T4 – Phonons

In the beginning of the course, we learned about the microscopic structure of crystals (lattice
structures, reciprocal lattice, x-ray diffraction). We developed the language that would allow us to
study solid-state physics based on geometrical periodic arrangements, both in the real space and in k-
space (the reciprocal space). So far, we talked about static structures. In reality, for every temperature
above 0 kelvin we expect to see atomic motion. Unlike gasses, in which we assume each atom to be a
free particle, in a solid matter, the dynamic is dictated by interactions between the atoms. We are
going to study this dynamic, assuming small changes around the equilibrium (the static crystal
structure).

1. Dispersion relation, Normal Modes, Phonons and Debye model

Dispersion relation (review of waves)

The E.O.M of a classical undamped oscillator (an ideal spring with no friction) is that of a plane
wave:

𝐴(𝑥, 𝑡) = 𝐴0 𝑒 𝑖(𝑘𝑥−𝑤𝑡) [1.1]

𝑨𝟎 is the amplitude - maximum displacement from equilibrium.

2𝜋
k is the wavenumber (𝑘 = 𝜆
) - how many complete cycles (or wavelengths) are contained within a

given unit of length (spatial frequency).

𝒘 is the is the angular frequency (𝜔 = 2𝜋𝑓) - how many complete cycles (or wavelengths) are
contained within a given unit of time.

The dispersion relation is the connection between 𝜔 and k – 𝜔(𝑘)

1 1
Note: The wavenumber k has units of [𝑚], the angular frequency has units of [𝑠 ]. A linear dispersion
𝑚
relation has the units of velocity [ 𝑠 ], but 𝜔(𝑘) doesn’t have to be linear. For instance, the dispersion

relation for deep water waves is:

𝜔 = √𝑔𝑘 [1.2]

g is the acceleration due to gravity. This 𝜔(𝑘) is not linear.

Normal modes (review of waves)

Normal modes are collective oscillations where all particles move at the same frequency 𝜔. The
normal modes are the eigenmodes of the systems. When we extract the eigenmodes of a system (by
diagonalizing the Hamiltonian) we can find the dispersion relation.

1
Phonons

An eigenmodes of the system has known temporal and spatial frequencies (𝜔, 𝑘), but this is not the
whole story. Another important parameter is the amplitude (eq.1.1). The amplitude defines the total
energy of the mode (the eigenvalue of the eigenmode). In classical systems, the energy can have any
value from 0 to infinity (or practically – until the spring brakes). In crystals, the energy value isn’t
continuous, as there are minimal allowed quantum exactions. Those quantum exactions of the
vibrational modes are called phonons (similar to the quantum exactions of the electromagnetic modes,
which are called photons).

The study of phonons is an important part of solid-state physics. They play a major role in many of
interesting physical properties such as thermal conductivity and electrical conductivity.

Debye model

Debye model, in its 1D version, assume the atoms to be bound by a harmonic potential (connected to
each other by springs)

Figure 1 - Atoms in a harmonic potoential

The Debye model treats waves in a solid as sound waves, and therefore assumes a linear dispersion
relation:

ω(k) = vsound k [1.3]

This dispersion relation is inaccurate and holds only for small k values. To complete Debye’s solid
model, let's find the normal modes, as well as the general dispersion relation that Debye missed.

2. A Chain of 3 Atoms
For simplicity, let's find the normal modes of a chain with only 3 atoms.

𝑋𝑖 𝑒 is the equilibrium point of the 𝑖 𝑡ℎ atom, and 𝑋𝑖 is its real position. 𝑞𝑖 = |𝑋𝑖 − 𝑋𝑖 𝑒| is the
deviation from equilibrium.

2
 Figure 2 – A chin of 3 atoms.

As usual, we will look for the Hamiltonian of the problem. the kinetic energy:
3
1
𝑇 = ∑ 𝑚𝑖 𝑞𝑖̇ 2 [2.1]
2
𝑖=1

As for V, the potential energy, we know it will be quadratic. Therefore, we can expand V to second
order around an equilibrium point:
𝜕𝑉 1 𝜕2𝑉
𝑉 = 𝑉0 + ∑ ( ) 𝑞𝑖 + ∑ ( ) 𝑞 𝑞 +…. [2.2]
𝜕𝑞𝑖 0 2 𝜕𝑞𝑖 𝜕𝑞𝑗 0 𝑖 𝑗
𝑖 𝑖,𝑗

Since the potential is quadratic, the first order term is zero. We can also ignore the zero-order
constant, so practically:
1 𝜕2𝑉
𝑉=∑ ( ) 𝑞𝑞 [2.3]
2 𝜕𝑞𝑖 𝜕𝑞𝑗 0 𝑖 𝑗
𝑖,𝑗

Let's change coordinates:

𝜂𝑖 = 𝑞𝑖 √𝑚𝑖 [2.4]

And rewrite the kinetic and potential energies accordingly:

3
1
𝑇 = ∑ 𝜂𝑖̇ 2 [2.5]
2
𝑖=1
1
𝑉 = ∑ 𝑎𝑖,𝑗 𝜂𝑖 𝜂𝑗 [2.6]
2
𝑖,𝑗

Whereas we denote:

𝜕2𝑉 1
𝑎𝑖,𝑗 = ( ) [2.7]
𝜕𝜂𝑖 𝜕𝜂𝑗 0 √𝑚𝑖 𝑚𝑗

From Newton's second law:

⃗⃗𝐹𝑖 = 𝑚𝑖 𝑞𝑖̈ = √𝑚𝑖 𝜂𝑖̈ [2.8]

On the other hand, the force is the change in potential:

3
 𝜕𝑉 √𝑚𝑖 𝜕𝑉 √𝑚𝑖 𝜕𝑉
⃗⃗𝑖 = −
𝐹 =− = [2.9]
𝜕𝑞𝑖 √𝑚𝑖 𝜕𝑞𝑖 𝜕𝜂𝑖

From [2.8] and [2.9] we get the following equation of motion:

𝜕𝑉 [2.6] 1 𝜕 1
−𝜂𝑖̈ = → −𝜂𝑖̈ = ∑ 𝑎𝑖,𝑗 𝜂𝑗 𝜂𝑖 = ∑ 𝑎𝑖,𝑗 𝜂𝑗 [2.10]
𝜕𝜂𝑖 2 𝜕𝜂𝑖 2
𝑗 𝑗

In order to solve this E.O.M, we will guess a periodic solution:

𝜂𝑖 (𝑡) = 𝐴𝑠𝑖𝑛(𝜔𝑡) [2.11]

𝜂𝑖̈ (𝑡) = −𝜔2 𝐴𝑠𝑖𝑛(𝜔𝑡) = −𝜔2 𝜂𝑖 (𝑡) [2.12]

Plugging [2.12] into the E.O.M in [2.10] we obtain:

1
∑ 𝑎𝑖,𝑗 𝜂𝑗 = 𝜔2 𝜂𝑖 [2.13]
2
𝑗

Or in matrix form:

𝑎11 𝑎12 𝑎13 𝜂1 𝜂1

𝑎
[ 21 𝑎22 𝑎23 ] [𝜂2 ] = 𝜔2 [𝜂2 ] [2.14]
𝑎31 𝑎32 𝑎33 𝜂3 𝜂3

Now that we have an eigenvalue problem, we can diagonalize the matrix and find the normal modes
(𝜔, 𝑡ℎ𝑒 𝑒𝑖𝑔𝑒𝑛𝑣𝑎𝑙𝑢𝑒𝑠) and the corresponding eigenvectors.

𝑎𝑖.𝑗 are given inexplicably in [2.7]. To find them, we can use the explicit form of V.

The 3 atomic problem (figure 2) is simple enough to do so. The potential:

𝐾
𝑉= [(𝑞2 − 𝑞1 )2 + (𝑞3 − 𝑞2 )2 ] [2.15]
2

𝐾
𝑉= (2𝑞22 + 𝑞12 + 𝑞32 − 2𝑞1 𝑞2 − 2𝑞3 𝑞2 ) [2.16]
2

Reminder –

𝜕2𝑉 1
𝑎𝑖,𝑗 = ( ) [2.7]
𝜕𝜂𝑖 𝜕𝜂𝑗 0 √𝑚𝑖 𝑚𝑗

Using [2.16] we can find 𝑎𝑖,𝑗 by coefficients comparison:

𝜕2𝑉 𝜕2𝑉 𝐾
2 = 2 = 𝐾 → 𝑎11 = 𝑎33 = 𝑚 ≡ 𝑎
[2.17]
𝜕𝑞1 𝜕𝑞3

4
 𝜕2𝑉 𝜕2𝑉 𝐾
= = −𝐾 → 𝑎12 , 𝑎21 , 𝑎23 , 𝑎32 = − = −𝑎 [2.18]
𝜕𝑞1 𝜕𝑞2 𝜕𝑞2 𝜕𝑞3 𝑚

𝜕2𝑉 2𝐾
2 = 2𝐾 → 𝑎22 = 𝑚 = 2𝑎
[2.19]
𝜕𝑞2

𝜕2𝑉
= 0 → 𝑎13 , 𝑎31 = 0 [2.20]
𝜕𝑞1 𝜕𝑞3

Plugging these results into [2.14]:

𝑎 −𝑎 0 𝜂1 𝜂1
[−𝑎 2𝑎 −𝑎] [𝜂2 ] = 𝜔2 [𝜂2 ] [2.14]
0 −𝑎 𝑎 𝜂3 𝜂3

This problem has 3 eigenmodes:

1
• 𝜔 = 0. The corresponding eigenstate is [1], meaning that all masses are moving in the same
1
direction – this is translation:

Figure 3 – Translation

−1
𝐾
• 𝜔 = √𝑚. The corresponding eigenstate is [ 0 ]:
1

Figure 4 – Symmetric stretch

1
𝐾
• 𝜔 = √3 𝑚. The corresponding eigenstate is [−2]:
1

Figure 5 – Anti-symmetric stretch

3. A Chain of N Atoms

Figure 6 – A Chain of N atoms centered around the 𝒊𝒕𝒉 atom.

5
As we did in the previous case, we can write an E.O.M for the 𝑖 𝑡ℎ atom in terms of his nearest
neighbors (in the previous case, there were only nearest neighbors. This time, we take it as an
approximation):

𝑚𝑢𝐼̈ = 𝐾(𝑢𝐼+1 − 𝑢𝐼 ) − 𝐾(𝑢𝐼 − 𝑢𝐼−1 ) = 𝐾𝑢𝐼+1 − 2𝐾𝑢𝐼 + 𝐾𝑢𝐼−1 [3.1]

we will guess a periodic solution of the following form:

𝑢𝐼 (𝑡) = 𝐴𝐼 𝑒 𝑖𝜔𝑡 [3.2]

Plugging [3.2] into [3.1]:

−𝜔2 𝑚𝐴𝐼 = 𝐾𝐴𝐼+1 − 2𝐾𝐴𝐼 + 𝐾𝐴𝐼−1 [3.3]

Or in the matrix form:

Figure 7 – The eigenvalues problem in the matrix form

Assuming 𝐴𝑖 is complex:

𝐴𝐼 = 𝐴𝑒 𝑖𝑞𝐼 [3.4]
And plugging [3.4] into [3.3]:

−𝜔2 𝑚𝐴𝑒 𝑖𝑞𝐼 = 𝐾𝐴𝑒 𝑖𝑞(𝐼+1) − 2𝐾𝐴𝑒 𝑖𝑞𝐼 + 𝐾𝐴𝑒 𝑖𝑞(𝐼−1) [3.5]
From [3.5] we can extract the following dispersion relation:
𝐾 4𝐾 2 𝑞
𝜔2 = (2 − 2𝑐𝑜𝑠𝑞) = sin [3.6]
𝑚 𝑚 2

𝐾 𝑞
𝜔 = 2√ |sin ( )| [3.7]
𝑚 2

Figure 8 – The dispersion as function of q..

6
 2𝜋
As usual, we want the period to be equal to 𝑎
(remember what we said about the reciprocal
lattice vector and K-space), therefore:

𝑞 = 𝑘𝑎 [3.8]
k is the momentum. Notice that the period of |sin| is half of the period of sin.

𝜋 𝜋
Figure 9 – The dispersion relation. The first Brillouin zone is [− , ].
𝑎 𝑎

So 𝜔(𝑘) is given by:

𝐾 𝑘𝑎
𝜔 = 2√ |sin ( )| [3.8]
𝑚 2

For small k values, we can expand the dispersion relation to first order:
𝐾
𝜔 ≈ 𝑎√ 𝑘 [3.9]
𝑚

For small k values, we got a linear dispersion. In this regime, our results are in complete agreement
with those of Debye. For this reason, excitations of this band (fig.9) are called acoustic phonons.
In 3 dimensions, there are 3 branches of acoustic phonons (as there are 3 kinds of vibrational waves –
2 transverse waves and 1 longitudinal wave).

4. Diatomic Solid and Optical phonons

Figure 10 - Diatomic chain

Now, consider a periodic arrangement of two different types of atoms. For simplicity, let us
focus on the case where all the masses along our chain are the same
(m1 = m2 = m) but the two spring constants K and G are different.

The E.O.M for the two atoms:

𝑚𝑢1̈ = −𝐺(𝑢1𝐼 − 𝑢2𝐼 ) − 𝐾(𝑢1𝐼 − 𝑢2𝐼−1 ) [4.1]

7
 𝑚𝑢2̈ = −𝐾(𝑢2𝐼 − 𝑢1𝐼+1 ) − 𝐺(𝑢2𝐼 − 𝑢1𝐼 ) [4.2]

Our ansatz this time will be:

𝐼 (𝑡)
𝑢1,2 = 𝐴1,2 𝑒 𝑖𝜔𝑡 𝑒 𝑖𝑞𝐼 [4.3]

Plugging [4.3] into the [4.1], [4.2] yields (after some algebra):

𝜔2 𝑚𝐴1 = 𝐺(𝐴1 − 𝐴2 ) + 𝐾(𝐴1 − 𝐴2 𝑒 −𝑖𝑞 ) [4.4]

𝜔2 𝑚𝐴2 = 𝐾(𝐴2 − 𝐴1 𝑒 −𝑖𝑞 ) + 𝐺(𝐴2 − 𝐴1 ) [4.5]

Or in matrix form:

[ 𝐺+𝐾 −𝐺 − 𝑘𝑒 −𝑖𝑞 ] [𝐴1 ] = 𝜔2 𝑚 [𝐴1 ] [4.6]

−𝐺 − 𝐾𝑒 𝑖𝑞 𝐺+𝐾 𝐴2 𝐴2
The solution of this eigenvalue problem gives the following dispersion relation:

(𝐺 + 𝐾) 1
𝜔± = √ ± √𝐺 2 + 𝐾 2 + 2𝐺𝐾𝑐𝑜𝑠(𝑘𝑎) [4.7]
𝑚 𝑚

We obtain 2 different 𝜔’s for every k. This is due to the new degree of freedom in which the
different masses are off-phase and moves in opposite directions.
Figure 11 illustrates [4.7]. The upper branch is called the "optical branch" and the lower branch
is called the "acoustic branch".

Figure 11 - The Diatomic dispersion curve. The upper branch is the "optical branch" and the lower is the "acoustic
branch"

Optical Phonons

The number of vibrational branches in a lattice is equal to the number of degrees of freedom:
𝑁𝑚𝑜𝑑𝑒𝑠 = 𝑁𝑢.𝑐. 𝑑 [4.8]
Whereas 𝑁𝑢.𝑐. is the number of atoms in a Wigner-Seitz unit cell and d is the dimension of the
problem (1d,2d,3d..)
In 3D, we expect to see a total of 6 branches – 3 acoustic branches and 3 “optical phonons”
branches.

8
Solids can absorb light, but energy and momentum must both be conserved.

The dispersion relation of light (in vacuum) is given by:

𝜔𝑙𝑖𝑔ℎ𝑡 = 𝑐𝑘 [4.9]
The energy of light is given by:
𝐸 = ℏ𝜔 = ℏ𝑐𝑘 [4.10]
𝜔
As we saw (eq.3.9), for acoustic phonons with small k: = 𝑣𝑠𝑜𝑢𝑛𝑑 . Since the speed of light is much
𝑘

bigger than the speed of sound, light (with small k) cannot excite an acoustic phonon due to energy
conservation. On the other hand, phonons in the optical band (fig.11) have high energy values even
for 𝑘 → 0.