Professional Documents
Culture Documents
The Maxwell-Boltzmann distribution of molecular speeds in a gas is actually a probability density function
of a continuous variable, v, the speed of a molecule. You may be familiar with probability distribution
functions for discrete variables. For example, the probability of getting heads by flipping a fair coin is 12 ;
the probability of rolling snake-eyes (two ones) with a pair of dice is
seven is
1
6
1
36
. For continuous variables, you can not obtain the probability of a molecule having a specific
velocity from the probability density function. Rather, you must determine the probability of a molecule's
speed being in a certain range by integrating the probability density function over that range.
In other words, you can not evaluate the Maxwell-Boltzmann distribution function for v = 2.0 m/s and get
the probability of a molecule travelling at that speed. What you can do is determine the probability that a
molecule's speed is between 1.9 m/s and 2.1 m/s by integrating the Maxwell-Boltzmann distribution
function using 1.9 m/s and 2.1 m/s as the limits.
However, you can calculate expected values for v and v2, which you can use to calculate the average speed
and average kinetic energy of a molecule, respectively.
The basic properties of probability density functions (pdfs), and the definition of expected value can be
found in any probability textbook, such as Introduction to Probability by Bertsekas and Tsitsiklis, or A
First Course in Probability by Sheldon Ross.
First, the value of the integral of the pdf over all possibilities must be one.
+
pdf (x )dx = 1
In the case of the Maxwell-Boltzmann distribution, f(v) = 0 for v less than zero, because speed is never
negative. Our integrals over all possible speeds will be from zero to infinity. Also, the expected value of a
given function of x is the integral of that function weighted by the probability density function:
g (x ) =
g (x )pdf (x )dx
= x =
x pdf (x )dx
2 = ( x )2 =
(x )
pdf ( x )dx
We will be using the mean value equation, but not the variance equation. We will be finding the expected
value of the square of the speed, which is not the same as the square of the expected value of the speed.
v2 v
Triton College
PHY 107
John Baliga
1 of 8
2kT
32
v 2 e mv
2 kT
m
f (v )dv = 4
2kT
32
v 2 e mv
2 kT
dv
There are many ways to perform this integration. One way commonly used in statistical mechanics
applications is detailed below.
A substitution will be used that changes the integration variable to one that is dimensionless:
u=v
m
2kT
f (v )dv =
v e
2kT
2 mv 2 2 kT
f (v )dv =
f (v )dv =
0
u e
m
dv
2kT
2 u 2
du
f (v )dv = 1
0
By choosing a variable that simplifies the argument of the exponential, the overall integral is simplified.
Also, by choosing a dimensionless integration variable, any dimensioned quantities can be factored out of
the integral. Before the integration in terms of u was even performed, the factor outside of it was found to
be a dimensionless constant. Since the value of the integral was supposed to be the dimensionless constant
1, we have preliminary confirmation that the distribution function has the correct dimensions. In other
words, all the "physics" was factored out first, and the integral was just a dimensionless factor.
Evaluation of this integral:
u e
2 u 2
du =
is covered later.
Triton College
PHY 107
John Baliga
2 of 8
d
m
f (v ) = 4
dv
2kT
m
4
2kT
32
2ve mv
2 kT
2v
32
d 2 mv 2
v e
dv
2 kT
)= 0
2mv mv 2
+ v2
e
2kT
2 kT
= 0
m 3
v =0
kT
2kT
2kT
v 0,
,
m
m
2kT
is outside the domain, we exclude it. The minimum value of f(v) is zero, and since f(0) = 0,
m
we will exclude v = 0. Since the function is zero at the zero end of its domain, approaches zero as v goes to
Since
infinity, and is positive-valued for all non-zero v, we can safely assume that v =
2kT
corresponds to a
m
maximum value for f(v). We will forego the second derivative test. Therefore:
2kT
m
vmp =
v = vf (v )dv
m
v = 4 v
2kT
0
m
v = 4
2kT
0
Triton College
PHY 107
John Baliga
32
32
v 2 e mv
v 3e mv
2 kT
2 kT
dv
dv
3 of 8
Again, letting u = v
m
:
2kT
m
v = 4
2kT
0
32
0 2kT
v =
v =
2kT
m
v 3e mv
32
v 3e mv
2 kT
dv
2 kT
dv
m 1 2 mv 2
v e
2kT
2 kT
m
dv
2kT
2
2kT
v =4
u 3e u du
m
v =4
v =
u e
3 u 2
du =
2kT
m
1
2
8kT
m
1
will be shown later.
2
v 2 = v 2 f (v )dv
m
v 2 = 4 v 2
2kT
0
m
v 2 = 4
2kT
0
Triton College
PHY 107
John Baliga
32
32
v 2 e mv
v 4 e mv
2 kT
2 kT
dv
dv
4 of 8
Again, letting u = v
m
:
2kT
m
v 2 = 4
2kT
0
32
0 2kT
v 4 e mv
32
2 kT
dv
2 kT
dv
v2 =
v2 =
4 2kT m m
m 0 2kT 2kT
v 4 e mv
12
32
v 4 e mv
2 kT
m
dv
2kT
12
4 2kT m
mv 2
=
v e
m 0 2kT
2 kT
m
dv
2kT
v2 =
4 2kT 4 u 2
u e du
m 0
v2 =
4 2kT 3
m 8
v2 =
vrms =
u e
4 u 2
du =
3kT
m
v2 =
3kT
m
3
will be shown in the next section.
8
In summary:
vmp =
8kT
2kT
, vav =
, vrms =
m
m
3kT
m
Note that:
Triton College
PHY 107
John Baliga
5 of 8
Since we need to evaluate more than one improper integral containing the term e x , let's start with a
general integral of this form and see if we can save some time.
2
Let Ik represent the integral from zero to infinity with xk in the integrand:
I k = x k e x dx
2
To evaluate this integral, we will integrate by parts. We will assume that k is non-negative.
Letting u = e x and dv = x k dx , we also have du = 2 xe x dx and v =
2
1 k +1
x :
k +1
I k = x k e x dx
2
2
1
1 k +1 x 2
Ik =
x e
x k +1 2 xe x dx
k +1
0 k + 1 0
Ik =
2
1
[0 0] + 2 x k + 2e x dx
k +1
k +1
Ik =
2
Ik+2
k +1
Assuming that k is non-negative is important when evaluating the uv term at x = 0, and for avoiding
division by zero at k = 1. This restriction does not conflict with our purposes. This does not give us the
value of any integral, but it does give us something very valuable. The relation derived here is a recurrence
relation. If we know any Ik, then we can quickly evaluate the corresponding Ik+2, if k is non-negative.
Actually, we are going to use this recurrence relation in a different form by substituting k = n 2.
Ik =
I n -2 =
2
In
n 2 +1
I n -2 =
2
In
n 1
In =
Triton College
PHY 107
John Baliga
2
Ik+2
k +1
n 1
I n -2
2
6 of 8
With our recurrence relation, we need only to evaluate the integral for two non-negative consecutive
integer values of n, and we can easily find evaluate the integral for any such n. Of course, we will have to
use methods other than integration by parts, since that method did not give us direct answers.
We will evaluate I0 and I1, and use our recurrence relation to get the values through I4.
To evaluate I0, we will use a common trick:
I 0 = e x dx
2
I0 = e
2
x2
dx e y dy
I0 =
2
x2 + y2
)dxdy
0 0
I0 =
2
r 2
rdrd
0 0
e
2
I 02 =
r 2
rdr
1
From here, we simply use the substitution: u = r 2 du = 2rdr du = rdr
2
I0 =
2
I0 =
2
r 2
rdr
r =
u
e du
2 2
r =0
I02 =
I02 =
Triton College
PHY 107
John Baliga
I02 =
I0 =
[e ]
u r =
r =0
[e ]
4
r 2
[0 1] =
7 of 8
Evaluating I1 is straightforward:
I1 = xe x dx
2
1
Let u = x 2 du = 2 xdx du = xdx
2
I1 = xe x dx
2
x =
1
I1 = e u du
2 x =0
[ ]
1
I1 = e u
2
x =
x =0
[ ]
2
1
= e x
2
1
1
I1 = [0 1] =
2
2
1
1
n 1
and I1 = , we can use our recurrence relation I n =
I n - 2 to
2
2
2
determine the required values.
I 0 = e x dx =
2
I1 =
xe
x2
dx =
I2 =
2 x2
I3 =
x e
3 x2
I4 =
x
0
4 x2
1
2
2 1 1
1
dx =
=
2 2
4
3 1 1 1
dx =
=
2 2 2
3
4 1 1
dx =
=
8
2 4
Values of In for integer values of n greater than four can be found in a straightforward manner if desired.
Triton College
PHY 107
John Baliga
8 of 8