Homework 4 Sol
Homework 4 Sol
1. Ahlfors pg. 88 #1: Find the fixed points of the linear transformations
z
2z
,
w=
.
3z 1
2z
Is any of these transformations elliptic, hyperbolic, or parabolic?
w=
Solution: For the first one, we find the fixed points by solving z(3z 1) = 2z, or
3z 2 3z = 0, so we get z = 1 and z = 0. Now express
w0
=
w1
2z
3z1
2z
3z1
2z
2z
=
.
2z (3z 1)
z1
z
2z
z
2z
2(z 1)
z (2 z)
=
,
z
z
Solution: Observe first that we can compute k as S 0 (z ) where z is either fixed point.
This is because if we write w = S(z) when there are two distinct fixed points and ,
we get
z
S(z)
=k
,
S(z)
z
and thus
S 0 (z) S(z)
S 0 (z)
1
z
.
2 = k
S(z)
z (z )2
S(z)
Hence plugging in z = gives S 0 () = k. Similarly we get S 0 () = 1/k.
Now lets compute those fixed points; we obviously have az + b = (cz + d)z with
solutions
p
a d (d a)2 + 4bc
,
z =
2c
assuming that c 6= 0.
First of all, if c = 0 then the fixed points are and z =
b
,
da
and we have
b
a(d a)z + b(d a) bd
a(d a)z ba
a
b
az + b
wz =
=
=
=
z
,
d
da
d(d a)
d(d a)
d
da
so that k = a/d. Now ad = 1 by the normalization assumption, so that k = a2 .
1
Hyperbolic: k is real and positive if and only if a is real and then a + d = a + 1/a
which is either larger than 2 (if a > 0) or smaller than 2 (if a < 0).
Elliptic: k has unit modulus if and only if a does, and therefore a = ei for some
, which means we have a + d = a + 1/a = 2 cos , which is in (2, 2). Note that
the only way |a + 1/a| = 2 is if a = 1 or a = 1, but in either case we have d = a
and thus the transformation is parabolic.
Parabolic: the fixed points are equal if and only if a = d (so that they are both
infinity), and since d = 1/a we must have a = 1 so that a + d = 2.
Next consider the case when c 6= 0. We then have (using the normalization adbc = 1)
that
p
p
2 + 4(ad 1)
(d
a)
a
(a + d)2 4
a
=
.
z =
2c
2c
Now S 0 (z) = 1/(cz + d)2 so that
k 1 = S 0 (z ) =
d+
ad
(a+d)2 4
2
2 =
4
1
,
p
2 =
( 2 1)2
a + d (a + d)2 4
where = (a + d)/2. Note that the conditions of being hyperbolic or elliptic (depending on k) are symmetric with respect to k 7 1/k, so we can consider whichever is
convenient. For me 1/k will be more convenient. Notice that the fixed points are equal
if and only if = 1, which is the parabolic case; we will therefore assume 6= 1.
x2 + 1
.
2x
1 + e2i
= cos .
2ei
Furthermore the number k must have magnitude 1, since otherwise under iteration
of S points would converge either to a or b (and this certainly does not happen for
rotations). Hence the LFT must be elliptic.
It remains to show that every such elliptic LFT is a rotation. But observe that a
map fixing two points on the sphere is a rotation if and only if it preserves all circles
equidistant between the two fixed points (that is, the latitudes) and moves all circles
through the two fixed points (the longitutes) to each other, and thats exactly what
an elliptic LFT does (by the general properties of the C1 and C2 Steiner circles).
Hence the general LFT which corresponds to rotation must have the form w = S(z)
za
wa
= ei
w + 1/a
z + 1/a
for some a C and R. Of course we can solve this for w if desired.
4. Ahlfors pg. 96 #1: Find a conformal map from the common part of the disks |z| < 1
and |z 1| < 1 on the inside of the unit circle. Choose the mapping so that the
horizonal and vertical reflection symmetries are preserved.
3
2
= ei/3 .
z ei/3
z ei/3
zi
.
z+i
Solution: The singular endpoints are 1 and +1, so using the usual trick we send 1
to and 1 to 0 using
z1
.
S1 (z) =
z+1
This map will send the upper semicircle to a ray from 0 to . Since S(i) = i this ray
is the upper imaginary axis. Taking the square root gives a half-plane; our angles will
be from 3
to 2 and the square root will give angles from 3
to 4 . To get this onto
2
4
3i/4
the upper half-plane we multiply by e
. So
S2 (z) = e3i/4 z.
Now there are a variety of ways to get the upper half-plane onto the unit disc (or its
complement), and we need to figure out what happens to . We have S1 () = 1 and
S2 (1) = e3i/4 , so we need a map that sends the upper half-plane to the complement
of the unit disc and sends e3i/4 to . We can do this with an LFT; such a map must
necessarily send e3i/4 to 0 (since it preserves conjugation through circles). Thus
S3 (z) = k
z e3i/4
z e3i/4
for some k C. Since S3 (0) = ki we should choose |k| = 1 to end up on the unit circle;
k = 1 would be fine.
6. Show that every linear fractional transformation that maps the unit disc onto itself is
of the form
z+a
,
R, |a| < 1.
Sz = ei
1 + az
(One way is to do what I said in class and reduce it to the case of an LFT that maps
R to itself. A smarter way is to figure out what must happen to 1, 0, and under S.)
Solution: An LFT that preserves the unit disc must take 0 to some C with
|| < 1. By writing S(z) = ei S(z) we may assume that S(0) = is positive and real
(and obviously less than 1). As a consequence we must have S() = 1/ since is
the conjugate of 0 through the circle and 1/ is the conjugate of .
We must therefore have S(z) = (az + b)/(cz + d) where b = d and a = c/, so we can
write
az + d
S(z) =
.
az + d
Write a = d and we obtain
z +
S(z) =
.
z + 1
Since |S(1)| = 1 we must have
| + |2 = | + 1|2 ,
or in other words we have
||2 + 2Re() + 2 = 2 ||2 + 2Re() + 1.
4
S(z) =
and we conclude that
S(z) = ei
i
ei z +
i(+) z + e
=
e
,
ei z + 1
ei z + 1
ds2 =
2 |v0 |E
,
1 |z0 |2
where |v|E is its usual Euclidean length. Show that this metric is invariant under any
mapping of the form S in problem 6, in the sense that
|S 0 (z0 )(v0 )|P = |v0 |P ,
where the left side is evaluated at S(z0 ) and the right side is evaluated at z0 .
z+a
Solution: With S(z) = ei 1+az
, we have
S 0 (z) = ei
1 |a|2
.
(1 + az)2
We therefore have
|S 0 (z0 )(v0 )|2S(z0 ) =
|S 0 (z0 )v0 |2
(1 |S(z0 )|2 )2
(1|a|2 )2 |v0 |2
|1+az0 |4
|z0 +a|2 2
1 |1+az
2
0|
2 2
(1 |a| ) |v0 |2
.
|1 + az0 |2 |z0 + a|2 )2
|v0 |2
= |v0 |2P .
(1 |z0 |2 )2