Unit 3

Ideal systems in equilibrium.

3.1 Introduction
In this Unit we will consider some example of ideal systems, i.e. systems in which we
can neglect the interactions between their components to obtain their macroscopic
properties, although they are necessary to achieve the thermal equilibrium of the
system. We will focus in two systems: ideal gases, which are the idealization of
very dilute real gases, and ideal paramagnets, which model the behaviour of real
paramagnetic systems at high temperatures. In both cases, we will assume that we
can apply the Equilibrium Statistical Physics in its canonical formulation.

3.2 Distribution functions of ideal gases

Consider a molecule of mass m in a dilute gas. The gas may possibly consist of
several different kind of molecules; the molecule under consideration may also be
polyatomic or have additional internal degrees of freedom. Let us denote the position
of the center of mass of the molecule by r and the momentum of its center of mass
by p. If external fields (e.g. gravity) are neglected, the Hamiltonian of the molecule
is equal to
p2
h(r, p, qint , pint ) = + Hint (qint , pint ) (3.1)
2m
where the first term on the right-hand side is the kinetic energy of the center of mass
motion, and the second term corresponds to the energy associated to the possible
internal degrees of freedom. If the gas is diluted, we can neglect the interactions
between molecules, so we can treat it as an ideal gas. In this case, the (marginal)

1
Ideal systems in equilibrium. 2

canonical probability density of the molecule can be obtained from the total canon-
ical probability density by integration over the other molecules degrees of freedom
as  
p2
−β 2m
+Hint (qint ,pint )
Pm (r, p, qint , pint ) ∝ e (3.2)
Note that this expression is the same for all the molecules of the same species. Now,
if we are interested only on the center-of-mass variables irrespective of its internal
state, the marginal probability density is again by integrating (or summing) over
the internal degrees of freedom as
p2
P (r, p) ∝ e−β 2m (3.3)
which, again, is the same for all the molecules of the same species. Thus, if we mul-
tiply Eq. (3.3) by the total number N of molecules of the type under consideration,
we obtain the mean number of molecules density in the µ-space, i.e. the phase space
corresponding to a single particle center of mass. Let us express the result in terms
of the position and velocity v = p/m of the molecule center of mass. Thus, we define
the distribution function of velocities f (r, v) in the following way. Let’s consider a
region of the (r, v) space centered at the position r in a range d3 r and the velocity
v in a range d3 v. Then
f (r, v)d3 rd3 v (3.4)
is the mean number of molecules in the region defined above. By using Eq. (3.3),
this expression can be recast as
mv 2
f (r, v)d3 rd3 v = Ce−β 2 d3 rd3 v (3.5)
where C is a constant which will be determined by the normalization condition
Z Z
d r d3 vf (r, v) = N
3
(3.6)

Since f does not depend on r, the integration over this variable yields simply the
volume V . On the other hand, the integration on the velocities can be obtained in
a similar way we obtained the Eq. (2.49) in the Unit 2 with the use of the Gaussian
integrals I(h), so
Z ∞ 3  3
−β mv
2 2π 2
CV e 2 dv = CV =N (3.7)
−∞ βm
So   32
m mv 2
f (r, v) = n e−β 2 (3.8)
2πkB T
Ideal systems in equilibrium. 3

with n = N/V . This is the so-called Maxwell (also known as Maxwell-Boltzmann)

distribution function of velocities. Note that it is independent of the variable r
(which can also obtained from translational symmetry), so it is usual to drop the
dependence on this variable. Furthermore, it depends on the velocity through its
modulus (again, this is expected from the isotropy of space, since there is no preferred
direction).
Once we have defined the distribution function, we can define various other
distribution functions of physical interest. For example, g(vx ) is defined in such a way
that g(vx )dvx is the mean number of molecules per unit volume with x component
of the velocity between vx and vx + dvx , irrespective of the values of their other
velocity components. So, g(vx ) is the marginal distribution of vx obtained from the
Maxwell distribution function
Z ∞ Z ∞   32 Z ∞ Z ∞
m mv 2 mv 2 2
mvz
−β 2 x −β 2 y
g(vx ) = dvy dvz f (v) = n e dvy e dvz e−β 2
−∞ −∞ 2πkB T −∞ −∞
  32 r !2 r
m 2
mvx 2πkB T m 2
mvx
= n e−β 2 =n e−β 2 (3.9)
2πkB T m 2πkB T

Note that this expression is a Gaussian distribution of mean vx = 0 and variance

vx2 − vx 2 = kB T /m. From this result we obtain that

1 2 kB T
mv = (3.10)
2 x 2
which we obtained previously by the equipartition theorem. The normalization
condition for g(vx ) is that Z ∞
g(vx )dvx = n (3.11)
−∞

Obviously, the same results hold for vy and vz since all velocity components are com-
pletely equivalent by symmetry. Note also that the Maxwell distribution function
can be expressed in terms of g(vx ), g(vy ) and g(vz ) as
   
f (v) g(vx ) g(vy ) g(vz )
= (3.12)
n n n n
Thus the individual velocity component behave like statistically independent quan-
tities.
Other quantity of physical interest is the distribution function F (v) for the mod-
ulus of the velocity or speed v = |v|. So, F (v)dv is the mean number of molecules
per unit volume with speed between v and v + dv. This distribution function can
Ideal systems in equilibrium. 4

F(v)

_
v~ v vrms
v
Figure 3.1: Maxwellian distribution function of molecular speeds. The typical molec-
ular velocities are highlighted.

be obtained by integrating the Maxwell distribution function over the different di-
rections of the velocity. If we express d3 v in spherical coordinates, and noting that
the Maxwell distribution function only depends on the speed, then
Z   32
3 2 m mv
− 2k
2

F (v)dv = d vf (v) = 4πv f (v)dv = 4πn e B T

v 2 dv (3.13)
v<|v|<v+dv 2πkB T

which again is normalized as Z ∞

F (v)dv = n (3.14)
0
This distribution function is plotted in Fig. 3.1. It has a maximum at v = ṽ where
dF/dv = 0, i.e.  
mṽ 2
− 2k 2 m − mṽ
2

2ṽe B T + ṽ − ṽ e 2kB T = 0 (3.15)

kB T
so r
2kB T
ṽ = (3.16)
m
Other typical velocity is the mean speed v, defined as
R∞   32 Z ∞ r
vF (v)dv m − mv 2 8 kB T
v = R0 ∞ = 4π dvv 3 e 2kB T = (3.17)
0
F (v)dv 2πkB T 0 π m
Ideal systems in equilibrium. 5

dS

θ
vdt
v

Figure 3.2: Volume occupied by the molecules with velocity between v and v + d3 v
which collide the element area dA of the wall in a time dt.

p
where we used that the Gaussian integral I(3) = 1/(2a4 ),p
with a = m/2kB T .
Finally, we define the root-mean-square speed vrms = v 2 , where v 2 is given by
R∞ 2   32 Z ∞
v F (v)dv m − mv
2
3kB T
v2 0
= R∞ = 4π dvv 4 e 2kB T = (3.18)
0
F (v)dv 2πkB T 0 m
√ p
where we used that the Gaussian integral I(4) = 3 π/(8a5 ), with a = m/2kB T .
Alternatively, we may have used that v 2 = vx2 + vy2 + vz2 = 3vx2 = 3kB T /m. We note
p
that all these speed scales are proportional to kB T /m, multiplied a a factor of
order of 1.

3.3 Number of molecules striking a surface. Ef-

fusion.
Now, we would like to know how many molecules per unit time strike a unit area of
the wall of the container which confines the ideal gas. For this purpose, we consider
an element area dA of the wall. We choose the z axis to point along the outward
normal of this element area (see Fig. 3.2). Consider first those molecules (in the
immediate vicinity of the wall) whose velocity lies between v and v + d3 v (i.e. it is
in an infinitesimal volume d3 v centered at v). Molecules of this type which are in
a tilted cylinder of side vdt and cross-section area dA, where the tilt angle θ is the
angle of the velocity v with respect to the surface normal, will strike the wall area
Ideal systems in equilibrium. 6

dA in a time dt (i.e. between the instants t and t + dt, where t is arbitrary). This
quantity can be obtained by integrating the distribution function of velocities in the
tilted cylinder. Dividing this expression by the area dA and the time interval dt, we
get that the flux of particles of velocity v, Φ(v), is given by

Φ(v)d3 v = d3 vf (v)v cos θ (3.19)

Let Φ0 be the total flux of particles, i.e. the total number of molecules which strike
a unit area of the wall per unit time. This quantity can be obtained by integrating
Φ(v) over all the possible velocities which will cause the molecule to collide with
the element of area. It is easy to check that the condition for collision is that the
normal component of the velocity vz is positive, i.e. 0 ≤ θ < π/2. So,
Z
Φ0 = d3 vf (v)v cos θ (3.20)
vz >0

Since f (v) depends only on the speed |v|, we can evaluate this integral in spherical
coordinates
Z Z 2π Z π Z ∞
2
2
Φ0 = v sin θdθdφf (v)v cos θ = dφ dθ sin θ cos θ dvv 3 f (v)
vz >0 0 0 0
Z ∞
= π dvv 3 f (v) (3.21)
0

This result can be expressed in terms of the mean speed v, which can be computed
as
Z Z 2π Z π Z ∞
1 3
v = d vvf (v) dφ dθ sin θ dvv 3 f (v)
n 0 0 0
4π ∞
Z
= dvv 3 f (v) (3.22)
n 0
Now, comparison of Eqs. (3.21) and (3.22) leads to the following expression
1
Φ0 = nv (3.23)
4
which is valid for any distribution function of velocities which
p depends only on the
molecular speed v = |v|. For the Maxwell distribution, v = 8kB T /πm, and taking
into account that p = nkB T , we obtain that
p
Φ0 = √ (3.24)
2πmkB T
Ideal systems in equilibrium. 7

These expressions are useful to study the phenomenon of effusion. Suppose that
we make a hole in the wall of the container. If it is large, a hydrodynamic flow of
molecules will escape from the container, disturbing its equilibrium state. However,
if it is small enough (i.e. smaller than the molecular mean free path, which is
the mean distance the molecule in the gas travels before colliding with another
molecule), the equilibrium of the gas inside the container is disturbed to a negligible
extent. So, although the properties of gas in the container become time-dependent,
its variation can be regarded to be quasi-static. This process is called effusion. In
this case, the number of molecules which emerge through the small hole is the same
as the number of molecules which would strike the area occupied by the hole if the
latter were closed off. The molecular beam obtained by collimating the effusing
molecules of a container with a small hole with a vacuum maintained outside the
container can be used to check experimentally the Maxwell velocity distribution, as
well as a method for the separation of isotopes.
Another example is a container divided into two parts by an adiabatic partition
containing a hole. The container is filled with gas, but one part of the container
is maintained at temperature T1 , the other part at temperature T2 . What is the
relation between gas pressures of both sides? If the hole is large, the equilibrium is
reached when p1 = p2 , which is the mechanical equilibrium condition. However, if
the hole is small (so one deals with effusion), the equilibrium condition is reached
when the mass transfer from one container to the other is balanced with its coun-
terpart. As the total number of particles which effuse from each part is given by
Eq. (3.24), the equilibrium condition becomes
p p
√1 = √2 (3.25)
T1 T2
So, pressures are then not at all equal, but higher gas pressure prevails in the part
of the container at higher temperature.

3.4 Kinetic interpretation of the pressure

It is of interest to consider from a detailed point of view how a gas exerts a pressure.
The basic mechanism is clear: the mean force exerted on a wall of the container is
due to the many collisions of molecules with the wall, which transfer momentum to
the wall. So, the mean force dF exerted by the gas on a small element of area dA
of the container wall can be calculated as the mean rate of change of momentum of
this element of the wall. Let us denote C (+) as the mean momentum transported
by the molecules which collide the surface dA per unit time, and C (−) as the mean
Ideal systems in equilibrium. 8

momentum transported by the molecules which have collided the surface dA per
unit time. So
dF = C (+) − C (−) (3.26)
This expression can also applied to evaluate the mean force which the gas on one
side of a imaginary surface dividing the gas exerts on the gas on the other side.
Now, C (+) can be obtained by multiplying the flux Φ(v) by dA and the molecular
momentum mv, and integrating over all the velocities with vz > 0, i.e.
Z
(+)
C = dA d3 vf (v)vz (mv) (3.27)
vz >0

Analogously, C (−) can be obtained as

Z
(−)
C = dA d3 vf (v)|vz |(mv) (3.28)
vz <0

since now we have to integrate over the molecules with normal component of the
velocity vz < 0. As |vz | = −vz for vz < 0, by substitution in Eq. (3.26) we get that
Z Z
(+) (−) 3
dF = C − C = dA d vf (v)vz (mv) + dA d3 vf (v)vz (mv)
vz >0 vz <0
Z
= dA d3 vf (v)vz (mv) (3.29)

This expression is valid for any distribution function, even if the gas is not in equi-
librium. If the gas is in equilibrium, the distribution function of velocities depends
only on |v|. So, by symmetry
Z
dFx = dAm d3 vf (v)vz vx = 0 (3.30)

and analogously dFy = 0. This result expresses the fact that there can be no mean
tangential force on the wall in equilibrium. On the other hand, the normal force
does not vanish Z
dFz = dAm d3 vf (v)vz2 = pdA (3.31)

Thus, the mean pressure is given by

1 2
p = nmvz2 = nmv 2 = nK (3.32)
3 3
Ideal systems in equilibrium. 9

µ
θ

φ
x
Figure 3.3: Relative orientation of the dipolar moment µ with respect to the magnetic
induction field B.

where we used that v 2 = vx2 + vy2 + vz2 = 3vz2 and that the mean molecular kinetic
energy K = mv 2 /2. Again, this expression is completely valid for any distribu-
tion function of velocities which depends only on the molecular speed |v|. For the
Maxwellian distribution function of velocities, v 2 = 3kB T /m and then

p = nkB T (3.33)

which is the equation of the state of the ideal gas.

3.5 Classical paramagnetism

Finally, we will discuss another class of ideal systems, which are paramagnetic mate-
rials. The magnetic behaviour of materials can be classified into five major groups:
diamagnetism, paramagnetism, ferromagnetism, antiferromagnetism and ferrimag-
netism. In the first two cases the material has a magnetic response only under
the application of an external magnetic field. On the other hand, the other three
cases show magnetic properties even in absence of magnetic field. Diamagnetic and
Ideal systems in equilibrium. 10

paramagnetic systems can be treated as ideal systems, while in the other cases inter-
actions between atoms are important to understand the emergence of the magnetic
order. However, the Bohr-van Leeuwen theorem prevents the emergence of magnetic
behaviour (i.e. magnetization) in the framework of Classical Statistical Physics (not
in Quantum Statistical Physics). Nevertheless, it is possible to study paramagnetism
in Classical Statistical Physics if we assume the existence of permament magnetic
dipoles associated to the atoms of the paramagnetic system (which can be explained
by quantum physics). Under the action of an applied magnetic induction field B,
the magnetic dipoles have an energy given by
N
X N
X
H=− µi · B = −µB cos θi (3.34)
i=1 i=1

where µi is the magnetic dipolar moment of atom i (all of them with modulus µ),
and θi is the angle formed between µi and B (see Fig. 3.3). In this expression, we
are neglecting the interactions between magnetic dipoles, which is a good approxi-
mation for paramagnetic systems. We can study this system in the framework of the
canonical ensemble, so the probability of the magnetic dipole of an atom i to have
an orientation centered around the spherical angles θ and φ within a differential
solid angle dΩ = sin θdθdφ is given by
ρi (θ, φ)dΩ ∝ eβµB cos θ sin θdθdφ (3.35)
Note that this expression is independent of i, so all the atoms have the same prob-
ability distribution.
The average of cos θ can be obtained as
R 2π R π
dφ 0 dθ sin θ cos θeβµB cos θ
R
dΩ cos θeβµB cos θ 0
cos θ = R = R 2π R π (3.36)
dΩeβµB cos θ dφ 0 dθ sin θeβµB cos θ
0

Performing the integral of φ, we get a 2π-factor up and down, which cancel each
other. On the other hand, we can do a change of variable x = cos θ, so
R1
dxxeαx 2 cosh α
− 2 sinh α
1
cos θ = R−11
2
= α 2 sinh α α = coth α − = L(α) (3.37)
dxe αx
α
α
−1

where α = βµB and L(α) is the Langevin function. So, the magnetization M can
be obtained as
N
µX

M= cos φi sin θi ux + sin φi sin θi uy + cos θi uz (3.38)
V i=1
Ideal systems in equilibrium. 11

1.0

0.8

0.6
L(α)
0.4

0.2

0.0
0 5 10 15
α=βµΒ

Figure 3.4: The Langevin function.

Since cos φi sin θi = sin φi sin θi = 0, and cos θi is independent of i, we get that

M = uz nµL(α) ≡ M uz (3.39)

and the average magnetic energy is

E = −N µBL(α) (3.40)

Note that α is the ratio between the magnetic energy scale, which is µB, and the
thermal energy scale, which is kB T . For small temperatures, α  1 and L(α) ≈ 1.
So, M = nµ, i.e. all the magnetic dipoles are aligned along the B direction. On the
other hand, for large temperatures, α  1 and L(α) ≈ α/3. So,
nµ2 B nµ2 µ0 H H
M= ≈ =C (3.41)
3kB T 3kB T T
where we related B and the magnetic field H by the relation B ≈ µ0 H, where µ0 is
the vacuum magnetic permeability. The expression Eq. (3.41) is known as Curie’s
law, with a Curie’s constant C = nµ2 µ0 /(3kB ).
This system has an energy spectrum bounded from above by N µB. This means
that the magnetic system can have negative temperatures. Since the temperature
appears in the combination α = βµB, it is possible to generate negative temperature
states by instantaneously reversing the magnetic field (which is equivalent to change
sign to T for a fixed B). This is the population inversion procedure, in which the
Ideal systems in equilibrium. 12

higher energy states will be more populated that the lower energy ones. However,
this new state is not stable, since the magnetic system is coupled to other systems
with positive temperatures, so after some time the system will reach a common
equilibrium characterized by a positive temperature. Nevertheless, in some cases
the coupling is so weak that this state can be observed experimentally.

3.6 Bibliography
1. F. Reif., Fundamentals of Statistical and Thermal Physics (Waveland Pr. Inc.,
2008).

2. J. Javier Brey Abalo, J. de la Rubia Pacheco y J. de la Rubia Sánchez,

Mecánica Estadı́stica (UNED, 2011).