Lecture 2: The Microcanonical Ensemble

2.1 Recommended textbook chapters for this section

• McQuarrie - Chapter 1, Chapter 2.1

• Chandler - Chapter 3.1-3.2

2.2 Topics in this lecture

• Thermodynamics review

• Posulates of statistical mechanics

2.3 Thermodynamics review

2.3.1 The fundamental relation of thermodynamics
We ended the last lecture by discussing the combined first and second law of thermodynamics
(assuming a one-component system), which we write as:

dE = T dS − P dV (2.1)

This expression is also called the fundamental relation of thermodynamics in the energy
representation. We can write an equivalent expression in terms of the entropy:

1 P
dS = dE + dV (2.2)
T T
This is called the fundamental relation in the entropy representation. Each pair of variables
that appear in this expression is called a pair of conjugate variables, consisting of an intensive
prefactor and extensive differential. The four most common sets of conjugate pairs that we will
discuss are:

• Temperature (T , intensive) and entropy (S, extensive)

• Temperature (T , intensive) and energy (E, extensive; from entropy representation of funda-
mental relation)

• Pressure (P , intensive) and volume (V , extensive)

1
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

• Chemical potential (µ, intensive) and number of particles (N , extensive)

Because the energy is a state function, we can also write an exact differential of dE to obtain
expressions for intensive variables:
   
∂E ∂E
dE = dS + dV (2.3)
∂S V ∂V S
The subscripts indicate that the partial derivative is taken with the subscripted variable held
constant. We can think of the exact differential as essentially saying that the change in energy
as a function of S and V (obtained by integrating this expression) can be obtained by computing
the energy change along a two-part path consisting of a constant V path in which S varies (corre-
sponding to the first term) and a constant S path in which V varies. We could similarly write an
exact differential for E in terms of any variables thermodynamic variables. However, this particular
exact differential is useful because we can compare it to the fundamental relation in the energy
representation to obtain:

 
∂E
=T (2.4)
∂S V
 
∂E
= −P (2.5)
∂V S

In other words, the temperature is defined as the partial derivative of the internal energy
with respect to entropy at constant volume. Using these relations, we find that for each pair of
conjugate variables that appears in the fundamental relation, the variable that is always intensive
can be expressed as a partial derivative of either the energy or the entropy with respect to its
conjugate extensive variable and with all other extensive variables held constant.

2.3.2 Natural variables

We can extend the fundamental relation to include additional work terms (i.e., additional conjugate
pairs) by writing:

dE = T dS − P dV + µdN + other work terms (2.6)

We will typically omit work terms other than pressure-volume and chemical work unless required
to include them by a specific problem. Note again that each intensive variables (T, −P, µ) can be
written as a partial derivative of the internal energy with respect to its conjugate extensive variable.
As a result, we could integrate the fundamental relation to obtain a function that is only a function
of the extensive variables - that is, a function of the form E(N, V, S) since each intensive variable
is fully specified in terms of these variables. N , V , and S are called the natural variables of
the internal energy. Similarly, N , V , and E are the natural variables of the entropy, S(N, V, E),
based on the fundamental relation in the entropy representation. By inspection of the fundamental
relation, we see that dE = 0 for any process in which N , V , and S are held constant (such that
dN = dV = dS = 0) and therefore E(N, V, S) is at an extremum. If we were to control the natural
variables via suitable experimental manipulation, or by imposing the appropriate boundaries on
a system, we could then predict that the corresponding thermodynamic potential would go to an
extreme. We will take advantage of this behavior later.

2
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

2.3.3 Equilibrium
A system is said to be at equilibrium when there is no driving force to change intensive variables
in the system and the system’s properties do not change as a function of time. A system that is at
equilibrium is described as being in an equilibrium state which is fully specified by the values of
the thermodynamic parameters described above. The three most relevant conditions of equilibrium
that we will discuss are:

• Thermal equilibrium: if a system has diathermal walls, at equilibrium the temperature of

the system will be equivalent to the temperature of the surroundings, or T sys = T surr .
• Mechanical equilibrium: if a system has movable walls, at equilibrium the pressure of the
system will be equivalent to the pressure of the surroundings, or P sys = P surr .
• Chemical equilibrium: if a system has open walls, at equilibrium the chemical potential of
each component of the system will be equivalent to the chemical potential of each component
in the surroundings, or µsys surr .
i = µi

In each case, the system is free to exchange an extensive variable and the corresponding intensive
variable is then constant. These pairs of variables are again the conjugate pairs mentioned above.
A given system can satisfy multiple conditions of equilibrium simultaneously.

2.3.4 Additional thermodynamic potentials

It is possible to write other expressions, called thermodynamic potentials, that are equivalent
in information content to the fundamental relation but have a different set of natural variables. In
the third unit of this class, we will show that these relations are equivalent; here, we will only focus
on the method for defining them. The procedure is as follows:

• Identify the desired natural variables of the new potential (e.g., N, V, T )

• For any natural variable that is different from the natural variables of the internal energy
(i.e., N, V, S), subtract the product of the desired natural variable and its conjugate from the
internal energy, E. The new expression defines the equivalent thermodynamic potential.

Examples of this protocol follow for several typical thermodynamic potentials. First, we define
the enthalpy, H, as the thermodynamic potential that is a function of the natural variables N, P, S.
This differs from the internal energy by the variable P , so we subtract the product of P and its
conjugate (V ) to generate the expression for the enthalpy. The slight caveat here is that in our
sign convention it is −P , not P , that is used in the conjugate pair, so we subtract −P instead:

H = E − (−P V ) (2.7)
= E + PV (2.8)
We can now write an expression for the total derivative of the enthalpy and simplify using the
fundamental relation to identify what constant variables lead to an extremum of the enthalpy:

dH = dE + P dV + V dP (2.9)
= T dS − P dV + µdN + P dV + V dP (2.10)
= T dS + V dP + µdN (2.11)

3
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

The total derivative clearly illustrates that the natural variables of the enthalpy are N , P , and S
as desired - if these variables are constant, the enthalpy reaches an extremum. We can now identify
various relationships for thermodynamic parameters by comparing with the exact differential of the
enthalpy:

     
∂H ∂H ∂H
dH = dS + dP + dN (2.12)
∂S P,N ∂P S,N ∂N S,P
 
∂H
=T (2.13)
∂S P,N
 
∂H
=V (2.14)
∂P S,N
 
∂H
=µ (2.15)
∂N S,P

We can repeat this protocol to identify a thermodynamic potential that is a function of the
natural variables N, V, T , called the Helmholtz free energy, F (here ignoring the chemical
potential term for simplicity):

F = E − TS (2.16)
dF = dE − T dS − SdT
= (T dS − P dV ) − T dS − SdT
= −SdT − P dV (2.17)
 
∂F
= −S (2.18)
∂T V
 
∂F
= −P (2.19)
∂S T

Finally, we can identify a thermodynamic potential that is a function of natural variables

N, P, T , called the Gibbs free energy, G:

G = E − TS + PV (2.20)
dG = dE − T dS − SdT + P dV + V dP
= (T dS − P dV ) − T dS − SdT + P dV + V dP
= −SdT + V dP (2.21)
 
∂G
= −S (2.22)
∂T P
 
∂G
=V (2.23)
∂P T

We will return to these expressions repeatedly throughout our discussion of statistical mechanics,
and will discuss thermodynamic relations in more detail later in the course. If the importance of
natural variables is not clear, you can first see that different thermodynamic potentials will reach
extremum values when different conditions of equilibrium occur. For example, a system with

4
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

impermeable, rigid, diathermal walls that reaches thermal equilibrium with the surroundings will
have constant N , V , and T , and thus the Helmholtz free energy (with corresponding natural
variables) will reach an extremum value. Analysis of the Helmholtz free energy can then be used to
understand how changes to this system properties affect its equilibrium behavior. We will further
use the natural variables to connect to statistical mechanics as will be discussed in future lectures.
For now, we conclude our review of thermodynamics.

2.4 Postulates of statistical mechanics

Recall that our goal in statistical mechanics is to develop a mathematical framework that allows
us to calculate the macroscopic, thermodynamic properties of a system from a description of the
properties of individual molecules. For example, we might try to derive the value of a given
thermodynamic property such as the temperature or pressure from the knowledge of the quantum
mechanical energy levels available to a system with N particles. We will take the first steps toward
accomplishing that goal in this lecture.
Imagine a system composed of a single-component gas enclosed within a sealed, rigid, insulated
container. The walls of the container therefore prevent a change in the number of particles (N ),
the volume (V ), or the energy (E) of the gas; the system is isolated. Equivalently, we could
describe the system in terms of what macroscopic thermodynamic parameters are constant - this
gas is defined as having constant N V E, where these particular variables are constant due to the
constraints imposed by the walls, as opposed to being constant due to a condition of equilibrium.
At the macroscale, these three parameters define the system - if I were to characterize the gas
with no knowledge of the existence of individual gas particles, I would characterize it strictly in
terms of the values of N , V , and E. We therefore define the macrostate of the system as the
particular values of N V E (and any thermodynamic parameters that are not fixed, such as the
pressure or temperature in this example) that describe the current state of the gas. However, we
know that at the microscale there are an enormous number of particle configurations that can fulfill
these requirements. For example, assume that the gas is ideal and therefore the only contribution
to the internal energy is the kinetic energy of each particle. Even if the total internal energy (i.e.,
the sum of the kinetic energy of each particle) is a constant value, there are an enormous number
of ways that we could arrange particle positions in space that would still have the same internal
energy, since the kinetic energy does not depend on particle positions.
If we were to take a snapshot of the gas at a particular instant in time and record the positions of
all N particles, we would define a microstate of the system - that is, one particular configuration
out of the enormous set of configurations of particle positions that have the same macroscopic
values of N , V , and E. There are thus many, many microstates that are compatible
with a given macrostate. Oftentimes, authors will interchangably use the term “state” to mean
“microstate,” which can be a bit confusing, but should be apparent from context.

5
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

Given this definition of a microstate, we can then define the concept of an ensemble of mi-
crostates. An ensemble is a mental collection of all possible unique microstates that are consistent
with a specific macrostate; each member of the ensemble is considered identical macroscopically,
but unique microscopically. We will discuss four primary ensembles in this course that are given
names reflecting the choice of thermodynamic variables that are held constant. The ensemble for
which all microstates have a fixed number of particles, volume, and internal energy, is called the
microcanonical ensemble, or the N V E ensemble. The ensemble for which all microstates have
a fixed number of particles, volume, and temperature, is called the canonical ensemble, or N V T
ensemble. The ensemble for which all microstates have a fixed chemical potential, volume, and
temperature, is called the grand canonical ensemble, or µV T ensemble. Finally, the ensemble
for which all microstates have a fixed number of particles, pressure, and temperature is called the
isothermal-isobaric ensemble, or N P T ensemble.
Many more ensembles are possible based on different thermodynamic constraints. There are a
couple of caveats to note. First, note that each microstate of an ensemble will also have thermo-
dynamic parameters that are not fixed, and can vary with each microstate. For example, we can
define an instantaneous energy, Ei , for the ith microstate in the canonical ensemble, and this energy
will generally vary between microstates. Also note that in several of these ensembles parameters
are fixed not due to the walls of the system but rather due to conditions of equilibrium, a point
that we will return to in the next lecture.
The number of microstates in an ensemble is an important quantity referred to as the degener-
acy of the ensemble, which we write as Ω(N, V, E). The parentheses indicate the fixed variables in
the corresponding ensemble, so this is the degeneracy of a microcanonical ensemble; the degeneracy
of a canonical ensemble would be written as Ω(N, V, T ). For shorthand, I will occasionally refer to
the degeneracy of the microcanonical ensemble as Ω(E) which is equivalent to Ω(N, V, E) unless
otherwise specified. Here, we assume that the energy of a system is a discrete quantity such that
the number of states can be counted explicitly. Systems for which the energy is nearly continuous
are better described by a related concept called the density of states, ω(N, V, E), which describes
the number of states with an energy within a small interval E + dE. We will largely deal with the
degeneracy as our chosen quantity for most of this class. Other terms are often used to describe
the degeneracy of an ensemble - for example, the microstates in an ensemble are often called the
“phase space” of a system so that the degeneracy would be the “extent of phase space.”
Having defined an ensemble of microstates as a collection of thermodynamically identical but
microscopically distinct configurations of a system, we can now define two fundamental postulates
of statistical mechanics.

6
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

2.4.1 Postulate 1:
The time-average value of a thermodynamic parameter, Y , of a system is equal to the ensemble-
average value of Y .
The ensemble average is defined as the value of a parameter averaged across all microstates of
a statistical ensemble, weighted by the probability that the system is in each microstate. We can
formulate the first postulate mathematically as:

Z τ
1
Yobs = Y (t)dt Time average (2.24)
τ 0
microstates
X
hY i = p i Yi Ensemble average (2.25)
i
Yobs = hY i (2.26)

Equation 2.24 defines the time-average value of Y , written as Yobs . This notation indicates
that this is the value of Y that I would observe if I were to measure the parameter experimentally
(assuming a typical macroscopic measurement); in other words, this would be the volume/pres-
sure/etc. of the macrostate. Equation 2.25 defines the ensemble-average value of Y , written as
hY i. The sum runs over the entire ensemble of microstates where Yi is the Pvalue of Y in microstate
microstates
i, and pi is the probability that the system is in microstate i (such that i pi = 1). Equa-
tion 2.26 therefore states that the value of a parameter for a macrostate can be written as the value
of that parameter averaged across all microstates in the corresponding statistical ensemble. This
postulate is the foundation of statistical mechanics!
This postulate, also called the ergodic hypothesis, illustrates the connection between statisti-
cal mechanics and thermodynamics by linking the macroscopic (observed) value of a thermodynamic
variable to an average across a large number of particle configurations. We can picture this con-
nection by again imagining a gas. Assume that the gas were allowed to exchange energy with the
environment, such that it maintains a constant temperature but its energy can fluctuate. If we
were to observe this gas for a long period of time and periodically take snapshots of molecular con-
figurations, each snapshot by definition would be a microstate in the canonical (N V T ) ensemble,
and we could calculate the value of some non-fixed thermodynamic variable Y for each microstate
(e.g., the energy or the pressure). As time progresses, we record more and more configurations of
the system - we say that the system samples microstates. Some microstates may be sampled
more often than others, and therefore we could record identical values of Y multiple times. If we
waited a sufficiently long amount of time, the system would sample each microstate i according to
the corresponding probability pi . Therefore, the time-average value of Y would be equivalent to
taking the average of all the values I recorded - that is, the time-average value would be equivalent
to the ensemble-average value according to the ergodic hypothesis.

7
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

There are two critical aspects of this postulate that relate to the observation time, τ . First, the
properties of the system cannot change with time during the period of observation; that is, the sys-
tem must be at equilibrium. Second, the observation time, τ , must be very long so that the system
properly samples a representative number of microstates. “Very long” is obviously a qualitative
statement; the relevant timescale to compare to is the autocorrelation time of the observable of
interest. If we imagine placing the system in a single microstate and then observe the positions and
energies of all particles as a function of time, the relevant autocorrelation time would be the time
necessary for the system to reach a new microstate such that particle positions/energies are not
correlated with the previously recorded positions/energies (we will return to this point in Unit 2 of
this class). Autocorrelation times are generally very small - for example, the autocorrelation time
for particle positions in a gas is on the order of picoseconds, so any macroscopic observation time
will be sufficiently long to observe a representative set of microstates from the ensemble. However,
some systems may have long autocorrelation times (e.g., glasses) and therefore may not sample a
representative number of states during an expermient; such systems would then be non-ergodic.
Because the ergodic hypothesis tells us that ensemble averages can be related to macroscopic
thermodynamic parameters, all of the rules of thermodynamics that you have learned and that we
covered in the preceding lecture can be applied to ensemble-average parameters. For example, from
the fundamental relation in the energy representation we can show:

 
∂E
=T (2.27)
∂S V

These rules have been derived assuming some macrostate with values E, S, V , and T . However,
the ergodic hypothesis now tells us that we can relate macroscopic parameters to microscopic
parameters. For example, for an ensemble in which V and T are fixed and E is allowed to vary, I
could equivalently write:

 
∂hEi
=T (2.28)
∂S V

Note that I have not commented on the entropy, S, which we will return to in Postulate 2.
Equation 2.25 writes the ensemble average in terms of a summation over discrete microstates,
with each microstate occurring in the ensemble with probability pi . In principle, however, we do
not know the values of pi . The second postulate addresses this point.

2.4.2 Postulate 2:
In the microcanonical ensemble, all microstates are equally probable.
This postulate, also referred to as the principle of equal a priori probabilities, states that
each value of pi is equivalent in Equation 2.25 for microstates in the microcanonical ensemble, as
there is no a priori reason to assign different probabilities. Using the definition of the degeneracy,
Ω(N, V, E), from earlier, this postulate defines

1
pi = (2.29)
Ω(N, V, E)
microstates Ω(N,V,E)
X 1 X
hY i = pi Yi = Yi (2.30)
Ω(N, V, E)
i 1

8
University of Wisconsin-Madison Lecture 2
CBE 710, Fall 2019 - Prof. R. C. Van Lehn September 10, 2019

Note that we can define a degeneracy for any ensemble (e.g., Ω(N, V, T ) as the degeneracy of the
canonical ensemble), but Postulate 2 only applies to the microcanonical ensemble.
In combination, the two postulates state that if an isolated system is observed for a long period
of time, all possible microstates of the system will be visited with equal probability and the time-
average value of a thermodynamic property will be equivalent to the average of that quantity
across all microstates. As an aside, it is likely that there will always be some uncertainty in
the measurement of E due to the quantum properties of a system (i.e., due to the Heisenberg
uncertainty principle); however, this small uncertainty is effectively inconsequential.

2.4.3 The Boltzmann entropy

We will close this discussion of the microcanonical ensemble with a third “postulate” that will be
stated here (although it can be derived from future results). Specifically, we define the entropy
for the microcanonical ensemble, S, as:

S = kB ln Ω(N, V, E) (2.31)
where kB = 1.381 × 10−23 J/K is Boltzmann’s constant. This definition is also referred to as
the Boltzmann entropy to distinguish from an alternative (but equivalent) definition given by
Gibbs that will be discussed later. The entropy as defined is an extensive variable, as we know
from thermodynamics, because in general the degeneracy of an ensemble will increase with system
size. The entropy is also a property of the ensemble itself, unlike other thermodynamic variables
such as the pressure which can be given a specific value for each microstate - in other words, we
cannot write Si as the entropy of a particular state.
Conceptually, this microscopic definition of the entropy agrees with the concept of entropy as a
measure of the “disorder” of a system, as is often described in thermodynamics. Here, we quantify
disorder as the number of equivalent microstates - we can imagine that for a gas, there are a much
larger number of ways of repositioning particles to yield equivalent macroscopic properties than
for a solid. Alternatively, we can think of the entropy as quantifying the volume of the phase
space accessible to a system. Regardless of how you think of it, this definition of entropy provides
another connection between the microscopic features of a system (i.e., the degeneracy or number
of accessible microstates) and macroscopic thermodynamic parameters.