MATHEMATICAL TRIPOS Part II 2024

List of Courses

Algebraic Geometry
Algebraic Topology
Analysis of Functions
Applications of Quantum Mechanics
Applied Probability
Asymptotic Methods
Automata and Formal Languages
Classical Dynamics
Coding and Cryptography
Cosmology
Differential Geometry
Dynamical Systems
Electrodynamics
Fluid Dynamics
Further Complex Methods
Galois Theory
General Relativity
Graph Theory
Integrable Systems
Linear Analysis
Logic and Set Theory
Mathematical Biology
Mathematics of Machine Learning
Number Fields
Number Theory
Numerical Analysis
Principles of Quantum Mechanics
Principles of Statistics
Probability and Measure
Quantum Information and Computation
Representation Theory
 2
Riemann Surfaces
Statistical Modelling
Statistical Physics
Stochastic Financial Models
Topics in Analysis
Waves

Part II, 2024 List of Questions

 3
Paper 1, Section II
25F Algebraic Geometry
In this question, all algebraic varieties are over an algebraically closed field k.
What does it mean for a topological space to be irreducible? Show that if X ⊆ An
is a Zariski closed subset, then X can be written as a finite union of irreducible closed
subsets of An .
Write the closed subset Z(x3 − x22 , x21 − x22 − x42 + x23 ) of A3 as a union of irreducible
closed sets.
Now take k = C to be the field of complex numbers. Show that the set

Z := {(x, ex ) | x ∈ C} ⊆ A2

is dense in A2 in the Zariski topology.

Let X be an affine algebraic variety, and let U = {Ui | i ∈ I} be an open cover of
X. Show that U has a finite subcover. [Hint: Define for any regular function f on X the
distinguished open set
D(f ) := {x ∈ X | f (x) ̸= 0}.
You may use without proof the fact that the collection of distinguished open sets form a
basis for the topology on X.]

Paper 2, Section II
25F Algebraic Geometry
In this question, all algebraic varieties are over an algebraically closed field k.
Let X be an affine variety. Define the tangent space of X at a point P ∈ X. Define
the dimension of X in terms of (i) the tangent spaces of X and (ii) Krull dimension. Say
what it means for the variety to be singular at P .
Assume the characteristic of the field k is not 2. Let X := Z(x21 − x32 , x23 − x34 ) ⊆ A4 .
Calculate the tangent space of X at each point of X.
Consider the subset Y ⊆ P4 consisting of points with homogeneous coordinates
(y0 : y1 : y2 : y3 : y4 ) such that the matrix
 
y0 y1 y2
y2 y3 y4

has rank one. Show that Y is a closed subset of P4 in the Zariski topology. You may now
assume Y is irreducible in the Zariski topology, and hence is a projective variety. What is
the dimension of Y ? Show that Y is non-singular.

Part II, 2024 List of Questions [TURN OVER]

 4
Paper 3, Section II
24F Algebraic Geometry
In this question, all algebraic varieties are over an algebraically closed field k.
State the Riemann–Roch theorem.
Let X be a non-singular projective curve of genus 1, and let P0 ∈ X be a point.
Show that if D is a divisor of degree 0 on X, then there exists a unique P ∈ X such that
D is linearly equivalent to P − P0 .
Use this to describe a group law on X with P0 as the identity element. [You do not
need to prove this law satisfies the group axioms.]
Consider the group law on the non-singular cubic curve

X = Z(y 2 z − x3 + 4xz 2 − z 3 ) ⊆ P2

with identity element P0 = (0 : 1 : 0) ∈ X. Let A = (2 : 1 : 1) and B = (−2 : −1 : 1) be

two points on X. Find A + B.
Let X ⊆ P2 be a non-singular cubic curve. A point P0 ∈ X is an inflection point of
X if there exists a line L ⊆ P2 such that X ∩ L = {P0 }. Let P0 be an inflection point of
X, and give X the group structure with identity element P0 . A point P ∈ X is said to
be 3-torsion if P + P + P = P0 . Prove that P is a 3-torsion point if and only if P is an
inflection point.

Paper 4, Section II
24F Algebraic Geometry
In this question, all algebraic varieties are over the complex numbers C.
Let φ : P2 → P1 be a morphism. Show that φ is constant. [You may use without
proof the fact that any two curves in P2 have non-empty intersection.]
Show that there is no closed subvariety of P1 × P1 × P1 isomorphic to P2 .
State the Riemann–Hurwitz theorem.
Consider a non-singular projective curve X ⊆ P1 × P1 defined by a bihomogeneous
equation f (x0 , x1 , y0 , y1 ) = 0 which is homogeneous of degree 2 in x0 , x1 and homogeneous
of degree 3 in y0 , y1 . Here x0 , x1 are coordinates on the first P1 and y0 , y1 are coordinates
on the second P1 . Compute the genus of X. Deduce that X is not isomorphic to a non-
singular projective curve in P2 . [You may use without proof the fact that a non-singular
projective curve in P2 of degree d has genus g = (d − 1)(d − 2)/2.]

Part II, 2024 List of Questions

 5
Paper 1, Section II
21J Algebraic Topology
e → X is a covering map? If γ : [0, 1] → X is
What does it mean to say that p : X
e
a path and x̃0 ∈ X is such that p(x̃0 ) = γ(0), prove carefully that there is a unique path
e such that
γ̃ : [0, 1] → X

(i) γ̃(0) = x̃0 , and

(ii) p ◦ γ̃ = γ.

[You may use the Lesbegue number lemma.]

Let Y be a topological space, A ⊆ Y and B ⊆ Y be open subspaces with disjoint
closures, and ϕ : A → B be a homeomorphism. Let Y /ϕ denote the quotient of Y by the
equivalence relation generated by a ∼ ϕ(a) for all a ∈ A. Show that the function

Y ×Z
p : Yd
/ϕ := −→ Y /ϕ
(a, i) ∼ (ϕ(a), i − 1) for a ∈ A, i ∈ Z
[(y, i)] 7−→ [y]

is continuous and is a covering map.

Assume now that Y is path-connected. Let a0 ∈ A be a basepoint, which
determines a basepoint [a0 ] ∈ Y /ϕ. Show that Yd /ϕ is path-connected, that the subgroup
G ⩽ π1 (Y /ϕ, [a0 ]) associated to the covering space p : Yd
/ϕ → Y /ϕ is normal, and that the
quotient group π1 (Y /ϕ, [a0 ])/G is isomorphic to Z.

Paper 2, Section II
21J Algebraic Topology
State the Seifert–van Kampen theorem.
If (X, x0 ) is a based topological space, and f : (S n−1 , ∗) → (X, x0 ) is a map of based
spaces, define the space X ∪f Dn obtained by attaching an n-cell to X along f . For n = 2,
carefully prove a formula describing π1 (X ∪f D2 , x0 ) in terms of the group π1 (X, x0 ) and
the element [f ] ∈ π1 (X, x0 ).
Writing S 1 ∨ S 1 for the wedge of two circles, calculate π1 (S 1 ∨ S 1 , ∗).
Explain how to attach 2-cells to S 1 ∨ S 1 to obtain a space whose fundamental group
is the symmetric group on 3 letters, proving carefully that this is indeed the fundamental
group obtained.
[You may use any description of the group π1 (S 1 , ∗), provided it is clearly stated.
You should justify any presentation of the symmetric group on 3 letters that you use.]

Part II, 2024 List of Questions [TURN OVER]

 6
Paper 3, Section II
20J Algebraic Topology
Let L be a simplicial complex and K ⩽ L be a subcomplex, with associated chain
complexes C• (L) and C• (K). Setting

Ck (L)
Ck (L, K) := ,
Ck (K)

show that the boundary map of L descends to give C• (L, K) the structure of a chain
complex. Describe a long exact sequence relating H∗ (K), H∗ (L), and the homology
H∗ (L, K) of C• (L, K).
In the following, ∆n+1 denotes the standard (n + 1)-simplex and ∂∆n+1 denotes its
boundary, i.e. the union of all its simplices of dimension < n + 1. Suppose that n ⩾ 2.
Using that ∆n+1 has the same homology as a point, calculate the homology of
∂∆n+1 .
Recall that the rank rk A of an abelian group A is the maximal r such that Zr ⩽ A.
If K ⩽ ∂∆n+1 is a sub-simplicial complex, by considering H∗ (∂∆n+1 , K) show that for
each 0 < k ⩽ n we have
 
n+2
rk Hk (K) ⩽ − #(k + 1)-simplices of K.
k+2

[You may use that the rank of a subgroup or quotient of A is at most that of A.]
If K < ∂∆n+1 is a proper sub-simplicial complex, show that Hn (K) = 0.

Paper 4, Section II
21J Algebraic Topology
State the Mayer–Vietoris theorem for a simplicial complex K which is the union of
two subcomplexes M and N .
For each k = 2, 3, 4, . . ., construct an explicit simplicial complex K having

Z
 i=0
∼
Hi (K) = Z/k i = 1


0 else.

Justify your answer.

Part II, 2024 List of Questions

 7
Paper 1, Section II
23G Analysis of Functions
(a) Set R+ = (0, ∞), and let Lp = Lp (R+ , dx), 1 < p < ∞, where dx is Lebesgue
measure on R+ . Let F : R+ × R+ → R be integrable for the product measure dx ⊗ dx on
R+ × R+ . Set Z
G(y) = F (x, y)dx, y ∈ R+ .
R+
Show that if ∥g∥Lq ⩽ 1 for 1 < q < ∞ conjugate to p, then
Z Z hZ i1/p
|G(y)g(y)| dy ⩽ |F (x, y)|p dy dx.
R+ R+ R+

(b) Now let K : R+ × R+ → R be integrable and such that

Z ∞
−1
K(λx, λy) = λ K(x, y), λ, x, y > 0; and |K(x, 1)|x−1/p dx = 1.
0
R∞
Define T f (y) = 0 K(x, y)f (x) dx. Show that for f ∈ Lp we have

∥T f ∥Lp ⩽ ∥f ∥Lp .

[Hint: Consider fz (y) = f (yz) and show first that ∥fz ∥Lp = z −1/p ∥f ∥Lp .]
[You may use the identity
nZ o
∥f ∥Lp = sup |f (x)g(x)|dx : g ∈ Lq , ∥g∥Lq ⩽ 1 , q conjugate to p,

without proof.]

Part II, 2024 List of Questions [TURN OVER]

 8
Paper 2, Section II
23G Analysis of Functions
State (without proof) the Hahn–Banach theorem for linear functionals on a normed
real vector space X.
Now consider the topological dual space X ′ . For each x ∈ X, define x̂ : X ′ → R by
the action
x̂(f ) = f (x), f ∈ X ′ .
Show carefully that x 7→ x̂ defines a linear isometry from X into the bidual space X ′′ (the
topological dual space of X ′ ), and that

∥x∥X = sup |f (x)|.

∥f ∥X ′ ⩽1

Let I = [0, 1] and denote by L∞ the Banach space of µ-essentially bounded functions
on I, where µ is Lebesgue measure. Show that (L∞ )′ does not coincide with L1 (µ). [Hint:
Extend the functional ℓ(f ) = f (0) from the subspace C(I) of continuous functions on I to
L∞ .]

Paper 3, Section II
22G Analysis of Functions
State and prove the Sobolev embedding theorem H s (Rn ) ⊂ L∞ (Rn ).
Show that H 1 (R3 ) contains an unbounded function.
[You may use results from Fourier analysis without proof, provided they are carefully
stated.]

Paper 4, Section II
23G Analysis of Functions
(a) For f ∈ H s (Rn ), s ∈ R, show that there exists a unique weak solution u ∈ S ′ (Rn )
to the equation
∆2 u + u = f. (⋆)

(b) Show that if f has compact support in Rn and is infinitely differentiable, then
the solution u is also infinitely differentiable.
(c) Now let n = 3 and let f = δ0 be the Dirac measure at zero. Show that there
exists a unique continuous function u solving (⋆).
[You may use results about Fourier transforms and Sobolev spaces from the course
without proof, provided they are clearly stated.]

Part II, 2024 List of Questions

 9
Paper 1, Section II
35E Applications of Quantum Mechanics
Consider the quantum mechanical scattering of a particle of mass m in three-
dimensions, with Hamiltonian
|p|2
H= + V (r) .
2m
Here the potential V (r) is spherically symmetric, and it is localised in some region of
space. Using a partial wave decomposition,
∞
X ul (r)
ψ(r) = Pl (cos θ) ,
r
l=0

where Pl (cos θ) are Legendre polynomials, and boundary condition ul (0) = 0, the time-
independent Schrödinger equation for the wavefunction ψ(r) of the particle reduces to
 
d2 l(l + 1) 2m 2 ℏ2 k 2
− 2+ + V (r) ul (r) = k ul (r) , with E = .
dr r2 ℏ2 2m

(a) The asymptotic behaviour, for large r, of the wavefunction can be written
∞
X  −ikr

2l + 1 l+1 e eikr
ψ(r) ∼ (−1) + Sl (k) Pl (cos θ) .
2ik r r
l=0

For real values of k, show that the coefficients Sl (k) satisfy

Sl (k)∗ Sl (k) = 1 , Sl (k) Sl (−k) = 1 .

Hence deduce that Sl (k) = e2iδl (k) for some real function δl (k), and that δl (k) = −δl (−k).
(b) Focus now on the low-momentum behaviour, where the l = 0 mode dominates.
For some choice of potential V (r),

(k + 3iλ)(k + 2iλ)
S0 (k) = ,
(k − 3iλ)(k − 2iλ)

where λ is a real positive constant. Evaluate the scattering length as , and give an estimate
for the total cross-section σT at low energies.
Briefly explain the significance of the poles of S0 (k). Are there any resonances in
S0 (k)? Why is it important that λ is positive?

Part II, 2024 List of Questions [TURN OVER]

 10
Paper 2, Section II
36E Applications of Quantum Mechanics
Consider a Hamiltonian H that has a discrete spectrum and a ground state with
energy E0 .
(a) Describe briefly how to use the variational method to provide an upper bound
on E0 .
(b) Suppose that the trial wavefunction in the variational method is given by
N
X
|ψ⟩ = αn |ϕn ⟩ ,
n=1

where αn are complex variational parameters, and the |ϕn ⟩ form an orthonormal set, i.e.,
⟨ϕm |ϕn ⟩ = δmn , for m and n = 1, 2, . . . N .
Apply the variational method to H with trial wavefunction |ψ⟩, and show that the
lowest eigenvalue of the matrix H, which has entries Hnm = ⟨ϕn |H|ϕm ⟩, gives the optimal
upper bound on the ground state energy E0 .
(c) Consider a particle of mass m in an infinite one-dimensional square well of width
a, with a linear potential in the well,
( x
V0 0 ⩽ x ⩽ a,
V (x) = a
∞ otherwise ,

9ℏ2
where V0 = . Determine an upper bound for the ground state energy of this system
ma2
using ( q

2
sin nπx 0 ⩽ x ⩽ a,
ϕn = a a
0 otherwise ,
with n = 1, 2, as trial wavefunctions.

Part II, 2024 List of Questions

 11
Paper 3, Section II
34E Applications of Quantum Mechanics
(a) Consider a one-dimensional system with a periodic potential V (x) = V (x + a),
where a is a positive constant. The Schrödinger equation for the system is

ℏ2 d2 ψ
− + V (x)ψ(x) = E ψ(x) , (⋆)
2m dx2
where E is the energy of the system. The Floquet matrix F (E) is given by
   
ψ1 (x + a) ψ1 (x)
= F (E) ,
ψ2 (x + a) ψ2 (x)

where ψ1 and ψ2 are two linearly independent solutions to (⋆).

(i) Show that det(F ) = 1.

(ii) Explain why the trace TrF is real.

(iii) Explain what the implications on the band structure are when (TrF )2 < 4
and when (TrF )2 > 4.

(b) Consider a tight-binding Hamiltonian that acts upon a single band of localised
states in one dimension,
X 
Ht = t |j⟩⟨j + 1| + |j⟩⟨j − 1| + (−1)j |j⟩⟨j| ,
j∈Z

where t is a positive constant. The integer j should be thought of as indexing sites along a
chain of atoms separated from each other by a distance a; the state |j⟩ locates an electron
on atom j.

(i) Consider the translation operator Tℓ which acts on the states |j⟩ by shifting
them to another site, Tℓ |j⟩ = |j + ℓ⟩, with ℓ ∈ Z. Determine the values of
ℓ for which Tℓ commutes with the Hamiltonian Ht . Hence determine the
lattice spacing of the system.

(ii) Using Bloch’s theorem, show that the Hamiltonian of the system can be
reduced to a 2×2 matrix. Find the eigenvalues of this matrix.

(iii) What is the range of the first Brillouin zone? Plot the energy bands for
the first Brillouin zone. In your plot you should clearly label the axes and
indicate the values of the energies at the boundaries of the Brillouin zone.
What is the bandwidth of each of the bands?

Part II, 2024 List of Questions [TURN OVER]

 12
Paper 4, Section II
34E Applications of Quantum Mechanics
Consider a particle of charge −e and mass m moving in three dimensions with an
electric field E and a magnetic field B. The time-dependent Schrödinger equation for the
particle’s wave function ψ(x, t) is
   2
∂ ie ℏ2 ie
iℏ − ϕ ψ=− ∇+ A ψ , (⋆)
∂t ℏ 2m ℏ

where A is the vector potential and ϕ is the electric potential, with B = ∇ × A and
E = −∇ϕ − ∂A∂t .
(a) The Schrödinger equation (⋆) should be invariant under the gauge transforma-
tions
A(x, t) → A(x, t) + ∇f (x, t) ,
∂
ϕ(x, t) → ϕ(x, t) − f (x, t) ,
∂t
where f is an arbitrary smooth function, together with a suitable transformation of ψ(x, t).
State such a suitable transformation of ψ(x, t) and check explicitly that (⋆) is invariant
under these transformations.

(b) In the presence of electric and magnetic fields, the probability current is modified
to
iℏ e
J (x, t) = −(ψ ∗ ∇ψ − ψ∇ψ ∗ ) + ψ ∗ ψA .
2m m
Show that J is gauge invariant and that it satisfies the conservation equation
∂ρ
+∇·J =0 ,
∂t
where ψ(x, t) satisfies the Schrödinger equation and ρ = ψ ∗ ψ.

(c) Consider the situation with constant electromagnetic fields, where B = (0, 0, B),
E = (0, E, 0), and B and E are constants. Verify that the vector and electric potential can
be chosen to have the form

A = B(−y, 0, 0) , ϕ = −Ey .

Restrict the motion to the (x, y) plane. Show that stationary states are given by

ψ(x, t) = eikx e−iEt/ℏ φ(y) ,

where φ(y) is a harmonic oscillator wavefunction (i.e. it satisfies the time-independent

Schrödinger equation with a harmonic oscillator potential), E is the energy and k is a real
constant. At what position is the oscillator centered?
Find the allowed energies, and show that they can be written as
 
1
E = ℏω1 n + +W ,
2

where n is a non-negative integer, and you should determine both ω1 and W explicitly.
Briefly discuss what happens to the Landau levels in the presence of this electric field.
Part II, 2024 List of Questions
 13

Part II, 2024 List of Questions [TURN OVER]

 14
Paper 1, Section II
28K Applied Probability
(a) Let (Xt )t⩾0 ∼ Markov(Q) be a continuous time Markov chain with generator
matrix Q on a countable state space I, jump times Jn and jump chain (Yn )n⩾0 . Show that
for n ⩾ 1 and i1 , i2 , . . . , in ∈ I,

qin P(Jn ⩽ t < Jn+1 |Y0 = i0 , . . . , Yn = in ) = qi0 P(Jn ⩽ t < Jn+1 |Y0 = in , . . . , Yn = i0 ) .

(b) Now let (Xt )t⩾0 be irreducible. Fix any h > 0 and let Zn = Xnh for
n = 0, 1, 2, . . .. Show that (Zn )n⩾0 is a discrete-time Markov chain and give its transition
matrix. Show that (Xt )t⩾0 is recurrent if and only if (Zn )n⩾0 is recurrent.
(c) Finally, let I be finite and f : I → R be a function, identified with the vector
(f (x))x∈I . Show that
Ex (f (Xt )) − f (x)
Qf (x) = lim ,
t→0+ t
and Z t
Ex (f (Xt )) = f (x) + Ex (Qf (Xs )) ds.
0

Part II, 2024 List of Questions

 15
Paper 2, Section II
28K Applied Probability
(a) Let X = (Xt )t⩾0 be a simple birth process with rate λn = nλ > 0 for all n ⩾ 1,
and X0 = 1.

(i) Show that X is non-explosive.

(ii) Let n ⩾ 1. Show that conditional on the event {Xt = n + 1}, the times of
the n births have the distribution of the order statistics of n i.i.d. random
variables with probability density function

λeλx
f (x) = , 0 ⩽ x ⩽ t.
eλt − 1

(b) Let Xt be a birth and death process with rates λn = nλ and µn = nµ for n ∈ N,
λ > 0, µ > 0, and assume that X0 = 1. Let h(t) = P(Xt = 0).

(i) Show that h(t) satisfies

Z t
h(t) = e−(λ+µ)s {µ + λ(h(t − s))2 } ds .
0

(ii) Show that h′ (t) satisfies

h′ (t) = (h(t) − 1)(λh(t) − µ) .

(iii) Find h(t) for λ ̸= µ.

Paper 3, Section II
27K Applied Probability
(a) State and prove Little’s formula for a regenerative process.
(b) What is an M/G/1 queue? Find the expected length of the busy period for an
M/G/1 queue.
(c) Suppose customers arrive at a single server at rate λ > 0 and require an
exponential amount of service time with rate µ > 0. Customers not being served are
impatient and will leave at rate δ > 0, independently of their position in the queue. Let
Xt denote the length of the queue at time t.

(i) Show that the system X = (Xt )t⩾0 has an invariant distribution.

(ii) Is X positive recurrent? Justify your answer.

Part II, 2024 List of Questions [TURN OVER]

 16
Paper 4, Section II
27K Applied Probability
(a) A train station has a platform where passengers arrive according to a renewal
process with rate µ > 0 (i.e. inter-arrival times have mean 1/µ). As soon as N passengers
arrive, a train dispatches with all N passengers on board. The process continues. For
c > 0, the company incurs a cost at the rate of nc per unit time when exactly n passengers
are waiting, and a fixed cost K each time a train departs. What is the optimum value of
N?
(b) A factory manufactures bulbs with independent lifetimes that are uniformly
distributed in [0, 3] (months). Whenever a bulb from this factory dies, it is immediately
replaced by a new bulb from the same factory. You observe a bulb from this factory at
the instant t. Let At be the time the bulb has been running (until the instant t) and let
Et be the remaining time the bulb will last (after instant t). For large t, find P(At ⩽ x),
P(Et ⩽ x) and P(At + Et ⩽ x) for x ∈ [0, 3].
(c) Let Π be thePpoints of a non-homogeneous Poisson process on Rd with intensity
functionR λ. Let S = x∈Π g(x) where R g is a2smooth non-negative function. Show that
E(S) = Rd g(u)λ(u) du and var(S) = Rd g(u) λ(u) du, provided these integrals converge.

Part II, 2024 List of Questions

 17
Paper 2, Section II
32C Asymptotic Methods
(a) Suppose f (x) is a real-valued function and the set of functions ϕn (x), where
n = 0, 1, 2, 3, . . . , forms an asymptotic sequence as x → x0 . What is meant by the
statement “f (x) has an asymptotic expansion as x → x0 , with respect to the ϕn (x)”?
Given that
∞
X
f (x) ∼ an ϕn (x) as x → x0 ,
n=0

show that
Pn−1
f (x) f (x) − k=0 ak ϕk (x)
a0 = lim and an = lim . (†)
x→x0 ϕ0 (x) x→x0 ϕn (x)

(b) Consider the asymptotic sequence ϕn (x), defined by ϕ0 = x−1 and ϕn (x) =
x−n+1 e−x for n ⩾ 1, as x → ∞, and the function

1 xe−x
f (x) = + .
x x−1
Find the asymptotic expansion of f (x) with respect to the ϕn (x) as x → ∞.
Verify explicitly that your coefficients satisfy (†) for all n.
What is the asymptotic expansion of f (x) with respect to the asymptotic sequence
ψn (x) = x−n as x → ∞?
(c) Consider the sine-integral function,
Z ∞
sin(xt)
si(x) = dt.
1 t

Using integration by parts, show that

∞
X ∞
X
−2n−1
si(x) ∼ cos x an x + sin x bn x−2n−2 as x → ∞,
n=0 n=0

where you should determine the coefficients an and bn .

Part II, 2024 List of Questions [TURN OVER]

 18
Paper 3, Section II
30C Asymptotic Methods

(a) State Watson’s lemma.

(b) Consider the integral

Z 1√
I(x) = t eixt dt ,
0
where x is real and positive. Use the method of steepest descent to show that
√ X ∞
π
I(x) ∼ (−ix)−3/2 − eix an (−ix)−n−1 ,
2
n=0

as x → ∞, where the an are coefficients that you should determine.

R ∞ √ −u √
[Hint: 0 ue du = π/2.]

Part II, 2024 List of Questions

 19
Paper 4, Section II
31C Asymptotic Methods
Consider the eigenvalue problem
 
d2 y λ2
ϵ2 2 − 1− 2 y = 0, (⋆)
dx x

defined for x ∈ [1, ∞), with eigenvalue λ > 1 and small parameter 0 < ϵ ≪ 1, and with
boundary conditions y(1) = 0 and y → 0 as x → ∞.

(a) Find the single turning point of this problem, x = xtp .

Write down the WKBJ form of the inner solution y1 , valid in the region xtp − x ≫ ϵ,
and the outer solution y2 , valid in the region x − xtp ≫ ϵ. You may quote relevant WKBJ
results from the lectures.
(b) Derive an approximate solution to (⋆) near the turning point. By matching it
to y1 , obtain the quantisation condition
Z λ 2 1/2  
λ 3
−1 dx = ϵ n + π, (†)
1 x2 4

for integer n. [You may quote the following asymptotic behaviour of the Airy function:
h i
Ai(t) ∼ π −1/2 (−t)−1/4 cos 23 (−t)3/2 − 14 π as t → −∞. ]

(c) We seek approximations to the smallest eigenvalues λ, which are close to 1.

Using the Taylor-expansion method in (†), or otherwise, show that λ ≈ 1 + αϵ2/3 , where
you should determine α. [Hint: Consider the substitution x = λ(1 − u).]
(d) Show that the largest eigenvalues, λ ≫ 1, can be approximated by solutions to
the transcendental equation  
3
λ ln λ = ϵ n + π.
4

Part II, 2024 List of Questions [TURN OVER]

 20
Paper 1, Section I
4J Automata and Formal Languages
In this question, let Σ = {a, b} and V = {S, A} be the set of terminal and non-
terminal symbols, respectively.
(a) Give an example of a context-free grammar G = (Σ, V, P, S) and a word w ∈ W
such that w has a unique G-parse tree starting from S, but exactly two different G-
derivations. Justify your claim.
(b) Suppose that G is a context-free grammar such that there is a G-parse tree
starting from A producing baba and the following two trees are G-parse trees:

S S
A S A A
a b A A a a b b
a b b b a b

For each of the following words w, prove that w ∈ L(G):

(i) w = abbabaab and

(ii) w = abbbaa.

(c) Prove that the language L := {an bm amin(n,m) : n, m > 1} is not context-free.
[You may use the context-free pumping lemma without proof.]

Paper 2, Section I
4J Automata and Formal Languages
(a) Define what an index set is and when it is called non-trivial.
(b) Define the index set Inf .
(c) State Rice’s theorem.
(d) Let X ⊆ W and let Inf X := {w ∈ Inf : ran(fw,1 ) ⊆ X}. Show, by modifying
the proof of Rice’s theorem or otherwise, that for each nonempty X, the set Inf X is not
computable.

Part II, 2024 List of Questions

 21
Paper 3, Section I
4J Automata and Formal Languages
(a) Provide an example of a grammar G = (Σ, V, P, S) such that for

Q := P ∪ {S → ε} and
H := (Σ, V, Q, S),

we have L(G) ∪ {ε} =

̸ L(H). Justify your claim.
(b) Let G = (Σ, V, P, S) be an arbitrary grammar. Let S ∗ ∈
/ V and V ∗ := V ∪ {S ∗ }.
If α → β ∈ P , let α → β be the rule where all instances of S are replaced by S ∗ (both
∗ ∗

on the left and on the right of the arrow). Let

P ∗ := {α∗ → β ∗ : α → β ∈ P } and
P + := P ∗ ∪ {S → S ∗ }.

Show that for G+ := (Σ, V ∗ , P + , S), we have that L(G) = L(G+ ). [You may use without
proof that isomorphic grammars produce the same language.]
(c) Again, let G = (Σ, V, P, S) be an arbitrary grammar and use the notation
from (b). Let

Q+ := P + ∪ {S → ε} and
H + := (Σ, V ∗ , Q+ , S)

and prove that L(G) ∪ {ε} = L(H + ).

Paper 4, Section I
4J Automata and Formal Languages
(a) Let D = (Σ, Q, δ, q0 , F ) and D′ = (Σ, Q′ , δ ′ , q0′ , F ′ ) be two deterministic
automata. Define what it means that f is a homomorphism from D to D′ .
(b) Prove that if f is a homomorphism from D to D′ , then L(D) = L(D′ ).
(c) Define the following partial order on deterministic automata: we write D ⩽ D′ if
and only if there is an injective homomorphism from D to D′ . Check whether the following
statements are true or false. Justify your answers.

(i) There is a deterministic automaton D such that, up to isomorphism, there

are only finitely many D′ such that D ⩽ D′ .

(ii) There is a deterministic automaton D such that, up to isomorphism, there

are only finitely many D′ such that D′ ⩽ D.

Part II, 2024 List of Questions [TURN OVER]

 22
Paper 1, Section II
12J Automata and Formal Languages
(a) Let M = (Σ, Q, P ) be a register machine and w ⃗ a finite sequence of words.
⃗ : t ∈ N}, the computation sequence
Define the upper register index of M and {C(t, M, w)
of M upon input w.
⃗ [When defining the computation sequence, you may assume that “M
transforms C into C ′ ” is already defined.]
(b) Let M = (Σ, Q, P ) be a register machine such that fM,1 = χL for some language
L ⊆ W. Show that for every n ∈ N there is a word w such that the computation sequence
of fM,1 (w) uses at least n remove instructions of the form −(0, q, q ′ ) for some q, q ′ ∈ Q.
A register machine with upper register index n is called an n-register machine (i.e.
a register machine using n + 1 registers). A language L ⊆ W is called n-computable if its
characteristic function is computable by an n-register machine. In the following, let us
assume that Σ = {a, b}.
(c) Show that L = {an bn : n > 0} is 1-computable. [You may use constructions
from the course, as long as you state them precisely and justify the scratch space that
they use.]
(d) Let M = (Σ, Q, P ) be a 0-register machine such that fM,1 = χL for some
language L. Show that there are natural numbers t, t′ , k, and ℓ and q ∈ Q such that k ̸= ℓ
and for all x ∈ W, we have

C(t, M, xbk ) = (q, x) = C(t′ , M, xbℓ ).

(e) Using part (d) or otherwise, show that L = {an bn : n > 0} is not 0-computable.

Paper 3, Section II
12J Automata and Formal Languages
(a) Define what it means for a language L ⊆ W to satisfy the regular pumping
lemma.
(b) Prove that every regular language satisfies the regular pumping lemma. [You
may assume that “regular” is equivalent to “accepted by a deterministic automaton”
without proof.]
(c) Let L be a finite language such that there is a w ∈ L with |w| = 100. Show that
there can be no deterministic automaton D with at most 100 states such that L = L(D).
(d) Let Σ = {0, 1, 2} and let L ⊆ {0, 1}∗ be an arbitrary language. Show that
b := {u2v : u ∈ W, v ∈ L} ∪ {0, 1}∗
L

satisfies the regular pumping lemma.

(e) Using part (d) or otherwise, provide an example of a language L such that L
satisfies the regular pumping lemma, but for any grammar G, we have that L ̸= L(G).

Part II, 2024 List of Questions

 23
Paper 1, Section I
8E Classical Dynamics
A rigid circular hoop of mass m and radius a hangs from a fixed point on its
circumference, is constrained to lie in a vertical plane and is free to oscillate within this
plane. A bead, also of mass m, can slide without friction around the hoop. [You may
assume that the moment of inertia of a circular hoop of mass m and radius a about an
axis through its circumference and perpendicular to the plane of the hoop is I = 2ma2 .]
(a) Choose a set of generalised coordinates and write down the Lagrangian for the
system.

p (b) Show that

pthe frequencies for small oscillations around equilibrium are ω1 =
c1 g/a and ω2 = c2 g/a, where c1 and c2 are positive real numbers that you should
determine.

Paper 2, Section I
8E Classical Dynamics
(a) In Lagrangian mechanics, explain what is meant by the generalised momentum
associated to a generalised coordinate q.
(b) What does it mean for a generalised coordinate to be ignorable? Show that the
generalised momentum associated to an ignorable coordinate is conserved. [You may state
the Euler–Lagrange equations without proof.]
(c) A certain system has generalised coordinates (q1 , q2 , q3 ) and Lagrangian
1 2
1 2

L= q̇1 + q̇22 + q̇32 − q1 + q22 + q32 − α (q1 q2 + q2 q3 + q3 q1 ) ,
2 2
where α is constant. Show that L is invariant under rotations around the (1, 1, 1) axis in
q-space. Hence find two conserved quantities.

Paper 3, Section I
8E Classical Dynamics
(a) Using spherical polar coordinates, (r, θ, ϕ), write down the Hamiltonian for a
particle of mass m moving in a spherically symmetric potential.
(b) Show that pϕ and p2θ +(p2ϕ / sin2 θ) are each conserved. Interpret them physically.
[You may state Hamilton’s equations without proof.]
(c) The particle executes circular motion in a plane through the origin inclined
at angle ψ to the plane θ = π/2. Evaluate pθ (θ) and show that it vanishes when
sin θ = ± cos ψ. Interpret this result physically.

Part II, 2024 List of Questions [TURN OVER]

 24
Paper 4, Section I
8E Classical Dynamics
(a) What is meant by the phase space of a mechanical system with n degrees of
freedom?
(b) Write down Hamilton’s equations in terms of a Poisson bracket on phase space.
(c) A particle of charge e and mass m moves through a magnetic field B in three
dimensions with electric field E = 0. Show that its equations of motion can be obtained
using the non-standard Poisson brackets

{qi , qj } = 0 , {qi , pj } = δij , {pi , pj } = e ϵijk Bk ,

|p|2
but with a free Hamiltonian, H = 2m , where qi and pi are generalised coordinates and
momenta, respectively.

Paper 2, Section II
14E Classical Dynamics
A mass m1 is suspended from a fixed point with coordinates (x, y, z) = (0, 0, 0) by
a spring with spring constant k1 . A second mass m2 is suspended from the first mass
by a spring with spring constant k2 . Each spring has natural length ℓ. The motion of
the masses is restricted to the (x, y)-plane, with gravity acting in the −y direction. The
position of the first mass is (x1 , y1 , 0) and the position of the second mass is (x2 , y2 , 0).
(a) Write down the Lagrangian of the system and hence determine the equations of
motion.
(b) Find the equilibrium position of each mass that has y2 < y1 < 0.
(c) For the remainder of this question suppose that the x-coordinate of mass m2 is
held fixed at its equilibrium value, and consider the case m1 = m2 = m and k1 = k2 = k.
One of the system’s normal modes has the first mass moving in the x direction with no
motion in the y direction and the other mass stationary. Show that this mode’s frequency
ω1 satisfies !
k 1 1
ω12 = 2 − 2mg − mg .
m kℓ + 1 kℓ + 1

Find the other normal modes and corresponding frequencies, showing that they are
independent of the strength of gravity.

Part II, 2024 List of Questions

 25
Paper 4, Section II
15E Classical Dynamics
Torque free rotation of a rigid body with principle moments of inertia (I1 , I2 , I3 ) is
described by Euler’s equations

I1 ω̇1 = (I2 − I3 )ω2 ω3 ,

I2 ω̇2 = (I3 − I1 )ω3 ω1 ,
I3 ω̇3 = (I1 − I2 )ω1 ω2 .

(a) Write down expressions for the kinetic energy T and total angular momentum
L, and show that they are each conserved.

√(b) Suppose that√ I1 < I2 and I3 = I1 + I2 , and that initially ω2 (0) = 0 while
ω1 (0) I2 − I1 = ω3 (0) I2 + I1 . Show that subsequently
  
2 2T 2 I2 − I1
ω̇1 = − ω1 ω12 .
I2 I2 + I1

(c) Hence show that

ω1 (t) = A sech(Ωt)
for constants A and Ω which you should find in terms of the kinetic energy and moments
of inertia.
(d) Describe the motion as t → ∞.

Part II, 2024 List of Questions [TURN OVER]

 26
Paper 1, Section I
3K Coding and Cryptography
Briefly describe the binary Huffman code for encoding symbols 1, 2, . . . , m occurring
with probabilities p1 ⩾ p2 ⩾ · · · ⩾ pm > 0.
Consider the discrete random variable X taking seven values xi (1 ⩽ i ⩽ 7) with
the following probabilities:
 
x1 x2 x3 x4 x5 x6 x7
X= .
0.49 0.26 0.12 0.04 0.04 0.03 0.02

Find a binary Huffman code for X. What is its expected word length? [You do not
need to simplify the expression.]

Paper 2, Section I
3K Coding and Cryptography
(a) Show that Hamming’s original code is perfect.
(b) Consider the code obtained by using Hamming’s original code for the first 7 bits
and the final bit as a check digit, so that

x1 + x2 + · · · + x8 ≡ 0 (mod 2).

Find the minimum distance for this code. How many errors can it detect? How many
errors can it correct?
(c) Given a code of length n which corrects e errors, can you always construct a
code of length n + 1 which detects 2e + 1 errors? Give a brief justification of your answer.

Paper 3, Section I
3K Coding and Cryptography
In the following, equivalent definitions of the Reed–Muller code can be used without
justification.

(a) Let n = 2d (d ⩾ 1). For 0 ⩽ r ⩽ d, state and prove a formula for the rank of the
Reed–Muller code RM(d, r) of length n. State its minimum distance.

(b) Consider the Mariner 9 code RM(5,1). What is its information rate? What
proportion of errors can it correct in a single codeword? How do these two properties
compare to the Hamming code of length 31?

(c) Show that all but two codewords in RM(d, 1) have the same weight.

Part II, 2024 List of Questions

 27
Paper 4, Section I
3K Coding and Cryptography

(a) (i) Consider a cryptosystem (K, M, C). Let e, d be the respective encryption
and decryption functions. Model the key and messages as independent
random variables K, M taking values in K, M, respectively and such that
M = d(C, K) ∈ M and C = e(K, M ) ∈ C. Show that H(M |C) ⩽ H(K|C).
(ii) Let M = C = A, where A is a finite alphabet. Suppose we send n messages
(M1 , . . . , Mn ) encrypted as (C1 , . . . , Cn ) using the same key. Define the
unicity distance. By making some reasonable assumptions, give a closed
formula for the unicity distance as a function of |K|, |A| and a certain constant.

(b) Suppose we model English text by a sequence of random variables (Xn )n⩾1 taking
values in A = {A, B, . . . , Z, space}. We define the entropy of English to be

HE = lim H(X1 , . . . , Xn )/n.

n→∞

Assuming HE exists, show that 0 ⩽ HE ⩽ log2 27. [You may assume Gibbs’
inequality.]

Paper 1, Section II
11K Coding and Cryptography
(a) Consider the use of a binary [n, m]-code to send one of m messages through a
binary symmetric channel (BSC) with error probability p, making n uses of the channel.
Define the following decoding rules: (1) ideal observer, (2) maximum likelihood, and (3)
minimum distance. Show that if all the messages are equally likely then (1) and (2) agree.
If p < 21 show that (2) and (3) agree.
1
(b) Show that a BSC with error probability p < 4 has non-zero operational capacity.
(c) State Shannon’s second coding theorem. Consider a discrete memoryless channel
with input X taking values over the alphabet {0, 1}. For a, b ∈ Z, let Z be a random
variable that is independent of X, taking values over the alphabet {a, b} with distribution
P(Z = a) = P(Z = b) = 12 . The output of the channel is Y = X + Z. What is the capacity
of this discrete memoryless channel? [Hint: The capacity depends on the value of b − a.]

Part II, 2024 List of Questions [TURN OVER]

 28
Paper 2, Section II
12K Coding and Cryptography

(a) (i) Describe briefly the Rabin cryptosystem, including how to encrypt and decrypt
messages. Show that breaking the Rabin cryptosystem is essentially as
difficult as factoring the public modulus, N .
(ii) Criticise the following authentication procedure:
Alice chooses N as the public modulus for the Rabin cryptosystem. To be sure
you are in communication with Alice, you send her a “random item” r = m2
(mod N ). On receiving r, Alice proceeds to decode using her knowledge of
the factorisation of N , and finds a square root m1 of r. She returns m1 to
you and you check that r = m21 (mod N ).

(b) (i) Describe briefly the RSA cryptosystem with public modulus N .
A budget internet company decides to provide each of its customers with
their own RSA ciphers using a common modulus N . Customer j is given
the public key (N, ej ) and sent secretly their decrypting exponent dj . The
company then sends out the same message, suitably encrypted, to each of its
customers. You intercept two of these messages to customers i and j where
ei and ej are coprime. Explain how you would decipher the message.
You are one of the customers, and so also know your own decrypting exponent.
Can you decipher any message sent to another customer?
(ii) Explain why it might be a bad idea to use RSA with public modulus N = pq
with |p − q| small.
A user of RSA accidentally chooses the public modulus N to be a large prime
number. Explain why this system is not secure.

Part II, 2024 List of Questions

 29
Paper 1, Section I
9D Cosmology
Consider a flat (k=0) FLRW universe dominated by the potential energy of a scalar
field ϕ given by
λ n
V (ϕ) = ϕ , where λ > 0,
n
and n is a positive integer. The evolution equations for the scale factor a(t) and the field
ϕ(t) in the slow-roll approximation are respectively

8πG
H2 = V (ϕ) ,
3c2
dV
3H ϕ̇ = −c2 ,
dϕ

where H = ȧ/a and a dot denotes differentiation with respect to time t.

(a) By considering the chain rule ȧ = (da/dϕ)ϕ̇, or otherwise, solve the slow-roll
equations to find the scale factor as a function of ϕ(t),
 
4πG 2

a (ϕ(t)) = exp 4 ϕi − ϕ(t)2 ,
c n

where ti is the initial time with a(ti )=1 and ϕi =ϕ(ti ), with ϕi assumed to be large enough
to ensure inflationary expansion.
(b) By determining the Hubble parameter H, show that during inflation we have

1 2 c4 n2 V (ϕ)
ϕ̇ ≈ .
2c2 48πG ϕ2
Deduce the approximate value of ϕ = ϕend when inflation ends, that is, when the slow-roll
approximation breaks down. If n = 6, roughly estimate the initial value ϕi relative to ϕend
that would be required to solve the flatness problem of the standard cosmology.

Part II, 2024 List of Questions [TURN OVER]

 30
Paper 2, Section I
9D Cosmology
A non-relativistic particle species in equilibrium with mass m, temperature T and chemical
potential µ satisfying kB T ≪ mc2 and µ ≪ mc2 , is described by the Maxwell-Boltzmann
distribution. This can be integrated over momenta to give the total number density
 3/2
2πmkB T  
n = gs exp (µ − mc2 )/(kB T ) ,
h2

where gs is the degeneracy.

(a) Deuterium is in equilibrium with non-relativistic protons and neutrons at around

t ≈ 100 seconds (kB T ≈ 0.1 MeV) through the interaction D ←→ n + p. Show that the
ratio of the number densities can be expressed as
 −3/2
nD πmp kB T
≈ eBD /(kB T ) ,
np nn h2

where the deuterium binding energy is BD = (mp + mn − mD )c2 = 2.2 MeV, and
mp , mn and mD are the proton, neutron and deuterium masses, respectively. [Hint:
The degeneracy factor for Deuterium is gD = 4.]

(b) Now use fractional densities relative to the baryon number density nB (e.g.
Xp = np /nB ) to find an expression for XD /(Xp Xn ). In this case, replace nB = η nγ where
η is the baryon-to-photon ratio and the photon number is

16πζ(3)
nγ = (kB T )3 ,
(hc)3

where ζ is the Riemann zeta function. Briefly explain how the fractional density ratio
XD /(Xp Xn ) offers insight into the “deuterium bottleneck”, that is, the delay in forming
deuterium nuclei to temperatures well below the binding energy, kB T ≪ BD ?

(c) In an alternative cosmology, the baryon-to-photon ratio η is larger. Assuming

that the decoupling of neutrons and protons is unaffected by this change, would the helium
abundance YHe be larger or smaller in this scenario than the standard result YHe ≈ 0.25?
Explain your reasoning.

Part II, 2024 List of Questions

 31
Paper 3, Section I
9D Cosmology
The Friedmann and acceleration equations for a universe without a cosmological
constant (Λ = 0) are given by

8πG 2 4πG
H2 + kc2 = ρa , H′ = − (ρ + 3P )a2 ,
3c2 3c2
where ρ is the energy density, P is the pressure, k is the curvature and primes denote
differentiation with respect to conformal time τ (defined by dτ = dt/a(t)). Here, a is the
scale factor and H is the conformal Hubble parameter defined by H ≡ a′ /a.
(a) Assume that the universe is filled with a single-component fluid with equation
of state P = wρ, where w is a constant. By introducing the density parameter
Ω = 8πGa2 ρ/(3c2 H2 ), show that the two evolution equations can be recast as

kc2

= Ω − 1, 2H′ + (1 + 3w) H2 + kc2 = 0 .
H2
Hence, find the following evolution equation for the density parameter,

Ω′ = (1 + 3w)HΩ(Ω − 1) . (†)

(b) Using the time evolution of Ω from equation (†), qualitatively describe the
flatness problem of an expanding universe (H > 0) for models with an equation of state
parameter in the range 0 ⩽ w ⩽ 1. In particular, roughly sketch the time evolution of Ω
in a radiation-filled universe with w = 1/3 taking initial values for both Ω < 1 and Ω > 1.
(c) Briefly discuss how the flatness problem can be alleviated by an early epoch of
inflation during which w ≈ −1.

Part II, 2024 List of Questions [TURN OVER]

 32
Paper 4, Section I
9D Cosmology
The energy density of photons ρ is given by the Stefan–Boltzmann law,
4σ 4
ρ= T ,
c
where T is the temperature and σ a constant. As the volume V of the Universe slowly
expands, the first law of thermodynamics relates the photon entropy S to the energy
E = ρV and the pressure P = ρ/3 (for vanishing chemical potential µ = 0),

dE = T dS − P dV .

(a) By substituting the Stefan–Boltzmann law, show that the entropy differential
becomes
16σ  3 
dS = T dV + 3T 2 V dT ,
3c
which should be integrated to find an expression for the photon entropy density s = S/V .
(b) If the interaction rate Γ maintaining the photons in equilibrium is much greater
than the Hubble expansion rate H (i.e. Γ ≫ H), briefly give the key reason why the
photon number and entropy are conserved as the Universe expands (provided the effective
number of degrees of freedom g∗ of particle species in equilibrium also does not change).
Why does the photon temperature fall as T ∝ 1/a, where a is the scale factor?
(c) Electrons and positrons annihilate and fall out of equilibrium after neutrino
decoupling at around kB T ≈ 1 MeV. By counting the effective number of degrees of freedom
g∗ in equilibrium before and after this process, provide a brief explanation for why the
photon temperature Tγ and neutrino temperature Tν are related today by
 1/3
Tν 4
= .
Tγ 11

Part II, 2024 List of Questions

 33
Paper 1, Section II
15D Cosmology
Consider a uniformly expanding universe with energy density ρ(t) and pressure P (t)
which obey the continuity equation
ȧ
ρ̇ = −3 (ρ + P ) , (⋆)
a
where a dot denotes a derivative with respect to time t.
(a) Consider the conserved mass M of matter inside a uniform expanding sphere
of radius r(t) = a(t) x0 , with fixed comoving radius x0 . Suppose that the radius of the
sphere satisfies
dΦ GM 1
r̈ = − , where Φ(r) = − − Λ r2 c2 ,
dr r 6
with Λ a constant. By multiplying the acceleration r̈ by the velocity ṙ and integrating,
show that the scale factor obeys the Friedmann equation
 2
ȧ 8πG kc2 1 2
= ρ − + Λc , (†)
a 3c2 a2 3

where k is a constant.

(b) Now differentiate the Friedmann equation (†) and substitute the continuity
equation (⋆) to find the acceleration equation for ä/a in terms of ρ, P and Λ. Briefly note
two of the shortcomings of this Newtonian analysis.

(c) Consider a flat (k=0) universe with a positive cosmological constant Λ > 0 that
is filled with radiation pressure PR = ρR /3, measured to have energy density ρR (t0 ) = ρR0
at given time t = t0 . Use the Friedmann equation (†) to show that the Hubble parameter
H = ȧ/a can be expressed as

1 8πG ρR0
H 2 = H02 ΩR0 a−4 + Λc2 , where ΩR0 ≡ ,
3 3c2 H02

with H(t0 ) =H0 and a(t0 ) = 1. By considering the substitution b = a2 (or otherwise) find
the solution for the scale factor

a(t) = α [ sinh(β t) ]1/2 ,

where α and β are constants you should determine in terms of H0 and ΩR0 . [You may
assume that the universe started with a big bang.]
Show that the scale factor a(t) gives anticipated results at early and late times.
Estimate the transition time tΛ that separates the decelerating and accelerating epochs.
R √
[Hint: dx/ 1 + κ2 x2 ) = (1/κ) sinh−1 (κx) + const, where κ > 0 is a constant.]

Part II, 2024 List of Questions [TURN OVER]

 34
Paper 3, Section II
14D Cosmology
In a flat expanding universe dominated by non-relativistic matter with energy
density ρ(x, t), we represent small density inhomogeneities δ(x, t), |δ(x, t)| ≪ 1, relative to
the homogeneous background ρ̄(t) = ρ̄0 /a(t)3 by the relation ρ = ρ̄(1+δ), with scale factor
a(t) (take a(t0 )=1 today). The continuity, Euler and Poisson equations are respectively

δ̇ + ∇ · [v(1 + δ)] = 0 ,
ȧ c2
v̇ + 2 v + (v · ∇)v = −∇ϕ − ∇P ,
a ρ
4πG
∇2 ϕ = 2 ρ̄ δ ,
c
where P (x, t) is the pressure and ϕ the perturbation of the Newtonian potential. Here, we
will neglect rotational modes and assume the velocity v is solely compressional, that is,
the divergence is the only non-vanishing part and θ ≡ ∇ · v. We also assume that pressure
effects can be described using the sound speed c2s defined by c2s ≡ c2 (dP/dρ).
(a) Show that for an equation of state given by P = P (ρ), the pressure term
in the linearised Euler equation can be written as (c2 /ρ)∇P ≈ c2s ∇δ. Linearise both
the continuity equation and the divergence of the Euler equation (using θ = ∇ · v) and
substitute the Poisson equation in the latter. Transforming to Fourier space with comoving
wavemodes k (i.e. ∇ → ik/a), combine these to obtain the evolution equation for density
perturbations in an expanding universe,
 
ȧ 4πG c2s k 2
δ̈ + 2 δ̇ − ρ̄ − 2 δ = 0, (†)
a c2 a

where k = |k|. Briefly explain the qualitative implications of each term in this equation.
Define the Jeans length λJ and discuss its significance.

(b) Suppose that the non-relativistic pressure is given by an equation of state

P = αρ4/3 where 0 < α ≪ 1; assume that this does not influence the background evolution
of a matter-dominated FLRW universe with ρ̄/c2 = 1/(6πGt2 ). Show that the varying
sound speed has the following time-dependence,

c2s (t) L20

≈ ,
a2 t2
where L20 is a constant. Using this sound speed, seek power law solutions δ ∝ tγ of the
density perturbation equation (†) to find the general solution
"  #1/2
1 5 2
δ(k, t) = Ak tn+ + Bk tn− , where n± = − ± − L20 k 2 .
6 6

Given k̃J = 5/(6L0 ), describe these solutions at late times in the asymptotic limits where
k ≪ k̃J and where k ≫ k̃J . What would be the consequence of choosing α such that
L0 ≈ 50 Mpc today?

Part II, 2024 List of Questions

 35

Part II, 2024 List of Questions [TURN OVER]

 36
Paper 1, Section II
26J Differential Geometry
Let Σ ⊂ R3 be a smooth surface.
(a) For a point p ∈ Σ, define the first fundamental form IpΣ and the second
fundamental form IIpΣ . Give the definitions of the shape operator, the mean curvature
and the Gaussian curvature at p. What does it mean for Σ to be minimal ?
Now let Ω be a non-empty open subset of R2 and let h : Ω → R be smooth with
non-degenerate differential. Let

ϕ : Ω → R3
(x, y) 7→ (x, y, h(x, y))

and let S = ϕ(Ω).

S
(b) Calculate Iϕ(x,y) at an arbitrary point (x, y) ∈ Ω in terms of h.

(c) Write down a Gauss map N : Ω → S2 for S. Let

ϕt : Ω → R3
(x, y) 7→ ϕ(x, y) + tN (x, y)

and assume that there is ϵ > 0 so that St := ϕt (Ω) is a smooth surface for any t ∈ (−ϵ, ϵ).
Prove that the second fundamental form of S satisfies
1d
S
IIϕ(x,y) =− IϕStt(x,y)
2 dt t=0

S
at any point (x, y) ∈ Ω. Calculate IIϕ(x,y) in terms of h.
(d) Derive a differential equation in h that characterises when the surface S is
minimal.
(e) Calculate the area of S in terms of the height function h. Assume that for any
η : Ω → R smooth and compactly supported in Ω, the area Aη (t) of

Stη := {(x, y, h(x, y) + tη(x, y)) | (x, y) ∈ Ω}

is locally minimal at t = 0. Use the Euler–Lagrange equation to recover the differential

equation in h from part (d).

Part II, 2024 List of Questions

 37
Paper 2, Section II
26J Differential Geometry
Consider a smooth closed curve α : I → S2 on the sphere, parametrised by arc
length.
(a) Define the curvature κ, torsion τ , Frénet trihedron (t, n, b) and geodesic
curvature κg of a general curve on a general surface. Prove in the particular case of
the sphere that α = −κ−1 n − τ −1 κ−2 κ̇b and κg = −κ−1 τ −1 κ̇.
(b) State the local Gauss–Bonnet theorem for the curve α on S2 . Deduce that,R given
a fixed length I = [0, L], the curve α maximising the enclosed area also minimises I κg .
(c) Consider φ : [0, L] → R smooth with compact support in {κg ̸= 0}, and
β : [0, L] → R3 defined by β = −φt − κ−1 φ̇n − κ−1 κ−1
g φ̇b. Prove that β⊥α and β̇⊥α̇.

(d) Consider the curve γ ϵ := (α + ϵβ)/|α + ϵβ| for small ϵ. You may assume that,
RL
if the value ϵ = 0 is a critical point of the enclosed area, then 0 κ−1
g φ̇ = 0. Deduce from
this equation that area-maximising curves have constant geodesic curvature.
(e) Prove that a curve α on S2 with constant geodesic curvature is planar by showing
that the vector e(s) := α(s) × α̇(s) + κg α(s) is constant.
(f) Deduce that, if a curve on S2 of length L encloses an area A, then L2 ⩾ A(4π−A)
(with the convention that we always choose the smaller of the two areas enclosed by the
curve).

Paper 3, Section II
25J Differential Geometry
Let n ⩾ 1 and 1 ⩽ k ⩽ N be integers. Let Id denote the identity matrix.
(a) State the definition of a k-dimensional smooth manifold X ⊂ RN . Define the
tangent space Tx X at a point x ∈ X in terms of a parametrisation about x, and prove
that the tangent space is independent of the parametrisation.
2
(b) Let GL(n) ⊂ Rn be the set of real n × n invertible matrices. Prove that it is a
manifold, give its dimension, and compute TId GL(n).
2
(c) Let SL(n) ⊂ Rn be the set of real n × n matrices with determinant 1. Prove
that it is a manifold, give its dimension, and compute TId SL(n).
2
(d) Let SO(n) ⊂ Rn be the set of real n × n matrices with determinant 1 and
orthonormal column vectors. Prove that it is a manifold, give its dimension, and compute
TId SO(n).

Part II, 2024 List of Questions [TURN OVER]

 38
Paper 4, Section II
25J Differential Geometry
(a) Let X be a smooth, compact, connected manifold. State the homotopy lemma
for smooth maps f, g : X → X and the homogeneity lemma for X. Prove that the number
of pre-images modulo 2 of a smooth map f : X → X is the same at every regular value.
Deduce a proof of the smooth Brouwer fixed-point theorem.
(b) Let n ⩾ 1 and A := {1 ⩽ ∥x∥ ⩽ 2} ⊂ Rn+1 . Consider a smooth map φ : Sn → Sn
such that φ(x)⊥x for all x ∈ Sn , and extend it to φ̃ : A → A by φ̃(x) := φ(x/∥x∥)∥x∥.
For ϵ > 0, define ψϵ : A → A by

x + ϵφ̃(x)
ψϵ (x) := √ .
1 + ϵ2
Prove that, for ϵ small enough, ψϵ is a diffeomorphism from A to A with det Dψϵ > 0.
R
(c) Prove A det(Id + ϵDφ̃) = vol(A)(1 + ϵ2 )(n+1)/2 , where vol(A) is the volume of
A.
(d) Deduce that, when n is even, there is no smooth map φ : Sn → Sn such that
φ(x)⊥x for all x ∈ Sn .

Part II, 2024 List of Questions

 39
Paper 1, Section II
32A Dynamical Systems
Let F : I → I be a continuous map of an interval I ⊂ R.

(a) (i) Define what it means for F to be chaotic, according to Glendinning.

(ii) Define what it means for F to be chaotic, according to Devaney.

(b) Suppose now that F has a periodic orbit of period 3.

(i) Show that F also has periodic orbits of period n for all positive integers n.
(ii) Explain briefly why F must have at least four distinct 7-cycles.
(iii) How many distinct 8-cycles must F have?

[Relevant theorems that you use from the course should be stated clearly.]

Part II, 2024 List of Questions [TURN OVER]

 40
Paper 2, Section II
33A Dynamical Systems
(a) State the normal form for a transcritical bifurcation in terms of the time t,
the dependent variable x and parameter µ. Illustrate using diagrams why this type of
bifurcation is not structurally stable, making sure that your diagrams are clearly labelled.
(b) Consider the system given by

ẋ = y − x + ax3 ,
ẏ = rx − y − zy,
ż = −z + xy,

where a and r are constants.

(i) Show that the fixed point at the origin of the system is non-hyperbolic at
r = 1.

(ii) Find the stable, unstable and centre subspaces of the linearised system of
the fixed point at the origin at r = 1.

(iii) Set r = 1 and change to new coordinates (v, w, z) where v = (x + y)/2,

w = (x − y)/2 and z is unchanged. Seek the (non-extended) centre manifold
by writing w = wc (v) and z = zc (v). Find wc and zc to fourth order in v.
[Hint: By considering symmetries, some of this calculation can be simpli-
fied.]

(iv) Show that the evolution equation on the centre manifold is of the form

a − 1 3 (3a − 1)(a + 3) 5
v̇ = v + v + ...
2 8

(v) For what values of a is the origin asymptotically stable when r = 1?

Part II, 2024 List of Questions

 41
Paper 3, Section II
31A Dynamical Systems
(a) A dynamical system ẋ = f (x) in R2 has a periodic orbit x = X(t) with
period T . The linearised evolution of a small perturbation x = X(t) + η(t) is given
by ηi (t) = Φij (t)ηj (0). Obtain the differential equation and initial condition satisfied by
the matrix Φ(t).
Explain how the stability of the orbit X(t) is connected to the quantity

Z T 
exp ∇ · f (X(t)) dt .
0

(b) Using the energy-balance method for nearly Hamiltonian systems, find a
condition on the parameter a for a limit cycle to exist in the system given by

ẋ = y,
ẏ = −x + ϵ(1 − x2 + ay 2 )y,

where 0 < ϵ ≪ 1. Determine the stability of the limit cycle.

R 2π R 2π
[Hint: 0 sin2 θ cos2 θ dθ = π/4 and 0 sin4 θ dθ = 3π/4. ]

Part II, 2024 List of Questions [TURN OVER]

 42
Paper 4, Section II
32A Dynamical Systems
(a) Consider a dynamical system ẋ = f (x) in R2 .

(i) State the Poincaré–Bendixson Theorem.

(ii) Explain why every periodic orbit must enclose at least one fixed point.

(b) Consider the system in R2 given by

ẋ = −x + ay + x2 y,
ẏ = b − ay − x2 y,

where a and b are real, positive parameters.

(i) Find the fixed point of the system. On a sketch of the (a, b)-plane (for the
quadrant where a, b > 0), show where the fixed point is stable.

(ii) A closed region D of the (x, y)-plane is given by the filled polygon with
vertices at (0, 0), (0, b/a), (b, b/a), (x∗ , y ∗ ) and (x∗ , 0), where (x∗ , y ∗ ) is
such that ẋ = 0 at that point and the line segment between (b, b/a) and
(x∗ , y ∗ ) has gradient −1.
Show that trajectories do not leave D.
[There is no need to determine (x∗ , y ∗ ) explicitly, but you may use that
b < x∗ .]

(iii) Use your results above to give conditions on the parameters a and b for the
system to have a periodic orbit.

Part II, 2024 List of Questions

 43
Paper 1, Section II
37D Electrodynamics
Consider a spacetime with coordinates xµ = (ct, x) and metric ηµν =
diag(−1, 1, 1, 1), where µ, ν = 0, 1, 2, 3 and c is the speed of light. A 4-vector potential
Aµ (x) fills spacetime and is described by the action
Z  
1 µν m2
S[Aµ ] = − 1
4 Fµν F + 2 Aµ A − µ0 Aµ J d4 x ,
µ µ
µ0 c

where Fµν = ∂µ Aν − ∂ν Aµ is the field strength tensor, Jµ (x) is a conserved 4-current

density, m and µ0 are constants and m ⩾ 0.
(a) Show that the equations of motion for the field,

∂µ F µν − m2 Aν = −µ0 J ν ,

follow from the principle of stationary action.

(b) Clearly state the conditions for the action to be invariant under Lorentz
transformations and under gauge transformations of the form Aµ → Aµ + ∂µ χ, where
χ is a scalar field.
(c) Show that for m > 0 the equations of motion imply the identity ∂µ Aµ = 0.
(d) Writing the vector potential as Aµ = (ϕ/c, A), show that for m ⩾ 0 the equations
of motion for ϕ and A can be written as
∂α
□ϕ + − m2 ϕ = −cµ0 J 0 , (†)
∂t

□A − ∇α − m2 A = −µ0 J ,
2
where α = 1 ∂ϕ
c2 ∂t
+ ∇ · A and □ = − c12 ∂t
∂ 2
2 + ∇ is the wave operator.

(e) For a point charge q at rest, the 4-current density is J 0 = cqδ(x) and J = 0,
where δ denotes the 3-dimensional δ function. By applying a Fourier transform in the
spatial coordinates to the equation of motion (†), show that for m > 0, a time-independent
field solution for a point charge at r = 0 is given by

exp(−mr)
ϕ=λ , A = 0,
r
where r = |x| and λ is a constant you do not need to determine. [Hint: You may use
without proof that the inverse Fourier transform of 1/(|k|2 + m2 ) is
Z
eik·x d3 k e−m|x|
= C ,
|k|2 + m2 (2π)3 |x|

where k is the wave vector and C is a non-zero constant.]

Provide a brief physical interpretation of this result, including the limiting case
where m → 0, and connect this interpretation to the result of part (b).

Part II, 2024 List of Questions [TURN OVER]

 44
Paper 3, Section II
36D Electrodynamics
The Maxwell equations for charge and current densities ρ(x, t) and J (x, t) are

ρ ∂B
∇·E = , ∇×E =− ,
ϵ0 ∂t
∂E
∇ · B = 0, ∇ × B = µ0 J + µ0 ϵ0 .
∂t
The energy and momentum density of the electromagnetic field are defined by
 
1 2 B2
E= ϵ0 E + and g = ε0 E × B .
2 µ0

(a) Use the Maxwell equations to show that g obeys the local conservation law

∂t gj + ∇i σij = −(ρE + J × B)j ,

with
σij = ã Ei Ej + b̃ E 2 δij + c̃ Bi Bj + d˜B 2 δij ,
where δij is the Kronecker delta and ã, b̃, c̃, d˜ are constants you should determine. [Hint:
A vector field a satisfies (∇ × a) × a = (a · ∇)a − 12 ∇(a2 ) ]
(b) Provide a brief physical interpretation of the tensor σij and, in particular, its
diagonal and off-diagonal components.
(c) The electric field of a homogeneously charged sphere with total charge q and
radius R centered on the origin is given by
(
q r
3 for r ⩽ R
E = E(r)er with E(r) = 4πϵ 0 R
q 1
4πϵ0 r2 for r > R ,

where r = |x| and er is the unit vector in the radial direction. The magnetic field B is
zero. Calculate the energy density E and, thus, the total energy Wem contained in the
electromagnetic field.
Under a Lorentzp transformation with velocity v = (0, 0, v) in the z′ direction and
2 2 ′
Lorentz factor γ = 1/ 1 − (v /c ), the electromagnetic field changes to E and B given
by

E ′|| = E || , E ′⊥ = γ(E ⊥ + v × B) ,
 
v E
B ′|| = B || , B ′⊥ = γ B ⊥ − × ,
c c

where the subscripts || and ⊥, respectively, denote field components parallel and perpendic-
ular to v. Compute the linear momentum P = (0, 0, Pz ) contained in the electromagnetic
field of a homogeneously charged sphere moving with v in the slow-velocity limit, i.e. ig-
noring terms of order (v/c)2 or higher.
R π [Hint: Vector fields a, b and c satisfy the relation
a × (b × c) = (a · c)b − (a · b)c and 0 sin θdθ = 43 .]
3

Part II, 2024 List of Questions

 45
Paper 4, Section II
36D Electrodynamics
(a) Define a dielectric medium and qualitatively explain how the polarization P
and magnetization M result in bound charge and bound current densities
∂P
ρbd = −∇ · P , J bd = +∇×M.
∂t

(b) Starting from the microscopic Maxwell equations for the electric and magnetic
fields E, B,
ρ ∂B
∇·E = , ∇×E =− ,
ϵ0 ∂t
∂E
∇ · B = 0, ∇ × B = µ0 J + µ0 ϵ0 ,
∂t
derive the macroscopic Maxwell equations for a dielectric medium in terms of the electric
displacement D and the magnetising field H, which you should define, as well as the free
charge and current densities.
(c) Consider a spherical shell with radial extent R1 < r < R2 , that consists of a
linear magnetic medium with constant permeability µ such that B = µH. This shell is
placed inside a magnetic field that approaches B = B0 ez , B0 = const, as r → ∞. The
electric field, the polarization, and the free charge and current densities are zero. Using
the ansatz B = ∇ψ, show that a solution for the magnetic field in all space is given by


 (a1 r + b1 /r2 ) cos θ for r < R1


ψ(r, θ) = (a2 r + b2 /r2 ) cos θ for R1 < r < R2



(a r + b /r2 ) cos θ
3 3 for r > R2 .

Using the boundary conditions at infinity, the origin, r = R1 and r = R2 , derive six
conditions for the free parameters a1 , b1 , a2 , b2 , a3 and b3 . Express the parameters b1 , a2 ,
b2 and a3 in terms of B0 , µ, R1 , R2 and a1 . Briefly describe how you would calculate the
remaining parameters a1 and b3 (you do not need to compute a1 and b3 ).
[Hint: In polar coordinates, the Laplace operator and the derivative in the Cartesian
z direction are
 
2 1 ∂2 1 ∂ ∂ 1 ∂2
∇ f = (rf ) + sin θ f + f,
r ∂r2 r2 sin θ ∂θ ∂θ r2 sin2 θ ∂ϕ2

∂ ∂ sin θ ∂
f = cos θ f− f .]
∂z ∂r r ∂θ

Part II, 2024 List of Questions [TURN OVER]

 46
Paper 1, Section II
39C Fluid Dynamics II
A solid cylinder of density ρ + ∆ρ, length L and radius a < L is placed with its axis
vertical at the centre of a long, closed, vertically oriented cylindrical container of radius
a + h, where h ≪ a. The container is otherwise filled with an incompressible fluid of
density ρ and dynamic viscosity µ, through which the solid cylinder falls with speed U .
Ignoring end effects, show that the downwards velocity field in the thin gap between the
solid cylinder and the container can be approximated as
∆p y
u=− y(h − y) + U ,
2µL h
where y is the coordinate directed inwards across the thin gap from the wall of the
cylindrical container and ∆p is the dynamic pressure difference between the fluid below
and above the cylinder. Determine the associated volume flux along the gap and the
viscous shear stress on the solid cylinder.
Use global mass conservation to show that ∆p ≈ 6µaLU/h3 . Show that the
associated form drag is much larger than the viscous force on the cylinder and hence
determine the speed of fall U .

Paper 2, Section II
39C Fluid Dynamics II
A two-dimensional incompressible Stokes flow has stream function ψ such that the
velocity u = ∇×(ψk), where k is the unit vector normal to the plane of the flow. Show
that
∇4 ψ = 0.

In plane polar coordinates (r, θ) the velocity is

1 ∂ψ ∂ψ
u= er − eθ .
r ∂θ ∂r
Given that the stream function has the form ψ = r2 f (θ), determine the rate-of-strain
tensor in terms of f and its derivatives. Hence write down the corresponding deviatoric
stress tensor for a fluid of dynamic viscosity µ.
Fluid with dynamic viscosity µ fills the two-dimensional region −α < θ < 0, r > 0,
where α > 0 is a constant. The boundary θ = −α is rigid, while a tangential stress S
is applied to the horizontal surface θ = 0. Given that the stream function has the form
ψ = r2 f (θ), write down the boundary conditions that apply to f (θ). Hence, determine
f (θ) and show that the surface velocity

Sr 1 − cos 2α − α sin 2α
U (r) = u(r, 0) = .
µ sin 2α − 2α cos 2α
 
2 1 ∂ ∂ψ 1 ∂2ψ
[Hint: In plane polar coordinates, ∇ ψ = r + 2 2 .]
r ∂r ∂r r ∂θ

Part II, 2024 List of Questions

 47
Paper 3, Section II
38C Fluid Dynamics II
A long, thin organism can be modelled as a body that occupies the region y = 0,
x > 0 in two dimensions, surrounded by an incompressible fluid. The organism is at rest
relative to the fluid that is far away from it, and it exerts a tangential stress Sx−1/2 , with
S > 0 a constant, on the surrounding fluid.

(a) Write down the boundary-layer equations describing steady high-Reynolds-

number flow around the organism. What boundary conditions apply to this flow?
(b) Use scaling to show that the width δ of the boundary layer in the y direction
satisfies δ ∼ (ρν 2 /S)1/3 x1/2 , where ρ and ν are the density and kinematic viscosity of the
fluid, respectively. What is the associated velocity scale in the direction parallel to the
organism?
(c) Determine the form of a similarity solution to the boundary-layer equations in
which the stream function is proportional to a dimensionless function f (η) of a suitable
similarity variable η. Determine the differential equation and boundary conditions satisfied
by f (η). Explain briefly how one could solve the differential equation numerically to
determine the velocity field and, specifically, the velocity of the fluid adjacent to the
organism.

Paper 4, Section II
38C Fluid Dynamics II
Given a material interface r = R(θ, t) between two fluid regions in plane polar
coordinates (r, θ), explain why
∂R v ∂R
+ = u,
∂t r ∂θ
where uer + veθ is the fluid velocity at the interface.
An ocean gyre is modelled as a two-dimensional, circular patch of water of radius a
in solid-body motion with angular velocity ω, surrounded by stationary water. Consider
small, sinusoidal perturbations to the edge of the gyre r = a + η(θ, t), where η =
ϵ exp(ikθ + σt) and ϵ ≪ ka. Assuming potential flow, determine the relationship between
the growth rate σ and the wave number k.
Briefly describe the subsequent motion of the gyre. Do disturbances propagate
upstream or downstream relative to the original motion of the gyre?
 
2 1 ∂ ∂ϕ 1 ∂2ϕ
[Hint: In plane-polar coordinates, ∇ ϕ = r + 2 2 .]
r ∂r ∂r r ∂θ

Part II, 2024 List of Questions [TURN OVER]

 48
Paper 1, Section I
7D Further Complex Methods
(a) Let f (x), x ∈ R, be a function with a finite number of singular points x = ck ,
k = 1, 2, . . . , N , where −∞ < a < c1 <R c2 < · · · < cN < b < ∞. Define what is meant by
∞
the Cauchy Principal Value integral P −∞ f (x)dx. [You may assume that the improper
Ra R∞
integrals −∞ f (x)dx and b f (x)dx exist.]
(b) What is the Hilbert transform H(f )(y) of a function f ?
1
(c) Let f (x) = x2 +1
. Evaluate H(f )(−1).

Paper 2, Section I
7D Further Complex Methods
Consider the differential equation

d3 y(x)
x + 2y(x) = 0 (†)
dx3
on the domain x ∈ (0, ∞).
(a) Write Z
y(z) = ezt f (t) dt,
γ

where γ is a contour in the complex plane, and substitute this expression for y into the
differential equation (†). Explain how the resulting integral equation can be solved by
finding an appropriate function f (t) and contour γ. Determine this function f (t) and
clearly state any required conditions on γ.
(b) Express the solution y(x) in integral form. [You do not have to evaluate this integral,
but you should simplify it as far as possible.] [Hint: Consider a subset of the real axis for
your choice of the contour γ.]

Part II, 2024 List of Questions

 49
Paper 3, Section I
7D Further Complex Methods 
Let F (z) be defined on the half plane H = z ∈ C : Re(z) > 0 such that

1. F (z) is analytic on H.

2. F (z + 1) = zF (z) for all z ∈ H,


3. F (z) is bounded in the strip z ∈ C : 1 ⩽ Re(z) ⩽ 2 ,

4. F (1) = 1.

(a) By using F (z + 1) = zF (z) for all z ∈ C, extend F meromorphically to z ∈ C

with Re(z) ⩽ 0. Identify and characterize the singular points of the extended function.
(b) Now consider a function Γ : H → C that also satisfies the four conditions listed
above. By meromorphically extending Γ in the same way as done for F in part (a), show
that the function f (z) = F (z) − Γ(z) is entire.
(c) Assuming additionally that f (z) is bounded on the strip 0 ⩽ Re(z) ⩽ 1, show
that the function S(z) = f (z)f (1 − z) is entire and bounded on C. You may conclude
without further proof that S(z) is therefore constant. Use this result to prove Wielandt’s
theorem, i.e. that F (z) = Γ(z) for z ∈ C.

Part II, 2024 List of Questions [TURN OVER]

 50
Paper 4, Section I
7D Further Complex Methods
The Papperitz equation is a second-order linear differential equation given by
 
d2 w 1 − α − α′ 1 − β − β ′ 1 − γ − γ ′ dw
+ + + (†)
dz 2 z−a z−b z−c dz
 
(b − c)(c − a)(a − b) αα′ ββ ′ γγ ′
− + + w=0
(z − a)(z − b)(z − c) (z − a)(b − c) (z − b)(c − a) (z − c)(a − b)

with the constraint α + α′ + β + β ′ + γ + γ ′ = 1 where α, α′ , β, β ′ , γ, γ ′ , a, b, c ∈ C and a, b, c

are pairwise distinct.
(a) State the general conditions for z0 ∈ C to be a regular singular point for a second-
order linear differential equation. Determine the regular singular points of the Papperitz
equation.
(b) Consider the Papperitz equation with non-integer α − α′ , β − β ′ and γ − γ ′ .
Derive the leading-order exponents of two linearly independent solutions of this equation
near each of its regular singular points.
(c) What is the Papperitz symbol for equation (†)?
(d) Consider now the hypergeometric equation
 
d2 w C 1 + A + B − C dw AB
+ + + w = 0,
dz 2 z z−1 dz z(z − 1)

where A, B, C ∈ C. This equation can be obtained as a special case of the Papperitz

equation by setting a = 0, b = 1 and taking the limit c → ∞. In the following you
may set c = ∞ without further proof or derivation. By comparing the coefficients of
the hypergeometric and the Papperitz equations, and setting α = 0, β = 0, establish a
Papperitz symbol for the hypergeometric equation.

Part II, 2024 List of Questions

 51
Paper 1, Section II
14D Further Complex Methods
(a) Consider a linear input–output system

Ly(t) = f (t),

where f (t) is the input, y(t) is the output and L is a linear operator. Define what it means
for the system to be causal and stable.
(b) Consider a linear ordinary differential equation for t > 0,

αy ′ (t) + y(t) = f (t) , (†)

with α ∈ R, α ̸= 0, and initial condition y(0) = 0.

(i) Using a Laplace transform, show that the transfer function of the linear
1
system (†) is G(s) = αs+1 . Determine for which values of α the system is
stable.

(ii) A negative feedback loop with H(s) = k, k ∈ R, is introduced into the

system for stable values of α such that the closed-loop transfer function is
G(s)
GCL (s) = 1+H(s)G(s) . By direct inspection of GCL (s), determine the values
of k for which the closed-loop system is stable. [You do not have to consider
the limiting cases for k.]
Explain the Nyquist criterion to determine the stability of a closed-loop
system. By calculating appropriate winding numbers, use the Nyquist
stability criterion to determine the values of k for which the closed-loop
system is stable. Compare the result with that obtained by direct inspection
of GCL (s). [Hint: You may use without proof that the number P of poles and
the number Z of zeros, counting multiplicities, of a meromorphic function
f (z) inside a clockwise simple closed contour γ obey the relation
I ′
1 f (z)
dz = P − Z ,
2iπ γ f (z)

if f (z) has no zeros or poles on γ.]

Part II, 2024 List of Questions [TURN OVER]

 52
Paper 2, Section II
13D Further Complex Methods
(a) When is a function f : C → C called elliptic? Define a fundamental cell explicitly
and state the property of an elliptic function regarding the number of its zeros and poles
in a fundamental cell.
(b) Show that an elliptic function without poles is constant. [You may use without
proof that an entire and bounded function is constant.]
(c) Let zj denote the poles of an elliptic function f in a fundamental cell. Show
that X
Res(f ; zj ) = 0.
j

Can there exist an elliptic function with a single pole of order one in a fundamental cell?
(d) If h is a meromorphic function on and inside a simple closed clockwise contour
γ and h has no zeros or poles on γ, then
I ′
1 h (z)
dz = P − Z ,
2πi γ h(z)

where P and Z denote respectively the number of poles and zeros, counting multiplicities,
of h(z) inside the contour γ. Using this relation, show that a non-constant elliptic function
f takes each value the same number of times in a cell, counting multiplicities.
(e) An example of an elliptic function is the Weierstrass function

1 X  1 1

P(z) = 2 + − ,
z (z − wm,n )2 wm,n2
(m,n)

where w1 , w2 ∈ C \ {0} with w w2 ∈

1
/ R, wm,n = mw1 + nw2 and the sum extends over
(m, n) ∈ Z × Z \ {(0, 0)}. Identify and characterize the singularities of the Weierstrass
function.
(f) The Laurent series of the Weierstrass function about z0 = 0 can be written as

X ∞
1
P(z) = 2 + a2k z 2k .
z
k=0

Calculate the coefficients a2k .

Part II, 2024 List of Questions

 53
Paper 1, Section II
18H Galois Theory
(a) Let α, β ∈ C be algebraic over Q. Show that if α and β have the same minimal
polynomial over Q then Q(α) ∼ = Q(β). Let k ⩾ 1 be the number of such isomorphisms.
Give an example where 1 < k < [Q(α) : Q]. Must k always divide [Q(α) : Q]? Justify
your answer.
(b) Let L/K be a finite extension of degree coprime to n. Clearly stating any
properties of the norm that you use, show that if α ∈ K is an nth power in L then it is
an nth power in K.
(c) Let K = Q(α) where α has minimal polynomial f over Q. Let p be a prime.
Show that f (X p ) is irreducible in Q[X] if and only if α is not a pth power in K.

Paper 2, Section II
18H Galois Theory
(a) Let L/K be a field extension. Explain what it means to say that

(i) L/K is finite;

(ii) L/K is separable;

(iii) L/K is simple.

Which pairs of these properties together imply the third? In each case give a proof or
counterexample.
(b) Let L be the splitting field of f (X) = X 3 − X − 1 over Q. Compute√Gal(L/Q).
Show that L has a unique quadratic subfield, and write it in the form √ Q( d) for d a
squarefree integer. Show also that if α is a root of f then L = Q(α + d).

Paper 3, Section II
18H Galois Theory
(a) Let M/L1 /K and M/L2 /K be finite extensions of fields. Define the composite
L1 L2 . Show that if L1 /K is Galois then L1 L2 /L2 is Galois, and that there is an injective
group homomorphism Gal(L1 L2 /L2 ) → Gal(L1 /K).
(b) Let K be a field of characteristic not dividing n, and with algebraic closure K.
Let ζn ∈ K be a primitive nth root of unity. Prove that [K(ζn ) : K] divides ϕ(n), and
that equality holds if K = Q.
(c) Let L1 = Q(ζ9 ) and L2 = Q(ζ15 ). Write each of the fields L1 ∩ L2 and L1 L2 in
the form Q(ζn ) for suitable n. Justify your answers.

Part II, 2024 List of Questions [TURN OVER]

 54
Paper 4, Section II
18H Galois Theory
(a) Give an example of a pair of monic quartic polynomials f, g ∈ Z[X] whose
splitting fields over Q are isomorphic, but whose Galois groups over Q are not conjugate
as subgroups of S4 .
(b) Let p be a prime and q = pd with d ⩾ 1. Show that there exists a field with q
elements, and that it is unique up to isomorphism. [Standard results about splitting fields
may be quoted without proof.] Show also that any irreducible factor of X q − X ∈ Fp [X]
has degree at most d.
(c) State a theorem which explains how reduction mod p may be used to help
compute Galois groups. The polynomial f (X) = X 4 − 21X 2 + 3X + 100 has discriminant
5472 . Compute Gal(f /Q).

Part II, 2024 List of Questions

 55
Paper 1, Section II
38B General Relativity

Consider a massive test particle moving in the Schwarzschild metric of a black hole
with mass m (in units with c = G = 1):
 2m  2  2m −1 2
ds2 = − 1 − dt + 1 − dr + r2 ( dθ2 + sin2 θ dϕ2 ) .
r r

(a) Assuming that the motion lies in the equatorial plane θ = π/2, justify briefly
why
1 2 m h2 mh2
h = r2 ϕ̇ and ṙ − + 2− 3
2 r 2r r
are constants of the motion, where dots denote derivatives with respect to the proper time
of the particle.
(b) For a circular orbit with a fixed value of r, determine h and hence deduce that
(i) r > 3m and (ii) (dϕ/dt)2 = m/r3 .
(c) Now consider a nearly circular orbit with shape given by u(ϕ) = 1/r. Let prime
denote differentiation with respect to ϕ, so that u′ = du/dϕ. Given that ṙ = −hu′ , and
assuming if needed that u′ ̸= 0, show that
m
u′′ + u = + 3mu2 .
h2
For m/h ≪ 1, this equation has an approximate solution of the form
m
u = (1 + α) + A cos[(1 + β)ϕ] ,
h2
where the constant A obeys |A| ≪ 1 but is otherwise arbitrary. The constants α and β
are small for m/h ≪ 1. Verify this solution by working to first order in A and determining
the constants α and β to leading non-trivial order in m/h.
Comment briefly on the significance of your result for β.

Part II, 2024 List of Questions [TURN OVER]

 56
Paper 2, Section II
38B General Relativity

(a) Define the Einstein tensor Gµν in terms of the Riemann tensor Rα βµν and use
the Bianchi identity ∇ρ R α βµν + ∇µ R α βνρ + ∇ν R α βρµ = 0 to show that ∇ν Gµ ν = 0.
Comment briefly on the significance of this result for consistency of the Einstein equations
(including a cosmological constant).
(b) For a universe described by the line element

ds2 = −dt2 + a(t)2 (dx2 + dy 2 + dz 2 ) ,

the Einstein tensor is diagonal with Gt t = −3ȧ2 /a2 and Gx x = −2ä/a − ȧ2 /a2 , where dots
denote differentiation with respect to t. Verify by direct computation that ∇ν Gµ ν = 0,
justifying the steps that you make and computing any metric connection components Γνµρ
that you may need.
Solve the vacuum Einstein equations with a cosmological constant Λ > 0 to obtain
a result for a(t) that corresponds to an expanding universe.

Part II, 2024 List of Questions

 57
Paper 3, Section II
37B General Relativity

(a) Consider a curve xα (λ) with tangent vector T α = dxα /dλ, in a spacetime with
metric gµν . For any vector fields U α and W α , show that, on the curve,

d
( gµν U µ W ν ) = (∇T U )α Wα + Uα (∇T W )α ,
dλ
where ∇V = V α ∇α denotes the covariant directional derivative along a vector V α .
(b) Now consider a one-parameter family of geodesics defined by the functions with
two arguments, xα (τ, σ). Fixing a value of σ in these functions gives a timelike geodesic
parametrised by proper time τ , with tangent vector T α = ∂xα /∂τ satisfying T α Tα = −1,
while fixing a value of τ gives a curve parametrised by σ, with tangent vector S α = ∂xα /∂σ.
Show that the commutator [T, S]α = (∇T S)α − (∇S T )α = 0 and hence derive the
equation of geodesic deviation in the form

(∇T ∇T S)α + E α β S β = 0 ,

where E α β is a tensor to be defined in terms of T α and the Riemann tensor Rα βµν .

(c) Suppose that in part (b),

Rαβµν = K( gαµ gβν − gαν gβµ ) ,

where K is a scalar. Show that, if T α Sα = 0 at some point on a particular geodesic, then

(∇T ∇T S)α − KS α = 0

at all points on that geodesic.

[Hint: You may use, without proof, standard properties of the metric connection,
symmetries of the Riemann tensor, and the Ricci identity in the form

(∇T ∇S V )α − (∇S ∇T V )α = (∇[T,S] V )α + Rα βµν V β T µ S ν

for vector fields T α , S α and V α .]

Part II, 2024 List of Questions [TURN OVER]

 58
Paper 4, Section II
37B General Relativity

(a) For a spacetime with metric gαβ , write down an explicit expression for the
metric-preserving Levi-Civita connection Γβαγ .
For a spacetime that is nearly flat, the metric can be expressed in the form

gαβ = ηαβ + hαβ ,

where ηαβ is a flat metric with constant components (though not necessarily diagonal in
the coordinates used) and the components hαβ and their derivatives are small. Show that
to leading order in these small quantities,

2Rαβ = A hα γ , γβ + B hβ γ , γα + C hγ γ , αβ + D hαβ, γρ η γρ ,

for constants A, B, C, D which you should determine. Indices are raised and lowered using
ηαβ .
(b) Consider the following metric

ds2 = 2 du dv + dx2 + dy 2 + H(u, x, y) du2 ,

where H(u, x, y) is a smooth function that is not necessarily small. You may assume that
the only nonzero connection coefficients are Γxvu , Γyvu , Γuxu , Γuyu , Γuvu , and the coefficients
related to these by symmetry. You may also assume that Ruu is the only coefficient of the
Ricci tensor that is not trivially zero. Compute Ruu and hence obtain the Ricci scalar.
Show that the full nonlinear vacuum field equations, with no cosmological constant, reduce
to a partial differential equation for H, that you should determine.
[You may assume in both parts of the question that

R αβ µ ν = Γναβ, µ − Γµαβ, ν + Γµαγ Γνγβ − Γναγ Γµγ β . ]

Part II, 2024 List of Questions

 59
Paper 1, Section II
17I Graph Theory
(a) Let G be a graph on n ⩾ 3 vertices with minimum degree at least n2 . Prove that
G is Hamiltonian.
(b) Now let G be a bipartite graph of minimum degree at least k ⩾ 2. Prove that G
contains either a path of length 2k or a cycle of length 2k. Give an example to show that
G need not contain a path of length 2k. Show also that G must contain either a path of
length 4k − 3 or a 4-cycle.

Paper 2, Section II
17I Graph Theory
State and prove Turán’s theorem.
A rhombus is the graph formed by two triangles sharing an edge. Prove that if G is
a graph on n ⩾ 4 vertices that has more edges than T2 (n), then G contains a rhombus.
Find a graph G on 6 vertices such that e(G) = e(T2 (6)) and G does not contain a
rhombus, but G is not isomorphic to T2 (6).

Paper 3, Section II
17I Graph Theory
(a) Show that every graph of average degree at least d contains a subgraph of
minimum degree at least d/2.
(b) Let δ, g ⩾ 3 be fixed. Using random graph methods, or otherwise, show that
there exists a graph G on at most n = (100δ)g vertices such that the minimum degree of
G is at least δ and the girth of G is at least g.

Part II, 2024 List of Questions [TURN OVER]

 60
Paper 4, Section II
17I Graph Theory
What does it mean to say that a graph G is strongly regular with parameters (k, a, b)?
Let G be a strongly regular graph with parameters (k, a, b) on n vertices such that b ⩾ 1
and G ̸= Kn . Prove the rationality condition, namely that the numbers
!
1 (n − 1)(b − a) − 2k
n−1± p
2 (a − b)2 + 4(k − b)

are integers.
What are the eigenvalues (with their multiplicities) of the Petersen graph (shown
below)?

Show that the set of edges of K10 cannot be partitioned into the edges of three
copies of the Petersen graph. [Hint: Suppose that it can be, and that the three Petersen
graphs have adjacency matrices A, B and C. You may wish to consider an appropriate
common eigenvector of A and B.]

Part II, 2024 List of Questions

 61
Paper 1, Section II
33C Integrable Systems
Consider the initial boundary value problem for a function u = u(x, t),

ut = iuxx , 0 < x < ∞, t > 0,

u(x, 0) = u0 (x), u(0, t) = h(t) .

Show that the equation ut = iuxx has a formulation as the consistency condition
for the following pair of equations for ψ = ψ(x, t) ∈ C,

ψt + ik 2 ψ = iux − ku,
ψx − ikψ = u.
2
By means of the integrating factor e−ikx+ik t , or otherwise, deduce that
Z t
ik2 t 2
û(k, t)e − û0 (k) = eik τ [ku(0, τ ) − iux (0, τ )] dτ ,
0

where
Z ∞ Z ∞
−ikx
û(k, t) = e u(x, t)dx, and û0 (k) = e−ikx u0 (x)dx .
0 0

Hence, by considering also û(−k, t), find a function G = G(k, τ ) such that
Z +∞
1 2
u(x, t) = e−ik t+ikx [û0 (k) − û0 (−k)] dk
2π −∞
Z Z
1 +∞ t −ik2 (t−τ )+ikx
+ e G(k, τ )dτ dk .
π −∞ 0

[You may assume that u is smooth and rapidly decreasing so that û(k, t) is holomorphic
for Im{k} < 0, and satisfies lim|k|→∞ û(k, t) = 0.]

Part II, 2024 List of Questions [TURN OVER]

 62
Paper 2, Section II
34C Integrable Systems
(a) Explain what it means to say the KdV equation ut + uxxx − 6uux = 0, where
u = u(x, t), has a Lax pair formulation in terms of the linear operators

L = −∂x2 + u and A = 4∂x3 − 3u∂x − 3∂x u .

(b) Consider now the case of periodic boundary conditions for the KdV equation, so
that at each time t the unknown u is a real-valued function satisfying u(x+2π, t) = u(x, t).
At each fixed time t, introduce a basis φ+ , φ− of solutions to the scattering equation
Lϕ = k 2 ϕ (for k 2 > 0) determined by the initial conditions at x = 0,

φ± (0) = 1, ∂x φ± (0) = ±ik .

(i) Show that there exists a matrix

 
a b
T̂ = ,
b a

where a(t) and b(t) are functions of time and a, b denote the complex
conjugates, such that
   
φ+ (x + 2π) φ+ (x)
= T̂ .
φ− (x + 2π) φ− (x)

Show further that |a|2 − |b|2 = 1.

(ii) Now as t varies let u evolve in time according to the KdV equation. Show
that there exists a matrix
 
λ µ
Λ= ,
µ λ

where λ(t) and µ(t) depend on time, such that

   
Aφ+ + ∂t φ+ φ+
=Λ .
Aφ− + ∂t φ− φ−

Prove that ∂t T̂ = [Λ, T̂ ].

 
φ+
[Hint: Consider ∂t (T̂ Ψ) = (∂t T̂ )Ψ + T̂ ∂t Ψ , with Ψ = .]
φ−
Deduce that Re[a] = 12 (a + a) is independent of t.

Part II, 2024 List of Questions

 63
Paper 3, Section II
32C Integrable Systems
Explain how to find Lie symmetries for an ordinary differential equation
∆(x, u, u′ , . . . , u(N ) ) = 0 by considering the action of the N th prolongation

pr (N ) V = V1 ∂x + ϕ ∂u + ϕ1 ∂u′ + · · · + ϕN ∂u(N )

of the vector field V = V1 (x, u)∂x + ϕ(x, u)∂u . Give the inductive formula for the
computation of the ϕj = ϕj (x, u, u′ , . . . , u(j) ) which determines the prolongation.
For the special case in which V1 = f (x) depends only on x, compute the third
prolongation, and show that

ϕ3 = ϕxxx +u′ (αϕxxu −f ′′′ )+3(u′ )2 ϕxuu +(u′ )3 ϕuuu +β(ϕxu −f ′′ +u′ ϕuu )u′′ +(ϕu −3f ′ )u′′′ ,

for integers α, β which you should determine.

Find the Lie symmetries of the equation

d3 u 1
3
= 3
dx u
that are generated by a vector field of the form V = f (x)∂x + ϕ(x, u)∂u .

Part II, 2024 List of Questions [TURN OVER]

 64
Paper 1, Section II
22G Linear Analysis
State and prove the Closest Point Theorem. Deduce that if F is a closed subspace
of a Hilbert space H then H is the direct sum of F and F ⊥ .
Let H be a separable Hilbert space. An operator T on H is called a shift if there
exists an orthonormal (Hilbert) basis (en )∞
n=1 of H such that T (en ) = en+1 for all n.
Show that T is a shift if and only if T is an isometry with ∩∞ n
n=1 Im (T ) = {0} and
dim (Im T )⊥ = 1.

Paper 2, Section II
22G Linear Analysis
State and prove the Baire Category Theorem.
Let X be a Banach space, and let S be a non-empty subset of X that is closed, convex
and symmetric (S symmetric means x ∈ S implies −x ∈ S). Show that if ∪∞ n=1 nS = X
then S is a neighbourhood of the origin.
Give an example to show that the condition that S is convex cannot be omitted.

Paper 3, Section II
21G Linear Analysis
State and prove the Stone–Weierstrass theorem. [You may assume that the function
x1/2 is uniformly approximable by polynomials on [0, 1].]
Let C(R) denote the space of all bounded continuous functions from R to R,
equipped with the uniform norm. Explain briefly why C(R) is a Banach space.
If A is a subalgebra of C(R) that contains the constants and separates the points,
must A be dense in C(R)? Justify your answer.

Part II, 2024 List of Questions

 65
Paper 4, Section II
22G Linear Analysis
In this question all spaces are complex.
(a) Let T be an operator on l2 . Prove that the spectrum of T is non-empty. [If you
use a power series expression for the resolvent function then you must prove it.]
(b) Let (en )∞n=1 be the usual basis of l2 . In each case below, an operator on l2 is
given; determine for each operator whether or not it is compact, and find its spectrum.
[Any results about spectra from the course may be used without proof, as long as they
are stated clearly.]

(i) T (en ) = en+1 for all n.

(ii) S(en ) = en−1 for all n > 1, with S(e1 ) = 0.

(iii) R(en ) = n1 en+1 for all n.

(iv) Q(en ) = n1 en−1 for all n > 1, with Q(e1 ) = 0.

Part II, 2024 List of Questions [TURN OVER]

 66
Paper 1, Section II
16I Logic and Set Theory
State the Soundness Theorem for Propositional Logic.
Let S be a consistent set of propositions. Explain briefly why the function
v : L → {0, 1} on the set L of all propositions defined by
(
1 if S ⊢ t
v(t) =
0 if S ̸ ⊢ t

need not be a model of S. Show that S is contained in a consistent, deductively closed

set T ⊂ L such that the definition of v above with S replaced by T is a model of S. [You
need not prove that v is a model. The set of primitive propositions here is arbitrary. You
may assume Zorn’s lemma.]
Show that if every finite subset of an arbitrary set S of propositions has a model,
then S has a model.
Let X, Y be infinite sets such that there is an injection from X to Y . For each
x ∈ X, let Ax be a non-empty, finite subset of Y . Let P be a set consisting of pairwise
distinct primitive propositions px,y for all x ∈ X and y ∈ Y . For a valuation v on L, set

fv = {(x, y) ∈ X × Y : v(px,y ) = 1} .

For each of the following statements, either write down a set S ⊂ L that makes the
statement true or prove that no such set S exists.

(i) {fv : v is a model of S} is the set of all injective functions from subsets of
X to Y .

(ii) {fv : v is a model of S} is the set of all injective functions g : X → Y such

that g(x) ∈ Ax for all x ∈ X.

(iii) {fv : v is a model of S} is the set of all injective functions from X to Y .

Part II, 2024 List of Questions

 67
Paper 2, Section II
16I Logic and Set Theory
State and prove Hartogs’ Lemma.
Define ordinal exponentiation. Show that αβ+γ = αβ · αγ for all ordinals α, β,
γ. Given ordinals γ and α, show that there exist unique ordinals β and δ such that
γ = ω α · β + δ with δ < ω α . [You may assume standard properties of ordinal addition and
multiplication. Other results used must be proved.]
Let X be a well-ordered set. Say x ∈ X is a limit point of X if Ix ̸= ∅ and Ix has
no greatest element, where Ix = {y ∈ X : y < x}. Let X ′ denote the set of limit points of
X and define X (α) for all ordinals α by recursion as follows:

X (0) = X

′
X (α+1) = X (α)
\
X (λ) = X (α) (for non-zero limit ordinal λ)
α<λ

Show that if X ̸= ∅, then X ′ ̸= X. Deduce that X (α) = ∅ for some α. The least such α is
the index of X.
Show that if ξ is an ordinal, then

ξ ′ = {γ < ξ : ∃ β > 0 such that γ = ω · β}.

Describe ξ ′′ . Find the index of ω and the index of ω 2 .

Paper 3, Section II
16I Logic and Set Theory
(a) State Zorn’s Lemma, the Axiom of Choice and the Well-ordering Principle, and
prove that they are equivalent.
(b) State Gödel’s Completeness Theorem and the Compactness Theorem for First-
order Logic.
Let T0 , T1 , . . . , Tn be first-order theories in some language L that partition the
collection of all L-structures in the sense that every L-structure is a model of exactly
one Ti . Show that each Ti is finitely axiomatisable.

Part II, 2024 List of Questions [TURN OVER]

 68
Paper 4, Section II
16I Logic and Set Theory
(a) In this part of the question, work within ZF.
Define the rank of a set. For a non-zero limit ordinal α, what is the rank of the set
of all functions from α to α?
Define the von Neumann hierarchy and prove that it exhausts the set-theoretic
universe, in the sense that every set is an element of some member of the von Neumann
hierarchy.
(b) In this part of the question, work within ZFC.
Define the cardinal numbers ℵα . Is every infinite cardinal equal to some ℵα ? Justify
your answer.
Prove that ℵα · ℵα = ℵα for every ordinal α. Deduce that

ℵα + ℵβ = ℵα · ℵβ = ℵβ

for ordinals α ⩽ β.
Prove that the following sentence is a theorem of ZFC, where t ≡ x is taken to mean
‘t and x have the same cardinality.’

(∀x)(¬(x = ∅) ⇒ ¬(∃y)((∀t)((t ≡ x) ⇒ (t ∈ y))))

Part II, 2024 List of Questions

 69
Paper 1, Section I
6A Mathematical Biology
Consider a model for population growth in which the population n(t) evolves
according to
dn
= αn − βn3 ,
dt
where α, β > 0.
(a) Find and analyse the stability of the non-negative fixed points.
(b) Sketch the solution starting from a range of (non-negative) population sizes.
(c) Suppose a population either follows the model above, or it follows logistic growth.
Explain briefly how these possibilities may be distinguished by observing population
dynamics starting from a very small population. Illustrate your answer with a sketch
showing the difference between the two models.

Paper 2, Section I
6A Mathematical Biology
The model of a viral infection in a population is given by the system
dX
= µN − βXY − µX,
dt
dY
= βXY − (µ + ν)Y,
dt
dZ
= νY − µZ,
dt
where µ, β and ν are positive constants and X, Y , and Z are respectively the number of
susceptible, infected and immune individuals in a population of size N , independent of t,
where N = X + Y + Z.
(a) Interpret the biological meaning of each of the parameters µ, β and ν.
(b) Show that there is a critical population size Nc (µ, β, ν) such that if N < Nc
there is no steady state with the infection maintained in the population. Show that in
this case the numbers of infected and immune individuals decrease to zero for all possible
initial conditions.
(c) Show that for N > Nc there is a steady state (X, Y, Z) = (X ∗ , Y ∗ , Z ∗ ) with
0 < X ∗ , Y ∗ , Z ∗ < N . Show that this steady state is stable.

Part II, 2024 List of Questions [TURN OVER]

 70
Paper 3, Section I
6A Mathematical Biology
Consider a birth-death process in a population of size n. The birth rate per
individual is λ and the death rate per individual is γ + βn, where λ, γ and β are positive
constants.
Let pn (t) be the probability that the population has size n at time t. Write down
the master equation for the system. Show that
d
⟨n⟩ = (λ − γ) ⟨n⟩ − β⟨n2 ⟩ ,
dt
where ⟨.⟩ denotes the mean.
From this result, deduce a bound on the mean in a steady state in the case that
λ > γ.
What can be said about the mean at steady state when λ ⩽ γ?

Paper 4, Section I
6A Mathematical Biology
A discrete-time model of alien cell population dynamics considers the coupled
dynamics of immature and mature cells. At each time step, a proportion λ of the immature
cells mature and a proportion µ of mature cells divides. When a mature cell divides it is
replaced by four new immature cells. Finally, a proportion k of mature cells die at each
time step.
(a) Explain briefly how this model may be represented by the equations

at+1 = (1 − λ)at + 4µbt ,

bt+1 = λat + (1 − µ − k)bt

where 0 ⩽ µ, λ, k ⩽ 1 and µ + k ⩽ 1.
(b) What is the expected number of offspring per cell?
(c) By considering the total population of cells, show that population growth is only
possible if 3µ > k. How does this relate to your answer to part (b)?
(d) At time T , the population is treated with a chemical that completely stops cells
from maturing for all t ⩾ T , but otherwise has no direct effects. Explain what will happen
to the population afterwards. In terms of aT and bT , what will be the total number of
cells in the long term?

Part II, 2024 List of Questions

 71
Paper 3, Section II
13A Mathematical Biology
An experiment is run with bacteria in a thin channel. The concentration of bacteria
at time t is given by C(x, t), where x is position in the channel. The ends of the channel
are at x = 0 and x = L. The bacteria cannot survive outside the channel. Within the
channel, the concentration of bacteria is modelled by

∂C ∂2C
= D 2 + µC ,
∂t ∂x
where D and µ are positive constants.
Initially, the bacteria are at a uniform concentration C0 for 0 < x < L.
(a) Suppose that the ends of the channel are open so that bacteria may diffuse out
of the ends.

(i) Find the concentration C(x, t) for t > 0.

(ii) What is the flux of bacteria out of the channel? Comment on this flux at
t = 0.

(iii) Show that the population will grow if µ > µc , where µc (D, L) should be
given. Give a brief explanation for the dependence of µc on D and L.

(iv) Sketch the bacterial concentration as a function of x for a range of times

(on the same sketch), paying particular attention to early and late times.
Consider separately the cases when µ < µc and µ > µc (i.e. two sketches
are required).

(b) The experiment is run again with the x = L end of the channel closed (x = 0
remains open). Again initially C(x, t) = C0 . What is the condition now for population
growth in the long term? Comment briefly on how this compares to the condition for
growth in part (a).
(c) The experiment is run once more with the channel ends both open, but now the
per capita growth rate is a function of time (as the experimental conditions fluctuate each
day), given by µ(t) = µ0 + µ1 cos(t). Find the condition for population growth in the long
term. Give a brief interpretation of this result.

Part II, 2024 List of Questions [TURN OVER]

 72
Paper 4, Section II
14A Mathematical Biology
An activator-inhibitor system is described by the equations

∂u ∂2u
= D1 2 + u − uv + u2 ,
∂t ∂x
∂v ∂2v
= D2 2 + αu2 − βuv,
∂t ∂x
where α, β, D1 , D2 > 0.
(a) Find conditions on α and β for the spatially homogeneous system to have a
stable stationary solution with u > 0 and v > 0. Sketch this region in the α–β plane (for
the quadrant with α, β > 0).
(b) Consider spatial perturbations of the form cos(kx) about the solution found in
part (a), and set λ = D1 /D2 . Find conditions for the system to be unstable. For fixed β,
sketch in the λ–α plane the region where spatial instability is possible for some k ∈ R.
(c) Consider a general system of the form

∂u ∂2u
= D1 2 + f (u, v),
∂t ∂x
∂v ∂2v
= D2 2 + g(u, v).
∂t ∂x

A Turing instability of this system is when a stationary solution is stable to

homogenous perturbations but is unstable to some spatial perturbation. Explain why
a Turing instability is not possible when D1 = D2 .
Verify that this is consistent with your answer for part (b) .

Part II, 2024 List of Questions

 73
Paper 1, Section II
31K Mathematics of Machine Learning

(a) (i) Let Z be a non-empty set. Given z1 , . . . , zn ∈ Z and a class F of functions

f : Z → R, what is meant by the empirical Rademacher complexity
R̂(F(z1:n ))? Given i.i.d. random variables Z1 , . . . , Zn taking values in Z,
what is meant by the Rademacher complexity Rn (F)?

(ii) Suppose H is a class of functions h : Rp → {0, 1} with |H| ⩾ 2. Define the

shattering coefficient s(H, n) and the VC dimension VC(H) of H.
 Qp
(iii) Let H = {1A : A ∈ A} where A := j=1 (−∞, aj ] : a1 , . . . , ap ∈ R . Show
that VC(H) ⩽ p.

(0) (1)
(b) A new painkilling drug is tested on n patients. Let Yi and Yi be the
pain levels, on a scale from 0 (no pain) to M > 0 (maximum pain), of the ith patient,
before and after taking the painkiller respectively. Suppose the vector Xi ∈ Rp records
p additional
 characteristics
 of the ith patient, such as their age, weight, height, etc. We
(0) (1)
treat Yi , Yi , Xi ∈ [0, M ]2 × Rp for i = 1, . . . , n as independent copies of a random


triple Y (0) , Y (1) , X . Let A and H be defined as in part (a) (iii) above. We wish to
determine a region A ∈ A where if X ∈ A, we expect the drug to be effective. To this
end, let h∗ and ĥ minimise
h  i 1 X  (1)
n 
(1) (0) (0)
Q(h) := E Y −Y h(X) and Q̂(h) := Yi − Yi h(Xi )
n
i=1

respectively, over h ∈ H.

(i) Using any results from the course that you need, show that

EQ(ĥ) ⩽ Q(h∗ ) + 2Rn (F)

for an appropriate class of functions F that you should specify.

(ii) Using any results from the course that you need, show that
r
2p log(n + 1)
EQ(ĥ) ⩽ Q(h ) + 2M
∗
.
n

Part II, 2024 List of Questions [TURN OVER]

 74
Paper 2, Section II
31K Mathematics of Machine Learning
(a) Carefully P describe the construction of the regions {R̂1 , . . . , R̂J } of a decision
J ′
tree x 7→ T̂ (x) = j=1 γ̂j 1R̂j (x) trained on data D consisting of input–output pairs
(Xi′ , Yi′ ) ∈ Rp ×R, i = 1, . . . , n. [You need not explain how computations may be performed
in a computationally efficient manner.]
(b) In the following, consider the data D′ as deterministic (i.e. not random). Let
data D := (Xi , Yi )ni=1 consist of i.i.d. input–output pairs, and let (X, Y ) ∈ Rp × R have
the same distribution as (X1 , Y1 ) and be independent of D. Set
J
X
T̃ (x) := γ̃j 1R̂j (x),
j=1
n
X n
X
1
where γ̃j := Yi 1R̂j (Xi ) and Nj := 1R̂j (Xi ).
Nj + 1
i=1 i=1

Let γj := E[γ̃j | X1:n ]. Show that

n
X
  1
E (γ̃j − γj )2 | X1:n = Var(Yi | Xi )1R̂j (Xi ).
(1 + Nj )2
i=1

Now suppose further that Var(Y | X = x) is bounded from above by σ 2 for all
x∈ Rp . Show that
h
2i σ2J
E T̃ (X) − E T̃ (X) | X, X1:n ⩽ .
n
[Hint: You may use without proof, the fact that if N ∼ Binomial(n, q), for success
probability q ∈ (0, 1], then E[1/(N + 1)] ⩽ 1/(nq).]

Part II, 2024 List of Questions

 75
Paper 4, Section II
30K Mathematics of Machine Learning
(a) Let f : Rn → R be a convex function. What is meant by the subdifferential
∂f (α) of f at α ∈ Rn ?
Prove that α minimises f if and only if 0 ∈ ∂f (α).
(b) Consider a classification setting with input–output pairs (x1 , y1 ), . . . , (xn , yn ) ∈
Rp × {−1, 1}. What is meant by the empirical ϕ-risk R̂ϕ (h) of hypothesis h : Rp → R
given convex surrogate loss ϕ?
(c) In the following, we consider hypothesis class H := {hβ : β ∈ Rp }, where
hβ : x 7→ x⊤ β. Fix λ > 0 and let β̂ be the minimiser (assumed to exist) of q : Rp → R
given by
q(β) = R̂ϕ (hβ ) + λ∥β∥22 .
Let X ∈ Rn×p be the matrix with ith row xi for i = 1, . . . , n and suppose that XX ⊤
is invertible. Writing P := X ⊤ (XX ⊤ )−1 X ∈ Rp×p , show that q(β) ⩾ q(P β). [Hint:
Consider the decomposition β = {P + (I − P )}β.]
Hence conclude that β̂ = X ⊤ α̂ where α̂ is the minimiser of r : Rn → R given by

r(α) = q(X ⊤ α).

(d) Now take ϕ to be the hinge loss. Writing ki ∈ Rn for the ith column of XX ⊤ ,
show that
n
1X
ki ti = 2λXX ⊤ α̂,
n
i=1

where ti is related to yi x⊤i β̂ in a way you should specify. Hence conclude that whenever
⊤
yi xi β̂ > 1, then α̂i = 0.

Part II, 2024 List of Questions [TURN OVER]

 76
Paper 1, Section II
20F Number Fields
Define algebraic integer.
Prove that the set of algebraic integers is closed under multiplication. [You may use
without proof any characterisation of algebraic integers, provided it is properly stated.]
√
What is the ring of integers in the number field Q( 2)? Prove your claim.
For a polynomial f ∈ C[x], we define
d
Y
M (f ) = |ad | max{1, |αj |},
j=1

where d is the degree, ad is the leading coefficient and α1 , . . . , αd are the roots of f in C.
Suppose f ∈ Z[x] is irreducible, M (f ) = 2, and f has a real root α1 > 1.
Prove that α1 is an algebraic integer and |NQ(α1 )/Q (α1 )| = 2. [Hint: Consider the
number |NQ(α1 )/Q (α1 )|/2, and show that it is a rational integer.]

Paper 2, Section II
20F Number Fields
State Dirichlet’s unit theorem.
√
Let K = Q( 5), and determine the units in OK . [You may use without proof the
description of OK , as long as you state it clearly.]
√
For K = Q( 5), what are the possible degrees of extensions L/K of number fields
× ×
such that 1 < |OL /OK | < ∞? Give an example for each possible degree.

Paper 4, Section II
20F Number Fields
Let K be a number field.
Define the norm N (I) of an ideal I in OK .
Let d = [K : Q]. Define the discriminant of a tuple (α1 , . . . , αd ) ∈ K. Define the
discriminant of K.
State a formula for the norm of an ideal in terms of discriminants without proof.
Prove that N (αOK ) = |NK/Q (α)| for all α ∈ OK .
Now let L/K be an extension of number fields, and let P ⊂ OK and Q ⊂ OL be
non-zero prime ideals.
(a) Prove that Q|P OL if and only if P = Q ∩ OK .
(b) Suppose that P = Q ∩ OK and [OK : P ] = [OL : Q]. Prove that if Q = αOL for
some α ∈ OL , then P = NL/K (α)OK .

Part II, 2024 List of Questions

 77
Paper 1, Section I
1F Number Theory
Let N ∈ N be an odd composite integer, and let b ∈ (Z/N Z)× . Define what it means
for N to be a Fermat pseudoprime to the base b, and for N to be an Euler pseudoprime
to the base b.
Now let N = 105. Determine the proportion of bases b ∈ (Z/N Z)× such that N
is a Fermat pseudoprime to the base b. Determine the proportion of bases b ∈ (Z/N Z)×
such that N is an Euler pseudoprime to the base b.

Paper 2, Section I
1F Number Theory
Let d be a positive integer.
Define what it means for a positive definite binary quadratic form f (x, y) =
ax2+ bxy + cy 2 to be reduced. If d is congruent to 0 or 3 mod 4, define the class number
h(−d).
Show that if d is odd and has k distinct prime factors, then h(−4d) ⩾ 2k−1 .
Give an example with d > 1 to show that the inequality h(−4d) ⩾ 2k−1 can be
strict.

Paper 3, Section I
1F Number Theory
Let N be an odd positive integer, and let a be an integer.



Define the Jacobi symbol Na , and write down a formula for 2
N
.
State

the law of quadratic reciprocity for the Jacobi symbol, and use it to com-
3
pute N in terms of the value of N modulo 12.
Show that if d is a positive integer such
that d
≡ 0 or 3 mod 4, and a, b are positive
odd integers such that a ≡ b mod d, then −d −d
a = b .

Part II, 2024 List of Questions [TURN OVER]

 78
Paper 4, Section I
1F Number Theory
Define the Möbius function µ : N → C, and state the Möbius inversion formula.
Let k > 1 be an integer. We say that n ∈ N is k th -power free if for all d ∈ N with
d > 1, dk does not divide n. Show that
X 
1 n is k th -power free,
µ(d) =
0 otherwise.
1⩽d⩽n
dk |n

Prove that for n ∈ N we have

X
µ(n) = e2πid/n .
1⩽d⩽n
(d,n)=1

Paper 3, Section II
11F Number Theory
Let k ∈ N.
Let p be an odd prime and Hk = {a + pk Z | a ≡ 1 mod p}. Show that Hk is a cyclic
group under multiplication, and determine its order.
In the following, I denotes the k × k identity matrix.
Let p be an odd prime and let A be a k × k matrix with integer entries. Suppose
that A = I + pm B for some m ∈ N and some k × k matrix B with integer entries, and
that An = I for some n ∈ N. Show that A = I.
Now find the smallest integer r ⩾ 1 such that if A is a k × k matrix with integer
entries satisfying A = I + 2r B for some k × k matrix B with integer entries, and An = I
for some n ∈ N, then in fact A = I.

Part II, 2024 List of Questions

 79
Paper 4, Section II
11F Number Theory
Let θ ∈ R be irrational.
Define the convergents (pn , qn )n⩾0 of the continued fraction expansion [a0 , a1 , a2 , . . . ]
of θ. Prove that pn /qn → θ as n → ∞, and that if n ⩾ 1 is odd then
pn−1 pn
<θ< .
qn−1 qn
√
Compute the continued fraction expansion of 7.
For each of the following equations, either find a solution in strictly positive integers
x, y, or prove that no such solution exists:

(i) x2 − 7y 2 = 1.

(ii) x2 − 7y 2 = −1.

(iii) x2 − 7y 2 = 5.

(iv) px /qx = 34/13, where (pn , qn )n⩾0

√ is the sequence of convergents of the
continued fraction expansion of 7.

Part II, 2024 List of Questions [TURN OVER]

 80
Paper 1, Section II
41A Numerical Analysis
(a) The Fourier transform of an infinite sequence (vm )m∈Z is defined as
X
vb(θ) = vm e−imθ , −π ⩽ θ ⩽ π.
m∈Z

(i) Prove Parseval’s identity:

X Z π
2 1
|vm | = v (θ)|2 dθ.
|b
2π −π
m∈Z

(ii) Consider the following two-step recurrence in n for unm with n ∈ Z+ and
m ∈ Z:
1 h i
un+1
m = (1 − µ)un−1
m + µ(u n
m+1 + u n
m−1 ,
)
1+µ
with µ ⩾ 0. Use Fourier analysis to determine the range of µ for which the
method is stable.

(b) The linear system dy/dt = Ay is discretised by the scheme

yn+1 = (I − kB)−1 (I − kC)−1 yn ,

with B + C = A and k = ∆t, where y ∈ RM and A, B and C are M × M square matrices.

(i) Define the exponential of a matrix. Show that for k ≪ 1,

exp[kB] exp[kC] = exp[k(B + C)] + 12 k 2 (BC − CB) + O(k 3 ).

(ii) Find the order of the local truncation error of the scheme.

(iii) In the special case when the matrices B and C commute, does the order of
the scheme change?

Part II, 2024 List of Questions

 81
Paper 2, Section II
41A Numerical Analysis P
(a) Let h : R → R be a 2-periodic function with Fourier series h(x) = n∈Z b
hn eiπnx .
R 1
Let I(h) = 12 −1 h(x) dx. For N ⩾ 1, consider the approximation

N
X
1
IN (h) = h(k/N ).
2N
k=−N +1

Find the error |IN (h) − I(h)| in terms of the b

hn .
Assuming |b
hn | ⩽ M c|n| for all n ∈ Z, where M > 0 and c ∈ (0, 1), show that the
error decays exponentially fast with N .
(b) Let w : R → R be a 2-periodic function with a finite Fourier expansion
X
w(x) = bn eiπnx .
w
|n|⩽d

Consider the partial differential equation for u(x, t)

∂u ∂ 2 u dw ∂u
= −
∂t ∂x2 dx ∂x
with initial condition u(x, 0) = u0 (x) which is 2-periodic. Seek an approximate solution
for u(x, t) for all t ⩾ 0 that is 2-periodic in x with an expansion of the form
X
u(x, t) = bn (t)eiπnx ,
u
|n|⩽D

bn (t) of the
where D is the truncation level. Write down a differential equation for the u
form
db
un (t) X
= Bnm u bm (t),
dt
|m|⩽D

for a matrix B that you should specify.

Assume that w(x) = cos(πx). Show that in this case, the eigenvalues of B have
non-positive real part. Is B invertible?

Part II, 2024 List of Questions [TURN OVER]

 82
Paper 3, Section II
40A Numerical Analysis
Let A be an n × n real symmetric positive definite matrix and b ∈ Rn . Consider
the linear system
Ax = b.

(a) An iterative method is defined by xk+1 = Hxk + v with n × n matrix H and

v ∈ Rn .

(i) State, giving a brief justification, necessary and sufficient conditions for xk
to converge to x∗ = A−1 b as k → ∞.

(ii) Give H and v in terms of A and b for the Jacobi method. Prove that the
Jacobi method is convergent if A is strictly diagonally dominant. [You may
use Gershgorin’s circle theorem without proof.]

(b) Define the function f : Rn → R by f (x) = 12 xT Ax − bT x.

(i) Show that the unique global minimizer of f is at x∗ = A−1 b. Show also
that for any x we have the identity

f (x) − f (x∗ ) = 21 ∇f (x)T A−1 ∇f (x) ,

where ∇f (x) is the gradient of f at x.

(ii) Consider the gradient method with exact line search to find the x that
minimises f (x):

xk+1 = xk − tk ∇f (xk ) where tk = arg min f (xk − t∇f (xk )) .

t∈R

Give an explicit expression for tk in terms of the residual rk = b − Axk .

(iii) Deduce that

 
∗ (rTk rk )2
f (xk+1 ) − f (x ) = 1 − (f (xk ) − f (x∗ )) .
(rTk A−1 rk )(rTk Ark )

Conclude that
 
l k
f (xk ) − f (x ) ⩽ 1 −
∗
(f (x0 ) − f (x∗ )) ,
L

where l and L are respectively the smallest and largest eigenvalues of A.

What does this inequality tell us in the case when n = 1?

Part II, 2024 List of Questions

 83
Paper 4, Section II
40A Numerical Analysis
(a) Describe the power method, and the inverse iteration method with shift s ∈ R.
Briefly explain the purpose of these algorithms.
(b) Let A be an n × n real symmetric matrix. The QR algorithm is as follows. Set
A0 = A. For k = 0, 1, . . ., set Ak+1 = Rk Qk where Rk and Qk are determined by the QR
factorisation Ak = Qk Rk . Here, Qk is an n × n orthogonal matrix and Rk is an n × n
upper triangular matrix with all of its diagonal elements strictly positive.

(i) Briefly explain the purpose of the QR algorithm.

(ii) Show that for all k, Ak is symmetric and has the same eigenvalues as A.
A matrix M is r-banded if Mij = 0 whenever |i − j| > r. Show that if Ak
is r-banded then so is Ak+1 .
e k = Q0 . . . Qk−1 and R
(iii) For any k ⩾ 1, let Q ek = Rk−1 . . . R0 . Show that
k e e
A = Qk Rk .
e k . How do these relate to the inverse
(iv) Consider the first and last columns of Q
iteration method and the power method?

Part II, 2024 List of Questions [TURN OVER]

 84
Paper 1, Section II
34B Principles of Quantum Mechanics
(a) A two-dimensional Hilbert space is spanned by two normalised vectors |ϕ⟩ and
|ψ⟩, which are not necessarily orthogonal. Consider the linear operator H defined by

H |ψ⟩ = g |ϕ⟩ , H |ϕ⟩ = g ∗ |ψ⟩ ,

where g is a complex constant. Determine the condition on ⟨ψ|ϕ⟩ under which H is

Hermitian. Henceforth assume this condition is satisfied and find the eigenvectors |±⟩ of
H and the corresponding eigenvalues. Verify that the distinct eigenvectors are orthogonal.
(b) Now assume g = 1. The Hamiltonian of a two-dimensional system is given by
H ′ = H + ∆(t) where H is as in part (a) and

∆(t) |ϕ⟩ = ∆(t) |ψ⟩ = V (t) (|ϕ⟩ + |ψ⟩) ,

with V (t) a real and time-dependent function. Working in the |±⟩ basis, or otherwise,
determine the exact probability to find the system in state |ϕ⟩ at time t > 0 if it was in
state |ψ⟩ at t = 0.

Paper 2, Section II
35B Principles of Quantum Mechanics
(a) State the defining properties of a density operator ρ in a Hilbert space of finite
dimension N , and state the number of real free parameters that determine ρ. Starting
from ρH in the Heisenberg picture, derive the time-dependent ρS (t) in the Schrödinger
picture.
(b) State the commutation relations for the spin operators S with each other and
with S · S . For the pure state |ψ⟩ of a spin- 21 qubit, you are given ⟨Sx ⟩ψ , ⟨Sz ⟩ψ , and the
sign of ⟨Sy ⟩ψ . Determine the normalised state |ψ⟩. [Hint: Recall that the state need only
be determined up to an overall phase, so that the normalised state can be parametrised by
a single complex number.]
(c) For a general mixed state of a spin- 21 qubit, you are given ⟨Sx ⟩, ⟨Sz ⟩, and ⟨Sy ⟩.
Determine the mixed state. Are the expectation values of any three linearly independent
Hermitian operators sufficient to fully specify a general mixed state? Present a proof or a
counterexample.

Part II, 2024 List of Questions

 85
Paper 3, Section II
33B Principles of Quantum Mechanics
(a) Let the components of the angular momentum operator be Ji , for i = 1, 2, 3.
Define the states |j, m⟩ in terms of the action of certain angular momentum operators and
state the possible values of j and m. Let the operators J± ≡ J1 ± iJ2 and let C± (j, m)
be defined by J± |j, m⟩ = C± (j, m) |j, m ± 1⟩. Considering particular cases, or otherwise,
explicitly derive the four constants λ± ±
1 and λ2 appearing in
q
C± (j, m) = j(j + 1) + λ± ±
1 m(m + λ2 ) .

Hence, or otherwise, prove that if O is a linear operator that commutes with all
components of the angular momentum operator, then

(i) ⟨j, m| O |j ′ , m′ ⟩ ∝ δjj ′

(ii) ⟨j, m| O |j ′ , m′ ⟩ ∝ δmm′

(iii) ⟨j, m| O |j, m⟩ is independent of m.

(b) Now consider a two qubit system with states |j1 , m1 ; j2 , m2 ⟩, with j1 = j2 = 1/2.
Let |j, m⟩ be the eigenstates of the total angular momentum of this system. State which
values of j and m occur and write down a resolution of the identity in terms of the |j, m⟩
eigenstates. Write down the values of the Clebsch–Gordan coefficients

C(j, m; j1 , m1 , j2 , m2 ) ≡ ⟨j, m|j1 , m1 , j2 , m2 ⟩

for all appropriate values of j, m, m1 , m2 .

Suppose that an operator O commutes with all the components of the total angular
momentum of this two qubit system. Using the results derived above and the results from
part (a), or otherwise, write down the most general expression for

j1 , m1 ; j2 , m2 O j1 , m′1 ; j2 , m′2

for all distinct cases where it does not vanish.

Part II, 2024 List of Questions [TURN OVER]

 86
Paper 4, Section II
33B Principles of Quantum Mechanics
(a) For a harmonic oscillator A of unit mass and frequency ω, write down the
Hamiltonian in terms of position and momentum operators (XA , PA ) as well as in terms
of creation and annihilation operators (aA , a†A ). Briefly describe the Hilbert space of the
system.
(b) Now consider the AB system comprising oscillator A and a second harmonic
oscillator B with the same frequency and unit mass. The two oscillators are coupled by
an interaction Hamiltonian Hint = λXA XB , where λ is a real parameter.

(i) Compute the corrections to both the ground state of the AB system and
its energy to linear order in λ.

(ii) Using the corrected ground state obtained in part (i), compute the density
operator of the full system and hence the reduced density operator ρA , by
tracing over B. Your final answer should be a density operator ρA that is
normalised up to and including O(λ2 ). [You may assume that corrections
to the ground state beyond linear order do not affect the normalised density
operator up to and including O(λ2 ).]

(iii) For any properly

normalised density operator ρ, the purity is defined by
γ = Tr ρ2 . Determine an upper and lower bound on γ for a general state
and determine its value for a pure state.

(iv) Compute the purity of ρA to order λ2 . For what values of λ does it satisfy
the upper and lower bounds? Interpret your findings.

Part II, 2024 List of Questions

 87
Paper 1, Section II
29L Principles of Statistics
(a) Let θ̂ denote the maximum likelihood estimator based on i.i.d. observations
X1 , . . . , Xn from a parametric statistical model {f (·, θ) : θ ∈ Θ}, where θ0 ∈ Θ is the true
√
parameter. Write down the limiting distribution of n(θ̂ − θ0 ) under standard regularity
conditions. [You should define any quantities involved in the expression of the limiting
distribution.]

Now suppose X1 , . . . , Xn are i.i.d. Uniform[−θ, θ] random variables, for a parameter

θ > 0.
(b) Derive an expression for the maximum likelihood estimator θ̂ of θ.
(c) Let θ0 denote the true parameter. By calculating the cumulative distribution
d
function of θ̂ or otherwise, show that n(θ0 − θ̂) → Z for a random variable Z whose
distribution you should specify.
(d) Find, with brief justification, the limiting distribution of n(θ̂2 − θ02 ).
√ d
(e) Do we have n(θ̂ − θ0 ) → W for a random variable W with positive variance?
Justify your answer.

Paper 2, Section II
29L Principles of Statistics
Let X1 , . . . , Xn be i.i.d. observations from a statistical model {f (·, θ) : θ ∈ Θ}
satisfying the usual regularity conditions, where Θ ⊆ Rp . Suppose we wish to test

H0 : θ = θ 0 vs H1 : θ ̸= θ0 .

(a) Define the Wald statistic Wn (θ) and write down a test for H0 based on Wn (θ0 )
with asymptotic type-I error bounded by a given α ∈ (0, 1).
(b) Define the likelihood ratio statistic Λn and write down a test for H0 based on
Λn with asymptotic type-I error bounded by a given α ∈ (0, 1).

(c) Suppose now that p = 1, i.e. θ is a scalar parameter, and we are under the null
H0 . By considering an appropriate Taylor expansion, show that the two test statistics
above are asymptotically equivalent, in the sense that
Λn P
→ 1.
Wn (θ0 )

[You may use, without proof, a uniform law of large numbers, as long as it is clearly
stated.]

Part II, 2024 List of Questions [TURN OVER]

 88
Paper 3, Section II
28L Principles of Statistics
Consider a binomial model X ∼ Binomial(n, θ), where θ ∈ Θ = [0, 1], and let
δMLE (X) = Xn denote the maximum likelihood estimator. In all of the following, we use
the weighted quadratic loss
(a − θ)2
L(a, θ) = .
θ(1 − θ)

(a) Compute the risk R(δMLE , θ), for each θ ∈ (0, 1).
(b) Suppose π is a prior for θ. What is meant by the π-Bayes risk of an estimator
δ of θ? Prove that any estimator of θ which minimises the posterior risk also minimises
the π-Bayes risk.
(c) Now suppose π is the uniform prior on [0, 1]. By differentiating the expression for
the posterior risk with respect to δ(x), prove that δMLE is the unique π-Bayes rule. [In your
calculations, you may interchange R 1 differentiation and integration without justification.
a−1 b−1
You may also use the formulas 0 θ (1 − θ) dθ = Beta(a, b) and Beta(a + 1, b) =
a
Beta(a, b) · a+b without proof.]
(d) Prove that the estimator δMLE is minimax. What does it mean for an estimator
δ of θ to be admissible? Is the estimator δMLE admissible? Justify your answer. [Results
from the course should not be used without proof.]

Part II, 2024 List of Questions

 89
Paper 4, Section II
28L Principles of Statistics
Let X1 , . . . , Xn be i.i.d. draws from a distribution on R, with µ = E(X1 ) and
2
σ = Var(X1 ) > 0.
P P
(a) Show that n1 ni=1 (Xi − X̄n )2 → σ 2 , where X̄n is the sample mean of the
observations.

Let the function g : R → R be differentiable at µ with g ′ (µ) ̸= 0. Consider the

plug-in estimator Tn = Tn (X1 , . . . , Xn ) = g(X̄n ) for the parameter θ = g(µ).
(b) Using the Delta method, provide a formula for σn such that

Tn − θ d
→ N (0, 1).
σn

Consider now the jackknife variance estimator

 2
n n
n − 1 X 1X
vJACK = Tn−1,i − Tn−1,j  ,
n n
i=1 j=1

where Tn−1,i = Tn−1 (X1 , . . . , Xi−1 , Xi+1 , . . . , Xn ), for i = 1, . . . , n, is the leave-one-

out estimator. In the following suppose further that g is twice-differentiable with
supx∈R |g ′′ (x)| = 2M < ∞.
(c) By considering a Taylor expansion of Tn−1,i − Tn , show that

n−1
vJACK = (An + Bn + 2Cn ),
n
where, for some Rn,i satisfying |Rn,i | ⩽ M (X̄n−1,i − X̄n )2 that you should specify,
 2
n
X n
X n
X
An = (g ′ (X̄n ))2 (X̄n−1,i − X̄n )2 , Bn = Rn,i − 1 Rn,j 
n
i=1 i=1 j=1

√
and |Cn | ⩽ An Bn . Here X̄n−1,i is the sample mean of the observations with Xi excluded.
P P P
Further assuming that n1 ni=1 (Xi − X̄n )4 → c < ∞, show that vJACK /σn2 → 1.
1
[Hint: Use the fact (which you need not derive) that X̄n−1,i − X̄n = n−1 (X̄n − Xi ).]
(d) Give, with justification, an asymptotically valid (1 − α)-level confidence interval
for θ based on the jackknife estimator.

Part II, 2024 List of Questions [TURN OVER]

 90
Paper 1, Section II
27G Probability and Measure

(a) (i) What does it mean for a measure µ on a measurable space (Ω, F) to be
σ-finite? State the uniqueness of extension theorem for a σ-finite measure.

(ii) Let B be the Borel σ-algebra on R. Show that Lebesgue measure λ is

translation invariant, i.e., for x ∈ R and B ∈ B,

λ(B) = λ(B + x)

where B + x = {b + x : b ∈ B}.

(iii) Show that Lebesgue measure λ is the unique translation invariant σ-finite
measure on B such that λ((0, 1]) = 1.

(b) Let X, (Xn )n∈N be real-valued random variables with distribution functions FX ,
(FXn )n∈N respectively.

(i) State what it means to say that Xn → X in distribution in terms of their

distribution functions.

(ii) Now assume that Xn → X in distribution. Let B(0,1) and λ|(0,1) be

the Borel σ-algebra and the Lebesgue measure on (0, 1) respectively. On
((0, 1), B(0,1) , λ|(0,1) ), define for all ω ∈ (0, 1),

X̃n (ω) = inf{x ∈ R : ω ⩽ FXn (x)} , X̃(ω) = inf{x ∈ R : ω ⩽ FX (x)} .

Show that X̃ has the same distribution as X and X̃n has the same
distribution as Xn for all n, and X̃n → X̃ almost surely.
[You may use the fact that for a non-constant, right-continuous, non-
decreasing function g, f (ω) := inf{x ∈ R : ω ⩽ g(x)} is left-continuous
non-decreasing and f (ω) ⩽ x if and only if ω ⩽ g(x). You may also use the
fact that a non-decreasing function has at most a countable set of points of
discontinuity.]

Part II, 2024 List of Questions

 91
Paper 2, Section II
27G Probability and Measure
(a) State the two Borel–Cantelli lemmas.
(b) Let (Xn )n∈N be independent exponential random variables with rate 1. Let
Mn = max1⩽m⩽n Xm . Show that

(i) lim sup Xn / log n = 1 almost surely;

(ii) lim inf Mn / log n ⩾ 1 almost surely.

[Hint: You may use without proof the inequality ex ⩾ 1 + x for all x ∈ R.]
(c) Let µ, ν be two measures on a measurable space (Ω, F) such that µ(Ω) < ∞. We
say µ ≪ ν if for any A ∈ F, ν(A) = 0 implies µ(A) = 0. Show that µ ≪ ν if and only if
for all ε > 0 there exists δ > 0 such that for any A ∈ F, µ(A) < ε whenever ν(A) < δ.

Paper 3, Section II
26G Probability and Measure

(a) (i) State Lévy’s convergence theorem for characteristic functions.

(ii) Let (Xn )n∈N , X be random variables on Rd . Show that Xn → X weakly

if and only if for all u = (u1 , . . . , ud ) ∈ Rd , ⟨u,P Xn ⟩ → ⟨u, X⟩ weakly. [For
a = (a1 , . . . , ad ), b = (b1 , . . . , bd ) ∈ Rd , ⟨a, b⟩ := di=1 ai bi .]

(b) Let X be a real-valued random variable with characteristic function ϕ(t) =

E(eitX ).

(i) Show that for any u > 0

Z  
1 u sin(uX)
(1 − ϕ(t)) dt = 2E 1 − .
u −u uX

(ii) Show that Z u

1
P(|X| > 2/u) ⩽ (1 − ϕ(t))dt .
u −u

(iii) Now let (Xn )n∈N be a sequence of real-valued random variables with
characteristic functions (ϕn )n∈N . If for all t ∈ R, ϕn (t) → ϕ(t) for some
function ϕ that is continuous at 0, show that given any ε > 0, there exists
M > 0 such that P(|Xn | > M ) ⩽ ε for all n.

[You may use convergence results for integrals given in the course and Fubini’s
theorem, without proof.]

Part II, 2024 List of Questions [TURN OVER]

 92
Paper 4, Section II
26G Probability and Measure
(a) Define what it means for a sequence of real-valued random variables (Xn )n∈N
to be uniformly integrable. Show that if (Xn )n∈N is bounded in Lp for some p > 1, then
(Xn )n∈N is uniformly integrable. Give, with justification, a counterexample to show that
a sequence of random variables bounded in L1 need not be uniformly integrable.
(b) Let (Xn )n∈N be an identically distributed sequence of real-valued random
variables in L1 . Let Sn = X1 + · · · + Xn . Show that the sequence (Sn /n)n∈N is uniformly
integrable. Assuming that Sn /n → E(X1 ) in probability, prove that Sn /n → E(X1 ) in L1 .
(c) Now assume that (Xn )n∈N is an identically distributed sequence of random
variables in L2 . Let Mn = maxk⩽n |Xk |.

(i) Show that n−1/2 Mn → 0 in probability as n → ∞.

(ii) Show that n−1/2 E(Mn ) → 0 as n → ∞.

Part II, 2024 List of Questions

 93
Paper 1, Section I
10E Quantum Information and Computation
Let Uf denote a quantum oracle which acts on (n+1) qubits as follows: ∀ x ∈ {0, 1}n
and y ∈ {0, 1},
Uf |x⟩ |y⟩ = |x⟩ |y ⊕ f (x)⟩ ,
where the addition is taken modulo 2 and
n
X
f (x) = a.x ⊕ b := ai xi ⊕ b for some a ∈ {0, 1}n and b ∈ {0, 1}.
i=1

(a) (i) Write down an expression for the state |Φ1 ⟩ obtained when the state |0⟩⊗n |1⟩
of (n + 1) qubits is acted on by (Uf H ⊗(n+1) ), where H denotes the Hadamard
gate.
(ii) Let |Φ2 ⟩ := H ⊗(n+1) |Φ1 ⟩. Write an expression for this state.
(iii) Let |Φ3 ⟩ := Uf |Φ2 ⟩. Write an expression for this state.

(b) (i) |Φ3 ⟩ is a state of n + 1 qubits. What is the probability of obtaining the n-bit
string a by doing a measurement of the first n qubits in the computational
basis?
(ii) What is the state of the last (i.e., the (n + 1)th ) qubit after the above
measurement?
(iii) Find the probability that a measurement on this qubit yields the value of b
when a contains an odd number of 1s and when a contains an even number
of 1s.

Part II, 2024 List of Questions [TURN OVER]

 94
Paper 2, Section I
10E Quantum Information and Computation
Consider a function f : Z12 → Z9 defined by

f (x) = 4x mod 9.

(a) The function f is periodic. Find its period r.

(b) Suppose we are given the following quantum state of 2 registers:
1 X
|f ⟩ := √ |x⟩ |f (x)⟩ ,
12 x∈Z
12

and a measurement of the second register yields a value y.

What are the possible values of y and what are the corresponding probabilities?
(c) If y = 4, find the resulting state |α⟩ of the first register after the above
measurement.
(d) Let QFT12 denote the quantum Fourier transform modulo 12. How does it act
on a state |x⟩ for x ∈ Z12 ?
(e) Suppose a measurement of the state QFT12 |α⟩ yields a value c. What are the
possible values of c and what are the corresponding probabilities?

Part II, 2024 List of Questions

 95
Paper 3, Section I
10E Quantum Information and Computation
Recall the P
Schmidt decomposition theorem for bipartite states: For any bipartite
n Pm
j=1 Xij |i⟩|j⟩ of an n−dimensional system A ≃ C
state |ψAB ⟩ = n and an
i=1
m−dimensional system B ≃ C , there exist (non-unique) orthonormal bases {|αi ⟩}ni=1 ⊂
m

Cn , {|βj ⟩}m
j=1 ⊂ C , and a (unique) set of non-negative numbers λ1 , λ2 , . . . , λd (for
m

d = min{n, m}) such that

Xd
|ψAB ⟩ = λi |αi ⟩|βi ⟩.
i=1

The number of non-zero λi is called the Schmidt rank of |ψAB ⟩ and is equal to rank(X).

(a) Using the Schmidt decomposition theorem, show that a bipartite state |ψAB ⟩ is
entangled if and only if its Schmidt rank is at least 2.

(b) Consider the following tripartite states of 3 qubits:

(1)
|χABC ⟩ = |0A ⟩|0B ⟩|0C ⟩ ,
(2)
|χABC ⟩ = |0A ⟩|ϕ+
BC ⟩ ,
(3) 1
|χABC ⟩ = √ (|0A ⟩|0B ⟩|0C ⟩ + |1A ⟩|1B ⟩|1C ⟩) .
2

Here, |ϕ+
BC ⟩ =
√1 (|0B ⟩|0C ⟩
2
+ |1B ⟩|1C ⟩) is a Bell state of two qubits. Show that

(1)
(i) |χABC ⟩ is product across all three bipartitions,
(2)
(ii) |χABC ⟩ is product across the A − BC bipartition, and is entangled across the
B − AC and the C − AB bipartitions,
(3)
(iii) |χABC ⟩ is entangled across all three bipartitions.
(3)
(c) Suppose that Alice (A), Bob (B), and Charlie (C) share either the state |χABC ⟩
(4)
or |χABC ⟩ = √13 (|0A ⟩|0B ⟩|1C ⟩ + |0A ⟩|1B ⟩|0C ⟩ + |1A ⟩|0B ⟩|0C ⟩). Charlie performs a
measurement in the computational basis and obtains the result 0. Show that the
resulting state of Alice and Bob will be
(3)
(i) product in the case they share |χABC ⟩,
(4)
(ii) entangled in the case they share |χABC ⟩.

Part II, 2024 List of Questions [TURN OVER]

 96
Paper 4, Section I
10E Quantum Information and Computation
(a) Verify that for any 2 × 2 matrix, A, the following relation holds

(I ⊗ A) Φ+ = (AT ⊗ I) Φ+ ,

where |Φ+ ⟩ = √12 (|00⟩ + |11⟩), I denotes the 2 × 2 identity matrix and T denotes
transposition taken with respect to the standard basis {|0⟩ , |1⟩}.

(b) Let θ ∈ [0, 2π),  

cos θ − sin θ
Uθ = ,
sin θ cos θ
|ψ1 ⟩ = Uθ |0⟩ and |ψ2 ⟩ = Uθ |1⟩ .

(i) Show that ZX |ψ2 ⟩ = |ψ1 ⟩, where X and Z are the usual one-qubit gates.

(ii) Show that the Bell state |Φ+ ⟩ can be written as √1 (|ψ1 ⟩ |ψ1 ⟩ + |ψ2 ⟩ |ψ2 ⟩).
2
[Hint: Use part (a).]

(iii) Suppose Alice and Bob initially share the Bell state |Φ+ ⟩, with the first
qubit being with Alice and the second qubit being with Bob. Alice then
applies Uθ−1 to her qubit and then measures it in the computational basis.
What are Alice’s possible outcomes and what are the corresponding states
of Bob’s qubit after Alice’s measurement?

(iv) Suppose Alice knows the value of θ but Bob does not. Give a protocol
by which Alice can transmit |ψ1 ⟩ to Bob by sending just one classical bit
to him, given that they share the Bell state |Φ+ ⟩. (This is in contrast to
general teleportation of an unknown qubit, which uses one Bell state and
two classical bits of communication.)

Part II, 2024 List of Questions

 97
Paper 2, Section II
15E Quantum Information and Computation
(a) Consider a circuit which uses a controlled unitary gate with unitary operator U :

|0⟩ H H

|ψ⟩ U

Here H represents a Hadamard gate, |ψ⟩ denotes a single qubit state which is an eigenstate
of U , U |ψ⟩ = e2πiθ |ψ⟩, and the measurement is done in the computational basis. Express
the phase θ in terms of the probability of the measurement giving zero.
(b) Consider the following circuit which uses a controlled SWAP gate. A SWAP
gate acts as: SWAP |i⟩ |j⟩ 7→ |j⟩ |i⟩ ∀ |i⟩ , |j⟩ ∈ C2 .

|0⟩ H H

|ϕ1 ⟩
SWAP
|ϕ2 ⟩ V

Here V is a unitary operator, |ϕ1 ⟩ , |ϕ2 ⟩ are single qubit states, and the measurement is
done in the computational basis. Find the probability of the measurement giving zero.
(c) Consider the operator U and the state |ψ⟩ introduced in part (a). It is given
that θ = j/2m for some j ∈ {0, 1, 2, . . . , 2m − 1} and some m ∈ N. Define the controlled
unitary operator Θm (U ) which acts on a state |k⟩ |ψ⟩ of m + 1 qubits as:

Θm (U ) |k⟩ |ψ⟩ = |k⟩ U k |ψ⟩ ,

where U k |ψ⟩ is the state obtained by k successive applications of the operator U on the
state |ψ⟩ and k ∈ {0, 1, 2, . . . , 2m − 1}.

(i) Find an expression for the following state of (m + 1) qubits:

|Φ⟩ := Θm (U ) H ⊗m ⊗ I |0⟩⊗m |ψ⟩ .

[You should write out the result of applying the operators to |0⟩⊗m |ψ⟩.]

(ii) Write an expression for the corresponding state |ϕj ⟩ of the first m qubits.
m
Show that {|ϕj ⟩}2j=0−1 is an orthonormal basis.

(iii) Let F be an operator acting on m qubits as F |j⟩ = |ϕj ⟩. Justify why F is

a unitary operator.

(iv) State the 2 sequential operations that you can do on |ϕj ⟩ to find the value
of j. What is the probability pj of finding the value of j (and hence θ)?

[QUESTION CONTINUES ON THE NEXT PAGE]

Part II, 2024 List of Questions [TURN OVER]

 98
15E Quantum Information and Computation
(a) Consider a circuit which uses a controlled unitary gate with unitary operator U :

|0⟩ H H

|ψ⟩ U

Here H represents a Hadamard gate, |ψ⟩ denotes a single qubit state which is an eigenstate
of U , U |ψ⟩ = e2πiθ |ψ⟩, and the measurement is done in the computational basis. Express
the phase θ in terms of the probability of the measurement giving zero.
(b) Consider the following circuit which uses a controlled SWAP gate. A SWAP
gate acts as: SWAP |i⟩ |j⟩ 7→ |j⟩ |i⟩ ∀ |i⟩ , |j⟩ ∈ C2 .

|0⟩ H H

|ϕ1 ⟩
SWAP
|ϕ2 ⟩ V

Here V is a unitary operator, |ϕ1 ⟩ , |ϕ2 ⟩ are single qubit states, and the measurement is
done in the computational basis. Find the probability of the measurement giving zero.
(c) Consider the operator U and the state |ψ⟩ introduced in part (a). It is given
that θ = j/2m for some j ∈ {0, 1, 2, . . . , 2m − 1} and some m ∈ N. Define the controlled
unitary operator Θm (U ) which acts on a state |k⟩ |ψ⟩ of m + 1 qubits as:

Θm (U ) |k⟩ |ψ⟩ = |k⟩ U k |ψ⟩ ,

where U k |ψ⟩ is the state obtained by k successive applications of the operator U on the
state |ψ⟩ and k ∈ {0, 1, 2, . . . , 2m − 1}.

(i) Find an expression for the following state of (m + 1) qubits:

|Φ⟩ := Θm (U ) H ⊗m ⊗ I |0⟩⊗m |ψ⟩ .

[You should write out the result of applying the operators to |0⟩⊗m |ψ⟩.]

(ii) Write an expression for the corresponding state |ϕj ⟩ of the first m qubits.
m
Show that {|ϕj ⟩}2j=0−1 is an orthonormal basis.

(iii) Let F be an operator acting on m qubits as F |j⟩ = |ϕj ⟩. Justify why F is

a unitary operator.

(iv) State the 2 sequential operations that you can do on |ϕj ⟩ to find the value
of j. What is the probability pj of finding the value of j (and hence θ)?

[QUESTION CONTINUES ON THE NEXT PAGE]

Part II, 2024 List of Questions

 99
Paper 3, Section II
15E Quantum Information and Computation
Suppose Hn ≃ (C2 )⊗n can be partitioned into 2 mutually orthonormal subspaces Sg
and Sb . Let Pg and Pb denote the orthogonal projections onto Sg and Sb , respectively. We
call Sg the good subspace and Sb the bad subspace. All kets in this question are normalised.
(a) Show that any |φ⟩ ∈ Hn can be uniquely decomposed as follows:

|φ⟩ = a |ψ1 ⟩ + b |ψ2 ⟩ ,

where |ψ1 ⟩ ∈ Sg and |ψ2 ⟩ ∈ Sb . Express a, b, |ψ1 ⟩ and |ψ2 ⟩ in terms of |φ⟩, Pg and Pb .

For any |α⟩ ∈ Hn , consider a unitary operator R|α⟩ which acts as follows:

R|α⟩ |α⟩ = − |α⟩ ,

R|α⟩ |β⟩ = |β⟩ ∀ |β⟩ such that ⟨β|α⟩ = 0 .

In particular define R0 := R|0n ⟩ , where |0n ⟩ := |0⟩⊗n . Further, let A denote a unitary
operator which acts on |0n ⟩ as follows:
√ p
|Ω⟩ := A |0n ⟩ := p |ψg ⟩ + 1 − p |ψb ⟩ ,

where |ψg ⟩ ∈ Sg , |ψb ⟩ ∈ Sb , and p ∈ (0, 1). Let Rg := R|ψg ⟩ .

Suppose you are given quantum circuits which implement R0 , Rg and A. Below, we
√
consider a protocol that increases the amplitude p of the good state |ψg ⟩.
(b) Show that AR0 A−1 |Ω⟩ = R|Ω⟩ |Ω⟩.
(c) Consider the 2-dimensional real Euclidean plane P spanned by |ψg ⟩ and |ψb ⟩,
and let θ denote the angle between |Ω⟩ and the horizontal axis |ψb ⟩. Express θ in terms
of p.
(d) How can you geometrically interpret the actions of Rg and (−R|Ω⟩ ) on any vector
|ϕ⟩ in P?
(e) Justify that Rg and (−R|Ω⟩ ) leave the plane P invariant.
(f) Start with the state |Ω⟩ and apply the operator Q := −R|Ω⟩ Rg to it. What is
the action of a single use of Q on the state |Ω⟩ and k successive uses of Q on the state
|Ω⟩? [Hint: Draw clear diagrams.] What is the final state |Ωk ⟩ obtained after k successive
applications of Q on |Ω⟩?
(g) What is the probability that a measurement of |Ωk ⟩ will yield the good state
√
|ψg ⟩? Show that k = O(1/ p) iterations suffice to make the amplitude of the good state
close to 1.

Part II, 2024 List of Questions [TURN OVER]

 100
Paper 1, Section II
19H Representation Theory
What is a complex representation (ρ, V ) of a group G? What does it mean to say
that a representation (ρ, V ) of G is faithful ? Show that every finite group has a faithful
representation over the complex numbers.
Let G be a finite group, (ρ, V ) be a complex representation of G, and g ∈ G. Writing
S(g) for the set of eigenvalues of ρ(g), show that if g is conjugate to g k in G then

λ ∈ S(g) =⇒ λk ∈ S(g).

Deduce that if p is a prime number and (ρ, V ) is a faithful complex representation of Sp ,

then dim V ⩾ p − 1.
Consider the group G of all invertible functions σ : N → N, under composition. Does
G have a faithful complex representation?

Paper 2, Section II
19H Representation Theory
Let G be a finite group. What is the character χV of a complex representation
(ρ, V ) of G?
Suppose that (ρ, V ) and (σ, W ) are complex representations of G. Show that the
vector space HomC (V, W ) of C-linear maps α : V → W can be made into a representation
of G × G via


((g, h) · α)(v) = σ(h) α(ρ(g −1 )v) for (g, h) ∈ G × G, α ∈ HomC (V, W ) and v ∈ V.

Show that the character χHomC (V,W ) satisfies

χHomC (V,W ) (g, h) = χV (g)χW (h).

Consider the permutation representation CG of G × G arising from the action of

G × G on G via
(g, h) · x = gxh−1 for (g, h) ∈ G × G, x ∈ G
What is χCG ?
Suppose that V1 , . . . , Vr are all the simple representations of G (up to isomorphism).
Show there is an isomorphism
r
M
CG ∼
= HomC (Vi , Vi )
i=1

of representations of G × G.

Part II, 2024 List of Questions

 101
Paper 3, Section II
19H Representation Theory
Let p be a prime number and Fp = Z/pZ be the field with p elements. Consider the
group of unitriangular 3 × 3-matrices with coefficients in Fp ,
  
 1 x z 
G := 0 1 y  : x, y, z ∈ Fp .
 
0 0 1

(a) Describe the conjugacy classes of G.

(b) Show that G has precisely p2 complex representations of degree 1 and describe
them explicitly.
(c) Find an abelian subgroup A of G of order p2 and construct p − 1 irreducible
complex representations of G of degree p by induction from A.
(d) Determine the character table of G.

Paper 4, Section II
19H Representation Theory
What is the topological group S 1 ? Assuming any necessary facts about continuous
homomorphisms with domain (R, +), show that every irreducible complex representation
of S 1 is of the form
z 7→ z n : S 1 → GL1 (C)
for some n ∈ Z.
Let ρV : SU (2) → GL(V ) be a complex representation of the topological group
SU (2) and let χV be its character.
(a) Show that χV is determined by its restriction to a subgroup
P T of SU (2)
isomorphic to S 1 and deduce that P
χV may be written in the form n∈Z an z n for some
non-negative integers an such that n∈Z an < ∞. Show moreover that an = a−n for all
n ∈ Z.
(b) Let Vn be the (n + 1)-dimensional irreducible representation of SU (2). Write
down χVn in the form given in part (a). Decompose V4 ⊗ V4 , S 2 V4 and Λ2 V4 as a direct
sum of irreducible representations up to isomorphism.

Part II, 2024 List of Questions [TURN OVER]

 102
Paper 1, Section II
24H Riemann Surfaces
Given two suitable topological spaces Y and X, define a covering map π : Y → X.
What does it mean to say that X is simply connected? Write down the simply connected
Riemann surfaces (up to analytic isomorphism).
What is a lattice of C? Prove that for any lattice L there exists a non-constant
analytic function f : C → C∞ where f (z + l) = f (z) for all z ∈ C and l ∈ L.
Assuming now that the quotient space C/L is a Riemann surface where the natural
projection q : C → C/L is analytic, deduce the existence of a unique analytic function
f : C/L → C∞ such that f = f q for your function f above.
Show that neither f nor f are covering maps. Does there exist a covering map from
C/L to C∞ ? Justify your answer, stating clearly any results which you use.

Paper 2, Section II
24H Riemann Surfaces
For a non-constant analytic map f : R → S between compact Riemann surfaces
and a point z ∈ R, let mf (z) denote the multiplicity of f at z and deg(f ) the degree of f .
State the valency theorem. For the Riemann surface C∞ and a non-constant analytic
function f : C∞ → C∞ , which you may assume is of the form f (z) = p(z)/q(z) for non-
zero polynomials p, q, explain how to find deg(f ). Which f are the analytic isomorphisms
of C∞ ?
If h : C∞ → C∞ is the Möbius transformation that swaps ∞ with 1 and swaps 0
with −1, write down a formula for h, as well as a quadratic equation satisfied by the fixed
points of h.
Now consider the rotational symmetry group G of a regular octahedron P . You may
assume that G is realised as a group of Möbius transformations isomorphic to S4 with the
six vertices of P corresponding to the points 0, ∞, ±1, ±i ∈ C∞ . Write down the possible
sizes of the orbits under this action of G on C∞ .
Consider the function F : C∞ → C∞ given by the formula

(z 4 + 1)2 (z 4 + 6z 2 + 1)2 (z 4 − 6z 2 + 1)2

F (z) = .
z 4 (z 4 − 1)4

Which points in C∞ are mapped to ∞ by F and with what multiplicities? What is

deg(F )?
You may now assume that F is constant on orbits, namely that if z1 and z2 are in the
same orbit of this action of G on C∞ then F (z1 ) = F (z2 ). By using the valency theorem,
or otherwise, show that F distinguishes orbits, namely if F (z) = F (w) for z, w ∈ C∞ then
z and w are in the same orbit.

Part II, 2024 List of Questions

 103
Paper 3, Section II
23H Riemann Surfaces
State and prove the open mapping theorem for Riemann surfaces R and S. [You
may assume the identity theorem.] Suppose that f : R → S is analytic and non-constant.
Deduce that if R is compact then f is surjective.
Suppose that f : D → D is non-constant and analytic with f (0) = 0. Show that
g(z) = f (z)/z is analytic on D, and that for the open disc Dr of radius r < 1 and centre
0 we have |g(z)| ⩽ 1/r for z ∈ Dr .
Describe all of the analytic isomorphisms h : D → D such that h(0) = 0.
Suppose that F : C → C is an analytic isomorphism with F (0) = 0. Show that
|F (z)| → ∞ as |z| → ∞ and that F extends to an analytic isomorphism C∞ → C∞ .
Given an analytic isomorphism f : D → D with f (0) = 0, does f extend to an
analytic isomorphism C → C? Does every analytic isomorphism D → D extend to an
analytic isomorphism C → C? Justify your answers.

Part II, 2024 List of Questions [TURN OVER]

 104
Paper 1, Section I
5L Statistical Modelling
(a) What are the three main components of a generalised linear model for observa-
tions (Y1 , x1 ), . . . , (Yn , xn )?
(b) Assuming the model holds with the canonical link function, give expressions for
the mean and variance of the Yi as functions of the covariates xi .
(c) Define the Poisson generalised linear model with the canonical link function.

Paper 2, Section I
5L Statistical Modelling
Suppose we observe the proportion Y ∼ n−1 Binomial(n, p) where n ∈ N is known
and p ∈ (0, 1) is unknown. Compute the score function and the Fisher information for
this statistical model.
State the Newton–Raphson and Fisher scoring algorithms for computing the max-
imum likelihood estimator.
How many steps do these algorithms take to converge to the maximum likelihood
estimator when initialised at p(0) = Y ? How many steps does Fisher scoring take when
initialised at some p(0) ̸= Y ?

Paper 3, Section I
5L Statistical Modelling
Explain mathematically why the two results returned in the R output below are as
they are.

> n <- 50
> Y <- rnorm(n)
> p <- 2
> Z1 <- matrix(rnorm(n*p), nrow=n)
> lm1 <- lm(Y ~ Z1)
> sum(lm1$residuals)
[1] 0
>
> p <- 49
> Z2 <- matrix(rnorm(n*p), nrow=n)
> lm2 <- lm(Y ~ Z2)
> summary(lm2)$r.squared
[1] 1

Part II, 2024 List of Questions

 105
Paper 4, Section I
5L Statistical Modelling
Let the concentrations of insecticides Y1 , . . . , Yn relative to covariates x1 , . . . , xn ∈
R follow the model
p

Yi = µ + xTi β + εi ,

where µ ∈ R and β ∈ Rp are unknown and the εi are i.i.d. with distribution N (0, σ 2 ) for
known σ 2 > 0. Suppose that it is not possible to measure Yi directly, but we only have
access to a random variable Zi , satisfying
(
1 if Yi > τ,
Zi =
0 otherwise,

where τ ∈ R is some fixed known tolerance.

(a) Find a binary regression model with a suitable link function and some linear
predictor for the data (Z1 , x1 ), . . . , (Zn , xn ). How can the parameters µ and β be estimated
from these data?
(b) Given a new data point x∗ ∈ Rp , find an asymptotic 1 − α confidence interval
for the expected mean response at x∗ in the binary regression model.

Part II, 2024 List of Questions [TURN OVER]

 106
Paper 1, Section II
13L Statistical Modelling
During spawning season, female horseshoe crabs lay clusters of eggs which are
fertilised externally by a number of nearby male crabs. We are given a dataset with
information on n = 173 female crabs. It contains, among other variables, the weight and
width of each crab, as well as the number y of male crabs in the vicinity. Consider the
following (shortened) R output from an analysis of this dataset.

> head(Crabs[c("y", "weight", "width")])

y weight width
1 8 3.05 28.3
2 0 1.55 22.5

> crabs.lm <- lm(y ~ weight + width, data=Crabs)

> summary(crabs.lm)

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -4.5721 4.2132 -1.085 0.2794
weight 1.6817 0.9015 1.865 0.0639
width 0.1446 0.2218 0.652 0.5153

> anova(lm(y ~ 1, data=Crabs), crabs.lm)

Model 1: y ~ 1
Model 2: y ~ weight + width
Res.Df RSS Df Sum of Sq F Pr(>F)
1 172 1704.9
2 170 1477.7 2 227.2 13.069 5.252e-06 ***

> cor(Crabs$weight, Crabs$width)

[1] 0.8769373
> plot(crabs.lm, add.smooth=FALSE, which=c(1,2))

[QUESTION CONTINUES ON THE NEXT PAGE]

Part II, 2024 List of Questions
 107
(a) Write down the statistical model fitted by crabs.lm.
(b) State mathematical formulas for what the columns displayed in the output of
crabs.lm describe.
(c) State the null and alternative hypotheses for the hypothesis test performed using
the anova command. What can we conclude from the output?
Explain this in relation to the final column of the summary command output by
referring to the output of the cor command.
(d) What would you expect to see in the two diagnostic plots if the model crabs.lm
were to fit the data well? Which model assumptions appear to be violated according to
these plots?
(e) Suggest two modifications that may help to improve the quality of the fit.

Part II, 2024 List of Questions [TURN OVER]

 108
Paper 4, Section II
13L Statistical Modelling
A researcher has collected a dataset on 53 patients to study the occurrence of a
certain disease. The dataset contains for each patient, a variable (dis) indicating if
they are suffering from the disease (dis = 1) or not (dis = 0), a continuous variable
(pollution) describing the level of air pollution near their home, as well as a variable
(risk) recording which of two risk categories the patient falls under. Below is given the
(shortened) R output of the analysis performed by the researcher.

> head(disease)
dis pollution risk
1 0 58.44204 2
2 0 45.29248 1

> binom.lm1 <- glm(dis ~ pollution + risk, family=binomial, data=disease)

> summary(binom.lm1)

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -11.83686 3.62821 -3.262 0.00110 **
pollution 0.18243 0.05679 3.212 0.00132 **
risk2 1.53893 0.74665 -2.061 0.03929 *

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 73.5 on ? degrees of freedom

Residual deviance: 55.3 on ? degrees of freedom
AIC: 61.3

(a) Write down the generalised linear model being fitted. Why is there no coefficient
for risk1?
(b) Give an interpretation for the coefficient of pollution.
(c) What are the missing degrees of freedom in the output?
(d) How should the R code above be changed to fit the model that corresponds to
the ‘Null deviance’ in the output? What is the AIC value of that model?
(e) State the null and alternative hypotheses for the test performed in the code
below and describe the form of the test.
What can the researcher conclude from the following ouput?

> pchisq(73.5-55.3, df = 2, lower.tail = FALSE)

[1] 0.0001142191
> binom.lm2 <- glm(dis ~ pollution * risk, family=binomial, data=disease)
> pchisq(binom.lm1$deviance-binom.lm2$deviance, 1, lower.tail=FALSE)
[1] 0.8705801

Part II, 2024 List of Questions

 109

Part II, 2024 List of Questions [TURN OVER]

 110
Paper 1, Section II
36B Statistical Physics
(a) What systems are described by a canonical ensemble? If the energy of the ith
microstate is Ei , with i = 0, 1, 2, . . ., write down an expression for the partition function
Z in terms of temperature T and the Boltzmann constant kB .
(b) Calculate the partition function Z1 for a single classical ultra-relativistic spinless
particle moving in three-dimensional space in a potential U (x). [Ultra-relativistic means
that the energy-momentum relation is E = pc, where c is the speed of light.]
(c) A system of a large number N of identical, non-interacting particles of the type
described in part (b) is in equilibrium at temperature T in a potential

(x2 + y 2 + z 2 )n
U (x) = ,
V 2n/3
where n is a positive integer and V > 0 is an external parameter analogous to volume.

(i) Calculate the partition function and hence show that the Helmholtz free
energy is
h i
F = −N kB T ln V + A ln(kB T ) + ln In + B ,

where Z ∞
2n
In = u2 e−u du ,
0
and you should determine A and B.
∂F


(ii) Considering the conjugate pressure to V , p = − ∂V T,N , derive the
equation of state.

(iii) Compute the average energy E, the variance of energy (∆E)2 and the heat
capacity CV for the system. Comment on the behaviour of (∆E)/E in the
thermodynamic limit.

(iv) Obtain the local particle number density as a function of x and hence
determine the most likely |x| to find a particle.

Part II, 2024 List of Questions

 111
Paper 2, Section II
37B Statistical Physics
(a) Explain what is meant by an intensive quantity and what is meant by an
extensive quantity. Give two examples of each.
(b) A real-valued homogeneous function f of degree k satisfies

f (λx1 , λx2 , . . . , λxn ) = λk f (x1 , x2 , . . . , xn ) ,

for any real λ. Show that

n
X ∂f
xi = kf .
∂xi
i=1

(c) Explain why the energy E(S, V, N ) is a homogeneous function of degree 1, where
S is the entropy, V is the volume and N is the number of particles. Hence, using the
first law of thermodynamics, find an expression for E in terms of S, V , N , µ, p and T ,
where µ is the chemical potential, p is the pressure and T is the temperature. Show that
dµ = (V dp − SdT )/N .
(d) Consider a chemical reaction at constant T and p where each molecule of
chemical A can change into two molecules of chemical B and one molecule of chemical C,
and vice-versa, i.e. A ↔ 2B + C. By minimising the Gibbs free energy G, derive a relation
between the chemical potentials of the three chemicals at equilibrium, where the chemical
∂G
potential of chemical i is µi = ∂N i
.

Part II, 2024 List of Questions [TURN OVER]

 112
Paper 3, Section II
35B Statistical Physics
(a) What systems are described by a grand canonical ensemble? Use the grand
partition function to derive the formula for the Fermi-Dirac distribution for the mean
occupation numbers nr of discrete single-particle states r with energies Er ⩾ 0 in a gas of
non-interacting identical Fermions in terms of β = 1/(kB T ) and the chemical potential µ.
(b) Show that the density of states g(E) for a gas of non-interacting non-relativistic
identical Fermions with spin degeneracy gs in a large d-dimensional volume V , where
d ⩾ 2, is
g(E) = gs BV E (d−2)/2 ,
where B is a constant that you should determine. Compute the Fermi energy EF , i.e. the
chemical potential at zero temperature and with N particles. Find an expression at zero
temperature for pV in terms of N and EF , where p is the pressure. [You may denote the
surface area of a unit (d−1)-dimensional sphere by Sd−1 .]
(c) Consider a semi-infinite block of metal in 3-dimensional space occupying z ⩽ 0
with infinite extent in the x and y directions, and empty space for z > 0. An electron
in the metal with momentum (px , py , pz ) will escape from the metal if pz is large enough,
p2z
2m ⩾ EF + V0 , where m is the mass of the electron and V0 is the potential barrier.
The flux density, Jz , of electrons escaping from the metal is the number of electrons
escaping per unit area per unit time. Model the electrons as an ideal non-relativistic gas
of Fermions, and assume that kB T ≪ V0 and kB T ≪ EF . You may assume that µ = EF
in this limit. Thereby derive an expression for the flux density of electrons escaping from
the metal.
Z ∞ p
2
[Hint: e−a x dx = π/a for a > 0.]
−∞

Part II, 2024 List of Questions

 113
Paper 4, Section II
35B Statistical Physics
(a) What is meant by a first-order phase transition and what is meant by a second-
order phase transition?
(b) Derive the Clausius-Clapeyron equation for dp/dT , with p the pressure, along the
first-order phase transition curve between a liquid and a gas in terms of the temperature
T , the latent heat L, and the volume per unit particle in the two phases vliquid and vgas .
[You may assume that the chemical potentials of the two phases are equal on the phase
transition curve.]
(c) Assuming that L is constant, vgas ≫ vliquid , and that the gas obeys the ideal
gas law, derive an equation for the phase transition curve p(T ). Determine the change in
volume of the gas with T along the phase transition curve (i.e. dVgas /dT along the curve).
(d) Consider the Dieterici equation of state
 
kB T a
p= exp − .
v−b kB T v

Here v is the volume per unit particle.

(i) Provide a brief physical interpretation for the constants a and b.

(ii) Calculate the second virial coefficient.

(iii) Find the critical temperature Tc for the Dieterici equation of state in terms
of a, b and kB .

Part II, 2024 List of Questions [TURN OVER]

 114
Paper 1, Section II
30L Stochastic Financial Models
(a) In the context of a one-period market model, define the terms arbitrage and
risk-neutral measure. State the one-period fundamental theorem of asset pricing.
(b) Given an n × d matrix P , let

A = {φ ∈ Rd : (P φ)i ⩾ 0 for all i and (P φ)i > 0 for some i}

and ( )
n
X
Q= q ∈ Rn : P ⊤ q = 0, qi = 1, qi > 0 for all i .
i=1

Prove that Q = ∅ if and only if A =

̸ ∅.

Consider a one-period market model with d risky assets, where Sti is the price of asset
i at time t ∈ {0, 1} and r is the interest rate. Assume that there exists at least one risk-
neutral measure for the model, and that the random variables {S1i − (1 + r)S0i : 1 ⩽ i ⩽ d}
are linearly independent.
(c) Let Y be a random variable such that EQ (Y ) > 0 for all risk-neutral measures
Q. Show that there exists a vector θ ∈ Rd such that

Y ⩾ θ⊤ [S1 − (1 + r)S0 ] almost surely,

and we have strict inequality with positive probability.

(d) Let Z be a random variable such that EQ (Z) ⩾ 0 for all risk-neutral measures
Q. Show that there exists a vector ϕ ∈ Rd such that

Z ⩾ ϕ⊤ [S1 − (1 + r)S0 ] almost surely.

Part II, 2024 List of Questions

 115
Paper 2, Section II
30L Stochastic Financial Models
Let (Fn )n be a filtration such that F0 is trivial. Let (Mn )n be a martingale and let
T be a stopping time with respect to the filtration.
(a) Show that the stopped process (Mn∧T )n is a martingale. [Results on the
martingale transform may not be assumed without proof.]
(b) Assuming (Mn∧T )n is bounded and T is finite, show that E(MT ) = M0 . [Versions
of the optional stopping theorem may not be assumed without proof.]
For the rest of the problem, let X0 = 0 and Xn = ξ1 + · · · + ξn for n ⩾ 1, where
(ξn )n are IID and generate the filtration. Suppose P(ξn = +1) = P(ξn = −1) = 1/2 for all
n.
(c) Given a real number w > 1, show that there exists a real number z > 0 such
that the process

Mn = AwXn + Bw−Xn z n
is a martingale for all constants A and B.
(d) Fix positive integers a, b and define

T = min{n ⩾ 0 : Xn ∈ {−a, b}}.

For any 0 < z < 1, compute E(z T ). [You may use the fact that T is a finite stopping time
without proof.]

Part II, 2024 List of Questions [TURN OVER]

 116
Paper 3, Section II
29L Stochastic Financial Models
Consider a discrete-time market model with interest rate r > −1 and one stock with
time-n price Sn . Suppose S0 > 0 is given and that Sn = Sn−1 ξn for all n ⩾ 1, where the
sequence (ξn )n⩾1 is IID, and

P(ξ1 = 1 + b) = p = 1 − P(ξ1 = 1 + a)

for constants −1 < a < b and 0 < p < 1. Assume that there exists a risk-neutral measure
Q for the model.
(a) Show that a < r < b, and that the sequence (ξn )n⩾1 is IID under Q. Find
Q(ξ1 = 1 + b).

Fix a maturity date N > 0 and a payout function g. For all s > 0 define

V (N, s) = g(s),
q 1−q
V (n − 1, s) = V (n, s(1 + b)) + V (n, s(1 + a)) for all 1 ⩽ n ⩽ N,
1+r 1+r
where q = Q(ξ1 = 1 + b).
(b) Consider the case where g(s) = log s. Find, with justification, a formula for
V (n, s) of the form V (n, s) = An log s + Bn , for families of constants (An )0⩽n⩽N and
(Bn )0⩽n⩽N which should be specified.
(c) Prove that an investor with initial capital X0 = V (0, S0 ) can replicate the payout
of the vanilla European claim with time-N payout g(SN ). How many shares of the stock
should the investor hold during the time interval (n − 1, n] ?

Now assume a ⩽ 0, and fix a barrier B > S0 . For all s > 0 define

U (N, s) = g(s),
q 1−q
U (n − 1, s) = U (n, s(1 + b))1{s(1+b)<B} + U (n, s(1 + a)) for all 1 ⩽ n ⩽ N.
1+r 1+r

(d) In terms of the function U , derive an explicit formula for the number of shares of
the stock an investor should hold during the time interval (n−1, n] in order to replicate the
payout of the up-and-out European claim with time-N payout g(SN )1{max0⩽n⩽N Sn <B} .

Part II, 2024 List of Questions

 117
Paper 4, Section II
29L Stochastic Financial Models
Let (Wt )t⩾0 be a Brownian motion.
(a) Using the definition of Brownian motion, show that (Wt )t⩾0 is a Gaussian process
with E(Wt ) = 0 and E(Ws Wt ) = s for all 0 ⩽ s ⩽ t.
(b) Let
Z t
It = Ws ds for all t ⩾ 0.
0

Show that E(It ) = 0 and E(Is It ) = 16 s2 (3t − s) for all 0 ⩽ s ⩽ t.

Consider a continuous time market with interest rate r and a stock whose time-t
price is
1 2
St = S0 e(r− 2 σ )t+σWt ,
for a given volatility parameter σ > 0.
(c) Show that the discounted stock price (e−rt St )t⩾0 is a martingale with respect to
the filtration generated by the Brownian motion.
(d) Show that the time-0 Black–Scholes price of a European call with strike K and
maturity date T is S0 F (σ 2 T, Ke−rT /S0 ), where
1 √
F (v, m) = E[(e− 2 v+ vZ
− m)+ ] for all v, m ⩾ 0

and Z is a standard normal random variable.

(e) Show that the time-0 Black–Scholes price of a European claim with time-T
 +
R
1 T


payout exp T 0 log St dt − K is

1 2 )T 1 2 )T
S0 e−( 2 r+ασ F ( 13 σ 2 T, Ke−( 2 r−ασ /S0 ),

for a constant α to be determined. [You may assume that the process (It )t⩾0 defined in
part (b) is Gaussian.]

Part II, 2024 List of Questions [TURN OVER]

 118
Paper 1, Section I
2G Topics in Analysis
(a) Define the nth Chebychev polynomial Tn . Show that it is indeed a polynomial
and that −1 ⩽ Tn (x) ⩽ 1 for all x ∈ [−1, 1].
(b) Show that, if n ⩾ 1, the leading coefficient of Tn is 2n−1 .
(c) Show that Tn (x) = Tn (−x) if n is even, and Tn (x) = −Tn (−x) if n is odd,
explaining why these results hold for all x ∈ R.
(k)
(d) By looking at the roots of Tn (x), or otherwise, show that, if 0 ⩽ r ⩽ n − 1,
(r)
then Tn (x) is increasing for x ⩾ 1.
(e) Compute Tn′ (1) and show that Tn (x) ⩾ n(x − 1) + 1 for all x ⩾ 1.

Paper 2, Section I
2G Topics in Analysis
State Baire’s category theorem. Define an isolated point for a metric space.
Which of the following statements are true and which are false? Give a proof or a
counterexample. (By a metric space we mean a non-empty metric space.)

(i) A countable metric space must have isolated points.

(ii) A complete metric space cannot have isolated points.

(iii) An uncountable metric space without isolated points must be complete.

(iv) A complete metric space must be uncountable.

(v) A countable complete metric space must have isolated points.

(vi) All the points in a countable complete metric space are isolated.

(vii) A countable complete metric space containing at least two points must
contain at least two isolated points.

(viii) A countable complete metric space containing infinitely many points must
contain infinitely many isolated points.

Part II, 2024 List of Questions

 119
Paper 3, Section I
2G Topics in Analysis
In this question, a contour Γ will be given by a piecewise smooth map γ : [0, 1] → C
that is injective on [0, 1) and satisfies γ(0) = γ(1). You may assume the Cauchy integral
formula for a square contour.
If Γ is a contour and f : Γ → C is continuous, show that
I
1 f (w)
F (z) = dw
2πi Γ w − z

defines a continuous function on C \ Γ. Give an example to show that, even if f is analytic,

F cannot always be extended to a continuous function on C.
Suppose Ω is an open subset of C and K is a compact non-empty subset of Ω. If
f : Ω → C is analytic, show that we can find a finite collection of contours Γ1 , Γ2 , . . . , Γn
lying in Ω \ K such that
n
X I
1 f (w)
f (z) = dw.
2πi Γj w − z
j=1

Paper 4, Section I
2G Topics in Analysis
(a) Show that the collection of algebraic numbers is countable.
P
(b) Suppose that aj > 0 and ∞ j=1 aj converges. Show that we can find θj ∈ {0, 1}
so that
X∞
θj aj
j=1

is transcendental.
(c) Suppose θj ∈ {0, 1} and θj = 1 for infinitely many values of j. Show that
∞
X θj
j!
j=1

is irrational.

Part II, 2024 List of Questions [TURN OVER]

 120
Paper 2, Section II
11G Topics in Analysis
Suppose that aj , bj are real and strictly positive for all j ⩾ 0, and we have p0 = a0 ,
p−1 = 1, q0 = 1, q−1 = 0 with

pj = aj pj−1 + bj−1 pj−2

qj = aj qj−1 + bj−1 qj−2

for j ⩾ 1. Show that

pn b0
= a0 +
qn b1
a1 +
b2
a2 +
b3
a3 +
..
.
bn−1
an−1 +
an
and that       
pn bn pn−1 a0 b0 a1 b1 an bn
= ... .
qn bn qn−1 1 0 1 0 1 0

We now specialise to the case when bj = 1 and the aj are strictly positive integers
for all j. Show that pn qn−1 − qn pn−1 = (−1)n+1 for all n ⩾ 1.
Show that pn /qn tends to a limit x and

pn pn+1 1
−x + −x =
qn qn+1 qn qn+1

for each n ⩾ 0.
Now specialise still further to the case when aj = 1 for all j. Show that pn = Fn+2 ,
qn = Fn+1 for all n ⩾ 1 where F0 = 0, F1 = 1 and Fn+2 = Fn+1 + Fn . Solve√ this difference
equation to obtain an expression for Fn in terms of powers of ϕ = (1 + 5)/2. Hence show
that, in this case, the limit x discussed in the previous paragraph is ϕ. Show also that

Fn+1 ϕ Fn+2 1
Fn Fn+1 −ϕ → √ and Fn Fn+1 −ϕ → √
Fn 5 Fn+1 ϕ 5
as n → ∞.

Part II, 2024 List of Questions

 121
Paper 4, Section II
12G Topics in Analysis
(a) Suppose we have n points A1 , A2 , . . . , An , in order, along a straight line with
each point coloured red or green. If A1 is coloured red and An green, show that there
must be an odd number of segments Aj Aj+1 with Aj and Aj+1 of different colours.
(b) Consider an equilateral triangle DEF divided up into a grid of small equilateral
triangles by 3 sets of n equidistant lines parallel to each side. We refer to the lines joining
neighbouring points of the grid as segments.

(i) Suppose that each vertex of the grid is coloured red, green or blue, that every
vertex on DE is coloured red or green, every vertex on EF is coloured green
or blue and every vertex on F D is coloured blue or red. Show that there is
an odd number of triangles in the grid with vertices of different colours.

(ii) Suppose instead that each vertex of the grid is coloured red, green or blue,
and that the three vertices D, E and F have different colours. Must there
be a triangle in the grid with vertices of different colours? Give reasons.

(c) Let us write

△ = {λ1 d + λ2 e + λ3 f : λ1 + λ2 + λ3 = 1, λu ⩾ 0},

where d, e, f are the position vectors of D, E and F , and write I for the closed line
segment DE, J for the closed line segment EF , K for the closed line segment F D.
Which of the following statements are true and which are false? Give a proof or a
counterexample.

(i) There does not exist a continuous function f : △ → ∂△ with

f (x) ∈ I for all x ∈ I, f (x) ∈ J ∪ K for all x ∈ J ∪ K.

(ii) There does not exist a continuous function g : △ → ∂△ with

g(x) ∈ I for all x ∈ I, g(x) ∈ J for all x ∈ J, g(x) ∈ K for all x ∈ K.

Part II, 2024 List of Questions [TURN OVER]

 122
Paper 1, Section II
40D Waves
(a) Starting from the linearized mass and momentum conservation equations gov-
erning sufficiently small and smooth perturbations of a compressible homentropic inviscid
fluid at rest with constant reference density ρ0 , pressure p0 and sound speed c0 , show that
the pressure perturbation p̃ satisfies a wave equation. How is p̃ related to the velocity
potential ϕ?
(b) Consider a semi-infinite straight duct of uniform cross-section, aligned along the
x−axis for −L ⩽ x < ∞. There is a piston at the end of the duct which performs
oscillations ϵeiωt about its equilibrium position at x = −L. The duct is filled with
compressible fluid of density ρ− and sound speed c− in the region −L < x < 0 and
with compressible fluid of density ρ+ and sound speed c+ in the region 0 < x < ∞.
The piston’s oscillations are sufficiently small so that you may assume 0 < ϵ ≪ L and
|ϵω| ≪ min(c− , c+ ).

(i) Show that the complex amplitude of the velocity potential in x > 0 is given
by

i ρρ−
+
sin λ − cc−
+
cos λ ωL
ϵc−  2  2 , where λ= .
ρ+ 2 c− 2λ c−
ρ− sin λ + c+ cos

(ii) Consider the two sets of frequencies of oscillation such that λ = nπ and
λ = (n + 12 )π for integer n. Calculate the time-averaged acoustic energy
flux in x > 0 for each set, and briefly comment on the behaviour in the case
where ρ+ ≪ ρ− and c+ ≈ c− .

Part II, 2024 List of Questions

 123
Paper 2, Section II
40D Waves
The linearized Cauchy momentum equation governing small and smooth displace-
ments u(x, t) in a uniform, linear isotropic elastic solid of density ρ is

∂2u
ρ = (λ + µ)∇(∇ · u) + µ∇2 u,
∂t2
where the constants λ and µ are the Lamé moduli.
(a) Show that this equation supports two distinct classes of wave-like motion: P-
waves for the dilatation ϑ = ∇ · u with phase speed cp ; and S-waves for the rotation
Ω = ∇ × u with phase speed cs . You should express cp and cs explicitly in terms of the
Lamé moduli.
(b) Now consider a region of this solid with a horizontal plane boundary at z = 0
in which plane waves propagate with wave vector k = κ(sin θ, 0, cos θ), i.e. θ is the angle
the wave vector makes with the vertical z−direction and κ is the magnitude of the wave
vector. Explain briefly why such a domain can in general support:

(i) harmonic P-waves: u = A exp[iκ(x sin θ + z cos θ) − iωt];

(ii) harmonic SV-waves: u = BV exp[iκ(x sin θ + z cos θ) − iωt];

(iii) and harmonic SH-waves: u = BH exp[iκ(x sin θ + z cos θ) − iωt].

You should define explicitly the orientations of the complex vector amplitudes A, BV and
BH .
(c) Now consider a region of the solid between a rigid plane boundary at z = 0 and
a free surface at z = h > 0.

(i) Show that this region can support propagating SH-waves with wave vector
in the x−direction (i.e. with θ = π/2), calculating explicitly the dispersion
relation.

(ii) Deduce that there is a cut-off frequency ωn for each mode in the vertical,
given by
(2n + 1)π
ωn = cs ,
2h
for non-negative integers n = 0, 1, 2 . . . .

(iii) Express the phase velocity c and the group velocity cg of each mode in terms
of cs , ωn and κ.

(iv) Deduce that, for any given wave number κ > 0 and mode with n ⩾ 0,
c = mcg with m > 1, where you should express m in terms of n, κ and h.

(v) Calculate m explicitly for the specific wave with horizontal wavelength h
and n = 1.

[ Hint: You may find it useful to recall that ∇2 q = ∇(∇ · q) − ∇ × (∇ × q). ]

Part II, 2024 List of Questions [TURN OVER]

 124
Paper 3, Section II
39D Waves
Let u = (u, 0, w) denote the components of a solenoidal two-dimensional velocity
field with ∇ · u = 0. Consider small-amplitude vertical-velocity perturbations w(x, z, t)
about a state of rest in a two-dimensional stratified fluid of sufficiently slowly varying
background density ρ0 (z) such that

1 dρ0 ∂w ∂2w
≪ .
ρ0 dz ∂z ∂z 2

(a) Show that w(x, z, t) satisfies the equation

 2 
2 ∂ w 2 ∂2w g dρ0
∇ + N (z) = 0 where N 2 (z) = − ,
∂t2 ∂x2 ρ0 dz

and N is the buyoancy frequency.

(b) Consider a semi-infinite region of fluid with constant buoyancy frequency N ,
flowing with uniform velocity U > 0 in the x−direction and bounded below by a sinusoidal
range of hills at h = h0 sin kx, where k > 0.

(i) Determine the frequency of the motion induced in a frame of reference in

which the background fluid is stationary.

(ii) You may assume that the vertical component of the fluid velocity at z = 0 is
given by the vertical component of the velocity of fluid particles that follow
the undulations of the lower boundary, i.e. w = U ∂h/∂x at z = 0. Hence
establish that for z > 0, the vertical velocity satisfies

w = w0 exp[i(kx + mz − ωt)],

where w0 = U kh0 , and m satisfies a dispersion relation which you should

express in terms of N , U and k.

(iii) The time- and horizontally-averaged vertical wave energy flux at z = 0 is

given by ⟨ Iz ⟩0 = ⟨ p̃w ⟩0 , where p̃ is the pressure perturbation, the angle
brackets denote an average over the wave period and the overline denotes
an average over the hill wavelength. Calculate ⟨ Iz ⟩0 for sufficiently large
wavelengths k ⩽ kc = N/U making clear the reasoning for selecting the sign
of all terms.

(iv) Show that ⟨ Iz ⟩0 = 0 for sufficiently short wavelengths k > kc .

(v) Give a brief physical interpretation of the cut-off wavenumber kc = N/U .

Part II, 2024 List of Questions

 125
Paper 4, Section II
39D Waves
Consider one-dimensional homentropic flow in an ideal gas with ratio of specific
heats γ and equilibrium sound speed c0 .
(a) Starting from the conservation equations for mass and momentum, show that
the Riemann invariants R± , defined as

(c − c0 )
R± = u ± 2 ,
(γ − 1)

are constant on characteristics C± given by

dx
= u ± c,
dt
p
where u is the fluid velocity, and c = dp/dρ is the associated sound speed.
(b) Consider a semi-infinite tube filled with such an ideal gas in the region x > X(t)
to the right of a piston at x = X(t). At time t = 0, the piston and the gas are at rest,
X = 0, and the gas is uniform with c = c0 . For t > 0, the piston accelerates smoothly in
the positive x−direction.

(i) Show that, prior to the formation of a shock, the motion of the gas can be
written in terms of a parameter τ by
 
1
u(x, t) = Ẋ(τ ) on x = X(τ ) + c0 + (γ + 1)Ẋ(τ ) (t − τ ),
2

in a region which you should specify carefully.

(ii) Suppose X(t) is given by

c0 t3
X(t) = ,
T2
where T is a positive constant. Show that a shock first forms in the gas
when t/T = f (γ), for some function f (γ) which you should determine.

Part II, 2024 List of Questions [TURN OVER]

 126
END OF PAPER

Part II, 2024 List of Questions