Mathematical Tripos Part IB Michaelmas term 2020 Professor E.P.S.

Shellard
Course C10—No. B6

METHODS — EXAMPLES I

Fourier series
1. Fourier coefficients (full-range series). For the periodic function f (x) = (x2 − 1)2 on the interval −1 ≤ x < 1,
show that it has the Fourier series
∞
8 48 X (−1)n+1
f (x) = + cos nπx .
15 π 4 n=1 n4

Sketch the function f (x) and comment on its differentiability and the order of the terms in its Fourier series as
n → ∞.

2. Fourier coefficients (half-range series). Suppose that f (x) = x2 for 0 ≤ x < π. Express f (x) as (a) a Fourier
sine series, and (b) a cosine series, each having period 2π. Sketch the functions represented by (a) and (b) in the
Copyright © 2020 University of Cambridge. Not to be quoted or reproduced without permission.

range −6π to 6π. If the series (a) and (b) are differentiated term-by-term, how are the answers related (if at all) to
the Fourier series for g(x) = 2x and h(x) = 2|x| each in the range [−π, π)?
3. Series summation. Find the Fourier series of the 2π-periodic function f (x) = ex on [−π, π). Deduce that
∞
X 1 1
2
= (π coth π − 1) .
n=1
1 + n 2

4. Parseval’s identity and a low pass filter. (i) Given that a function f (t) defined over the interval (−T, T ) has the
Fourier series
∞   T ∞
1 nπt nπt 1 1 2 X 2
X Z
f (t) = a0 + an cos( ) + bn sin( ) , show that [f (t)]2 dt = a0 + (an + b2n ) ,
2 n=1
T T T −T 2 n=1

where you may assume f (t) is such that this series is convergent.
(ii) A unit amplitude square wave of period 2T is given by f (t) = 1 for 0 < t < T and f (t) = −1, for −T < t < 0.
Suppose this is the input for a system which permits angular frequencies less than 29 πT −1 to be perfectly transmitted
and frequencies greater than 29 πT −1 to be perfectly absorbed. Calculate the form of the output. The power is
proportional to the mean value of f 2 (t); what fraction of the incident power is transmitted?
5. Discontinuities and the Wilbraham-Gibbs phenomenon*. (i) Suppose that f is a 2π-periodic square wave given
by
 ∞
1 0<x<π 1 2 X sin(2n − 1)x
f (x) = Sketch f and show that its Fourier series is + .
0 π < x < 2π . 2 π n=1 2n − 1

N
1 2 X sin(2n − 1)x
(ii) Now define the partial sum of this series as SN (x) = + ,
2 π n=1 2n − 1

x N
1 1 sin 2N t
Z X
and find the following expression SN (x) = + dt . [Hint: consider cos(2n−1)x .]
2 π 0 sin t n=1

(iii) Deduce that SN (x) has extrema at x = mπ/2N, m = 1, 2, ...2N − 1, 2N + 1, ..., (i.e. all integer m except
m = 2kN , k integer) and that the height of the first maximum for large N is approximately
π
π 1 1 sin udu
Z
SN ( )= + (≃ 1·089) .
2N 2 π 0 u

Comment on the accuracy of Fourier series at discontinuities.

[This question takes you through some important steps which are used in the proof of Fourier’s theorem - refer, for
example, to chapter 14 of Jeffreys & Jeffreys. Henry Wilbraham was a fellow of Trinity who wrote a paper (when he
was 22) deriving this result 50 years before Gibbs, after whom it is commonly named.]
 Sturm-Liouville theory
6. Eigenfunctions and eigenvalues. Prove that the boundary value problem
′′ ′
y + λy = 0; y(0) = 0, y(1) + y (1) = 0,

has infinitely many eigenvalues λ1 < λ2 < λ3 ... Indicate roughly the behaviour of λn as n → ∞.
7. Recasting in Sturm-Liouville form. Express the following equations in Sturm-Liouville form:
′′ ′ ′′ ′
(i) (1 − x2 )y − 2xy + λy = 0 , (ii) x(x − 1)y + [ax − b]y + λy = 0 ,

where a > b and λ are constants.

(iii) Find the eigenvalues and eigenfunctions for
′′ ′
y + 4y + (4 + λ)y = 0, y(0) = y(1) = 0.
Copyright © 2020 University of Cambridge. Not to be quoted or reproduced without permission.

What is the orthogonality relation for these eigenfunctions?

8. Bessel’s equation. (i) Show that the eigenvalues of the Sturm-Liouville problem
d du
(x ) + λxu = 0 , 0 < x < 1,
dx dx

with u(x) bounded as x → 0 and u(1) = 0, are λ = jn2 (n = 1, 2, ...), where the jn are the zeros of the Bessel function
d
J0 (z), arranged in ascending order. [Note: Bessel’s equation of order zero is dz (z dy
dz ) + zy = 0 , (z > 0), which you
may assume has one solution J0 (z) defined as
∞
X z 2m
J0 (z) = (−1)m
m=0
22m (m!)2

and a second solution that is the sum of a regular function and J0 (z) log z.]
(ii) Using integration by parts on the differential equations for J0 (αx) and J0 (βx), show that

1 ′ ′
βJ0 (α)J0 (β) − αJ0 (β)J0 (α)
Z
J0 (αx)J0 (βx)xdx = (β 6= α)
0 α2 − β 2
Z 1 Z 1
1 ′
J0 (jn x)J0 (jm x)xdx = 0 , (n 6= m) , [J0 (jn x)]2 xdx = [J0 (jn )]2 . [Hint: Consider β = jn + ǫ as ǫ → 0.]
0 0 2

(iii) Assume that the inhomogeneous equation

d du
(x ) + λ̃xu = xf (x),
dx dx

where λ̃ is not an eigenvalue, has a unique solution such that u(x) is bounded as x → 0 and u(1) = 0. Assuming also
that f (x) satisfies the same boundary conditions as u and the completeness of the eigenfunctions J0 (jn x), obtain the
eigenfunction expansion of u.
9. Higher order self-adjoint form*. Show that the fourth-order differential operator
4
X dr
L= pr (x) ,
r=0
dxr

where the pr (x) are real functions, is self-adjoint (for appropriate choices of boundary conditions, which you do not
′ ′ ′′′
need to determine) only if p3 = 2p4 , p1 = p2 − p4 .
Considering a specific example, show that the boundary value problem
′′′′ ′ ′
−y + λy = 0; y(0) = y(1) = y (0) = y (1) = 0,

where λ > 0 is real, has infinitely many eigenvalues λ1 < λ2 < λ3 ... . Indicate roughly the behaviour of λn as n → ∞.

†
If you find any errors in the Methods Examples sheets, please inform your supervisor or email [email protected].