Chapter 7

Quantum Gases

7.1 THE GRAND CANONICAL ENSEMBLE

The calculations done in Chapter 4 should make it clear that the canonical ensemble, in which the energy
of the system is not fixed exactly, is much more convenient to use than the microcanonical ensemble. The
canonical ensemble describes a system that is in thermal contact with a reservoir and can therefore exchange
energy with it. The average energy of the system is determined by the reservoir temperature. Although the
system described by the canonical ensemble does not have an exactly predictable energy, it does have an
exact and definite number of particles. We will now introduce another ensemble that is designed to describe
a system that can exchange both energy and particles with a much larger reservoir. It is called the grand
canonical ensemble. In the grand canonical ensemble, both the number of particles and the energy are allowed
to fluctuate, their average values being determined by the given values of the affinity and temperature.
In order to derive the probability distribution appropriate for such a situation, we will consider a system,
like that shown in Fig. 7.1, composed of a small but macroscopic sample volume and a much larger particle
reservoir. Both can exchange energy with a still larger heat reservoir at inverse temperature β. The system,
composed of the sample and the particle reservoir (but not the heat reservoir),

sample particle
reservoir

heat reservoir

Fig. 7.1 A sample and a much larger reservoir that can exchange energy and
particles. The two parts of ‘the system’ can exchange energy (but not particles) with
a heat reservoir.

can be described by a canonical ensemble. It contains a total of NT particles. We introduce a more compact
symbol, zi , for the six-dimensional phase-space coordinates (ri , pi ) of the ith particle. We assume that the
particles in the sample volume do not interact with the particles in the reservoir volume. Thus, if particles
1 through N are in the sample volume while particles N + 1 through NT are in the reservoir volume, then
QUANTUM GASES 7.1 THE GRAND CANONICAL ENSEMBLE 139

the energy can be written as

E = HS (z1 , . . . , zN ) + HR (zN +1 , . . . , zNT ) (7.1)

where HS is the Hamiltonian function for the sample particles and HR is that for the reservoir. Now, using
a canonical probability density for the system, we can easily calculate the probability of finding particles
numbered 1 through N in the sample volume at phase-space points, (z1 , . . . , zN ), with the other NT − N
particles being somewhere in the reservoir. It is [see Eq. (4.56)]
∫ /
P̃N (z1 , . . . , zN ) = e−βHS (z1 ...zN ) e−βHR d 6zN +1 · · · d 6zNT h3NT NT !ZT (7.2)
R

where ZT is the partition function of the system. The probability of finding any N particles at phase space
points z1 to zN within the sample and all other particles in the particle reservoir is obtained by multiplying
this by the number of ways of choosing N particles from NT particles, that is, by the binomial coefficient.
( ) ∫ /
NT −βHS (z1 ...zN )
PN (z1 , . . . , zN ) = e e−βHR d 6zN +1 · · · d 6zNT h3NT NT !ZT
N
∫ R
/ (7.3)
= e−βHS (z1 ...zN ) e−βHR d 6zN +1 · · · d 6zNT h3NT N !(NT − N )!ZT
R

The integral over the reservoir phase space can be written in terms of the canonical potential of the reservoir
with NT − N particles, which can then be expanded to first order in N . According to Eq. (4.55),
∫
e−βHR d6(NT −N ) z = h3(NT −N ) (NT − N )! exp[ϕR (NT − N, β, VR )]
R (7.4)
3(NT −N )
≈h (NT − N )! exp[ϕR (NT , β, VR )] exp(−αN )

Using this in the equation for PN (z1 , . . . , zN ), and lumping together all factors that are independent of N into
one normalization constant that we will call Λ−1 , we get the desired grand canonical probability distribution

e−αN −βHS (z1 ,...,zN )

PN (z1 , . . . , zN ) = (7.5)
Λh3N N !

Thus, if a system can exchange particles and energy with a reservoir at affinity α and coldness β, then, at
equilibrium, the probability of finding the system to contain exactly N particles with positions (r1 , . . . , rN )
and momenta (p1 , . . . , pN ) is given by

e−αN −βH(r1 ,p1 ,...,rN ,pN )

PN (r1 , p1 , . . . , rN , pN ) = (7.6)
Λh3N N !

The normalization constant Λ, called the grand partition function, must be chosen so that the integration
and sum of PN over all possible states of the system is one. This gives
∞
∑ ∫
e−αN
Λ(α, β, V ) = e−βH(1,...,N ) d3N x d3N p (7.7)
h3N N !
N =0

The sum on N has been extended from what would be its natural limit, NT , to infinity by using the fact
that, for very large N , the N ! in the denominator makes the terms diminish very rapidly. We will see that
the probability of obtaining a value of N that is much larger than the average value of N is negligibly small.
An equivalent calculation for a quantum mechanical system shows that the probability of finding the
system with exactly N particles and in the N -particle quantum state ψK (r1 , . . . , rN ) of energy EN K is

PN K = Λ−1 e−αN −βEN K (7.8)

QUANTUM GASES 7.2 SURFACE ADSORPTION 140

The normalization constant Λ must then be chosen so that the sum of PN K over all N and K is unity.
∑∑
Λ= e−αN −βEN K (7.9)
N K

It can be shown that, as h̄ → 0, the quantum mechanical expression for Λ reduces to the classical formula
given in Eq. (7.7).
As expected (see Problem 7.7), the logarithm of the grand partition function gives the grand potential
in rational units.
log Λ(α, β, V ) = ψ(α, β, V ) = S o − αN − βE (7.10)
Using the thermodynamic identity S o = αN + βE + γV , ψ can be written in the simpler form

log Λ = ψ(α, β, V ) = βpV (7.11)

The probability of finding the system with exactly N particles, irrespective of the quantum state, is
∑ /
PN = e−αN −βEN K Λ (7.12)
K

This shows that the average number of particles is given by the expected thermodynamic relation.
∑∑
⟨N ⟩ = Λ−1 N e−αN −βEN K
N K
∂ ∑ ∑ −αN −βEN K
= −Λ−1 e
∂α
N K
∂Λ/∂α (7.13)
=−
Λ
∂ log Λ
=−
∂α
∂ψ
=−
∂α

In an almost identical way, one can show that

∂ψ
⟨E⟩ = − (7.14)
∂β

Fig. 7.2 A surface containing adsorption sites that can each accept a single particle.

7.2 SURFACE ADSORPTION

The surfaces of real vessels containing gases do not have the convenient but artificial properties of the hard-
wall or periodic boxes we have been using in our calculations. Even when cleaned and polished, they are
usually covered with adsorption sites, at which particles of the gas may temporarily attach themselves, due
to interatomic forces between the gas molecules and the surface atoms (see Fig. 7.2). The disappearance of
particles onto the surface or their reappearance in the gas as the temperature is raised causes a change in the
QUANTUM GASES 7.2 SURFACE ADSORPTION 141

apparent properties of the gas, which must be corrected for in order to interpret experimental measurements
in terms of the true properties of the gas alone.
When investigating the phenomenon of adsorption, it is convenient to use a grand canonical ensemble,
since the number of adsorbed particles is not fixed. The model that will be used is a surface containing K
independent adsorption sites. Each site may be unoccupied or it may be occupied by a single gas particle,
which may be in any of a finite number of discrete quantum states of energies ε1 , ε2 , . . .. In this calculation,
the system is only the set of adsorbed particles. The particles in the gas act as the reservoir and do not
enter explicitly into the calculation. In summing over all states of the surface, we need only sum over the
occupation number (0 or 1) at each site, and, for the occupied sites, over the set of energy states.
As usual, the most efficient procedure is to calculate the normalization constant Λ, which will then yield
the thermodynamic potential ψ, from which other physical properties can be calculated. The easiest way of
deriving the correct expression for Λ is to consider first the trivial cases of a single adsorption site and two
adsorption sites. If K = 1, then N can be either 0 or 1, and the sum over quantum states only appears if
N = 1.
∑1 ∑ ∑
Λ= e−αN e−βεk = 1 + e−α e−βεk (7.15)
N =0 k k

For K = 2 there are two adsorption sites and therefore two occupation numbers, N1 and N2 . Again, the
sum over quantum states at a particular site only occurs when that occupation number is not zero.

∑
1 ∑
1 ∑∑
Λ= e−αN1 −αN2 e−βεk1 −βεk2
N1 =0 N2 =0 k1 k2
(∑
1 ∑ )( ∑
1 ∑ )
= e−αN1 e−βεk1 e−αN2 e−βεk2 (7.16)
N1 =0 k1 N2 =0 k2
( ∑ )2
= 1 + e−α e−βεk
k

For general K,
( ∑ )K
Λ = 1 + e−α e−βεk (7.17)
k
∑
The sum za = exp(−βεk ) is just the canonical partition function for a single occupied adsorption site.
Using the grand potential,
ψ(α, β) = K log(1 + e−α za ) (7.18)
it is easy to calculate the fraction of occupied adsorption sites, fa = Na /K.

∂ e−α za
fa = − log(1 + e−α za ) = (7.19)
∂α 1 + e−α za

This can be solved for α as a function of fa .

(1 − f )
a
α = log + log za (7.20)
fa

It will be assumed that the adsorbing surface is in equilibrium with an ideal gas at pressure p. For any ideal
gas, not necessarily monatomic, Eq. 5.46 says that

∂α(p, T ) 1
=− (7.21)
∂p p

which can be integrated to give

α = − log p + log g(T ) (7.22)
QUANTUM GASES 7.3 QUANTUM IDEAL GASES 142

where the function g(T ) cannot be determined by the ideal gas equation alone and is different for different
ideal gases. Eliminating α between Eqs. (7.20) and (7.22) gives a relation between the gas pressure and the
adsorption fraction.
(g ) (1 − f )
a
log = log + log za (7.23)
p fa
which can easily be solved for fa as a function of p and T .
p
fa = (7.24)
p + g/za

It is clear that, for fixed T , the adsorption site occupation probability goes from 0 to 1 as the pressure goes
from 0 to infinity.
In order to proceed further, one must determine the functions za (T ) and g(T ). Let us consider the
simplest case by assuming that the gas is a monatomic gas and that there is only one possible quantum
state at each adsorption site. We assume that the state has a negative energy, ε = −kTo , indicating a bound
state. Then
za (T ) = eTo /T (7.25)
The affinity of the gas can be calculated as a function of the pressure and temperature by first computing
the partition function. ∫
1 ∑ 2
−β pn /2m 3N
Z= e d r d3N p
N !h3N
( )N (7.26)
1 V
=
N ! λ3
where λ is the thermal de Broglie wavelength. The canonical potential of the gas is
( )
ϕ = N log V − log N + 1 − log λ3 (7.27)

and the affinity is equal to ∂ϕ/∂N .

α = log(V /N ) − log λ3
[ ] (7.28)
= − log p + log (kT )5/2 (2πm)3/2 /h3

This gives
g(T ) = (kT )5/2 (2πm)3/2 /h3 ≡ γT 5/2 (7.29)
and
p
fa = (7.30)
p+ γT 5/2 e−To /T
For fixed p, fa goes from 1 to 0 as T goes from 0 to infinity.

7.3 QUANTUM IDEAL GASES

The distribution functions for Bose–Einstein and Fermi–Dirac ideal gases were derived in Chapter 2 by the
method of determining the most probable macrostate. It will be instructive to apply the grand canonical
ensemble to a study of the same systems. For both bosons and fermions, the many-particle quantum states
are defined by a sequence of occupation numbers N1 ,N2 , . . ., where Nk is the number of particles in the kth
single-particle quantum state. The total number of particles and the total energy of the state are given by
∑
N= Nk (7.31)
k

and ∑
E= εk Nk (7.32)
k
QUANTUM GASES 7.3 QUANTUM IDEAL GASES 143

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1.0
Fig. 7.3 The fraction of occupied adsorption sites as a function of the binding energy
(in K), To , for the case of argon at STP. The vertical axis gives fa and the horizontal
axis is To /TSTP .

In using the grand canonical ensemble, the first thing one should do is calculate the grand partition
function Λ. From Λ, the grand potential, ψ = log Λ, can be easily calculated. Then the thermodynamic
relations N = −∂ψ/∂α, E = −∂ψ/∂β, and βp = ∂ψ/∂V will give the equations of state of the system.
According to Eq. (7.9), Λ is equal to the sum of exp(−αN − βE) over all values of N and, for each value
of N , over all N -particle quantum states of the system. Expressing N and E by means of Eqs. (7.31) and
(7.32), we can write Λ as
∑∞ ∑ ( ∑ ∑ )
Λ= exp −α Nk − β εk Nk
N =0 {Nk }N k k
∞ ( ∑ ) (7.33)
∑ ∑
= exp − (α + βεk )Nk
N =0 {Nk }N k

where {Nk }N indicates a sum over all all combinations of occupation numbers that add up to N . But
summing over all values of N and, for each value of N , summing over all N -particle quantum states is
equivalent to summing independently over all single-particle occupation numbers. Therefore, writing {Nk }
for the infinite sequence of occupation numbers {N1 , N2 , . . .}, we get
∑ ( ∑ )
Λ= exp − (α + βεk )Nk
{Nk } k
∑ ∑ ∑ (7.34)
= e−(α+βε1 )N1 e−(α+βε2 )N2 e−(α+βε3 )N3 . . .
N1 N2 N3

and, because the logarithm of a product is equal to the sum of the logarithms of its factors
∑ (∑ )
ψ(α, β, V ) = log Λ = log e−(α+βεk )Nk (7.35)
k Nk

For bosons, Nk = 0, 1, 2, . . ., which means that the sum over Nk in Eq. (7.35) is of the form
∑ 1
xNk =
1−x
Nk

with x = exp[−(α + βεk )]. That is

∞
∑ 1
e−(α+βεk )Nk = (7.36)
Nk =0
1− e−(α+βεk )
QUANTUM GASES 7.5 THE IDEAL FERMI GAS AT LOW TEMPERATURE 144

and therefore ∑ [ ]
ψ(α, β, V ) = − log 1 − e−(α+βεk ) (7.37)
k

For fermions, the sum has only two terms, giving

∑
1
e−(α+βεk )Nk = 1 + e−(α+βεk ) (7.38)
Nk =0

and ∑ [ ]
ψ(α, β, V ) = log 1 + e−(α+βεk ) (7.39)
k

Both cases can be treated simultaneously by the formula

∑ [ ]
ψ=ζ log 1 + ζe−(α+βεk ) (7.40)
k

where {
+1, for FD
ζ= (7.41)
−1, for BE
N and E are then given, in terms of α and β, by the thermodynamic relations

∂ψ ∑ e−(α+βεk ) ∑ 1
N =− = −(α+βε
= (7.42)
∂α 1 + ζe k ) eα+βεk + ζ
k k

and
∂ψ ∑ εk e−(α+βεk ) ∑ εk
E=− = = (7.43)
∂β 1 + ζe−(α+βεk )
k
eα+βεk + ζ
k

These equations are in agreement with the Fermi–Dirac and Bose–Einstein distribution functions obtained
in Chapter 2. The average number of particles that occupy any single-particle quantum state of energy ε is

1 1
fBE (ε) = or fFD (ε) = (7.44)
eα+βε − 1 eα+βε + 1

for bosons and fermions, respectively.

7.4 BOLTZMANN STATISTICS

If the average occupation number of each single-particle quantum state is much less than one, then eα+βε ≫ 1,
and both the Bose–Einstein and the Fermi–Dirac density formulas can be approximated by the Maxwell–
Boltzmann formula
fMB (ε) = e−α−βε (7.45)
When this is done, one says that one is using Boltzmann statistics. Because fFD < fMB < fBE , quantities
calculated using this approximation usually are intermediate between those for bosons and fermions at the
same values of α and β.

7.5 THE IDEAL FERMI GAS AT LOW TEMPERATURE

In Chapter 2 it was shown that the quantum mechanical distribution functions reduced to their classical
equivalents at sufficiently high temperatures. Therefore, the specifically quantum mechanical aspects of
quantum gases are revealed by studying their low-temperature behavior. This must be done separately for
Fermi–Dirac and Bose–Einstein systems, because their low-temperature properties are very different. In
the next two sections the low-temperature properties of a system of nonrelativistic spin- 12 fermions will be
analyzed. Following this, we will consider an equivalent system of bosons.
QUANTUM GASES 7.5 THE IDEAL FERMI GAS AT LOW TEMPERATURE 145

It is assumed that the particles are in a cubic periodic box of side L. It was shown in Section 2.15
that the single-particle energy eigenstates are then plane waves with allowed momentum values that form a
cubic lattice of spacing h/L in momentum space, where h = 2πh̄. The energy of a particle with momentum
p is εp = p2 /2m. Each allowed momentum eigenstate can be occupied by two particles of opposite spin.
Therefore, the sum in Eq. (7.39) for the grand potential ψ can be converted to an integral of the form
∫
log[1 + e−(α+βp
2
ψ = 2(L/h)3 /2m)
]d3 p (7.46)

√ √
By converting to an energy variable, ε = p2 /2m, and using the facts that p = 2mε, dp = dε 2m/ε/2, and
L3 = V , this integral can be transformed as follows.
∫ ∞
8πV
log[1 + e−(α+βp
2
/2m)
ψ= 3 ]p2 dp
h
∫ o
∞ (7.47)
= 2γV log[1 + e−(α+βε) ]ε1/2 dε
o

where the constant γ = 2π(2m/h2 )3/2 . The same transformation of Eqs. (7.42) and (7.43) gives
∫ ∞
ε1/2 dε
N = 2γV (7.48)
o eα+βε +1

and ∫ ∞
ε3/2 dε
E = 2γV (7.49)
o eα+βε + 1
The logarithmic function in Eq. (7.47) can be eliminated by a partial integration. If that is done, and the
thermodynamic relation ψ = βpV is used, one finds that the classical relationship between pressure and
energy density, p = 23 (E/V ), is also valid for a quantum mechanical ideal gas (see Problem 7.8).
In analyzing the low-temperature properties of the Fermi–Dirac gas, it is best to change variables from
the affinity α to the chemical potential µ = −ατ , where τ = kT . The integrals for N and E, in terms of µ
and τ , are
∫ ∞
N ε1/2 dε
= 2γ (7.50)
V o e(ε−µ)/τ + 1
and ∫ ∞
E ε3/2 dε
= 2γ (7.51)
V o e(ε−µ)/τ +1
These integrals are both of the general form
∫ ∞
g(ε) dε
I= (7.52)
o e(ε−µ)/τ +1

where g(ε) is a smooth function that is independent of τ . Therefore, we will present a general method of
evaluating integrals of this type as a power series expansion in the temperature. It is easy to verify the
following limit formula for the Fermi–Dirac function fFD (ε).
( ) {
1 1, if ε < µ
lim = (7.53)
τ →0 e(ε−µ)/τ + 1 0, if ε > µ

This has a simple physical interpretation. The Fermi–Dirac function gives the probability that a single-
particle state of energy ε is occupied. At zero temperature, the N -particle system goes to a state in which
the N single-particle states of lowest energy are occupied while all higher states are unoccupied. Thus at
zero temperature the chemical potential has a value equal to the cutoff energy separating the occupied states
QUANTUM GASES 7.5 THE IDEAL FERMI GAS AT LOW TEMPERATURE 146

T=0

1
T>0

ε
εF
Fig. 7.4 The average occupation of a Fermi–Dirac single-particle state of energy ε at
zero temperature and at some small but nonzero temperature.

from the unoccupied ones. This is called the Fermi energy and is written εF . The Fermi energy is defined
by the relation [see Eq. (7.50)]
∫ εF
N 3/2
n= = 2γ ε1/2 dε = 34 γεF (7.54)
V o

which gives εF = (3n/4γ)2/3 . At finite temperature fFD (ε) deviates from its zero-temperature limit, as
shown in Fig. 7.4.
In the Mathematical Appendix, the following low-temperature expansion is derived for the general
Fermi–Dirac integral shown in Eq. (7.52).
∫ ∞ ∫ µ
g(ε) dε π2 ′ 7π 4 ′′′
(ε−µ)/τ
= g(ε)dε + g (µ)τ 2 + g (µ)τ 4 + · · · (7.55)
o e +1 o 6 360

where the primes indicate derivatives with respect to ε, evaluated at ε = µ. Using this expansion to order
τ 2 in Eq. (7.50) gives the following relation for the particle density, n = N/V .

3n π2 τ 2
≈ µ3/2 + 1/2 (7.56)
4γ 8µ

We can determine the dependence of the chemical potential µ on the temperature τ for small τ as follows.
In the second term, which is of order τ 2 , we will make a negligible error in replacing µ by εF . The left-hand
3/2
side is equal to εF . Therefore
3/2 π2 τ 2
εF ≈ µ3/2 + 1/2 (7.57)
8εF
3/2
Dividing through by εF , solving for µ/εF , and expanding the solution as a power series in τ 2 gives

µ ( π 2 τ 2 )2/3 π2 τ 2
≈ 1− 2 ≈1− (7.58)
εF 8εF 12ε2F

or
π2 τ 2
µ ≈ εF − (7.59)
12εF
which shows that, to first order in τ , the chemical potential is independent of temperature for fixed density.
The reason is that, in going from zero temperature to some low temperature, the change in the Fermi–Dirac
distribution function is an odd function of ε − µ. Thus for fixed µ the same number of particles are added
to the region ε > µ as are removed from the region ε < µ (see Fig. 7.4).
The expansion of the energy equation to second order in τ is

E π 2 1/2 2
= 25 µ5/2 + µ τ (7.60)
2γV 4
QUANTUM GASES 7.7 THE IDEAL BOSE GAS AT LOW TEMPERATURE 147

The energy to order τ 2 at fixed n is obtained by using µ = εF − π 2 τ 2 /12εF in this equation and retaining
terms of order τ 2 . Using the definition of εF , the result can be expressed in the form

E π2 τ 2
= 35 εF + (7.61)
N 4εF

The pressure as a function of temperature and density is then given by the relation p = 23 (E/V ).
The dimensionless expansion parameter in these expressions is clearly τ /εF . If τ /εF ≫ 1, then the ideal
Fermi gas behaves as a classical ideal gas. In that case it is said to be nondegenerate. For τ /εF ≪ 1 the gas
is called a degenerate Fermi gas, and the exclusion principle is the most important factor determining the
properties of the system.

7.6 THE ELECTRON GAS

A metallic crystal is distinguished from other types of crystals by the existence of a large system of conduction
electrons within it. The number of conduction electrons is equal to the number of atoms in the sample times
the valence of the atomic species (1 for sodium, potassium, and copper, 2 for zinc, etc.). The conduction
electrons occupy single-particle quantum states in what is called the conduction band. Semiconductors also
contain conduction electrons, but these are either electrons that have been thermally excited from the valence
band to the conduction band (in which case the material is called an intrinsic semiconductor) or they are
electrons that are associated with donor impurities (in a doped semiconductor). In either case, the number
of conduction electrons is much smaller than it is for a typical metal. At zero temperature the states in the
conduction band are filled (two electrons per state, with opposite spins) up to the Fermi energy, εF . Each
state in the conduction band is also an eigenstate of crystal momentum with eigenvalue p = h̄k, where k is
the wave vector of the state. For some metals the relationship between the energy and momentum of a state
in the conduction band can be approximated by that for a free particle.

h̄2 k 2
εk = (7.62)
2m∗
where the effective mass m∗ is a parameter that is of the order of an electron mass but is not equal to me .
We can get an estimate of the contribution made by the conduction electrons to the specific heat of a
typical metal by taking m∗ = me and V /N = a3 , where a = 4 Å is the diameter of a sodium atom. Using
these values to calculate γ and εF , we find that εF ≈ 1 eV. This energy corresponds to a temperature
of about 10,000 K. Therefore, at room temperature, the expansion parameter τ /εF ≈ 0.03, which justifies
neglecting all but the first nonvanishing term in the expansion for E/N . The specific heat per particle at
room temperature is then given by

∂(E/N ) π2 k2 T
Cel = = ≈ 0.15 k (7.63)
∂T 2εF

where k is Boltzmann’s constant. The Debye temperature of sodium is about 158 K and, therefore, at room
temperature we can use the Dulong–Petit law, which states that the vibrational specific heat per particle is
equal to 3 k.
Cvib = 3 k (7.64)
Comparing these equations shows that at room temperature the electronic motion makes only a slight
contribution to the specific heat of a metal. However, at temperatures lower than the Debye temperature of
the metal, the vibrational specific heat goes to zero as T 3 while the electronic specific heat is proportional
to T . Thus at sufficiently low temperature the electronic specific heat dominates.

7.7 THE IDEAL BOSE GAS AT LOW TEMPERATURE

As T → 0 the Pauli exclusion principle becomes the most important factor determining the properties of
a system of fermions. Since bosons do not satisfy an exclusion principle, their physical characteristics at
low temperature will obviously be very different from those of fermions. In this section we will analyze the
low-temperature properties of a gas of noninteracting spinless bosons. The basic equations for N and E in
QUANTUM GASES 7.7 THE IDEAL BOSE GAS AT LOW TEMPERATURE 148

terms of α and β can be taken from the fermion case by dropping the factor of 2 that is due to the spin
degeneracy, and making a change of sign in the Fermi–Dirac distribution function. When this is done, the
equation for N [Eq. (7.50)] becomes
∫ ∞
N ε1/2 dε
n= =γ (7.65)
V o eα+βε −1

The temperature dependence of this equation can be made explicit by transforming the variable of integration
to x = βε. ∫ ∞ 1/2
3/2 x dx
n = γτ (7.66)
o eα+x − 1
The affinity is limited to the range 0 ≤ α < ∞. If α is chosen to be negative, then fBE (ε) would be negative
at small ε, which is physically meaningless, because it is the average occupation of a single-particle state.
It is easy to see that n(α, τ ) is a monotonically decreasing function of α (at each value of x, the integrand
decreases if α is increased). The maximum value, which occurs at α = 0, is
∫ ∞
x1/2 dx
nc (τ ) = γτ 3/2 (7.67)
o ex − 1

nc is called the critical density. This integral is of a type that can be evaluated in terms of the gamma
function Γ(x) and the Riemann zeta function ζ(x). The general formula is
∫ ∞ ν
x dx
= Γ(ν + 1)ζ(ν + 1) (7.68)
o ex − 1

The maximum particle density is therefore given by

( ) ( )
nc (τ ) = Γ 32 ζ 32 γτ 3/2 (7.69)

But this result is clearly absurd! Since the particles do not interact, there is no physical mechanism preventing
one from constructing a system with a density greater than nc . A mistake has been made somewhere. The
mistake is not hard to find. The lowest-energy single-particle state is the zero-momentum state (which has
zero energy). According to the Bose–Einstein distribution, the number of particles in that state at affinity
α is
1
Np=0 = α → ∞ as α → 0 (7.70)
e −1
Thus, in making the transformation from a sum over discrete momentum states to an integral, the zero-
momentum state must be singled out for special treatment. In Problem 7.22, the reader is asked to show
that this is sufficient to fix the difficulty; it is not necessary to take into account, individually, the other very
low-energy states neighboring the zero-momentum state. Therefore, Eq. (7.66) must be modified to read
∫ ∞
1 3/2 x1/2 dx
n= + γτ (7.71)
V (eα − 1) o eα+x − 1

With this modification, for any value of the density n, there is a corresponding value of the affinity α that
satisfies this equation. When n is larger than the critical density, then α is of order 1/N and, in the integral
part of Eq. (7.71), α may be set equal to zero. In this case, a finite fraction of all of the particles in the
system occupy the discrete zero-momentum quantum state.
The collection of particles in the zero momentum state is called the Bose–Einstein condensate. It
exists whenever n > nc (τ ). The number of particles in the condensate is just the difference between the
total number of particles and nc V . Since the condensate, when it exists, accounts for a finite fraction of
the particles in the system, the gas with the condensate has macroscopically different properties than the
gas without it. The change from one state to the other is a thermodynamic phase transition, called Bose
condensation. The phase diagram is shown in Fig. 7.5. The phase without the condensate (that is, n < nc )
is called phase I, while the phase that contains a condensate (n > nc ) is called phase II.
QUANTUM GASES 7.7 THE IDEAL BOSE GAS AT LOW TEMPERATURE 149

20

15

Phase II
10

5 Phase I

0
0 1 2 3 4
Fig. 7.5 The phase diagram for the ideal Bose gas. The vertical axis gives n/γ and
the horizontal axis is τ = kT . Recall that γ = 2π(2m/h2 )3/2 .

In phase I the function α(n, τ ) must be determined by numerically solving the Eq. (7.71) in the form
∫ ∞
x1/2 dx n
= (7.72)
o eα+x − 1 γτ 3/2

In phase II the function α(n, τ ) has the simple form (see Problem 7.24)

1
α(n, τ ) = Phase II (7.73)
V [n − nc (τ )]

For phase II, in any integral, the value of α may be replaced by zero, since it is of order 1/N . The condensate
particles all have zero energy; therefore, the condensate makes no contribution to the energy and so the
integral expression for E can be used in both phases.
∫ ∞ ∫ ∞
ε3/2 dε x3/2 dx
E(n, τ, V ) = γV = γV τ 5/2 (7.74)
o e α+βε −1 o eα+x − 1

where α is understood to mean α(n, τ ). In the condensed phase, α ≈ 0, and thus, the energy becomes
independent of n.
∫ ∞
5/2 x3/2 dx ( ) ( )
E = γV τ = Γ 52 ζ 52 γV τ 5/2 Phase II (7.75)
o e −1
x

The constant volume specific heat (in rational units) in Phase II is

dE ( ) ( )
CV = = 52 Γ 52 ζ 52 γV τ 3/2 Phase II (7.76)
dτ
The entropy is the integral of dQ/τ .
∫ τ
o CV dτ ( ) ( )
S = = 53 Γ 52 ζ 25 γV τ 3/2 Phase II (7.77)
o τ

For a given density, the transition from one phase to the other occurs at a critical temperature τc that is
given by solving Eq. 7.69 for τ .
[ ( ) ( ) ]2/3
τc = n/Γ 32 ζ 32 γ (7.78)
In terms of τc , the specific heat per particle can be written in the simple form
( )
CV 15 ζ 52 ( τ )3/2
= ( ) Phase II (7.79)
N 4 ζ 32 τc
QUANTUM GASES 7.8 FLUCTUATIONS IN THE IDEAL BOSE GAS⋆ 150

2.0
CV
1.5

1.0

0.5

0
0 0.5 1.0 1.5
Fig. 7.6 The constant volume specific heat per particle as a function of τ /τc . In
rational units the specific heat is dimensionless.

In Phase I, E(n, τ ) and CV (n, τ ) must be calculated from Eq. (7.74) using the function α(n, τ ). A graph
of CV /N as a function of τ /τc is given in Fig. 7.6. Notice that Bose condensation is a very subtle phase
transition. Across the phase transition line the density and the energy per particle are continuous. Even the
specific heat is continuous. One must go to dCV /dτ in order to find a quantity that jumps discontinuously
at the phase transition.

7.8 FLUCTUATIONS IN THE IDEAL BOSE GAS⋆

A study of fluctuations in the ideal Bose gas below the critical temperature yields some surprising results.
In phase II, α = 0, and therefore, the fundamental relation for S o becomes
E + pV
S o = βE + βpV = (7.80)
τ
Noninteracting bosons also satisfy the ideal gas relation pV = 23 E (see Problem 7.8), which gives

S o = 53 E/τ (7.81)
Equation (7.75) gives τ as a function of the energy density E/V , which can be used in Eq. (7.81) to write
S o in terms of E and V .
5 ( ( 5 ) ( 5 ) )2/5 2/5 3/5
So = Γ 2 ζ 2 γ V E (7.82)
3
We notice that this entropy function does not contain the variable N . This is reasonable, because α =
(∂S o /∂N )E = 0 in phase II. If more particles are added to the system, they all go into the condensate,
leaving E and S o unchanged. We will now see that this has disastrous consequences for the thermodynamic
properties of the system. Letting so , n, and ε be the entropy, number of particles, and energy per unit
volume, Eq. (7.82) can be written
so (n, ε) = 35 Aε3/5 (7.83)
[ ( 5 ) ( 5 ) ]2/5
where A = Γ 2 ζ 2 γ . The fact that so is not a function of n means that, in the situation depicted in
Fig. 7.7, the condition of maximum entropy would not determine the equilibrium density in the two halves.

Fig. 7.7 Two volumes containing an ideal Bose gas are connected by a pinhole.
QUANTUM GASES 7.10 THE STATISTICS OF COMPOSITE PARTICLES 151

This peculiar phenomenon is easy to understand. Because the condensate particles all have zero
momentum, they make no contribution to the pressure in a container. Thus, if the condensate densities
in the left and right sides were unequal, this would not create a pressure difference that would bring the
system back to equal densities. Because of this characteristic, the ideal Bose gas cannot be considered as
an acceptable model of any physical system. An example of the bad features of the model is that, in the
presence of a weak gravitational field in the z direction, if the temperature is below the critical temperature,
all of the condensate, which contains a finite fraction of the particles in the system, becomes smeared against
the bottom of the vessel in a single quantum state that has a scale height of (h̄2 /m2 g)1/3 . For the earth’s
gravitational field, which is weak for thermodynamic purposes, this scale height is about one micrometer (see
Problem 7.28). Also, if the periodic boundary conditions are replaced by hard-wall boundary conditions,
then a macroscopic change would take place in the particle density throughout the container, not only near
the walls (see Problem 7.29). This extreme sensitivity of the particle density to the boundary conditions is
never exhibited by real systems.

7.9 THE MEAN-FIELD INTERACTION⋆

In order to obtain a reasonable model of a system of Bose–Einstein particles, some interaction between the
particles must be introduced. Even a very weak interaction eliminates the bizarre properties of the model. In
order to construct a model that has normal physical properties without seriously complicating the analysis,
we will introduce an interaction potential, v(r), with the following characteristics:
1. v(r) is a steadily decreasing positive function. Therefore, it creates purely repulsive interparticle forces.
2. The range of v(r) is much ∫larger than the average distance between particles in the system but the
∞
volume integral of v(r), 4π o v(r)r2 dr = σ, is finite.
An interaction potential with these characteristics, called a mean-field interaction, can be constructed
by starting with any positive, decreasing potential function V (r) that has a finite volume integral σ and
reducing its amplitude while increasing its range by scaling it according to the formula

v(r) = ϵ3 V (r/ϵ) (7.84)

where ϵ is a dimensionless adjustable parameter. The integral of v(r) is then independent of ϵ, but its range
gets larger and larger as ϵ is made smaller and smaller.
For very small values of ϵ the interaction between any two particles is negligible but a particle finds
itself in a very smooth potential field due to the combined interactions of thousands of particles in its large
“interaction volume.” If the density of particles is uniform, then that potential field will be constant. But
adding a constant to the Hamiltonian will not change the energy eigenfunctions of the system. Thus the
energy eigenstates for the uniform ideal Bose–Einstein gas are still energy eigenstates of the system, but
their energies have been shifted by an amount that depends upon the particle density. Being careful not to
double count the interaction energies, one can easily evaluate the total energy of a state with occupation
numbers Np .
∑ p2
E= Np + 12 σnN ≡ EK + 21 σnN (7.85)
p
2m

where the sum is over all the momentum eigenstates of the system, σ is the volume integral of v(r), and
n = N/V . With fixed N , a shift in all energy levels has no effect on their occupations. Therefore, EK and
S o , as functions of T , are unaffected by the shift. Thus we can use Eq. (7.83) for S o with E replaced by EK .
5
S o (N, E, V ) = AV 2/5 (E − 21 σn2 V )3/5 (7.86)
3
The entropy per unit volume is then
5
so (n, ε) = A(ε − 21 σn2 )3/5 (7.87)
3
The reader will be asked to show (see Problem 7.30) that this entropy function is strictly convex and therefore
leads to reasonable predictions for macroscopic properties. (Note that, because EK is always positive, ε has
the range 21 σn2 < ε < ∞.)
QUANTUM GASES 7.11 SUPERFLUIDITY AND LIQUID HELIUM 152

7.10 THE STATISTICS OF COMPOSITE PARTICLES

The elementary particles of nature are either bosons or fermions. Those that contribute to the structure
of atoms, namely protons, neutrons, and electrons, are all spin- 12 Fermi–Dirac particles. The atom, when
treated as a single particle, inherits its statistics from the elementary particles that compose it. If it contains
an even number of elementary fermions, then it is itself a Bose–Einstein particle; if it contains an odd
number, it is a Fermi–Dirac particle. The reasoning behind this rule is easily understood by considering a
simple example. A hydrogen molecule can be treated either as a system composed of two hydrogen atoms
or as a system of two protons and two electrons. In the first case, it has a wave function ϕ(QA , QB ), where
QA and QB are each a set of coordinates, sufficient to define the position and state of a hydrogen atom. In
the second case it has a wave function ψ(xA , yA , xB , yB ), where xA and xB are coordinates appropriate to
the two protons and yA and yB are each space and spin coordinates for an electron. The transformation

ψ(xA , yA , xB , yB ) −→ ψ(xB , yB , xA , yA ) (7.88)

which describes an exchange in position of the protons and the electrons, is equivalent to the combination
of the two separate transformations

ψ(xA , yA , xB , yB ) −→ ψ(xB , yA , xA , yB ) −→ ψ(xB , yB , xA , yA ) (7.89)

But, because protons and electrons are Fermi–Dirac particles, each of the separate transformations multiplies
the wave function ψ by −1. Thus, the combined transformation leaves ψ unchanged. When this fact is
expressed in atomic coordinates it takes the form

ϕ(Qa , QB ) = ϕ(QB , QA ) (7.90)

This line of reasoning clearly leads to the conclusion that the exchange of the full sets of coordinates for two
identical atoms multiplies the wave function by +1 or −1 depending upon whether the atoms contain an
even or odd number of elementary fermions.

7.11 SUPERFLUIDITY AND LIQUID HELIUM

There is no real system that is well approximated by the ideal Bose gas. Photons are Bose–Einstein particles,
and they are noninteracting to a very high degree, but the fact that the number of photons is not a conserved
quantity has a strong effect on the properties of the photon gas. In particular, it eliminates the phase
transition associated with Bose condensation, which is the most striking characteristic of the ideal Bose gas.
The photon gas will be discussed in a later section.
Liquid helium is the system that is closest (although not very close) to the ideal Bose gas. Helium
exists in two isotopic forms, 3 He and 4 He. The two isotopes, although they are chemically almost identical,
have very different physical properties at low temperatures. The differences in their physical characteristics
are not due to the difference in their masses, which are approximately in a ratio of 3 to 4. Their most
fundamental difference is that the 3 He atom, composed of 5 fermions (2 protons, 1 neutron, and 2 electrons),
is itself a Fermi–Dirac particle, while the 4 He atom is a Bose–Einstein particle. We will restrict our attention
to the heavier isotope, which incidentally constitutes the bulk of naturally occurring helium, the lighter
isotope contributing less than one part per million. The phase diagram of bulk 4 He is shown in Fig. 7.8.
It differs substantially from that of other inert elements, such as neon and argon. It has no triple point, in
the ordinary sense of having liquid, solid, and gas simultaneously in equilibrium, because the liquid phase
extends down to zero temperature. The λ line separates two different liquid phases. (The isotopes of helium
are the only pure substances that have more than one liquid phase.) The phase transition separating the
two liquid phases, called the lambda transition, derives its name from the appearance of the specific heat
singularity that occurs at the transition. As can be seen in Fig. 7.8, the specific heat curve resembles the
Greek letter λ. The phase transition curve in the p -T plane is called the λ line and is written Tλ (p). If no
pressure is specified, then Tλ will always mean its lower endpoint, where the λ line meets the liquid–vapor
phase transition curve. Extremely careful measurements have shown that, very close to Tλ (p), along the
whole length of the λ transition line, Cp can be accurately represented by a function of the form
√
Cp = B − At−α (1 + D t) for T > Tλ (p) (7.91)
QUANTUM GASES 7.11 SUPERFLUIDITY AND LIQUID HELIUM 153

p
B
3.0
solid A
30 C
2.0
20

He II He I 1.0
10

gas
1 2 3 4 5 6 1.0 2.0 3.0
T(K) T(K)

Fig. 7.8 (A) The phase diagram for 4 He. At zero temperature, it requires about
25 atmospheres of pressure to solidify helium. (B) The specific heat of liquid helium in
equilibrium with its own vapor (in J/g-K).

and √
Cp = B ′ − A′ t−α (1 + D′ t) for T < Tλ (p) (7.92)
where t = |T − Tλ |/Tλ is a dimensionless temperature variable. The parameters A, A′ , B, B ′ , D, and D′
have values that vary substantially along the λ transition line, but their ratios, A/A′ , B/B ′ , and D/D′ , are
independent of position on the λ line. The values of these universal amplitude ratios and the value of the
critical exponent α (all given in Table 7.1) can be calculated using the renormalization group theory of critical
phase transitions, which will be discussed in a later chapter. There is, at present, no way of calculating the
individual amplitudes, A(p), B(p), and D(p). They must be taken from experiment.

Table 7.1
Specific heat parameters for liquid He*
α A/A′ B/B ′ D/D′
−.016 1.07 1.00 1.03
* These values are “best estimates.” There is still substantial
experimental and theoretical uncertainty in them.

Although liquid helium is a system of strongly interacting particles, and therefore has properties very
different from the ideal Bose gas, it has one thing in common with the ideal Bose gas; it is clear that the
λ transition is associated with a Bose–Einstein condensation of the helium atoms into a macroscopically
occupied single-particle quantum state. Liquid helium has another striking property, namely superfluidity,
that the ideal Bose gas, if it existed, would not have. The phenomenon of superfluidity, in which a fluid can
flow through extremely narrow, crooked channels with absolutely no resistance or viscosity, seems to defy
one’s physical intuition. In this section we will explore the basic mechanism that leads to superfluidity.
In trying to elucidate the mechanism of superfluidity, we will consider the properties of the substance
near zero temperature, where the theoretical analysis is simpler than it is close to the λ line. The most
essential aspect of the analysis is the form of the quantum mechanical excitations near the ground state.
According to Eq. (7.85), the total energy and momentum of a Bose–Einstein gas with mean-field interactions
can be written in terms of a set of occupation numbers as
∑
E = Eg + Np εp (7.93)
p

and ∑
P= Np p (7.94)
p
QUANTUM GASES 7.11 SUPERFLUIDITY AND LIQUID HELIUM 154

ε
2

0.5 1.0 1.5 2.0 2.5

p
Fig. 7.9 The relation between energy and momentum for the elementary excitations
in liquid helium near zero temperature. ε is given in units of 10−22 J and p in units of
10−24 kg-m/s. The slope of the straight line is defined as Vc .

where the ground-state energy, Eg , depends on the total density and εp = p2 /2m. Each occupation number
can take the values 0, 1, 2, . . .. Any many-particle system whose total energy and momentum eigenvalues
satisfy formulas of this general form is said to have a Bose–Einstein quasiparticle spectrum. For the ideal
Bose gas the quasiparticles are simply the actual particles in the system, and the prefix “quasi” is therefore
illogical. However, for liquid 4 He close to T = 0, it can be shown, both experimentally and theoretically,
that the energy and momentum eigenstates are given by formulas like Eqs. (7.93) and (7.94) in which the
quasiparticles are quantized sound waves or phonons. These excitations are similar to the phonon excitations
that account for the vibrational specific heat of crystals at low temperatures, with the one difference that, for
a liquid, there are only longitudinally polarized waves. The energy and momentum carried by a phonon are
given in terms of its angular frequency and wave vector by ε = h̄ω and p = h̄k. For very long wavelengths
(very small k) the usual dispersion relation for sound waves, ω = ck, where c is the speed of sound, gives
rise to a linear energy function.
εp = cp (7.95)
As the wavelength decreases, becoming comparable to the average interparticle spacing in the liquid, the
function εp begins to differ substantially from its long-wavelength linear form. The actual energy-momentum
relation for quasiparticle excitations in liquid 4 He is shown in Fig. 7.9. The quasiparticle excitations in the
linear part of the curve are called phonons; those near the local minimum are called rotons. There is no
sharp borderline between phonons and rotons. In Fig. 7.9, a line is drawn whose slope defines what we will
call the critical velocity, Vc . The critical velocity is the largest velocity that has the property that Vc ≤ εp /p
for all quasiparticle excitations.

v=0

Fig. 7.10 A stationary sample of liquid helium at T = 0 fills an infinite pipe that is
initially moving to the right.

To see how the details of the excitation spectrum are associated with the existence of superfluidity, we
consider the system shown in Fig. 7.10. An infinite circular pipe, whose linear mass density is M , is filled
with liquid 4 He in its ground state. The pipe, but not the fluid, is initially moving to the right at a speed
v, less than Vc . We do not assume that the surface of the pipe is perfectly smooth. Rather, we assume
QUANTUM GASES 7.11 SUPERFLUIDITY AND LIQUID HELIUM 155

that it has microscopic irregularities that would, because of the relative motion of the pipe and the fluid, be
expected to transfer energy and momentum from the pipe to the fluid until both are moving with the same
velocity. This would mean that, at some later time, the fluid would no longer be in its ground state but
would be in some excited state of energy E and momentum P . Naturally, the energy and momentum picked
up by the fluid must have come from the pipe, since the total system is assumed to be isolated. The pipe
is assumed to be a heavy, classical system. We will consider what happens within a unit length of the pipe.
Initially, the pipe has an energy M v 2 /2 and the fluid has energy Eg . At the later time, the pipe velocity has
decreased to v − ∆v and the fluid is in an excited state described by occupation numbers {Np }. The energy
and momentum conservation laws demand that
∑
1
2 M v 2
+ Eg = 1
2 M (v − ∆v) 2
+ Eg + Np εp (7.96)
p

and ∑
M v = M (v − ∆v) + Np px (7.97)
p

If we assume that ∆v ̸= 0, then

∑
Np εp = M v ∆v − 21 M (∆v)2 < M v ∆v (7.98)
p

and ∑ ∑
M ∆v = Np px ≤ Np p (7.99)
p p

Using the second inequality to eliminate M ∆v in the first, we get

∑ ∑
Np εp < v Np p (7.100)
p p

or ∑
Np (εp − vp) < 0 (7.101)
p

But this is clearly impossible, since εp − vp > 0 for all p and Np ≥ 0. (Remember that we have assumed that
v is less than Vc .) Thus, the only solution to the energy and momentum conservation laws is the trivial one,
∆v = 0 and Np = 0 for all p. In spite of its rough surface, there is no way that the pipe can transfer energy
and momentum to the fluid without violating the fundamental conservation laws. Therefore, according to
our argument, no energy and momentum will be transferred and the fluid will simply remain stationary,
sitting there inside a rough, moving pipe! It would be very reasonable for the reader to react with extreme
skepticism toward this bizarre conclusion. To overcome some of that skepticism, we describe the following
real experiment.
A toroidal (doughnut-shaped) hollow tube is tightly packed with a fine powder, such as ground glass,
saturated with liquid helium, and sealed (see Fig. 7.11). It is mounted on a fine wire so that it can oscillate
about its axis of symmetry as a torsional pendulum within an evacuated, temperature-controlled chamber.
The resonant frequency of the torsional pendulum is monitored as the temperature is gradually reduced. It
remains constant until the temperature falls below Tλ . Then the oscillation frequency begins to increase
and, near T = 0, reaches the value that it would have if there were no liquid helium present, that is, if only
the container and the ground glass were contributing to the moment of inertia. The process is completely
reversible. If the temperature is raised, the full moment of inertia returns. The only obvious explanation is
that, below Tλ , only a part of the fluid (called the normal fluid component) is participating in the oscillatory
motion, while the rest of it (the superfluid component) avoids moving along with the tube by winding its
way, without resistance, through the narrow convoluted channels between the pieces of ground glass! One
should not attempt to picture the superfluid motion as the ordinary motion of fluid particles. It is something
closer to quantum tunneling on a massive scale. A way to appreciate just how “quantum mechanical” the
superfluid motion is, is to remember that, if a helium atom is in a single-particle quantum state that extends
QUANTUM GASES 7.12 QUANTIZED VORTICES 156

Fig. 7.11 A hollow toroid, filled with packed powder and liquid helium, hangs on a
wire so that it can oscillate about its axis of symmetry as a torsional pendulum.

throughout the free volume within the torus, then that atom is not at any particular place in the torus at
all—it is simultaneously everywhere in that macroscopic volume! But, with Bose–Einstein condensation, a
substantial fraction of the particles are in just such a macroscopically extended state.
Another real experiment, using the same ground glass filled torus, illustrates the wonders of superfluid
motion even more clearly. By rotating the wire, the torus is spun like a top about its symmetry axis. This
is done at a temperature above Tλ . Keeping the rotation constant, the temperature is gradually reduced to
near 0 K. Then a small amount of friction is applied to the wire so that the torus slows down and stops.
Everything looks completely quiescent. Let us call the symmetry axis (the wire axis) the z axis and choose
perpendicular x and y axes in the plane of the torus. Now a small amount of torque is applied about the
x axis. Due to that torque the torus rotates about the y axis, like a spinning top! The torus clearly carries
angular momentum. The fluid inside has never stopped moving and is now flowing freely through the narrow
channels. Remember, the fluid was put into rotation while it was normal (T > Tλ ), kept rotating as it was
cooled to T = 0, and only then was the torus slowed down.

7.12 QUANTIZED VORTICES

There is an important flaw in the theoretical analysis given here that we have to discuss now. The experiments
that we have described all involved tubes filled with ground glass. This was not intended only to make the
phenomenon seem still more spectacular. If the same experiments are done using a hollow tube, filled only
with liquid helium, they will not work as described. In general, when liquid helium near zero temperature
flows through a channel, it will behave as a superfluid as long as the flow velocity is less than some critical
velocity Vc . But the critical velocity will not be equal to the slope of the line drawn in Fig. 7.9. Instead it
will depend on the size of the channel, and for macroscopic channels it will be so small that the superfluid
phenomena will be very difficult to observe. The source of the difficulty is not that the basic logic of the
theoretical analysis is wrong, but rather that we have not taken into account all possible excited states of
the system. The value of the critical velocity is determined by the ratio of the energy of an excitation to
its momentum. There exist other excitations, whose nature we will describe shortly, that have much higher
energies than the phonon or roton modes shown in Fig. 7.9 but that nonetheless have much lower energy-to-
momentum ratios. The creation of those excitations allows energy and momentum to be transferred to the
fluid at a much lower critical velocity than that determined from the phonon–roton spectrum.
To describe the nature of the excited states that we have missed, it is best to consider the case of helium
within a small rotating bucket rather than the case of helium flowing through a tube (Fig. 7.12). If a bucket
containing liquid helium at T = 0 is slowly brought into rotation, the liquid will behave in the following
way. (We are now describing real experiments, not theoretical predictions, although the two things agree.)
If the angular velocity of the bucket is maintained at some small value, the liquid will remain stationary as
the bucket rotates around it. If the angular velocity of the bucket is increased beyond some critical angular
velocity, then the stationary fluid will make an abrupt transition to a state in which there is a vortex line
near the axis of rotation. Actually, there is a certain energy barrier (intermediate states of higher energy)
between the stationary state and the state with the vortex line, so that the experimenter may be required
to tap the experimental apparatus gently in order to help along the transition. But once the line has been
created, further taps will not create new lines nor get rid of the one that is there. If the angular velocity
QUANTUM GASES 7.13 THE TWO-FLUID MODEL 157

Fig. 7.12 A cylindrical bucket of liquid helium can be rotated about its symmetry
axis. Actually, the bucket must be closed on top; otherwise it will empty by the liquid
creeping up and down over the surface of the bucket—a characteristic of liquid helium
that has not been discussed in the text.

of the bucket is below the critical angular velocity no amount of tapping will create a stable vortex line.
The vortex line, running from top to bottom, is a very miniaturized version of the whirlpool created in an
emptying sink or bathtub. The center of it, which is almost entirely empty, has a diameter of only a few
Ångstroms. The fluid flows around the line with a velocity that decreases with distance from the line. The
vorticity, which is defined as the line integral of the local fluid velocity along any path encircling the line, is
exactly equal to h/m, where h is Planck’s constant and m is the mass of a helium atom.
I
h
v · dℓ = (7.102)
m

The existence of this quantum of vorticity is well understood. In fact, it was predicted theoretically long
before it was experimentally observed. It can be derived by assuming that the superfluid velocity is related
to the condensate wave function by the equation mv(r) = h̄∇ϕ, where the macroscopically occupied state
is of the form ψ = Reiϕ (see Problem 7.32).
Returning to the experiment, if the angular velocity is increased further, at a second critical value a
second vortex line appears. The two lines do not coalesce, but remain separated. Vortex lines of vorticity
2h/m are not fundamentally forbidden, but they are energetically unfavorable in comparison with two single
lines. As the angular velocity is increased, more and more quantized vortex lines are created. Of course,
the flow around each vortex line is in the direction of the bucket rotation. Once there are many lines,
they will distribute themselves so that the average fluid velocity (averaged over a volume containing many
vortex lines) will be the same as would exist in rigid body motion. Since an ordinary fluid would rotate
(in a rotating bucket) with rigid body motion, the liquid helium, if not observed very carefully, would seem
to be rotating like an ordinary fluid. Actually, once the first vortex line is created, the macroscopically
occupied single-particle state is no longer a zero-momentum eigenstate. Rather, it is a quantum state with
a macroscopic flow pattern that includes the vortex line. That is, the excited states that we have neglected
in our analysis are not states in which a small vibration has been set up in an otherwise uniform stationary
fluid, but ones in which all the fluid participates in a quantized flow with a velocity that is everywhere
proportional to Planck’s constant. For liquid helium flowing through a pipe, the excited states that are
important in determining the critical flow velocity are states containing quantized vortex rings, as illustrated
in Fig. 7.13. The critical velocity associated with the creation of quantized rings goes to zero as the channel
size increases. However, for very narrow channels, Vc increases to about 50 m/s, a value comparable to the
slope of the line in Fig. 7.9.
QUANTUM GASES 7.13 THE TWO-FLUID MODEL 158

fluid velocity

vortex ring

Fig. 7.13 If liquid helium is flowing through a pipe at a velocity larger than its
critical velocity, it will transfer energy and momentum to the pipe by creating vortex
rings, which are quantized flow patterns in which the singular vortex line is in the form
of a closed ring, similar to a smoke ring (which the reader has probably never seen, since
nobody smokes these days).

7.13 THE TWO-FLUID MODEL

Liquid 4 He, below the λ transition, behaves as if it were composed of two interpenetrating fluids that can
pass through one another without friction. The first, a normal fluid, has properties similar to helium above
the λ transition. That is, it has a small but finite viscosity and interacts normally with the surface of the
container. The second, a superfluid, flows without friction in a flow pattern determined by the phase of the
condensate wave function. The mass density of the helium, which depends only slightly on temperature, is
the sum of the mass densities of the normal and the superfluid components.

ρ = ρn (T ) + ρs (T ) (7.103)

The superfluid and normal fractions are shown in Fig. 7.14 as functions of T at saturated vapor pressure.
Notice that, below 1 K, liquid helium is almost a pure superfluid.
Now we need to clarify a subtle and easily confused point. We called the collection of particles in the
macroscopically occupied single-particle state the condensate. (For liquid helium at rest the condensate state
is simply the zero-momentum state, but in the presence of superfluid flow it is more complicated.) No one
has yet succeeded in measuring the fraction of particles in the condensate, but it is possible to make fairly
reliable theoretical estimates of the condensate fraction. At zero temperature it is estimated that about 10%
of the particles are in the zero-momentum state. But, at the same temperature, all of the particles are in
the superfluid. Clearly, the superfluid is not to be identified with the condensate. At zero temperature the
whole system of N particles becomes locked into a complicated N -particle ground-state wave function that
extends over the full macroscopic volume. If that wave function is expanded in terms of single-particle plane
waves, then the zero-momentum plane wave has an average occupation of about 0.1 N . But it is the complete
system of particles in the N -particle ground state that moves coherently and constitutes the superfluid, not
only those particles in the heavily occupied single-particle state.
For liquid helium, at close to zero temperature, it is not hard to answer the question: “What is the
normal fluid?” The superfluid is then, by definition, everything else. We consider a system composed of
liquid helium, in a very narrow stationary pipe, at equilibrium at some small but nonzero temperature. At
time zero the pipe is instantaneously given some small velocity v that is less than the critical velocity. If
the temperature were zero, then, according to our previous analysis, no change could take place in the fluid.
However, at a finite temperature, thermal excitations already exist within the fluid.
If an excitation of momentum p is scattered, at the wall, into a state of momentum p′ , then the change
in energy of the liquid will be ∆E = εp′ − εp and the ratio of energy change to momentum change will
be (εp′ − εp )/|p′ − p|. This quantity is not necessarily larger than Vc . In fact, it can have any value at
all. For example, if p′ ≈ −p, then the ratio of ∆E to |∆p| will be very small. Therefore the energy and
momentum conservation laws do not prevent the scattering of existing thermal excitations, although they
do still prevent the transfer of momentum from the pipe to the fluid by the creation of new excitations. The
QUANTUM GASES 7.15 THE AFFINITY OF THE QUASIPARTICLE GAS 159

1.0
ρn /ρ

0.5
ρs /ρ

0 T(in K)
0 1.0 2.0
Fig. 7.14 The superfluid and normal fluid fractions, ρs (T )/ρ and ρn (T )/ρ, between
T = 0 and Tλ .

“quasiparticle gas” will absorb energy and momentum from the pipe, and its momentum distribution will
shift until it comes into equilibrium with the pipe. Equilibrium will be established when the average group
velocity of the quasiparticles is equal to the velocity of the pipe. But if the temperature is very low then the
density of thermal excitations will also be very low, and equilibrium can be established with a very small
transfer of energy and momentum from the pipe. Thus the system behaves as if the pipe were filled with a
very low-density normal fluid. That is, the gas of thermal excitations is the normal fluid. The normal fluid
mass density can be defined in terms of the density of momentum carried by the quasiparticle gas ⃗π and the
velocity of the pipe v, which is the same as the average group velocity of the quasiparticles.

⃗π = ρn v (7.104)

A detailed calculation of ρn at low temperature is left to the exercises in the supplement to this chapter.

7.14 THERMODYNAMICS OF HELIUM NEAR T = 0

At low temperatures, the thermodynamic properties of liquid helium, and in particular its specific heat,
can be calculated by treating the system as an ideal gas of Bose–Einstein quasiparticles with the energy–
momentum relation shown in Fig. 7.9. The energies of quantized vortex states are so high that those states
can be completely neglected at low temperatures. Assuming that the system is contained in a cubic box of
side L, the density of allowed quasiparticle momentum values is (L/h)3 = V /h3 . The grand potential of the
system of quasiparticles is given by Eq. (7.40), with ζ = −1 and the sum converted to an integral by using
the momentum state density. ∫
V ( )
ψ=− 3 log 1 − e−α−βεp d3 p (7.105)
h

7.15 THE AFFINITY OF THE QUASIPARTICLE GAS

The greatest difference between the quasiparticle gas and an ordinary ideal Bose–Einstein gas is not the
altered relationship between energy and momentum, but the fact that the number of quasiparticles is not a
conserved quantity. The creation of a quantized sound wave does not change the number of helium atoms
in the system. Thus these quantized vibrations can be freely created and absorbed at the boundary of the
container. If a macroscopic variable is not constrained by a conservation law, then its value at equilibrium is
determined by the maximum entropy principle of thermodynamics. That principle states that, at equilibrium,
∂S
=0 (7.106)
∂N
But, by definition, ∂S/∂N = α, and thus, for a gas of nonconserved particles, such as the quasiparticles
of liquid helium, the affinity is zero at equilibrium. (The same condition pertains to the set of quantized
vibrations in a solid.) With α = 0, the grand potential for the system is
∫ ( )
4πV
ψ=− 3 log 1 − e−βεp p2 dp (7.107)
h
QUANTUM GASES 7.15 THE AFFINITY OF THE QUASIPARTICLE GAS 160

The assumption that the quasiparticles do not interact is only valid when their density is small. As
the temperature approaches Tλ , the density of quasiparticles increases to the point where their interactions
strongly affect the excitation spectrum of the system. Near Tλ the system can no longer be treated as an
ideal quasiparticle gas. Therefore, our present analysis is restricted to a range of temperature of about
0 < T < 2 K. The energy at the roton minimum is εo , where εo = 8.6 K. Thus, below 2 K, only the states
on the linear portion of the phonon curve and the roton states close to the minimum are significant. We
can therefore make the approximation of replacing the single set of quasiparticles, which has a somewhat
complicated spectrum, by two independent sets of quasiparticles. One set, which we will call phonons, has a
purely linear energy–momentum curve with a slope given by the zero-temperature sound velocity. That is,
εp = cp (7.108)
with c = 244 m/s. The other set, the rotons, have a parabolic relationship between energy and momentum,
with the parameters chosen to match the experimental dispersion curve at the roton minimum. This requires
that, for the rotons,
(p − po )2
εp = εo + (7.109)
2µ
where
εo /k = 8.6 K, po = 2.0 × 10−24 kg m/s, and µ = 1.0 × 10−27 kg
The grand potential is the sum of the grand potentials due to the phonons and the rotons.
ψ = ψphon + ψrot (7.110)
The integral for ψphon is easily calculated
∫ ∞
4πV
ψphon = − 3 log(1 − e−βcp )p2 dp
h o
∫ ∞
4πV
=− log(1 − e−x )x2 dx (with x = βcp)
(hβc)3 o
∫ ∞ 3
4πV x dx
= (by partial integration) (7.111)
3(hβc)3 o ex − 1
4πV
= Γ(4)ζ(4) [by Eq. (7.68)]
3(hβc)3
4π 5 V τ 3 ( π4 )
= 3
ζ(4) = and τ = β −1
45 (hc) 90
The roton grand potential is
∫ ∞ ( )
4πV
log 1 − e−βεo e−β(p−po ) /2µ p2 dp
2
ψrot = − (7.112)
h3 o

For 0 < T < 2 K, we have e−βεo < e−4 = 0.018, and therefore one can use the approximation log(1−x) ≈ −x,
which is equivalent to using a Maxwell–Boltzmann distribution for the rotons, rather than a Bose–Einstein
distribution. ∫
4πV −βεo ∞ −β(p−po )2 /2µ 2
ψrot = 3 e e p dp (7.113)
h o
The maximum of the exponential factor in the integral, which occurs at p = po , is unity. If p < 0 and
T < 2 K, then β(p − po )2 /2µ > 144, which means that exp[−β(p − po )2 /2µ] < e−144 . Thus we can extend
the integration to −∞.
∫ ∞
4πV
ψrot = 3 e−εo /τ e−(p−po ) /2µτ p2 dp
2
(τ = kT = 1/β)
h −∞
∫ ∞
4πV
= 3 e−εo /τ e−q /2µτ (q 2 + 2po q + p2o ) dq
2
(q = p − po ) (7.114)
h −∞
4πV √
= 3 2πµτ (p2o + µτ )e−εo /τ
h
QUANTUM GASES 7.17 THE FREQUENCY DISTRIBUTION 161

These formulas for ψphon and ψrot will be used in Problem 7.31 to calculate the specific heat of liquid helium
at low temperatures.

7.16 BLACKBODY RADIATION

If a cavity, whose walls are kept at a finite temperature, is completely evacuated of ordinary matter, its
energy density does not become zero, because it is impossible to eliminate from the cavity the electromagnetic
radiation that is constantly being emitted and absorbed by the walls. This section is devoted to a study
of the physical properties of such cavity radiation. A perfectly black body (one that absorbed all radiation
that was incident upon it) would, at equilibrium within the cavity, have to emit radiation with the same
intensity and frequency distribution as the cavity radiation. Therefore, the type of electromagnetic radiation
found within a cavity at equilibrium is also referred to as blackbody radiation. The electromagnetic energy
within the cavity may be treated as an ideal gas of massless Bose–Einstein particles, called photons. Since
the relation between the momentum and the wave vector, p = h̄k, which is valid for nonrelativistic particles,
is also true for zero mass photons, the allowed momentum states for photons in a periodic box of volume
V = L3 are the same as those for nonrelativistic particles. However, for each allowed momentum state, there
are two possible values of the photon spin, corresponding to the two possible transverse polarization states
of an electromagnetic wave. The energy of a photon of momentum p is

εp = cp (7.115)

Since photons are not a conserved particle species, as are electrons or neutrinos, the argument given in
Section 7.15 applies, and one may conclude that the affinity α is exactly zero. Using the facts that the
density of single-particle quantum states in momentum space is 2(L/h)3 and that α = 0 for the photon gas,
Eq. (7.40) for the grand potential can be written as
∫
2V
ψ(β, V ) = − 3 log[1 − e−βcp ] d3 p (7.116)
h

Except for the factor of 2, this is the integral that is worked out in Eq. (7.111).

8π 5 V τ 3
ψ(β, V ) = (7.117)
45 (hc)3

The pressure and energy density of the radiation gas can now be obtained from the grand potential by
the thermodynamic identities
∂ψ 8π 5 τ 4
p=τ = (7.118)
∂V 45 (hc)3
and
E 1 ∂ψ 8π 5 τ 4
=− = (7.119)
V V ∂β 15 (hc)3
Notice that, because of the modified relationship between the energy and the momentum of the particles,
the photon gas obeys a relation p = 13 E/V , rather than p = 23 E/V .
Equation (7.119) shows that the electromagnetic energy density in a cavity is proportional to the fourth
power of the temperature. This relationship is called Stefan’s law.

7.17 THE FREQUENCY DISTRIBUTION

In order to determine the frequency distribution function of the radiation, we write E/V = −(∂ψ/∂β)/V as
an integral, using Eq. (7.116).
∫
E 2 cpe−βcp 3
= 3 d p
V h 1 − e−βcp
∫ (7.120)
2 cp 3
= 3 d p
h ecp/τ − 1
QUANTUM GASES 7.17 THE FREQUENCY DISTRIBUTION 162

Changing to an angular frequency variable, h̄ω = cp, and using the fact that d3 p → 4πp2 dp = 4π(h̄/c)3 ω 2 dω,
gives ∫ ∞
E h̄ ω3
= 2 3 dω (7.121)
V π c o e h̄ω/τ −1
The frequency distribution function, D(ω, T ) is defined by saying that the energy density of radiation with
angular frequency within the interval dω at temperature T is D(ω, T ) dω. It is clear from Eq. (7.121) that

h̄ ω3
D(ω, T ) = (7.122)
π 2 c3 eh̄ω/kT −1

At any given temperature, the dominant frequency is determined by finding the value of ω that
maximizes D(ω, T ). The equation ∂D/∂ω = 0 gives

eω 2 (h̄/kT )ω 3
− =0 (7.123)
eh̄ω/kT − 1 (eh̄ω/kT − 1)2

which can be written in the form

h̄ω
= 3(1 − e−h̄ω/kT ) (7.124)
kT
The solution of the equation x = 3(1 − e−x ) is x = 2.82144, and therefore at temperature T the maximum
in the frequency distribution function occurs at

kT
ωmax = 2.82144 (7.125)
h̄
That ωmax is directly proportional to T is Wien’s displacement law. These predictions are all in excellent
agreement with experimental results.
QUANTUM GASES PROBLEMS 163

PROBLEMS
7.1 Do the ‘equivalent calculation’ mentioned just prior to Eq. (7.8).

7.2 For a system of noninteracting one-dimensional classical particles in a harmonic oscillator potential:
(a) Calculate the grand potential, ψ(α, β). Note that there is no fixed “volume” for this system. (b) Obtain
N and E as functions of α and β and show that they satisfy the equipartition theorem.

7.3 (a) Using the grand canonical ensemble, verify the fluctuation formulas used in Chapter 5, namely
(∆N )2 = ∂ 2 ψ/∂α2 and 2 2 2
√ (∆E) = ∂ ψ/∂β . (b) Show that, for a macroscopic system, ∆N/N and ∆E/E
are both of order 1/ N and are therefore completely negligible.

7.4 If the adsorbed particles on a surface are free to move across the surface, then they may sometimes
be approximated as a two-dimensional ideal gas in which the energy of a particle of momentum (px , py )
is ε(p) = (p2x + p2y )/2m − εo , where εo is the binding energy of the particle to the surface. Using this
approximation, calculate the surface density of adsorbed particles if the pressure in the three-dimensional
gas, in equilibrium with the adsorbed particles on its surface, is p.

7.5 Diatomic molecules are sometimes decomposed, upon adsorption, into their constituent atoms, attached
to the surface. Consider a gas of diatomic nitrogen (N2 ), and assume that the vibrational excitations can be
ignored, so that Eq. (4.65) is applicable. Assuming that the gas is in equilibrium with K atomic adsorption
sites, derive the equivalent of Eq. (7.24) for the fraction of occupied sites.

7.6 For a surface with K adsorption sites and a single bound state at each site, show that the entropy (in
rational units) of the adsorbed particles can be written in the form
[ ]
S o = −K f log f + (1 − f ) log(1 − f )

where f = N/K is the fraction of sites that are filled.

7.7∑It is clear from Eq. (7.7) that the grand partition function is given by Λ(α, β, V )
= e−αN ZN (β, V ), where ZN is the canonical partition function for a system of N particles. Using
the fact that ZN (β, V ) = exp ϕ(N, β, V ), where ϕ is the canonical potential, approximate the sum by an
integral and use the method described in Section A.9 to show that

ψ ≡ log Λ = ϕ − αN = S o − αN − βE

where N (α, β, V ) is the solution of α = ∂ϕ(N, β, V )/∂N .

7.8 (a) Equation (7.47) gives the grand potential ψ = βpV for a gas of spin- 12 fermions. By a partial
integration, show that pV = 23 E for the Fermi–Dirac ideal gas. (b) Do the same for a gas of spinless bosons.

7.9 If a hydrogen atom is placed in a magnetic field, its nucleus (a proton) can have its spin, and therefore its
magnetic moment, oriented either parallel or antiparallel to the field. The energy levels of the two orientations
are ±µB, where µ = 2.79255 µN . (µN = eh̄/2Mp = 5.05 × 10−27 J/T is the Bohr nuclear magneton.) (a) In
a field of 2 Tesla, at a temperature of 300 K, what is the ratio R(T ) = (N− − N+ )/(N− + N+ ). (N− and N+
are the numbers of parallel and antiparallel spins, respectively.) (b) If a substance containing hydrogen is
placed in a magnetic field and subjected to electromagnetic radiation at a frequency ν = 2µB/h, energy will
be absorbed from the radiation field, causing transitions between the two nuclear magnetic energy states.
The absorption rate is proportional to R(T ). This is the phenomenon of nuclear magnetic resonance. (c) For
a field of 2 T, in what range is the nuclear magnetic resonance frequency (infrared, microwave, etc.)?

7.10 Shown in Fig. 7.15 is an impurity atom in a cubic crystal of lattice constant a. The atom has become
ionized by transferring an electron to one of its six nearest neighbors. Assuming that, in the absence of an
electric field, the six positions are equivalent, determine the average polarization of the ion-electron system
when an electric field E is imposed in the z direction.
QUANTUM GASES PROBLEMS 164

+ - y

Fig. 7.15

7.11 In the grand canonical ensemble, the probability of finding the system with exactly N particles in
some particular state is given by Eq. (7.6). Assume that the system is an ideal monatomic gas in a volume
V , at temperature T . (a) Calculate N̄ , the average number of particles in the system, as a function of the
affinity α. (b) Show that ∆N , the uncertainty in N , satisfies Eq. (5.73). That is, that (∆N )2 = −∂ N̄ /∂α.
(c) For one mole of gas at STP, calculate ∆N/N̄ .

Fig. 7.16

7.12 Consider the system shown in Fig. 7.16. The volume on the left is 43 of the total volume. In the
left-hand side is a Fermi–Dirac gas at zero temperature. In the initial state εF = 1 eV. The right-hand side
is completely empty, and the system is totally isolated. At t = 0 the plug is pulled. What is the temperature
when the system comes to equilibrium?

7.13 If the Fermi energy is much larger than the rest energy of an electron (εF ≫me c2 ), then the relationship
between energy and momentum can be approximated by that for zero-mass particles; that is, E = cp. For
such an ultrarelativistic electron gas, obtain an expansion for the energy and pressure up to order τ 2 .

7.14 Massless particles have the energy–momentum relationship ε = cp. (a) For a system of spinless
one-dimensional Fermi–Dirac massless particles, determine the exact relationship involving N , µ, and τ .
(b) Show that, at τ = 0, µ = εF .

7.15 Neutrinos are zero-mass spin- 21 Fermi–Dirac particles. Although neutrinos have spin- 21 , there is only
one neutrino state for each momentum, because of the fact that the spin angular momentum of a neutrino is
always antiparallel to its momentum vector. Treating the neutrino number as a strictly conserved quantity,
obtain expansions for p and E/N , up to order τ 2 , for a neutrino gas.

7.16 The antiparticle to the neutrino is the antineutrino, a particle with the same properties as the neutrino,
but with its spin angular momentum always parallel to its linear momentum. In fact, the neutrino number
Nν and the antineutrino number Nν are not separately conserved, but only the lepton number, L = Nν − Nν ,
QUANTUM GASES PROBLEMS 165

is conserved. Due to creation of neutrino–antineutrino pairs, the total number of particles, N = Nν + Nν ,

can change spontaneously. A neutrino–antineutrino gas may be treated as two interpenetrating Fermi–Dirac
ideal gases with a chemical reaction, ν + ν ↔ 0. Using the chemical equilibrium equation, calculate the
pressure of such a system as an expansion in τ , separately, for the two cases L = 0 and L > 0. (Note: This
problem requires the solution of Problem 7.15.)

7.17 Calculate the specific heat at low temperature of a two-dimensional Fermi–Dirac ideal gas of spin- 21
particles.

7.18 The baloneyon is an imaginary Fermi–Dirac particle with spin- 21 and the following relationship between
energy and momentum.
E = B|p|4

where B is the baloney constant. What is the Fermi energy of a system of baloneyons as a function of the
particle density?

7.19 Consider a mixture of N α particles and 2N electrons within a volume V . When the average kinetic
energy of the electrons, due to the Pauli Exclusion Principle, is much larger than the binding energy of an
electron bound to an α particle, then the electrons will be stripped off the α particles and the system can be
approximated as an interpenetrating Fermi–Dirac electron gas and a classical gas of α particles, rather than
as a gas of helium atoms, which it would become at lower densities. This situation exists in high-density
stars. (a) Determine the mass density of the system when the average kinetic energy of the electrons (at 0 K)
is ten times the binding energy of a single electron to an α particle in free space. (b) How does it compare
with the estimated mass densities at the center of the sun (105 kg/m3 ) and in a white dwarf (2 × 109 kg/m3 )?
(c) The internal temperature of white dwarf stars is about 107 K. What is the ratio of τ to εF in a white
dwarf? (This ratio determines whether the zero-temperature approximation is accurate.)

7.20 Copper has a density of about 8.9 g/cm3 . Copper has one conduction electron per atom. (a) Calculate
the pressure of the conduction electron gas at 0 K. (b) Calculate the increase in pressure when the temperature
is raised to 300 K. (c) What is holding copper together against the pressures obtained in (a) and (b)?

7.21 The pressure in a finite-density Fermi–Dirac ideal gas does not go to zero as the temperature ap-
proaches zero. Show that the entropy satisfies the limit equations mentioned in Section 5.18; namely
( ∂S ) ∂ ( )
lim = ∞ and lim S = 0
E→Eo ∂V ∂V E→Eo

7.22 In this problem we want to show that, below the Bose–Einstein condensation temperature, it is not
necessary to treat any state except the zero momentum state individually. Consider the thermodynamic
limit, N = nL3 → ∞, with fixed n and β. Show that the number of particles in any fixed finite number of
momentum states that does not include the zero-momentum state, divided by N , approaches zero, even for
α = 0. (Hint: The most heavily occupied nonzero momentum states are the ones next to the zero momentum
state.)

7.23 Show that, in an ideal Bose gas, the Bose–Einstein transition occurs when( the
) number of particles in
a cubic thermal de Broglie wavelength is equal to 2.612. That is, when nλ3 = ζ 23 .

7.24 Using Eqs. (7.68) and (7.71), derive the function α(n, τ ) in phase II of an ideal Bose–Einstein gas.

7.25 For an ideal Bose–Einstein gas, calculate the ratio of the condensate density to the total density as a
function of τ /τc . [τc is defined in Eq. (7.78).]

7.26 (a) For a two-dimensional gas of spin-zero particles in a periodic box of area L2 , calculate the number
of momentum eigenstates with energies within the interval ε to ε+dε. (b) Use the result of (a) to obtain
an integral formula for N/A as a function of α and τ for a 2D Bose–Einstein gas. (c) Show that the
QUANTUM GASES PROBLEMS 166

result obtained in (b) implies that no Bose–Einstein condensation occurs for the two-dimensional ideal
Bose–Einstein gas.

7.27 (a) Consider a system of conserved bosons for which the relationship between energy and momentum
is not ε = p2 /2m, but ε = cp, where p = |p|. For this system, calculate the Bose–Einstein condensation
temperature τc as a function of the particle density. (b) For a system of particles in which ε = cp3 , show
that there is no Bose–Einstein condensation.

7.28 For a particle in a gravitational potential U = mgz, the Schrödinger equation for the ground state is

h̄2 d2 ψ(z)
− + U ψ(z) = Eo ψ(z)
2m dz 2

Assume that a “hard floor” exists at z = 0, so that ψ has the boundary condition ψ(0) = 0. (a) Show
that the ground-state wave function has the form ψ(z) = u(z/s), where s = (h̄2 /m2 g)1/3 and u(x) satisfies
an eigenvalue equation that does not contain the parameters h̄, m, or g. Thus the parameter s gives the
scale height for a quantum particle in a gravitational potential. For an ideal Bose–Einstein gas at zero
temperature, all the particles will go into this single-particle ground state. (b) Evaluate s for the earth’s
gravitational field and m equal to the mass of a 4 He atom.

7.29 (a) Write the ground-state wave function for a single particle of mass m in a hard-walled box of volume
L3 . (b) For an N -particle Bose–Einstein ideal gas in such a hard-walled box, determine the particle density
as a function of position at temperature zero.
7.30⋆ (a) It is a mathematical theorem that a function f (x, y) is strictly convex if and only if ∂ 2 f /∂x2 < 0,
∂ 2 f /∂y 2 < 0, and (∂ 2 f /∂x2 )(∂ 2 f /∂y 2 ) > (∂ 2 f /∂x ∂y)2 . Prove that so (n, ε), given in Eq. (7.87), is a strictly
convex function. (b) Show that the peculiar characteristic of the ideal Bose gas illustrated in Fig. 7.7
disappears for the Bose gas with mean-field interactions.

7.31 Using Eq.s 7.111 and 7.114, separately calculate and plot the contributions to the specific heat,
C = ∂E/∂τ , of liquid helium due to phonons and rotons.

7.32 Assume that the condensate wave function has the form ψ(r) = R exp[iϕ(r)] and that the superfluid
velocity at point r is related to ψ by mv = h̄ ∇ϕ(r). From the single valuedness of ψ, derive the quantization
of vorticity [Eq. (7.102)].

7.33 Calculate the number of photons in a 1 cm cube at 300 K.

7.34 Consider a small area dA on the inside surface of a cavity at temperature T . Show that the
electromagnetic energy that falls upon that area, per second, can be written as I(ω, T ) dA, where the spectral
intensity function I(ω, T ) = (c/4)D(ω, T ). Thus, the frequency distribution function D(ω, T ) also gives the
frequency distribution of the radiation that would pour out of a hole in a cavity at temperature T . This is
not true for the speed distribution for nonrelativistic particles. The speed distribution for particles within a
box (see Problem 2.8) is not the same as the speed distribution of particles exiting from a small hole in the
box. Explain what is the essential characteristic of photons that makes I proportional to D.