Contents

Statistical and Low Temperature Physics (PHYS393) 2.1 Fermi-Dirac statistics

2.2 Electronic heat capacity

2. Electrons in metals 2.3 Effective mass, and a few other things

Kai Hock
2011 - 2012
University of Liverpool

Electrons in metals 1

Electrons in a metal

The Drude model was propose in 1900 by Paul Drude to

explain the transport properties of electrons in a metal.

This treats the electrons like an ideal gas.

2.1 Fermi-Dirac statistics

It explains very well the DC and AC conductivity in metals, the

Hall effect, and thermal conductivity.

However, it greatly overestimates the heat capacities.

http://en.wikipedia.org/wiki/Drude_model

Electrons in metals 2 Electrons in metals 3

 Why did the electron gas model fail? The exclusion principle

When we derived the Maxwell-Boltzmann distribution for the Electrons are not allowed to occupy the same energy states.
ideal gas, we assumed that it is extremely unlikely for two
atoms to occupy the same energy level. So they have to be stacked up from bottom to top.

The reason is that the energy levels are very close together,
When heated, most of the electrons are stuck - there is no
compared to the average energy of the atoms.
space above to move up in energy !

Unfortunately, this is no longer true for electrons in a metal at

Only those few at the very top can. As a result, the heat
room temperature. (Prove it.)
capacities are much smaller than expected of a gas.

Electrons in metals 4 Electrons in metals 5

Fermi-Dirac statistics Fermi-Dirac distribution

As a result, we cannot use the Maxwell-Boltzmann distribution According to the exclusion principle, exactly ni states must be
for the ideal gas. filled and exactly gi − ni states unfilled.

The total number of ways the electrons can be arranged in this

Fortunately, we can use the same statistical methods that we
bundle is therefore:
have learnt so far.
gi!
Ωi =
ni!(gi − ni)!
Lets start by looking at the states in an energy interval dε.
For the whole system - i.e. all the bundles - we get
Y gi!
Ω=
i n i !(gi − ni)!
Applying the Lagrange multiplier method again, we would get
ni 1
=
gi exp(−λ1 − λ2εi) + 1
where λ1 and λ2 are the Lagrange multipliers.
We have gi energy states in the interval. Suppose the interval
We have applied the same constraints as before on the particle
contains ni electrons.
number N and the energy U .
Electrons in metals 6 Electrons in metals 7
 Fermi-Dirac distribution Fermi-Dirac distribution

We have obtained the Fermi-Dirac distribution function.

Previously, we have used n(ε) for the number density, g(ǫ) for
density of states, and dε for the energy interval of the bundle. g(ε)dε
n(ε)dε = .
exp(−λ1 + ε/kB T ) + 1
We also know that one of the Langrange multipliers would be The total number of electrons is fixed and is given by
related to temperature: integrating the normalisation condition:
g(ε)dε
Z ∞
λ2 = −1/kB T N =
0 exp(−λ1 + ε/kB T ) + 1
The distribution of the electrons may then be written as In principle, we could solve for the Lagrange multiplier λ1. It is
g(ε)dε common practice to express it in terms of µ as follows:
n(ε)dε = .
exp(−λ1 + ε/kB T ) + 1
exp(−λ1) = exp(−µ/kB T )
This is called the Fermi-Dirac distribution function.
µ is called the ”chemical potential.”

Electrons in metals 8 Electrons in metals 9

Occupying the energy states Occupying the energy states

For convenience, we define the following function:

The graph for f (ε) looks like this
n(ε) 1
f (ε) = =
g(ε) exp((−µ + ε)/kB T ) + 1
Since g(ε)dε is the number of states and n(ε)dε is the number
of particles, f (ε) would be the fraction of states that are
occupied.

So f (ε) is called the occupation number. We need to get a feel

as to what this looks like and how it changes with temperature.

Lets start with the simplest case: T = 0K. If we allow T to

approach zero, we will find: The shape shows that the energy levels are occupied below a
certain energy, and unoccupied above that.
f (ε) = 1 for ε < µ
f (ε) = 0 for ε > µ This feature is characteristic of the Fermi-Dirac distribution
that we are studying.
This means that all states with energy below µ are fully
occupied. All states with energy above µ are empty.
Electrons in metals 10 Electrons in metals 11
 The Fermi energy The Fermi energy

At T = 0K, The highest energy in the stack of electrons is µ. In terms of the occupation number,
This energy is also called the Fermi energy, EF . n(ε) = g(ε)f (ε)
So the normalisation condition can be written as:
Z ∞ Z ∞
N =2× n(ε)dε = 2 × g(ε)f (ε)dε
0 0
We know that at 0K, f (ε) = 0 for ε > µ. So So the integration
We can find the Fermi energy EF by integrating the would stop at ε = µ:
normalisation condition: Z E
F
Z ∞ N =2× g(ε)f (ε)dε
0
N =2× n(ε)dε
0 since µ = EF at 0K.
The factor of 2 must be added because each energy level can
be occupied by 2 electrons - spin up and spin down. We also know that at 0K, f (ε) = 1 for ε < µ. So
Z E
F
We shall solve this for the Fermi energy EF . N =2× g(ε)dε
0

Electrons in metals 12 Electrons in metals 13

The Fermi energy Chemical potential

We have previously derived the density of states:

4mπV We have previously mentioned that the chemical potential in
g(ε) = (2mε)1/2
h3 thermodynamics is the Gibbs free energy at equilibrium. E.g. it
In the topic on ideal gas, this is obtained by counting the could be the heat change when 1 mole of vapour condenses:
number of energy states of a particle in a 3-D box. In this
topic on electrons, we have used the same particle in a box. ∆µ = ∆U + p∆V − T ∆S.
The difference from ideal gas only arises later on, when we
We have just seen that at 0 K, the chemical potential for an
make different assumptions about the energy levels.
electron gas is the maximum energy of the electrons - the
So the same formula for the density of states can be used for Fermi energy. We may think of this as the energy change if we
both the ideal gas and the electrons. We can therefore remove 1 electron from the gas.
substitute the formula into the normalisation integral
Z E
F
In both cases, the chemical potential is the energy change
N =2× g(ε)dε when the number of particles in a system is changed.
0
and solve for the Fermi energy. The result is http://en.wikipedia.org/wiki/Chemical_potential
!2/3
~2 3π 2N
EF = .
2m V
Electrons in metals 14 Electrons in metals 15
 Heat capacity

We want to use the ideas and formulae that we have developed

to calculate the heat capacity of electrons.

To see how to do this, recall that the Drude model would

predict a heat capacity for that is the same as that of an ideal
gas:
2.2 Electronic heat capacity
3
C = N kB .
2
It is known from experiments that the actual heat capacity of
the electrons is much smaller. This refers to measurements
that are done at room temperature.

Electrons in metals 16 Electrons in metals 17

Heat capacity Heat capacity

We can understand this if we think that the electrons that are
stacked up to the Fermi energy do not have enough energy to
jump out of the stack.
We start by assuming that the thermal energy is indeed much
smaller than the Fermi energy. We shall derive an expression
for this thermal energy, and then calculate it at room
temperature to see if the assumption can be justified.

At a temperature above 0 K, the occupation number

1
f (ε) =
exp((−µ + ε)/kB T ) + 1
would no longer have a sharp step at the Fermi energy.

This would only be true if the thermal energy is much smaller

than the Fermi energy.

In order to find out if this is true, we need to estimate this

If kB T is much smaller than µ, the graph would remain close to
thermal energy. the step, as if the step has become smoother.
Electrons in metals 18 Electrons in metals 19
 Heat capacity Heat capacity

The ”smoothened” slope of the graph tells us that electrons
just below the Fermi energy (µ = EF ) is excited just above it.
We are assuming that kB T is much smaller than µ.
For energy ε higher than µ by a few times of kB T , The
exponential function exp[(ε − µ)/kB T ] would quickly become
large.

The occupation number is then approximately

1 1
f (ε) = →
exp[(ε − µ)/kB T ] + 1 exp[(ε − µ)/kB T ]
which is just the exponential function with negative argument
So we can estimate gain in thermal energy of the excited
electrons by the width δ of this slope. f (ε) = exp[−(ε − µ)/kB T ].

In order to do this, we take a closer look at the occupation

number
1
f (ε) =
exp[(ε − µ)/kB T ] + 1

Electrons in metals 20 Electrons in metals 21

Heat capacity Heat capacity

There is a corresponding behaviour just below the Fermi energy.

This means that the part of the graph to the right of µ falls off
exponentially. For energy ε lower than µ by a few times of kB T , The
exponential function exp[(ε − µ)/kB T ] would quickly become
small.

The occupation number is then approximately

1
f (ε) = → 1 − exp[(ε − µ)/kB T ]
exp[(ε − µ)/kB T ] + 1
where we have used the binomial expansion and kept only the
first order term.

It falls by a fraction of 1/e over an energy range of kB T .

Electrons in metals 22 Electrons in metals 23

 Heat capacity Heat capacity

This means that the part of the graph to the left of µ tends to At this point, we should justify our assumption that kB T at
the line f (ε) = 1 exponentially. room temerature is much smaller than the Fermi energy
µ = EF , which is defined at 0 K.

We shall take sodium metal as an example, and calculate kB T

and EF for this metal.

In sodium, each atom has one valence electron. This electron

is mobile and forms the electron gas that we are talking about.
In order to calculate the Fermi energy, we need the number
density N/V . We can calculate this from the following data:
density = 0.97 g cm−3
It reaches within f (ε) = 1 by a fraction of 1/e, over an energy relative atomic mass = 23.0
range of kB T .

This means that it is the electrons within this energy range So the volume for one mole of atoms is
that is excited. So the thermal energy of the excited electrons
23 ÷ 0.97 = 23.71 cm3,
is of the order of kB T .
Electrons in metals 24 Electrons in metals 25

Heat capacity Heat capacity

In contrast, at room temperature 298 K, we can calculate that

and the number density is
1
N kB T ≈ eV.
= NA ÷ 23.71 40
V This is about 120 times smaller than the Fermi energy.
where NA is Avogadro’s number. This gives an answer of
2.54 × 1028 m−3. We can repeat this for other typical metals, and we would get
similar answers.
Using the Fermi energy formula,
!2/3
~2 3π 2N This justifies our assumption that at room temperature kB T is
EF = . much smaller than the Fermi energy.
2m V
where m is the mass of the electron, we can find that the
We are now a step closer to estimating the electronic heat
Fermi energy is 3.16 eV.
capacity. Next, we need to understand the behaviour of the
excited electrons.

Electrons in metals 26 Electrons in metals 27

 Heat capacity
When temperature increases above 0 K, the step in the
Fermi-Dirac distribution becomes smoother as electrons just
below the Fermi energy are excited above it.
Notice that the ”tail end” of the distribution - to the right -
looks exponential. This is because there are relatively few
electrons above the Fermi energy. So these can behave like the
ideal gas and approximately obey the Maxwell-Boltzmann
distribution.
Electrons in metals 28
Heat capacity
As we have seen, these electrons close to the Fermi energy
behave like the ideal gas. We know that the energy of an ideal
gas is
3
U1 = N1kB T
2
where N1 is the number of particles in the ideal gas.
In the case of the electrons, N1 should refer to the number of
electrons above the Fermi energy, and not the total number of
electrons. We can estimate this number as follows.
Electrons in metals 30
Heat capacity
It is clear from the graph that for a small increase in
temperature, only electrons close to the Fermi energy are
excited. Most of the electrons are below the Fermi energy and
are not excited at all. Since they cannot be excited, these
electrons would not contribute to the heat capacity .
So it is mainly the electrons close to the Fermi energy that
would contribute to the heat capacity. So we can use these to
estimate the heat capacity and ignore the rest.
Electrons in metals 29
Heat capacity
The electrons that do get excited are in the small energy range
of order kB T from the Fermi energy EF .
So we can estimate N1 with the number of electrons that are
within the energy interval of
dε = kB T
from the Fermi energy.
Electrons in metals 31
 Heat capacity Heat capacity

From the definitions of the density of state and the Fermi

energy, it can be shown (see Exercise) that
3N
g(EF ) =
4EF
where N is the total number of electrons. This gives
!
3N 3 k T
N1 ≈ 2 kB T = N B
4EF 2 EF
Substituting into the energy for ideal gas
We know that the number of particles in a given energy
3
interval is U1 =
N1 kB T
2
n(ε)dε = 2g(ε)f (ε)dε we get the energy for the excited electrons
where the factor of 2 again comes from the spin states of the k T2
!
3 3 kB T 9
electrons. U1 = N k B T = N kB B
2 2 EF 4 EF
Differentiating with respect to T , we get the electronic heat
At 0K, the energy states below EF are fully occupied, i.e.
capacity
f (ε) = 1. So the number would given by
9 k T
N1 ≈ 2g(EF )kB T C = N kB B
2 EF
Electrons in metals 32 Electrons in metals 33

Heat capacity Heat capacity

1. Only electrons within kB T of the Fermi level can jump up.

We summarise what we have learnt about the heat capacity: Those further down cannot jump because the states above are
still mostly occupied.
Some states below Fermi level becomes empty, and some
states above becomes occupied. 2. The number of electrons that can get excited is 3k BT
2EF N .
This is a fraction 3k
2E
B T of the total.
F

3. Each would gain an energy of about 3kB T /2. So the

increase in total energy is obtained by multiplying the number
of excited electrons by this energy
3kB T 9 k T2
∆U ≈ N × 3kB T /2 = N kB B .
2EF 4 EF
4. Differentiating with respect to T , we get the electronic heat
For a temperature T , an electron that gets excited would be
capacity
able to gain on average an energy of about 3kB T /2.
9 k T
C = N kB B
2 EF

Electrons in metals 34 Electrons in metals 35

 Heat capacity of a metal Heat capacity of a metal

Notice that the heat capacity is directly proportional to T .

π2 T
We have obtained the electronic heat capacity C= N kB
2 TF
9 k T
C= N kB B This is often written in the form
2 EF
C = γT.
More detailed calculations show that the factor of 9/2 should
really be π 2/2: There is another contribution to the heat capacity. This comes
from the vibrations of the atoms, and it is proportional to T 3.
π2 k T
C= N kB B
2 EF We could imagine writing the total heat capacity in the form:
This equation is often written in the form
cV = γT + AT 3.
π2 T We can measure this heat capacity to check if the formula is
C= N kB
2 TF correct. Suppose that we have obtained a table of values for T
where TF = EF /kB is called the Fermi temperature. and cV . To check if the formula is correct, we can rewrite it in
this form:
cV /T = γ + AT 2
If we plot cV /T against T 2, we should get a straight line.
Electrons in metals 36 Electrons in metals 37

Heat capacity of a metal

In 1955, William Corak and his fellow co-workers measured the

heat capacities of copper, gold and silver from 1K to 5K in
their laboratory in Pittsburgh ...

2.3 Effective mass, and a few other things

W. S. Corak, et al, Physical Review, vol. 98 (1955) pp. 1699-1707

and got the straight lines. This shows that the predictions of
the Fermi-Dirac statistics are correct.
Electrons in metals 38 Electrons in metals 39
 Heat capacity of a metal Heat capacity of a metal

From the formula,

cV /T = γ + AT 2 Does this mean that our theory is wrong?

we know that we can find γ from the y-intercept of the graph.

Yes and no. The theory tells us that the electronic heat
The following table shows the values for a few metals.
capacity is proportional to T . The measurement shows that
there is indeed such a contribution.

So we are not completely wrong. Perhaps the theory needs

refining. We can be optimistic and go back and try and
understand what we have missed.

Recall that we have started with a particle in a 3-D box and

The second column contains the values of γ from the formula: calculated the energy levels. Then we just fill these up with
π2 k T electrons and calculated the heat capacity. All we have is a gas
C= N kB B = γT
2 EF of electrons in empty space.
The third column contains the values actually measured. They
are obviously different.
Electrons in metals 40 Electrons in metals 41

Heat capacity of a metal Heat capacity of a metal

But what about the atoms? The ”empty space” is really filled If we imagine that the electrons in a metal has a different
with atoms. The electrons must surely interact with the atoms. effective mass m∗ than its natural mass, we can ”explain” the
This then is the reason for the difference between theory and difference in γ.
measured γ.
According to the formula for heat capacity, calculated γ is
However, the measured C is proportional to T . This agrees given by
with theory, and it should mean something. One possibility is π2 k
γ= N kB B
that, for some reason, the electron interacts only weakly with 2 EF
the atoms. This idea has been shown to be correct by other This is inversely proportional to the Fermi energy
types of measurements. !2/3
~2 3π 2N
EF = .
2m V
According to this idea, the behaviour of the electrons in the
which is in turn inversely proportional to the mass m of the
presence of the atoms is essentially the same. The difference is
electron.
that the interaction with the atoms make the electrons behave
as if they have a different mass.
So γ is directly proportional to the mass.

Electrons in metals 42 Electrons in metals 43

 Heat capacity of a metal Behaviour at high temperatures

At very high temperatures, all the electrons could get excited.

This means that we can get the effective mass if we divide the
measured γ (in the third column) by the calculated γ (in the
second column).
This is shown in the last column of the table.
Then they would start to behave like an ideal gas. There are
many more energy state they can reach and they are much less
likely to be forced towards the same energy states.
This would happen only if the electrons are excited far above
the Fermi energy. Therefore we can use Fermi temperature as
a reference. A temperature would be high if it is high compared
to the Fermi temperature.

At the high temperature, the heat capacity would therefore

change to that of the ideal gas. So instead of
π2 T
C= N kB
2 TF
which is very small, it would become
3
C= N kB
2
which is much larger.
Electrons in metals 44 Electrons in metals 45

A note on the chemical potential A note on fermions

We have seen many times that the chemical potential µ is

equated to the Fermi energy EF . It is important to note that
this is true only at 0 K. Note that electrons belong to a larger family of particles called
fermions.
Recall the chemical potential µ is determined by the number of
particles: A fermion is a particle with a half integer spin. Other examples
Z ∞
g(ε)dε are
N = the proton,
0 exp[(ε − µ)/kB T ] + 1
the neutron,
At T = 0 K, the integral is just
Z µ the helium-3 (3He) atom and
N = g(ε)dε the oxygen-13 (13O) atom.
0
which is why µ is the Fermi energy then.
All fermions are known to show the kind of properties we have
seen for electrons. They obey the Fermi-Dirac statistics.
At higher temperature, µ would change. But for temperatures
well below the Fermi temperature, as with electrons in metals
at room temperature, we may assume that µ is still quite close
to the Fermi energy.
Electrons in metals 46 Electrons in metals 47
 What we have learnt so far

1. Electrons follow the Fermi-Dirac distribution:

g(ε)dε
n(ε)dε = .
exp[(ε − µ)/kB T ] + 1
2. At 0 K, the electrons fill up the energy states from the
lowest level. The highest energy of the electrons is called the
Fermi energy:
!2/3
~2 3π 2N
EF = .
2m V 2.4 Exercise
3. When thermal energy kB T is small compared to the Fermi
energy, only electrons near the Fermi energy would get excited.
So only these would contribute to the electronic heat capacity:
π2 T
C=N kB
2 TF
4. The Fermi temperature is defined as
EF
TF =
kB
Electrons in metals 48 Electrons in metals 49

Exercises Exercises

We can make use of the following relations for the density of

states.

Exercise 1 Relation with the total number of states:

G(ε)
Show that the density of states at the Fermi energy EF is given g(ε) = .
dε
by
Formula for the total number of states:
3N
g(EF ) = 4πV
4EF G(ε) = (2mε)3/2
3h3
where N is the number of electrons. so
G(ε) ∝ ε3/2.
Relation with the Fermi energy and N :
Z E
F
N =2 g(ε)dε.
0

Electrons in metals 50 Electrons in metals 51

 Exercises Exercises

From
G(ε)
g(ε) =
dε
and 3G(EF )
Z E g(EF ) = .
F 2EF
N =2 g(ε)dε,
0 Substituting the above result for N :
we have
N = 2G(EF ),
N = 2G(EF ).
we get
From
3N
g(EF ) = .
G(ε) ∝ ε3/2, 4EF
we have
g(ε) 3
= .
G(ε) 2ε
At the Fermi energy EF , we get ...

Electrons in metals 52 Electrons in metals 53