Complex Analysis: Problems on Residues, Rouché's Theorem, and Conformal Mappings, Assignments of Mathematics

Solutions to selected problems from a university-level complex analysis course. Topics covered include residues, rouché's theorem, and conformal mappings. Students will learn how to find residues, apply rouché's theorem to determine the number of zeroes of a function, and analyze conformal mappings using schwarz's lemma and the fundamental theorem of algebra.

Typology: Assignments

Pre 2010

Uploaded on 10/01/2009

koofers-user-t18
koofers-user-t18 🇺🇸

8 documents

1 / 2

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Patrick Corn
Math 185, 8/9/05
Solution Set 12
Problem 1: Suppose that f(z) = (zz0)mg(z) where g(z) is holomorphic and nonzero in a
neighborhood of z0. Then, as we have seen a few times already, the residue of f/f at z0is m.
(Just use the product rule.) So if fhas a zero of order k, the residue at it is k, and if fhas a pole
of order , the residue at it is . Now just apply the residue theorem.
Problem 2: Let z0be a zero of f. Let Crbe the circle of radius raround z0, where ris small
enough so that fhas no zeroes other than z0on or inside Cr. Let ǫ= inf{|f(z)|:zCr}. Then
ǫ > 0, and we can find NNsuch that |f(z)fn(z)|< ǫ |f(z)|for all n > N and zCr. So
then fand fnhave the same number of zeroes inside Cr(with multiplicities), by Rouch´e’s theorem.
Doing this process at each of the mzeroes, we see that fnhas at least mzeroes if fhas at least m
zeroes, for ngreater than the maximum of the numbers Nwe obtain via the above process.
As for the corollary, fix wG, and consider the functions fn(z)f(w), which converge to
f(z)f(w). If the latter function has at least two zeroes, then fn(z)f(w) has at least two zeroes
for sufficiently large n, but this contradicts the hypothesis that fnis univalent. So f(z)f(w) has
one zero for all wG, so fis univalent.
Problem 3: Let L1be the line segment {it :|t| R}. Let L2be the arc {Reit :π/2t3π /2}.
Let Kbe the closed half-disk bounded by L1and L2. Let f(z) = z+a. Let g(z) = z+aez.
Suppose Ris large enough so that the closed disk |z+a| 1 is contained in the interior of K.
Then on L1and L2, we have
|f(z)g(z)|=|ez| 1<|z+a|=|f(z)|,
and f(z) has exactly one zero in K, so g(z) has exactly one zero there as well. As we let Rget
bigger, Kincreases to cover the entire left half plane, so the result follows. (If there was another
zero, it would lie inside Kfor large enough R, which would be a contradiction of the above Rouch´e’s
Theorem argument.)
To see that this zero is on the real axis, note that g(a) = ea<0 and g(0) = a1>0, so g
has a zero in (a, 0) by the Intermediate Value Theorem.
Problem 4: (i) The LFT φ1(z) = i(z1)
z+1 should work.
(ii) The function φ2(z) = φ(z)2should work, since squaring sends the upper half plane to C
minus the nonnegative real axis.
(iii) Try φ3(z) = Log(φ2(z))+πi
2π.
(iv) Try φ4(z) = pφ1(z) = eLog(φ1(z))/2. You might also use eπφ3(z)/2, if you liked doing things
sequentially.
(v) Try φ5(z) = 1φ4(z)
1+φ4(z).
(vi) Try φ6(z) = φ5(z)2. If you want an explicit formula for φ6, you can get it if you want; knock
yourself out.
Problem 5: Suppose f(0) = 0, f(0) is real and positive, and fmaps the open unit disk onto
itself. Well, then g=f1has the same properties (because gand fare reciprocals). In fact, by
Schwarz’s lemma, both f(0) and g(0) are 1, but they are reciprocals, so they must both equal
1. Since equality holds, f(z) = λz, but λ=f(0) = 1, so we are done.
Problem 6: Let fbe a conformal automorphism of C, and consider the singularity of fat .
We have seen that if it is removable, then fis a constant, which is no good. We have also seen
that if it is a pole, then fis a nonconstant polynomial. But the univalence of fimplies that it
factors as a(zb)nfor some n(because it can’t have more than one zero, but it has to factor
1
pf2

Partial preview of the text

Download Complex Analysis: Problems on Residues, Rouché's Theorem, and Conformal Mappings and more Assignments Mathematics in PDF only on Docsity!

Patrick Corn Math 185, 8/9/ Solution Set 12

Problem 1: Suppose that f (z) = (z − z 0 )mg(z) where g(z) is holomorphic and nonzero in a neighborhood of z 0. Then, as we have seen a few times already, the residue of f ′/f at z 0 is m. (Just use the product rule.) So if f has a zero of order k, the residue at it is k, and if f has a pole of order ℓ, the residue at it is −ℓ. Now just apply the residue theorem.

Problem 2: Let z 0 be a zero of f. Let Cr be the circle of radius r around z 0 , where r is small enough so that f has no zeroes other than z 0 on or inside Cr. Let ǫ = inf{|f (z)| : z ∈ Cr }. Then ǫ > 0, and we can find N ∈ N such that |f (z) − fn(z)| < ǫ ≤ |f (z)| for all n > N and z ∈ Cr. So then f and fn have the same number of zeroes inside Cr (with multiplicities), by Rouch´e’s theorem. Doing this process at each of the m zeroes, we see that fn has at least m zeroes if f has at least m zeroes, for n greater than the maximum of the numbers N we obtain via the above process. As for the corollary, fix w ∈ G, and consider the functions fn(z) − f (w), which converge to f (z) − f (w). If the latter function has at least two zeroes, then fn(z) − f (w) has at least two zeroes for sufficiently large n, but this contradicts the hypothesis that fn is univalent. So f (z) − f (w) has one zero for all w ∈ G, so f is univalent.

Problem 3: Let L 1 be the line segment {it: |t| ≤ R}. Let L 2 be the arc {Reit^ : π/ 2 ≤ t ≤ 3 π/ 2 }. Let K be the closed half-disk bounded by L 1 and L 2. Let f (z) = z + a. Let g(z) = z + a − ez^. Suppose R is large enough so that the closed disk |z + a| ≤ 1 is contained in the interior of K. Then on L 1 and L 2 , we have

|f (z) − g(z)| = |ez^ | ≤ 1 < |z + a| = |f (z)|,

and f (z) has exactly one zero in K, so g(z) has exactly one zero there as well. As we let R get bigger, K increases to cover the entire left half plane, so the result follows. (If there was another zero, it would lie inside K for large enough R, which would be a contradiction of the above Rouch´e’s Theorem argument.) To see that this zero is on the real axis, note that g(−a) = −e−a^ < 0 and g(0) = a − 1 > 0, so g has a zero in (−a, 0) by the Intermediate Value Theorem.

Problem 4: (i) The LFT φ 1 (z) = −i z(z+1− 1)should work. (ii) The function φ 2 (z) = −φ(z)^2 should work, since squaring sends the upper half plane to C minus the nonnegative real axis.

(iii) Try φ 3 (z) = Log(φ^22 (πz ))+πi. (iv) Try φ 4 (z) =

φ 1 (z) = eLog(φ^1 (z))/^2. You might also use eπφ^3 (z)/^2 , if you liked doing things sequentially. (v) Try φ 5 (z) = (^1) 1+−φφ^44 ((zz)). (vi) Try φ 6 (z) = φ 5 (z)^2. If you want an explicit formula for φ 6 , you can get it if you want; knock yourself out.

Problem 5: Suppose f (0) = 0, f ′(0) is real and positive, and f maps the open unit disk onto itself. Well, then g = f −^1 has the same properties (because g′^ and f ′^ are reciprocals). In fact, by Schwarz’s lemma, both f ′(0) and g′(0) are ≤ 1, but they are reciprocals, so they must both equal

  1. Since equality holds, f (z) = λz, but λ = f ′(0) = 1, so we are done.

Problem 6: Let f be a conformal automorphism of C, and consider the singularity of f at ∞. We have seen that if it is removable, then f is a constant, which is no good. We have also seen that if it is a pole, then f is a nonconstant polynomial. But the univalence of f implies that it factors as a(z − b)n^ for some n (because it can’t have more than one zero, but it has to factor 1

by the Fundamental Theorem of Algebra), and it is clear that this function is only univalent in a neighborhood of b if n = 1 (you can see it directly, or use the Local Mapping Theorem). Since it is clear that nonconstant linear functions are indeed conformal automorphisms of C, we must only show that the singularity at ∞ cannot be essential. Suppose it is; fix w 0 ∈ C, and for some z 0 ∈ C we know that f (z 0 ) = w 0. If ǫ is a sufficiently small positive number, then in some open neighborhood N of z 0 , all the values w in V = {|w−w 0 | < ǫ} are assumed by f exactly once. (This is the local mapping theorem.) But now let U = {|z| > R}, and take R sufficiently large so that U ∩N = ∅. Then f (U ) must intersect V by the Casorati-Weierstrass criterion, which contradicts the univalence of f. So we are done.

Problem 7: This is obvious from problem 5; if we look at h(z) = f −^1 (g(z)), then h is a conformal automorphism of the unit disk with h(0) = 0 and h′(0) = g′(0)/f ′(0), which is real and positive. So then we can conclude that h is the identity, by problem 5.

Problem 8: Let g be the Riemann map from G to the unit disk. Suppose g(z 0 ) = w 0. Let φ(z) = 1 z−−ww 00 z , and let h(z) = φ(g(z)). Then h(z 0 ) = 0, and h is still a Riemann map from G to

the unit disk (why?). Let k = h−^1. Then k(0) = z 0. Now we just let ℓ(z) = k(λz) for some λ on the unit circle to adjust the argument of the derivative at 0. (So if arg k′(0) = θ, simply let λ = ei(θ^0 −θ).)

Problem 9: First note that

sin^3 x =

(eix^ − e−ix)^3 − 8 i

8 i

(e^3 ix^ − e−^3 ix) +

8 i

(eix^ − e−ix) =

sin x −

sin 3x

and then consider f (z) =

(^34) eiz (^) − 14 e 3 iz z^3.^ Integrating around the contour on p.^ 122, we get from Cauchy’s theorem that the whole integral is 0, and the sum of the two integrals on the real axis is ∫ (^) R

ǫ

3 4 e

ix (^) − 3 4 e

−ix (^) − 1 4 e

3 ix (^) + 1 4 e

− 3 ix x^3

dx

(watch those negative signs!) So the limit as ǫ → 0 +^ and R → ∞ of this sum is ∫ (^) ∞

0

2 i

sin^3 x x^3

dx.

Standard arguments show that the integral around CR tends to 0. I’ll leave them to you; they’re exactly the same as in the book. (You’ll have to demonstrate on the exam that you can do something like this, though!) What about the integral around Cǫ? Well, the integral of a function which is holomorphic near 0 around Cǫ must tend to 0, so we just need to figure out the principal part of the Laurent series of our function and integrate that. A computation shows that 3 4 e

iz (^) − 1 4 e

3 iz

z^3

2 z^3

4 z

  • g(z),

where g(z) is holomorphic near 0. Now, the integral of (^21) z 3 around Cǫ equals 0 by the Fundamental Theorem of Calculus (the antiderivative is even, and the endpoints of the arc are negatives of each other!) And the integral of (^43) z around Cǫ is just −^3 πi 4 , either by the Fundamental Theorem of Calculus or direct parameterization. (Check this!) Putting it all together yields

2 i

0

sin^3 x x^3

dx −

3 πi 4

and so we are done.

2