CHAPTER FIVE

FUNDAMENTAL CONCEPTS OF STATISTICAL PHYSICS"

INTRODUCTION

In the previous chapters we have discussed classical thermodynamic principles which can be used to
predict relationships among the various macroscopic properties of a system. These thermodynamic principles can
be used to predict such things as the difference between the heat capacities GT and GZ , and to show how these
quantities might vary with other macroscopic parameters, such as pressure and temperature. It is not possible,
however, to derive from thermodynamic considerations alone such things as the absolute magnitudes of the heat
capacities. The same is also true for the internal energy and the entropy of a system. Only differences, or changes
in these quantities can be calculated using the techniques of classical thermodynmaics.
We can make significant progress and go beyond the limitation of pure thermodynamics only by making
some assumptions regarding the fundamental nature of matter. We did this in chapter two when we made use of
one of sciences most fruitful assumptions: that matter is not continuous, but is composed of a large number of
discrete particles called molecules. By applying the basic laws of Newtonian physics regarding the motion and
interaction of these microscopic particles, we were able to derive an equation of state for an “ideal gas” along
with an expression for the internal energy of that gas. We were then able to derive the heat capacity at constant
volume and pressure for this ideal gas. Although our treatment of the motion of molecules in a gas was very
elementary, the molecular theory of gases has actually been very thoroughly developed. Progress in the treatment
of liquids and solids, however, has been much slower because of the complexity of the molecular interactions
within liquids and solids.
A typical approach to modeling a system on the atomic scale is to make certain “reasonable” assumptions
about the motion and interaction of the individual molecules. The subject of kinetic theory attempts to precisely
apply the laws of mechanics to individual molecules of a system in an attempt to understand the behavior of the
system as a whole. In this approach, it is assumed that the laws of mechanics, deduced from the behavior of
matter in bulk, can be applied, without change, to particles like molecules and electrons. This is a formidable task
and has many inherent difficulties - not the least of which is keeping up with the position and momentum of each
of the nearly 10"* individual particles found in a cubic centimeter of air at standard temperature and pressure.
Statistical thermodynamics (or statistical mechanics), on the other hand, ignores the details of the
interactions among individual particles and attempts to use statistical principles to predict the average behaviour
of the system. In fact, unlike simple kinetic theory, it is the large number of individual elements of the system
which makes this particular approach so useful. Both approaches, however, make the basic assumption that the
molecules of the system obey the classical laws of mechanics. As progress has been made in both of these areas,
it has now become evident that this principle assumption is not correct. Macroscopic predictions based upon
these two models, although very promising, have not been in complete accord with experimental facts. This
failure of small-scale systems to obey the same mechanical laws as large-scale systems has led to the development
of what we now call quantum mechanics. Today, the subject of statistical thermodynamics is best treated from the
viewpoint of quantum mechanics, and we will typically make use of quantum mechanical arguments throughout
our treatment of statistical thermodynamics.

CHARACTERISTIC FEATURES OF MACROSCOPIC SYSTEMS

AND THE APPLICABILITY OF STATISTICS

The world that we experience through our senses consists of objects that are macroscopic, i.e., large
compared to atomic dimensions. In fact, it is almost inconceivable just how many atoms or molecules make up
even the smallest visible object. As we begin our study of statistical thermodynamics, we want to investigate how
just a few unifying concepts of atomic theory can lead to an understanding of the observed behavior of
macroscopic systems, how the quantities which describe the directly measurable properties of such systems are
interrelated, and how these quantities can be deduced from a knowledge of atomic characteristics.
Although a macroscopic system consists of very many atoms, the laws of quantum mechanics describing
the dynamical behavior of atomic particles are well established. In addition, the electromagnetic forces
responsible for the interactions between these atomic particles are also very well understood. Ordinarily these are
the only forces relevant because gravitational forces between atomic particles are usually negligibly small
Chapter Five: Fundamentals of Statistical Physics 2

compared to electromagnetic forces. In addition, a knowledge of nuclear forces is usually not necessary since the
atomic nuclei do not get disrupted in most ordinary macroscopic physical systems and in all chemical and
biological systems. Thus, our knowledge of the laws of microscopic physics should be quite adequate to allow us,
in principle, to deduce the properties of any macroscopic system from a knowledge of its microscopic
constituents. One might think, for example, that if we knew the position and momentum of each particle of a
system, and the interaction force acting on each individual particle, we could predict the motion of each individual
particle as they interact with one another. However, a typical macroscopic system of the type encountered in
everyday life contains about 1025 interacting atoms. Determining the three-dimensional position and momentum
of each of these 1025 interacting particles dwarfs the capabilities of even the most fanciful of future computers;
furthermore, unless one asks the right questions, reams of computer output are likely to provide no insight
whatever into the essential features of a problem. In addition, these "simple" interactions often produce
unexpected results. For instance, consider a gas of identical simple atoms (e.g., helium atoms) which interact with
each other through simple known forces. It is by no means evident, from simple microscopic information, that
such a gas can condense very abruptly so as to form a liquid. Yet this is precisely what can happen.
The discovery of concepts sufficiently powerful to gain a clear understanding of the relationships between
microscopic interactions and measurable macroscopic parameters clearly represents a major intellectual
challenge. But, over the years, clever individuals have been able to utilize relatively simple reasoning to gain
substantial progress in understanding macroscopic systems. And, surprisingly, it is the very presence of large
numbers of particles that allows us to effectively use statistical methods to gain some insight into how
microscopic interactions manifest themselves in macroscopic ways.
As an example of how this may be accomplished, we shall look at a few simple examples dealing with gas
particles confined inside a container.

Fluctuations in Equilibrium

Consider a gas of identical molecules, e.g., helium (He) or nitrogen (N2 ). If the gas is dilute (i.e., if the
number of molecules per unit volume is small), the average separation between the molecules is large and their
mutual interaction is correspondingly small. If the gas is sufficiently dilute so that the interaction between its
molecules is essentially negligible this gas is called an ideal gas. Each of the molecules of this ideal gas spends
most of its time moving like a free particle uninfluenced by the presence of the other molecules or the container
walls; only rarely does it come sufficiently close to the other molecules or the container walls so as to interact (or
collide) with them. In addition, if the gas is sufficiently dilute, the average separation between its molecules is
much much larger than the average de Broglie wavelength of a molecule, so that quantum-mechanical effects are
of negligible importance and it is permissible to treat the molecules as distinguishable particles moving along
classical trajectories.
We now consider such an ideal gas of R molecules confined within a container or box which is isolated
(i.e., it does not interact with any other system) and which has been left undisturbed for a very long time (thus, the
system is in thermodynamic equilibrium). Now imagine that we can use a movie camera to record the motion of
the molecules as they move around inside the container without disturbing them in any way. Successive frames
would show the positions of each of the molecules at regular intervals separated by some short time 7 .
As we watch this movie we would observe the gas molecules in constant motion. Any given molecule
would move along a straight line until it collides with some other molecule or with the walls of the box and then it
would continue moving along some other straight line until it collides again; and so on and so forth. Each
molecule would move strictly in accordance with the laws mechanics. Nevertheless, R molecules moving
throughout the box and colliding with each other would represent a situation so complex that the picture on the
screen would appear rather chaotic (unless, of course, R is very small).
We now focus our attention only on the positions of the molecules and their distribution in space. (We
ignore entirely the momentum of the molecules.) We thus define a limited amount of information which we really
want to know (i.e., the precise state of the system in which we are really interested). To be more precise, let's
consider the box to be divided by some imaginary partition into two equal parts (see Fig. 5.1). We will denote the
number of molecules in the left half of the box by 8 and the number in the right half by 8´. Of course
8  8´ œ R (5.1)
the total number of molecules in the box. If R is large, we would ordinarily find that 8 ¶ 8´, i.e., that roughly
half of the molecules are in each half of the box (altough this statement is only approximately true). As we watch
Chapter Five: Fundamentals of Statistical Physics 3

the movie, we see the molecules colliding occasionally with each other or with the walls, some of them enter the
left half of the box, while others leave it. The number 8 of molecules actually located in the left half of the
container fluctuates constantly in time, but these fluctuations (for large R Ñ are usually small enough so that 8 does
not differ too much from R /#.
You might wonder if it is possible for all the molecules to be found on one side of the box at a particular
instant in time. There is, in fact, nothing that prevents all molecules from being in the left half of the box (so that
8 œ R , while 8´ œ 0). But just how likely is it that such an event might occur? To gain some insight into this
question, let us ask in how many ways the molecules can be distributed between the two halves of the box. We
shall call each distinct way in which the molecules can be distributed between these two halves a configuration.
First, consider a single molecule. This molecule can be found in the box in two possible configurations,
i.e., it can be either in the left half or the right half. Since the two halves have equal volumes and are otherwise
equivalent, we suspect that the molecule is equally likely to be found in either half of the box. If we consider #
molecules, each one of them can be found in either of the 2 halves. Hence the total number of possible
configurations (i.e., the total number of possible ways in which the 2 molecules can be distributed between the
two halves) is equal to # ‚ # œ ## œ % since there are, for each possible configuration of the first molecule, 2
possible configurations of the other. If we consider 3 molecules, the total number of their possible configurations
is equal to 2 ‚ 2 ‚ 2 œ 23 œ 8. Similarly, if we consider the case of 4 molecules, the number of possible
configurations is given by 2% œ "'. (These 16 configurations are listed explicitly in the appendix in Table 5.1.)
[NOTE: In our treatment of this simple situation, we assume that the likelihood of finding a particular molecule
in any half of the box is unaffected by the presence there of any number of other molecules. This should be true if
the total volume occupied by the molecules themselves is negligibly small compared to the volume of the box.]
In the general case of R molecules, the total number of possible configurations is 2 ‚ 2 ‚ â ‚ 2 œ 2R .
Note that there is only one way of distributing the N molecules so that all N of them are in the left half of the box.
This arrangement represents only one special configuration of the molecules compared to the 2R possible
configurations of these molecules. Hence we would expect that, among a very large number of frames of our
movie, on the average only one out of every 2R frames would show all the molecules to be in the left half. If cR
denotes the fraction of frames showing all the R molecules located in the left half of the box, i.e., if cR denotes
the relative frequency, or probability, of finding all the R molecules in the left half, then
"
cR œ (5.2)
#R
Similarly, the case where no molecule at all is in the left half is also very special since there is again only one such
configuration of the molecules out of 2R possible configurations. Thus the probability c! of finding no molecule
located in the left half should also be given by
"
c! œ (5.3)
#R
More generally, consider a situation where 8 of the R molecules of the gas are located in the left half of the
box and let us denote by VÐ8Ñ the number of possible configurations of the molecules in this case. [That is, VÐ8Ñ
is the number of possible ways the molecules can be distributed in the box so that 8 of them are found in the left
half of the box.] Since the total number of possible configurations of the molecules is 2R , one would expect that,
among a very large number of frames of our movie, on the average VÐ8Ñ out of every 2R such frames would show
8 molecules to be in the left half of the box. If c8 denotes the fraction of frames showing 8 molecules located in
the left half, i.e., if c8 denotes the relative frequency, or probability, of finding 8 molecules in the left half, then
VÐ8Ñ
c8 œ (5.4)
#R
Example: Consider the special case (Fig. 5.1 and Table 5.1) where the gas consists of only four
molecules. Suppose that a movie of this gas consists of a great many frames. Then we expect that
the fraction c8 of these frames showing 8 molecules in the left half (and correspondingly
8´ œ R  8 molecules in the right half) is given by:
1 4 6
c! œ c% œ 16
c$ œ c" œ 16
c# œ 16
Chapter Five: Fundamentals of Statistical Physics 4

Figure 5.1
Four Particles in a Divided Box

Three “pictures” of four particles in a box at different instances of time. The box is divided into two
equal parts and the number of particles in each half of the box is indicated below that half of the
box. The short line segment eminating from each particle indicates the direction of the particle's
velocity.

Table 5.1
Enumeration of the 16 possible ways that 4 particles can be located in the two halves of a
container.

1 2 3 4 n n' C(n)
L L L L 4 0 1
L L L R 3 1
L L R L 3 1 4
L R L L 3 1
R L L L 3 1
L L R R 2 2
L R L R 2 2
L R R L 2 2 6
R L L R 2 2
R L R L 2 2
R R L L 2 2
L R R R 1 3
R L R R 1 3 4
R R L R 1 3
R R R L 1 3
R R R R 0 4 1

The first four columns indicate the location of each of the four particles in the box. Each is
labeled as to which side of the box the particle is in. The columns designated by 8 and 8w
indicate the total number of particles in each half of the box, and the column designated by GÐ8Ñ
indicates the number of different microstates which correspond to a given macrostate.
Chapter Five: Fundamentals of Statistical Physics 5

As we have seen, a situation where 8 = R (or where 8 = 0) corresponds only to a single possible molecular
configuration. More generally, if R is large, then VÐ8Ñ << 2R if 8 is even moderately close to R (or even
moderately close to 0). In other words, a situation where the distribution of molecules is so nonuniform that 8 <<
R /2 (or that 8 >> R /2) corresponds to relatively few configurations. A situation of this kind, which can be
obtained in relatively few ways, is rather special and is accordingly said to be relatively nonrandom or orderly; it
occurs relatively infrequently. On the other hand, a situation where the distribution of the molecules is almost
uniform, so that 8 ¶ 8´, corresponds to many possible configurations; indeed, as is illustrated in Table 5.1, VÐ8Ñ
is a maximum if 8 œ 8´ œ R /2. A situation of this kind, which can be obtained in many different ways, is said to
be random or disordered; it occurs quite frequently. In short, the case where the molecules are distributed more
randomly (or uniformly) occur more frequently than the less random ones. The physical reason is clear: All
molecules must move in a very special way if they are to concentrate themselves predominantly in one part of the
box; similarly, if they are all located in one part of the box, they must move in a very special way if they are to
remain concentrated there.
To gain a greater appreciation for just how often we might expect to find all the molecules on one side of
our container, let us consider some specific examples. If the gas consisted of only 4 molecules, all of them would,
on the average, be found in the left half of the box once in every 16 frames of our movie. A fluctuation of this
kind would thus occur with moderate frequency. On the other hand, if the gas consisted of 80 molecules, all of
these would, on the average, be found in the left half of the box in only one out of 280 ¶ 1024 frames of our
movie. In other words, if we could view a million frames every second, we would have to watch our movie
considerably longer than the age of the universe before we would have a reasonable chance of obtaining just one
frame which would show all the molecules in the left half of the box. [NOTE: There are about 3.15 ‚ 107 seconds
in a year and the estimated age of the universe is of the order of 1010 years.] Finally, suppose that we consider as a
realistic example a box having a volume of 1 cm3 and containing air at atmospheric pressure and room
temperature. Such a box contains about 2.5 ‚ 1019 molecules. A fluctuation where all of these are located in one
19 18
half of the box would, on the average, appear in only one out of 22.5‚10 ¶ 107.5‚10 frames of our movie.
Fluctuations where not all, but a predominant majority of the molecules are found in one half of the box, would
occur somewhat more frequently; but this frequency of occurrence would still be exceedingly small. Hence we
arrive at the following general conclusion: If the total number of particles is large, fluctuations corresponding to
an appreciably nonuniform distribution of these molecules occur almost never.
Now, let's again consider our isolated ideal gas which has been left undisturbed for a long time. The
number 8 of molecules in one half of the box will fluctuate in time about the value R /2 which occurs most
frequently. The frequency of occurrence of any particular value of 8 decreases rapidly the more 8 differs from
R /2, i.e., the greater the difference |?8| where
R
?8 ´ 8  2
(5.5)

Indeed, if R is large, only values of 8 with |?8|<< R occur with significant frequency. Positive and negative
values of ?8 would, of course, occur equally often. The time dependence of 8 for the case where R œ % and
R œ %! are shown for a particular sequence of time in Figure 5.3. If R were much greater than 40, we would
expect that 8ÎR would be almost always 1/2, with very small fluctuations about this value.
Chapter Five: Fundamentals of Statistical Physics 6

Figure 5.2
Fifteen Frames of a Computer Simulated Movie of Four Particles Moving in a Container, Separated into
Two Equal Halves, Showing the Number of Particles in Each Half and the Velocity Direction of Each
Particle

Fig. 5.2 Computer-made pictures showing 4 particles in a box. The fifteen successive frames (labeled
4 œ !ß "ß #ß âß "%Ñ are pictures taken a long time after the beginning of the computation with assumed initial
conditions. The number of particles located in each half of the box is printed directly beneath that half. the short
line segment emanating from each particle indicates the direction of the particle's velocity.
Chapter Five: Fundamentals of Statistical Physics 7

Figure 5.3
A Plot Showing the Number of Particles (and Relative Number of Particles) in the Left Half of the
Container as a Function Time for R œ % and R œ %!.

Fig. 5.3 A plot of the number of particles (and the relative number of particles) in the left half of the box as a
function of the frame index 4 or the elapsed time > œ 479 . The number 8 of the 4th frame (or the relative number)
is indicated by a horizontal line extending from 4 to 4  ". The top pair of graphs are of the case where R œ %,
while the bottom pair of graphs are for the case where R œ %!. You will notice that the information for the case
where R œ % is identical to that shown in Figure 5.2 for the first 15 frames.
Chapter Five: Fundamentals of Statistical Physics 8

Conclusion
The microscopic state, or microstate of a gas confined within a container can be described in greatest detail
by specifying the position and velocity of each molecule of the system at any particular time (giving the maximum
possible information about the gas molecules at that particular instant). From this microscopic point of view, the
motion of the motion of the gas particles appears very complex since the locations and the momenta of each
individual molecule are constantly changing from one instant to another.
From a large-scale or macroscopic point of view, however, one is not typically interested in the behavior
of each and every molecule, but in a much less detailed description of the gas. Thus the macroscopic state, or
macrostate, of the gas, for example, might be quite adequately described by specifying merely the number of
molecules located in any part of the box at any particular time. [NOTE: To be specific, we could imagine that the
container is subdivided into many equal cells, each having the same volume and each cell's volume being large
enough to contain many individual molecules. The macroscopic state of the gas could then be described by
specifying the number of molecules located in each cell of the container.]
From this macroscopic point of view, an isolated gas which has been left undisturbed for a long time
represents a very simple situation since its macroscopic state does not tend to change significantly over time.
Indeed, suppose that, starting at some time >1 , we observed the gas over some moderately long period of time T by
taking a movie of it. Alternatively, suppose that, starting at some other time >2 , we again observed the gas over
the same time T by taking a movie of it. From a macroscopic point of view these two movies would ordinarily
look indistinguishable. For the case where we have divided our container into two equal parts, we would find that
in each case the number 8 of particles in the left half of the box would ordinarily fluctuate about the same value
R /2, and the magnitude of the observed fluctuations would ordinarily also look alike. Disregarding very
exceptional occurrences the observed macrostate of the gas is thus independent of the starting time of our
observations; i.e., we can say that the macrostate of our gas does not tend to change in time. In particular, the
value about which 8 fluctuates (or, more precisely, its average value) does not tend to change in time. A system
of many particles (such as our gas) whose macroscopic state does not tend to change in time is said to be in
equilibrium. This seems to be consistent with out understanding of the concept of thermodynamic equilibrium as
presented in the first part of this course as we attempted to describe classical thermodynamics, and thus justifies
our attempt to describe a complex molecular system using simple statistical arguments.

"
Much of the material in this chapter (in particular the figures) is taken from the first chapter
of Berkeley Physics Course, Vol. 5, "Statistical Physics", by F. Reif, McGraw-Hill, 1967.