Electrons in a Periodic Potential Bloch’s Theorem

5.1 Bloch’s Theorem

We have learned that atoms in a crystal are arranged in a Bravais lattice. This
arrangement gives rise to a periodic potential that has the full symmetry of the Bravais
lattice to the electrons in the solid.
Qualitatively, a typical crystalline potential may have the form shown in Fig. 5.1,
resembling individual atomic potential at the atomic site, but flattening off in the region
between adjacent ions. It turns out irrespective of the details of the crystalline potential,
the mere fact that the potential is periodic enables many important conclusions to be
drawn.

Fig. 5.1 (From A&M)

Let’s continue to consider the independent electron picture. Then the Schrödinger
equation for a single electron in a periodic potential, U(r), is:

2 2 
H  (   U (r ))  E (5.1)
2m

The resulting electron eigenstates can be shown to have the following important
properties according to Bloch’s Theorem:
 

 nk (r )  e ik r u nk (r ), (5.2)

where

unk(r + R) = unk(r) (5.3)

for all R in the Bravais lattice. With eqns. 5.2 and 5.3, an alternative form of Bloch’s
Theorem is:

(r + R) = eikR(r) (5.4)

1
Electrons in a Periodic Potential

real number.

2
Electrons in a Periodic Potential

wavevector

3
Electrons in a Periodic Potential Bloch’s Theorem

Recall that b1 · (b2 × b3) is the volume of a reciprocal lattice primitive cell. For a cubic
lattice, one readily see that Eqn. 5.22 gives
3k = (2)3/V , (5.23)

where V is the volume of the specimen. One can show that this result in fact applies to
any crystalline structures.
To gain further insight, we examine another proof of the Bloch’s Theorem. We start
with the observation that we can always expand any function obeying the Born-von
Karman boundary condition (eqn. 5.17) in the set of all plane waves that satisfy the same
boundary condition. This means that the expansion comprises terms with wavevectors
equal to the allowed Bloch wavevectors:
 
 (r )  

c e
k
ik r
. (5.24)
k

Because U(r) is periodic in the Bravais lattice, its plane wave expansion (i.e. Fourier
expansion) will only contain components with wavevectors that are the reciprocal
wavevectors.
  
U (r )  

U G
 exp(iG  r ) (5.25)
G
Obviously, for U(r) to be real,
U-G = UG* (5.26)

Substitute eqns. 5.24 and 5.25 in each term of the Schrödinger equation (eqn. 5.1):

p2 2 2  2 2  ikr
    k ck e .
2m 2m k 2m

      
U  (

U G eiG r )(

ck eik r ).  

U G ck ei ( k  G )r

G k k ,G

Put k’ = k + G then replace k’ by k,


U  
 
U G c k G e ik r (5.27)
k ,G

The Schrödinger equation becomes:

2 2 

 {( 2m
k   )c k  

U  c   }e
G k G
ik  r
0 (5.28)
k G

2 2
Or, ( k   )c k   U G ' c k G '  0 (5.29)
2m 
G'

It is convenient to limit the wavevectors k in eqn. (5.29) to those lying in the first
Brillouin zone (BZ). We can do so by writing k = q – G, where q lies in the first BZ.
2   2
( (q  G )   )c q G   U G ' c q G G '  0
2m 
G'

4
 Electrons in a Periodic Potential

Next, replace q by k, and G’ by G’-G:

2   2
( (k  G )   )c k G   U G ' G c k G '  0 , (5.30)
2m 
G'

where k lies in the first BZ. Equation (5.29) represents a set of equations relating ck with
the other coefficients ck-G, ck-G’, …. , corresponding to wavevectors differing from k by a
reciprocal lattice vector. It may help to write eqn. (5.30) in a matrix form:

· · · · · · · ·
· 2   2
(k  G2 )    U 0
U G G
1 2
U G
2
U G  G
1 2
U 2G
2
· c k  G
2
2m

· U G G
1 2
2   2
(k  G1 )    U 0 U G
1
U 2G
2
U G
2  G1
 · c k G
1
2m

· U G
2
U G
1
2 2
k   U0
U G
1
U G
2
· c k =0
2m

· U (G 
2  G1 )
U  2G
1
U G
1
2   2
(k  G1 )    U 0
U G
2  G1
 · c k G
1
2m

· U  2G
2
U (G G
1 2)
U G
2
U  (G
2  G1 )
 2   2
(k  G2 )    U 0
· c k G
2
2m

· · · · · · · ·

(5.31)

The matrix equation (5.31) provides an explicit form for the eigenvalue problem with the
plane wave states, eik·r, ei(k-G)·r, etc. constituting the working basis. Notice that there are
N different values of k in the first BZ and so there are N different problems being
expressed by eqn. (5.31), where N is the number of primitive unit cells in the specimen.

Given the above scheme, in which all the wavevectors are reduced to the ones lying in
the first BZ, any wavevectors k’ differing from k by a reciprocal lattice vector, i.e., k’ =
k + G, would end in the same eigenvalue problem as that found for k. Therefore, (k) is
periodic in k.

One may also perceive that there are many eigenvalue solutions k to eq. (5.30) or (5.31)
for each k. We may thus label the eigenvalues k by a “band index”, n, analogous to the
quantum number used to label the energies of a particle confined in a box. For a given n,
the set of k or (k) constitutes an energy band.

5
Electrons in a Periodic Potential

The solutions to each of the problems with a given k and n would give eigenfunctions
that are linear combinations of the basis functions, ei(kG)·r and the corresponding
eigenvalues, . Explicitly, we may write
  
nk(r) = G ck-Gei(k-G)·r  e ik r (

c k G e iGr ). = eik·r unk(r),
G
  
where u nk (r )  

c  e
k G
 iG  r
. , which clearly fulfills the condition, unk(r + R) = unk(r).
G
This proves the Bloch’s Theorem.

5.2 Some Consequential New Concepts

5.2.1 Crystal Momentum
For free electrons (for which U(r) = 0 in the Hamiltonian), the electron wavefunctions,
 
nk(r), are simultaneous eigenstates of the momentum operator, p  i , with k being
the momentum eigenvalue. But for Bloch electrons, their wavefunctions (which are of the
form shown in eqn. (5.2)) are in general NOT eigenstates of p since
 

 i nk (r )  i(e ik r u nk (r ))
  

 k  nk (r )  (e ik r )(i )u nk (r ) . (5.35)

Nevertheless, k is in many ways a natural extension of p for electrons in a periodic
potential, which you would appreciate later. It has therefore been called the crystal
momentum to emphasize this similarity, but it should always be borne in mind that it is
not the electron momentum.

Note that sincek(r) = G ck-Gei(k-G)·r, the probability of any scattering event where a k
state is scattered into a k’ state would involve the integral of k’(r)k(r) ~ G ck-Gei(k-G-
k’)·r
. This leads to the selection rule: k’ = k – G.

5.2.2 Energy Bands

As pointed out above, there are many eigenvalue solutions k to eqn. (5.30) for each k
that we may label by a “band index”, n. And the set of (k) for a given n constitutes an
energy band.

In addition, k is periodic in units of a BZ. It is customary to represent all k in the first

BZ by translating those “outside pieces” through a suitable reciprocal lattice vector back
into the first BZ. Such presentation scheme is called the reduced zone scheme. An
example is shown immediately below. We shall discuss the other common schemes later.

6
Electrons in a Periodic Potential

Example 1: Find the low-lying free-electron (U

= 0) energy bands, (k), in the [100] direction
of a simple cubic lattice with lattice constant a.

(k - G) = (ħ2/2m)((kx – Gx)2 + Gy2 + Gz2)

G (000)/(ħ2/2m) (kx00) Band

index, n
2
000 0 kx 1
100, -100 (2/a)2 (kx±2/a)2 2, 3
010, 0-10, (2/a)2 kx2 + (2/a)2 4, 5, 6, 7
001, 00-1
110, 101, 1- 2(2/a)2 (kx+2/a)2 + 8, 9, 10,
10, 10-1 (2/a)2 11
-110,-101,-1- 2(2/a)2 (kx-2/a)2 + (2/a)2 12, 13,
10,-10-1 14, 15
011, 0-11, 2(2/a)2 kx2 + 2(2/a)2 16, 17,
011, 0-1-1 18, 19

5.2.3 Mean Velocity

The mean velocity of an electron with a non-plane-wave wavefunction is determined by

its group velocity:
 1 
v n (k )   k  n (k ) (5.39)


This result is remarkable. It asserts that electrons in a periodic potential can forever
maintain a constant mean velocity that depends on which stationary eigenstates the
electrons are in despite of the strong interactions they have with the positive ions of the
lattice, as accounted for by V(r).

5.3 Density of States and van Hove Singularity

In Chapter 2 (pp. 6-7), we have derived the expression for the density of states, g() for
Sommerfeld’s free electron model, and used it to obtain the average energy and heat
capacity of the electronic system. The algebra can be extended in a straightforward
manner to the present problem of electrons in a periodic potential. Like before, the
average value of a quantity Q that is a function of n and k (or ) is given by:

Q  2 Qn (k ) (5.40)
n ,k

An intensive quantity q can be defined from Q:

7
Electrons in a Periodic Potential


Q d 3k 
q  lim  2  Q ( k ) (5.41)
(2 ) 3
n
V  V
n

We can re-write eqn. 5.41 as:

q   dg ( )Q( ) (5.42)
if

d 3k 
g ( )   g n ( )    3  ( n (k )   ), (5.43)
n n 4

where the integral is over a primitive cell. Clearly, gn()d is the number of allowed
states of the nth band, lying within the shell in k-space enclosed by the energy surfaces
n=  and n = +d. From eqn. (5.43), we can write,

dS n (k )  
g n ( )   S n ( ) k  (k ) (   n (k )), (5.44)
4 3


where dS n (k ) is the surface area element of the energy surfaces at point k (see Fig. 5.2)

and k  (k ) is the local distance between the two energy surfaces n=  and n = +
   
dat point k (such that d   k  n (k )  dk |  k  n (k ) | k  (k ) ). By substituting

Fig.5.2

5.55

Electrons in Weak Periodic Potential

 
k  (k )  d / |  k  n (k ) | in eqn. (5.44) and integrating over d, one obtains


dS n (k ) 1
g n ( )   S n ( )  . (5.55)
4 |  k  n (k ) |
3

At points where n(k) vanishes (i.e. maximum or minimum, whether local or global, of
the energy band in k-space), the integrand of eqn. (5.55) diverges although in three
dimensions the integral of gn() remains finite. Nevertheless, the local slope of the
density of states, dgn/d, may still diverge. Such divergence is called van Hove

8
Electrons in a Periodic Potential

singularity. Since dgn/d enters into all terms except the first one in the Sommerfeld
expansion (eqn. 2.44), van Hove singularities at the Fermi surface may lead to anomalies
in some thermodynamic quantities at low temperatures.

5.4 Electrons in a Weak Periodic Potential – A Simple Example

We shall examine the application of the Bloch’s theorem to the simple case of a weak
periodic potential. At first sight, this may appear to be a purely academic problem. But in
reality, this model provides good approximation to metals of groups I to IV (i.e. those
that only have s and p electrons outside the noble gas configuration.) There are two
reasons for this Nearly-Free Electron behavior:
(1) Due to the Pauli’s exclusion principle, the conduction electrons are excluded by
the core electrons and hence cannot go very near to the positive ion where the
potential is strong.
(2) In the regions where conduction electrons are allowed, their mere existence helps
screen out the electrical potential of the positive ions (see A&M Ch. 17), making
the potential much shorter range than 1/r.

In the presence of a periodic potential, eqn. (5.30) applies and says that
   
 k (r )  

c k G e i ( k G )r , (5.56)
G

2   2
and ( (k  G )   )c k G   U G ' G c k G '  0 (5.57)
2m 
G'

With a weak potential, we may treat the problem as a small perturbation to the
(unperturbed) free electron problem that has eigenvalue and eigenfunction:

2  2  
   k0  (k ) , and  0 k (r )  e ik r (5.58)
2m

By using the perturbation theory and assuming UG << k, we have:

| U G | 2
   k0  U 0    (U 3 ). (5.59)

G 0  
0

k
0
 
k G

Eqn. (5.59) is valid only if 0k  0k-G. In the case where there is a reciprocal lattice vector
G such that the states k and k-G are degenerate, we must diagonalize the part of the
Hamiltonian that couples these degenerate states. Below we show that degeneracy always
occurs at the Brillouin zone boundary:

Proof: Recall that any wavevector k on the Brillouin zone boundary satisfies:

9
Electrons in a Periodic Potential

k·G = G2/2 (5.60)

Hence, 0 k-G = (k – G)2/2m = (k2 + G2 – 2k·G)/2m = k2/2m = 0k (5.61)

Suppose there are only two degenerate states, such that

0k = 0k-G + O(UG)
and 0k - 0k-G’| >> UG if G’  G.

Then only those terms in (5.59) involving the coefficients ck and

ck-G dominate.
(0k ck  UG ck-G = 0 (5.62)

U-G ck + (0k-G ck-G = 0 (5.63)

These equations will have consistent solutions only if

0k UG
=0 (5.64)
U-G 0k-G

+ UG2

We consider the case where U(r) is attractive such that UG < 0. Equation (5.67) becomes:

10
Electrons in a Periodic Potential

|-(r)|2 |+(r)|2

Note that d/dk along the direction of G is zero at the BZ boundary. That means kk at
the BZ boundary k = G/2 is zero, or k is perpendicular to G there.

___________________________________________

Let’s now consider a 2-dimensional rectangular lattice, where ay = a’ < ax = a as shown at

right. The energy contour for  = A(ħ2/2m)(/a)2, where 1 <  < (a/a’)2, is shown in the
left panel of Fig. 5.4 below. The right panel shows the energy dispersion (k) along the
[10] direction. (Figure credit: NPO).
As the derivation of eqn. (5.66) is general, the effect of a small perturbation from
a weak periodic potential, U(r), is to produce a splitting in (k) at a Brillouin zone
boundary. For a BZ boundary at G/2, the amount of splitting is 2|UG| (see Fig. 5.4).
Hence, if any particular UG vanishes due to the structure of the Bravais lattice, energy
curves (k) crossing the BZ boundary along G will not suffer any splitting.

11
Electrons in a Periodic Potential

Fig. 5.4. The two arrows drawn at the BZ boundary indicates the direction of the electron
velocity at the respective points on the BZ boundary.

Fig. 5.5

Figure credit: NPO

12
Electrons in a Periodic Potential

Other zone schemes (z.s.): We have discussed the reduced z.s. There are also other
common z.s. such as the extended z.s. and repeated z.s. The figure below illustrates how
various z.s. are constructed and related.

Fig. 5.6 (A&M)

Extended zone
scheme

Reduced zone
scheme

Repeated zone
scheme

The physical motivation for the extended zone scheme is that, when an electron in
an applied E or B field reaches a Bragg plane (at k, say), an Umklap scattering process
takes it to k – G. If we strictly restrict our view to the first BZ (i.e., the reduced z.s.), the
U scattering causes a large discontinuity in the electron trajectory in the k space. But in
the repeated z.s., the trajectory continues smoothly into the next replica as shown at right
in the extended z.s. for the 2D rectangular lattice discussed above. The advantage of
working in the repeated z.s. becomes apparent when we consider the equation of motion
in the next chapter.

13
Electrons in a Periodic Potential

Fig. 5.7

(Figure from: NPO)

Quasi-1-D behavior: In the above example, the Bragg planes normal to bx are brought
close together by the assumption ax >> ay. If F lies within the gap between the side
pockets in the 2nd BZ, all the occupied states are in the first BZ, leaving the 2nd BZ
unoccupied. In the repeated z.s., the Fermi surface (FS) consists of two sinusoidal curves
(Fig. 5.7). Because the group velocities on the FS (arrows) are predominantly parallel to
ky, the conductivity along the y-axis is much larger than that along the x-axis. By simply
arranging for the ions to be closer along the y-axis, we attain quasi-1D electronic
behavior even though the starting SE is fully isotropic. Such FS are found in the organic
conductor (TMTSF)2PF6, a member of the Bechgaard Salts.

5.4 Fermi Surface

As in the Sommerfeld theory, the electron ground state (i.e the T = 0 state) of the Bloch
electrons is obtained by progressively filling the lowest available (i.e. unoccupied) energy
level until all the electrons are filled. With this procedure, two kinds of configurations
may result: (1) The uppermost populated energy band n is completely filled. (2) The
uppermost populated energy band n is partially filled. Since each energy level n(k) can
accommodate two electrons (each of spin +1/2 or –1/2), and there are N k-states in each
band (where N is the number of primitive unit cells), an energy band can accommodate
up to 2N electrons. Therefore, a completely filled uppermost band can occur only if the
number of electrons per primitive cell is even. If configuration (1) occurs, there will be an
energy gap between the highest occupied band (the valence band) and the lowest

14
Electrons in a Periodic Potential

unoccupied band (the conduction band). A large energy gap (such as several eV), that is
much bigger than kBTroom results in the solid behaving like an insulator. On the other
hand, a small energy gap (< ~1 eV) results in the behavior of an intrinsic semiconductor.
In case (2) where the uppermost populated energy band is partially filled, a constant
energy surface in k-space can be defined whereat the energy equals that of the highest
energy level occupied. This surface serves as the Fermi surface of the Bloch electrons.
Note that in the case where the uppermost band is completely filled, such definition of the
Fermi surface as the dividing surface between occupied and unoccupied states become
ill-defined since it can be anywhere within the energy gap. In this case, the chemical
potential is often employed to define the Fermi level, which can be shown, for intrinsic
semiconductors, to be near the middle of the band gap.

5.4.1 Fermi Surface in the Extended and Reduced Zone Scheme

For metals with valence > 1, the Fermi surface often lies outside the 1st BZ, and intercepts
several other BZ’s of higher order. Below we shall discuss how the different “pieces” of
the Fermi surface within each BZ of the same (higher) order can be reconstructed back
into the 1st BZ. In essence, the procedure reconstructes the Fermi surface in the reduced
zone scheme. We shall discuss how presenting the Fermi surface in various zone schemes
may give better insight to the energy band structure, and transport properties of the metal.

We first discuss how the higher order BZ’s are defined. Like the 1st BZ, they are also
defined by first drawing perpendicular bisectors to the lines joining any two points in the
reciprocal lattice. The nth BZ is the region containing points that can be reached from the
origin by crossing (n 1) perpendicular bisectors, but no fewer. Fig. 5.8 illustrates how
this definition applies to a 2D square lattice.

Fig. 5.8 (left: A&M,

right: NPO)

As one can see, the boundary of the nth BZ gets progressively more complex as n
increases. Nonetheless, various pieces of a BZ of order n can be reconstructed into the 1st
BZ through translations by a set of suitably chosen reciprocal lattice vectors. Figs. 5.9 -
5.13 show examples of how a Fermi surface may be reconstructed for various lattices.

15
Electrons in a Periodic Potential

Fig. 5.9 (Kittel)

5.9.

Fig. 5.10 (Kittel)

5.7

Fig. 5.11 (Kittel) 5.7,

16
Electrons in a Periodic Potential

Fig. 5.12 (Kittel)

Fig. 5.11.

Fig. 5.13 (Kittel)

5.11.

5.4.2 Hole Pockets (based on NPO)

Consider the square lattice and assume that we have a population of 2N electrons (two
valence electrons per ion) to occupy the 1st and 2nd BZ as shown in Fig. 5.8. If U(r) is
absent, the (unperturbed) Fermi surface is a circle whose area A equals that of the 1st BZ
(A contains N k-states). This is plotted as a faint circle in the left panel of Fig. 5.14
(credit to NPO). The perturbed FS (bold arcs) shows discontinuities reflecting the energy
gap at each intersection with Bragg planes. As shown, occupied states spill into the 2nd
BZ while an equal number of unoccupied states remain at the corners of the 1st BZ. By
translating the triangular segments by some G, we may rearrange the 2nd BZ patches so

17
Electrons in a Periodic Potential

that the FS re-assemble into

two elliptical volumes (upper
right panel). Because the
group velocities point
outwards, the lowest energy
state in the 2nd BZ is at the
center of each elliptical
volume. Similarly, we may
rearrange the patches in the
1st BZ to get a nominally
circular FS in the 1st BZ. In
this pocket, the group
velocity points inwards,
indicating that the k state Fig. 5.14
with the highest energy
within the 1st BZ is at the
center of the hold pocket. The
volume encloses an “air
bubble” of unoccupied states
trapped in the 1st BZ.
In the excitation
picture, we focus on the small
pockets in the 2nd BZ and the vacant states in the 1st BZ, and ignore the remaining ~N
occupied k states. We identify each unoccupied state k in the bubble as a hole excitation,
whose properties we shall discuss shortly.
At the start, we had a population of 2N potentially itinerant electrons. However,
because the unperturbed FS embraces almost the entire 1st BZ, most of these electrons are
actually precluded from responding to an applied E and/or H field by Pauli’s Exclusion
Principle. What remain are a small electron minority occupying states in the 2nd BZ and
an equal no. of vacant states in the 1st BZ.

18