Chapter 14

Ideal Bose gas

In this chapter, we shall study the thermodynamic properties of a gas of non-interacting

bosons. We will show that the symmetrization of the wavefunction due to the indis-
tinguishability of particles has important consequences on the behavior of the system.
The most important consequence of the quantum mechanical symmetrization is the Bose-
Einstein condensation, which is in this sense a special phase transition as it occurs in a
system of non-interacting particles. We shall consider, as an example, a gas of photons
and a gas of phonons.

14.1 Equation of state

We consider a gas of non-interacting bosons in a volume V at temperature T and chemical
potential . The system is allowed to interchange particles and energy with the surround-
ings. The appropriate ensemble to treat this many-body system is the grand canonical
ensemble.

Our bosons are non-relativistic particles with spin s, whose one-particle energies (~k)
include only the kinetic energy term and are given by

~ 2k2
h
(k) = (k) = .
2m
Since the lowest one-particle energy 0 is zero,

0 = (0) = 0,

the chemical potential must fulfill the condition:

< < 0,

as we discussed in chapter 12.

We consider a situation when the gas is in a box with volume V = Lx Ly Lz and subject
to periodic boundary conditions. In the thermodynamical limit (N , V , with

177
178 CHAPTER 14. IDEAL BOSE GAS

n= N V
= const), the sums over the wavevector ~k can be replaced by integrals as in the
case of the Fermi gas. However, here we have to be careful when happens to approach
the value 0. In order to see what kind of trouble we then might get into, let us calculate
the ground state occupation. The expectation value of the occupation number is given by
1
<n
r >= f+ (r ) = .
e(r ) 1
For << 1, the occupation of the lowest energy state (0) = 0 can be approximated
as
1 1 1
<n 0 ms >= = ,
e 1 1 + . . . 1
which means that < n 0 ms > can assume any macroscopic value. On the other hand, the
density of states (E), which has the same expression as for fermionic systems (eq. 13.8),

(E) E 0 as E 0.
P R
This is where we are going to encounter a problem: if we replace r by dE (E), we
will get that the ground state has zero weight even though, as we have just shown, it can
be macroscopically occupied. In the case of a fermionic system, we would not have this
problem due to the Pauli principle: < n r >: 0 < n r > 1.

In order to avoid the problem in the bosonic system, we treat the ground state separately.
The grand canonical potential (T, V, z) (do not confuse it with the DOS (E)) can be
written then as
Z
V 2 (k)


(T, V, z) = (2s + 1) 4 dk k ln 1 ze
(2)3 0
+ (2s + 1) ln(1 z) . (14.1)
| {z }
term corresponding to
the ground state (0) = 0

With the substitution r

x=h
k
2m
and the definition of the thermal de Broglie wavelength
s
2h2
= ,
m
expression (14.1) becomes
Z  
2s + 1 4V 2
(T, V, z) = dx x2 ln 1 zex + (2s + 1) ln(1 z). (14.2)
3 0

We use the Taylor series expansion of the ln,

X yn
ln(1 y) = (|y| < 1),
n=1
n
14.1. EQUATION OF STATE 179

for the integral:

Z X

2

x2
 zn
dx x ln 1 ze = .
0 4 n=1 n5/2
We define the functions
Z
4
2

x2
 X zn
g5/2 (z) = dx x ln 1 ze =
0 n=1
n5/2
and
X zn
d
g3/2 (z) = z g5/2 (z) = 3/2
.
dz n=1
n

With these definitions, the grand canonical potential (eq. (14.2) takes the form

2s + 1
(T, V, z) = V g5/2 (z) + (2s + 1) ln(1 z) . (14.3)
3

Note that except for the extra term on the right-hand side, it has the same form as
the expression for the Fermi gas.

From = P V , we get

2s + 1 2s + 1
P = 3
g5/2 (z) ln(1 z) . (14.4)
V

We express the fugacity z in terms of the density of particles n:

>  
<N 2s + 1 2s + 1 z
n= =z P = 3
g3/2 (z) + (14.5)
V z T,V V 1z

and then, by combining eq. (14.4) and (14.5), we obtain the thermal equation of state.

The term
2s + 1 z
n0 =
V 1z
in eq. (14.5) describes the contribution of the ground state to the particle density n as
can be seen from
1 1 1 1 z n0
0 ms >=
<n = ,
V V z 1
1 V 1z 2s + 1
with
ms = s, s + 1, . . . , +s.
n0 can be macroscopically large, which corresponds to the Bose-Einstein condensation.
180 CHAPTER 14. IDEAL BOSE GAS

The caloric equation of state can be obtained by calculating the internal energy U :
   

U = ln Z(T, z, V ) = (T, V, z)
z,V z,V

By considering eq. (14.3), one obtains

3 2s + 1
U = kB T V g5/2 (z) . (14.6)
2 3
This expression is the same as the one for the Fermi gas when f5/2 (z) is substituted by
g5/2 (z). Combining eq. (14.6) and eq. (14.5), one can derive the caloric equation of state.
Now, we combine (14.4) and eq. (14.6) to get

3 3
U = P V + kB T (2s + 1) ln(1 z) .
2 2

Note that the expression for the Bose gas has an extra term
3
kB T (2s + 1) ln(1 z),
2
which was absent in the case of the Fermi gas as well as in the case of the ideal gas
of particles.

14.2 Classical limit

We first investigate the classical limit (non-degenerate Bose gas) corresponding to low
particle densities and high temperatures. In this limit,

z = e << 1

so that
1
<n
r >= zer << 1
z 1 er
1
and the Bose-Einstein distribution function reduces to the Maxwell-Boltzmann one. All
energy levels are only weakly occupied. The probability of double occupancy is practically
zero. It is therefore to be expected that in this limit the differences between Bose-, Fermi-
and Boltzmann statistics are negligible.

As z << 1, it is sufficient to retain only the first two terms of the series for g5/2 (z) and
g3/2 (z):

z2
g5/2 (z) z + ,
25/2
z2
g3/2 (z) z + 3/2 .
2
14.2. CLASSICAL LIMIT 181

With that, the particle density takes the following form:

2s + 1  z  2s + 1 z
n z 1 + +
3 23/2 V 1z
2s + 1  z  2s + 1
z 1 + 3/2 + z(1 z).
3 2 V
In the zeroth approximation, we thus have that
 
3 (0) 3
n (2s + 1)z 1+
V

and therefore
n3
z (0) 3

.
(2s + 1) 1 + V

As in the case of the Fermi gas, here the classical limit corresponds to n3 << 1, i.e.,
small particle densities and small thermal de Broglie wavelengths. The condition
3
<< 1
V
allows us to neglect the correcting term coming from the explicit consideration of the
ground state contribution so that we can write

n3
z (0) ,
2s + 1

which is the same result as we had for fermions. The equation of state reduces then to
the equation of state for classical particles:
2s + 1 n3
P = ,
3 2s + 1
VP = <N > kB T.

In order to obtain the next order corrections, we make use of the equation
2s + 1  z 
n z 1 +
3 23/2
where we neglect the term corresponding to the ground state contribution, which can be
rewritten in terms of z (0) as
 
n3 (0) (1) z (1)
z z 1 + 3/2
2s + 1 2
and then as
z (0)
z (1)  . (14.7)
z (1)
1+ 23/2
182 CHAPTER 14. IDEAL BOSE GAS

In the right-hand side of eq. (14.7), we approximate z (1) by z (0) :

 
(1) z (0) (0) z (0)
z (0)
z 1 3/2 .
1 + z3/2 2
2

We substitute this expression in the pressure equation for the ideal Bose gas,
2s + 1 2s + 1 (1)
P 3
g5/2 (z) z ,
3
where the ground state contribution has been neglected. This gives us
 
n3
P V =< N > kB T 1 .
4 2(2s + 1)

The last term in this expression are the quantum corrections.

Note that if we consider now, for comparison, the equation of state for a real gas
(the van der Waals equation, for instance), we would observe that similar additive
corrections appear with respect to the ideal case due to the interaction between
particles. Here, however, the additive terms originate from the indistinguishability
principle and not from the interaction among particles.

The correcting term for the ideal Fermi gas was positive and therefore contributed as a
repulsion among particles. For the Bose gas, the additive term is negative and therefore
contributes as an attraction among particles.
Quantitatively, the quantum corrections are much smaller than terms coming from the
interaction among particles.

14.3 Bose-Einstein condensation

We consider now the limit of high particle densities and low temperatures (quantum limit),
where one finds important qualitative differences between bosons, fermions and classical
particles.

For bosons, we have the restriction

< 0,

which determines the range of value that z assumes:

0 < z < 1.

The functions
X zn
g (z) =
n=1
n
14.3. BOSE-EINSTEIN CONDENSATION 183

are related to one another through the recursive formula:

d
g1 (z) = z g (z).
dz
In the interval 0 z 1, they are positive and monotonically growing functions of z (see
Fig. 14.1).

Figure 14.1: g family.

For z = 1, g coincide with the Riemann function , defined as

Z
X 1 1 y n1
(n) = = dy .
p=1
pn (1 21n )(n) 0 ey + 1

Thus,
 
5
g5/2 (1) = = 1.342,
2
 
3
g3/2 (1) = = 2.612
2
and g1/2 (z) diverges at z = 1. From the expression for the particle density for bosons n,
2s + 1 2s + 1 z
n= 3
g3/2 (z) + ,
V z1
we have that the particle density of the (2s + 1)-degenerate ground state (k = 0) = 0,
defined as
2s + 1 z
n0 = ,
V z1
can be written in the following form:
2s + 1
n0 = n g3/2 (z) .
3
Since, as can be observed in Fig. 14.1, g3/2 (z) for 0 < z 1 varies in the interval
[0, 2.612], there should be temperatures T and densities n where
2s + 1
n> g3/2 (1).
3
184 CHAPTER 14. IDEAL BOSE GAS

In this case, n0 > 0, that is, a finite fraction of bosons occupies the ground state. When
|| 0, n0 becomes macroscopically large. This phenomenon is called Bose-Einstein
condensation.

This macroscopic state would not be that spectacular if it occured only for temperatures
kB T < 1 (k = 0) = 1 , (14.8)
where 1 is the first excited state. For macroscopic systems, this condition is fulfilled for
temperatures T < 1020 K so that at T = 0 all particles are in the ground state.

What is so spectacular about the Bose-Einstein condensation is that it sets in at much

higher temperatures than those imposed by eq. (14.8). In fact, the phase transition to
the condensed state occurs when
n3 = (2s + 1)g3/2 (1),
which defines the critical value for the de Broglie wavelength c for a given particle density
n:  3/2
3 2s + 1 2h2
c = g3/2 (1) =
n mkB Tc
and hence the critical temperature Tc at which the phase transition occurs:
 2/3
2h2 n
kB T c = .
m (2s + 1)g3/2 (1)

For a given temperature T , we can find the critical density nc :

 3/2
2s + 1 mkB T
nc (T ) = g3/2 (1) = (2s + 1) g3/2 (1) .
3 2h2

Since the particle density is proportional to the inverse the average interparticle distance
|~r| to the third power,
1
n 3,
|~r|
we can argue that the Bose-Einstein condensation occurs when the thermal de Broglie
wavelength approaches |~r|:
|~r|.

Comparison of the critical temperature Tc of the ideal Bose gas with the Fermi
temperature TF of the ideal Fermi gas
Consider s = 1/2 fermions with mass mf and s = 0 bosons with mass mb , both system
with equal densities nf = nb = n. The TF /Tc is then given as

TF 1 mb mb
= (3 2 g3/2 (1))2/3 1.45 .
Tc 4 mf mf
14.3. BOSE-EINSTEIN CONDENSATION 185

mb
For mf
1, both temperatures are of the same order of magnitude.

Let us consider now a gas of conduction electrons and a gas of 4 He atoms as representative
fermionic and bosonic systems. In this case,
mb
8 103 ,
mf

Tf 104 K and therefore

Tc 3.13 K.
This value of Tc is much bigger that what one would expect for a normal occupation of
the ground state, T 1020 K. In this example, we demonstrate that the Bose-Einstein
condensation is a phase transition.

Fugacity
In order to understand the phenomenon, let us represent n,

2s + 1 2s + 1 z
n= 3
g3/2 (z) + ,
V 1z

graphically. For that, we plot the function y1 (z)

3 z
y1 (z) = g3/2 (z) +
V (1 z)

and the constant

n3
y2 =
2s + 1
on the same graph. The intersection between these two curves provides z as a function of
T and n.

With this graphical representation, we can obtain z(T, N ) at a given V .

186 CHAPTER 14. IDEAL BOSE GAS

The solution zb corresponds to the condensate region since there

 
n3
> g3/2 (1).
2s + 1

In the thermodynamic limit (N , V , n = const) the term coming from the

ground state contribution
3 z
V (1 z)
has to remain finite. This implies that V (1 z) has to remain finite, i.e., 1 z has to
behave as V1 . Therefore, for V , we can approximate the fugacity as

n3 n3
solution of = g3/2 (z) for < g3/2 (1),
z= 2s + 1 2s + 1

1 n3
for g3/2 (1).
2s + 1
Then, since
2s + 1
n0 = n g3/2 (z),
3

n0 n3
n 0 if < g3/2 (1),


2s + 1
3
 3/2
n 0 2s + 1 c T n3
n 1 n3 g3/2 (1) = 1 3 = 1 T if g3/2 (1).


c 2s + 1
For 0 < T < Tc , we have a mixture of two phases (see Fig. 14.2): the condensate phase
formed by the macroscopic fraction of N0 bosons out of N occupying the lowest energy
level: "  3/2 #
T
N0 = N 1
Tc

and the gas phase formed by the rest of the particles N1 which are in excited (~k 6= 0)
states:  3/2
T
N1 = N N0 = N .
Tc
14.4. THERMODYNAMICS FOR THE IDEAL BOSE GAS 187

Figure 14.2: Occupation density of the ground state of the ideal Bose gas as a function
of temperature.

At T = 0,
N0 = N, N1 = 0.
The Bose-Einstein condensation takes place at T = Tc .

14.4 Thermodynamics for the ideal Bose gas

With the Bose-Einstein condensation and the concept of a mixed state of gas and con-
densate, we have a strong analogy with the gas-liquid phase transition that we discussed
when studying properties of a real gas. As in that case, the abrupt change of n0 at Tc
leads to discontinuities in the thermodynamic potentials of the ideal Bose gas, which have
different analytical expressions above and below the critical point (Tc , nc ).

Thermal equation of state

We consider the thermodymanic limit (N , V , n = const). In the equation for
pressure for the Bose gas given as
2s + 1 2s + 1
P = 3
g5/2 (z) ln(1 z),
V
the second term approaches zero:
2s + 1 V
ln(1 z) 0
V
for z < 1, i.e., n < nc . For the condensation region (n nc ), where z 1, the above
limit is not trivial. Here we make use of the fact that
1
(1 z) ,
V
as discussed previously, and also get
2s + 1 V
ln(1 z) 0.
V
188 CHAPTER 14. IDEAL BOSE GAS

Summarizing, we get the following equation for pressure:

2s + 1
3 g5/2 (z) for n < nc ,


P =


2s + 1
g5/2 (1) for n > nc .
3
In the condensation region, the pressure is independent of the volume and particle
number; it only depends on temperature. This was also the case in the gas-liquid
phase transition. Considering this analogy, the Bose-Einstein condensation can be
described as a first-order phase transition.

The Bose-Einstein condensation was predicted by Satyendra Bose and Albert Einstein
in 1924-1925. It took almost 70 years to have an experimental corroboration of this
phenomenon with the ultracold gas systems. Previous experiments had been done with
4
He as well as with hydrogen.