Dept of Phys

M.C. Chang

NFE model is good for Na, K, Al… etc,

in which the lattice potential is only a
small perturbation to the electron sea.
Energy bands (Nearly-free electron model)
• Bragg reflection and energy gap
• Bloch theorem
• The central equation
• Empty-lattice approximation

• For history on band theory, see 半導體的故事, by 李雅明, chap 4

Bloch recalled,
The main problem was to explain how the electrons could sneak by
all the ions in a metal so as to avoid a mean free path of the order of
atomic distances. Such a distance was much too short to explain the
observed resistances, which even demanded that the mean free path
become longer and longer with decreasing temperature.

By straight Fourier analysis I found to my delight that the wave differed

from the plane wave of free electrons only by a periodic modulation.
This was so simple that I didn't think it could be much of a discovery,
but when I showed it to Heisenberg he said right away: "That's it!"

ψ k (r ) = eik ⋅r uk (r )

Hoddeson L – Out of the Crystal Maze, p.107

Nearly-free electron model
Free electron plane wave
2π
ψ k (r ) = eik ⋅r , k x = nx
... etc
L
• Consider 1-dim case, when we turn on a lattice potential with period a,
the electron wave will be Bragg reflected when k=±π/a, which forms
two different types of standing wave. (Peierls, 1930)

⎛πx ⎞ G
ψ + = c ( eiπ x / a + e − iπ x / a ) =
2
cos ⎜ ⎟, k ⋅G =
ˆ
a ⎝ a ⎠ 2
⎛πx ⎞
ψ − = c ( eiπ x / a − e − iπ x / a ) = i
2
sin ⎜ ⎟.
a ⎝ a ⎠
• Density distribution of the two standing waves
 • These 2 standing waves have different electrostatic energies.
This is the origin of the energy gap.
If potential V(x)=Vcos(2πx/a), then Note: Kittel use
a potential energy U (=eV)
Eg = ∫ dxV ( x)(−e) (| ψ − ( x) |2 − | ψ + ( x) |2 )
0

⎛ 2π x ⎞⎛ 2 π x 2 πx⎞
a
2e
= − V ∫ dx cos ⎜ ⎟⎜ sin − cos ⎟ = eV
a 0 ⎝ a ⎠⎝ a a ⎠
• Lattice effect on free electron’s energy dispersion
ψ−

ψ+

Electron’s group velocity is zero near the boundary of the 1st BZ

(because of the standing wave).
Q: where are the energy gaps when U(x)=U1 cos(2πx/a)+U2 cos(4πx/a)?
A solvable model in 1-dim: The Kronig-Penny model (1930)
(not a bad model for superlattice)

• Electron energy dispersion calculated from the Schrodinger eq.

Energy bands
(Kittel, p.168)
Bloch theorem (1928)

The electron states in a periodic potential is of the form

ψ k (r ) = eik ⋅r uk (r ) , where u (r+R)= u (r) is a cell-periodic function.
k k

• A simple proof for 1-dim:

|Ψ(x)|2 is the same in each unit cell. Ψ(x+a)=CΨ(x)
Consider periodic BC, ψ ( x + Na) = C Nψ ( x) = ψ ( x)
⎛ s ⎞
→ C = exp ⎜ 2π i ⎟ , s = 0,1, 2 , ( N − 1)
⎝ N⎠
⎛ s ⎞ s
write ψ k ( x) = exp ⎜ 2π i x ⎟ uk ( x), 2π =k
⎝ Na ⎠ Na
then uk ( x + a ) = uk ( x)

Similar proof can be extended to higher dimensions.

 important

• Schrodinger eq for ψ:
Lattice
p2 potential
Hψ k = ε kψ k , H= + U (r )
2m
ψ k ( r ) = e ik ⋅ r u k ( r )
uk(x) depends on the form of the periodic lattice potential.

1023 times less effort than

• Schrodinger eq for u: the original Schrodinger eq.

H (k )uk = ε k uk within one unit cell

where H (k ) ≡ e − ik ⋅r Heik ⋅r ← Effective Hamiltonian for uk(r)

⎛∇
2
⎞
2
= ⎜ + k ⎟ + U (r )
2m ⎝ i ⎠
 important
Allowed values of k are determined by the B.C.

Periodic B.C. ψ k (r + N i ai ) = ψ k (r ), i =1,2,3

(3-dim case)
→ eik ⋅ Ni ai = 1, ∀i N3

→ k ⋅ N i ai = 2π mi , mi ∈ Z , ∀i N2

m1 m m N1
→ k= b1 + 2 b2 + 3 b3
N1 N2 N3

b1 ⎛ b2 b3 ⎞ 1 b1 ⋅ (b2 × b3 )
Δk=
3
⋅⎜ × ⎟ = b1 ⋅ (b2 × b3 ) =N
N1 ⎝ N 2 N 3 ⎠ N Δk
3

N =N1 N 2 N 3
V
i a1 ⋅ (a2 × a3 ) = v = ,
N
(2π )3
i b1 ⋅ (b2 × b3 ) = ,
v
(2π )3
i Δk=3
, as in the free-electron case.
V
Therefore, there are N k-points in a BZ (a unit cell in reciprocal lattice),
where N = total number of primitive unit cells in the crystal.
Countings in r-space and k-space

r-space k-space

a crystal with PBC: 1st BZ:

N points here

N points here

• N unit cells • Infinite reciprocal lattice points

(N lattice points x 1 atom/point) • N k-points in 1st BZ

• If each atom contributes q • N k-points in an energy band

conduction electrons, then Nq
electrons in total.
Q: what if there are p atoms per lattice point?
Difference between conductor and insulator (Wilson, 1931)

• There are N k-points in an energy band, each k-point can be

occupied by two electrons (spin up and down).
∴ each energy band has 2N “seats” for electrons. conductor
E
• If a solid has odd number of valence electron per primitive
cell, then the energy band is half-filled (conductor).
For example, all alkali metals are conductors.
k
• If a
even
solidnumber
has even
of electrons
number ofper
valence
primitive
electron
cell, then
per primitive
there arecell,
2
then
possibilities:
the energy(a) no
band
energy
mightoverlap
be filledor(if(b)
filled,
energy
thenoverlap.
insulator).
E.g., alkali earth elements can be conductor or insulator.

insulator conductor
E E

k k
How do we determine uk(r) from lattice potential U(r)?

(
⎡ p+ k
) ⎤
2
Schrodinger equation ⎢ + U ( r ) ⎥ uk ( r ) = ε k uk ( r )
⎢ 2m ⎥
⎢⎣ ⎥⎦
Keypoint: go to k-space to avoid derivatives and simplify the calculation

Fourier transform
1. the lattice potential
U (r ) = ∑ U G eiG⋅r G=2πn/a
G

2. the wave function uk (r ) = ∑ Ck (G )e − iG⋅r k=2πn/L

G

2
k2
Schrod. eq. in k-space
(ε 0
k −G )
− ε k Ck (G ) + ∑ U G '−G Ck (G ') = 0, ε ≡
2m
0
k
aka. the central eq. G'

Kittel
uses λk
Matrix form of the central eq. (in 1D) G=ng (g≣2π/a)

⎛ ε k0+ 2 g − ε k Ug U2g U3g U4g ⎞ ⎛ Ck (−2 g ) ⎞

⎜ ⎟⎜ ⎟
⎜ U−g ε k0+ g − ε k Ug U2g U 3 g ⎟ ⎜ Ck ( − g ) ⎟
⎜ U −2 g U−g ε k0 − ε k Ug U 2 g ⎟ ⎜ Ck (0) ⎟ = 0
for a
⎜ ⎟⎜ ⎟ particular k
⎜ U −3 g U −2 g U−g ε k0− g − ε k U − g ⎟ ⎜ Ck ( g ) ⎟
⎜
⎝ U −4 g U −3 g U −2 g U−g ε k0− 2 g − ε k ⎟⎠ ⎜⎝ Ck (2 g ) ⎟⎠

• For a given k, there are many eigen-values εnk, with eigen-vectors Cnk.

• The eigenvalues εn(k) determines the energy band.

• The eigenvectors {Cnk(G), ∀G} determines the Bloch states.

Example: U(x) = 2U cos2πx/a

= U exp(2πix/a)+U exp(-2πix/a) (Ug=U-g=U)

⎛ ε k0+ 2 g − ε k U 0 0 0 ⎞ ⎛ C k ( −2 g ) ⎞
⎜ ⎟⎜ ⎟
⎜ U ε k0+ g − ε k U 0 0 ⎟⎜ k C ( − g ) ⎟
⎜ 0 U ε k0 − ε k U 0 ⎟ ⎜ Ck (0) ⎟ = 0
⎜ ⎟⎜ ⎟
⎜ 0 0 U ε k0− g − ε k U ⎟ ⎜ Ck ( g ) ⎟
⎜
⎝ 0 0 0 U ε k0− 2 g − ε k ⎟⎠ ⎜⎝ Ck (2 g ) ⎟⎠
 important
• What are the eigen-energies and eigen-states when U=0?

U(x)=0 εk C(-g)=1 C(0)=1 C(g)=1

ε3k
ε2k
ε1k
-2g -g 0 g 2g k
1st BZ

• when U(x)≠0, for a particular k, unk is a linear combination

of plane waves, with coefficients Cnk : unk ( x) = ∑ Cnk (G )e − iGx
G

• From the central eq., one can see that Ck +G ' (G + G ') = Ck (G )

unk ( x) = ∑ Cnk (G )e − iGx

G

⇒ un ,k +G ( x) = e − iGx unk ( x) ψ n ,k (r + R) = eik ⋅Rψ n,k (r );

ψ n , k + G ( x ) = ψ nk ( x ) ψ n ,k +G (r ) = ψ n ,k (r )

• Bloch energy εn,k+G = εnk (∴ info in the 1st BZ is enough)

Approximation of the central equation

• The Bloch state ψ nk ( x) = ∑ Cnk (G )e

i ( k −G ) x

G
is a superposition of … exp[i(k-g)x], exp[ikx], exp[i(k+g)x] …

• If k ～ 0, then the most significant component of ψ1k(x) is

exp[ikx] (little superposition from other plane waves).

• If k ～ g/2, then the most significant components of ψ1k(x) and

ψ2k(x) are exp[i(k-g)x] and exp[ikx], others can be neglected.

Truncation:

⎛ ε k0+ 2 g − ε k U 0 0 0 ⎞ ⎛ Ck (−2 g ) ⎞
⎜ ⎟⎜ ⎟
⎜ U ε k0+ g − ε k U 0 0 ⎟⎜ C k ( − g ) ⎟
⎜ 0 U ε k0 − ε k U 0 ⎟ ⎜ Ck (0) ⎟ = 0
⎜ ⎟⎜ ⎟
cutoff for
⎜ 0 0 U ε 0
k−g − εk U ⎟⎜ k C ( g ) ⎟ k～g/2!
⎜
⎝ 0 0 0 U ε k − 2 g − ε k ⎟⎠ ⎜⎝ Ck (2 g ) ⎟⎠
0
Energy levels near zone boundary k ～ g/2

• Cut-off form of the central eq. ⎛ ε k0 − ε U ⎞ ⎛ C (k ) ⎞

⎜⎜ ⎟⎟ ⎜ ⎟ =0
⎝ U ε k0− g − ε ⎠ ⎝ C (k − g ) ⎠

• Energy eigenvalues • Energy eigenstates

(
1 0
ε k + ε k0− g ) ±
1
(ε − ε k0− g ) + ( 2U )
2
ε ± (k ) = 0 2

2 2
k ψ ± k ( x) = C± (k )eikx + C± (k − g )ei ( k − g ) x
g
define k = k −
2 ψ + ,k
k 2 ⎛ 2ε g /2 ⎞
2 0

then ε ± (k ) = ε 0
±U + ⎜1 ± ⎟
2m ⎜⎝ U ⎟⎠
g /2

parabola
ψ −,k

⎧⎪cos (π x / a )
⇒ ψ ( x) ⎨
⎪⎩ sin (π x / a )
g
±,
2
 Kittel, p.225
3 ways to plot the energy bands:

(k ' ∈1st BZ)

ε n,k ' → ε k
(k = k '+ G )
1st
Brillouin
zone

Sometimes it is convenient to
repeat the domains of k

Fig from Dr. Suzukis’ note (SUNY@Albany)

 Nearly-free-electron model in 2-dim (energy bands)
• 0th order approx.: empty lattice (U(r)=0)
• 1st order approx.: energy gap opened by Bragg reflection

k
G
Laue condition k ⋅G =
2
G/2

→ Bragg reflection whenever k hits the BZ boundary

“Empty lattice” in 2D
2D square lattice’s reciprocal lattice

• Free electron in vacuum:

2
k2
εk =
2m

• Free electron in empty lattice:

(k′ + G )
2 2

ε k = ε nk ′ =
2m
k = k′ + G
k ′ ∈1st BZ

M
2π/a
• How to fold a parabolic “surface” back to the first BZ?

Γ X
Folded parabola along ΓX (reduced zone scheme)

• In reality, there are

energy gaps at BZ
boundaries because of
the Bragg reflection

• The folded parabola

along ΓM is different
M

Γ X

• Usually we only plot

the major directions, for
2D square lattice, they
are ΓX, XM, MΓ

2π/a
Empty Lattice in 3D

Simple cubic lattice

(k′ + G )
2 2

ε k = ε nk ′ =
2m

k = k′ + G
k ′ ∈1st BZ
Empty FCC lattice
1st Brillouin zone:

Energy bands for empty FCC lattice

along the Γ-X direction.
Comparison with real band structure

The energy bands for Actual band structure for

“empty” FCC lattice copper (FCC, 3d104s1)

d bands

From Dr. J. Yates’s ppt

Origin of energy bands - an opposite view

Tight binding model

(details in chap 9)

• Covalent solid • Alkali metal

• d-electrons in • noble metal
transition metals