Statistical Physics.

HS 2014
Solutions Sheet 6. Prof. Manfred Sigrist

Exercise 1. Ideal bosons in a harmonic trap.

In this exercise we consider an ideal gas of bosons confined in a three-dimensional harmonic

potential. Observe the differences to the cases of classical and fermionic particles in the same
harmonic potential, which we discussed in Sheet 2, Exercise 3 and in Sheet 5, Exercise 1, re-
spectively.

The energy states of the bosons are given by

Ea = ~ω(3/2 + ax + ay + az ) , (1)
including the zero point energy of E0 = 3 ~ω/2. The occupation numbers of the oscillator modes
of the state with energy Ea is given by a = (ax , ay , az ) with ai ∈ {0, 1, 2, ...}.

(a) Consider the high-temperature, low-density limit (z  1). Derive the grand canonical
partition function for the phonons, Zb , and compute the grand potential Ωb . Take into
account also the zero-point energy of the harmonic oscillators. Show that
 
Ωb ∝ g4 ze−βE0 , (2)

where the function gs (z) is defined as

∞
X zl
gs (z) = . (3)
ls
l=1

Solution. We begin with the general definition of the grand canonical partition function within the
occupation number formalism (section 2.5 of the lecture notes) and find
∞ 
Y X  na Y
Zb = ze−βEa = (1 − ze−βEa )−1 (S.1)
a na =0 a

In order to compute the grand potential Ω = −1/β log Z, we use the series expansion
∞
X (−x)`
log(1 + x) = − for − 1 < x ≤ 1 . (S.2)
`
`=1

This expansion is applicable for the logarithm of the partition function in (S.1) as 0 < ze−βEa ≤ 1 (as
µ ≤ Emin ).
With this replacement we can exactly calculate log(Zb ) in the high-temperature limit (β → 0):
∞ ∞ ∞
!3
X −βEa
XX z ` −`βEa X z ` X −`β~ω/2 −`β~ωa
log Zb = − log(1 − ze )= e = e e
a a `=1
` ` a=0
`=1
(S.3)
∞  −`β~ω/2 3 ∞ −βE0 `


X z` e β→0 X ze 1 1  
= −`β~ω
≈ = g4 ze−βE0
` 1−e ` (`β~ω) 3 (β~ω) 3
`=1 `=1

The grand potential is then given by

1 1  
Ωb = − 3
g4 ze−βE0 . (S.4)
β (β~ω)
Notice that, even though the zero-point energy E0 shows up in the expression for Ωb explicitly, it has no
physical consequences. It merely leads to a shift of the chemical potential by E0 .

(b) Derive an expression for the internal energy U and the average particle number hN i. Follow
the same approach as in Sheet 5, Exercise 1 in order to obtain U in terms of N .

1
 Solution. First, we compute the internal energy of the system,

∂(β Ωb ) 
Ub = , (S.5)
∂β z

where the derivative has to be taken at constant fugacity z = eβµ . Starting from (S.4) we find
3 1 
−βE0
 E0 
−βE0

Ub = g4 ze + g3 ze , (S.6)
β (β~ω)3 (β~ω)3
which shows that the internal energy is proportional to the grand potential plus a term representing the
zero-point energy of the harmonic oscillators.
The average particle number can be computed in a similar way,
∂
hNb i = z log Zb . (S.7)
∂z
We have
∂ 1   1  
hNb i = z 3
g4 ze−βE0 = 3
g3 ze−βE0 , (S.8)
∂z (β~ω) (β~ω)
where we used
∂
z g4 (z) = g3 (z). (S.9)
∂z
We immediately see, that the second term in equation (S.6) is just given by hNb iE0 .
In order to simplify the following calculation, we define a renormalized fugacity
z̃ = ze−βE0 (S.10)
Now we want to relate the internal energy to the particle number and start with the high-temperature
expansion of the particle number equation,
z̃ 2
 
1 1
hNb i = g3 (z̃) ≈ z̃ + . (S.11)
(β~ω)3 (β~ω)3 8

The parameter ρ is given by

3
~ωN 1/3

ρ≡ . (S.12)
kB T
The condition z  1 also implies ρ  1. Expanding in ρ allows us to deal with the particle number instead
of the chemical potential. Inverting the series ρ = z + z 2 /8, we find
ρ2
z̃ = ρ − . (S.13)
8
To interpret the condition ρ  1 we note that ~ω is the characteristic energy scale for a particle in
the harmonic potential. Since every particle can carry energy in three degrees of freedom (one for each
quantum number describing the state) the characteristic average energy scale for the N particles is ~ωN 1/3 .1
Therefore, the condition ρ  1 requires that the characteristic energy scale of the system is much smaller
than the thermal energy kB T (high-temperature limit). This means that we consider temperatures at which
the average occupation of the states is much smaller than one (low-density limit).
We write the internal energy up to second order in ρ as
z̃ 2
 
3 1 3 1
Ũ := U − N E0 = g 4 (z̃) = z̃ +
β (β~ω)3 β (β~ω)3 16
2
 
3 1 ρ
= ρ−
β (β~ω)3 16
   3 !
3 2 3 1 ~ω 1
= N − N (β~ω) = 3N kB T 1 − N , (S.14)
β 16 kB T 16

where we recover the equipartition law in leading order and the (negative) first order quantum corrections
∝ N (~ω/kB T )3  1 distinguishing the bosons from the ideal classical gas.

1
The reason for this is that in three-dimensional (configuration-)space the radius of a ball grows with the
volume as V 1/3 . Think of N as the volume.

2
(c) Compute the specific heat CN for constant particle number.
Compute the thermal expansion coefficient α. Use the average square displacement of
the harmonic oscillator reff2 = x2 ha + a + a i in order to define an effective volume
0 x y z
3
Veff = 4π/3 reff . Give an interpretation of Veff .

Solution. Since our system does not really have a volume as thermodynamic variable we have to compute
the specific heat CN by fixing the number of particles. Hence, as a starting point we use the expression
(S.14) for the internal energy, where we can keep N fixed:
   3 !
∂U 1 ~ω
CN = = 3N kB 1 + N . (S.15)
∂T N 8 kB T

Finally, we compute the thermal expansion coefficient α = V −1 (∂V /∂T ) at fixed N . For this we have to
2
define an effective volume of the system by introducing an effective radius reff ≡ hr2 i . From basic quantum
2 2
mechanics we know that reff = x0 hax + ay + az i, where x0 represents the characteristic length scale which
2
we fix to unity. Hence, we can relate reff to the internal energy of the system via
2
~ω reff N = Ũ . (S.16)

Therefore, we find
 3/2
4π 3 4π Ũ
Veff ≡ reff = . (S.17)
3 3 ~ωN
For the thermal expansion coefficient we obtain

∂ Ũ 3/2
 
3
α = Ũ −3/2 = Ũ −1 CN
∂T N 2
 3 !
3 −1 1 ~ω
= Ũ 3kB N 1 + N
2 8 kB T
ρ  
3 1 1+ 8 3 1 3
= ρ = 1 + ρ , (S.18)
2 T 1 − 16 2 T 16

which agrees (in leading order) with the result for a classical gas in the harmonic trap (α = 3/(2T )).

(d) Plot your results for U , C, and α for the classical, the fermionic, and the bosonic case and
note the differences.

Solution. In summary, in Sheet 5, Exercise 1 and in this exercise we have found up to first order in ρ:
 3 !
~ω 1
U = 3N kB T 1 ± N , (S.19)
kB T 16
 3 !
1 ~ω
CN = 3N kB 1 ∓ N , (S.20)
8 kB T
 3 !
31 3 ~ω
α= 1∓ N , (S.21)
2T 16 kB T

where the upper and lower sign corresponds to fermions and bosons, respectively. For the classical case,
the corrections to 1 vanish for all three formulas.
These results as a function of temperature are plotted in Fig. 1; each for the classical, the fermionic, and
the bosonic case. Note that our expansions up to first order in ρ are only valid for ρ  1. We can still
plot these expansions for larger values of ρ (that is, lower temperatures) to observe the trends, keeping in
mind that these results are not exactly valid.
We see that

• In zeroth order in ρ the results for the classical (Boltzmann) gas in a harmonic trap are recovered.

3
 10 1.2

8
1.1

CN @kBTDH3NL
U@kBTDH3NL

6
1.0
4
0.9
2

0 0.8
2 4 6 8 10 2 4 6 8 10
kB T kB T

0.5

0.4
Α@kBTD

0.3

0.2

3 4 5 6 7 8 9 10
kB T

Figure 1: Thermodynamics of fermionic and bosonic gases compared to the classical gas. Note that these quantities
are computed within the high-temperature, low-density approximation and are therefore not exact results. Still,
they can be used to observe trends. The dashed (blue) line is for the fermions, the dotted (red) line for the bosons,
and the continuous (black) line for the classical gas. We set N ~ω = 100.

• Due to quantum corrections the internal energy U for fermions (bosons) is higher (lower) than the
ideal classical gas.
This can be understood by taking quantum statistics into account. Fermions are not allowed to
occupy the same state (Pauli) while bosons tend to gather in the same quantum state. Lowering
the temperature, the system tends to occupy low energy states with growing probability. While the
classical system is not influenced at all by double occupancy, in the fermionic system the double
occupancy is forbidden and occupation of low-energy states is thus reduced, increasing the internal
energy Ub compared to the classical gas. In the bosonic case, the opposite happens: the probability
of occupying low energy states is enhanced, reducing the internal energy Ub .
• The thermal expansion coefficient is lowered (enhanced) for fermions (bosons) compared to classical
gas. This feature represents the fact that with decreasing temperature the bosons tend to occupy
more low energy states than the fermions, thus reducing the effective volume Veff more strongly with
temperature which enhances the thermal expansion coefficient.

(e) Find the critical temperature Tc at which Bose-Einstein condensation occurs. How can
this be reconciled with the high-temperature, low-density limit?
Hint. The chemical potential can not be larger than the lowest energy level of the particles.

Solution. We now analyse the number equation (S.8) for the bosonic system in detail. If we decrease
the temperature, in order to keep the particle number constant, the function g3 (z) has to increase. For a
bosonic system, the chemical potential must always be smaller than the state with lowest energy. In this
case, this corresponds to

z ≤ e3β~ω/2 . (S.22)

The maximum of the monotonically increasing function g3 (z̃) (see Fig. 2) is taken when z = e3β~ω/2 .

4
 1.2
1.0
0.8

g3HzL
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
z

Figure 2: Plot of the function g3 (z) for values of z between 0 and 1.

Hence, the number equation cannot hold for arbitrary small temperature. There exists a critical tempera-
ture Tc at which the chemical potential is equal to the ground-state energy, µ = E0 , and thus ze−βE0 = 1.
This temperature can easily be computed as

~ωN 1/3
Tc = (g3 (1))−1/3 . (S.23)
kB
The physical interpretation of this critical point is simple: For the derivation of the number equation,
the zero energy state has been neglected. For temperatures above Tc this is a negligible error, but for
temperatures below Tc it becomes energetically favorable to occupy this single state with a macroscopically
large particle number. Hence, the approximation to neglect the occupation of the zero energy state becomes
bad for temperatures below Tc . Macroscopically large occupation of a single quantum state, or equivalently
a density matrix where a single eigenvalue is by far the dominant one, is a definition of a condensate. Hence,
Tc represents the critical temperature for the bosonic gas to form a Bose-Einstein condensate.
With equation (S.23) we show that our high temperature approximation holds even until Tc if the number
of particles N is large: when we compress enough particles, they will condense at a high temperature and
we can use the instability of this approximation to find our Tc . Of course, at Tc itself the approximation
breaks down.

5
Exercise 2. Behavior of excitations in a semiconductor.

In this exercise we analyze the properties and behavior of electron-excitations of a semiconductor

at finite temperature. In solid state theory, electronic states |k, αi are usually labeled by a
pseudomomentum k = (kx , ky , kz ) and a band-index α ∈ {1, 2, . . . }. For a crystal with lattice
constant a, the pseudomomentum takes values in the so-called Brillouin zone {−π/a, π/a}3 .
Assuming a cubic crystal with side-length L there exist (L/a)3 equally distributed k-vectors in
this Brillouin zone. Each of the states is doubly degenerate due to the spin, such that there are
in total 2(L/a)3 states for each band.

In order to simplify the treatment, we only

take into account two bands whose energies E
are approximated as parabolic, conduction band
~2 k2 ~2 k2
v (k) = − c (k) = Eg + , (4)
2mv 2mc
as shown in the figure. Here the indices v and
c stand for valence and conduction band, re- µ Eg
spectively. The parameters mv and mc which
define the curvature of the two bands are k
called effective masses and can in general be
different from one another and from the elec-
tron mass. The bandgap, Eg , is the energy
valence band
difference between the bottom of the conduc-
tion band and the top of the valence band.

Assume for this exercise that the bandgap is much larger than the thermal energy and the
chemical potential lies within the gap, βEg  βµ  1.

(a) Assume first that the particle number is not fixed and calculate the grand potential Ω of
this system.
Hint. For large L, a sum over k can be approximated by an integral:
X Z π/a L3
≈ d3 k 3 (5)
−π/a 8π
k

Solution. In the grandcanonical ensemble, each of the states can be either occupied or empty, such that
we can write the partition function in the occupation number formalism:
hY Y 1
X  nk,α i2
Z(β, z) = ze−βα (k)
k α=v,c nk,α =0
(S.24)
Y h  i2
β~2 k2 /2mv −β(Eg +~2 k2 /2mc )
= 1 + ze 1 + ze
k

Here the exponent 2 is present due to the spin degeneracy. The grand potential is now given by
Xh  2 2
  2 2
i
Ω(β, z) = −1/β log(Z(β, z)) = 2 log 1 + zeβ~ k /2mv + log 1 + ze−β(Eg +~ k /2mc ) .
k
(S.25)

Due to the assumption βEg  βµ  1 we know that

2 2 2 2
zeβ~ k /2mv
1 and ze−β(Eg +~ k /2mc )
1 (S.26)

6
 for all k, leading to the approximations

~2 k2
   
2 2 2 2
log 1 + zeβ~ k /2mv ≈ β µ + + z −1 e−β~ k /2mv
2mv (S.27)
 2 2
 2 2
log 1 + ze−β(Eg +~ k /2mc ) ≈ ze−β(Eg +~ k /2mc ) .

At this point we can replace the sum over k in (S.25) by an integral over k. This leads to

1 π/a 3 L3 β~2 k2
Z  
2 2 2 2
Ω=− d k 3 βµ + + z −1 e−β~ k /2mv + ze−β(Eg +~ k /2mc )
β −π/a 4π 2mv
(S.28)
3 2 2
(L/a) ~ π L3 (kB T )5/2  3/2 −βµ 
≈ −2µ(L/a)3 − 2
− √ mv e + m3/2
c e
β(µ−Eg )
,
mv a 3/2
2π ~ 3

where we have replaced the integration interval [−π/a, π/a]3 by R3 for the Gaussian integrals (last two
terms in the integral).

(b) In a realistic system, the particle number is fixed, as every atom in the solid contributes
a specific number of electrons. We assume here a particle number, such that the lower
band is completely filled at zero temperature, i. e. N = 2(L/a)3 . Calculate the chemical
potential µ(T ) at finite temperature.

Solution. In the grandcanonical ensemble the average particle number is given by

∂Ω
hN i = − . (S.29)
∂µ

Here we assumed that hN i = 2(L/a)3 . This leads to the equation

L3 (kB T )5/2 h −βµ

i
2(L/a)3 = 2(L/a)3 + √ −βm3/2
v e + βm3/2
c e
β(µ−Eg )
. (S.30)
3/2
2π ~ 3

This is equivalent to
 3/2
mv
= eβ(2µ−Eg ) , (S.31)
mc
leading to the result
 
Eg 3 mv
µ= + kB T log . (S.32)
2 4 mc

(c) Starting from your result for Ω, calculate the internal energy U (T, N ) − U (T = 0, N ) (for
N = 2(L/a)3 ), using a Legendre transform.
Hint. The final result is given by
 3/2
kB T 3/4
U (T, N ) − U (T = 0, N ) = N e−βEg /2 (mv mc ) (3kB T + Eg ) . (6)
2π~3

Solution. The internal energy is given by

U = Ω + T S + µN . (S.33)

We already calculated µ as a function of T and N . The entropy is given by the partial derivative

L3 (kB T )5/2
 
∂Ω 5  3/2 −βµ  1  3/2 −βµ
S=− = √ mv e + m3/2
c eβ(µ−Eg )
− −µmv e + (µ − Eg )m3/2 β(µ−Eg )
c e
∂T 2π 3/2 ~3 2T kB T 2
(S.34)

7
 ∂ 1 ∂
where we used ∂T
= −k 2 ∂β
. Using equation (S.32), we find
BT
 
mv3/2 e−βµ + mc3/2 eβ(µ−Eg ) = 2 e−βEg /2 (mv mc )3/4 (S.35a)
 
−µmv3/2 e−βµ + (µ − Eg )mc3/2 eβ(µ−Eg ) = Eg e−βEg /2 (mv mc )3/4 , (S.35b)

and with N = 2(L/a)3

3/2
L3 (kB T )5/2 a2 kB T

√ =N kB T . (S.35c)
2π 3/2 ~3 2π~2

Inserting equations (S.35) and (S.34) into equation (S.33), we now obtain the result
3/2
a2 kB T

U (T, N ) − U (T = 0, N ) = N e−βEg /2 (mv mc )3/4 (3kB T + Eg ) . (S.36)
2π~2
The zero-temperature energy

N ~2 π 2
U (T = 0, N ) = − (S.37)
2mv a2
that we subtracted is just the energy of the completely filled valence band.

(d) A picture that is frequently used in solid state theory is that of electrons and holes: When
an electron is excited to the conduction band, it leaves an empty state in the valence
band. This empty state now behaves like a particle itself and is called a hole. Therefore
an excitation can be regarded as a creation of two particles, similar to the creation of
particle-antiparticle pairs in particle physics.
Use this scheme to interpret the calculated internal energy in terms of the equipartition
law for an ideal gas. How many electrons are in the conduction band?

Solution. We call Nc the number of electrons in the conduction band. In the particle-hole picture this
is also equal to the number of holes in the valence band. Assuming now the behavior of ideal gas particles
for both electrons and holes, the internal energy of electrons and holes is given by
3 3
Ue = Nc kB T + Nc Eg Uh = N c kB T , (S.38)
2 2
where Eg describes the energy offset of the conduction band. The total internal energy is then

U = Nc (3kB T + Eg ) . (S.39)

By comparing this with equation (S.36) we find the number of electrons in the conduction band
3/2
a2 kB T

Nc = N e−βEg /2 (mv mc )3/4 , (S.40)
2π~2
which is exponentially suppressed by the size of the bandgap.