Further Complex Methods, Examples sheet 3 (C8c) Lent Term 2025

Mathematical Tripos Part II - C Course Professor Peter Haynes

Comments and corrections: e-mail to phh1@cam.ac.uk.
Starred questions or parts of questions are intended as extras: attempt them if you have time,
but not at the expense of unstarred questions.

1 (a) Prove that for Rez > 1,

Z ∞ z−1 Z
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

1 t Γ(1 − z) tz−1
dt = dt,
Γ(z) 0 et − 1 2iπ γ e
−t − 1

where γ denotes the Hankel contour. Hence, deduce that the RHS of the above equation
provides the analytic continuation of Riemann’s zeta function.
(b) The Bernoulli numbers Bn are defined by

1 X
∞
tm−1
= B m ,
et − 1 m!
m=0

and B0 = 1, B1 = − 21 , B2m+1 = 0 for m = 1, 2, . . . .

Use (a) and the residue theorem to compute ζ(−n), n = 0, 1, 2, . . . in terms of Bn . Hence,
deduce that the negative even integers are zeros of ζ(z).

2 Show that for Re z > 1

Z
1−z −z −z −z −z 1 ∞
tz−1
(1 − 2 )ζ(z) = (1 −2 +3 −4 ···) = dt.
Γ(z) 0 et + 1

[Note: This result is actually valid for Re z > 0.]

3 Show that Z (0+)

ln t
dt = 0.
−∞ e−t − 1
Hence show that
lim (ζ(z) − (z − 1)−1 ) = γ,
z→1

and √
ζ ′ (0) = − ln 2π.
 4 The psi-function is defined to be
d
ψ(z) = ln Γ(z).
dz
Show that
X
∞
1
′
ψ (z) = , (z 6= 0, −1, −2 · · · ).
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

(s + z)2
s=0

[Recall Q9 on Example Sheet 2.]

Then show that when z is real and positive, that Γ(z) has a single minimum which lies between
z = 1 and z = 2.

Show that, for |z − 1| sufficiently small,

X
∞
ζ(s)
ln Γ(z) = −γ(z − 1) + (−1)s (z − 1)s .
s
s=2

What is the radius of convergence of this series?

5 Find two independent solutions of the Airy equation w′′ − zw = 0 in the form
Z
w(z) = ezt f (t) dt,
γ

where γ is to be specified in each case. Show that there is a solution for which γ can be chosen
to consist of two straight line segments in the left half t-plane (Re t ≤ 0).
1
For this solution show that, if w(z) is normalised so that w(0) = iA 3− 6 Γ(1/3), where A
1
is a constant, then w′ (0) = −iA 3 6 Γ(2/3).
R∞
[Note: Γ(z) = 0 e−t tz−1 dt for Re z > 0.]

6 By writing w(z) in the form of an integral representation with the Laplace kernel show
that the confluent hypergeometric equation zw′′ + (c − z)w′ − aw = 0 has solutions of the form
Z
w(z) = ta−1 (1 − t)c−a−1 etz dt,
γ

provided the path γ is chosen such that [ta (1 − t)c−a etz ]γ = 0.

In the case Re z > 0, find paths which provide two independent solutions in each of the
following cases (where m is a positive integer):
(i) a = −m, c = 0;
(ii) Re a < 0, c = 0, a is not an integer;
(iii) a = 0, c = m;
(iv) Re c >Re a > 0, a and c − a are not integers.
 7 Use the Laplace transform to solve the ordinary differential equation

d2 y
− k 2 y = f (t), k > 0, y(0) = y0 , y ′ (0) = y0′ .
dt2
Let f (t) = e−k0 t , k0 6= k, k0 > 0, so that the Laplace transform of f (t) is
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

1
fˆ(s) = .
s + k0
Show that
k0
y0′ e−k0 t cosh kt k
y(t) = y0 cosh kt + sinh kt + 2 − + sinh kt. (1)
k k0 − k 2 k02 − k 2 k02 − k 2

Now suppose that f (t) is an arbitrary continuous function that possesses a Laplace
transform. Use the convolution theorem for Laplace transforms, or otherwise, to show that
Z t
y′ sinh k(t − t′ ) ′
y(t) = y0 cosh kt + 0 sinh kt + f (t′ ) dt .
k 0 k

Put f (t) = e−k0 t and re-obtain your answer to the first part of this question. Suppose
now that k0 = k. What is y(t)? Could you have found this solution by taking the limit in (1)
as k0 → k?

8 The Schrödinger equation is

∂u ∂ 2 u
i + 2 = 0.
∂t ∂x
Suppose that u(x, 0) = f (x).
Fourier transform this equation with respect to x to find
Z ∞
e−iπ/4 i(x−x′ )2
u(x, t) = √ e 4t f (x′ )dx′ .
2 πt −∞
R∞ iu2 du iπ √
(You may it useful to recall that −∞ e =e4 π.)
Now use Laplace transform methods to find the same solution to this problem.
 9 ∗ A simple version of the Klein-Gordon equation is

ψtt = ψxx − ψ. (2)

(This equation describes, amongst other things, the propagation of large-scale variations
in the height of the sea surface in the presence of rotation.)
(a) Solve this equation subject to the initial conditions ψ(x, 0) = 0, ψt (x, 0) = δ(x) using
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

Laplace transform methods. Show that, for t < |x|, ψ(x, t) = 0, and, for t > |x|,
Z
1 ds
ψ(x, t) = est exp(−(1 + s2 )1/2 |x|)
2πi γ 2(1 + s2 )1/2

where γ, followed anticlockwise, encloses a branch cut along the imaginary axis from s = −i
to s = i.
(b) Show that, defining the variable w by (t2 − x2 )1/2 w = st − (1 + s2 )1/2 |x|, the above
integral may be transformed to give
Z
1 dw
ψ(x, t) = exp((t2 − x2 )1/2 w)
2πi γ 2(1 + w2 )1/2

with γ defined in the w-plane as in the s-plane.

(c) Show using Laplace’s method that J0 (z), which is the solution of zy ′′ + y ′ + zy = 0
such that y(0) = 1 and y ′ (0) = 0 can be represented as
Z
1 ezs
J0 (z) = ds
2πi γ (1 + s2 )1/2

with γ again as defined above.

(d) Deduce that the solution of (2) specified above for t > |x| may be written as
1
ψ(x, t) = J0 ((t2 − x2 )1/2 ).
2
Draw a sketch of the solution.