Solutions Practice Final
Solutions Practice Final
Hint. A holomorphic function has a primitive if the integral on any triangle in the domain
is zero.
Solution: Call the given function f (z). Consider a triangle in . Either it encloses
the circle |z| = 2 or it doesnt. In the latter case, f (z) is holomorphic in a neighborhood
of the triangle and so by Goursats theorem the integral of f over such a triangle is
zero. In case the triangle, which we denote by T , does enclose the circle |z| = 2, then
by generalized Cauchys theorem
Z Z
f (z) dz = f (z) dz,
T |z|=2
where the circle is given positive orientation. Since the given circle encloses all poles
of f (z) when thought of as a meromorphic function, again by generalized Cauchys
theorem, we see that for any R > 2,
Z Z
f (z) dz = f (z) dz.
|z|=2 |z|=R
But the latter integral goes to zero as you let R since the function grows like
1/R2 and the length of the circle grows like R. So the original integral must have
been zero. Hence in both cases integrals on triangles is zero. Hence there must exist
a primitive.
Remark. To show that the integral on |z| = 2, you could have also calculated the
residue at infinity, and shown that that residue is zero.
z2
?
(z 1)(z 2 + 1)
1
Solution: No. If there was a primitive, then since the circle |z| = 3 lies in we would
have to have Z
f (z) = 0.
|z|=3
On the other hand the integral equals Resz= f (z) = Resz=0 f (1/z)/z 2 . But
1 1 1
Resz=0 f (1/z)/z 2 = lim z 2
f = lim = 1 6= 0.
z0 z z z0 (1 z)(1 + z 2 )
Solution: Since f (z) is entire it is given by a power series that converges on the entire
complex plane to the function. We show that f n (0) = 0 for all n 1. This would show
that the the power series just a constant. By CIF, for any radius R > 1,
Z
(n) n! f (z)
f (0) = dz,
2i |z|=R z n+1
|f (z)|
Z
n!
|f (n) (0)| |dz|
2 |z|=R |z|n+1
n! (1 + R)1/2
C 2R. (Since on |z| = R, by given estimate, |f (z)| (1 + R)1/2 )
2 Rn+1
R R
2Cn! n 0,
R
as long as n 1. This completes the proof.
3. If f : D C is defined by
ew cos w
Z
1
f (z) = dw,
2i |w|=1 (w z)2
where the circle is given the positive orientation, evaluate f 0 (0).
2
So f 0 (z) = 2ez sin z, and so f 0 (0) = 0.
Solution: The integral can be evaluated using the residue theorem since tan z is a mero-
morphic function with the only poles inside |z| = 2 being at z = /2 and z = /2. But
we evaluate it using the argument principle.
(cos z)0
tan z = sin z/ cos z = ,
cos z
and so
Z
tan z = 2i number of zeroes of cos z inside |z| = 2 = 4i.
|z|=2
For the second integral, inside |z| = 2, the function has a simple pole only at z = /2.
But
Resz=/2 z /2 tan z = lim z + /2 z /2 tan z = ,
z/2
Solution: The poles of cot z are at the zeroes of sin z, namely all the integers. The
poles are all simple, and so the residue is given by
cos z zn 1
Resz=n cot z = lim (z n) = cos n lim = ,
zn sin z zn sin z
where we used LHospitals rule in the last equation.
3
since the degree of the denominator is at least two more than the degree of the numerator.
The integral is of type III(a). So consider a contour made up of a large semi-circle of radius
R traversed in the anticlockwise direction (denoted by CR ) and a straight line path from
x = R to x = R along the x-axis, and the complex valued function
zeiz
f (z) = .
(z 2 + 1)2
The only pole of the function inside the contour is at z = i, and ordz=i f (z) = 2. By the
residue theorem,
R
xeix zeiz
Z Z
dx + dz = 2iResz=i f (z).
R (x2 + 1)2 CR (z 2 + 1)2
7. Evaluate
x2 + 1
Z
dx.
0 x4 + 1
Solution: This is standard of Type II. Since the integrand is even, the integral we need,
which we denote by I is
R
x2 + 1 x2 + 1
Z Z
1
I= dx = lim dx.
2 x4 + 1 R R x4 + 1
Again, using the same contour as above, since the degree of the denominator is two bigger
than the degree of numerator, the integral on the semi-circular will approach zero as R
and we obtain X
I = i Resz=p f (z),
p an interior pole
where
1 + z2
f (z) = .
1 + z4
The poles of the function are at k = ei/4+ik/2 , k = 0, 1, 2, 3. Since the contour is in the
upper half plane, the only poles in the interior of the contour are 0 = ei/4 and 1 = e3i/4 .
4
Since both poles are simple, the residues can be calculated using LHospitals rule, namely
(z k ) 2 1 + k2
Resz=k = lim (1 + z ) = .
zk 1 + z4 4k3
8. Let {fn } be a sequence of entire functions that converges compactly to a f with f not identically
zero. If all fn have only real roots, show that all roots of f must be real.
Solution: This is a direct consequence of the generalized Hurwitzs theorem from the final
homework (Why?). We give a direct proof. Let p be a zero of f , whcih must be isolated.
Suppose p is not real. Let r > 0 such that Dr (p) does not contain any other root and does
not intersect the x-axis. Denoting the boundary circle by C, by the argument principle,
f 0 (z)
Z
1
dz 1,
2i C f (z)
But the integrals on the right are integers since they also count the number of zeroes of fn
inside C. But then there must exist an N large enough so that the integrals on the right
are greater than or equal to one for n > N . This is a contradiction since all roots of fn are
real and the real axis does not intersect Dr (p).
9. Suppose f (z) is holomorphic in an open set containing the closure of the unit disc {|z| < 1},
such that |f (z)| < 1 whenever |z| = 1. Show that the equation f (z) = z 3 has exactly three
solutions (counting multiplicities) inside the unit disc.
5
Hence h and g have same number of roots in |z| < 1 counting multiplicities, which is three.
10. Does there exist a holomorphic function on the unit disc D such that
(
1
1 , n is even
f = n1
n n+1 , n is odd?
Solution: No. Since f (1/2m) = 1/2m for natural numbers m, and since 1/2m 0
as m , by the principle of analytic continuation, f (z) = z. But his contradicts
f (1/(2m + 1)) = 1/(2m + 2).
11. Let f be holomorphic on a domain containing the closure of the unit disc, such that |f (z)| 5
for all |z| = 1. If f (0) = 3 + 4i, find f 0 (0).
Solution: We claim that f is constant, and hence f 0 (0) = 0. To see this, note that
|f (0)| = 5 |f (z)| for all |z| = 1. This contradicts the maximum principle unless f (z) is a
constant.
Conformal maps
12. Find a fractional linear transformation that maps the circle |z 1| = 1 onto the x-axis.
Solution: A circle is characterized by three points on it, which in this case, we can take
as z = 0, z = 2 and z = 1 + i. The x-axis is specified by w = 0, w = and w = 1. So the
required fractional linear transformation is
z i1
Tz = .
z2 i+1
Remark It is a good exercise in manipulating complex numbers and conjugates to show
that T z is purely real if |z 1| = 1. To see this note that |z 1| = 1 is equivalent to
|z|2 = z + z (Why?).
6
Solution: First map the disc to upper half plane by
1+z
zi ,
1z
and then rotate by /2. So
z+1
(z) =
z1
Solution: A biholomorphism of the unit disc that sends zero to zero is a rotation.
That is
f (z) = ei z.
But then = /2.
(c) {z | 0 < arg(z) < 3/2} and {z | 0 < arg(z) < /2}.
Solution: (z) = z 1/3 = elog z/3 , where we use the branch cut as the positive real axis
(that is, we restrict 0 < argz < 2).
(d) {z | |z| < 1, Im(z) > 0} and {z | |z| < 1}. Hint. Note that you cannot simply use z 2 .
Find a fractional linear transformation that will map the given region conformally onto
the first quadrant {z | Re(z) > 0, Im(z) > 0}.
7
Gamma and Zeta functions
14. (a) Use a change of variables to show that for all t > 0,
1 Z
=t evt (vt)1/2 dv.
2 0
Solution: We will just prove the last part. Let us first assume that 0 < s < 1. Then
Z
(1 s) = ex xs dx.
0
For t > 0, consider the change of variables x = tv. Then dx = tdv and so
Z
(1 s) = t etv ts v s dv.
0
So
Z
(s)(1 s) = et ts1 (1 s) dt
Z0 Z
= et(v+1) v s dt dv
Z0 0 s
v
= dv.
0 v+1
In class we showed (Type IV integrals; integrating around branch cuts and residue theorem)
that Z s
v
dv = .
0 1+v sin s
Plugging in s = 1/2 gives
(1/2)2 = ,
and so (1/2) = .
8
Miscellaneous
Solution: 0 = (ef (z) )0 = f 0 (z)ef . Since ef is never zero, f 0 (z) = 0 on , and hence f is
constant.
16. If f (z) is an entire function such that |f (z)| as z , show that there exists a constant
c > 0 and R > 0 such that
|f (z)| > c|z|
for all |z| > R.
Solution: This directly follows from the fact that any entire function with a pole at infinity
is a non-constant polynomial. But we can give a direct proof. Since f (z) has a pole at
infinity, g(w) = f (1/w) has a pole at w = 0 of order M 1. In particular, there exists
c > 0 such that
c
|g(w)|
|w|m
c
for all |w| < r for some r > 0. We can choose r < 1 so that, |w|m < |w|, and so |g(w)| > |w|
on |w| < r. Or equivalently
1 c
f > ,
w |w|
for |w| < r. Putting z = 1/w, this is the same as saying