NUS Complex Analysis 2.2.2
NUS Complex Analysis 2.2.2
H. Logemann
Contents
The subject matter is the theory of complex functions of one complex variable. We
begin with a study of (complex) differentiation - remarkably, differentiability throughout
a region ensures the existence of derivatives of all orders, and the validity of the Taylor
expansion.
A large part of the course is devoted to contour integration. Contour integrals are a
special kind of line integral, and are the subject of many unexpected theorems. The
main result in this context is the Cauchy integral theorem. Many seemingly intractable
real definite integrals can be transformed into complex contour integrals, that can be
evaluated using the residue theorem without doing any integration. Contour integration
also provides many insights into the behaviour of differentiable complex functions. One
such application we give is the fundamental theorem of algebra.
These methods have many applications beyond the scope of the course. Inversion of
Laplace and Fourier transforms, stability of feedback systems (Nyquist criterion) and the
study of the distribution of prime numbers, are just three.
Section 1 Preliminaries
Section 2 Complex differentiation
Section 3 Curves, paths and contour integrals
Section 4 The local Cauchy theorem
Section 5 Consequences of local Cauchy theory
Section 6 The global Cauchy theorem
Section 7 Isolated singularities and the residue theorem
Bibliography
Below there is a list of some of the text books on complex analysis which can be found
in the library. The books marked by a ∗ are more advanced than the other books, but
should be accessible to good undergraduate students.
L. Ahlfors. Complex Analysis
J. Bak & D.J. Newman. Complex Analysis
R.B. Burckel. An Introduction to Classical Complex Analysis, Vol. 1 ∗
H. Cartan. Elementary Theory of Analytic Functions of One or Several Variables ∗
W. Fischer & I. Lieb. A Course in Complex Analysis
R.E. Greene & S.G. Krantz. Function Theory of One Complex Variable ∗
J.M. Howie. Complex Analysis
K. Knopp. Theory of Functions, Part 1
S. Lang. Complex Analysis
N. Levinson & R.M. Redheffer. Complex Variables
J.E. Marsden & M.J. Hoffman. Basic Complex Analysis
R. Narasimhan & R. Nievergelt. Complex Analysis in One Variable ∗
T. Needham. Visual Complex Analysis
H.A. Priestley. Introduction to Complex Analysis
R. Remmert. Theory of Complex Functions ∗
W. Rudin. Real and Complex Analysis ∗
S. Maad Sasane & A. Sasane. A Friendly Approach to Complex Analysis
E. Wegert. Visual Complex Functions
1 Preliminaries
We collect a number of preliminary definitions and results which will be needed in the
course. A large part of this section is devoted to elementary topological concepts and
results.
1.1
Convergent sequences in the complex plane
We say that a sequence (zn ) in C converges to z ∈ C as n → ∞ and write limn→∞ zn = z
if, for every ε > 0, there exists N ∈ N such that
n≥N =⇒ |z − zn | ≤ ε .
The proof follows easily from the completeness of R and the observation that if (zn ) is a
Cauchy sequence in C, then (Re zn ) and (Im zn ) are Cauchy sequences in R.
A sequence (zn ) of complex numbers is called bounded if supn∈N |zn | < ∞.
Theorem 1.2 (Bolzano-Weierstrass theorem for the complex plane) Every bounded
sequence of complex numbers has a convergent subsequence.
Proof. Let (zn ) be a bounded sequence in C. Then there exists M > 0 such that
|zn | ≤ M for all n ∈ N. Hence |Re zn | ≤ |zn | ≤ M and |Im zn | ≤ |zn | ≤ M for all n ∈ N.
By the Bolzano-Weierstrass theorem for R, there exists a convergent subsequence (Re znj )
of (Re zn ) with limit x say. Another application of the Bolzano-Weierstrass theorem for R
shows that (Im znj ) has a subsequence (Im znjk ) converging to y say. It follows that znjk
converges to x + iy as k → ∞. ✷
D(z, r) := {w ∈ C : |z − w| < r}
is called the (open) disc with centre z and radius r. The set
1.2
Let {Sj } be a family of subsets of C, where J is an arbitrary index set. If Sj is open for
all j ∈ J, then ∪j∈J Sj is open; if Sj is closed for all j ∈ J, then ∩j∈J Sj is closed (see
Problem 2). Moreover, the intersection of finitely many open sets is open and the union
of finitely many closed sets is closed (see Problem 2)
Let T ⊂ S ⊂ C. The set T is said to be relatively open in S if T = U ∩ S for some open
set U ⊂ C. It is readily shown that T is relatively open in S if, and only if, for every
z ∈ T , there exists ε > 0 that S ∩ D(z, ε) ⊂ T . Note that, if T ⊂ S is an open set, then
it is relatively open in S. Note further that, if S is an open set, then T ⊂ S is relatively
open in S if, and only if, T is open.
A set T ⊂ S ⊂ C is said to be relatively closed in S if its complement S\T is relatively
open in S; equivalently, T is relatively closed in S if T = V ∩ S for some closed set V .
Let z ∈ C. A neighbourhood of z is any set containing D(z, ε) for some ε > 0; a punctured
neighbourhood of z is any set containing D× (z, ε) for some ε > 0. Obviously, a set S ⊂ C
is open if, and only if, S is a neighbourhood of z for every z ∈ S.
Theorem 1.3 A set S ⊂ C is closed if, and only if, for all convergent sequences (zn ) in
S, limn→∞ zn ∈ S.
Proof. We prove this theorem by contraposition, i.e., we show that a set S ⊂ C is not
closed if, and only if, there exists a convergent sequence (zn ) in S such that limn→∞ zn 6∈ S.
Assume first that S is not closed, that is, C\S is not open. Then there exists z ∈ C\S,
such that D(z, ε) ∩ S 6= ∅ for all ε > 0. For each n ∈ N, choose zn ∈ D(z, 1/n) ∩ S. Then
zn → z as n → ∞, and so (zn ) is a convergent sequence in S the limit of which is not in
S.
Conversely, assume that there exists a convergent sequence in (zn ) in S such that z :=
limn→∞ zn 6∈ S. For every ε > 0 there exists N ∈ N such that zn ∈ D(z, ε) for all n ≥ N .
But since z ∈ C\S and zn ∈ S for all n ∈ N, it follows that C\S is not open, and thus, S
is not closed. ✷
The closure of a set S ⊂ C, denoted by S̄, is the intersection of all closed sets in C
containing S. It is clear that S̄ is closed and that S = S̄ if, and only if, S is closed.
We can think of A(S) as the set of points which can be arbitrarily well approximated by
points in S.
Proof of Theorem 1.4. To show that S̄ ⊂ A(S), it is sufficient to show that A(S) is
closed. To this end, let (zn ) be a sequence in A(S) with limn→∞ zn = z. We show that
z ∈ A(S) (it follows then from Theorem 1.3 that A(S) is closed). For each n ∈ N, there
n
exists a sequence (wm ) in S such that
n
lim wm = zn .
m→∞
1.3
n
Hence, for each n ∈ N, there exists mn ∈ N such that |wm n
− zn | ≤ 1/n. Consequently,
n n
|wm n
− z| ≤ |wm n
− zn | + |zn − z| ≤ 1/n + |zn − z| ,
n
and therefore, since limn→∞ zn = z, we may conclude that wm n
→ z as n → ∞. Now
n
wmn ∈ S and so, z ∈ A(S).
Conversely, to prove that A(S) ⊂ S̄, let z ∈ A(S). Then there exists a sequence (zn ) in
S ⊂ S̄ such that zn → z as n → ∞. Since S̄ is closed, it follows from Theorem 1.3 that
z ∈ S̄. ✷
A point z ∈ C is called boundary point of a set S ⊂ C if, for each ε > 0, we have that
D(z, ε) ∩ S 6= ∅ and D(z, ε) ∩ (C\S) 6= ∅ .
Note that a boundary point of S may or may not belong to S. The set of all boundary
points of S is denoted by ∂S and is called the boundary of S. Clearly, if S is open, then
S ∩ ∂S = ∅ and if S is closed, then ∂S ⊂ S. Moreover, it can be shown that S̄ = S ∪ ∂S
(see Problem Sheet 1).
A set S ⊂ C is called bounded if supz∈S |z| < ∞. A compact subset of C is a set which is
bounded and closed.
Example. Let z ∈ C and r > 0. Clearly, the disc D(z, r) = {w ∈ C : |z − w| < r} and
the punctured disc D× (z, r) = {w ∈ C : |z − w| < r}\{z} are open sets with
D̄(z, r) = D× (z, r) = {w ∈ C : |z − w| ≤ r} .
Obviously, D(z, r) and D× (z, r) are not closed, whilst D̄(z, r) is closed. Moreover,
∂D(z, r) = ∂ D̄(z, r) = {w ∈ C : |z − w| = r} , ∂D× (z, r) = {w ∈ C : |z − w| = r} ∪ {z} .
The sets D(z, r), D̄(z, r) and D× (z, r) are bounded, D̄(z, r) is compact, whilst D(z, r) and
D× (z, r) are not compact.
Let w, z ∈ C with w 6= z. The line segment
L = {(1 − t)w + tz : 0 ≤ t < 1}
is neither open nor closed, it is bounded, but it is not compact. The boundary of L is
given by ∂L = L ∪ {z} = L̄. ✸
Corollary 1.5 A set S ⊂ C is compact if, and only if, every sequence in S has a con-
vergent subsequence with limit in S.
1.4
Lemma 1.6 (Decreasing sets lemma) If Kn , n ∈ N, are non-empty compact sets in
C such that K1 ⊃ K2 ⊃ K3 ⊃ . . ., then ∩∞
n=1 Kn 6= ∅.
Example. Let x, y ∈ C. We show that the line segment Λ := [x, y] is connected. To this
end, let D1 , D2 ⊂ C be open sets such that D1 ∩ D2 ∩ Λ = ∅ and Λ ⊂ D1 ∪ D2 . It is clear
that there exists i ∈ {1, 2} such that x ∈ Di , and so, Di ∩ Λ 6= ∅. Let k ∈ {1, 2} be such
that k 6= i. To establish connectedness of Λ, we need to show that Dk ∩ Λ = ∅. Seeking
a contradiction, suppose that Dk ∩ Λ 6= ∅. Then there exists z ∈ Dk ∩ Λ. It is clear that
the line segment [x, z] is contained in Λ. Since z ∈ Dk ∩ Λ, it follows that z 6∈ Di ∩ Λ, and
thus, the set T := {t ∈ [0, 1] : (1 − t)x + tz 6∈ Di ∩ Λ} is non-empty. Moreover, x ∈ Di ∩ Λ
and so, by the openness of Di , we have that τ := inf T > 0. Setting ξt := (1 − t)x + tz for
all t ∈ [0, 1], it follows from the definition of τ and openness of Di that ξt ∈ Di ∩ Λ for
all t ∈ [0, τ ) and ξτ 6∈ Di ∩ Λ. Therefore, ξτ ∈ Dk ∩ Λ and, by the openness of Dk , there
exists 0 ≤ σ < τ such that ξt ∈ Dk ∩ Λ for all t ∈ [σ, τ ]. Consequently, ξt ∈ Di ∩ Dk ∩ Λ
for all t ∈ [σ, τ ) which is impossible since D1 ∩ D2 ∩ Λ = ∅. ✸
Let S ⊂ C and s1 , s2 ∈ S. We say that s1 is connected to s2 (in S), and write s1 ∼ s2 ,
if there exists a connected set T ⊂ S such that s1 , s2 ∈ T . We claim that ∼ defines an
equivalence relation on S: indeed, it is obvious that ∼ is reflexive and symmetric, and
the transitivity property is an immediate consequence of the following lemma.
1.5
Lemma 1.7 Let J be an arbitrary index set and Sj ⊂ C for all j ∈ J. If, for every
j ∈ J, the set Sj is connected and ∩j∈J Sj 6= ∅, then the set ∪j∈J Sj is connected.
Dk ∩ S = ∪j∈J (Dk ∩ Sj ) = ∅,
Corollary 1.8 Let S ⊂ C be non-empty and let s ∈ S. The equivalence class [s] is
the largest connected subset of S containing s, that is, [s] is connected and if R ⊂ S is
connected and s ∈ R, then R ⊂ [s].
Proof. For every t ∈ [s], there exists a connected set St ⊂ S such that s, t ∈ St .
Obviously, ∩t∈[s] St 6= ∅, and so, invoking Lemma 1.7, we have that T := ∪t∈[s] St ⊂ S is
connected. Therefore, to prove that [s] is connected, it is sufficient to show that [s] = T .
If z ∈ T , then there exists t ∈ [s] such that z ∈ St , and so z ∼ s, showing that z ∈ [s].
Consequently, T ⊂ [s]. It is trivial that [s] ⊂ T , whence [s] = T . Finally, let R ⊂ S
be connected and such that s ∈ R. For every z ∈ R we have that z ∼ s, implying that
z ∈ [s] and thus R ⊂ [s], completing the proof. ✷
Note that if S is a domain, then, for every s ∈ S, the equivalence class [s] is also a domain.
A subset T ⊂ S is said to be a connected component of S if T is an equivalence class of
the relation ∼ . A non-empty set S ⊂ C is connected if, and only if, S has exactly one
connected component, namely S. Furthermore, a non-empty set S ⊂ C is disconnected if,
and only if, it has more than one connected component. Every non-empty subset S ⊂ C
is the union of its connected components and any two different connected components are
disjoint.
Recall that a non-empty set C ⊂ C is said to be convex if, for all w, z ∈ C, it follows that
[w, z] ⊂ C.
Example. If C ⊂ C is convex, then C is connected. Indeed, for arbitrary c1 , c2 ∈ C, the
connected set [c1 , c2 ] is contained in C, and so c1 ∼ c2 , implying that C is connected. ✸.
Example. Let z1 , z2 ∈ C and r1 , r2 > 0. Define Sj := D(zj , rj ), j = 1, 2. Clearly, S1
and S2 are convex and hence connected. Set S3 = S1 ∪ S2 . Suppose that z1 6= z2 and set
ρ := |z2 − z1 | > 0. If r1 , r2 > ρ/2, then S1 ∩ S2 6= ∅ and S3 is connected (by Lemma 1.7).
1.6
If r1 , r2 < ρ/2, then S1 ∩ S2 = ∅ and S3 is not connected. In this case, S3 has precisely
two connected components, namely S1 and S2 . ✸
A polygonal contour P is a set in C of the form
n−1
P = ∪j=1 [zj , zj+1 ] , (∗)
Theorem 1.9 A domain D ⊂ C is connected if, and only if, for all w, z ∈ D, there exists
a polygonal contour in D connecting w and z.
Continuous functions
Consider a function f : S → C, where S ⊂ C is non-empty. Let z ∈ S̄. We say that
f (w) converges to a complex number l as w → z if, for all ε > 0, there exists δ > 0 such
that |f (w) − l| ≤ ε for all w in S with |w − z| ≤ δ. We write f (w) → l as w → z or
limw→z f (w) = l. A routine exercise shows that limw→z f (w) = l if, and only if, for every
sequence (wn ) in S with limn→∞ wn = z we have that limn→∞ f (wn ) = l. It is clear that
f (w) → l as w → z if, and only if, Re f (w) → Re l and Im f (w) → Im l as w → z. If
‡
See the next subsection for the definition of continuity for a complex-valued function.
1.7
S is unbounded, then we say that f (w) converges to a number l ∈ C as |w| → ∞ if, for
all ε > 0, there exists η > 0 such that |f (w) − l| ≤ ε for all w in S with |w| ≥ η. We
write f (w) → l as |w| → ∞ or lim|w|→∞ f (w) = l. The algebra of complex limits (sums,
products, etc.) can be developed as in the real case.
We say that f is continuous at z ∈ S if limw→z f (w) = f (z). Equivalently, f is continuous
at z ∈ S if, and only if, for every sequence (wn ) in S with limn→∞ wn = z, we have that
limn→∞ f (wn ) = f (z). We mention that f is continuous at z if, and only if, Re f and Im f
are continuous at z. If f and g are continuous at z, then f + g and f g are continuous at z.
Let T ⊂ C. As in the real case it can be easily proved that if g : T → C is a function with
g(T ) ⊂ S, g is continuous at z ∈ T and f is continuous at g(z), then f ◦ g is continuous
at z.
If f is continuous at z for every z ∈ S, then we call f continuous (on S). The set of all
continuous functions f : S → C is denoted by C(S). If f, g ∈ C(S), then f + g and f g
are in C(S). Furthermore, if f ∈ C(S), g ∈ C(T ), where T ⊂ C, and g(T ) ⊂ S, then
f ◦ g ∈ C(T ).
Example. The functions z 7→ z̄ and z 7→ |z| are continuous on C. The argument
function z 7→ Arg(z) is defined for every z in the punctured plane C× := C\{0}. It is
continuous at every point z in the slit plane C− := C\(−∞, 0] and it is discontinuous at
every z ∈ (−∞, 0). ✸
For a function f : S → C and subsets T ⊂ S and U ⊂ C we define the image of T under
f and the pre-image of U under f by
respectively. The following result links continuity to the topological concepts of openness
and connectedness.
1.8
showing that f is continuous at z. Since z was an arbitrary element in S, it follows that
f is continuous.
(ii) Let U1 , U2 ⊂ C be open and such that
f (S) ∩ U1 ∩ U2 = ∅, f (S) ⊂ U1 ∪ U2 .
S1 ∩ S2 = ∅, S = S1 ∪ S2 .
Since f is piecewise continuous, Re f and Im f are piecewise continuous, and so the two
integrals on the right-hand side of the above equation do exist.
Lemma 1.12 (Triangle inequality for integrals) Let a, b ∈ R with a < b and let
f : [a, b] → C be piecewise continuous. Then
Z b Z b
f (t)dt ≤ |f (t)|dt .
a a
1.9
Rb
Proof. Set λ := a f (t)dt. If λ = 0, there is nothing to show; so assume that λ 6= 0.
Rb
Setting µ := λ̄/|λ|, we have |µ| = 1 and |λ| = µλ. Then a µf (t)dt = µλ = |λ| is real and
hence
Z b Z b Z b Z b
|λ| = Re(µf (t))dt ≤ |µf (t)|dt = |µ| |f (t)|dt = |f (t)|dt .
a a a a
✷
Let S ⊂ C be non-empty and let f , fn (n ∈ N) be complex-valued functions defined on
S. We say that fn converges uniformly to f (on S) as n → ∞ if, for all ε > 0, there exists
N ∈ N such that |fn (z) − f (z)| ≤ ε for all n ≥ N and all z ∈ S. Note that N does not
depend on z – the definition requires the existence of a N which “works” for all z ∈ S.
The following result is known from real analysis. The extension to the complex case is
straightforward, and, therefore, the proof is omitted.
The next result shows that if a sequence of continuous complex-valued functions defined
on a real interval converges uniformly, then integral and limit can be interchanged.
Rb
Proof. Note that by Theorem 1.13, f is continuous and so the integral a f (t)dt is well
defined. Using Lemma 1.12 we obtain
Z b Z b Z b
f n (t)dt − f (t)dt ≤
|fn (t) − f (t)|dt ≤ (b − a) sup |fn (t) − f (t)| .
a a a t∈[a,b]
Rb Rb
By hypothesis, supt∈[a,b] |fn (t) − f (t)| → 0 as n → ∞, and so, a
fn (t)dt → a
f (t)dt as
n → ∞. ✷
Let J ⊂ R be an open interval, let t0 ∈ J and consider a function f : J → C. We say
that f is differentiable at t0 if the limit
f (t) − f (t0 )
lim =: f ′ (t0 )
t→t0 t − t0
exists. Clearly, if f is differentiable at t0 , then f is continuous at t0 . The function f
is differentiable at t0 if, and only if, Re f and Im f are differentiable at t0 and we have
f ′ (t0 ) = (Re f )′ (t0 ) + i(Im f )′ (t0 ). The fundamental theorem of calculus for real-valued
functions extends to complex-valued functions in a straightforward way.
1.10
Power series
The material in this subsection is very similar to the real case considered in first and
second year analysis, and therefore the proofs are omitted.
Let S ⊂ C be non-empty and Plet fn , where n ∈ N0 , be complex-valued functions defined
∞
on S. We
Pm say that the series n=0 fn converges uniformly (on S) if the sequence of partial
sums ( n=0 fn ) converges uniformly (on S) as m → ∞.
Let an ∈ C for n ∈ N0 and let z0 ∈ C. A complex power series (at z0 ) is a series of the form
P ∞ n
n=0 an (z − z0 ) , where z is a complex variable. We say that the power series converges
for z = w, if the sequence of partial sums ( m n
P
n=0 n (w − z0 ) ) converges as m → ∞. As
a
usual, if the power series does not converge for z = w, then we say it diverges for z = w.
The
Pmpower series isncalled absolutely convergent for z = w if the sequence of partial sums
( n=0 |an (w − z0 )| ) converges as m → ∞. We recall that absolute convergence implies
convergence, but the converse is not true.
Theorem 1.16 If ∞ n
P
n=0 an (z − z0 ) is a complex power series, then there
P∞exists 0 ≤ ρ ≤n
∞ (called its radius of convergence) such that Pif |z − z0 | < ρ, then n=0 an (z − z0 )
∞ n
converges absolutely, and if |z − z0 | > ρ, then n=0 an (z − z0 ) diverges.
1.11
2 Complex differentiability
Throughout this section, let D ⊂ C be a domain. The following definition is fundamental.
Definition. A function f : D → C is said to be complex differentiable (or C-differentiable)
at z0 ∈ D if the limit
df f (z) − f (z0 )
f ′ (z0 ) = (z0 ) = lim
dz z→z 0 z − z0
exists. ✸
If f is C-differentiable at z0 , then f is continuous at z0 . We note that f : D → C is C-
differentiable at z0 ∈ D with f ′ (z0 ) = a if, and only if, there exists a function g : D → C
such that
f (z) = f (z0 ) + a(z − z0 ) + |z − z0 |g(z) and lim g(z) = 0,
z→z0
2.1
Example. The complex exponential function exp : C → C is defined by
Obviously, for real arguments, the complex exponential function coincides with the real
exponential function. It can be shown that exp(z1 +z2 ) = exp(z1 ) exp(z2 ) for all z1 , z2 ∈ C
(this is the result of a routine calculation). Furthermore, exp is C-differentiable at every
point in C, and
d
exp′ (z) = ez = exp z = ez ∀ z ∈ C,
dz
−
see Problem 7. Set C := C\(−∞, 0] (the slit plane) and consider the function
where Arg is the principal value of the argument introduced in Section 1. It is straight-
forward to show (see Problem 10) that Log is the inverse function of exp |H , the complex
exponential function restricted to the horizontal strip H := {z ∈ C : −π < Im z < π}.
We claim that Log is C-differentiable at every point in the slit plane and Log ′ (z) = 1/z
for every z ∈ C− . To show this, we note that Log is continuous and exp′ (z) = exp(z) 6= 0
for every z ∈ C which, together with Lemma 2.1, yields that Log is C differentiable at
exp(s) for every s ∈ H and
1 1
Log ′ (exp s) = ′
= ∀ s ∈ H.
exp (s) exp s
Now exp maps H onto C− and thus,
1
Log ′ (z) = ∀ z ∈ C− .
z
Warning. If z1 , z2 ∈ C− are such that z1 z2 ∈ C− , then Log z1 , Log z2 and Log (z1 z2 ) are
well defined, but, in general, Log (z1 z2 ) 6= Log z1 + Log z2 . ✸
Obviously, a function f : D → C can be considered as an R2 -valued function of two
real variables. If we take this point of view, we indicate this by writing fR , that is,
fR : D ⊂ R2 → R2 with fR (x, y) := (Re f (x + iy), Im f (x + iy)) for all x + iy = (x, y) ∈ D.
We say that f is real differentiable at a point z0 = x0 + iy0 ∈ D, where x0 , y0 ∈ R, if fR
is differentiable at (x0 , y0 ) in the sense of real analysis. We recall that fR is said to be
differentiable at (x0 , y0 ) with derivative fR′ (x0 , y0 ), a linear map from R2 to R2 , if there
exists a function g : D → R2 such that g(x, y) → (0, 0) as (x, y) → (x0 , y0 ) and
′ x − x0
x − x0
fR (x, y) = fR (x0 , y0 ) + fR (x0 , y0 ) +
g(x, y).
y − y0 y − y0
As usual, we identify the linear map fR′ (x0 , y0 ) with its matrix representation with respect
to the canonical basis of R2 . If fR is differentiable at (x0 , y0 ), then the partial derivatives
of the component functions of fR exist at (x0 , y0 ) and fR′ (x0 , y0 ) is equal to the Jacobian
matrix.
The next theorem explains how complex differentiability is related to real differentiability.
2.2
Theorem 2.2 Let z0 = x0 + iy0 ∈ D (where x0 , y0 ∈ R). The function f : D → C is
complex differentiable at z0 with f ′ (z0 ) = b + ic (where b, c ∈ R) if, and only if, f is real
differentiable at (x0 , y0 ) with Jacobian matrix J(x0 , y0 ) at (x0 , y0 ) given by
b −c
J(x0 , y0 ) = .
c b
The above theorem shows that complex differentiability is a stronger property than real
differentiability: real differentiability at (x0 , y0 ) alone is not sufficient to guarantee com-
plex differentiability at x0 + iy0 . Indeed, if f is real differentiable at (x0 , y0 ), but the
Jacobian matrix does not have the above structure, then f is not complex differentiable
at x0 + iy0 .
√
Expressing (b, c) in polar coordinates, that is, (b, c) = r(cos θ, sin θ), where r = b2 + c2 ,
the matrix J(x0 , y0 ) can be written in the form
cos θ − sin θ
J(x0 , y0 ) = r ,
sin θ cos θ
∂u ∂v ∂u ∂v
(x0 , y0 ) = (x0 , y0 ), (x0 , y0 ) = − (x0 , y0 ).
∂x ∂y ∂y ∂x
These partial differential equations are called the Cauchy-Riemann equations. Note that
if f is C-differentiable at z0 = x0 + iy0 , then the derivative f ′ (z0 ) is equal to
∂u ∂v
f ′ (z0 ) = (x0 , y0 ) + i (x0 , y0 ),
∂x ∂x
or, equivalently, by the Cauchy-Riemann equations,
∂v ∂u
f ′ (z0 ) = (x0 , y0 ) − i (x0 , y0 ).
∂y ∂y
Proof of Theorem 2.2. Assume that f is complex differentiable at z0 with f ′ (z0 ) = b+ic.
Then there exists a function g : D → C such that
2.3
Therefore, the above first-order Taylor formula for f (z) can be written in real terms as
follows:
b −c x − x0
x − x0
fR (x, y) = fR (x0 , y0 ) +
y − y0
gR (x, y),
+
c b y − y0
This is a first-order Taylor formula for fR at (x0 , y0 ), and, consequently, f is real differ-
entiable at (x0 , y0 ) with Jacobian matrix J(x0 , y0 ) given by
b −c
J(x0 , y0 ) = . (‡)
c b
Conversely, assume that f is real differentiable at (x0 , y0 ) with Jacobian matrix J(x0 , y0 )
given by (‡). Then (†) holds, which, by (∗∗), can be re-written as (∗). This is a first order
Taylor formula for f (z), and so, f is complex differentiable at z0 with complex derivative
f ′ (z0 ) = b + ic. ✷
Example. We claim that the function f : C → C, z 7→ z̄ is nowhere C-differentiable.
To this end, write z = x + iy and f (z) = f (x + iy) = u(x, y) + iv(x, y), where x, y ∈ R,
u(x, y) = x, and v(x, y) = −y. Noting that, for all (x, y),
∂u ∂v
(x, y) = 1 6= −1 = (x, y),
∂x ∂y
we see that the first of the Cauchy-Riemann equations does not hold, and thus, by The-
orem 2.2, f is nowhere C-differentiable. ✸
Definition. Let D ⊂ C be a domain, let z0 ∈ D and let f : D → C be a function. We
say that f is holomorphic (or analytic) if f is C-differentiable at all points in D. The
function f is said to be holomorphic (analytic) at z0 if there exists ε > 0 such that f is
C-differentiable at all points in D(z0 , ε) (or, equivalently, if f |D(z0 ,ε) is holomorphic). ✸
The set of all holomorphic functions f : D → C is denoted by H(D). If f, g ∈ H(D),
then f + g and f g are in H(D) (it follows that H(D) is a ring). Polynomials and the
complex exponential functions are holomorphic on C and Log is holomorphic
P∞ on the slit
−
plane C . In Section 4 we will see that any power series P (z) = n=0 an (z − z0 )n with
radius of convergence ρ > 0 defines a holomorphic function on D(z0 , ρ).
Recall that a region in C is a non-empty connected open set.
Proof. Let w, z ∈ R be fixed, but arbitrary. Since R is open and connected, it follows
n−1
from Theorem 1.9 that there exists a polygon P = ∪k=1 [sk , sk+1 ] in R with w = s1 and
z = sn . For every k = 1, . . . , n − 1 we have
d
f ((1 − t)sk + tsk+1 ) = f ′ ((1 − t)sk + tsk+1 ))(sk+1 − sk ) = 0 ∀ t ∈ [0, 1].
dt
2.4
Hence, f (sk ) = f (sk+1 ) for all k = 1, . . . , n − 1, implying that f (w) = f (z). Since w and
z were arbitrary elements in R, we conclude that f is constant. ✷
We now describe a process which manufactures holomorphic functions of a certain struc-
ture. Special cases will be of importance later.
The above theorem says in particular that Fn′ is obtained by differentiating under the
(n)
integral, F1 is arbitrarily often C-differentiable and F1 = n!Fn+1 .
Proof of Theorem 2.4. It is useful to indicate the dependence of Fn on g by writing
Fng := Fn . We proceed in three steps.
Step 1. We claim that F1g is continuous. Let z0 ∈ D and choose δ > 0 such that
D(z0 , 2δ) ⊂ D. Then
and, since
b
g(t)dt
Z
F1g (z) − F1g (z0 ) = (z − z0 ) ∀ z ∈ D, (∗)
a (h(t) − z)(h(t) − z0 )
we obtain
b
|z − z0 |
Z
|F1g (z) − F1g (z0 )| ≤ |g(t)|dt ∀ z ∈ D(z0 , δ).
δ2 a
This estimate shows that F1g is continuous at z0 and since z0 ∈ D was arbitrary, we
conclude that F1g is continuous.
Step 2. We prove that F1g ∈ H(D) and dF1g /dz = F2g . To this end, let z0 ∈ D and define
a continuous function ϕ : [a, b] → C by
g(t)
ϕ(t) := ∀ t ∈ [a, b].
h(t) − z0
Appealing to (∗), we obtain
2.5
showing that dF1g /dz = F2g .
Step 3. We complete the proof by an induction argument. Let m ∈ N and assume that,
g
for every continuous g : [a, b] → C, the function Fmg is holomorphic and dFmg /dz = mFm+1 .
g
Let z0 ∈ D and define ϕ as in Step 2. Since Fmϕ (z0 ) = Fm+1 (z0 ), we have, for all z ∈ D,
b
g(t)(h(t) − z) + (z − z0 )g(t)
Z
ϕ g
Fmϕ (z) − Fmϕ (z0 ) + (z − z0 )Fm+1 (z) = dt − Fm+1 (z0 )
a (h(t) − z)m+1 (h(t) − z0 )
b
g(t)dt
Z
g
= − Fm+1 (z0 ),
a (h(t) − z) m+1
and thus,
g g ϕ
Fm+1 (z) − Fm+1 (z0 ) = Fmϕ (z) − Fmϕ (z0 ) + (z − z0 )Fm+1 (z) ∀ z ∈ D. (∗∗)
ϕ
By induction hypothesis, Fmϕ is holomorphic and hence continuous. Moreover, Fm+1 is
bounded in a neighbourhood of z0 . It now follows that the right-hand side of (∗∗) con-
g
verges to 0 as z → z0 , showing that Fm+1 is continuous at z0 . Since z0 ∈ D was arbitrary,
g
we see that Fm+1 is continuous. In this argument, g is an arbitrary continuous function
ϕ
[a, b] → C, and consequently, Fm+1 is also continuous. Dividing (∗∗) by z − z0 and letting
z → z0 gives
g g
Fm+1 (z) − Fm+1 (z0 ) dFmϕ ϕ ϕ
lim = (z0 ) + Fm+1 (z0 ) = (m + 1)Fm+1 (z0 ),
z→z0 z − z0 dz
ϕ
where we have made use of the induction hypothesis and the continuity of Fm+1 . Finally,
ϕ g
by the definition of ϕ, Fm+1 (z0 ) = Fm+2 (z0 ), and so
d g g
F (z0 ) = (m + 1)Fm+2 (z0 ),
dz m+1
completing the induction argument and the proof. ✷
2.6