UNIVERSITY COLLEGE LONDON

DEPARTMENT OF PHYSICS AND ASTRONOMY

PHAS0025 MATHEMATICAL METHODS III

Practice problems on solving partial differential equations
(2022–23)

Please try these problems at home or in the PST sessions. Feel free to discuss any with
the PST tutors. These will not be marked, but the mark allocation is indicative.
Solutions will be posted at the end of week 2 when all PSTs have taken place.

1. The temperature T (x, t) at time t satisfies the equation

∂2T ∂T
− =0,
∂x2 ∂t
where the coordinate x lies in the range 0 ≤ x ≤ 2π.
By separating the variables x and t, show that one solution for T is given by
2
T (x, t) = {A cos(λx) + B sin(λx)} e−λ t ,

where A, B and λ are constants. [5 marks]

If the temperatures at x = 0 and x = π are kept at T = 0 K, show that the

temperature must have the form
∞
X
T (x, t) = Bn sin(nx) exp(−n2 t) ,
n=1

where n is an integer. [3 marks]

Solution
Assume that a solution of the equation

∂2T ∂T
− =0
∂x2 ∂t
exists in the form
T (x, t) = X(x) × Θ(t) .
Straightforward differentiating then leads to

1 d2 X 1 dΘ [1]
= .
X dx2 Θ dt

The left hand side is a function of x whereas the right hand side depends purely
upon t. This can only be true if both are equal to the same constant. To agree
with the notation in the question, take the separation constant to be −λ2 to find
the two ordinary differential equations

d2 X
= −λ2 X ,
dx2
dΘ [1]
= −λ2 Θ .
dt
1
 The solutions of the first equation are of the form cos(λx) and sin(λx), whereas
the time equation has the single solution exp(−λ2 t). Putting these together with
arbitrary constants A and B, we find
2
T (x, t) = {Aλ cos(λx) + Bλ sin(λx)} e−λ t .

where A, B and λ are constants. [2]

A special case of λ = 0 leads to a non-oscillatory, time-independent solution [1]
T (x, t) = A0 + B0 x,
which (as can be seen) does not satisfy the boundary conditions:

T (0, t) = T (π, t) = 0

and can be omitted.

If T (0, t) = 0 for all time, the coefficient of the cosines must vanish, i.e. Aλ = 0.
If, further, T (π, t) = 0 then sin(λπ) = 0. This requires that λπ = nπ, where n
is an integer. Thus the solution becomes

T (x, t) = Bn sin(nx) exp(−n2 t) . [2]

Since the heat diffusion equation is linear and homogeneous, any linear super-
position with different values of n is also a solution:
∞
X
T (x, t) = Bn sin(nx) exp(−n2 t) , [1]
n=1
where the Bn must be fixed by other boundary conditions, such as the tempera-
ture distribution at t = 0.
2. The function u(x, t) satisfies the differential equation
 2   2 
∂ u 2 2 ∂ u
+ α u = c ,
∂t2 ∂x2
where c and α are real constants. By seeking a solution of the equation in the
separable form u(x, t) = X(x) × T (t), find the most general solution for which
u(0, t) = 0, u(L, t) = 0, and u(x, 0) = 0. [10 marks]

Solution
Assume that one solution of
 2   2 
∂ u 2 2 ∂ u
2
+α u=c
∂t ∂x2
can be written in the form

u(x, t) = X(x) × T (t) .

Inserting this ansatz into the equation gives

 2   2 
d T 2 2 d X [1]
X 2
+α XT =c T .
dt dx2
2
 After dividing both sides by the product X T , we find that the left-hand side is
a function of t whereas the right-hand side depends purely upon x. They must
therefore both be equal to some constant, which we denote by −λ2 :

1 d2 T
   2 
2 2 1 d X
+ α = c = −λ2 .
T dt2 X dx2

This leads to the two ordinary differential equations

 2 
d T
+ (α2 + λ2 ) T = 0 ,
dt2
 2 
d X λ2 [2]
2
+ 2 X = 0.
dx c

The X equation has solutions of the form

X(x) = cos(λx/c) or sin(λx/c) . [1]

However, for all values of t the function u(x, t) vanishes, which means that
the cosine must be discarded. Furthermore, in order that the function vanish at [1]
x = L, one requires that
sin(λL/c) = 0 ,
which means that
λ = ncπ/L ,
where n is a (positive) integer. [1]
The general solutions of the T equation are
p  p 
T (t) = cos α2 + λ2 t or sin α 2 + λ2 t . [1]

The fact that u vanishes at t = 0 for all values of x means that, once again, the
cosine solution can be thrown away. [1]
One solution of the original differential equation is then
 p 
u(x, t) = An sin t α2 + (n2 c2 π 2 /L2 ) sin(nπx/L) . [1]

However, the differential equation is linear and homogeneous. As a conse-

quence, any linear superposition of these solutions is also a solution. Hence
∞
X  p 
u(x, t) = An sin t α2 + (n2 c2 π 2 /L2 ) sin(nπx/L) . [1]
n=0

3. Laplace’s equation in two-dimensional plane polar coordinates is

 2     2 
2 ∂ F ∂F ∂ F
r 2
+r + =0.
∂r ∂r ∂θ2

3
Assuming that the function F can be separated as

F (r, θ) = R(r) × Θ(θ) ,

derive ordinary differential equations for R(r) and Θ(θ).

Solve the radial equation by trying for a solution of the form R(r) = rα . Hence
show that the most general single-valued solution may be written as
∞
X
An rn + Bn r−n (Cn cos nθ + Dn sin nθ) .

F (r, θ) =
n=0
[6 marks]

Find the form of the solution for which F (r, θ) = F1 cos θ when r = a, where
F1 is a constant. Is the solution unique? [4 marks]

Solution
Inserting
F (r, θ) = R(r) × Θ(θ)
into Laplace’s equation
 2     2 
∂ F ∂F ∂ F
r2 + r + =0,
∂r2 ∂r ∂θ2

we obtain
d2 R dR d2 Θ [1]
r2 Θ 2
+ rΘ +R 2 =0.
dr dr dθ

Divide through by F and take the angular term over to the right hand side.

d2 R 1 d2 Θ
   
1 dR
r2 2 + r =− ·
R dr dr Θ dθ2

Since the left hand side is a function of r and the right hand side of θ, the only
way that they can be equal is if they both equal a constant. Let that constant be
−n2 . We then find two ordinary equations

d2 R dR
r2 2
+r − n2 R = 0,
dr dr
d2 Θ [1]
+ n2 Θ =0.
dθ2

The angular equation is just that of SHM, so that

Θ(θ) = Cn cos nθ + Dn sin nθ ,

where Cn and Dn are constants. In order that the solution be single-valued,

Θ(θ + 2π) = Θ(θ), the separation constant n must be an integer. [1]
α
By trying for a solution to the radial equation of the form R(r) = r , we obtain

α(α − 1)r2 rα−2 + αr rα−1 − n2 rα = (α2 − n2 ) rα = 0 .

4
 This is possible only if α = ±n, and the radial function then is

R(r) = An rn + Bn r−n ,

where An and Bn are constants. [2]

Putting the angular and radial parts together, one solution of Laplace’s equation
is
F (r, θ) = An rn + Bn r−n (Cn cos nθ + Dn sin nθ) .


However, since the equation is linear and homogeneous, any linear superposition
of such solutions is also allowed, leading to
∞
X
An rn + Bn r−n (Cn cos nθ + Dn sin nθ) . [1]

F (r, θ) =
n=0

We are given that

∞
X
An an + Bn a−n (Cn cos nθ + Dn sin nθ) = F1 cos θ .

F (a, θ) =
n=0

Since the right hand side has a cosine angular dependence, all the Dn vanish, as
do all the Cn except for C1 . For ease of notation, let a1 = A1 C1 and b1 = B1 C1 .
The general solution is then

F (r, θ) = {a1 r + b1 /r} cos θ ,

with boundary condition [2]

a1 a + b1 /a = F1 .

Inserting this [1]

F (r, θ) = a1 r + (F1 a − a1 a2 )/r cos θ .


The solution still has one free parameter (a1 ) here and so it is not unique. This
is because we have only applied one boundary condition to a second order equa-
tion. [1]

4. A function u(x, y) of two independent variables x and y satisfies the first order
partial differential equation
∂u 1 ∂u
x − 2 y =0.
∂x ∂y

By first looking for a separable solution of the form

u(x, y) = X(x) × Y (y), find the general solution of the equation. [8 marks]

Determine the u(x, y) which satisfies the boundary condition

u(1, y) = 1 + sin y. [3 marks]

5
Solution:
Inserting the ansatz u(x, y) = X(x) × Y (y) into the partial differential equation

∂u 1 ∂u
x = 2 y .
∂x ∂y
leads to
x dX y dY [1]
= .
X dx 2Y dy

Since the left hand side is a function of x and the right of y, this means that
they must both equal some separation constant k. We then obtain two ordinary
differential equations

dX k
= X,
dx x
dY 2k [1]
= Y ,
dy y
where k is an arbitrary separation constant.
The first equation can be written as
Z Z
dX dx
=k ,
X x
which may be integrated to give

`n(X/Ak ) = k `n(x) ,

where Ak is an arbitrary constant.

Taking exponential of both sides,

X = Ak x k .

The y equation is similarly

Z Z
dY dy
= 2k ,
Y y
which may be integrated to give

`n(Y /Bk ) = 2k `n(y) ,

and hence
Y = Bk y 2k .

The corresponding solution for u(x, y) is

u(x, y) = Ck (xy 2 )k .

The most general solution to the differential equation is a linear superposition of

such solutions: X [6]
u(x, y) = Ck (xy 2 )k ,
k
6
 where the sum may actually be an integral over continuous values of k (though
students would not be expected to stress this point).
It is crucial to note that the solution is only a function of z = xy 2 and it doesn’t
matter how x and y are varied separately provided that this combination is kept
fixed.
Now we are given the boundary condition that
√
u(1, y) = 1 + sin y = 1 + sin( z) .

because x = 1. Having got the solution now in terms of z, we can read off the
general solution to be √ [3]
u(x, y) = 1 + sin(y x) .

5. Solve the “diffusion equation”

∇2 f (x, y, z, t) = α ∂t f (x, y, z, t)

by separation of variables and find the general periodic solution in terms of x, y

and z and show that it can be represented as the following expansion
X
f (r, t) = a(k)eik·r−κt , with ακ = |k|2 ,
k

where the vectors r and k are given by r = (x, y, z)T and k = (kx , ky , kz )T . [16 marks]

Solution:
To separate the variables, assume f (x, y, z, t) = X(x)Y (y)Z(z)T (t), insert it
into the diffusion equation, and divide by X(x)Y (y)Z(z)T (t) to obtain:

X 00 Y 00 Z 00 T0 [1]
+ + =α
X Y Z T
where 0 stands for ordinary differentiation. This equation is comprised of four
terms, each of which depends on a different, independent variable. Each of these
terms must hence be equal to a constant, so that one can write

X 00 = cx X,
00
Y = cy Y,
Z 00 = cz Z,
0
T = ct T

with the additional proviso:

cx + cy + cz = αct . (1) [2]

The solution of a linear differential equation with constant coefficients can be

found as one of the following forms: eikx , eκx , cos(kx) and sin(kx), which
should lead to equivalent general solutions. Here we choose the eikx x , eiky y and [2]
eikz z representations for X, Y, Z, respectively, and e−κt for representing the T
function.

7
Substituting these representations into the ordinary differential equations, we
obtain the following relations for the separation constants:
−kx2 = cx ,
−ky2 = cy ,
−kz2 = cz ,
−κ = ct ,
connected by the relation (see Equation (1))
[2]
kx2 + ky2 + kz2 = ακ.

Then the general solutions of these ordinary differential equations can be ex-
pressed, in terms of exponentials, as
X(x) = akx eikx x + bkx e−ikx x ,
Y (y) = cky eiky y + dky e−iky y ,
Z(z) = ekz eikz z + fkz e−ikz z ,
[1]
T (t) = gκ e−κt .

A special case is when kα = 0, which leads to the solutions of the following

forms:
X(x) = a0 + b0 x,
Y (y) = c0 + d0 y,
Z(z) = e0 + f0 z, [1]

which we will ignore as non-periodic.

The general solution now takes the form:
X
f (x, y, z, t) = [akx eikx x +bkx e−ikx x ][cky eiky y +dky e−iky y ][ekz eikz z +fkz e−ikz z ]gκ e−κt ,
kx ,ky ,kz

where [2]
ακ = (kx2 + ky2 + kz2 ).

Notice that in the infinite sum above, the contributions for a given value of the
separation constant, say kx , are equivalent to those for the opposite value −kx .
For example the following two contributions
akx eikx x + bkx e−ikx x
and
a−kx e−ikx x + b−kx eikx x
are equivalent and can be combined. Hence, the general solution can be re-
arranged into the following, more compact form
X
f (x, y, z, t) = eikx x eiky y eikz z akx cky ekz gκ e−κt ,
kx ,ky ,kz

8
 where [3]
ακ = (kx2 + ky2 + kz2 ).

Defining the vectors r = (x, y, z)T and k = (kx , ky , kz )T and the new coeffi-
cients
a(k) = akx cky ekz gκ with ακ = (kx2 + ky2 + kz2 ),
one gets the general solution in the form:
X
f (r, t) = a(k)eik·r−κt , with ; ακ = |k|2 . [2]
k

Since k can be any continuous value one can also write the general solution in
the integral form: Z
f (r, t) = a(k)eik·r−κt dk.

6. A uniform bar of length L is clamped at x = 0 and x = L such that it lies in

a horizontal plane. Provided that the deviations from equilibrium are small, the
vertical vibration u(x, t) satisfies the fourth-order equation

∂2u 4
2 ∂ u
+ c =0,
∂t2 ∂x4
where c is a constant. Look for a separable solution of the form

u(x, t) = X(x) × T (t) ,

find the ordinary differential equations satisfied by X(x) and T (t). [3 marks]

Show that the general solution for X(x) can be given by

X(x) = A cos(βx) + B sin(βx) + C cosh(βx) + D sinh(βx) .

Find the most general solution of the equation which satisfies the boundary con-
ditions

u(0, t) = 0 , u(L, t) = 0 ,

∂ 2 u(x, t) ∂ 2 u(x, t)
= 0, =0,
∂x2 x=0 ∂x2 x=L
u(x, 0) = 0.

[10 marks]

Solution:
Assume that one solution of
∂2u ∂4u
2
+ c2 4 = 0 ,
∂t ∂x

9
can be written in the form

u(x, t) = X(x) × T (t) .

Inserting this ansatz into the equation gives

 2   4 
d T 2 d X [1]
X 2
= −c T .
dt dx4

After dividing both sides by the product X T , we find that the left-hand side is
a function of t whereas the right-hand side depends purely upon x. They must
therefore both be equal to some constant, which we denote by −λ2 :

1 d2 T
   4 
2 1 d X [1]
= −c = −λ2 .
T dt2 X dx4

This leads to the two ordinary differential equations

 2 
d T
+ λ2 T = 0 ,
dt2
 4 
d X λ2 [1]
4
− 2 X = 0.
dx c

By inspection or by seeking a solution of the form epx , it is seen that the X

equation has solutions

X(x) = A cos(βx) + B sin(βx) + C cosh(βx) + D sinh(βx) ,

where β 2 = λ/c. [3]

2 2
The boundary conditions at x = 0 require that A+C = 0 and −β A+β C = 0,
so that A = C = 0. [2]
The boundary conditions at x = L then require B sin(βL) + D sinh(βL) = 0
and −β 2 B sin(βL) + β 2 D sinh(βL) = 0. From this we deduce that

B sin(βL) = 0,
D sinh(βL) = 0.

Since the sinh function does not vanish except at the origin, the second equation
requires D = 0. For the first, we cannot put B = 0 or otherwise the whole
solution would vanish. We therefore require that sin(βL) = 0, so that β =
nπ/L, where n is a non-negative integer. [3]
The t equation is much more standard and has solutions sin(λt) and cos(λt).
However, we must discard the second because it doesn’t vanish at t = 0. Putting [1]
everything together, the most general solution to the original equation is then
∞
cn2 π 2
X    nπx 
u(x, t) = Bn sin t sin . [1]
n=0
L2 L

10
7. In atomic units the Schrödinger equation for the hydrogen atom can be written:
 
1 1
− ∇2 − ψ = Eψ
2 r

where, in spherical polar coordinates:

∂2
   
1 ∂ ∂ 1 ∂ ∂ 1
∇2 = 2 r2 + 2 sin θ + 2 2 ·
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂φ2

(a) By writing ψ = R(r)Θ(θ)Φ(φ) show that Φ(φ) must satisfy the equation

d2 Φ
= −m2 Φ .
dφ2
[4 marks]

(b) What are the solutions of this equation? Explain why m, the constant of
separation, must take integer values in this physical system. [3 marks]

(c) Hence show that R(r) must satisfy the radial equation
 
1 d 2 dR
r + Rr + ERr2 = λR ,
2 dr dr

where λ is another constant of separation. Obtain the corresponding equa-

tion for Θ(θ). [6 marks]

(d) For the special case of λ = 0, show that the radial equation can be written

1 d2 U
 
1
+ + E U =0
2 dr2 r

where U (r) = rR(r). [3 marks]

(e) Show that for large values of r

U (r) = A exp(αr) + B exp(−αr)

1
where α = (−2E) 2 . How can this solution be simplified for bound states
of hydrogen? [4 marks]

Solution:

(a)
ψ(r, θ, φ) = R(r) × Θ(θ) × Φ(φ) .

   
1 d dR 1 1 d dΘ
ΘΦ r2 + RΦ 2 sin θ
2r2 dr dr 2 r sin θ dθ dθ
 2   
1 1 d Φ 1
+ RΘ 2 2 + + E RΘΦ = 0 .
2 r sin θ dφ2 r
[2]

11
 Divide by R Θ Φ and multiply by r2 sin2 θ
sin2 θ d
     2 
dR 1 d dΘ 1 d Φ
r2 + sin θ sin θ +r sin2 θ(1+Er)+ = 0.
2R dr dr 2Θ dθ dθ 2Φ dφ2
The first 3 terms here depend on r and θ but the last is a function purely of [1]
φ. Since r, θ and φ are independent variables, the third term must be some
constant, denote by −m2 : [1]
2
d Φ
= −m2 Φ .
dφ2
(b) This has solutions e±imφ or, alternatively, cos mφ and sin mφ. When φ
increases by 2π, the solution returns to the same point, so we expect the
same physical solution. Thus Φ(φ + 2π) = Φ(φ). This can only be ac-
complished if m is a real integer. Then Φ(φ) is clearly a periodic function.
[3]
2
(c) The remainder of the equation can be manipulated (divide sin θ and rear-
range) into
m2
   
1 d 2 dR 1 1 d dΘ
r + r(1 + Er) = − sin θ .
2R dr dr 2 sin2 θ 2Θ sin θ dθ dθ
The left hand side is a function of r only, while the right hand side depends [2]
only on θ. This means that both sides must be equal to some constant,
which we denote by λ. This results in two ordinary differential equations: [2]
 
1 d dR
r2 + r(1 + Er)R = λ R ,
2 dr dr
m2
   
d dΘ
sin θ + 2λ sin θ − Θ = 0.
dθ dθ sin θ
[2]
(d) Let U = rR and λ = 0.
R0 = −r−2 U + r−1 U 0
[1]
r2 R0 = −U + rU 0
 
d 2 dR
r = −U 0 + U 0 + rU 00
dr dr
[1]
rU 00 + 2U + 2ErU = 0
gives [1]
1 d2 U
 
1
+ +E U =0
2 dr2 r
1
(e) For large r, r tends to zero so
U 00 = −2EU
1
which has solution for α = (−2E) 2 [1]
U (r) = A exp(αr) + B exp(−αr).
For bound states this solution must be normalisable, however exp(αr) is [1]
infinite at r = ∞ so cannot be normalised, therefore A = 0. [2]

12