Professional Documents
Culture Documents
df
(if such a limit exists). We also denote f 0 (c) by dz
(c). A necessary condition
for the existence of f 0 (c) is that
∂f 1 ∂f
= = f 0 (c)
∂x i ∂y
at c, because restriction of the difference quotient to the horizontal line y = b
and to the vertical line x = a when passing to limit would give the same
result, where c = a + ib with a, b ∈ R. The reason for the factor 1i is that
z − c = i(y − b) with the factor i on x = a whereas z − c = x − a without
the factor i on y = b. This necessary condition
∂f 1 ∂f
=
∂x i ∂y
is known as the Cauchy-Riemann equations. When f (z) = u(x, y) + iv(x, y),
separation of the Cauchy-Riemann equations into real and imaginary parts
yields its alternative formulation
∂u ∂v
∂x = ∂y
∂u ∂v
∂y
= − ∂x .
with
Ej (x, y)
lim p = 0 for j = 1, 2.
(x,y)→(a,b) (x − a)2 + (y − b)2
and ∂u ∂v
∂x = ∂y
∂u ∂v
∂y
= − ∂x ,
we can add i times the second equation of (∗) to the first to get
which means
f (z) = f (c) + f 0 (c)(z − c) + (E1 + E2 )
when we set f 0 (c) = ux (a, b) + ivx (a, b).
with
Ej (x, y)
lim p = 0 for j = 1, 2
(x,y)→(a,b) (x − a)2 + (y − b)2
means that the map Φ : (x, y) 7→ (u, v) can be approximated at (x, y) = (a, b),
to an order higher than the first, by the linear map dΦ : R2 → R2 defined by
the 2 × 2 Jacobian matrix
ux (a, b) uy (a, b)
vx (a, b) vy (a, b)
vanishes.
Math Math 213a (Fall 2021) Yum-Tong Siu 5
(where o (`c ) is the Landau symbol for lower order), in which one order of `c
is from the length of ∂Rc and the other from Ec (z). We divide up each side
of R into m equal parts so that we end up with m2 equal closed subrectangles
Rcj with a point cj inside (for 1 ≤ j ≤ m2 ).
Math Math 213a (Fall 2021) Yum-Tong Siu 6
and
m2 Z
Z
X
2 1
f (z)dz ≤ f (z)dz = m o = o(1),
m2
∂R j=1
∂Rcj
R
which implies, as m → ∞, the vanishing of ∂R f (z)dz. This argument is
not yet rigorous, because there is a question of uniformity in the use of the
Landau symbol of lower order. The usual argument for uniformity by com-
pactness is by reductio ad absurdum of selecting a sequence of points where
the uniformity fails and considering the situation at one of their accumula-
tion points. We now use the following rigorous argument of combining the
above intuitive analysis with the uniformity argument by compactness.
R
Suppose ∂R f (z)dz is nonzero. Let A be its absolute value. Let ` be the
length of the boundary ∂R of R. Divide R into 4 equal closed subrectangles.
Then one closed subrectangle R(1) out of the four must satisfy
Z
A
f (z)dz ≥ .
∂R (1)
4
Divide up R(1) into closed subrectangles. Then one closed subrectangle R(2)
out of the four must satisfy
Z
≥ A.
f (z)dz 42
∂R(2)
and the estimate that for any ε > 0 there exists mε ∈ N such that |Ec (z)| ≤
ε 2`m on R(m) and
Z 2
`
Ec (z)dz ≤ ε m
for m ≥ mε .
(m)
∂R 2
This yields the contradiction
2 Z
` A
ε m ≥
f (z)dz ≥ m
2 ∂R(m) 4
The reason for the piecewise smooth condition is that for a curve-segment
γ defined by t 7→ z = ϕ(t) ∈ Ω for t ∈ [a, b],
Z Z b
f (z)dz = f (ϕ(t))ϕ0 (t)dt
γ t=a
for any closed curve-segment γ in Ω with initial point c1 and terminal point
c2 .
Integral of Power Function along a Circle. For any ρ > 0 and any m ∈ Z,
Explicit computation with the parametrization θ 7→ ρeiθ (0 ≤ θ ≤ 2π) for
the circle |z| = ρ yields
Z Z 2π
m
z dz = i ρm+1 ei(m+1)θ dθ
|z|=ρ θ=0
i
(m+1)θ 2π
m+1 e θ=0
= 0 for m 6= −1
=
2πi for m = −1
is equal to
f (z) − f (c)
Z Z Z
f (z) dz
dz = dz + f (c)
|z−c|=ρ z−c |z−c|=ρ z−c |z−c|=ρ z−c
= 2πi f (c)
Math Math 213a (Fall 2021) Yum-Tong Siu 10
by letting ρ → 0, because
Z
f (z) − f (c) f (z) − f (c)
dz ≤ 2πρ sup
|z−c|=ρ z − c |z−c|=ρ z − c
and
f (z) − f (c)
sup < ∞.
0<|z−c|≤ρ0 z−c
for τ ∈ (c, d). By differentiating with respect to τ and using the fundamental
theorem of calculus we get with τ = t
Z b
∂ b
Z
∂f
(x, t)dx = f (x, t)dx.
x=a ∂t ∂t x=a
for any nonnegative integer n the n-th complex derivative f (n) (z) of f is given
by Z
(n) n! f (ζ)
f (z) = dζ for z ∈ G.
2πi ∂G (ζ − z)n+1
To prove it, we first verify that for any complex-valued continuous function
g on ∂G and for any nonnegative integer n, the function
Z
g(ζ)
z 7→ h(z) = n+1
dζ
∂G (ζ − z)
∂ p+q ∂ p+q
Z
1
(†)p,q h(z) = g(ζ) p q dζ
∂xp ∂y q ∂G ∂x ∂y ζ −z
This concludes the proof of Cauchy’s integral formula for derivatives of holo-
morphic functions. -A corollary is that any holomorphic function of z = x+iy
is infinitely differentiable as a function of x, y.
means that f (z) can be interpreted the limit of the Riemann sum
X 1 f (ζj )
(ζj+1 − ζj ),
j
2πi ζj − z
which means that any holomorphic function f (z) is the limit of a finite lin-
ear combination, with constant coefficients, of the translates of the special
Math Math 213a (Fall 2021) Yum-Tong Siu 13
function z1 , which is the Cauchy kernel.. The word “kernel” carries the mean-
ing that all holomorphic functions germinate from the Cauchy kernel . The
1 1
Cauchy kernel is alternatively taken to mean 2πi ζ−z
so that any holomorphic
function is reproduced by integration against it along a surrounding contour.
where ∂~n is the derivative in the normal direction of the boundary |z| = 1 of
the unit disk D. To facilitate the discussion of the relation between Cauchy’s
kernel and Newton’s potential, we now introduce two differential operators
∂
∂z
and ∂∂z̄ .
∂f
to ∂z
, because
∂f ∂f ∂f ∂f
f0 = = + = .
∂x ∂z ∂ z̄ ∂z
∂ ∂ 2 2 ∂ 2
The Laplacian ∆ = ∂x 2 + ∂y 2 is equal to 4 ∂z∂ z̄ . More precisely, the equation
Since
i
dx ∧ dy = dz ∧ dz̄,
2
the equation for the Newton potential can be rewritten as
∂2
1 i
4 log |z| dz ∧ dz̄ = δ0 .
2π ∂z∂ z̄ 2
or
−1 ∂ 2
2
log |z| dz ∧ dz̄ = δ0 ,
2πi ∂z∂ z̄
which, after using
∂ ∂ ∂ 1
log |z|2 = (log z + log z̄) = log z = ,
∂z ∂z ∂z z
becomes
−1 ∂ 1
dz ∧ dz̄ = δ0
2πi ∂ z̄ z
1
where the Cauchy kernel z
appears. Note that
∂ 1
log z =
∂z z
holds in the sense of distributions, because the function z1 is locally inte-
grable on C so that integration by parts can be justified when both sides are
integrated against a smooth function on C with compact support.
Math Math 213a (Fall 2021) Yum-Tong Siu 15
where χ(−∞,x) is the characteristic function of the interval (−∞, x). The
rewriting is motivated by the fact that the derivative of the Heaviside function
is the Dirac delta function, which is analogous to the equation for Cauchy
kernel derived from Newton’s potential.
Cauchy’s integral formula for smooth functions states that
∂f
1
Z
f (ζ)dζ 1
Z
∂ ζ̄
(ζ)dζ ∧ dζ̄
f (z) = + .
2πi ∂G ζ −z 2πi G ζ −z
1
R f (ζ)dζ
The term 2πi ∂G ζ−z
is analogous to the term g(z). The operator ∂∂ζ̄ is anal-
ogous to dtd . The Cauchy kernel 2πi 1 1
ζ−z
is analogous to χ(−∞,x) . Analogous
to Z b
d
χ(−∞,x) (t)ϕ(t) dt = ϕ(x)
dx t=a
from the fundamental theorem of calculus, the Cauchy integral formula for
smooth functions can be used to yield
h(ζ)dζ ∧ dζ̄
Z
∂ 1
= h(z)
∂ z̄ 2πi G ζ −z
for z ∈ G when h is a smooth function on Ω. Suppose we know that there
exists some smooth function F on a neighborhood U of Ḡ such that
∂F
= h(z)
∂ z̄
on U . Then we can apply Cauchy’s integral formula for smooth functions to
F to get
Z Z ∂F (ζ)dζ ∧ dζ̄
1 F (ζ)dζ 1 ∂ ζ̄
F (z) = +
2πi ∂G ζ −z 2πi G ζ −z
h(ζ)(ζ) dζ ∧ dζ̄
Z Z
1 F (ζ)dζ 1
= + ,
2πi ∂G ζ −z 2πi G ζ −z
Math Math 213a (Fall 2021) Yum-Tong Siu 17
∂
which, upon applying ,
yields immediately
∂ z̄
h(ζ)(ζ) dζ ∧ dζ̄
Z
∂ 1
h(z) =
∂ z̄ 2πi G ζ −z
for z ∈ G. However, we still have the problem of coming up with some such
smooth solution F of
∂F
= h(z)
∂ z̄
on U .
For a ∈ C, when |z − a| > |ζ − a|, with the exchange of the roles of z and
ζ, the geometric series expansion
1 −1 −1 1
= = ζ−a
ζ −z (z − a) − (ζ − a) z − a 1 − z−a
∞ −1
X (ζ − a)n X (z − a)m
=− = −
n=0
(z − a)n+1 m=−∞
(ζ − a)m+1
(by using n = −(m + 1) in the change of dummy index)
where for the last step the Cauchy integral formula for derivatives of holo-
morphic functions is used.
(n)
so that cn = f n!(a) by the Cauchy integral formula for derivatives of holo-
morphic functions.
Math Math 213a (Fall 2021) Yum-Tong Siu 19
PnR1 < |z − a| k< R2 instead of just the convergence of the principal sum
of
k=−n ck (z − a) as n → ∞. Then for R1 < r < R2 and any integer k,
Z ∞ Z
f (z)dz X
= cn (z − a)n−k−1 dz = 2πick ,
|z−a|=r (z − a)k+1 n=0 |z−a|=r
on any open disk |z − a| < r whose closure is contained in Ω. Either all the
coefficients of the power series vanish, in which case f is identically zero on
|z − a| < r and Z contains an open neighborhood of a or f (z) = (z − a)k g(z)
with ∞
X
g(z) = cn (z − a)n−k
n=k
is called the principal part of the Laurent series. The isolated singularity of
f at a is called
because (z − a)k f (z) is a power series whose (k − 1)-th power term has coef-
ficient c−1 . Residues are introduced for the computation of definite integrals
by techniques of complex analysis.
A B
≤ |f (z)| ≤
|z − a|k |z − a|k
for some positive constants A, B when z is in a sufficiently small deleted
1
neighborhood of a, because |f (z)| ∼ |z−a| k means that the isolated singularity
(z − a1 )p1 · · · (z − ak )pk
f (z) = g(z)
(z − b1 )q1 · · · (z − b` )q`
Then
k `
f0 X f0 X f0
Z
1
= Resaµ + Resbν
2πi ∂G f µ=1
f ν=1
f
k `
X pµ X qν
= Resaµ − Resbν
µ=1
z − aµ ν=1 z − bν
k
X `
X
= pµ − qν .
µ=1 ν=1
Math Math 213a (Fall 2021) Yum-Tong Siu 24
f0
Since f
= (log f )0 , it follows that
f0
Z Z
1 1
= d log f
2πi ∂G f 2πi ∂G
Z
1
= (d log |f | + i arg f )
2πi ∂G
Z
1
= d arg f
2π ∂G
1
= ∆∂G arg f.
2π
Here arg f is the argument of f which means θ when f = |f |eiθ . Though
arg f can be defined as an angle or direction, yet its numerical value is not
well-defined and can only be chosen up to an integral multiple of 2π.
The function
f 0 + tg 0
Z
1 1
t 7→ = ∆∂G arg(f + tg)
2πi ∂G f + tg 2π
for 0 ≤ t ≤ 1 is continuous with integral values so that it is independent of
t ∈ [0, 1], because f (z) + tg(z) is nowhere zero for t ∈ [0, 1] and z ∈ ∂G from
the assumption that |f | > |g| on ∂G.
Proof. Since f 0 is not identically zero, the zeroes of f 0 are isolated. Let G
be an open subset of Ω and z0 be any point of G. If f 0 is nonzero at z0 , the
inverse mapping theorem guarantees that f can be locally inverted at z0 and
some open neighborhood Uz0 of z0 in G is mapped by f to some open subset
of C. If f 0 is zero at z0 , the vanishing order of f 0 at P is some positive integer
k. Write f (z) = f (z0 ) + (z − z0 )` g(z) with g(z) holomorphic and nonzero at
z0 . Then ` = k + 1.
Math Math 213a (Fall 2021) Yum-Tong Siu 26
Let g be the inverse map of f . By the chain rule g 0 (0)f 0 (0) = 1. The
Schwarz lemma applied separately to f and g implies that |f 0 (0)| ≤ 1 and
|g 0 (0)| ≤ 1. Hence |f 0 (0)| = 1 and f (z) = eiα z for some α ∈ R by the Schwarz
lemma.
where R (x, y) is a rational function (with the value of R (cos θ, sin θ) assumed
finite at every point θ ∈ [0, 2π]), The substitution
iθ z + z1 z − z1 dz
z = e , cos θ = , sin θ = , dθ =
2 2i iz
is used to convert the integral into a contour integral
Z
f (z)dz
|z|=1
Math Math 213a (Fall 2021) Yum-Tong Siu 28
of a meromorphic function f (z) of z over the contour |z| = 1, which can then
be evaluated by using Cauchy’s residue theory.
The formula is
Z ∞
P (x)dx X P (z)
= 2πi Resa .
x=−∞ Q(x) Im a>0
Q(z)
over the contour of the boundary of the upper half disk of radius R centered
at the origin as R → ∞.
P (z)eiz dz
Z
lim = 0,
R→∞ C
R
Q(z)
the degree of the denominator is at least 2 more than that of the numerator.
the contour used is the boundary of the domain which is the annulus
n o
z ∈ C ε < |z| < R
with inner radius ε and outer radius R minus the right half-strip
n o
z ∈ C Re z ≥ 0, −ε ≤ Im z ≤ ε
The meromorphic function used involves choosing branches for the loga-
rithmic function and the fraction power function which can also be expressed
in terms of the logarithmic function.
Math Math 213a (Fall 2021) Yum-Tong Siu 31
of the integral over ∂G by letting ∂G recede to infinity and at the same time
keep these additional partial fraction terms from the pole singularities of f .
ζ2
f (ζ) −1 1 ζ 0 1 00 2
= p+1 + + + ··· f (0) + f (0)ζ + f (0)ζ + · · ·
ζ p+1 (ζ − z) ζ z z2 z3 2
Math Math 213a (Fall 2021) Yum-Tong Siu 33
1
and the coefficient of ζ
is
Note that
1 z2 zp
+ 2 + · · · + p+1
an an an
is the p-th Taylor polynomial for the function
1
z − an
at z = 0 from geometric series expansion.
and hence uniformly bounded. The function cot z is bounded on the line
joining 21 (1 − i)π to 12 (1 + i)π and we can use periodicity
P and conclude that
cosec z is uniformly bounded on Cn . Hence cot z − z1 = n∈Z−{0} z−nπ1 1
+ nπ
because the residue of cot z at nπ is 1.
Since
d z z 1 1
log 1 − en = + ,
dz n z−n n
when we integrate from z = 0 to z, we get
Z ζ=z
1 1 z z
+ dζ = log 1 − en .
ζ=0 ζ −n n n
After exponentiating, we obtain
sin πz Y z z
=C 1− en .
πz n
n∈Z−{0}
If we use cosec πz instead of cot πz we can obtain the sum of the series
P ∞ n
n=−∞ (−1) f (n).