hmwk4 Sol
hmwk4 Sol
hmwk4 Sol
Solution: I’ll just make a comment on this one. You close the contour
with a semicircle in either the upper or lower half plane depending on the
sign of ǫ. Most people worked out both cases. But note that the change of
variables x → −x in the original integral shows that the integral is equal to
the integral with ǫ replaced by −ǫ. In other words the integral is an even
function of ǫ. So you only have to compute it for one case, e.g., ǫ ≥ 0.
Solution: Inside the given circle there is a second order pole at −u and
first order poles at k for |k| ≤ N. The residue from the pole at −u leads to
the π 2 / sin2 (πu) and the sum of the other residues leads to the partial sum
X
N
1
n=−N
(u + n)2
This part was pretty straightforward and I won’t write it out. The harder
part is showing that the integral around the contour converges to zero as
N → ∞.
We will show that there is a constant M (independent of N) such that
| cot(πz)| ≤ M for all N and |z| = N + 1/2.
e2iπx−2πy + 1
i
e2iπx−2πy − 1
Since cot(πz) is an odd function, we need only consider z in the upper half
plane, ie., y ≥ 0 . If x is within 1/4 of N + 1/2 for some integer N, then
cos(2πx) ≤ 0. So the real part of e2iπx−2πy is ≤ 0. Hence |e2iπx−2πy − 1| ≥ 1.
The numerator is trivially bounded in modulus by 2. So | cot(πz)| ≤ 2 on
the vertical strips given by |Re(z) − (N + 1/2)| < 1/4 for some N. The parts
1
of the circle that do not lie in these strips are bounded away from the real
axis, i.e., there is a δ > 0 such that they lie in {z : |Im(z)| ≥ δ}. For y ≥ δ,
Let
Prove that
1
= lim sup |an |1/n
R2 n→∞
Solution: This problem was so poorly stated that everyone missed the
point of I was trying to get at. So I gave everyone full credit on the problem
and wrote comments on most papers. I will reassign this problem (hopefully
better stated) in the problem set after the midterm. Here is an attempt at a
better statement of the problem:
Suppose that f is analytic on the annulus {z : ρ1 < |z| < ρ2 }. From what
we did in class we know that it has a Laurent series of the form
X
∞
f (z) = an z n
n=−∞
2
meaning that the series converges to f (z) on the annulus. Moreover the
convergence is absolute.
Define
1
= lim sup |an |1/n
R2 n→∞
4. Let f and g be analytic on an open set containing the closed disc |z| ≤ 1.
Suppose f has a simple zero at z = 0 and has no other zeroes in the closed
disc. Define for complex w,
Prove that there is an ǫ > 0 such that for |w| < ǫ, fw has a unique zero zw
in the closed disc and the mapping w → zw is continuous.
Solution: Let M = sup |g(z)|, m = inf |f (z)| where the sup and inf
are over the unit circle. Since the circle is compact and both functions are
continuous, the sup and inf are attained and so M < ∞ and m > 0. Assume
M > 0. (Otherwise g vanishes on the unit circle and so must be the zero
function.) Define ǫ = m/M. Let w be such that |w| < ǫ. We apply Rouché’s
theorem to the functions f (z) and wg(z). We have |wg(z)| < ǫM = m and
|f (z)| ≥ m on the circle, so |wg(z)| < |f (z)| on the circle. By the theorem
f (z) and f (z) + wg(z) have the same number of zeros inside the circle. So
f (z) + wg(z) has exactly one zero, zw .
To prove zw is continuous in w, fix w with |w| < ǫ and let wn be a
sequence converging to w. Suppose zwn does not converge to zw . Then there
3
is an ǫ > 0 and a subsequence of zwn whose distance to zw is always at least ǫ.
This subsequence is in the unit disc, a bounded set, so it has a subsequence
which converges to something, call it z ′ , in the closed unit disc. Note that z ′
cannot be zw . Let uk be the corresponding subsequence of the subsequence of
wn . So zuk converges to z ′ . Now wuk converges to w, so fwuk (zuk ) converges
to fw (z ′ ). But fwuk (zuk ) = 0 for all k, so z ′ is a root of fw . By the bounds
in the first paragraph, fw is not zero on the boundary of the disc, so z ′ is
in the open unit disc. But fw has only one root in this disc, and it is zw .
Contradiction.
Solution: Let
X
∞
f (z) = an z n
n=−∞
be the Laurent series of f for |z| > R. Then for |z| < 1/R,
X
∞
f (1/z) = an z −n
n=−∞
So
X
∞
−2
−z f (1/z) = − an z −n−2
n=−∞
The residue of this is the coef of the 1/z term. This is when n = −1 and the
residue at ∞ is −a−1 .
(b) What is the relation of the residue at ∞ to the integral
Z
1
f (z)dz
2πi γ
4
Solution: The Laurent series converges uniformly on the circle of radius r,
so we can integrate it term by term to get
Z
1
f (z)dz = a−1
2πi γ
Solution: The contour in part (b) encloses all the singularitesPmwith winding
number 1, so by the residue theorem the integral in part (b) is k=1 Res(f, zk ),
and the equation in (c) follows.
6. Let
cos z
f (z) =
z(1 + z 2 )
Solution:
z2 z4
cos(z) = 1 − + −···
2! 4!
1 1 1 1
= 1− 2 + 4 − 6 ···
1+z 2 z z z
So the power series of cos z/(1 + z 2 ) about the origin is
z2 z4 1 1 1
(1 − + − · · ·)(1 − 2 + 4 − 6 · · ·)
2! 4! z z z
The z 5 term in the Laurent series we want will be 1/z times the z 6 term in this
power series which works out to be (−1 − 1/2 − 1/24 − 1/720)z 5 = − 1111 720
z5 .
−5
(b) Find the z term in the Laurent series of f in the annulus
{z : 1 < |z| < ∞}.
5
Solution: If we take r > 1, the coefficient of z 5 in the Laurent series is
given by
Z Z
1 4 1 z 3 cos(z)
z f (z)dz = dz
2πi |z|=r 2πi |z|=r 1 + z2
You can compute this integral by the residue theorem and find cos(i) =
cosh(1).
Solution: Let h(z) = f (z)/g(z). Since g does not vanish on D = {z : |z| <
R}, h is analytic on D. On the boundary of D we have |h(z)| = 1. So by the
maximum-modulus theorem we have |h(z)| ≤ 1 on D, i.e., |f (z)| ≤ |g(z)| on
D. The same argument with f and g interchanged shows |g(z)| ≤ |f (z)| on
D. So |g(z)| = |f (z)| on D. Thus the analytic function h attains it maximum
modulus at an interior point (in fact at every interior point) and so must be
a constant. So f (z) = cg(z) and |c| must be 1.