Notes on the Distribution Function

Prof. Marco Panesi

Center for Hypersonics and Entry Systems Studies (CHESS)
University of Illinois

Contents
1 Distribution Function: The Basics 1

2 Molecular Collisions 4
2.1 Collision Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Definition of Cross Section . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Definition of Collision Rate . . . . . . . . . . . . . . . . . . . . . 5
2.4 Hard Sphere (HS) Model . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Variable Hard Sphere (VHS) Model . . . . . . . . . . . . . . . . 6
2.6 Elastic Collision Definition . . . . . . . . . . . . . . . . . . . . . . 6
2.7 Scattering Angles and Cross Sections . . . . . . . . . . . . . . . . 7

3 Maxwellian Distribution 8

In kinetic theory, the distribution function is a crucial concept that is used

to describe how the velocities of particles in a dilute gas are distributed at any
given time. These notes introduce the fundamentals of the distribution func-
tion and its applications in the context of particle motion and collisions. Key
topics covered include the mathematical formulation of the distribution func-
tion, differential fluxes of molecules, momentum, and energy, and the treatment
of molecular collisions, such as elastic collisions and their role in determining
equilibrium states like the Maxwell-Boltzmann distribution. Furthermore, mod-
els like the hard-sphere and Variable-Hard-Sphere (VHS) models are discussed,
providing insights into how collision rates and cross sections are computed in
gas dynamics. For more in-depth details, please refer to the provided references.

1 Distribution Function: The Basics

In this so-called velocity space the molecule is thus represented by a point located
at the end of the velocity vector Ci laid off from the origin. The entire gas in
the volume V is represented correspondingly by a cloud of N points in this space.

1
 Figure 1: Cartesian and Spherical Coordinates

The number of particles number of particles of class Ci in the volume element

dVC is
dn = nf (Ci )dVC
The distribution function f (Ci ) gives the fraction of molecules in class Ci , that
is, the fraction with velocity in the range (Ci , Ci + dCi ). This function is fun-
damental in kinetic theory to describe how molecules are distributed in terms
of their velocities.

f (Ci )dVc = fraction of molecules in class Ci with velocities in the range (Ci , Ci +dCi )

Here, Ci is a vector representing the components of the velocity in three

Cartesian coordinates:

Ci ≡ (C1 , C2 , C3 )
The volume element dVc in velocity space can be written in two coordinate
systems:

• In Cartesian coordinates, it is given by:

dVc = dC1 dC2 dC3

• In Spherical polar coordinates, the volume element is:

dVc = C 2 dC sin(θ) dθ dϕ

where C is the magnitude of the velocity, θ is the polar angle, and ϕ is

the azimuthal angle.

2
 (n)
Differential number flux The flux of molecules is denoted by ϕk , which
is defined as the number of molecules crossing a unit area per unit time in the
direction xk . Mathematically:

(n) # of molecules
ϕk =
time × area in direction xk
The differential flux of molecules traveling in the direction k is:
(n)
d(3) ϕk = n (Ci · îk ) f (Ci ) dVc
where:
• n is the number density,

• îk is a unit vector in the k-direction,

• f (Ci ) is the distribution function of molecules with velocity Ci ,
• dVc is the differential volume element in velocity space.

For example, the differential flux in the x2 -direction is:

(n)
d(3) ϕ2 = n C2 f (Ci ) dVc
(n)
Here, C2 is the x2 -component of the velocity vector Ci , and d(3) ϕ2 repre-
sents the differential flux of molecules in class Ci traveling in the x2 -direction.

Differential Fluxes of Momentum and Energy One can construct dif-

ferential fluxes of momentum and energy. The number flux and the energy flux
are both vectors, where the direction of the vector represents the direction in
which the transport occurs. Momentum transport occurs in a similar way and
can be represented using similar expressions.
Number flux and energy flux are scalar vectors, where the direction of the
vector indicates the direction of transport. Momentum flux, on the other hand,
is a tensor. One ”direction” is the momentum component, and the other is the
direction in which this vector quantity is being transported.

• Example 1 :
(ϵ)1
d(3) ϕ2 = mC 2 nC2 f (Ci ) dVc
2
This represents the differential kinetic energy flux in the x2 -direction being
transported by molecules of class Ci . Here, 21 mC 2 is the kinetic energy of
the molecules of class Ci , and C2 is the x2 -component of velocity Ci .
• Example 2 :
(p)
d(3) ϕ12 = mC1 nC2 f (Ci ) dVc
This represents the differential x1 -momentum flux in the x2 -direction,
transported by molecules of class Ci .

3
2 Molecular Collisions
2.1 Collision Process
A general collision between two molecules can be represented as follows:

(Ci , Zi ) −→ (Ci′ , Zi′ )

Here, Ci and Zi represent the initial states (i.e., velocities) of the two
molecules before the collision, and Ci′ , Zi′ represent their final states after the
collision. The equilibrium distribution is the velocity (or energy) distribution
of particles that does not change over time, despite collisions occurring contin-
uously. This occurs when the system has reached a state in which collisions
do not further alter the statistical distribution of velocities or energies in the
system. For instance, in a gas at equilibrium, the velocity distribution is given
by the Maxwell-Boltzmann distribution.

2.2 Definition of Cross Section

Consider a uniform beam of test particles (labeled A) incident on a field particle
(labeled 2), as illustrated schematically. Let nB denote the number density of
test particles in the beam, and g the relative velocity between the test particles
and the field particle. The flux of test particles, Γ1 , is defined as the number of
test particles crossing a unit area normal to the beam per unit time:

ΓB = nB g.
The total collision cross section σAB (g) between a test particle (A) and a
field particle (B) is defined by the ratio of the number of collisions per unit time
between the test particles and the field particle to the test particle flux:

Number of collisions with field particle per unit time

σAB (g) = .
ΓB
The dimensions of σAB (g) are:

[time−1 ]
[σAB ] = = [L2 ],
[ΓAB ]
indicating that the total cross section has units of area. Physically, the cross
section represents the effective geometrical ”blocking area” that the field par-
ticle presents to the beam of test particles. This quantity is a function of the
relative speed g and depends on the types of particles involved in the collision.
For hard-sphere interactions, the cross section is independent of the order of the
subscripts, meaning σAB (g) = σAB (g).

The cross section is often interpreted geometrically as the area around a field
particle within which a test particle must pass to result in a collision, and it

4
 Figure 2: Cartesian and Spherical Coordinates

is commonly used to compute collision rates and related properties in kinetic

theory.

2.3 Definition of Collision Rate

The collision rate quantifies how frequently collisions occur between two types
of molecules, A and B, within a given volume and time. The rate is given by:
Number of collisions
ZAB =
Time × Volume
This is an important quantity for calculating macroscopic properties like
reaction rates and transport coefficients (e.g., viscosity or thermal conductivity)
in gases.

2.4 Hard Sphere (HS) Model

To model molecular collisions, a straightforward approach is to treat molecules
as hard spheres, similar to billiard balls. In this model, the molecules of type A
and B have fixed radii, and a collision occurs when their centers come within a
certain distance of each other. In a real (quantum) world of nuclei surrounded by
swarms of electrons, things are not so clear-cut, and one must define probabilities
of particular outcomes given specific inputs.
Let’s define the parameters:

• mA , mB : masses of molecules A and B.

• dA , dB : diameters of molecules A and B.
• gi = Zi − Ci : the relative velocity of molecule B with respect to molecule
A before the collision.

We can imagine a sphere of radius dAB surrounding the center of molecule

A. If the trajectory of molecule B brings its center inside this sphere, a collision
will occur between A and B. The cross-sectional area of this sphere is the target

5
area for the collision, and if the relative velocity gi is directed inside this area,
a collision will occur. A collision occurs when the distance between the centers
of the molecules is less than the sum of their radii, dAB , given by:
dA + dB
dAB =
2
The collision cross-section, denoted σt , represents the effective area for a
collision to occur. For hard spheres, the total collision cross-section is:

σt = πd2AB
This is a simplified model that is useful for understanding basic concepts in
kinetic theory.

While the hard-sphere model gives valuable insights, it often fails to accu-
rately predict specific details, especially when dealing with real gases or plasmas
where interactions are more complex.

2.5 Variable Hard Sphere (VHS) Model

An improvement on the hard-sphere model is the Variable Hard Sphere (VHS)
model, which accounts for the fact that the collision cross-section depends on
the relative velocity of the molecules. In the VHS model, the cross-section is a
function of the relative speed g:

σt = σ(g)
This model is more accurate for aerodynamics and gas dynamics calcula-
tions, especially in non-equilibrium conditions, such as those encountered in hy-
personic flows or plasmas. For more complex cases, such as chemical reactions
or quantum mechanical effects, even more sophisticated models are required.

2.6 Elastic Collision Definition

An elastic collision is one in which the total kinetic energy of the system remains
constant. There is no loss of kinetic energy to promote changes in energy (such
as molecular vibrations or rotations), and momentum is conserved.
In an elastic collision between two molecules:

g′ = g

where g ′ is the relative speed after the collision, and g is the relative speed
before the collision. However, the relative velocity vector can change direction.
The scattering angles govern this change in direction.

6
2.7 Scattering Angles and Cross Sections
Two angles characterize the change in the direction of the relative velocity vec-
tor:
• χ: the polar angle.

• ϵ: the azimuthal angle.

These angles define the orientation of the velocity vector after the collision
gi′ , relative to its direction before the collision gi . The differential solid angle
dΩ̂, which describes the infinitesimal changes of the angles χ and ϵ, is given by:

dΩ̂ = sin χ dχ dϵ
dσ
The differential scattering cross-section dΩ represents the probability that a
collision will result in scattering into a particular solid angle dΩ̂:
dσ
dΩ
i.e., that gi′ has orientation (χ, ϵ) or within dχ, dϵ of (χ, ϵ)

Total Cross Section

The total cross-section σT integrates the differential cross section over all pos-
sible scattering directions:
Z Z 2π Z π
dσ dσ
σT = dΩ = sin χ dχ dϵ
Ω dΩ 0 0 dΩ
This quantity gives the overall likelihood of a collision occurring, regardless
of the outcome (elastic, inelastic, or reactive).
If the scattering is azimuthally symmetric (independent of ϵ):
Z π  
dσ
σT = 2π sin χ dχ
x=0 dΩ

If the scattering is totally isotropic (independent of both χ and ϵ):

 
dσ
σT = 4π
dΩ

(Isotropic scattering is commonly assumed in aerodynamics calculations us-

ing Monte Carlo methods but is not very realistic.)
Hard spheres are isotropic scatterers. Using σT = πd2AB , we get:

d2
 
dσ
= AB
dΩ HS 4

7
 Using the cross-section ideas, one can write down the differential collision
rate for molecules A of class Ci with molecules B of class Zi .
Consider first a single B molecule of class Zi . Then it is moving at speed
(velocity) gi relative to A molecules of class Ci and sweeps a volume:
 
dσ
dV = g dt dΩ̂
dΩ
in time dt.
The number of A molecules of class Ci in this volume is:

nA fA (Ci ) dVc
Thus, the differential collision frequency for this B molecule of class Zi is:
 
dσ
dZBA = nA fA (Ci ) dVc g dt dΩ̂
dΩ
But there are nB fB (Zi ) dVZ such B molecules per unit volume, so:
 
dσ
d(8) ZBA = d(8) ZAB = nA fA (Ci ) dVc nB fB (Zi ) dVZ g dΩ̂
dΩ
(A colliding with B is the same as B colliding with A.)
To get the macroscopic rate, one has to do the required integrations. One
can immediately integrate over all dΩ̂:

d(6) ZAB = nA fA (Ci ) dVc nB fB (Zi ) dVZ g σT

where we use:
Z
dσ
σT = dΩ̂
Ω dΩ
and note that σT = σT (g) in general.

3 Maxwellian Distribution
On depleting and replenishing rates. Replenishing rates are most easily handled
by using the concept of inverse collisions.

The inverse collision (Ci′ , Zi′ ) → (Ci , Zi ) is paired with the direct collision
(Ci , Zi ) → (Ci′ , Zi′ ), which is a depleting collision.

Then,
∂
[nA fA (Ci )dVc ]colls =
∂t
Z Z  
dσ
nA nB [fA (Ci′ )fB (Zi′ ) − fA (Ci )fB (Zi )] g dΩ̂ dVz dVc′
Vz Ω dΩ

8
 where we have used (without proof) that:
   
dσ dσ
dΩ̂ = dΩ̂inverse
dΩ dΩ direct
Recall that g ′ = g for elastic collisions.

dVz′ dVc′ = dVz dVc

Detailed balance requires that the equilibrium condition:
∂
[nA fA (Ci )dVc ]colls = 0
∂t
is satisfied not because the integral:
Z Z
[. . . ] dΩ̂ dVz = 0
Vz Ω

but because the integrand is zero. Thus:

fA (Ci′ )fB (Zi′ ) = fA (Ci )fB (Zi )

As shown in the text, this leads ultimately to the equilibrium (Maxwellian)
distribution:

mCi2
 m 3/2  
fM (Ci ) dVc = exp − dVc
2πkT 2kT
where m = mA or mB (as required) and Ci is a dummy variable (could be
Zi too!).

Maxwellian velocity distribution:

mCi2
 m 3/2  
fM (Ci ) dVc = exp − dVc
2πkT 2kT
Maxwellian speed distribution:

χM (C) dC

mC 2
Z  m 3/2  
χM (C) dC = fM (Ci ) dVc = 4πC 2 exp − dC
4π 2πkT 2kT

Showed in class that if a gas had two components A and B, each of which was
Maxwellian distributed with parameter T , i.e., each of which was at equilibrium
at temperature T , then for pairs A and B selected at random, the **relative
velocity distribution function** f (gi ) is also Maxwellian and is characterized by
temperature T and reduced mass:

9
 mA mB
m∗AB =
mA + mB
 ∗ 3/2
m∗ g 2
 
mAB
f (gi ) = fM (gi ) = exp − AB i dVg
2πkT 2kT
The relative speed distribution:
Z
χM (g)dg = fM (gi )dVg
orientations of gi
3/2
m∗AB m∗ g 2
  
= 4πg 2 exp − AB dg
2πkT 2kT
These results can be used to simplify the expression for d(6) ΘAB in the case
that fA (Ci ) and fB (Zi ) are Maxwellian:

d(6) ZAB = nA fA (Ci ) nB fB (Zi ) g σT (g) dVc dVz

The relative speed distribution:
Z
χM (g)dg = fM (gi )dVg
orientations of gi
3/2
m∗AB m∗ g 2
  
= 4πg 2 exp − AB dg
2πkT 2kT
These results can be used to simplify the expression for d(6) ZAB in the case
that fA (Ci ) and fB (Zi ) are Maxwellian:

d(6) ZAB = nA fA (Ci ) nB fB (Zi ) g σT (g) dVc dVz

Transform to the center of mass frame:
m∗AB
Ci = Gi − gi
mA
m∗AB
Zi = Gi + gi
mB
where gi = Zi − Ci and
mA Ci + mB Zi
Gi =
mA + mB
One gets finally, after integrating over all gi :
3/2
m∗AB m∗ g 2
  
dZAB = nA nB exp − AB 4πg 2 · gσT (g) dg
2πkT 2kT

= nA nB gσT (g) χM (g) dg

10
 One can now pose the question: What is the relative speed distribution of
molecules that **collide**? This should be contrasted to the earlier result for
the relative speed distribution for molecules chosen at random. The specifica-
tion that the molecules collide changes the result (in general) because one is
now answering a different question.

Let us call the relative speed distribution of colliding molecules χc (g).

Then by definition of a distribution function:

χc (g) dg ≡ fraction of molecules that collide with speeds in the range (g, g+dg)

This is defined as:

# of collisions per unit vol. with speeds in range (g, g + dg) dZAB
=
# of collisions per unit vol. ZAB
Thus,

nA nB gσT (g) χM (g) dg

χc (g) dg = R∞
nA nB 0 gσT (g) χM (g) dg
Simplifying:

gσT (g)χM (g) dg

χc (g) dg = R ∞
0
gσT (g) χM (g) dg
We see immediately that even if σT is constant, independent of g (which
corresponds to the hard sphere (HS) model), then:

gχM (g) dg
χHS
c (g) dg = R ∞
0
gχM (g) dg
where:
 1/2
8kT
ḡ =
πm∗AB
Thus, χHS HS
c (g) dg ̸= χM (g) dg. In other words, χc (g) ̸= χM (g).
We see immediately that even if σT = const, independent of g, corresponding
to the hard sphere (HS) model, then:

gχM (g) dg gχM (g) dg

χHS
c (g) dg = R ∞ =
0
gχM (g) dg ḡ
where:
 1/2
8kT
ḡ =
πm∗AB

11
 Thus, χHS HS
c (g) dg ̸= χM (g) dg, i.e., χc (g) ̸= χM (g).
The relative speed distribution function of colliding molecules is **not the
same** as the relative speed distribution of molecules chosen at random.

χM (g) (the relative speed distribution of molecules chosen at random).

Explicitly:
1/2  ∗ 3/2
m∗AB m∗
 2
mAB AB g
χHS
c (g) dg =g e− 2kT · 4πg 2 dg
8πkT 2πkT
   ∗ 2
1 mAB m∗
AB g
2

= e− 2kT g 3 dg
8 kT
Verify normalization:
∞ 2 Z ∞
1 m∗AB m∗
Z  2
AB g
χHS
c (g) dg = e− 2kT g 3 dg = I3 (ν)
0 2 kT 0
 ∗ 2  2
1 mAB 1 2kT
= ·
2 kT 2 m∗AB

=1 as required.
K
For the special case σT (g) = g where K = σ0 gref .
K
g χM (g)dg χM (g)dg
then χc (g)dg = R ∞ K = R∞ = χM (g)dg.
0 g χM (g)dg 0
χM (g)dg
Only in this case χc (g)dg = χM (g)dg. Such molecules are called ”Maxwellian”
molecules. Use of σT (g) = K g simplifies many calculations, and Maxwell made
extensive use of this model. It may be shown [Chap IX, Sec 8, 9] that to have
σT (g) = Kg , molecules must have an inverse power repulsive potential V (r) that
varies as
1
V (r) ∼
.
r4
Hard sphere molecules behave (in a certain limit) as though
1
V (r) ∼ , α → ∞.
rα
Looking at the behavior of real gases, it is found that the hard sphere model
is too ”hard,” whereas the Maxwell molecules are too ”soft.” This statement is
based on the temperature dependence of the coefficient of viscosity (µ).
Solutions of the Boltzmann equation for inverse power model cross-sections
(see chap IX) show that

12
   21 + α2
T
µ = µref .
Tref
1
Where V (r) ∼ rα . Thus for hard spheres (α → ∞):
 1/2
T
µ = µref ,
Tref
whereas for Maxwell molecules (α = 4):
 
T
µ = µref .
Tref
The temperature dependence of µ for N2 (for example) suggests α = 9 − 11.
Note, however, that an inverse-power model for N2 -N2 interactions is by no
means a perfect model, though it works for many aerodynamic calculations.

13
 For the more general case:
 −ω
g
σT (g) = σref
gref
This is the Variable Hard Sphere (VHS) model. The negative sign is usually
explicit.
 −ω
g
gref χM (g) dg
g
χc (g) dg = R  −ω
∞ g
0
g gref χM (g) dg

g 1−ω χM (g) dg g 3−ω

= R ∞ 1−ω = R∞ m∗ g 2
0
g χM (g) dg g 3−ω e− 2kT dg
0
Using the result for the gamma function from the handout:
Z ∞  
n −βv 2 1 n+1
v e dv = n+1 Γ
0 2β 2 2
(complete Γ function rather than incomplete because the lower limit of in-
tegration is 0).
So:
Z ∞
m∗ g 2
g 3−ω e− 2kT dg
0
m∗
has n = 3 − ω and β = 2kT .
 
1 4−ω
=
4−ω Γ
2 m∗ 2 2
2kT
Finally:
 4−ω
m∗
 2
1 m∗ g 2
χc (g) dg = 2 4−ω

g 3−ω e− 2kT dg
2kT Γ 2

14
 Change of variable.
Ex. Kinetic energy distribution function.
Let ϵ be KE (Kinetic Energy), and F (ϵ)dϵ = fraction of molecules with KE
in range (ϵ, ϵ + dϵ).
Want an explicit expression for F (ϵ)dϵ.
Now ϵ = 21 mC 2 , so we could (or should!) use the speed distribution function:
 m 3/2 mC 2
χ(C)dC = e− 2kT dC · 4πC 2
2πkT
 1/2
1 2ϵ
ϵ = mC 2 ⇒ C=
2 m
dϵ
dϵ = mCdC ⇒ dC =
mC
m

3/2 − ϵ dϵ 2ϵ 1/2

2ϵ


Fraction in range (C, C + dC) = 2πkT e kT m m · 4π m

 m 1/2 2  2ϵ 1/2 ϵ
= · e− kT dϵ
2πkT m m
 ϵ 1/2 ϵ dϵ
=2 e− kT
πkT kT
By definition, this is:
 
1 ϵ 1/2 − ϵ dϵ
F (ϵ)dϵ = 2 √ e kT
π kT kT
Check normalization:
Z ∞ Z ∞
2 ϵ 1/2 − ϵ dϵ
F (ϵ)dϵ = √ e kT
0 π 0 kT kT
Check normalization:
Z ∞ Z ∞
2 ϵ 1/2 −ϵ/kT dϵ
F (ϵ)dϵ = √ e
0 π 0 kT kT
ϵ dϵ
Let kT = x, so dx = kT .
Z ∞ Z ∞ Z ∞
2 2
F (ϵ)dϵ = √ x1/2 e−x dx = √ x3/2−1 e−x dx
0 π 0 π 0
   
2 3 2 1 1
=√ Γ =√ · Γ
π 2 π 2 2
2 1 √
= √ · · π=1 as required.
π 2

15