Topic 2 Notes: Jeremy Orloff
Topic 2 Notes: Jeremy Orloff
Topic 2 Notes: Jeremy Orloff
Jeremy Orloff
2 Analytic functions
2.1 Introduction
The main goal of this topic is to define and give some of the important properties of
complex analytic functions. A function f (z) is analytic if it has a complex derivative f 0 (z).
In general, the rules for computing derivatives will be familiar to you from single variable
calculus. However, a much richer set of conclusions can be drawn about a complex analytic
function than is generally true about real differentiable functions.
In calculus we defined the derivative as a limit. In complex analysis we will do the same.
∆f f (z + ∆z) − f (z)
f 0 (z) = lim = lim .
∆z→0 ∆z ∆z→0 ∆z
Before giving the derivative our full attention we are going to have to spend some time
exploring and understanding limits. To motivate this we’ll first look at two simple examples
–one positive and one negative.
Example 2.1. Find the derivative of f (z) = z 2 .
answer: We compute using the definition of the derivative as a limit.
That was a positive example. Here’s a negative one which shows that we need to a careful
understanding of limits.
Example 2.2. Let f (z) = z. Show that the limit for f 0 (0) does not converge.
answer: Let’s try to compute f 0 (0) using a limit:
1
2 ANALYTIC FUNCTIONS 2
The limits don’t agree! The problem is that the limit depends on how ∆z approaches 0. If
we came from other directions we’d get other values. There’s nothing to do, but agree that
the limit does not exist.
We’ll there is something we can do: explore and understand limits. Let’s do that now.
Definition. The open disk of radius r around z0 is the set of points z with |z − z0 | < r,
i.e. all points within distance r of z0 .
The open deleted disk of radius r around z0 is the set of points z with 0 < |z − z0 | < r.
That is, we remove the center z0 from the open disk.
r r
z0 z0
then
• lim f (z)g(z) = w1 · w2 .
z→z0
• If h(z) is continuous and defined on a neighborhood of w1 then lim h(f (z)) = h(w1 )
z→z0
(Note: we will give the official definition of continuity in the next section.)
We won’t give a proof of these properties. As a challenge, you can try to supply it using
the formal definition of limits given in the appendix.
2 ANALYTIC FUNCTIONS 4
A function is continuous if it doesn’t have any sudden jumps. This is the gist of the following
definition.
Definition. If the function f (z) is defined on an open disk around z0 and lim f (z) = f (z0 )
z→z0
then we say f is continuous at z0 .
If f is defined on an open region A then the phrase ‘f is continuous on A’ means that f is
continuous at every point in A.
As usual, we can rephrase this in terms of functions of (x, y):
Fact. f (z) = u(x, y) + iv(x, y) is continuous iff u(x, y) and v(x, y) are continuous as
functions of two variables.
Since continuity is defined in terms of limits, we have the following properties of continuous
functions.
Suppose f (z) and g(z) are continuous on a region A. Then
• f (z)g(z) is continuous on A.
Using these properties we can claim continuity for each of the following functions:
2
• ez
• If P (z) and Q(z) are polynomials then P (z)/Q(z) is continuous except at roots of
Q(z).
By definition the extended complex plane = C ∪ {∞}. That is, we have one point at infinity
to be thought of in a limiting sense described as follows.
A sequence of points {zn } goes to infinity if |zn | goes to infinity. This “point at infinity”
is approached in any direction we go. All of the sequences shown in the figure below are
growing, so they all go to the (same) “point at infinity”.
R
Re(z)
1
• lim f (z) = ∞ ⇔ lim =0
z→∞ z→0 f (1/z)
Example 2.6. lim ez is not defined because it has different values if we go to infinity in
z→∞
different directions, e.g. we have ez = ex eiy and
lim ex eiy = 0
x→−∞
lim ex eiy = ∞
x→+∞
lim ex eiy is not defined, i.e. x is constant, so ex eiy loops in a circle indefinitely.)
y→+∞
This is a lovely section and we suggest you read it. However it will be a while before we
use it in 18.04.
One way to visualize the point at ∞ is by using a (unit) Riemann sphere and the associated
stereographic projection. The figure below shows a sphere whose equator is the unit circle
in the complex plane.
Stereographic projection from the sphere to the plane is accomplished by drawing the secant
line from the north pole N through a point on the sphere and seeing where it intersects the
plane. This gives a 1-1 correspondence between a point on the sphere P and a point in the
complex plane z. It is easy to see show that the formula for stereographic projection is
a b
P = (a, b, c) 7→ z = +i .
1−c 1−c
The point N = (0, 0, 1) is special, the secant lines from N through P become tangent lines
to the sphere at N which never intersect the plane. We consider N the point at infinity.
In the figure above, the region outside the large circle through the point z is a neighborhood
of infinity. It corresponds to the small circular cap around N on the sphere. That is, the
small cap around N is a neighborhood of the point at infinity on the sphere!
The figure below shows another common version of stereographic projection. In this figure
the sphere sits with its south pole at the origin. We still project using secant lines from the
north pole.
2.6 Derivatives
The definition of the complex derivative of a complex function is similar to that of a real
derivative of a real function: For a function f (z) the derivative f at z0 is defined as
f (z) − f (z0 )
f 0 (z0 ) = lim
z→z0 z − z0
Provided, of course, that the limit exists. If the limit exists we say f is analytic at z0 or f
is differentiable at z0 .
Remember! The limit has to exist and be the same no matter how you approach z0 .
If f is analytic at all the points in an open region A then we say f is analytic on A.
As usual with derivatives there are several alternative notations. For example, if w = f (z)
we can write
dw f (z) − f (z0 ) ∆w
f 0 (z) = = lim = lim
dz z→z0 z − z0 ∆z→0 ∆z
Challenge. Use polar coordinates to show the limit in the previous example can be any
value with modulus 1 depending on the angle at which z approaches z0 .
It wouldn’t be much fun to compute every derivative using limits. Fortunately, we have the
same differentiation formulas as for real-valued functions. That is, assuming f and g are
differentiable we have:
d
• Sum rule: (f (z) + g(z)) = f 0 + g 0
dz
d
• Product rule: (f (z)g(z)) = f 0 g + f g 0
dz
d f 0g − f g0
• Quotient rule: (f (z)/g(z)) =
dz g2
d
• Chain rule: g(f (z)) = g 0 (f (z))f 0 (z)
dz
df −1 (z) 1
• Inverse rule: = 0 −1
dz f (f (z))
To give you the flavor of these arguments we’ll prove the product rule.
d f (z)g(z) − f (z0 )g(z0 )
(f (z)g(z)) = lim
dz z→z0 z − z0
(f (z) − f (z0 ))g(z) + f (z0 )(g(z) − g(z0 ))
= lim
z→z0 z − z0
f (z) − f (z0 ) (g(z) − g(z0 ))
= lim g(z) + f (z0 )
z→z0 z − z0 z − z0
0 0
= f (z0 )g(z0 ) + f (z0 )g (z0 )
Here is an important fact that you would have guessed. We will prove it in the next section.
Theorem. If f (z) is defined and differentiable on an open disk and f 0 (z) = 0 on the disk
then f (z) is constant.
The Cauchy-Riemann equations are our first consequence of the fact that the limit defining
f (z) must be the same no matter which direction you approach z from. The Cauchy-
Riemann equations will be one of the most important tools in our toolbox.
Before getting to the Cauchy-Riemann equations we remind you about partial derivatives.
If u(x, y) is a function of two variables then the partial derivatives of u are defined as
∂u u(x + ∆x, y) − u(x, y)
(x, y) = lim ,
∂x ∆x→0 ∆x
2 ANALYTIC FUNCTIONS 9
The Cauchy-Riemann equations use the partial derivatives of u and v to do two things:
first, check if f has a complex derivative and second, how to compute that derivative.
We start by stating the equations as a theorem.
Theorem 2.10. (Cauchy-Riemann equations) If f (z) = u(x, y) + iv(x, y) is differentiable
then
∂u ∂v ∂v ∂u
f 0 (z) = +i = −i
∂x ∂x ∂y ∂y
In particular,
∂u ∂v ∂u ∂v
= and =− .
∂x ∂y ∂y ∂x
This last set of partial differential equations is what is usually meant by the Cauchy-Riemann
equations.
Here is the short form of the Cauchy-Riemann equations:
ux = vy
uy = −vx
We’ll compute f 0 (z) by approaching z first from the horizontal direction and then from the
vertical direction. We’ll use the formula
f (z + ∆z) − f (z)
f 0 (z) = lim ,
∆z→0 ∆z
where ∆z = ∆x + i∆y.
Horizontal direction: ∆y = 0, ∆z = ∆x
f (z + ∆z) − f (z)
f 0 (z) = lim
∆z→0 ∆z
f (x + ∆x + iy) − f (x + iy)
= lim
∆x→0 ∆x
(u(x + ∆x, y) + iv(x + ∆x, y)) − (u(x, y) + iv(x, y))
= lim
∆x→0 ∆x
u(x + ∆x, y) − u(x, y) v(x + ∆x, y) − v(x, y)
= lim +i
∆x→0 ∆x ∆x
∂u ∂v
= (x, y) + i (x, y)
∂x ∂x
2 ANALYTIC FUNCTIONS 10
It turns out that the converse is true and will be very useful to us.
Theorem. Consider the function f (z) = u(x, y) + iv(x, y) defined on a region A. If u and v
satisfy the Cauchy-Riemann equations and have continuous partials then f (z) is diffentiable
on A.
The proof of this is a tricky exercise in analysis. It is somewhat beyond the scope of this
class, so we will skip it. With a little effort you should be able to grasp it.
The Cauchy-Riemann equations provide us with a direct way of checking that a function is
differentiable and computing its derivative.
Example 2.11. Use the Cauchy-Riemann equations to show that ez is differentiable and
its derivative is ez .
answer: We write ez = ex+iy = ex cos(y) + iex sin(y). So
We see that ux = vy and uy = −vx , so the Cauchy-Riemann equations are satisfied. Thus,
ez is differentiable and
d z
e = ux + ivx = ex cos(y) + iex sin(y) = ez .
dz
Example 2.12. Use the Cauchy-Riemann equations to show that f (z) = z is not differen-
tiable.
2 ANALYTIC FUNCTIONS 11
ux = 1, uy = 0, vx = 0, vy = −1
Since ux 6= vy the Cauchy-Riemann equations are not satisfied and therefore f is not
differentiable.
Theorem. If f (z) is differentiable on a disk and f 0 (z) = 0 on the disk then f (z) is constant.
Proof. Since f is differentiable and f 0 (z) ≡ 0, the Cauchy-Riemann equations show that
We know from multivariable calculus that a function of (x, y) with both partials identically
zero is constant. Thus u and v are constant, and therefore so is f .
In this section we’ll look at many of the functions you know and love as functions of z. For
each one we’ll have to do three things.
Most often, we can compute the derivatives of a function using the algebraic rules like the
quotient rule. If necessary we can use the Cauchy-Riemann equations or, as a last resort,
even the definition of the derivative as a limit.
Before we start on the gallery we define the term “entire function”.
Definition. A function that is analytic at every point in the complex plane is called an
entire function. We will see that ez , z n , sin(z) are all entire functions.
The following is a concise list of a number of functions and their derivatives. None of the
derivatives will surprise you. We also give important properties for some of the functions.
The proofs for each follow below.
2. f (z) ≡ c (constant)
Domain = all of C (f is entire).
f 0 (z) = 0.
3. f (z) = z n (n an integer ≥ 0)
Domain = all of C (f is entire).
f 0 (z) = nz n−1 .
4. P (z) (polynomial)
A polynomial has the form P (z) = an z n + an−1 z n−1 + . . . + a0 .
Domain = all of C (P (z) is entire).
P 0 (z) = nan z n−1 + (n − 1)an−1 z n−1 + . . . + 2a2 z + a1 .
5. f (z) = 1/z
Domain = C − {0} (the punctured plane).
f 0 (z) = −1/z 2 .
7. sin(z), cos(z)
eiz + e−iz eiz − e−iz
Definition. cos(z) = , sin(z) =
2 2i
(By Euler’s formula we know this is consistent with cos(x) and sin(x) when z = x is
real.)
Domain: these functions are entire.
d cos(z) d sin(z)
= − sin(z), = cos(z).
dz dz
eiw − e−iw
z= ⇒ e2iw − 2izeiw − 1 = 0
2i
Solving the quadratic in eiw gives
√
iw 2iz + −4z 2 + 4 p
e = = iz + 1 − z 2 .
2
Taking the log gives
p p
iw = log(iz + 1 − z 2 ) ⇔ w = −i log(iz + 1 − z 2 ).
Choosing a branch is tricky because both the square root and the log require choices.
We will look at this more carefully in the future.
For now, the following discussion and figure are for your amusement.
Sine (likewise cosine) is not a 1-1 function, so if we want sin−1 (z) to be single-valued
then we have to choose a region where sin(z) is 1-1. (This will be a branch of sin−1 (z),
2 ANALYTIC FUNCTIONS 15
i.e. a range for the image,) The figure below shows a domain where sin(z) is 1-1. The
domain consists of the vertical strip z = x + iy with −π/2 < x < π/2 together with
the two rays on boundary where y ≥ 0. The figure indicates how the regions making
up the domain in the z-plane are mapped to the quadrants in the w-plane.
2.8.2 Proofs
Here we prove at least some of the facts stated in the list just above.
1. f (z) = ez . This was done in Example 2.11 using the Cauchy-Riemann equations.
Now
f (z) − f (z0 ) z n − z0n
f 0 (z0 ) = lim = lim
z→z0 z − z0 z→z0 z − z0
= lim (z n−1
+z n−2
z0 + z n−3 z02 + . . . + z 2 z0n−3 + zz0n−2 + z0n−1 )
z→z0
= nz0n−1 QED.
Since we showed directly that the derivative exists for all z, the function must be
entire.
4. P (z) (polynomial). Since a polynomial is a sum of monomials, the formula for the
derivative follows from the derivative rule for sums and the case f (z) = z n . Likewise
the fact the P (z) is entire.
7. sin(z), cos(z). All the facts about sin(z) and cos(z) follow from their definition in
terms of exponentials.
8. Other trig functions cot(z), sec(z) etc. Since these are all defined in terms of cos and
sin, all the facts about these functions follow from the derivative rules.
9. sinh(z), cosh(z). All the facts about sinh(z) and cosh(z) follow from their definition
in terms of exponentials.
10. log(z). The derivative of log(z) can be found by differentiating the relation elog(z) = z
using the chain rule. Let w = log(z), so ew = z and
d w dz
e = = 1.
dz dz
dew dw
So, = 1.
dw dz
dw dw 1
So, ew =1 ⇒ = w
dz dz e
Using w = log(z) we get
d log(z) 1
= . QED
dz z
11. z a (any complex a). The derivative for this follows from the formula
z a = ea log(z)
dz a a az a
⇒ = ea log(z) · = = az a−1
dz z z
We often compose functions, i.e. f (g(z)). In general in this case we have the chain rule to
compute the derivative. However we need to specify the domain for z where the function is
analytic. And when branches and branch cuts are involved we need to take care.
2 2
Example 2.13. Let f (z) = ez . Since ez and z 2 are both entire functions, so is f (z) = ez .
The chain rule gives us
2
f 0 (z) = ez (2z).
Example 2.14. Let f (z) = ez and g(z) = 1/z. f (z) is entire and g(z) is analytic everywhere
but 0. So f (g(z)) is analytic except at 0 and
df (g(z)) −1
= f 0 (g(z))g 0 (z) = e1/z · 2 .
dz z
Example 2.15. Let h(z) = 1/(ez − 1). Clearly h is entire except where the denominator is
0. The denominator is 0 when ez − 1 = 0. That is, when z = 2πni for any integer n. Thus,
h(z) is analytic on the set
C − {2πni, where n is any integer}
2 ANALYTIC FUNCTIONS 17
The quotient rule gives h0 (z) = −ez /(ez − 1)2 . A little more formally: h(z) = f (g(z)).
where f (w) = 1/w and w = g(z) = ez − 1. We know that g(z) is entire and f (w) is analytic
everywhere except w = 0. Therefore, f (g(z)) is analytic everywhere except where g(z) = 0.
Example 2.16. It can happen that the derivative has a larger domain where it is analytic
than the original function. The main example is f (z) = log(z). This is analytic on C minus
d 1
a branch cut. However log(z) = is analytic on C − {0}. The converse can’t happen.
dz z
√
Example 2.17. Define a region where 1 − z is analytic.
√
answer: Choosing the principal branch of argument, we have w is analytic on
C − {x ≥ 1, y = 0}
√ √
Note. A different branch choice for z would lead to a different region where 1 − z is
analytic.
The figure below shows the domains with branch cuts for this example.
Im(w) Im(z)
Re(w) Re(z)
1
√ √
Left: domain for right: domain for 1 − z
w,
√
Example 2.18. Define a region where f (z) = 1 + ez is analytic.
√
answer: Again, let’s take w to be analytic on the region
C − {x ≤ 0, y = 0}
So, f (z) is analytic except where 1 + ez is real and ≤ 0. That is, except where ez is real
and ≤ −1. Now, ez = ex eiy is real only when y is a multiple of π. It is negative only when
y is an odd mutltiple of π. It has magnitude greater than 1 only when x > 0. Therefore
f (z) is analytic on the region
C − {x ≥ 0, y = odd multiple of π}
2 ANALYTIC FUNCTIONS 18
The figure below shows the domains with branch cuts for this example.
Im(w) Im(z)
3πi
πi
Re(w) Re(z)
−πi
−3πi
√ √
Left: domain for w, right: domain for ez + 1
The intuitive idea behind limits is relatively simple. Still, in the 19th century mathemati-
cians were troubled by the lack of rigor, so they set about putting limits and analysis on
a firm footing with careful definitions and proofs. In this appendix we give you the formal
definition and connect it to the intuitive idea. In 18.04 we will not need this level of for-
mality. Still, it’s nice to know the foundations are solid, and some students may find this
interesting.
lim zn = a.
n→∞
Again, the definition just says that eventually the sequence is within of a, no matter how
small you choose .
Example 2.19. Show that the sequence zn = (1/n + i)2 has limit -1.
answer: This is clear because 1/n → 0. For practice, let’s phrase it in terms of epsilons:
given > 0 we have to choose N such that
One strategy is to look at |zn + 1| and see what N should be. We have
2
1 1 2i 1 2
|zn − (−1)| = + i + 1 = 2 + < 2 +
n n n n n
Sometimes we need limits of the form lim f (z) = a. Again, the intuitive meaning is clear:
z→z0
as z gets close to z0 we should see f (z) get close to a. Here is the technical definition
Definition. Suppose f (z) is defined on a punctured disk 0 < |z − z0 | < r around z0 . We
say lim f (z) = a if for every > 0 there is a δ such that
z→z0
This says exactly that as z gets closer (within δ) to z0 we have f (z) is close (within ) to a.
Since can be made as small as we want, f (z) must go to a.
Notes. 1. Using the punctured disk (also called a deleted neighborhood) means that f (z)
does not have to be defined at z0 and, if it is then f (z0 ) does not necessarily equal a. If
f (z0 ) = a then we say the f is continuous at z0 .
2. Ask any mathematician to complete the phrase “For every ” and the odds are that they
will respond “there is a δ . . . ”
Here’s an equivalent way to define limits of functions: the limit lim f (z) = a if, for every
z→z0
sequence of points {zn } with limit z0 the sequence {f (zn )} has limit a.