Mathematical Biology: Example Sheet 2

David Tong, January 2025

1. Let x, y be the normalised populations of phytoplankton and zooplankton respectively.

The system is modelled by the following differential equations, where the constants !, b
and c are positive,
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

dx x2
= b x(1 → x) → y 2 ,
dt ! + x2
dy x2
= cy 2 → y.
dt ! + x2
Briefly explain the meaning of each term in these equations.
Assume that c > 2 and ! " 1. By finding the nullclines and carefully considering how
they intersect, show that there is one fixed point where both x > 0 and y > 0 and that it
is stable.
The system is perturbed by the intrinsic growth rate b increasing as a result of a rise
in the temperature of the sea. Consider how this changes the nullclines. If the system
was at the stable fixed point before, what happens after the temperature rise? Deduce the
possibility of excitable behaviour in which there can be a spike in the plankton population
sizes.
2. Circulating red blood cells (RBCs) are destroyed in the spleen and created in the bone
marrow in proportion to the number destroyed on the previous day.

a) Treat days as discrete time units, and let Rn be the number of RBCs in circulation
on day n, let Mn be the number produced by marrow on day n, let f be the fraction
of RBCs removed by the spleen every day, and let γ be the number produced on day
n for each cell lost on day n → 1.

Write down equations for Rn+1 and Mn+1 in terms of Rn and Mn , and show that

Rn = Aλn+ + Bλn−
where !
1→f ± (1 → f )2 + 4γf
λ± = .
2
Deduce that the RBC count may fluctuate, but that it will remain constant at large
times if and only if γ = 1.

1
 b) Start again, but this time formulate the model for RBC count as a delay differential
equation for R(t), where the production of new cells is proportional to R(t→T ). Now
f should be a rate and γ is still the ratio of cells made for each one lost. Show that
there exists a solution of the form R(t) = Ceλf t if
1
(λ + 1) = e−λf T .
γ
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

Show graphically that this equation has a positive real root if and only if γ > 1, and
interpret this result.

3. Consider the difference equation with delay (note the n → 1),

xn+1 = rxn (1 → xn−1 ).

This is the discrete equivalent of the delay differential equation for population.
Show that the fixed point x∗ = 1 → r−1 is unstable if r is sufficiently large. Show
that when r takes its marginal value for instability, the linearised stability problem has a
periodic solution (to be determined).

4. A discrete-time model for breathing is given by

Vn+1 = α Cn−k ,
Cn+1 → Cn = M → β Vn+1 ,

where Vn is the volume of each breath at time step n and Cn is the concentration of carbon
dioxide in the blood at the end of time step n. (This model was presented in lectures and
we found and analysed the stability of the steady state when k = 0 and k = 1.)
For general (integer) k > 1, by seeking parameter values when the modulus of a per-
turbation to the steady state is constant, show that the range of parameters where the
solution is stable is " #
π
0 < αβ < 2 sin .
4k + 2
Notice how this range shrinks as the time lag k increases. What is the periodicity of the
constant-modulus solution at the upper end of this range?

2
 5. The concentration of bacteria C(x, t) in a thin channel of length L obeys the diffusion
equation in one dimension with constant diffusivity D. Initially, the concentration of
bacteria is C(x, 0) = C0 + (C1 → C0 ) x/L (this was achieved by keeping the concentration
at the ends x = 0 and x = L at C = C0 and C = C1 respectively and waiting for the
system settle to its steady state).
At t = 0 the ends of the channel are both suddenly sealed. There is no flux (i.e.
∂C/∂x = 0) through x = 0 or x = L for t > 0. Use separation of variables to find the
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

bacterial concentration in 0 ≤ x ≤ L for t > 0, and hence show that

$ " 2 2 #(
1 % 4 & nπ ' nπ
C(x, t) = C0 + (C1 → C0 ) → cos x exp → Dt .
2 n≥1, odd n2 π 2 L L2

Sketch the bacterial distribution for a range of times, paying particular attention to the
shape at very early and late times. Explain what is happening.

6a) In an axisymmetric cylindrical geometry, the diffusion equation is

" #
∂C D ∂ ∂C
= r
∂t r ∂r ∂r

where D > 0 is a constant. Find and sketch the similarity solution of the diffusion equation
which satisfies C → 0 as r → ∞ and
) ∞
2πr C(r, t) dr = M
0

where M > 0 is a constant, by assuming that the solution is of the form

M r
C(r, t) = ηF (ξ) with η = and ξ = .
Dt (Dt)1/2

Hence show that C is Gaussian.

b) Find the analogous similarity solution in a spherically symmetric geometry, where
C(r, t) satisfies
" # ) ∞
∂C D ∂ 2 ∂C
= 2 (r with 4πr2 C(r, t) dr = M .
∂t r ∂r ∂r 0

and C → 0 as r → ∞. [Hint: Start by finding appropriate η (independent of r) and ξ

(proportional to r)]

3
 Mathematical Biology: Example Sheet 3

David Tong, January 2025

1. The concentration of a chemical C(x, t) satisfies the nonlinear diffusion equation

! " # ∞
∂C ∂ ∂C
= D(C) and C(x, t) dx = N
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

∂t ∂x ∂x −∞

with D(C) = k C p , for positive constants N , k and p. Use dimensional analysis to find
a suitable space-like ξ and space-independent η for the similarity solution of the form
C(x, t) = η F (ξ). Use this to seek the solution initially localised to the origin, and show
that F is of the form
! "1/p
p 2
F (ξ) = A → ξ for |ξ| < ξ0
2(2 + p)

and F (ξ) = 0 for |ξ| ≥ ξ0 , for some A and ξ0 . For the case when p = 2, find A and ξ0 .

2. A simple model of the spreading of an animal population n(x, t) in a spatial domain is

given by the nonlinear reaction-diffusion equation
! "
∂n ∂ ∂n
=D n + αn
∂t ∂x ∂x

with n(x, 0) = N δ(x) and n → 0 as |x| → ∞. Here D and N are positive constants and
α is a constant which may be positive or negative.
By setting n(x, t) = R(x, τ ) eαt , where τ (t) is some time-like variable satisfying τ (0) = 0,
show that a suitable choice of τ yields the non-linear diffusion equation
! "
∂R ∂ ∂R
= R
∂τ ∂x ∂x

with R(x, 0) = N δ(x).

By making a suitable similarity ansatz, show that the population is confined to a region
|x| < x0 , where ! αt "
3 9 e →1
x0 = N D .
2 α
Describe the evolution of the population in the cases α = 0, α > 0 and α < 0.

1
 3. A bistable system with diffusion is given by
∂p ∂ 2p
= → p(p → r)(p → 1),
∂t ∂x2
where 0 < r < 1. Seek a travelling wave solution by setting ξ = x → ct and p(x, t) = f (ξ),
and find the differential equation satisfied by f .

a) Rewrite your differential equation as two first order equations. Suppose that c takes
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

the exact value that allows a travelling wave solution (you will need to consider
r < 1/2 and r > 1/2 separately). Sketch the phase plane for the system, marking
the trajectory that corresponds to the travelling wave.

b) Impose the (slightly odd requirement) that the solution to the original second-order
differential equation also satisfies f # = af (f →1). What values of a and c yield a valid
solution? By solving this first-order equation for f , give the corresponding solution
for p(x, t).

4. The SIR epidemic model can be extended to be a spatial model for the spread of an
infectious disease:
∂S ∂ 2S
= →βIS + D 2
∂t ∂x
∂I ∂ 2I
= +βIS → νI + D 2 .
∂t ∂x
Suppose that an epidemic wave arrives in a previously uninfected region (so S ≈ N and
I ≈ 0). Consider the dynamics near this wave front by taking

S = N → u(ξ) and I = v(ξ)

with ξ = x → ct, and linearise in u and v. You may assume that the system will settle to
the slowest possible wave speed. Find the wave speed of the epidemic, and show that it is
√
proportional to R0 → 1.

5. Investigate the possibility of Turing instability for the reaction-diffusion system.

∂u u2
= ∇2 u + → bu
∂t υ
∂υ
= d ∇2 υ + u2 → υ.
∂t
In particular, find the region of the parameter space (b, d) in which Turing instability can
occur, and give the value for the critical wavenumber at the onset of instability in terms
of d.

2
 6. A space-dependent phytoplankton and zooplankton model can be reduced to the fol-
lowing equations,
∂u
= ∇2 u + u + u2 → γuυ,
∂t
∂υ
= d∇2 υ + βuυ → υ 2 .
∂t
Find the regions in the β → γ plane (a) in which there is a stable, homogeneous state
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

(u0 , υ0 ) in which neither u0 nor υ0 is zero and (b) in which that state may be unstable to
a Turing instability. In case (b), for what values of d will the instability occur? Find the
critical wavenumber at the onset of the instability in terms of β and d.

7. Consider the chemotactic system

∂ 2n
! " ! "
∂n ∂ ∂a n
=D 2 → nχ(a) + bn 1 → ,
∂t ∂x ∂x ∂x n0
∂a ∂ 2a
= DA 2 + hn → da,
∂t ∂x
where
χ 0 a0
χ(a) = .
(a0 + a)2
Find a rescaling such that this reduces to

∂ 2u
$ %
∂u ∂ u ∂υ
= 2 →β + u(1 → u),
∂τ ∂ξ ∂ξ (α + υ)2 ∂ξ
∂υ ∂ 2υ
= δ 2 + γ(u → υ).
∂τ ∂ξ

[Hint: do the rescaling over a few steps, keeping an eye on the intended final form.]
Show that the uniform, steady solution u = υ = 1 is unstable to a spatial perturbation
if
βγ √ √
> ( γ + δ)2 .
(1 + α) 2

Find the critical wavenumber in the case when α = γ = δ = 1.

3
 Mathematical Biology: Example Sheet 4

David Tong, January 2025

1. Consider a birth and death process in which births can give rise to either one or
two offspring, with probability λ1 and λ2 respectively, while the probability of death per
individual is β.
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

Write down the master equation for the probability pn (t).

∞
!
Show that the generating function φ(s, t) = sn pn satisfies the equation
n=0
" #
∂φ ∂φ
= (s → 1) (λ1 + λ2 (s + 1)) φ → β .
∂t ∂s
Use this equation in the steady state to show that
1 1
"n# = (λ1 + 2λ2 ) and σ 2 = (λ1 + 3λ2 ).
β β

2. Consider an experiment where two or three individuals are added to a population with
probability λ2 and λ3 respectively per unit time. The death rate in the population is a
constant β per individual per unit time.
Write down the master equation and derive an equation for ∂φ/∂t, where φ is the
generating function (as above). Find the solution for φ in steady state.
Show that for given target mean but otherwise a free choice of λ2 and λ3 , the exper-
imenter can minimise the variance by only adding two individuals at a time. Find this
minimum variance in terms of the mean.

3. Consider a birth-death process described by the master equation

dpn
= λ(pn−1 → pn ) + β [f (n + 1) pn+1 → f (n) pn ]
dt
with f (n) = n(n → 1)

i) Give an explanation of the terms on the right hand side.

ii) Show that the generating function φ(s, t) satisfies

∂φ ∂ 2φ
= λ(s → 1)φ + βs(1 → s) 2
∂t ∂s

1
 iii) Use the equation for φ in the steady state, or the master equation directly, to obtain
equations for "n2 # and "n3 # , in terms of µ = "n# and r = λ/β (do not try to evaluate
µ itself).

iv) With the mean µ unknown this system of equations is not closed. Nonetheless show
that the variance σ 2 = "n2 #→"n#2 ≤ r+ 14 . Show also, using the inequality "n2 # ≥ "n#2 ,
√
that "n# ≤ (1 + 1 + 4r)/2.
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

4. Consider a stochastic model of a population where the death rate is β per capita (so
total rate βn), and M individuals are added at the same time at rate λ (where M is a
positive integer).

i) Give the master equation and find the mean and variance of the population size at
steady state.

ii) Write down the Fokker-Planck equation for this system. Use this to find the mean
and variance of the population.

5. A particle starts at the origin (0, 0) at time t = 0. In each of the cases below, derive
the corresponding Fokker-Planck equation (in x, y). The particle moves in a random walk,
where it takes steps with the step sizes below, each with probability rate λ.

i) A square grid:
(→1, 0), (+1, 0), (0, →1) and (0, +1)

ii) A triangular grid: √ √ √ √

(+1, 0) , (→1, 0) , ( 21 , 23 ) , ( 12 , → 23 ), (→ 21 , + 23 ) and (→ 12 , → 23 )

iii) Square grid with bonus diagonal steps (breaking isotropy):

(→1, 0), (+1, 0), (0, →1) , (0, +1), (+1, +1) and (→1, →1)

For case iii), write down the master equation for the probability Pm,n (where m, n
are the discretised x, y) and use it to find "m2 + n2 #. Also calculate "x2 + y 2 # from the
Fokker-Planck equation.

6. A two-population dynamic model has the transition probability rates

(m, n) → (m + 1, n) : µ + λ1 n,
(m, n) → (m → 1, n) : β1 m,
(m, n) → (m, n + 1) : λ2 m,
(m, n) → (m, n → 1) : β2 n.

2
 i) Construct a master equation for Pm,n and use it to derive equations for the time
evolution of "m# , "n# . Find conditions on the parameters µ, λ1 , λ2 , β1 and β2 for
there to be a stable fixed point with "m# , "n# > 0.

ii) Write down the Fokker-Planck equation. Now consider small fluctuations near the
fixed point found above, and approximate u as linear in x and D as constant. Show
that the covariance matrix C satisfies
Copyright © 2023 University of Cambridge. Not to be quoted or reproduced without permission.

dC
= a C + C aT + b
dt
where a and b are matrices which should be given.

iii) For the special case when λ1 = λ2 = λ and β1 = β2 = β consider the equation for
dC/dt in components. Show that there is a fixed point for C (which need not be
explicitly found) and that it is stable. Explain what this means for this model.

* iv) As in iii), but with general λ1 , λ2 , β1 and β2 .