PII - Quantum Condensed Matter Physics - Ritchie (2019) 399pg PDF

Quantum Condensed Matter Physics
Lecture 1
David Ritchie
QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 1.1

Quantum Condensed Matter Physics: synopsis (1)
1. Classical and Semi-classical models for electrons in solids (3L)
Lorentz dipole oscillator, optical properties of insulators. Drude model and optical
properties of metals, plasma oscillations. Semi-classical approach to electron
transport in electric and magnetic fields, the Hall effect. Sommerfeld model, density
of states, specific heat of; electrons in metals, liquid 3He/4He mixtures. Screening
and the Thomas-Fermi approximation.
2. Electrons and phonons in periodic solids (6L)
Types of bonding; Van der Waals, ionic, covalent. Crystal structures. Reciprocal
space, x-ray diffraction and Brillouin zones. Lattice dynamics and phonons; 1D
monoatomic and diatomic chains, 3D crystals. Heat capacity due to lattice
vibrations; Einstein and Debye models. Thermal conductivity of insulators. Electrons
in a periodic potential; Bloch’s theorem. Nearly free electron approximation; plane
waves and bandgaps. Tight binding approximation; linear combination of atomic
orbitals, linear chain and three dimensions, two bands. Pseudopotentials. Band
structure of real materials; properties of metals (aluminium and copper) and
semiconductors.
Semi-classical model of electron dynamics in bands; Bloch oscillations, effective
mass, density of states, electrons and holes in semiconductors
QCMP Lent/Easter 2019 1.2

3. Experimental probes of band structure (4L)

Photon absorption; transition rates, experimental arrangement for absorption
spectroscopy, direct and indirect semiconductors, excitons. Quantum oscillations;
de Haas-Van Alphen effect in copper and strontium ruthenate. Photoemission; angle
resolved photoemission spectroscopy (ARPES) in GaAs and strontium ruthenate.
Tunnelling; scanning tunnelling microscopy. Cyclotron resonance.
Scattering in metals; Wiedemann-Franz law, theory of electrical and thermal
transport, Matthiessen’s rule, emission and absorption of phonons. Experiments
demonstrating electron-phonon and electron–electron scattering at low
temperatures.

4. Semiconductors and semiconductor devices (5L)

Intrinsic semiconductors, law of mass action, doping in semiconductors, impurity
ionisation, variation of carrier concentration and mobility with temperature - impurity
and phonon scattering, Hall effect with two carrier types.
Metal to semiconductor contact. P-n junction; charge redistribution, band bending
and equilibrium, balance of currents, voltage bias. Light emitting diodes; GaN,
organic.
Photovoltaic solar cell; Shockley-Queisser limit, efficiencies, commercialisation. Field
effect transistor; JFET, MOSFET. Microelectronics and the integrated circuit.
Band structure engineering; electron beam lithography, molecular beam epitaxy.
Two-dimensional electron gas, Shubnikov-de Haas oscillations, quantum Hall effect,
conductance quantisation in 1D. Single electron pumping and current quantisation,
single and entangled-photon emission, quantum cascade laser.

5. Electronic instabilities (2L)
The Peierls transition, charge density waves, magnetism, local magnetic moments,
Curie Law. Types of magnetic interactions; direct exchange, Heisenberg
hamiltonian, superexchange and insulating ferromagnets, band magnetism in
metals, local moment magnetism in metals, indirect exchange, magnetic order and
the Weiss exchange field.
6. Fermi Liquids (2L)
Fermi liquid theory; the problem with the Fermi gas. Liquid Helium; specific heat
and viscosity. Collective excitations, adiabatic continuity, total energy expansion for
Landau Fermi liquid, energy dependence of quasiparticle scattering rate.
Quasiparticles and holes near the Fermi surface, quasiparticle spectral function,
tuning of the quasiparticle interaction, heavy fermions, renormalised band picture
for heavy fermions, quasiparticles detected by dHvA, tuning the quasiparticle
interaction. CePd2Si2 ; heavy-fermion magnet to unconventional superconductor
phase transitions.
•Course material will be useful for several part III courses.

•Printed overheads & problem sheets provided.
•All available in pdf on web: http://www.sp.phy.cam.ac.uk/drp2/home
Books
1. Band Theory and Electronic Properties of Solids,
Singleton J (OUP 2008)
2. Optical properties of Solids, Fox M

(2nd edn OUP 2010)
3. The Oxford Solid State Basics, Simon S H

(OUP 2013)
4. Introduction to Solid State Physics, Kittel C

(8th edn Wiley 1996)
5. Solid State Physics, Ashcroft N W and Mermin N D,

(Holt, Rinehart and Winston 1976)


Lorentz dipole oscillator, optical properties of insulators. Drude model
and optical properties of metals, plasma oscillations. Semi-classical
approach to electron transport in electric and magnetic fields, the Hall
effect. Sommerfeld model, density of states, specific heat of; electrons
in metals, liquid 3He/4He mixtures. Screening and the Thomas-Fermi
approximation.

Optical properties of insulators
• Response to high frequency electric field in electromagnetic waves.
• Wavelength long compared to interatomic spacing.
• Classical picture – Lorentz dipole oscillator model
No EM wave EM wave – electron cloud
oscillates about nucleus
electron cloud displacement u.
+u
nucleus -u
1
t =0 t=
2f
• Model atoms as nucleus + electron cloud.

• Applied electric field causes displacement of electron cloud,
• Assume restoring force is proportional to displacement.

• Electron cloud behaves as damped harmonic oscillator
mu + mγ u + mωT2u =
qE
• ωT natural frequency, determined by force constant and mass; γ damping
rate (no model for this yet…)
− iωt
• Consider oscillating electric field E (t ) = Eω e , which induces oscillating
− iωt
displacement u (t ) = uω e
• Resulting dipole moment per atom at angular frequency ωT : pω = quω
• Polarisation = dipole moment per unit volume
=Pω p=
ωN /V npω
• From Pω = 0 χω Eω and the equation of motion we obtain for the
polarisability:
N q2 q2
χω = n
V m0 (ωT − ω − iωγ )
2 2
m0 (ωT2 − ω 2 − iωγ )
• Frequency dependence of 𝜒𝜒𝜔𝜔 typical of harmonic oscillator
• Permittivity from model: q2
1 χω =+
ω =+ 1 n
• Figure shows typical m0 (ωT2 − ω 2 − iωγ )
frequency dependence
• Analogy with damped SHO
tells us that power absorbed
by electron cloud is
determined by Im(𝜖𝜖𝜔𝜔 ):
P= 1
2 ω 0 | Eω |2 Im(ω )
• Simple way to think of
absorption lines in optical
spectra
• At low frequencies:
 (ω → 0 ) =1 + n q 2 m0ωT2
• Explains why different materials have very different static permittivities
− 
• Reflectivity between media of different permittivities: r = 1 2,
1 + 2
Power reflection coefficient given by: R = r
2
•
• In a solid polarization fields of other atoms can alter resonant frequency
Example – atomic absorption line
• Absorption coefficient α and

refractive index n of sodium
gas
−3
• Atom density is 1× 10 m
7
• n0 is the off resonance

refractive index
• Absorption due strongest
hyperfine component of the
D2 line at λ = 589nm
so =ν 5.1×1014 Hz
• Linewidth ∆ν ≈ 100MHz
hence ∆ν / ν = 2 ×10−7
• Very narrow…..
Taken from Optical properties of solids by M Fox

• Lorentz oscillator model, superposition of spectra
• May have a number of allowed transitions at energies  ωT1 ......ωT2
• Usually high frequency – terahertz to ultraviolet
• Resulting frequency-dependent permittivity from adding responses
associated with each transition:  (ω ) = 1 + ∑
χ (ω )i
Permittivity

Optical properties of insulators – comparison with
experiments
• Fused silica glass SiO2
• Expected general
characteristics observed
• Transparent in visible light
• Strong absorption peaks in
infrared and ultraviolet
• IR peaks due to vibrations of
SiO2 molecules
• UV absorption background
across 10eV bandgap
• UV absorption peaks caused
by inner core electron
transitions in Si and O Visible
light
For weakly n ≈ Re (  ) Infra-red Ultra-
absorbing violet
Im (  )
X-rays
medium with
nκ κ≈
2n
QCMP Lent/Easter 2019 Optical properties of Solids, M Fox 1.13
Dispersion of light
• SiO2 refractive index in more detail
• n increases with frequency – ‘normal
dispersion’ caused by tails of
absorption peaks in IR and UV
• Can be used to separate colours of
light using a prism
• Short light pulses of duration t p have
spread of frequencies δ f ≈ 1/ t p
• Different velocities of frequency
components causes problems in high
speed optical fibre communications
• Temporal broadening of pulse Optical properties of Solids, M Fox
∆τ ∝ d 2ω / dk 2 ∝ d 2 n / d ω 2 ∝ d 2 n / d λ 2
• Lorentz model gives d n / dλ < 0 above one
2 2
absorption line and d n / dλ > 0 below next

2 2
absorption line
• Choosing d n / dλ = 0 we minimise dispersion
2 2
• For SiO2 this is at λ  1.3µ m – a preferred optical fibre wavelength

Lorentz oscillator model: connection to quantum mechanics
• This classical oscillator model cannot be the whole story – works as a
phenomenological description of optical response function
• For two sharp levels, from time-

dependent perturbation theory
χω ∝ ( Eb − Ea − ω ) −1 + iδ ( Eb − Ea − ω )
• Where the imaginary part is the
transition rate
• This can also be written
χω ∝ 1
Eb − Ea − ω −iγ / 2
where γ is very small
• As energy levels broaden into bands and γ increases, this expression

becomes similar to Lorentz model close to resonance where ω= T Eb − Ea
• We multiply by ωT + ω on top and bottom and approximate: ωT + ω ≈ 2ωT
hence:
ωT + ω 2ωT
χω ∝ 2  2
ωT − ω − i (ωT + ω )γ / 2 ωT − ω 2 − iωγ
2

Drude Model
• Assume we have a gas of
electrons free to move between
positive ion cores
• These electrons are only scattered
by the ion cores
• Between collisions with the ion
cores the electrons do not interact
with each other.
• Collisions are instantaneous
resulting in a change of electron
velocity.
• The probability an electron has a
collision in unit time is τ , the
−1
scattering rate. taken from wikipedia

• Electrons achieve thermal
equilibrium with their surroundings
only through collisions.

Drude Model - relaxation time approximation
• The current density J due to electrons of number density n, mass m, of

average velocity v and momentum p is given by:
ne
J= − nev = − p
m
• Consider the evolution of p in time δ t under the action of an external
force f ( t )
• Probability of a collision during δ t is δ t / τ where τ is the average time
between collisions.
• Probability of no collision during δ t is 1 − δ t / τ
• For electrons that have not collided momentum increases:
δ p f (t )δ t + Ο(δ t ) 2
=
• So the contribution to the average momentum during δ t for electrons that
have not collided is:
p(t + δ t ) =(1 − δ t / τ )(p(t ) + f (t )δ t + O(δ t ) 2 )
Drude Model - relaxation time approximation
• Electrons which have collided are a fraction δ t / τ of the total.
• The momentum they will have acquired since colliding (where their
momentum was randomised) is  f (t )δ t
• So contribution to average momentum for electrons which have collided is
of order (δ t ) - small
2
• For momentum as before:

p(t + δ t ) =(1 − δ t / τ )(p(t ) + f (t )δ t + O(δ t ) 2 )
• If δt → 0 we can rearrange this to give
dp(t ) p(t )
=
− + f (t )
dt τ
• Hence the collisions produce frictional damping.
• Apply to electrical conductivity σ defined by J = σ E , assume steady
state so: dpd(tt ) = 0 and the force on an electron is f (t ) = −eE
• From above J = − ne m p
• Hence p(t ) m ne 2τ
f (t ) = =− J=−eE ⇒ σ =
τ neτ m
Summary of Lecture 1
• Introduction to course and recommended text books

• Lorentz oscillator model for optical absorption in solids
• Comparison with atomic absorption
• Comparison of Lorentz model with experimental results
• The Drude model of electron motion in solids
• The relaxation time approximation, scattering and electrical
conductivity

Lecture 1
The End

Lecture 2
David Ritchie


approximation.

Drude model
dp(t ) p(t )
• From above: =
− + f (t )
dt τ
• In magnetic and electric fields: f= q ( E + v × B ) and j = nqv
• Hence the same equation in different forms:
dp(t ) p(t )
=− + q (E + v × B)
dt τ
dv (t ) v (t ) q
=− + (E + v × B)
dt τ m
dj(t ) j(t ) nq 2
=− + (E + v × B)
dt τ m
dj(t ) j(t ) nq 2 E q
=− + + ( j× B)
dt τ m m

Drude model: Frequency dependant conductivity
dj(t ) j(t ) nq 2 E q
• From above =
− + + ( j× B)
dt 2
τ m m
dj(t ) j(t ) nq
• If B=0: + =E and if j = jω exp ( −iωt ) , E = Eω exp ( −iωt )
dt τ m 2 2
τ Eω
(
τ − iω j=
−1
)m Eω ⇒ j=
nq E
=
nq
2
nq ω
( )
• Then
m (1 − iωτ )
ω ω
m τ − iω
−1
nq 2τ ε 0ω pτ2
• And given j = σ E we
= have σ ω =
m (1 − iωτ ) (1 − iωτ )
2
nq
• Where ω p =
2
defines the Plasma frequency.
m0
nq 2τ qτ
• At low frequencies ωτ  1 we get σ 0 = ,µ = ⇒ σ 0 = nq µ
m m
• Where µ is defined as the carrier mobility

Optical properties of metals: connection to ac
conductivity
• Current density j = u  nq (velocity x density x charge)
• Polarisation P = nuq Hence, P  = j for the conduction electrons
c
• Adding in the polarisation of the core electrons as χ ∞ - the background
polarisability we get: P = j + 0 χ ∞ E

• At angular frequency ω, substituting:
E = Eω exp(−iωt ), j = jω exp(−iωt ), P = Pω exp(−iωt ), Pω = 0 χω Eω
• We get jω =−iω (0 χω Eω − 0 χ ∞ Eω ) = −iω0 Eω ( χω − χ ∞ ) =
σ ω Eω
• Hence
iσ ω
σω =
−iω0 ( χω − χ ∞ ), σ ω =
−iω0 (ω − ∞ ) ⇒ ω = + ∞
0ω
• This relates the imaginary part of the permittivity to the real part of the
frequency dependant conductivity.
σ ω ε 0ω p2τ (1 − iωτ ) from above we
−1
=
• Combining this with our expression
obtain: iω 2τ ω2
ω =
∞ + =
∞ − 2
p p
QCMP Lent/Easter 2019

ω (1 − iωτ ) ω + iω / τ 2.5
Optical properties of metals
• Use Lorentz oscillator model for bound electrons giving background ∞
• Use Drude model for conduction electrons, zero restoring force ⇒ ωT = 0
nq 2
χω = , ωT 0,
m0 (ωT − ω − iωγ )
2 2
Drude Peak
Resonance due
to core electrons
nq 2
⇒ ω = ∞ −
m0 (ω 2 + iωγ )
ω p2 nq 2
=
∞ − 2 , ω p2 =
(ω + iωγ ) m0
• |  (ω ) | diverges as ω →0 so metals are highly reflecting at low frequency

• Peak in Im(ω ) at low ω, due to enhanced absorption, ‘Drude peak’:
•  (ω ) crosses zero at ω P and approaches 1 at a high frequency so metals
∗
become transparent in the ultraviolet.

Optical properties of metals
• Reflectivity at the interface between two media using the Drude model
• If permeability is unchanged at interface: 1 − 2
• Power reflection coefficient R = r
2
r=
1 + 2
Power reflection coefficient R
∞ = 1
∞ = 5
∝ (ω p / ω ) 4
• Plateau in R at low frequency, related to conductivity of material

• R ∝ω
−4
at high frequency – blue line on figure
• If background permittivity ∞ - due to polarisability of core electrons is
significant, R can go to zero at finite frequency – red line
Experimental reflectivity of Aluminium
• Experimental Reflectivity of
Aluminium as a function of photon Reflectivity of Aluminium
energy - green curve, (solid) Theory
1.0
• Reflectivity above 80% for visible
0.8 ω p = 15.8eV
Reflectivitiy
region of spectrum - aluminium
0.6 Experimental data
coating is used for commercial
0.4
mirrors
0.2
• Plasma frequency in ultra-violet
0.0
• Both theory curves assume we 0 5 10 15 20
have ω p = 15.8eV Energy (eV)
• Red (dashed) no damping Data from Ehrenreich et al (1963)
• Blue (Dotted) with=τ 8.0 ×10 s value deduced from DC conductivity

−15
values - slightly better fit

• Two unexplained features in experimental results: (1) reflectivity is smaller
than predicted (2) small dip in reflectivity around 1.5eV
• Both explained by considering interband absorption rates
QCMP Lent/Easter 2019 Optical properties of Solids, M Fox 2.8
Optical properties of metals: plasma oscillations
• Part IB electromagnetism – “Plasma Oscillations”
• electrons moving in a positively charged environment - model for metal.
• Consider probing a slab of material by applying an oscillating field D:
D = 0ω E = 0 E + P ⇒ P = D (1 − ω−1 )
ω p2
ω= ∞ −
(ω 2 + iωγ )
−1 ω 2
+ iωγ
∞ ≈ 1 ⇒ ω = 2
ω − ω p2 + iωγ
• In metals ω p  γ = τ
−1
and usually ∞ = 1 so 𝜖𝜖𝜔𝜔
−1
peaks and ω ≈ 0
p
• More generally as ω → 0 at ω = ω p∞ we get Plasma Oscillations at a

−1/2
frequency defined by ω p2 = ne 2 / m0

• Polarisation causes build-up of surface charge, which generates the
restoring force driving the oscillations. The electrons slosh back and forth.

Optical properties of metals: plasma oscillations
and electron energy loss spectroscopy in Ge and Si
EELs
• Incoming electron at
wavevector 𝑞𝑞 , energy ℏ𝜔𝜔.
Outgoing electron at
wavevector 𝑘𝑘 , energy ℏ𝜈𝜈 .
Plasmon generated with From optical
energy ℏ 𝜔𝜔 − 𝜈𝜈 absorption
−1 ω 2
+ iωγ
• Response to oscillating applied D field given by ω = 2
2
ω − ω p2 + iωγ
ne
• Resonance at plasma frequency ω p2 = width 𝜏𝜏 −1 .
QCMP Lent/Easter 2019 m0 2.10
Transport in electric and magnetic fields
2
dj(t ) j(t ) nq E q
• Previously =
− + + ( j× B)
dt τ m m
• If B is parallel to z-axis, taking components
( ddt + τ −1 ) j y = q
m (nqE y − Bjx )
( ddt + τ −1 ) jx = m
q
(nqEx + Bj y )
• Steady state =0 dj
dt
j y = qmτ (nqE y − Bjx ), jx = qτ
(nqEx + Bj y )
m
• Hall effect: current confined to x-axis ( j y = 0) by transverse field (E y ) due
to charge build up.
Ey 1
• Hall coefficient =RH = gives carrier density and sign of charge
jx B nq
qτ qτ
If B = 0 then jx =m nqEx ⇒ nqEx =Ex =m =µ carrier ‘mobility’
j v
•
x x
• Mobility defined as velocity of charge carrier per unit electric field.
nq 2τ
• since j = σ E ⇒ σ = m = nq µ
• Cyclotron frequency defined as ωc =eB m ⇒ ωcτ = µB
Hall effect in metals
R Lück, Phys Stat. Sol. 18, 49(1966)
• First measured 1879 E H Hall
−1 / Rh ne
• From the last slide RH = 1/ nq aluminium
• Drude theory predicts RH is ωτ = µ B
independent of B and τ 0.01 0.1 1.0 10 100 1000
• In metals however it is found that RH

does vary with magnetic field as well
−0.33
as temperature and sample purity
• Measuring pure samples at low Metal Valence -1/Rhne
temperatures and high magnetic Li 1 0.8
fields (1T) limiting values are Na 1 1.2
obtained. K 1 1.1
• Comparison made between observed Cu 1 1.5
and predicted number of free Ag 1 1.3
electrons per atom for a range of Au 1 1.5
metals
Be 2 -0.2
• Theory seems to work quite well for Mg 2 -0.4
alkali metals & noble metals but not
In 3 -0.3
for the others - new theories needed!
Al 3 -0.3
Scanning Hall Probe Microscopy
• First report Chang et al Appl Phys Lett 61, 1974 (1992)
• Hall sensor mounted on scanning system to measure local magnetic field
Magnetic field sensitivity 10 T spatial resolution 0.35µ m
−5
•
• Now commercial product from Attocube, Nanomagnetics, Magcam……

Problems with Drude Theory
• Drude model predicts the electronic head capacity to be from equipartition
of energy Cel = 32 nk B independent of T.
• Measured heat capacity falls far below that expected from equipartition
theorem and is temperature dependent.
• Correct description: degenerate Fermi gas (cf. Stat. Phys. course).

• Interpret velocity u in Drude model as drift velocity, averaged over many
particles. Individual electrons actually travel at up to 1% of speed of light!
• The Drude model – frequency dependant conductivity

• Optical properties of metals
• Experimental reflectivity of aluminium
• Plasma oscillations, electron energy loss spectroscopy
• Hall effect from Drude model
• Hall effect in metals
• Applications of Hall effect – scanning Hall probe microscopy
• Problems with Drude theory

Lecture 2
The End

Lecture 3
David Ritchie


approximation.

Sommerfeld Model – density of states
• Free electron gas - Schrodinger equation: − 2m ∇ 2ψ (r ) = Eψ (r )
2
• Introduce eigenstates
ψ k (r )= A exp(ik ⋅ r ), Ek=  k / 2m
2 2
satisfying periodic boundary conditions ψ ( x + L, y, z=

) ψ ( x, y, z ) etc.
• Allowed values of momentum are discrete: k = 2Lπ ( nx ,n y ,nz ) where
𝑛𝑛𝑥𝑥 , 𝑛𝑛𝑦𝑦 , 𝑛𝑛𝑧𝑧 are positive or negative integers
• At zero temperature
fill up Fermi sphere
to the Fermi energy EF
• Each triplet of
quantum numbers
corresponds to 2
states – electron
spin degeneracy k F = (3π 2 n)1/3
• volume in k-space
(2π / L)3
Sommerfeld Model – density of states
4 / 3π k F3
• Number of occupied states in Fermi sphere: N = 2 ⋅
(2π / L)3
• Hence if n= N / V= N / L3 then k F = (3π 2 n)1/3
• Also
EF=  2 k F2 / 2m=  2 (3π 2 n) 2/3 / 2m ⇒ ln( EF )= 2
3 ln(n)+const
dEF 2 dn dn 3 n
⇒ = ⇒ = g ( EF ) =
EF 3 n dEF 2 EF
• Density of states
Volume of shell in k − space 4π k 2 dk
g ( E )dE =
2⋅ =
2⋅
Volume of k − space per state (2π )3 / V
• Hence 1
V 2 dk V m  2mE  2
= g ( E ) 2= 4π k 2  2 
(2π ) 3
dE π    
2
• Factor of 2 for spin degeneracy

• Often g ( E ) is given per unit volume so V disappears.
Sommerfeld Model – electronic specific heat
• Occupancy of states in thermal equilibrium – Fermi distribution:
1
f (E) =
e( E −µ )/ kBT + 1
• Chemical potential T =0 ⇒ µ = EF
1
• Number density of particles n = N / V = ∑ f ( Ei ) = ∫ g(E)f(E) dE
• Energy density u = ∫ Eg(E)f(E) dE V i
• At room temperature k BT ≈ 0.025eV For metals EF ≈ a few eV hence
k BT  E F
∂f ( E )
• From above cv = ∂u / ∂T |v = ∫ Eg ( E ) dE
∂T
∂f ( E )
• Since the Fermi function is a step function is sharply peaked at the
chemical potential ∂T
• Contributions to the specific heat only come from states within k BT of the
chemical potential, with each state having specific heat k B and we can
guess that kT
cv ≈ n B k B
QCMP Lent/Easter 2019 EF 3.5
• To calculate this more accurately…. We take the density of
states as a constant so ∂f ( E ) ∂f ( E )
=cv ∫ Eg ( E )
∂T
dE ≈ g ( E ) ∫ E
∂T
dE
∂f ex x 1 ∂µ 
• changing variables x = ( E − µ ) / k BT ⇒ = x × +
∂T (e + 1)  T k BT ∂T 
2
• The number of particles is conserved so

dn ∂f ( E ) ∞ ex x 1 ∂µ 
=0 =g ( EF ) ∫ dE =g ( EF )k BT ∫ × +  dx
dT ∂T −∞ (e + 1)  T k BT ∂T 
x 2
• The first term in the square brackets is odd, xe 2 is even so ∂µ  0

x
( e +1) ∂T
• To the same level of accuracy:
∂f ( E ) ∞ ex x
cv )∫ E
g ( EF= dE g ( E F ) k B T ∫ ( µ + k B T x ) x dx
∂T −∞ (e + 1) T
2
∞ x 2 x
e π 2
( EF )k B2T ∫
g= d x k 2
BTg ( EF )
−∞ (e x + 1) 2 3
3 n
• Since g ( EF ) = we can write
2 EF
π2 2 π 2 2 3 n π 2 k BT π2 T
cv =k BTg ( EF ) =k BT = nk B nk B
3 3 2 EF 2 EF 2 TF
• This result is of the same form as the equation above obtained from a
simple argument but with a different prefactor - π / 2 as opposed to 1.
2
• This calculation is the leading order term in an expansion in powers of

. ( k BT / E F ) 2
• To next order the chemical potential is temperature dependent (see below)
but because for metals k BT  EF we can usually ignore it.
1 π k BT 2




µ =E 1 − (
F  ) + O ( k BT / E F ) 4 

3 2 EF 
 
• Examples:
• Electron gas in solids – often much smaller than lattice specific heat
• Liquid helium mixtures of 3He in 4He – near ideal Fermi gas

Specific Heat of mixtures of 3He and 4He
• Experimental procedure:
4K
• Cool helium mixtures to mK temperatures,
• Isolate from surroundings
• Input heat for given time
• Measure temperature rise Helium dilution
• Calculate specific heat at particular refrigerator
1.5K
temperatures and pressures
0.7K
50mK
5mK
From wikipedia
QCMP Lent/Easter 2019 Polturak and Rosenbaum JLTP, 43, 477 (1981) 3.8
Specific heat of mixtures of liquid 3He and 4He
π2 T 2
EF
( 3π n )
2/3
• From above = = = 2
cv nk B , TF
2 TF k B 2mk B
• Linear behaviour in Fermi gas regime T  TF
• So knowing cv , T , n we can calculate TF the Fermi temperature

and m the effective mass
• Effective mass 2.44 to 3.07 time bare 3He mass – due to interactions
QCMP Lent/Easter 2019 Polturak and Rosenbaum JLTP, 43, 477 (1981) 3.9
Screening and Thomas-Fermi approximation
• Placing a positive charge in a metal will result in electrons moving around
to screen its potential resulting in zero electric field.
• This is quite different from a dielectric where electrons are not able to
move freely and the potential is reduced by dielectric constant
• In a classical picture electrons can move anywhere, but quantum
mechanics dictates this is not possible - an electron cannot sit right on top
of a nucleus.
• In metals a balance is reached between minimising potential and kinetic
energy, screening over a short but finite distance.
• We estimate the response of a free electron gas to a perturbing potential.
𝑉𝑉0 (𝐫𝐫) is the electrostatic potential, 𝜌𝜌0 (𝐫𝐫) the charge distribution.
ρ0 (r )
∇ V0 (r ) =
2
−
0
• Consider the positive background charge to be homogeneous with the
electron gas moving around - plasma or “Jellium” model and in this case
𝜌𝜌0 (𝐫𝐫) = 0 everywhere (this does not include the charges used to set up the
perturbing potential).
Screening and Thomas-Fermi approximation
• In the presence of a perturbing potential Vext (r ) the electron charge
density redistributes ρ= (r ) ρ0 (r ) + δρ (r ) which changes the potential
,=
V (r ) V0 (r ) + δ V (r ) . The changes are related by: ∇ 2δ V (r ) = − δρ( r )
0
• We link the charge redistribution to

the applied potential by assuming the
perturbing potential shifts free
electron energy levels – the same as
assuming a spatially varying Fermi δρ ( r )
energy. This is the “Thomas-Fermi”
approximation.
• The potential is the total produced by

the added external charge and the
induced “screening” charge
,=
V tot Vext + δ V hence:
2 2
− ∇ ψ (r ) + (−e)(δ V (r ) + Vext (r ))ψ (r ) =Eψ (r )
2m
Screening and the Thomas-Fermi approximation
• Assume the induced potential is slowly varying on the scale of the
Fermi wavelength 2π / k F so the energy eigenvalues are just shifted
by potential as a function of position:
E (k , =
r ) E0 (k ) − eVtot (r )
2k 2
where E0 (k ) has a free electron parabolic dispersion 2m
• Keeping the electron states filled up to a constant energy µ means we
adjust the local Fermi energy EF (r ) as measured from the bottom of
the band so:
=µ EF (r ) − eVtot (r )
• A small shift in the local Fermi
Energy leads to a change in the
local electron number density, n.
δn V ( EF )δ EF
g= egV ( EF )Vtot
• And from above V= tot Vext + δ V
so we have:
=δ n egV ( EF ) (Vext + δ V )
• Since the added potential and induced electron number density are small
we can use Poisson’s equation to write:
2
e e gV ( EF )
∇ δ V (r ) = δ n =
2
(δ V (r ) + Vext (r ))
0 0
• We can calculate the induced potential and density response using Fourier
transformation. Assume an oscillatory perturbing potential : Vext = Vext (q)eiq .r
and a resulting oscillatory induced potential: δ V = δ V (q)e
iq .r
substituting
into the equation above: 2
∇ (δ V (q)e ) e gV ( EF )
2 iq⋅r
=
− q δ2
V (q ) e = iq⋅r
(δ V (q ) + V (q )) e iq⋅r
0
ext
⇒ δ V (q)  q 2 + e 2 gV ( EF ) / 0  =
−Vext (q)e 2 gV ( EF ) / 0
e 2 gV ( EF ) / 0 2
qTF
⇒ δ V (q) =
−Vext (q) 2 2 =
− 2 Vext (q)
q + e gV ( EF ) / 0 q + qTF2
• Where we define the Thomas–Fermi wavevector:

1
 2
=qTF 

e gV ( EF ) / 
2

0
1
 2
=
• The Thomas-Fermi wavevector qTF 

e gV ( EF ) / 
2
0  , and given that for
the free electron gas:
( )
3/2  2 k F2
gV ( E ) = 1
2π 2
2m
2
1/2
E , EF = 2m ⇒ gV ( EF ) = m
k
π 22 F
2
1 me 4 k F  2.95 −1 
2
• We obtain:= = =  1/2 Å 
2
q k
π 0 
TF 2 2 F
π aB  rs 
aB 4πme 20  0.53Å
2
=
• where the Bohr radius is
and the Wigner-Seitz radius, rs is defined by (4π / 3) rS =n −1
3
• To find the induced electron number density – from above we have

2
qTF
egV ( EF ) (Vext + δ V ) , δ V (q) =
δn = − 2 2
Vext (q ), q 2
TF =
e 2
gV ( EF ) / 0
q + qTF
2
qTF 0  2
qTF  0  q2 
⇒ δ n(q) = nind (q) = Vext (q) 1 − 2 2  = Vext (q)  2 2 
e  q + qTF  e  q / qTF + 1 
The Thomas-Fermi dielectric function
• The wavevector dependent dielectric function  (q) relates the electric
displacement D to the electric field E by 0 (q)E(q) = D(q)
• given:
∇Vext =
− D(q) / 0 , ∇Vtot =
−E(q) ⇒ ∇Vext =
 (q)∇Vtot
• Since from above
Vtot = Vext + δ V ⇒ Vext (q) =  (q) (δ V (q) + Vext (q) )
• Using 
Vtot
2
qTF q2
δ V (q) =
− 2 2
Vext (q) ⇒ Vtot (q) =
Vext (q) 2 2
q + qTF q + qTF
• And hence the “Thomas Fermi dielectric function” is given by:
2
q
TF (q) = 1 + TF2 
q
−1
• qTF is the Thomas-Fermi screening length, for copper where the electron
density =n 8.5 ×1022 cm −3 we have 1/ qTF = 0.055nm .
Thomas-Fermi screening
• From last slide  (q) = 1 + qTF / q 
TF 2 2
0 5 r
−2
• For small q (long distances)  ∝ q
TF
Potential energy
exp(−kr )
• Long range part of Coulomb potential -0.1 r
−2
also ∝ q so it is exactly cancelled screened
• In real space if Vext= Q / r (Coulombic -0.2
1
unscreened
and long range) then V ( r=) (Q / r )e − qTF r r
is the short range screened potential. -0.3
(problem sheet 1 question 6) In this graph screening length 1/k=1

• The screened potential is known as the “Yukawa potential” in particle physics
• Exponential factor reduces range of Coulomb potential – screened over
distances comparable to inter-particle spacing
• Mobile electron gas highly effective at screening external charges.
• Application to resistivity of alloys – atoms of Zn (valency 2) added
substitutionally to metallic copper, (valency 1) has an excess charge.
• Foreign atom scatters conduction electrons with interaction given by
screened Coulomb potential – scattering contributes to increase in in
resistivity, theory and experiment in agreement.
• The Sommerfeld model – electrons in a degenerate Fermi gas

• Free electron gas in three dimensions
• Fermi surface and density of states
• Thermal properties of the Fermi gas – specific heat
• Experimental measurements of specific heat in liquid helium.
• Screening and the Thomas-Fermi approximation,
• Thomas-Fermi wavevector and dielectric function
• Effect of screening on a Coulomb potential

Lecture 3
The End

Lecture 4
David Ritchie


Types of bonding; Van der Waals, ionic, covalent. Crystal structures.
Reciprocal space, x-ray diffraction and Brillouin zones. Lattice
dynamics and phonons; 1D monoatomic and diatomic chains, 3D
crystals. Heat capacity due to lattice vibrations; Einstein and Debye
models. Thermal conductivity of insulators. Electrons in a periodic
potential; Bloch’s theorem. Nearly free electron approximation; plane
waves and bandgaps. Tight binding approximation; linear combination
of atomic orbitals, linear chain and three dimensions, two bands.
Pseudopotentials.. ……….


From atoms to solids – the binding of crystals
• Cohesion due to interaction between electrons and nuclei giving rise to
effective interaction potential between atoms
• Several different types of bonds – firstly Van der Waals
• Example: Inert gases – filled electron shells, large ionization energies
• Electron configuration in solid similar to that in separated atoms
• Atoms neutral – interaction weak, due to Van der Waals interaction
• Consider atom as oscillator – electrons fluctuate around nucleus
• Zero point fluctuations cause dipole moment p1, a second atom at distance R
experiences induced electric field E ∝ p1 / R
3
• This field induces a dipole at the second atom p2 ∝ α p1 / R where α

3
is the atomic polarizability

• The second atom induces an electric field at the first
E1 ∝ p2 / R 3 ∝ α p1 / R 6 .
• The energy of the system changes by: ∆U = − p1 ⋅ E1 ∝ −α p1 / R .
2 6
• Induced interaction ∝< p12 > and is always attractive

• Note - Energy depends on < p1 > and not < p1 > 2 which is zero
2

From atoms to solids – Van der Waals bonding
• If atoms move together so the electron charge distributions overlap they
repel each other
• Repulsion due to electrostatic forces and the Pauli exclusion principle
which prevents electrons having the same quantum numbers
• For example if we try to force 2 spin parallel electrons into the same 1s
state in H, one will go into the 2s state at a large energy cost
• Calculations of repulsive interactions is complex
there is a short range (hard core) potential
• Common empirical “Lennard-Jones” potential
 σ 12  σ 6   Rmin 12  Rmin  

6
) 4ε   −   =
V (r=  ε   − 2  
 r   r    r   r  
• ε , σ are constants depending on atoms involved

• Except He, Inert gases form close-packed face-centered cubic solids with
high coordination nos. (10-12), low cohesive energies and melting points
• He is special – due to zero point motion does not solidify even at absolute
zero, unless pressurised to 30 Bar!

Van de Waals bonding for low-dimensional
structures
Liu et al, Nature Materials 1, 16042 (2016)

From atoms to solids – Ionic bonding
• Atoms with electronic configuration close to a filled shell will tend to gain or
lose electrons to fill the shell
+ −
• Energy for reaction M → M + e in gas phase is ionization energy I
− −
• Energy for reaction X + e → X in gas phase is electron affinity A
• To form ionic molecules it costs an energy I + A but the electrostatic
potential energy between charges is reduced by a greater amount.
1
∑ ∑
• Electrostatic interaction for a diatomic crystal:
U electrostatic = U ij
2 i j
• Where U ij = ± q / 4πε 0 Rij is sum of all Coulomb energies between ions.
2
• If the system is on a regular lattice with constant R then we can write:

U electrostatic = −α M q 2 / 4πε 0 R
where the Madelung constant depends on structure NaCl (α M = 1.7476)
CsCl (α M = 1.7627) cubic ZnS or Zincblende (α M = 1.6381)
• We must add repulsive short range force, since ions are different sizes we
get different energies - explains why NaCl has rocksalt structure and not
CsCl structure
• Intermediate coordination numbers (6-8)
Crystal structures
QCMP Lent/Easter 2019 From Kittel 4.7

From atoms to solids – Covalent bonding
• Consider covalent bonding in hydrogen molecule – due to electron pair
• Overlapping orbitals on neighbouring atoms hybridise
• Hamiltonian symmetric about point between nuclei, hence eigenstates have
odd or even parity about this point
• With a basis of atomic states φ (r − R ) where the nucleus is at R we get
two states, of odd and even parity
ψ ± (r ) = φ (r − R a ) ± φ (r − R b )
• ψ + is non zero and ψ − has a node between the nuclei
• For an attractive potential we will have E+ < E− and the two electrons of
opposite spin will fill the lower ‘bonding’ state ψ + .
• The ‘anti-bonding’ state will be separated from the bonding state by an
energy gap E= g E− − E+ and will be unfilled
E−
φa φb
Va (r ) Vb (r )
Ra Rb E+

• An example of degenerate perturbation theory
• We calculate single-electron energy levels, ignore electron-electron
repulsion and use Dirac notation, so:
φ (r − R a )
a = φ (r − R b )
b =
• Look for energy eigenfunction within subspace spanned by orthonormal
ψ α a +β b
basis functions:=
• Apply Hamiltonian, and since a and b have the same eigenenergy:
H ψ =E ψ ⇒ α H a + β H b =α E a + β E b
• Left multiply by a and b we obtain two equations:
α a H a +β a H b =
αE
α b H a +β b H b =
βE
∗
• From this, given H ba = H ab we can obtain an eigenvector equation for the
coefficients α , β which only has a non trivial solution if the determinant is
zero:
H aa − E H ab
=0
∗
H ab Hbb −E
• From previous slide: H aa − E H ab

=0
∗
H ab Hbb −E
• From this we can find the energy eigenvalues which are distributed around
the average of H aa and H bb
1/ 2
 
H aa + H bb ∆E  H aa − H bb 
2
∆E  2
=
E ± , = 
   +H 
ab 
2 2 2 

 2  

• For H aa  H bb , ∆E / 2 =
| H ab | we get covalent bonding
• If H aa << H bb or H aa >> H bb then E = H aa or E = H bb and H ab is

irrelevant - we have ionic bonding
• See Problem Sheet 1 question 7

Covalent and ionic semiconductors
• With s electrons molecules are formed
which form a weakly bound molecular
solid.
• With p and d orbitals, bonds become
directional such as sp3 in C, Si and Ge
• These hybrids point in tetrahedral
directions (111),(1 11),(11 1),(1 1 1)
and each atom donates 1 electron
• This forms an open tetrahedral network
– “diamond structure” Tetrahedral bonding in diamond
• In GaAs and cubic ZnS the total number of electrons satisfies “octet rule”
• Same tetrahedral structure as diamond with alternating atoms - “zincblende”
• Zincblende structure cohesion is partly covalent and partly ionic
• Hexagonal crystal structure based on local tetrahedral network - “wurtzite”
• Wurtzite structure favoured in more ionic systems.
• With increasing ionicity we get : Group IV Ge (diamond), III-V GaAs (zincblende),
II-VI ZnS (zincblende or wurtzite), II-VI CdS (wurtzite), I-VII NaCl (rocksalt)
Metals
• Band forms from atomic states. Partially filled which
implies energy gain generalisation of covalent bond
to ‘giant molecule’
• Electrons in band states delocalised, high
conductivity. Bonding is isotropic, like van der Waals.
• Close packing. to maximise density while keeping
atomic cores far apart: fcc or hcp. High coordination
numbers ≈ 10 − 12
• Screening by conduction electrons. Screening length
of order atomic spacing  1Å
• Within a row in periodic table: ion core potential
grows, density increases, crossover to covalent
semiconductors, then insulating molecular structures.
• Transition metals. d-electrons more localised, inside
s- and p- orbitals. Are often spin-polarised hence
magnetism in 3d elements. 4d and 5d orbitals
overlap giving high binding energy (e.g. W melts at
3700K).

Crystal Lattices
• An ideal crystal is an infinite repetition of structural unit in space
• Repeating structure called the lattice
• Group of atoms which is repeated is called the basis
• The basis may be a single atom or very complex – such as a polymer
Lattice Basis Crystal

• Lattice is defined by primitive translation vectors: ai , i = 1, 2, 3
• An arbitrary lattice translation is: T =∑ na
i i i
• The atomic arrangement looks the same from equivalent points in the unit
r + ∑ ni a i ∀ integer ni
cell:
r′ =
i
• The lattice formed is called a Bravais lattice
Wigner-Seitz cell
• The region in space around a particular lattice point closer to it than any
other lattice point
• To construct – draw lines from a given lattice point to all of its neighbours.
• Draw planes perpendicular with each line intersecting at the line midpoint.
• The smallest volume enclosed is the Wigner-Seitz primitive unit cell
Bcc lattice, Wigner-Seitz cell fcc lattice, Wigner-Seitz cell

Space and point groups
• Symmetry operations which map lattice onto itself Space group

• Map lattice onto itself, but keep one point fixed Point group.
• Point group operations: reflections, inversions, rotations
• Seven point groups for Bravais lattices = Seven crystal systems
• cubic, tetragonal, orthorhombic, monoclinic, triclinic, trigonal, hexagonal.
• Fourteen space groups for Bravais lattices
• General crystal structures (Bravais lattice + basis): 32 point groups, 230
space groups.

The Bravais lattice types
See also p.170 Singleton

Index system for crystal planes
• Coordinates of three lattice points enough to define crystal plane

• Label plane by coordinates where it cuts axes.
{ri } = { xa1, ya 2 , za3 }
• (hkl ) denotes plane which cuts axes at a1 / h, a 2 / k , a3 / l
= yk= zl= integer
or multiple thereof, so that xh
• ( hkl ) is the index of the plane

• For set of planes equivalent by
symmetry use { }
• e.g.{100} For a cubic crystal has
equivalent symmetry planes:
(100),(010),(010),
(100),(0 10),(00 1)
• Note that overbar 1 denotes negation
Crystal plane indices
From Kittel

• The binding of crystals

• Van der Waals
• Ionic
• Covalent
• Metals
• Crystal lattices
• Wigner –Seitz cell
• Space and Point groups
• Index system for crystal planes

Lecture 4
The End

Lecture 5
X-ray
detector sample source
monochromator
David Ritchie




The reciprocal lattice and diffraction
• Reciprocal lattice concept arises from scattering of waves by crystals.

• Builds on Fraunhofer diffraction from a grating generalised to scattering from
a 3D periodic lattice.
• Consider scattering of a plane wave off a single atom or more generally the
basis forming the unit cell
Bragg scattering k
• Incoming wave of wavevector k 0 is incident on
k0
potential centred at R i
• At large distances scattered wave is circular
• Total field taken as a scalar
ik0 |r − R i | k
e
ψ ∝ e ik 0 ⋅ ( r − R i ) + f G
| r − Ri |
k0
• Details of scattering buried in Form Factor f
• This is a function of scattering angle, type and arrangement of atoms etc.
• Total scattered intensity assumed small

The reciprocal lattice and diffraction
r⋅R
• At large distance from the scattering centre k0 | r − R i |≈ k0 r − k0 r i ( )
• Defining scattered wavevector k = k0 ( rr ) momentum transfer q= k − k 0
 e − iq⋅R i

• The waveform is given by: ψ ∝ e − ik 0 ⋅Ri  eik 0 ⋅r + ceik0r f 
 r 
θ ) f exp[−iq ⋅ R i ]
• Effective scattering amplitude f (=
• Summing over identical lattice sites the scattered intensity is proportional to
the differential scattering cross-section:
dσ 2
= f (θ= f ∑ R exp[−iq ⋅ R i ]
2
)
dΩ i
• We add terms with different phases q ⋅ R i leading to cancellation unless

Bragg condition q ⋅ R = 2π m is satisfied for all R with m an integer
• If two vectors satisfy the Bragg condition then so will their sum hence the
special values of q ≡ G satisfying this lie on a reciprocal lattice
• Primitive vectors of the reciprocal lattice in terms of real space lattice
primitive vectors:
a 2 ∧ a3
b1 = 2π and cyclic permutations
QCMP Lent/Easter 2019 a1 ⋅ a 2 ∧ a3 5.4
Diffraction conditions
• For elastic scattering conservation of energy requires k = k 0 and Bragg
condition requires k − k o = G where G is a reciprocal lattice vector
• Combining these two conditions k ⋅ G / 2 = ( )
2
G / 2
• This defines a plane perpendicular to G which intersects G at its midpoint
• The set of all such planes defines the incident wavevectors that satisfy the
conditions for diffraction
• Ewald construction – in reciprocal space

• Reciprocal lattice points shown
• k 0 incident wavevector with origin chosen
(point a ) so it ends on a reciprocal lattice point G
k
• Sphere radius k 0 drawn about origin of k 0
2θ
• Diffracted beam forms if sphere intersects any
θ k0
other reciprocal lattice point k0
a
• Origin of k 0 is on perpendicular bisector of G
• Angle θ is the Bragg angle and we obtain
2 k sin θ = G
Diffraction conditions
• From last slide 2 k sin θ = G

• The spacing between parallel lattice planes perpendicular to
G = hb1 + kb 2 + lb3 is given by d ( hkl ) = 2π / G
• Given that k = 2π / λ we can write
2 ( 2π / λ ) sin θ =2π / d ( hkl ) ⇒ 2d ( hkl ) sin θ =λ

• Where θ is the angle between the incident beam and the crystal planes and
equals half the angle of deflection
• The indices defining an actual crystal plane may contain a common factor n
and we can generalize this equation to give the conventional form of Bragg’s
law:
2d sin θ = nλ

High resolution x-ray diffraction
• Find thickness and composition of
SiGe thin films on Si substrates
X-ray
• From Bragg’s law thickness detector sample source
λ ( 2cos ω∆ω )
−1
t
• Vertical lattice constants cL , cS monochromator
found from cL sin ωS
=
cS sin ωL
• Hence concentration of Ge found
sample
ω ω
X-ray source
detector
View from above

X-ray analysis of superlattice structure

Diffraction conditions and Brillouin zones
• The set of reciprocal space planes satisfying
the Bragg condition is constructed by finding
those planes which are perpendicular
bisectors of every reciprocal lattice vector G .
• The planes so constructed divide reciprocal
space up into cells. The cell closest to the
th
origin is called the first Brillouin zone. The n
Brillouin zone consiststh of all the fragments
exterior to the ( n − 1) plane but interior to
the nth plane.
• The first Brillouin zone is the Wigner-Seitz Reciprocal lattice point

Reciprocal lattice vector
cell of the reciprocal lattice which has an vector perpendicular bisector
important role in discussion of electronic 1st Brillouin zone
states in a periodic potential. 2nd Brillouin zone
• The volume of each Brillouin zone (adding up the fragments) is equal to the
volume of the primitive unit cell of the reciprocal lattice, which is (2π ) / Ω cell
3
where Ω cell is the volume of the primitive unit cell of the crystal.
Brillouin zone in 3D
a ∧a
• Using b1 = 2π a ⋅a2 ∧3a etc we can calculate primitive vectors in reciprocal
1 2 3
space for a number of lattices
• Reciprocal lattice of simple cubic lattice is also simple cubic
• Reciprocal lattice of face centred cubic is body centred cubic
• Reciprocal lattice of body centred cubic is face centred cubic
• Example: face centered cubic lattice
Primitive vectors of FCC lattice Reciprocal lattice of FCC is BCC

Brillouin zones shown
QCMP Lent/Easter 2019 (Taken from Kittel) 5.10
Lattice dynamics and phonons – 1D monatomic chain
• Consider a row of identical atoms

distance a apart connected by springs
of constant K and free to move
horizontally
• Displacement of n atom at point rn = na is given by un , equation of

th
motion: ∂ 2 un
m = K (un+1 − un ) + K (un−1 − un )
∂t 2
• We guess wave solution:= un (t ) uo cos(qrn − ω (q )t )

wavelength = λ 2π / q and period = T 2π / ω (q )
• Substituting into equation of motion we obtain
mω 2 (q ) =
2 K (1 − cos(qa )) =
4 K sin 2 (qa / 2)
• Hence dispersion relation between frequency and wavevector
ω=
(q ) 2( K / m)1/ 2 sin(qa / 2)
Lattice dynamics and phonons – 1D monatomic chain
• From last slide ω=
1/ 2
(q ) 2( K / m) sin(qa / 2)
• Periodic in q with period 2π / a
• Long wavelength modes ( qa → 0)
linear dispersion ω ( q ) = qa K / m
• Same as for a wire with tension Ka
and density m / a
• compressive waves with velocity
=v a ( K / m)1/ 2
• These waves behave like sound
waves - acoustic mode
• For larger q dispersion is periodic, phase shift between neighbouring

atoms given by eiqa
• So for qa = 2π neighbouring atoms move in phase, for qa = π they
move in anti-phase
• We simplify things by only considering a range of πa ≤ q ≤ πa
• This corresponds to the first Brillouin zone
Lattice dynamics and phonons – 1D diatomic chain
• The monoatomic chain contains only acoustic modes. The spectrum
becomes more complicated with more atoms per unit cell.
• Assume two different atoms with different masses and spring constants
• There are two equations of motion

∂ 2unA ′
mA = K ( u − u ) + K (un−1,B − unA )
∂t 2 nB nA
∂ 2unB ′
mB = K (un+1 A − unB ) + K (un, A − unB )
∂t 2
• The solution is quite complicated – so look at a limit

• Suppose the atoms are the same mass so mA = mB and are quite strongly
′
bonded in molecular pairs so K  K
• Every ‘molecule’ will have a mode where the atoms oscillate out of phase at
a frequency ω = 2K / m
2
o
1D diatomic chain – different spring constants
= 0)
• The coordinate undergoing oscillation is uopt ( q = u A − uB
• We have assumed this is at q = 0 where each molecule undergoes the
oscillation in phase with the next
′
• Since K  K the restoring force and frequency is largely independent of q
• There are two branches to the
dispersion curve
• The acoustic branch the (low)
frequency vanishes linearly
with wavevector.
• The optical branch has finite
(high) frequency as q → 0 and
is referred to as optical a2
𝜋𝜋/a2
because of interaction with
light Acoustic
• Atomic displacements for two
modes quite different
Optical
QCMP Lent/Easter 2019 Note that a2 is the distance between two ‘molecules’ 5.14
1D diatomic chain – different masses
• Using the equations of motion assuming K ′ = K but mA ≠ mB
∂ 2unA ∂ 2unB
mA = K (unB + un−1,B − 2unA ), mB = K (un+1 A + un, A − 2unB )
∂t 2
∂t 2
• With lattice constant a , substituting in travelling wave solutions with

amplitudes α , β on alternate planes:
= unA α exp(inka ) exp(−iωt ), = unB β exp(inka ) exp(−iωt )
and solving the simultaneous equations leads to
mA mBω 4 − 2 K (mA + mB )ω 2 + 2 K 2 (1 − cos ka ) =
0
ω
• May be solved for ω 2- see diagram
• Assuming k = ±π / a or  1 1  Optical phonons
2K  + 
ka  1 ⇒ cos ka =
1 − k 2a2 / 2  A mB 
m
2 K / mB
we obtain the limiting results shown mA > mB
• Notice there are no solutions for ω 2 K / mA
between 2 K / mA and 2 K / mB
• If we look there for real ω solutions
K
we find k is complex – so wave is ka acoustic phonons
2mA + 2mB k
damped π /a
Lattice dynamics and phonons – 3D crystal
• Extending to 3D requires a dispersion relation ω (k ) describing waves
propagating in different directions
• As well as compressional waves there are transverse waves resulting in
three branches of phonons, two transverse and one longitudinal.
• There are always 3 acoustic modes and for a solid with m atoms per unit cell
there will be 3( m − 1) optical modes – again split into two transverse and
one longitudinal.
• Phonon dispersion
in Ge - each panel
a different direction
• Diamond structure
• 2 atoms per unit
cell
• Measured by
inelastic neutron
scattering

Inelastic neutron scattering ε (k ) =
 2k 2
k ′, ε (k ′)
• Incident neutron transfers 2m
some of its (well controlled)

Q =k − k ′ + G
energy ε and momentum k k , ε (k )
Q) ε (k ) − ε (k ′)
E (=
to phonon Incident n
• Measurement made of Q, E (Q)

outgoing neutron energy and
momentum ε (k ′), k ′
• Most common method for
measuring phonon dispersion
curves E (Q)
• Can also measure phonon
lifetimes
• Quantitative technique, can
measure throughout Brillouin
zone
• Neutrons produced by nuclear
reactor – so large facilities
Multi axis crystal spectrometer at the National Institute for
required. Standards and Technology Centre for Neutron Research in
Phys Rev Lett 110, 157401 (2013) Gaithersburg, Maryland, USA –said to be the most intense
QCMP Lent/Easter 2019 neutron beam in the world. 5.17
• Reciprocal lattice and diffraction

• Diffraction conditions
• High resolution x-ray diffraction
• Diffraction conditions and Brillouin zones
• Brillouin zones in 3D
• Lattice dynamics and phonons – 1D monatomic chain
• 1D diatomic chains – different spring constants
• 1D diatomic chains – different masses
• Inelastic neutron scattering for determination of phonon dispersion

Lecture 5
X-ray
detector sample source
monochromator
The end

Lecture 6
David Ritchie




Lattice dynamics and phonons – density of states
• We need to calculate density of states for phonons

• In 1D monoatomic chain of N atoms (N large), there are N degrees of
freedom and hence N modes.
• So the allowed points in k-space are
2π  N N N
kn = n ; n =  − , − + 1,..., 
L  2 2 2
• Here −π / a < k < π / a and the lattice constant a = L / N
• Same spacing of k-states as for electrons, but discrete nature of atoms leads
to maximum momentum – from counting degrees of freedom on Brillouin
zone boundary
• In 3D each branch of phonon spectrum has N states but N= L / Ω cell
3
where Ω cell is the volume of the unit cell and L = V the volume of the
3
crystal
• Volume associated with each allowed k-point is ∆k = (2π )3 / L3
• There are 3 acoustic branches and 3( m − 1) optical branches ( m atoms per
unit cell)
Lattice dynamics and phonons – Density of states
• We start with the optical branches and use the Einstein Model which
assumes a flat dispersion: ω (k ) = ω0 In this case the density of states is
E (ω ) N δ (ω − ω0 )
simply D=
• For the acoustic modes, which disperse linearly
as ω → 0 we have ω = vk where v is the speed
of the mode and following similar arguments as
for electrons we obtain the Debye model result
4π k 2 dk Vω2
DD (ω ) =
=
(2π / L)3 d ω 2π 2 v 3
• This cannot apply as the dispersion approaches the zone boundary, in the
Debye model the spectrum is cut off at a frequency ω D and wavevector k D
• This frequency is determined by counting the total number of states and
setting this equal to N ,
ωD 6π 2 3
v N
hence:
∫0
DD (ω )dω = N =
Which gives ωD3
V
.
• This corresponds to replacing the correct cutoff (determined by intersecting

Brillouin zone planes) by a sphere of radius k=
D ωD / v
Heat capacity due to lattice vibrations- Einstein model
• Phonons obey Bose-Einstein statistics, their number is not conserved so the
chemical potential is zero and we use the Planck distribution
1
n(ω ) =
exp(ω / k BT ) − 1
• Note that n → 0 for k BT < ω and n → k BT / ω for k BT > ω
• The internal energy is given by: U = ∫ D (ω ) n(ω )  ω dω
• For the Einstein model given that
E (ω )
D= N δ (ω − ω0 )
N ω0
• We obtain U E = ωo / k BT
and for the specific heat capacity
e −1 2 ω0 / k BT
 ∂ U    ω  e
= CV =  Nk B  0  ω0 / kBT
 ∂T V − 2
 B 
k T ( e 1)
• At low temperatures this varies as exp( − ω0 / k BT ) and is very small, but it
saturates at a value of Nk B (the Dulong and Petit law) above the
characteristic temperature = θ E ω 0 / k B
Heat capacity due to lattice vibrations – Debye model
• At low temperatures the contribution of optical modes is small and the Debye
spectrum is appropriate giving for each mode:
Vω2
ωD ω
U D ∫= D(ω )n(ω ) ω dω ∫ ω
dω
0 2π v e
2 3  / k BT
−1
• Defining the Debye temperature= θ D ωD / k B and using ωD3 = 6π 2 v3 N / V
from above, we obtain θ D = (6π 2 N / V )1/3 v / k B
• By writing x = ω / k BT and including a factor of 3 for different modes we
obtain for the internal energy
3
3V  ωD ω3
T  θ D /T x3
2 3 ∫0 ∫
UD = ω / k BT
dω 9 Nk BT   dx
2π v e −1  θD  0 e −1
x
• and by differentiating with

respect to temperature, the
heat capacity – plotted right
3
 T  θ D /T x 4 e x
CV = 9 Nk B   ∫ dx
 θD  (e − 1)
0 x 2
For T  θ D , x → 0 ⇒ CV ≈ 3 Nk B

2016 6.6
1.6
Heat capacity due to lattice vibrations - Debye model
3
 T  θ D /T x 3
• From last slide U D = 9 Nk BT   ∫0 dx
 θD  e −1
x
• We approximate for very low temperatures by allowing θ D / T → ∞ and

using the standard integral ∞ x3 π4
∫
0 e −1
x
dx =
15
• Hence U D = 3π 4 Nk BT 4 / 5θ D3 .
• Differentiating with respect to temperature
θ D = 92K
3
12π 4
T 
CV = Nk B  
5  θD 
3
• T temperature dependence
• Experimental results for solid Argon
• Clearly a good approximation
L Finegold and N E Phillips Phys Rev 177,

QCMP Lent/Easter 2019 1383 (1969) 6.7
Heat capacity due to lattice vibrations and electrons
• In metals we have both electrons and
lattice vibrations giving rise to a low
temperature heat capacity of the form
α T + β T 3 ⇒ Cv / T =+
Cv = α βT 2
2
• If we plot Cv / T vs T we obtain a
straight line.
• α the T = 0 intercept is determined
by the electron specific heat
• β the gradient is determined by lattice
specific heat
W S Corak et al Phys Rev 98, 1699, (1955)
W H Lien and N E Phillips Phys Rev 133, A1370

QCMP Lent/Easter 2019 (1964) 6.8
Einstein and Debye models
• Comparison of Einstein and

Debye models of specific
Heat
• At low temperatures difficult
to thermally excite optical
modes hence Einstein is
below Debye
• Comparison of (a) Debye

density of States with (b)
density of states of a real
material
• Complex shape due to
adding different branches of
phonon spectrum and van
Hove singularities
Thermal conductivity of insulators
• The thermal conductivity, κ is defined by J q =−κ ∇T where J q is the flux
of heat (energy per unit area per unit time)
=
• Kinetic theory gives κ 13= Cv lv 13 Cv v 2τ ,Cv the phonon specific heat per
unit volume, v phonon velocity, l = vτ mean free path and τ scattering time
• Debye theory: Cv ∝ T at low T and constant at high T . v is velocity of
3
sound and almost constant, τ depends on several scattering mechanisms

First Brillouin zone
• Phonons can be scattered by (i) ky ky
other phonons – in “normal” and
“umklapp” processes (ii) point k1
kx k2 k1 kx
defects (iii) sample boundaries (iv)
crystal dislocations. k3
k2 k3 G k1+k2
• The overall scattering time given
by: τ −1 =τ ph−1
+ τ def
−1
+ τ boun
−1
+ τ dis
−1
Normal: k3=k1+k2 Umklapp: k3+G=k1+k2
• At low T , l is large and τ is often determined by sample boundaries –

constant with T so κ ∝ T
3
• In middle range of T , κ often peaks and impurity scattering is important.

• As T increases up to and beyond,TDebye , , phonon numbers and scattering
increases (umklapp particularly important) κ decreases rapidly
Thermal conductivity of insulators
• Rapid increase with κ ∝ T at low temperatures – boundary scattering can
3
dominate so size of sample important

• κ peaks in middle range of T and impurities determine peak value
• κ rapidly decreases at higher T as more phonons are excited with umklapp
particularly important due to with large changes in wavevector
• Diamond has a very high thermal conductivity at room T - 5 times copper
Thermal conductivity of LiF
Effect of boundary scattering Thermal conductivity of LiF
(A) 7.3mm Effect of impurities (isotopes) Thermal conductivity of diamond
(B) 4mm
0.1%
(C) 2.1mm 13C
0.02% 6Li
(D)1.1mm
0.01% 6Li
4.6% 6Li
1%
90.4% 6Li 13C
25% 6Li
50.1% 6Li
Curves
Curves correspond
correspond to
to different isotopic
different widths
purity of specimens
of specimen
Proc. R. Soc. Lond. Wei et al PRL 70, 3674 (1993)
Phys Rev 156,
A 1965 289 66-80
975 (1967)

Electrons in a periodic potential
• So far we have ignored the electrostatic potential due to the positively
charged ions.
• We are looking to find the set of plane waves which satisfy the periodic
boundary conditions
• Hamiltonian Ĥ ψ = ( − 2m2
∇ 2 + V (r )) ψ = E ψ where V (r ) is periodic
• Because V (r ) has the same periodicity as the lattice it can be Fourier
expanded as V (r ) = ∑
V
G G
e iG⋅r
• We define its Fourier components at reciprocal lattice vectors G as

1 1
∫ ∫
− iG⋅r − iG⋅r
VG = e V (r )d 3
r e V (r )d 3
r
Vol. Vol. per cell unit cell
• Since the potential is real VG∗ = V− G

• For G = 0 the Fourier component VG = V0 is the average of the potential
which we set to zero
• Construct eigenstate ψ from plane wave states k = e (which form a
ik ⋅r
complete set of basis vectors obeying periodic boundary conditions) giving

=ψ ∑=
ck k
k
∑k
c
k
e ik ⋅r

2 2
• Apply the Hamiltonian Ĥ =− ∇ + ∑ GVG eiG⋅r to ψ = ∑ ck ei k⋅r
obtaining 2m k
 iG⋅r   ik ⋅r 
∑k k k
E 0
c e ik ⋅r
+ ∑ G
G
V e  ∑ k
 k
c e 

=
E ∑
k
ck e ik ⋅r
• where Ek0 =  2 k 2 / 2m. The potential energy term can be re-written as

V (r )ψ
= ∑Gk
=V c
G ,k
e i ( G +k )⋅r
∑ G k −G
V c
G ,k
e ik ⋅r
since the summation is over all possible values of G , k . We can now write
 0 
∑e
k
ik ⋅r
( Ek − E )ck + ∑VG ck −G  =
 G 
0
ik ⋅r
• Since k = e form an orthogonal set of functions the coefficient of each
term in the sum must vanish and we get the eigenvalue equation:
( Ek0 − E )ck + ∑VG ck −G =

0
G

• From the last slide ( Ek − E )ck + ∑ G VG ck −G =
0
0
• In this equation k can be anywhere in reciprocal space
• The first Brillouin zone contains all of the useful information about k-space
• We write q= k + G ′ relating the general wavevector k to q which lies in the
first Brillouin zone, G ′ is a reciprocal lattice vector
• Hence we can write= Ek0  2 k 2 /=2m  2 (q − G ′) 2 / 2m
• And thus
 2

 2m (q − G ′) − E  cq−G′ + ∑VG cq−G′−G =
2
0
  G
• Finally writing G=′′ G + G ′ we obtain

 2 
 2m (q − G ′) − E  cq−G′ + ∑VG′′−G′cq−G′′ =
2
0
  G′′
• This equation can be used to specify the coefficients ck which are used to
∑= ∑k
construct the wavefunctions
=ψ ck k c e ik  r
k k
Electrons in a periodic potential - Bloch’s Theorem
 2 
• From the last slide  (q − G ′) − E  cq−G′ + ∑VG′′−G′cq−G′′ =
2
0
 2m  G′′
• This equation only involves coefficients ck in which k= q − G with G being

general reciprocal lattice vectors
• If we choose a value of q then the only ck that feature in the equation above
are those of the form cq−G and these specify the wavefunction
• Hence for each distinct value of q there is a wavefunction
ψ q (r ) = ∑ cq−G ei ( q−G )⋅r
G
written ψ q (r ) e=
• Which can be = cq−G e eiq⋅r
∑
G
iG⋅r iq⋅r
u j ,q (r )
− iG⋅r
• Where u j ,q (r ) is built from the periodic function e and must have the
same periodicity as the lattice
• Bloch’s theorem: “Eigenstates of the 1D Hamiltonian can be chosen as a

plane wave multiplied by a function with the periodicity of the Bravais lattice”
Bloch’s theorem
• We now join this all up and apply it to electrons subject to a periodic
potential
• The energy eigenstates of electrons in a lattice:
 pˆ 2 
Ĥψ (r ) =+ V (r ) ψ (r ) =
Eψ (r )
 2m 
where V (r + R ) =V (r ) for all R in a Bravais lattice.
• The ψ k can be chosen so that
(n)
ψ k( n ) (r ) eik⋅r uk( n ) (r )=
where uk( n ) (r + R ) uk( n ) (r )
• Note that potential is periodic and ψ = plane wave x periodic function
• n is the band index – necessary because there may be several distinct
eigenstates of Ĥ with the same symmetry label k
• Or, alternatively if we apply a translation r → r + R then we obtain
ψ k( n ) (r + R ) =
eik⋅Rψ k( n ) (r )
Bloch’s Theorem
• Bloch states (plane wave x periodic function) are similar to eigenstates of
free electrons (just plane waves), but the choice of periodic function gives
additional freedom in labelling states.
ig⋅r
• Remember that e is periodic with same periodicity as the Bravais lattice
(follows from definition of reciprocal lattice vectors g ).
• We can use this to relabel a Bloch state k with a different wavevector k − g
ig⋅r ( m )
by introducing a different periodic function u = e u
(n)
:
ψ k( m ) (r ) e=
= uk (r ) eik⋅r e − ig⋅r =
ik ⋅r ( m )
eig⋅r uk( m( r))  ei ( k −= uk −g (r ) ψ k( n−)g (r )
g )⋅r ( n )
• The ( m),( n) labels are called band indices. In this case, there are two
different states ψ k , ψ k labelled with the same k vector, but
(m) (n)
(
belonging to different bands m, n ψ k (r ) = e ψ k (r )
(m) − ig⋅r ( n )
)
• For every state labelled with a k vector outside the first Brillouin zone, there
is an identical state which can be labelled with a vector q= k − g inside the
first Brillouin zone.
• Corollary: any quantity that depends on the wavefunction, in particular
energy, is periodic in wavevector space.
• Lattice dynamics and phonons – density of states,

• Einstein and Debye models of lattice specific heat capacity
• Comparison with experimental results
• Electrons in a periodic potential,
• Bloch’s theorem

Lecture 6
The end

Lecture 7
David Ritchie




Nearly Free Electron Approximation
• To find approximate single electron states in a periodic potential by
hybridising nearly degenerate plane wave states
• Recall that the single-electron state was obtained from plane wave
=
expansion ψk ∑c
G k −G
k −G
• If the lattice potential is weak compared to kinetic energy the eigenstates are
constructed from k plus a small number of lattice harmonics states k − G
• we use second order perturbation theory to calculate the degree of
admixture of the harmonic states. The energy shift due to admixing k − G
is given by
=
∆Ek VG / ( Ek0 − Ek0−G )
2
• VG is the Fourier component of the lattice potential at reciprocal lattice

vector G given by V = ∑ GVG eiG⋅r
• This energy shift and the associated admixture of lattice harmonics is largest
when the states are nearly degenerate so Ek0  Ek0−G
• To work out the perturbed energy levels we use the equation from above
ck + ∑ GVG ck −G =
 
E −E



0
k
0
k −G



0
• restricting G vectors to those linking nearly degenerate states with Ek ≈ Ek −G
0 0

Nearly Free Electron approximation
• One-dimensional chain
• States k and k − G = k − 2π / a
are nearly degenerate
• Potential VG admixes k − G with k
• ψ =ck k + ck −G k − G
• VG admixes other states, but further
away in energy so less important 2
∑G G
2
2 k  2 k −G iG⋅r
=
• Apply Hamiltonian using Ek0
=
2m , E 0
k −G =,V2m V e
ˆ ψ =
H ( pˆ 2
/ 2m + V ) ( ck k + ck −G k − G ) = E ( ck k + ck −G k − G )
= ck Ek0 k + ckV k + ck −G Ek0−G k − G + ck −GV k − G
k and k − G with k=
• Left multiplying by V k V0 , k V = k − G VG
and k − G V k =VG∗ =
V− G for a real potential we obtain
ck E = ck Ek0 + ckV0 + ck −GVG , ck −G E = ckV− G + ck −GV0 + ck −G Ek0−G
Ek0 − E  ck + ∑ GVG ck −G =
 
• A special case of equation from above: 


0
• From the last slide if we set V0 = 0 and rearrange
=ck ( E − Ek0 ) ck −GVG , ck −G=( k −G ) ckV−G
E − E 0
 E − Ek0 −VG   ck 
⇒   =
0
 −V− G E − Ek −G   ck −G 
0
⇒ (E − E )( E − E ) − V =E − E ( E + E )+ E E − VG =0
0 0 2 2 0 0 0 0 2
k k −G G k k −G k k −G
 0 
( k −G ) ± (E ) (E + Ek0−G ) ± VG
2
E= + −E + 4 VG ≈
0 0 0 2 0
1
2  kE E k k −G
1
2 k
 
• Which gives a symmetrical splitting of levels about
0 0
the free electron value where Ek , Ek −G cross
E
close to the Brillouin zone boundary at k = ±π / a
• At the zone boundary the Bloch states are formed 2VG 2VG
from either the sum of or difference between the
k
two unperturbed states k , k − G
• Both sum and difference wavefunctions are − πa π
a
standing waves, in one case nodes are located
near atomic cores in the other between them.
• Another approach - start from an equation with reciprocal space periodicity
built in:  
2
 
 q−G′ ∑ G′′−G′ q−G′′  =
 

 (q − G ′) 2
− E c + V c  0

 2m
  G′′ 
• q is in the first Brillouin zone and is obtained from k by subtracting G ′if

necessary
• If the potential is zero then =E 2m (q − G ′) 2
2
and we obtain a set of
parabolic bands with energies given by
 2
E0n (k ) = (k − 2π n / a ) 2 , n = ..., −2, −1, 0,1, 2,...
2m
• Introduce a small periodic perturbation V ( x ) = 2V2π / a cos ( 2π x / a )
• Perturbation only important for momenta where we have degenerate states,
for example n = 0,1 at k = π / a and from the equation above
 2 2 
 k −E V2π / a  c 
 2m  c
k
=0
 ∗  2
2π 2   k − 2π / a 
 V2π / a (k − ) − E 
QCMP Lent/Easter 2019  2m a  7.6
• The solution of this equation is
2
 1 2  2π   1   2   2π  
2 2 2 2 2
±
E= (k )  k +  k − 
  ±  k −  k −    + 4V 2
2m 2   a   2  2 m 2 m  a  
2π / a

• And at k = π / a the energy levels are E ± (π / = a ) Eπ0/ a ± | V2π / a |
• By differentiating E ± ( k ) with respect to k and evaluating at k = π / a we
find that ∂E / ∂k = 0 so the constant energy surfaces are perpendicular to
the Bragg plane.
• Assuming an attractive potential so V2π / a < 0 then we find ( α , β constants)
2
ψ ( x)= α 2 sin 2 (π x / a ) , E (π / =
+
a ) Eπ0/ a + | V2π / a |
2
ψ ( x)= β 2 cos 2 (π x / a ) , E (π / =
−
a ) Eπ0/ a − | V2π / a |
• Which implies that for the lower energy state the electron density is highest
at x =... − 2a, − a,0, a, 2a,3a... corresponding to the ionic core “s-type”
• The upper energy state has the highest electron density between ionic cores
at values of x = ... − 32 a, − 12 a, 12 a, 23 a... “p-type”
Nearly Free Electron approximation – real space
Periodic potential V ( x )
V ( x) = 2V2π / a cos ( 2π x / a )
0 x
V2π / a < 0
Ion cores
Square of wavefunctions
2 β2
ψ − ( x)
= β 2 cos 2 (π x / a )
E (π / =
a ) Eπ0 / a − | V2π / a |
low energy “s-type” 0 x
2
α2
ψ + ( x)
= α 2 sin 2 (π x / a )
E (π / =
a ) Eπ0 / a + | V2π / a |
high energy “p-type” 0 x
−2a −a 0 a 2a
Nearly Free Electron
approximation - reciprocal
space
• (a) and (b) show free electron
parabolas
• (c) introduces distortion due to a
Bragg plane splitting the
degeneracy
• (d) the parts of (c) corresponding to
the original free electron parabola
• (e) the effect of the additional
Bragg planes on the free electron
parabola – extended zone scheme
• (f) as (e) but reduced zone scheme
• (g) as (f) repeated zone scheme
From Singleton
NFE approximation: calculations for
3D metals
• Top figure energy vs k x , k y for k z = 0
within first Brillouin zone
• Energy contours circular near bottom of
band but distort as band approaches
Brillouin zone boundary (BZB)
• Near top of band they enclose corners of
Brillouin zone
• Bottom figure equal energy surface in 3D
reciprocal space
• Shape of surface shows energy is
periodic in reciprocal space
• Bloch’s theorem: for every ψ kn+G there
is an identical state ψ km so Ek is
periodic in reciprocal space
• Equal energy surface intersects BZB at
right angles – hybridisation and band
distortion strongest
Tight Binding – Linear combination of atomic orbitals
• Builds up wavefunctions in the solid from wavefunctions of individual atoms
• Diatomic molecule – for identical atoms Hˆ =T + Va + Vb where T is the
kinetic energy and Va ,Vb are the (identical) potentials on the two atoms
• Basis set is two states a , b that satisfy
(T + Va =
)a E0 a , (T + Vb=
)b E0 b
• With E0 as the eigenenergy of the atomic state
• We look for solutions = ψ α a + β b , substitute into Ĥ ψ = E ψ and
premultiply by a , b
• Define E 0 = Hˆ aa = a T + Va + Vb a = Ea + a Vb a - shift of the atomic
energy of atom a by the crystal potential of the other atoms
• Define t = Hˆ ab = a T + Va + Vb b the hopping matrix element coupling
the two atomic states together – the sign depends on the orbital symmetry
e.g for two s-states with an attractive potential Vi < 0 then t < 0 but for two
px-states with atoms aligned along x then t > 0
 E 0 − E t α 
• We obtain simultaneous equations:  ∗    =0
 t E 0 − E   β 
Tight Binding – Linear combination of atomic orbitals
1 
• For t < 0 the eigenstates are: ψ=  a  b  , E= E0 ± | t |
2
• For the lower energy (bonding state) the electron density has a maximum
between the atoms. For the higher energy (antibonding state) the electron
density has a node between the atoms
• We next consider what happens when we bring together more atoms in a
linear chain
• The 1s band of a ring of 20

hydrogen atoms
• The one electron energies
are calculated in the tight-
binding approximation using
nearest neighbour overlap (From Kittel)
integrals

Tight binding – Linear chain
• Generalising this approach to a linear chain of atoms subject to periodic
boundary conditions
• Bloch’s theorem dictates that a wavefunction using one orbital per unit cell
must be of the form ψ = ∑
e ik ⋅R n
n
n φ ( r − R n ) are the position and orbital of atom n
n
• Rn , =
• We check ψ obeys Bloch’s Theorem using Tâψ (= r ) ψ (r + a) which will
map =n φ (r − R n ) onto m = φ (r − R m ) where R n − a = Rm
Tâφ (r − R n ) = φ (r − R n + a) = φ (r − (R n − a)) = φ (r − R m )
• Applying the translation operator to ψ
a ∑n ∑m ∑m
ˆ ψ
T ˆ =
T e ik ⋅R n
n = e ik ⋅R n
m e ik ⋅a
= e ik ⋅R m
m e ik ⋅a
ψ
a
• So ψ obeys Bloch’s Theorem

• We apply the Hamiltonian Ĥ to ψ and left multiply by 0 to find an
equation for the dispersion E ( k )
=
E (k ) 0 Hˆ ψ
= ∑n
e ik ⋅R n
0 Hˆ n
Tight binding – Linear chain
=
• From last slide E (k ) =0 Hˆ ψ ∑n e ik ⋅R n
0 Hˆ n
• Neglect matrix elements between orbitals not next to each other
• Define ‘transfer integrals’ t ∗ = t = n Ĥ n + 1 and E 0 = n Hˆ n
• Hence R n= 0, ±a and we obtain for the energy dispersion relation
E=
k E 0 + 2t cos ( ka )
• Diagram shows eigenvalues confined to a band of energy centred on E 0
where we assume t < 0 - applies to s-orbitals
• The weaker the wavefunction overlap the narrower the band
• Because we apply periodic boundary conditions the allowed values of k are
discrete but close together spaced by ∆k = 2π / L where L = Na
• If N is very large the energies form a continuous band
Note – some
textbooks will
define:
t=
− n Ĥ n + 1

Tight binding – three dimensions
• To extend to three dimensions again use ψ =
sum over all atoms in the 3D solid
n ∑
e ik ⋅R n
n but extend the
• The dispersion E (k ) =
n ∑
e ik ⋅R n
0 Hˆ n now contains contributions from
nearest neighbour interactions in all three directions
• Assume that 0 Hˆ n = 0 if atom n is not a nearest neighbour to atom 0.
• If atom n is a nearest neighbour to atom 0 then 0 Hˆ n = t
• For simple cubic crystal then R n = (± a,0,0); (0, ± a,0); (0,0, ± a )
• Hence summing over all values of R n
E 0 + 2t cos ( k x a ) + cos ( k y a ) + cos ( k z a )  k
Ek =
y
• Energies are
confined to a
band of width 0
12t
• Figures
assume
=
k z 0,=a 1 0 kx
Tight binding - 3D simple cubic crystal
• Three dimensional constant energy surface in reciprocal space

• Calculated using tight binding for a single orbital per atom in a simple cubic
lattice
E0 + 2t cos ( k x a ) + cos ( k y a ) + cos ( k z a ) 

Ek =  
• For FCC and BCC lattice bands calculated by tight binding see problem 2.4

• Nearly free electron approximation,

• Splitting of energy levels due to periodic potential
• Real space potential and wavefunctions,
• Reciprocal space band structure,
• Calculations for 3D metals
• Tight binding approximation – Linear combination of atomic
orbitals
• Linear chain
• Three dimensional cubic crystal

Lecture 7
The end

Lecture 8
David Ritchie


………. Nearly free electron approximation; plane waves and

bandgaps. Tight binding approximation; linear combination of atomic
orbitals, linear chain and three dimensions, two bands.
Pseudopotentials. Band structure of real materials; properties of
metals (aluminium and copper) and semiconductors.
Semi-classical model of electron dynamics in bands; Bloch
oscillations, effective mass, density of states, electrons and holes in
semiconductors


Tight binding – number of orbitals per atom
• To now we have calculated Bloch states with a single orbital for each atom –
this only gives a single energy band
• The number of orbitals in the band that correspond to a non-degenerate
atomic level is 2N for N atoms - explained below
• Values of k within the first Brillouin zone define independent wavefunctions
• The simple cubic Brillouin zone has −π / a < k x < π / a etc and the volume
in k-space is 8π / a
3 3
• The number of orbitals (counting both spin orientations) per unit volume of k-
space is V / 4π 3 where V is the volume of the crystal
( ) (
• Hence the number of orbitals is 8π 3 / a 3 × V / 4π 3 =
3
) 2V / a 3
• Since V is the volume of the crystal and 1/ a the number of atoms per
unit volume the number of atoms N = V / a
3
• Hence the number of orbitals is twice the number of atoms

Tight binding – two bands
• To understand many materials we must extend this method so it produces
several bands – we need to build Bloch states from several orbitals per site.
• We combine two orbitals an , bn to form a hybridised local orbital and
combine these to form a Bloch state - coefficients α k , β k can depend on k
=ψ ∑ ik ⋅R n
α
k an + β k bn
 
e 




n
• Inserting ψ into Schrodinger equation Ĥ ψ = E ψ left multiplying by
basis states a0 , b0 and turning into eigenvector problem we obtain
 Ea (k ) − E Vk  α k 
 V∗   =0
 k Eb (k ) − E   β k 
• Ea = ∑ e n a0 Hˆ an is the dispersion of a band formed exclusively
ik ⋅R
n
from atomic orbitals an
• Eb = ∑ e n b0 Hˆ bn is dispersion of a band formed from bn orbitals
ik ⋅R
• The off-diagional matrix elements: Vk = ∑ eik⋅Rn a0 Hˆ bn and with

n
R n
some relabelling
Vk∗ = ∑ R eik⋅Rn b0 Hˆ an
n

Tight binding – two bands
• The eigenvalue problem gives rise to two possible energies at each
wavevector – forming two bands from two orbitals
• The effect of the off-diagonal elements is to hybridise the two bands (formed
from each set of atomic orbitals) where they become nearly degenerate.
• We could have approached this problem by forming two Bloch states, one
from each set of atomic orbitals
ψa ∑n
= e ik . R n
an , ψb ∑n e ik . R n
bn
• Then combining the Bloch states using = ψ α k ψ a + βk ψ b we obtain
the same form as above for ψ
• So the two routes are the same – (1) bandstructure arising from hopping
between hybridised molecular orbitals or (2) hybridisation between bands
arising from atomic orbitals
• To extend to multiple orbitals per unit cell we generalise the summation:
ψ = ∑ n,ν eik⋅R cνk nν
n
ν
• Where ν labels the different orbitals n which exist in the n unit cell and
th
ν
.ck is the associated coefficient determining the level of mixing of the orbital
• The number of bands is equal to the number of local orbitals per unit cell
Tight Binding versus Nearly Free Electron approximation
• Tight Binding (TB) and nearly free electron (NFE) approx. do similar things:
• Both construct a Bloch state from a reduced set of basis functions
• Coefficients found by solving eigenvector/eigenvalue equation
• NFE basis set is selected from plane wave states k − G for which matrix
elements VG are large and which are nearly degenerate with k
• For smooth potential we disregard high G Fourier components - set of basis
functions is small. Computers can use large plane wave basis sets.
• In the NFE approximation kinetic energy appears on diagonal of energy
matrix and potential on off-diagonal terms – very efficient if potential is weak.
• TB allows reasonable answers with relatively little computation
• Atomic orbital states associated with different unit cells of crystal form a
complete basis set in which Bloch states can be expanded
• Number of bands equals the number of atomic orbitals used per unit cell
• If the hopping matrix elements are less than separation between bands then
bands do not cross – strength of the potential is larger than kinetic energy
• In TB potential energy appears on diagonal. Hopping elements – equivalent
of kinetic energy, form off-diagonal terms
Pseudopotentials
• The NFE and TB methods are not accurate predictors of band structure
• Commonly used to form simple models with parameters determined by
experiment or more complex calculations
• Band gaps in semiconductors often small – only a few Fourier components
of potential required because effective scattering potential for valence
electrons is much smaller than full atomic potential
• Effective potential called pseudopotential which reproduces valence states
as lowest eigenstates of problem – we can forget about core states.
• The true potential V ( r ) has a wavefunction
for valence electrons Φ ( r ) that oscillates
near the core
• Pseudopotential Vs ( r ) has a wavefunction
Φ
. s ( r ) that is smooth near the core
• Pseudo potential wavefunction approximates
wavefunction far from core
• Very successful empirical pseudopotential
method - band structure calculated using a
few VG obtained from fits to measurement of
optical reflectance and absorption
Band structures of real materials
• Band gaps arise because of interference between forward and backward
going degenerate plane waves mixing to form standing waves
• Band gap in 3D arises from splitting of degeneracy due to scattering from
Fourier component of lattice potential so we require E0=
(k ) E0 (k − G )
for a given G we find the k so
|k | =
| k − G | ⇒ k ⋅G / 2 =
2 2 2
G/2
• Satisfied by any k lying in a plane perpendicular to and bisecting G
• This is the boundary of a Brillouin zone and the Bragg scattering condition
• Energy eigenstates form discrete bands En (k ) which are continuous
functions of momentum k and labelled by a band index n
• Bandstructure is periodic in reciprocal lattice En= (k ) En (k − G )
• Eigenstates are in form given by Bloch’s theorem ψ nk (r ) = e unk (r )
ik ⋅r
which continues to hold if k → k + G so k can always be chosen in the

first Brillouin zone
• k is ‘crystal momentum’, enters conservation laws for scattering processes
e.g., if an electron absorbs the momentum of a phonon of wavevector q, the
final state will have a Bloch wavevector k ′ = k + q + G where G is the
reciprocal lattice vector necessary to keep k ′ inside the first Brillouin zone
Band structures of real materials
• The spacing between k-points in 1D is 2π / L where L is the linear

dimension of the crystal. Generalising to 3D, the volume associated with
each k-value is: ( ∆k ) = (2π )3 / V with V the volume of the crystal.
3
• Within each Brillouin zone the number of k-states allowed by periodic

boundary conditions equals the number of unit cells in the crystal.
• This is so large bands are continuous and we use the density of states
( )
• Electrons are fermions so each k -point is occupied by two ↑, ↓ electrons
• In a system with one electron per unit cell (e.g. lattice of sodium atoms) half
the states will be filled in the first Brillouin zone – hence a metal
• With a system with two electrons per unit cell the first Brillouin zone will be
filled meaning that it may be an insulator if the band gap is large enough, but
if the Fermi energy lies within a band the material will be a metal.
• If bands overlap so there are 2 or more partially filled bands we have a metal
• When the energy between electrons due to Coulomb repulsion is larger than
the bandwidth then materials can be ‘Mott insulators’. They can undergo a
metal to insulator transition if electron density or temperature changes.

Band structures of real materials - notation
• The bandstructure En (k ) defines a 3D function which is difficult to visualise
• By convention, cuts through this function are plotted along particular
directions in k-space
• A shorthand is used to label
X
points in k-space
• Diagram shows 1st Brillouin
zone for several crystal
lattices with labelled points X
K
• The zone centre Γ =(0,0,0)

• Example: for FCC the X point Face centred cubic Body centred cubic
2π
.a (1,0,0) etc is at the zone
edge in 6 (100) directions
• L is on the zone boundary at
K
.πa (1,1,1) in 8 (111) directions
• K is at zone edge in (110)
direction
• W is at square/hexagon Simple cubic Hexagonal
intersection
QCMP Lent/Easter 2019 https://wiki.fysik.dtu.dk/gpaw/tutorials/bandstructures/bandstructures.html 8.10
Examples of band structures - Metals
• With an odd number of electrons per primitive unit cell, chemical potential
must lie within a band, hence no energy gap. Because low-energy electronic
excitations are possible, the system is a metal.
• In a simple metal such as Na (3s1 with 1 valence electron) or Al (3s23p1 with
3 valence electrons) Fermi surface is close to a free-electron sphere
• In other cases (e.g. Cu 3d104s1) the sphere Fermi
extends in some directions to meet the Brillouin surface of
zone boundary surface. There can be situations copper
where several bands are cut by the Fermi
energy, and the topology of Fermi surfaces is
sometimes complicated – studied by de Hass -
Van Alphen effect.
• Even with the right number of electrons to fill bands and make a
semiconductor, the bands may still overlap. Consequently, the fermi surface
will intersect more than one band, making a pocket of electrons in one band
and removing a pocket of electrons from the band below (making holes).
This accounts for the metallicity of Ca and Mg and (which have two
electrons per unit cell), and also As, Sb, Bi,. These are known as semimetals
Aluminium band structure
2 1
• Has three valence electrons in configuration 3s 3 p
• FCC crystal structure hence BCC reciprocal lattice – shown below
• First Brillouin zone full, valence electrons spread into 2nd,3rd and 4th
• Band structure close to free electron parabola except when near Brillouin
zone boundaries, bands fill up to Fermi level
• No clear band gap in all the directions - hence a metal
Energy (eV)
Brillouin zone of FCC lattice
QCMP Lent/Easter 2019 Wavevector k 8.12

Aluminium band structure
• Transitions from filled states below EF to empty states above EF
• Fermi’s golden rule states density of states determines
transition rates - depends on both initial and final density of states
• High density of states for parallel bands - Hence reflectivity dip at 1.5eV
• Band structure suggests transitions possible over range of energies > 1.5eV
not parallel bands so less significant but does reduce reflectivity
Transition energy ~ 1.5eV
From Lecture 1
Reflectivity of Aluminium
Theory
1.0
0.8
Energy (eV)
Reflectivitiy
ω p = 15.8eV
0.6 Experimental data

0.4
0.2
0.0
0 5 10 15 20
Energy (eV)
Data from Ehrenreich et al (1963)
QCMP Lent/Easter 2019 Wavevector k B Segall Phys Rev 124, 1797 (1961) 8.13
Copper band structure
Electron configuration [Ar]3d104s1
• 3d tightly bound, narrow band, high DoS, full – 10 electrons

• 4s free electron like, lower DoS, broader band - 1 electron
• Fermi surface in the 4s band and above 3d band
• Transitions possible - filled 3d to 4s states above EF
• Well defined threshold for transitions about 2eV
• Reflectivity above 2eV reduced due to interband absorption
well below plasma frequency at 10.8eV
• Reflectivity falls in visible part of spectrum – hence colour
QCMP Lent/Easter 2019 From Optical properties of solids Mark Fox 8.14
Semiconductors
• If there is an even number of
electrons per unit cell, it is possible GaAs
(with no band overlap) for all of the
occupied states to fill bands, with an
energy gap to the empty states.
• The system will be a semiconductor
or insulator – the case for the group
IV elements, C, Si, Ge as well as III-
V compounds such as GaAs, InP.
• These semiconductors have 2 atoms
per unit cell (diamond or zincblende
structure) and 8 valence electrons Si
per unit cell — 4 filled bands.
• Maximum in valance band for both
Si and GaAs is at Γ .
• Minimum in conduction band is also
at Γ for GaAs (direct bandgap) but is
at X for Si (indirect bandgap).
Bandgap shown in blue
• Tight binding - number of orbitals,

• Tight binding - two bands,
• Tight binding versus nearly free electron approximations
• Pseudopotentials
• Bandstructure of real materials
• Metals, semi-metals
• Example - Aluminium
• Example - Copper
• Example – Si and GaAs Semiconductors

Lecture 8
The end

Lecture 9
David Ritchie


………. Nearly free electron approximation; plane waves and

bandgaps. Tight binding approximation; linear combination of atomic
orbitals, linear chain and three dimensions, two bands.
Pseudopotentials. Band structure of real materials; properties of
metals (aluminium and copper) and semiconductors.
Semi-classical model of electron dynamics in bands; Bloch
oscillations, effective mass, density of states, electrons and holes in
semiconductors


Semiclassical model of electron dynamics
• We now discuss the dynamics of electrons in energy bands
• The bandstructure is dispersive so we should treat particles as wave-
packets, a superposition of different k
• The band energy  (k ) is the frequency associated with the phase rotation of
the wavefunction ψ k ∝ e
− i ( k ) t / 
• For the motion of a wave in a dispersive band we should use the group
−1
velocity vg = dω / dk which as a vector is v g= r=  ∇ k  (k )
• Note that the effects of the lattice potential are contained in  (k )
• If a force is applied to a particle, the rate of doing work on that particle is
given by dk dk dk
= ⋅∇k  (k )= F ⋅ v ⇒ F= 
dt dt dt
• And introducing electric and magnetic fields
F= dk / dt = −e(E +  −1∇k  (k ) × B)
− e(E + v × B ) =
• An electric field shifts the crystal momentum k in the direction of the field
• A magnetic field causes an electron to move in k-space in a plane
perpendicular to the field on a path of constant energy
• This is the basis of techniques to measure the Fermi surface of metals

Semiclassical model of electron dynamics
• The diagram shows the energy,  group velocity
∗
. d / dk and effective mass m =  / (d  / dk )
−1 2 2 2
of an electron for a 1D band

• Applying an electric field k=(t ) k (0) − eEt /  
.k increases steadily with time throughout, initially
kx
its velocity increases as for a free electron. −
π Γ π
As the electron moves up the band d 2 / dk 2

a a
•
decreases and the effective mass increases until vg =
1 d
 dk
the acceleration stops and the velocity is constant.
As d  / dk changes sign the effective mass
2 2
• Γ
kx
π
becomes negative and the velocity starts to fall π
− a
a
until it reaches zero at the zone boundary,
becoming negative as it passes to the next zone
• As k increases further moving up and down the
band structure, the velocity oscillates with time and
Γ
the electron’s position oscillates with time −1
kx
d  2
m∗ =  2  2 
• So applying a DC voltage results in an AC current  dk 
π π
• No net current flows and we have an insulator! −
a a

Bloch oscillations
• So far all attempts to observe Bloch oscillations in a crystal have failed.
• This is due to scattering off impurities and phonons in the solid as the
momentum approaches π / a , however it is possible with artificial structures
• Using alternating thin layers of GaAs (9.7nm) and Al0.3Ga0.7As (1.7nm)
repeated 35 times, an artificial periodic potential is created with a periodicity
40 times longer than the atomic spacing, known as a superlattice.
• The momentum at the zone boundary for this superlattice is 40 times lower
than in a crystal and the wavepacket does not need such large velocities.
• The diagram shows energy versus position of the conduction and valence
bands of the superlattice
• The tilting is produced by the applied
electric field. The levels shown form a
Wannier-Stark ladder for electron
wavepackets made by excitation from
the valence band in one quantum well, 
either vertically ( n = 0) or to neigh- GaAs
bouring ( n = ±1) or next-neighbouring
r
(.n = ±2) wells of the electron lattice. AlGaAs
C Waschke at al Phys Rev Lett 70, 3319 (1993)

Bloch Oscillation measurements
• Experiment used terahertz time

domain spectroscopy
• A technique only sensitive to
coherent emission processes
• Experimental setup to observe the
dipole radiation from oscillating
charge is shown in the figure
• An unfocused optical femtosecond
beam from a Ti-sapphire laser with
a pulse duration of 100 fs , strikes H G Roskos et al Phys Rev Lett 68, 2216 (1992)
the sample at an angle of 45° C Waschke at al Phys Rev Lett 70, 3319 (1993)
• The sample is held in a cryostat at 10 K.

• Terahertz radiation, emitted collinearly with the reflected optical excitation
beam leaves cryostat and is collected with a pair of off-axis paraboloid
mirrors and detected by a photoconductive dipole antenna
• This photoconductive dipole antenna is gated by a second, time-delayed
portion of the femtosecond laser beam to measure electric field emitted
from sample as a function of time on ps timescale.
Bloch Oscillation measurements
• In the experiment, electron and

hole pairs are excited optically by a
100
. fs pulsed laser with energy just
above the band gap of GaAs
• The coherent terahertz radiation is
measured as a function of time (left
panel) for different DC electrical
biases, more oscillations are seen
as the bias becomes more negative
• The spectral content is determined
by taking a Fourier transform of the
results (right panel)
• The frequency of the Bloch
oscillation peak increases as the
bias becomes more negative than
−2.4V

Waschke at al Phys Rev Lett 70, 3319 (1993) 9.7
Density of states – energy bands
• Earlier in the course we discussed the density of states of a free electron

gas.
• The maxima Emax and minima Emin of all bands have a locally quadratic
dispersion with respect to momentum measured from the maxima or minima
• We can define the effective masses with α referring to Cartesian
−1
coordinates ∗ 2  2 
mα =
 ∂ E (k ) / ∂kα2
 k min 
• The 3D density of states near the minimum is given by
3/2
V m  ∗
g (=
E ≥ Emin )   E − Emin
2π 2   2 
∗ ∗ ∗ ∗ 1/ 3
• The bare mass has been replaced by an effective mass m = ( mx m y mz )
which averages the curvature of the bands in 3D
• A similar form applies near the band maxima with g ( E ) ∝ Emax − E
• The flatter the band the higher the effective mass and the larger the density
of states

Density of states
• For any form of E (k ) the density of states is given by
dk
=
g (E) ∑
=
n
gn (E ) ∑n ∫ 4π 3 δ ( E − En (k ))
• Because of the δ function the momentum integral is over a surface in k-
space S nwhich depends on the energy E , e.g. S n ( EF ) is the Fermi surface
• We can separate the integral into a 2D surface integral along a contour of
constant energy and an integral perpendicular to this surface dk⊥ , hence
dS
gn (E ) ∫
Sn ( E )
4π 3 ∫
δ ( E − En (k ))dk⊥
dS 1
=∫
Sn (E) 4π 3 | ∇ E (k ) |
⊥ n
δ ( x − x0 ) / f ′( x0 )
given δ ( f ( x) − f ( x0 )) =
• ∇ ⊥ En (k ) is the derivative of the energy in the normal direction which
becomes zero at the edges of the bands and the saddle points which can
exist in 2D and 3D bands, hence the cusps in the DoS
Density of states
• Maxima, minima, and saddle points are all
generically described by dispersion
(measured relative to the stationary point) of
2 2 2 2 2 2
E (k ) = E0 ± kx ± ky ± kz
2mx 2m y 2mz
• If all the signs are positive, this is a band
minimum; if all negative, a band maximum;
when the signs are mixed there is a saddle
point.
• In the vicinity of each of these critical points,
(van Hove singularities) the density of states
or its derivative is singular. In two dimensions,
a saddle point gives rise to a logarithmically
singular density of states, whereas in three
dimensions there is a discontinuity in the
derivative.
• Examples of the generic behaviour of the density of states in 1D,2D and 3D
are shown in the figure
Electrons and holes in semiconductors
• Filled bands are inert, if all the states in a Brillouin zone are occupied, total
current is obtained by integrating the group velocity over the whole zone
−1
• The group velocity is vg =  d  / dk and  ( k ) is a periodic function so the
integral yields zero and there is no net current
• All insulators have even valence or a lattice containing an even number of
atoms in the basis hence filled bands
• Consider what happens if we move an 
electron from the valence band to the
conduction band by absorption of a photon
• This creates an electron hole-pair
• For typical values of energy the photon
momentum is very small compared to the
electron momentum and it adds negligible
momentum to the system
• Hence the hole momentum is the negative of
the momentum of the empty electron state
so k h = −k e

• Hole energy. If the zero of energy is at the top of the band. The lower in the
band the missing electron lies the higher the energy of the system. The
energy of the hole is opposite in sign to the energy of the missing electron
because it takes more work to remove an electron from a low orbital than a
high orbital. h (k h ) = −e (k e )
• Hole velocity. if we combine these two rules together then since there are
two sign changes v h =  −1∇k h h (k h ) =  −1∇k e e (k e ) = v e and the velocity
of the hole equals the velocity of the electron
•
∗
Effective mass. The effective mass m =  / ∂  / ∂k
∗
2
∗
( 2
)
2
and substituting
in the equations for ,k means that mh = − me so the electron mass is
negative and the hole mass is positive at the top of the valence band
• Hole charge. We take the equation of motion for the electron
dk e
 = − e(E + v e × B )
dt dk h
• Make the replacements ke → − kh , ve → vh hence  = e(E + v h × B )
and the effective charge of a hole is positive, dt

From Kittel
• (a) At t=0 all states are filled except F at the top of the band the velocity ve is
=
zero at F because ve = −1
dω / dk 0
• (b) An electric field Ex is applied in the + x direction, force on the electrons is
in the − k x direction and all electrons make transitions together in the − k x
direction moving the hole to state E. A current flows
• (c) After more time the electrons move again in k-space, the hole is now at D
• Motion of electrons in the conduction band and E

holes in the valence band in an electric field E je
ve e
• The drift velocities are in opposite directions
jh
but currents are in the direction of E h vh
• Semi-classical model of electron dynamics

• Bloch oscillations – experiment using semiconductor superlattice
• Density of states in energy bands
• Comparison of electrons and holes in semiconductors: wavevector,
energy, velocity, mass, charge

Lecture 9
The end

Lecture 10
GaAs
David Ritchie

Photon absorption; transition rates, experimental arrangement for

absorption spectroscopy, direct and indirect semiconductors, excitons.
Quantum oscillations; de Haas-Van Alphen effect in copper and
strontium ruthenate. Photoemission; angle resolved photoemission
spectroscopy (ARPES) in GaAs and strontium ruthenate. Tunnelling;
scanning tunnelling microscopy. Cyclotron resonance. Scattering in
metals; Wiedemann-Franz law, theory of electrical and thermal
transport, Matthiessen’s rule, emission and absorption of phonons.
Experiments demonstrating electron-phonon and electron–electron
scattering at low temperatures.

Experimental probes of the band structure
• One way to investigate the band structure is the excitation of an electron by
a photon from an occupied state to an empty state in the conduction band
leaving behind a hole in the valence band and creating an electron-hole pair
• Photons cause nearly vertical transitions - the wavevector of a photon with
energy close to the band gap is much smaller than is possible for an electron
• In a direct band gap semiconductor (GaAs) the lowest energy available
states for electron and hole are at the same wavevector, the optical
threshold is at the vertical transition
• For indirect semiconductors
(Si, Ge) the valence band
maximum is at a different
wavevector to the
conduction band minimum
• For excitation at the
minimum energy a phonon
must be excited as the
photon is absorbed, 2nd =
For phonons and photons
5
v ω= / k E / 2π k the
order process, much less factor of around 10 between the two velocities is the
difference between vertical & horizontal transitions
likely than a direct transition
Transition rates for photon absorption
• The initial and final electron states are related by  f= i + ω
• optical absorption will have a threshold where  f − i = Eg above which
there is a continuous range of possible transitions determined by the upper
and lower limits of the bands
• The optical absorption coefficient is determined by the quantum mechanical
transition rate Wi→ f for exciting an electron in an initial quantum stateψ i to a
final stateψ f by absorption of a photon of frequency ω
Transition rate is given by Fermi’s golden rule Wi − f =  M g ( ω ) which
2π 2
•
depends on the dipole matrix element M and the density of states g ( ω )
• The matrix element is calculated using the initial and final states and the
perturbation due to the interaction of the electric field of the photon E0 eik⋅r
and the electric dipole moment p e = −er
M =ψ f Hˆ ′ ψ i , Hˆ ′ =−p e ⋅ E0 eik⋅r
• The electron states are described by Bloch functions which are a plane
wave with wavevectors k i , k f multiplied by functions ui , u f with the
periodicity of the lattice ik ⋅r
=ψ i (r ) =
1
V
ui
ik ⋅r
e i
, ψ f (r ) 1
V
uf e f

• We calculate the matrix element by integrating over the whole crystal
e ∗
∫
− ik f ⋅r ik ⋅r ik i ⋅r 3
M u f (r ) e ( E 0 ⋅ re )ui (r ) e dr
V
• This can be simplified using conservation of momentum k f − k i = k
so the phase factor is zero – if this is not so, the integral will sum to zero
• ui , u f are functions with the periodicity of the lattice so we can separate the
integral over the whole crystal into a sum over identical unit cells – which are
in phase
• Hence M ∝∫ ui∗ (r ) x u f (r )d 3r where we assume the light is polarised
unit cell
along the x-axis
• To evaluate this matrix element we need to know the form of ui , u f which
are derived from the atomic orbitals of the constituent atoms, so vary from
one material to another.
is k 2π / λ ≈ 10 m but the wavevector of
−1
• The wavevector of the photon= 7
the electrons are much larger – less than π / a ≈ 10 m so we can neglect

10 −1
the photon momentum and k f = k i hence photon absorption is vertical on

the E − k diagram
• We need to consider the atomic states involved in the transition and how
they affect the dipole matrix element
• The semiconductors we consider all have 4 valence electrons
• Obvious for Si, Ge (group VI), for GaAs 3 electrons from Ga (Group III) and
5 electrons from As (group V) are shared forming zincblende crystal
• The valence electron configuration of the 4 electron atom (Ge) is 4s24p2
• Diagram shows evolution of s- and p- atomic states to s- and p-bonding and
antibonding molecular states to the valence and conduction bands of the
crystal. The level ordering is correct for GaAs and Ge but different for Si
• Diagram suggests that the top of
the valence band is derived from
p-bonding orbitals while the
bottom of the conduction band is
from s-antibonding orbitals
• Hence optical transitions between
the valence band and the
conduction band are electric-
dipole allowed
M Fox Optical properties of solids
GaAs band structure
• GaAs band structure shown in diagram for
Empty states
2 directions in reciprocal space (100), (111)
• GaAs is a direct bandgap semiconductor EF
• 3 valence bands - occupied states
corresponding to 3 p-bonding orbitals Filled states
• Single empty conduction band corresponds
to s anti-bonding state
• Strictly valid only at zone centre Γ (111) (100)
• Atomic character only well defined at high

symmetry points - changes away from Γ, Χ, L
• Lower diagram shows simplified structure
for small k with parabolic dispersion
• One electron band, three hole bands
• Bandgap given by Eg, split-off hole band
lower in energy by ∆
• Heavy hole and light hole transitions to
conduction band shown
QCMP Lent/Easter 2019 M Fox Optical properties of solids 10.7

• The factor of g (ω ) that appears in Fermi’s golden rule is the joint density of
states due to the fact that both initial and final states lie in continuous bands
• E-k relations for conduction, heavy-hole, light-hole, split-off hole bands
2k 2 2k 2 2k 2 2k 2
c (k ) = Eg + ∗
, hh (k ) = − ∗ , lh (k ) = − ∗ , so (k ) = −∆ − ∗
2me 2mhh 2mlh 2mso
• Conservation of energy during a heavy-hole or light hole transition requires
2k 2 2k 2
ω = Eg + ∗
+ ∗
where mh
∗
is the heavy hole or light hole mass
2me 2mh
1 1 1
• We define reduced electron-hole mass by = + hence we can
write ω
= Eg +  2 k 2 / 2 µ µ me mh∗ ∗
• Using formula for density of states per unit volume we find

ω < Eg , g (ω ) =
0
 1  2 µ 
3/2
ω > Eg , = g (ω )  2  2  ω − Eg
 2π   
• So the density of states factor rises as ω − Eg for energies greater than
the bandgap
Experimental arrangements for optical absorption
spectroscopy
M Fox Optical properties of solids
• Beer’s law: δ I = −αδ z × I ( z ) ⇒ I ( z ) =

I 0 exp ( −α z )
• α defined as fraction of intensity I (power per unit area) absorbed per unit length
Photon absorption – direct semiconductors
• Having discussed the matrix element and DoS contributions to Fermi’s
golden rule, Wi − f =  M g ( ω ) , we can deduce frequency dependence
2 π 2
of the absorption coefficient α (fraction of power absorbed per unit length)

• We expect that for ω < Eg there will be no absorption and α = 0
• For ω > Eg we expect that α ∝ ω − Eg with the absorption increasing
above the bandgap
Also since g ( ω ) ∝ µ we expect transitions with larger reduced masses
3/2
•
give rise to stronger absorption
• These results can be compared to
experimental data
• The diagram shows the square of
the absorption coefficient plotted
versus photon energy for III-V direct
band-gap semiconductor InAs
• The linear dependence confirms
α
. ∝ ω − Eg , it gives a bandgap
2
of 0.35eV in good agreement with

electrical measurements
QCMP Lent/Easter 2019 E D Palik (1985) Handbook of the optical constants of solids 10.10
Photon absorption
• In many III-V semiconductors including GaAs the frequency dependence of
absorption α ∝ ω − Eg is only approximately obeyed for a number of
reasons
• We neglect Coulomb attraction between the electrons and holes which can
enhance absorption and lead to formation of a bound pair – an exciton
• These effects get larger as bandgap increases and temperature is lowered –
hence previous example is InAs narrow BG semiconductor (0.35eV) at 298K
• The semiconductor crystal may include
impurities and defects with energies GaAs
within the bandgap – hence additional
absorption below the bandgap energy
• The parabolic band approximation is
only valid near k = 0 - as the photon
energy increases above the bandgap
the joint DoS is no longer ∝ ω − Eg
• In this case we need to use the full
band structure to evaluate the DoS
MD Sturge Phys Rev 127, 768 (1962)

Photon absorption – indirect semiconductors
• Important semiconductors e.g. Si and Ge have their conduction band
minimum away from Brillouin zone centre and the valence band maximum
• A transition between band edges needs a big change in electron wavevector
• Transition must involve a phonon to conserve momentum
• Consider an indirect transition exciting a valence band electron (i , k1 ) to a
state ( f , k2 ) in the conduction band. Photon energy ω , phonon energy Ω
• Conservation of energy  f = i + ω ± Ω and momentum k =f k i ± q
• The ± allows for the possibility of phonon absorption (+) or emission(-)
• Have neglected photon’s momentum Silicon Eg GaAs Eg
which is small on these scales
• Indirect transitions which involve both
photons and phonons are second
order process which are much less
likely than the absorption of a photon
be a direct and semiconductor
• Hence absorption rate for indirect
semiconductors much smaller above
band edge –for Si/GaAs one tenth
• Derivation of QM transition rate for indirect band gap semiconductor gives
for absorption coefficient α ( ω ) ∝ ( ω − Eg  Ω) - see Yu and Cardona
2
• Frequency dependence different to that for direct bandgap semiconductors –

can be used to determine if bandgap is direct or not .
• Threshold close to bandgap, depends if the phonon is absorbed or emitted
• Band structure for Ge similar to GaAs but
indirect with band gap of 0.66eV - less than
0.8eV direct gap at 298K
• Figure (a) shows indirect absorption in Ge
near bandgap with α ∝ ω Ge band
structure
• Expect contributions from phonon emission
and absorption – latter only possible at high
temperatures where phonons are excited (111) (100)
• (b) shows band edge absorption

above 0.8eV with α ∝ ω − Eg
• Direct absorption much larger
than indirect absorption
(a) GG McFarlane and V Roberts Phys Rev 97, 1714-6
(1955)
(b) WC Dash and R Newman Phys Rev 99, 1151 (1955)
• Figure shows interband absorption of Si to
relatively high energy - E =ω ≈ 10eV
Very large absorption α ≈ 10 m above
8 −1
•
3eV with two peaks E1 , E2
• Compares to indirect region just above
bandgap with α ≈ 10 -10 m
2 6 −1
• Band structure of silicon shown, indirect

bandgap value - Eg  1.1eV
• Minimum direct separation between
conduction and valence bands near L
point 3.5eV (E1 )
• Second absorption near Χ 4.2eV (E2 )
• In both regions conduction and valence
bands parallel and dE / dk very small
Silicon band
• Hence very high joint density of states and structure
very high direct transition rate and
absorption
(111) (100)
Excitons
• Peak in GaAs absorption spectrum at low
GaAs
temperatures signifies presence of excitons
• An exciton is a bound electron-hole pair
analogous to the hydrogen atom
• In an exciton electron and heavy hole
masses are combined in a reduced
effective mass 1/ =µ ∗ 1/ me∗ + 1/ mhh∗
• GaAs has permittivity ε r = 12.8

• We can use modified equation for the T=1.2K
energy levels of a hydrogen atom to
GW Fehrenbach et al J Lumin. 30, 154-61 (1985)
calculate energy levels of excitons
• Figure shows absorption by three
µ ∗ 13.6eV Rx
En = − ⋅ = − 2 exciton energy levels below the
meε r
2
n 2
n conduction band in pure GaAs
• n is the energy level quantum number • Using µ ∗ = 0.05me we obtain
• Energy of exciton is equal to energy . x = 4.2meV in good agreement
R
required to create an electron-hole pair with the results
minus the binding energy or • Quantum confined structures
=
n E g − Rx / n 2
greatly enhance exciton effects
• Transition rates for photon absorption, Fermi’s golden rule,

matrix elements, density of states
• GaAs band structure
• Photon absorption in direct semiconductors, InAs, GaAs
• Photon absorption in indirect semiconductors, contribution of
phonons in Ge, above bandgap absorption in Si.
• Excitons

Lecture 10
GaAs
David Ritchie

Lecture 11
David Ritchie



Quantum oscillations
• For pure samples, many material properties have been found to oscillate as
a function of applied magnetic field
• The form of these quantum oscillations can be used to infer the shape of the
Fermi surface and other key electronic properties
• A full quantum mechanical treatment of the motion of electrons in a strong
magnetic field is problematic
• When the lattice potential can be neglected - for free electrons the
Schrodinger equation can be solved directly
• For real materials the lattice potential is essential to the band structure and
cannot be neglected
• We use a semi-classical treatment using the Bohr-Sommerfeld quantisation
condition

• The canonical momentum for a particle in a magnetic field, p , (conjugate to
the position r ) is the sum of the kinetic momentum mv = k and the field
momentum qA so = p k + qA
• We assume that orbits in a magnetic field are quantized according to the
∫ ( )
Bohr-Sommerfeld relation p ⋅ dr = n + 12 2π  where n is an integer
• Particles with charge q moving in a strong magnetic field orbit along a path
determined by the Lorentz force m= v qr × B .
• This relation connects the components of velocity and acceleration of the
particle at r in the plane perpendicular to B and can be integrated to give
.mv=⊥ qr × B where r is measured from the centre of the orbit
• ∫ ∫
Hence p = mv + qA = q (r × B + A ) ⇒ p ⋅ dr = q r × B ⋅ dr + q A ⋅ dr
integrating around a loop
∫
• ∫ ∫
Now A ⋅ dr =∇ × A ⋅ dS =B ⋅ dS =
S ∫S
Φ by Stokes’ theorem, and
∫ r × B ⋅ dr =−B ⋅ ∫ r × dr =−2BA r =−2Φ since ∫ r × dr equals twice
the area Ar enclosed by the loop
• Putting this together: ∫ p ⋅ d r = − 2 q Φ + q Φ = − q Φ = ( n + 2 ) 2π 

1

• From the last slide p ⋅ dr =− qΦ = ( n + 12 ) 2π 
∫
• We arrive at the conclusion that the flux threading the real space orbit for an
electron is quantised Φ n = Arn B = ( n + 12 ) 2π  / e
• Can we relate motion of the electron in real space to motion in k-space?
• From earlier mv= ⊥ k =
⊥ qr × B so with constant B , k is perpendicular
to r and its magnitude is a factor Bq /  different from r .
• So the k-space orbit has the same shape as the real space orbit, but is
turned by 90 degrees and stretched by Bq / 
• This means that the area enclosed by the k-space orbit Ak is
Ak = ( e /  ) B 2 Ar
2
• Where q has been replaced by the electron charge e

• Combining this result with A=
n
r ( n + 2 ) 2π  / Be from above
1
• We obtain
2π e
=Ak B ( n + 12 )

• Which is an expression for the area of an orbit in k-space as a function of
integer n and suggests that the area is quantised
Density of states oscillations
• In a magnetic field, the allowed k-
states no longer form a regular lattice
as k is not a good quantum number
• All the k-states in the vicinity of a k-
orbit superimpose to form the orbital
motion of the electrons
• Electrons now live in a set of cylinders “Landau Fermi surface
=
tubes” with area Ak 2π e B ( n + 12 ) which cut
through the zero field Fermi surface
• What is the B field dependence of the DoS g ( EF )?
• Consider a slice ⊥ B through Fermi surface area
• This will only contribute to g ( EF ) if its area
coincides with the area of a Landau tube Ak
• As B increases one tube after another will satisfy this condition at field
values of=Bn−1 2πAke (n + 12 )
• Consequently the contribution of this slice to g ( EF ) oscillates with a period
given by the Onsager relation: 1 1 1 2π e 1
∆  = − =
QCMP Lent/Easter 2019  B  Bn+1 Bn  Ak 11.6
Density of states oscillations
• Energy of the band electrons is Kittel Singleton
completely quantised into
ladder of Landau levels in the
plane perpendicular to B
• Motion parallel to B is
unconstrained
• The DoS is an infinite ladder of
Landau levels each with a 1D
density of states function
superimposed
• As each of the sharp peaks in
the DoS moves through the
chemical potential there is a
modulation of the density of
states and chemical potential
• This affects a number of
different properties of the
material
Fermi surface orbits
• We can only measure quantum
oscillations associated with extremal
orbits – a Landau tube touches rather
than cuts through the Fermi surface
• In these regions are many close lying
orbits with nearly identical cross-
section causing DoS to add coherently
Singleton
• For the rest of the Fermi surface the
oscillations attributed to each orbit
have different periods and add
incoherently wiping out the effect
• Several different frequencies may be
superimposed corresponding to
different possible extremal orbits –
e.g. neck and belly orbits in Cu and Au
• Measurement of observed frequencies
as a function of B allows Fermi surface
to be mapped Singleton
• Open orbits do not give rise to quantum oscillations

de Haas-Van Alphen effect
• Many observable properties depend on the DoS at the Fermi level
• Magnetic susceptibility χ ( B ) is proportional to g ( EF ) , measurements
at low temperatures exhibit oscillations, when plotted against 1/B allow the
determination of extremal Fermi surface cross-section – the de Hass-Van
Alphen effect
• Experiments require high purity samples – electronic mean free path must
be long enough to allow the electrons to complete one orbit before scattering
• High magnetic field: which makes the cyclotron orbits tighter and helps to
fulfil the mean free path condition
• Low temperature: the DoS oscillations are smeared out when the Fermi
surface is smeared by thermal broadening, typically T<1K for transition
metal compounds and <100mK for heavy fermion compounds
• Measurement system consists of sample coil and compensating coil in
series opposition - without a sample the voltages induced cancel
• Voltage induced V = α (dM / dB )(dB / dt ) where α is a constant
• If we know dB / dt we can find dM / dB in which dHvA oscillations occur
• Use either large static field up to 20 T plus small modulation field of a few
mT or pulsed fields up to 60T rising to that value in a few mS
De Haas Van Alphen effect in Copper
• First observed at the Cavendish

by David Shoenberg in 1959
• Impulsive field – discharged
4000
. µ F capacitor at 2500V
through liquid air cooled magnet
rising to 2 × 10 gauss in 10mS
5
• Magnetic moment oscillations

measured by coil around sample
• Whisker of Cu oriented along
(111) axis, oscillation period of
−9 −1
, × 10 gauss effective
1.7
mass of 1.3 times electron mass
• Other whiskers along (100) and
(110) directions did not show the
effect – not clear why
1T = 104 Gauss  D Shoenberg Nature 183, 171 (1959), Proc Roy Soc
79, 1 (1962)
De Haas Van Alphen effect in Sr2RuO4
• 2D layered perovskite structure
• Metallic properties – superconducting at
temperatures <1K and well described by
Fermi liquid theory – similar to 3He
• High purity crystals prepared by floating
zone method – moving vertical
polycrystalline rod down through focussed IR
radiation from halogen lamps
• Only small region is molten (~2100°C) at any
one time, at liquid/solid boundary impurities
John Passaneau, Penn State
diffuse into liquid region
• As liquid region moves through crystal,
impurities move with it leaving pure single
crystal behind
• Material not in contact with container – often
main source of contamination
• Impurities collect at one end of crystal which
can be cut off
QCMP Lent/Easter 2019 Y Maeno et al Phys Today 54, 42 (2001) 11.11
• Field modulation technique used to

measure susceptibility at temperatures
between 20mK and 1.2K in static fields up
to 18T
• Beating in oscillations clearly visible
• Fourier transform of data plotted as a
function of 1/ B reveals three main peaks
α
. , β , γ plus harmonic at 2α
• Splitting in β causes beats in long field
sweeps
• Amplitude of oscillations rises dramatically
below 1K following predictions of Fermi
liquid theory
A P Mackenzie et al Phys Rev Lett
76, 3768 (1996)
• Observed frequencies are proportional
to k-space area
• Consistent with a Fermi surface with
two electron cylinders ( β , γ ) centered
on ΓZ line and one hole cylinder (α )
running along corners of Brillouin zone
• Results compared with calculated
Fermi surface parameters
α
β γ α
α
QCMP Lent/Easter 2019 A P Mackenzie et al Phys Rev Lett 76, 3768 (1996) 11.13
• Quantum oscillations
• Density of states oscillations
• de Haas-van Alphen effect (dHvA)
• Fermi surface orbits
• dHvA in copper
• dHvA in strontium ruthenate

Lecture 11
The end

Lecture 12
David Ritchie



Photoemission
• The most direct way to measure the electron spectral

function is by photoemission
• In a photoemission experiment photons are incident
on a solid sample. Electrons are excited from
occupied states in the band structure to states above
the vacuum energy
• The excited electron leaves the crystal and is
collected in a detector that analyses both its energy
and momentum
• The incident photon carries very little momentum
compared to the crystal momentum, so the
momentum of the emitted electron parallel to the
surface is close to that of its original state in the band
structure of the solid
• The perpendicular component of the momentum is
not conserved – changes as electrons escape
through surface Density of states

Photoemission
• We relate the energy of the outgoing electrons, E f , to the energy of the
incoming photons ω , the work function φ and the initial energy of the
electron in the solid Ei 2k 2
Ef = =Ei + ω − φ ,k f  =ki
f
2m
• In this equation Ei is referenced to the Fermi energy EF , but E f is
referenced to the vacuum ground state energy
• We use the detector angle θ to find k with k = k f sin θ
• Problems occur if sample surface is rough as momentum parallel to the
surface is changed
• Photoemission data is easiest to interpret when there is little dispersion of
electron bands perpendicular to the surface – as in anisotropic layered
materials
• Analysing both the energy and momentum of the outgoing electron allows
the determination of the band structure directly. Integrating over all angles
gives a spectrum proportional to the total density of states.
• Photoemission gives information only about the occupied states – inverse
photoemission involves injecting an electron into a sample and measuring
the ejected photon, allowing the mapping of unoccupied bands
Angle Resolved Photoemission Spectroscopy (ARPES)
equipment
wikipedia
• ARPES Systems use ultrahigh

−9
vacuum techniques P ≤ 10 mbar so
electrons travel to detector without
encountering a gas atom
• 3-axis sample rotation
• Cryogenic temperatures for samples
• Detectors available for electron spin Scientaomicron DA30 ARPES system
direction measurements
Angle resolved photoemission - GaAs GaAs
• Sample is GaAs 50nm thin film doped with Be
protected by 1nm thick As cap layer
• Soft x-rays of different polarizations (893eV)
give photo-electrons enough energy to escape.
• ADRESS beamline at Swiss Light Source used
• UHV conditions – 5x10-11mbar, T=11K
• Thermal broadening 50-150meV
• Results show Band dispersion E ( k ) including
light hole, heavy hole and split-off hole bands
M Kobayashi et al Appl Phys Lett 101, 242103 (2012)
V N Strocov et al J Synch. Rad. 21, 32 (2014)

Angle resolved Photoemission - layered metal Sr2RuO4
• 28eV photons, electron energy
resolution <21meV
• Bands nearly 2D in character
• (a) energy scans from Γ to centre of
zone face M (b) from X to corner M
• Several bands sharpen as they
approach and cross EF
• (c) shows positions of peaks as a
function of momentum at EF
Fermi surface
• (c) should be compared to band map
structure calculation of Fermi
surfaces in (d)
(d )
Experiment: A Damascelli et al Phys Rev

Lett 85, 5194 (2000)
Theory: I I Mazin and D J Singh Phys Rev

QCMP Lent/Easter 2019 John Passaneau, Penn State Lett 79, 733 (1997) 12.7
Tunnelling
• Tunnelling spectroscopies, which inject or remove electrons through a
barrier have now evolved to be very important probes of materials
• A potential barrier allows a probe (usually a simple metal) to be maintained
at an electrical bias different from the chemical potential of the material
• The current passed through the barrier comes from non-equilibrium injection
– tunnelling
• Model for tunnelling from a metal into more complex material shown in figure
• Current is given by integrated area
between two chemical potentials –
provided the matrix element for
tunnelling is taken into account
• If DoS for metal (or probe) – labelled 1 1
is slowly varying, the differential metal 2
conductance dI / dV is proportional to sample
DoS of material itself at the bias eV
above the chemical potential µ 2

Tunnelling
• With the metal/probe (1) and sample (2) maintained at different electrical
potentials separated by a bias voltage, the current through the junction can
be predicted to be of the form µ
I∝
µ
∫
+ eV
g1 (ω )g 2 (ω )T (ω )dω
• Where T (ω ) is the transmission through the barrier for an electron of

energy ω and g1 , g 2 are the densities of states
• If the barrier is very high so T (ω ) is not a strong function of energy and if the
density of states in the contact/probe, g1, is approximately constant the
energy dependence comes from the density of states, g 2 inside the sample
being investigated
• Hence the differential conductivity is proportional to the density of states in
the sample
dI
∝ g 2 ( µ + eV )
dV
• It is difficult to maintain large biases so most experiments are limited to
probing electronic structure within a volt or so of the Fermi energy

Scanning tunnelling microscopy
• A scanning tunnelling microscope (STM) uses a sharp metal tip positioned
by 3 piezoelectric transducers with vacuum as the tunnel barrier.
• The tunnelling probability is an exponential function of the barrier thickness
• High spatial resolution possible - 0.1nm lateral and 0.01nm depth, individual
atoms can be imaged and manipulated despite nm or larger tip
• Tip close to surface - electron
wavefunctions overlap
• On applying bias to sample a
tunnel current is measured
• Current converted to a
voltage and fed back to the z-
piezo controller to keep the
current constant
• Z piezo voltage gives surface
topography when scanned
• Invented by Binnig and Roher
at IBM labs in Zurich - won
Nobel prize in 1986 Introduction to scanning tunnelling microscopy C J Chen
G Binnig et al Appl Phys Lett 40, 178 (1982)
QCMP Lent/Easter 2019 G Binnig et al Phys Rev Lett 49, 57 (1982) 12.10
Scanning tunnelling microscopy
catalysts
• Scanning tunnelling microscopy used to study
processes on single crystal surfaces
• E.g. design of new catalysts - in this case to
produce hydrogen from hydrocarbons and water
• STM image shows the surface of a Ni single
crystal
• Some of the Ni atoms are substituted by Au
atoms. The Au atoms are darker and the Ni
atoms around a Au atom are brighter
• This is not because the Au atoms are
depressed but they have a lower local DoS and
the Ni atoms adjacent to a Au atom have
enhanced DoS due to perturbation by Au atom
• Because DoS is closely related to catalytic
reactivity the perturbed Ni atoms are more
highly reactive and act as a better catalyst
QCMP Lent/Easter 2019 F Besenbacher et al Science 279, 1913 (1998) 12.11

Scanning tunnelling microscopy – positioning single atoms
• The tip is positioned above the
adatom to be moved
• The tunnelling current is increased
lowering the tip until the tip/adatom
interaction energy reaches diffusion
activation – where adatom can move
across ridge between stable positions
• Pull atom to desired location, reduce
current to move tip away
• In UHV at 4K Fe atoms can be moved
on Cu(111) surface to form a 48 atom
ring with a diameter of 7.13nm
• Atom ring acts to confine Cu surface
state electrons
• Tunnelling spectroscopy shows
discrete resonances in local density
of states indicating size quantisation
QCMP Lent/Easter 2019 M F Crommie et al Science 262, 291 (1993) 12.12

Cyclotron resonance
• It is possible to make a direct measurement of the cyclotron resonance
frequency ωc = eB / m and hence effective mass using millimetre waves or
∗
far infrared radiation to excite transitions between Landau levels

• This experiment is known as cyclotron resonance
• For semiconductor samples which have much lower carrier density than
metals, the radiation can easily penetrate samples
• Measurements are usually made in
transmission, either by fixing the
magnetic field and varying the energy
of the radiation or using a fixed
frequency source such as a far infra-
red laser (shown here) and sweeping
the magnetic field detecting the
radiation with a bolometer.
• The linewidth of the resonance gives
information about the scattering rate
• In lightly doped samples carriers must be excited into bands by raising the
temperature or illuminating the samples with above bandgap radiation
Cyclotron resonance in Ge
• Figure shows absorption by cyclotron
resonance in a single crystal of Ge at
4K
• Electrons and holes present because
of above bandgap illumination
• Microwave frequency 24GHz and
magnetic field applied in (110) plane
at 60 degrees to [100] axis
• Resonance due to light and heavy
holes visible as are three electron
resonances
• The three electron resonances occur
because the anisotropic band minima
lie along [111] axes and the static
magnetic field makes three different
angles with these 4 axes
• Experimental and calculated effective
masses are shown in the figure
QCMP Lent/Easter 2019 G Dresselhaus et al Phys Rev 98, 368 (1955) 12.14
• Photoemission
• Angle resolved photoemission spectroscopy (ARPES)
• ARPES equipment
• ARPES applied to GaAs and strontium ruthenate
• Tunnelling, scanning tunnelling microscope (STM)
• STM applied to catalysts and positioning atoms
• Cyclotron resonance - example germanium

Lecture 12
The End

Lecture 13
Cu
David Ritchie



Scattering in metals
• First we consider the effect of scattering in a metal with isotropic bands
∗
characterised by an effective mass m and a spherical Fermi surface
• Earlier in the course we discussed the relaxation time approximation finding
an expression for electrical conductivity in terms of the electron number
density, n , effective mass and scattering time for electrical conduction,τ σ
σ = ne 2τ σ / m∗
• The thermal conductivity, κ is defined by J q =−κ ∇T where J q is the flux
of heat (energy per unit area per unit time)
We also have from kinetic theory κ = 13 v τ κ Cel , τ κ is the scattering time
2
•
for thermal transport,Cel is the electronic specific heat given by:
Cel = 12 nπ 2 k B2T / EF
• If we assume v
2
= vF2 where vF is the Fermi velocity then we obtain for
the electronic thermal conductivity
T 2
κ = nπ k
1
6
2 2
Bv Fτ κ
EF
∗ 2
• Assuming EF = 12 m vF , taking the ratio of thermal to electrical conductivity
and dividing by T we can define: κ π 2 k B2 τ κ
=
L =
QCMP Lent/Easter 2019 σT 3e 2 τ σ 13.3
Wiedemann-Franz law
κ π 2 k B2 τ κ
• From the previous slide =
L =
σT 3e 2 τ σ
• If we assume the scattering times are the same τ κ = τ σ we have the
Wiedemann-Franz law, which states that for metals the ratio of the thermal
conductivity to the electrical conductivity is proportional to the temperature
L0 is the Lorentz number with value L0 =(π k B / 3e ) =2.45 × 10 W ΩK
2 2 2 −8 −2
•
• Experimental results for Kittel
T=0 and 100°C (see table)
are in good agreement
with this value for a range
of different metals
• So by measuring the
electrical conductivity we
can estimate the thermal
conductivity
• As the temperature is lowered the Lorentz number decreases, e.g. for pure
Cu at 15K it is an order of magnitude smaller than L0
• Attributed to a difference between the thermal and electrical scattering times
Electrical transport in metals
• Previously we have seen that=
v g (1/ )∇k E (k ) - so electron velocities
are always perpendicular to the surface of constant energy
• Consider electrical conduction; with no electric field the velocities are in
random directions and no net current flows
• When a electric field is applied, electrons at the Fermi surface acquire a
small amount of extra velocity in a particular direction - top part of diagram
• Electrons on RHS of fermi surface move into
Singleton
slightly higher energy states, other electrons fill
the states vacated. On LHS states are vacated
Phonon scattering
• Fermi surface moves by a small amount where
δ. k 1 m∗vd , vd ≈ 10−3 ms −1 is the drift velocity
• Note that vd  vF ≈ 3 × 106 ms −1
• The lower diagram shows the Fermi-Dirac
distribution for left heading and right heading
electrons (dotted) and equilibrium situation (grey)
• τ σ is time to randomise an electron’s forward
velocity, a scattering process sends an electron
heading to the right into empty state heading left
Black circles filled states,
white circles empty states
Thermal transport in metals
• For thermal conductivity Singleton
• In kinetic theory, when conducting heat, gas
molecules travelling in one direction have a
higher thermal energy (or temperature) than
those travelling in the other direction
• In the diagram the electrons travelling left are
cooler with a less smeared out Fermi-Dirac
distribution than those travelling to the right –
in the direction of the heat flow
• The long arrow represents phonon scattering
events at room temperature which affect both
electrical and thermal conduction.
• The short arrows are phonon scattering processes at low temperatures which
only affect thermal conduction – by warming up cold electrons and cooling
hot electrons
• τ κ is the time to randomise an electron’s thermal energy, a scattering process
can cause an electron to lose or gain  k BT of energy and move into an
empty state close by or move from hot side of Fermi surface to cold side by
scattering off a high momentum phonon
Matthiessens’s rule
• We assume that electronic scattering rates are additive
1 1 1 1
= + + + ....
τ τ1 τ 2 τ 3
Where the τ are scattering rates due to different processes e.g. collisions
−1
• j
with or emission/absorption of phonons, scattering from impurities etc
• This equation suggests that the scattering process with the shortest
scattering time will dominate – so we can predict regions of temperature
where we can ignore all forms of scattering but one
• E.g. it is reasonable to assume that the scattering of electrons at high
temperature will be almost entirely due to phonons
• Matthiessen’s rule is only an approximation, it fails when the outcome of one
scattering process influences another and when one or more scattering time
is a function of k
• In the second case κ , σ will involve total scattering times due to all
processes averaged over k whereas the equation implies the summation of
reciprocals of each scattering time individually averaged over all k
• An application of Matthiessen’s rule is the calculation of electron mobility in
two-dimensional electron gases for different scattering processes
Emission and absorption of phonons
• We picture phonons as propagating local distortions of the crystal which may
scatter an electron in two different ways:
• Elastic processes, where both electron and phonon change wavevector and
energy, constrained by conservation of energy and momentum
• Inelastic processes where the phonon is emitted or absorbed by an electron
causing the electron’s wavevector and energy to change
• So far we have assumed that the positions of ions are not affected by the
presence of mobile electrons – unrealistic since they are both highly charged
• In reality the passage of an electron will result in the distortion of the lattice
around it. The electron can be scattered by this – it has emitted a phonon
• Phonons behave as massless bosons – they can be created and destroyed
in a similar way to photons
• They have a black body distribution with energy at temperature T, ω  k BT
• So when an electron scatters from or absorbs a phonon, the phonon will
have an energy  k BT
• Electrons are distributed within about ± k BT of the Fermi energy. An
electron can only emit a phonon up to energies of roughly k BT , since there
are no unoccupied states for it to fall into at lower energies
Electron-phonon scattering at room temperature
• The probability of emitting a phonon of energy  k BT will have a similar
temperature dependence to the probability of absorbing a phonon since:
• Emission depends on density of available phonon states with energy  k BT
• Absorption depends on the number of phonons around with an energy  k BT
• The Debye temperature, θ D of most metals is within about 100K of room
temperature (298K) (e.g. Cu 315K, Fe 420K, Al 394K, Pt 230K, Ag 215K)
• k Bθ D is roughly the energy of the most energetic phonons, so at room
temperature phonons with energy ω  k Bθ D will have wavevectors similar
to the width of the Brillouin zone and the Fermi wavevector q  k F
• So 1 phonon can scatter an electron to the other
side of the Fermi surface q
• Hence at room temperature the electrical and
thermal scattering rates are approximately equal k1 k2
τ. σ−1 ≈ τ κ−1 (and hence the Weidemann-Franz law is 1 k  k k
2 F
obeyed) and are proportional to the number of

phonons with ω  k BT Fermi surface

Electron-phonon scattering at low temperatures
• At low temperatures T  θ D phonons will have energies k BT  k Bθ D so q
is much less than the size of the Brillouin zone and the Fermi wavevector
• One inelastic scattering event will be able to change the
electrons energy by  k BT and hence the thermal k  k k
scattering rate τ κ ∝ no. of phonons with ω  k BT
−1
1 2 F
k2
which is ∝ T at low temperatures from Debye theory
3 q
qk F
θ
θ small k1
• Hence one phonon scattering event (elastic or inelastic)
will be unable to knock the electron to the other side of
the Fermi surface (required for electrical scattering) and Fermi surface
so the electrical scattering rate is much less than the
thermal scattering rate τ σ  τ κ
−1 −1
• This is the reason for the failure of the Weidemann-Franz law at low T
• To take account of the fact that that many scattering events through a small
angle θ are required before the excess forward velocity of the electron is
randomised the scattering rate must included a weighting factor
1 − cos θ ≈ θ 2 / 2 ≈ q 2 / 2k F2 ≈ ω 2 / 2k F2 vφ2 ∝ T 2
where ω = vφ q is the phonon dispersion relation, vφ the speed of sound
• Hence we have τ σ ∝ T , τ κ ∝ T and so τ σ−1 is much smaller at low T
−1 5 −1 3

Electron-phonon scattering at low temperatures
• The τ σ ∝ T dependence is rarely exactly obeyed
−1 5
• The periodicity of k-space allows phonons Umklapp

scattering
with small q to scatter electrons at the
Fermi surface (A) into empty states with
energy  EF in an adjacent Brillouin zone
(B). These states may have a velocity
almost opposite to that of the original
state, This is called umklapp scattering
• Complicated Fermi surfaces may allow scattering of electrons by phonons
with small q into parts of the Fermi surface with very different velocities
−θ /T
Both of these effects give a scattering rate τ ∝ e F where θ F is a
−1
•
characteristic temperature depending on the Fermi surface geometry
• For very low temperatures the phonon scattering becomes negligible and
scattering of electrons by impurities and defects becomes dominant
• Impurities have a different ionic core from the host metal and therefore will
often appear to be charged with respect to the background.
• Scattering of electrons by impurities deflects them through large angles. One
scattering event degrades thermal and electrical transport in the same way
and τ κ = τ σ hence σ  const, κ ∝ T and the Weidemann-Franz law holds
Scattering – electrical resistivity
• Table gives summary of
the temperature
dependence of scattering
times and electrical and
thermal conductivities
• (a) Resistance relative to
D. K. C. MacDonald, K. Mendelssohn
room temperature value Proc. R. Soc. Lond. A 202 103-126
for three samples of Na (1950)
of different purity. The (a) Resistance of Na

results show a constant (b)
resistance at low
temperature determined
by impurity scattering.
Resistance rises for
T>~10K
• (b) shows normalised electrical resistivity data for several metals as a
function of normalised temperature. They all fit the same form as a function
of temperature
At low T the resistivity is constant, at high T, σ ∝ T ⇒ ρ ∝ T
−1
•
Scattering - thermal conductivity
• Thermal conductivity κ
rises linearly with T,
before reaching a peak
then decreasing to a
values independent of T
• Peak in κ higher for
High purity
samples with fewer
Cu
impurities – confirmed by
results for both Li and Na
Lower purity
• Similar results for Cu
• A good illustration of the
competition between
scattering times High purity
τ=
−1
τ imp
−1
+ τ ph
−1
Na
• At low T, few phonons so τ imp < τ ph , impurity
scattering dominates and κ ∝ T . As T rises to 10D
θ Lower purity
phonon scattering time becomes shorter τ ph < τ imp

and phonon scattering dominates with κ ∝ T
−2
Low temperature solid state physics
H M Rosenberg OUP 1963
Electron-electron scattering
• In metals with simple Fermi surfaces e-e scattering relatively unimportant
• Initial and final state for both electrons have E  EF , k  k F
• Energy and momentum must also be conserved
• This combination makes electron-electron scattering quite unlikely
• It becomes important when (a) the Fermi surface is complicated so
conservation of E , k becomes easier (b) density of states at EF is very large
due to large effective mass, increasing number of initial and final states
• Examples: transition metal elements and heavy fermion compounds
• Consider a filled Fermi sphere and a single excited electron with 1 > EF
• To be scattered it must interact with an electron with 2 < EF since only
states with energies less then the Fermi energy are occupied
• Pauli exclusion principle requires that these electrons can only scatter into
unoccupied levels where 3 > EF , 4 > EF
• Energy conservation requires 1 + 2 = 3 + 4
• If 1 = EF these conditions can only be fulfilled if 2= 3= 4= EF
• The allowed wavevectors for electrons 2,3,4 must then occupy zero volume
in k-space which makes the probability of this process very small at zero T
• Hence the electron scattering lifetime= at  E=F, T 0 is infinite
Electron-electron scattering
• When 1 is a little different from EF some
phase space becomes available for the
process since the other 3 energies can now
vary within a shell of thickness 1 − EF about
the Fermi surface
• This gives a scattering rate ∝ (1 − EF ) 2
because once 2 , 3 have been chosen, energy
conservation allows no choice for 4
• If the excited electron is superimposed on a
thermal distribution of electrons at non-zero T,
there is an additional range of choice in
energies available for the scattering process
with rate ∝ ( k BT ) 2
• So overall τ =α (1 − EF ) + β ( k BT ) with α , β
−1 2 2
constant T(K)
• At finite temperature 1 − EF  k BT so we can say τ −1 ∝ T 2 , τ ∝ T −2

and hence σ ∝ T −2 , ρ ∝ T 2
• Figure shows resistivity of Pd, Fe, Nb with low temperature T2 dependence
visible
• Scattering in metals
• Wiedemann-Franz law
• Electrical and thermal transport in metals
• Matthiessens’s rule
• Emission and absorption of phonons
• Electron-phonon scattering at room and low temperatures
• The effect of scattering on electrical resistivity and thermal conductivity
• Electron-electron scattering

Lecture 13
Cu
The end

Lecture 14
David Ritchie


Intrinsic semiconductors, law of mass action, doping in semiconductors,

impurity ionisation, variation of carrier concentration and mobility with
temperature - impurity and phonon scattering, Hall effect with two
carrier types.
Metal to semiconductor contact. P-n junction; charge redistribution,
band bending and equilibrium, balance of currents, voltage bias. Light
emitting diodes; GaN, organic…….


Semiconductors – intrinsic carriers
• In semiconductors the energy gap is small enough
so thermal excitation of the carriers across the gap
is important

• Diagram shows density of states for electrons and
holes as well as the Fermi function determining
occupancy of thermally excited states
• The chemical potential lies mid-gap c
• We calculate the thermal intrinsic carrier
f ( )
concentration in a model semiconductor with µ
parabolic electron and hole bands
• The conduction and valence band dispersions are v
given by 2k 2 2k 2
c (k ) =
c + ∗
, v (k ) =
v −
2me 2mh∗
• Density of states for the conduction band
3/ 2 g ( )
1  2m  ∗
1/ 2
= g e ( ) 
e
 ( −  )
2π 2   2  3/ 2
c
1  2m  ∗
1/ 2
=
• Density of states for the valence band g h ( ) 
h
 v
( −  )
2π 2   2 
• We can calculate the carrier density when the chemical potential µ is known
∞
•
function c ∫
For electrons in the conduction band n = g e ( ) f ( )d with the Fermi
1 − (−µ )/ ( k T )
=f ( ) ≈e B
e(−µ )/ ( kBT ) + 1
•  − µ >> k BT - a non-degenerate gas
Where the approximation is valid for
∗ 3/ 2 3/ 2
• Hence 1  2me  ∞  me k BT  − (c −µ )/( kBT )
∗
2  ∫
− (  − µ ) / ( k BT )
1/ 2
n 2 
( − c ) e =
d2 2 
e
2π    c  2π  
• And a similar calculation for holes gives 3/ 2
 m k T  − ( µ −v )/( kBT )
∗
p  2 2 
e h B
 2π  
• We define temperature dependent concentrations representing the number
of states within k BT of the band edge for the conduction and valence bands
3/ 2 3/ 2
m k T  m k T 
∗ ∗
=nc (T ) 2=
 2 
, nv (T ) 2 e B
2 
h B
 2π    2π  
• Hence
− ( c − µ )/( k BT ) − ( µ −v )/( k BT )
n n=
c (T ) e , p nv (T ) e
• The concentrations of electrons and holes in terms of the chemical potential
− ( c − µ )/( k BT ) − ( µ −v )/( k BT )
• =
From the last slide n n=
c (T ) e , p nv (T ) e
− g /( k BT )
• Hence we can write np = nc (T ) nv (T )e where g= c − v
• This result is known as the law of mass action and is independent of µ - as
yet unknown
• In this derivation we have nowhere assumed that the material is intrinsic and
the result holds in the presence of impurities and dopants
• The only assumption made is that the distance of the Fermi level from the
edge of both bands is large in comparison to k BT
• A simple kinetic argument shows why np is constant at a given temperature
• Suppose the equilibrium population of electrons and holes is maintained by
blackbody radiation
• The photons generate electron-hole pairs at a rate A(T ) while B (T ) np is
the rate of recombination e+h=photon
dn dp
=A(T ) − B (T )np =
dt dt
• In equilibrium d=
n / dt d=p / dt 0 hence np = A(T ) / B (T ) - a constant at
a given temperature T
• Since the product pn is a constant at a given temperature, the introduction of
a small amount of a suitable impurity to increase n will decrease p
• This is important because we can reduce the total carrier concentration n+p
in an impure crystal by controlled introduction of suitable impurities – known
as compensation.
• In an intrinsic semiconductor the number of electrons equals the number of
holes and we can write
( nc (T ) pv (T ) ) e
1/ 2 − g /(2 k BT )
n=
i p=
i
• The intrinsic carrier concentration depends exponentially on g / (2k BT )
Note 12 g and not g - because the creation of an electron also creates a hole
• From a previous slide
3/ 2 3/ 2
 m k T  − (c −µ )/( kBT )
∗
 m k T  − ( µ −v )/( kBT )
∗
n  2 e B
2 
e , p  2 2 
h B
e
 2π    2π  
• If we set n = p
 /( k T )
e 2 µ /( kBT ) = (mh∗ / me∗ )3/2 e g B ⇒ µ = 12 g + 34 k BT ln(mh∗ / me∗ )
• Which gives the position of the chemical potential.
• If mh∗ = me∗ ⇒ µ = 12 g the chemical potential is in the middle of the bandgap
Doped semiconductors
• Carriers can be created in semiconductors by adding impurity atoms – a
process known as doping
• Consider effect on a GaAs crystal of replacing a Ga atom by a Si atom
• Si provides 4 electrons instead of the normal 3 from Ga so it appears like a
Ga atom with an extra electron and an extra positive charge in the nucleus
• Suppose the electron wanders away from the impurity site, it will experience
an attractive force from the charged Si impurity
• The donor energy levels can be calculated as for a hydrogen atom
• We take into account the influence of the surrounding material by making
two corrections
(1) The Coulomb potential is screened by the dielectric constant, so it is
much weaker than in free space (for GaAs ε  13.1 )
∗
(2) Use effective mass of the electron (for GaAs me / me = 0.067 )
• The net effect is that the binding energy for the 1s state is now
e 4 me∗ me∗ / me
=
∆d = ×13.6 eV
2(4πεε 0  ) 2
ε 2
• Which for GaAs is 5.3meV – much smaller than the bandgap of 1.4eV
• From the last slide ∆ d = me∗ / (ε 2 me ) × 13.6 eV
• For GaAs the Si donor ionisation energy of about 5.3meV is equivalent to a
temperature of 50K. So Si donors in GaAs will all be ionised at room
temperature
Si R A Stradling
• The figure shows far infra-red
absorption due to P impurities in Si
(ionization energy ~45meV). The peaks
correspond to hydrogen like transitions
between the n=1 ground state and
higher levels
• The hydrogen-like bound states are referenced to the bottom of the
conduction band, because the electron unbinds from the donor atom by
occupying a conduction band state
• A Beryllium atom in GaAs acts as an acceptor dopant if it sits on a Ga site.
• Be only donates 2 electrons as opposed to the normal 3 from Ga (donating a
hole) and appears like a Ga atom with a negative charge
• So we have a positively charged hole circling a negatively charged nucleus
• The hydrogenic binding energy can be calculated in the same way as for
donors – taking into account the hole effective mass
• When the hole unbinds from its Be atom the impurity accepts an electron
from the valence band
• The accepted electron is used to complete the covalent bonding with
neighbouring atoms and renders the site negatively charged
• So while ionising a donor atom releases an electron into the conduction
band, ionising an acceptor atom absorbs an electron from the valence band
creating a hole in the valence band
• Even for very low densities of impurities, since the donor or acceptor
energies are much smaller than the band gap, impurities are often the main
source of electrically active carriers
• If donors predominate the carriers are mostly electrons and the material is
said to be n-type.
• If acceptors dominate carriers are mostly holes and the material is p-type
• In most materials there are both donors and acceptor impurities – in GaAs Si
on a Ga site is a donor and on an As site is an acceptor, whether you get n
or p-type depends on the crystal growth technique
• Experimentally the different carrier regimes may be distinguished by
measuring the Hall effect, the sign of which depends on the carrier type
Doped semiconductors – Impurity ionisation
• As long as the no. of donors/acceptors is low enough so the chemical
potential lies in the bandgap then the law of mass action holds
• Given the densities of ionised donors and acceptors N D , N A we can use
( ) ( )
3 − E /( k T )
∗ ∗ 3/2
.np  4 and the conservation law n − p = N D − N A
k BT
2π  2 m me h e g B
to find values for n, p if we know the effective masses and the bandgap
• Diagram shows the temperature
dependence of the electron density in Si
with a net donor density N D − N A = 1015 cm −3
• At T<100K the extrinsic electrons freeze out
onto the donors, the gradient depends on
the donor ionisation energy
• For 150K<T<300K all of the donors are
ionised – the saturation range, n is constant
• For T>500K the intrinsic contribution to n
becomes larger than extrinsic contribution,
the gradient depends on the main bandgap
• The intrinsic contribution, ni is around
3x109cm-3 at RT – negligible
Electrical conductivity with two carrier types
• The electrical conductivity of a semiconductor is a sum of contributions from
all carrier types, usually electrons and heavy holes written in terms of carrier
densities and mobilities= σ neµe + peµhh
• We write electron and hole mobilities in terms of scattering times for
holes µe eτ µ τ
∗ ∗
electrons and heavy = = e / me , hh e hh / m hh
• The temperature dependency of the electrical conductivity is determined by
convolutions of the temperature dependences of the carrier concentrations
and scattering times. There are two important types of scattering:
• Impurity scattering is similar to Rutherford scattering. The scattering
−2
crossection varies as  and since in the non-degenerate case   k BT
the crossection varies as T and the mean free path as λ ∝ T
−2 2
The carrier speed v ∝ 1/2 ∝ T 1/2 hence scattering time τ= λ ∝ 3/2

• imp / v T
• Contrast to metals where τ imp is independent of T - carriers in metals have
similar energy, but in semiconductors a Boltzmann energy distribution
Phonon scattering with T  θ D - the number of phonons is ∝ T so λ ∝ T
−1
•
and with v ∝  ∝ T we get τ= λ ∝ −3/2
1/2 1/2
pho / v T
• As a consequence impurity scattering dominates the mobility at low T and
phonon scattering at high T with a peak in mobility in between
Electron mobility versus D Steigerwald et al, JOM 49, 18 (1997)
µ ∝ T −3/2 µ ∝ T 3/2
temperature
3 ×1016 cm −3
• Top figure shows electron mobility of 5 1×1017 cm −3
samples of GaN with different doping 3 ×1017 cm −3
2 ×1018 cm −3
levels and carrier concentrations GaN 1.5 ×1019 cm −3
• For high temperature variation close to

expected µ ∝ T
−3/2
due to phonon
scattering
• At low temperature drop off slower than
expected µ ∝ T
3/2
for impurity scattering
– measurements at lower temperatures
may reach this dependence
• Lower figure shows highest mobility
sample of GaAs ever reported GaAs
• At high T close to predicted variation for
µ ∝ T 3/2 µ ∝ T −3/2
phonon scattering
• At low T faster drop off than expected –
possibly due to metal-insulator transition n 2.8 ×1013 cm −3
=
C R Stanley et al Appl Phys

QCMP Lent/Easter 2019 Lett 58, 478 (1991) 12.12
Hall effect with two carrier types
• Earlier in the course we studied the Hall effect for a single carrier type
• ( )
Assuming dj / dt =− j / τ + ne / m ( E + v × B )
2
• in the steady state - for electrons

=
je ( ne τ 2
e / me ) ( E + v e ×=
B ) neµe ( E + v e × B )
• If both electrons and holes are present we have
=j neµe ( E + v e × B ) + peµh ( E + v h × B )
• Assume current is flowing in the x-direction, current in the y-direction is zero,
B is in the z-direction, vxe , vxh are opposite signs and µ = v / E
j=
x eEx (nµe + p µh ), =
0 eE y (nµe + p µ h ) − eB (nµe vex + p µ h vhx )
⇒ 0 eE y (nµe + p µh ) + eBEx (nµe2 − p µh2 )
=
• Eliminating Ex we get
jx B (nµe2 − p µh2 ) Ey (nµe2 − p µh2 )
Ey =
− ⇒ RH = = −
e( n µ e + p µ h ) 2
jx B e( n µ e + p µ h ) 2
• A minority carrier can determine the Hall coefficient sign if the mobility is
– e.g. GaAs has µe 8800cm V s , µh 400cm 2V −1s −1
2 −1 −1
high enough= =
at room temperature so electrons are likely to dominate
Hall effect with electrons and holes
• Diagrams show Hall coefficient RH in InSb plotted
−1 extrinsic
as a function of T
• Upper diagram - InSb is doped with donors, lower
diagram doped with acceptors
• Ratio of mobilities µe / µ h  100
• At high T InSb is in the intrinsic regime, electrons
dominate RH and the slope of the Hall coefficient
can be used to determine the bandgap giving a
value of 0.24eV
• At low T in the upper diagram, no intrinsic carriers
. H = − ne
R 1
remains negative giving an electron
concentration = n 1.1×1016 cm −3 extrinsic
• In the lower diagram as the temperature is
lowered the intrinsic electrons freeze out they no
longer dominate RH and RH changes sign
• In the extrinsic region RH is positive indicating
holes RH = pe 1
⇒ p =2 ×1016 cm −3
• Intrinsic semiconductors, law of mass action

• Doping in semiconductors
• Variation of carrier concentration with temperature, impurity ionisation
• Electrical conductivity, impurity and phonon scattering
• Electron mobility versus temperature
• Hall effect with two carrier types – example InSb

Lecture 14
The end

Lecture 15
David Ritchie


Intrinsic semiconductors, law of mass action, doping in semiconductors,

impurity ionisation, variation of carrier concentration and mobility with
temperature - impurity and phonon scattering, Hall effect with two
carrier types.
Metal to semiconductor contact. P-n junction; charge redistribution,
band bending and equilibrium, balance of currents, voltage bias. Light
emitting diodes; GaN, organic.
Photovoltaic solar cell; Shockley-Queisser limit, efficiencies…….

Semiconductor devices
• We consider the general properties of surfaces and interfaces between
materials with applications to semiconductor devices
• We use the semiclassical approximation, treating electrons as classical
=
particles with Hamiltonian H En (k ) − eφ (r ), momentum p = k and a
spatially varying potential φ (r )
• The potential arises from external applied fields, charges induced by doping
and changes in the material composition
• When discussing narrow quantum wells we will need to modify this
approximation to quantise the energy levels
• For an isolated solid in equilibrium the energy difference between the
chemical potential, µ and the vacuum level is the work function Φ - the
energy required to remove an electron from the Fermi level and place it in a
state of zero kinetic energy in free space
• 2 different isolated materials with different Φ will have different µ
• When placed in contact their chemical potentials must equalise. Electrons
flow to charge the more electronegative material, its potential changes and
an overall balance will be established
• In general there will be internal inhomogeneous electric fields
Metal to semiconductor contact
• Consider this process for ideal metal
in contact with a semiconductor
• (a) metal and semiconductor not in
contact – in equilibrium with vacuum
level. n-type semiconductor, µ close
to conduction band edge
• (b) In contact, electrons transferred
from semiconductor to metal
producing electric potential φ ( x )
which eventually equilibrates so µ is
a constant over whole system,
combined function µ + φ ( x ) called
electrochemical potential
• (c) energy level diagram relative to
constant chemical potential.
Semiconductor band bends upwards,
donor levels emptied of electrons
leaving +ve charged depletion region
and Schottky barrier Φ b
Metal to semiconductor contact
• Barrier set up between metal and semiconductor inhibits current flow.
• An electron must either tunnel through barrier (at low T) or be thermally
excited over it (thermionic emission)
• When a large enough bias is applied junction may act as a rectifier, (with a
bias applied µ is different for the metal and semiconductor)
• Upper diagram, applying a +ve voltage to
the metal relative to the semiconductor
lowers the barrier for electrons to enter the
metal. Can eventually tip the electron bands
so much that the barrier disappears
• Current grows rapidly as +ve bias increases
• Middle diagram, if the metal bias is at a
negative voltage relative to the
semiconductor the depleted region grows in
width and current remains small
• Lowest diagram shows IV characteristic of
Silicon-tungsten diode for high frequency
(mm wave) rectification
p-n junction – charge redistribution
• A p-n junction is formed by
inhomogeneous doping when a
layer of n-type material is placed
next to p-type material
• Inside n-type µ n just below bottom of
conduction band
• Inside p-type µ p just above top of
valence band
• Joining p-n gives step in µ
• Current flows because of different
chemical potentials, electrons from
n-side fill holes on p-side
• No mobile charges left in depletion
region around junction – a thickness
of between 10nm and 1µm
• Ionised donors have +ve charge on
n-side, ionised acceptors have –ve
charge on p-side
• Charge neutrality ⇒ N a w p = N d wn
p-n junction – band bending
• Poisson’s law – charge redistribution
causes electrostatic potential φ
• Energy levels shift E → E − eφ ( z )
φj
• Far away from junction in n-type µ is
z
close to top of bandgap and in p-type E − eφ
it is close to bottom of band gap
• Charge flows until energy levels have
moved enough for µ to line up across
µ
junction and equilibrium is reached z
• Junction potential eφ= j µn − µ p  Eg
• Charge neutrality and Poisson’s
1/ 2
equation lead to expressions relating  2ε 0εφ j N a N d 
junction potential, φ j thicknesses of n- N ( a ,d ) w( p ,n ) = 
and p-type depletion regions wn , w p  e N a + N d 
and doping densities of n- and p-type
regions N d , N a 1/ 2
 2ε 0εφ j N a + N d 
=φj
e
2ε 0ε
( a p d n)
N w 2
+ N w 2
wp + wn =
 
 e Na Nd 
p-n junction in equilibrium
• Top: 2 ways of representing energy levels in
a p-n junction (a) includes electrostatic
potential in the electrochemical potential
(b) chemical potential constant and potential
shifts energy levels
• When the donor or acceptor levels pass
through µ levels are ionised and annihilate
– impurity levels now charged
• Bottom: (a) carrier and (b) charge densities
near the depletion region of a p-n junction
• When T is low the carrier density changes
abruptly at the point where µ passes
through the donor or acceptor level
• Close to the barrier the carriers are depleted
– the system is physically charged
• Charge density of + eN d on the n-type side and −eN a on the p-type side
• This dipole layer produces a potential φ ( x ) shown in (b)
• Potential self consistently determines charge flow and depletion region width
p-n junction in equilibrium - summary
• Overview of p-n junction in

equilibrium
• Mismatch in µ causes charge
transfer across junction building
contact potential, resulting in band
bending until µ equal on both sides
• Electrons from n-side cross junction
annihilating holes causing carrier
free depletion region
• Charge transfer results in space
charge, maximum charge density
given by dopant concentration
• Space charge causes in-built
junction field E j and contact
potential ϕ j which builds until charge
transfer stops

p-n junction – balance of currents
• Majority holes diffuse from p to n-
side, recombine with electrons
• Hole recombination or diffusion
( h ) − eφ j / ( k BT )
(h)
current J rec = J0 e ∝
probability that holes climb the
potential barrier, J 0( h ) constant
(h)
• Electrons do the same in the J rec
opposite direction. Currents add hole
• Small no. of minority holes
(. np = ni2 ) thermally generated
on n-side drift to p-side under
influence of in-built field.
• Resulting hole drift or generation
(h)
current J gen depends on T, details
of band structure and doping
• Electrons do same in opposite (h)
direction. Currents add J gen
hole
Equilibrium: J tot = J rec − J gen = 0
(h) (h) (h)
•
p-n junction – voltage biased
Bias voltage (+ve on p-side) modifies effective junction potential φ j = φ j − V
eff
•
• This changes the recombination current (barrier height changes) but leaves
generation current unchanged – minority carrier density on both sides
remains the same
( h ) eV / k T
Forward bias J rec = J gen e B recombination current outstrips generation
(h)
•
current exponentially giving diode action (at zero bias J rec = J gen )
(h) (h)
( h ) eV / k BT
• (h)
Hence J tot = J rec
(h)
− J gen
(h)
= J gen e − J gen
(h)
= J gen
(h)
(eeV / kBT − 1)
• In reverse bias:
(h)
J rec → 0 ⇒ J tot
(h)
→ − J gen
(h)
current saturates at low level

• Diode equation – sum of hole and
=
electron currents (
I I sat eeV / kBT − 1 )
• Saturation current
− Eg / k BT
I sat = J gen
(h)
+ J gen
(e)
∝ ni2 ∝ e
• Reverse breakdown when reverse
bias gets too large
• Mechanism - tunnelling of majority
carriers across depletion zone
Light emitting diodes
• The inverse process to the
photovoltaic effect powers light
emitting diodes (LEDs)
• A current is injected into a p-n diode
in a non-equilibrium situation where
electron and hole chemical
potentials differ by a large bias
potential
• Electrons are injected from n-side to
p-side of junction and holes in p-n junction in forward bias, electrons
reverse direction and holes injected into junction,
photon emitted on recombination
• Recombination of electron-hole pair
occurs with emission of a photon • Efficient LEDs have been
with energy close to bandgap developed across the visible
• Not efficient for indirect bandgap spectrum - up to six times more
semiconductor (Si, Ge) efficient than incandescent bulbs
• Direct bandgap III-V or II-VI • Nobel prize in 2014 awarded for
materials used development of efficient blue GaN
LEDs.
M S Slur et al IEEE trans Elect Dev.
GaN LEDs 57, 12 (2010)
• Red and green LEDs developed

since 1970s, used mainly as indicator
lights
• Development of efficient blue GaN
based LEDs over past 20 years has
led to many applications in lighting
• High efficacy – 300lm/W (100lm/W in
2010) compared to 16 lm/W for
incandescent lights and 70lm/W for
fluorescent lights
• But for ‘White’ LEDs which combine
blue light with yellow light from a
phosphor only 80lm/W efficacy
• Lifetime upto 10x longer than
Fluorescent lights
• The future:
• Developing efficient green LEDs for red/green/blue spectral control
• Developing efficient UV LEDs for water purification and health applications
Organic LEDs
• Electroluminescent layer is a film of
organic semiconductor
• Main applications in displays for TVs
and mobile phones
• Works without back light - can be
thinner and lighter than an LCD and
achieve a greater contrast ratio and
wider viewing angle
• Response times up to 1000x faster
than LCDs
• Can be deposited on flexible
substrates
• Based on either small molecules or
polymers
• Problems with water damage and
lifetime – gradual degradation
• Currently expensive but costs likely to
fall with mass production
• Metal to semiconductor contact

• p-n junction:
• Charge redistribution
• Band bending – equilibrium - balance of currents
• Effects of voltage bias
• Light emitting diodes: GaN, organic

Lecture 15
The End!

Lecture 16
David Ritchie


…. bending and equilibrium, balance of currents, voltage bias. Light

emitting diodes; GaN, organic.
Photovoltaic solar cell; Shockley-Queisser limit, efficiencies,
commercialisation. Field effect transistor; JFET, MOSFET.
Microelectronics and the integrated circuit.
Band structure engineering; electron beam lithography, molecular beam
epitaxy. Two-dimensional electron gas, Shubnikov-de Haas oscillations,
quantum Hall effect, conductance quantisation in 1D..……

Photovoltaic solar cell
• Operation of a p-n junction based
solar cell
• When illuminated each photon
generates an electron-hole pair
• Pairs generated away from junction
will recombine rapidly
• Pairs generated near junction
separated by in-built electric field
• Electrons flow towards n-side, holes
towards p-side Electron-hole pair generated in
• This is equivalent to increasing the junction region then separated
by junction field Ej
generation current, which flows in
reverse (as opposed to forward)
direction (n to p)
• Separation of charges across depletion region adds an extra dipole to
system - like charging a capacitor and generating an overall electrical bias
• Induced voltage is in forward direction – because it is opposite in sign to the
built in potential
Photovoltaic solar cell
• Model operation as a current
source in parallel with a diode
• Current delivered I ph depends on
amount of light falling on junction
area
• Consider IV characteristics
• For zero load resistance (short
circuit, I load = I ph but V = 0 ) and
infinite load resistance ( I load = 0 )
no power is extracted IV characteristic for
a solar cell. Amount
• What is open circuit Vd ? of power that can
• Upper limit given by band gap Eg be extracted is
- if Vd exceeds φ j  Eg / e then
limited by bandgap
and photocurrent
the in-built junction field vanishes
and photo generated carriers no
longer swept out of junction area
• Maximum power extracted for
ideally chosen load resistance is
determined by quantity I ph Eg / e
Solar cells: Shockley-Queisser limit
• How far is it possible to optimise solar
cells by tuning the bandgap energy?
• Shockley and Queisser analysed the
maximum efficiency of solar cells
• They considered matching between
semiconductor band gap and intensity
spectrum of sunlight
• Photons can only be captured if the
bandgap is lower than the photon energy
• Power extracted depends on bandgap
• Also excited carriers in excited states well away from band edges will lose
energy by decaying to lower lying states before leaving device
∞
• Ratio of the energy extracted ∞ from sunlight E∫ I (ω ) Eg dω to the total
energy incident on device ∫0
I (ω )ω dω
intensity) can be optimised as a function of Eg
g
(where I (ω ) is the spectral
• When combined with other limitations, optimum efficiency of single –

junction solar cell is 33% for a bandgap of 1.2eV. Shockley and Queisser, J
• Si solar cells achieve up to 22% efficiency Appl Phys, 510 (1961)
Solar cells- best research efficiencies
Conversion efficiencies of best research solar cells worldwide from 1976 through
2016 for various photovoltaic technologies. Efficiencies determined by certified
agencies/laboratories
Sarah Kurtz and Keith Emery - National Renewable Energy Laboratory (NREL), Golden, CO, USA
Solar cells – commercialisation (1)
Device
Cost
Global capacity
World’s largest solar farm in Rosamond,

California. 1.7 x106 solar panels over
13km2. 579MW, power for 255,000 homes
QCMP Lent/Easter 2019 wikipedia 16.7
Solar cells – commercialisation (2)
Reduction in Si cost reduces module cost to <$0.5/W
U.S. Solar Photovoltaic System Cost Benchmark: Q1 2017

Ran Fu et al.
National Renewable Energy Laboratory Slide courtesy Dr Louise Hirst
Field effect transistor
• Field effect transistors (FET) are the mainstay of the semiconductor
industry
• Their principle of operation is based on our ability to manipulate the carrier
density in a channel between two electrodes via a controlling voltage
applied to a third electrode
• This controlling electrode is called the gate
• The electrode at which the mobile carriers (usually the electrons)
originates is called the source
• The electrode towards which the carriers move is called the drain.
• A very readable account of the operation of FETs can be found at:
http://www.freescale.com/files/rf_if/doc/app_note/AN211A.pdf.
• We distinguish between two types of FETs
• (a) junction based FETs (JFET), which use p-n junctions to control the
width of the conducting channel
• (b) FETs in which the gate is separated from the rest of the device by an
insulating layer, the metal-oxide semiconductor FET (MOSFET)

Junction field effect transistor (JFET)
• JFET varies the current between source
and drain by changing the width of the
conducting channel
• Between source and drain in n-type
semiconductor, electrical conductivity is
high because of the high carrier density
• Adding p-type regions between source
and drain contacts connected to gate
electrodes, allows control over the
current flow between source and drain
• At junction between p-type and n-type
regions of the device, depletion zones
form and conducting width of channel
between source and drain is reduced
• By applying a voltage to the gate electrodes, depletion zone width can be
controlled, altering the width of conducting channel:
• Positive gate voltage reduces the size of the depletion zone, and increases
the current in the conducting channel
• Negative gate voltage widens the depletion zone and reduces the current
JFET operation
• Current voltage characteristics of an n-type
JFET with p-type gate
• With increasing drain-source voltage VDS, the
drain-source current ID rises roughly linearly,
controlled by the gate-source voltage VGS
• Increasing VDS causes depletion regions to
grow until they meet
• In this saturation region any further increase in
VDS is counterbalanced by an increase in the
depletion region towards the drain
Saturation
• The effective increase in channel resistance region
Linear
Breakdown region
prevents any increase in ID as VDS increases region
• VDS that causes the current limiting condition is
known as the pinch-off voltage VP
• If VDS is too high breakdown region is entered
and ID increases rapidly
• ID controlled by VGS in saturation region is
operational mode for JFET in amplifier circuit
Metal oxide semiconductor field effect transistor (MOSFET)
• MOSFET – very commonly used in modern electronics
• Width of the conducting channel between source and drain controlled by
electric fields using a gate electrode insulated from the rest of the device
• No current flow from the gate electrode - extremely high input impedance
• In contrast in the JFET - where a small current flows across depletion region
Manufacturing process for enhancement mode n- channel MOSFET
• (a) p-doped substrate (b) n-doped
source and drain contacts by ion
implantation (a) (b)
• (c) insulating silicon oxide layer

deposited followed by insulating silicon
nitride – stops sodium diffusion
• (d) metallic contact made to source and
drain through holes in insulator
(c) (d)
• Gate electrode insulated from substrate
• Applying a +ve voltage to the gate pulls electrons into the depleted zone
and establishes a conducting channel between source and drain
MOSFET operation
• Several MOSFET designs used, operation relies on two principles:
• (1) By changing gate voltage, depleted regions between source and drain
electrodes can be filled with carriers or depleted of carriers. Allows variation
of the resistance of the source-drain channel
• (2) As for the JFET, pinch-off occurs near the drain electrode, causing the
source-drain current to saturate making the device useful as an amplifier
(a) (b) (c)

• (a) enhancement mode MOSFET, +ve voltage pulls minority carriers
towards surface forming high conductivity inversion layer channel
• (b) depletion – enhancement mode MOSFET: -ve voltage depletes channel,
increasing resistance, +ve voltage enhances channel, reduces resistance
• (c) Typical IV characteristic for depletion-enhancement mode MOSFET, note
pinch-off at high source – drain voltage
MOSFET- inversion layer
• Band bending in a MOSFET
• Applying a +ve voltage to the gate electrode creates
an electric field across the insulating oxide layer
• This field penetrates some distance into the
semiconductor
ϕ ( z) z
• This field sets up a varying potential ϕ ( z ) close to the
surface of the semiconductor
• If the resulting band-bending at the z
semiconductor/oxide interface
becomes larger than the band gap Eg
the conduction band edge falls below
the chemical potential at the surface
causing an inversion layer to form
• Width of inversion layer can be
controlled by gate voltage but is
narrow enough so quantisation
effects are observed

Microelectronics – the start
• First integrated circuit developed by Jack Kilby at
Texas Instruments in the summer of 1958 (he wasn’t
allowed a summer holiday….)
• Developed combination of a transistor, resistors and
capacitor on a single piece of germanium which
produced a sinewave voltage – an oscillator
• Also inventor of handheld calculator and thermal
printer………….
Patent
First integrated circuit – an

electrical oscillator, Jack Kilby
Texas InstrumentsI1958 (Nobel
prize 2000)

Microelectronics – now
• No. of transistors in an IC risen by a
factor of 106 in 40 years
• Development being slowed by
difficulty of making small enough
features – currently 10nm process
being developed, 7nm has been
demonstrated, 5nm being thought
about….
• Will run into quantum tunnelling
problems at 7nm – may be the end of
Moore’s Law…..
• 3D devices and architectures being 42nm
developed to increase density
• Materials with higher mobility than Si
required to increase speed and
reduce power dissipation – Ge or
InGaAs
QCMP Lent/Easter 2019 14nm trigate transistor 16.16

• Photovoltaic solar cell - Shockley-Queisser limit, efficiencies,

commercialisation
• Field effect transistor - JFET, MOSFET
• Microelectronics and the integrated circuit

Next term – Final 6 lectures
• Semiconductor devices and low-dimensional physics

• Electronic instabilities - charge density waves, magnetism
• Fermi liquids - collective excitations and heavy Fermion
materials

Lecture 16
Device
Have a good vacation!

Lecture 17
David Ritchie


….Photovoltaic solar cell; Shockley-Queisser limit, efficiencies,

quantum Hall effect, conductance quantisation in 1D. Single electron
pumping and current quantisation, single and entangled-photon
emission, quantum cascade laser.


Band structure engineering
• The spatial control of band structure using different materials, can result in
the confinement of electrons and/or holes to 2D,1D or 0D
• Relies on the development of advanced crystal growth techniques such as
molecular beam epitaxy (MBE) and metal organic chemical vapour
deposition (MOCVD) to grow near perfect crystals on single crystal
substrates at growth rates around 1 monolayer/s
• Epitaxial heterostructures of different materials can be grown with near
monolayer control over semiconductor composition and doping
• Heated substrate creates large lateral motion of atoms before incorporation
into the growing crystal and leads to very smooth layers
• Ultra-high purity materials can be grown with impurity levels as low as 1 part
in 1010 leading to electron transport mean free paths as high as 0.1mm
• Both growth techniques used for routine production of lasers for CD, DVD
players (mainly MOCVD) and high frequency transistors (mainly MBE)
• Electron beam lithography is used to produce laterally patterned structures
on 100nm scale
• A very wide range of semiconductors can be used to produce many device
structures – several examples to follow
Electron beam lithography
• Technique used to form very high
resolution patterns (down to a few
nm) in thin layer of polymer resist
• Patterns are transferred to sample
by etching or deposition of metals or
insulators followed by lift-off
• Widely used technique with many
applications in research as well as
the real world.
A micrograph of a grid formed in ZnO

demonstrating a sub-10nm resolution
A SAW driven electron interferometer
QCMP Lent/Easter 2019 Images from GAC Jones and J Griffiths 17.4
Molecular beam epitaxy (MBE) Growth Chamber
• MBE system capable of sub-
monolayer precision growth
of ultra-high purity III-V
semiconductors
• System comprises ultra-high
vacuum growth, preparation
and loadlock chambers made
from stainless steel and using
copper gasket seals
• Ion, turbo and cryo-pumps
• LN2 cryoshield (400L/day)
• System baked at 200oC to
remove impurities
• Ga, Al, As, In, Si, C sources
VG Semicon V80H
• Shutters control desposition
• Ultra-high vacuum: 10-11 mbar total – mostly hydrogen
• Pressure of impurities:10–15 mbar – around 1 part in 1010 in crystal
• Growth of thick layers to bury contamination – up to 6 months to reach
highest purity.
VEECO GEN III MBE growth chamber
• MBE growth system used in Semiconductor Physics research group for

production of ultra-high purity low-dimensional structures including, high
mobility 2D electron and hole gases as well as self-assembled quantum dots
Reflection high energy electron diffraction (RHEED)
• RHEED technique used to measure surface structure of growing crystal
[ll0] Azimuth (2x) [Īl0] Azimuth (x4)
I Farrer

RHEED Oscillations
• Observation of semiconductor growth monolayer by monolayer using
measurement of intensity of specularly reflected electrons
8s
GaAs 1µm/hr
2x4 pattern (110) direction
AlAs 0.5µm/hr
AlGaAs
1.5µm/hr

Transmission electron microscopy (TEM) of
GaAs/AlGaAs Epilayers
• Provides very high resolution images of epitaxially grown structures
GaAs
AlGaAs
GaAs
2DEG
GaAs
2.5nm AlGaAs
/2.5nm GaAs
AlGaAs superlattice
High Resolution TEM GaAs/AlGaAs TEM of GaAs/AlGaAs 2DEG

interface (T Walther, Materials Science) structure with superlattice buffer
(W M Stobbs, Materials Science)
Two-dimensional electron gas (2DEG)
Dingle et al APL 33, 665 (1978), Stormer et al SSC 29, 705 (1979)
2DEG mobility in modulation

doped structures
QCMP Lent/Easter 2019 Pfeiffer et al APL 55,1888 (1989) 17.10

Two-dimensional electron gas (1)
• Density of states in two dimensions
• Electrons fill states up to the Fermi
energy. 2π / L
• In 2D the density of states is calculated:
• Consider a lattice of points in k-space,
• 2D gas L × L in size. δk
ky
• Spacing between points is 2π / L ,area k
in k-space occupied by each mode:
.( 2π / L )
2
kx
• If states filled to k= k x + k y , area in k-
2 2 2
space of occupied states between k

and k + δ k is 2π kδ k .
Area = ( )
• No. of occupied states 2π 2
k= k + k
2 2
x
2
y L
• in this region:
2π kδ k kδ kL2
=δn =
( 2π / L ) 2π
2

Two-dimensional electron gas (2) eB
• From the last slide no. of states between and k + δ k
k ω L =
2m
2π kδ k kδ kL 2
dn kL δ k
2
= δn = hence =
( 2π / L )
2
2π dE 2π δ E
• For a free electron: E =  k / 2m so
∂k m dn L2
m
= 2 ⇒ =
2 2
∂E  k dE 2π  2
• Hence per unit area, with 2 ∂n m
a factor of 2 for spin, =
g (E) =
the density of states is: L2
∂E π  2
(independent of E )
• Application of a magnetic field in

the z-direction.
2ωL
• Constant density of states splits
into pairs of Landau levels.
• Difference in average energy of
adjacent pairs: ∆E =2 ωL g ωL
E
• Pair splitting due to spin- ∆Es =g ω L B=0
determined by electron g-factor: B=B
• As magnetic field increases , energy splitting between Landau levels
increases: eB
E 2ω=
∆= L (ω=
L eB / 2m )
m
• As field increases highest Landau levels become depopulated one-
by-one and the electrons distributed to other levels.
• If occupation of a Landau level per unit area is: nL
• Taking into account spin degeneracy the average density of states in
presence of a field is 2nL / 2ω L = nL / ω L
• Equating this to density of states at B = 0 : g ( E )  m / 
2
mωL eB
=
nL =
π  2π 
• If there are υ filled Landau levels at a field B1 the total density of
electrons per unit area is given by:
υ eB1
ne =
2π 
• Suppose there are υ occupied Landau levels: ne = υ eB1 / 2π 
• If the field is increased from B1 to B2 and the highest Landau level is
depopulated.
• The electrons are re-distributed among υ − 1 levels.
eB1 eB
• Hence: ne     1 2
2 2
• On eliminatingυ : 1
e  1 1 
ne    
2  B1 B2 
1
• This means that the depopulation of the Landau levels is periodic in
B
• At low temperatures where KT  ω the depopulation of the Landau levels
is seen in the resistance of a high mobility 2D electron gas.
• Fermi level between Landau levels - behaves like an insulator
• Fermi level in a Landau level – behaves like a conductor.
1
• Resistance oscillates with - Shubnikov-de-Haas effect.
B 17.14
Shubnikov-de-Haas effect
800
• Oscillations in resistance
υ=
700
• of a high mobility 2D 600 Landau Level no. T=300mK
• electron gas. 500
10 7
400
• Periodic in 1/B ↓ ↓
300
• We can calculate 200

8
6 5 4 3 2
electron density 100
0
−1
• 10 oscillations in 1.05T 0 1 2 3 4 5 6 7 8 9
ρ ( Ω / ) B(T)
• Each oscillation is two 800
levels - no spin splitting 700

V
(yet) 600 10 osc.
1 1 0.105 1 500 No spin splitting
  T 400
I
B1 B2 2
300
1
e 1 
1 200
ne    
2  B1 B2  100
 ne  4.601015 m2
0.0 0.5 1.0 1.5 2.0 2.5 3.0
1/B (1/T)
Quantum Hall effect
• Measure voltage perpendicular to
current - Hall effect : V 1
R=
h =
• ne density/unit area IB ne e
• In high mobility 2D electron gas

Hall voltage deviates from straight I
V
line forming plateaux when there
are full Landau levels - where:
V 2π  25812.807
= = Ω
I υe 2
υ
• with υ being the no. of filled
Landau levels at the plateaux T=1.5K
• When Rh on plateau resistance → 0

• Since 1990 used as standard of
resistance (5 parts in 108)
• Effect very insensitive to sample
disorder, astonishing accuracy for
solid state experiment, Nobel prize
1985 awarded to von Klitzing
QCMP Lent/Easter 2019 von Klitzing, Dorda and Pepper Phys Rev Lett. 45, 494 (1980) 17.16
Quantum Hall effect (2)
• Best explanation uses edge states
• In a high magnetic field classically
electrons will move in circles and
EF
‘skip’ along the edge of the sample.
• Landau Level picture – levels rise at
edges of sample forming edge states
• With υ full Landau levels only υ edge Skipping
Landau Levels
orbits
states contribute to conduction.
• Regard edge states as 1D conductors a b
∂I e2 e2 e2 edge states
g= = υ ⇒ I1 = υ V1 , I2 = υ V2
∂V 2π  2π  2π 
• With net current flowing into sample: I1,V1
f I ,V c
e 2 I 2 2 I
⇒ I = I1 − I 2 = υ (V1 − V2 )
2π 
V − V V − Ve 2π  25812.807
⇒ 1 2 =a = 2 = Ω e d
I1 − I 2 I υe υ
• With this picture quantum Hall plateaux very narrow - to observe them
we must introduce disorder and localize some electrons!
Fractional Quantum Hall Effect
• Observed in very high
mobility 2D electron and
hole gases Willett et al PRL 59,1776 (1987)
• Collective behaviour with
electron gas condensing
into a liquid state
• Most useful explanation:
composite fermion theory
• Vortices (flux quanta)
captured by each electron
forming quasi-particles
• Renormalises magnetic
field with SdH oscillations
developing at higher and
lower magnetic fields
µ=1.6x106cm2V-1s-1
• Fractional states develop n= 3.0x1011cm-2
in Hall Voltage T=150mK
1998 Nobel Prize: Laughlin Stormer and Tsui

Quantized Conductance in 1D
• 1D potential well defined in 2D gas by surface ‘gate’ fingers with
negative (-1V) voltage w.r.t electron gas
-1V −∆V
• ∆V ≈ −100 µV electrons travel
from 2D(ii) to 1D to 2D(i) region. 0V
2D(ii)
1D
• In 1D region electrons travel in 2D(i)
different energy levels in -1V
parabolic potential well – no
• Making gate voltage more negative
scattering occurs between levels
decreases well width and increases
at low temperatures (<1.5K)
level spacing – depopulating energy
levels one by one.
Number of occupied subbands
Waveguide:
Fermi 2 lateral modes
energy 1 vertical mode
4 3 2 1
QCMP Lent/Easter 2019 Vg more negative® 17.19
Quantized Conductance in 1D (2)
1D 2D
• Each 1D level acts as a
2D µf
waveguide µb ∆V
• If a voltage of ∆V ≈ 100 µV is
applied then a current starts to
1D levels
flow in each 1D channel.
• We assume T  0 K . Spin degeneracy
µf µf
 1 dE   2 dk  2e 2
I = ∫ e vg
dn
dE = ∫ e    dE =
 dk   2π dE 
2e
2π 
( µ f − µb ) = ∆V
µb 
µb
dE h
So for υ filled levels conductance: Factor of 2 for current

direction
I 2e 2
g = υ
=
∆V h Length in k-space Density of states
• Result only dependant on N=

2k
=
kL k
⇒n= ⇒
dn 1 dk
=
fundamental constants - effects ( 2π / L ) π π dE π dE
of velocity and DoS cancel. Spacing of states

Quantized Conductance in 1D (3)
• Making gate voltage more negative decreases E / ω
ψ ( x)
2
well width and increases level spacing –

depopulating energy levels one by one.
-1V
−∆V
0V x
2D(ii)
1D
2D(i)
-1V
Thomas et al APL
67, 109 (1995)
Wharam et al J
Phys C 21, L209
(1988)

• Band structure engineering, use of electron beam lithography and molecular

beam epitaxy
• Two-dimensional (2D) electron gas – density of states splitting into Landau
levels with increasing magnetic field.
• Depopulation of Landau levels with increasing magnetic field, oscillations in
resistance of electron gas – Shubnikov-de Haas effect. Measurement of 2D
electron density.
• The quantum Hall effect – observed in 2D electron gas, plateau in Hall
voltage when Landau level is full, provides resistance standard accurate to 5
parts in 108.
• Fractional quantum Hall effect
• Energy levels in a one-dimensional wire
• Quantised conductance through a one-dimensional wire

Lecture 17
Inbox (22,028)
The End

Lecture 18
David Ritchie


….Photovoltaic solar cell; Shockley-Queisser limit, efficiencies,

quantum Hall effect, conductance quantisation in 1D. Single electron
pumping and current quantisation, single and entangled-photon
emission, quantum cascade laser.


The Single-Electron Pump
• Single-electron pump - device that transfers
one electron per cycle at a well known
frequency
• Can be used to generate a quantized
current
• First realised using arrays of Al/Al-oxide
tunnel junctions between metallic islands
Geerligs et al, PRL 64, 2691 (1990)
• Frequency can be measured very
accurately – traceable back to caesium
atomic clock f
• Fundamental electron charge is well known
e = 1.602176565x10-19 C ±0.022 ppm
• Can potentially be used for a new quantum dQ
current standard for the Ampère I = ⇒ I pump =
e× f
dt
Single-electron pump slides – Dr Joanna Waldie Can also pump 𝑛𝑛 electrons: 𝐼𝐼𝑝𝑝 = 𝑛𝑛𝑛𝑛𝑛𝑛
Why Do We Need a New Ampère?
• S.I. definition of the Ampère: • In practise we realise the Ampère using
• The constant current which, if quantum resistance and voltage standards
maintained in two straight parallel combined by Ohm’s law
conductors of infinite length, of
negligible circular cross-section, Quantum Hall
and placed 1 m apart in vacuum, resistance
would produce between these
conductors a force equal to 2×10−7 RH=
h 1
×
newtons per metre of length. e2 n
Best accuracy: a few parts in 107
Josephson
Voltage
h
V=
j × nf
2e
We need a new standard for the Ampère based on fundamental constants

Semiconductor Quantum Dot Electron Pump
• Uses a quantum dot defined by • Electron Pumping Mechanism
gates in a GaAs/AlGaAs 2DEG
VG1 (dc)
f ~ 1GHz
VG2 (dc)
entrance exit
Ip
Quantum Dot
~ 100 nm
Blumenthal et al, Nature Phys 3, 343 (2007)

A Kaestner et al, APL 92, 192106 (2008)

Semiconductor Quantum Dot Electron Pump
• Uses a quantum dot defined by • Electron Pumping Mechanism
gates in a GaAs/AlGaAs 2DEG
VG1 (dc)
f ~ 1GHz
VG2 (dc)
Ejection to drain
entrance exit Single electron

in dot
Back-tunnelling
Ip
Electron capture
Quantum Dot from Fermi sea
~ 100 nm
Blumenthal et al, Nature Phys 3, 343 (2007)

A Kaestner et al, APL 92, 192106 (2008)

Current Quantization
• Origin of quantization accuracy • Pumped current at f = 120 MHz
• Quantized energy levels in dot
implies large separation of tunnelling 𝑰𝑰𝒑𝒑 = 𝒆𝒆𝒆𝒆 ≈ 𝟏𝟏𝟏𝟏 pA
rates back to source reservoir
Γ3 e-
Γ2
Γ1
Fermi
level
VG1(t) VG2(fixed)
entrance exit
Kashcheyevs + Kaestner PRL 104, 186805 (2010)

Current pump – accuracy
• To measure the error rate in ppm:
∆𝑰𝑰𝒑𝒑 = 𝟏𝟏𝟏𝟏𝟔𝟔 ×
𝑰𝑰𝒑𝒑 −𝒆𝒆𝒆𝒆 Average ∆𝐼𝐼𝑝𝑝 (-1.19±1.4) ppm
𝒆𝒆𝒆𝒆
• Gives a value of 1.2ppm
• Quantisation accuracy improved
by:
(a) Low temperature (< 1 K)
(b) Small dot – large energy level
spacing 𝐼𝐼𝑝𝑝 = 𝑒𝑒𝑒𝑒 ≈ 151 pA
(c) Perpendicular magnetic field
(d) Custom waveforms to drive
pumping, designed to avoid
B = 14 T, T = 300 mK
excited states in dot
• UK National Physical Laboratory Giblin et al Nature Comms 3, 930 (2012)
most accurate semiconductor
pump so far: 150pA with error
rate 1.2ppm
But to redefine the Ampère need < 0.1 ppm!!

Single Photon Source
• Recent developments in semiconductor Atomically resolved structure
growth technology have enabled the of InAs quantum dots by STM
production of single photon sources
based on self-assembled InAs quantum
dots.
• InAs has a lattice constant 7% greater
than GaAs.
• InAs grown on GaAs wets surface and
forms 3D islands.
• Burying islands in GaAs changes
properties due to interdiffusion. J. Márquez et al, Appl. Phys. Lett.
• Dot diameter 10-25nm, height 4-8nm, 78, 2309 (2001)
density 109-1011cm-2.
electron
• Dots coherently strained – display good
photoluminescence and electrical
properties. n-type InAs p-type
• Dots tend to store electrons and holes GaAs GaAs
due to narrow InAs bandgap.
hole
Single Photon Source (2)
Yuan et al Science 295, 102 (2002)
• Layered structure made from GaAs,
grown by molecular beam epitaxy
(MBE).
• GaAs p-type/ undoped/n-type
structure with InAs self assembled
quantum dots in undoped region.
• Low dot density 5x108cm-2 and
micron sized aperture ensures
optical emission observed from only
a few quantum dots.
electron
n-type p-type
GaAs GaAs • InAs dots have a narrower bandgap than
GaAs – electrons and holes collect in
hole them and recombine emitting a photon.
• Near ideal diode IV
characteristics with turn on at
1.5V.
• Luminescence emitted due to
one electron and one hole
recombining in single dot giving
out photon with well defined
energy.
• Emission from both ‘exciton’
(1 electron and 1 hole) or ‘bi-
exciton’ (2 electrons and 2
holes) states exciton
• Expected variation of intensity

with current for both exciton and
bi-exciton
biexciton – similar to
dependence on laser power for
photoluminescence.
• ‘Hanbury-Brown and
2D ‘wetting layer’(WL)
Twiss’ setup for
photon pair correlation
measurements.
• Beam splitter, two
spectrometers and exciton recombination
photon counting Si
avalanche
photodiodes.
• Time delay between
photons recorded.
• Sample driven with 1.5V dc plus square pulses of 0.4ns width repeated
every 12.5ns (80MHz).
• Exciton lifetime =1.3ns > pulse width - only one exciton created per
pulse.
• Low correlation between two photon events for dot (in contrast to WL).
• Single photons emitted – uses in quantum cryptography and computing.
Entangled photon-pair sources
• Polarization entangled pair photon sources Biexciton State
from biexciton/exciton cascade wavefunction:
• Non - zero ‘fine structure’ splitting of exciton So Exciton

level S due to dot anisotropy enables ‘which States
path information’ and suppresses entanglement

• Solved by (1) picking the right dot (2) using Ground State
annealing (3) in plane magnetic field
• Correlation measurements between
polarization of exciton and biexciton photons –
require good correlations in rectilinear, diagonal
bases, anti-correlation in circular basis states
• Best fidelity to ideal state >90%
• Both optically and electrically driven entangled
photon sources available
• Applications in quantum teleportation and
quantum relays for quantum key distribution,
optical quantum computing. Stevenson et al, Nature 439, 179-182 (2006)
QCMP Lent/Easter 2019 Salter et al Nature, 465, 594–597 (2010).18.13
Quantum Cascade Laser
• Most semiconductor lasers consist of
forward biased pn junctions where
electrons recombine with holes to emit
light.
• The quantum cascade laser however
makes use of intersubband transitions
in quantum wells. unbiased structure
• One difficulty is in pumping the
system, this is solved by electrically
biasing the structure which allows
electrons to tunnel from one well to e- 2
the next.
1 2
• This is a 4 level laser with the lowest 0
level of one well forming the upper -ve 3
1 e-
level of the next.
0
• The transition between levels 1 and 0
is phonon assisted – to make it much biased structure
+ve
faster than that between 2 and 1.

Quantum cascade laser (2)
• New type of semiconductor laser –
invented at Bell laboratories in 19941
• Semiconductor structure - a sequence
of quantum wells and barriers 45nm
Sequence is repeated many ( ≈ 80)
times - up to 800 individual layers
• In each sequence electrons undergo a
transition between subbands of the
quantum well, emitting a photon
• With 80 or more of these quantum
wells in a laser each electron causes
the emission of 80 photons – highly
efficient - in conventional
semiconductor (diode) lasers each
electron causes emission of 1 photon
• Wavelength 3-27µ m (infrared) powers
up to 1W (pulsed) at room temp
1J Faist et al Science 264,553 (1994)
THz Quantum Cascade Laser (1)
• Since 2002 rapid progress Injection barrier
in development of cascade
lasers at THz frequencies –
wavelengths up to 150µm
E (eV)
• GaAs/AlGaAs quantum
well structure – energy E(eV)
level separation controlled
by quantum well width
• Control of barrier widths
(a)
and tunnelling time gives
correct lifetimes for states –
allows population inversion 100 repeats
Kohler et al Nature 417(2002) 156-159
• Quantum well structure
repeated 100 times - 1
electron passing through • Cascade lasers require waveguide
device produces 100 structure to confine radiation and
photons –high efficiency generate optical mode structure
THz Quantum Cascade Laser (2) – Device structure
• Multilayered active region etched into waveguide structure
• Ohmic contacts made to top and bottom of device
Top ohmic
• Care taken with heat dissipation
contact and
waveguide
cladding layer
Bottom ohmic
contact
n+ doped GaAs
SI GaAs substrate
Active region and waveguide Figure:J.Alton

THz Quantum Cascade Laser (3)
• Performance
Driven by: 200ns pulses, 2s
intervals
2mW output power at 4K and
4.4THz, 0.1mW at 40K,
maximum operating T=50K
• Applications:
THz spectroscopy
Imaging - security and
pharmaceuticals
Gas sensing – emissions
Astronomy- Local oscillators
Quantum control experiments
• Current state-of-the-art - Koher et al Nature 417,156 (2002)

200K operation at 3.2Thz/94 µm, Fathololoumi et al Opt Ex 20, 3866 (2012),
1W emission at 3.4THz and T=10K, Li et al Elec Lett 50, 309 (2014)
Quantum Cascade Lasers – wavelength coverage
Atmospheric windows
CW InP
450
Pulsed
Temperature (K)
CW GaAs
Pulsed
300
•Very wide range of
wavelengths available
due to the use of
150 ‘band-gap engineering’
of multilayered
structures to define
0 energy levels.
2 5 10 20 50 100
Wavelength (µm)
Figure: J Faist, University of Neuchatel

• Single electron pumping with application to a current standard

• Single photon sources, made from InAs quantum dots
• Quantum cascade laser

Lecture 18
The End

Lecture 19
David Ritchie

phase transitions.

Electronic instabilities
• Crystal structure of solids much more complex
than expected
• Few solids are simple close packed structures
• e.g. Ga metal has several different phases as a
function of temperature and pressure
• Se crystallises in a structure which is an array of
spiral chains with 3 atoms per unit cell
• As, Sb, Bi, have puckered sheets where each
atom has 3 nearest neighbours
• Due to chemical bonding and balance of forces
• Fundamental principle of bonding – by placing As
chemical potential in a gap, occupied states
lowered in energy (and unoccupied states go up)
• Getting chemical potential to lie in a gap involves
the Brillouin zone boundary being in the right Se
place – at a momentum containing exactly the
correct number of states to account for all the
electrons in a solid
The Peierls Transition
• Consider a 1D chain of atoms
• Lattice constant a , electron density chosen so the Fermi wavevector k f lies
in the middle of the band - a metal
• We can turn this metal into an insulator by applying an external potential
with periodicity 2π / Q where Q = 2k f
• From previous lectures we know a periodic potential V0 cos Qx produces
Bragg scattering at a wavevector Q / 2 - a new Brillouin zone boundary
• If Q / 2 = k f there is an energy gap induced on the Fermi surface
• Rather than applying an external potential we can get the same effect by
making a periodic lattice distortion (PLD) with the same periodicity
• So move the n atom in the chain to a new position R= na + u0 cos(Qna )
th
n
• Assume the PLD amplitude is small u0  a
• We have already met this situation in the diatomic chain……
• If the atoms have a PLD with periodicity 2π / Q they will also produce a
new potential with the same period which is seen by the electrons
• The amplitude of the Fourier components is proportional to the displacement
and we can write VQ = g Q u0 with g Q the electron-phonon coupling constant
• Energy gap on the zone boundary is VQ
1D chain
• Hence an energy level at a momentum
just below k f is lowered by an energy
proportional to atomic displacement u0
- the unoccupied level just above k f is
raised by the same amount
• Period chosen to be 2π / 2k f so that a
band gap of 2 g Q u0 is introduced at
exactly the chemical potential
• Overall there is an energy lowering as a 1D chain with lattice modulation
result of the PLD, the magnitude can be
calculated (see problem 4.4) by adding
up the energy changes of all occupied
states giving in the limit u0 / a  1
where A is a constant:
Eelec = A(u0 / a ) 2 ln u0 / a
• The ln varies a bit faster than the
square, it is negative, the energy goes
down with the distortion
• So by an extension of the standard band structure result there is an
electronic charge modulation accompanying the periodic lattice distortion
• This is usually called a charge density wave (CDW)
• The result Eelec = A(u0 / a ) ln u0 / a is just the electronic contribution to
2
the energy from states close to the Fermi surface

• We can model other interactions between atoms as springs and we can add
an elastic energy of the form EElas = K (u0 / a ) giving a potential of the
2
form:
=
E ( x) Ax ln x + Bx
2 2
• This always has a minimum at

non-zero displacement
• The system lowers its energy by
distorting to produce a PLD and
CDW with a period determined
by the Fermi wave vector: 2π / 2k f
• This spontaneous lattice
distortion is a broken symmetry
phase transition named after its
discoverer Rudolf Peierls
Charge density waves
• Materials that are strongly anisotropic in electronic structures are prone to
spontaneous lattice formation and accompanying charge density wave
• Phase transitions occur on lowering T, Real space
corresponding to onset of ordering
• Can be monitored by measuring Bragg
peaks in crystal structure using electron,
neutron or x-ray scattering
• Figures show electron diffraction images
from CDW in La0.29Ca0.71Mn03
• Top figure – real space shows short scale
atomic lattice with periodic modulation K-space
• Bottom figure Fourier transform – widely
spaced bright peaks from small unit cell
• Less intense peaks from CDW
• Two periods not related since CDW
period determined by fermi surface size
and shape, depends on electron
concentration
Image from J Loudon, PA Midgeley, ND Mathur
Charge density waves
• The onset of a charge density wave can be seen in the phonon spectrum
• The coefficient of the quadratic term of energy as a function of displacement
gives the phonon stiffness for a mode of wavevector 2k f
• The onset of CDW is when the stiffness becomes zero (-ve below transition)
• At this point there is no restoring force and the phonon spectrum ω ( q )
shows a sharp minimum close to q = 2k f
• Figure shows Phonon dispersion
curves (measured by inelastic neutron
scattering at room T) for quasi 1D
organic compound TTF-TCNQ
(tetrathiofulvalene tetracyanoquinone)
• Dispersion curve is along the direction
of the chains in which there is a soft
phonon that turns into a periodic lattice
distortion at low temperature.
• Figure also shows many non-
dispersive optical modes
H A Mook and CR Watson Phys Rev
QCMP Lent/Easter 2019 Lett 36, 801 (1976) 19.8
Magnetism
• Defined as capacity of materials to change local magnetic field
• Not possible for classical systems in thermal equilibrium to be magnetic:
Bohr-van Leeuwen theorem – see Feynman lectures volume 2
• All magnetic phenomena rooted in protection of orbital or spin angular
moment afforded by quantum mechanics
• Distinguish between materials which are diamagnetic, paramagnetic or
magnetically ordered
• Diamagnetic materials - magnetism induced is opposite to applied field, a
relative permeability µ < 1 and magnetic susceptibility χ = µ − 1 < 0
• Due to motion of charged quantum mechanical particles, a weak effect,
negligible in many materials exhibiting paramagnetism or magnetic ordering
Examples: Bismuth (χ = -16.6x10-5), copper (χ = -1x10-5), water (-9.1x10-6),
superconductors (-1) ‘perfect’ diamagnets
• Diamagnets can be levitated and held in a stable
equilibrium in a strong magnet with no power
consumption, total (magnetic+gravitational) energy
having a minimum.
• Experiment – high water content allow a frog to be
levitated in a 16T field
QCMP Lent/Easter 2019 M V Berry and A K Geim Eur J Phys 18, 307 (1997) 19.9
Magnetism
• Paramagnetic materials - magnetism induced is in
direction of applied field
• Observed in systems with partially filled atomic
shells or unpaired electrons. Orbital and spin
angular momentum give rise to paramagnetic
response significantly greater than any diamagnetic
effect due to paired electrons.
• Permeability µ > 1 Magnetic susceptibility χ > 0
• Paramagnetic effects quite small 10 > χ > 10
−1 −6 wikipedia
• Examples: Oxygen, sodium (χ = 7.2x10-6), Motion of liquid oxygen in

aluminium (χ = 2.2x10-5), calcium, uranium. a magnetic field
• As temperature is reduced in a paramagnet, below a critical value

magnetic dipoles may be ordered
• Ferromagnetism – dipoles aligned giving a permanent magnetic field
• Antiferromagnetism – dipoles anti-aligned
• Magnetic ordering due to quantum mechanical exchange interaction -
magnetic dipole-dipole interaction much smaller
Magnetic moments - paramagnetism
• A semiclassical description of paramagnetism (corresponds to J = ∞ )
• Consider magnetic moments μ at an angle θ to the applied field
• Energy E =−μ ⋅ B =− µ B cos θ Boltzmann factor exp ( µ B cos θ / k BT )
• Probability of angle between θ and θ + dθ is proportional to
exp ( µ B cos θ / k BT ) × 12 sin θ dθ
• Average moment given by:
π
µ cos θ exp ( µ B cos θ / k T ) sin θ dθ
∫=
1
µz 0 B
1
2
µ
∫−1
x exp( yx)dx
π
∫ exp ( µ B cosθ / k T ) sin θ dθ
1
0 B
1
2 ∫ exp( yx)dx
−1
• Where y µ= B / k BT , x cos θ
M / Ms
• This gives the Langevin function
µz 1 M
= coth y −= µB
µ y Ms
• The saturation magnetisation M s = nµ k BT
occurs when all n moments per unit
volume are aligned which will happen
when µ B / k BT >> 1
1 y 1 y
• As = y µ B / k BT → 0, coth y  + ⇒ coth y − 
y 3 y 3
M µz y µB
= =
• Hence 
Ms µ 3 3k BT
• Since in small magnetic fields
χ=
M / H ≈ µ 0 M / B, nµ
Ms =
nµ 0 µ 2
• We obtain χ =
3k BT
• This is known as the Curie law of
susceptibility, note the inverse
temperature dependence.
• Figure shows χ plotted against
−1
temperature measured for paramagnetic

ions in a gadolinium salt
Gd(C2H3SO4).9H2O
• The straight line is the Curie law
QCMP Lent/Easter 2019 Jackson and Omnes, Leiden comm. 168a 19.12
• Calculation repeated for quantum system
• Classical moments replaced by quantum spins with J = 2 ⇒ mJ =
1
± 12
• Magnetic moments µ B , − µ B corresponding energies − µ B B, µ B B
assuming g = 2 hence:
− µ B exp ( µ B B / k BT ) + µ B exp ( − µ B B / k BT )  µB B 
g µ B mJ = µ B tanh  
exp ( µ B B / k BT ) + exp ( − µ B B / k BT )  B 
k T
M mJ
• If y = µ B B / k BT = g µ B JB / k BT ⇒ = = tanh y
Ms J
• In small fields and with M / Ms
 µB  µB nµ0 µ B2
tanh  ≈ ⇒χ=
 k BT  k BT k BT µB
k BT
• Can be extended to the general
case where J can take any integer
or half integer value.
• Can also be derived using free energy – see problem sheet 4 Q6
• We have looked at the semiclassical Langevin theory for J = ∞ and the
case for J = 12 . The general solution is known as the Brillouin function
BJ ( y ) =
M / Ms = 2 J +1
2J coth ( 2 J +1
2J y ) − 1
2 J coth ( ),
y
2J ng J µ B J
Ms =
M / Ms
Gd3+ J=7/2
1 B1/2
B2
Fe3+ J=5/2
B∞ = L( y )
Cr3+ J=3/2
0
y = µ B / k BT 10
QCMP Lent/Easter 2019 W E Henry Phys Rev 88, 559 (1952) 19.14
Types of magnetic interactions
• In many materials a finite magnetism is often observed in absence of
magnetic field
• Must be produced by interactions coupling to electron magnetic moment
• First idea – moments couple through dipole magnetic fields
• Interaction energy of two magnetic dipoles is of order µo m 2 / 4π r 3
• Using a magnetic moment of order a Bohr magnetron:
µo  e  1
2 3
2  aBohr 
U dipolar ≈   ≈ πα   Ryd.
4π  2m  r 3  r 
• Where α ≈ 1 / 137 is the fine structure constant
−5
• At typical atomic separations of 2nm this is about 4 × 10 eV
corresponding to a temperature of less than 1K
• This is far too small to explain ordering magnetic temperatures for Co -
1388K, Fe - 1043K, Ni - 627K
• The real explanation revolves around the symmetry of wavefunctions and
the Pauli exclusion principle, the large energies are due to the Coulomb
interaction between electrons

Types of magnetic interactions

• Peirels transition
• Charge density waves
• Magnetism
• Local magnetic moments and paramagnetism
• Curie Law
• Types of magnetic interactions

Lecture 19
The End

Lecture 20
David Ritchie

phase transitions.

Types of magnetic interactions – Direct exchange
• Consider two electrons in two orthogonal orbitals a , b which are
eigenstates of the single-particle Hamiltonian Ĥ 0
• The electrons are indistinguishable and the two-body wavefunction is
antisymmetric under particle exchange
Ψ (r1 , r2 ) = −Ψ (r2 , r1 )
• The two-body antisymmetric and symmetric spatial wavefunctions can be
represented by
Ψ (r=
1
1 , r2 ) ( ab − ba ) ,
1
Ψ (r=
1 , r2 ) ( ab + ba )
2 2
• Where the first slot in the Dirac-ket vector is the state occupied by electron 1
and the second slot that occupied by electron 2
• Combining with symmetric and antisymmetric spin wave functions to
maintain overall antisymmetry on particle exchange gives
 ↑↑ 
 
1
2
( ab + ba ) ( ↑↓ − ↓↑ ,) 1
2
( ab − ba )  ↑↓ + | ↓↑ 
 
 ↓↓ 
 
• The state on the left is a spin singlet and the one on the right a spin triplet
Direct exchange
=
ˆ Hˆ + Hˆ ,interaction part H=
• The full Hamiltonian is H 0 1,2
ˆ
1,2 V (r1 − r2 )
• Introduce shorthand notation E0 = ab Hˆ ab = Ea + Eb + ECoul
ECoul = ab Hˆ 1,2 ab = ∫ a 1 b 2
| ψ | | ψ | −
2 2 3 3
(r ) (r ) V (r1 r2 )d r d
1 r2
∫ b 1 a 1 a 2 b 2 1 2 1 r2
ψ ψ ψ ψ
∗ ∗
Eex =
ba Hˆ 1,2 ab (r ) (r ) (r ) (r )V (r − r ) d 3
r d 3
• ECoul looks like the Coulomb repulsion between charge densities, Eex is of a
similar form but the electrons have exchanged places
• For short range interactions such as V = δ (r1 − r2 ), ECoul → Eex
• Singlet state energy is 1
Es = ab + ba H ab + ba =E0 + Eex
ˆ
2
• Triplet state energy is lower 1
Et = ab − ba Hˆ ab − ba =E0 − Eex
2
• There is a spin dependant effective interaction in this system.
• The interaction arises because the electrons are constrained to single
occupancy of the two orbitals allowing spin flips as the only degree of
freedom
Direct exchange
• This simple example reflects a general phenomenon
• The spin triplet is symmetric under electron exchange, its associated spatial
wavefunction must be antisymmetric to maintain overall antisymmetry.
• An antisymmetric spatial wavefunction must have nodes when two spatial
coordinates are equal i.e. ψ (...., ri = r ,...rj = r ,...) = 0
• Hence the particles stay further apart in an antisymmetrised spatial state
than in an symmetric state
• This reduces the effect of the repulsive Coulomb interaction
• Thus the combination of the Pauli principle and Coulomb repulsion means
that states with antisymmetric spatial states (symmetric spin states which
generally have high spin) have lower energy.
• When the orbitals are orthogonal, Eex is positive and the lowest energy state
contains the triplet spin state
• If the overlapping orbitals are not orthogonal – as happens for orbitals in
neighbouring atoms the interaction may be negative and the singlet spin
state lowest in energy – an example is molecular hydrogen
• The involvement of the Coulomb interaction explains the high ordering
temperatures in Ferromagnetism
Heisenberg Hamiltonian
• We can express the spin-dependant interaction between the electrons in
terms of their spin states Sˆ1 , Sˆ2
• Triplet and singlet states differ in the expectation value of the magnitude of
total spin S=ˆ Sˆ + Sˆ
1 2
• The eigenvalue for Ŝ in a spin state S is given by S ( S + 1)
2
• We can differentiate between singlet and triplet states:
( ) 3
2
Sˆ 2 =Sˆ1 + Sˆ2 =S1 + S 2 + 2 S1 ⋅ S 2 = + 2 Sˆ1 ⋅ Sˆ2
ˆ 2 ˆ 2 ˆ ˆ
2
where for a spin-half state Sî = 12 ( 12 + 1) = 34
2
• So for the singlet state S = 0 and the eigenvalue of Sˆ1 ⋅ Sˆ2 is − 34

• For the triplet state S = 1 and the eigenvalue of Sˆ1 ⋅ Sˆ2 is + 14
• We can write the spin Hamiltonian as
1
H spin =
ˆ ( Es + 3Et ) − ( Es − Et ) Sˆ1 ⋅ Sˆ2
4
• Where Es , Et are the energies of the singlet and triplet states
• Defining J = ( Es − Et ) / 2 and shifting the zero of energy: Hˆ spin =
−2 JSˆ1 ⋅ Sˆ2
• Model favours parallel spins if J > 0 , antiparallel spins if J < 0
Heisenberg Hamiltonian
• From the previous slide
Hˆ spin =
−2 JSˆ1 ⋅ Sˆ2
• We can extend this to cover a
collection of spins through the
Heisenberg model
−∑ J ij Sî ⋅ Sˆ j
Hˆ Heisenberg =
ij
• This takes into account pairwise

interactions between the total spins (a) (b)
on neighbouring atomic sites Sî , Sˆ j
• The coupling only depends on the The diagram above shows the
relative orientation of the spins and not ground state of the Heisenberg
model for:
their direction relative to the crystal
(a) A ferromagnet with aligned spins
• If the angular momentum of an ion corresponding to J ij > 0
contains both orbital and spin parts, the (b) An antiferromagnet with anti-
spin Hamiltonian will include a term aligned spins on neighbouring
that depends on the absolute spin sites corresponding to J ij < 0
direction
Superexchange and insulating antiferromagnets
• With strong overlap between orbitals (covalent bond), advantageous for the
system to form fully occupied hybridised molecular orbitals. Singlet state has
far lower energy than triplet state and system has no magnetic character
• But - weaker interactions can be important e.g. 2 magnetic moments (Mn)
separated by a non magnetic ion (often O2-) in an insulator such as MnO
• Direct exchange is not important - too far apart Ground state
• Ground state - each magnetic ion is singly

occupied, non-magnetic ion doubly occupied
• If the 2 spins on O2- are anti-parallel, it is possible Mn(d-orbital) O(p-orbital) Mn(d-orbital)
for an electron to hop onto one of the Mn ions and
Excited state 1
be replaced by an electron from the other Mn
• The excited state created can be admixed to the
initial ground state and will, by 2nd order
perturbation theory, cause the new, perturbed Mn(d-orbital) O(p-orbital) Mn(d-orbital)
ground state energy to be lowered Excited state 2
• The admixture is not possible if the 2 magnetic
moments were aligned, hence the ground state
depends on orientation of 2 magnetic moments
Superexchange and insulating antiferromagnets
• Second order perturbation theory suggests that this superexchange
interaction is of order J ~ −t 2 / U < 0 where t is the matrix element
governing hopping between the magnetic moment (Mn) and the non
magnetic ion (O). U is the Coulomb repulsion energy on the magnetic
moment
• When extended to a lattice it favours an antiferromagnetic state in which
alternate sites have antiparallel spins
• On complicated lattices, very complex arrangements of spins can result
• The magnitude of this interaction is often quite small, from a few K to a few
hundred K.
• This system will often exhibit phase transitions from a magnetic ordered
state to a disordered paramagnetic state at room temperature or below

Magnetic order in MnO – Neutron scattering
• Neutron magnetic moments can couple to atomic
magnetic moments
• Amplitude of scattering depends on direction of atomic
moment so arrangement in crystal can be determined.
• At 293K two peaks are seen due to different nuclear
scattering for Mn and O ions in FCC lattice – confirms
structure determined by x-ray diffraction
• Below 118K no change in nuclear
scattering peaks but additional peaks
due to Mn magnetic moment alignment T=80K
• In crystal (111) planes of Mn spins
point in the same direction which
reverses in adjacent (111) planes
• The magnetic unit cell is double the MnO
size of the chemical unit cell T=293K
Tc=118K
• Amplitude of peaks can be used to
determine magnetic order as a function
of temperature
QCMP Lent/Easter 2019 Shull et al Phys Rev B 83, 333 (1951) 20.10
Band magnetism in metals
• Pauli paramagnetism is the response of a metal to a magnetic field
• We have a Fermi gas with energy dispersion k in magnetic field Ba = µ0 H
• spin up and spin down will be split by the Zeemen effect
k µ B Ba ,
k ↑ =− k µ B Ba
k ↓ =+
• The chemical potential (and EF ) is the same for both spins so there is a
transfer of carriers from the minority spin band to the majority spin band
• This produces a population imbalance between the 2 spin species:
µ B Ba g v ( EF )
n↑ − n↓ =
• Where g v ( EF ) is the density of states at
the Fermi level per unit volume
• We assume that the splitting is small
enough so the DoS is constant
• Magnetisation is= M µ B (n↑ − n↓ ) and
.Ba = µ0 H giving static spin susceptibility:
M
= χ=
σ µ µ 2
0 B g ( EF )
H
• From the last slide the Pauli susceptibility is χσ = µ0 µ B g ( EF )
2
• We compare this result to the Curie law susceptibility obtained from local
moments obtained before χ (T ) = 13 µ0 µ VN kB1T
2
• The Curie law susceptibility is strongly temperature dependent

• The Pauli susceptibility is largely temperature independent – particularly at
low temperatures where the thermal broadening of the Fermi distribution is
not important
• Similarities – both depend on the square of the fluctuating moment and both
have an energy scale in the denominator – given that g ( EF )  VN E1F
• For Curie law energy scale is thermal energy k BT
• For Pauli susceptibility energy scale is the Fermi energy EF
• From a different angle – remember that when calculating the electronic
specific heat, only a fraction k BT / EF contribute to the specific heat
• Here we argue that this fraction of electrons can act like local moments,
multiplying the Curie law by the fraction k BT / EF transforms it (to a
constant of order 1) into the Pauli susceptibility for metals

• Now to include interactions
• The Stoner-Hubbard model includes an energy penalty U for lattice sites
that are doubly occupied – those holding both an up- and down-spin electron
Hˆ int = ∑ Un
sitesi
n
i↑ i↓
• Treating this interaction as a mean-field approximation we get a shift in the

energies of the two spin bands
k + U n ↓ − µ0 µ B H ,
k ↑ = k + U n ↑ + µ0 µ B H
k ↓ =
• So the presence of spin-down electrons increases the energy of spin-up
electrons in the same way as a magnetic field pointing down would
• The interactions between electrons appears in the same way as an
additional magnetic field
• This is called an exchange field, it is not physical in the sense that it could
deflect a compass needle, it is a ‘book-keeping’ device to handle the effects
of the Coulomb interaction between electrons

• Using the same approximation as before – the DoS is taken to be a
constant, we determine the average spin density
N 1
) [U (n↑ − n↓ ) + 2 µ0 µ B H ] g v ( EF )
(n↑ − n↓ =
V 2
=
• The magnetisation is M µ B (n↑ − n↓ ) which then gives us the static spin
susceptibility
µ B2 g ( EF )
χσ = µ 0
1 − Ug (2E )
F
• In this expression g is the density of states per atom, in contrast to g v = VN g

which is density of states per unit volume.
• In comparison to the non-interacting state the magnetic susceptibility is
enhanced and will diverge if U is large enough so the Stoner criterion is
satisfied:
Ug ( EF ) / 2 > 1
• This marks the onset of Ferromagnetism in this model

• The Stoner criterion for ferromagnetic order Ug ( EF ) / 2 > 1 has a
fundamental interpretation
• Because the density of states per atom is of order 1/ EF the Stoner
criterion expresses the balance between the interaction energy U and the
kinetic energy EF
• If the kinetic energy of the electrons is high, they will not form a magnetically
ordered state
• If the interaction strength is higher than the kinetic energy the electron
system can lower its energy by aligning its spins
• Variations on this criterion appear in many other areas of correlated electron
physics

Local moment magnetism in metals – indirect exchange
• In a d-band metal such as iron or f-band metals such as gadolinium or
erbium there are both localised electrons with a moment derived from the
tightly bound orbitals as well as itinerant electrons from the sp bands
• The itinerant bands are only weakly spin-polarised, if at all because the
exchange interactions are small and the kinetic energy large
• But – the itinerant electrons acquire an induced spin polarisation due to its
interaction with the core spins on one atom
• This spin polarisation can be transmitted to a neighbouring ion where it
attempts to align the neighbouring spin
• There is then an interaction between the localised electron spins which is
mediated by itinerant electrons - often called RKKY
In metals, a local moment will polarise the

conduction electron spins, producing a spin
density that decays with distance and oscillates
in sign with period 1/2kF . The interaction of the
induced spin density with a neighbouring local
moment produces the RKKY interaction.

Magnetic order and the Weiss exchange field
• An alternative approach to the problem of magnetism could start with a
phenomenological equation of state, linking magnetism and magnetic field
=H aM + bM 3
where we neglect the vector nature of H , M
• In this equation of state, a takes the role of the inverse susceptibility
χ.=
−1
dH / dM and b ensures that the magnetisation bends over towards
saturation for high fields
• For a materials to be ferromagnetic, we require a finite M, a remanent
magnetisation, even for zero H.
• This appears to be possible only if a is negative
• But - systems of non-interacting electrons do not exhibit a negative
susceptibility – the Curie law for isolated moments gives a ∝ T
• In metals the Pauli susceptibility is positive and only very weakly
temperature dependant – so we need to introduce a further term which
captures the effect of interactions between the electrons
• The simplest way to incorporate these interactions is to introduce an
exchange molecular field, h into the equation of state
H + h= aM + bM 3
Magnetic order and the Weiss exchange field
• The exchange molecular field, h is not a real magnetic field (which could
deflect a compass needle….etc)
• It represents, using a mean field technique the effect of the exchange
interaction produced by many electrons on a particular test electron
• We assume that the exchange field is proportional to the overall
magnetisation with constant of proportionality λ, this is known as the Weiss
molecular field concept
• We arrive at a feedback equation:
H + λ M = aM + bM 3
• This can be re-written in the form of the original equation of state with a
modified linear coefficient a = a − λ as:
∗
(a − λ ) M + bM 3 =
H= a∗ M + bM 3
• Although the non-interacting susceptibility is finite, interactions between the
electrons give a feedback effect which boosts the magnetic susceptibility
χ =1 / a∗ =1/ (a − λ ) =χ 0 / (1 − λχ 0 )
• Where χ 0 = 1 / a is the non-interacting susceptibility.
• This leads to a magnetic instability if λχ 0 > 1 which is a more general form
of the Stoner criterion
• Types of magnetic interactions

• Direct exchange
• Heisenberg hamiltonian
• Superexchange and insulating ferromagnets
• Band magnetism in metals
• Local moment magnetism in metals – indirect exchange
• Magnetic order and the Weiss exchange field

Lecture 20
The End

Lecture 21
David Ritchie

phase transitions.

Fermi liquid theory – the problem with the Fermi gas
• Up to now our modelling of electrons in solids has been fairly simple-minded
• We have not looked for eigenstates of many particle systems.
• We have calculated the eigenstates of single particle Hamiltonians (subject
to a periodic lattice potential, filled these states with electrons, treating our
system as a degenerate Fermi gas
• This separation of a many particle problem into single-particle states relies
on being able to separate the many-body wavefunction into an
antisymmeterised product of single particle wavefunctions
• e.g. for a two-particle wavefunction:
=
Ψ (r1 , r2 ) 1
2
(ψ a (r1 )ψ b (r2 ) −ψ a (r2 )ψ b (r1 ) )
• These product states ignore correlations between electrons – we find that
the expectation values of products such as r1r2 decompose into the
products of expectation values r1 r2
• In the presence of strong electron-electron interactions the electron motion
must be correlated
• For example in the Thomas Fermi approximation, the electrons in a Fermi
gas react to the introduction of a charged impurity so as to screen the
impurity at long distances
Fermi liquid theory – the problem with the Fermi gas
• Taking this to the next level – for any electron the other electrons in a metal
execute a correlated screening motion, reducing their density in the vicinity
of the first electron
• This will reduce the electron density on the vicinity of the first electron and
reduces the effective range of the Coulomb potential due to the first electron
and implies that the electrons undergo correlated motion…..
• But – correlations are not included in the single particle picture
• The band structure approach has been very successful explaining many
properties of materials – transport properties, thermodynamic properties
such as the heat capacity, semiconductors, quantum oscillations etc etc
• The extreme case of 3He at low temperatures is an illustration of this…..

Liquid Helium 3 – specific heat
• The isotopes of 4He (bosons) and 3He(fermions) do not solidify down to zero
temperature unless placed under significant pressure (around 33bar)
• This is caused by very weak mutual interaction and low mass boosting
quantum mechanical zero point motion.
• They are examples of quantum fluids and 3He is an uncharged analogue of
electrons in solids
• Because of hard-core repulsion at small
separation and weak Van der Waals
interaction, picture helium atoms at low T as
a close-packed assembly of hard spheres -
not a solid but clearly strongly correlated
• Measurements of all key properties show
good agreement with Fermi gas theory
• e.g. the molar heat capacity of a degenerate
fermi gas Cm / T → const at low
temperatures – just as found in 3He
• Figure shows measurements of heat capacity
of 3He at low temperatures for different
pressures Greywall Phys Rev B 27, 2747 (1983)
Liquid Helium 3 – specific heat
• Looking more closely – while the general form of the heat capacity and
other properties – thermal conductivity, viscosity are the same as a Fermi
gas the detailed prefactors may be different
• In a Fermi gas as we saw earlier in the course the specific heat
π2 T 2
EF
( 3π n )
2/3
=
cv =
nk B , TF = 2
2 TF k B 2mk B
• where TF is the Fermi temperature – which is determined by the number
density, n , and mass, m, of the particles
• The specific heat per mole, Cm is given by Cm / T  π2 TR where R is the
2
F
gas constant
• These equations predict an increase in TF with
increasing pressure as the number density n rises
• Gives a reduction in Cm / T with increasing pressure
• Opposite to dependence observed in experiments
• ‘Explained’ by change in effective mass of 3He atoms
from 2.76m3 at low pressure to 5.65m3 at 33 bar
• Changes in m3∗ with pressure larger than in dilute
mixtures of 3He in 4He (L3) – greater interactions
Liquid Helium - viscosity
• Torsional oscillator measurements of liquid
helium viscosity
• Cylindrical volume filled with liquid helium,
hanging from hollow torsion rod
• Oscillation about vertical axis, motion
damped by liquid helium
• Frequency 2kHz, background Q factor >105
• Oscillator driven by voltage applied to fixed
electrode, motion detected by measuring
voltage on second electrode (oscillator vane
at constant 400V)
• Driven at peak of resonance with constant
amplitude (10-100nm) of motion, drive
voltage required ∝ dissipation
• Experiment mounted on 3He/4He dilution
refrigerator. Measurements over a
temperature range 3mK-1K
• Period and dissipation measured as a
function of T
Liquid Helium - viscosity
• From kinetic theory, viscosity
1
η = nmv f λ 100% 3He
3
• Number density n , mass m Fermi 96&97% η∝
1
velocity v f , mean free path λ = v f τ η 3He
T2
between collisions where τ η is the
quasiparticle viscosity scattering
time 5%3He
• For gas or liquid of fermions,

scattering time
1 1
τη ∝ ⇒ η ∝
T2 T2 1.3%3He
• Correct temperature dependence
observed for pure 3He
• Mixtures of 3He and 4He don’t give
correct T dependence due to
surface effects
• Internal dimensions of oscillator  λ Ritchie et al Phys Rev Lett 59, 466 (1987)
3He – viscosity mean free path
• So what happens if we reduce
internal dimensions of oscillator?
η −1 ∝ T 2
• Height reduced from 6mm to 45µm
• Inverse viscosity plotted vs T2
• Linear as expected but deviation
from linearity at lowest T
• Passes through minimum at a
temperature of 1.26mK before
passing through superfluid
transition at 1.13mK
• Minimum in inverse viscosity is
signature of mean free path being
Knudsen minimum
equal to size of oscillator so
λ ≈ 45µ m
• Known as Knudsen minimum
• Very long mean free path given Td2 = 2.13mK 2
close packed nature of 3He atoms
QCMP Lent/Easter 2019 Parpia and Rhodes Phys Rev Lett 51, 805 (1983) 21.9
Collective excitations – remember phonons
• In a lattice we concentrate on deviations of the system from its ground state

• These excitations are deformations of the lattice, we decompose them into
normal modes labelled with a wavevector and frequency of vibration
• Since we have harmonic oscillations we can use creation and annihilation
operators to generate an excited state – ‘creating’ a phonon
• Commutation relations between creation and annihilation operators cause
phonons to follow Bose statistics
• We can create multiple excitations of the same k-state
• So a collection of atoms - which may be fermions or bosons can form a
lattice and the low energy excitations can behave like a Bose gas
• This is a tremendous simplification of original problem
Collective excitations in a Fermi liquid
• Perhaps something analogous happens when fermions form a liquid

• Could the low energy excitations of the liquid formed from strongly
interacting fermions (e.g. electrons) behave like a gas of weakly interacting
fermions?
• This would explain why a single-particle description works so well in many
materials
• In many cases the properties of the electrons making up the liquid carry over
with only slight modifications to the properties of the fermionic excitations of
the Fermi liquid
• The idea behind Landau’s Fermi liquid theory…..
Adiabatic continuity
• Imagine an assembly of non-
interacting electrons
• Electron system forms a Fermi gas,
ground state of system represented by
filled electron states within Fermi
surface, empty electron states above it
• Gradually turn on interaction between
electrons and follow evolution of
energy levels in the system
• Principle of adiabatic continuity
suggests we can continue to label the
energy eigenstates in the same way as
for non-interacting system
• When the system is tuned energy
levels shift but labels remain useful
• We assume the excitations of the
interacting Fermi system follow the Changing a square well potential into an
same basic rules as the Fermi gas SHO potential, Schofield, Contemp.
Phys. 40, 95 (1999)
Adiabatic continuity
• Important consequence of one-to-one
correspondence between quasiparticle
states and states of non-interacting
system is that the volume of the Fermi
surface is unchanged as interaction is
turned on - Luttingers’s Theorem
• There are a number of pitfalls with this
argument
• It only holds if the energy levels do not
cross as the interaction is turned on
• This is not guaranteed – difficult to find
a non-trivial example where this is true
for an interacting system
Changing a square well potential into an
• We can say that for interacting Fermi SHO potential, Schofield, Contemp.
systems, a Fermi liquid state is possible Phys. 40, 95 (1999)
but not every Fermi system will be
described in this way

Total energy expansion for Landau Fermi liquid
• Landau’s Fermi liquid: label excited states of
the interacting system by quantum nos. of
non-interacting system: wavevector k , spin,
band index, etc
• In analogy with phonons for excited states of
lattice vibrations, we refer to an excited state
quasiparticle of wavevector k
• The quasiparticle lies outside the Fermi
surface
• Also quasiholes – excitation at wavevectors
inside Fermi surface
• We can express the total energy of the
interacting system as a function of the
occupation number of the various states
• At low temperatures with few excitations this
can be approximated by a Taylor expansion
E[nk ]= ∑k  (k ) n (k ) + 2 ∑ kk ′ f (k , k ) n (k ) n (k )
1 ′ ′
Total energy expansion for Landau Fermi liquid
• From last slide
E[nk ]= ∑ k
 (k ) n (k ) + 2 ∑ kk ′ f (k , k ) n (k ) n (k )
1 ′ ′
• First term - the energy of having quasiparticles in band states of energy  (k )

• The next term is second order in occupation number n(k ) and accounts for
interactions between excited states or quasiparticles
• The interaction function is given by f (k , k ′)
• Hence by using a relatively simple expression for the total energy Landau
found a variety of expressions linking material properties such as heat
capacity, magnetic susceptibility and compressibility to properties of
quasiparticles and the interaction function

Energy dependence of the quasiparticle scattering rate
• Quasiparticle scatters by creating electron-hole pair

• Particle with energy e, momentum p comes in, 2 particles and 1 hole come
out
• By Pauli exclusion principle all outgoing particles must have energy
(referenced to EF ) >0. Remember the hole energy = −3
• By conservation of energy  = 1 + 2 − 3 is fixed
• 1 and 2 can be chosen freely from the range 0 to  which fixes 3
• Hence number of available final states for scattering event rises with
increasing incident quasiparticle energy and is ∝  2

Energy dependence of quasiparticle scattering rate
• Hence the scattering rate Γ  
2
• Therefore quasiparticles with energy  → 0 i.e. close to the Fermi surface,

scatter extremely rarely and are long-lived
• Despite the fact that they travel through a strongly interacting and dense
liquid, these quasiparticles can travel long distances before they scatter.
• Their free motion is protected by the Pauli exclusion principle because there
are very few empty states into which they can scatter
• This is a key result from Landau’s Fermi liquid theory, particles interact but
do not scatter and therefore are long-lived
• In metals this effect explains why the mean free path can be very long in
copper if the temperature is low enough
• The electrical current is carried by a quasiparticle excitation that is a
collective mode of the Fermi system
• In the language of perturbation theory the quasiparticle is a dressed
excitation involving a correlated motion of the added electron with the many
particle background

Energy dependence of quasiparticle scattering rate
• Conversely with increasing = E − EF the scattering rate grows more
quickly than 
• When the scattering rate exceeds  /  the quasiparticles are no longer well
defined because they scatter before their wavefunction can undergo a full
oscillation
• Hence the quasiparticles become overdamped
• Landau’s Fermi liquid theory only applies to low energy excitations and only
works well when k BT  EF
• This is still as very wide range – in most metals EF is thousands of Kelvin

• Fermi liquid theory – the problem with the Fermi gas

• Liquid Helium – specific heat and viscosity
• Collective excitations
• Adiabatic continuity
• Total energy expansion for Landau Fermi liquid
• Energy dependence of quasiparticle scattering rate

Lecture 21
The End

Lecture 22
David Ritchie

phase transitions.

Quasiparticles and holes live near the Fermi surface
• Fermi liquid quasiparticles
are only defined close to
Fermi surface
• We cannot have a filled
Fermi sphere of
quasiparticles as for the
Fermi gas
• For a Fermi liquid the
wavevector k labels the
quasiparticle states which
are only well defined
close to the Fermi surface
because only then is the
lifetime large enough
• We have an unusual ‘vacuum’ ground state in which the low energy
excitations are fermionic particles and holes which sit on either side of a
boundary surface in momentum space

Quasiparticle spectral function
• Think about the fermi liquid in another way - consider the response of the
many-particle system to the addition or removal of a quasiparticle
• For a non-interacting system, place a particle into a single-particle
eigenstate of the Hamiltonian labelled by momentum k
• The wavefunction evolves in time according to the Schrodinger equation
ψ k (r, t ) =
ψ k (r )e − i t / 
k
• Where ψ k is the Bloch wavefunction satisfying the time-independent

Schrodinger equation
• The time-dependent solution oscillates in time with a frequency ω = k / 
• Looking at this in Fourier space
ψ
= k (r, ω ) 2πψ k (r )δ (ω − k / )
• So the wavefunction has spectral weight only at ω= k / 
• We can say that the probability amplitude of finding an electronic state with
energy ω and momentum k is
A(k , ω ) = δ (ω − k / )
• Where A(k , ω ) is called the electron spectral function for the non-
interacting system
• In an interacting system an electron may take part in a number of
eigenstates – so we expect the spectral function to spread out in energy
• Interactions modify the precise form of the dispersion relation so we replace
.k by a renormalised k
∗
• This is often called a mass renormalisation m / m = k / k
• If we add a quasiparticle it will scatter from other quasiparticles or create
other electron-hole pairs
• Because of this the probability amplitude of finding a quasiparticle of
momentum k at time t will decay exponentially at the quasiparticle scattering
rate Γ k
( i /  −Γk ) t
• These effects change the time dependence of the state to e k
• After a Fourier transform this leads to an ansatz for the spectral function in
an interacting system of the form
1  1 
A(k , ω ) = − ℑ
π  ω − k / + iΓ(k ) 
• This describes quasiparticles with a dispersion relation k and decay rate Γ k
• Notice that if the inverse lifetime Γ k → 0 the spectral function reduces to the
non-interacting form A(k , ω ) = δ (ω − k / )
Fermi gas – non interacting
• We must take care with the chemical potential
• In equilibrium (and at T=0) we cannot add
fermionic excitation at an energy ω < µ
• So we will infer that for ω > µ the spectral
function we have is for particle like excitations but
for ω < µ it is the spectral function for holes
• If Γ k is small the quasiparticles are long-lived
and have some real meaning – a quasiparticle Fermi liquid - interacting
resonance at ω = k can be resolved if the
peak width in the spectral function is less than the Fermi liquid
centre peak’s position
• For interacting system (b) A(k , ω ) has width 2Γ k
• Phase space argument shows Γ k ∝  2 quality
factor k / Γ k ∝ k / k which diverges as  → 0
2
• Quasiparticles overdamped at high energies but well defined for E → EF

• Quasiparticle concept is self consistent – possible to have quasiparticles that
scatter so rarely so their lifetime is long enough to observe them !
Tuneability of the quasiparticle interaction
• The interaction term in the Landau expansion for the total energy
1
2 ∑ kk ′
f (k , k ′)n(k )n(k ′)
causes quasiparticle properties to be changed w.r.t. the free electron value.
∗
• Most importantly - the effective mass m can be orders of magnitude larger
than the bare electron mass
• The quasiparticle interaction f (k , k ′) can be completely different from the
Coulomb repulsion, which acts on the underlying electrons
• In particular, f (k , k ′) can be spin-dependent
• This is an example of the tuneability of correlated electron systems: although
the underlying Coulomb interaction is fixed, the effective interaction in the
low energy model is highly dependent on details of the system, and can
therefore be tuned over a wide range by changing, for example, magnetic
field, pressure or doping.

Heavy fermions
• In Fermi liquid theory, quasiparticle mass differs from bare electron mass.
• Strong interactions can cause high effective masses.
• Heavy Fermion materials (at very low temperature) have very high
Sommerfeld coefficient of the heat capacity, and high, weakly temperature-
dependent magnetic susceptibility.
• This suggests they follow Fermi liquid theory, but the effective quasiparticle
masses are strongly enhanced: in some cases up to 1000 times .
• Usually, heavy Fermion materials contain Cerium, Ytterbium or Uranium,
which contribute partially filled f-orbitals to the band structure.
• These highly localised states are important, because in a lattice, they lead to
very narrow bands. The strong Coulomb repulsion prevents double
occupancy of these states.
• There are hundreds of heavy Fermion materials
• Examples: CeCu Si , CeCu , CeCoIn , YbCu Si , UPt .

Heavy fermions – CeCu6
• Figure shows Sommerfeld coefficient
of heat capacity C / T for Heavy
Fermion material CeCu6
• In the low-T limit C / T approaches
850mJ(molK2)
• Comparing this value to pure copper,
the heat capacity per atom has been
boosted by a factor of 150
• So… replacing 14% of the copper
atoms in a sample of pure Cu by Ce
we increase the low temperature heat
capacity by a factor of 150!!!
• We find that (assuming k F the same)
∗
g ( EF ) ∝ 1 / v=
F , m vF k F  const.
∗
• Hence m ∝ g ( EF ) - so an enhanced DoS at the fermi surface crucial for
the properties of these materials
• Note strong upturn in C / T at low T
• Heavy fermion state only develops fully at low temperatures
Heavy fermions – UPt3
• At low T the behaviour of CeCu6 and
other heavy Fermion materials
consistent with Fermi liquid theory
• At high T the picture is quite different
• An isolated magnetic moment displays
a T dependent Curie law susceptibility
χ −1 ∝ T
• Similar form seen in many heavy
fermion materials at high T, often with Frings et al J. Mag Mat
140, 31 (1983)
a slope consistent with a Curie
constant expected from electronic
configuration of magnetic ions
• Figure: UPt3, electrons in partially filled f-orbitals act as local moments at
high T – and obey Curie law susceptibility.
• Curie susceptibility does not extend to zero T but crosses over to a constant
value, just as the Sommerfeld coefficient of the heat capacity does
• As T is lowered, free spins start to bind to conduction electrons to make
heavy fermions, according to the Kondo model
Renormalised band picture for heavy fermions
• To find a qualitative understanding of the origin of the heavy fermion state
• Consider the hybridisation between the bands associated with the more
extended s,p,d-orbitals and the bands which arise from the tightly localised f-
orbitals
• A partially filled f-orbital will always lie inside filled s,p, and even d-orbitals
with a higher major quantum number
• There is negligible hybridisation between f-orbitals on neighbouring atoms –
too far apart
• The result is a very flat f-band
• If we consider single-electron states naively, then we find that the f-band
formed from the atomic f-orbitals is well below the chemical potential, and
should therefore be completely full
• In such a scheme there would be no local moments at high temperature and
no heavy fermion behaviour at low temperature
• The scheme fails, because it ignores the strong Coulomb repulsion between
electrons sharing the same f-state
• Instead of this naïve picture….

Renormalised band picture for heavy fermions
• Once single electron has occupied an f-
orbital, energy cost for 2nd electron
hopping onto same orbital is very high
• Modify single particle picture to ensure
f-orbitals average occupancy is one
• Renormalise energy of the f-states
close to the chemical potential, near EF
• Renormalised, narrow f-band and
broad band from atomic s, p, and d-
orbitals hybridise, producing anti-
crossing very close to EF
• New dispersion E ( k ) crosses EF at much reduced slope compared to the
broad s, p, d-bands. As the slope dE / dk gives the Fermi velocity and is
inversely proportional to the effective mass, this scheme explains enhanced
effective masses observed in materials with partially occupied f-orbitals
• Note that k F is larger for hybridised system than it would be without
hybridisation. Volume of the Fermi surface in heavy fermion systems is
large enough to contain not only the electrons on s, p, and d-bands but also
the f-electrons
Quasi-particles detected in de Hass-van Alphen experiments
• Quantum oscillations – evidence for
heavy fermion liquid state
• Magnetic fields up to 16T
• Several fermi surface sheets
• Volume enclosed by Fermi surface is
very stringent criterion, used to
decide between the heavy Fermi
liquid scenario – in which f- electrons
contribute to the Fermi surface – and
local moment models
• Temperature dependence of
oscillation amplitude leads to
effective mass
• Measured effective masses
consistent with heat capacity results
• % reflect contributions from
quasiparticles in each sheet to
specific heat UPt3
QCMP Lent/Easter 2019 McMullan et al arXiv:0803.1155 22.13
Tuning the quasiparticle interaction
• We cannot tune the Coulomb interaction – but the effective interaction,

which governs the behaviour of a system at low temperatures, depends on
many details and is thereby highly tuneable
• By varying the effective interaction, we can select different ground states
(magnetism, superconductivity, etc.)

CePd2Si2 : heavy-fermion magnet to unconventional
superconductor
• Antiferromagnet below
TN=10K
• TN depends on pressure.
• Magnetism suppressed
near 2.8 GPa
• Anomalous resistivity
temperature dependence Antiferromagnet
• Quantum phase transition:
paramagnet
can cross phase boundary
at low T, using control
parameter other than
temperature.
Mathur et al, Nature 394, 39 superconductor

(1998)

Heavy fermion materials summary
• Usually 4f (Ce, Yb) or 5f (U) intermetallic compounds, such as UPt3 or
CeCu2Si2
• Low temperature properties consistent with Fermi liquid theory, but very high
effective carrier masses
• What is the role of electrons in partially filled f-orbitals? Do they behave like
local moments or like conduction electrons?
• Where quantum oscillation studies have been successful, they indicate that
the electrons contribute to the Fermi surface, as in a normal metal
• Because g ( EF ) is so high in these materials, they tend to order
magnetically or even become superconducting.
• There are many different ordered low temperature states in these metals,
some simple, some very exotic
• There is an increasing number of materials (e.g., YbRh2Si2 ), which do not
follow Fermi liquid theory at low T. Are they ’non-Fermi’ liquids? Do the f-
electrons remain as local moments down to absolute zero in this case, not
contributing to the Fermi surface? This is being investigated at the moment

• Quasiparticles and holes near the Fermi surface

• Quasiparticle spectral function
• Tuning of the quasiparticle interaction
• Heavy Fermions
• Renormalised band picture for heavy fermions
• Quasiparticles detected by dHvA
• Tuning the quasiparticle interaction
• CePd2Si2 : heavy-fermion magnet to unconventional superconductor
• Heavy fermion materials summary

Lecture 22
The End!!!

Quantum Condensed Matter Physics 2018-9
Lorentz dipole oscillator, optical properties of insulators. Drude model and optical properties of
metals, plasma oscillations. Semi-classical approach to electron transport in electric and magnetic
fields, the Hall effect. Sommerfeld model, density of states, specific heat of; electrons in metals,
liquid3 He/4He mixtures. Screening and the Thomas-Fermi approximation.

Types of bonding; Van der Waals, ionic, covalent. Crystal structures. Reciprocal space, x-ray
diffraction and Brillouin zones. Lattice dynamics and phonons; 1D monoatomic and diatomic chains,
3D crystals. Heat capacity due to lattice vibrations; Einstein and Debye models. Thermal conductivity
of insulators. Electrons in a periodic potential; Bloch’s theorem. Nearly free electron approximation;
plane waves and bandgaps. Tight binding approximation; linear combination of atomic orbitals,
linear chain and three dimensions, two bands. Pseudopotentials. Band structure of real materials;
properties of metals (aluminium and copper) and semiconductors. Semi-classical model of electron
dynamics in bands; Bloch oscillations, effective mass, density of states, electrons and holes in
semiconductors

Photon absorption; transition rates, experimental arrangement for absorption spectroscopy, direct
and indirect semiconductors, excitons. Quantum oscillations; de Haas-Van Alphen effect in copper
and strontium ruthenate. Photoemission; angle resolved photoemission spectroscopy (ARPES) in
GaAs and strontium ruthenate. Tunnelling; scanning tunnelling microscopy. Cyclotron resonance.
Scattering in metals; Wiedemann-Franz law, theory of electrical and thermal transport,
Matthiessen’s rule, emission and absorption of phonons. Experiments demonstrating electron-
phonon and electron–electron scattering at low temperatures.

Intrinsic semiconductors, law of mass action, doping in semiconductors, impurity ionisation,
variation of carrier concentration and mobility with temperature - impurity and phonon scattering,
Hall effect with two carrier types. Metal to semiconductor contact. p-n junction; charge
redistribution, band bending and equilibrium, balance of currents, voltage bias. Light emitting
diodes; GaN, organic. Photovoltaic solar cell; Shockley-Queisser limit, efficiencies. Field effect
transistor; JFET, MOSFET. Microelectronics and the integrated circuit. Band structure engineering;
electron beam lithography, molecular beam epitaxy. Two-dimensional electron gas, Shubnikov-de
Haas oscillations, quantum Hall effect, conductance quantisation in 1D. Single electron pumping,
single and entangled-photon emission, quantum cascade laser.

The Peierls transition, charge density waves, magnetism, local magnetic moments, Curie Law. Types
of magnetic interactions; direct exchange, Heisenberg hamiltonian, superexchange and insulating
ferromagnets, band magnetism in metals, local moment magnetism in metals, indirect exchange,
magnetic order and the Weiss exchange field.

Fermi liquid theory; the problem with the Fermi gas. Liquid Helium; specific heat and viscosity.
Collective excitations, adiabatic continuity, total energy expansion for Landau Fermi liquid, energy
dependence of quasiparticle scattering rate. Quasiparticles and holes near the Fermi surface,
quasiparticle spectral function, tuning of the quasiparticle interaction, heavy fermions, renormalised
band picture for heavy fermions, quasiparticles detected by dHvA, tuning the quasiparticle
interaction. CePd2Si2 ; heavy-fermion magnet to unconventional superconductor phase transitions.

PII - Quantum Condensed Matter Physics - Ritchie (2019) 399pg PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PII - Quantum Condensed Matter Physics - Ritchie (2019) 399pg PDF

Uploaded by

Copyright:

Available Formats

Quantum Condensed Matter Physics

QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 1.1

QCMP Lent/Easter 2019 1.2

3. Experimental probes of band structure (4L)

QCMP Lent/Easter 2019 1.3

4. Semiconductors and semiconductor devices (5L)

QCMP Lent/Easter 2019 1.4

•Course material will be useful for several part III courses.

2. Optical properties of Solids, Fox M

3. The Oxford Solid State Basics, Simon S H

4. Introduction to Solid State Physics, Kittel C

5. Solid State Physics, Ashcroft N W and Mermin N D,

QCMP Lent/Easter 2019 1.6

1. Classical and Semi-classical models for electrons in solids (3L)

QCMP Lent/Easter 2019 1.7

• Model atoms as nucleus + electron cloud.

QCMP Lent/Easter 2019 1.8

• Absorption coefficient α and

• n0 is the off resonance

Taken from Optical properties of solids by M Fox

QCMP Lent/Easter 2019 1.12

absorption line and d n / dλ > 0 below next

• For SiO2 this is at λ  1.3µ m – a preferred optical fibre wavelength

• For two sharp levels, from time-

• As energy levels broaden into bands and γ increases, this expression

QCMP Lent/Easter 2019 1.15

scattering rate. taken from wikipedia

QCMP Lent/Easter 2019 1.16

• The current density J due to electrons of number density n, mass m, of

• For momentum as before:

• Introduction to course and recommended text books

QCMP Lent/Easter 2019 1.19

QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 1.20

QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 2.1

1. Classical and Semi-classical models for electrons in solids (3L)

QCMP Lent/Easter 2019 2.2

QCMP Lent/Easter 2019 2.3

QCMP Lent/Easter 2019 2.4

QCMP Lent/Easter 2019

• |  (ω ) | diverges as ω →0 so metals are highly reflecting at low frequency

become transparent in the ultraviolet.

• Plateau in R at low frequency, related to conductivity of material

• Blue (Dotted) with=τ 8.0 ×10 s value deduced from DC conductivity

values - slightly better fit

• More generally as ω → 0 at ω = ω p∞ we get Plasma Oscillations at a

frequency defined by ω p2 = ne 2 / m0

QCMP Lent/Easter 2019 2.9

• In metals however it is found that RH

QCMP Lent/Easter 2019 2.13

• Correct description: degenerate Fermi gas (cf. Stat. Phys. course).

• The Drude model – frequency dependant conductivity

QCMP Lent/Easter 2019 2.15

QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 2.16

QCMP Lent/Easter 2019 http://www.sp.phy.cam.ac.uk/drp2/home 3.1

1. Classical and Semi-classical models for electrons in solids (3L)

QCMP Lent/Easter 2019 3.2

satisfying periodic boundary conditions ψ ( x + L, y, z=

• Factor of 2 for spin degeneracy

• The number of particles is conserved so

• The first term in the square brackets is odd, xe 2 is even so ∂µ  0

• This calculation is the leading order term in an expansion in powers of

QCMP Lent/Easter 2019 3.7

• So knowing cv , T , n we can calculate TF the Fermi temperature