CUST 2023-2024
Maths for Physics 2
Session 2
Tutorial
Problem 1: Let f (t) be the quadratic polynomial function
f (t) = t2 . (1)
1) Construct the complex Fourier series of f on the interval t ∈ (−π, π).
2) Deduce the corresponding real Fourier series of f on the same interval t ∈ (−π, π).
Solution: 1) Since the function f as defined by (1) is a priori nonperiodic, we first
construct a periodic function fper (t). We hence require the latter to satisfy
fper (t) = f (t) = t2 , ∀t ∈ (−π, π) , (2)
and then we require fper to repeat itself periodically outside of the interval (−π, π). By
doing this, we construct a function fper that is periodic of period
T = 2π . (3)
Now, being periodic, we can write fper as a Fourier series, namely a complex one here,
and we have in view of the expression (3) of the period here
∞
X ∞
X
2niπt/T
fper (t) = cn e = cn e2niπt/2π ,
n=−∞ n=−∞
that is
∞
X
fper (t) = cn eint , (4)
n=−∞
where the complex Fourier coefficients cn are then given by, also using (2) and (3),
Z T /2 Z π
1 −2niπt/T 1
cn = dte fper (t) = dte−2niπt/2π t2
T −T /2 2π −π
that is
Z π
1
cn = dt t2 e−int , ∀n ∈ Z , (5)
2π −π
which we write in the form
1
cn = In , (6)
2π
2
where In is the integral (remember that −i = −i2 /i = 1/i)
Z π Z π
2 −int
In ≡ dt t e = dt t2 ent/i , ∀n ∈ Z . (7)
−π −π
Let’s first compute this integral for n = 0, we have
Z π 3 π
t 1 3
2
1 3
π − (−π)3 = π + π3 ,
I0 = dt t = =
−π 3 −π 3 3
that is
2π 3
I0 = . (8)
3
Let’s now compute the integral In for n ̸= 0. We do this by means of an integration
by parts, which we recall reads, for two functions u and v,
Z b Z b
′
dt u(t)v (t) = [u(t)v(t)]ba − dt u′ (t)v(t) , (9)
a a
where u′ , v ′ denote the derivatives of u, v. Choosing our function u to be t2 and our
function v ′ to be ent/i , so that we have
i nt/i
u′ (t) = 2t and v(t) = e ,
n
we hence get for (7)
π Z π
i 2 −int i
In = te −2 dt tent/i , (10)
n −π n −π
where we compute separately the two terms in the right-hand side. First we have
π
i 2 −int i 2 −inπ
− π 2 einπ
te = π e
n −π n
iπ 2 2π 2
= [cos(nπ) − i sin(nπ) − cos(nπ) − i sin(nπ)] = sin(nπ) ,
n n
that is
π
i 2 −int
te = 0. (11)
n −π
We now compute the remaining integral in the right-hand side of (10) with another
3
integration by parts of the form (9), where this time we choose u = t (that is u′ = 1) and
again v ′ = ent/i (that is v = (i/n)ent/i ), so that we get
Z π π Z π
nt/i i −int i
dt te = te − dt ent/i
−π n −π n −π
i i −int π
i −inπ inπ
= πe + πe − e
n n n −π
iπ
= [cos(nπ) − i sin(nπ) + cos(nπ) + i sin(nπ)]
n
i2 2iπ 2i
− 2
[cos(nπ) − i sin(nπ) − cos(nπ) − i sin(nπ)] = cos(nπ) − 2 sin(nπ) ,
n n n
that is
Z π
2iπ
dt tent/i = (−1)n . (12)
−π n
Substituting now (11) and (12) into (10) hence yields for In
i 2iπ 4π
In = −2 (−1)n = 2 (−1)n . (13)
n n n
Therefore, substituting first (8) into (6) yields for c0
1 2π 3
c0 = ,
2π 3
that is
π2
c0 = . (14)
3
Then, substituting (13) into (6) yields for cn
1 4π
cn = (−1)n ,
2π n2
that is
2
cn = (−1)n , ∀n ̸= 0 . (15)
n2
The expressions (14) and (15) of the complex Fourier coefficients hence completely
4
determine the complex Fourier series (4), so that we have with (2)
π2 X (−1)n
t2 = +2 2
eint , (16)
3 n̸=0
n
on the interval t ∈ (−π, π).
2) We now use the fact that we know that the coefficients an and bn of the real Fourier
series of f are related to the complex coefficients cn through
a0 = c 0 and an = cn + c−n and bn = i(cn − c−n ) , ∀n > 0 . (17)
Therefore, from (14) we readily get
π2
a0 = . (18)
3
Then, from (15) we get, since (−1)−n = (−1)n ,
2 n 2 −n 2 n 2
an = (−1) + (−1) = (−1) + (−1)n ,
n2 (−n)2 n2 n2
that is
4
an = (−1)n , (19)
n2
and
2 n 2 −n 2 n 2 n
bn = i 2 (−1) − (−1) = i 2 (−1) − 2 (−1) ,
n (−n)2 n n
that is
bn = 0 , (20)
which was expected since our periodic function fper is by construction even here.
Therefore, from (18), (19) and (20) we readily see that f admits the real Fourier series
π2 X (−1)n
t2 = +4 cos(nt) , (21)
3 n>0
n2
on the interval t ∈ (−π, π).
5
Problem 2: Let f (t) be a periodic function of period T . Consider then the function
F (t0 ) defined by the integral
Z t0 +T
F (t0 ) ≡ dt f (t) , t0 ∈ R . (22)
t0
1) Determine the derivative dF/dt0 of the function F .
2) Deduce that
Z T /2
F (t0 ) = dt f (t) , ∀t0 ∈ R . (23)
−T /2
Solution: 1) Let’s first recall the so-called Leibniz integral rule for differentiation under
the integral sign. For a function g(t, t0 ) of the two independent variables t and t0 , and
given two functions a(t0 ) and b(t0 ) of the variable t0 , we have
Z b(t0 ) Z b(t0 )
d db da ∂
dt g(t, t0 ) = g[b(t0 ), t0 ] − g[a(t0 ), t0 ] + dt g(t, t0 ) . (24)
dt0 a(t0 ) dt0 dt0 a(t0 ) ∂t0
Therefore, applying (24) to the derivative dF/dt0 obtained upon differentiating (22)
with respect to t0 , i.e. setting in (24) g(t, t0 ) = f (t) (which is hence independent of the
variable t0 , so that ∂g/∂t0 = 0), a(t0 ) = t0 and b(t0 ) = t0 + T , we get
dF db da d(t0 + T ) dt0
= f [b(t0 )] − f [a(t0 )] = f (t0 + T ) − f (t0 )
dt0 dt0 dt0 dt0 dt0
that is
dF
= f (t0 + T ) − f (t0 ) . (25)
dt0
Now, we use the fact that f is assumed to be periodic of period T , which means that
f (t0 + T ) = f (t0 ) , ∀t0 ∈ R . (26)
Combining (25) with (26) hence readily shows that
dF
=0 , ∀t0 ∈ R . (27)
dt0
2) We just showed with (27) that the derivative of F vanishes identically for any
6
t0 ∈ R: we can thus immediately conclude that the function F itself must be constant
for any t0 ∈ R, that is
F (t0 ) = constant , ∀t0 ∈ R . (28)
Therefore, in order to determine this constant, we merely need to compute F for a
particular value of t0 . Let’s take this value to be
T
t0 = − ,
2
we hence get from (28)
T
F (t0 ) = F − , ∀t0 ∈ R . (29)
2
Since we have from the definition (22) of F that
Z −T /2+T Z T /2
T
F − = dt = dt f (t) ,
2 −T /2 −T /2
we indeed get from (29) that
Z T /2
F (t0 ) = dt f (t) , ∀t0 ∈ R . (30)
−T /2
Problem 3: Let f (t) be the linear polynomial function
f (t) = t . (31)
We recall that it can be written, on the interval t ∈ (−π, π), in the form of the (real)
Fourier series
∞
X (−1)n+1
f (t) = t = 2 sin (nt) . (32)
n=1
n
7
Apply Parseval’s theorem to the function f to show that
∞
X 1 π2
= . (33)
n=1
n2 6
Solution: We recall that, for a function f written as the real Fourier series
∞
X ∞
X
f (t) = a0 + an cos (nt) + bn sin (nt) (34)
n=1 n=1
on the interval t ∈ (−π, π), the Parseval theorem reads
Z π ∞
1 2 1X 2
dt f (t) = a20 + (a + b2n ) . (35)
2π −π 2 n=1 n
Let’s hence first explicitly write down the expressions of the Fourier coefficients an
and bn of the function f (t) = t: upon comparing (32) with (34), it is clear that
(−1)n+1
a0 = an = 0 and bn = 2 , ∀n > 0 . (36)
n
Therefore, combining (35) with the definition (31) of f and the expression (36) of the
Fourier coefficients readily yields
Z π ∞ ∞ ∞
1 21X 2 1X 4 X 1
dt t = bn = 2
=2 2
,
2π −π 2 n=1 2 n=1 n n=1
n
so that we get
∞ Z π
X 1 1
2
= dt t2 . (37)
n=1
n 4π −π
Now, computing the integral in the right-hand side of (37) yields
π 3 π
2π 3
Z
2t 1 3
dt t = = π − (−π)3 = ,
−π 3 −π 3 3
so that we get for (37)
∞
X 1 1 2π 3 π2
2
= = ,
n=1
n 4π 3 2 × 3
8
that is indeed
∞
X 1 π2
2
= . (38)
n=1
n 6
This result (38) is actually a historical problem: indeed, to actually show that the
series in the left-hand side of (38) admits the exact, closed form expression π 2 /6 is referred
to as the Basel problem. This name comes from the city of Basel in Switzerland, where
the Bernoulli [who tried, unsuccessfully, to prove (38)] lived, but also where Leonhard
Euler was born. Euler was the first to provide a solution to this problem, namely to
give an exact value (i.e. π 2 /6) to the series n 1/n2 . It’s certainly noteworthy that
P
Fourier series, and Parseval’s theorem in particular, provide such a simple proof of such
a historical problem!
Furthermore, the series n 1/n2 is actually only a particular case of a much more
P
general mathematical object: the so-called Riemann zeta function ζ(z), which is defined
for complex numbers z ∈ C by the series
∞
X 1
ζ(z) ≡ , z ∈ C. (39)
n=1
nz
If we compare (38) with (39), it’s clear that the left-hand side in (38) is nothing but ζ(2),
so that (38) also reads
π2
ζ(2) = . (40)
6
The Riemann zeta function is related to the distribution of prime numbers, and as such
is one of the most important functions of mathematics. Its properties are at the heart
of an important conjecture: the so-called Riemann hypothesis. Whether the Riemann
hypothesis is true or wrong remains, to this day, an open question: it was one of Hilbert’s
famous problems (namely the 8th one), and is one of the Clay Mathematics Institute’s
Millenium Prize Problems.