Solution to Homework 4, Math 7651, Tanveer

1. Show that the function w(z) = 21 (z + 1/z) maps the exterior of a unit circle
centered around the origin in the z-plane to the exterior of a straight line cut
from −1 to 1 in the w-pane. What is the inverse of this mapping. Be specific
on choice of branch if the inverse function involves branches.
Solution: If we express z = reiθ , then the domain of interest in the z-plane
corresponds to r > 1. We note than that
   
1 1 1 1
w = ξ + iη = r+ cos θ + r− sin θ (1)
2 r 2 r

So, r = const > 1 corresponds ellipses with major axis r + 1r and minor axis
r − 1r . We note that as r → ∞, w → 12 z and as r → 1+ , w → cos θ, and therefore
the boundary |z| = 1 is mapped to the slit [−1, 1] in the w-plane. We note that
since 2zw = z 2 + 1, solving the quadratic and seeking the root that matches
asymptotic behavior w ∼ z2 , as z → ∞, we have
p
z = w + w2 − 1 (2)

where we define
 
p i i
w2 − 1 = |w2 −1| exp arg(w − 1) + arg(w + 1) , with arg (w±1) ∈ (−π, π]
2 2
(3)
2. Solve the potential problem ∆φ = 0 in the strip region 0 < y < 1 with
boundary condition φ(x, 1) = 1, φ(x, 0) = 0. What a priori assumption is
needed on the growth of φ as x → ±∞ to assure that there is only one solution
to this problem.
Solution: Clearly on inspection φ(x, y) = y is a solution of ∆φ = 0 and satisfies
stated boundary conditions. We now address the question fo uniqueness. If φ̃
is another solution, we then have for Φ = φ̃ − φ satisfy ∆Φ = 0 with homoge-
nous boundary conditions Φ(x, 0) = Φ(x, 1) = 0. Uniqueness revolves around
question of use of maximimum principle. Let Ψ be the harmonic conjugate of
Φ. Then, consider the analytic function

χ = eΦ+iΨ , (4)

We note that maximimum principle applied to χ and 1/χ relate to maximimum

principle and minimum principles for Φ. Now, note with z = x + iy,

ζ = ζ(z) := eπz (5)

maps the strip in the z-plane to the upper-half plane, with {z : =z = 0} mapped
to {ζ : =ζ = 0, <ζ > 0} and {z : =z = 1} mapped to {ζ : =ζ = 0, <ζ < 0}. In
order to apply the Phragmen-Lindeloff principle to χ, we have to assume χ is
not singular at ζ = 0 (corresponding to z = −∞ and that that
β βπx
χ ≤ eC|ζ| = eCe (6)

1
for some β ∈ [0, 1). This means that we have to assume we are in the class of
functions for which
φ(x) ≤ Ceβπx as x → ∞ (7)

Now consider the a priori condition on Φ + iΨ that rules out a singularity at

ζ = 0. Claim it is enough to require a priori that Φ + iΨ is weakly singular at
ζ = 0. To prove this consider γ(ζ) = iζ(Φ + iΨ). it is analytic in the half-plane
and continuous upto the real axis, including ζ = 0. Since =γ = 0, Schwartz
reflection principle implies that γ is an analytic function in the lower-half plane
as well, and in particular analytic at ζ = 0. But from a priori condition on
weak singularity of Φ + iΨ, it follows that γ(0) = 0, and therefore Φ + iΨ is
analytic at ζ = 0. Now, we claim that the a priori weak singularity hypothesis
of Φ + iΨ is implied by requiring

Φx , Φy ≤ Ce−βπx as x → −∞ (8)

for β ∈ [0, 1). To prove this we note that from Cauchy Riemann condition
d
dz (Φ + iΨ) = Φx − iΦy . Therefore, using ζ = eπz , ζ 0 (z) = eπx , (8) may be
replaced by
d C
(Φ + iΨ) ≤ β+1 (9)
dζ |ζ|
near ζ = 0 (i.e. x = −∞). On integration, we obtain

C
Φ + iΨ ≤ (10)
|ζ|β

which gives the a priori weak singularity condition. Thus the a priori condition
on the class of solution φ to the originally posed problem should include

φx , φy ≤ Ce−βπx as x → −∞ (11)

for some β ∈ [0, 1), in addition to (7).

Note: It is possible to replace (8) by |Φ| ≤ Ce−βπx , by using a more refined
argument. This is expected from the symmetry of the problem between x = −∞
and x = +∞ and the result (7)
3. Solve the potential problem for φ(x, y)

∆φ = 0 in y > 0

with
∂φ
φ(x, 0) = 1 for x > 1, φ(x, 0) = −1 for x < −1 and (x, 0) = 0 for|x| < 1
∂y
State and prove conditions that will make the solution you find unique. Hint:
Think of mapping to a semi-infinite strip and Schwarz Reflection Principle.

2
Solution: We want to map the upper-half z = x + iy plane H into the semi-
infinite half-strip
S := {ζ : =ζ > 0 : <ζ ∈ (−1, 1)}
We already know from class that the map from H to the rectangular domain

Sc := {ζ : c > =ζ > 0 : <ζ ∈ (−1, 1)}

with z = ±1 corresponding to ζ = ±1 and z = ±1 + ic corresponding to ζ = ± k1

is given from Scwartz reflection principle to1
Z z
dz 0
ζ=A p p
0 1 − z 0 2 1 − k2 z 0 2

and that in the limit k → 0+ , c → ∞ corresponding to the semi-infinite strip

S. Therefore, for a strip. ζ = A sin−1 z. Or z = sin A z
. In order for z = ±1
2 πζ
to correspond to ζ = ±1, A = π , or z = g(ζ) = sin 2 maps S into H. We
can check in particular that real axis segement [−1, 1] is mapped to real axis
segment [−1, 1] This map is conformal (i.e. preserves angles) except at ζ = ±1
where g 0 = 0. Therefore, the boundary condition

∂y φ(x, 0) = 0 for x ∈ (−1, 1)

becomes in the ζ = ξ + iη plane the following equation

∂η Φ(ξ, 0) = 0 for ξ ∈ (−1, 1) (12)

where Φ(ξ, η) = φ (x(ξ, η), y(ξ, η)) where x(ξ, η) and y(ξ, η) is defined by x+iy =
g(ξ + iη). Translating the conditions on φ on different parts of ∂H, the other
boundary conditions on Φ(ξ, η) in S are

Φ(−1, η) = −1 , Φ(1, η) = 1 for η > 0 (13)

We note that Φ(ξ, η) = ξ solves ∆ξ,η Φ = 0 as well as satisfy boundary conditions

(12), (13). This implies
2  −1
φ(x, y) = ξ(x, y) = <g −1 (x + iy) = < sin (x + iy) (14)
π
To explore whether or not this is the only solution take the difference between
two solutions Φ̃ in S. Then, define analytic function W so that

W (ξ + iη) = Φ̃(ξ, η) + iΨ̃(ξ, η)

Therefore, if we assume that the solution we are seeking has continuous first
derivatives as η → 0+ for ξ ∈ (−1, 1), then from Cauchy Riemann conditions
for ξ ∈ (−1, 1)
∂η Φ(ξ, 0) = ∂ξ Ψ(ξ, 0)
1 Variables z and ζ are switched here from class notes

3
implying
Ψ̃(ξ, 0) = 0 without loss of generality
From Schwartz reflection principle W (ζ) extends to an analytic function to the
entire-strip
S := {ζ : −1 < <ζ < 1}
Further, since the solutions are assumed continuous upto the boundary, then
repeated reflection using Schwartz reflection principle gives W to be an entire
function of ζ, and therefore using Phragmen-Lindeloff principle on eW and e−W ,
it follows that if we assume |Φ(ξ, η)| ≤ c1 + c2 |ξ 2 + η 2 |α/2 , i.e. α ∈ [0, 1), then
the difference of two solutions Φ̃(ξ, η) = 0, and we have uniqueness.
4. a. Find a Mobius transformation that maps the unit disk centered at 0
to the upper-half plane. b. Find a 1-1 analytic mapping of the interior of
|z − 1 − 2i| = 2 to the interior of the unit disk centered at i/2.
z+1
Solution: Consider ζ = f (z) = −i z−1 . It is readily checked that f (eiθ ) =
− cot θ2 maps the unit circle to the real line with z = ±1 corresponding to
ζ = ∞ and ζ = 0 respectively. This is a fractional linear map and therefore 1-1
on inspection. Further, we may check that f (0) = i confirming that f : D1 → H,
where H is the upper-half plane.
b. Find a 1-1 analytic mapping of the interior of |z − 1 − 2i| = 2 to the
interior of the unit disk centered at i/2.
Solution Note S0 (z) = 12 (z − 1 − 2i) maps D, which is the interior of the
circle |z − 1 − 2i| of radius 2 into the D1 , the unit circle centered at the origin.
Further note that S1 (w) = w − 2i also maps the unit circle centered at i/2 into
D1 . Therefore, it is clear that

f ≡ S1−1 ◦ S0 : D → D1

in a 1-1 analytic manner. So

i 1
f (z) = + (z − 1 − 2i)
2 2
5. Extend the proof of solution to Theodersen’s integral equation to the space
Ḣ1 , i.e. space ofq2π-periodic functions f so that both f and f 0 ∈ L2 (0, 2π) with
norm kf kḢ1 = kf k2L2 + kf 0 k2L2 .
Solution: Recall that if we define φ(ν) = ν + χ(ν), then χ is 2π periodic, and
Theodorsen’s integral equation becomes

χ(ν) = H [ρ] (ν) (15)

where
ρ(ν) = log [R (ν + χ(ν))] (16)
It is convenient to define

ρ1 (ν) = log [R (ν + χ1 (ν))] , ρ2 (ν) = log [R (ν + χ2 (ν))] (17)

4
Then, if |R0 (ν)| ≤  and R(ν) ≥ R0 , we note from meanvalue theorem that

ρ1 (ν) − ρ2 (ν) ≤ χ1 (ν) − χ2 (ν) (18)
R0
Also, on taking derivative we have
 
0 0d
ρ (ν) = (1 + χ (ν)) log R (19)
dφ φ=ν+χ

Defining
0 0
(log R) (ν + χ1 ) = P1 (ν) , (log R) (ν + χ2 ) = P2 (ν) , (20)
We note that both P1 and P2 satisfy

|P1,2 (ν)| ≤ (21)
R0
while if we define
d2
M= sup log R , (22)
φ∈[0,2π] dφ2
then from mean-value theorem ,

|P1 (ν) − P2 (ν)| ≤ M |χ1 (ν) − χ2 (ν)| . (23)

From (19), it follows that

ρ02 (ν) − ρ01 (ν) ≤ |P2 | χ02 (ν) − χ01 (ν) + |P2 − P1 ||1 + χ01 (ν)|
 0
≤ χ (ν) − χ01 (ν) + M (1 + |χ01 |) χ2 (ν) − χ1 (ν) (24)
R0 2
Recall Hilbert transform operator commutes with derivative and that for any
ψ ∈ L̇2 , kH[ψ]kL2 ≤ kψkL2 . Therefore, the estimate (24) implies that for χ1 ,
χ2 in a ball of size B around the origin in the Ḣ1 space,

d 
k H [ρ2 − ρ1 ] kL2 ≤ kχ0 − χ01 kL2 + M (1 + B)kχ2 − χ1 k∞
dν R0 2

≤ kχ0 − χ01 kL2 + CM (1 + B)kχ2 − χ1 kḢ1 , (25)
R0 2
where we used the fact that in 1-D there exists some constant so that kχk∞ ≤
CkχkḢ1 . From class notes, we already know that

kρ2 − ρ1 kL2 ≤ kχ2 − χ1 kL2 (26)
R0
Therefore, it follows from (25)-(26) that in a ball of size B,

kH [ρ2 ] − H [ρ1 ] kḢ1 ≤ 1 (1 + B) kχ1 − χ2 kḢ1 , (27)

5
where  

1 = max , CM (28)
R0
If we write the Theodorsen integral equation as

χ = H [ρ0 ] + H [ρ − ρ0 ] , (29)

where
ρ0 = log R(ν) (30)
it follows from the estimate (27) that in a Ball B = 2kρ0 kḢ1 for sufficiently
small 1 , we have a contractive map in the Banach space Ḣ1 and therefore the
integral equation (29) has a unique solution.
Note that the assumption of small 1 implies that not only does R0 and R00
exist and R has a lower limit R0 that stays away from 0, but also that R00 is
not large.