§4 – Vibrations

§4 – Vibrations
4.1 Vibrations

This is a sizeable subject, which leads to powerful predictions about the thermal properties of
solids. When considering the counting of modes, we will see how reciprocal space notation is a
useful tool.

We will start by considering simple models of vibrations and progress towards a more general case.
Although strictly necessary, we will not study elasticity theory.

Probably the simplest model of lattice vibrations is a 1-dimensional monoatomic infinite chain of
identical particles.

M M M

us-1 us us+1
Figure 4.1: Monoatomic crystal comprising
atoms of mass M connected by force constant
C separated by repeat distance a.

a) 1D Monoatomic Chain

From Fig 4.1, we can see that the force (to the right) of a mass at a point s on the chain, M S is
given by:

2
d us
F s= M 2
=C u s−1−u s C u s 1−u s=C u s1u s−1−2u s  (4.1)
dt

This equation has a wave solution. As usual, we have an amplitude equal to the value at t = 0 and a
complex exponential term:

u s= u s 0 e−i  t (4.2)

We can drop the “0”after substituting into equation (4.1) without losing information:

2
−M  u s=C u s1u s−1−2 u s (4.3)

–1–
Last Modified: 04/12/2006
 §4 – Vibrations :

This is now simply a recurrence relation. If we assume that u s has a wave-like space
dependence, we can simplify this further:

i K  sa
u s=u e (4.4)

Where K is some wavenumber, s is an integer (the number of the particle along the chain) and a is
the lattice spacing. In other words, s a corresponds to distance along the chain.

Thus, equation (4.3) becomes:


−M  2 ei K s a=C e i K s a e e

iKa −i K a
−2
(4.5)
(s+1) (s-1)

1 i x −i x
Using the standard identity, cos x≡  e e  gives:
2

−M  2=2 C cos K a−1

2 C
⇒  =2 1−cos K a
M

 2=
4C
M
1
sin 2 K a
2   (4.6)

This is known as a dispersion relation. It tells us how the frequency of the wave changes with the
wavenumber. Every value of K can be a solution.

4.2 First Brillouin Zone

The wavenumber has units [L]-1, which suggests that there may be a link between the wavenumber
and reciprocal space.

If we plot ω against K, we can see that  K  is periodic in K.

We can see the First Brilluoin Zone as the first set of unique solutions of  K  about K = 0.
The value of K at the zone boundary can be found by writing e i K s a as:

e i K sa
=e i K sa
e 2i s
=e

i K
2
asa

–2–
Last Modified: 04/12/2006
 §4 – Vibrations :

− 
So, we have unique solutions in the range ≤K ≤ . But, K can be positive or negative, so the
a a
zone boundary occurs at:


K ZB =± (4.7)
a

This is a very important result.

We can see from the picture that the high-frequency wave, corresponding to K beyond the zone
boundary, is indistinguishable from a lower frequency wave within the FBZ:

Figure 4.2: The wave represented by the solid curve conveys no

further information than the dashed curve at the atom positions.
Only wavelengths longer than 2a are needed to represent the
motion.

It is possible to distinguish the positive and negative states at the zone boundary, but only because
of the time-dependence of u s :

u s t = u s 0 e i K s a e−i  t= u s 0 ei  K s a −t  (4.8)

The phase velocity is given by:


v p= (4.9)
K

The phase velocity can be positive or negative.


However, at the zone boundary, K =± , we lose this sense of direction:
a

s
u s t K =/ a= u s 0 exp  i ± s−t = u s0 −1  e−i t (4.10)

–3–
Last Modified: 04/12/2006
 §4 – Vibrations :

Equation (4.10) represents a standing wave with alternate atoms oscillating in opposite phase. So, at
the zone boundary, we no longer have a travelling wave.

4.3 Group Velocity

A generally more useful measure of velocity is the group velocity, which represents the movement
of the wave packets. This is defined as the rate of change of angular velocity with wavenumber:

[    ]    
1 1
d d C 2 Ka C 2 Ka
v g= = 2 sin =a cos (4.11)
dK dK M 2 M 2

The group velocity falls smoothly to zero at the zone boundary.

If we take the long wavelength limit, then K ≪


a
and cos
Ka
2
≈1 .  
Thus, the long-wavelength group velocity is:

 
1
C 2
v g ≈a =const.
M
(4.12)

Figure 4.3: Group velocity vg versus K, for

a 1-dimensional monoatomic chain.

This can also be interpreted as the linear region of the dispersion relation.

The angular frequency in this case is given by:

  ∣  ∣  
1 1
C 2 Ka C 2 a
=2 sin ≈2 ∣K∣
M 2 M 2

–4–
Last Modified: 04/12/2006
 §4 – Vibrations :

The long-wavelength group velocity is then:

 
1
 C 2
v p= =a =v g (4.13)
K M

This is the velocity of long-wavelength elastic waves in the solid, commonly called the speed of
sound in the material:

v s≈v p≈v g (4.14)

A non-dispersive medium has v g =v p .

Dispersion effects set in near the zone boundary. We can consider this as interference with a Bragg-
diffracted wave travelling in the opposite direction, slowing down the wave packets until they
eventually form a standing wave at the zone boundary.

It is easy to generalise these results to three dimensions by replacing the wavenumber K with the 3-
dimensional wavevector K  :

Dispersion Relation:


 K   K

Group Velocity:

v g  vg =∇ K   K


Where: ∇ K is the gradient (del) operator in wavevector space, which is a vector.

In the long-wavelength limit, an isotropic medium (e.g. cubic structure) has the same v g value in
all directions.

There are two types of polarization: longitudinal and transverse, denoted by L and T. Altogether,
there a 3 modes, 1 L mode and 2 T modes.

Typical values for the constants are:

M =Z m p~50×10−27 kg

a~0.3 nm

v s~104 m s−1

–5–
Last Modified: 04/12/2006
 §4 – Vibrations :

Using these values gives a spring constant of:


2
v
C=M s =5×10−26⋅3×10 132=50 N m −1
a

This is a typical macroscopic value for a light spring.

4.4 Structure with a Basis

So far, we have considered the 1-dimensional primitive lattice. The next level of complexity is a
two-atom basis, or equivalently, a basis with two spring constants. An example of this is the [1 1 1]
planes of a NaCl structure.

This time, we have two coupled equations of motion:

d 2 us
M1 =C v sv s−1−2 us  (4.15)
d t2

d2 v s
M2 =C u s1us −2 v s  (4.16)
d t2

Again, the solution will be a wave, but there will be different amplitudes for the two masses. There
will be the same frequency and wavenumber for each wave-mode of the system:

−i t i K sa 
u s=u e e (4.17)

−i t i K s a 
v s=v e e (4.18)

M1 M2 M1

us vs us+1

Figure 4.4: Diatomic crystal structure with

masses M1 and M2 connected by force
constant C. The repeat distance a is twice
that of the monoatomic crystal.

–6–
Last Modified: 04/12/2006
 §4 – Vibrations :

We can substitute and cancel the common factors of the exponentials:

− 2 M 1 u=C vve−i Ka−2 u  (4.19)

− 2 M 2 v=C  u e i K au−2v  (4.20)

We can write the coupled bilinear equations (4.19) and (4.20) in terms of a matrix:

 2C− 2 M 1
−C exp[i K a ]1
−C 1exp[−i K a ] u
2
2C− M 2 v
= 0
0     (4.21)

This is an eigenvalue problem. As with all eigenvalue problems (in matrix form), the determinant of
the matrix must be zero is order to obtain solutions:

 4 M 1 M 2− 2 2 C M 1 2 C M 24 C 2−C 2  1exp [−i K a ] exp [i K a]1 =0

Using the identity 2 cos x=e x e−x :

M 1 M 2  4−2CM 1M 2  22C2  1−cos  K a =0

Finally, completing the square gives solutions to  2 as:

22 C  M 1M 2 ±4 C 2  M 1 M 22 −8 M 1 M 2 C 2 1−cos K a 

= (4.22)
2M1M2

There are two frequencies for each K wavenumber. This corresponds to the two “branches” of the
dispersion relation.

Before we plot this result, it will be useful to examine the limits of this solution.

a) Small-K Limit

In this case, K ≪  , so 1−cos K a≈ 1 K 2 a 2 .

a 2

–7–
Last Modified: 04/12/2006
 §4 – Vibrations :

So, we have:
4 2 2 2 2
M 1 M 2  −2C M 1M 2  C K a =0

Let us look at the general equation for the roots of a quadratic in the limit c 0 :

For the negative solution:

lim
c 0
[ −b− b 2−4 a c −b− b 2
2a
=
2a
=−
b
a ] (4.23)

The positive solution is slightly more complicated, but we can use a series expansion of the form

    
1 /2 2
1/ 2 1 /2 y 1/ 2 1 y y
 x y  =x 1 =x 1 O :
x 2 x x

−b b 2−4 a c −bb  1−4 a c/2 b   c

2
≈ = (4.24)
2a 2a b

Thus the two small-K solutions are:

12=
2 C  M 1M 2 
M1M2
=2 C
1

1
M1 M 2   (4.25)

2 2 2
C K a C
 22= = K 2 a2 (4.26)
2 C  M 1M 2  2 M 1M 2 

M1M2
Equation (4.25) is proportional to the reciprocal of the reduced mass = . There is
 M 1M 2 
also no dispersion in this branch.

–8–
Last Modified: 04/12/2006
 §4 – Vibrations :

Equation (4.26) is like the solution to the monoatomic problem, where the group velocity
incorporates both masses:

   
1/ 2 1/ 2
C C
v g =a a
M 2 M 1M 2 

b) Zone Boundary

±
At the zone boundary, we know that K = . Since the solutions at these two limits will be
a

equivalent, but with opposite sign, we will just consider the positive limit. At K = we have
a
cos K a=−1 , which gives:

M 1 M 2  4−2 C  M 1M 2 24 C 2=0

So, solving for  2 gives:

2 C M 1M 2 ± 4C2  M 1M 22−16 C 2  M 1 M 2

 2=
2 M1M 2

Take the term in the square root:

4C 2 M 1 M 22−16 C 2 M 1 M 2 =4C2  M 21 2M 1 M 2 M 22−4 M 1 M 2 

4C 2 M 1 M 22−16 C 2 M 1 M 2 =4C2  M 21 −2M 1 M 2M 22=4C2  M 1−M 2 2

Thus, the solutions of  2 are:

2C M 1M 2 ±2C M 1 −M 2  C

 2=
2M1M 2
=
M1M2
[ M 1M 2 ± M 1−M 2 ] (4.27)

–9–
Last Modified: 04/12/2006
 §4 – Vibrations :

This clearly has two solutions, given by:

1 =
 2C
M1
(4.28)

 2=
 2C
M2
(4.29)

So, at the zone boundary, the branches separate into solutions related to an individual mass.

We can use these limit solutions to project the final curves:

Figure 4.5: Optical and acoustic branches of the

dispersion relation of a diatomic linear lattice in
the positive half of the first Brillouin zone. Here,
a is the lattice constant.


As before, we do not consider states outside ± , since for discrete atoms, these will be identical
a
to those shown.

4.5 Nature of Branches

The splitting of the dispersion relation into optical and acoustic branches is a fundamental paradigm
of solid state physics.

We can determine the nature of the modes from the eigenvalues of the matrix equation.

In the optical mode, u and v have opposite sign.

In the acoustic mode, u and v have the same sign.

– 10 –
Last Modified: 04/12/2006
 §4 – Vibrations :

The ratio of u and v varies with K.

We saw before that a wave appears to have a low-frequency and high-frequency modes. For
transverse waves, these correspond to the acoustic and optical modes respectively.

Figure 4.6: Transverse optical and acoustic waves

in a diatomic linear lattice. Both modes have the
same wavelength.

At the zone centre (K = 0) of the optical mode, the wave is flat.

At the zone boundary, only one of the ions oscillates.

We have not shown how the 2-branch dispersion evolves into a single branch when the two masses
are equal.

The problem is that we used a in both cases. However, these are not the same. Once the masses
become the same, the diatomic length is twice the monoatomic length as defined in Fig. 4.1 and 4.4.

We can shift the negative section of the optical branch into a doubled Brillouin Zone.

We can close the gap between the acoustic branch and shifted part of the optical branch by setting
M 1=M 2 .

Let us check the frequencies:

 c= 2C
  1

1
M1 M2  for a diatomic chain.

If M 1=M 2 , then:

 c=
 4C
M
for a monoatomic chain.

– 11 –
Last Modified: 04/12/2006
 §4 – Vibrations :

General Case:

In a general 3-dimensional monoatomic system, there will be always be 3 acoustic branches:

Transverse – 2 modes

Longitudinal – 1 mode

This means there are three angular frequencies for each value of the wavevector 
k .

For a 2-atom basis, there will be twice this number of modes. However, since there are always 3
acoustic modes, these extra 3 modes with be optical modes.

For a p-atom basis, there will be p×3 number of total modes.

It is clear that the most general case of an n-dimensional system with a p-atom basis, will have:

n  p−1 optical modes

n acoustic modes

Where the optical modes will be of mixed polarisation.

4.6 Quantisation of Vibrations

A mechanical oscillator has discrete excitation energies. We have shown in quantum mechanics,
that the energy, ε of an oscillator, with quantum number n (corresponding to the energy level) and
angular frequency ω is given by:

 
= n
1
2
ℏ (4.30)

Where n is a non-negative integer.

Mechanical vibrations correspond to bosons, which means that any number of them can occupy an
energy state, i.e. n can be any non-negative integer.

The quanta of mechanical excitation are called phonons, by an obvious analogy with the quanta of
optical oscillation: the photons.

If n = 0, then we have the lowest energy state, which has zero-point energy, which we can see from
1
equation (4.30) is equal to ℏ .
2

– 12 –
Last Modified: 04/12/2006
 §4 – Vibrations :

The zero-point energy is a result of Heisenberg's uncertainty principle: the angular frequency has
units of inverse time, and we know that the uncertainty between energy and time is given by:
ℏ
  t≥ .
2

Each K-value of the excitation of the crystal corresponds to a different mode, like an independent
oscillator.

We will show later that the eigenvectors for all values of K are orthogonal to one another. We can
write the energy as a function of K, since ω is a function of K:

 
K = n
1
2
ℏ  K 

We can consider the phonon to be a particle inside the crystal (because of wave-particle duality).
Thus, we can consider the phonon to have a momentum:


pK =ℏ K (4.31)

Note:

This is not a physical momentum in the standard sense. We call it the crystal momentum and can be
seen as a “potential momentum” which can be gained or lost by interacting with an external probe,
such as a photon or neutron.

Since neutrons are electrically neutral, they are much more effective at measuring the phonon
dispersion relation than other methods, such as photon scattering.

1 2
A neutron's energy is given by n= m v . This energy can be gained or lost when interacting
2
with a crystal.

1 ℏ2 2
m v 2= k (4.32)
2 2m

This is an inelastic scattering, meaning that the neutron's energy is not conserved.

However, since energy cannot be created or destroyed, we can use the conservation of energy and
momentum to relate the initial  E , k  and final  E ' , 
k '  energy-momentum:

– 13 –
Last Modified: 04/12/2006
 §4 – Vibrations :

From the conservation of energy:

E ' =E±ℏ  K  (4.33)

And, from the conservation of “momentum”:


k '=k G±
 K (4.34)

Where lowercase k represents neutron wavenumber and uppercase K represents phonon

wavenumber.

 .
 and  K
Thus, we can reconstruct both K

We can use the results from the measurement of the neutron's final energy and momentum, with
equations (4.33) and (4.34) to plot the experimental dispersion curves shown below for Sodium,
Silicon and Germanium.

– 14 –
Last Modified: 04/12/2006