23 Cauchy Integral Formula II
Cauchy Integral Formula for general curves We have studied Cauchy’s Integral Formula on a circle:
f(z) = 1 2πi
I
∂∆(a;r)
f(ζ
ζ− zdζ, ∀z ∈ ∆(a; r).
One natural question is: instead of a circle ∂∆(a; r), if we consider any closed curve C inside D, can we have
f(z) = 1 2πi
I
C
f(ζ) ζ− zdζ?
where C is defined by a parametric function φ : [0, 1] → C.
This may not be true. See examples below.
[Example]
❅❅
■
(a) C is given by φ : [0, 2π], θ 7→ φ(θ) = a + eiθ,
1 2πi
H
C dζ
ζ−a = 2πi1 R2π 0
deiθ
eiθ = 2πi1 R2π 0
ieiθdθ
eiθ = +1.
❅❅❘ (b) C is given by φ : [0, 2π], θ 7→ φ(θ) = a + e−iθ,
1 2πi
H
C dζ
ζ−a = 2πi1 R2π 0
de−iθ
e−iθ = 2πi1 R2π
0 −ie−iθdθ
e−iθ = −1.
❅❅
■❅■❅
(c) C is given by φ : [0, 2π], θ 7→ φ(θ) = a + e2iθ,
1 2πi
H
C dζ
ζ−a = 2πi1 R2π 0
de2iθ
e2iθ = 2πi1 R2π 0
2ie2iθdθ
e2iθ = +2.
❅❅❘
❅❅❘ (d) C is given by φ : [0, 2π], θ 7→ φ(θ) = a + e−2iθ,
1 2πi
H
C dζ
ζ−a = 2πi1 R2π 0
de−2iθ
e−2iθ = −2.
Therefore, in general, we don’t have a formula such as f(z) = 1
2πi I
C
f(ζ) ζ− zdζ for any closed curve C.
Rotation number Think about f (z) = 1, from above, f (z) = 2πi1 H
C f(ζ)
ζ−zdζ is always true, which depends on the parametric function of the curve.
Let a ∈ C and C be a closed curve that does not pass through a. The rotation number of C around a is defined by
n(a; C) := 1 2πi
I
C
dζ ζ− a which is an integer.
Remarks
1. In the above examples (a)(b)(c)(d), the rotation numbers of C are n(0; C) = 1, −1, 2, −2.
2. If C is a Jordan curve (no self-intersection point), it excludes the example (c) and (d) so that n(a; C) = ±1. If, in addition, C has counterclockwise orientation, then n(a; C) = 1; otherwise, n(a; C) = −1.
3. If a is outside of the closed curve C, then the rotation number n(0; C) = 0 by Cauchy’s Integral Theorem.
❅❅❘
a
s
n(a; C)=2πi1 H
C dζ ζ−a = 0
4. On December 18, 1811, Gauss wrote to Bessel,
“If we define log x viaR 1
xdxstarting at x = 1, then arrive at logx having gone around the point x = 0 one or more times or not at all, every circuit adding the constant +2πi or −2πi; thus the fact that every number has multiple logarithms becomes quite clear.” 71
General Cauchy Integral Formula
Theorem 23.1 (General Cauchy Integral Theorem) Let f be a holomorphic function in a domain D ⊂ C. Let C ⊂ D be a closed curve that is homotopic to a point. Let z ∈ D − C.
Then
n(z; C)f (z) = 1 2πi
I
C
f(ζ) ζ− zdζ.
In particular, if, in addition, C is a Jordan curve with counter-clockwise orientation, then f(z) = 1
2πi I
C
f(ζ) ζ− zdζ.
Proof: Fixing any z ∈ D − C. Let g(ζ) := f(ζ)
ζ− z − f(z)
ζ− z, ∀ζ ∈ D − {z}.
We claim: g(ζ) is holomorphic on D.
If ζ ∈ D with ζ 6= z, g(ζ) is holomorphic at ζ.
To show that g is holomorphic at z.
71R. Remmert, Theory of Complex Functions, Springer, 1991, p.168.
Let us consider a disk ∆(z; r) ⊂ D and
g(ζ) = f(ζ)
ζ− z − f(z) ζ− z =
P∞
n=0an(ζ − z)n− a0
ζ− z =
P∞
n=1an(ζ − z)n
ζ− a =
∞
X
n=1
an(ζ − z)n−1,
which is holomorphic at z. Claim is proved. Then applying Cauchy Integral Theorem, H
Cg(ζ)dζ = 0, i.e., I
C
f(ζ) ζ− zdζ −
I
C
f(z) ζ− zdζ =
I
C
f(ζ)
ζ− zdζ − n(z; C)f(z) = 0. The mean value equality and inequality
Theorem 23.2 Let f be a holomorphic function in a domain D ⊂ C with a disk ∆(a, r) ⊂⊂
D. Then
f(a) = 1 2π
Z 2π 0
f(a + reiθ)dθ.
Consequently,
|f(a)| ≤ sup
∂∆(a,r)|f|.
Proof: By Cauchy integral formula, f(a) = 1
2πi I
∂∆(a,r)
f(ζ)
ζ− adζ = 1 2πi
Z 2π 0
f(a + reiθ)
(a + reiθ) − aireiθdθ = 1 2π
Z 2π 0
f(a + reiθ)dθ.
Morera’s theorem A complex-valued function f is holomorphic if and only if f is analytic; if and only if f is C1-smooth satisfying the Cauchy-Riemann equation. Now we have one more equivalent definition.
Theorem 23.3 (Morera’s theorem) Let f be a continuous complex-valued function on a domain D ⊂ C. Then
f is holomorphic ⇐⇒
I
∂∆
f(z)dz = 0, for any triangle ∆ ⋐ D.
Proof: (=⇒) By Cauchy-Goursat Theorem.
(⇐=) ∀a ∈ D, to show: f is a holomorphic function in ∆(a; r0) ⋐ D.
For any point z ∈ ∆(a; r0), we denote by Cz the line segment from a to z.
Cz
s
s
a
✒ z
❅❅
Γ ■ s
z+ h Let F (z) :=R
Czf(ζ)dζ.
F(z+h)−F (z)
h =
R
Cz+hf −R
Czf h
=
R
Γf
h → f(z) as h → 0.
Since F′ = f , f is holomorphic in ∆(a; r0).
[Example] Let T be a triangle inside a domain D ⊂ C and w be a point inside this triangle. Let f be a function defined in D and be holomorphic in D − {w}. If f is bounded near w, prove that
I
T
f(z)dz = 0.
Proof: Consider a disk ∆(w, ǫ0) inside the triangle T . By connecting T with ∂∆(w, ǫ) with a line segment L, we consider a closed curve Tǫ := T + L − ∂∆(w, ǫ) − L. Then by applying Cauchy Integral Theorem, H
Tǫf(z)dz = 0, i.e., I
T
f(z)dz = I
∂∆(w,ǫ)
f(z)dz = lim
ǫ→0
I
∂∆(w,ǫ)
f(z)dz.
Since f is bounded near w, we have |f(z)| ≤ M for any z ∈ ∆(w, ǫ) for any 0 < ǫ ≤ ǫ0 for some ǫ0. Then
| I
∂∆(w,ǫ)
f(z)dz| ≤ M I
∂∆(w,ǫ)|dz| = 2πǫM → 0 as ǫ → 0. Therefore, we obtain
I
T
f(z)dz = lim
ǫ→0
I
∂∆(w,ǫ)
f(z)dz = 0.
In fact by the above argument, we nearly prove the following result.
Appendix: The Fresnel Integral The integrals RR
0 cost t2dtand RR
0 sin t2dtare called Fresnel integrals after Augustin-Jean Fresnel that are used in optics, which are closely related to the error function 72
Theorem 23.4 For any a ∈ R with |a| ≤ 1, Z ∞
0
e−(1+ia)2t2dt= 1 2· 1
1 + ia
√π.
Splitting into the real part and the imaginary part, it gives Z ∞
0
e(a2−1)t2cos(2at2)dt = 1 2(1 + a2)
√π,
Z ∞ 0
e(a2−1)t2sin(2at2)dt = a 2(1 + a2)
√π, −1 ≤ a ≤ 1.
In particular, when a = 1, we obtain Z ∞
0
cos t2dt= Z ∞
0
sin t2dt= 1 2
r π 2.
Proof: When a = 0, R∞
0 e−t2dt= √2π. It suffices to show: I :=R∞
−∞e−t2dt=√
π, which is so-called Gauss error integral.
Using the polar coordinates x = r cos φ and y = r sin φ, the double integral is evaluated I2 =
Z ∞
−∞
e−x2dx Z ∞
−∞
e−y2dy= Z Z
R2
e−x2−y2dxdy
= Z 2π
0
Z ∞
0
e−r2rdrdφ= π Z ∞
0
e−tdt= π.
Assume 0 < a ≤ 1 here. Define f(z) := e−z2.
For a triangle determined by the three points: 0, r and r + ari. Denote by γ1, γ2, γ3 the line segments from the origin to r, from r to r + ira, from the origin to r + ira, respectively.
Then γ1+ γ2− γ3 forms a closed curve with counter-clockwise orientation.
72A.J. Fresnel (1788-1827) is a French engineer and physicist.
By Cauchy Integral Theorem, Z
γ3
f(ζ)dζ = Z
γ1
f(ζ)dζ + Z
γ2
f(ζ)dζ.
We claim that lim
r→∞
R
γ2f(ζ)dζ) = 0. In fact, γ2(t) = r + it for t ∈ [0, ar]. Then
|f(γ2(t))| = |e−(r+it)2| ≤ e−r2+t2 ≤ e−r2 · ert f or 0 ≤ t ≤ r so that
| Z
γ2
f(ζ)dζ| ≤ Z ar
0 |f(γ2(t))||γ′2(t)|dt ≤ e−r2 Z r
0
ertdt≤ 1 r. Our claim is proved.
Write γ3(t) = (1 + ia)t for t ∈ [0, r]. Then γ′3(t) = 1 + ia and (1 + ia)
Z ∞
0
e−(1+ia)2t2dt= lim
r→∞
Z
γ3
f(ζ)dζ = lim
r→∞
Z
γ1
f(ζ)dζ = Z ∞
0
e−t2dt= π 2.