0% found this document useful (0 votes)
45 views10 pages

Supplement: Laurent Series Development (Theorem V.1.11)

The document provides a detailed proof of the Laurent series development theorem. It begins by recalling previous results needed for the proof, including theorems about power series convergence and analytic functions having power series representations. It then defines functions f1 and f2 that together make up the original function f within different annuli. It shows the coefficients an have the claimed form for positive powers of (z-a) by relating f2 to its Taylor series. It defines a new function g in terms of f1 and shows g is analytic, from which the coefficients an for negative powers are obtained by relating f1 to g's power series.

Uploaded by

TOM DAVIS
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
45 views10 pages

Supplement: Laurent Series Development (Theorem V.1.11)

The document provides a detailed proof of the Laurent series development theorem. It begins by recalling previous results needed for the proof, including theorems about power series convergence and analytic functions having power series representations. It then defines functions f1 and f2 that together make up the original function f within different annuli. It shows the coefficients an have the claimed form for positive powers of (z-a) by relating f2 to its Taylor series. It defines a new function g in terms of f1 and shows g is analytic, from which the coefficients an for negative powers are obtained by relating f1 to g's power series.

Uploaded by

TOM DAVIS
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Supplement: Laurent Series Development 1

Supplement: Laurent Series Development


(Theorem V.1.11)

Note. In this section, we give a detailed proof of Theorem V.1.11, the “Laurent
Series Development” theorem. First, we recall some previous results which are
needed in the proof of this theorem.


X
Theorem III.1.3. If an (z − a)n , define the number R as 1
R
= lim |an |1/n (so
n=0
0 ≤ R ≤ ∞). Then

(a) if |z − a| < R, the series converges absolutely,

(b) if |z − a| > R, the series diverges, and

(c) if 0 < r < R then the series converges uniformly on |z − a| ≤ r. Moreover,


R is the only number having properties (a) and (b). R is called the radius of
convergence of the power series.


X
Theorem IV.2.8. Let f be analytic in B(a; R). Then f (z) = an (z − a)n for
n=0
(n)
|z − a| < R where an = f (a)/n! and this series has radius of convergence ≥ R.

Lemma IV.5.1. Let γ be a rectifiable curve and suppose ϕ is a function defined


R
and continuous on {γ}. For each m ≥ 1 let Fm(z) = γ ϕ(w)(w − z)−m dw for
/ {γ}. Then each Fm is analytic on C \ {γ} and Fm0 (z) = mFm+1 (z).
z∈
Supplement: Laurent Series Development 2

Corollary IV.5.9. (Theorem 5.8 with one curve.) Let G be an open set and
f : G → C analytic. If γ is a closed rectifiable curve in G such that n(γ; w) = 0 for
all w ∈ C \ G then for a ∈ G \ {γ}
Z
(k) k! f (z)
f (a)n(γ; a) = dz.
2πi γ (z − a)k+1

Theorem V.1.2. If f has an isolated singularity at a then z = a is a removable


singularity if and only if lim (z − a)f (z) = 0.
z→a

Note. Now for our main result.

Theorem V.1.11. Laurent Series Development.


Let f be analytic in ann(a; R1, R2). Then

X
f (z) = an (z − a)n
n=−∞

where the convergence is absolute and uniform over the closure of ann(a; r1, r2 ) if
R1 < r1 < r2 < R2. The coefficients an are given by
Z
1 f (w)
an = dw (1.12)
2πi γ (w − a)n+1

where γ is the circle γ(t) = a + reit where t ∈ [0, 2π] for any r with R1 < r < R2 .
Moreover, this series is unique.

Proof. If R1 < r1 < r2 < R2 and γ1 , γ2 are the circles γ1 (t) = a + r1 eit for
t ∈ [0, 2π] and γ2 (t) = a + r2 eit for t ∈ [0, 2π], then γ1 ∼ γ2 in ann(a; R1, R2 ).
Then by Cauchy’s Theorem (Third Version, Theorem IV.6.7) for any analytic g on
Supplement: Laurent Series Development 3

R R
ann(a; R1, R2 ), we have γ1
g= γ2
g. So the integral in (1.12) is independent of r,
so for each n ∈ Z, an is a constant (i.e., independent of γ).
Claim. Definition of f1 and f2 ; f (z) = f1 (z) + f2 (z).
Moreover, function f2 : B(a; R2) → C given as
Z
1 f (w)
f2 (z) = dw,
2πi γ2 w − z
where |z − a| < r2 and R1 < r2 < R2, is a well defined function (independent of
r2 ). By Lemma IV.5.1 (with m = 1 and ϕ(w) = f (w)), f2 is analytic in B(a; r2 ).
Similarly, if G = {z | |z − a| > R1} then f1 : G → C defined as
Z
−1 f (w)
f1 (z) = dw,
2πi γ1 w − z
where |z − a| > r1 (so r1 is chosen such that z ∈
/ {γ1 }) and R1 < r1 < R2 is analytic
in G.

If R1 < |z − a| < R2, fix r1 and r2 so that R1 < r1 < |z − a| < r2 < R2 . Let
γ1 (t) = a + r1 eit for t ∈ [0, 2π] and γ2 (t) = a + r2 eit for t ∈ [0, 2π], as above. Let
λ be a line segment from a point on γ1 radially to a point on γ2 which does not
Supplement: Laurent Series Development 4

contain z:

Since γ1 ∼ γ2 on ann(a; R1, R2 ), we have that the closed curve γ = γ2 − λ − γ1 + λ


is homotopic to 0. Next, n(γ2 ; z) = 1 and n(γ1 ; z) = 0 gives:
Z
1 f (w)
f (z) = dw by the Cauchy’s Integral Formula
2πi γ w − z
Z Z
1 f (w) 1 f (w)
= dw − dw since γ is piecewise smooth
2πi γ2 w − z 2πi γ1 w − z
= f2 (z) + f1 (z) by the definitions of f1 and f2.

We now create series for f1 (with negative powers of (z − a)) and f2 (with the usual
positive powers of (z − a)). 
Claim. The an are as claimed for n ≥ 0.
Since f2 is analytic in B(a; R2) then by Theorem IV.2.8

X
n f (n) (a)
f2(z) = an(z − a) and an = .
n=0
n!
1
R f(w)
Since f2 (z) = 2πi γ2 w−z dw, in Lemma IV.5.1 with m = 1 we have F1 (z) =
1
R ϕ(w)
2πi γ2 w−z dw where ϕ(w) = f (w), and therefore by induction
Z
(n) n! ϕ(w)
f2 (z) = n!Fn+1 (z) = dw.
2πi γ2 (w − z)n+1
Supplement: Laurent Series Development 5

We then have
Z
f (n) (a) 1 f (w)
an = = Fn+1 (a) = dw,
n! 2πi γ2 (w − a)n+1
as claimed. 
Claim. Definition of g(z); g(z) is analytic for z ∈ B(0; 1/R1 ).
Now define g(z) for 0 < |z| < 1/r1 as g(z) = f1 (a + 1/z). Notice that a + 1/z =
a − (−1/z) and | − 1/z| > r1 , so a + 1/z ∈ ann(a; r1 , ∞). Since f1 is defined as
Z
−1 f (w)
f1 (z) = dw
2πi γ1 w − z
where γ1 (t) = a + r1 eit with t ∈ [0, 2π], then f1(a + 1/z) is defined as long as
/ {γ1 }. We have:
a + 1/z ∈

Since f1 (z) is analytic for all z with |z − a| > r1 (from the definition of f1), then
g(z) is analytic for all a + 1/z with |(a + 1/z) − a| > r1 , or |1/z| > r1 or 0 < |z| <
1/r1 . So g has an isolated singularity at z = 0. We now show g has a removable
singularity at z = 0 by showing limz→0 g(z) exists (and hence limz→0 zg(z) = 0,
as required by Theorem V.1.2 for g to have a removable singularity at z = 0).
Suppose R2 > r > r1 , z ∈ ann(a; r, ∞), and ρ(z) = d(z, C) where C is the circle
Supplement: Laurent Series Development 6

{w | |w − a| = r}:

Define M = max{|f (w)| | w ∈ C} and give C a counterclockwise orientation. Then


for such z “outside” of C,
Z
−1 f (w)
|f1(z)| = dw
2πi γ1 w − z
Z
−1 f (w)
= dw since γ1 ∼ C on ann(a; R1, R2 )
2πi C w − z
Z
1 |f (w)| 1 M 2πr Mr
≤ |dw| ≤ = .
2π C |w − z| 2π ρ(z) ρ(z)
Now limz→∞ ρ(z) = ∞ and M, r are constants, so limz→∞ f1 (z) = 0. Therefore,

lim g(z) = lim f1 (a + 1/z) = lim f1 (z) = 0.


z→0 z→0 z→∞

Since limz→0 g(z) = 0, g has a removable singularity at z = 0 by Theorem V.1.2. We


define g(0) = 0 and then g is analytic on B(0; 1/r1 ) (by the definition of removable
singularity). Notice that r1 > R1 was arbitrary above, and we can conclude that
each claim is valid for r1 replace with R1 (as in the text). So g can be written
P P∞
g(z) = ∞ n=1 Bn z n
(B0 = 0 since g(0) = 0). So, f 1 (a + 1/z) = g(z) = n=1 Bn a
n
Supplement: Laurent Series Development 7

for z ∈ B(0; 1/R1 ). 


Claim. The an are as claimed for n ≤ −1.
−1
R f(x)
We have by definition that f1(z) = 2πi γ1 w−z dw for |z − a| > R1 where γ1 (t) =

a + r1 eit for t ∈ [0, 2π]. So for |a| < R1 we have


  Z
1 −1 f (w)
f1 a + =  dw.
z 2πi γ1 w − a + z1

Replacing w with a + 1/w, dw with −w −2 dw, and using the inverse of this mapping
to take γ1 to γ10 (so that the integrand receives the same values of w) we have

1 1 1 −it
γ10 (t) = = = e for t ∈ [0, 2π].
γ1 (t) − a (a + r1 eit ) − a r1

Then f1 (a + 1/z) becomes


  Z Z
1 −1 f (a + 1/w) −2 1 f (a + 1/w)
f1 a + =   [−w ] dw =  dw
z 2πi γ10 a + w1 − a + 1z 2πi γ10 w 1 − wz
Z Z
1 zf (a + 1/w) 1 zf (a + 1/w)
= dw = dw
2πi γ10 w(z − w) 2πi −γ10 w(w − z)
Z
z f (a + 1/w)/w
= dw = g(z) for 0 < |z| < 1/R1 .
2πi −γ10 w−z
P
As established above, g(z) is analytic on B(0; 1/R1 ), so g(z) = ∞ n
n=1 Bn z (notice

that B0 = 0 since g(0) = 0). Also, by Theorem IV.2.8, Bn = g (n) (0)/n!. We now
calculate g (n) (0) using Lemma IV.5.1 and the representation of g(z) as
Z
z f (a + 1/w)/w
g(z) = dw
2πi −γ10 w−z
1 it
(notice that −γ10 (t) = r1
e for t ∈ [0, 2π]). From the Product Rule, Lemma IV.5.1
(with ϕ(w) = f (a + 1/w)/w), and Math Induction we can show that
Z Z
(n) n! f (a + 1/w)/w n!z f (a + 1/w)/w
g (z) = n
dw + dw for |z| < R1.
2πi −γ10 (w − z) 2πi −γ10 (w − z)n+1
Supplement: Laurent Series Development 8

n!
R f(a+1/w) g(n)(0) 1
R f(a+1/w)
So g (n) (0) = 2πi −γ10 w n+1
dw, so Bn = n!
= 2πi −γ10 w n+1
dw. Replacing w
with 1/(w − a) and dw with −(w − a)−2 dw and using the inverse of this mapping
to take γ10 to γ1 (similar to above)
Z Z  
1 f (a + 1/w) 1 f (w) −1
Bn = dw =  dw
2πi −γ10 w n+1 2πi −γ1 1 n+1 (w − a)2
Z Z w−a
−1 1 f (w)
= f (w)(w − a)n−1 dw = dw.
2πi −γ1 2πi γ1 (w − a)−n+1
Defining a−n as Bn (and replacing n with −n in Bn ) we have
Z
1 f (w)
a−n = dw for − n ≥ 1
2πi γ1 (w − a)n+1
P
and f1(a + 1/z) = ∞ n
n=1 Bn z for 0 < |z| < 1/R1 or

X
f1 (z) = Bn (z − a)−n for z ∈ ann(a; R1, ∞)
n=1
X∞ −1
X
−n
= a−n (z − a) = an (z − a)n.
n=1 n=−∞

So

f (z) = f1 (z) + f2(z)


−1
X ∞
X
n
= an (z − a) + an (z − a)n
n=−∞
| {z } |n=0 {z }
for |z−a|>R1 for |z−a|<R2

X
= an (z − a)n for z ∈ ann(a; R1, R2 )
n=−∞

where the an are as claimed. 


Claim. The Laurent series converges absolutely and uniformly on r1 ≤ |a| ≤ r2 .
P
Since f2(z) = ∞ n
n=0 an(z −a) for z ∈ B(a; R2 ) then this series converges absolutely

for |z − a| < R2 by Theorem III.1.3(a) and converges uniformly for |z − a| ≤


Supplement: Laurent Series Development 9

P∞ n
r2 for any 0 < r2 < R2 by Theorem III.1.3(c). Similarly, g(z) = n=1 Bn z

converges absolutely for |z − a| < 1/R1 and converges uniformly for |z − a| ≤ 1/r1
P
for any r1 > R1 by Theorem III.1.3(a) and (c). So f1(z) = ∞ n=1 a−n (z − a)
−n

converges absolutely for |z − a| > R1 and uniformly for |z − a| ≥ r1 . Therefore


P
f (z) = ∞ n −
n=−∞ an (z − a) converges absolutely and uniformly on ann(a; r1 , r2 ) if

R1 < r1 < r2 < R2. 


Claim. The Laurent series representation is unique for given f .
Now for the uniqueness. If we were dealing with an analytic function on B(a; R),
we could deduce that the coefficients are unique, in fact an = f (n) (a)/n! as given
in Theorem IV.2.8. For a Laurent series, we integrate. Let R1 < r1 < r2 < R2
and let γ = ((r1 + r2 )/2)eit for t ∈ [0, 2π]. Notice that the Laurent series converges
P
uniformly on ann(a; r1 , r2 )− as given above. So for f (z) = ∞ n
n=−∞ an (z − a) we

have for k ≥ 0:
Z Z X ∞
f (w)
k+1
dw = an (w − a)n−(k+1) dw
γ (w − a) γ n=−∞
Z −1 ∞
!
X X
= an (w − a)n−(k+1) + an (w − a)n−(k+1) dw
γ n=−∞ n=0
since the convergence is absolute—see Definition V.1.10
Z −1
! Z X ∞
!
X
= an (w − a)n−(k+1) dw + an (w − a)n−(k+1) dw
γ n=−∞ γ n=0
−1 Z
X  ∞ Z
X 
= an (w − a)n−(k+1) dw + an (w − a)n−(k+1) dw
n=−∞ γ n=0 γ
since the convergence is uniform; Lemma IV.2.7
∞ Z  X ∞ Z 
X a−n
= n+k+1
dw + an(w − a)n−(k+1) dw
n=1 γ (w − a) n=0 γ
Z
ak
= 0+0+ dw since each integrand has a primitive on
γ w−a
Supplement: Laurent Series Development 10

G = ann(a; r1 , r2 ) for k 6= n

= ak 2πi n(γ; a) = ak 2πi.

1
R f(w)
So for k ≥ 0 it must be that ak = 2πi γ (w−a)k+1 dw. The arbitrary nature of γ is
explained above. For k ≤ −1, consider
Z Z X ∞
f (w)
k+1
dw = an (w − a)n−k−1 dw
γ (w − a) γ n=−∞
Z −1 ∞
!
X X
= an (w − a)n−k−1 + an(w − a)n−k−1 dw
γ n=−∞ n=0
Z −1
! Z ∞
!
X X
= an (w − a)n−k−1 dw + an (w − a)n−k−1 dw
γ n=−∞ γ
n=0
−1
X Z  ∞ 
X Z 
n−k−1) n−k−1
= an (w − a) dw + an (w − a) dw
n=−∞ γ n=0 γ
∞ Z  ∞ Z 
X a−n X
= dw + an(w − a)n−k−1 dw
n=1 γ (w − a)n+k+1 n=0 γ
Z
ak
= 0+ dw + 0 = ak 2πi n(γ; a) = ak 2πi.
γ w−a

1
R f(w)
So for k ≤ −1 it must be that ak = 2πi γ (w−a)k+1 dw. The arbitrary nature of γ is

explained above.

Revised: 4/4/2018

You might also like