CHAPTER 5

STATISTICAL MECHANICS

5.1 INTRODUCTION

of concern systems
Most the problems of condensed-matter physics having
large number of physically identical constituents--electrons, atoms, mass particles,
a

harmonic oscillators, or other basic entities. In such problems, it is hopeless to try to

trace the motion of each of these individual constituents. Instead, you must settle
for knowing averages of relevant dynamical quantities over the entire range of pos-
sible system configurations. The macroscopic properties of the system can then be
understood in terms of these average values. The questions that must be resolved
are how to determine the probability that a given system configuration will occur,
and what configuration is of maximum probability. Also, it is important to
system
understand how energy, velocity, and momentum are distributed among the sys-
tem's particles or other constituents.
is referred to statistical
as
The area of physics that addresses these questions
asserts that randomness is
mechanics or statistical physics. Since quantum theory
mechanics and statistical phys-
inherent feature of microscopic physics, quantum
an the macroscopic behavior of systems
ics are intimately related. On the other hand,
methods can be related to the
internal dynamics is expressed by probabilistic
whose and therefore statistical physics is also
thermodynamic properties of the system, it can be thought of in
associated with classical thermodynamics. Indeed,
closely the microworld of quantum theory and
the
Some sense as a connecting link between
domain classical mechanics and thermodynamics.
of
macroscopic

CONFIGURATIONS
5.2 PHASE SPACE AND SYSTEM
describe the
this subject one must initially about how to
think or
To approach unrestricted by any constraint
motion is completely
a1or of particles whose motion. All you can say classically
in these circum-
of
any preassigned equation initial position
coõrdinates and
momentum

Ces 1s that if you know a particle's

components, you can precisely determine its motion at all later times. The position
coördinates and momentum components can theretore be thought of as quantities
fundamental to a description of system dynamics. T particles are free from any and
all constraints, what can you say about these quantities without any more detailed
specification of the system and its dynamics? Well, only that any given set of posi-
tion coördinates is as likely as any other, and that all conceivable sets of momentum
(or velocity) components are of equal probability. One says, more precisely, that the
a priori probability of all possible states of motion--as defined by position coördinates
and momentum components--are the same. The words a priori in this context
mean "in the absence of any dynamical constraints on the system or forces acting on
its particles". This is the fundamental postulate of statistical mechanics.
This postulate suggests that very high particle velocities and kinetic energies
are inherently as likely as those that are quite low. This seems counterintuitive, but
of any constraint whatever. There
you must remember that it refers to a system free
is no "a priori" requirement that the system have a finite total energy, for example,
or that it be a closed system having a finite mass and a finite and constant number of
particles. So, under these circumstances, this odd conclusion may seem less strange.
When constraints such as these are introduced, they change the probabilities.
iated with possible configurations of the system, so that they are no longer unitorm.
For example, when one imposes a requirement that the system have a fixed number
of particles and a constant total energy, the effect is to change the probabilities so that
low energy configurations are increased relative to those of very high energy. In
such a system, it is clear that no single particle can have more energy than the fixed
total energy assigned to the system as a whole. The probability of system configur-
ations in which any particle possesses more than this preassigned energy is therefore
zero. Clearly, probabilities have changed not in respose to the fundamental statis-
tical principle stated above, but in response to constraints that have been introduced.
The probabilities will change, in fact, precisely so that these constraints are satisfied.
It is convenient to describe the possible system configurations as points in an
orthogonal multi-dimensional coördinate space referred to as phase space. In the
simplest example, that of a one-dimensional system in which particles move only
along the x-direction, the motion of a particle is completely determined when the
quantities x and p, are given as functions of time. The dynamical state of the system
at any time is thus described by the coördinates of a point in a phase plane whose co-
ordinate axes are x and p, and the motion of the particle can be completely specified
by a curve in the phase plane whose points are indexed by these coördinates. For
example, if the system is composed of independent (i.e., uncoupled) one-dimension-
al harmonic oscillators of mass m and frequency o, we can write such an equation or
motion (using energ8y conservation) as

Pz4mo*
2
(5.2-1)
where e is the constant total energy. The simple harmonic motion is represented in
the phase plane as an
ellipse, since the above equation, regarded as a
relationshipP
between the relevant-or, in the jargon of theoretical physics, "canonical"--variables,
has the form associated with an ellipse whose axes coincide with those of the coórd-
inate system. Indeed, if you put the above equation into standard form by dividing
both sides by e, the semi-axes a and b of the ellipse can be expressed by

a 2e and b= 2me (5.2-2)

mw
This situation is illustrated in Figure 5.1. Statistical mechanics postulates that
in the absence of constraints all states of motion have equal probability. The a priori
probability that a point describing the state of the system will be in any area element
ApAr is therefore the same for all area elements in the phase plane. The number of
possible system configurations in a given region of the phase plane is likewise di-
rectly proportional to the area of that region.

V2me
2e/mo?

|Energy e

Area AA
Energy e+Ae

one-dimensional harmonic oscillator of constant energy e.

Figure 5.1 Phase space trajectory for a oscillator of somewhat larger energy. The area between the
A slightly larger trajectory is shown for an
two curves is proportional to the
number of possible states of the system in this energy interval.

of course, the equation of motion (5.1 be viewed as

1) must
In this
a constraint example,
restricting the motion to an elliptical path, around which the phase
physically, is imposed by forces that
point circulates as time elapses. This constraint, oscillator rather than a free particle.
act on the particle and cause it to behave
as an
In the phase plane, it confines the particle to system contigurations described by
are not of probability.
points on the ellipse. Now, all system configurationslie on the equalin which case
Indeed, such states have probability zero unless they curve,
determined how fast the phase point travels along the ellipse.
by
their probability is of constant total energy E, however, the ellipse can be of
constraint
l you relax the area element with equal probability. The
arbitrary size, and the particle can be in any
between the two elliptical curves shown in
probability that it will be in the region
Figure 5.1 is now directly proportional to the phase plane area enclosed between
in three dimensions, three posi-
them. For systems whose constituents can move the state of
tion coördinates and three momentum components are needed to define
whose points are
motion, and the phase space is a six-dimensional orthogonal space
190

indexed by coördinates (x, y, z, Pz, Py P). This restricts our

ability to visualize it
geometry, but in no way limits its usefulness as a tool for the further developmens
of statistical mechanics. The basic postulate of statistical mechanics ent
can now
understood as stating that every volume element of this six-dimensional space ha be
in the absence of any constraint, equal probability for any constituent of the svsfes

5.3 THE DISTRIBUTION FUNCTION AND THE DENSITY OF STATES

In order to calculate the number of persons in the USA under 30 years of

ge,
you need to know the age distribution of the population--that is, what fraction of the
entire population is of age 0 to 1 years, 1 to 2,2 to 3,..nto n +1, etc. Likewise, to
determine most of the macroscopic properties of particle ensembles and similar
systems, one must know how particles or other constituents are distributed, either
in phase space or in
energy. Generally speaking, it is simplest to determine averages
over energy, so we shall discuss this case first.
It either the classical viewpoint of
is possible treat this subject
to
from
cles having continuously distributed energies and momenta, or from a quantum
parti
perspective, in which there may be discrete quantum states and energy eigenvalues.
There are significant learning advantages in choosing the quantum formulation
from the beginning, and we shall therefore pursue this course. In the limit of mac
roscopic systems, however, quantum states and energy levels are so dense and close
y spaced in energy and phase space that they can beregarded as a quasi-continuum.
Tt is also important, therefore, to be able to express our results in this limit where
quantum and classical statistical physics coinide.
From the quantum perspective, the fundamental postulate of statistical me
chanics states that in the absence of any constraint, all quantum states of the system
can be occupied with equal probability. The words "quantum states" are emphasized
because it is important to remember that quantum states and energy levels are not
the same when degeneracy is present. Under these circumstances there can be
many
different independent solutions of the wave equation, each defining a distinct quan-
tum state with a unique
probability density, all belonging to the same energy eigen-
value. It is the quantum states, not the energy levels, that are
occupied with equal
priori probability.
As mentioned previously, when the usual constraints are introduced, this a
priori probability will be modified in some as yet unspecified fashion. If we now de-
fine a quantity fle) as the average number of
state at energy e, and if we let
particles occupying a single quantum
g(e)de be the number of quantum states of the system
in an energy range de about
energy e, the number N(E)de of particles (or other
constituents) in this energy range will be the product of these two factors, or

N(Ede = fle)g(e)de
(5.3-1)
The function fle) as defined above is referred to as the distribution function. The
distribution function expresses the statistical
average of the number of particles in a
quantum state of given energy. For the present, the Pauli exclusion principle wil
not be imposed, so this number can be
very small, of the order of unity, or very
 19
large. One of our objectives is to calculate this
ity. We shall find that it is function, using the
laws of probabil
essentially
of particles or other constituents that
a statistical
quantity, independent of the kind
these entities, and of system
comprise the system, of interactions between
dynamics
the other hand, is a reflection of how
in general. The density of states function, on
energy levels are spaced, and of how manyY
quantum states are allocated to them. These quantities are determined ultimately
from Schrödinger's equation and the
the dynamical properties
boundary conditions, and thus depend only on
of the system. The
quantity, and is independent of density of states is therefore a dynamical
any probabilistic phenomena.
Equation (5.3 - 1) is written in the context of the classical limit, where there is
a
quasi-continuum of energies, in
which, though energy levels are discrete, there are
always many quantum states in any experimentally resolvable energy range de. A
similar equation can be written, however, for the case of discrete and
rated energy levels. In this case, we consider the n-th widely sepa-
which there belong gn distinct quantum states; the number
level, of energy
energygn is referred En, to
to as the
degeneracy factor of the level. The average number of particles N, occupying this
level can now be written (recalling the definition of the distribution function) as

N, =8f(e,), (5.3-2)
It is clear from the two above
expressions that the degeneracy factor and the density
of states function are essentially similar
quantities.
If we know the distribution function and the
density
of states or the
degener-
acy factor, we can calculate the average of any energy-dependent parameter , as

EeNede
(6) - Ne)de, (5.3-3)
NE)de
where N, the total number of particles or other constituents in the system, can be
expressed as the integral of Ne) over the range of energies accessible to system con-
stituents. This is simply the standard procedure for averaging over a dístribution;
you take an energy range, figure the number of particles in it, multiply that number
by the value of f at that energy, sum over energies, and divide by the total number
of particles. It differs nevertheless from the prescription for
finding an expectation
value in quantum theory; in the equation above, f is presumably already an expec-
tation value. A similar calculation is appropriate for widely spaced discrete energy
levels, but in this case the integral can be represented as a discrete sum over the
allowed energies, so that

()- (5.3-4)

where N, is given by (5.3- 2).

 It is also possible to compute averages Such as these by integrating or sum.
ming over the phase space coördinates. We have mentioned in the preceding sec
tion that the possible states of motion--defined by
the possible values of the e
and momentum components--are uniformly distributed in phase space, Acc
to quantum theory, allowed quantum states are also uniformly distributed in nhnase
space, for essentially the same reasons. This means that the density of states factor i
s
constant in phase space. The formula for averaging over phase space coördinates is
thus written

Ie_JJ SP.a)SP.a)dpd
(5)= (5.3-5)
SP,4)dpda
where f is the distribution function expressed in terms of phase space coördinates.
and p. 4, dp, and dq merely abbreviations for (p,, Py» P), (x, y, z), (dp», dpy, dp:), and
(dx, dy, dz), respectively. The density of states factor, being constant in both numer-
ator and denominator, cancels. Several examples of these averaging procedures will
be presented in subsequent sections.
We must now turn our attention to the problem of calculating the distribu-
tion function and the density of states function. We shall start with the density of
states, since it is simpler, and show how it is obtained for a system of independent
free particles. As indicated previously, the density of states can be calculated from
the solutions of Schrödinger's equation under appropriate boundary conditions.
We shall assume that our "free" particles are confined by a rectangular container
whose sides extend from the origin to distances Lz, Ly and L, along the respective
axes, the container being treated simply as a three-dimensional potential well of
infinite depth. Under these circumstances, within the well, the potential energyis
zero, which allows us to write Schrödinger's equation for any particle insideas

V+ky =0, (5.3-6)

where
=Zme (5.3-7)

The solutions to this equation are, as one can verify by direct substitution,

y(x,y,2) = Ae***"e*= Aeir)

(5.3-8)

where k,k, behave

tion
as components of a propagation vectork, and satisfy the rela-
2me
k+k+k =kk= (5.3-9)

As in the mathematically similar case of acoustic waves in crystal lattices, We

shall require these solutions to obey periodic boundary conditions, in which values
of the wave function on
opposite parallel faces of the enclosure are equal at all
times. This is not
strictly proper boundary condition for a wave function i n a Tg
a

rectangular enclosure; we should actually require the wave function to be zero on

 all boundaries to preserve its continuity. As in the case of sound waves, however,
imposition of periodic boundary conditions suggests that a "topological transtor-
the
mation" has been made, in which all opposite faces are placed in contact
wth one
another. In one dimension, we have seen that this(ampounts to bending a line into
a ring. In a two-dimensional
system, it can be accomplished by bending and joining
opposite sides ot a square to form first a cylinder, and subsequently bending and
stretching the free edges until a toroid is formed. In three dimensions, ie's geomet-
rically impossible, though we shall use its consequences anyway. We have seen in
Chapter 3 that such transformations have essentially no effect on the number of
allowed states or on their disposition in frequency or energy. For this reason, the use
of these boundary conditions is standard practice in solid-state theory; we shall use
them not only here, but elsewhere, as a matter of course. In this case, we shall set

y(0,y,z) = y(L,y,z)
vx,0,2) = Y(x,L,z) , (5.3-10)
and y(x,y,0) = y(x,y,L,) .

If this set of boundary conditions is substituted into (5.3 8), it will be seen that they
can be satisfied only if k, ky, andk, have certain sets of discrete values, given by

27Tn (n, = 0,t1,#2,...)

k,2 n,/L, (n, =0,+t1,t2,...), (5.3-11)

2Tn./L, (n, = 0,t1,t2,...)

Since p and k are related by de Broglie's relation, this equation in effect defines a set
of quantum states associated with the momentum eigenvalues,

(n, = 0,tl,+2,...),
Phn,/L
(n, = 0,t1,t2,...), (5.3-12)
P, hnL (n, = 0,tl1,+2,...)
Phn,/L,
It will be observed that, as mentioned previously, these quantum states are uni-
formly distributed throughout phase space. There is also a set of energy eigenvalues,
which we can obtain from (5.3 -9) with the aid of (5.3 - 11); they can be written as

En,n,)=: (5.3-13)

Using de Broglie's relation, we can write (5.3 9) as an energy-momentum re-

lation of the form

p++p= 2me. (5.3-14)

In this three-dimensional problem, phase space has six coördinates, x, y, z, Px Py» P:
Since the above equations are independent of the coördinates x, y, and z, we can plot
Our results in a
three-dimensional orthogonal momentum space (p,, Pu P) without
oss of
generality. In such a space, the above equation defines a spherical surface of
radius p or (2me)1/2 for a free particle of constant energy e, as illustrated in Figure 5.2.
For a slightly greater energy e + de, corresponding to a larger momentum p + dp, the
constant-energy sphere will increase in radius by an amount dp, as shown. From
(5.2 14), it is clear that the relationship between dp and de is,

pap = m de
(5.3-15)
The allowed momentum eigenvalues given by (5.3 12) can be plotted in this
diagram as a rectangular orthogonal lattice of points corresponding to all possible
integral values of the quantum numbers n,, ny, and n,. In such a lattice, the "unit
cell is a rectangular volume as illustrated in the figure. Each side of the cell corre
sponds to a unit difference between quantum numbers in the respective coördinate
directions. The lengths of these sides are thus as labeled in the diagram, according to
(6.3- 12). Note that the side lengths involve Planck's constant--a very small number
on the scale of macroscopic units-divided by a macroscopic length. The cells are
therefore very small for macroscopic systems, far smaller than any experimentally
resolvable energy difference. For this reason it is possible to determine how many
quantum states there are in any energy range de--which is, by definition, the density
P
Energy e+ de
Momentump+ap

Energy&
|hlL
Momentum p

hiL
hlLy
P

Figure 5.2 Spherical surfaces of constant energy in momentum space. The allowed momentum states
are arranged in a rectangular orthogonal lattice of closely spaced points whose "unit cel" is shown at
the right. In reality, this cell is far smaller than any experimentally resolvable energy difference de.
The
sides of theunit cells are as labeled in this diagram. It must be noted that the black dots at the
ners of this cell do not represent atoms, but simply allowed momentum eigenvalues. The free particles
cor
themselves are distributed randomly in real space throughout the volume of the container. Only a
single octant of the spherical surfaces is illustrated for simplicity, but in reality the surfaces are com-
plete spheres.

of states--simply by calculating the phase space volume between the two spherical
shells shown in the diagram and dividing by the volume allotted to each quantum
 195

ctate as given by the volume of the above unit

cell. The volume of momentum
space dV, enclosed by the shells is

dV, = 4mpdp = 4 p mde = 4Tmv2me de

(5.3-16)
Tf we divide this by the cell volume
h3/(L,L,L,) h/V, where V is the volume of the
=

container, we obtain the quantity previously defined as gle)de, whose form in this
case is

se)de= 42
h3 e
de (5.3-17)

Note that the container localizes particles within a region whose dimensions
in coördinate space are (Ax, Ay, Az) (Lx, Ly, L,). The volume of momentum space
per quantum state, however, is Ap,ApAp, = h/(L,L,L) This can now be recognized
as a relationship stemming from the Heisenberg uncertainty principle, of the form

(Ap, Ax)Ap, Ay(Ap, Az) = * (5.3-18)

The left side of this expression can be interpreted as the volume of phase space per
quantum state, and the uncertainty principle requires it to be constant, in this case
equal to h3. The Heisenberg principle is obeyed, of course, irrespective of the dynam
ical properties of the system, or of the mechanisms that localize its constituents in
coördinate or momentum space. Its effect is to guarantee that quantum states are
distributed uniformly in phase space, in the sense that the same phase space volume
is associated with each quantum state.
If the particles, like electrons, have spin s =1/2, the quantum states will have
spin degeneracy, in which two states of opposite spin (m, = t1/2) appear for every
quantum state identified in the absence of spin. The density of states is now multi
plied by a factor of two, the result being

5y2zV/2
&le)de =

h de. (5.3-19)

The above derivation assumes a rectilinear container, but the fact that only the total
volume aPpears in the density of states suggests that gle) is independent of the shape
of the enclosure, a conjecture verifiable by detailed calculation. If the particles exert
torces on one another, the eigenfunctions and energy eigenvalues are altered, which
in turn changes the density of states function. If the system constituents are entities
such as harmonic oscillators) having dynamical properties different from those of
free particles, the eigenfunctions and energy eigenvalues will also differ, leading to
altered density of states, even when the constituents are totally independent of
One another. In all cases, however, the density of states function can be calculated by
a
procedure similar to the
given above.
one
In condensed-matter physics, the harmonic oscillator plays an important part
4many different situations. It is thus important to know the density of states func-
on for an assembly of harmonic oscillators. For simplicity, we assume initially that
e r e 1s no coupling between oscillators, thus that they are entirely independent of
one another. For a system of one-dimernsional oscillators, the energy levels are
states in an interval
equally spaced and non-degenerate, so the number of quantum
de is proportional to de so long as de is much larger than the spacing ha between
levels. In fact. we may conclude from this that g(e)de must have the value de/ hw.
Alternatively, since all states are allotted equal amounts of phase space, you
can find the shaded area between the two constant energy contours in Figure 5.1,
and divide by the area per quantum state to find the number of states in de. In this
one-dimensional system, the Heisenberg uncertainty principle has the form--in
than h3, Since the
analogy to (5.3 18), AxAp, =h, so one must divide by h rather
area of an ellipse is ab, where a and b are the semi-axes, from (5.2 2), we can Write

27te
A mab E
T 2me = : (5.3-20)
Then,
21 de, dA2 de =
dA de
de
and
h ha hw

and 8(e) de = E (5.3-21)

h

For a system of three-dimensional oscillators, the above method works, but

of six-dimensional figures in
requires the calculation of areas and included volumes
a direct extension of the method1
phase space. To avoid this mind-boggling exercise,
so useful for free particles can be used. First, we must note that the potential energy
of an isotropic harmonic oscillator in three dimensions can be written as

Vx,y,z) = mo'(r*+y+z)= }mo? (5.3-22)

The Hamiltonian operator for this system thus has the form

H+Vay,2)= 2m oztmo'

ma- = H,(x)+H,()+H,(7) (5.3-23)

This Hamiltonian splits nicely into three parts, each of which has the form of
a one-
dimensional oscillator Hamiltonian depending on one and only one coördinate.
Each of the parts, therefore, has eigenfunctions V» Vy, and y,, of the form (4.12 - 19),

that likewise depend on only one coördinate, and that satisfy one-dimensional wave

equations of the form

H,v.(x)=(n, +})ho v,t)

(5.3-24)
H, v,(y)=(n,+i)ho v,(y),
and H,v,(2) = (n, + i)ho v,(2) .

let the full

Now, consider the product wave function ylx, y, z) =
v,(x)v,(y) v,(z), and
Hamiltonian (5.3 23) operate on it. In each part of the resulting expression, two of
 three factors are
regarded
as constants
with the aid of (5.3 - 24) one may finally obtain,by the
these
operators H, H,, and H,, and

(H+H,+ H,) V,(x) V,(y) v,(2) = ¬(n,,n,n,) V,(x) y,y) v,(2)

(5.3-25)
where
en,n,n,) =(", tn, t n, + 3)ho
(n,,n,n, =0,1,2,3,...) (5.3-26)
The wave functions are products of functions of x,
y, and z, having the form
(A12 19), and the energy eigenvalues can be written as the sum of those of three
ane-dimensional Oscillators with the three independent quantum numbers n,, n.,,
and n, each of which can be zero or a positive integer. This result--and the algebraic
form of the Hamiltonian--suggests, correctly in most cases, that the three-dimen-
sional oscillator can be viewed as three independent one-dimensional oscillators.
There is no way of plotting quantum states of this system in real coördinate
space or in momentum space as we did for free particles, since neither position or
momentum is a conserved quantity. Nevertheless, the three independent quantum
numbers, n y, and n2, provide the basis for a similar geometric argument. If these
three quantum numbers are envisioned as the coördinates of an orthogonal three-
dimensional system, and if the energy e is constant, the surface defined in this space
by the above equation is a plane, as illustrated in Figure 5.3. For a slightly greater
energy, the surface is a parallel plane as shown. If we assume that the energy levels

Energy e&
Energy e+ de

n
y
dl=dehw

Figure 5.3 Planar surfaces of constant energy in an space whose coöordinates are the three
orthogonal
oscillator. The density of states follows
quantum numbers of the isotropic three-dimensional harmonic
from calculatingthe volume between these planesand dividing by the unit cell volume, which in this
case is unity.

for the
e so closely spaced form a quasi-continuum, an assunption necessary
as to
can be ne
uensity of states concept to be meaningful, the quantity 3/2 in (5.3 26)
8lected, and the equation of the constant-energy surface written as
 &
n+n,+n, =
h (5.3-27)
where e is constant. The
intercepts of this plane on all three axes axes are at
tance= E/ ho from the origin. The equation of the second plane iss
a dis.

E+de
(n,+dn )+ (n, + dn,)+ (n, + dn,)=
h (5.3-28)
and its intercepts are thus indicated in the
in the this system form a
as
diagram. The quantum states plotted
simple cubic lattice of points, spaced at distances An 1
=

along all three axes. They therefore form a lattice of cells having unit volume, so
that the density of states is given directly
by the volume dV, included between the
two surfaces in the
figure. Since the quantum numbers are all positive or zero, only
the first octant, as illustrated, is relevant. The area of the
is that of an equilateral
planar surface at energy e
triangle whose side, from the figure, is lW2. The area of an
equilateral triangle of side s is s?\3/4, and for side W2, therefore, 1213/2. The volumne
dV, is found as the product of this area and the thickness of the element, which
must be measured
along the normal to the planes, not along the axes. Along the
normal, the numbers n,, ny, and n, are all equal, their common value being, from
(5.3-27),
n,n,n, = 3h
also, dn, = dn, = dn, =
de along normal)
3hw (5.3-29)
The normal vectors n and dn can thus be written

n=i,n, +i,n, +i,n, =i,hi +ih

de de de
and dn 3ho +i,
3h +i,- (5.3-30)
From this, it is clear that the magnitude of dn, which is the thickness of the volume
element, is
de
dn 3h (5.3-31)

The density of states now follows as the product of dn and the area calculated above;

dv=V3. de de
a 2 3 hw 2(hw) hw
&(e)de=ede
or, (5.3- 32)
2(ho
This expression is useful in studying the thermal properties of crystals, particularly
their specific heats, and also in certain other contexts. Similar methods can be used
to find density of states functions for more complex systems. The reader should
review Section 3.8 (Chapter 3) for still another example of this type of
problem.
 199
5.4 THE
MAXWELL-BOLTZMANN DISTRIBUTION FUNCTION
The problem yet unsolved is that of
determining the distribution function
H). As noted previousiy, 1t there are no constraints like those
SVstem with a constant number of particles and constant total imposed by a closed
associated with all quantum states is the same, and the average
as.
energy, the probability
number of particles
ner quantum state--and thus the distribution
function--is constant and independent
of energy. This case is really too simple to be of much
practical value, and it is more
realistic to consider a thermally isolated system with a fixed number of particles or
other constituents, and a fixed total energy. In this case, it is found that the distribu-
tion function is no longer donstant, but
decreases exponentially with energY.
We shall proceed initially along classical lines, envisioning a system of N
identical particles that can (like billiard balls) be distinguished from one another by
an assigned number or other identifier. [ However, the framework of
quantum
and energy levels will be used from the beginning, though we shall refrain states
from im-
posing the Pauli exclusion principle until later. We shall thus assume an isolated
system of N identical (but identifiable) particles, with constant total energy U. These
particles populate n energy levels with energies e, e , Es,.. . E. There will be a
statistical distribution of particles among energies so that the number of particles in
each level is given by Ni, Nz, Na, .. . N, The laws of probability indicate that some
of these possible distributions of particles among energies can be arrived at by many
different and independent chains of events, while some can be formed statistically
by only a few independent paths, and are thus far less probable.
The situation can easily be understood by considering a simultaneous toss of5
identical coins. Intuitively, it is obvious that a distribution of 3 heads and 2 tails
occurs much oftener than 5 heads. But why? Well, the distribution 5 heads requires
that, independently, each coin turn up heads. The configuration 3 heads, 2 tails,
however can happen in several different ways. There is, for example, no require-
ment that coin No. 1 turn up heads. If it does not, coin No. 2 need not turn up
heads, either, necessarily. All that is needed is tor any three of the five to fall heads
up, and there are ten independent ways for this to occur: 123, 124, 125, 134, 135, 145,
34, 235, 245, and 345, to be exact. So there are ten ways to obtain the outcome 3
heads, 2 tails, while 5 heads can occur in only one way. The probability of the first
distribution is therefore ten times greater than that of the second.| The distribution
of particles among energy levels is governed by similar considerations, though the
number of "coins'" and "possible end results" is vastly greater. Finally, one must ask
what distribution of particles among levels has maximum probability, and also
whether one may assume that this distribution satisfactorily represents the thermal
equilibrium state of the system|
The calculation of the number of ways of distributing identifiable particles
among energy levels is equivalent to calculating the number of ways of distributing
umbered objects among a set of numbered boxes Adopting this sinmple model, ob-
we

are of distributing N identifiable

ask how
Snall many independent ways there
ects among n boxes so that there are finally N, objects in the first box, Nz in the sec-

ond,..N, in the i-th,.. N. in the n-th. The probability that this final distribution
.

will ultimately occur is, as in the previous example concerning the toss of five coins,
Proportional to this number.
 First let us assume that there are only two boxes, as shown in Figure 5.4.

Then, in general,there can be in box 1 and N2 in box 2, where

N, objects
N+N2 = N (5.4-1)
Let us now denote by Q(N,, N,) the number of independent ways of arriving at a
configuration in which there are N objects in box 1 and N2 in box 2. }If N; is zero,
thereis only one such way--to have all the objects in box 2. Thus, Q0, N) =1, as
illustrated in Figure 5.4(a).
For Ng= 1, there are N ways of choosing which particle to put into box 1, as
shown at (b) in the figure, so Q(1, N) = N. |For Na=2, there are N ways of choosing
the first particle and N - 1 ways of selecting the second, which leads to N(N - 1)

ways of arriving at this configuration. But some of these choices lead to the same
final state, in which the same two particles turn up in box 1. The duplication can be
eliminated, as illustrated at (c) in the figure, by dividing by the number of ways of
permuting the two particles in box 1 among themselves; this number is 2--or, more

N N2
Box 1 2

N2 (a) N=0

(b) N=1
Particle
1 N2 N2 N N2

(c) N=2
Particles
a, b N2 b, a N 2 Nways of choosing particle a
N-1 ways of choosing particleb

(d) N=3
Particles N
a,b.c N2 a,c,b b,a,c N2
JN ways of choosing particle a
N-1 ways of choosing particle b
|N-2 ways of choosing particle c
b,c.a N2 C,a,b N c,b,a N2
Figure 5.4 Configurations in which there are (a) zero, (b) one, (c) two, and (d) three identifiable objects in
thefirst of two containers which hold a total of N such objects. As illustrated, configurations which
differ only in that the same particles are interchanged within a given container are not distinguishable
physically, and therefore do not count as different configurations.

precisely, 2!. The answer is therefore Q(2, N) = N(N 1)/2!. When N2 = 3, the first

particle in box 1 can be selected ways,

the second in N 1
in N and the third in
-

ways,
 N 2 ways. But the three particles in box 1 can be permuted among themselves in
31 6 ways, as shown at (d) in the figure. As a result, Q(3, N) = N(N - 1)(N - 2)/3!.
Clearly, when there are N, particles in the first box. we shall find

Q(N,N) =N(N-1)(N-2) (N-N,+1)_ N! N!

= (5.4-2)
N! (N-N)!N,! N!N!
the final form of this result being obtained with the aid of (5.4 - 1).
Now suppose that the second box is divided by an internal partition into two
into
compartments, as illustrated in Figure 5.5, and let the quantity N2 be partitioned
two parts V1 and V2 representing the number of particles in each compartment.
in
According to the number of ways of
(5.4- 2), realizing the distribution (V1, V2)
these internal compartments is

Q(V V)= N (5.4-3)

V!V2
and designate
But now, we can view the system as having three distinct containers,
there are N1 particles in
the number of ways of arriving at a distribution in which
second as QN, V» Va).
box 1, V in the first compartment of box 2, and va in the

N2

N1 V1 V2

Box 1 2

Figure 5.5 A system in which the second box is divided

into two compartments by an internal partition.

distribution (N1, N) between the

The number of ways of arriving at the original
each of these, there are Q(Vi, v2) ways of
original boxes is still given by (5.4 2). For of box 2, as given by (5.4 3).
arranging N2 particles among the internal subdivisions
The total number of ways of realizing the distribution (N1, V1, V2) among three
distinct containers is clearly the product of these two expressions, so that

N N! N!
(5.4-4)
QN,V1V)N!N,! vilVa N!v!v2!
We may now simply relabel the three compartments, writing vi as N2 and vz as Na
and writing, for three separate containers, the expression

N!
Q(N,NN) N,!N,!N (5.4-5)
It is easy to see, moreover, that this process of subdivision can be iterated as often as
desired, and that it will lead ultimately to a result of the form

N! N!
QNN2,N, N)N,IN,IN!-- N,! (5.4-6)

i=l

as the number of ways of realizing a distribution (N, N2, N3,.. . N,) in a system of N
objects distributed among n boxes. The I1-notation above is used for products in the
same way as the -notation for sums. The actual probability of this distribution is
obtained by dividing by the number of ways of assigning N objects ton boxes with-
out regard for what final arrangement is obtained, which is equal to nN. In our
work, there is no urgent need for doing this, so we shall simply work with the quan-
tity QN, Na, N3,.. . N,) as written above. There is, however, a more important
modification to (5.4 6) that must be made before it can be applied to our problems.
We wish ultimately to associate the "objects'" in the boxes with the particles or
other constituents of our physical system, and the boxes themselves with the energy
levels of the system. If there were no such thing as quantum degeneracy, there
would be no problem, since then there would be one quantum state associated with
each box. In the presence of degeneracý--and in three-dimensional systems, you will
recall, degeneracy is to be expected--there will be g; quantum states associated with
the i-th level, where g. is a degeneracy factor that follows from Schrödinger's equa-
tion and the boundary conditions. It is these quantum states, not the energy levels,
that are occupied with equal a prori probability.
This complication can be addressed by the already familiar approach of sup-
posing that the i-th energy level is subdivided internally into g, compartments. I It is
now necessary to know how many ways there are of assigning the N;: particles in the
energy level to these various compartments. (n determining this number, there is
no requirement that any specific distribution of particles among compartments be
realized, nor any limit (as yet) as to how many particles can be placed in any given
compartment. The first particle can be assigned with equal probability to any of the
8 compartments, as can the second, the third, etc. The total number of possibilities
is thus gi multiplied by itself N, times, or gNi, There is a factor of this form for every
energy level, so a product of all these factors over all levels must be included in the
final expression for Q(N1, N2,Na,.. N,), which now becomes

N!
QNN2N N)= (5.4-7)
i=l

i=l

We shall assume that the thermal equilibrium distribution corresponds to

that which is most probable, in other words to the one which--among all possible
configurations--can be achieved in the maximum number of statistically indepen-
dent ways. The problem is now a purely mathematical exercise of maximizing the
above expression for Q with respect to the parameters N1,Nz, N3, . .. Nn, that is,
finding what set of values for these parameters will make Q as large as possible,
subject to the constraints of a constant number of particles and constant total energy.
These constraints can be stated algebraically as

N= N =
const. (5.4-8)
i=1

and eN=U =const. (5.4-9)

i=l

If there were no constraints, the problem.could be solved by setting aQ/ON, = 0

The
(= 1, 2, 3,.. .
n) and resulting equations for the set of values {NJ.
solving the
the subset of values for
fact that the parameters must be maximized only within
which the equations of constraint are obeyed introduces an
added complexity which
Consider a func-
is most easily resolved by the method of Lagrangean multipliers.
to a
tion f several variables (x1, X2. xn), which is to be maximized subject
of
. .

remain constant. For f to bee

constraint requiring that a given function plXi, Xz.. x,)
.

Under
a maximum, df must be zero, and if p
is held constant, dp mnust also vanish.
these circumstances it is clear that
(5.4-10)
df+adp =0
This can also be writ-
where a is an arbitrary, and as yet undetermined, multiplier.
ten as
(5.4-11))
+ a , ++ lrt +a,=0.
terms in parentheses above are zero,
This equation will clearly be satisfied if all the
so that
(5.4-12)
+ - d 0, (i=1,2,3,.. n).
well as the unde-
unknown values of the variables X,
as
In this problem, there are n
defines
+ 1 unknowns. Equation (5.4
12) n
termined multiplier a, for a total ofn
between these quantities, and
the equation of constraint
relationships
(5.4-13)
P 2 . X ) = const.

which can be solved for the n + 1 required

provides another, giving n + 1 equations
is referred to as a Lagrangean
a
multiplier.
answers. The quantity x2 + y2 +z2 is to be a maxi-
case where f(x, y, z)
=

Consider an example the

256. Now, the set of equa-
as
z) x* + y4 + z
=

to the constraint that plx, y,

=

mum subject
tions (5.4- 12) with x,
=
X,y, and z yield
2x+4ox=0 4ox-2
2y+4ay=0 or, 4ay=-2
2+43 = 0 4a2=-2

x + y + z = 256
while
The first three of the four above require tha
equations
= y= 2 = -2/a

Substituting this into the last equation gives 12/ = 256, an equation that can lead to
either a positive or negative value for a. If the positive value is chosen, the values
of x, y, and z as given by the equation immediately above will be imaginary, so you
must take the negative value. One then obtains

16
8
and ==f= 3

It is easily seen that the above answers satisfy the conditions of the problem as initi-
ally stated. It is also apparent that they are correct, since f can be interpreted as the
square of the distance from the origin in an (x, y, z) coördinate system, and the equa-
tion of constraint restricts the points (x, y, z) to lie on a surface which when plotted
looks like a cube with rounded edges and somewhat convex sides. The points fur
thest from the origin are, in view of the cubic symmetry of this surface, those on the
"cube" diagonal, which is exactly what is found above. Though this discussion has
been presented in the context of maximizing a certain function, it is clearly applic-
able also to problems in which minimization is sought.
In the specific problem of finding the most probable distribution of particles
among the allowed energies of a system we must satisfy not one but two equations
of constraint, (5.4 - 8) and (5.4 -9). This is easily accomplished by introduing two

multipliers instead of one. Thus, if there are two equations of constraint, in which
two functions, plx1, X2...x) and vlxu X2.. . x,), are held constant, under the stated
conditions df, d p and dy must all be zero, and one can write

df+ado+Bdy = 0 (5.4-14)

If this is expanded in the same way as (54 12), n equations of the form

+ 0, (=1,2,3,...) (5.4-15)

can be written, which with the two equations of constraint,

p , 2 .. ,)=N= N; = const. (5.4-16)

and

3)=U (5.4-17)
VA =
>e,N, = const.

provide n +2 relations solvable for the n +2 unknowns (x1,X2.x,), a and p.

Returning now to the problem at hand, we must find the values of the inde-
pendent variables (N,N, Ns,..N,) that maximize Q in (5.4 7), subject to the two
constraints (5.4 - 8) and (5.4 - 9). The problem is made much simpler if we settle for

maximizing the logarithm of Q rather than Q itself. Then, since the logarithm of a
productis the sum the logarithm of the factors, the products will be
of to
converted
sums, which are much easier to handle mathematically. The procedure is valid be-
cause a set of independent variables that maximizes any quantity also maximizes its
logarithm, or any other function of it monotonic with respect to all variables.
If we take the natural logarithm of both sides of (5.4 7), we obtain

InQ(N,N,.. N,) = In N+N,Ing-In N! (5.4-18)

I=l

It will be necessary to find derivatives of the quantities N! in the above expression

it is
to carry out the maximization procedure outlined above. In order to do this,
use the Stirling series expansion for the factorial function, which ex-
convenient to
the logarithm of the factorial of a number z as
presses
1 1
2 (5.4-19)
In(z) = In
+(+)Inz-7t 12360
360 1260
126023
This can be differentiated term by term to get

1 1
In z!= Inz + (5.4-20)
dz 12 1202 2522
The series expansion (5.4 - 19) is not easy to obtain, but there is nothing particularly
illuminating about the difficult but fairly standard mathematical work involved in
getting it. We shall therefore not attempt to derive it here.
Ttis assumed in these calculations that there is an extremely large number of
in the
system, so many that the nymber N; in any given energy level, or in
particles
any resolvable energy interval is also very large. In (5.4 - 20), all terms in the series
except the first approach zero under these circumstances, which suggests that they
may be neglected. We shall therefore retain only the first term in (5.4 20). We
must now find the derivatives called for in (5.4 - 15). In doing this, it is important to
note that the quantities N; are independent variables, and that if you differentiate a
function of N, with respect to the variable Ni. you will get zero unless i and j are the
same. Therefore, in (5.4 18), for example, in differentiating with respect to Ni, there
is no contribution to the result from any summation term other than those for
which i =j. Also, it is important to recall that the degeneracy factors gi are constant.
Remembering these facts, and using only the first term in the series in (5.4 20), one
finds,
2(,In -(In)=Ing,- In N, (5.4-21)
aN
Also, from (5.4 - 16) and (5.4 17), it follows that

d N, =1and
N N, ZN,=e, (5.4-22)
N aN,H i=1
Substituting all these results into (5.4 - 15), we get

Ing-InN+a +Be, = 0 (5.4-23)

The quantity N, in this expression is the most probable number of particles to
be found in the j-th energy level, which we identity with the average number there
in the thermal equilibrium state. The number gi is the number of quantum states
associated with the j-th energy level. It follows that Ni/l8; is the average number of

particles per quantum state at that energy, which is by definition the distribution
function fle,). If we remove the first two terms above to the right side of the equa-
tion and exponentialize both sides, we can write the distribution function explicitly
as

N f¬) =ee. (5.4-24)

8
In this expression, the multipliers a and ß are still undetermined. They can be
evaluated from the equations of constraint (5.4 16) and (5.4 17), as we shall soon
see. The distribution function is an exponential function of energy, and since f is
negative, it decreases exponentially with energy. We shall show in the next section
that
1
KT
(5.4-25)
where T is the absolute temperature and k is Boltzmann's constant, equal to the
ideal gas constant per particle. The distribution function (5.4 - 24), calculated under

the assumption of identifiable particles and without reference to the Pauli exclusion
principle, is referred to as the Maxwell-Boltzmann distribution function.
From (5.4-24) and (5.4 25), it is evident that N,can be written as

N8,e"e (5.4-26)
From (5.4 16) and the above equation, however, we can write

N=, ", (5.4-27)

which can be solved for the quantity expa to obtain

N
e
-E/kT (5.4-28)

and
-E /kT
N,8,ee/T Nge
-E/KT (5.4-29)
28,e
 For systems having a
a.will be be replaced by the
quasi-continuum of energy levels, the degeneracy factor
density of states
function gle)de, the population factors
Nj
are replaced by the function
N(¬)de and the summations by
integrations over the
range of allowed energies. This allows the four expressions above to be written as

N(E) de =
fle)g(¬) de e"etlk" g(E)de
=
, (5.4-30)

N=Ne) de =e" ste)e " de (5.4-31)

N
(5.4-32)
&(e)ee/de
and

N(e) de =*
NeElk g(e) de (5.4-33)
8(eeeTde

The quantity QN1, Na, N3,.. . N,) is closely related to the thermodynamic
is a measure of the randomness
entropy of the system. You will recall that entropy
or disorder in a thermodynamic system.
The parameter Q has similar properties.
ordered state, for example,
When the quantities (N1, N2, N3,... N,) describe a highly
there are relatively few
one in which all the particles are in a single energy level,
such a distribution, and Q is therefore
small. The distributions of
ways of realizing be realized in many
conversely, are random configurations that can
high probability, state
which Q is relatively large. The equilibrium state--a
independent ways, and for to evolve over time--is a distribution of maxi-
toward which a closed system tends
which Q has the largest possible value. This is also
a property
m u m probability, for
it can be shown (though we shall not bother to
associated with the entropy. lndeed,
between these two quantities can be written as
do so here) that the relationship

S kln n" (5.4-34)

where S is the thermodynamic entropy.

5.5 MAXWELL-BOLTZMANN STATISTICS OF AN IDEAL GAS

We shall now apply the results derived in the preceding section to a mona-
tomic ideal gas. We envision such a gas as a system of free particles--particles that
between which there are no
may undergo elastic collisions with one another, but
long-range forces. In such a system, the potential energy of interaction is zero, and
the entire energyof the system resides in the kinetic energy of the particles. The
quantum states and energy levels of this system have been described in Section 5.3
and the density of states function has already been exhibited as equation (5.3 - 17) for

particles with no spin degeneracy, which is what we shall assume for the present.
 Substituting this density of states function into (5.4 32), we can express the
parameter expa as

e (5.5-1)
4V2rVm/2
h veetlde
Jo

The
lowest energy particle in this system
a have is zero, when it is at rest. The
can
range of allowed energies, which defines the integration limits, is thus as shown
above. The above integral may now be transformed by the substitution

X = e/kT
(5.5-2)
to the form
N
e" =.
4 2 r vk T ) 2
(5.5-3)

The integral in this

equation can be expressed as a T-function, which is a generaliz-
ation of the factorial function to numbers that need not be
series (5.4 19) and (5.4 20) are
integers. The Stirling
actually expansions of this function. The T-function
and its relationship
to the factorial function, however, is most definedsimply by

r(n)=(n-1)!= |xd (5.5-4)

From this, it is apparent that the above integral is N(3/2), which can be shown to
have the value Vtf2. Subsituting this value into
(5.5 3), we can write
-

3/2

2 70mkT (5.5-5)

The Maxwell-Boltzmann distribution function for this system can now be written

3/2
fe)=e"= eElkT

27mkT (5.5-6)

It is
important to note that the value of expa is
one must observe
that these expressions are applicabletemperature-dependent. Also,
AIso,
only free-particle systems
to
whose density of states is
given by (5.3 17), and do not apply to
density of states functions
differs from this form. A plot of the systems whose
tion (5.5 6) for several
-
Boltzmann distribu-
ticle densities in temperatures is shown in Figure 5.6. The
distribution of par-
energy can be written as

N(e)de =f(¬)g(e)de=. 2 7N
(7tkT)377 de
Ve e"t/RT
(5.5-7)
 The total energy U, which represents the thermodynamic internal energy of
the system, and the specific heat, can also be calculated quite simply. The internal
energy is clearly expressable as

U=V(e)de "ds, (5.5-8)

where, as in (5.5 - 3), x = e/kT. The integral can be identified from (5.5 - 4) as I(5/2),

or (3/2), which must equal (3/2)(1/2)!. Since we have seen above that (1/2)! has the
value T/2, the above integral must be equal to 3Vz/4, which allows us to write a
result that may already be familiar,

U=NKT (5.5-9)

The internal energy per particle is thus 3kT/2. The heat capacity C. (at constant vol-
ume) is defined as the increment of internal energy needed to produce unit tempe-
rature rise. This can be written as

(5.5-10)

The specific heat c, is simply the heat capacity per unit mass, per mole, per particle,
or other unit quantity. Per particle, the specific heat is clearly 3k/2. In a mole of
substance, the number of atoms is Avogadro's number, N, so that the molar specific
heat is
,N. (5.5-11)

This result indicates that the specific heat of an ideal gas should be independent of
temperature, a result that is in agreement with experiment, as is also the predicted
magnitude of c,.
It is also important to know how particles are distributed in velocity or mo-
mentum, as well as in energy. The distribution (5.5 7) can be converted to a veloc-
ity distribution by recalling that

e=mv+v+v)=4m and de =mvdv. (5.5-12)

Making these substitutions in (5.5 7) yields

3/2
Nvdv=4nv 2 7TkT-m'n dy (5.5-13)

The quantity N(»)dv can be interpreted as the number of particles in a velocity in-
terval dv about the velocity v. Since only the magnitude of the velocity appears, it
is Teally a distribution of particle speeds, though the term "velocity distributíon" has
become ingrained by usage. Another interpretation is that it expresses the number
of particles in a spherical shell of radius v and thickness dv in a three-dimensional
orthogonal velocity or momentum space.
 Finally, it is sometimes useful to know how the cartesian components of
velocity are distributed. It is easy to write the distribution function (5.5 - 6) in terms
of these velocity components as

3/2
f,v,)= -m(v+v2+y2kr
fvV, v2mkT (5.5-14)

Also, according to (5.4 30), we can write

-

N(v,v,v,)dv, dv, dv, f(v,v,v,)8(v,v, v,) dv, dv, dv,

= ,
(5.5-15)
where g(v,, vy, v,) dv,dv,dv, represents the number of quantum states in a rectang-
ular volume element. However, velocity and momentum differ only by a factor m,
and since from Figure 5.2 it is evident that each quantum state occupies a volume of
momentum space k/V, the volume of velocity space per quantum state is

dv,dv,dv, = P9,4P;_ (5.5-16)

m

This means that the number of quantum states per unit volume of velocity space is
m3V/h3, and that
8(V,)dv_dv,dv, = dv,dv,dv, (5.5-17)
h
Substituting (5.5 - 14) and (5.5 -17) into (5.5 -15), we obtain finally

N(vv,v,)dv,dv, dv = N m
m+v)/2" dv,dv,dv,
e * d», dv,dv,
(5.5-18)
27tkT
It is also of interest to inquire about the distribution in velocity space of each
cartesian component of velocity. This distribution may be obtained merely by inte
grating the above distribution over two of the three velocity components. Thus, if
we integrate (5.5 18) over v, and v, from - to +oo, we obtain an expression for
N(v,)dv, the number of particles whose x-velocity components are in an interval
dx about the velocity Vx Using the table below to evaluate the required integrals, we
find
m
N,)dv, | v,,)dv.| dv,dv, = 2.7kT a "T dv, e m

(5.5-19)

The Boltzmann distribution function and the free-particle velocity distribu-

tions (5.5-13) and (5.5 19) are plotted for several temperatures in Figure 5.6. t5
easily observed from the figure that the most probable value of v, (and the otner
cartesian components as wel1) is zero, while the least probable particle speed v 1s
zero. This seemingly paradoxical result can be explained by recalling that for a
particle to be at rest it isn't sufficient that its x-component of velocity be zero; what is
required is that all three cartesian components be zero simultaneously. This is a tar
less probable event, and despite the high relative probability that a single compon-
ent vanishes, the probability that all of them are zero simultaneously is very low.
The situation can also be understood with reference to the density of states function

ge), which goes to zero as e approaches zero, for similar reasons.

N)
| Se) T

(b) veo

TT T

(c)

(a)

Figure 5.6 (a) The Maxwell-Boltzmann distribution of energies at three different temperatures fora
monatomic ideal gas. (b) The distribution of molecular speeds and (c) the distribution of velocity
components along one of the cartesian axes for this system.

When the distribution function and the velocity distributions that follow
from it are known, it is possible to calculate average values of many quantities of
interest over the distributions, using the basic rules (5.2-3, 4, and 5). For example,
the average thermal speed <U> for a particle in an ideal Maxwell-Boltzmann distri-
bution can be calculated from (5.5 13) as

VN(v)dv
emv/2kT
N
N(v)dv

or )=v=8T
(5.5-20)

Table 5.1 is helpful in evaluating the definite integral. Both bracket and "v-bar"
notation are frequently used in reference to this quantity. For helium, this formula
yields an average velocity of 1260m/sec at 300K. The root mean square velocity can