Chapter 7

Electrons in periodic potentials

7.1 Schr odinger equation for periodic potentials and Blochs theorem
So far we have only considered free electrons. The energy dispersion of the allowed states was given by the
dispersion:
=

2
k
2
2m
, k
2
= |k|
2
= k
2
x
+k
2
y
+k
2
z
(7.1)
Until now we assumed the Schrodinger equation of a free particle to be valid

2
2m

2
= . (7.2)
The solutions of this Schrodinger equation are plane waves

k
(r) =
1

V
e
i

kr
. (7.3)
And the momentum was given by p =

k, the eigenvalues of the momentum operator. But now we want to take

into account a potential V (r) so we have to solve the following Schrodinger equation:

2
2m

2
+V (r)(r) = (7.4)
We assume the potential has the same periodicity as the underlying Bravais lattice:
V (r) = V (r +

R)

R of the appropriate Bravais lattice vector. (7.5)
What can we say about the solutions of this Schrodinger equation? The solutions of this equation satisfy Blochs
theorem which may be written in two equivalent forms. As a rst denition, we state that the solutions must
take the form:

k
(r) = exp(i

kr)u
n

k
(r) (7.6)
whereby u
n

k
(r) is a function with the same period as the lattice itself, i.e.
u
n

k
(r) = u
n

k
(r +

R)

R (7.7)
Consider now the solution at positive (r +

R):

k
(r +

R) = e
i

k(r+

R)
u
n

k
(r +

R) = e
i

R
e
i

kr
u
n

k
(r) = e
i

k
(r) (7.8)
Obviously the solutions change by a phase function exp(i

R) between lattice points; this is the second and

equivalent statement of Blochs theorem and serves to remind us of the translational invariance of the direct
lattice (as found before for phonons).
7.1.1 Proof of Blochs theorem
First we dene an operator

T

R
which has the following property,

R
f(r) = f(r +

R) (7.9)
38
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 39
i.e. it is simply a translation operator. Now we operate with

T

R
on our Hamiltonian:

H(r)(r) =

H(r +

R)(r +

R) =

H(r)(r +

R) =

H

T

R
(r) (7.10)
In other words the translation operator

T

R
commutes with the Hamiltonian as long as

H(r) is periodic:
[

R
,

H] = 0 (7.11)
Furthermore:

R
1

R
2
=

T

R
2

R
1
= (r +

R
1
+

R
2
) =

T

R
1

R
2
(7.12)
Thus

T(

R)

R and

H for a set of commuting operators. It is known from a basic theorem of quantum mechanics,
that in this case, the eigenstates of

H can be chosen to be simultaneous eigenstates of

T

R


R.

H = as usual and

T

R
= c(

R) (7.13)
c(

R) is quite generally some complex constant which we shall now determine. Furthermore

R
1

R
2
=

T

R
1
c(

R
2
) = c(

R
1
)c(

R
2
) and =

T

R
1
+

R
2
= c(

R
1
+

R
2
). (7.14)
Thus the eigenstates must satisfy the multiplicative property:
c(

R
1
)c(

R
2
) = c(

R
1
+

R
2
) (7.15)
Clearly this property can be described using an exponential form, and in particular for our primitive lattice
vectors a
i
we write:
c(a
i
) = exp(2ix
i
), with x
i
C (7.16)
Thus for some general

R = n
1
a
1
+n
2
a
2
+n
3
a
3
we nd:
c(

R) = c(a
i
)
n
i
c(a
2
)
n
2
c(a
3
)
n
3
(7.17)
Now consider an arbitrary reciprocal lattice vector:

k = x
1

b
1
+x
2

b +x
3

b
3
=

K (7.18)
which is not equal to some reciprocal lattice vector (i.e. x
i
/ Z). Then the scalar product between this vector
and

R from above yields

R = (x
1
n
1
+x
2
n
2
+x
3
n
3
) 2 (7.19)
using a
i

b
j
= 2
ij
. We have thus found the appropriate argument for our exponential form:
c(

R) = exp(i

R). (7.20)
Thus we conclude that
T

R
= (r +

R) = c(

R)(r) = exp(i

R)(r) which is exactly the statement of Bloch

s theorem given before.

(7.21)
7.1.2 Bloch states and Bragg reections
For a more physical picture of what is happening, let us consider electron scattering.
An electron is a wave with wave vector

k. What happens if this wave is scattered from a Bragg plane of atoms
as we considered for X-ray photons? For photons:
_

k

K
_
2
=

with the initial state

k, nal state

k

and

K a reciprocal lattice vector. Alternatively

k

K =
1
2
K denes a Bragg plane. (equation 4.9) (7.22)
In one dimension for a direct lattice (real space) of equally spaced points with separation a, the reciprocal lattice
is a set of equally spaced points, with spacing 2/a.
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 40
0

4
a

2
a
2
a
4
a
k
Figure 7.1: As discussed in the text for a 1D periodic lattice, the Bragg planes are separated by /a and correspond to
the boundaries of the Brillouin zones.
Scattering is clearly possible for electrons with wave vectors

k = n

a
_
=
1
2
K
_
. If we scatter elastically in one
dimension we clearly go from

k to

k

k. The rst reection occurs with wave vectors

k corresponding to to

k
BZ
.

a
+

a

Figure 7.2: What happens for our scattered free electron at /a as shown above?
If we consider a superposition of the travelling waves

k, and

k we nd that we no longer have purely travelling

waves but that standing waves become possible:
= c
1
exp
_
i
x
a
_
+c
2
exp
_
i
x
a
_
(7.23)
Clearly there are two orthogonal solutions, which are standing waves:
c
1
= c
2
or c
1
= c
2
. (7.24)
(+) = c exp
_
i
x
a
_
+c exp
_
i
x
a
_
= 2c cos
_
x
a
_
(7.25)
and
() = c exp
_
i
x
a
_
c exp
_
i
x
a
_
= 2ic sin
_
x
a
_
(7.26)
Both are standing waves but they clearly localise our free electrons at dierent locations in space:
(+) = |(+)|
2
cos
2
_
x
a
_
(7.27)
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 41
and
() = |()|
2
sin
2
_
x
a
_
(7.28)
x
U(x)
a
+ + +
Figure 7.3: Both the attractive potential as well as the distribution of the electron density are plotted schematically as
a function of position.
(+) solutions localise the electrons close to the attractive potential, hence decreasing the potential energy.
() solutions localise the electrons away from the attractive potential, hence increasing the potential energy.
For U(x) = U cos (2x/a) as a simple model for the potential energy, (+) =
_
2/a cos (x/a) and () =
_
2/a sin(x/a), where we have normalised our wave functions over one lattice period. We now calculate the
dierence in energy for the normalised wave functions over a period of the lattice.
E
g
=
_
a
0
dx U(x)
_
|(+)|
2
|()|
2

= 2
_
a
0
dx U(x) cos (2x/a)
_
cos
2
(x/a) sin(x/a)

= U (7.29)
This integral simply yields the value U and hence we conclude, that there is a gap in our energy dispersion! Our
two states are no longer degenerate.
7.1.3 Density of k-space in a periodic potential
The Born von Karman boundary condition in a periodic crystal gives:
(r +N
i
a
i
) = (r) i = 1, 2, 3
_

_
N
1
a
1
= L
1
,
N
2
a
2
= L
2
,
N
3
a
3
= L
3
.
(7.30)
with N
1
N
2
N
3
= N the number of primitive cells in the crystal.
We now apply Blochs theorem:

k
(r +N
i
a
i
) = exp(iN
i

k a
i
)
n

k
(r) =
n

k
(r) exp(iN
i

k a
i
) = 1 (7.31)
As before

k = x
1

b
1
+x
2

b
2
+x
3

b
3
and thus
exp(i2N
i
x
i
) = 1 (7.32)
or we require x
i
= m
i
/N
i
with m
i
integer. For the allowed values of Bloch wave vectors we then nd the following
relation:

k =
3

i=1
m
i
N
i

b
i
(7.33)
Thus the volume of

k-space per allowed value of k is given by

k =

b
1
N
1
_

b
2
N
2

b
3
N
3
_
=
1
N

b
1
(

b
2

b
3
)
. .
=V
r
(7.34)
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 42
or

k =
(2)
3
V
=
(8
3
)
NV
d
(7.35)
7.2 The meaning of band structure
Bloch theory introduces a wave vector

k but does this wave vector play the same role as the wave vector

k
of the Sommerfeld theory? It can easily be shown, that the wave vector

k introduced by Blochs theory is
not proportional to the particle momentum p/ as is the case for free electrons.

k
=

i

_
exp(i

kr)u
n

k
(r)
_
=

k
n

k
+ exp(

kr)

u
n

k
(r) = some multiple of
n

k
(7.36)
or in other words,
n

k
is not a momentum eigenstate.

k is referred to as the crystal momentum and

k is
a quantum number characteristic of the (discrete) translational symmetry of the periodic crystal.
The wave vector

k can always be chosen to lie in the rst Brillouin zone:

k +

K exp(i

R) = exp(i

R) exp(i

R)
. .
=1 by denition
(7.37)
And therefore
(r +

R) = exp(i

R)(r) = exp(i

R)(r) . (7.38)
The second index n of the wave function
n

k
is known as the band index. Expanding our Hamiltonian for
some arbitrary but xed

k gives

H =
_

2
2m

2
+U(r)
_
=
_
_

2
2m
_

i
+

k
_
2
+U(r)
_
_
u

k
(r) =

k
u

k
(r) (7.39)
with the usual periodic boundary condition
u

k
(r) = u

k
(r +

R). (7.40)
So we are dealing with an hermitian eigenvlaue problem, conned to a nite volume of space. (i.e. our
primitive cell - as given above) Thus we expect a set of discretely spaced eigenvalues for a given

k and they
are characterised by the quantum number n (band index).

k is only a parameter of the Hamiltonian.
The variation of the parameter

k will lead to a continuous variation of the eigenvalues
n

k
and this contin-
uous variation of
n

k
is what is meant by the term band structure. Conning

k to a single primitive cell
gives us a complete set of wave functions (for given n).

n,

k+

K
(r) =
n

k
(r) ;
n,

k+

K
=
n

k
(7.41)
For a given n, there is some

k
1
such that
n

k
1
is minimum. For a given n, there is also some

k
2
such that

k
2
is maximum. In between there is a continuous variation of
n

k
- the band structure.
What is the meaning of the dispersion?
1

n
(

k) (7.42)
We already encountered the dispersion relation of free elcetrons
=

2
k
2
2m

1

k
=

k
m
=
p
m
= v. (7.43)
Hence the gradient of the dispersion is the velocity of a free electron! Similarly the above quantity is used
(in the semiclassical interpretation of band structure) as the velocity (group velocity) for an electron of
momentum

k (superposition of states above

k). For the stationary states of

k there is a nite velocity v(

k)
which, despite scattering with the ions, remains constant for all times!
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 43
7.3 1D Model of band structure - the Kronig-Penney-Model
Consider the Schrodinger equation:
_

2
2m
d
2
dx
2
+U(x)
_
(x) = (x) (7.44)
with U(x) the periodic potential energy shown in Figure 7.4:
U(x)
0 -b a a+b
U
0
Figure 7.4: Model for a periodic potential
U(x) is a periodic potential with period (a + b). There are two separate solutions. For 0 < x < a we have
U(x) = 0 and the solution is a combination of plane waves:

1
(x) = AexpiKx +Bexp(iKx) (7.45)
The energy of the electrons in this region is given by =
2
K
2
/2m.
For b < x < 0 U(x) = U
0
and assuming < U
0
the solution of the Schrodinger equation in this area is
given by:

2
(x) = C exp(Qx) +Dexp(Qx) (7.46)
Obviously
2
(x) is a decaying and expanding solution. The energy can now be calculated as follows:
U
0
=

2
Q
2
2m
(7.47)
Blochs theorem now requires the periodicity of these solutions:
(a < x < a +b) is related to (b < x < 0) (7.48)

k
(r +

R) =

k
exp(i

R) (a < x < a +b) = exp(iK(a +b))(b < x < 0) (7.49)

This denes our wave vector k which characterises our solutions. How do we chose the amplitudes A,B,C and
D? For mathematical reasons one has to demand the continuity of the wave function and the rst derivative at
x = 0 and at x = a. This leads us directly to the following two conditions:
A+B = C +D (from (x = 0)) (7.50)
iK(AB) = Q(C D)
_
from
d
dx

x=0
_
(7.51)
And at x = a, using (a) = (b) exp(ik(a +b)) we nd
Aexp(iKa) +Bexp(iKa) = (C exp(Qb) +Dexp(Qb)) exp(ik(a +b)) (from (x)) (7.52)
iK (Aexp(iKa) Bexp(iKa)) = Q(C exp(Qb) Dexp(Qb)) exp(ik(a +b))
_
from
d
dx
_
(7.53)
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 44
The above four equations can be solved by writing them as a matrix, calculate the determinant of the matrix and
setting the determinant equal to zero. We already performed precisely the same procedure during our treatment
of phonons. Therefore we only state the nal result:
_
(Q
2
K
2
)
2Qk
_
sinh(Qib) sin(Ka) + cosh(Qb) cos (Ka) = cos (k(a +b)) (7.54)
A special case of the Kronig-Penney is found for a periodic set of -functions:
a x
Figure 7.5: In the limit of a periodic set of -functions the Kronig-Penney model reduces to the form discussed in the
text.
b 0: U
0
such that Q
2
ba/2 = P remains nite. Q K and QB 1:
Q
2
K2
2QK
Qb a sin(Ka) + 1 cos (Ka) = cos (ka) (7.55)

Q
2
ba
2

1
Ka
sin(Ka) + cos (Ka) = cos (ka) or
P
Ka
sin(Ka) + cos (Ka) = cos (ka) (7.56)
Transcendental equation for cos (ka) which must always be less than 1!
ka = 0
ka =
ka =
ka = 2
+1
1
Figure 7.6: The left hand side of Equation 7.56 above is plotted as a function of the parameter Ka. Solutions of the
transcendental equation can only be found with real values of k when the plotted function lies within the region [1, 1].
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 45
1BZ. 2BZ. 3BZ.

ka
0
2
3
Figure 7.7: The dispersion of the 1D periodic -function potentials are plotted in an extended zone scheme.
The above dispersion is plotted in the extended zone scheme. Usually the band structure is plotted in the reduced
zone scheme, i.e. we must map everything back into the rst Brillouin zone using appropriate translations

K.
Figure 7.8: The band structure (3D) of silicon is plotted along the important symmetry directions of reciprocal space.
In 1D band gaps always exist but become smaller for increasing energy (P/ka0) as can be seen above.
The general features of band structures in three dimensions are quite similar but can (and do) dier because of
extra dimensions! In 3D band gaps can exist but will depend explicitly upon the details of the calculation.
What is important, is how we ll the states according to the Pauli exclusion principle taking account of the
fact that a single band contains exactly N electrons with a spin degeneracy of two.
7.4 Band structure - part 2
7.4.1 The meaning of the Fermi surface
As seen before we expect to nd a set of functions
n
(

k) in k-space which dene the allowed energies of our elec-

trons in a periodic potential. We now have to ll these states according to the Pauli-principle for non-interacting
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 46
fermions. Put simply for the N states per band within the rst Brillouin-zone (which we remember completely
describes our solutions), we have in fact 2N states per band (g
s
the spin degeneracy when B = 0 is 2)
Depending upon how the bands align relative to each other we may thus expect one of the following solu-
tions. (remembering we always have nN electrons to t into the bands with n N)
Insulator
F
= (T = 0) lies midway between two bands, a Fermi surface does not exist, this case is shown in
a) of Figure 7.9.
(=0)=
F
]
]
empty
full
]
empty
full
1/2 full
a) b)
c)
empty
full
two partially filled bands
empty
full
almost empty , almost full
d)

Figure 7.9
Metal: For metals
F
= (T = 0) lies within one or more bands. A Fermi-surface exists, and in fact this is the
denition of a metal.
A metal is a solid with a Fermi-surface:
n
(

k) =
F
for one or more values of n.
If this is true for more than one value of n, we speak of branches of the Fermi-surface.(taken together the branches
build a set of constant energy surfaces). For completeness we now will discuss two special cases:
Semiconductor: If the energy gap E
g
is small compared with some energy scale (e.g. k
B
T), the insulator
with a small band gap is called a semiconductor. (For example silicon at room temperature (25 meV) has a band
gap of 1.1 eV)
Semimetal: If there is a small overlap of adjacent bands, then one band is almost full and the other almost
empty. The resulting Fermi-surface is small (in area of k-space), therefore the density of states is small and one
speaks of a semimetal.(Figure 7.9)
In free electron theory we found that to calculate important quantities (e.g. (T)) we had to know the density
of states g() which we weighted according to i) Fermi-Dirac statistics and ii) X, the quantity of interest.
X(T) =
_

0
g()X()f(, T) d , X(0) =
_

F
0
g()X() d (7.57)
The rst integral extends over all states. For T = 0 only the states beneath the Fermi-energy are occupied, it
therefore suces to integrate from zero to the Fermi-energy.
What can we say about the density of states g() for an arbitrary band
n
(

k)? And what consequences can this

have fore the evaluation of interesting physical properties? In principal we must evaluate simple sums:
X = 2

n,

k
X
n
(

k) , where the sum extends over all occupied states (7.58)

CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 47
For a large crystal this sum may be replaced by an integral for the

k-space distribution of states:
x = lim
V
X
V
= 2

n
_
d
3

k
(2)
3
X
n
(

k) (7.59)
Assuming that X
n
(

k) depends upon n and

k only through the energy, we may dene:
x =
_
d g()X() (7.60)
Where g() is the density of states, which is given be summing the individual densities of states of each band:
g() =

n
g
n
() and g
n
() =
_
d
3

k
(4
2
)
(
n
(

k)) (7.61)
An alternative and important picture can be found by considering, as in the case of phonons, the denition of
g()
g() d = no. of allowed states between energy and +d.
or more correctly (here without proof);
g() d = no. of allowed states between constant energy surfaces S
n
() and S
n
( +d).
S
n
( )
S
n
( + d) k( k)
g
n
( )
a)
b)
Figure 7.10: a) Surface of constant energy in k-space; b) typical density of states with characteristic van Hove singular-
ities.
g
n
() d =
_
S
n
()
dS
4
3
k(

k) (7.62)
But what is k(

k) ?
+d = +|

n
(

k)|k(

k) k(

k) =
d
|

n
(

k)|
(7.63)
g
n
() =
_
S
n
()
dS
4
3
1
|

n
(

k)|
(7.64)
As seen before for phonons the important point is the existence of van Hove singularities when
n
(

k) = 0. (at
the beginning and end of each band, and maybe in between) In therre or two dimensions these singularities lead
to discontinuities in the gradient of g() but not to a devergence. In 1D however quantities can become divergent
as seen by considering g
1D
() itself. The question that now arises, is how to calculate
n
(

k). But how does one

go about actually calculating the details of a particular crystal band structure? Do we actually know the periodic
potential U(r) in our periodic lattice?
Even if we can calculate the periodic potential arising from the ions, we seem to have completely neglected the
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 48
Coulomb interactions between electrons, but with what justication? If Coulomb interactions between electrons
and ions are important, then why not electron-electron interactions? In fact we havent done anything as bad as
this. We have assumed U(r) = U(r +

R) and found single electron levels
n

k
(r) which change between adjacent
cells of the direct lattice only by an unimportant phase-factor (

). Thus the
n

k
(r) are also periodic in

R and their Coulomb interactions may be, in some sense, be subsumed into U(r). But do we then know U(r)?
For our persued purposes it will suce to consider two simple models; one which emphasises the free electron
character of the band structure and one which shows the relations to atomic structure (tight binding).
7.4.2 Nearly free electrons
Mathematics: For a one dimensional lattice, the periodic potential can be expressed as follows;
U(x) =

K
U

K
exp(i

Kx) (7.65)
this is a Fourier expansion in functions periodic with

K from the reciprocal lattice. U(x) is real by virtue of the
fact that
U

K
= U

K
, (7.66)
we therefor can write Equation 7.65 as
U(x) =

K>0
U

K
_
exp(i

Kx) + exp(i

Kx)
_
=

K>0
2 U

K
cos

Kx. (7.67)
We now insert Equation 7.65 into the Schrodinger equation:
_
1
2m
p
2
+U(x)
_
(x) =
_
_
1
2m
p
2
+

K
U

K
exp(i

Kx)
_
_
(x) = (x) (7.68)
Now we expand the wave function in an Fourier series with
=

k
c(

k) exp(i

kx). (7.69)
Which values of

k do we take - only those compatible with Blochs theorem, i.e. which give us a solution compat-
ible with the Bloch form. This expansion of can also be inserted into the Schrodinger equation Equation 7.68.

2
2m
k
2
c(

k) exp(i

kx) +

k
U

K
c(

K) exp(i(

k +

k)x) =

K
c(

K) expi

kx (7.70)
This equation must be solved for all Fourier components:
(

k
)c(

k) +

K
U

K
c(

k

K) = 0 (7.71)
So we have to solve this set of algebraic equations.

K
=
2
k
2
/2m XXX Yes, but in practice often easier to XXX
e.q. consider the point at k = K/2 as before:
( )C
_
K
2
_
+UC
_

K
2
_
= 0 (7.72)
k =
K
2
, =

2
K
2
4.2m
only coecients in
K
2
considered, everything else ignored (7.73)
k =
K
2
:
( )C
_

K
2
_
+UC
_
K
2
_
= 0 (7.74)

U
U

= 0 or ( )
2
= U
2
(7.75)
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 49
= U =

2
2m
_
K
2
_
2
U (7.76)
equivalent to considering the periodic potential 2U cos (kx).
As before we expect the splitting of free electron dispersion into two bands:
Extension of nearly free electron model to two and three dimensions:
2U
k
x
1. B.Z.
2. B.Z.
2. B.Z.
3. B.Z.
3. B.Z.

a
a)
b)
Figure 7.11: XXX
almost full
high enery XXX in the corners are missing be-
cause there are states with lower energy just
outsider the first Brillouin-zone.
= ; surface of constant energy

F
1. ..
2. ..
surface of constant
energy
Figure 7.12: XXX
Increasing the density increasing the number of occupied branches of our free-electron band-structure.
What is wrong with this picture? Not much, only the neglect of some weak periodic potential!
energy gaps exist at zone boundaries
Fermi-surface will intersect zone boundaries perpendicular
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 50
1.
2.
3.
4.
full
Figure 7.13: XXX

toal volume enclosed by the Fermi-surface = density independent of details of electron interactions
Figure 7.14: XXX
Free electrons in a fcc. Bravais-lattice (direct) reciprocal lattice is bcc. (Wigner-Seitz) cell is 1. Brillouin-zone
is a truncated octahedron.
7.4.3 Tight binding method
Form an alternative point of view, a solid can be seen as a collection of weakly bound atoms and the Bloch states
are weakly modied atomic orbitals.
weighted radial wave-functions.
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 51
Figure 7.15: XXX
Figure 7.16: default
Signicant overlap exist only for the 3s states. Atomic levels are presumably a good description for the core
electrons (1s,2s,2p) but the valence electrons 3s will be modied from the simple atomic orbitals. Tight-
Binding-Approximation
Hamiltionian at lattice point is assumed to be that of a single atom and bound states are localised.

H
at

n
=
n

n
for the bound states. (r) for the crystal must full Bloch condition and also should be expandable from atomic
wave functions:

H(r) =
_

H
at
+ U(r)
_
r = (

k)(r) , with (r) =

R
exp(i

kr)(r

R), (r) =

n
b
m

n
(r) LACO approximation
(7.77)
N-fold degenerate levels
overlapping bands.
N values of k (N = no. of
primitive cells)
Figure 7.17: XXX
Points about tight binding:
n extends over atomic states close in energy ... to the atomic states XXX (e.g. XXX especially for dgenerate
CHAPTER 7. ELECTRONS IN PERIODIC POTENTIALS 52
atomic state)
low lying atomic states; localised; narrow bands... high energy states broad bands; spatially extended
v(

k) = 1//

k = 0 for bands which are not at. narrow bands v 0 localised (XXX)
comlicated when basis is complicated XXX