An introduction to solid-state physics

International summer School in electronic structure Theory: electron

correlation in Physics and Chemistry

Centre Paul Langevin, Aussois, Savoie, France

Xavier Blase

Institut Néel, CNRS, Grenoble, France.

June 16, 2017

Bibliography

I Introduction to Solid State Physics, 8th Edition, Charles Kittel, John

Wiley and Sons Eds.
I Solid State Physics, N.W. Ashcroft and N.D. Mermin, Saunders
College Publishing.
I P. Yu and M. Cardona, Fundamentals of semiconductors, Springer.
I Ibach and Lüth, Solid-state physics, Springer (with sections devoted
to experimental techniques).

More advanced:
I Theoretical Solid State Physics, W. Jones and March, Dover Eds.
I Quantum Theory of Solids, Charles Kittel, John Wiley and Sons.
Introduction

There are typically 1022 electrons per cm3 in condensed matter systems.
This is somehow too much for considering a solid as a big molecule.
Fortunately, translation symmetries can help a lot in particular in the case
of one-body Hamiltonian of the kind met in mean-field approaches (DFT,
Hartree-Fock, tight-binding or extended-Hückel, etc.)

Figure: (Left) First silicon transistor (Bell Labs, Murray Hill, NJ) announced in
1951 by William Shockley; (Right) Silicon band structure. Shaded area
indicates energy domains with no allowed states.
The unit cell and the periodic vectors
The unit cell is a portion of space that repeated periodically can
reconstruct the entire crystal. A unit cell can contain several atoms (the
motif). The lattice vectors: R ~ijk = i~a1 + j~a2 + k~a3 allow to reconstruct
the crystal from the atoms in the unit-cell with (~a1 , ~a2 , ~a3 ) the basis
vectors. The minimum volume cell is a primitive cell.

Figure: (Left) 2D square lattice with one atom per cell. A unit-cell is shaded in
blue. (Right) 2D hexagonal cell with one atom per cell. Two different unit cells
are represented. For the blue cell, 1/4th of each connected atom belong to this
cell. The yellow one is called the Wigner-Seitz cell that is invariant with respect
to the crystal symmetry point group.
The 3D Bravais lattices and the motif
Depending on the shape of the unit-cell, one can categorize 3D
crystal under 7 different ”lattice systems”, that yield 14 Bravais
lattices depending on the disposition of atoms in the unit cell
(the motif).

For example, a cubic lattice can be ”simple”, body-centered (BCC) or

face-centered (FCC). Silicon and diamond are FCC lattices with 2 atoms per
primitive cell.

Courtesy: http : //chemwiki.ucdavis.edu/Wikitexts/UC Davis/UCD Chem 2B/UCD Chem 2B

Symmetries, commutators and quantum numbers

Reminder: in the case of a spherically symmetric potential, the

Hamiltonian commutes with the angular momentum Lz and L2 operators
(and L2 commutes with Lz ) which implies that there exists a common
basis of eigenstates:

H|ψnlm >= Enlm |ψnlm >

2
L |ψnlm >= l(l + 1)~|ψnlm >
Lz |ψnlm >= m~|ψnlm >

We know that the solutions are of the type:

ψnlm (~r ) = ψnlm (r , θ, φ) = Rnl (r )Ylm (θ, φ)

and that (nlm) are ”good” quantum numbers: the Hamiltonian H acting
on such states preserve the symmetry character of these eigenstates.
Bloch theorem in 1D: introduction

We consider now the case of a crystal

with discrete (not infinitesimal)
translation properties.

Assume that the potential is periodic: V(x+R)=V(x) (with R=na) and

call (TR ) the translation operator. Then:

TR [V (x)ψ(x)] = V (x − R)ψ(x − R) = V (x)TR ψ(x)

which means that the potential, and thus the Hamiltonian, commute
with the translation operator: [TR , H] = 0. Then quantum mechanics
says that one can find a common eigenbasis for the two operators.

H|ψk > = Ek |ψk >

TR |ψk > = Ck (R)|ψk >
Bloch theorem in 1D (II)

We can find the expression of the Ck by simple considerations. The

translation operator should preserve the normalisation of ψ:
Z Z Z Z
dx|ψ(x − R)|2 = dx|TR ψ(x)|2 = dx|C (R)|2 |ψ(x)|2 = dx|ψ(x)|2

so |C (R)|2 = 1 and C (R) = e iθ(R) . Further:

Ta Ta ψ(x) = ψ(x − 2a) = T2a ψ(x) ⇒ C (a)C (a) = C (2a)

The only mathematical function satisfying such conditions is:

C (a) = e −ika ⇒ C (2a) = C (a)C (a) and C (R = na) = e −ikR .

The quantum number (k) is associated with the translation operator.

Bloch theorem

We know therefore (generalizing to 3D) that one can find an eigenbasis

of the Hamiltonian and of translation operators such that:

~ = T ~ ψ~ (~r ) = e −i ~k·R~ ψ~ (~r ) or ψ~ (~r + R)

ψ~k (~r − R) ~ = e i ~k·R~ ψ~ (~r )
R k k k k

This is a first formulation of Bloch theorem. A second formulation comes

~
when considering the properties of u~k (~r ) = e −i k~r ψ~k (~r ):
~ ~ ~ ~ ~ ~ ~
~ = e −i k·(~r −R) e −i k·R ψ~ (~r ) = e −i k·~r ψ~ (~r ).
TR~ u~k (~r ) = e −i k·(~r −R) ψ~k (~r − R) k k

Namely, the fonction u~k (~r ) is periodic and:

~
ψ~k (~r ) = e i k~r u~k (~r ), with u~k (~r ) periodic.
Bloch states with bare hands (literally)
~
The e i k~r phase term can be regarded as an ”envelope function” that
modulates the periodic function u~k (~r ). In the 1D example here below,
assume that each atom has one (p z ) orbital. One can create different
Bloch states by changing the magnitude of the k-vector (we represent
e.g. the real part of the wavefunctions).

For the first/second Bloch state, (k = π/2a) and (k = π/4a).

The reciprocal space

The ~k-vectors are homogeneous to the inverse of a distance and lives in

the ”reciprocal space”. If (~a1 , ~a2 , ~a3 ) are the periodic vector of the
crystal, we choose to represent the ~k-vectors as a function of the
reciprocal space basis: (~b1 , ~b2 , ~b3 ) vectors such that:

~bi = 2π ~aj × ~ak

⇒ ~ai · ~bj = 2πδij
~ai · (~aj × ~ak )

Defining the reciprocal space vectors: G~ = l1 ~b1 + l2 ~b2 + l3 ~b3 , then the
~ ·~
iG r
e vectors form a basis for periodic functions since for any lattice
~ = n1 ~a1 + n2 ~a2 + n3 ~a3 ,
vector in real space R
~ ~ ~ P ~ ~ P ~
e i G ·(~r +R) = e i G ·~r +i α nα lα~aα ·bα = e i G ·~r +i α nα lα 2π = e i G ·~r
~
For example, u~k (~r ) = G~ u~k (G~ )e i G ·~r , the planewave expansion of u~k (~r ).
P
The Brillouin zone
From the definition of (G):
h i
~ ~ ~ ~ ~ ~ ~ ~
e i(k+G )·R = e i k·R and ψ~k+G~ (~r ) = e i k·~r e i G ·~r u~k+G~ (~r ) = e i k·~r ũ(~r )

~
where ũ is periodic: the Bloch states associated with the e i k·~r and
~ ~ ~ ~ ~
e i(k+G )·~r phase factors are the same, and e i k·R and e i(k+G )·R are the
same eigenvalues of TR~ .

The Brillouin zone (BZ) is the ensemble of independent ~k-vectors (not

connected by any G~ -vector). It is the primitive cell of the reciprocal
vectors lattice.
Figure: 2D hexagonal lattice with its
first Brillouin-zone. The important
k-points bear specific names (Γ is the
zone-center, K is at the corner, etc.)
The Brillouin zone (II)
The Brillouin zone is usually taken to be the highest-symmetry primitive cell of
the reciprocal lattice, namely the ”Wigner-Seitz” primitive cell. Plot the planes
normal to reciprocal lattice vectors cutting them ”in the middle”. The volume
that such planes will define is the Brillouin zone. High symmetry directions and
k-points have ”standard” names.

Figure: (Left) 2D square and hexagonal BZ. (Middle) The BZ of a face-centered

cubic (FCC) lattice is a truncated octaedron. (Right) The BZ of a body-centered
cubic lattice is a rhombic dodecahedron.
Number and nature of the k-vectors

For illustration, let’s go back to 1D. The values

of (k) are governed by the boundary conditions.
Solid-state physicists adopt usually the Born
and von Karman periodic boundary conditions
where the solid ”closes” onto itself. This means
that with N cells, one has the condition:

ψ(x + Na) = ψ(x) ⇒ e ikNa = 1

⇒ k = integer × (2π/Na)

The first BZ is: − πa < k ≤ π

a.

There are thus N k-vectors in the BZ, as many as unit cells.

~ ~ 1−e ikNa
PN−1
e i k·R = e ikna =
P
Sum rule: ~
R n=0 1−e ika
= 0 for k 6= 0 in the BZ.
Orthogonality of Bloch states
Let’s demonstrate that Bloch states for different (~k)-points in the BZ are
orthogonal. One can write:
Z
~0 ~
hψ~k 0 |ψ~k i = d~r e i(k −k)·~r u~k∗0 (~r )u~k (~r )

which can be expressed in terms of Fourier components:

Z
1 ~0 ~ ~
d~r e i(k −k+G )·~r = δ(~k 0 − ~k + G~ ) = 0
8π 3

since by definition for (~k) and (~k 0 ) in the BZ, they cannot differ by a
reciprocal-lattice vector G~ . Alternatively by Bloch theorem:
!
X ~0 ~ ~ Z
i(k −k)·R
hψ~k 0 |ψ~k i = e ψ~k∗0 (~r )ψ~k (~r ) = 0
R Ωcell

unless ~k 0 = ~k + G~ (including the condition: ~k 0 = ~k ).

Block diagonalisation of the Hamiltonian

Since the Hamiltonian is periodic, then the arguments developed here

above hold and:

hψn0~k 0 |Ĥ|ψn~k i = hψn0~k |Ĥ|ψn~k iδ(~k − ~k 0 )

where the indices (n,n’) serve to distinguish Bloch states with the same
~k-vector (not necessarily eigenstates of Ĥ, e.g. basis vectors).

The Hamiltonian does not couple Bloch states

with different Bloch vectors.

This is the central result.

Since there are as many k-point in the BZ as

there are unit cells in the crystal, in the absence
of orthogonality and block diagonalization of
the Hamiltonian, the Bloch representation
would not have helped much.
Setting up H(~k)
~
We start from Bloch theorem: ψn~k (~r ) = e i k~r u~k (~r ), where un~k (~r ) is
periodic. Then:

−~2 ∇2
 
+ V (~r ) ψn~k (~r ) = En~k ψn~k (~r )
2m

yields straighforwardly:

" #
p + ~~k)2
(~
+ V (~r ) un~k (~r ) = En~k un~k (~r ), with p~ = −i~∇.
2m

This is a ~k-specific Hamiltonian: one has to set-up and diagonalize a

different Hamiltonian for each ~k-point of interest in the BZ.

Compare e.g. to the hydrogen case where we had a specific radial

equation for each l-quantum number, with the (L2 /2mr 2 ) centrifugal
term coming from the kinetic operator in spherical coordinates.
Setting up H(~k) in a basis
LCAO basis: if each atom is described by a set of atomic orbitals
αnlm (~r ), then one can create a periodic crystal basis {φnlm } to describe
the periodic function u~k (~r ) as follows:
X
φJ,nlm (~r ) = ~
αJ,nlm (~r − ~τJ − R)
~
R

where ~τJ runs over the unit cell atoms.

PW basis: the planewave (PW)

representation uses the Fourier expansion of
un~k (~r ) over the reciprocal lattice G~ -vectors:
~
X
u~k (~r ) = c(G~ )e i G~r , with:
~
G
Z
d~r ~
u~k (G~ ) = u~ (~r )e −i G~r
Ω k
Linear combination of atomic orbitals versus planewaves

The question: ”which is the best basis ?” (planewaves, Gaussians,

real-space grid, wavelets, etc.) has probably no answer besides ”it
depends on the system you study !”.

Atomic-like orbitals are extremely

compact and allow a natural description
of the variations of wavefunctions close
to the atoms.

Plenewave basis are on the contrary not

very good for describing strong
variations of the density, but are more
systematic and allow to sample the
density far away from the atoms (e.g.
diffuse orbitals, interstitial sites, etc.)
Courtesy:
http://www.iue.tuwien.ac.at/phd/osintsev/disserch4.html
The full space group
Besides the translations, the crystal is invariant under various operations
{g } containing rotations, reflexions, inversion, etc. with respect to a
point, an axis, or a plane, and combinations of such operations with
translations: (g |a)~r = g ?~r + ~τ , with ~τ not necessarily a lattice vector.
−1
ψ~k (g ?~r + ~τ ) = e ik·(g ?~r +~τ ) uk (g ?~r + ~τ ) = e ig k·~
r
ũ(~r ), with ũ(~r ) periodic.

The eigenstates ψ~k (g ?~r + ~a) are also the eigenstates of Ĥg −1 k . In
particular, the eigenvalues of Ĥg −1 k and Ĥk are the same.

The irreducible Brillouin zone defines the set of non-equivalent ~k-points.

Irreducible Brillouin zone (in red) for an

hexagonal lattice. The irreducible BZ represents
here 1/24 of the full Brillouin zone. Letters
specify the name of the ”high symmetry”
~k-vector and directions. All other ~k-vector in
the BZ can be obtained by symmetry.
What is a band-structure ?
The plot of the En~k eigenvalues of the Hamiltonian Ĥ~k for ~k-vectors
moving along specific directions of the BZ represents the bandstructure.
10/05/15 08:54

Γ15
Ec
Γ'25 Χ1 Ev
0
E in eV

−6

Χ1

−10
Γ1 Γ1

L Λ Γ Δ Χ U,K Σ Γ
Si k

Above: band structure of silicon (FCC lattice). The zero of energy has
been set to the top of the valence bands. The shaded area above the
(occupied) valence bands is the gap. Above the gap lie the conduction
bands empty at 0 Kelvin. The shaded area below the valence bands is
the gap in energy between the valence bands and the core levels.
http://upload.wikimedia.org/wikipedia/commons/0/04/Band_structure_Si_schematic.svg Page 1 sur 1
Reading a band-structure
In a metal, there is no band gap: valence bands (VB) and conduction
bands (CB) overlap. There is no need of energy for electrons to jump in
empty energy levels.

(From G.Sun (2010), N. Costa and A. Cartaxo (Ed.), ISBN:

978-953-307-088-9, DOI: 10.5772/8672).

In a direct band gap, the VB top and CB bottom are at the same
~k-vector: a photon carrying close to zero momentum can promote an
electron or be emitted by ”hot” electron relaxation. In an indirect band
gap semiconductor such as silicon, one need the momentum of a phonon
to make transitions between the VB top and CB bottom.
The minimal band structure and the effective masse
Let’s look at the band structure of Si and in
particular the top of the valence bands at
10/05/15 08:54

6
zone-center Γ and the bottom of the conduction
Γ15 bands along ΓX. Close to these minima/maxima,
Ec

0
Γ'25 Χ1 Ev the bands are parabolic. In the vicinity of the
extrema, we write the Bloch vectors as (k + q)
E in eV

where k is the Bloch vector at the band extremum

−6 and q is small. Then (with i,j=x,y,z):
Χ1
1 X ∂ 2 εn
−10
εn (k + q) = εn (k) + qi qj + O(q 3 )
Γ1 Γ1
2 ij ∂ki ∂kj
L Λ Γ Δ Χ U,K Σ Γ
Si k

For electrons or holes close to the bands maxima (remember than room
temperature is 26 meV), then one can define an effective mass ”tensor”:

~2 1 ∂ 2 εn
=
http://upload.wikimedia.org/wikipedia/commons/0/04/Band_structure_Si_schematic.svg
2mij∗
Page 1 sur 1
2 ∂ki ∂kj
Usefulness of effective mass and interactions in crystal

The effective mass tensor accounts for the effect of the interaction with
the lattice and the other electrons that renormalize the electron mass: to
lowest order, electrons and holes close to the band edges are free-electron
with renormalized mass. Due to anisotropy, this is not as simple as the
free-electron model, but one can keep in mind that an average effective
mass in Si is roughly: m∗ ' 0.2me , where me is the ”vacuum” mass of
the electron.

Before exploring a few examples of the usefulness of such an effective

mass, we will admit here - but that will be the subject of many
discussions these two weeks - that another effect of the electron-electron
interaction is to renormalize the long-range Coulomb interaction in
semiconductors by an constant M named the macroscopic dielectric
constant. Namely:
1 1
⇒ beyond a (very) few nanometers.
|r − r0 | M |r − r0 |
The hydrogenoid model of impurity

In e.g. n-type doped Ge by P,

the crystal has an extra ionic
charge (ZP = ZGe + 1) and an
extra electron that jumps into
the conduction bands at room
temperature. It is an effective
hydrogenoid problem.
  ∗
e 4 me 1 m
Rydb = ⇒ Ebinding = Rydb 2 ' 9 meV
2~2 M me
~2 m∗
 
∗
bohr = ⇒ a = ( M ) bohr ' 80 bohrs.
me e 2 me

The experimental binding energy value for P:Ge is: Ebinding = 12 meV, in
nice agreement with the estimated value. The small binding energy and
the large effective bohr radius justify the approximations used.
The density of states

The density of states g (ε)dε counts the

number of Bloch states with energy within
[ε, ε + dε] and per unit volume. It is a crucial
quantity since it tells you how many states are
available at a given energy to absorb a photon,
conduct electricity or heat (assuming that by
doping or other means you can put your Fermi
level at this energy).

The density of states requires a summation over the entire Brillouin zone
(BZ), namely:

8π 3 8π 3
Z
1 dk X ΩBZ
g (ε) = δ(ε − εnk ), with ∆k = = = ,
Ω BZ ∆k n N NΩcell Ω

where ∆k is the elementary volume in reciprocal space allowing for one

Bloch state (2 if adding spin variable).
Parabolic bands density of states (0D,1D,2D,3D)
In the case of parabolic bands, it is a standard exercise to
count the number of allowed Bloch vectors up to an energy
E and its corresponding k-vector amplitude such that
E = ~2 k 2 /2me :

(4πk 3 /3) N
N(E ) = 2spin ⇒ ∝ E 3/2
(8π 3 /Ω) Ω
√
so that g (ε) = Ω1 dN(E )/dE ∝ E .

(Exercise: the kinetic energy per volume scales as ρ5/3 , ρ the density.)

It is also standard to redo the exercise at

various space dimension to show that:
I g2D (ε) ∝ constante,
√
I g1D (ε) ∝ 1/ ε,
P
I g0D (ε) ∝ n δ(ε − εn ).
Figure courtesy Lorenzo Mino et al 2013 J. Phys. D: Appl. Phys. 46 423001
A well-known example: the nanotubes

Nanotubes have allowed exploring in much

better details physics at ”1D” (together
with nanowires, etc.)

Band structure (DFT/LDA)and electronic

density of states (eDos) of a semiconducting
(10,0) nanotube. At the onset of parabolic
√
bands, the eDos varies clearly as 1/ .

Band structure and eDos of a metallic (9,9)

nanotube. At the Fermi level, the dispersion
is linear, not parabolic: the eDos is constant
around EF .

Courtesy Charlier et al. Rev. Mod. Phys. 2007

Formation of bands and band gaps

The quasi-continuum of {εnk } energies produces a manifold a states. But

how can we interpret the formation of bands and band gaps ? Why is
graphene semi-metallic and diamond insulating ? Why is hexagonal
boron-nitride insulating ? Why does the band gap close from diamond to
silicon, both crystallizing in the same FCC structure ?
Bands from the free-electron gas model

It is instructing to start from the 1D free-electron

model in a box of length L (L √→ +∞) of which the
eigenstates are ψ(k) = e ikr / L and the energies
are ε(k) = ~2 k 2 /2m.
(Figure from Jones and March, Theoretical Solid State Physics, p. 41, Dover Eds.)

We can always say that the free-electron gas is invariant by a translation

T (a) and define a reciprocal-space with a Brillouin zone: |k| ≤ π/a and
reciprocal-lattice vectors G = n(2π/a).

Scheme: take a k-vector in the BZ and get the energy of all e ik·r unk (r)
eigenvectors of Ĥk (u periodic) ⇒ these are the e ik·r e iG·r states !!

For a given (k) in the BZ, plot all the ~2 |k + G|2 /2m energy levels.

In the Bloch states band-structure picture, all free-electron states are

”folded back” into the BZ.
Bands from the free-electron gas model (II)
We can now switch the periodic potential V yielding:
~2 2 0 0
P
2m (k + G ) cn (k + G )δG ,G + G 0 cn (k + G − G )V (G ) = Enk cnk (G ).
0

Close to the (k = π/a) zone-boundary, the e ikr

and e i(k+G )r (with G = −2π/a) states -
degenerate in energy without V - mix strongly
through U = V (2π/a) ⇒ solve:
 2 
2
~
k − ε c(k) + Uc(k + G ) = 0
 2m2 
2
2m (k + G ) − ε c(k + G ) + Uc(k) = 0
~

yielding to lowest order (see Kittel, chapter 7):

2
h
2 δk2 i
δk2 ~2 δk2 π 2
ε( πa + δk) = ε0 + ~2m ± U 1 + 2 Uε02 ~2m , with ε0 = 2m ( a ) .
Complement: the Peierls distorsion

The above discussion can be applied to understand the Peierls distorsion,

which can be viewed as a solid-state physics version of the molecular
Jahn-Teller effect. Assume a simple 1D band model at half-filling with
the Fermi wavevetor at halfway the zone-boundary (kF = π/2a).

A dimerisation of the lattice will double the

unit-cell size (a⇒2a) divide by 2 the BZ
(2π/a ⇒ 2π/2a), introducing Fourier
components of the potential V at G = π/a
that can couple the two states at ±kF .

The crucial argument now is that the penalty in elastic energy is

quadratic in the (δa) distortion, while the electronic energy gain is linear
with U (see previous slide), that is with the distortion strength ⇒ the
distortion always win at small (δa).
From the atomic limit to the solid

Rather than starting from the free-electron gas model, it is very intuitive to
start from the atomic limit in the standard picture of the atomic-orbitals
interaction and the formation of bonding-antibonding states, generalizing this
approach to a large number of atoms in interaction. In many important cases,
its is however difficult to recognize the solid limit from the atomic limit !!

Figure: evolution of the energy

bands from the atomic limit to the
solid as a function of (decreasing)
interatomic distance. (From:
”Electronic structure and the
properties of solids”, Walter A.
Harrison, Dover Ed.)

In the solid limit for silicon, one starts from well defined (2s) and (2p) atomic
states, to intertwined hybridized sp 3 bands.
The tight binding (or extended Hückel) formalism
Assume atomic orbitals φm (~r ) per site, typically eigenstates of the
isolated atom Hamiltonian. ”m” here can represent (nlm) indexes.

One can form crystal extended Bloch basis

P states as a linear combination
of atomic orbitals (LCAO): ψm (~r ) = R~n am (R ~n )φm (~r − R ~n ).
~ ~
~l ) = e i k·Rl ψ(~r ), one finds
For ψ(~r ) to satisfy the Bloch theorem: ψ(~r + R
i ~
k·R~
~n ) = e n am (R ~0 = ~0).
that: am (R

By normalization, assuming that: hφm (~r − R~n )|φm (~r − R

~l )i = δnl , that is
√
~
we assume localized orbitals, then: am (R0 ) ' 1/ N and:

1 X i ~k·R~n ~n ).
ψm~k (~r ) ' √ e φm (~r − R
N ~
Rn

Such a basis Bloch state can be e.g. an s-orbital repeated periodically

~ ~ ~
with a e i k·Rn phase factor. The phase factor is the standard e i k·~r Bloch
phase taken to be zero only at atomic sites.
The 2D square lattice
Assume one orbital φ(~r ) per site and interatomic distance (a):
Onsite energy set to zero:

E0 = hφ(~r )|Ĥ|φ(~r )i = 0,

First-nearest neighbor
hopping energy:

t = −hφ(~r )|Ĥ|φ(~r + ~a1/2 )i.

Then the tight-binding approach can be set up as follows:

1 X i ~k·R~nm ~nm )
E (kx , ky ) = hψ~k (~r )|Ĥ|ψ~k (~r )i with ψ~k (~r ) = √ e φ(~r − R
N n,m

where R ~nm = n~a1 + m~a2 , and yields the following band structure:

E (kx , ky ) = −t e ±ikx a + e ±iky a = −2t [cos(kx a) + cos(ky a)] .

Square lattice at half filling: Fermi surface nesting

Assume one electron per site, namely a situation of

half-filling where only half of the band is occupied.
Then the Fermi surface is a square ! The Fermi
surface is said to be nested: specific wavevectors Q
can connect large portions of the Fermi surface.

This may lead to instabilities of the Fermi surface. Assume a

perturbation δVQ (phonon, magnon, etc.) with wavector Q. Then a very
large number and initial and final states are available for scattering,
inducing a huge response of the system to the perturbation.

Such a model is a minimal model for the cuprate (CuO) planes close to
the Fermi level with strong Cu (dx 2 −y 2 ) character.
The graphene

Close to EF the graphene band structure is dominated by pz orbitals.

With two atoms per cell at positions τA and τB in the unit-cell one can
A/B
build two basis Bloch states per k: ψk = n φz (r − τA/B − Rn )e ik·Rn .
P

We define:
I hφz (r − τA/B )|Ĥ|φz (r − τA/B )i = 0
(energy reference)
Courtesy J Gttinger et al 2012 Rep. Prog. Phys. 75 126502 I hφz (r − τA )|Ĥ|φz (r − τB )i = γ

A/B A/B
One then diagonalize the 2x2 Hamiltonian with the hψk |Ĥ|ψk i
matrix elements, keeping only first-nearest-neigbour interactions, yielding:

E ± (k) = ±γ|α(k)|2 , α(k) = 1 + e −ik·a1 + e −ik·a2

which cancels at the (k = K) BZ corners.

(see The band theory of graphite, P. R. Wallace Phys. Rev. 71, 622, 1947).
The ionicity gap of hexagonal BN

Hexagonal BN cristallized as graphene, but atom ”A” and ”B” are

different : one is boron, the other nitrogen. We follow the very same
treatment than graphene, but the onsite energies must be different for
boron and nitrogen: hφz (r − τA/B )|Ĥ|φz (r − τA/B )i = ∆/2 for boron and
-∆/2 for nitrogen which is more electronegative.

Then solving again the 2x2 Hamiltonian

problem, one finds:
p
E ± (k) = ± ∆2 /4γ 2 + |α(k)|2

Even when α(k) cancels due to the

hexagonal geometry, as in graphene, the
difference of ionicity ∆ leaves the gap
opened.
Courtesy Arenal et al, Adv. Phys. 2010
Complement: insulator-to-metal transition by doping
A nice illustration of the formation of a band is given by the physics of
”degenerate” semiconductors which are semiconductors, or insulators,
than turned metallic due to high doping: upon increasing dopant
concentration, the impurity states start overlapping and form a band
which becomes larger and larger. When the impurity band overlap with
the valence of conduction bands, the system becomes metallic !

Figure. (Left) Insulator to metal transition by doping (Blase et al., Nature Mater. 8, 375, 2009) (Right) Superconducting transition in
B-doped diamond (Ekimov et al., Nature 428, 542-545, 2004).
Vibrational modes in solids: phonons

Another important use of periodicity concerns the vibrational modes in crystals:

the phonons. The theory starts from the Taylor expansion of the total energy
around the atomic equilibrium positions:

∂2E
 
n o 1 X
E ({RI }) = E ( R0I ) + uI α uJβ + O(u 3 )
2 ∂uI α ∂uJβ u=0
I α,Jβ

where e.g. (uI α ) is the displacement along the direction (α = x, y , z) around

its equilibrium direction of atom (I) in the crystal. Using Newton’s equation
with classical ions of mass MI , namely: MI d 2 uI α /dt 2 = −∂E /∂uI α , the
√
search for harmonic solutions: uI (t) = vI exp(iωt)/ MI yields:

∂2E
 
2
X 1
ω vI α = DI α,Jβ vJβ with: DI α,Jβ = √
Jβ
MI MJ ∂uI α ∂uJβ

∂ 2 E /∂uI α ∂uJβ is the force constant matrix, namely ”the spring constant”

between the two atoms (I) and (J), and DI α,Jβ is the dynamical matrix.
The frozen phonon approach
In a molecule, the dynamical matrix (DI α,Jβ ) can be calculated by finite
differences displacing explicitely all atoms by a small distance ±δu (say 0.1
bohr) in each direction of space. This comes at the price of 6Nat total energy
calculations, where Nat is the number of atoms.

In a solid, Nat → +∞ ! The textbook approach is to look for solutions that are
planewave like, or propagating waves, namely: u(R, t) = ~ × e i(q·R−ωt) with
u(R, t) the displacement of atom R at time (t). ~ is the polarization vector.
The reason for that really is that the dynamical matrix depends only on the
distance (RI − RJ ) thanks to periodicity:
XX
ω 2 vα (R) = Dα,β (R − R0 )vβ (R)
R0 β

where the right-hand-side is a convolution product that yields a direct product

by Fourier transform:
X X
ω 2 vα (q) = Dα,β (q)vβ (q), Dα,β (q) = Dα,β (R)e −iq·R
β R
Phonon band structures: acoustic and optical modes
Phonon modes with different q-vectors ”cannot mix” (in the harmonic
approximation) and only q-vector in the Brillouin zone should be
considered, since again:

uq+G (R, t) = ~ × e i((q+G)·R−ωt) = ~ × e i(q·R−ωt) = uq (R, t)

Periodicity has allowed to block-diagonalize the dynamical matrix

allowing the efficient calculation of vibrational modes in solids.

Figure (Left). DFT and experimental (dots) phonon band structure for GaAs (Baroni
et al. Rev. Mod. Phys. 2001). There are Natcell =2 atoms/cell and thus 6 modes per

q-point. (Right) Schematic representation of phonons with various wavevectors.

Complement: The Fermi surface

The Fermi surface is the energy surface in the Brillouin zone such that
εnk = EF , where EF is the Fermi level. At zero Kelvin, all states at lower
energy are occupied, all states above are empty.

FCC Brillouin zone.

Figure (Left) The cover of Ashcroft and Mermin, Solid State Physics, Saunders
College Publishing, showing the Fermi surface of silver. (Right) An historical band
structure of silver using relativistic augmented planewave method (N.E. Christensen,
TU Denmark, Phys. Stat. sol. (b) 54, 651 (1972).
The Fermi surface

The Fermi surface characterizes the charge carriers at room temperature

(kB T =25 meV), in connection with the electronic and thermal
conduction properties. Any perturbation bringing little energy to the
system, as compared to electronic band dispersions (namely a few eV),
such as phonons or magnons, will scatter electron from one-point of the
Fermi surface to another.

Figure
(Left) LiFeAs Fermi ”multiple” surface: several bands cross the Fermi level. (Centre
and left) Do you recognize the materials with such Fermi surfaces ?
Complement: Impulsion operators as generators of
translations
Start with a 1D infinitesimal translation by a distance (da):

 
dψ i ~ d
T̂da ψ(x) = ψ(x − da) = ψ(x) − da = Iˆ − da ψ(x)
dx ~ i dx

so that: T̂da = Iˆ − ~i (da)p̂x . This leads to a general relation:

 
iˆ
T̂~a = exp − ~p · ~a .
~

If a system is invariant by infinitesimal translations, then the Hamiltonian

commutes with T̂da , that is with the impulsion operator. By Ehrenfest,
the impulsion is conserved.

dhp̂x i ∂ p̂x 1 h i
=h i + h p̂x , Ĥ i = 0 + 0
dt ∂t i~