Complex Differentiation: Umberto L. Hryniewicz April 4, 2023
Complex Differentiation: Umberto L. Hryniewicz April 4, 2023
Umberto L. Hryniewicz
April 4, 2023
p
The space of complex numbers C with the absolute value |x+iy| = x2 + y 2
is a normed real vector space. Note that {1, i} is a basis of C seen as a real
vector space.
Let U ⊂ C be an open set and f : U → C be a function. We say that f has
a complex derivative at z0 ∈ U , or that f is complex differentiable at z0 , if the
limit
f (z) − f (z0 )
f 0 (z0 ) = lim (1)
z→z0 z − z0
exists in (C, | · |). This means that
f (z) − f (z0 ) 0
∀ > 0 ∃δ > 0 0 < |z − z0 | < δ ⇒
− f (z0 ) <
z − z0
or equivalently
Df (z0 ) : h 7→ f 0 (z0 )h
where f 0 (z0 )h stands for complex multiplication of the complex numbers f 0 (z0 )
and h. Hence, if f 0 (z0 ) = a + ib then the linear map Df (z0 ) is represented with
respect to the basis {1, i} as the matrix
a −b
Df (z0 ) = (2)
b a
∂x f1 − ∂y f2 = 0
(3)
∂y f1 + ∂x f2 = 0
1
With ∂x f = ∂x f1 + i∂x f2 and ∂y f = ∂y f1 + i∂y f2 this is the same as
∂x f + i∂y f = 0
Hence
z n − z0n
f 0 (z0 ) = lim = nz0n−1
z→z0 z − z0
as expected.
As noted above, if a function has a complex derivative at a given point
then the differential at this point has the form (2), in other words the Cauchy-
Riemann equations are satisfied. This makes it very simple to give examples of
functions that do not have a complex derivative. A basic example is the complex
conjugate function
f (z) = z̄
It satisfies
1 0
Df (z) =
0 −1
which violates the Cauchy-Riemann equations. Another source of examples are
real-valued functions f : U → R ⊂ C. For these we have
∂x f ∂y f
Df =
0 0
and hence can only satisfy the Cauchy-Riemann equations at points where ∇f
vanishes.
2
To get more interesting holomorphic functions we need to consider complex
power series. These are sums of the form
∞
X
an (z − z0 )n {an }n≥0 ⊂ C, z0 ∈ C (4)
n=0
The point z0 is called the center of the series. The radius of convergence is
defined as
1
R= p (5)
lim supn→∞ n |an |
with the following convention: R = 0 if the lim sup is +∞, and R = +∞ if
the lim sup is 0. The reason for this terminology is that the series (4) converges
locally uniformly on BR (z0 ). To prove this, let us consider the case where
0 < R < +∞; the case R = +∞ is left to the reader. Fix any t ∈ [0, 1) and let
> 0 satisfy (1 + )t < 1. From (5) we find N ∈ N such that
p
n ≥ N ⇒ R n |an | ≤ 1 +
Thus we can estimate
p n
z ∈ BtR (z0 ), n ≥ N ⇒ |an (z − z0 )n | ≤ tR n |an | ≤ ((1 + )t)n
Since (1 + )t < 1 and a geometric series is absolutely convergent, we can use
the above estimate and Weierstrass’ Test to conclude that (4) is absolutely con-
vergent in BtR (z0 ), for every t ∈ [0, 1). This shows that (4) defines a continuous
function on BR (z0 ).
Now note that the general term an (z − z0 )n of (4) is holomorphic since it is
a polynomial, and that the radius of convergence of the series
∞
X ∞
X
nan (z − z0 )n−1 = (n + 1)an+1 (z − z0 )n (6)
n=1 n=0
obtained by differentiating (4) term by term is the same as the one for the
original series since
p p
lim sup n |an | = lim sup n (n + 1)|an+1 | (7)
n→+∞ n→+∞
References
[1] Thema1-2.pdf, available in Moodle.