2 and π, plus an arbitrary integer multiple of 2π.
Let’s find first an angle between 0 and 2π radians. Given that

sin(arg z) = 12/13, cos(arg z) = −5/13,

and using the identity sin θ = sin(π − θ), we get that

arg z = arccos(−5/13) = π − arcsin 12/13;

recall that the real function arcsin x ranges from −π/2 and +π/2, but arccos x ranges from 0 to
π. It finally follows that log(−5 + i12) = ln 13 + i(2k + 1)π − i arcsin 12/13. ◭

◮Example 8 Find all the solutions of cos z = 2.

Solution: We write
eiz + e−iz
= 2.
2
Substituting eiz = λ, it follows:
λ + λ−1
= 2,
2
λ2 + 1 = 4λ.
Complex Differentiation 9

√
This√is a quadratic
√ equation in λ; solving
√ it, we find
√ that λ = 2 ± 3. Note, however, that
2 − 3 = 1/(2 + 3): hence, ln(2 − 3) = − ln(2 + 3). Going back to the variable z, we find
that  √
iz 2 + 3√
e =
1/(2 + 3).
Taking the complex logarithm of both sides, we get:
√
iz = ± ln(2 + 3) + i2kπ [k = 0, ±1, ±2, . . .]

(the term i2kπ may be inserted on either side of the equality, because k may be positive or
negative). So, dividing through by i, we find that
√
z = ±i ln(2 + 3) + 2kπ.

Note that there is no real solution; note also that all solutions differ by multiples of 2π: the
cosine remains a periodic function with period 2π, even when complex values are allowed. ◭

POWERS
As a rule, the expression az , where a and z may be complex numbers, is always interpreted as

a z = ez log a
,

where log a is defined up to an integer multiple of i2π. It follows immediately that

a z = ez ln |a|+iz arg a+i2zkπ

.

Now we have three possibilities:

• If z is an integer, say z = m, then 2mk is also an integer, and so ei2mkπ = 1 (regardless of
k). Hence, am has only one value, the one obtained multiplying a by itself m times.
• If z is a rational fraction, say z = m/n, where m and n are integers without common factors,
then ei2mkπ/n has exactly n different values. This case was illustrated in example 3.
• If z is not an integer and is not rational, then the expression az has an infinite number of
values.

◮Example 9 Find (1 + i)3 , using example 6.

1
Solution: In example 6 we found that log(1 + i) = 2 ln 2 + iπ/4 + i2kπ. It follows that

(1 + i)3 = e(3/2) ln 2+i3π/4+i6kπ .

But then we see immediately that

√ √
e(3/2) ln 2 = e(ln 8)/2 = 8, ei3π/4 = (−1 + i)/ 2, ei6kπ = 1.

So, finally, √ √
(1 + i)3 = 8 · (−1 + i)/ 2 = −2 + i2.
As an exercise, verify that (1 + i) · (1 + i) · (1 + i) = −2 + i2. ◭
10 Complex Differentiation

◮Example 10 Find (1 + i)1/3 .

Solution: Proceeding like in the previous example, we get immediately that

(1 + i)1/3 = e(ln 2)/6 + iπ/12 + i2kπ/3.

It follows that
√
e(ln 2)/6 = eiπ/12 = cos π/12 + i sin π/12.
6
2 and

However,
1 if k = . . . 0, 3, 6, 9, . . .

1
√
ei2kπ/3 = 2 (−1 + i 3)
if k = . . . 1, 4, 7, 10, . . .
√
1
2 (−1 − i 3)
if k = . . . 2, 5, 8, 11, . . .

√
So, the possible values of (1 + i)1/3 are determined by multiplying 6 2 (cos π/12 + i sin π/12) by
each of the (three) possible values of ei2kπ/3 listed above. ◭
√
2
◮Example 11 Find all the values of the complex power 2 .
Solution: By definition,
√ √ √ √ √
2 2 log 2 2 (ln 2+i2kπ) 2 ln 2
2 =e =e =e · ei 8kπ
. [k = 0, ±1, ±2, . . .]
√
The expression ei 8kπ
has an infinite number of different values, as k varies:
√
ei 8kπ
=1 if k = 0,
√ √
= cos 8π ± i sin 8π, if k = ±1,
√ √
= cos 2 8π ± i sin 2 8π, if k = ±2,
√
and so on. Because 8 is irrational, the numbers generated in this √ way—and there is an infinity
of them—are√
all different. Multiplying each one of them by exp(
√
2 ln 2), we get all the complex
2 2
values of 2 . One of them coincides with the real power 2 that you met in high school. ◭

◮Example 12 Find ii .
Solution: Note that |i| = 1 and i = eiπ/2 . Hence, log i = iπ/2 + i2kπ. We find immediately that

ii = e−π/2−2kπ ; [k = 0, ±1, ±2, . . .]

in this case the expression on the right is infinite-valued, but all its values are real. ◭

If you are observant, you’ll have noticed that, according to the definitions given so far, technically,
the expression ez is multi-valued unless z is an integer. If you spotted this, congratulations.
However, virtually every author since the days of Euler has used ez as a convenient alternate
for exp z, and we’ll follow this (wrong) practice in these notes.
3. Limits, Continuity and Derivatives
Complex limits, continuity and differentiation are defined exactly as you would expect, and they
mirror the definitions given in real analysis.
Complex Differentiation 11

There are, however, some points to note. First of all, always bear in mind that the complex
notation is actually a short-hand way of expressing two real functions of two real variables. In
other words, 
z = x + iy,
w = f (z) ⇐⇒
w = u(x, y) + i v(x, y).
Therefore, if z0 = x0 + iy0 is a point in the xy plane, then by definition

lim f (z) = lim u(x, y) + i lim v(x, y), (6)

z→z0 x→x0 x→x0
y→y0 y→y0

where the limits on the right-hand side are limits of functions of two variables, which you’ll find
explained in your old calculus textbook.

◮Definition: We say that

lim f (z) = A + iB [A, B real]
z→z0

if, and only if, both limits appearing on the right-hand side in (6) exist finite, and

lim u(x, y) = A, lim v(x, y) = B,

x→x0 x→x0 ◭
y→y0 y→y0

The concept of infinity may now be introduced, but carefully. Most books don’t attach any
meaning to expressions like lim f (z) = ∞ if f is a complex function, and neither shall we. On
the other hand, writing
lim |f (z)| = ∞
z→z0
p
simply means that the real function u2 (x, y) + v 2 (x, y) gets larger than any arbitrary bound,
for all (x, y) sufficiently close to (x0 , y0 ), and in this sense it’s acceptable. It is also possible to
give a meaning to “limit as z goes to infinity”:
def
◮Definition: lim f (z) = lim f (1/z), if the limit on the right-hand exists. ◭
z→∞ z→0

Continuity is defined like for real functions.

◮Definition: A complex function f (z) is said to be continuous at a point z0 if f (z0 ) is defined

and lim f (z) = f (z0 ). ◭
z→z0

All the theorems on limits of real functions of two variables are applicable here, so we shall not
spend any time on them. Derivatives of complex functions are introduced in a natural way.

◮Definition: A complex function f is said to be differentiable at a point z0 if, and only if,

f (z0 + ∆z) − f (z0 )

lim exists finite.
∆z→0 ∆z
The value of such a limit is then called the derivative of f and denoted f ′ or also df /dz. ◭

Again, it may be shown that the differentiation rules you learnt in first year are formally ap-
plicable to complex derivatives as long as these exist: the sum rule, the product rule, the chain
12 Complex Differentiation

rule and so on. There’s even an equivalent form of de l’Hospital’s theorem which may be applied
to complex limits.
For functions expressed by simple formulas, the existence of the derivative may be estab-
lished in a simple way, not unlike what you saw in first year calculus.

◮Example 13 Find the derivative of f (z) = z 4 at an arbitrary point z0 .

Solution: By definition,
(z0 + ∆z)4 − z04
f ′ (z0 ) = lim .
∆z→0 ∆z
Rather than separate the real part u and imaginary part v of the right-hand side and then
use (6), we simplify it by expanding (z0 + ∆z)4 by the binomial formula:

z04 + 4z03 · ∆z + 6z02 · (∆z)2 + 4z0 · (∆z)3 + (∆z)4 − z04

f ′ (z0 ) = lim =
∆z→0 ∆z
4z03 · ∆z + 6z02 · (∆z)2 + 4z0 · (∆z)3 + (∆z)4
= lim .
∆z→0 ∆z

The common factor ∆z may now be cancelled from numerator and denominator, yielding

4z03 + 6z02 · ∆z + 4z0 · (∆z)2 + (∆z)3

f ′ (z0 ) = lim .
∆z→0 1

Now it’s clear that there’s no need to look into u(x, y) and v(x, y) for this specific example.
Since ∆z = ∆x + i∆y, and both ∆x and ∆y go to zero in the numerator only, we may simply
replace ∆z in the equation above with 0 + i0. It follows that

f ′ (z0 ) = 4z03 .

The right-hand side may be separated into a real and imaginary part, if one so wishes, but there
is absolutely no need to do it. Actually, it’s better to leave the answer in this form because the
result and the procedure mirror what you saw in first-year calculus. ◭

Proceeding in the same way, it’s easy to show that f (z) = z n is differentiable at any point in the
plane if n is a positive integer, and that df /dz = n z n−1 : exactly like in real calculus. Moreover,
we have mentioned that the usual rules for calculating real derivatives are applicable to complex
functions, as long as such derivatives exist. In particular, by the “sum rule”, we deduce that

d
a0 + a1 z + a2 z 2 + · · · + an z n = a1 + 2a2 z + 3a3 z 2 + · · · + nan z n−1 ,


dz

for every integer n and arbitrary parameters a0 , a1 , . . ., an .

◮Conclusion: Polynomials in z are differentiable at any point in the complex plane and their
derivative may be found by the same rules as in real calculus. ◭

However, differentiability is a very strong requirement for complex functions, much stronger
than continuity. This is a completely different scenario from what you saw in first-year calculus.

◮Example 14 Where is the function f (z) = z (i) continuous? (ii) differentiable?

Complex Differentiation 13

Solution: By definition, f (z) = x − iy, hence both the real and the imaginary part, u = x and
v = −y, are continuous at every point in the plane, so f (z) is continuous everywhere.
However, for differentiability we need to consider

∆f x + ∆x − i(y + ∆y) − (x − iy)

= ;
∆z ∆x + i∆y

simplifying, we get immediately that

∆f ∆x − i∆y (∆x)2 − (∆y)2 − i2∆x∆y

= = .
∆z ∆x + i∆y (∆x)2 + (∆y)2

Let’s consider the real part of the last expression, and let ∆x and ∆y go to zero. We’ll now see
that
(∆x)2 − (∆y)2
lim does not exist.
∆x→0 (∆x)2 + (∆y)2
∆y→0

This may be proven as you learnt in the first semester, by letting ∆y = m∆x, where m =
constant, and then ∆x → 0. The result depends on the parameter m:

(∆x)2 − m2 (∆x)2 1 − m2
lim = ,
∆x→0 (∆x)2 + m2 (∆x)2 1 + m2

therefore the limit of the real part doesn’t exist. No need to consider the imaginary part: it
follows immediately that f (z) = z is not differentiable anywhere.
In real analysis, there are indeed functions that are continuous everywhere but differentiable
nowhere; however, they are constructed by fairly advanced methods and, by engineering stan-
dards, “look weird”. The function z, on the other hand, is probably one of the easiest examples
that come to mind, and yet it is everywhere continuous and nowhere differentiable. ◭

◮Example 15 Show that f (z) = |z|2 is not differentiable except at z = 0.

Solution: By definition, f (z) = x2 + y 2 ; hence the imaginary part of this function is identically
zero. The real part represents a paraboloid with vertex at the origin and, as a real function of
x and y, is differentiable everywhere. However, in the sense of complex functions, we find that

∆f (x + ∆x)2 + (y + ∆y)2 − x2 − y 2 2x ∆x + (∆x)2 + 2y ∆y + (∆y)2

= = .
∆z ∆x + i∆y ∆x + i∆y

Proceeding like in the preceding example (let ∆y = m∆x and then ∆x → 0) one finds that the
right-hand side does not tend to any limit, except for the special case where x = y = 0. ◭

4. Cauchy-Riemann Equations
It’s easy to find a simple, necessary condition for complex differentiability. Suppose df /dz exists
at a certain point. In other words, suppose that

∆u + i∆v
lim = A + iB
∆x→0 ∆x + i∆y
∆y→0
14 Complex Differentiation

where u and v are, as usual, the real and imaginary part of f (z). Recall that the existence of
the limit implies that the limit is always the same, if we allow ∆x and ∆y to approach zero in
any way we please.
Very well then, let’s allow ∆z to approach zero along a straight, horizontal path: let ∆y = 0,
and let only ∆x be variable. In other words, consider the limit

∆u + i∆v ∆u ∆v
lim = lim + i lim .
∆x→0 ∆x ∆x→0 ∆x ∆x→0 ∆x

This leads us back to the definition of partial derivatives of real functions, which you saw in the
first semester. Hence, by definition:

∆u + i∆v ∂u ∂v
lim = +i .
∆x→0 ∆x ∂x ∂x
Next, we allow ∆z to approach zero along a straight, vertical path. This means we let ∆x = 0,
∆y = variable, and allow ∆y → 0. Proceeding exactly like before, we find:

∆u + i ∆v ∂u ∂v
lim = −i + .
∆y→0 i ∆y ∂y ∂y

The two limits must be equal (because we are assuming that df /dz exists). Therefore,

∂u ∂v ∂u ∂v
+i = −i + .
∂x ∂x ∂y ∂y

Separating real and imaginary part, we finally get that

∂u ∂v ∂u ∂v
= , and =− . (7)
∂x ∂y ∂y ∂x

These simple equations, called Cauchy-Riemann Equations, are a direct consequence of the
existence of df /dz. Hence, they are a necessary condition for differentiability. In other words,
a function f (z) is differentiable at a certain point only if its real part u and imaginary part v
satisfy, at that point, Cauchy-Riemann equations. If they don’t, we deduce immediately that f
is not differentiable.

◮Example 16 Show that f (z) = z is not differentiable anywhere (this is example 14).
Solution: The real and imaginary part of f are, respectively, u = x and v = −y. We find that

∂u ∂v
=1 = −1.
∂x ∂y

The first Cauchy-Riemann equation is always violated: ux 6= vy . Hence, f is not differentiable:

no need to examine the second Cauchy-Riemann equation. Compare this method with that of
example 14; which one is easier? ◭

Cauchy-Riemann equations provide a necessary condition for differentiability: where they are
not satisfied, the corresponding function is not differentiable. But obviously a sufficient condition
would be more desirable, i.e., a simple way to determine where a function is indeed differentiable,
rather than where it is not.
Complex Differentiation 15

◮This is a difficult point. Cauchy-Riemann equations alone are not sufficient. However, it may
be shown that if
(i) the real part u and the imaginary part v of a complex function f (z) satisfy the Cauchy-
Riemann equations, and
(ii) the partial derivatives ux , uy , vx and vy are continuous at all points in an open neighbor-
hood, then f (z) is differentiable in such neighborhood. ◭

The proof of this theorem hinges on the concept of differentiability of a function of several real
variables, which is no longer taught to engineering students at this University. It may be found
in any good book on advanced calculus: see, for instance, Courant & John, Introduction to
Calculus and Analysis, Vol. 2, Sect. 8.2, Springer-Verlag (Berlin), 1989.
Recall, an open neighborhood in the xy plane is a set that consists only of internal points.
A good example is a circle with the boundary removed. Take, for instance, the circle with radius
1 and center at the origin: the point (1, 0) sits on the boundary therefore it is not “in” the open
neighborhood. On the other hand, the point (0.999999999, 0) is internal, even though it’s very
close to the previous one: it is “in” the neighborhood. Every internal point, no matter how close
it is to the boundary, is always surrounded by a neighborhood where every point is internal—
such a neighborhood may be very small, but it would still contain an infinity of points, and they
would all be internal.
If you have followed so far, you’ll realize that the shape of an open neighborhood doesn’t
have to be a circle: it may be a triangle, a cardioid, a star. You may think in terms of circles,
if it helps to draw pictures in your mind, but you don’t have to.
Finally, you may quibble that the requirement (ii) is over-stated: if (i) holds, then ux = vy
and uy = −vx (Cauchy-Riemann), hence if ux and uy are continuous, then vx and vy must also
be continuous. Did you notice that? Good. However, as they said in Latin, repetita juvant: to
repeat things, helps.
One last remark before we look at examples: observe that from the partial derivatives in (7),
one may actually derive, with no extra work, the derivative df /dz. The point is that—again—
where lim∆z→0 ∆f /∆z exists, ∆z may approach zero from any direction, including (but not
necessarily) vertically and horizontally. In the first case, ∆x = 0 and hence ∆z = i∆y; in the
second, ∆y = 0 and hence ∆z = ∆x. Therefore

df ∆u + i∆v
= lim = ux + ivx ,
dz ∆x→0 ∆x
∆u + i∆v
= lim = −iuy + vy .
∆y→0 i∆y

Obviously, both formulations are equivalent (Cauchy-Riemann, again).

Having sorted out the technical points, we’ll now see that in practice, for most functions of
everyday use in engineering, Cauchy-Riemann equations are necessary and sufficient.

◮Example 17 Show that f (z) = ez is differentiable everywhere and find its derivative.
Solution: We note that
ez = ex+iy = ex (cos y + i sin y);

hence
u = ex cos y, v = ex sin y.
16 Complex Differentiation

It follows that

ux = ex cos y, uy = −ex sin y, vx = ex sin y, vy = ex cos y.

Hence, we find that ux = vy and uy = −vx . Cauchy-Riemann equations are satisfied for every
x and y. Furthermore, it is obvious that the four partial derivatives listed above are continuous
functions of x and y in the whole plane, and so the second requirement is also satisfied. Therefore
ez is differentiable everywhere.
Its derivative may be found as follows:

d ez
= ux + i vx = ex (cos y + i sin y) = ex eiy = ex+iy .
dz
The last term is clearly equal to ez , hence df /dz = ez . ◭

◮Example 18 Show that sin z is differentiable everywhere and d(sin z)/dz = cos z.
Solution: We note that

sin z = sin(x + iy) = sin x cos iy + cos x sin iy =

= sin x cosh y + i cos x sinh y;

the last step follows from (4). Hence

u = sin x cosh y, ux = cos x cosh y, uy = sin x sinh y,

and
v = cos x sinh y, vx = − sin x sinh y, vy = cos x cosh y.
Clearly u and v are continuous at all points in the xy plane, and their derivatives satisfy Cauchy-
Riemann equations (7), so sin z is differentiable everywhere. It follows immediately that

ux + ivx = cos x cosh y − i sin x sinh y.

Substituting cosh y = cos iy and i sinh y = sin iy, we get

ux + ivx = cos x cos y − sin x sin iy = cos(x + iy).

The last term on the right-hand side is, of course, cos z, and so finally:

d sin z
= cos z,
dz
as expected. ◭

In a very similar way, one may show that ecz is also differentiable, c being a (possibly complex)
constant. And from here one may go on and find the derivatives of all trigonometric and
hyperbolic functions.

◮Example 19 Find the derivative of f (z) = z 4 using the Cauchy-Riemann equations (this is
example 13 revisited)
Complex Differentiation 17

Solution: By the binomial formula,

(x + iy)4 = x4 + i4x3 y − 6x2 y 2 − i4xy 3 + y 4 .

It follows that u = x4 − 6x2 y 2 + y 4 and v = 4x3 y − 4xy 3 ; hence that

ux = 4x3 − 12xy 2 = vy , vx = 12x2 y − 4y 3 = −uy ,

which incidentally shows that Cauchy-Riemann conditions hold. We deduce that

ux + ivx = 4x3 − 12xy 2 + i(12x2 y − 4y 3 ) = 4(x3 + i3x2 y − 3xy 2 − iy 3 ) = 4(x + iy)3 ,

and the last term on the right-hand side is clearly 4z 3 , as expected. ◭

d
In this way one may derive again that dz z n = nz n−1 , but the method outlined in example 13
is perhaps more direct.

◮Example 20 Examine the differentiability of f (z) = log z.

Solution: By definition, log z = ln |z| + i arg z; note that ln |z| = 21 ln(x2 + y 2 ).
Suppose initially that z lies in the first or fourth quadrant: then, x is positive and y ranges
from −∞ to +∞. Therefore, clearly, arg z = arctan y/x + 2kπ. We find that

u= 1
2 ln(x2 + y 2 ), v = arctan y/x + 2kπ,

and hence that

x y −y/x2 1/x
ux = , uy = , vx = , vy = .
x + y2
2 x + y2
2 1 + y 2 /x2 1 + y 2 /x2

Simple manipulations show that ux = vy and uy = −vx (Cauchy-Riemann). If z lies in the

second or third quadrant, then arg z = arctan y/x + (2k + 1)π (convince yourself of this). From
here one may proceed exactly like before.
Continuity of the partial derivatives, however, fails at the origin, because, for instance,
x
lim ux = lim doesn’t exist;
x→0 x→0 x2 + y2
y→0 y→0

a similar problem occurs with uy . Therefore f (z) = log z is differentiable everywhere except at
the origin.
To calculate the derivative, we write

x −y/x2 x − iy 1
ux + ivx = 2 2
+ i 2 2
= 2 2
= .
x +y 1 + y /x x +y x + iy

Hence, finally,
d log z 1
= .
dz z
Note the difference between
d ln |x| 1 d log z 1
= and = .
dx x dz z
18 Complex Differentiation

The modulus bars are correct on the left-hand side, because the formula is applied only to
the real function ln |x|. On the right-hand side they would be wrong, because log |z| is not
differentiable! Convince yourself of this. ◭

You’ll be surprised to learn that Cauchy-Riemann equations were not discovered by Cauchy nor
by Riemann, but by d’Alembert (1717-1783) in the course of his research on hydrodynamics.
Much later Cauchy (1789-1857), and especially Riemann (1826-1866), understood the role of
these equations in the theory of complex functions, and this is probably why they are still
named after them.
5. Holomorphic Functions
We have seen that for a function f = u + iv to be differentiable, it is sufficient that the partial
derivatives of u and v be continuous in an open neighborhood and satisfy Cauchy-Riemann
equations. Most functions of practical interest in engineering enjoy this property for all z, with
the possible exception of isolated points: for example, the logarithm is not differentiable only at
z = 0.

◮Definition: A complex function f = u+iv is said to be holomorphic at a point z0 if there is an

open neighborhood containing z0 such that u and v are continuous and satisfy Cauchy-Riemann
at every point in the neighborhood. ◭

If you followed so far, you understand that “being holomorphic” implies “being differentiable”.
But there is a reason for introducing a new abstruse word. Holomorphic means more than just
differentiable. It means that every point where f (z) is holomorphic is surrounded (at least up
to a certain distance) by a region where at every other point f (z) is holomorphic too. That’s
where the concept of open set plays its part, but we won’t get into more details.
It’s not difficult to create complex functions that have the opposite property of being differ-
entiable only at scattered, isolated points. We’ve already seen one such function, in example 15.
We shall also see, in the tutorial problems, an example of a function that is differentiable along
a whole straight line. More complicated examples of this kind may be constructed, but this
aspect of the theory is not essential in this course.
Holomorphy, on the other hand, is always extended over open sets; a function cannot be
holomorphic at just one point. In practice, almost all the complex functions used in engineering
are holomorphic almost everywhere.
For example, log z is holomorphic everywhere except at z = 0; tan z is holomorphic every-
where except at x = ± 21 π, ± 32 π, ± 52 π, etc. Here we have an infinity of points where the function
tan z is not holomorphic, but each one is surrounded by a sea of other points where holomorphy
holds. Therefore they are isolated singular points.
The only non-holomorphic functions you need to keep in mind for this course are f (z) = z,
which we have already discussed, and f (z) = |z|: and this one is actually a real function of a
complex variable.
HARMONIC FUNCTIONS
Not every function u(x, y) of two real variables may represent the real part of a holomorphic
function.
For example, there is no holomorphic function f (z) such that its real part is u = x3 + y 3 .
How can we be so sure? Elementary: suppose a function v exists such that u+iv is holomorphic.
Complex Differentiation 19

Then it must be ux = vy and uy = −vx (Cauchy-Riemann). Differentiating the Cauchy-Riemann

equations, we get: uxx = vyx = vxy = −uyy . It follows that

uxx + uyy = 0.

For the function we have in mind, which is u = x3 + y 3 , we find that uxx + uyy = 12xy, and so
there is no open neighborhood where uxx + uyy = 0.

◮Definition: A real function u(x, y) is called harmonic if it satisfies the equation uxx +uyy = 0,
which is called the Laplace equation. ◭

Therefore, a function u(x, y) may represent the real part of a holomorphic function only if it is
harmonic. Vice-versa, if a function u(x, y) is harmonic in an open neighborhood, then it’s the
real part of some holomorphic function. Similar conditions hold for the imaginary part.

◮Example 21 Let u = x2 − y 2 . Show that it is harmonic and find a function v(x, y) such that
u + iv is holomorphic. Identify the corresponding complex function f (z).
Solution: Note that uxx = 2 and uyy = −2; hence uxx + uyy = 0 everywhere, as required. For
the second part of the problem, we must find a function v(x, y) that pairs up with this u. Note
that ux = 2x and uy = −2y, hence vy = 2x and vx = 2y (Cauchy-Riemann). Integrating vy
with respect to y, we get that v = 2xy + g(x), where g(x) is a real function of x only. Similarly,
integrating vx with respect to x we get that v = 2xy + h(y), where h(y) is a real function of y
only. By comparison, g = h = constant. It follows that v = 2xy + c, where c is a real constant;
hence, finally:
f (x + iy) = x2 − y 2 + i (2xy + c) = (x + iy)2 + ic.

We immediately recognize that f (z) = z 2 + ic, where c is an arbitrary real constant. ◭

◮Example 22 Show that u(x, y) = y 3 − 3x2 y is harmonic and find a function v(x, y) such that
u + iv is holomorphic. Identify the corresponding complex function f (z).
Solution: Note that uxx = −6y and uyy = 6y; hence uxx + uyy = 0 everywhere, as required. For
the second part of the problem, we observe that, by the first Cauchy-Riemann equation,

ux = −6xy = vy =⇒ v(x, y) = −3xy 2 + g(x),

where g(x) is a function of x alone. Similarly, the second Cauchy-Riemann equation yields

uy = 3y 2 − 3x2 = −vx =⇒ v(x, y) = −3xy 2 + x3 + h(y),

where h(y) is a function of y alone. By comparison, we deduce that g(x) = x3 and h(y) =
constant. It follows that v = x3 − 3xy 2 + c, where c is a real constant; hence, finally:

f (x + iy) = y 3 − 3x2 y + i (x3 − 3xy 2 + c) = i(x + iy)3 + ic.

We immediately recognize that f (z) = iz 3 + ic, where c is an arbitrary real constant. ◭

20 Complex Differentiation

ANALYTIC FUNCTIONS

◮Definition: A function f (z) is called analytic at a point z0 if, in a neighborhood of z0 , it may

be expanded as a power series of the form
∞
X
f (z) = ak (z − a)k = a0 + a1 (z − a) + a2 (z − a)2 + a3 (z − a)3 + · · · .
◭
k=0

The simplest example of an analytic function is probably the geometric series:

∞
1 X
= zk . [ |z| < 1]
1−z
k=0

The proof of this identity (for complex z) is exactly like the one that you saw in high school
with real numbers, so we won’t go into it—but do it, as an exercise. Other examples include
the exponential series, and the series for sine and cosine (both hyperbolic and trigonometric).
Analytic functions may be visualized as “polynomials of infinite degree”. As it happens,
analytic functions have many properties in common with polynomials, but not all of them. For
instance, the geometric series already shows that the fundamental theorem of algebra does not
apply to analytic functions because the equation 1/(1 − z) = 0 has no roots.
Analytic functions have several important properties. Unfortunately their proofs require
some knowledge of the theory of series convergence, which would lead us too far away from
the core of this course. We shall mention—without proof—one property only, because of its
importance in applications.
◮Theorem If a function f (z) is analytic at z0 , then there is an open circle C with center at z0 ,
such that at every point in C [note that all points in C are internal points] f (z) is analytic and
differentiable. Circle C may have an infinite radius, in which case it coincides with the whole
complex plane.
Corollary: If a function is analytic at z0 , then it is holomorphic in a circle centered at z0 . ◭

Furthermore, the derivative of a power series may be found by term-by-term differentiation;

power series, in this respect, may be treated like polynomials. This procedure yields another
power series, which obviously (by the same theorem) will be analytic and differentiable in an
open circle. Therefore the second derivative may be differentiated term-by-term, and so on and
so forth.

◮Conclusion: If a function is analytic at z0 , then it may be differentiated any number of times,

and all its derivatives are analytic at z0 . ◭

◮Example 23 Differentiating the geometric series, we get

∞ ∞ ∞
1 X 2 X 3! X
= kz k−1 , = k(k−1)z k−2 , = k(k−1)(k−2)z k−3 ,
(1 − z)2 (1 − z)3 (1 − z)4
k=0 k=0 k=0

and so on. All the power series generated in this way converge in the open circle with radius 1
and center at the origin. ◭
Complex Differentiation Tutorial Problems 21

PROBLEMS

Revision
1. Show that |z1 + z2 |2 + |z1 + z 2 |2 = 2 |z1 |2 + 2 |z2 |2 + 4 Re(z1 ) Re(z2 ).
2. Find all the z such that |z − 1| = |z − 9| = 5.
√
3. Find all the values of 6 8.
4. Show that | sin z|2 = sin2 x + sinh2 y.
5. Show that | cos z|2 + | sin z|2 ≥ 1.
√ √
6. Calculate: (a) (−1 + i 3)60 , (b) [(1 + i 3)/(1 − i)]40 .
7. Find all the roots of the equation z 3 + 3z 2 + 3z + 3 = 0.

Cauchy-Riemann Equations
8. Find the real and imaginary part of f (z) = z 3 − i z.
9. Find the real and imaginary part of f (z) = z 4 − |z|4 .
10. Find the real and imaginary part of f (z) = (1 + i z)/(1 + z).
11. Find the function f (z) such that its real part is u(x, y) = ey cos x and its imaginary part is
v(x, y) = ey sin x. Show that for this function Cauchy-Riemann equations are nowhere satisfied.
12. Find the function f (z) such that its real part is u(x, y) = −2(xy + x) and its imaginary part
is v(x, y) = x2 − y 2 − 2y. Show that for this function Cauchy-Riemann equations are satisfied
at all points in the xy plane.
13. Find the function f (z) such that its real part is u(x, y) = x/(x2 + y 2 ) and its imaginary part
is v(x, y) = −y/(x2 + y 2 ). Show that f (z) is holomorphic at all points in the xy plane except
the origin.
14. Show that the function u(x, y) = ax2 + bxy + cy 2 , where a, b and c are real constants, is
harmonic if and only if a = −c (with no restriction on b).
15. Show that the function u(x, y) = −ex sin y is harmonic and hence find a function v(x, y)
such that u + i v is holomorphic. Hence, express the result as a function of z alone.
16. Show that the function v(x, y) = arctan y/x is harmonic and hence find a function u(x, y)
such that u + i v is holomorphic. Hence, express the result as a function of z alone.
17. Show that the function v(x, y) = sin 3x cosh 3y is harmonic and hence find a function u(x, y)
such that u + i v is holomorphic. Hence, express the result as a function of z alone.
18. Show that the function f (x, y) = x2 + i y 2 is differentiable along the straight line x = y but
is not holomorphic anywhere. Express this function in terms of z.

ANSWERS

2 z =5± √ i3. √ √ √ √
3 z = ± 2, z = (−1 ± i√ 3)/ 2, z = (1 ± i√3)/ 2. √ √
6 (a) 260 , (b) −219 (1 + i 3). 7 z = −(1 + 3 2), z = −1 + (1 ± i 3)/ 3 4.
8 u = x3 − 3xy 2 − y, v = 3x2 y − y 3 − x. 9 u = −8x2 y 2 , v = 4xy(x2 − y 2 ).
10 u = [x − 2xy − y + 1]/[(x + 1)2 + y 2 ], v = [x2 − y 2 + x + y]/[(x + 1)2 + y 2 ].
11 f (z) = ei z . 12 f (z) = iz 2 − 2z. 13 f (z) = 1/z + ic.
22 Tutorial Problems Complex Differentiation

p
15 v(x, y) = ex cos y + c; f (z) = i(ez + c). 16 u(x, y) = ln x2 + y 2 ; f (z) = log z + c.
17 u(x, y) = − cos 3x sinh 3y; f (z) = i sin 3z + c [or also f (z) = sinh i3z + c].
18 Hint: You need to show that: (i) if x = y, then

(x + ∆x)2 + i (y + ∆y)2 − (x2 + i y 2 )

lim exists
∆x→0 ∆x + i∆y
∆y→0

and is equal to 2x; (ii) Cauchy-Riemann equations do not hold if x 6= y. Hence, f ′ (z) exists
at all points on the straight line x = y, and only there. However, since no point on this line is
internal (every disk with center on the line contains points that are not on the line), f (z) is not
holomorphic anywhere.
f (z) = (z + z)2 /4 − i (z − z)2 /4.
Complex Integration 23

Chapter Two

COMPLEX INTEGRATION

1. Definite Integration
In real analysis you saw the fundamental theorem of calculus, which links integration and differ-
entiation. Complex analysis does have its own version of the fundamental theorem, but there
are some subtle differences.
First of all, you will recall that a real function of one variable doesn’t even have to be
continuous—never mind differentiable—in order to be integrable. Therefore, the question of
whether a given real function has an “anti-derivative” is, so to speak, separated from whether
it has a derivative.
The corresponding problem in complex analysis is not so simple. We have seen that differ-
entiability comes with strong requirements (i.e., Cauchy-Riemann equations, real and imaginary
parts must be harmonic). It shouldn’t come as a surprise that anti-differentiation—indefinite in-
tegration, if you wish—has similar requirements. Definite integration, by comparison, is straight-
forward, we’ll start with this one.

◮Definition: Let f (z) = u(x, y) + i v(x, y) be a complex function, and γ a smooth path in the
domain of definition of f . The expression
Z Z Z
   
f (z) dz = u(x, y) dx − v(x, y) dy + i u(x, y) dy + v(x, y) dx (8)
γ γ γ

is called the definite integral of f over the path γ. ◭

The integrals appearing on the right-hand side above are ordinary line integrals of real functions
in the xy plane, which you’ll find explained in your old calculus book. If you find this definition
confusing, consider the formal expansion

f dz = (u + i v) (dx + i dy) = u dx − v dy + i (u dy + v dx),

and then it should become obvious. The actual calculation of the definite integral may be done
either using a pair of real parametric equations for x and y, or a single complex parametric
equation for z. (
x = x(t)
⇐⇒ z(t) = x(t) + i y(t).
y = y(t)
The first method is what you learnt in first-year calculus; the second method is entirely equiva-
lent, it’s more compact but may need some practice to get used to, which is how we’ll use the
rest of this section.
24 Complex Integration

R
◮Example 24 Evaluate γ (2z + z) dz, where γ is the semicircle with center at the origin,
radius = 1, cut in the upper half-plane by the x axis, covered right-to-left.
Solution: A possible parametric representation of the
semicircle is
( y
x = cos t
[ 0 ≤ t ≤ π].
y = sin t
γ
But instead of substituting

dx = − sin t dt and dy = cos t dt −1 1 x

into (8) and evaluate two real integrals (which wouldn’t be wrong), we write
(
it
dz = i eit dt,
z = cos t + i sin t = e , =⇒
z = e−it .

It follows immediately that

Z Z π Z π
it −it it
2ei2t + 1 dt = i π



(2z + z) dz = 2e + e i e dt = i
γ 0 0

It should be stressed at this point that the choice of a parametric representation

√ for the path
is not unique. For instance, in this problem one could have used x = t, y = 1 − t2 , with t
ranging from 1 to −1. The final result would have been the same, but the calculations would
have been, perhaps, less simple. ◭
R
◮Example 25 Evaluate γ (2z + z) dz, where γ is formed by the straight path from (1, 0) to
(0, 1) and the straight path from (0, 1) to (−1, 0).
Solution: We break the problem into two halves. From (1, 0) to (0, 1) we let
x = 1 − t, y = t, [0 ≤ t ≤ 1] y
and from (0, 1) to (−1, 0) we let 1

x = t − 1, y=t [1 ≥ t ≥ 0] γ

It follows immediately that dz = (−1 + i) dt for the

first part, and dz = (1 + i) dt for the second part. The −1 1 x
rest
Z is straightforward:
(2z + z) dz =
γ
Z 1 Z 0
= [[2(1 − t + it) + (1 − t − it)]] (−1 + i) dt + [[2(t − 1 + it) + (t − 1 − it)]] (1 + i) dt =
0 1
Z 1 Z 1
= (−1 + i) [[3(1 − t) + i t]] dt − (1 + i) [[3(t − 1) + it]] dt =
0 0
Z 1 Z 1
= i6 (1 − t) dt − i 2 t dt = i2.
0 0
Complex Integration 25

Note that this example has the same integrand as the preceding one, and the same endpoints,
but a different path joins the endpoints. ◭
R
◮Example 26 Calculate γ (|z|/z) dz if (i) γ is the portion of Archimedes’ spiral x = t cos t,
y = t sin t extending from the origin to the point B(2π, 0); (ii) γ is the segment of the x axis
extending from the origin to B.
Solution: (i) Parametric equations for the spiral are given, so we simply let
z = t (cos t + i sin t) = t eit , y
which gives immediately |z| = t, dz = (1 + it) eit dt
[and arg z = t + 2kπ, which is not needed]. Therefore,
(ii) B
2π
|z| t(1 + it) eit dt
Z Z
O x
dz = = 2π(1 + i).
γ z 0 t eit
(i)
(ii) On the positive real axis, z = x and |z| = x too,
because x > 0. Hence dz = dx, and finally:
2π
|z|
Z Z
dz = dx = 2π.
γ z 0

This is another example showing that in general definite complex integration is path-dependent
even if the endpoints are kept fixed. ◭

◮Example 27 Calculate γ exp(z 2 − z 2 ) dz, where γ is the straight vertical segment leading
R

from A( 14 , 0) to B( 14 , π).
Solution: First of all, we simplify the exponential:

z 2 − z 2 = (x + iy)2 − (x − iy)2 = i 4xy.

The path γ is a straight vertical segment in the first quadrant. We substitute

x = 41 , y = t, =⇒ z= 1
4 + i t, dz = i dt, 4xy = t,

and let t range from 0 to π. It follows immediately that

Z Z π
z 2 −z 2
 π
e dz = i eit dt = i eit 0 = −i 2.
γ 0 ◭

◮Example 28 Calculate γ 1/ | cos z|2 + 21 dz, where γ is the whole vertical straight line
R

z = 14 π + iy, travelled in the direction of increasing y.

Solution: In the addition formula for the cosine,

cos z = cos(x + iy) = cos x cos iy − sin x sin iy,

applying (4), we substitute cos iy = cosh y and sin iy = i sinh y. We get that

cos z = cos x cosh y − i sin x sinh y,

26 Complex Integration

and hence that

| cos z|2 = cos2 x cosh2 y + sin2 x sinh2 y.
Now we introduce parametric equations for the path, for instance

x = 14 π, y = t. [−∞ < t < ∞]

It follows immediately that cos x = sin x = 1/√2, and also that dz = i dt. Therefore,

| cos z|2 = 1
2 cosh2 t + 1
2 sinh2 t,

and finally
∞ ∞ i∞
dz i dt i dt
Z Z Z h
1 = 2 2 = = i tanh t = i 2.
γ | cos z|2 + 2 −∞
1 1
2 cosh t + 2 sinh t +
1
2 −∞ cosh2 t −∞ ◭

◮Example 29 Calculate γ dz/|z|2 , given that γ is the whole right-hand branch of the hyper-
R

bola x2 − y 2 = 1, travelled upwards.

Solution: Parametric equations for the righ-hand branch of the hyperbola are
x = cosh t, y = sinh t.
y
It follows immediately:

z = cosh t + i sinh t, dz = (sinh t + i cosh t) dt

x
|z| = cosh t + sinh2 t
2 2

−∞ < t < ∞
γ
Therefore,
∞ ∞ ∞
dz (sinh t + i cosh t) dt sinh t dt cosh t dt
Z Z Z Z
= = +i .
γ |z|2 −∞ cosh2 t + sinh2 t −∞ cosh2 t + sinh2 t −∞ cosh2 t + sinh2 t

We see immediately that the first integral on the right-hand side (the real part) is zero by
symmetry, because the integrand is odd in t:

0 ∞
sinh t dt sinh t dt
Z Z
=− .
−∞ cosh2 t + sinh2 t 0 cosh2 t + sinh2 t

Conversely, the integrand in the imaginary part is even in t, and we have:

0 ∞
cosh t dt cosh t dt
Z Z
= ,
−∞ cosh2 t + sinh2 t 0 cosh2 t + sinh2 t

which yields immediately

∞ ∞
cosh t dt cosh t dt
Z Z
=2 .
−∞ cosh2 t + sinh2 t 0 cosh2 t + sinh2 t
Complex Integration 27

Substituting sinh t = u, and noting the identity cosh2 t + sinh2 t = 1 + 2 sinh2 t, we get:

∞ ∞ √ i∞
cosh t dt du h 1 π
Z Z
=2 = 2 √ arctan 2u =√ .
−∞ cosh2 t + sinh2 t 0 1 + 2u 2
2 0 2
√
dz/|z|2 = i π/ 2.
R
So, finally, γ
◭

2. Cauchy-Goursat Theorem

You’ll probably recall that in real vector analysis you were introduced to the concept of con-
servative vector fields, i.e., vector fields that lead to a certain class of path-independent line
integrals.
As complex definite integrals are basically pairs of line integrals in the xy plane, it’s not
surprising that path-independence may also be found in complex integration.

REVISION

Before we continue, let us revise some concepts and theorems that you perhaps saw in first-year
calculus, and will play a major role in the following sections.

◮Definition: An open set of points D in the xy plane is called connected if any pair of points
in D may be joined by a path that lies entirely in D. (In plain words, a set is connected if it is
all in one piece.) ◭

◮Definition: An open set D is called simply connected if it’s connected and every closed path
in D surrounds only points that belong in D. (In plain words, a simply connected domain is a
domain without holes.) ◭

The two definitions given above may be extended to sets that are not open, but we’ll not need
them it this course.

Disconnected Connected, but not simply connected Simply connected

◮Theorem Let P (x, y) and Q(x, y) have continuous first partial derivatives in a simply-
connected open domain D. If
∂P ∂Q
=
∂y ∂x
R
everywhere in D, then the line integral [P dx + Q dy] is path-independent in D ◭
28 Complex Integration

This theorem is a direct consequence of Green’s Theorem, and has a famous three-dimensional
counterpart in Stokes’ Theorem—which is of no use here because complex numbers are inherently
two-dimensional (as we noted on page 1).
For a better revision of these ideas, see W. Kaplan, Advanced Calculus Fourth Edition,
Addison-Wesley (USA), 1991, or also N. Piskunov, Differential and Integral Calculus, Mir
(Moscow), 1974.
THE CAUCHY-GOURSAT THEOREM
Consider the definite integral of a holomorphic function f = u + i v: by definition [go back to
equation (8), page 23] this is
Z Z Z
   
f (z) dz = u(x, y) dx − v(x, y) dy + i u(x, y) dy + v(x, y) dx .
γ γ γ

Assume that u and v satisfy Cauchy-Riemann equations (7), and that ux and uy are continuous.
Let’s look first at the real part of the right-hand side, which is
Z
 
u(x, y) dx − v(x, y) dy .
γ

If in this expression we let P ≡ u and Q ≡ −v, we find (by Cauchy-Riemann) that

uy = −vx =⇒ Py = Qx ,

so all hypotheses of the path-independence theorem recalled above are satisfied. Similarly, the
imaginary part is Z
 
u(x, y) dy + v(x, y) dx ;
γ

here we let P ≡ v and Q ≡ u. It follows that

ux = v y =⇒ Py = Qx .

Again, the path-independence theorem is applicable.

◮Conclusion: If f (z) is holomorphicR in a simply-connected open domain D with continuous

partial derivatives ux and uy , then f (z) dz is path-independent for every smooth path con-
tained entirely in D. ◭

The elegant proof outlined here is generally called Cauchy’s proof, but its genesis was as tortuous
and difficult, as this one is simple.
For a long time, after Bombelli introduced “imaginary” numbers and used them to solve
cubic equations (around 1560), complex numbers were poorly understood. The apparently
simple definition of complex numbers as points in the xy plane, which today is the norm,
was given in 1806, by Argand. In between, for√ two and a half centuries, the mathematical
community was not sure about the meaning of −1, while accepting the beatiful results that
could be obtained through it.
Indeed, when Cauchy presented this theorem to the French Academy in 1814, he avoided
complex numbers, which however appeared in his first book on the subject (1825). He didn’t
mention Cauchy-Riemann equations either, although they had been found by d’Alembert half a
Complex Integration 29

century before. Eventually Cauchy achieved a clear view of complex functions theory, expanding
on the work of Gauss, Green and Laurent, among others. But only in a paper of 1846 did he
come fairly close to the modern perspective; the same paper contains also Cauchy’s integral
formula, which we’ll discuss soon. Meanwhile Green (1790-1841) died ignored by most of his
contemporaries. For a vivid account of this story, see I. Grattan-Guinness, The Norton History
of the Mathematical Sciences, Norton (USA), 1998.
Much later, in 1900, Goursat (1858-1936) found a different (but difficult) proof that avoided
the assumption that ux and uy be continuous. His famous Course of Analysis, a model of clarity
and rigor, was still in print in the 1970s.

⊲Corollary: An immediate consequence of Cauchy-Goursat theo-

rem is that if γ is a smooth closed path contained entirely in a
simply-connected domain where f (z) is holomorphic, then
I (ii) B
f (z) dz = 0. (9)
γ
(i)
γ
To prove this, choose two arbitrary points on γ, call them A and B,
A
and consider the two Rsections of γ delimited by A and B, as shown
in the picture. Since f (z) dz is path-independent, we find that D
Z Z
f (z) dz = f (z) dz
(i) (ii)

On the other hand, obviously,

Z Z I
f (z) dz − f (z) dz = f (z) dz,
(i) (ii) γ

because γ is oriented in the same direction as the arrow along path (i), but in the opposite
direction along path (ii): hence the negative sign. Combining the last two results we get that
I
f (z) dz = 0,
γ

as stated in equation (9). ⊳

3. Indefinite Integration
Another important consequence of Cauchy-Goursat theorem is that holomorphic functions pos-
sess an indefinite integral.
To see this, we only need to consider the expression
Z z
def
F (z) = f (ζ) dζ
z0

and show that, if f (z) is holomorphic, then F ′ (z) exists and is equal to f (z).
The outline of the proof is almost trivial. Consider ∆F , that is,
Z z+∆z Z z
∆F = f (ζ) dζ − f (ζ) dζ,
z0 z0
30 Complex Integration

where ∆z is at this point arbitrary, but not zero. Because f (z) is by assumption holomorphic,
the two integrals on the right-hand side are path-independent and hence may be combined:
Z z+∆z
∆F = f (ζ) dζ.
z

Now we may replace f (z) with u(x, y) + i v(x, y) and use the fact that u and v are continuous
real functions with continuous partial derivatives. It then follows (proof omitted) that

∆F = f (z) ∆z + ε ∆z

where ε is a complex quantity that goes to zero as ∆z → 0. We may finally divide the last
equation by ∆z, and deduce that

∆F ∆F
= f (z) + ε, =⇒ lim = f (z).
∆z ∆z→0 ∆z

The last equation establishes that F ′ (z) indeed exists and equals f (z); in other words, F (z) is
the anti-derivative of f (z).
This means that it is possible to extend to holomorphic functions all the techniques of
anti-differentiation developed in real analysis; formulas are formally identical.

z 3 dz, where γ is the upper half of the Archimedes’ spiral defined in

R
◮Example 30 Find γ
example 26.
1 4
Solution: f (z) = z 3 is holomorphic in the whole plane, and its anti-derivative is F (z) = 4z
(plus a constant, if you wish). The endpoints of γ are 0 and 2π, on the x axis. Therefore
Z h i2π
z 3 dz = 1 4
4z = 4π 4 .
γ 0

No need to use the parametric equations for the path; this integral is path-independent. ◭

This may all look simple, except for a small detail: how can we recognize that a function is
holomorphic? Here are a few simple rules.
• We have shown that f (z) = z in not holomorphic anywhere. Hence, if in the expression for
f (z) you spot z, you’ll know that f is not holomorphic.
• We have shown that polynomials in z are holomorphic in the whole plane.
• Analytic functions of z are holomorphic in a circle centered at the center of expansion. Some
of them, in particular ez , and sine and cosine (hyperbolic/trigonometric), are analytic in
the whole plane.
• Rational functions of z are holomorphic everywhere except at the points where the denom-
inator is zero.
• The complex logarithm is holomorphic in the whole plane except at the origin; but is multi-
valued. If a point z moves over a path that loops around the origin, then arg z increases by
multiples of 2π.
• The composition of two holomorphic functions is generally holomorphic, with the possible
exception of some isolated points, as next example shows.
Complex Integration 31

◮Example 31 Consider f (z) = log z and g(z) = cosh z.

Their composition is h(z) = log(cosh z), which is holomorphic everywhere except at those points
where cosh z = 0, because log 0 is undefined. Hence, h(z) is holomorphic everywhere except at
z = ±i π/2, ±i 3π/2, ±i 5π/2, ±i 7π/2, . . . ◭

As a rule, multi-valued functions must be integrated with some care, because their value must
be “followed” along the path.

◮Example 32 Find γ dz/z, where γ is the portion of the Archimedes’ spiral z = t eit extending
R

from z = 2π to z = 8π.
Solution: The anti-derivative of the integrand is

d−1
 
1
= log z + C,
dz −1 z
so, one is tempted to write
ifinish 8π
dz h
Z
= log z 2π
γ z start

and move on. But recall that

def
log z = ln |z| + i arg z.

Now, how does arg z change along the path?

The picture, produced by the free package gnuplot, provides the answer. Following z as it
moves along γ, we clearly see that γ wraps itself around the origin three times counter-clockwise;
therefore arg z grows by 6π radians (1080◦ in high-school parlance). So, finally,

dz
Z
= ln |8π| − ln |2π| + i ∆ arg z = ln 4 + i 6π.
γ z

Still confused? Let’s do this problem again, using parametric equations like in example 26. Let
z = t eit , dz = (1 + it) eit dt. It follows:

8π 8π
dz (1 + it) eit dt 1 + it i8π
Z Z Z h
= = dt = ln |t| + i t = ln 4 + i 6π.
γ z 2π teit 2π t 2π

We got the same answer, of course, but with much more work.
In this simple example the initial and final values of arg z do not influence the result, which
depends only on their difference. In general, however, one must bear this possibility in mind.
As an exercise, convince yourself that if γ were replaced by the spiral z = (2π + t) ei t/3 ,
which also starts at z = 2π and finishes at z = 8π, but wraps itself around the origin only once,
as you can see using gnuplot, then the final result would be ln 4 + i 2π. ◭

4. Cauchy’s Integral Formula

In this section we’ll see one of the most important results in all fields of mathematics. But before
we get into its derivation, which is very easy, we need another preliminary result.
32 Complex Integration

◮Example 33 Show that

I
dz y
= i 2π, R
CR z − z0
z0
where z0 is an arbitrary point in the xy plane and CR is a
circle with radius R and center at z0 , covered counterclock- x
wise; show also that this result does not depend on where
O CR
the integration starts/ends.
Solution: A convenient parametric representation for the circle CR is z = z0 + R eit , where the
parameter t is the angle arg(z − z0 ) and ranges from a certain initial value t1 to t1 + 2π. It
follows that
1 1
= , dz = i Reit dt,
z − z0 Reit
and finally that
t1 +2π
dz i Reit
I Z
= dt = i 2π,
CR z − z0 t1 Reit
independently of t1 , as required. Note that this result does not depend on z0 and R either. The
value of the integral remains equal to i 2π even if R is taken arbitrarily small (but not zero).
Method 2: the integrand has an anti-derivative, as we saw in example 20:

d−1
 
1
= log(z − z0 ) + a constant.
dz −1 z − z0

If z moves around z0 , then the expression (z − z0 ) returns to its initial value (the constant is
irrelevant, of course), and so does ln |z − z0 |, but arg(z − z0 ) increases by 2π radians: hence
I
dz h ifinish
= ln |z − z0 | + i arg(z − z0 ) + C = i 2π.
CR z − z0 start

It clearly does not matter where the integration starts. ◭

Suppose now that f (z) is holomorphic in an open domain D, and let C be a closed smooth contour
in D, such that f is holomorphic inside the region enclosed by C and also on the boundary C.
Recall that, conventionally, counter-clockwise contour
integration is always regarded as positive.
The domain D does not have to be simply-connected,
but it is crucial that the region enclosed by the contour C
is “hole-free”, and f is holomorphic on the boundary C as CR
well.
C
z0
Consider
f (z)
Z
dz, (10)
C z − z0
D
where z0 is an arbitrary point inside C. Since z0 is internal to
C, it’s possible to draw a circle CR with center at z0 , radius
R, that is contained entirely in C, as shown in the picture.
Complex Integration 33

Since f (z) is holomorphic on and inside C, then there is only one point inside C where
f (z)/(z − z0 ) may be non-holomorphic: this is the point z0 (because the denominator becomes
zero).
Clearly, the value of the integral (10) does not change if C is altered in such a way that
the new, deformed contour lies entirely in D, and the deformation does not carry it across the
point z0 .
In particular, we may modify C into the circle C R without affecting the value of the integral,
and we have that
f (z) f (z)
Z Z
dz = dz.
C z − z0 CR z − z0

Combining this equation with the result of example 33, we find that
f (z) f (z) dz
Z Z Z
dz − i 2π · f (z0 ) = dz − · f (z0 ) =
C z − z0 CR z − z0 CR z − z0
f (z) − f (z0 )
Z
= dz.
CR z − z0
It follows immediately that:
f (z) f (z) − f (z0 ) |f (z) − f (z0 )|
Z Z Z
dz − i 2π · f (z0 ) = dz ≤ |dz|.
C z − z0 CR z − z0 CR |z − z0 |

But obviously |z − z0 | = R because the integration is over a circle, and |dz| = R dt. Also,
|f (z) − f (z0 )| may be made as small as we please by taking R sufficiently small (but not zero),
because f is continuous. So, finally,
Z t1 +2π
f (z) ε
Z
dz − i 2π · f (z0 ) ≤ R dt = ε · 2π.
C z − z0 t1 R
We see that the left-hand side of this equation is not negative, but is smaller than any arbitrary
positive constant, no matter how small. Therefore it must be
f (z)
Z
dz − i 2π · f (z0 ) = 0.
C z − z0
◮Conclusion: Removing the modulus bars, which are no longer needed, and dividing through
by i 2π, we obtain:
1 f (z)
Z
dz = f (z0 ), (11)
i 2π C z − z0
which is Cauchy integral formula. ◭

One can see immediately that this is a remarkable formula, even without knowing its applications
in science and engineering. For example, it shows that the values of a holomorphic function at
points inside a contour C are completely determined by the values on the boundary. With an
abuse of language, one may say that holomorphic functions are “tamper-proof”: once their value
is assigned on a contour C, the values inside C are automatically determined.
Furthermore: let the contour be the circle CR with center at z0 , radius R. Then, as we have
seen in example 33, z = z0 + Reit , dz = iReit dt, and (11) yields
Z 2π
1
f (z0 ) = f (z0 + Reit ) dt.
2π 0
34 Complex Integration

We see that the value of a holomorphic function at the center of any circle is the average
of the values on the circumference. Another interesting consequence comes if we take the
real/imaginary part of the last equation. Since f = u + i v, and u and v are harmonic, this
means that the value of a harmonic function of x and y at the center of any circle in the xy
plane is the average of the values on the circumference.
NUMERICAL EXAMPLES
◮Example 34 Find C (3z 2 − 5z + 4) dz/(z − 1) where C is the circle with center at the origin,
H

radius = 7.
Solution: Let f (z) = 3z 2 − 5z + 4, which is holomorphic everywhere because it’s a polynolmial,
and let z0 = 1. The contour C encloses z0 . Hence,

3z 2 − 5z + 4
I
dz = i 2πf (1) = i 4π.
C z−1

Note that the radius of the circle does not have to be equal to 7; it may be equal to any number
greater than 1. If the radius is exactly 1, then C goes over the point z0 = 1, and the integral
becomes meaningless. ◭

◮Example 35 Find C eiz dz/(z 2 + 1) where C is the cardioid given in polar coordinates by the
H

equation ρ = 1 + sin φ.
Solution: The function
y
eiz eiz
=
z2 + 1 (z + i)(z − i)

is holomorphic everywhere except at the points z = i 1

and z = −i. The cardioid C, however, surrounds only
the first one of them. Therefore, we let

eiz x
f (z) = and z0 = i.
(z + i)
−1
It follows immediately that

eiz dz eiz dz f (z)

I I I
= = dz,
C z2 + 1 C (z + i)(z − i) C z−i

and hence, by Cauchy integral formula (11), that

eiz dz e−1
I
= i 2π = πe−1 .
C (z + i)(z − i) i 2

Once again, the shape of C may be altered. Any other contour enclosing i but not −i would
have given the same result. Make sure you understand this point. ◭

◮Example 36 Find C tan z dz/(3z 2 + i 5z + 2), where C is the unit circle.

H

Solution: The integrand has an infinity of singular points, i.e., points where it is not holomorphic,
but only one such point lies inside the unit circle.
Complex Integration 35

First of all, tan z fails to be holomorphic at z = ±π/2, ±3π/2, ±5π/2, ±7π/2, and so on.
All these points are on the x axis. However, π/2 ≈ 1.57, so even the nearest ones to the origin
lie outside the unit circle.
There are two more points where the integrand is not holomorphic: they are the points
where
3z 2 + i 5z + 2 = 0.
Solving this equation, we find
√ 
−i 5 ± −25 − 24 −i 5 ± i 7 −i 2,
z= = =
6 6 i/3.

Having found the roots of the denominator, we may factorize it. Hence, we write the integrand
in a convenient form:
tan z tan z
2
= .
3z + i 5z + 2 3(z + i2)(z − i/3)
Clearly −i 2 is outside the unit circle and i/3 is inside. Therefore, we let

tan z
f (z) = ,
3(z + i2)

and the integral may now be found by Cauchy integral formula:

f (z) tan i/3 2π tan i/3

I
= i 2πf (i/3) = i 2π = .
C z − i/3 3(i/3 + i 2) 7

HIt may be shown 2

that tan i/3 = i tanh 1/3: verify this (using equations (4), perhaps). So, finally:
C
tan z dz/(3z + i 5z + 2) = i (2π/7) tanh 1/3. ◭

5. Extensions of Cauchy Integral Formula

Cauchy integral formula may be immediately
extended to contour integrals like C

f (z) dz
I
z2 C
,
C (z − z0 )(z − z1 ) · · · (z − zn )
C2
C0
where f (z) is holomorphic inside a closed con- z0
tour C, as well as on the boundary. C1
The method is illustrated by the picture
on the right, for the case where there are only z1
three points, z0 , z1 and z2 . We break the
contour C at any point and add a path leading C
toward the point z0 . Such a path does not C
have to be straight, though it looks straight
in the picture. We then add a contour C 0 that
surrounds the point z0 , and trace back our steps toward C. We repeat the procedure for the
point z1 : we break C adding a path leading toward z1 , then a contour C 1 surrounding z1 , then
another path leading back toward C. Finally, we do the same for the point z2 : we break C
adding a path leading toward z2 , then a contour C 2 surrounding z2 , then another path leading
36 Complex Integration

back toward C. If there are only three points, like in the picture, no more detours are needed.
If there are more points, we must repeat the procedure for each of the points included in C.
In this way we create a modified region γ ( shaded in the picture) that does not include
any of the points z0 , z1 , z2 . Therefore, by (9),

f (z) dz
I
= 0.
γ (z − z 0 )(z − z1 )(z − z2 )

We note that γ consists of the original contour C, of the contours C 0 , C 1 , C 2 , and of the paths
joining C to C 0 , C 1 , C 2 . However:
(i) While C is covered counter-clockwise, the inner contours C 0 , C 1 , C 2 are covered clockwise.
Recall that clockwise contour integrals count as negative.
(ii) Paths joining the various contours are covered twice, in opposite directions, and therefore
the corresponding integrals cancel out.
In summary, we get
2
f (z) dz f (z) dz
I X I
− = 0;
C (z − z0 )(z − z1 )(z − z2 ) Ck (z − z0 )(z − z1 )(z − z2 )
k=0

the negative sign account for clockwise integration. It follows immediately that
2
f (z) dz f (z) dz
I X I
= . (12)
C (z − z0 )(z − z1 )(z − z2 ) Ck (z − z0 )(z − z1 )(z − z2 )
k=0

This result shows that the integral on the left-hand side, over a contour surrounding the points
z0 , z1 and z2 , may be written as the sum of an integral over C 0 surrounding only z0 , plus an
integral over C 1 surrounding only z1 , plus an integral over C 2 surrounding only z2 . Each of the
integrals on the right-hand side may then be evaluated by Cauchy integral formula (11).
3
◮Example 37 Find C sin πz dz/(z 3 − z 2 + 16
H
z), where C is the unit circle.
3 3 3
Solution: First of all, we factor z 3 −z 2 + 16 z = z(z 2 −z+ 16 ). Solving the equation z 2 −z+ 16 = 0,
1 3
we find the roots z = 4 and z = 4 . Hence,

z3 − z2 + 3
16 z = z(z − 14 )(z − 43 ).

We note that the integrand is non-holomorphic at three points, namely 0, 41 and 34 , all of which
are inside the contour of integration. Therefore, applying (12), we write
sin πz
I
1 3 dz =
C z(z − 4 )(z − 4 )

(sin πz) (z − 14 )(z − 43 ) (sin πz) z(z − 43 ) (sin πz) z(z − 14 )

I  I  I 
= dz + dz + dz,
C0 z C1 z − 41 C2 z − 43

where C 0 surrounds only z = 0, C 1 surrounds only z = 14 and C 2 surrounds only z = 43 . The

size and shape of these contours don’t matter, as long as they surround only one point where
the integrand is non-holomorphic. It follows by Cauchy integral formula (11) that

(sin πz) (z − 14 )(z − 43 )

I 
dz = i 2π · (sin 0) (− 14 )(− 34 ) = 0,
  
C0 z
Complex Integration 37

and similarly that

(sin πz) z(z − 43 ) √
I 
dz = i 2π (sin π4 ) 41 (− 12 ) = −i 8π 2,
  
1
C1 z−4
(sin πz) z(z − 14 ) √
I 
dz = i 2π (sin 3π
 3 1 
3 4 ) 4 ( 2 ) = i (8/3)π 2.
C2 z−4
3
√ √
So, finally, C sin πz dz/(z 3 − z 2 + 16
H
z) = i 8 · (−1 + 1/3)π 2 = −i (16/3)π 2. ◭

Naturally, if there are more (or less) than three points inside C where the integrand is non-
holomorphic, the sum in formula (12) extends to all such points and only to them.

◮Example 38 Find C sin z dz/(z 2 + 1), where C is the circle |z| = 2.

H

Solution: The function f (z) = sin z is holomorphic in the whole plane. The denominator, z 2 + 1,
is zero only at z = ±i; C is the circle with radius = 2 and center at the origin, and so it encloses
both points. Therefore, applying (12) and then (11), we find immediately:
sin z dz sin(−i) sin i i 4π sin i
I
··· = = i 2π + i 2π = = i 2π sinh 1.
C (z + i)(z − i) −i − i i+i i2 ◭

◮Example 39 Find C z 2 dz/(z 3 − 8), where C is the square |x| + |y| = 3.

H

Solution: First of all, we find the roots of

y
3
z − 8 = 0, 3
which are
√
z=2 and z = 2e±i 2π/3 = −1 ± i 3.
−3 −1 2 3 x
Their position is shown in the picture. Simple calculations
show that the square C surrounds them all. Before we ap-
C
ply (12), we write: −3

z 2 dz z 2 dz
I I
3
= ,
C z −8 C (z − z0 )(z − z1 )(z − z2 )
where √ √
z0 = 2, z1 = −1 + i 3, z2 = −1 − i 3.
Then, we apply (12):
z 2 /(z − z1 )(z − z2 ) z 2 /(z − z0 )(z − z2 ) z 2 /(z − z0 )(z − z1 )
I I I
··· = dz + dz + dz.
C0 z − z0 C1 z − z1 C2 z − z2
Each of the integrals on the right-hand side may be now be found by Cauchy integral for-
mula (11). We find:
z 2 dz i 2πz02 i 2πz12 i 2πz22
I
3
= + + .
C z −8 (z0 − z1 )(z0 − z2 ) (z1 − z0 )(z1 − z2 ) (z2 − z0 )(z2 − z1 )
√ √
Now one must substitute z0 = 2, z1 = −1 + i 3 and z2 = −1 − i 3, and then simplify. The
calculations are time consuming, though not difficult (do it, as an exercise); in the end one finds
that the right-hand side is equal to i 2π. ◭
38 Complex Integration

◮Example 40 Find C (z 2 − i) dz/(z 4 + 64), if C is the circle |z + 3| = 5.

H

Solution: C has radius 5 and center on the x axis at y

x = −3, as pictured on the right.
The integrand is a rational function, hence it is 4
holomorphic everywhere except at the points where
the denominator is zero. Hence, we need to solve the z0 z3
equation 5
4
z + 64 = 0. −8 −3 0 2 x
This is easy: writing it in the form z1 z2

z 4 = −64, C −4

we find immediately that

z = (−64)1/4 = (64 eiπ )1/4 .
In the complex field the fourth root is a four-valued function. Hence, the denominator has four
distinct roots, shown in the picture. They are:
√ √
z0 = 8 ei 3π/4 = −2 + i 2, z1 = 8 ei 5π/4 = −2 − i 2,
√ √
z2 = 8 ei 7π/4 = 2 − i 2, z3 = 8 ei 9π/4 = 2 + i 2.

It’s easy to see that z2 and z3 lie outside the circle C, so we don’t need to consider them. Writing
the integrand as follows:

z2 − i z2 − i
I I
dz = dz,
C (z 4 + 64) C (z − z0 )(z − z1 )(z − z2 )(z − z3 )

considering that C surrounds only z0 and z1 , applying (12), and substituting for z0 , z1 , z2 , z3
their numerical values, we find immediately that

z02 − i z12 − i
· · · = i 2π · + i 2π · =
(z0 − z1 )(z0 − z2 )(z0 − z3 ) (z1 − z0 )(z1 − z2 )(z1 − z3 )
π π
= −i . ◭
32 4
CAUCHY FORMULA FOR DERIVATIVES
The letter z appearing under the symbol of integration in Cauchy integral formula (11) is
a dummy variable, which may be replaced by any other symbol. So, for instance, we may
rewrite (11) in the form
1 f (ζ)
Z
f (z) = dζ,
i 2π C ζ − z
which is entirely equivalent, but shows the right-hand side as a function of z.
It can be rigorously shown that the equation above may be formally differentiated with
respect to z: in other words, it may be shown that
Z  Z  
d f (ζ) ∂ f (ζ)
dζ = dζ
dz C ζ −z C ∂z ζ − z
Complex Integration 39

Put in a different way, differentiation with respect to z and integration with respect to ζ may
be swapped around. This is precisely what one would expect, but remember, differentiation of
complex functions is in general more subtle than its real counterpart. Actually, the proof is not
difficult but it would take us too far away.
Using this result, and noting that
 
∂ 1 1
= ,
∂z ζ − z (ζ − z)2

we find that, for a function f that is holomorphic on and inside a contour C,

1 f (ζ)
I
′
f (z) = dζ.
i 2π C (ζ − z)2

This is Cauchy formula for first derivatives. It may be differentiated again under the same
assumptions ; one gets (verify this)

2! f (ζ) 3! f (ζ)
I I
f ′′ (z) = dζ, f ′′′ (z) = dζ,
i 2π C (ζ − z)3 i 2π C (ζ − z)4

which are Cauchy formulas for second and third derivatives, respectively. Formulas for higher
derivatives are obtained in the same way. More generally, after a suitable relabelling of variables,
one obtains:
n! f (z)
I
dz = f (n) (z0 ), (13)
i 2π C (z − z0 )n+1
where n is the order of the derivative of f . The case n = 0 corresponds to the standard Cauchy
integral formula (11).
It is possible to use these formulas to evaluate contour integrals in a way that is entirely
analogous to (12), even when more than one singular point is surrounded by C; the details are
the same as before.

(z 2 − 3z + 4) dz/z 2 , if C is the unit circle.

H
◮Example 41 Find C
Solution: To apply (13), we let f (z) = z 2 − 3z + 4 and z0 = 0. It follows that f ′ (z) = 2z − 3
and f ′ (0) = −3; hence, applying (13), we find:

z 2 − 3z + 4 i 2π ′
I
2
dz = f (0) = −i 6π.
C z 1!

Method 2: For this problem, there is also a direct method that does not call for (13): simplifying
the integrand, write

z 2 − 3z + 4 3 4
I I I I
dz = dz − dz + dz.
C z2 C C z C z2

Then note that

3 4
I I I
dz = 0, dz = i 6π, dz = 0.
C C z C z2
40 Complex Integration

The first result follows from the fact that f (z) ≡ 1 is holomorphic everywhere; the second result
follows from example 33; the third result comes from the fact that z −2 has an anti-derivative,
which is −z −1 , that returns to its
H initial value if z covers any closed contour (whether it surrounds
the origin or not). So, finally, C (z 2 − 3z + 4) dz/z 2 = −i 6π, as expected. ◭

◮Example 42 Find C ez ln 2 dz/(z − 5)3 , where C is a contour surrounding the point z = 5.

H

Solution: Using (13), we find immediately that

ez ln 2 i 2π d2 h z ln 2 i
I
dz = e = i π(ln 2)2 e5 ln 2 = i 32π(ln 2)2 .
C (z − 5)3 2! dz 2 z=5 ◭

◮Example 43 Find C (z 4 + 1) dz/z(z − 1)2 , where C is the square |x| + |y| = 3 pictured in
H

example 39.
Solution: The contour encloses both z = 0 and z = 1. Following exactly the same steps of the
derivation of (12), we deform C into two contours C 0 and C 1 , such that C 0 surrounds only the
origin and C 1 surrounds only the point z = 1.
It follows that

(z 4 + 1) (z 4 + 1) (z 4 + 1)
I I I
2
dz = 2
dz + 2
dz.
C z(z − 1) C0 z(z − 1) C1 z(z − 1)

Now, by Cauchy integral formula (11) we find that

z4 + 1
 4 
z +1
I
dz = i 2π = i 2π;
C0 z(z − 1)2 (z − 1)2 z=0

by Cauchy formula for first derivatives [see (13)] we find that

z4 + 1 i 2π d z 4 + 1
I  
dz = = i 4π.
C1 z(z − 1)2 1! dz z z=1

Therefore, finally, C (z 4 + 1) dz/z(z − 1)2 = i 2π + i 4π = i 6π.

H

Method 2: By Heaviside’s cover-up method, we find that

1 1 1 1
2
= − + .
z(z − 1) z z − 1 (z − 1)2

Hence,
z4 + 1 z4 + 1 z4 + 1 z4 + 1
I I I I
dz = dz − dz + dz.
C z(z − 1)2 C z C z−1 C (z − 1)2
No need to modify the contour, here. The first two terms are found directly by Cauchy integral
formula (11), and the third term by Cauchy formula for first derivatives [see (13) with n = 1].
It follows that
h i h i h i
· · · = i 2π z 4 + 1 − i 2π z 4 + 1 + i 2π 4z 3 = i 2π − i 4π + i 8π.
z=0 z=1 z=1

In the end, we recover the result i 6π, as expected. ◭

Complex Integration 41

◮Example 44 Find C z 3 dz/(z − 1)(z − 2)3 if C is:

H

(i) the circle |z − 2| = 1/2; (ii) the circle |z − 2| = 2.

Solution: (i) The contour C surrounds only the point z = 2. Hence, by Cauchy formula for
second derivatives [see (13)], we get immediately:

z3 i 2π d2 z3
I  
dz = = i 4π.
C (z − 1)(z − 2)3 2! dz 2 (z − 1) z=2

(ii) In this case C surrounds both z = 2 and z = 1. Proceeding like in the preceding example,
we get that
z3 z3
I I I
··· = 3
dz + 3
dz,
C C2 (z − 1)(z − 2) C1 (z − 1)(z − 2)

where C 2 surrounds only the point z = 2 and C 1 surrounds only the point z = 1. The first term
has been found in part (i), and it is equal to i 4π. For the second term we may use Cauchy
integral formula (11), which yields

z3 z3
I  
dz = i 2π = −i 2π.
C1 (z − 1)(z − 2)3 (z − 2)3 z=1

Therefore, finally, in this case we have:

z3
I
dz = i 4π − i 2π = i 2π.
C (z − 1)(z − 2)3 ◭
42 Tutorial Problems Complex Integration

PROBLEMS

Revision Examples on Green’s Theorem

Green’s theorem in the plane says that if P (x, y) and Q(x, y) are continuous with continuous
partial derivatives on a simple contour C and in the region R inside C, then
I ZZ
   
P dx + Q dy = Qx − Py dx dy.
C R

You are strongly exhorted to revise Green’s theorem in your old calculus book.
H
19. Show that if C is a simple closed path, then C z dz = i 2A, where A is the area of the region
enclosed by C.
20. Show that if C is a simple closed path, then C |z|2 z dz = i 4 I, where I is the moment of
H

inertia about the origin of the plane region enclosed by C (assuming unit density σ).
Path-dependent Integration
R 1+i
21. Calculate 0 (x2 − iy 2 ) dz along the straight line from the origin to 1 + i.
R
22. Find γ dz from −i to i if:
(i) γ is the polygonal path with vertices at −i, −1 − i, −1, i.
(ii) γ is the left half of the circle |z| = 1.
R
23. Find γ x dz from −4 to +4 if:
(i) γ is the polygonal path with vertices at −4, −4 + i4, 4 + i4, +4.
(ii) γ is the upper half of the circle |z| = 4.
H
24. Calculate γ z dz, where γ is the whole circle |z| = 4, covered counter-clockwise.
R −1
25. Calculate 1 |z − 1|2 dz along the upper half of the circle |z| = 1.
H
26. Calculate γ dz/z, if γ is the whole circle |z| = 2, covered counter-clockwise.
R1
27. Calculate −1 z · Im(z 2 ) dz along the lower half of the circle |z| = 1.
R 1+i
28. Calculate 0 |z|2 dz along an arc of the parabola y = x2 .
29. Calculate γ dz/|z|2 , if γ is the whole right-hand branch of the hyperbola x2 /9 − y 2 /16 = 1.
R
R 1+i 2
30. Calculate 0 e|z| x dz along the straight line joining the origin and the point 1 + i.
R
31. Calculate γ (1/z − 1/z) dz along the whole straight horizontal line z = i.
32. Calculate γ z dz along the whole asteroid |x|2/3 + |y|2/3 = 1.
H

Path-independent Integration
R 1+i
33. Find 0
z 3 dz
R i ln 2
34. Find 0
sin z dz
R i π 2 z/2
35. Find 0
z e dz
R 1+i
36. Find 1−i
2 sin z cosh z dz
R √3+i 3

2
37. Find 0
z/(z + 8) dz
Complex Integration Tutorial Problems 43

R 1+i 2
38. Find i
z ei z /2
dz
R
39. Find γ
dz/(z + 12), where γ is the path pictured in example 32.

Cauchy Integral Formula

40. Find:
(a) C (ez + cos z) dz/z where C is the circle |z| = 1.
H
H
(b) C dz/(z + 1)(z − 1) where C is the circle |z − 1| = 1.
(c) C dz/(z + 1)(z − 1)2 where C is the circle |z − 1| = 1.
H

(d) C dz/(z 3 − 1) where C is the circle |z − 1| = 1.

H
H
(e) C dz/z cos z where C is the circle |z| = 1.
(f) C sin z dz/z 4 where C is the circle |z| = 1.
H

(g) C sinh z 2 dz/z 3 where C is the circle |z| = 1.

H

(h) C dz/(z 4 + 4) where C is the circle |z − 1 − i| = 1.

H

41. Find:
(a) C ez dz/(z 2 + 2z) where C is the circle |z| = 1.
H

(b) C sin(zπ/2) dz (z 2 + 2z − 3) where C is the circle |z − 1| = 2.

H 

(c) C e−1/(z+2) tan z dz/z where C is the circle |z| = 1.

H

(d) C dz /(z 2 + 16) where C is the circle |z| = 5.

H

(e) C sinh((z + i)π/2) dz (z 2 − 2z) where C is the circle |z| = 1.

H 

(f) C (sin πz 2 + cos πz 2 ) dz/(z 2 − 3z + 2) where C is the circle |z| = 3.

H

(g) C sin6 z dz/(z − π/6) where C is the circle x2 − 0.6x + y 2 = 0.

H

(h) C sin6 z dz/(z − π/6) where C is the circle x2 − 0.5x + y 2 = 0.

H

42. Calculate |z|=1 z dz/ez .

H

43. Calculate |z|=2 ez dz/(z − iπ/2).

H

44. Calculate C (z + 6) dz/(z 2 − 4), where C is the circle (x − 2)2 + y 2 = 1.

H
H
45. Calculate C (10z − 50) dz/(z − 1)(z − 2)(z + 3)(z + 4), where C is the circle with center at
z = 12 and radius equal to 3.
46. Calculate C 8 dz/(z 2 − i10z/3 − 1), where C is the unit
H
y
circle.
47. Calculate C ez ln 3 dz/(z 3 − z), where C is the triangle
H
C
with vertices at A = − 21 − i3, B = 2 and C = − 12 + i3,
covered counterclockwise. Simplify the result as much as
possible.
2
48. Calculate |z|=2 ez −3z+2 dz/(z − 1)3 .
H x
49. Calculate |z|=1 dz/(5z 2 − i 26z − 5).
H

50. Calculate C cos 4z dz/z 3 , where C is the contour shown

H

in the picture on the right (the origin is marked).

44 Tutorial Problems Complex Integration

ANSWERS

21 2/3. 22 The value is i2 in both cases.

23 (a) −i 32, (b) −i 2π.
24 i 32π. 25 −4 − i π. 26 0. 27 −π/2.
28 8/15 + i 5/6; use the parametric equations x = t, y = t2 .
29 i 4π/15; use the parametric equations x = 3 cosh t, y = 4 sinh t.
1 2
30 4 (1 + i)(e − 1); use the parametric equations x = y = t.
31 i 2π.
32 i 3π/4; use the parametric equations x = cos3 t, y = sin3 t.
33 −1. 34 −1/4. 35 8π − 16 − i(2π 2 − 16).
36 i(cosh 2 − cos 2). 37 (1 + i)/48. 38 sin 1/2 + i (cos 1/2 − e−1 ).
4π + 6
39 ln + i 4π.
π+6
40 (a) i 4π (b) iπ (c) −i π/2 (d) i 2π/3 (e) i 2π (f) −i π/3 (g) i 2π (h) π − i π)/8.
41 (a) iπ (b) iπ/2 (c) 0 (d) 0 (e) π (f) i4π (g) iπ/32 (h) 0.
42 0 43 −2π 44 i 4π 45 i 2π. 46 −6π. 47 i π. 48 i 3π. 49 −π/12. 50 0.