MPHY2002: Introduction to Biophysics

Chapter 3: Diffusion

3.1 Introduction

What would it be like if we could shrink ourselves to the size of a cell

and enter into the human body? There is a wonderful science fiction
film called “Fantastic Voyage” that explored this concept. In the film, a
scientist with a crucial piece of knowledge is rendered unconscious by
a blood clot in his brain. With new technology, an intrepid crew
miniaturises themselves and a nuclear submarine in order to navigate
within the blood vessels of the scientist and destroy the blood clot. The
crew encounters a world that looks much like our own in many ways,
and their submarine can glide smoothly.

In fact, reality is much more interesting! If we were really able to

shrink ourselves to cellular size scales, we would observe a world
much different from our own, in which particles change their speed
and direction seemingly at random as they collide with others.

We might imagine that random motion is detrimental to us. For

instance, particles that are actively transported to target locations
within cells could be impeded by collisions with particles traveling in
the opposite direction. Randomness can also be of critical importance,
however. In cases where particles are not actively transported, they
can still move passively within a cell due to collisions with particles.
This type of movement is called diffusion. With diffusion, individual
particles do not follow straight trajectories, but we might expect that
they eventually reach locations far from their starting points if they
experience a sufficient number of collisions.

1
MPHY2002: Introduction to Biophysics

We are familiar with many physics problems in which randomness is

not present (or more accurately, in which we choose to ignore it). An
example question might involve calculating the trajectory of a ball that
is initially at a specific location that is thrown at a specific velocity in a
specific direction. When randomness is prominent and there are many,
many particles all whizzing around in different directions, it might be
tempting to throw up our hands and say “we don’t have the slightest
clue about the positions and velocities of any of the particles, so there’s
nothing we can do!” Fortunately, it’s actually possible to make a lot of
progress with understanding random behaviour – but we’ll need to
think carefully about what questions we’d like to answer. This chapter
is about taking first steps in this direction.

3.2 Discrete diffusion model (1D)

In the spirit of a Biophysicist, let’s try to simplify the situation as much

as possible, and then build up complexity as necessary. Here are a
series of simplifications:

Our particle moves in a one-dimensional environment, so that its

motion is constrained along a line of infinite length.

• Our particle moves at constant velocity until it collides with

another particle, at which time there are only two possible
outcomes: the particle continues with the same velocity; b) the
direction of the particle changes but the magnitude of its
velocity is unchanged. These two outcomes take place with
equal probability and are independent of the position of the
particle.
• The time intervals between collisions are all equal.
• The speed (which is the magnitude of the velocity) stays
constant but that the direction at which the particle travels is
either the same or opposite in sign.

So what do these particle trajectories look like? Between collisions, the

distance travelled by a particle is constant, so we can represent a
trajectory as a series of “steps” on a line. Our particle might move like
this:

Our particle might also move like this:

2
MPHY2002: Introduction to Biophysics

Right away, we can see that if we ask the question: “what is the precise
location of our particle after n collisions,” we’d come up blank, because
we don’t know the positions and velocities of all other particles and
consequently we don’t know in which direction our particle will travel
after each collision. So this is not a particularly useful question to ask!

If we ask about the probability that the particle achieves a particular

displacement after n collisions, we can make progress. To answer this,
let’s start with an iterative equation that describes all random walks:

xn xn −1 + kn L
= (0.1)

where xn is the displacement after n collisions, L is the length of each

step (a constant in this model), and kn is a random variable that can be
either -1 or 1 with equal probability:

1
Pr [kn =−1] =
2 (0.2)
1
Pr [kn= 1=
]
2

Furthermore, let’s start our random walk at the origin: x0 = 0. To see

how Equation (0.1) works, imagine our particle at its starting point
( n = 0) . After one step, its displacement x1 is either L or –L, depending
on whether the step was to the right or to the left. If the first step was
to the right, we could either have x2 = 0 or x2 = 2L, depending on the
value of k2. Likewise, if the first step was to the left, we could either
have x2 = -2L or x2 = 0. In summary, for x2, we have:

Pr [ x2 =
−2 L ] =
0.25
Pr [ x2 =− L] =0
Pr [ x=
2 ] 0.5
0= (0.3)
Pr [ x=
2 ] 0
L=
Pr [= L ] 0.25
x2 2=

For a random walk of n steps, we can calculate the probability of any

particular displacement xn. To do so, it’s useful to notice that the
displacement is determined by the total number of steps to the left and

3
MPHY2002: Introduction to Biophysics

to the right. The order in which these steps occur does not matter.
Denoting the number of steps to the right as the random variable M,
the total displacement is:

xn = ML + ( n − M )( − L ) = ( 2M − n ) L (0.4)

Let’s consider a walk with three steps, so that n = 3 . To calculate the

probability that the displacement is xn = −1 , we notice that there
must be two steps to the left and one to the right, i.e. M = 1 . The three
walks that satisfy these conditions are:

RIGHT → LEFT → LEFT

LEFT → RIGHT → LEFT

LEFT → LEFT → RIGHT

The probability of the first walk is given by:

Pr [k1 =+1] ⋅ Pr [k2 =−1] ⋅ Pr [k3 =−1] =0.5 ⋅ 0.5 ⋅ 0.5 =0.125 . With the
same calculation method, we can see that the probabilities of the
second and third walks are also equal to 0.125. So we have
Pr [ xn =−1] =0.125 + 0.125 + 0.125 =0.375 .

3.3 Descriptive statistics for the discrete diffusion model (1D)

Calculating the probabilities of specific displacements is a starting

point for understanding diffusion, but it is impractical in almost all
experimental contexts because the number of collisions is so large and
it is impractical to keep track of individual particles. However, we can
calculate descriptive statistics that relate to experimentally measurable
quantities. One example of a descriptive statistic is the mean
displacement after n steps. Using Equation (0.1), we can write, using
the bracket notation (< >) introduced in Chapter 2:

xn
= xn −1 + kn L (0.5)

Bracket notation is useful because of its efficiency, but it is worth

reminding ourselves what it represents. On the LHS, we have a
summation of all possible values of xn, weighted by their respective
probabilities.

xn = ∑ Pr [ x=
n α ]⋅α (0.6)
α

Likewise, on the RHS, we have a summation of all possible values of

xn −1 + kn L , weighted by their respective probabilities.

4
MPHY2002: Introduction to Biophysics

∑ Pr [ xn−1 + kn L =α ] ⋅ α
xn −1 + kn L =
α
(0.7)

Using the additive property of the mean, and factoring out the constant
L, we can expand Equation (0.5) as:

xn
= xn −1 + kn L (0.8)

Now we can work backwards to n = 1 . The first term, x0 is identically

zero by definition. For the second term, we have:

kn =Pr [ kn =1] ⋅ (1) + Pr [ kn =−1] ⋅ ( −1)

1 1
= (1) + ( −1) (0.9)
2 2
=0

This means that x1 = 0 . The same arguments can apply to x2 , and

in general:

xn = 0 (0.10)

The fact that the average displacement is zero is quite reasonable: in

our model, there is no preferred direction – a particle is just as likely to
travel to the left as it is to travel to the right – so by symmetry the only
possible answer for xn is zero.

We arrive at a more interesting result if we ask the question: “what is

the mean-squared displacement of our particle after n collisions?” The
square of a value is always positive, and after many collisions it’s
rather unlikely that our particle will be at a displacement of exactly
( xn )
2
zero, so we may expect that the quantity will be non-zero.

Returning to Equation (0.1), we calculate the mean-squared

displacement:

( xn−1 + kn L )
2
xn 2
=

=xn −12 + 2 xn −1kn L + kn 2 L2 (0.11)

=xn −12 + 2 L xn −1kn + L2 kn 2

In the last step of Equation (0.11), we made use of the additive

property of the mean operator. Let us consider each term of the RHS of
Equation (0.11) in turn. There’s very little that we can do with first
( xn −1 )
2
term, , as it’s almost the quantity that is the same as the LHS

except that it corresponds to the previous step. For the second term,

5
MPHY2002: Introduction to Biophysics

2 L xn −1kn , we make use of the result that for independent random

variables, the mean of a product is the product of the means:

xn −1kn = xn −1 kn . (0.12)

We know from Equation that the second term of the product is 0, so

that the second term in the product is identically zero. This leaves the
final term in the LHS of equation (x). Noting that since kn can assume
2
only two values, -1 and 1, kn is always equal to unity. Therefore:

kn 2 = 1 (0.13)

As a result, we have

xn 2
= xn −12 + L2 (0.14)

Once again, we have a recursive equation; using the fact that x0 = 0 ,

we arrive at:

x12 = L2
(0.15)
x2 2 = 2 L2

so that in general:

xn 2 = nL2 (0.16)

Beyond 1D, the displacement is a vector quantity, which we denote as

d n . It is straightforward to extend Equation (0.16) to 2 and 3
dimensions if we assume that each Cartesian component of d n behaves
in the same way as the displacement in 1D.

It is straightforward to simulate 2D or 3D diffusion processes with a

computer. The plots below correspond to one random walk in 2D in
which 1,000,000 steps were taken. At each step, the simulated particle
moved a distance of 1 in the x- direction, and 1 in the y- direction, with
the directions of the steps chosen at random.

6
MPHY2002: Introduction to Biophysics

In two dimensions, we have d=

N xn i + yn j , where i and j are the unit
vectors in the x and y dimensions, respectively. Using the Pythagorean
theorem, the magnitude of d n is given by:

2
d n= xn 2 + yn 2 (0.17)

Taking mean values and using the additivity of means, we have:

2
d=
n xn 2 + yn 2 (0.18)

Since there is no difference between x and y dimensions, we

2
dn = 2nL2 (0.19)

Using the same type of calculation, we find that for three dimensions,
the mean-squared displacement is given by:

2
dn = 3nL2 (0.20)

Equations (0.16), (0.19), and (0.20) tell us how the mean-squared

displacement of a particle is related to the number of collisions that it
experiences. With biological experiments, our independent variable is
usually the time travelled by a particle. The number of collisions n and
the time t are related by the length of time between collisions, ∆t :
n= t ∆t so that using Equation (0.16), we have for 1-dimensional
diffusion:

 t 
xn 2 =   L2
 ∆t 
(0.21)
 L2 
=t 
 ∆t 

7
MPHY2002: Introduction to Biophysics

In the second line of Equation (0.21), the quantity in brackets is a

constant, which tells us that the mean-squared displacement is directly
proportional to time. The environment in which the particle is situated
determines the constant: it determines the distance that a particle
travels between collisions (the spatial scale of the diffusion process)
and the time between collisions (the temporal scale of the diffusion
process). It is conventional to define a quantity called the diffusion
constant that encapsulates both spatial and temporal scales:

L2
D≡ , (0.22)
2∆t

so that we have:

xn 2 = 2 Dt
(0.23)
( Diffusion in 1 dimension )
The factor of 2 in Equation (0.23) may seem like an unnecessary
complication in this derivation, but as we will see later in this chapter,
it actually leads to simpler equations when the diffusion process is
described in terms of differential equations.

Equations (0.19) and (0.20) can be rewritten as:

2
dn = 4 Dt
(0.24)
( Diffusion in 2 dimensions )
2
dn = 6 Dt
(0.25)
( Diffusion in 3 dimensions )
3.3 Binomial distributions and the discrete diffusion model

In this section we start with the discrete diffusion model in 1-D, and
explore what happens to the probability distributions when there are a
large number of steps.

The displacement of a particle in a particular 1-D random walk is

determined by the number of steps in a given direction. In particular, if
there is a total of n steps and M of them are to the right, there must be
n − M steps to the left. If each step is of length L, the displacement is
given by:

xn = ML − ( n − M ) L . (0.26)

8
MPHY2002: Introduction to Biophysics

Conversely, the number of steps in a particular direction is determined

by the displacement. Indeed, if we are given xn , we can calculate M by
rearranging Equation (0.26):

1
=M ( nL + xn ) . (0.27)
2L

In the case where steps to the right and to the left have equal
probabilities, and all steps are independent, the probability of a
particular random walk with M steps to the right is given by:
(1 2 ) ⋅ (1 2 ) (1 2 ) .
M n−M n
=

The number of walks in which there are M steps to the right is given by
the binomial coefficient, which is spoken as “n choose M” and written
as:

n n!
 = . (0.28)
M
  M !( n − M ) !

It follows that the probability that a random walk has M steps to the
right is:

n
 n  1
Pr [ M ]  ⋅ 
= m= (0.29)
M  2

As an example, we could ask: in the case that n = 10 and L = 1 , what is

the probability that xn = 2 ? Using equation (0.27), M = 6 .
Substituting this value into equation (0.29), the probability is
therefore:

10
 10  1
Pr [ x=
n ] Pr [ M= 6=]     = 0.205 .
2=
 6  2 

The probability distributions defined in Equation (0.29) belong to a

particular class of distributions called Binomial distributions. In
general, we obtain a Binomial distribution when there are n
independent events, and at each event either one outcome takes place
with probability p, or another outcome takes place with probability
q = 1 − p . It turns out that if the random variable M has a Binomial
distribution, the mean, µ M , and the variance, σ M 2 , are given by:

µM
= M
= np
. (0.30)
σ M 2 =( M − M )
2
npq
=

9
MPHY2002: Introduction to Biophysics

As an example of a Binomial distribution is a series of coin tosses: if we

were to toss one coin 16 times and if the coin were unbiased, we would
have p= q= 1 2 so that we would have 8 heads on average, and a
standard deviation of 2.

Using Equation (0.29), and the particular case p= q= 1 2 for our

random walk, we can calculate histograms in which the probabilities of
individual displacements of are plotted, with the number of steps held
constant. In the following plots, a step length of unity was assumed. At
the beginning of a random walk, before any steps are taken, the
probability is unity that the displacement is non-zero.

After one step, the distribution broadens.

After two steps, the distribution assumes a peak at zero.

10
MPHY2002: Introduction to Biophysics

After six steps, the distribution starts to assume a bell-like shape.

The plots above suggest that as n gets larger, the probability

distribution tends to get even more bell-shaped. It turns out that if n
and the product np are both large, the binomial distribution can be
approximated by a continuous Gaussian distribution:

1  ( m − µ M )2 
=p ( m) exp  −  (0.31)
2πσ M 2  2σ M 2 

where the values σ M and µ M are defined in Equation (0.30). Equation

(0.31) can be interpreted in terms of probabilities with the equation:

Pr [ m < M < m + dm ] =p ( m ) dm .

Equation (0.31) provides the probability density function

corresponding to the number of steps to the right. To transform this to
the corresponding PDF for particle displacement, we have:

x = mL − (n − m) L = ( 2m − n ) L (0.32)

11
MPHY2002: Introduction to Biophysics

Equation (0.32) implies that dx = 2 Ldm . Furthermore, for the

unbiased random walk, we have: p= q= 1 2 , t = n∆t , and
D L
= 2
( 2∆t ) . With these substitutions, Equation (0.31) becomes:

1  − x2 
p ( x) = exp  . (0.33)
4π Dt  4 Dt 

3.5 Continuous diffusion model

In the previous section, we analysed the diffusion process in terms of a

random walk by individual particles that start at one location. In many
biophysics contexts, the particles of interest are distributed
throughout a region of space. Imagine, for instance, that we are
interested in the diffusion of oxygen molecules from the exterior to the
interior of a cell. In these contexts, it is convenient to consider a scalar
quantity called the number density. In one-dimension, the number
density is the number of particles per unit length; likewise in two or
three dimensions, it is the number of particles per unit area or volume,
respectively.

To start, let’s consider an arbitrary distribution c ( x, t ) . What does our

intuition tell us about how c ( x, t ) will change as time increases? One

way to think about this is to consider two equally sized adjacent
regions of space in which the region on the left has more particles than
the region on the right. As a result of this difference in number density,
it will be more likely that more particles will move from left to right
than from right to left. In general, we can imagine particles will tend to
flow from regions of high concentration to regions of low
concentration, so that the effect of diffusion is to “flatten” or “blur”
c ( x, t ) as time increases. Let’s make those concepts quantitative.

As with the one-dimensional discrete diffusion model, we partition

space into bins of length L, and denote the number of particles in a bin
centred at point x at time t as N(x,t).

12
MPHY2002: Introduction to Biophysics

At every time step ∆t , each particle moves a distance L either to the

right or to the left, at random. Approximately half of a given bin’s
population hops to the left, and half hops to the right. With this
approximation, the net number of particles that move from left to right
across the boundary between the bins centred at x − L and at x
(denoted as “a” in the Figure above) is the difference between the
number crossing from the bin centred at x − L to bin centred at x (left
to right) and the number crossing from the bin centred at x to the bin
centred at x − L (right to left): N ( x − L, t ) 2 − N ( x, t ) 2 .

The net number of particles moving from left to right per unit time is a
useful quantity: it is called the flux, j. Based on the above calculations,
we have:

N ( x − L, t ) 2 − N ( x , t ) 2
j=
∆t
(0.34)
1 N ( x − L, t ) − N ( x , t )
=
2 ∆t

Equation (0.34) can be written in terms of number density, using the

relation N ( x ) = c ( x ) L :

c ( x − L, t ) L − c ( x , t ) L
j=
2∆t
(0.35)
 L   c ( x − L, t ) − c ( x , t ) 
2
=  ⋅ 
 2∆t   L 

In the last line of equation (0.35), both the denominator and

numerator were multiplied by L. The first bracketed term in Equation
(0.35) was already defined in Equation (0.22): it is the diffusion
constant. In the limit that L is very small, the second bracketed term of
that line is approximately equal to the partial derivative of c(x,t) with
respect to x:

c ( x − L, t ) − c ( x , t ) ∂c
≈− (0.36)
L ∂x

This leads us to Fick’s law:

∂c
j = −D (0.37)
∂x

Fick’s law tells us that the flux is directly proportional to the spatial
derivative of the concentration. While this is useful, it is not the end of
the story: it does not say what happens to particles as they flow from
one region to another. In our model, the total number of particles is

13
MPHY2002: Introduction to Biophysics

conserved. In particular, if there is a difference in flux at the

boundaries of a spatial region, that difference will directly affect the
number of particles that bin centred at x, N ( x, t ) . We can now see
why the Diffusion coefficient was defined as it was in Equation (0.22):
it leads to a simple expression for Fick’s law.

Looking again at Figure above, we see that the number N ( x, t )

changes in one time step for two reasons: Particles can cross the
boundary “a,” and they can cross boundary “b” – but particles are not
created or destroyed. Recalling that j refers to the net flux from left to
right, we find the net change to be:

∂N ( x, t )  L  L
= j x − − j x +  . (0.38)
∂t  2  2

Using the fact that N ( x ) = c ( x ) L , we have:

∂  c ( x, t ) L   L  L
= j x − − j x +  (0.39)
∂t  2  2

Once again, we may take the bins to be narrow, whereupon the right-
hand side of this formula becomes (-L) times a derivative. Dividing by
L then gives:

∂c ∂j
= − (0.40)
∂t ∂x

a result known as the continuity equation. Fick’s law and the

continuity equation refer to separate but related ideas: the former
relates the flux is related to the numbers of particles, whereas the
latter expresses the idea that that the number of particles is conserved.
We can combine them to obtain a differential equation for c ( x, t ) . By
taking the spatial derivative of both sides of Fick’s law and substituting
into the continuity equation, we arrive at:

∂c ∂ 2c
=D 2 (0.41)
∂t ∂x

In three dimensions, the single spatial derivative is replaced by the

Laplacian (del-squared) operator.

∂c 2 ∂ 2c ∂ 2c ∂ 2c
=∇ c= 2 + 2 + 2 (0.42)
∂t ∂x ∂y ∂z

Let’s consider whether the diffusion equation is consistent with our

intuition. Consider a distribution of particles. In areas with a higher

14
MPHY2002: Introduction to Biophysics

concentration, the second spatial derivative of c is negative, so the

concentration will decrease with time. Likewise, in areas with lower
concentration, the second spatial derivative of c is positive, so that the
concentration will increase with time. If Robin Hood had an equation,
this might be it: “take from the rich and give to the poor!”

As you can verify in Question 2, a solution to Equation (0.41) is:

n
c ( x, t ) =
2
/4 Dt
e− x . (0.43)
4π Dt

It is satisfying that Equation (0.43) is consistent with the equation

derived from the Binomial model, Equation (0.33). If we compare
Equation (0.43) to the generic equation for a Gaussian probability
distribution:

1  − ( x − µ )2 
p ( x) = exp  
2πσ 2  2σ 2 
 

we find that:

σ 2 = 2Dt ,
which tells us that the concentration spreads out with time at a rate
that is proportional to the diffusion constant D. Once again, we find
D L2 2∆t and
agreement with the Binomial model: substituting =
n= t ∆t , we obtain once again σ 2 = nL2 .

Additionally, with this comparison, we find that:

µ =0,

which means that the concentration of particles is distributed equally

about x = 0 . This is reasonable, because in this model there are no
external forces that and therefore there is no preferred direction in
space.

Questions

1) Consider a particle that undergoes a random walk in one

dimension.

i) For the first n steps, the particle takes steps of length

L1 (in the positive direction with probability ½ and in
the negative direction with probability ½). What is the
RMS distance travelled?

15
MPHY2002: Introduction to Biophysics

ii) After the n steps travelled with steps of length L1

described in part i), the particle takes another n steps
with a different length L2 (in the positive direction
with probability ½ and in the negative direction with
probability ½). What is the total RMS distance travelled
after these n+n=2n steps?

2) Note: this problem is purely supplemental; it is not

necessary to solve it for the purposes of the MPHY0006
module in 2021 and beyond.

Consider an event in which particles are released and undergo

diffusion in a medium. For simplicity, consider the case where
the geometry is one dimensional, so that the concentration of
the particles as a function of distance from the origin and time
is given by c(x,t).

i) At t=0, there are n particles concentrated at the origin.

Show explicitly that the concentration profile:
𝑛𝑛 2
𝑐𝑐 (𝑥𝑥, 𝑡𝑡) = 𝑒𝑒 −𝑥𝑥 ⁄4𝐷𝐷𝐷𝐷
√4𝜋𝜋𝜋𝜋𝜋𝜋
is a solution to the diffusion equation. Sketch the
solution for t=1/D and t=1/(2D), indicating on each
graph the x coordinates at which c reaches 1/e2 of its
maximum value.

ii) Now consider a different initial distribution for the

particles: at t = 0, n/2 particles are concentrated at x = -
1 and n/2 particles concentrated at x = 1. Using the
answer to i), derive the expression for c(x,t). Find the
time t=tBlurred at which the concentration profile no
longer has two distinct peaks (tBlurred should be solely a
function of D).

16
MPHY2002: Introduction to Biophysics

Solution to 2) above

2 i)

Note: this problem is purely supplemental; it is not

necessary to solve it for the purposes of the
MPHY0006 module in 2021 and beyond.

To show that c(x,t) is a solution, it suffices to show that when we plug it

into the diffusion equation, the right hand side of the equation is equal to
the left hand side. The diffusion equation is:

𝜕𝜕𝜕𝜕 𝜕𝜕 2 𝑐𝑐
= 𝐷𝐷 2
𝜕𝜕𝜕𝜕 𝜕𝜕𝑥𝑥
Using the chain rule, we have:
3
1 𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 𝑥𝑥 2 −2
LHS = − 𝑡𝑡 −2 𝑒𝑒 −𝑥𝑥 + 𝑒𝑒 −𝑥𝑥 � 𝑡𝑡 �
2 √4𝜋𝜋𝜋𝜋 √4𝜋𝜋𝜋𝜋𝜋𝜋 4𝐷𝐷

𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 1 𝑥𝑥 2 −2
= 𝑒𝑒 −𝑥𝑥 �− 𝑡𝑡 −1 + 𝑡𝑡 �
√4𝜋𝜋𝜋𝜋𝜋𝜋 2 4𝐷𝐷

For the RHS, we have:

𝜕𝜕𝜕𝜕 𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 2x
= 𝑒𝑒 −𝑥𝑥 �− �
𝜕𝜕𝜕𝜕 √4𝜋𝜋𝜋𝜋𝜋𝜋 4Dt

𝜕𝜕2 𝑐𝑐 2 𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 2x 2

RHS = 𝐷𝐷 = 𝐷𝐷 �− � 𝑒𝑒 −𝑥𝑥 + 𝐷𝐷 𝑒𝑒 −𝑥𝑥 �− �
𝜕𝜕𝑥𝑥 2 4𝐷𝐷𝐷𝐷 √4𝜋𝜋𝜋𝜋𝜋𝜋 √4𝜋𝜋𝜋𝜋𝜋𝜋 4Dt

𝑛𝑛 2 ⁄4𝐷𝐷𝐷𝐷 1 𝑥𝑥 2 −2
= 𝑒𝑒 −𝑥𝑥 �− 𝑡𝑡 −1 + 𝑡𝑡 �
√4𝜋𝜋𝜋𝜋𝜋𝜋 2 4𝐷𝐷

LHS = RHS, so the function c(x,t) is a solution to the diffusion equation.

For the sketch, note that at any time, the maximum value of c, cmax, occurs
at x=0. We therefore have:
𝑛𝑛
cmax = .
√4𝜋𝜋𝜋𝜋𝜋𝜋

When the function reaches 1/e2 = e-2 of its maximum value we have:

e-2cmax = c(x,t)
𝑛𝑛 𝒏𝒏 𝟐𝟐 ⁄𝟒𝟒𝟒𝟒𝟒𝟒
e-2 = 𝒆𝒆−𝒙𝒙
√4𝜋𝜋𝜋𝜋𝜋𝜋 √𝟒𝟒𝟒𝟒𝟒𝟒𝟒𝟒

so that by setting the exponents equal:

−2 = − 𝑥𝑥 2 ⁄4𝐷𝐷𝐷𝐷

For t=1/D, x=±2√2

17
MPHY2002: Introduction to Biophysics

And for t=1/2D, x=±2

A sketch of the functions:

2 ii)

Note: this problem is purely supplemental; it is not

necessary to solve it for the purposes of the
MPHY0006 module in 2021 and beyond.

Note that if c-1(x,t) and c1(x,t) are solutions to the diffusion equation, then
c-1(x,t) + c1(x,t) is also a solution. This is because of the linearity of the
differentiation process, and it should be consistent with your intuition:
c(x,t) is just a measure of how many particles there are per unit distance
(per unit volume in three dimensions), and we can add up counts of
different particles to get the total number of particles.

For the distribution of particles centred at x=-1, it suffices to simply shift

the solution to the left and reduce the number of particles by half.
(Convince yourself that if c(x,t) is a solution, c(x+1,t) is also a solution.).
This distribution is therefore:

𝑛𝑛/2 2 ⁄4𝐷𝐷𝐷𝐷
𝑐𝑐−1 (𝑥𝑥, 𝑡𝑡) = 𝑒𝑒 −(𝑥𝑥+1)
√4𝜋𝜋𝜋𝜋𝜋𝜋

18
MPHY2002: Introduction to Biophysics

Similarly, for the distribution of particles centred at x=1, it suffices to shift

the solution to the right and reduce the number of particles by half:

𝑛𝑛/2 2 ⁄4𝐷𝐷𝐷𝐷
𝑐𝑐1 (𝑥𝑥, 𝑡𝑡) = 𝑒𝑒 −(𝑥𝑥−1)
√4𝜋𝜋𝜋𝜋𝜋𝜋

So the required expression for the total number of particles c(x,t) is:

𝑛𝑛/2 2 ⁄4𝐷𝐷𝐷𝐷 𝑛𝑛/2 2 ⁄4𝐷𝐷𝐷𝐷

c(x,t) = 𝑒𝑒 −(𝑥𝑥+1) + 𝑒𝑒 −(𝑥𝑥−1)
√4𝜋𝜋𝜋𝜋𝜋𝜋 √4𝜋𝜋𝜋𝜋𝜋𝜋

To find tblurred, note that initially, for small times, the concentration as a
function of x consists of two peaks. Moreover, the concentration increases
with increasing distance from x=0 (just sketch a solution to convince
yourself of this). This means that for small times,
𝜕𝜕2 𝑐𝑐
> 0.
𝜕𝜕𝑥𝑥 2
After time tblurred, the concentration as a function of x consists of just one
peak, so that
𝜕𝜕2 𝑐𝑐
< 0.
𝜕𝜕𝑥𝑥 2
So for tblurred, we seek a time t such that
𝜕𝜕2 𝑐𝑐
= 0.
𝜕𝜕𝑥𝑥 2
We could calculate the second derivative of c-1(x,t)+c1(x,t), but by
𝜕𝜕2 𝑐𝑐
symmetry we know that if = 0, each of the terms should equal zero.
𝜕𝜕𝑥𝑥 2
𝜕𝜕2 𝑐𝑐
By symmetry we also know that = 0 occurs for x=0.
𝜕𝜕𝑥𝑥 2

So we set:
𝜕𝜕 2 𝑐𝑐1
=0
𝜕𝜕𝑥𝑥 2
Using the result from part i), we have:
𝑛𝑛 2 1 (𝑥𝑥 − 1)2 −2
𝑒𝑒 −(𝑥𝑥−1) ⁄4𝐷𝐷𝐷𝐷 �− 𝑡𝑡 −1 + 𝑡𝑡 �� =0
√4𝜋𝜋𝜋𝜋𝜋𝜋 2 4𝐷𝐷 𝑥𝑥=0
1 −1 1 −2
− 𝑡𝑡 + 𝑡𝑡 = 0
2 4𝐷𝐷

Setting this equal to zero and solving for t gives tblurred:

1
tblurred=
2𝐷𝐷

19