Lecture4 Gen

• lectures accompanying the book: Solid State Physics: An
Introduction, by Philip Hofmann (2nd edition 2015,

ISBN-10: 3527412824, ISBN-13: 978-3527412822,
Wiley-VCH Berlin.
www.philiphofmann.net
1
Thermal properties
at the end of this lecture you should
understand....
• Lattice vibrations: harmonic oscillator, infinite and finite
chains of atoms
• The first Brillouin zone
• Heat capacity of the lattice: Einstein and Debye models
• Thermal conductivity
• Thermal expansion
• Allotropic phase transitions
• Melting
2
Thermal vibration of the atoms
• What is the frequency a

and amplitude of the
atomic vibrations?
• Consider the strongly
a-x a+x
simplified case of only
one oscillator.
• What is the spring F
constant?
3
interatomic potential (arb. units)
Lattice vibrations
interatomic force (arb. units)

repulsive
0
attractive
0
repulsive part
resulting
potential
grad (r)
attractive part
0 1 4 6 0 1 4 6
interatomic distance / equilibrium distance a interatomic distance r (equilibrium distance)
d (a) 1 d2 (a) 1 d 3
(a)
(x) = (a) + (x a) + (x a) +
2
(x a)3 + . . .
dx 2 dx 2 6 dx 3
d2 (a)
F (x a) ⇡ (x a) = (x a)
dx2
4
g how far one gets with this very simple picture!
A simple
e force constant harmonic
γ is equal to φ′′ (a)oscillator
and x is the displacem
mum position, the equation
equationofofmotion
motion is
d2 x
M 2 = −γx,
dt
solution
M is the mass of the vibrating ion. This leads to a harmon
x(t) = Acos(⇥t + )
43
1 angular frequency of the harmonic vibration
equency !
γ
ω= .
M
• Strong force constants and light
nergy for one harmonic oscillator is masses give rise to high
frequencies.
5
Estimate of vibrational amplitudes
Classical treatment using the equipartition theorem.
1 1 2
E = Mv +
2
x
2 2
6
⇥x amplitudes
Estimate of vibrational
⇤= 2
a0
Classical treatment using the equipartition theorem.
stress ⇥x a0 ⇥
Y = 1 = 1 =
strain
E = M v +0 x 2 a 2 x 2 a0
2 2
BT.⇥
This must be equal to k⇥ Y athe
When kinetic energy is zero
0
1 2
⇥xmax = kB T
2
⇥1/2
2kB T
xmax =
⇥
6
L 1
F =⇥x⇥x amplitudes
Estimate of vibrational
⇤ = ⇥x
2
a
⇤= 0 2
Classical treatment using theaequipartition theorem.
stress 0
⇥x a0 ⇥
Y = stress 1 = ⇥x 1
a = ⇥
YE=strain 2 0
=
= M v +02 x2 a x =2 a0
strain
2 a0 2x a0
BT.⇥
This must be equal to k⇥ Y athe
When kinetic energy is zero
0
⇥ ⇥ Y a0
1 1 22
⇥x max =
⇥xmax = kB T
T
22
⇥
⇥1/2
2kB
2k BTT 1/2
xmax==
xmax ⇥
⇥
L
and this is usually only a few percent 1
of the interatomic spacing.
= L 16
More sophisticated models
(but still all in 1D)
• An infinite chain with one atom per unit cell

• An infinite chain with two atoms per unit cell
• A finite chain of atoms
what are the allowed vibrational frequencies !?
7
The infinite one-dimensional chain
equ. distance a
a+u n -u n-1 a+u n+1 -u n
of 2π/a. The atoms at the lattice sites shall be connected with s
force constant
n-1 γ. If we take
n only nearest neighbour
n+1 interactions in
the equation of motion for atom n is
equ. of motion d2 u n
M 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
d2 u n
M 2 = −γ[2un − un−1 − un+1].
dt
where un is the displacement of the nth atom in the chain. T
solved by a kind of wave which is only defined on the lattice site
1
un (t) = √ uei(kan−ωt) ,
M
8 where k = 2π/λ is the one dimensional wave vector of the wav
equ. distance a
of 2π/a. The a+u atoms at the lattice
n -u n-1 a+usites
n+1 -ushall
n be connected wi
force constant
of 2π/a. Theγ. If weattake
atoms the only
latticenearest neighbour
sites shall interaction
be connected with s
theforce
equation
n-1 of γ.
constant motion forn atom
If we take n is neighbour
only nearest n+1 interactions in
2
d un
equ. of motion M d2 u2 n = −γ(un − un−1 ) + γ(un+1 − un )
M dt 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
or d22un
M d u2n = −γ[2un − un−1 − un+1].
M dt 2 = −γ[2un − un−1 − un+1].
dt
where
whereununisisthe
thedisplacement of the
displacement of the nth
nthatom
atomin inthethe chain.
chain. T
solved bybya akind
solved kindofofwave
wave which
which isis only
onlydefined
definedonon
thethe lattice
lattice sites
11 i(kan−ωt)
i(kan−ωt)
un(t)
(t) =
=√√ ueue , ,
M
M
equ. distance a
n -u n-1 a+usites
n+1 -ushall
n be connected wi
force constant
atoms the only
be connected with s
theforce
equation
n-1 of γ.
2
d un
M dt 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
or d22un
M d u2n = −γ[2un − un−1 − un+1].
M dt 2 = −γ[2un − un−1 − un+1].
dt
where
whereununisisthe
thedisplacement
displacement ofof the
the nth
nthatom
chain. T
solved bybya akind ofofwave which is only
onlydefined
defined
i(kanonon t)the lattice s
solved
ansatz kind wave un (t) = ue
which is the lattice site
11 i(kan−ωt)
i(kan−ωt)
un(t)
(t) ==√ √ ueue , ,
MM
equ. distance a
n -u n-1 a+usites
n+1 -ushall
n be connected wi
force constant
atoms the only
be connected with s
theforce
equation
n-1 of γ.
2
d un
M dt 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
or d22un
M d u2n = −γ[2un − un−1 − un+1].
M dt 2 = −γ[2un − un−1 − un+1].
dt
where
whereununisisthe
thedisplacement
the nth
nthatom
chain. T
onlydefined
defined
solved
11 wave number
i(kan−ωt)
i(kan−ωt)
un(t)
(t) ==√ √ ueue , ,
MM
equ. distance a
n -u n-1 a+usites
n+1 -ushall
n be connected wi
force constant
atoms the only
be connected with s
theforce
equation
n-1 of γ.
2
d un
M dt 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
or d22un
M d u2n = −γ[2un − un−1 − un+1].
M dt 2 = −γ[2un − un−1 − un+1].
dt position on chain
where
whereununisisthe
thedisplacement
the nth
nthatom
chain. T
onlydefined
defined
solved
11 wave number
i(kan−ωt)
i(kan−ωt)
un(t)
(t) ==√ √ ueue , ,
MM
equ. distance a
n -u n-1 a+usites
n+1 -ushall
n be connected wi
force constant
atoms the only
be connected with s
theforce
equation
n-1 of γ.
2
d un
M dt 2 = −γ(un − un−1 ) + γ(un+1 − un )
atom n dt
or
or d22un
M d u2n = −γ[2un − un−1 − un+1].
M dt 2 = −γ[2un − un−1 − un+1].
dt position on chain
where
whereununisisthe
thedisplacement
displacement of of the
the nth
nthatom
chain. T
onlydefined
defined
solved
11 wave number
i(kan−ωt)
i(kan−ωt)
u n (t)
(t) == √√ ueue , ,
inserting this gives MM
M ! 2 uei(kan !t)
= uei(kan !t)
[2 e ika
eika ] = 2 (1 cos ka)uei(kan !t)
2 i(kan !t) i(kan !t)
M ! ue = 2 (1 cos ka)ue
M !2 = 2 (1 cos ka) (works for any amplitude u)
9
ngth λ. Substituting this into the equation of motion gives
2 i(kan−ωt) −ika ika i(kan−ωt) i(k
e = γ[2
2 −e − e ]e
M ! = 2 (1 cos ka)
= −2γ(1 − coska)e
(works for any amplitude u)
s has a solution ifthis

we can
choose the ω = ω(k)
only be solved whensuch that
! !
2γ(1 − coska) γ ka
ω(k) = =2 |sin |.
M M 2
using that cos = cos2 sin2 =1 2sin2
ulting ω(k) are also plotted in Fig.
2 5.1.
2 The solutions
2 describ
ating along the chain. What is special about the waves is th
y defined on the actual lattice sites. The phase velocity of the
and the group velocity is ∂ω/∂k. Relations of the type (5.4)
a frequency or energy to a wave vector are called dispersion re
encounter them many more times, for example in connecti
nic states. For a long wave length, i.e. a small k, the sin in (
tituted
9
by its argument and one gets the linear relation
ngth λ. Substituting this into the equation of motion gives
2 i(kan−ωt) −ika ika i(kan−ωt) i(k
e = γ[2
2 −e − e ]e
M ! = 2 (1 cos ka)
= −2γ(1 − coska)e
(works for any amplitude u)
s has a solution ifthis

we can
choose the ω = ω(k)
only be solved whensuch that
! !
ω(k) = =2 |sin |.
M M 2
using that cos = cos2 sin2 =1 2sin2
ulting ω(k) are also plotted in Fig.2 5.1.
2 The solutions
2 describ
ating along the chain. What is special about the waves is th
y defined on in general
the actualwe have that
lattice sites.ω depends
The phase on velocity
k. of the
and the groupω(k) is called
velocity the dispersion
is ∂ω/∂k. relation
Relations of the type (5.4)
a frequency or energy to a wave vector are called dispersion re
encounter them many more times, for example in connecti
nic states. For a long wave length, i.e. a small k, the sin in (
tituted
9
by its argument and one gets the linear relation
solution un (t) = ue
i(kan t)
r
ka
dispersion relation !(k) = 2 | sin |
M 2
10
solution un (t) = uei(kan t)
r
ka
M 2
ω
0
-2π 0 2π
a a
10 k = 2π/λ
phase velocity and group velocity
general 1D wave A(x, t) = A0 e

i(kx !(k)t)
or here even A(x, t) = A0 cos(kx !(k)t)
in general we have that ω depends on k.

ω(k) is called the dispersion relation
11

i(kx !(k)t)

we define
!
the phase velocity vp =
k
11

i(kx !(k)t)

we define
!
k
@!
the group velocity vg =
@k
11

i(kx !(k)t)

we define
!
k
@!
the group velocity vg =
@k
11
the two are the same for !(k) = ck
examples of dispersion relations
vibrations in a 1D chain
r
ka
!(k) = 2 | sin |
2
ω
M
0
-2π 0 2π
a a
k = 2π/λ
12
vibrations in a 1D chain
r
ka
!(k) = 2 | sin |
2
ω
M
0
-2π 0 2π
a quantum mechanical particle a a
k = 2π/λ
~2 k 2
E(k) = ~!(k) =
2m
~k 2
!(k) =
2m
!
@!(k) ~k p
vg = = = k
@k m m
12
light in vaccum
⌫ =c
2⇡⌫ =c
2⇡
!
1
! =c
k
!(k) = kc k
in vacuum the dispersion relation of light is linear.

Light travels with c independent of the frequency.
13
light in vaccum
⌫ =c
2⇡⌫ =c
2⇡
!
1
! =c
k
!(k) = kc k
in vacuum the dispersion relation of light is linear.

Light travels with c independent of the frequency.
light in matter
!(k) = kc(!)
13
linear dispersion (acoustic phonons, photons)
!
one partial wave vp =
k
!
amplitude
k
distance
14
!
k
!
amplitude
k
distance
7 partial waves
partial wave #
15 distance
!
k
!
amplitude
k
distance
sum of partial waves (package) 7 partial waves

@!
vg =
@k partial wave #
amplitude
16 distance distance
!
k
!
amplitude
k
distance
• phase and group sum of partial waves (package)

velocities are the
same amplitude
• the wave packet

does not change
when travelling
17 distance
sine-like dispersion
!
k
!
amplitude
k
distance

@!
vg =
@k partial wave #
amplitude
!
k
!
amplitude
k
distance

@!
vg =
@k partial wave #
amplitude
!
k
!
amplitude
k
distance

@!
vg =
@k
partial wave #
amplitude
k2 dispersion
!
k
!
amplitude
k
distance

@!
vg =
@k
partial wave #
amplitude
i(kan t)
r
ka
M 2
22
i(kan t)
r
ka
M 2
0
-2π 0 2π
a a
22 k = 2π/λ
! !
= solution= 2 un (t) = ue
|sin |. i(kan (5.4)
t)
M M 2 r
ka
also plotted in
dispersion Fig. 5.1. The!(k)
relation solutions | sinwaves
= 2 describe |
M
chain. What is special about the waves is that they2
actual lattice sites. The phase velocity of the wave ν
velocity is ∂ω/∂k. Relations of the type (5.4), which
energy to a wave vector are called dispersion relations.
m many more times, for example in connection with
long wave length,
sound i.e. a small k, the sin in (5.4) can
rgument and onewave gets the linear relation
ω
!
γ
ω= ak = νk. (5.5)
M
0
ty and the phase velocity are -2πequal. This limit
0 cor- 2π
a
a
gation of sound waves with a velocity ν. The wave
k = 2π/λ
he 22atomistic structure becomes unimportant and the
! !
= solution= 2 un (t) = ue
|sin |. i(kan (5.4)
t)
M M 2 r
ka
also plotted in
dispersion Fig. 5.1. The!(k)
relation solutions | sinwaves
= 2 describe |
M
chain. What is special about the waves is that they2
actual lattice sites. The phase velocitystanding
of the wave ν
velocity is ∂ω/∂k. Relations of the typewave (5.4), which
energy to a wave vector are called dispersion relations.
m many more times, for example in connection with
long wave length,
sound i.e. a small k, the sin in (5.4) can
rgument and onewave gets the linear relation
ω
!
γ
ω= ak = νk. (5.5)
M
0
ty and the phase velocity are -2πequal. This limit
0 cor- 2π
a
a
gation of sound waves with a velocity ν. The wave
k = 2π/λ
he 22atomistic structure becomes unimportant and the
r
ka
!(k) = 2 | sin |
M 2
ω
0
-2π 0 2π
a a
k = 2π/λ
• The dispersion relation is periodic and the periodicity is one

reciprocal lattice vector!
23
=10a
2
k=
a 10a
• Changing k by one reciprocal lattice vector gives exactly the

same movement of the atoms.
24
=10a = 10 a
11
2 2 +2 2 11
k= k= =
a 10a 10a a 10a
• Changing k by one reciprocal lattice vector gives exactly the

same movement of the atoms.
24
r
ka
!(k) = 2 | sin |
M 2
ω
0
-2π 0 2π
a a
k = 2π/λ
25
The first Brillouin zone
0
-2π 0 2π
a a
first Brilloin zone
k = 2π/λ
The first Brillouin zone is the region of reciprocal space

which is closer to one reciprocal lattice point than to any
other (Wigner-Seitz cell in reciprocal space).
26
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
27
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
28
real reciprocal
space space
K
L
Γ
29
Two atoms per unit cell
b/2+u n -v n-1 b/2+v n -u n
n-1 n n+1
M1 M2
30
Two Atoms Per Unit Cell
We also brieflyb/2+u n -vthe

discuss b/2+v nwith
n-1situation -u n two atoms per unit cell as shown
in Fig. 5.3. The calculation of the vibrational frequencies is very similar to
n-1
the case of one atom per unitncell. The lattice constant
n+1 is now b, the length
of the reciprocal lattice vector
M 1is 2π/b. Of
M 2the two atoms per unit cell, one
is at the origin and one at b/2. We write down the forces on each atom in
a similar manner as above (see also Fig. 5.3) and obtain two equations of
motion
equations of motion
d2 u n d2 vn
M1 2 = −γ[2un − vn−1 − vn ] M2 2 = −γ[2vn − un − un+1],
dt dt
47
30

discuss b/2+v nwith
n-1
M 1is 2π/b. Of
motion
equations of motion
d2 u n d2 vn
dt dt
ansatz
47
un (t) = uei(kbn t)
vn (t) = vei(kbn t)
30

discuss b/2+v nwith
n-1
M 1is 2π/b. Of
motion
equations of motion
where un and vn are the displacements of the first and second atom in th
nth unit cell,
d2 unrespectively. This can again be
d2 vnsolved by wave-type function
of the formdt dt
1 ansatz
i(kbn−ωt) 47 1
un (t) = √ ue vn (t) = √ vei(kbn−ωt)
M1 M2
un (t) = uei(kbn t)
vn (t) = vei(kbn t)
When this is inserted into the equations of motion, we obtain a homogeneou
linear system of equations for equations,
two linear u and v two unknowns
2 −ikb 2 ikb
−ω M1 u = γv(1 + e ) − 2γu − ω M2 v = γu(e + 1) − 2γv.
The coefficient (system

matrix is of homogeneous
called the linearmatrix
dynamical equations)
of the system. A solu
30
When this is inserted into the equations of motion, we obtain a homogeneous
Two atoms
linear system of equations for u and v
per unit cell
2 −ikb 2 ikb
The coefficient matrix is called the dynamical matrix of the system. A solu-
2of dynamical matrixikb
(2 ! M1 )u
tion exists if the determinant (1 + e vanishes,
)v = 0 i.e.
! (e ikb 2
+ 1)u + (2 ! M! 2 )v = 0
! 2 ikb
!
2γ − ω M1 −γ(e + 1)
! −γ(1 + e−ikb ) 2γ − ω 2 M2 ! = 0
! !
This has a solution if the determinant of the coefficients vanishes. This

happens when
" # $" #2 %
2 1 1 1 1 4 2 kb
ω =γ + ± + − sin ,
M1 M2 M1 M2 M1 M2 2
and these solutions are also shown in Fig. 5.3. Again, the solutions have the
periodicity of the reciprocal lattice vector 2π/b, i.e. it is sufficient to know
them within the first Brillouin zone. Now we have two branches of solutions.
The
31 solution which goes to zero for small k is called the acoustic branch.
When this is inserted into the equations of motion, we obtain a homogeneous
Two atoms
linear system of equations for u and v
per unit cell
2 −ikb 2 ikb
The coefficient matrix is called the dynamical matrix of the system. A solu-
(2 ! M1 )u
tion exists if the determinant (1 + e vanishes,
)v = 0 i.e.
! (e ikb 2
+ 1)u + (2 ! M! 2 )v = 0
! 2 ikb
!
2γ − ω M1 −γ(e + 1)
! −γ(1 + e−ikb ) 2γ − ω 2 M2 ! = 0
! !
this has only a (non-trivial) solution when
This has a solution if the determinant of the coefficients vanishes. This
happens when 2 ! 2 M1 (e ikb + 1)
coefficient matrix # $" (1 + eikb# 2 % =0
"
1 1 1 1 ) 2 2 4 ! M 2
kb
ω2 = γ + ± + − sin2 ,
M1 M2 M1 M2 M1 M2 2
and these solutions are also shown in Fig. 5.3. Again, the solutions have the
periodicity of the reciprocal lattice vector 2π/b, i.e. it is sufficient to know
them within the first Brillouin zone. Now we have two branches of solutions.
The
When this is inserted into the equations of motion, we1obtain a homogeneous
Two
√
1
unof(t)equations
linear system =
atoms
ue i(kbn−ωt) per unit cell
for u and v vn (t) = √ ve i(kbn−ωt)
M1 M2
2 −ikb 2 ikb
When this is inserted into the equations of motion, we obtain a homogen
linear system ofmatrix
The coefficient equations for utheand
is called v
dynamical matrix of the system. A solu-
tion exists
2 if the (2 −ikb
determinant! M1 )u (1 + 2e vanishes,
)v = 0 ikb
i.e.
−ω M1 u = γv(1! + e ) − 2γu − ω M2!v = γu(e + 1) − 2γv.
! (e ikb 2 2 !
! 2γ − ω+M 1)u
1 + (2 +!1)M! 2 )v = 0
ikb
−γ(e
The coefficient matrix is called
! −γ(1 the
−ikb dynamical2 = 0 of the system. A s
matrix
+e ) 2γ − ω M2 !
tion exists if the determinant
this has only aof(non-trivial)
dynamical matrix
solution vanishes, i.e.
when
This has a solution !if the determinant of the coefficients ! vanishes. This
2 ikb
happens when ! 2γ − ω M2 1 −γ(e
! 2
M + 1) (e
! ikb
+ 1)
! 1 ! = 0
coefficient matrix
! −γ(1 + e −ikb
) 2γ − ω 2
ikb#2 2 M ! 2 % =0
(1 + e ) 2 4 ! Mkb
$
2
" # "
2 1 1 1 1 2
ω =γ + ± + − sin ,
This has a solution
M1 if the
M2 determinant
M1 Mof2 the coefficients
M1 M2 2vanishes.
happens when two solutions for every value of k
and these solutions are also shown $"in Fig. 5.3. Again, the solutions have
%1/2 the
periodicity of the1 reciprocal
" # lattice vector 2π/b,# 2i.e. it is sufficient to know
2 1 1 1 4 2 kb
ω = γthe first+Brillouin
them within ± zone.
γ Now+we have two− branches sin of solutions.
,
M1 M2 M1 M2 M1 M2 2
The
1 2
Twotheatoms
n this is inserted into perof unit
equations cell
motion, we obtain a homog
r system of equations for u and v
2 −ikb optical
2 branch ikb
−ω M1 u = γv(1 + e ) − 2γu − ω M2 v = γu(e + 1) − 2γ
coefficient matrix is called the dynamical matrix of the system. A

ω
exists if the determinant of dynamical matrix vanishes, i.e.

! !
! 2γ − ω 2 M1 −γ(eikb + 1) !
! −γ(1 + e−ikb ) 2γ − ω 2 M2 ! = 0
acoustic branch
! !
0
-2π 0 2π
b
has a solutionfirstifBrilloin
the zone determinant of the coefficients vanishes.
b
k = 2π/λ
pens when
" # $" #2 %1/2
2 1 1 1 1 4 2 kb
ω =γ + ±γ + − sin ,
M1 M2 M1 M2 M1 M2 2
32
Why “optical branch”?
typical vibrational frequency ~ 1012 Hz

typical vibrational energy ~ 10 meV
light:
2⇡c
= ⇡ 10-4 m
!
0
-2 0 2
b b
first Brilloin zone
k=2 /
33

light:
2⇡c
= ⇡ 10-4 m
!
0 2⇡
-2
b
0
first Brilloin zone
2
b k= ⇡ 6x104 m-1
k=2 /
33

light:
2⇡c
= ⇡ 10-4 m
!
0 2⇡
-2
b
0
first Brilloin zone
2
b k= ⇡ 6x104 m-1
k=2 /
2⇡
⇡ 6x1010 m-1
b
33

light:
2⇡c
= ⇡ 10-4 m
!
0 2⇡
-2
b
0
first Brilloin zone
2
b k= ⇡ 6x104 m-1
k=2 /
2⇡
⇡ 6x1010 m-1
b
33

light:
2⇡c
= ⇡ 10-4 m
!
0 2⇡
-2
b
0
first Brilloin zone
2
b k= ⇡ 6x104 m-1
k=2 /
2⇡
⇡ 6x1010 m-1
b
light E-field the same over very long
distance
+ - + - + - + - + - + -
33
A long chain of finite, but macroscopic length
What are the boundary conditions?
What do we do about the ends of the chain?
length l, fix the end atoms.
l
This gives only standing waves.

34
Periodic boundary conditions
Max Born and Theodore von Karman (1912)
chain with N atoms:
uN +n (t) = un (t)
35
chain with N atoms:
uN +n (t) = un (t)
N
1
A finite chain with no end!

35
• We want to get rid of the

surface restrictions, i.e. we L
want a solid which is finite in

size but has no surfaces (!).
• If we move by one crystal L
size L, we have to get the

same.
u(x, y, z) = u(x + L, y, z) = u(x, y + L, z) = u(x, y, z + L)
36


size L, we have to get the wrap

same. on
36


size L, we have to get the wrap

same. on
The surface was invented by the devil - Wolfgang Pauli

36
chain with N atoms:
uN +n (t) = un (t)
37
ce again but since it is made from small finite cry
scribe the properties of one of these finite crystal.
n-von Karman boundary conditions restrict the poss
chain with N atoms:
es in the crystal. First of all, it becomes clear tha
uN +n
velength for a chain of(t)
N= un (t)with a spacing of a
atoms
e precisely, webehave
wave must to require
the same that
when going N sites further
ikan ika(N +n)

e =e
ikN a
e =1
sible values of k become
37
2π
describe thePeriodic
propertiesboundary
of one of these finite crystal.
conditions
scribe the
Born-von Max
properties
Karman
of
boundary
one of these
conditions
finite
restrict
crystal.
Born and Theodore von Karman (1912) the possibl
n-von Karman boundary conditions restrict
aves in the crystal. First of all, it becomes clear that ththe poss
es in the crystal. chain
Firstwith N atoms:
wavelength for a chain of Nofatoms
all, itwith
becomes
a spacingclear
of atha
is
velength
More precisely, uN +n
for awechain
have to(t)
of N = unthat
atoms
require (t)with a spacing of a
the same that
ikan ika(N +n)
e =e
ikan ika(N +n)
e =e
ikN a
e =1
ikN a
e
possible values of k become =1
sible values of k become 2π
k= n
aN
37
2π
describe thePeriodic
propertiesboundary
of one of these finite crystal.
conditions
scribe the
Born-von Max
properties
Karman
of
boundary
one of these
conditions
finite
restrict
crystal.
Born and Theodore von Karman (1912) the possibl
n-von Karman boundary conditions restrict
aves in the crystal. First of all, it becomes clear that ththe poss
es in the crystal. chain
Firstwith N atoms:
wavelength for a chain of Nofatoms
all, itwith
becomes
a spacingclear
of atha
is
velength
More precisely, uN +n
for awechain
have to(t)
of N = unthat
atoms
require (t)with a spacing of a
the same that
ikan ika(N +n)
e =e
ikan ika(N +n)
e =e
ikN a
e =1
This restricts ikN
the a
possible k values
possible values of k become e =1
kN a = 2 m
sible values of k become 22π
k== mn
aN
aN
37
2π vibrations (m=0....N-1)
So there are N possible different
Finite chain with 10 unit cells and
one atom per unit cell
2
k=
ω
m
aN
0
-2π 2π 2π 2π 2π 2π
-8 -4 0 4 8
a 10a 10a 10a 10a a
first Brilloin zone
k = 2π/λ
• N atoms give N so-called normal modes of vibration.

38 • For long but finite chains, the points are very dense.
counting normal modes.....
2
boundary conditions k= m
aN
0
-2 2
0
chain with 1 atom / unit cell a
first Brilloin zone
a
N x 1 modes k=2 /
and N unit cells
(since we have N degrees of freedom)
39
2
aN # k-points
0
-2 2
0
first Brilloin zone
a
N x 1 modes k=2 /
and N unit cells
39
2
aN # k-points
0
-2 2
0
first Brilloin zone
a
N x 1 modes k=2 /
and N unit cells
# eigenvalues per k-point
39
2
aN # k-points
0
-2 2
0
first Brilloin zone
a
N x 1 modes k=2 /
and N unit cells
0
-2 0 2
chain with 2 atom / unit cell b
first Brilloin zone
b
N x 2 modes k=2 /
and N unit cells
(since we have 2xN degrees of freedom)

39
2
aN # k-points
0
-2 2
0
first Brilloin zone
a
N x 1 modes k=2 /
and N unit cells
# k-points
0
-2 0 2
chain with 2 atom / unit cell b
first Brilloin zone
b
N x 2 modes k=2 /
and N unit cells

(since we have 2xN degrees of freedom)
39
Quantised vibrations
• We have the vibrational frequencies and can treat the solid

as quantum mechanical oscillators of these frequencies.
• We shall call these quantized oscillations “phonons”.
40
One 1
harmonic 1
oscillator: quantum model
E = mvx + ⇥x
2 2
(7.27)
2 2
C = 3R (7.28)
⇥
The energy levels are quantized
E 1
En = n + h̄⌅E (7.29)
2
⇥
1
En = n + h̄⌅ (7.30)
2
⇥
⇤ h̄⌅E 1
Z= exp[ γ (n + )]
! (7.31)
k B T 2
ω=
hω n=0 .
M
En = nh̄⌅E (7.32)
one harmonic oscillator isT
pn = Ae
nh̄ E /kB
(7.33)
1ln Z 2 1 h̄⌅2 E
⇥E⇤E= = Mvx=+ h̄ γx . (7.34)
image source: wikimedia, author AllenMcC.
41 kT e E BT
/k 1
One 1
harmonic 1
oscillator: quantum model
E = mvx + ⇥x
2 2
(7.27)
2 2
C = 3R (7.28)
⇥
The energy levels are quantized
E 1
En = n + h̄⌅E (7.29)
2
⇥
1
En = n + h̄⌅ (7.30)
2
⇥
⇤ h̄⌅E 1
Z= exp[ γ (n + )]
! (7.31)
k B T 2
ω=
hω n=0 .
M
En = nh̄⌅E (7.32)
one harmonic oscillator isT
pn = Ae
nh̄ E /kB
(7.33)
1ln Z 2 1 h̄⌅2 E
⇥E⇤E= = Mvx=+ h̄ γx . (7.34)
image source: wikimedia, author AllenMcC.
41 kT e E BT
/k 1
long chain: quantum model
ω
0
-2π 2π 2π 2π 2π 2π
-8 -4 0 4 8
a 10a 10a 10a 10a a
first Brilloin zone
k = 2π/λ
42
have seen that a chain of N atoms can also be viewed as N ha
ors, each with long chain:
a different k =quantum
2πn/aN andmodel
vibrational frequenc
h a system the quantised energy levels have to be
! "
1
El (k) = l + h̄ω(k)
2
energy (arb. units)
ω(k) refers to the dispersion relation for the chain (see Fig. 5
} hω(k)
tations here lends itself to an alternative
and so on interpretation of k.
ω
up to
e viewed k as the one dimensional
l=2
l=1 2π wave vector. But here it b
N Na
nt that k takes also a role l=0 of a quantum number, just as l
2π 2π 2π 2π
ation of
1
Na k and
2
Na
3lNadescribes
4
Na one vibrational
k
0
state of the solid
-2π 2π
a
-8
10a
-4
2π
10a 0 4
2π
10a
8
2π
10a
2π
a
illator k which is excited to the level l. The interpretation

k = 2π/λ
of
first Brilloin zone
m number has a lot to do with the symmetry of the system. In

e spherical symmetry which gives rise to the quantum number
solids we have translational symmetry and the appropriate qu
is k.
42
have seen that a chain of N atoms can also be viewed as N ha
ors, each with long chain:
a different k =quantum
2πn/aN andmodel
vibrational frequenc
h a system the quantised energy levels have to be
! "
1
El (k) = l + h̄ω(k)
2
energy (arb. units)
ω(k) refers to the dispersion relation for the chain (see Fig. 5
} hω(k)
tations here lends itself to an alternative
and so on interpretation of k.
ω
up to
e viewed k as the one dimensional
l=2
l=1 2π wave vector. But here it b
N Na
nt that k takes also a role l=0 of a quantum number, just as l
2π 2π 2π 2π
ation of
1
Na k and
2
Na
3lNadescribes
4
Na one vibrational
k
0
a
state of the solid
-2π 2π
-8
10a
-4
2π
10a 0 4
2π
10a
8
2π
10a
2π
a
illator k which is excited to the level l. The interpretation

k = 2π/λ
of
first Brilloin zone
m number has a lot to do with the symmetry of the system. In

• The excitations
e spherical symmetryof which gives rise are
these oscillators to the quantum
called phonons. number
solids we have translational symmetry and the appropriate qu
•
is k.
The dispersion is often called a phonon dispersion curve.
42
Phonons <-> Photons
• Strong analogy: both bosonic excitations

• Both described by quantum mechanical harmonic
oscillators
• Wave-particle duality
43
!
k
!
amplitude
k
distance

@!
vg =
@k partial wave #
amplitude
Excitation of optical vibrations: particle picture
crossing point annihilation of photon

creation of phonon
! = ck
light dispersion (qualitatively)
Two conditions:
klight = kphonon
0
~!light = ~!phonon
-2 0 2
b b
k=2 /
light E-field the same over very long

distance
+ - + - + - + - + - + -
45
and the possibleThree
values of dimensional
k become solids
2π
1D k= n (5.6
aN
where n is an integer. We have seen that the vibrations are unaffected b
adding multiples of 2π/a (i.e. reciprocal lattice vectors) to k and therefor
we only get N possible different values for k and therefore also N differen
vibrational frequencies ω. The possible values of n or k can be chosen t
lie in the first Brillouin zone, i.e. −N/2 < n < N/2 or between 0 and N
if only positive values of k are desired. Physically, it does not make an
difference. The k values for a finite chain are illustrated in Fig. 5.4. Not
that for a real finite solid the number of atoms in one direction is very larg
Therefore the distances between the allowed k points is very small and th
finite vibrational frequencies resemble closely the continuum of states for th
infinite chain. This is quite intuitive: if we take N free atoms which ca
only move in one dimension, each atom has one degree of freedom. Th
total number of degrees of freedom is conserved when we put the atoms int
50
46
a three dimensional solid with two atoms per unit cell, the dynamical matrix
Three dimensional solids
and the possible values of k become
will be a six by six matrix, giving six vibrational frequencies ω for every value
of k. There will be three acoustic modes, 2π one with longitudinal polarisation
as in one dimension and1D two with k= n
transversal
aN polarisation. Similarly, (5.6
there
will also be three optical modes.
whereActually,
n is an integer.
the one We have
dimensional seen
k that
turns
cubic crystal with lattice spacing a and macroscopic the
into avibrations
real wave are unaffected
vector with b
three
adding multiples
components k inofthree
2π/adimensions
(i.e. reciprocal
but it lattice
does vectors)
of course to k its
keep andadditional
therefor
side length L and N unit cells in each direction:
weinterpretation
only get N possible
as a quantum different valuesFor
number. forakcrystal
and therefore also Nspacing
with a lattice differen
a
vibrational
in all threefrequencies
directions and ω. N TheL =
atoms aN
possible
in everyvalues of n orone
direction, k can
gets be chosen t
lie in the first Brillouin zone,2πi.e. −N/2 < n n<x 2π N/2nyor
2π between
nz 2π 0 and N
if 3D k = (k
only positive x , ky , k
values ofz ) k= are (n x , ny , nz Physically,
desired. )=( , it does not), make(5.8)
an
aN L L L
difference.
2
The k values for a finite chain are illustrated in Fig. 5.4. Not
The similarity is quite far-reaching. Even mathematically (in quantum field theory)
that for a real finite solid the number of atoms in one direction is very larg
photons can be viewed as excitations of quantum mechanical harmonic oscillators.
finite vibrational frequencies resemble53 closely the continuum of states for th
infinite chain. This is quite intuitive: if we take N free atoms which ca
only move in one dimension, each atom has one degree of freedom. Th
50
46
a three dimensional solid with two atoms per unit cell, the dynamical matrix
Three dimensional solids
and the possible values of k become
of k. There will be three acoustic modes, 2π one with longitudinal polarisation
as in one dimension and1D two with k= n
transversal
aN polarisation. Similarly, (5.6
there
whereActually,
n is an integer.
the one We have
dimensional seen
k that
turns
cubic crystal with lattice spacing a and macroscopic the
into avibrations
real wave are unaffected
vector with b
three
adding multiples
components k inofthree
2π/adimensions
(i.e. reciprocal
but it lattice
does vectors)
of course to k its
keep andadditional
therefor
side length L and N unit cells in each direction:
weinterpretation
only get N possible
as a quantum different valuesFor
number. forakcrystal
and therefore also Nspacing
with a lattice differena
vibrational
in all threefrequencies
directions and ω. N TheL =
atoms aN
possible
in everyvalues of n orone
direction, k can
gets be chosen t
lie in the first Brillouin zone,2πi.e. −N/2 < n n<x 2π N/2nyor
2π between
nz 2π 0 and N
if 3D k = (k
only positive x , ky , k
values ofz ) k= are (n x , ny , nz Physically,
desired. )=( , it does not), make (5.8)
an
aN L L L
difference. The k values for a finite chain are illustrated in Fig. 5.4. Not
that
2
real finite solid kNa = 2πn
The similarity is quite far-reaching. Even mathematically (in quantum field theory)
for acan the number of atoms
many indices (for example in the equations of motion)
photons be viewed as excitations of quantum
in oneharmonic
mechanical
direction is very larg
oscillators.
finite vibrational frequencies resemble53 closely the continuum of states for th
infinite chain. This is quite! intuitive: mβj if we take N free atoms which ca
only move in one M ẍα dimension,
nαi + each Φ
mβjatom x
nαi has one = 0.
mβjdegree of freedom. Th
3 x (#atoms / unit cell) x (unit cells) solutions
i.e. 3 x (#atoms / unit cell) solutions
50 for every k-point in the BZ
46
real reciprocal
space space
K
L
Γ
47
Phonons in 1D
optical branch
!i (k)
ω
acoustic branch branch

0
-2π 0 2π
b b
first Brilloin zone
k = 2π/λ
Phonons in 3D
3D wave
vector
!i (k)
branch
48
Phonons in 3D crystals: Aluminium
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
• Results from inelastic x-ray scattering / neutron scattering.
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
energy (meV)
30
K
L
Γ 20
10
0
Γ X K Γ L
k=0 k=0
49
40
X
energy (meV)
K 30
L a1
Γ 20
a2 10
a3
0
Γ X K Γ L
k=0 k=0
50
Phonons in 3D crystals: diamond
150
X
energy (meV)
K
100
L
Γ
50
Γ X K Γ L
k=0 k=0
• Acoustic and optical branches present.

51
Estimate ϒ from Young’s modulus
a
a +x a +x
a F
52
a
a +x a +x
a F
F
stress = 2
a
52
a
a +x a +x
a F
F
stress = 2
a
harmonic oscillator F = x
52
a
a +x a +x
a F
F
stress = 2
a
52
a
a +x a +x
a F
F
stress = 2
a
x
⇥= 2
a
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
Y =
✏
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
Y =
✏
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
Y =
✏
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
xa
Y = = 2 =
✏ a x a
52
a
a +x a +x
a F
F
stress = 2
a
x x
⇥= 2 ✏=
a a
xa
Y = = 2 =
✏ a x a
=Ya
52
=Ya
example with numbers:
Y=500x109 Nm-2
a=10-10 m
γ= 50 Nm-1
53
=Ya
example with numbers:
Y=500x109 Nm-2
a=10-10 m
γ= 50 Nm-1
and look again at the classical result for the amplitude (at 300 K)
⇥1/2
2kB T
xmax = ⇥ 1.3 10 11
m
53
Y = = =
strain a2 x a
quency1 Estimate ω from Young’s modulus 0
0
!
γ
ω= . ⇥ ⇥ Y a0
M
nergy for one harmonic oscillator is ⇥1/2
2kBremember
T
diamond
1 1 2
lead
xmax =
2
E = Mvx + γx . ⇥
mass 12 u2 2 207 u
1 vibration
Lof the
lassicalY motion, we can
950*10 9 Nmestimate
-2 15*10the =
amplitude
9 Nm -2
ω
13 12
L
e equipartition theorem of statistical mechanics. 0 T
This theorem
ω calc 8*10momentum
every generalised Hz 4*10 Hz coordinate which ap-
or position 0
2π/a0 4π/a0
ed in the Hamilton function (or in the total ⇧U

energy
0
in our case),
14 13C=
k
ω exp 0 to 2*10 0 to 1*10
ribute with a mean energy kB T /2 to the system. ⇧T Here we have
Hz Hz
coordinates and therefore the the mean energy of such an oscil-
• reasonable 1
the=highest 1
tact with a heat agreement
bath is kB T . AtE mvx + ⇥x2
2
displacement the
•
as only potential
most energy
important, and
again: so
light 2
elements and 2
strong forces give
high frequencies.
1 2
γx = k T 54
C = 3R
max B
Heat capacity of the lattice
55
General Theories
Thermodynamics
CV 0 for T 0 ⇥ ⇥
⌦Q ⌦E
no numerical value
at finite T CV = =
⌦T V ⌦T V
Classical Statistical
< E >= E = 3NA kB T
Mechanics
⇥
⌦E
CV = = 3R
⌦T V
Dulong-Petit law, numerical value
⌦U
independent of T
C=
⌦T
56 1 2 1 2
Heat capacity: classical statistical mechanics
⌦U
C=
so for a one classical dimensional ⌦Tharmonic oscillator we have
1 1 2
E = mvx + 2
x
2 2
so the mean energy is kBCT. = For3Ra three-dimensional oscillator
we have 3 kBT. For one mole of ions ⇥ we get
⇥ 1 ⇥
En = ⌦Qn + h̄⌥⌦E
<CEV >== E = 3N2= A kB T = 3RT
⌦T V ⌦T V
En = andnh̄⌥
so
< E >= E = 3NA kB T
pn = Ae nh̄⇥/kB T
⌦E
CV = h̄⌥ = 3R
⇥E⇤ = h̄⌦T V
e /kB T 1
⌦U
This is called the Dulong-Petit law.
57
C= h̄⌥
Comparison of the Dulong-Petit law to
experiment
-1 -1
77 K (JK ) 273 K (JK )
classical value 24.9 24.9
copper 12.5 24.3
aluminium 9.1 23.8
gold 19.1 25.2
lead 23.6 26.7
iron 8.1 24.8
sodium 20.4 27.6
silicon 5.8 21.8
values for one mole of substance
• At high temperatures the Dulong-Petit law works quite well,

also for metals.
58
Heat capacity of diamond
10 Dulong-Petit Value
Molar Heat Capacity, C (J K-1)
1
0.1
3
˜ T
0.01
0.001
ΘD=1880 K
2 3 4 5 6 2 3 4 5 6 2
10 100 1000
Temperature, T (K)
59
1 1 2
The Einstein model E = mv forx +the xheat capacity
2
2 2
C = 3R by independent oscillators
The solid’s vibrations are represented
⇥
1
En = n + h̄⌥E
2
⇥
⇤ h̄⌥E 1
Z= exp[ (n + )]
n=0
k B T 2
En = nh̄⌥E
pn = Ae nh̄ E /kB T
⌦ ln Z h̄⌥E
⇥E⇤ = = h̄
⌦kT e E /kB T 1
h̄⌥E
U = 3NA < E >= 3NA
eh̄ E /kB T 1
⇥ ⇥2 h̄ E /kB T
60 ⌦E h̄⌥E e
1 1 2
2
2 2
⇥
1
En = n + h̄⌥E
2
⇥
⇤
the mean energy for all the 3 Nh̄⌥ 1 in one mole is
A oscillators
E
Z= exp[ (n + )]
k T ⇥ 2
n=0 B
1
E⇥ = 3NA n⇥ + E
En = nh̄⌥2E
⌦ ln Z h̄⌥E
⇥E⇤ = = h̄
⌦kT e E /kB T 1
h̄⌥E
U = 3NA < E >= 3NA
eh̄ E /kB T 1
⇥ ⇥2 h̄ E /kB T
60 ⌦E h̄⌥E e
1 1 2
2
2 2
⇥
1
En = n + h̄⌥E
2
⇥
⇤
the mean energy for all the 3 Nh̄⌥ 1 in one mole is
A oscillators
E
Z= exp[ (n + )]
k T ⇥ 2
n=0 B
1
E⇥ = 3NA n⇥ + E
En = nh̄⌥2E
pn = Ae
and using the nh̄ E /kB T
Bose-Einstein distribution
⌦ ln Z 1 h̄⌥E
⇥n⇤ =
⇥E⇤ = =
e E /kBeTh̄
⌦kT 1 BT 1
E /k
⇥E
h̄⌥
U = 3N < E >= 13NA 1
⇥E⇤ = 3NA
A
eh̄+E /kB T E 1
e E /kB T 1 2
⇥ ⇥2 h̄ E /kB T
60 ⌦E h̄⌥E e
Z= exp[ (n +
The Einstein model for the heat
n=0
k T
capacity
B
En = nh̄⌥E
⇥ ⇥2 E /kB T
⇥⇥E⇤ e
C = 3NA = 3R
E
⇥T kB T (e E /kB T 1)2
V
⌦ ln Z h̄⌥E
⇥E⇤ = = h̄
⌦kT e E /kB T
3.0
h̄
• correct Dulong-Petit value for U =
2.5
3NA < E >= 3NA h̄
e E /k
molar heat capacity / R

high temperatures ⇥ ⇥
2.0 2
•
experiment
⌦E
zero heat capacity at zero CV = 1.5
h̄⌥E e
= 3R Einstein model h̄
temperature ⌦T V kB T (e
•
1.0
The Einstein frequency / h̄⌥E
E =
temperature is an adjustable 0.5
kB
parameter. 2
0.0
⌥ 2.5V 3.0
0.0 0.5
g(⌥) =
1.0 1.5 2.0
T/Θ 2⇧ 2 v 3
E
61 ⌅ max
⇥
1 low-temperature heat capacity
The EinsteinEnmodel:
= n+ h̄⌥E (7.22)
2 -1
10
⇥
⇤ 1
•
-5
h̄⌥E 10
log (molar heat capacity / R)

Z= exp[
but low temperature (n + )] 10
-9
(7.23)
k B T 2 -13
behaviour of the experiment
n=0 10
-17
experiment
10
En = nh̄⌥E T3
not correctly reproduced. 10
-21 (7.24)
Einstein model
behaviour inpexperiment,
n = Ae nh̄ E /kB T 10
-25
-29
(7.25)
10
exponential behaviour
⌦ ln Z hereh̄⌥E -33
⇥E⇤ = = 10
(7.26)
⌦kT eh̄ E /kB T 1 10 -37
10-41
h̄⌥E
U = 3NA < E >= 3NA 10-45 (7.27)
eh̄ E /kB T 1 0.01 0.1 1
⇥ ⇥2 h̄ E /kB T
log (T / ΘE)
⌦E h̄⌥E e
CV = = 3R (7.28)
⌦T V kB T (eh̄ E /kB T 1)2
⇥2 h̄⌥eEh̄ E /kB T
h̄⇤E E = (7.29)
= 3R 2h̄ E /kBkT
(7.38)
kB T (e 2
B
2eh̄ E /kB T + 1)
⇥ ⌥ V
g(⌥) = 2 2 3 (7.30)
h̄⇤E 2⇧ vh̄ E /kB T
= 3R ⌅ max e for very low T (7.39)
kB T g(⌥)h̄⌥ 62
Why does the Einstein model work at high T?
Why does it fail at low T?
• At high T the small spacing between the

energy levels is irrelevant.
• At sufficiently low temperature, the
energy level separation is much bigger
than kBT.
• Eventually all the oscillators are “frozen”

in the ground state. Increasing T a little
does not change this, i.e. it does not
change the energy.
63
The Debye model for the heat capacity
• General idea: use a better,
ω
more realistic description of
the crystal vibrations,
especially for those with 0
-2π 0 2π
a a
low energies. first Brilloin zone
•
k = 2π/λ
These are the acoustic,
very long wavelength
energy (arb. units)

vibrational models of the
crystal. } hω(k)
and so on
• The energy separation l=2

l=1
up to
2π
N Na
between the levels is l=0
2π 2π 2π 2π
VERY small for low k. 1
Na
2
Na
3
Na
4
Na k
64
kT e E B 1
The Debye model for the heat capacity
h̄⌅E
U = 3NA < E >= 3NA
eh̄ E /kB T 1
⇥ ⇥2
•
h̄ E /kB T
General idea: use a E
better, h̄⌅E e
CV = = 3R
(eh̄ 1)
ω
T kB T E /kB T
more realistic description of V
the crystal vibrations, h̄⌅E
especially for those with 0 E = 0 2π
low energies.
-2π
a kB
first Brilloin zone
a
•
k = 2π/λ
These are the acoustic, ⌅ = ⇥k
very long wavelength
nx ⇤ ny ⇤ nz ⇤
energy (arb. units)
k = (kx , ky , kz ) = (
vibrational models of the , )
crystal. L L L
} hω(k)
and so on
• The energy separation

g(⌅)d⌅ =
⌅2 V
d⌅
l=2
l=1
up to
2π
N Na
between the levels is 2⇤2π ⇥ 2π
2 3 l=0
2π 2π
VERY small for low k. 1
Na
2
⌅ DNa
3
Na
4
Na k
643N = g(⌅)d⌅
n E
(7.25)
ln Z h̄⌅E
⇥E⇤ = = h̄ /k T (7.26)
/2
and l = 0, 1, kT2, . . .e. These
E B 1
energy levels are dis-
h̄⌅E
U = 3NA < E >= 3NA h̄ /k T (7.27)
•
chain Now
of Nwe atoms
havecan
many also be
E viewed
eoscillators
B 1 asdifferent
with N harmonic
values of ω for
different ⇥
k = 2πn/aN
different values ofand ⇥2 vibrational
k. h̄ E /kB Tfrequencies ωn .
E h̄⌅E e
=
uantised = 3Rlevels have to
energy be (7.28)
T V! k T (eh̄ E /k B T 1)2
energy (arb. units)

" B
1
El (k) = l+ h̄⌅E
h̄ω(k) (5.7)
E2= (7.29)
} hω(k)
k B
and so on
up to
l=2
e dispersion relation for the chain (see Fig. 5.5(b)). l=1 2π
N Na
⌅ = ⇥k
s itself to an alternative interpretation 2π 2πof k.2π So 2π
l=0
far, (7.30)
1
Na
2
Na
3
Na
4
Na k
e one dimensional wavenx ⇤ ny ⇤ nBut
vector. z ⇤ here it becomes
k = (kx , ky , kz ) = ( , ) (7.31)
also a role of a quantum L number,
L L just as l. The
describes
65
one vibrational2
⌅ V
state of the solid. It is
frequencies given by (5.11). For one oscillator with frequency ω
the result is
h̄ω
mean energy for one oscillator ⟨E⟩ = h̄ω/k T
e B −1
(see 5.10). The zero point energy h̄ω/2 is neglected here sin
contribute to the heat capacity. For a three dimensional s
tempting to write the mean energy for all oscillators as
h̄ω ′
!
′
⟨E⟩ = 3 dω .
eh̄ω /kB T − 1
′
The factor of 3 stems from the fact that we want to work in thr
where there are three types of waves for a given ω: two differe
and one longitudinal. Here we assume that all three waves foll
sion ω = ν|k|. From the total energy we could calculate the h
But there are two problems: the first is that there may be m
in, say, the frequency interval ω1 +dω than in ω1 +dω. This mu
by a weighting factor g(ω) in the integral. This factor is called
states, for obvious reasons. The second problem is that we ha
the upper limit ωD for the integration, in other words, we hav
! ωD
66
g(ω)h̄ω
frequencies given by (5.11). For one oscillator with frequency ω
the result is
h̄ω
mean energy for one oscillator ⟨E⟩ = h̄ω/k T
e B −1
(see 5.10). The zero point energy h̄ω/2 is neglected here sin
contribute to the heat capacity. X For a three ~! dimensional s
i
mean energy for three dimensional = 3 for all
hEi energy
tempting to write the mean ~!
oscillators as
solid (sum over all phonons) ! i e i /k B T 1
h̄ω ′ ′
⟨E⟩ = 3 dω .
eh̄ω /kB T − 1
′
The factor of 3 stems from the fact that we want to work in thr
where there are three types of waves for a given ω: two differe
But there are two problems: the first is that there may be m
by a weighting factor g(ω) in the integral. This factor is called
the upper limit ωD for the integration, in other words, we hav
! ωD
66
g(ω)h̄ω
U =For
frequencies given by (5.11). 3None < E >=with
A oscillator 3NAfrequency
h̄ /k ω
the result is
e E B
⇥ h̄ω ⇥2 h̄
mean energy for one oscillator E
⟨E⟩ = h̄⌅E e
CV = e = 3R − 1
h̄ω/k B T
T
(see 5.10). The zero point energyVh̄ω/2 is neglected k B T (eh̄ E /
here sin
contribute to the heat capacity. X For a three ~! dimensional s
h̄⌅
i E
mean energy for three dimensional
tempting to write the mean = 3 for all
hEi energy ~! E =
oscillators as
solid (sum over all phonons) ! i e i /k B T kB1
h̄ω ′ ′
⟨E⟩ = 3 dω .
two transverse, one longitudinal wave,all with ⌅ = ⇥k
e h̄ω /kB T − 1
′
The factor of 3 stems from the fact that we want to nxwork

⇤ nin thr
y⇤ n
where there are three types k= (kx , kfor
of waves ) = ( ω: two
y , kaz given , differe
L
L
sion ω = ν|k|. From the total energy we could calculate 2 the h
⌅ V
But there are two problems: theg(⌅)d⌅ = there
first is that may
d⌅be m
2⇤ 2⇥3
by a weighting factor g(ω) in the integral. This⌅ Dfactor is called
3N =
g(⌅)d⌅
the upper limit ωD for the integration, in other 0 words, we hav
! ωD ⌅ max
66
g(ω)h̄ω g(⌅)h̄⌅
actor of 3 stems frequencies
from the given
factbythat Uwe=For
(5.11). 3N
want one
Ato <workE >=
oscillator 3N
inwith
three
Afrequency
dimens
h̄ /k ω
the result
e E B
there are three typesisof waves for a given ⇥ ω: two h̄ω different ⇥2transve h̄
mean energy for one oscillator
ne longitudinal. Here we assume that E three
⟨E⟩
all = waves h̄⌅
follow
E the e
dis
CV = e = 3R − 1
h̄ω/k B T
= ν|k|. From the total energy we couldT calculate
(see 5.10). The zero point energyVh̄ω/2 is neglected k
the B T
heat (e h̄ E /
capacit
here sin
ut there are two problems:
contribute to thetheheat
firstcapacity.
is that there
X For amay three
~!be more oscilla
dimensional
h̄⌅ s
i E
mean energy
y, the frequency for three
interval
tempting dimensional
ω1 +dω
to write hEiin
the than
mean =ω31 +dω.
energy for all~! This
E = must
oscillators be
as inclu
solid (sum over all phonons) e i /k B T k B 1
weighting factor g(ω) in the integral. This factor ! i is′ called ′the densit
h̄ω
⟨E⟩
, for obvious reasons. The second problem iseh̄ωthat = 3 /kB Twe
dω .
have to estab
two transverse, one longitudinal wave,all with ⌅ = ⇥k
′
− 1
pper limit ωDThe forfactor
the integration, in other words,
of 3 stems from the fact that we want to
we have work in thr
n x ⇤ n y ⇤ n
where there are three
! ωD types k= (kx , kfor
of waves
g(ω)h̄ω ) = ( ω: two
y , kaz given , differe
⟨E⟩ = 3
and one longitudinal. Here we assume dω that all three
L waves
L (5
foll
eh̄ω/k B T −1
0
2
⌅ V
e are looking for But there
g(ω) are ω
and two
D . problems: theg(⌅)d⌅first is that = there mayd⌅ be m
in, say, thethefrequency interval ω1 +dω than 2⇤ 2⇥3
e start by calculating density of states g(ω) forinaωthree
1 +dω.dimensio
This mu
by a weighting factor g(ω) in the integral. This
It is insufficient to do this in one dimension because g(ω) depends ⌅ D
factor is called
states, for obvious reasons. The second
mensionality of the problem and we want to compare to experime 3N =problem is that we ha
g(⌅)d⌅
the upper limit ωD for the integration, in other
0 words, we hav
Consider a cube of solid with a macroscopic! ωD ⌅sidemax length L. The
g(ω)h̄ω g(⌅)h̄⌅
nal solutions for such a cube are equivalent to those in one dimens
66
U ′= 3N < E >= 3N
actor of 3 stems from the fact that weFor
frequencies given by
⟨E⟩ (5.11).
= h̄ω
3 wantone
A
h̄ωdω
′ to
′
oscillator
work
/kB.T − 1
dωinwith
.three
Afrequency
e dimens
h̄ E /k ω
B
the ⟨E⟩
result =
is 3 e (5.1
there are three types of waves e forBa given
h̄ω ′ /k T
− 1⇥ ω: two h̄ω different ⇥2transve h̄
ne The
mean factor
energy
longitudinal. of
for 3
onestems
Here from
oscillator
we assume the fact
that E three
that
⟨E⟩
all = we want
waves to h̄⌅
work
follow
E inthe e
three
dis
actorwhere
of 3 stems
therethefrom the
are total fact
three energy C
that
V =
types ofwewaves we want to =
h̄ω/k
ework
forcalculate 3R
a given ω:
B T
in−three
1
two dimension
different
= ν|k|.
there are From (see types
three 5.10). ofThe
waveszerofor point a couldT
energyω:
given Vh̄ω/2
two k
the B
isdifferent T
heat
neglected (e h̄ E /
capacit
here sin
transvers
ut and one longitudinal. Here we assume that all three waves follow
ne theresion
are two
longitudinal.
ω =
problems:
contribute
Here
ν|k|. Fromwe the
theheat
to assume
the first
total that is that
capacity.
energyall there
X
three
we
Forwaves
could
amay three
~!be more
dimensional
follow
h̄⌅
calculate
i the oscilla
E the heatdispe s
y,=the mean
ν|k|.
energy
frequency
From
for three
interval
tempting
the
dimensional
to
total ω
write+dω
energy the than
wemeanhEi in
could = ω3
energy +dω.for
calculate all
~! This
E = must
oscillators
the heat be
as inclu
capacity.
B1 be more
1 1 /k T
solid
But(sum
thereover alltwo
are phonons)
problems: the first! isi that e ithere B k
may
weighting
t there arefactor
two g(ω) in the
problems: theintegral.
first is Thisthere
that factor may is′ be
h̄ω called
more the densit
oscillato
in, say, the frequency interval ω⟨E⟩ 1 +dω = 3than in′ ω1 +dω.dωThis ′
. must b
y,, for
thebyobvious
frequency reasons.
a weighting interval The
ω
factor one
two transverse, g(ω)
1
second
+dω than problem
in
in the integral.
longitudinal ω 1 +dω.iseh̄ωthat
wave,all with
B T we
This
/k
This ⌅factor − havebetoinclude
must
= ⇥kis called the
1 estab
pper limitfactor
weighting ωDThe for
g(ω)the inintegration,
factor of the integral.
3 stemsThe
in other
This
fromsecond
words,
factor
the factproblem is we
called
that we want
have
the
to density
work in thr
states, for obvious reasons. is thatn x we
⇤ have
n y ⇤ nt
, forthe obvious reasons.
upperwhere
limitthere The
ωD for second
are the
! ωD types
three k=
problem (kxis
in, k
of waves
g(ω)h̄ω
integration, that
for
other )words,
y , kazwe =have
given ( ω: we
to
two,establis
differe
have
pper limit ωDand for one ⟨E⟩ = 3
thelongitudinal.
integration,Here in other
we dω
words,
assume we
that have
all three
L waves
L (5
foll
e h̄ω/k T
BωD − 1
g(ω)h̄ω
!
sion ω = ν|k|. From 0 the total energy we could calculate the h
!⟨E⟩ ωD = 3 dω 2
g(ω)h̄ω ⌅ V
g(ω)
⟨E⟩ = are
and
3 ωtwo
D . problems: 0
h̄ω/k
ethedω
g(⌅)d⌅
B T
first is 1= there may
−that d⌅ be m
(5.1
Problems:
in, say, thethe frequency
0 e h̄ω/kB T − 1
interval ω1 +dω than 2⇤
inaωthree
2⇥3
1 +dω.dimensio
This mu
e start by calculating
and we are looking
1. Whatfor density
is g(ω) and ωD . of states g(ω) for
eItare by a weighting factor g(ω) in the integral. This
is insufficient toWhat
doandthis
is ωin
⌅ factor is called
looking
We startforby g(ω)
2. calculating D . one
the dimension
density of because
states g(ω) Dg(ω)
for adepends
three d
emensionality
start by
calculating
solid. It isof the the
problem
insufficient density
to and
do of
thiswestates
inwant
one 3N
g(ω)
to =
problem
for a
compare
dimension three
is
to
that
g(⌅)d⌅ we
dimension
becauseexperime
g(ω)
ha
d
the upper limit ωD for the integration, in other 0 words, we hav
It isthe
insufficient
Consider a cubetoofdosolid
dimensionality this
of the in problem
with onea dimension
macroscopic
and!we ωD
because
⌅
wantside to g(ω)
length
comparedepends
L. toTheexo
mensionality of the problem and we want to g(ω)h̄ω
compare
max
to g(⌅)h̄⌅
experiment
nal data.
66 solutions for such
Consider a cube
a cube of are
solidequivalent to those in side
with a macroscopic one length
dimens
U ′= 3N < E >= 3N
⟨E⟩ (5.11).
= h̄ω
3 wantone
A
h̄ωdω
′ to
′
oscillator
work
/kB.T − 1
dωinwith
.three
Afrequency
e dimens
h̄ E /k ω
B
the ⟨E⟩
result =
is 3 e (5.1
h̄ω ′ /k T
ne The
mean factor
energy
longitudinal. of
for 3
onestems
Here from
oscillator
we assume the fact
that E three
that
⟨E⟩
all = we want
waves to h̄⌅
work
follow
E inthe e
three
dis
actorwhere
of 3 stems
therethefrom the
are total fact
three energy C
that
V =
h̄ω/k
ework
forcalculate 3R
a given ω:
B T
in−three
1
two dimension
different
= ν|k|.
three 5.10). ofThe
energyω:
given Vh̄ω/2
two k
the B
isdifferent T
heat
capacit
here sin
transvers
ne theresion
are two
longitudinal.
ω =
problems:
contribute
Here
ν|k|. Fromwe the
theheat
to assume
the first
total that is that
capacity.
energyall there
X
three
we
Forwaves
could
amay three
~!be more
dimensional
follow
h̄⌅
calculate
i the oscilla
E the heatdispe s
y,=the mean
ν|k|.
energy
frequency
From
for three
interval
tempting
the
dimensional
to
total ω
write+dω
energy the than
wemeanhEi in
could = ω3
energy +dω.for
calculate all
~! This
E = must
oscillators
the heat be
as inclu
capacity.
B1 be more
1 1 /k T
solid
But(sum
thereover alltwo
are phonons)
may
weighting
t there arefactor
two g(ω) in the
first is Thisthere
h̄ω called
more the densit
oscillato
. must b
y,, for
thebyobvious
frequency reasons.
ω
factor one
1
second
+dω than problem
in
in the integral.
wave,all with
B T we
This
/k
must
= ⇥kis called the
1 estab
pper limitfactor
g(ω)the inintegration,
factor of the integral.
3 stemsThe
in other
This
fromsecond
words,
factor
called
that we want
have
the
to density
work in thr
⇤ have
n y ⇤ nt
upperwhere
limitthere The
ωD for second
are the
! ωD types
three k=
problem (kxis
in, k
of waves
g(ω)h̄ω
integration, that
for
other )words,
y , kazwe =have
given ( ω: we
to
two,establis
differe
have
thelongitudinal.
we dω
words,
assume we
that have
all three
L waves
L (5
foll
e h̄ω/k T
BωD − 1
g(ω)h̄ω
!
!⟨E⟩ ωD = 3 dω 2
g(ω)h̄ω ⌅ V
g(ω)
⟨E⟩ = are
and
3 ωtwo
D . problems: 0
h̄ω/k
ethedω
g(⌅)d⌅
B T
−that d⌅ be m
(5.1
Problems:
0 e h̄ω/kB T − 1
inaωthree
2⇥3
1 +dω.dimensio
This mu
and we are looking
1. Whatfor density
doandthis
is ωin
looking
We startforby g(ω)
the dimension
density of because
states g(ω) Dg(ω)
for adepends
three d
emensionality
start by
calculating
problem
to and
do of
thiswestates
inwant
one 3N
g(ω)
to =
problem
for a
compare
dimension three
is
to
that
g(⌅)d⌅ we
dimension
becauseexperime
g(ω)
ha
d
It isthe dN
insufficient(!)
the upper
a cubetoofdosolid
limit
this
ω D for
in problem
the integration,
onea dimension
in
because
other
0 words,
g(ω)
we
depends
hav
o
Consider
g(!) =dimensionality with
of the macroscopic
and!we ωD ⌅
wantside to length
compare L. toTheex
mensionality d!of the problem and we want to g(ω)h̄ω
compare
max
to g(⌅)h̄⌅
experiment
nal data.
Consider a cube
a cube of are
dimens
U ′= 3N < E >= 3N
⟨E⟩ (5.11).
= h̄ω
3 wantone
A
h̄ωdω
′ to
′
oscillator
work
/kB.T − 1
dωinwith
.three
Afrequency
e dimens
h̄ E /k ω
B
the ⟨E⟩
result =
is 3 e (5.1
h̄ω ′ /k T
ne The
mean factor
energy
longitudinal. of
for 3
onestems
Here from
oscillator
we assume the fact
that E three
that
⟨E⟩
all = we want
waves to h̄⌅
work
follow
E inthe e
three
dis
actorwhere
of 3 stems
therethefrom the
are total fact
three energy C
that
V =
h̄ω/k
ework
forcalculate 3R
a given ω:
B T
in−three
1
two dimension
different
= ν|k|.
three 5.10). ofThe
energyω:
given Vh̄ω/2
two k
the B
isdifferent T
heat
capacit
here sin
transvers
ne theresion
are two
longitudinal.
ω =
problems:
contribute
Here
ν|k|. Fromwe the
theheat
to assume
the first
total that is that
capacity.
energyall there
X
three
we
Forwaves
could
amay three
~!be more
dimensional
follow
h̄⌅
calculate
i the oscilla
E the heatdispe s
y,=the mean
ν|k|.
energy
frequency
From
for three
interval
tempting
the
dimensional
to
total ω
write+dω
energy the than
wemeanhEi in
could = ω3
energy +dω.for
calculate all
~! This
E = must
oscillators
the heat be
as inclu
capacity.
B1 be more
1 1 /k T
solid
But(sum
thereover alltwo
are phonons)
may
weighting
t there arefactor
two g(ω) in the
first is Thisthere
h̄ω called
more the densit
oscillato
. must b
y,, for
thebyobvious
frequency reasons.
ω
factor one
1
second
+dω than problem
in
in the integral.
wave,all with
B T we
This
/k
must
= ⇥kis called the
1 estab
pper limitfactor
g(ω)the in
factor of
integration,
the integral.
3 stemsThe
in other
This
fromsecond
words,
factor
called
that we want
have
the
to density
work in thr
⇤ have
n y ⇤ nt
upperwhere
limitthere The
ωD for second
are the
! ωD types
three k=
problem (kxis
in, k
of waves
g(ω)h̄ω
integration, that
for
other )words,
y , kazwe =have
given ( ω: we
to
two,establis
differe
have
thelongitudinal.
we dω
words,
assume we
that have
all three
L waves
L (5
foll
e h̄ω/k T
BωD − 1
g(ω)h̄ω
!
!⟨E⟩ ωD = 3 dω 2
g(ω)h̄ω ⌅ V
g(ω)
⟨E⟩ = are
and
3 ωtwo
D . problems: 0
h̄ω/k
ethedω
g(⌅)d⌅
B T
−that d⌅ be m
(5.1
Problems:
0 e h̄ω/kB T − 1
inaωthree
2⇥3
1 +dω.dimensio
This mu
and we are looking
1. Whatfor density
doandthis
is ωin
looking
We startforby g(ω)
the dimension
density of because
states g(ω) Dg(ω)
for adepends
three d
emensionality
start by
calculating
problem
to and
do of
thiswestates
inwant
one 3N
g(ω)
to =
problem
for a
compare
dimension three
is
to
that
g(⌅)d⌅ we
dimension
becauseexperime
g(ω)
ha
d
dN (!)
the upper limit ω for
D N (k) the integration, in other
0 words, we hav
It isthe
g(!) insufficient
Consider
mensionality
nal data.
a cubetoof
=dimensionality
solutionsd!of the
for such
dosolid
this
strategy:
problem
a cube
in problem
with
of the onea dimension
!(k)
and we
macroscopic
k(!)
and!we
want to
}N
ωD
because
(!)⌅
wantside to
g(ω)h̄ω
compare
max
g(ω)
length
compare
to
depends
L.
g(⌅)h̄⌅
experiment
toTheexo
66
Consider a cube of are
dimens
dimensions. We do not go into much detail because little additional physical
and we are
Vibrational density of states looking for g(ω)
insight is gained and the equations get quite messy because one has to keep
and ω
track of many indices. The ideaWe of thestart by
dynamical calculating
matrix
kx
introduced the de
before
nx
can be extended to three dimensional solids in order to accommodate not only
solid. It is insufficient to
more atoms per unit cell but also more directions of motion. For instance, for
n |k| do this i
thetwodimensionality
a three dimensional solid with atoms per unit cell, the ofdynamical
the problemmatrix
data. modes,
of k. There will be three acoustic ny
Consider
one with a cube
longitudinal ky
of solid
polarisationw
as in one dimension and two brational solutions
with transversal for Similarly,
polarisation. such a there cube
travelling
Actually, the one dimensional k turns waves. Thevector
into a real wave three with dime
three
1 2π/L
values
components k in three dimensions 1 which
but it does are
of given
course keep
2π/L (5.8),
its com
additional
interpretation as a quantum number. For a crystal with a lattice spacing a
(5.6)).
each
in all three directions and N state inWe
every can
atomscharacterised by think
direction, one getsof each vib
of
k = (kx , ky , kz ) =
2πk’s, (kx , ky ,nk
(nx , ny , nz ) = (
x 2π
z ),
nor, equivalently
y 2π nz 2π
), (5.8)
aN
possible states LN do L we L get for a g
2
The similarity is quite far-reaching. Even mathematically (in quantum field theory),
densely,
photons can be viewed as excitations of quantumthemechanical
totalharmonic
number of state
oscillators.
53
67 N
We start by calculating the density
ωD go into much detail because little additional physical
of states g(ω
dimensions. We do not g(ω)h̄ω
!
⟨E⟩ =Vibrational
3 solid.
insight is gained and the h̄ω/k
Itand
is
equations
we
density
insufficient
get
are
dωquite
looking
of
to do
messy states
this
because
for
in oneg(ω)
(5.14)
one
and
dimension
has to keep
ω
e T − 1
the
0 dimensionality
track of many indices. The ideaWe
B
of the of the
start
dynamicalby problem
calculating
matrix and we want
introduced the
beforeto
de
nx kx
okingcanfor g(ω) andtodata.
be extended ωthree
D . Consider
dimensional
solid. It
a cube
solids
is in of
order solid
insufficient to with
accommodate
to
a macroscopi
do not only
this i
more atoms the
y calculating per unitbrational
cell but
density solutions
alson
of states moreg(ω) for
forsuch
directions ofamotion.
a three cube |k| are
For equivalent
dimensional instance, for to
a three dimensionaltravelling
solid withthe twodimensionality
waves. atoms The
ufficient to do this in one dimension because g(ω) depends on
per three
unit cell, of
the the
dynamical
dimensional problem
wavematrixvect
will be a six by six values
matrix, giving
which six
are vibrational
given frequencies
(5.8), ω
completely for every value
analogou
ality of the problem and data.
we want
of k. There will be three acoustic modes, Consider
to
ny
compare
one with toa cube
experimental
longitudinal ky
of solid
polarisation w
er aascube of solid (5.6)).
with a We can
macroscopic think sideof each
length vibrational
L. The vi-state as
in one dimension and two
of are
brational
k’s,modes.
with transversal solutions
(kx , ky , kz )toor,those
polarisation.
equivalently
for such
Similarly,
by a triple of n’s
a cube
there
tionswillfor such
also a cube
be three optical equivalent in one dimension:
es. The three the
Actually, possible
dimensional
one travelling
states
wavekvector
dimensional N do waves.
turns we
can
into get
a only
real The
forwave
a given
have three
certain
vector with dime
highest |k| o
three
1 2π/L
arecomponents
given (5.8), densely,
k incompletely
three values
the
dimensions total
analogous
1
butwhich
number
it does
to oneare
of
of course given
states
dimension keep
2π/L
is (5.8),
just
its given
additional
(see com
by
interpretation as a
n think of each vibrational quantum number.
state as For
being a crystal
defined with bya lattice
a triple spacing a
in all three directions and each(5.6)).
N state We
atomscharacterised can by
in every direction, think one of
gets
N = many
4 each
πn , 3 vib
kz ) or, equivalently by a triple of n’s, (nx , ny , nz ). How 3
N do we get for a given of 2πk’s,|k|(k
highest
k = (kx , ky , kz ) = orx ,
n?k
(nx , ny , nz ) = ( y ,
If k
nthe
x 2π
z ) nor,
y 2π n
states
,
equivalently
z 2π
lie very
), (5.8)
or, expressed aN
possible in terms
states of |k|
LN doL we L get for a g
otal number of states is just given by the volume of the sphere
2
The similarity is quite far-reaching. Even mathematically (in quantum " field # theory),
3
total number of 4 densely,
photons can be viewed as excitations 3 of
total
quantumthe
number total
of
mechanical number
harmonic 4 of
L|k|
oscillators. state
states for a N = πn , states for a N= π ,
maximum n 3 maximum |k| 3 2π
53
in67terms of |k| N
total number of
" #3
4 L|k|
states for a N= π ,
maximum |k| 3 2π
61
68
total number of
" #3
4 L|k|
maximum |k| 3 2π
⇥
the dispersion ω(|k|) = ν|k| 61=
|k|
! "3
4 Lω V 3
N(ω) = π = 2 3ω ,
3 2πν 6π ν
L3 is the volume of the crystal. From this we can get the density
differentiation
dN ω2V
g(ω)dω = dω = 2 3 dω
dω 2π ν
have to address the question of the upper integral limit ωmax for
tever the nature of the excitations, the limit of the integral must
ch68 that we recover the correct number of normal modes, so for
total number of
" #3
4 L|k|
maximum |k| 3 2π
⇥
the dispersion ω(|k|) = ν|k| 61=
|k|
! "3
4 Lω V 3
N(ω) = π = 2 3ω ,
3 2πν 6π ν
L3 is the volume of the crystal. From this we can get the density
differentiation
dN ω2V
dω 2π ν
ch68 that we recover the correct number of normal modes, so for
total number of
" #3
4 L|k|
maximum |k| 3 2π
⇥
r,the dispersion
if we ω(|k|) = ν|k|
use the dispersion ω(|k|) = ν|k|
61=
|k|
"3 !
! "3
4 Lω 4 Lω V 3 V 3
N(ω) = N(ω)
π = π = 2 3
ω ,
= 2 3ω ,
3 2πν 3 2πν 6π ν 6π ν
3 3
here V = L
L is the volume is the
of volume of the From
the crystal. crystal. From
this we this
can we
getcan
theget the dens
density
f states by differentiation
differentiation
2
dN dN
ω 2
V ω V
g(ω)dω =g(ω)dω = dω =
dω = dω 2 3 dω2π 2 ν 3 dω
dω 2π ν
Now we have to address the question of the upper integral limit ωmax
have to address the question of the upper integral limit ω max for
5.14). Whatever the nature of the excitations, the limit of the integral m
tever thesuch
e chosen nature
thatofwe
therecover
excitations, the limit
the correct of the
number integralmodes,
of normal must so
Nchatoms
68 that we recover the correct number of normal modes, so for
in the solid we must have
total number of
" #3
4 L|k|
we use the dispersion
maximum ω(|k|)
|k| = ν|k| 3 2π
4
!
Lω
"3
V ⇥
the dispersion ω(|k|) =
N(ω) = π
r, if we use the dispersion ν|k|
ω(|k|) = ν|k|
61
|k| =
= 2 3ω , 3
3 2πν
"3 ! 6π ν
! "3
4 Lω 4 Lω V 3 V 3
3 N(ω) = N(ω) π =the =
π crystal. ω ,
=From ω , we can get the d
V = L is the volume 3 of
2πν 3 2πν 2 3
6π ν 6π ν 2 3this
es by differentiation
3 3
here V =
L is the volumeL is the
this we this
can we
getcan
theget the dens
density
2
differentiation dN ω V
dω 2π2 ν
dN dNω 2
V ω V
g(ω)dω = dω = dω
w we have g(ω)dω to address= thedωquestion
dω
= dω 2 3of
2π ν
dω2π
the2ν 3
upper integral limit ω m
Whatever
Now we have thetonature
addressofthe
thequestion
of the upper of the
integral integra
limit ωmax
have
sen to
suchaddress
that the
we question
recover of
the the upper
correct integral
number oflimit ω
normal for
modes,
max
tever
ms inthe
e chosen nature
thesuch
solid
thatofwe
we the excitations,
must have
recover the limit
the correct numberof the integralmodes,
of normal must so
Nchatoms
in the solid we must have#
total number of
" #3
4 L|k|
we use the dispersion
maximum ω(|k|)
|k| = ν|k| 3 2π
4
!
Lω
"3
V ⇥
the dispersion ω(|k|) =
N(ω) = π
r, if we use the dispersion ν|k|
ω(|k|) = ν|k|
61
|k| =
= 2 3ω , 3
3 2πν
"3 ! 6π ν
! "3
4 Lω 4 Lω V 3 V 3
3 N(ω) = N(ω) π =the =
π crystal. ω ,
=From ω , we can get the d
V = L is the volume 3 of
2πν 3 2πν 2 3
6π ν 6π ν 2 3this
es by differentiation
3 3
here V =
L is the volumeL is the
this we this
can we
getcan
theget the dens
density
2
differentiation dN ω V
dω 2π2 ν
dN dNω 2
V ω V
g(ω)dω = dω = dω
w we have g(ω)dω to address= thedωquestion
dω
= dω 2 3of
2π ν
dω2π
the2ν 3
upper integral limit ω m
Whatever
Now we have thetonature
addressofthe
thequestion
of the upper of the
integral integra
limit ωmax
have
sen to
suchaddress
that the
we question
recover of
the the upper
correct integral
number oflimit ω
normal for
modes,
max
tever
ms inthe
e chosen nature
thesuch
solid
thatofwe
we the excitations,
must have
recover the limit
the correct numberof the integralmodes,
of normal must so
Nchatoms
in the solid we must have#
U ′= 3N < E h̄ω>= 3N
⟨E⟩ (5.11).
= h̄ω
3 want
⟨E⟩
one
h̄ωdω′
A tooscillator
=B.Th̄ω/k
/k ′ work dω inwith
. threeAfrequency
e dimens
h̄ E /k B
ω
the ⟨E⟩
result =
is 3 e e − 1
BT − 1 (5.1
h̄ω ′ /k T
− 1⇥ ω: two h̄ω different ⇥2transve
(see 5.10). The zero point energy h̄ω/2 is neglected here h̄
sin
ne The
mean factor
energy
longitudinal. of
for 3
one
Herestems from
oscillator
we assume the fact
that E three
that
⟨E⟩
all = we want
waves to h̄⌅
work
follow
E inthethreee
dis
actorwhere of 3 stems
there from
contribute
are the
three tofactthe
types C
that
heat
V of = we want
capacity.
waves for toaFor=
h̄ω/k
ework
given 3R
a
B T
in−three
three
ω:
1 dimensional
two dimension
different s
=
there ν|k|. are From (see
three the
types total
5.10). ofTheenergy
waves zero we
point
for a could T
energy
given calculate
ω:
V h̄ω/2
two is thek B
neglected
different T
heat (e
here
transvers
h̄
capacit E
sin
/
and one tempting
longitudinal. to write
Here thewe mean
assume energy that for all all oscillators
three waves as follow
ut
ne there are two
longitudinal. problems:
contribute
Here we theheat
to assume
the firstthat is that
capacity.
all there
three! Forwaves amay three
′
be
followmore
dimensional
h̄⌅ the oscilla
dispe s
y,=the sion
tentativeω =
frequencymeanν|k|. From
energy
interval
tempting for
to the
ω
write +dω total
the energy
than
mean in ω
energywe could
+dω.for all
h̄ωcalculate
This = must
oscillators
E the heat
′ be as inclu
ν|k|. From the total energy
1 we ⟨E⟩
could = 3
calculate
1 Ethe
T −1k
dω
heat . capacity.
threeButdimensional
there aresolid two problems: the first! isethat h̄ω /kBthere
′
may
B be more
weighting
t there arefactor
two g(ω) in the
problems: the integral.
first is Thisthere
h̄ω calledmore the densit
oscillato
in, say, theThe frequency
factor of interval
3 stems ω
from 1 +dω
⟨E⟩ =
the 3than
fact thatin ωwe 1 +dω.
want dω This
to ′
. work must
in th b
y,, for
theby obvious
frequency reasons.
interval Theω +dωsecond than problem
in ω +dω. iseh̄ωthat
′
B T we
This
/k − havebetoinclude
must
= ⇥kω:
1 estab
a weighting
two factor
transversal,
where there are g(ω)
1
one in the
longitudinal
three types integral.
1
wave,all
of waveswith for a⌅
This factor
given is called
two differe the
pper limitfactor
g(ω)the
factor inintegration,
ofthe integral.
3 stemsThe from
in other
This
the
words,
factor is we
called have the density
states, forand obvious reasons.
one longitudinal. Here wefact
second assume thatthat
problem we want is that
all to
n
three x
work
we
⇤ wavesn in
have
y ⇤thr
n
foll t
upperwhere
sion
limit ωthere
= The
ωDν|k|. for second
are From
three
the
! ωD thetypes k=
problem
total
g(ω)h̄ω
integration, (kxis
of waves
energyin, kthat
for, kazcould
ywe
other we)words,
given=have
(calculate
ω: wetotwo ,establis
differe
the h
have
pper limit ωDand for But
one ⟨E⟩ =
thelongitudinal.3
integration, in
Here other
we words,
assume dω that we allhave
three
L waves
L (5m
foll
there are two problems: h̄ω/k T the first is that there may be
e !
0 the total energy
BωD − 1
g(ω)h̄ω
sion ω =the
in, say, ν|k|. From
frequency
!⟨E⟩ ωD = 3 interval ω +dω we
than could
indω ω1calculate
+dω.
2 Thisthemu h
g(ω)h̄ω 1 ⌅ V
e are looking by forBut
a g(ω)there
weighting
⟨E⟩ = are
and
3 ωtwo
D . problems:
factor g(ω) 0in the h̄ω/k
ethe integral.
dω B T
first is
g(⌅)d⌅ −that 1= there
This factormay d⌅
is be m
called
(5.1
in,
states,Problems:
say, the frequency
fortheobvious e
0 reasons.
h̄ω/kB T − 1
interval ω1 +dω
The second than inaωthree
problem 2⇤
1 +dω.
2⇥3
is that This
we muha
e start by
and we arecalculating
looking
1. Whatfor density
is g(ω) and ωD . of states g(ω) for dimensio
eItareis insufficient by
the a upper
weighting
toWhatdoandlimit
this factor
is ωin
ω D for g(ω) the inintegration,
the integral. in ⌅
This
other factor
words, is we
called
hav
looking
We start forby g(ω)
the dimension
density because
! ωD problem is thatthree
of states g(ω) Dg(ω)
for adepends d
emensionality
start by
solid. It is
calculating
of the
insufficient the
problem density
to and
do of
this we states
inwant
one g(ω) 3N
to for =
compare
dimension g(ω)h̄ω
a three g(⌅)d⌅
to
because dimension
experime
we
g(ω)
ha
d
⟨E⟩
the upper limit ωD for the integration, h̄ω/k = 3 in other 0T dω
words, we hav
It isthe
Consider insufficient
a cubetoofdosolid
dimensionality this
of the in problem
with onea dimension
macroscopic
and!we e⌅sidetolength
0 because
want B −
g(ω) 1 depends
compare L. The
to ex o
ωD
mensionality of the problem and we want to compare g(ω)h̄ωmax
to g(⌅)h̄⌅
experiment
nal data.
Consider
and alooking
we aarecube cube of arefor equivalent
solid with
g(ω) to
. those in side
anda ωmacroscopic one lengthdimens
the nature of the excitations, the limit of the integral must
What is the upper integration limit?
hat we recover the correct number of normal modes, so for
olid we must have
# ωD
3N = 3 g(ω)dω
0
3 2N 3
ωD = 6π ν
V
Debye frequency and the corresponding temperature ΘD =
the Debye temperature.
14) becomes
2 ωD
ω V h̄ω 3V h̄ ω2
#
dω = dω (5.15)
2π 2 ν 3 eh̄ω/kB T − 1 2π 2 ν 3 0 eh̄ω/kB T − 1
70
the nature of the excitations, the limit of the integral 2must
What is the upper integrationdN limit?⇤ V
hat we recover the correct number of normal modes, so for
g(⇤)d⇤ = d⇤ =
olid we must have
# ωD
d⇤ 2⇥ 2 3
3N = 3 g(ω)dω ⇤ Debye angular

D
1 0 frequency
3N =Θ D = h̄ωD
g(⇤)d⇤
/k Debye temperature
B
3 2N 3 0
ωD = 6π ν
V ⇤ kNamax= 2πn
g(⇤)h̄⇤
Debye frequency and the corresponding
<
E >= M3 ẍ +
!
temperature
Φ mβj
x h̄ =
/k0.
Θ
T
D =
14) becomes
α nαi
0
mβj nαi
e mβj
B
⇤
2 ωD
D
2 g(⇤)h̄⇤
ω V h̄ω 3V h̄
3 ω
#
< dω E
= >= dω (5.15)
2π 2 ν 3 eh̄ω/kB T − 1 2π 2 ν 3 0 eh̄ω/kB Te−h̄1 /kB T
70 0 1
the nature of the excitations, h̄⌅
the
E limit of the integral must
What is the upper E = integration limit? 2 (7
hat we recover the correct number kB dN of normal modes, ⇤soVfor
g(⇤)d⇤ = d⇤ =
olid we must have
use the dispersion ω(|k|)
# ωD = ν|k|
⌅ = ⇥k d⇤ 2⇥ 2 3(7
3Nk==(k z ) ="
3 x , ky ,!kg(ω)dω( , ⇤ ) Debye angular (7
3L L L D
1 Lω04 V frequency
3N =
3
N(ω) = π = ω ,
3 2πν ⌅ 2 V 6π
g(⌅)d⌅ =
2ν 3
Θ = h̄ω /k
d⌅ D D
g(⇤)d⇤
BDebye temperature (7
3 2N 3
2⇤ 2 ⇥ 3 0
3 ωD =of6π ⌅ν
= L is the volume the crystal.
V D ⇤
From this we can get the dens
kNa = 2πn
by (no
differentiation 3N
more vibrational = than
modes max
three times atoms in crystal) (7
g(⌅)d⌅ g(⇤)h̄⇤
<
E >=
⌅dN
0
M3 ẍ ω +
2
V
!
temperature ΘD =
Φ x h̄ =
/k
mβj
0. T
14) becomes <g(ω)dω
E >= =
3 dω
max
dω =
α nαi
g(⌅)h̄⌅
2π02Tν 3
dω d⌅ e
mβj nαi mβj
B
(7
1
0 e h̄
⇤
/k B
we 2have to address the question N ωthe

2of
D
3 upper2 integral g(⇤)h̄⇤
limit ωmax
= 6⇤
3 (7
3
ω V h̄ω ⌅D 3V h̄ ω
# D
⇥
Whatever the <
naturedω
of E
=
the >=
excitations,
V the limit of dω
the (5.15)
integral m
2π 2 ν 3 eh̄ω/k T − 1
B 2π
n such that we recover the correct
2ν 3
4
0
70
e
number
⇥
0
h̄ω/k T −
3 of eh̄ 1
normal
B/k B T
modes, so1
the nature of the excitations, h̄⌅
the
E limit of the integral must
What is the upper E = integration limit? 2 (7
hat we recover the correct number kB dN of normal modes, ⇤soVfor
g(⇤)d⇤ = d⇤ =
olid we must have
use the dispersion ω(|k|)
# ωD = ν|k|
⌅ = ⇥k d⇤ 2⇥ 2 3(7
3Nk==(k z ) ="
3 x , ky ,!kg(ω)dω( , ⇤ ) Debye angular (7
3L L L D
1 Lω04 V frequency
3N =
3
N(ω) = π = ω ,
3 2πν ⌅ 2 V 6π
g(⌅)d⌅ =
2ν 3
Θ = h̄ω /k
d⌅ D D
g(⇤)d⇤
BDebye temperature (7
3 2N 3
2⇤ 2 ⇥ 3 0
3 ωD =of6π ⌅ν
= L is the volume the crystal.
V D ⇤
From this we can get the dens
kNa = 2πn
by (no
differentiation 3N
more vibrational = than
modes max
three times atoms in crystal) (7
g(⌅)d⌅ g(⇤)h̄⇤
<
E >=
⌅dN
0
M3 ẍ ω +
2
V
!
temperature ΘD =
Φ x h̄ =
/k
mβj
0. T
14) becomes <g(ω)dω
E >= =
3 dω
max
dω =
α nαi
g(⌅)h̄⌅
2π02Tν 3
dω d⌅ e
mβj nαi mβj
B
(7
1
0 e h̄
⇤
/k B
we 2have to address the question N ωthe

2of
D
3 upper2 integral g(⇤)h̄⇤
limit ωmax
= 6⇤
3 (7
3
ω V h̄ω ⌅D 3V h̄ ω~!D = kB ⇥D
# D
⇥
Whatever the <
naturedω
of E
=
the >=
excitations,
V the limit of dω
the (5.15)
integral m
2π 2 ν 3 eh̄ω/k T − 1
B 2π
n such that we recover the correct
2ν 3
4
0
70
e
number
⇥
0
h̄ω/k T −
3 of eh̄ 1
normal
B/k B T
modes, so1
0
The Debye model
⌅ for the heat capacity
g(⌅)h̄⌅
max
< E >= 3 d⌅
0
so that
eh̄ /kB T 1
⌅ D
g(⌅)h̄⌅
< E >= 3 d⌅
0 eh̄ /kB T 1
2N
3
⌅D = 6⇤ ⇥3
V
⇥3
12R⇤ 4 T
CV =
5 D
16
71
0
0
The Debye
which results in model
⌅ for the heat capacity
g(⌅)h̄⌅ max
< E >= 3 3 2 N 3T d⌅
ω so that
0 D = 6πeh̄ /kνB 1 (
⌅ D V
ωD is called the Debye angular frequencyg(⌅)h̄⌅ and the corresponding temper
< E >= 3 d⌅
ΘD = h̄ωD /kB is called the Debye 0
h̄ /kB T
temperature.
e 1
With this (4.41) becomes
2N 3
! ωD 2 ⌅D = 6⇤
3
⇥ ! ωD
ω V h̄ω V 3V h̄ ω3
⟨E⟩ = 3 2 3 h̄ω/k T
dω = 2 3 h̄ω/k T
dω (
0 2π ν e B − 1 2π ν⇥ 3 0 e B − 1
12R⇤ T 4
C V =
5 B T andDxD = ΘD /T this becomes
and with the substitution x = h̄ω/k
4 4 xD 3
#3 ! xD
x3
"
3V kB T x T
!
⟨E⟩ = 3 x
dx =169NkB T x
dx. (
2 3
2π ν h̄ 0 e −1 ΘD 0 e −1
From this the heat capacity of the solid can be determined by differenti
with respect to temperature. Instead of writing down an expression fo
heat capacity, we focus on the low and high temperature limits.
71
For high temperatures x in (4.49) is small and and the exponential
0 ωD 0
0
#
nwhich The
results Debye
in
3N = 3model
0
⌅ for
g(ω)dω the heat
g(⌅)h̄⌅
capacity max
< E >= N3 3 2 N 3 d⌅
sults in
3
ωD = 6π 2 ω3 so
ν0 D = that
6π
e h̄ /kνB T 1 (4.47) (
V ⌅ V
3 2N 3 D
ωDebye
D is called the Debye ωDangular ν
= 6π frequency g(⌅)h̄⌅
and the corresponding temper
angular frequency < E >=and 3V the corresponding d⌅ temperature
ΘD = h̄ωD /kB is called the Debye
is
lledcalled
the Debye the Debye
frequency temperature. 0
and the corresponding
h̄ /kB T
temperature.
e 1
temperature ΘD =
is Withthe
called thisDebye(4.41) becomes
temperature.
4.41) becomes 2N 3
this (5.14) becomes ! ωD 2 ⌅D = 6⇤
3
⇥ ! ωD
ω V h̄ω ! ωD V 3V h̄ ω 3
D ⟨E⟩
2 = 3
ω# ωVD 2 h̄ω 2π 2 ν 3 eh̄ω/kB3V T −
dω =
h̄ 1 # ωD 2π 2 ν⇥ ω32
3
h̄ω/k T
dω (
ω V 0 h̄ω dω = 3V h̄ 4 ω 30 edω B (4.48)−1
= 32π 2 ν 3 eh̄ω/k BT − 1 dω = 2π 2 12R⇤
ν 3 e T
h̄ω/k B T −dω 1 (5.15)
0
2
2π ν e 3 h̄ω/k B T − 1 CV =2π ν 00 e2 3 h̄ω/k B T −1
and with the substitution x = h̄ω/k 5 B T andDxD = ΘD /T this becomes
ubstitution
substitute
h the substitution x = xh̄ω/k =!h̄ω/kB TB T
andandxx DD= = h̄ω
ΘDD /T/k B T becomes
this this
3
becomes
4 4 xD 3 xD
x3
" #
3V kB T x T
!
4 ⟨E⟩ 4 x = 4 x 3 3
3 dx! = 169Nk
" 33# BT xxDD 3 3 dx. (
3V kB T 2πx ν h̄ x 0 e −1 T xDx
#
x x
D
" #
T T Θ e − 1
! D 2 3
!
=
B dx = 9Nk T dx.0 (5.16)
2π
3 2 ν 3 h̄3
0x − 1 e x dx
− 1 = 9Nk B TB
Θ e x −
x−1 1 dx. (4.49)
3 h̄ e
From0this the heat capacity of the solid Θ D
D 0
can e
0 be determined by differenti
s the heat
with capacity
respect of the solid can
to temperature. be determined
Instead by down
of writing differentiation
an expression fo
eat capacity
pect of theInstead
solid can be determined by differentiation
heat capacity, we focus on the low and high temperatureforlimits.
to temperature. of writing down an expression the
temperature.
acity, we focus onInstead
the low of writing
and high 71 down an
temperature expression for the
limits.
0 ωD 0 ωD
0
#
The 3N = 3 3N
⌅ =
Debye model max3
for
g(ω)dω g(ω)dω
the heat capacity
nwhich results in 0 0 g(⌅)h̄⌅
< E >= N3 3 2 N 3 d⌅
which
sults in results inωD = 6π
3 2 3 so
ω
ν0 D = that
6π
e h̄ /kνB T 1 (4.47) (
V ⌅ V
3 2 N 33 D 2N 3
ωDebye
D is called the Debye ωDangular νωD = 6πg(⌅)h̄⌅
= 6π frequency and ν the corresponding temper
angular frequency < E >=and 3V the corresponding V d⌅ temperature
ΘD = h̄ωD /kB is called the Debye
is
lledωcalled the Debye temperature. 0
h̄ /kB T
temperature.
e 1
D is called the Debye frequency and the corresponding Θ
the Debye frequency and the corresponding temperature D =
temperature Θ
is Withthe
called this (4.41)
Debye becomes
temperature.
4.41)h̄ωDbecomes
/kB is called the Debye temperature. 2N 3
this (5.14) becomes
With thisωD(5.14) 2 becomes
⌅ 3
D = 6⇤ ⇥ ! ωD
ω V h̄ω V 3V h̄ ω 3
!
! ωD
D ⟨E⟩
2 =
ω# ωVD 2 h̄ω 3 2 3 h̄ω/k 3V
T h̄ dω = 2 ω3
3
h̄ω/k T
dω (
ω V 0# ωD2π h̄ωω 2νdω e = −
3V 1
h̄ ωD 2π ν 2 #
ω⇥ edω − 1
2 (4.48)
B
# B
3 ω
= 32π 2⟨E⟩ 3 h̄ω/k T V dω =h̄ω2 12R⇤ 3 4 3V h̄B T dω (5.15)
T
h̄ω/k
0 D
ω
ν e
2π=2 ν33 eh̄ω/k
B − B T1− 1
2 3 C 2π
h̄ω/k =2π
T
ν2 ν 3 dω
0 = e
eh̄ω/kB T −−
2 3 1 1 h̄ω/k T
dω (5
0 2π ν e V − 1 0 2π ν e −becomes
1
5
B B
and with the substitution 0 x = h̄ω/k B T and D
x D =0 Θ D /T this
ubstitution
substitute
h the substitution x = xh̄ω/k=!h̄ω/k B TB T and andxx DD= = h̄ω
ΘDD /T/k B T becomes
this this becomes
and with3V thek 4substitution
T 4 xD
x x
3 = h̄ω/kB T and xD = Θ
"
T
# 3 D /T
! xD this 3becomes
x
B
4 ⟨E⟩ 4 x =4 x 3 3
3 dx ! =169Nk" 33# BT xxDD 3"3 dx. (
3V kB T 2πx ν 4h̄ x 4 0 xDe −31 T xDx 3 0 xD e −
#
x x
D
" #
T T Θ 3 1
! D 2 3
! !
3V k T x T x
# #
=
B
3 B
x dx dx
= =
9Nk 9Nk TB T x
dx.dx. (5.16)(4.49)
3 h̄2π
3 2ν⟨E⟩
3 h̄ =
e e
0x −21 3 3 − 1 x
B dx Θ Θ= D 9Nk B
0 T e
e −
x−1 1 x
dx. (5
From0this the 2πheatν h̄ capacity
0 e − of1the solid
D can0 beΘdetermined
D 0 e by− differenti
1
s the heat
with capacity
respect of the solid can
to temperature. be determined
Instead of writingby down
differentiation
an expression fo
eat
pect capacity
From
to this theofheat
temperature. theInstead
solid can
capacityof be
thedetermined
ofwritingsolid
down by differentiation
cananbeexpression
determinedfor by
thedifferentia
heat capacity, we focus on the low and high
withwerespect
temperature. toInstead
temperature. Instead
of writing ~!
of writing =
temperature
D k B ⇥D limits.
down an expression
for the fo
acity, focus on the low and high 71 down an
temperature expression
limits.
The Debye
# ω D 2
ω V model h̄ω for the heat
3V h̄
# ωDcapacity
ω2
⟨E⟩ = 3 2 ν 3 eh̄ω/kB T − 1
dω = 2ν 3 h̄ω/kB T − 1
dω (5
0 2π 2π 0 e
and with the substitution x = h̄ω/kB T and xD = ΘD /T this becomes

4 4 xD 3
"3 # xD
x3
!
3V kB T x T
#
⟨E⟩ = dx = 9NkB T dx. (5
2π 2 ν 3 h̄3 0
x
e −1 ΘD 0
x
e −1
From this the heat capacity of the solid can be determined by differentia
with respect to temperature. Instead of writing down an expression fo
heat capacity, we focus on the low and high temperature limits.
For high temperatures x in (5.16) is small and and the exponential f
tion in the integral can be approximated by 1+x. The integral is then m
over x2 and the resulting energy is ⟨E⟩ = 3NkB T . For one mole of a
this is equal to 3RT , i.e. it is the Dulong Petit result. This was of co
expected from the result in the Einstein model: the Dulong Petit resu
always reached at sufficiently high temperatures.
72
62
ωD2
0 2π 2 ν 3 eh̄ω/kB T − 13N ν3 0
2π g(ω)dω eh̄ω/k BT − 1
The Debye
and with
#
ω V model
=3
ωDthe 2substitution
h̄ω 0 for
x = the
h̄ω/k heat
T
3V h̄
B and capacity
ωDx
#
D =ω Θ2D /T this be
⟨E⟩ = 3 dω = dω (5
substitution
hich results in 0 x = 2πh̄ω/k
4 4 B T
2 ν 3 eh̄ω/k B Tand
x
− 13
xD =2πΘ2 ν 3/T this
D 0 ! e becomes
h̄ω/k"BT − 1
3 xD 3
3V kB T x T x
# D
#
⟨E⟩ = 3 2 N dx
3 = 9Nk B T
3 2π 2 ν 3 h̄ ωx = 6π ν
3
4and4 with
# xDthe substitution = h̄ω/k T and x = Θ /T this becomes
e x −! 1 3 Θ ex−1
D B
" D x D D3
kB T x 0
T x 0
#
V D
is called
3 the Debye
x
dx = 9N
frequency k
andB T
the corresponding dx.
temperature
3 x Θ (5.16
=
ν h̄ 0 3Ve kB−T 1
D2 3 From this
4 the
4 xD capacity
heat 3 of ΘD
the solid can e −1 x
be xD
determined
3 D by di
! "
x T
# #
0
ωD /kB⟨E⟩ = respect
is with
called the Debye
3 dx
temperature.
2 ν 3 h̄ to temperature.
x
= 9Nk B T
Instead of writing down
x andx. (5
expres
2π
With this (5.14) becomes 0 e − 1 Θ D 0 e − 1
e heat capacity
heat capacity, of the we solid
focus can
on thebelow determined by differentiatio
and high temperature limits.
From this
to temperature.the
# ωFor
D heat
high
ω V Instead
2 capacity
temperatures
h̄ω of thex solid
of writingin can
# ωbe
(5.16)
3V h̄ down determined
Dis small and and
anω 2expressionby differentia
the for
expon
th
⟨E⟩ = respect
with 3tion in tothe temperature.
integral can dω
be =
Instead of
approximated writing bydown
1+x. an dω (5.15) fo
expression
The integral is
y, we focus 0 on the low and high temperature
heat capacity,
2π 2 ν 3 eh̄ω/k
we focus
B T − 1
on the low
2π 2 ν 3
0
and high
eh̄ω/k B
limits.
T −
temperature
1
over x2 and the resulting energy is ⟨E⟩ = 3NkB Tlimits. . For one mo
temperatures
nd high
withFor high
temperature,
the x
substitution in
temperatures(5.16)
x
this is equal to 3RT , i.e. = xis
in
h̄ω/k small
(5.16)
T and
is
small
and x
B it is the = and
smallΘ
D Dulong and
D
the
/T and
this exponential
Petit result. This wf
the fun
exponential
becomes
tegral can
tion in thebe
expected approximated
integral from can thebe resultbyin 1+x.
approximated the byThe1+x.
Einstein integral
The
model: is
integral
the then
is mere
then
Dulong m
Pe
2 k4 T 4 x 3
! " 3 x 3
3V x T x
# #
the resulting
over
⟨E⟩ =
x and
alwaysB theenergy
reached
D
resulting
atis ⟨E⟩
energy=
sufficiently
dx = 9NkB T 3N
is k
⟨E⟩
highB =T .
3Nk ForB one mole
D
T
temperatures. . For of
one mole
dx. (5.16) atom
of a
3 x x
tothis
3RTis2πequal
2 3
, νi.e. isethe
h̄ toit0 3RT , −i.e.
1 Dulong
it is the Petit ΘDresult.
Dulong Petit eThis
0 result. − 1 This
was was of co
of cours
expected from the result in the Einstein model: 62 Dulong the Dulong Petit resu
m the
this result
omalways the heat in the Einstein
capacity of the model:
solid can be the
determined by Petit result
differentiation
reached at sufficiently high temperatures.
ed at sufficiently
th respect high temperatures.
to temperature. Instead of writing down an expression for the
at capacity, we focus on the low and high
72 temperature limits.
62
ωD2
0 2π 2 ν 3 eh̄ω/kB T − 1 2π ν 3 ⇥
0 e h̄ω/kB T − 1
The
and with Debye
# ω D the 2 model
3N
substitution =
ω V E h̄ω=0 n + 3V
3 for
x g(ω)dω
= 1
the
h̄ω/k B heat
T
h̄ and
# ωDx capacity
D =ω Θ2D /T this be
⟨E⟩ = 3 n dω = h̄⌅E dω (5
substitution 0 x =2π h̄ω/k
2 ν 3 eh̄ω/k
B TB Tand
x
− 1 x D 2
= 2π Θ 2 ν 3/T this
D 0 e becomes
h̄ω/k BT − 1
3 xD
hich results in 4 4 3 3
! "
3V kB T x T x
# D
#
⟨E⟩ = ⇤⇥ N dx = 9Nk T
4and4 with 3 2π 2 ν 3 h̄ ωx
# xDthe substitution 3 3
=
= e6π 2
h̄ω/k
x −! 1 νT3
h̄⌅ and
E 3x B
=x 1 Θ /T
Θ this becomes
ex−1
Z = 0 exp[V T (n + D )] x 3
D B D D D
"
kB T x 0
#
is called
3 the Debye
x
dx
frequency= 9N k
and B T
the k B T
corresponding 2 temperature
3 x
dx. Θ (5.16
=
ν h̄ 0 3Ve kB−T 1
D2 3 From this4 the
4 xD capacity
heat n=0
3 of ΘD
the solid can e −1 x
be xD
determined 3 D by di
! "
x T
# #
0
ωD /kB⟨E⟩ is with= respect
called the Debye temperature. dx = 9Nk B T dx. (5
2π
With this (5.14) becomes
2 ν 3 h̄ to temperature.
3
0 e x − 1E n = Instead
nh̄⌅ E
of
Θ D
writing 0
down
e x − an
1 expres
e heat capacity heat capacity, of the we solid
focus can on the benh̄ determined
low and high temperature by differentiatio
limits.
toFrom thisωFor
temperature.
# theωhigh
D heat
2
V Instead h̄ω pn of
capacity
temperatures =the
of Aex solid
writingin
3V(5.16)
E
h̄ can
# /kB T
ωbe
down
D determined
is small and and
anω 2expression by differentia
the for
exponth
⟨E⟩with= respect
3tion in to the temperature.
integral can dω
be =
Instead of3 writingh̄ω/k
approximated by down
1+x. dω
anThe (5.15) fo
expression
integral is
y, we focus 0 on 2π the
2 ν 3 e low
h̄ω/kB T
and− 1 ln Z temperature
high 2π 2 ν 0 h̄⌅E e B
limits.
T − 1
heat capacity,
over x and 2 we focusthe =
⇥E⇤ on the low
resulting = high
and
energy is temperature
⟨E⟩ = 3Nk B T limits.
. For one mo
temperatures
high
nd with For high
temperature,
the x
substitution in
temperatures (5.16)
this is equal to 3RT , i.e. x = x is
h̄ω/kin small
kT
(5.16)
T small
and
B it is the
and
is
x eh̄ and
small
=
E /kB the
D DulongΘ Dand
/T
T
and 1
exponential
Petit result. This wf
this the
becomesexponential fun
tegral
tion in can thebe
expected approximated
integral from can thebe result by
approximated
in 1+x.
the by The1+x.
Einstein integral
h̄⌅ The
model:
E
is
integral
the thenis mere
then
Dulong m
Pe
2 k 4 T 4 U = 3N x 3 < E >= 3N 3 x 3
! "
3V x T x
# #
the
⟨E⟩ = resulting
over x and
alwaysB energy
the D
resulting
reached atis
A ⟨E⟩
energy
sufficiently =
dx = 9NkB T 3N
is ⟨E⟩k
high B=
A T .
3Nk For
D
B Bone
temperatures.
e h̄ E /k T .T For mole
one moleof
1dx. (5.16) atom
of a
3 x x
tothis3RT is2πequal
2 3
, νi.e. isethe
h̄ toit0 3RT , −i.e.
⇥ 1 Dulong
it is the Dulong
Petit ΘDresult.
⇥2 Petit 0
h̄
eThis
result.− 1 This
/kB T
was was
of of
coursco
expected from the result E in the Einstein h̄⌅ model: e the E Dulong Petit resu
m the
this result
omalways the heat
C V
reached
in
= the
capacity Einstein
at sufficiently of the= 3R
for
solid
high
model:
one
can E
mole
be the62
determined
temperatures. C Dulong
V /k T= by 3R Petit
differentiationresult
ed at sufficiently
th respect to temperature. T
high temperatures.
Instead
V k
of writing
B T (e h̄ E B
down an expression for the 1) 2
at capacity, we focus on the low and high temperature limits. h̄⇧max

72
h̄⌅E
62 =
ωD 2 ⇥ ωD Θ 3
ω V h̄ω 3V h̄ Dω h̄ωD /kB
=
! !
hich
3 results The in Debye
# ωD
ω 2
V model
dω
T −=1 3n + 2π h̄ω= 1for the 3V heat
h̄
# ωDcapacity
ω 2 2V
dω
⌅ (4.48)
⟨E⟩2π =2 ν33 eh̄ω/kBE
n 2 N2ν
dωh̄⌅33=
E e h̄ω/kB T =
g(⌅)d⌅ − 1 d⌅
dω (7(5
0
0 2π ωD B=T 6π
2 ν 3 eh̄ω/k −2 1 V
0
ν 2π 2 ν 3
0 e
γ 2⇤
h̄ω/kB T 2
Mω −⇥
2 3
1
Y = =
he substitution
isand
called substitute
the Debye x = ⇥ h̄ω/k
frequency
⇤ T
and and
the x = h̄ω
corresponding /k Ta⌅ this
temperature abecomes
ΘD =
with the substitution x = h̄ω/k B
h̄⌅E B D
T and 1 Dx D = B
Θ D /T Dthis becomes
D /kB is called the Z =
Debye exp[
temperature. (n + #3)]!3N = g(⌅)d⌅ (7
3V kB T4 this
4 (5.14)xD becomes 34 # x k T " 2 xDkNa =
3 2πn3
With 3V kBx T x0 D x
!
4 3 B ! " 3 x
T n=0 x T
D
#
⟨E⟩ 3 = x
dx = 9Nk dxB T= 9Nk B T ⌅ x
dx. (4.49)
dx. (5
2π ν h̄ ω0D2πω2eν2 V3 h̄− 10 h̄ω
2 3 # 3
En = nh̄⌅E3V<h̄ E >= 3 !
e x − 1 Θ D
# ωD Θ 0 D ω e2 0− 1e −
max x g(⌅)h̄⌅
1 (7
⟨E⟩ = 3 dω = 2 3Mα ẍnαih̄ω/k + Φdω
mβj
x (5.15)= 0.
2
2π ν e 3 h̄ω/k B T − 1 nh̄ E2π νT 0 e B
0
T −1 e
mβj nαi h̄ /k
mβjB T 1
theFromheatthiscapacity
0 the heatof nthe= Ae
pcapacity solid of the /k
cansolidbe determined
B can be determined ⌅ D by differentiation
by differentia(7
ct with
low
d with respect
to temperature.
temperature,
the to temperature.
substitution Instead
x = h̄ω/kof Instead
B Twriting
large
and xof
D = ΘD"/T
writing
down down
anthisexpression
3 becomes g(⌅)h̄⌅
an expression 4 for thefo
heat capacity, we focus ln
on Z the low and h̄⌅
< E >= ∞
3 x
E temperaturedxlimits.
high =
π
ity, we focus ⇥E⇤on =the low and = high temperature e xlimits.
0 1 e h̄ /kB T
15 (7
1
For high xD
4 4temperatures 3 kTx in (5.16) e h̄ /k
E is B T
small 3
1
0 xD
and and
− 3 the exponential f
! "
3V k T x T x
# #
h⟨E⟩temperatures
= B x in (4.49)dx = is small
9Nk B T and and the exponential
dx. N (5.16)func-
tion in2πthe integral
3
ecan
x − be 1 approximated Θ by 1+x. The
e x −integral
1 is3then m
⌅D = 6⇤
2 ν 3 h̄ 3 2
integral 2 can be 0
approximated by 1+x. h̄⌅
D
The E integral
0
is then ⇥ merely
over x U and=the 3Nresulting
A < E >= energy3NAis ⟨E⟩ = 3Nk T . For one V mole of(7a
1 mole
h̄ E /kB T B
domthe
this resulting
thisisthe
equalheatto energy
capacity
3RT , i.e. is ⟨E⟩
of the =
it solid 3Nk
is thecan e beTdetermined
B
Dulong . ForPetit one of atoms
by differentiation
result. This was ofthis co
th3RTrespect
, i.e.to it leads ⇥
temperature. to theInstead
Dulong of ⇥
writing
Petit down
result. an expression
This was for
of ⇥
the
course
3
expected fromEthe result in the h̄⌅
2
Einstein model:
e h̄ E /k the
B T12R⇤ Dulong4 Petit
T resu
at
rom V =
capacity,
Cthe
always we focus
result in
reached on
at the =for
the
sufficiently3Ronehigh
low
Einstein and Ehigh temperature limits.
mole
model:
temperatures.the =
CV Dulong Petit result (7is
For high temperatures T V x in (5.16) kBisTsmall(eandEandBthe exponential
h̄ /k T 1)5 2 func-
D
ched at sufficiently high temperatures.
n in the integral can be approximated by 73 1+x. The integral is then merely
h̄⌅ 62
Comparison Debye model - experiment
10 Dulong-Petit Value
Molar Heat Capacity, C (J K-1)
0.1
3
˜T
0.01
0.001
ΘD=1880 K
2 3 4 5 6 2 3 4 5 6 2
10 100 1000
Temperature, T (K)
74
⇥ ⇥2 h̄⌅E h̄ E /kB T
UE= 3NA < E >=h̄⌅
3NE e
Why does the Debye
CV =model work
= 3R betterA h̄
e /kat T low
1 E B
T V⇥ kB T⇥2 (eh̄ E/k TB T

h̄ /k
T than the Einstein
C =
E
model?
= 3R
h̄⌅
V
k h̄⌅
E e E B
E (e 1)2
T T h̄ E /kB T
B
V
E =
Einstein Debye=
k
h̄⌅ B
E
E
E E kB
⌅⌅ = ⇥k
= ⇥k
nn
x⇤x ⇤ny n
⇤ yn⇤ n ⇤
z⇤ z
k =k(k , xk,yk,yk, kzz)) =
=x(k = (( , , ) )
LL L LL L
⌅2 V
g(⌅)d⌅ = ⌅22 V d⌅
g(⌅)d⌅ = 2⇤ ⇥2 3 d⌅
3
⌅ D2⇤ ⇥
hωE 3N =⌅ g(⌅)d⌅
D
0
nx=1 3N
nx=2=
⌅ nmax
x=3 g(⌅)d⌅
g(⌅)h̄⌅
< E >= 3 0 d⌅
⌅ 0 eh̄ /kB T 1
g(⌅)h̄⌅
max
• < E >=
The Debye model gives a better 3 ⌅D3 = 6⇤2 h̄⇥ 3/k
representation
0
N
eV
for
B Tthe d⌅
very
1
low energy vibrations. ⇥3
4 N
3 12R⇤ 2 T 3
• At low temperatures, these vibrations
75
= = 6⇤ most.
CV⌅D
matter
5 VD
⇥
Debye temperatures of the elements
H (in K) He
Li Be B C N O F Ne
344 1440 2230 75
Na Mg Al Si P S Cl Ar
158 400 428 645 92
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
91 230 360 420 380 630 410 470 445 450 343 327 320 374 282 90 72
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
56 147 280 291 275 450 600 480 274 225 209 108 200 211 153 64
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
38 110 142 252 240 400 430 500 420 240 165 72 79 105 119
Fr Ra Ac Rf Db Sg Bh Hs Mt Ds Rg Uub
(taken from Kittel)
1
quency !
γ
ω= .
M
nergy for one harmonic oscillator is 76
Comparison of the Dulong-Petit law to
experiment
-1 -1
77 K (JK ) 273 K (JK )
classical value 24.9 24.9
copper 12.5 24.3
aluminium 9.1 23.8
gold 19.1 25.2
lead 23.6 26.7
iron 8.1 24.8
sodium 20.4 27.6
silicon 5.8 21.8
values of one mole of substance
• At high temperatures the Dulong-Petit law works quite well.

• At low temperatures, it does not. But we already know from
77 basic principles that it wouldn’t.
3
L is the volume of the crystal. From this we can get the density
differentiation
Limits of the Debye model
dN ω2V
dω 2π ν
g(ω) (arb. units)
ch that we recover the correct number of normal modes, so for
the solid we must have
experiment
# ωD
3N = 3 g(ω)dω
0
Debye model
s in
0 1 ω3 2 2 N 33 4 5
D = 6π ν
V 13 s-1)
ω (10
the Debye frequency and the corresponding
78 temperature ΘD =
a power Q̇. Once dynamic equilibrium is established
Thermal conduction
measure the temperature difference ∆T between two
of the rod which are separated by ∆x. The thermal
in equilibrium
A
defined as
1 ∂Q ∆T
heating =κ . heat bath T
A ∂t ∆x 1 ∂Q ∆x
κ=
this expression with the A ∂t usual
∆Tdefinition of the elec-
Here ∆T /∆x plays the1role ∂Q of∆xthe electric field and
κ =
of the current density.A ∂t ∆T
ΘD = h̄ω1D∂Q /kB∆x
r introduction of phonons,
thermal conductivity κ = travelling wave solutions
A ∂t ∆T
dic boundary conditions cannot be used to describe
Θ D
n fact, while the waves = h̄ω
γ areMω
1 @Q D /k 2 T
B
travelling, their average
Y = ΘAD=@t = 
= h̄ωD /kBx (
over the whole crystal sucha thataa temperature gradient
2
or thermal conduction we γ want
γ Mω vibrational
Mω 2 excitations
Y = Y == =
79 (
How to measure the
conductivity / resistivity
U
I
j I d
V = =
E AU
sample 1
dimension
m
Thermal conduction
of the rod which are separated by ∆x. The thermal
in equilibrium
defined as
1 ∂Q ∆T
A ∂t ∆x 1 ∂Q ∆x
κ=
this expression with the
A ∂t usual
∆Tdefinition of the elec-
Here ∆T /∆x plays the1role ∂Q of∆xthe electric field and
κ =
of the current density.A ∂t ∆T
ΘD = h̄ω1D∂Q /kB∆x
r introduction of phonons,
κ = travelling wave1/field
solutions
current density A ∂t ∆T
Θ D
n fact, while the waves= h̄ω
γ areMωD /k 2
B
electrical Y = ΘD== h̄ωD /kB (
over the whole crystal such
a thataa temperature gradient
j I d 2
=
or thermal= conduction weγ want
γ Mω vibrational
Mω 2 excitations
E AU Y = Y == =
81 (
Thermal conduction
ofo the
be rod
much more
which are important
separated bythan the thermal
∆x. The
in equilibrium lattice cont
not always
defined as the case. A classic example is the ins
ne of the highest
1 ∂Q ∆Tthermal conductivity of all ma
. It may A ∂tnot be ∆xclose
1 ∂Q ∆xto daily experience but i
ke κ =
thistea spoonswith
expression outthe of diamond.
usual
A ∂t ∆T definitionWeof will
the discuss
elec-
to the
Here thermal
∆T /∆x conductivity
plays the 1role
∂Q of∆xthe in a metal
electric later. F
field and
of the current κ =
density.
tions just add.ΘThe = A ∂t
total
h̄ω1
∆T
/k
∂Q thermal
∆x conductivity is
D D B
r introduction of phonons, κ = travelling wave1/field solutions
he electronic
current density thermal A conductivity.
∂t ∆T
Θ D
n fact, while the waves= h̄ω
γ areMω
D /k B2
electrical Y = ΘκD== = κ
h̄ω +
D /k B κ . (
over the whole crystal such
a thatapa temperature
e gradient
j I d 2
= = conduction we
or thermal γ want
γ Mω vibrational
Mω 2 excitations
have Eto state Y = Y precisely
A U more == = 81 what we mean by t(
he electronic thermal conductivity.
Thermal conduction (at room temperature)
κ = κp + κe .
• Nottoonlystate
ave more are
the electrons precisely what weκ mean -1
(Wm K )
by
-1 th
important.
ose you have a rod with a diamond cross sectional area A
2000
•d be
A few insulators are also
in a heat bath with temperature
copper T while
400 th
found to be good thermal
ated gold
with a power Q̇. Once dynamic equilibrium 310
conductors.
aluminium 230
• Thewe
me, rangemeasure the temperature
of electrical
silicon
difference
160
∆T
conductivity of materials is
middle of the rod
much bigger than the range
which are
sodium separated by
140 ∆x.
κ is then defined
of thermal as(24
conductivity glass 1.0
vs 5 orders of magnitude) polystyrene 0.02
1 ∂Q ∆T
=κ .
A ∂t ∆x
82
Thermal conduction by phonons
propagations of phonons
heat source
generation of lattice vibrations
generation of phonons
83
l with a certain group velocity. These wave p
propagations of phonons
ot side of the solid and then propagate to the c

hononsheatcan
source be viewed as a type of particles wh
generation of lattice vibrations
Indeed, it turns out that the thermal conductiv
generation of phonons
bed using kinetic gas theory, as if the phonons we

through theheat solid.
capacity For describing
mean freethe paththermal co
of phonons
honons, we take a result from kinetic gas theor
1
κp = cv λp vp ,
3 phonon (i.e. sound)
speed
t capacity of the solid per 83
unit volume, λp is the
gas of phonons, we take a result from kinetic
Thermal conduction by phonons gas theo
1
κp = cv λp vp ,
3
the heatat capacity of the solid per unit volume, λ
any temperature the mean free path is limited byp is the
phonons and vp is the phonon velocity. In order to e
• scattering
vp to be the velocity of sound
from defects ν and for the heat capac
or impurities
ults of•the previous
scattering fromsection. The
the sample only quantity which i
boundaries
free path of the phonons. We discuss it in the followi
honons propagate through the crystal, they can be sc
ns in the lattice, such as point defects, dislocations and
o grow crystals of such high perfection that these scatte
mportant. Then the scattering of the phonons at t
can be seen experimentally. Scattering from crystal im
84
1
κp = cv λp vp ,
3
• scattering
or impurities
boundaries
84
1
κp = cv λp vp ,
3
• scattering
or impurities
boundaries
84
1
κp = cv λp vp ,
3
• scattering
or impurities
boundaries
84
1
κp = cv λp vp ,
3
• scattering
or impurities
boundaries
84
1
κp = cv λp vp ,
3
the heat capacity of the solid per unit volume,
at high temperature the mean free path is limited byλ p is the
vp to be the velocity of sound ν and for the heat capac
ults of the previous section. The only quantity which i
honons• scattering
propagatefromthrough the crystal,
other phonons they
(but this is can be sc
strictly
spoken an anharmonic effect).
85
we take a result from kinetic gas theory:
1
κp = cv λp vp , 10000 (5.17)
Si
3
thermal conductivity κ (W m-1K-1)

ty of the solid per unit volume, λ is the mean free
• the mean free p
p ispath
thedecreases
phonon velocity. 1000 In order to evaluate κp ,
at high
locity of Tsound ν and for the heat capacity we can
• thesection.
ious heat capacity
The only quantity which is unknown
decreases at low
e phonons.
T
We discuss it
100 in the following. melting
point
ate through the crystal, they can be scattered by
• there is a
, such as point defects, dislocations and the like. It
maximum in the
of such high perfection
conductivity at that
10 these scattering effects
about 10%
en the scattering of θ 10 100 1000
D of the phonons at the
temperature, T (K)sample
erimentally. Scattering from crystal imperfections
86
Thermal expansion
87
1/2
2kB T
x1max2= Thermal expansion (7.10
⇥x
volume = k ⇥
BT
expansion (7.9)
2 max
⇥
1 ⇧V ⇥1/2 -5 -1
V = 2kB T α (10 K (7.11
)
xmax = V ⇧T P Lead 2.9 (7.10)
⇥
linear expansion
1 dL ⇥ Aluminium 2.4
1=1 L⇧V
⇥LdT Brass 1.9 (7.12
V = = 0 Copper 1.7 (7.11)
VL0 ⇥T
⇧T P
1 Steel 1.1
=1 3dL V Glass 0.9
(7.13
L = L=0 (1 + T) (7.12)
L0 dT diamond 0.12
⇧U
C 1= Invar 0.09 (7.14
= ⇧T (at room temperature)(7.13)
V
1 3 2 1
• Eα = mvx + ⇥x2
is temperature-dependent
2 ⇧U 2 88
and vanishes at T=0. (7.15
Thermal expansion: examples
not so smart design better design
89
Invar
• Invar is a 63% NiFe alloy with the

lowest coefficient of thermal expansion
of all metallic compounds.
• Commonly used in thermostats,
thermal switches, precision
instruments...
• Nobel prize to Charles-Edouard
Guillaume, 1920.
90
interatomic potential (arb. units)
Lattice vibratons
interatomic force (arb. units)

repulsive
0
attractive
0
repulsive part
resulting
potential
grad (r)
attractive part
0 1 4 6 0 1 4 6
interatomic distance / equilibrium distance a interatomic distance r (equilibrium distance)
d (a) 1 d2 (a) 1 d 3
(a)
(x) = (a) + (x a) + (x a) +
2
(x a)3 + . . .
dx 2 dx 2 6 dx 3
first
energy harmonic potential anharmonic
=0
offset linear force. term 91
Thermal expansion on the atomic scale
(classical)
interatomic potential φ (arb. units)
1
interatomic distance r (equilibrium distance)
92
(classical)
r1 r2
kBT1
0
1
92
(classical)
1
93
(classical)
r1 r2
kBT1
0
1
93
(classical)
r3 r1 r2 r4
kBT2
kBT1
0
1
93
(classical)
r3 r1 r2 r4
kBT2
kBT1
0
1
93
Thermal expansion on an atomic scale
• The bottom line: for the harmonic solid, there is no thermal

expansion. Thermal expansion is caused by anharmonicity.
• For a quantum treatment, we get the same qualitative
result.
• The shallower the interatomic potential, the more
anharmonic effects are expected.
94
R = ma + na (2.1)
Rmn von=der 1 + na (2.1)
mn
modynamischen 1 2
Potentialen gelangen. Die Grundgl
man maGrundgleichung 2 dE = T dS P dV + µd
Allotropic phase transitions
dass
mno dE = ma
modynamischen + nafestes
= 01 fuer 2 + oa(S,
Potentialen3 V, N ). Tatsaechlich
gelangen. Die (2.2) d
muss
Grundgl
mno = ma 1 + na 2 + oa 3 (2.2)
ass• dE = 0 fuer
chgewicht
phase ein Minimum
transition from one annehmen.
solid phase to Genauso
U Vfestes (S, V, N ). Tatsaechlich muss d
another solid gilt vo
+ (P/T sin
)dV
phase.
chgewicht =
ein U V
(µ/T
Minimum )dN , dass im
annehmen. Gleichgewicht
Genauso (2.3)
dS
gilt =
vo
sin = d (2.3)
, N • 2U
+ (P/T). Die Entropie
transitions
)dVV = (µ/Td
involving
n⇥
muss
)dNlatent
, ein
heat
dass Maximum
(first
im order annehmen.
phase
Gleichgewicht dS
(2.4)
Ma
=
dre Transformationen
transitions)
N ). Die2U VEntropie
with a wie
singularity F
in =
the
= n⇥ muss ein Maximum annehmen. E
heat T S noch
capacity. viele
(2.4)
Maa
•
odynamische
n⇥ = 2dsin
Potentiale finden.
dre Transformationen wie F = E T S(fornoch
ultimate reason: minimum in the Zum
free energyBeispiel (2.5)
constant
viele a
n⇥ =and
T,V,N) 2dsin
more practically the Gibbs free energy (for (2.5)
odynamische Potentiale
h finden. Zum Beispiel
me constant
⇥ = ⇥h
T,P,N). Di⇥erential (2.6)
rgie ⇥ = ⇥ 2mE dE = T dS P dV + µdN(2.6)
e 2mE Di⇥erential
Energie
gie F = U T S dF = SdT
dE = T dS P dV + µdN P dV + (2.7)
µdN
F = U TS
Enthalpie dG = SdT + V dP + (2.7) µdN
Energie dF = SdT P dV + µdN
skanonisches Potential dJ = SdT pdV N dµ
G = U + PV TS
Enthalpie dG = SdT + V dP + (2.8) µdN
skanonisches Potential dJ95 = SdT pdV N dµ
Gibbs free energy G (arb. units)
0
0 TC
temperature T
96
Allotropic phase transitions: example I
BCC Fe FCC Fe
(ferrite) (austenite)
912 oC
• In the fcc phase, a small amount of carbon atoms can be

put into the interstitial positions (C radius is 61% of Fe
radius)
• in the bcc phase this is not possible. The re-distribution of C

while cooling affects the strength of the steel (influenced by
cooling rate)
97
Allotropic phase transitions: example II
grey Sn white Sn
(brittle) (silvery, ductile)
13.2 oC
T
• The transition temperature can be lower for none-pure
samples.
• The transition is auto-catalytic.
• Problems for organ pipes in nordic countries (tin pest).
• ...and polar expeditions.
• The problem can be avoided by alloying Sb, Bi or Pb into Sn.

98
n⇥ = 2dsin
h
⇥= ⇥
2mE
F =U TS
G = U + PV TS
0
0 TC
temperature T
99
n⇥ = 2dsin
h
⇥= ⇥
2mE
F =U TS
G = U + PV TS
meta-stable phases are,

B
of course, not visible here!
0
0 TC
temperature T
99
h
⇥= ⇥
Melting
2mE
F =U TS

G = U + PV TS
solid
liquid
0
0 TC
temperature T
100
Melting temperature and cohesive energy
W
Ta
3 Mo
B
Cr TiVPt
2 Fe
Be Si
Ca Ag AuLa
melting temperature (K)
1000 Mg Sb Al
7 Zn Te
6 Cd Pb
5 Po Se Sn
Li In S
4 Na
3 RbK Ga
Br
Hg
2
Xe Cl
Kr
100
Ar
7
6
5
4
3
Ne
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8
0.1 1
cohesive energy per atom (eV)
101
stress ⇥x a0 ⇥
=
YLindemann = 2 criterion
= (F. Lindemann, (7.7)1911)
strain a0 x a0
The crystal melts when the vibrational amplitude
⇥⇥Y
reaches (7.8)
a0 fraction of the interatomic spacing
some
1 5%
(e.g. or so).
⇥xmax = kB T
2
(7.9)
2
⇥1/2
2kB T x2max
xmax = Tm = (7.10)
⇥ 2kB
L 1
= (7.11)
L0 T
⇧U
C= (7.12)
⇧T
1 1
E = mvx + ⇥x2
2
(7.13)
2 2
C = 3R (7.14)
⇥
1 102
uch W or Mo stressfor which
⇥x acovalent
0 ⇥ bonding is important, as well as covalent
=such as Si.
YLindemann
aterials = Simple =metals like
criterion (F. Lindemann,
the Alkali metals (7.7)
are1911)
found in the
strain 2
a0 x a0
iddle of the range.
The crystal
For the prediction melts
of the when
melting the vibrational
temperature of amplitude
a solid, the relation to
reaches⇥⇥Y a0 fraction of the interatomic spacing
some (7.8)
he cohesive energy is not very useful because the cohesive energy has to be
nown in the 1 first5%
(e.g. or so).
2 place. An alternative idea was developed by F. Lindemann
⇥xmax = kB T (7.9)
1910. He 2 suggested that melting would occur when the amplitude of
he interatomic vibration ⇥ becomes
1/2 too large, i.e. when it reaches is certain
2kB T 2
xmax
action of =
the interatomic
xmax spacing. Using (5.2), = try this
Tm we (7.10)
idea, stating that
he solid melts once x⇥ 2k
max reaches 5% of the interatomicBdistance a, i.e.
L 1 with the frequency1
= 2
(0.05a) γ (0.05a) ω M (7.11)
2 2 !
γ
L0 Tm T= = . with ω =
M
.
2kB 2kB
⇧U The total energy for one harmonic oscillator
= Debye angular frequency ωD and for a(7.12)
or ω we can useCthe the atomic1 radius.
⇧T 2
E = Mvx + γ
1
we now write the above expression in terms of the Debye temperature 2 we2
1 1
et E = mvx + ⇥x22 Assuming classical(7.13)
motion, we can estimate
2 2 (0.05a)2 ωD 2
kbyB using
M the equipartition theorem of statis
Tm = . every generalised momentum
2 states that (5.18) or
C = 3R 2h̄ pears squared in the (7.14)
Hamilton function (or i
he result of this is shown⇥ in Fig. 5.12(b). The shouldvery simplistic
contribute model
with a mean clearly
energy kB T /
1 102
two squared coordinates and therefore the th
ddle of the range.
uch W or Mo stressfor which⇥x acovalent
0 ⇥ bonding is important, as well as covalent
For Y =
Lindemann
the prediction
aterials such = of the
as Si. aSimple =
criterion
melting
metals (F. Lindemann,
temperature of a (7.7)
solid,
like the Alkali metals are found in the 1911)
the relation to
cohesive strain
energy is
2 x
not a
0 very useful 0 because the cohesive energy has to be
iddle of the range.
own in The
the crystal
first
For the prediction place. melts
ofAn
the when
alternative
melting the vibrational
idea
temperature was amplitude
developed
of a solid, bythe
F. relation
Lindemann to
reaches⇥⇥Y a0that
some fraction of the interatomic (7.8)
spacing
he cohesive energy is not very useful because the cohesive the
1910. He suggested melting would occur when energyamplitude
has to beo
interatomic
nown in the1 first
(e.g. 5% or so).
vibration
2 place. becomes
An too idea
alternative large,was i.e.developed
when it by reaches
F. is certain
Lindemann
⇥x = k B T (7.9)
ction
1910.of the
He2 interatomic
max
suggested that spacing.
melting Usingwould(5.2), we try
occur when thistheidea, stating that
amplitude of
hesolid melts once
interatomic vibration ⇥
xmax reaches
becomes
1/2 5% toooflarge,
the interatomic
i.e. when itdistance
reaches a,
is i.e.
certain
2kB T 2
xmax
action of xmax =
the interatomic spacing. 2Using (5.2), T m =
we
2 2
try this (7.10)
idea, stating that
he solid melts once xTmax ⇥
=
(0.05a)
reaches 5%γ of
=
(0.05a)
the ω
interatomic M2k . Bdistance a, i.e.
m
L 1 2kB 2kB frequency1
with the
= (0.05a) γ 2
(0.05a) ω M (7.11)
2 2 !
γ
r ω we can use theLDebye 0 Tm T=angular frequency= ω and .
for a the with ω =radius
atomic .
2kB 2kDB M
we now write the above ⇧U expression in terms of the
The total Debye
energy for onetemperature
harmonic oscillator we
or ω we can useCthe = Debye angular frequency ωD and for a(7.12) the atomic1 radius.
⇧T E = Mvx + γ 2 1
we now write the above expression (0.05a) in2terms
Θ2D kB M of the Debye temperature we
1 1Tm = . with = 2
k B(5.18)
2
D
et E = mvx + ⇥x2 2
2h̄ 2
Assuming classical(7.13)motion, we can estimate
2 2 (0.05a)2 ωD 2
kbyB using
M the equipartition theorem of statis
e result of this is shown T inmFig.
= 5.12(b).2 The states very
. every
that simplistic model
generalised momentumclearly
(5.18) or
roduces the trend C = correctly
3R even though 2h̄ pearsit neglects
squared in (7.14)
entropy
the effects
Hamilton as (or
function weli
he result of this is shown ⇥ in Fig. 5.12(b). The shouldvery simplistic
contribute model
with a mean clearly
energy kB T /
the influence of the detailed 1 atomic102structure. two squared coordinates and therefore the th
or ω we can use the Debye angular frequency ωD and for a the atomic radiu
Lindemann
we now write the abovecriterion
expression (F. Lindemann,
in terms 1911) w
of the Debye temperature
et
(0.05a)2 Θ2D kB M
Tm = 2 . (5.18
2h̄
he result of this is shown in Fig. 5.12(b). The very simplistic model W
clear
Ta Re
produces the trend correctly even though it neglectsVNbentropy
3 Mo effects as we
Hf
CrRh
Ru
2 Ti
the influence of the detailed atomic structure.Si
Cu
Ag
Fe
Mn Be
La Ca As Ge
melting temperature (K)
1000 Mg Sb
7 Te Zn
6 Tl Pb Bi Cd
5 Se Sn
Li In
4
Cs KNa Ga
3
Hg
2
Xe
Kr
100
Ar
7
6
5
4
3
Ne
10 100 1000
estimated melting temperature (K)
103

Lecture4 Gen

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Lecture4 Gen

Transféré par

Droits d'auteur :

Formats disponibles

• lectures accompanying the book: Solid State Physics: An

Introduction, by Philip Hofmann (2nd edition 2015,

• What is the frequency a

interatomic force (arb. units)

Classical treatment using the equipartition theorem.

• An infinite chain with one atom per unit cell

what are the allowed vibrational frequencies !?

s has a solution ifthis

s has a solution ifthis

general 1D wave A(x, t) = A0 e

or here even A(x, t) = A0 cos(kx !(k)t)

in general we have that ω depends on k.

general 1D wave A(x, t) = A0 e

or here even A(x, t) = A0 cos(kx !(k)t)

in general we have that ω depends on k.

general 1D wave A(x, t) = A0 e

or here even A(x, t) = A0 cos(kx !(k)t)

in general we have that ω depends on k.

general 1D wave A(x, t) = A0 e

or here even A(x, t) = A0 cos(kx !(k)t)

in general we have that ω depends on k.

a quantum mechanical particle a a

in vacuum the dispersion relation of light is linear.

in vacuum the dispersion relation of light is linear.

sum of partial waves (package) 7 partial waves

• phase and group sum of partial waves (package)

• the wave packet

sum of partial waves (package) 7 partial waves

sum of partial waves (package) 7 partial waves

sum of partial waves (package) 7 partial waves

sum of partial waves (package) 7 partial waves

• The dispersion relation is periodic and the periodicity is one

• Changing k by one reciprocal lattice vector gives exactly the

• Changing k by one reciprocal lattice vector gives exactly the

The first Brillouin zone is the region of reciprocal space

We also brieflyb/2+u n -vthe

We also brieflyb/2+u n -vthe

We also brieflyb/2+u n -vthe

The coeﬃcient (system

This has a solution if the determinant of the coeﬃcients vanishes. This

coeﬃcient matrix is called the dynamical matrix of the system. A

exists if the determinant of dynamical matrix vanishes, i.e.

typical vibrational frequency ~ 1012 Hz

typical vibrational frequency ~ 1012 Hz

typical vibrational frequency ~ 1012 Hz

typical vibrational frequency ~ 1012 Hz

typical vibrational frequency ~ 1012 Hz

This gives only standing waves.

chain with N atoms:

chain with N atoms:

A finite chain with no end!

• We want to get rid of the

want a solid which is finite in

• If we move by one crystal L

size L, we have to get the

u(x, y, z) = u(x + L, y, z) = u(x, y + L, z) = u(x, y, z + L)

• We want to get rid of the

want a solid which is finite in

• If we move by one crystal L

size L, we have to get the wrap

u(x, y, z) = u(x + L, y, z) = u(x, y + L, z) = u(x, y, z + L)

• We want to get rid of the

want a solid which is finite in