MA1024 – Discussion 5, Microscopic Spring-mass System

Microscopic things, when immersed in water, undergo Brownian motion – they move ran-
domly due to the collisions of the water molecules. A microscopic particle on a spring, for
example, does not swing smoothly back and forth, but rather fluctuates in an apparently
random fashion about its equilibrium position.

Although the motion of the particle cannot be precisely predicted, the probability density
of the particle can be. This is a function, ρ, which can be used to predict the probability
of finding the particle in a particular region of space. These probability densities play an
important role in connecting molecular properties to bulk properties, a focus of statistical
mechanics.

Here, we will consider a particle on a spring moving in two dimensions, and the probability
density function ρ(x, y).

1
For a microscopic particle on a spring, immersed in water, the probability density
2 +y 2 )
ρ(x, y) = N e−(x

where N is a constant, called the normalization constant. In this problem, you will estimate
and then calculate this constant.

A probability density, ρ(x, y) is related to probability in the following way. The probability,
P , of finding the particle in a small square of width dx and height dy centered at x, y is

P = ρ(x, y)dxdy = ρ(x, y)dA

We can use this equation to estimate the normalization constant.

θ = π/2

θ = π/3 0 0

0 5
r = 3.0 0 0
2.5 5 6
1 0
θ = π/6
2.0
1 28 10
21
23 21
0 44 43 0
0 39 31 2
6 5
24 28 29 19
33 20 13 37
θ=0 19 19
18 13
14
34 16 24 25
21
25 18
4 6
0 30 36 2
0 29 30 0
14 23

6 27 29 7

1 3
6 4
0 1
1 1

0 0

1a) Approximate the area, dA, of each of the three regions drawn in red. Note that ∆r = 0.5
and ∆θ = π/6.

1b) There are 1000 measurements of the particle position shown at left. The numbers, at
right, show the number of measurements in each region (i.e., there are 19 measurements in
the red region nearest the center). Use these numbers to estimate the probability, P , of
finding the particle in each of the three red regions.

1c) Using the equation above, transformed to polar coordinates, P ≈ ρ(r, θ)dA, use your
answers to parts 1a and 1b to estimate the normalization constant, N .

1d) Suppose you wanted to estimate the normalization constant N as accurately as possible
from these measurements. How could you do it? (There are many possible answers).

2
One way to increase the accuracy of our estimate would be to increase the resolution – that
is, make ∆r and ∆θ smaller. However, if you use the same procedure as part I, decreasing
∆r and ∆θ increases the error of your estimate of P .

The trick to increasing accuracy is to take one of the larger regions and then divide it into
smaller regions. To see how this works, let’s consider the region towards the top, with 21
measurements.

2a) Divide that region into fourths (i.e., make ∆r = 0.25 and ∆θ = π/12), and calculate
the areas of each of the four regions.

2b) Now, the probability P that you calculated for this region in problem 1b is
4
X
P = ρ(ri , θi )dAi
i=1

where ri and θi are r, θ pairs somewhere in the four regions you found in part 2a, and dAi
are the corresponding areas you calculated in part 2a.

2c) What would happen if you divide that region into smaller and smaller pieces, taking
the limit as ∆r → 0 and ∆θ → 0?

2d) Take the limit described in problem 2c, to end up with an answer similar to what you
got in 2b and 1c.

3
Another way to find the normalization constant is to solve the following equation
Z ∞Z ∞ Z ∞Z ∞
2 2
ρ(x, y)dxdy = N e−(x +y ) dxdy = 1
−∞ −∞ −∞ −∞

Note that you cannot solve this integral in Cartesian coordinates.

3a) Convert this equation to polar coordinates and solve for N .

3b) How does your answer compare to what you found in parts 1 and 2?

3c) Which one do you think is more accurate? Why?