Diffusion equation for the random walk

Random walk in one dimension

l = step length
= time for a single step
p = probability for a step to the right, q = 1 p is the probability for a step to the left
PN(m) = probability to find the walker at position x = ml at time t = N
A random walk is a Markov process.
Let j and k be states (in this case
positions) and let p(jk) be the
probability for a transition from j to k,
then the transition probabilities
1. are independent of time
2. depend only on the states j and k,
not on the history of the system
3. obey the sum rule P( j k ) = 1
(some state must k
be reached)

The probability PN(m) satisfies the

stochastic difference equation

PN +1(m) = pPN (m 1) + qPN (m + 1)

t/
PN+1(m)
N+1
p
N

Subtract PN(m) on both sides

in the limit of large N, the
differences become differentials

Diffusion equation

P
2P
=D 2
t
x

PN(m+1)

PN(m1)
m1

Specialize to the case p = q = 1/2

D is the (self) diffusion coefficient

boundary condition, for all times

P ( x, t ) 0 as x

initial condition (delta peak at the origin)

P ( x,0) = ( x)

P ( x, t ) =

with

2 = 2 Dt

x2

exp
2 2
2 2

x/l

1
1
PN +1(m) = PN (m 1) + PN (m + 1)
2
2
1
P +1(m) PN (m) = (PN (m 1) + PN (m + 1) 2 PN (m) )
N

 

2  

2
P
P

l 2
t
x 2
2 2
P 1 l P
l2
=
D
=
t N
2 x 2
2
D

Solve the diffusion equation with

solution

m+1

For any time t

P ( x) dx =

x2
1
dx
exp

4Dt
4 Dt

As time goes on, the probability packet spreads

3

P(x)

t1 > 0
2.5

t2 > t1

t3 > t2

1.5
1
0.5
0
-1

-0.5

0
x

0.5

Mean square displacement (in one dimension)

in three dimensions one finds

r 2 = x 2 + y 2 + z 2 = 6 Dt

P( x)dx = 2 Dt