AE/ME5102 Advanced Gas Dynamics: Notes1 Set 3

Instructor: J. Blandino
Worcester Polytechnic Institute, Fall 2015

Phase Space
At a time t, a particles position is specified by a position vector in Cartesian coordinates:

~r = x1i + x2j + x3 k
These coordinates (x1 , x2 , x3 ) define the particles location in physical space. The
corresponding velocity vector

~ = d~r = C1i + C2j + C3 k

C
dt
has coordinates (C1 , C2 , C3 ) which define the position in velocity space. The dynamic state of the particle is specified by the position and velocity, which collectively
define the particles location in phase space, with six coordinates (x1 , x2 , x3 , C1 , C2 , C3 ).
It will be useful to define a differential volume element in physical space (using the
notation from Ref [1]):

dVx = dx1 dx2 dx3

So a particle located at a position ~r, will reside in a volume element of physical
space, with an x coordinate that lies between x1 and x1 + dx1 , a y coordinate that
lies between x2 and x2 + dx2 , and a z coordinate that lies between x3 and x3 + dx3
1

Notes are based on material from the course text, Ref. 1: Introduction to Physical Gas
Dynamics, Vincenti, W. and Kruger, C., Krieger Pub., Copyright 1965. Any figures used from
Ref.1 are so noted and copyrighted by Krieger Publishing Co.

Figure 1: Volume element in physcial space [1]

Similarly, we can define a volume element in velocity space as:

dVC = dC1 dC2 dC3

Figure 2: Volume element in velocity space [1]

A volume element in phase space can therefore be written as:

dVx dVC = dx1 dx2 dx3 dC1 dC2 dC3

So particles in dVx dVC are a subset of those in dVx that have an x component of
velocity between C1 and C1 + dC1 etc.

Distribution Function
General Considerations
~ t) represents a phase space density. So if we define
The distribution function F 0 (~r, C,
~ t) as the number of particles with coordinates ~r, C
~ in a volume element
dNm (~r, C,
dVx dVC of phase space, then

~ t) = F 0 (~r, C,
~ t)dVx dVC
dNm (~r, C,
If F 0 is a function of ~r inhomogeneous (i.e. spatially non-uniform with a particle
number density n = n(~r)). Note: shortly we will see how the number density and
distribution function are related.
If F 0 is independent of ~r homogeneous (i.e. spatially uniform with n = n0 =
constant)
~ t) F (C,
~ t) and is referred to as a velocity
If F 0 is homogeneous, then F 0 (~r, C,
~
distribution function. Often, the t dependence is implied and we just write F (C).
~
For this case, dNm = F (C)dV
C represents the number of particles with a velocity
~
vector C in a volume element dVC .
Normalized Distribution Function
It is often more convenient to work with a normalized distribution function. We can
define this as

f (Ci ) =

F (Ci )
Nm

(1)

~
where we have written the velocity vector using indicial notation, i.e. Ci = C.
Also, note that Nm is the total number of particles, with all possible positions and
velocities in the region or system of interest.
With this definition, the number of particles in a differential element of velocity
space will be given by

dNm = Nm f (Ci )dVC

Since all the particles will have some velocity, if we integrate over all possible velocities, we have to recover the
number of particles present:
Z

Nm f (Ci )dVC = Nm

Note that since dVC = dC1 dC2 dC3 , the integral above is
actually three integrals, one over each velocity component.
Also, since Nm is just a constant, then the normalized
distribution function has the important property that
Z

f (Ci )dVC = 1

Mean Value of Particle Properties

We will see later that many properties of the gas are functions of the particle velocity
(e.g. energy and momentum fluxes). If Q(Ci ) represents some property which a
function of the particle velocity, then we can define the average or mean value of
this property using the normalized distribution function:

R
Q=

Nm

QdNm
=

Nm

Q(Ci )Nm f (Ci )dVC

Nm

Q(Ci )f (Ci )dVC

Q=

Relation to Number Density n

First, lets consider the case where the spatial distribution is nonuniform. The
number of particles in a volume element of phase space is then

dNm (ri , Ci ) = F 0 (ri , Ci )dVx dVC

where again, the t dependence of F 0 is not shown explicitly.

We can then write for the number density

dn(ri , Ci ) =

dNm (ri , Ci)

= F 0 (ri , Ci )dVC
dVx

Integrating over all possible velocities give the particle

number density at the particular location

F 0 (ri , Ci )dVC

n(ri ) =

If n(ri ) in turn, was integrated over all of the (physical)

volume V , then we would recover the total number of
particles present.

Z
Nm =

ZZ

F 0 (ri , Ci )dVx dVC

n(ri )dVx =
V

Now lets consider the case where the particle distribution is spatially uniform
n(ri ) = n0 .

Recall from Eq. 1

f (Ci ) =

F (Ci )
F (Ci )
=
Nm
n0 V

and the total number of particles in the system Nm can

be recovered from the normalized velocity distribution
f (Ci )

Z
Nm = n0 V = V

n0 f (Ci )dVC

In this course, we will generally assume that the particle

distributions are spatially uniform.

Maxwellian Velocity Distribution

The Maxwellian velocity distribution function is of particular importance because,
as we will see, it describes the velocity distribution for a gas in equlibrium.
The derivation of the Maxwellian velocity distribution is presented in detail in Ref
[1], Chapter 2, Sections 4 and 5. We will first consider an overview of the derivation
and then consider some of the new concepts and terminology introduced. The
detailed steps are covered as part of the reading assignment from Ref [1].
Outline of Derivation
1. Some new concepts and terms to be introduced:

- detailed balancing
- velocity class for a molecule
- depletion and replenishment collisions
2. Presentation of geometry used to describe collisions
- relative velocities
- sphere of influence
- line of centers

3. For a collision between two classes of molecules, write an expression for the
following:
- differential element of volume a molecule would have
to occupy to be involved in a collision with a molecule
of a different class
- frequency of collisions between molecules of two velocity classes
- net collision rate of molecules into a class
4. In equilibrium, the net rate at which molecules enter or leave a class is zero.
This requirement is imposed as a condition on f (Ci )
Principle of Detailed Balancing

The state of a gas is fully known when n and f (Ci ) are known.
Repeated collisions produce continual changes in f (Ci ).

After a sufficient time, if undisturbed, the system of gas molecules reaches a steadystate such that n and f remain statistically the same at every point for all time.
Analog birth and death rates do not alter the statistical age distribution of a large population.
This state of affairs is referred to as equilibrium

So we can state the problem we wish to solve as follows: For a population of billiard ball like, perfectly elastic molecules, undergoing continuous collisions, find a
relation that must be satisfied by the distribution function when the system is in
equilibrium.
To express this relation mathematically, we must first define a few more terms:
A molecule is said to be of a particular class, Ui if its
velocity lies in the range

U1 to U1 + dU1
U2 to U2 + dU2
U3 to U3 + dU3

i.e. the velocity lies in a velocity space volume element

dVU .

Similarly, a molecule is of class Yi if its velocity lies in

the range

Y1 to Y1 + dY1
Y2 to Y2 + dY2
Y3 to Y3 + dY3
and the velocity lies in a velocity space volume element
dVY .
A depletion collision for class Ui will remove a molecule
from the volume of velocity space dVU that defines the
class.
A replenishment collision for class Ui will add a molecule
to the volume of velocity space dVU that defines the
class.
In equilibrium, the frequency of depletion collisions is
balanced by the frequency of replenishment collisions.

Collision Geometry
We define a vector ~g to represent the relative velocity of molecules with velocities
Ui and Yi

The magnitude of this relative velocity can be expressed as

1

g = [(Y1 U1 )2 + (Y2 U2 )2 + (Y3 U3 )2 ] 2

10

As before, we can define a sphere of influence centered on molecule A

~ is a vector conThe line of centers L
necting the centers of molecule A and
B, originating in A
For now, we assume the molecules all
have the same diameter d. Molecule A
will undergo a collision whenever the
center of molecule B comes within a
distance d of A
The collision geometry is described in terms of the angles and velocities shown in
Figure 3. Note that in this figure, the sphere of influence is centered about the
molecule of class Ui .

Figure 3: Geometry used to describe collision between two molecules with arbitrary
relative velocity [1]

11

The overall strategy is as follows:

Step 1: find an expression for the frequency of collisions

between molecules of two classes, lying in a specified
range of angles and velocity space
Step 2: use the principle of detailed balancing to impose
a condition which the distribution function for the gas
in equilibrium must satisfy
Step 3: use this imposed condition to find the correct
mathematical form for the distribution function of a gas
in equilibrium.

We will discuss the Step 1 in lecture; you will review Steps 2 and 3 in the reading.
We refer to the angles shown in Figure 3 and consider a collision between a molecule
~ lies between and  to +
of class Ui and Yi , such that the line of centers L
d and  + d
This range of angles determines a differential area element d on the sphere of
influence

d = dsin d d d
d = d2 sindd

12

In a time interval dt, a collision will occur for any molecule of class Yi with its
center located in a differential volume

diff vol = d2 sincosddgdt

This is the volume of a slant cylinder

(shown in Figure 3).

Now we want to know, how many molecules of class Ui are present in a unit volume?
These will be the potential targets of the molecules of class Yi .
Using our previously defined, normalized, velocity distribution function f (we dont
know the mathematical form of it yet of course, that is what we are trying to
determine!), the number of molecules of class Ui that are present in a unit volume
can be expressed as

nf (Ui )dVU where dVU = dU1 dU2 dU3

And the total volume of all the slant cylinders (one for each of the above molecules)
per unit volume is

nd2 f (Ui )gsincosdddVU dt

How many molecules of class Yi have their centers within this collection of slant
cylinders?

# of molec. of class Yi per unit volume total vol. of slant cylinders per unit vol.
13

(nf (Yi )dVY ) (nd2 f (Ui )gsincosdddVU dt)

This expression gives the number of collisions occurring in a time dt per unit volume.
To get the rate, we divide the above by dt (and re-arrange) to obtain the following:

n2 d2 f (Ui )f (Yi )gsincosdddVU dVY

which has units of collisions per unit volume per unit time.
This expression concludes Step 1 of the three steps we listed earlier.
It is worth pausing to think about what the above result represents. It gives us
an expression, using the geometry we defined (Figure 3), for the number of collisions
between molecules in class Ui and class Yi , per unit time, per unit volume. But it
is restricted to only those molecules who lie in a small range of velocity specified
by dVU and dVY , and only those molecules whose line of centers lie within a narrow
range of angles d and d.
In the derivation of the Maxwellian distribution for a gas in equilibrium, this result
is used to find an expression for the net rate of collisions into a particular class.
Remember that the Principle of Detailed Balancing tells us that for a gas in equilibrium, this net rate will be zero.
The details presented in the reading assignment involve applying the constraint
imposed by the Principle of Detailed Balancing (Step 2), and then working out the
exact mathematical form of the velocity distribution function for a gas in equilibrium
(Step 3). This function, referred to as the Maxwellian distribution, is found to have
the following form:
f (Ci ) =

h m
i
 m  32
exp
(C12 + C22 + C32 )
2kT
2kT

14