You are on page 1of 15

Faststoffysik II - Eksamen

Sebastian Yde Madsen


August 6, 2023

1 The Tight-Binding model


• The Tight-Binding model (also called LCAO model - stands for Linear Combination of Atomic Orbitals), is an
approximate model used in solid state physics to describe the electronic states formed in crystalline structures, i.e.
metals, insulators and semi-conductors.

• Most fundamental classical description is given be the Drude model (inspired by Kinetic Gas Theory). However,
has shortcomings:
– No justification for neglecting electrostatic interactions, i.e. (electron-electron) and (electron-ion).
– No justification for ignoring collisions.
– Classical Equipartition Theorem suggests that electrons should contribute considerably more to heat capacity,
than is observed.
• Most fundamental quantum mechanical description is Free-electron model - QM analog to Drude model. Utilizes
Born-Oppenheimer Approx.1 and One-electron Approx.2 . Also still assumes no (electron-electron) or (electron-
ion) interaction. Effectively reduces to solving Time-independent (stationary) Schrödinger Wave equation:

−ℏ2 ∇2
Ĥψ(r) = − ψ(r) + V (r)ψ(r) = Eψ(r) (1)
2me

for V (r) = 0 and periodic/cyclic boundary conditions3 (”free particle in box”), which has plane-wave solutions:

ℏ2 ℏ2  n x π  2  n y π  2  n z π  2
 
1
ψ(r) = √ eik·r with E(k) = ||k||2 = + + (2)
L3 2me 2me L L L

• The Free-electron model gives more accurate estimate of electrons contribution to heat capacity, correct value
for Lorentz number in Wiedemann-Franz law (as opposed to Drude model), etc. However still has obvious
shortcomings such as ignoring collisions.

2 Graphene
• Define structure of Graphene: i.e. draw honeycomb structure of carbon atoms w. lattice vectors:
1
− 21
! !
2
t1 = a √3 , t2 = a √3 (3)
2 2

and unit-cell (2. atoms pr. unit cell) + basis vectors:


!
0
d1 = 0, d2 = a √
3
(4)
3

and also 1BZ in reciprocal space: see - fig. 1.

1 Quick def. here


2 Quick def. here.
3 ψ(x + L, y, z) = ψ(x, y + L, z) = . . .

1
Figure 1: Graphene structure a) is real space, b) is reciprocal space and corresponds to 1 BZ.

• To enable an accurate characterization of the properties of Graphene, we want to investigate the band structure. The
carbon atoms have an electron structure of 1s2 2s2 2p2 4 - where we are only concerned w. the 4 electrons in the
outermost shell n = 2:
– 3 of these electrons are in the 2px , 2py and 2s orbitals - in the x, y-plane.
– 1 of these electrons are in the 2pz orbitals - orthogonal to the x, y-plane.

The band wavefunctions originated from px , py and s orbitals (σ bands) are even under reflection in the plane of
graphite; they do not mix with band wavefunctions originated from pz orbitals (π bands), which are odd under
reflection in the plane of graphene. Thus σ-bands and π-bands can be studied separately.
Læs Magnus’ for resten

3 Hartree-Fock theory
Another approach for dealing with many electron-systems, i.e. another method for calculating dispersion relation E(k).

• A general model of a solid of many electrons and nuclei, is the one in which the coulomb-like interactions between
(electron-electron), (nuclei-nuclei) and (electron-nuclei) are considered alongside the kinetic energy of both. This
results in a Hamiltonian on the form:
X p̂2 X p̂2 X e2 X zi zj e2 X zi e2
e N
Ĥ = + + + + (5)
i
2me i
2mN i<j
|ri − rj | i<j |ri − rj | i<j |ri − rj |

However, due the large mass difference mN ≫ me and assumed rigidity of the solid, one utilizes the Born-Oppenheimer
approx., which neglects the kinetic energy term of the nuclei alongside of (nuclei-nuclei) interactions due to the as-
sumption of nearly static nuclei. As such, one is left with an effective Hamiltonian in the form:
X p̂2 X e2 X zi e2
e
Ĥef f = + + (6)
i
2me i<j |ri − rj | i<j |ri − rj |

• To derive the Hartree-Fock equations, one starts out by making the ansatz that the N -electron system can described
by an anti-symmetrized (according to Pauli principle) product wavefunction:
N! N
O 
1 X
Ψ0 ≡ √ cα P̂α ψi (si , ri ) , si ∈ {↑, ↓} (7)
N! α
| {z } i
Anti-symmetrizer

and then continuous to perform variation on:


⟨Ψ0 | Ĥef f |Ψ0 ⟩ (8)
which, is done by varying ψi (si , ri ) → ψi (si , ri ) + δψi (si , ri ), constrained by orthonormality, s.t.:

⟨ψi (si , ri )|ψj (sj , rj )⟩ = δi,j (9)

4 Remember notation is defined as nl# electrons

2
In practice, one utilizes Lagrange-multipliers, and then considers the functional5 :
X
G({ψi (si , ri )}) ≡ E({ψi (si , ri )}) − εi ⟨ψi (si , ri )|ψi (si , ri )⟩ (10)
i

instead. Now, varying G to linear order and setting = 0, one arrives at:
Uncover mysterious shitty notation from book def. of ⟨G1 ⟩ and ⟨G2 ⟩.

Write about jellium model

4 Electronic transport properties and the Boltzmann equation


We will now describe intraband transport process’s, specifically for metals with a partly filled conduction band using the
Boltzmann equation.
• In equilibrium, the probability of a state being occupied is given by the Fermi-Dirac distribution
1
f 0 (k) = . (11)
eβ(Ek −µ) + 1
and the number of electrons in a 6D phase-space hypercube of volume d3 kd3 r is given by

d3 rd3 k
N = 2f 0 (k) . (12)
(2π)3

• In non-equilibrium, if e.g. a field is applied, the probability of a level being occupied will also depend on space
and time, f (r, k, t). If the spatial dependence of the disturbance causing the system to go out of equilibrium varies
slowly compared to the average distance between the electrons 1/kF , the number of a states inside a 6D phase-space
hypercube with volume d3 rd3 k being occupied is

d3 kd3 r
N = 2f (r, k, t) . (13)
(2π)3

• To find f (r, k, t), we use the semi-classical equations of motion

1
vk = ∇k · E(k), Group velocity (14)

d(ℏk)
F(k) = , Force on electron (15)
dt
and the fact that the number of electrons is conserved, s.t.:

d3 k 3 d3 k′ 3 ′
2f (r, k, t) d r = 2f (r + v k dt, k + Fdt, t + dt) d r (16)
(2π)3 (2π)3

which, when combined w. Liouville’s theorem: d3 rd3 k = d3 r′ d3 k′ , yields:

f (r, k, t) = f (r + vk dt, k + Fdt, t + dt). (17)

Now, if we take into account the collisions, this will break the conservation of number of electrons. This is very hard
to deal with. However, this is handled by including an extra term:

f (r, k, t) = f (r + vk dt, k + Fdt, t + dt) + Icoll [f ]dt. (18)

Finally, by Taylor expansion, we arrive at the Boltzmann equation:

5 Quick def. here.

3
5 Optical properties of metals and semiconductors
We will now describe the optical properties of metals and semi-conductors.
• We employ Maxwell’s equations in matter. We take the current density (and the E-field) along the x-direction and
the B-field in the y-direction. Therefore the wave propagates in the z-direction. We assume no external charges or
currents. We can then find:
d2 E ω2 4πω
2
= − 2 E(z) − i 2 J(z) (19)
dt c c
N.B. and we can check that it holds in vacuum as well, by setting J(z) = 0 and considering plane wave E-field (eikz ),
in which case we get the expected linear dispersion ω = kc → E(k) = ℏω = ℏck.
• Inside a metal, we have to calculate J(z). We assume a linear response, so J ∝ E. To this end, we must have the
conductivity tensor σµ,ν (r − r′ , t − t′ ), which in general must be integrated over. We, however, assume homogeneous
and isotropic material (s.t. σ is no longer a tensor) and neglect local field effects induced by in-homogeneity of the
sample. Further, we make assumption of a local response;, s.t. σ(r − r′ , t − t′ ) = δ(r − r′ )σ(t − t′ ) which results in
neglecting wave vector dependence of the conductivity function. For this, we get:

J(r) = σ(ω)E(r) (20)

N.b. The local response assumption is justified when the average distance travelled by carries are smaller than the
variation of the field.
• Putting this into the differential equation for the electrical field:

d2 E(z) ω2 ω2
 
4πσ(ω)
= − 1 + i E(z) = − N (ω)2 E(z), E(z) = E0 eiω/cN (ω)z (21)
dt2 c2 ω c2

where N (ω) = n(ω) + iκ(ω) is the complex refractive index, and n(ω) is the usual refractive index and κ is the
extinction coefficient. From the above expression it is evident that the electric field is exponentially damped. (Note
also that the dielectric function is given as ε(ω) = N (ω)2 .)
Consider fig. 2

Figure 2: Generic depiction of partially transmitted electric field.

Utilizing continuity at the surface to get Er + Ei = Et alongside the above result we find the Reflection (at normal
incidence) to be given as:
2 2
Er 1−N (n(ω) − 1)2 + κ2
R= = = (22)
Ei 1+N (n(ω) + 1)2 + κ2
• We now turn to Drude theory for Jellium with a positive, uniform charge density. The carrier dynamics are described
by the following linear 2. order in-homogeneous differential equation:
m
mü = − u̇ − eE0 e−iωt (23)
τ
where τ is the phenomenological relaxation time. The first term is the damping term and the second is the
driving term. This has solution:
eτ 1
u(t) = u0 e−iωt , u0 = E0 (24)
m ω(i + ωτ )
which we can use to get the free carrier contribution to the current density:
 2 
ne τ 1
J = n(−e)u̇ = E0 e−iωt ≡ σ(ω)E(t) (25)
m 1 − iωτ

4
(N.B. here n is the nr. of carriers pr. unit volume.). With the knowledge that τ ∼ 10−14 in typical situations, and the
relation between the static conductivity and the free-electron plasma frequency ωp2 = 4πne2 /m, it becomes possible to
define a frequency scale that enables a distinction between 3 characteristic ranges for metals:
1. Low-Frequency Region: ω ≪ τ1 ≪ ωp - In the low-frequency region ω is negligible with respect to 1
τ, and the
material is strongly dissipative......
add more
1
2. High-Frequency Region τ ≪ ω ≪ ωp ......
add more
3. Ultraviolet Region ω ≈ ωp or ω > ωp ......
add more

6 Excitons
7 Screening
Static dielectric screening of Metals
We will see that although the screening of electrical fields is perfect on long length scales inside metals, as we have learned
in introductory electromagnetism, it is not complete on short length scales compared with the average distance between the
electrons. First, we will do this using Thomas-Fermi theory - but can also be done bu using Lindhard model.
• Lets begin by a brief reminder of how Maxwell’s equations may be formulated inside a medium (Macroscopic):
Microscopic Macroscopic
I ∇ · E = 4πρ ∇ · D = 4πρfree
II ∇·B=0 ∇·B=0
1 1
III ∇ × E = − ∂t E ∇ × E = − ∂t E
c c
1 1
IV ∇ × B = (4πJ + ∂t E) ∇ × H = (4πJfree + ∂t D) (26)
c c
with ρ = ρfree + ρbound , where Free charge density refers to the concentration of charge carriers that can move
relatively freely within a material. In the absence of an externally imposed field, these charges may still be present
within the material, without necessarily contributing to a net movement of charge. The Bound charge density, on
the other hand, arises due to the polarization of atoms or molecules within a material when subjected to an external
field. While these charges are not free to move independently like free charges, they still contribute to the overall field
within the material.6 The current densities are similarly defined J = Jfree + Jbound . Now, starting from the definition
of the Displacement field, taking the divergence on both sides, and utilizing I:
D = E + 4πP
⇔ ∇ · D = ∇ · E + 4π∇ · P
⇔ ρfree = ρ + ∇ · P
⇔ ∇ · P = −ρbound (27)
From this it can be deduced that the physical interpretation of the Polarization P is the electric dipole moment per
unit volume.
For many (linear) materials, the Polarization is proportional to the electric field
P = χe E, (28)
where χe is the electric susceptibility7 . Utilizing this in the definition of the Displacement vector:
D = (1 + 4πχe )E = ϵE, (29)

6 When an external field is applied to a material, the positive and negative charges within the atoms or molecules experience a force that

slightly shifts them. This shift creates temporary charge separations, resulting in localized regions of positive and negative charge. These bound
charges contribute to the overall field within the material, affecting its response to the external field.
7 A fundamental parameter used to describe the response of a material to an applied electric field. It quantifies the degree to which a material

becomes polarized (develops bound charges) when subjected to an external electric field. In other words, electric susceptibility indicates how
easily a material can be polarized by an electric field.

5
where ϵ is known as the dielectric function (sometimes also relative permittivity). As seen from this, if one can
determine the dielectric function (and the displacement vector - which can be done according to I), one can
determine the electrical field inside the material, and from this, alot of the important properties of the material.

• Now lets derive the dielectric function using Thomas - Fermi model. We consider a Jellium model with a initial
uniform charge distribution. Now imagine placing a positive charge Ze into a metal. The electrons will gather around
the positive charge - screening it in such a way that the electric field is zero far away. However, close to the charge
the electric field is different from zero.
The added charge will cause a electric potential to form, which in turn, will perturb the electron density by of the
metal by δn(r) and thus induce a (bound) charge density given by ρbound = −eδn(r).
Now, we will take the potential formed by the positive charge, to be of the form of a plane wave:
U (r) = U (q)eiq·r (30)
where we will assume q ≪ kF meaning that potential varies slowly on the length scale of the electrons, s.t. that it
can locally be regarded as homogeneous. This effectively induced a number of depleted electron states in the metal as
the fermi energy stays the same - see fig. 3

Figure 3: insert description.

The number of depleted states are D(EF − U (r))U (r), and now assuming that U (r) ≪ EF , we get:
D(EF − U (r))U (r) ≈ D(EF )U (r) = −δnbound (r). (31)
which then yields and expression for the bound charges:
ρbound = −eδnbound (r) (32)
= eD(EF )U (r) (33)
iq·r
= eD(EF )U (q)e (34)
= −∇ · P (35)
And this has the solution:
eq
P(r) = iD(EF )e U (q)eiq·r , (36)
q
where eq is the unit vector in the direction of q.
Now, if we utilize the general force definition, we find a relation between the potential and the electric field:
1
F = −∇r U (r) = −eE(r) ⇔ E(r) = ∇r U (r) (37)
e
and utilize the assumption of plane wave potential eq. (30):
iq
E(r) = U (q)eiq·r (38)
e

6
it is given from the linearity between the Displacement vector and the Electric field eq. (29), that we can finally
calculate the Thomas Fermi dielectric function:
e
D P iD(EF )e qq U (q)eiq·r 4πe2 n(EF ) kT2 F
ϵ= = 1 + 4π = 1 + 4π iq
= 1 + = = ϵ(q)T F (39)
E E e U (q)e
iq·r q2 q2

where kT F is the Thomas Fermi wave number.


Notice that
lim ϵ(q)T F = ∞+ (40)
q→0

i.e. for a slowly varying potential the dielectric function is infinite, corresponding c.f. eq. (29) to a zero electric field -
perfect screening. On the other hand:
lim+ ϵ(q)T F = 0 (41)
q→∞

i.e. when the potential varies extremely fast - the electrons doesn’t have ’time to re-adjust’ - we have no screening.
• If we utilize the density of states for a free electron gas, we find that Thomas Fermi wave number is on the
order of magnitude of the inverse Bohr Radius:

kT F ∼ a−1 −1
B ↔ aB ∼ kT F (42)

meaning that the screening cloud of electrons has a radius of ∼ aB - if longer away from Ze positive charge than aB
then perfect screening - if closer not perfect screening.
Talk about limitations of the model and why lindhard gives better predicition.

8 Hubbard model and Mott transition


9 Transition metal dichalcogenides and moire patterns
10 Experimental determination of electronic structure (Angle resolved pho-
toemission spectroscopy)
Appendix
Tight-Binding
N.B. The tight binding approach is usually best for semi-conductors and insulators, whereas the free and nearly-free
approach works okay for metals.

Lets consider a one-dimensional crystal of N atoms, w. lattice constant a, s.t. length is L = aN (N ≫ 1). We
have a periodic potential:
V (x) = V (x + a) (43)
s.t. it can be written as a Fourier series:
X 2π
V (x) = V (hn )eihn x , hn = n (Reciprocal lattice points) (44)
n
a

to ensure V (x) is real valued we require V (hn ) = V ∗ (−hn ). Now, solving the Time-Independent (stationary) Schrödinger
Wave equation TISE:
ℏ2 d2
Ĥψ(x) = − ψ(x) + V (x)ψ(x) = Eψ(x) (45)
2me dx2
under periodic/cyclic boundary conditions, s.t.:

ψ(x) = ψ(x + L) = ψ(x + aN ) (46)

which has Bloch Waves8 :


ψk (x) = eikx uk (x) (47)

8A lattice periodic function ’modulated’ by a plane wave.

7
where the function uk (x) = uk (x + a) is lattice periodic.
Notice that by evaluating eq. (46):
ψ(x + aN ) = eik(x+aN ) uk (x + aN ) = eik(x+aN ) uk (x) = eikaN uk (x) = eikaN ψ(x) (48)
we immediately find the requirement:
eikaN = 1 ⇒ kN a = n2π (49)
9
which gives the allowed k-values :

k= (50)
Na
N.B. still many k-values in BZ10

Free-electron case
In the Free-Electron Model we assume a zero potential V (x) = 0 (or constant - just shifts eigenvalues), and in this
case, the solution to TISE is simply normalized plane waves:
1 ℏ2 k 2
Wk (x) = √ eikx with E(k) = (51)
L 2me
which constitutes an orthonormal system (over the allowed k-values), consider e.g.:
Z L
⟨Wk′ (x)|Wk (x)⟩ = Wk∗′ (x)Wk (x)dx = δk′ ,k . (52)
0

Furthermore, note that the dispersion in this model E(k) (a parabola in k), obviously doesn’t full-fill requirement of
periodicity in reciprocal space:

Figure 4: Free electron parabolic dispersion.

(should have E(k) = E(k + hn )).

Nearly-free electron case


V (hn )eihn x , s.t. the hamiltonian becomes:
P
Now, if we exploit the periodicity of the lattice potential, V (x) = n

−ℏ2 d2 X
Ĥ = 2
+ V (hn )eihn x (53)
2me dx n

we see that the action on one of our normalized plane waves becomes:
1 ℏ2 k 2 ikx X
 
i(hn +k)x
ĤWk (x) = √ e + V (hn )e (54)
L 2me n
ℏ2 k 2 X
= Wk (x) + V (hn )Wk+hn (x) (55)
2me n

a sum of plane waves only shifted by Reciprocal lattice points.


Therefore the space spanned by {Wk (x), Wk+h1 (x), . . .} are closed under the action of the Hamiltonian, and we can restrict
ourselves to only considering k-values in the 1 BZ. The ansatz made for solving TISE is therefore on the form:
1 X X
ψk (x) = √ c(k + hn )ei(k+hn )x = c(k + hn )Wk+hn (x) (56)
L n n

9 the k-values are now quantized.


−π
10 BZ has width 2π
a
(1 BZ is usually defined on a
<k≤ π
a
) and k-values increments w. steps of size 2π
aN
, where N ≫ 1.

8
and by considering the action of Ĥ on this:
1 X ′ X
Ĥψk (x) = √ c (k + hn )ei(k+hn )x = c′ (k + hn )Wk+hn (x) (57)
L n n

we see that the eigenvalue form of TISE is only satisfied for:

c′ (k + hn ) = Ec(k + hn ) (58)

which is not generally gauranteed.


Notice that by setting uk = √1L n c(k + hn )eikhn the ansatz is a bloch wave, such that:
P

(
ψk (x) = eikx uk (x)
⇐⇒ ψk (x + a) = eik(x+a) uk (x + a) = eika ψk (x) (59)
uk (x) = uk (x + a)

Now, from Variational method, we know that solving the TISE can be reduced to solving the secular equation, which in
turn involves the matrix elements of the hamiltonian w. respect to the basis functions, and in our case it is clear that:
X 1
ψk (x) = c(k + hn ) √ ei(k+hn )x (60)
n
| {z } L
cn | {z }
ϕn (x)

and we therefore consider:


Z L
1
Hmn = ⟨Wk+hm (x)| Ĥ |Wk+hn (x)⟩ = e−i(k+hm )x Ĥei(k+hn )x dx (61)
L 0
L
ℏ2 (k + hn )2 −i(k+hm )x ik(k+hn )x 1 L −i(k+hm )x  X
Z Z
1 
= e e dx + e V (hl )eihl x ei(k+hn )x dx (62)
L0 2me L 0
l
2 2 Z L
ℏ (k + hn ) 1 X
= δm,n + e−i(k+hm )x V (hl )ei(k+hl +hn )x dx (63)
2me L 0
l

where, due to the orthogonality, the only terms in the sum in the second term, that survives, are the ones for which
hm = hl + hn , corresponding to hl = hm − hn , s.t.:

ℏ2 (k + hn )2
Hmn = δm,n + V (hm − hn ) (64)
2me
such that the secular equations reads:
 2
ℏ (k + hn )2
X 
c(k + hn ) δm,n + V (hm − hn ) − Eδm,n = 0, ∀m (65)
n
2me

w. the corresponding matrix form:


 ℏ2 (k + h )2 
n
− E δm,n + V (hm − hn ) = 0 (66)
2me

Now, lets consider the dispersion in the case of a weak periodic potential, and lets investigate what happens at the
BZ-boundary at cross-over.(see red dot on fig. 5)

Figure 5: Initial dispersion for weak periodic potential

9
here it turns out to be sufficient to use only two plane-waves in the expansion in eq. (56), and we therefore choose the
two plane-waves corresponding to the shown energy-parabolas:
1 1 2π
ϕ1 (x) = Wk+h0 (x) = √ eikx , ϕ2 (x) = Wk+h−1 (x) = √ ei(k− a )x (67)
L L
which mix at the BZ-boundary (lifts degeneracy and gives rise to two new energy states). Evaluating the two plane-waves
at the boundary:
1 π 1 π
ϕ1 (x) = Wk= πa +h0 (x) = √ ei a x , ϕ2 (x) = Wk= 2π
a +h−1
(x) = √ e−i a x (68)
L L
and the corresponding energies:
2 2
ℏ2 (k + h0 )2 ℏ2 πa ℏ2 (k + h−1 )2 ℏ2 −π
a
E1 = = , E2 = = ⇒ E1 = E2 ≡ E0 (69)
2me 2me 2me 2me
and choosing the expansion coefficients (given by taken Fourier transform of potential) in the potential as:

V (h0 ) ≡ 0, V (h−1 ) ≡ V1 , V (h1 ) ≡ V1 ∗ (70)

we get a secular equation (Note that this is exactly the characteristic polynomial that appears when trying to get
eigenvalues of some matrix operator) on the form:
 
E0 − E V1
=0 (71)
V1 ∗ E0 − E
E02 + E 2 − 2E0 E − |V1 |2 = 0 (72)
π
that, at k = a, has solutions:  π
E k= = E0 ± |V1 | (73)
a
s.t. the situation now corresponds that on fig. 6, w. a constant band gap opening of width 2|V1 |.

Figure 6: Pertubated dispersion for weak periodic potential

Notice how this corresponds to using time-independent perturbation theory to lifting degeneracy. Furthermore, note
that this is only correct at exactly k = πa , and not in the vicinity.
To determine how the dispersion looks in the vicinity, we add a little parameter, ki′ ≡ k + hi − ∆k, which, yields:
 2 
ℏ2 π
a −∆k
2me −E V1
=0 (74)
 2 
ℏ2 − π
 
∗ a −∆k
V1 2me −E

which, has solutions: s


ℏ2 ∆k 2 ℏ2 ∆k 2
E± (∆k) = E0 + ± 4E0 + |V1 |2 (75)
2me 2me
for ∆k = 0 gives same result as before. Now assuming small ∆k (only care about immediate vicinity), we can expand the
square root in powers of ∆k 2 and get:

ℏ2 ∆k 2 E0 ℏ2
 
E± (∆k) ≈ E0 + ± |V1 | + ∆k 2 + . . . (76)
2me |V1 |me

10
in which case we can obtain the effective mass11 in the immediate vicinity (differentiating w. respect to ∆k instead of k) as:

|V1 |me
m∗ = (77)
|V1 | ± 2E0

which, under the assumption of a weak potential, s.t. |V1 | ≪ E0 , is approx.:

|V1 |me
m∗ ≈ (78)
2E0
from which we see that the effective mass is proportional to the size of the band-gap opening (remember size of band gap
was given as 2|V1 |). Can also see qualitatively - the less curvature of bands at band gap opening the bigger the opening and
the bigger the effective mass.

Tight binding (LCAO) - 1D


Good for qualitative understanding. Typically not used to make super precise calculations.
Here we start out w. 1-dimensional atom hamiltonian:
ℏ2 d2
Ĥa = − + Va (x), with Ĥa ϕa (x) = Ea ϕa (x) (79)
2me dx2
here we will only be concerned w. the lowest energy orbital states (1s) - not super realistic, but will become more accurate
later.
We will consider one electron approximation12 , in the case of having one atom centered at each lattice site tn = an
(now not in reciprocal space, but real space), we get a real space potential curve given as on fig. 7

Figure 7: 1D potential curve in real space

w. each atom contributing w. some Coulomb like potential, and the total potential as a simple sum over each
(superposition principle): X
V (x) = V (x − tn ) (80)
n

and the total hamiltonian for solid:


ℏ2 d 2
Ĥ = − + V (x) (81)
2me dx2
Now, from the Bloch theorem, we know that in a periodic potential, the wave-function that solves the SE, must be Bloch
wave, and therefore use a linear combination of orbital wave-functions ϕa (x − tn ) (which we assume to form an orthonormal
basis ⟨ϕa (x − tm )|ϕa (x − tn )⟩ = δmn 13 ) centered on the individual lattice sites, as our ansatz:
1 X iktn
ψk (x) = √ e } ϕa (x − tn ) (82)
N n | {z
c n

Note that these Bloch waves form an orthonormal system (as they should) ⟨ψk (x)|ψk′ (x)⟩ = δk,k′ .
Now, again, from the variational method, we know that the secular equation involves the matrix elements of the wave-
function, which, in turn, must therefore involve matrix elements w. respect to these orbital wave-functions, and we therefore
initially consider:
⟨ϕa (x − tm )| Ĥ |ϕa (x − tn )⟩ . (83)
We will go about this by splitting the matrix element up into two different parts:

11 The effective mass is defined as m∗ = ℏ2 (∂kk E(k))−1


12 Only one electron moving in the effective potential of the atoms
13 Equivalent to assuming that these orbital wave functions, has practically no overlap with neighbouring orbital wave functions.

11
1. On-site the part of the matrix element for which tn = tm .
2. Off-site the part of the matrix element for which tn ̸= tm .
To further simplify, we initially consider the action of the Hamiltonian on one of the orbital wave-functions:

−ℏ2 d2 X
Ĥϕa (x − tn ) = ϕ a (x − tn ) + Va (x − tm )ϕa (x − tm ) (84)
2me dx2 m
−ℏ2 d2 X
= ϕ a (x − tn ) + V a (x − t n )ϕ a (x − tn ) + Va (x − tm )ϕa (x − tn ) (85)
2m dx2
| e {z } m: m̸=n
Ĥ ϕ (x−t )
| {z }
a a n
∆Vn ϕa (x−tn )

where we call ∆Vn the rest potential, which is the resulting potential that we get when we remove the contribution of the
atomic potential at lattice site tn (see fig. 8).

Figure 8: 1D potential curve in real space for rest potential at tn = t1

and the on-site matrix element then becomes:

⟨ϕa (x − tn )| Ĥ |ϕa (x − tn )⟩ = Ea + ⟨ϕa (x − tn )| ∆Vn |ϕa (x − tn )⟩ (86)


| {z }
−β

where −β cannot really be further calculated, but it is clear, from the fact that it is very small as ∆Vn is very small at tn
(see fig. 8). (we set it to −β, because β > 0 for s-orbitals.)
Now, lets calculate off-site matrix element, where we only take neighbouring orbital-wave functions into account (fair
approx. as overlap between wave functions that are further away from each other in lattice, is very very small.)

⟨ϕa (x − tn )| Ĥ |ϕa (x − tn±1 )⟩ = Ea ⟨ϕa (x − tn )|ϕa (x − tn±1 )⟩ + ⟨ϕa (x − tn )| ∆Vn±1 |ϕa (x − tn±1 )⟩ (87)
| {z }
−γ

where γ > 0 for s-orbitals.


Now, lets consider how the dispersion looks:
* +
1 X iktm X
iktn
E(k) = ⟨ψk (x)| Ĥ |ψx (k)⟩ = e ϕa (x − tm ) Ĥ e ϕa (x − tn ) (88)
N m n
1 X X ik(tn −tm )
= e ⟨ϕa (x − tm )| Ĥ |ϕa (x − tn ))⟩ (89)
N n m

and now because of the periodic boundary conditions we have the same sum N times:
1 X ik(tn −tm )
E(k) = N e ⟨ϕa (x − tm )| Ĥ |ϕa (x − tn ))⟩ (90)
N n

and then setting tm = 0 without any loss of generality:


X
E(k) = eiktn ⟨ϕa (x)| Ĥ |ϕa (x − tn ))⟩ (91)
n

and then utilizing the fact the we only regard nearest neighbours:

E(k) = eikt−1 ⟨ϕa (x)| Ĥ |ϕa (x − t−1 ))⟩ + eikt1 ⟨ϕa (x)| Ĥ |ϕa (x − t1 ))⟩ (92)
−ika ika
=e ⟨ϕa (x)| Ĥ |ϕa (x + a))⟩ + e ⟨ϕa (x)| Ĥ |ϕa (x − a)))⟩ (93)

12
and then using the calculated on-site and off-site matrix elements:
h i
E(k) = Ea − β − γ eika + e−ika = Ea − β − 2γcos(ka) (94)

see fig. 9.

Figure 9: 1D dispersion relation for tight binding model (in reciprocal space).

We see that we have a band width of 4γ, and as such, that it is proportional to the overlap/hopping integral γ.
Now, for small k, we can do a Taylor expansion (around k = 0) and approximate the cosine by a parabola:

E(k) ≈ Ea − β − γ + γak 2 (95)

which looks kinda like a free particle dispersion. From this we can also calculate the effective mass:
−1 ℏ2
m∗ = ℏ2 ∂kk = (96)
2|γ|a2

from which we can say that if γ (the overlap integral) becomes very small, corresponding to a very flat dispersion, the
electrons cannot move - their effective mass becomes very high. Note that as |γ| implicitly decreases exponentially w. a,
even if one very to increase a, the effective mass would still increase.

Furthermore, notice that by considering the wave-function eq. (82) (which is only for the s-orbital), it is clear that that the
relative phase between the orbital wave functions ϕa (x − tn ) is positive when k = 0 (because ei0tn = 1), and alternating
π
positive and negative for k = πa (because ei a tn = ±1) - see fig. 10

π
Figure 10: Relative phase at different 1d lattice points for s-orbital approx. in tight binding for k = 0 and k = a
.

Notice how k = 0 corresponds to bonding orbital and k = πa to anti-bonding orbital (k = 0 + tn has low energy and
k = πa + tn has high energy - see dispersion relation for s-orbital on fig. 9)

Dynamical aspects 1D
Lets consider the dynamics of Bloch electrons in 1D solids. Specifically we will see that Bloch electrons behave
similar to free electrons when exposed to an electric field, just with a different (effective) mass.

Lets start by considering a general Bloch wave function (1D):

ψk (x) = eikx uk (x) (97)

Here it is important to note that ψk (x) is not an eigenfunction of the momentum operator p̂ = −iℏ∂x 14 . We therefore note
that while free electrons has momentum ℏk, this is not the case for Bloch electrons. Instead, in this context, we denote

14 Can easily see by applying operator and noting result cannot be written as eigenvalue eq.

13
ℏk as the Crystal momentum.
Now, lets consider the group velocity of Bloch electron, which can be found from the dispersion relation:
1
v(k) = ∂k E(k) (98)

and note that in the case of dealing with an free electron, this gives a very reasonable result, as:

ℏ2 k 2
Free electron : E(k) = (99)
2me
such that:
1 ℏk
∂k E(k) = (100)
ℏ m
which makes good sense in the semi-classical sense, as p = mv ↔ v = p/m.
Now, the fact that this is actually also correct for Bloch electrons, can be seen by evaluating the expectation of the
”velocity operator” w. respect to the Bloch wave:

⟨ψk (x)| |ψk (x)⟩ (101)
m
which can be seen in video 1. week 2, or p. 39-40 in book. The result is that:
p̂ 1
v(k) = ⟨ψk (x)| |ψk (x)⟩ = ∂k E(k) (102)
m ℏ
from which is also becomes evident that the group velocity of the Bloch electron is directly proportional to the gradient
of the dispersion curve in consideration. We now discuss the rate of change of the crystal momentum under external fields.
To be specific, consider the effect of an external electric field Felectric (assumed uniform and directed in the positive x-
direction) on the dynamics of an electron in a given energy band. When the field acts on the electron for a small time d t ,
the carrier gains the energy:
1 dE 1
dE = (−e)Felectric vdt = (−e)Felectric dt ⇒ dk = (−1)Felectric dt (103)
ℏ dk ℏ
from which it is given that:
d(ℏk)
= −eFelectric (104)
| dt {z }
Acceleration theorem

even though this is derived in a quasi-classical manner, it still, in fact, holds for Bloch electrons.
Now from this we also find the general expression for the effective mass. Consider first the acceleration:

dv d 1 dE(k) 1 d2 E(k) dk 1 d2 E(k)


= = 2
= 2 (−e)Felectric (105)
dt dt ℏ dk ℏ dk dt ℏ dk 2
and then from F = ma ↔ a = F/m one must in general have:

dv 1
= ∗ (−e)Felectric (106)
dt m
and from this we see that we can define the effective mass, as:
−1
m∗ = ℏ2 ∂kk E(k) (107)

Metals and Insulators


The most accurate defining property of metals, is that the density of states15 at the Fermi energy, is zero:

Metal: D(EF ) = 0 (108)



Lets investigate the Conductivity in 1D, for a lattice periodic material of length L = N a, with k = n N a.

15 The energy dependent number of available states pr. energy interval dE.

14
Variational method
Lets consider the SE on the form of a eigenvalue equation:

Ĥψ(x) = Eψ(x) (109)

with the wave functions expanded in some discrete orthonormal basis {ϕn (x)}
X
ψ(x) = cn ϕn (x) (110)
n

s.t. ϕ∗m (x)ϕn (x)dx = δm,n .


R

Lets then taker inner product with some ϕm (x) (Left multiply with ϕ∗m (x) and integrate):
X Z XZ
cn ϕ∗m (x)Ĥϕn (x)dx = E ϕ∗m (x)ϕn (x)dx (111)
n n
X
cn Hmn = Ecm (112)
n
X
cn Hmn − Ecm = 0 (113)
n
X 
cn Hmn − Eδm,n = 0 (114)
n

Now, this is just for one specific m, the entire collection is a homogeneous system of linear equations, which can be
solved by the secular determinant (just determinant of matrix w. coefficients from equations):

Hmn − Eδm,n = 0 (115)

see simple example in video (week 1).

15

You might also like