Arithmetic, Geometry and Topology the Complex

Plane

In this chapter, we will study the basic properties of the field of complex numbers. We
will begin with a brief historic sketch of how the study of complex numbers came to be and
then proceed to develop tools needed to study calculus on the complex plane. We will also
give several applications of introducing complex numbers to solving classical problems from
geometry and trigonometry.

1 Arithmetic of the complex plane

1.1 Why complex numbers?
If you recall from your study of Real Analysis, the introduction of the real numbers as the
completion of the field of rational numbers is unavoidable if one wants to develop calculus in an
adequate manner. The question arises as to why we need to enlarge R further and introduce the
field C. Complex numbers were first introduced in 1545 by the Italian mathematician Cardano
in his Ars magna in connection with quadratic equations and he immediately discards them
commenting they were “as subtle as they are useless”. In fact, complex numbers were often
dismissed as “imaginary” or “impossible” even by prominent mathematicians such as Leibniz.
So why study them at all?
Recall from high school mathematics that the solution of the quadratic equation x2 = mx+c
is given by the expression
1 p 
x= m ± m2 + 4c . (1.1)
2
The quantity under the square root is called the discriminant, denoted D. According to most
textbooks, complex numbers were introduced to ensure that quadratic equations always have
solutions. This is not only historically inaccurate but also highly misleading.
Geometrically, solving the quadratic equation x2 = mx + c is same as finding the points
of intersection of the parabola P given by the equation y = x2 and the line L given by the
equation y = mx + c. Three possibilities can arise,

(i) L and P intersects at two points. This corresponds to (1.1) yielding two real solutions.
In this case D > 0.
(ii) L and P intersects at one point. In this case D = 0.
(iii) L and P do not intersect at all. In this case D < 0.

Thus, when D < 0, the fact that P and L do not intersect is reflected in the fact that (1.1)
are complex numbers. This shows that there is absolutely no reason to introduce complex
numbers in the study of quadratic equations. Cardano was perfectly justified in dismissing

1
 1 Arithmetic of the complex plane

complex numbers as “useless” in connection to solving quadratic equations. That complex

numbers were introduced to solve quadratic equations is a lie repeated blindly by many ill-
informed textbook authors!
The correct reason for introducing complex numbers is in connection with solving cubics.
We begin with the first theorem of the course.

Theorem 1 (Cardano). The solution of the cubic equation

x3 = 3px + 2q (1.2)

is given by q q
3 3
p p
x = q + q − p + q − q2 − p3 .
2 3 (1.3)
Any arbitrary cubic x3 + ax2 + bx + c can be transformed by a linear change of coordinates
to an equation of the form (1.2).

Proof. It is an exercise for you to prove that an arbitrary cubic can indeed be transformed to
an equation of the form (1.2).
To solve (1.2), we first set x = u + v. Expanding the LHS of (1.2), we get

u3 + v3 + 3uv(u + v) = u3 + v3 + 3uvx.

Equating the above equation with the RHS of (1.2), we see that p = uv and that u3 + v3 = 2q.
Eliminating v, we end up with the equation

p3
u3 + − 2q = 0.
u3
This is a quadratic in u3 whose solutions are given by
p
u3 = q ± q2 − p3 .

By symmetry, v3 has the exact same solutions. As u3 + v3 = 2q, without loss of generality, we
can take as the solutions for u3 and v3 as
p
u3 = q + q2 − p3
p
v3 = q − q2 − p3

Thus the required solutions are given by

q q
3 3
p p
x = q + q − p + q − q2 − p3 .
2 3

Geometrically, solving for the cubic (1.2) is equivalent finding the intersection of the cubic
with the line L given by the equation y = 3px + 2q. Note that a cubic equation always has at
least one real root (why?). This means that the formula (1.3) must always yield at least one

2
 1 Arithmetic of the complex plane

real number. It was Bombelli who realized that there is something strange about the formula.
He considered the cubic x3 = 15x + 4 which has solutions
√3
√
3
x = 2 + 11i + 2 − 11i,

here we are freely using complex notation which I am assuming you are already familiar with.
The above does not seem to be a real number at all. But, Bombelli had a “wild thought”. By
√
guessing, he realized that x = 4 solves the cubic. So he assumed that 3 2 + 11i is an expression
√
of the type 2+ui and 3 2 − 11i is an expression of the form 2−ui, so that x = 2+ui+2−ui = 4.
Of course, for this to make sense Bombelli assumed that the ordinary laws of algebra are true
for the complex numbers, i.e.,

a + ib + a 0 + ib 0 = (a + a 0 ) + i(b + b 0 ).

Next to determine u, he needed to evaluate (2+ui)3 . To do this, he assumed that multiplication

obeys the following obvious rule

(a + ib)(a 0 + ib 0 ) = (aa 0 − bb 0 ) + i(a 0 b + ab 0 ),

where we are using i2 = −1. Expanding out (2 + ui)3 using the above rule we get

−iu3 − 6u2 + 12iu + 8 = 2 + 11i,

which readily yields u = 1. Similarly, (2 − i)3 = 2 − 11i.

Bombelli’s “wild thought” shows that the “useless” complex numbers are unavoidable in the
solution of cubics. However, the study of complex numbers remained a mere curiosity and
were considered mysterious for almost 250 years.

1.2 Notation and terminology

We will identify the set of complex numbers, denoted C, with R2 . In this identification, the
complex number a + ib corresponds to the pair (a, b). The point 1 ∈ C corresponds to (1, 0)
and the point i corresponds to (0, 1). Geometrically, a complex number is nothing but a vector
in the so called Argand–Gauss complex plane.
The following picture and table summarizes all the relevant notation and terminology related
to complex numbers.

Terminology Meaning Notation

modulus of z length r of the vector z |z|
argument of z angle θ that the vector z makes with the x-axis arg(z)
real part of z x coordinate of the vector z Re(z)
imaginary part of z y coordinate of the vector z Im(z)
real axis the set of real numbers
imaginary number a number that is a real multiple of i
imaginary axis the set of imaginary numbers
complex conjugate of z reflection of z in the real axis z

3
 1 Arithmetic of the complex plane

z = x + iy
y = Im(z)
r = |z|

θ = arg(z)

0 x = Re(z)

z = x − iy

Figure 1: The complex plane

Note that for z = x + iy, z = x − iy and

1 1
Re(z) = (z + z), Im(z) = (z − z).
2 2i
The sum of two complex numbers z and w can be obtained geometrically using the paral-
lelogram law for addition of vectors.
In order to visualize the product, we need to introduce the polar representation of complex
numbers in terms of r andp θ. The modulus
√ of |z| is the distance from the origin to the point
(x, y). Explicitly, |z| = x + y = z · z. The modulus satisfies a number of simple and easy
2 2

to prove inequalities.

X
n X
n
zi 6 |zi |,
i=1 i=1
||z| − |w|| 6 |z ± w|,
|z1 · · · zn | 6 |z| · |w|.

With the sum and product defined as in the previous section, C is a commutative field. The
inverse of the number z 6= 0 is given by z−1 = 1z = |z|z2 .
We now want rigorously define the the argument arg(z). To do this we will have to first
introduce trigonometric functions. The correct way to do this is to use power series which
we shall indulge in at a later time. For the time being, we recall the definitions in terms of
circular functions that you are no doubt familiar with from calculus.
First of all, we define 2π to be the length of the circumference of the circle of radius 1.
Denote the unit circle by T. We now define a map 1t : R → T as follows: for t > 0, 1t is the
point on T obtained by starting at the point 1 ∈ T and moving a distance t counter-clockwise;
for t < 0 one does the same thing but clockwise. Note that, by definition, 1t is a 2π-periodic

4
 1 Arithmetic of the complex plane

function. We define
sin t := Im1t , cos t := Re1t .
All the familiar identities of trigonometry can easily by derived by using the above definition.
The fact that sin2 t + cos2 t = 1 is now obvious as (cos t, sin t) lies on the unit circle. Other
identities such as the following fundamental identities:

cos(t + s) = cos t cos s − sin t sin s

(1.4)
sin(t + s) = sin t cos s + cos t sin s,

can be proved using geometric arguments. These identities prove that

1s · 1t = 1t+s ,

which says that 1t is a group homomorphism from the additive group (R, +) to the multi-
plicative group (T, ·) with kernel 2πZ. We immediately get the de Moivre’s formula (cos t +
i sin t)n = cos nt + i sin nt. Classical trigonometric identities such as formulas for cos 2t, sin 2t,
etc., easily follow. One also gets the usual differentiation rules for sin and cos easily.
If 0 6= z ∈ C, then z/|z| ∈ T, which implies that we can find θ ∈ R such that 1θ = z/|z|.

Definition 2. The argument of 0 6= z ∈ C, denoted arg z, is any number θ ∈ R such that

1θ = z/|z|. Among the all the arguments of z, there is precisely one that belongs to the interval
(−π, π] called the principal argument of z and denoted Arg z.

Sometimes we will be sloppy with notation and let arg z denote the set {θ + 2πk, k ∈ Z}
the set of all possible arguments of z. The argument of z is uniquely determined modulo an
integer multiple of 2π. The definition of arg z combined with (1.4) easily implies that

1
arg zw = arg z + arg w, arg = arg z = − arg z.
z
We define the angle between z and w when z, w 6= 0 as the angle that goes from z to w.
i.e., arg w − arg z. This is the same as arg wz and arg wz . Note that these are oriented angles
and might be negative.
One can determine Arg z by using the inverse tangent function arctan as follows. Recall that
the arctan function is a bijective continuous function from R to (−π/2, π/2). If z = x + iy, x >
0, then Arg z = arctan yx . If z = iy, y > 0, then Arg z = π/2. In the second quadrant
z = x + iy, x < 0, y > 0, Arg z = π + arctan yx and in the third quadrant z = x + iy, x < 0, y <
0, Arg z = π − arctan yx . Finally, if z = iy, y < 0, then Arg z = −π/2.
We are now in a position to give a geometric interpretation of complex multiplication. We
represent the complex number z using the polar representation as rθ , where r = |z| and
θ = arg z. Then as |zw| = |z| · |w| and arg zw = arg z + arg w, it follows multiplying w z is same
as dilation by the real number |z| composed with a rotation by angle θ (the order in which
the dilation and rotation is performed is irrelevant). In particular, z itself can be written as
|z| · 1arg z . Very soon, we shall give a better way to represent complex numbers in terms of the
exp function.

5
 1 Arithmetic of the complex plane

z = x + iy

z x y
|z| = |z| + |z|

θ = arctan yx

Figure 2: Determining Arg z

1.3 Powers and n-th roots

The historic lie perpetuated in textbooks is that complex numbers were introduced to solve
quadratic equations. In fact, the fundamental theorem of algebra asserts that C is algebraically
closed. In particular, given a complex number z0 , the polynomial equation zn − z0 = 0 must
have solutions. We can explicitly obtain the n-roots using the polar representation of z0 . Write
z0 = r1θ , then zn = z0 means that

|z|n = r, n arg z = θ + 2πk, k ∈ Z.

This means that |z| = r1/n and arg z = nθ + 2π nk . Note that two different values of k, say
k1 , k2 will give the same value for arg z iff k1 − k2 is an integer multiple of 2π. Thus there are
n-distinct solutions, corresponding to k = 0, 1, . . . , n − 1.
The n-th roots of unity form a cyclic group of order n generated by the element 12π/n which
√
is called the n-th primitive root. For x > 0, x will always denote the positive square root.
√
If 0 6= z ∈ C, then z := |z|1 Arg z . This is called the principal branch of the square root.
p
√ 2
Similarly, n z := n |z|1 Arg z .
p
n

6
 1 Arithmetic of the complex plane

The integer powers of z are defined in the standard manner:

zn = z| ·{z
· · z}
n-times
−n −1
z = z| · · z−1}
·{z
n-times
 1 n
Taking rational powers is more complicated. If q = m n
, then zq := z m . Note that this a
set and not a single number. The set does not depend on the particular representation of q.
It is easy to see that as sets,  1 n 1
zm = (zn ) m .

Furthermore, if q = m n
is in its reduced form and z = |z|1θ then in polar representation, zq is
the set of m distinct values
p
m
|z|n 1 n θ+k 2πn , k = 0, . . . , m − 1.
m m

Note that the law of exponents is not true. For instance,

 n m
n
z 6= z m

simply because the LHS is a number whereas the RHS is a set. Think on how to rephrase the
law of exponents to make it true in our situations.

1.4 The field structure on C

As remarked before, C is a commutative field. It is natural to ask if it is ordered.

Definition 3. Let S be a set. A total ordering on S is a relation 6 that satisfies

1. Reflexivity: a 6 a for all a in S.

2. Antisymmetry: a 6 b and b 6 a implies a = b.
3. Transitivity: a 6 b and b 6 c implies a 6 c.
4. Comparability (trichotomy law): For any a, b in S, either a 6 b or b 6 a.

If a 6 b and a 6= b, we often write a < b.

Definition 4 (Ordered Field). A field K with a total ordering 6 is said to be an ordered field
if it satisfies

• if a 6 b then a + c 6 a + c, ∀c ∈ K,
• if 0 6 a and 0 6 b then 0 6 a · b.

Theorem 5. C cannot be given the structure of an ordered field.

Proof. Assume to the contrary that 6 makes C into an ordered field. Then either i > 0 or
i < 0. Suppose i > 0. Then i2 = −1 > 0. Adding 1 to both sides, we see that 0 > 1. On
the other hand −1 > 0 implies that (−1)(−1) = 1 > 0. This is a contradiction. An analogous
argument works for i < 0.

7
 2 Geometry and Topology of C

Another natural question is the following: view C as sitting in R3 as the first two coordinates;
Can we give R3 the structure of a commutative field so that C is a subfield?

Theorem 6. R3 cannot be given the structure of a commutative field such that C is a

subfield.

Proof. R3 is obviously a vector space. Assume also that we have defined a multiplication · on
R3 that makes it a commutative field that extends C. Denote the basis vectors (1, 0, 0), (0, 1, 0)
and (0, 0, 1) by 1, i, j, respectively. With this notation, i2 = −1. Let us compute ij. Suppose
ij = a + bi + cj, a, b, c ∈ R. Observe first that

−j = (i2 )j = i(ij), j2 = −(ij)2 .

So,
−j = ai − b + c(ij) = ai − b + c(a + bi + cj).
Equating coefficients, we see that c2 = −1 which is absurd.

2 Geometry and Topology of C

2.1 Conformal and anticonformal linear mappings
In this section, we consider a R-linear mapping T : C → C, i.e., T (λ1 z1 + λ2 z2 ) = λ1 T (z1 ) +
λ2 T (z2 ), ∀λ1 , λ2 ∈ R. Now, C is both a vector space over R of dimension 2 and vector space
over C of dimension 1. It is natural to ask for conditions on T that guarantee that T is C-linear.
Two obvious necessary conditions:

1. T is R-linear,
2. T commutes with multiplication by i, i.e., T (iz) = iT (z)∀z ∈ C.

It turns out that these two conditions are sufficient to guarantee C-linearity of T . Suppose
T is C-linear, then T (z) = T (1)z and therefore T is just multiplication by a scalar. Writing
T (1) = (a, c) and T (i) = (b, d), we see that the matrix of T under the standard basis is:
 
a b
c d

This means that T (x, y) = (ax + by, cx + dy), setting x = z+z 2 , y = 2i and performing a
z−z

straightforward computation yields that T (z) = αz + βz where α = 2 (a + d − ib + ic) and

1

β = 12 (a − d + ic + ib). Note that given arbitrary α, β ∈ C, we can find a, b, c, d that satisfy

the above relations and hence any map of the form αz + βz is automatically R-linear. Hence,
T is C-linear iff β = 0. This means that a = d and c = −b. In matrix form, C-linear maps
have a matrix of type  
a b
−b a
Now, if T is R-linear and commutes with i, then T (iz) = iT (z) = αiz + βiz = αiz + βiz which
means that iβz = −βiz and so β = 0.

8
 2 Geometry and Topology of C

Another easy computation shows that

det T = ad − bc = |α|2 − |β|2 ,
which means that T is invertible iff |α| 6= |β|. In this case, we can explicitly solve for the inverse
T −1 :
αw − βw
T −1 (w) = 2
|α| − |β|2
Definition 7. An R-linear and invertible mapping T : C → C is said to be a conformal linear
mapping if it preserves oriented angles, i.e, if z, w ∈ C \ {0} then
arg Tz − arg Tw = arg z − arg w.
Equivalently, arg Tz − arg z = arg Tw − arg w, or in other words, the function arg Tz − arg z is
constant modulo 2πZ. This means that arg Tz z lies on a ray originating from the 0. However,
if Tz = αz + βz, then
Tz z
=α+β .
z z
Note that the RHS above traces out a circle centred at α of radius |β| and thus can lie on a
ray iff β = 0. We have proved
Proposition 8. Let T : C → C be an invertible R-linear mapping. Then the following are
equivalent:
1. T is conformal.
2. T is of the form αz, α ∈ C.
3. T is C-linear.
4. T is a composition of a rotation followed by a dilation.
On the other hand, if α = 0, then T = βz and in this case T preserves angles but reverse
orientation. We say that T is anitconformal or C-anitlinear. In this case the matrix of T is
of the form  
a −c
c a
In addition, if T also preserves distances, then |T (z)| = |z| in which event either |α| or |β| = 1.
It is easy to see that in this case T must be an orthogonal matrix.

2.2 Analytic geometry

We want to express the usual notions from analytic geometry in complex notation. Changing
to complex coordinates often greatly simplifies problems of analytic geometry. We equip C
with the usual Hermitian inner-product hz, wi = z · w. It follows that the usual inner-product
on R2 is Rehz, wi.
The equation of line in analytic geometry is Ax + By + C = 0. Setting x = z+z 2 and
y = z−z
2i and computing shows that the general form a line L in complex notation is of the
form αz + αz + m = 0, α ∈ C, m ∈ R. In parametric form, the line joining z1 , z2 ∈ C is of the
form z1 + tz2 , t ∈ R.
The equation of a circle of radius r centred at a ∈ C is given by |z − a| = r. In parametric
form, z(t) = a + r1t .

9
 2 Geometry and Topology of C

2.3 Curves
A substantial portion of our course will involve path integrals. We recall here some basic
notions about curves and paths to set the stage for later.

Definition 9. Let U ⊂ C be a domain. A curve in U is a continuous mapping γ : [a, b] → U

where [a, b] is an interval in R. The curve γ is said to be closed if γ(a) = γ(b) and we say it
is simple if γ is injective.

We will often denote the image of the curve γ called the support of γ by γ∗ but often we
shall identify both γ and γ∗ . We imagine γ∗ to be a continuous thread in space but that
picture is incorrect as there are curves γ : [0, 1] → [0, 1] × [0, 1] such that γ∗ = [0, 1] × [0, 1]!
In this context, it is appropriate to state the Hahn–Mazurkiewicz theorem which characterizes
all possible continuous images of the unit interval.

Theorem 10 (Hahn–Mazurkiewicz). A non-empty Hausdorff topological space is a contin-

uous image of the unit interval if and only if it is a compact, connected, locally connected
second-countable space.

We will often call a closed curve that is injective expect at the endpoints a Jordan curve.
Such a curve has no self-intersections except at the endpoints. Note that that a Jordan curve
is homemorphic to the unit circle S1 .
Our main focus in this course will be curves that are “sufficiently nice” to allow us to integrate
continuous functions along the curve. To this end, we make the following

Definition 11. A curve γ : [a, b] → U is said to be differentiable if γ can be extended to

some open set G ⊃ [a, b] as a differentiable function, i.e., writing γ(t) = x(t) + iy(t), both
x(t) and y(t) are differentiable functions on (a, b) and the one-sided derivatives of x(t) and
y(t) exist at a and b respectively. If moreover, γ 0 (t) := x 0 (t) + iy 0 (t) are continuous on [a, b],
we say that γ is of class C1 . Furthermore, if the derivative is non-zero at each point, then we
say that γ is regular. The adjective piecewise is prefixed to indicate that the property is true
for all but finitely many points. A path is a curve that is piecewise of class C1 .

We will visualize paths as being trajectories of particles moving in the plane. In this visualiza-
tion, the derivative γ 0 (t) which is geometrically the tangent to γ at the point γ(t) is nothing
but the velocity of the moving particle at time t.

Example 12. We have already encountered one example of a Jordan curve. The curve 1t |[−π,π]
is Jordan curve that is also regular. This curve is in fact of class C∞ . Such a curve is called a
smooth curve. The derivative of this curve

1t0 (t) = − sin t + i cos t = iγ(t) (2.1)

Notice that geometrically γ 0 (t) is the tangent to γ at the point γ(t). Thus (2.1) says that the
tangent to 1t at the point t is orthogonal to the vector 1t (t). This is because multiplication
by i is rotation by π2 .

10
 2 Geometry and Topology of C

It is time to improve the notation 1t to one that is significantly better. We define

eit := 1t = cos t + i sin t. (2.2)
This is called Euler’s formula. In particular, we have
eiπ = −1,
which is widely considered one of the most beautiful equations in all of mathematics.
Recall that e is the unique real number with the property that the function ex : R → R has
derivative itself. It might seem quite mysterious that Euler’s constant appears out of thin air in
(2.2). Let us motivate the grand entrance of e into complex analysis. The property (ex ) 0 = ex
characterized the exponential uniquely. If we want to extend the definition to numbers of the
form it, it is natural to expect that (eit ) 0 = ieit . Let us temporarily denote our candidate
function for eit by Z(t). Then Z 0 (t) = iZ(t) and also Z(0) = 1. This says that at time 0,
the vector Z(0) = 1 and the velocity Z 0 (0) = i. A split second later the particle has moved
upwards a bit in the direction i. Now, the velocity is again orthogonal to the position Z(t)
and thus the curve turns slightly. In this way, we see that the curve Z(t) traces out the unit
circle which is exactly what the definition (2.2) is saying.
For convenience, we summarize some facts about eit that you have already seen when you
studied the function 1t :
(i) eit · eis = ei(t+s) .
(ii) Any 0 6= z ∈ C can be written as z = |z| · ei arg z .
(iii) z · w = |z| · |w| · ei(arg z+arg w) .
(iv) The reciprocal of eit is the function e−it .
it −it
(v) cos t = Re eit = e +e 2 .
eit −e−it
(vi) sin t = Im e =it
2i .
Often curves are described in polar form as follows. We specify two continuous functions
R : [a, b] → R+ and φ : [a, b] → R and set
γ(t) = R(t)eiφ(t) .
γ is clearly a continuous curve in C \ {0} and Γ is differentiable iff R and φ are. Furthermore,
γ 0 (t) = (R 0 (t) + iR(t)φ(t))eiφ(t) .
Question: Can every continuous curve be written this way?
Example 13. A cardioid is a plane curve traced by a point on the perimeter of a circle that is
rolling around a fixed circle of the same radius. To illustrate the power of complex notation,
let us compute the equation of the Cardiod traced by the point 1 which lies on the circle of
radius 1 centred at 2 that is rolling on the unit circle. After the circle has rolled a distance
t, its centre is now at 2eit and the point that is now touching the unit circle is the point
2eit + ei(π+t) . This means that that the image of the point 1 is 2eit + ei(π+2t) = 2eit − ei2t . This
is the complex form of the equation and an easy computations with this expression can be used
to obtain the polar form as well as the parametric equations of the cardioid. Please compare
this derivations with the usual derivations using classical analytic geometry to understand the
power of complex notation.

11
 2 Geometry and Topology of C

2eiθ θ

θ
θ image of 1
θ
θ 1

Figure 3: Equation of cardioid

Example 14. Another illustration off the power of complex notation is to find the equation of
a spiral. Let R(t) be an increasing continuous positivie function such that limt→−∞ R(t) = 0
and limt→∞ R(t) = ∞, then it is easy to see that

γ(t) = R(t)eit

traces a spiral which expands as it rotates counter-clockwise.

We will now talk about reparametrization of curves. Given a curve γ, there could be many
other curves whose image coincides with γ∗ . For, instance the unit circle is the image of both
2
the curves γ1 (t) = e2πit and γ2 (t) = e−2πit . γ1 traverses the circle counter-clockwise at unit
speed whereas γ2 clockwise with speed 2.

Definition 15. Let γ : [a, b] → U be a curve. We say that Γ : [c, d] → U is a reparametrization

of γ if

1. Γ ∗ = γ∗ ;
2. There exists a continuous strictly monotone map Φ : [a, b] → [c, d] such that Γ ◦ Φ = γ.

Remark 16. Such a map Φ is a homeomorphism of intervals. In fact, any homeomorphism of

intervals is either strictly increasing or strictly decreasing. This defines an equivalence relation
on the curves as follows: two curves γ1 and γ2 are equivalent if one is the reparametrization
of the other by a strictly increasing homeomorphism.

12
 2 Geometry and Topology of C

Given a curve γ : [a, b] → U, we define the map Φ : [a, b] → [a, b] by Φ(t) = b + (a − t)

and let γ be the curve γ ◦ Φ. Intuitively, γ is the same curve γ traversed in reverse. Any
parametrization of γ∗ is either equivalent to γ or γ. Thus there are only two equivalenence
classes of curves that parametrize γ∗ and an orientation of γ∗ is a choice of one of these
classes.
We will now recall notions about the length of curve. We consider a partition P, a = t0 <
t1 < · · · < tn = b of [a, b] and define

X
n−1
L(P) := |γ(ti+1 ) − γ(ti )|.
i=0

P should be thought of as an inscribed polygon on γ with length L(P). The length of the curve

L(γ) := sup{L(P) : P is a partition of [a,b]}.

If L(γ) < ∞ then γ is said to be rectifiable. Note that when if γ is piecewise C1 then it is
rectifiable and Zb
L(γ) = |γ 0 (t)|dt.
a
This integral is well-defined as is continuous on [a, b] except for finitely many points.
|γ 0 (t)|
Note also that the integral above is invariant under a change of parametrization.
If γ is regular then defining Zt
s(t) := |γ 0 (x)|dx,
a
we see by the fundamental theorem of calculus that s 0 (t) exists and is equal to |γ 0 (x)|. Fur-
thermore as |γ 0 | is a nowhere vanishing function, s(t) is a strictly increasing function. This
means that s : [a, b] → [0, L] is a strictly increasing C1 function and thus setting Ψ = s−1 , we
see that Γ = γ ◦ Ψ−1 is a reparametrization with the property that the length Γ |[0,t] = t. It
is also easy to see that Γ is a unit-speed curve. This parametrization is called the arc-length
parametrization.

Definition 17. Let U ⊂ C be open, γ : [a, b] → U be a path and h : U → C be a continuous

function. We define the path integral
Z Zb
h := h(γ(t))|γ 0 (t)|dt.
γ a

This is well-defined and independent of the parametrization of γ.

2.4 Branches of the argument function and the index of a closed curve
As we have mentioned before, the function Arg : C \ {0} → (−π, π] is not continuous on any
point of the negative real axis. However, it is continuous at any other point in C \ R− .

Definition 18. Let E ⊂ C be a set that does not contain 0. Then any continuous mapping
g : E → R that satisfies g(z) ∈ arg z is called a continuous branch of the argument in E.

13
 2 Geometry and Topology of C

If h, g : E → R are two branches of the argument, then h−g

2π is a continuous integer valued
function on E and is hence constant on each connected component of E. In particular, if E
were connected then any two branches of the argument differ by a constant integer multiple
of 2π. On C \ R− , any branch of the argument is of the form Arg z + 2πZ.

Example 19. On any circle C centred at the origin, there are no continuous branches of the
argument. If not, then let g : C → R be a branch of the argument and let a ∈ R be the point
on the negative real axis that intersects C. Then on the connected set C \ {a}, we can find a
k ∈ Z such that Arg z + 2πk = g(z). As the value of Arg jumps at the point a, we see that g
is discontinuous at a, a contradiction.

Example 20. If L is any ray starting from the origin, say L := Arg z = α, all branches of arg z
are given by
arg z = Arg(z · ei (π − α) − π + α + 2πZ.
This is derived as follows. We first rotate by π − α so that the ray L coincides with the
negative real axis. Now we find the value of Arg and subtract π − α to compensate for our
initial rotation.

Example 21. Let D := D(a, r) be a disk that misses the origin, then we can find a continuous
branch of arg on D. To see this, consider a (possibly) larger disk D(a, r 0 ) such that 0 ∈
∂D(a, r 0 ). Then let L be ray starting at the origin tangent to ∂D(a, r 0 ) at 0. Then any branch
of arg on C \ L works.

Definition 22. Let X be a connected topological space and let f : X → C \ {0} be a continuous
function. A continuous branch of the argument of f is any continuous function h : X → R
such that h(x) ∈ arg f(x) ∀x ∈ X.

One important case is when X = E ⊂ C and f is the identity function in which case we
recover the definition of a continuous branch of arg. We will also consider branches of the n-th
root of f : E → C \ 0 : any continuous function h : E → C such that hn = f. If g : E → R
is branch of the argument, then h = |f|1/n eig/n is branch of the n-th root of f. The following
example illustrates that the converse is not, in general, true.

Example 23. Let E = C \ [−1, 1] and let f(z) = z2 − 1. Note that the image √ of z − 1 is
2

contained in (−∞, 0] precisely when z ∈ [−1, 1] or the imaginary axis. We take z2 − 1 using
the principal square root on the right half-plane and the negative of the principal square root
on the left half-plane.

We come back to the question about polar representation of curves. γ : [a, b] → C is a

continuous curve that does not contain the origin. We want to write

γ(t) = R(t)eiφ(t) .

Of course, R(t) = |γ(t)| which is a continuous function as γ(t) 6= 0. For each t ∈ [a, b], we
must select φ(t) ∈ arg γ(t) in a continuous way. This is of course possible, if we can find a
branch of arg on γ∗ . But it actually suffices to find a branch of the argument of γ.

14
 2 Geometry and Topology of C

Theorem 24. If γ : [a, b] → C is a curve that misses the origin, then we can find a
continuous branch of the argument of γ, say φ(t). Any other branch of the argument of
γ is of the form φ + 2πZ. Moreover, if γ is differentiable at the point t ∈ [a, b], then so
0 (t)
is φ and φ 0 (t) = Im γγ(t) .

Proof. Observe that on any disk in C that does not contain the origin, we can define a
continuous branch of the argument. Let r := dist(γ∗ , 0). As γ∗ is compact, r > 0. Finitely
many sets of the form γ−1 (D(t, r) cover [a, b] by the compactness of [a, b]. Hence, we can
find a partition a = t0 < t1 < · · · < tn = b such that each γ([ti−1 , ti ) is contained in a disk
of the form Di := D(t, r). Let gi be a branch of arg on Di . Then φi = gi ◦ γ is a branch of
arg(γ(t), t ∈ [ti−1 , ti ]. At the point ti , φi (ti ) might not be equal to φi+1 (ti ). Nevertheless,
their difference is an integer multiple of 2π which we add to φi+1 . By gluing lemma for
continuous functions, this allows to construct a global branch of the argument of γ.
Now, suppose γ is differentiable at the point t. We may, without loss of generality, assume
that γ(t) lies on the upper half-plane. This in a neighbourhood of t, φ(t) = Arg γ(t) + 2πk
for some k ∈ Z. Thus, φ 0 (t) exists and is equal to

x2 (t) x(t)y 0 (t) − y(t)x 0 (t) γ 0 (t)

 
d d y(t)
(Arg γ(t)) = arctan = 2 · = Im ,
dt dt x(t) x (t) + y2 (t) x2 (t) γ(t)
where γ(t) = (x(t), y(t)).

Example 25. The curve γ(t) = eit , t ∈ [0, 2π] has a continuous branch of the argument given
by h(t) = t, but clearly no branch of arg can exist on the circle T.

Now, let γ : [a, b] → C \ {0} be a curve that misses 0 and let φ be a branch of the argument
of γ. Then φ(b) − φ(a) does not depend on the choice of φ as any two choices differ by an
element of 2πZ. This is called the variation of the argument and denoted ∆γ arg. If, in
addition, γ is a path, then
Zb 0
γ (t)
∆γ arg = Im dt,
a γ(t)
by the fundamental theorem of calculus and the previous theorem.

2.5 The index or winding number of a closed curve

φ(b)−φ(b)
If γ : [a, b] → C \ {0} is a closed curve, then for any branch φ of the argument of γ, 2π
is an integer as γ(a) = γ(b).

Definition 26. For a closed curve γ : [a, b] → C \ {0}, we define the index or the winding
∆ arg
number of γ with respect to 0, Ind(γ, 0) by the number γ2π .

Intuitively, the winding number measures the number of rotations around 0 a particle moving
along γ makes. One can easily define this notion for more general cases.

Definition 27. Let γ : [a, b] → C \ {z} be a closed curve. Then, we define the index of γ with
respect to z, Ind(γ, z) by
Ind(γ, z) = Ind(γ − z, 0).

15
 2 Geometry and Topology of C

Now, suppose γ = R(t)eiφ(t) is a path that misses the origin. Then γ 0 (t) = (R 0 (t) +
iR(t)φ(t))eiφ(t) . Thus
γ 0 (t)
 0 
R (t) 0
= + iφ (t) .
γ(t) R(t)
Evaluating
Zb Zb Zb
γ 0 (t) γ 0 (t) γ 0 (t) |γ(b)|
dt = Re dt + i dt = log + i∆γ arg .
a γ(t) a γ(t) a γ(t) γ(a)

This shows that for a path γ : [a, b] → C \ {z}

Zb
1 γ 0 (t)
Ind(γ, z) = dt.
2πi a γ(t) − z

One can rewrite the integral above in the following form

Z
1 dz
Ind(γ, z) = .
2πi γ 1 − z

The above integral expressions clearly show that whenever γ is a path, the function Ind(z, γ)
as a function of z on C \ γ∗ is a continuous function.
Now, let γ : [a, b] → C be a closed curve. Then C \ γ∗ is an open set and hence has only
countably many connected components. On each of these connected components, the function
Ind(·, γ) is a constant function. Moreover, C \ γ∗ has only one connected component that
is unbounded. This is because γ∗ is compact and is hence contained in some large disk D.
Thus the unique unbounded component is the one that contains C \ D. We claim that on the
unbounded component, Ind(·, γ) is zero. We know that it is constant, so it suffices to pick
z ∈ C \ D and show that Ind(z, γ) = Ind(γ − z, 0) = 0. But γ − z ⊂ D − z and D − z is a disk
centred at z that misses the origin. This means that we can find a branch of arg on D − z and
hence ∆γ−z arg = 0 proving our claim.

2.6 Applications
Our first application will be to find an expression for cos4 t. The straightforward method is to
just compute:

2 cos t = (eit + e−it ),

16 cos4 t = (eit + e−it )4
= (ei4t + e−i4t ) + 4(ei2t + e−i2t ) + 6,

which gives
1
cos4 t = (cos 4t + 4 cos 2t + 3).
8

16