Solid State Physics PDF

SOLID STATE PHYSICS
By definition, solid state is that particular aggregation form of matter characterized by strong
interaction forces between constituent particles (atoms, ions, or molecules). As a result, a
solid state material has an independent geometric form (in contrast to liquids, which take the
form of the container) and an invariant volume (in contrast to gases/vapors) in given
temperature and pressure conditions. As temperature increases, a solid state material can
evolve into another aggregation form (liquid or gas). Solid state physics studies the structural,
mechanical, thermodynamic, electrical, magnetic, and optical properties of (poly-)crystalline
and non-crystalline solids (for example, amorphous materials, such as glass).
Crystal structure
The properties of crystalline solids are determined by the symmetry of the crystalline lattice,
because both electronic and phononic systems, which determine, respectively, the electric/
magnetic and thermal response of solids, are very sensitive to the regular atomic order of
materials and to any (local or non-local) perturbation of it. The crystalline structure can be
revealed by the macroscopic form of natural or artificially-grown crystals (see the pictures
below), or can be inferred from the resulting debris after cleaving a crystalline material.
(a) (b) (c)
(d) (e) (f)
Crystals of (a) baryt, (b) salt, (c) hexagonal beryl, (d) trigonal quartz, (e) monoclinic gypsum,
and apatite (f)
Crystal Structure 2
Non-crystalline materials have no long-range order, but at least their optical properties
are similar to that of crystalline materials because the wavelength of the incident photons (of
the order of 1 μm) is much larger than the lattice constant of crystals and so, photons “see” an
effective homogeneous medium. Other properties of non-crystalline materials are derived
based on concepts proper to crystalline solids and, therefore, the crystal structure is extremely
important in understanding the properties of solid state materials.
The macroscopic, perfect crystal is formed by adding identical building blocks (unit
cells) consisting of atoms or groups of atoms. A unit cell is the smallest component of the
crystal that, when stacked together with pure translational repetition, reproduces the whole
crystal. The periodicity of the crystalline structure that results in this way is confirmed by X-
ray diffraction experiments. The figures below illustrate crystals in which the basis consists of
(a) one atom and (b) two atoms.
(a) (b)
The group of atoms or molecules that forms, by infinite repetition, the macroscopic
crystal is called basis. The basis is positioned in a set of mathematical/abstract points that
form the lattice (also called Bravais lattice). So, a crystal is a combination of a basis and a
lattice. Although usually the basis consists of only few atoms, it can also contain complex
organic or inorganic molecules (for example, proteins) of hundreds and even thousands of
atoms.
In three dimensions, all Bravais lattice points
Rmnp = ma1 + na 2 + pa 3 (1)
can be obtained as superpositions of integral multiples of three non-coplanar primitive

translation vectors a1 , a 2 and a 3 (m, n, and p are arbitrary integers). A basis consisting of s
atoms is then defined by the set of vectors r j = m j a1 + n j a 2 + p j a 3 , j = 1,2,…,s, that
describe the position of the centers of the basis atoms with respect to one point of the Bravais
lattice. In general, 0 ≤ m j , n j , p j ≤ 1 .
Crystal Structure 3
Every point of a Bravais lattice is equivalent to every other point, i.e. the arrangement
of atoms in the crystal is the same when viewed from different lattice points. The Bravais
lattice defined by (1) is invariant under the operation of discrete translation
Tqrs = qa1 + ra 2 + sa 3 along integer multiples q, r and s of vectors a1 , a 2 and a 3 ,
respectively, because
Tqrs ( Rmnp ) = Tqrs + Rmnp = Rq + m,r + n,s + p (2)
is again a Bravais lattice point. In fact, since the translation operation is additive, i.e.
TqrsTuvw = Tq +u ,r +v,s + w , associative, i.e. Tqrs (TuvwTmnp ) = (Tqrs Tuvw )Tmnp , commutative, i.e.
−1
Tqrs Tuvw = TuvwTqrs , and has an inverse, such that Tqrs = T−q, − r , − s and Tqrs T−q, − r , − s = I
with I the identity transformation, it follows that the translations form an abelian
(commutative) group. Because condition (2) is satisfied for all Bravais lattice points, a1 , a 2
and a 3 are called primitive translation vectors, and the unit cell determined by them is called
primitive unit cell. The modulus of these vectors, a1 =| a1 | , a 2 =| a 2 | and a3 =| a 3 | , are the
lattice constants along the respective axes, and the volume of the primitive unit cell, which in
this case is a parallelepiped, is Ω =| (a1 × a 2 ) ⋅ a 3 | .
It is important to notice that the set of vectors a1 , a 2 and a 3 is not unique (see the
figures below for a two-dimensional lattice), but all primitive unit cells have the same
volume.
The primitive unit cell covers the whole lattice once, without overlap and without
leaving voids, if translated by all lattice vectors. An equivalent definition of the primitive unit
cell is a cell with one lattice point per cell (each lattice point in the figures above belong to
Crystal Structure 4
four adjacent primitive unit cells, so that each primitive unit cell contains 4×(1/4) = 1 lattice
point). Non-primitive (or conventional) unit cells are larger than the primitive unit cells, but
are sometimes useful since they can exhibit more clearly the symmetry of the Bravais lattice.
Besides discrete translations, the Bravais lattice is invariant also to the point group
operations, which are applied around a point of the lattice that remains unchanged. These
operations are:
• Rotations by an angle 2π / n about a specific axis, denoted by C n , and its multiples,
C nj = (C n ) j . Geometric considerations impose that n = 1, 2, 3, 4 and 6, and that
repeating the rotation n times one obtains C nn = E , where E is the identity operation,
which acts as r → r . Moreover, C1 = 2π = E does not represent a symmetry element.
C D
θ θ
A B
The allowed values of n can be determined assuming that we apply a rotation with an
angle θ around an axis that passes first through a point A and then through an adjacent
lattice point B. The points A and B are separated by the lattice constant a. If C and D
are the resulting points, they should also be separated by an integer multiple of a. From
the requirement that CD = a + 2a sin(θ − π / 2) = a − 2a cos θ = ma, or − 1 ≤ cos θ =
(1 − m) / 2 ≤ 1 , with m integer, it follows that m can only take the values −1, 0, 1, 2, and
3, the corresponding n = 2π / θ taking the values specified above. As for translations,
the rotations also form an abelian group.
Examples of two-dimensional figures with different rotation symmetries:
C2 C3 C4 C6
• Inversion I, which is defined by the operation r → −r if applied around the origin.

Crystal Structure 5
• Reflection σ j , which can be applied around the horizontal plane (j = h), the vertical
plane (j = v), or the diagonal plane (j = d).
• Improper rotation S n , which consists of the rotation C n followed by reflection in

the plane normal to the rotation axis. Note that S 2 ≡ I .
When we combine the point group symmetry with the translational symmetry, we
obtain the space-group symmetry. It is important to notice that the basis can introduce
additional symmetry elements, such as helicoidal symmetry axes and gliding reflection
planes. The figure bellow represents several symmetry operations: (a) translations, (b)
rotation, (c) inversion, and reflection with respect to a (d) vertical, and (e) horizontal plane.
(a) (b) (c) (d) (e)
Crystal lattices are classified according to their symmetry properties at point group
operations. There are 14 three-dimensional Bravais lattices, which belong to 7 crystal
systems, as can be seen from the figure below, where the primitive translation vectors are
denoted by a, b, c (with respective lengths a, b, and c), and α, β, γ are the angles between b
and c, c and a, and a and b, respectively. These crystal systems, which are different point
groups endowed with a spherical symmetric basis, are:
• cubic, for which a = b = c, α = β = γ = 90°. It consists of three non-equivalent space-

group lattices: simple cubic, body-centered cubic, and face-centered cubic. This
crystal system has the highest symmetry and is characterized by the presence of four
C 3 axes (the diagonals of the cube)
• tetragonal, for which a = b ≠ c, α = β = γ = 90°. It encompasses the simple and body-

centered Bravais lattices and contains one C 4 symmetry axis.
• orthorhombic, for which a ≠ b ≠ c, α = β = γ = 90°. It incorporates the simple, body-

centered, face-centered, and side-centered lattices and has more than one C 2
symmetry axis or more than one reflection plane (actually, three such axes/planes,
perpendicular to each other).
Crystal Structure 6
• hexagonal, for which a = b ≠ c, α = β = 90°, γ = 120°. It

is characterized by the existence of a single C 6
symmetry axis. The conventional hexagonal unit cell
(see the figure at right) is composed of three primitive
cells.
• trigonal, for which a = b = c, α = β = γ ≠ 90°. It contains a single C 3 axis.
• monoclinic, for which a ≠ b ≠ c, α = γ = 90°≠ β . It includes the simple and side-

centered lattices, and has one C 2 symmetry axis and/or one reflection plane
perpendicular to this axis.
• triclinic, for which a ≠ b ≠ c, α ≠ β ≠ γ ≠ 90°. This is the crystal system with the
lowest symmetry. It has no rotation axis or reflection plane.
Crystal Structure 7
The different crystal systems have different numbers of unit cell types because other
possible unit cell types cannot represent new Bravais lattices. For example, both the body-
centered and the face-centered monoclinic lattices can be reduced to the side-centered lattice
by appropriately choosing the primitive translation vectors.
Examples of two sets of primitive translation vectors for a body-centered cubic (bcc)
lattice are represented in the figure below at left and center, while the figure at right displays a
set of primitive translation vectors for a face-centered cubic (fcc) lattice.
The primitive translation vectors for the left figure above can be expressed as
a1 = (a / 2)( x + y − z ) , a 2 = (a / 2)(− x + y + z ) , a 3 = (a / 2)( x − y + z ) , (3)
while those for the right figure are
a1 = (a / 2)( x + y ) , a 2 = (a / 2)( y + z ) , a 3 = (a / 2)( z + x ) (4)
and the angles between these vectors are 60°.

A simple lattice has lattice points only at the corners, a body-centered lattice has one
additional point at the center of the cell, a face-centered lattice has six additional points, one
on each side, and a side-centered lattice has two additional points, on two opposite sides. The
simple lattices are also primitive lattices and have one lattice point per cell, since the eight
sites at the corners are shared by eight adjacent unit cells, so that 8×(1/8) = 1. The non-simple
lattices are non-primitive. The volume of the primitive unit cell in these lattices is obtained by
dividing the volume of the conventional unit cell by the number of lattice points. In particular,
the body-centered lattices have two points per unit cell: the eight at the corners which
contribute with 8×(1/8) = 1, and the one in the center, which belongs entirely to the unit cell.
The face-centered lattices have 4 lattice points per cell: those in the corners contribute with
Crystal Structure 8
8×(1/8) = 1, and those on the faces contribute with 6×(1/2) = 3, since they are shared by two
adjacent cells. Finally, the side-centered lattices have two lattice points per cell: the points at
the corner contribute with 8×(1/8) = 1, and those on the faces with 2×(1/2) = 1. If each lattice
point is expanded into a sphere with a radius equal to half of the distance between nearest
neighbors, such that adjacent spheres touch each other, then a packing fraction can be defined
as the fraction between the volume of the spheres contained in the conventional unit cell and
the volume of the unit cell. Note that in the volume between the spheres one can always insert
smaller spheres, which can stand for other atom types.
The 14 Bravais lattices incorporate all possible crystalline structures; they result by
taking into consideration the space-group symmetry, i.e. the symmetry at translations and the
point group symmetry of the lattice (the symmetry with respect to rotation, reflexion or
inversion). When the basis consists of only one atom, the Bravais lattice is identical to the
crystalline structure. But when the basis is complex and consists of several atoms, say s, the
crystalline structure can be seen as formed by the interpenetration of s Bravais lattices. The
Bravais lattices have always an inversion center in one of the lattice points, whereas such an
inversion center can lack in crystals with complex bases.
By counting the point groups of the possible different crystals (which have bases with
different symmetries), one ends with 32 crystalline classes that can be accommodated by the 7
crystal systems. Also, there are 230 space groups that result from the combination of the 32
crystalline structures with the translational symmetry.
Index system for lattice points, directions and planes

When the origin of the primitive translation vectors is a lattice point, another lattice point
with a position Rmnp = ma1 + na 2 + pa 3 is simply specified by the set of numbers [[m,n,p]]. A
negative integer m, n or p is denoted by a − sign placed on top of it. For example, [[ mn p ]]
stays for the lattice point specified by the integers m, −n and p, with m, n and p positive
numbers. In particular, for the three-dimensional primitive Bravais lattices the coordinates of
the lattice point at the origin are [[0,0,0]], the other lattice points differing only through
discrete translations along the three coordinate axis. The number of non-equivalent lattice
points in a Bravais lattice is given by the number of lattice points per unit cell. In particular,
for the body-centered lattice, the position of the lattice point at the center of the cube is
denoted by [[1/2,1/2,1/2]], the three additional lattice points in face-centered lattices having
coordinates [[0,1/2,1/2]], [[1/2,0,1/2]], [[1/2,1/2, 0]]. In a similar manner, depending on the
Crystal Structure 9
set of opposite sites they can occupy, the additional site in a face-centered lattice has the
coordinates [[0,1/2,1/2]], [[1/2,0,1/2]] or [[1/2,1/2,0]].
A direction, by definition, passes through two lattice points. To specify a direction in
a crystalline lattice, one uses the symbol [mnp], where m, n and p are three integers
determined by the following rule: since one can specify a direction by the coordinates
[[ m1 , n1 , p1 ]] and [[ m2 , n 2 , p 2 ]] of two points through which it passes, the indices m, n and p
are defined as the smallest integer numbers that satisfy the proportionality relations
m m2 − m1 n n2 − n1 p p 2 − p1
= , = , = , (5)
n n2 − n1 p p 2 − p1 m m2 − m1
or
m : n : p = (m2 − m1 ) : (n 2 − n1 ) : ( p 2 − p1 ) . (6)
If one of the integers is negative, the − sign is placed on top of the integer. For example,
[ mn p ] stays for the direction specified by the integers m, −n and p. If the direction is not
considered as an oriented axis but as a simple line, the direction specified by the integers m, n,
and p is the same as that specified by −m, −n, and −p (otherwise, the change of all signs
means a change of direction of the same line). If there are several equivalent directions
(equivalent, from the point of view of crystal symmetry), they are denoted as 〈mnp〉 . A
particular situation is encountered in the hexagonal lattice, in which lattice directions are
labeled by four numbers (this situation is not further discussed in this course).
Examples: The a1 axis is the [100] direction. The − a 2 axis is the [ 0 1 0 ] direction. Other
examples are illustrated in the figure below.
a3
a2
[110]
a1 [011]
[101]
[011]
Crystal Structure 10
In three-dimensional lattices, the orientation of a crystal plane is determined by three

non-collinear points in the plane. If each point is situated on a different crystal axis, the plane
is specified by the coordinates of the points in terms of the lattice constants a1 , a 2 , and a 3 .
Another way to specify the orientation of a plane, which is more useful for structure analysis,
involves the determination of three indices, called Miller indices, according to the rule:
• Find first the intercepts of the plane on the axes in terms of lattice constants a1 , a 2 ,
and a3 , irrespective of the nature (primitive or non-primitive) of the unit cell.
• Take the reciprocal of these numbers.
• If fractional, reduce these numbers to the smallest three integers, say m, n, p, with the
same ratio. The result, symbolized by (mnp) (or (mn p) if the second index, for
example, is negative), is the Miller index system of the plane.
The Miller index for an intercept at infinity is zero. The faces of a cubic crystal, for
example, are denoted by (100), (010), (001), ( 1 00) , (0 1 0) , and (00 1 ) . Moreover, the plane
(200) is parallel to (100), but cuts the a1 axis at a / 2 . If, from the point of view of crystal
symmetry, there is a set of nonparallel equivalent planes, they are symbolized as {mnp}. For
example, the set of faces of a cubic crystal is {100}. Again, for the hexagonal lattice there are
four Miller indices instead of three. Examples of Miller indices are given in the figures below.
(001) (101) (111)

Note that the Miller indices determine not only one plane but a family of parallel
planes, since there is an infinite number of planes with the same indices, all of which cut the
coordinate axes at s / m , s / n , and s / p , with s integer. The plane that cuts the axes at 1 / m ,
1 / n , and 1 / p is the closest to the origin from the family of parallel planes.
Note also that the planes with Miller indices (sm,sn,sp) are parallel with the plane
(mnp), but the distance between them is s times smaller. For example, the set of planes (222)
is parallel to but twice as close as the (111) set of planes.
In cubic crystals, the plane (mnp)
is perpendicular to the direction [mnp]
with the same indices (see the example
in the figure at right), but this result
cannot be extended to other crystal
systems.
Simple crystal structures
One of the most simple crystal structures and, at the same time, of general interest, is that of
NaCl (sodium chloride). It is illustrated below. The lattice is face-centered cubic, with a basis
consisting of one Cl− ion (blue) at [[000]] and a Na+ ion (green) at [[1/2,1/2,1/2]]. As can be
seen from the figure below, a unit cube consists of four NaCl units, with Na+ ions at positions
[[1/2,1/2,1/2]], [[0,0,1/2]], [[0,1/2,0]], and [[1/2,0,0]] and Cl− ions at [[000]], [[1/2,1/2,0]],
[[1/2,0,1/2]], and [[0,1/2,1/2]]. Each atom has as nearest neighbors six atoms of opposite kind.
Example of crystals with this structure and their lattice constants are given below.
Crystal a(Å) Crystal a (Å) Crystal a (Å)

LiF 4.02 KBr 6.60 MgO 4.21
LiBr 5.50 AgBr 5.77 MnO 4.43
NaCl 5.64 AgF 4.92 MgS 5.20
NaI 6.47 CaSe 5.91 PbS 5.92
KCl 6.29 BaO 5.52 SrTe 6.47
Another common structure is that of CsCl (other crystals with the same structure are
given in the table below). The lattice is in this case simple cubic, with a basis consisting of
one Cs+ ion (red) at [[000]], and one Cl− ion (green) at [[1/2,1/2,1/2]]. The number of nearest
neighbors (of opposite kind) is eight.
Crystal a (Å) Crystal a (Å) Crystal a (Å)

AlNi 2.88 CsCl 4.12 TlCl 3.83
CuZn (β-brass) 2.94 CsBr 4.29 TlBr 3.97
AgMg 3.28 CsI 4.57 TlI 4.20
The crystal structure of diamond (and also of Si and Ge semiconductors) is

represented below.
Crystal a (Å)
C (diamond) 3.57
Si 5.43
Ge 5.66
α-Sn (grey) 6.49
It is a face-centered cubic (fcc) lattice with a basis consisting of two identical atoms, with
coordinates [[000]] and [[1/4,1/4,1/4]]. Alternatively, diamond can be viewed as being formed
from two interpenetrating fcc lattices, displaced by 1/4 of the volume diagonal. Since the
conventional unit cell of the fcc lattice contains 4 lattice points, the conventional unit cell of
diamond has 2×4 = 8 atoms. No primitive cell exists that contains only one atom. In diamond,
each atom has 4 nearest neighbors and 12 next nearest neighbors. It is usually encountered in
materials where the covalent bonding prevails. Note that, although a fcc lattice, the packing
fraction of the diamond structure is only 0.34.
A closely related crystal structure to that of the diamond is the cubic zinc sulfide (zinc
blende structure). It differs from diamond in that the two atoms of the basis are different (in
this case, Zn and S). The conventional unit cell contains four molecules, the Zn atoms (dark
blue in the figure below) being placed at the positions [[000]], [[0,1/2,1/2]], [[1/2,0,1/2]] and
[[1/2,1/2,0]], whereas the S atoms (green) occupy the positions [[1/4,1/4,1/4]], [[1/4,3/4,3/4]],
[[3/4,1/4,3/4]], and [[3/4,3/4,1/4]]. Each atom is surrounded by four equally distant atoms of
the opposite kind, placed in the corners of a regular tetrahedron.
Crystal a (Å) Crystal a (Å) Crystal a (Å)

SiC 4.35 AlP 5.45 InAs 6.04
ZnS 5.41 AlAs 5.66 InSb 6.48
ZnSe 5.67 GaAs 5.65 SiC 4.35
MnS (red) 5.60 GaSb 6.12 CuCl 5.41
CdS 5.82 GaP 5.45 CuBr 5.69
CdTe 6.48 AgI 6.47 HgSe 6.08
Unlike in the diamond structure, where there is a center of inversion at the midpoint of every
line between nearest-neighbor atoms, such inversion centers are absent in the zinc blende
structure. This is an example of additional symmetry operations related to the basis of the
crystal structure.
The hexagonal close-packed (hcp) crystal structure can be obtained from the
hexagonal Bravais lattice if the basis consists of two atoms (blue and red in the figure below,
left) and if the atoms in one plane, which touch each other, also touch the atoms in adjacent
planes. The packing fraction in this case is 0.74 (as in fcc lattices), and is maximum. This
crystal structure is found in the solid state of many elements, as can be seen from the table
below. The hcp structure can be viewed as vertical arrangement of two-dimensional
hexagonal structures, such as the spherical atoms in the second layer are placed in the
depressions left in the center of every other triangle formed by the centers of the spherical
atoms in the first layer. The third layer of atoms is then placed exactly above the first, the
fourth above the second, and so on. This kind of arrangement is called ABAB… In an ideal
hcp structure, the height between the first and the third layers (the height along the c axis in
the figure below) is c = 8 / 3a = 1.63a. Because the symmetry of the hcp lattice is
independent of the ratio c/a, in real hcp structures this ratio can take values close to, but not
exactly identical to the ideal 1.63 value (see the table below).
Crystal a (Å) c/a Crystal a (Å) c/a

He 3.57 1.63 Mg 3.21 1.62
Be 2.29 1.58 Ti 2.95 1.58
Nd 3.66 1.61 Zr 3.23 1.59
Zn 2.66 1.86 Y 3.65 1.57
Cd 2.98 1.88 Gd 3.64 1.59
α-Co 2.61 1.62 Lu 3.50 1.58
If the c/a ratio differs considerably from the ideal

1.63 value, the hexagonal structure is no longer close-
packed. This is the case of graphite, for example (see the
figure at right), with a = 1.42Å and c = 3.40 Å, which
implies that c/a = 2.39.
A structure closely related to hcp is wurtzite, generally encountered in binary com-
pound semiconductors such as ZnS (wurtzite), ZnO, BN,
CdS, CdSe, GaN, AlN, but sometimes also in ternary
compounds such as Al0.25Ga0.5N. In binary compounds
(see the figure at right), each element has a hcp structure,
and the crystal is formed by interpenetrating two such
structures, so that an atom in one hcp lattice is
equallydistanced from the atoms in the other hcp lattice.
The crystal structure of the elements in the periodic table is indicated in the figure
below. Note that several elements can suffer transitions from one crystalline structure to
another depending on the external conditions: temperature, pressure, etc. In the table below
dhcp stands for double hexagonal closed-packed (the height of the cell along the direction
normal to the hexagonal planes is twice that in the hcp structure)
Lattice constants of some elements that crystallize in the fcc crystal structure:
Crystal a (Å) Crystal a (Å) Crystal a (Å) Crystal a (Å) Crystal a (Å)
Ar 5.26 Au 4.08 Cu 3.61 Ni 3.52 Pt 3.92
Ag 4.09 Ca 5.58 Kr 5.72 Pb 4.95 Sr 6.08
Al 4.05 β-Co 3.55 Ne 4.43 Pd 3.89 Xe 6.2
Lattice constants of some elements that crystallize in the bcc crystal structure:
Crystal a (Å) Crystal a (Å) Crystal a (Å) Crystal a (Å)
Ba 5.26 Fe 4.08 Mo 3.61 Rb 3.52
Cr 4.09 K 5.58 Na 5.72 Ta 4.95
Cs 4.05 Li 3.55 Nb 4.43 V 3.92
W 6.08
Reciprocal lattice
The concept of reciprocal lattice is directly connected with the periodicity of crystalline
materials and of their physical properties (such as charge density, electric field distribution,
etc.). Since the crystal is invariant under any translation with a Bravais lattice vector
Rmnp = ma1 + na 2 + pa 3 (1)
for any integers m, n or p, any function ϕ with the same periodicity as the crystalline lattice
must satisfy the relation
ϕ (r ) = ϕ (r + Rmnp ) , (2)
where r = ( x1 , x 2 , x3 ) is an arbitrary position vector with coordinates x1 , x 2 , and x3 in the
(generally non-orthogonal) system of coordinates determined by a1 , a 2 , and a 3 . This implies
ϕ ( x1 , x2 , x3 ) = ϕ ( x1 + ma1 , x 2 + na2 , x3 + pa3 ) (3)
or, for a function that can be expanded in a Fourier series
ϕ ( x1 , x 2 , x3 ) = ∑ ϕG exp[i (G1 x1 + G2 x 2 + G3 x3 )] (4)

G1 ,G2 ,G3
it follows that, for any m, n, and p,
exp(imG1a1 ) = 1 , exp(inG2 a 2 ) = 1 , exp(ipG3 a3 ) = 1 . (5)
Thus, Gi , with i = 1, 2, 3, can only take discrete values Gi = 2πsi / ai , and (4) becomes
ϕ (r ) = ∑ ϕ G exp(iG ⋅ r ) (6)
s1 , s2 , s3
where
G = s1b1 + s 2 b2 + s3 b3 (7)
is a vector in a coordinate system defined by the vectors bi , i = 1,2,3, such that

Reciprocal lattice 2
bi ⋅ a j = 2πδ ij . (8)
Due to (6), the vectors G can be understood as wavevectors of plane waves with the periodicity
of the reciprocal lattice and wavelengths 2π / | G | , similar to wavevectors in optics that are
perpendicular to wavefronts and have wavenumbers 2π / λ , where λ is the optical wavelength.
Similar to the Bravais lattices that are constructed starting with the primitive vectors ai ,
one can define a reciprocal lattice in terms of the primitive vectors bi , such that G in (7) are
points in the reciprocal lattice. A reciprocal lattice can only be defined with respect to a given
direct lattice. As demonstrated in the following, the G vectors have dimensions (and meaning
of) wavevectors related to plane waves with the periodicity of the direct lattice.
If the vectors ai are chosen and the volume of the primitive cell in the direct space is
Ω =| (a1 × a 2 ) ⋅ a 3 | , the vectors bi can be chosen as
b1 = (2π / Ω)(a 2 × a 3 ), b2 = (2π / Ω)(a 3 × a1 ), b3 = (2π / Ω)(a1 × a 2 ) . (9)
It follows then that the volume of the primitive cell of the reciprocal lattice is given by
Ω rec =| b1 ⋅ (b2 × b3 ) |= ( 2π ) 3 / Ω . (10)
An example of direct and corresponding reciprocal lattices in two dimensions is given below.
For a1 = d ( x − y ) , a 2 = d ( x + y ) , from (8) we get b1 = (π / d )( x − y) , b2 = (π / d )( x + y ) .

In three dimensions, the reciprocal lattices for the Bravais lattices in the cubic system
are summarized in the table below. The reciprocal lattice of a cubic lattice is also cubic since,
in this case, if x, y, z are orthogonal vectors of unit length, a1 = ax, a 2 = ay, a 3 = az and
Ω = a 3 , from (10) it follows that b1 = (2π / a) x, b2 = (2π / a) y, b3 = (2π / a) z , i.e. the

reciprocal lattice is simple cubic with a lattice constant 2π / a .
Real space Reciprocal space

Lattice Lattice constant Lattice Lattice constant
SC a SC 2π / a
BCC a FCC 4π / a
FCC a BCC 4π / a
Analogously, the reciprocal lattice to the bcc lattice with (see the first course)
a1 = (a / 2)( x + y − z ) , a 2 = (a / 2)(− x + y + z ) , a 3 = (a / 2)( x − y + z ) , and Ω = a 3 / 2 has
primitive vectors b1 = (2π / a)( x + y ) , b2 = (2π / a)( y + z ) , b3 = (2π / a)( z + x ) , i.e. is a fcc
lattice with a volume (of the primitive unit cell) in reciprocal state of Ω rec = 2( 2π / a ) 3 ,
whereas the reciprocal lattice of the fcc lattice, with a1 = (a / 2)( x + y ) , a 2 = (a / 2)( y + z ) ,
a 3 = (a / 2)( z + x ) , and Ω = a 3 / 4 is a bcc lattice with Ω rec = 4(2π / a ) 3 and primitive vectors
b1 = (2π / a)( x + y − z ) , b2 = (2π / a)(− x + y + z ) , b3 = (2π / a)( x − y + z ) . In both cases the
cubic structure of the reciprocal lattice has a lattice constant of 4π / a .
Observation: The reciprocal lattice of a reciprocal lattice is the direct lattice.
Because the product of a primitive Bravais lattice vector and of a primitive vector of the
reciprocal cell is an integer multiple of 2π , i.e. that
G mnp ⋅ Rhkl = 2π (mh + nk + pl ) , (11)
for all integers m, n, p and h, k, l, it follows that exp(iG ⋅ R) = 1 for any vector R in the Bravais
lattice and any vector G in the reciprocal lattice. This implies that the function exp(iG ⋅ r ) has
the same periodicity as the crystal because exp[iG ⋅ ( r + R )] = exp(iG ⋅ r ) exp(iG ⋅ R ) =
exp(iG ⋅ r ) . As a consequence,
∫cell exp(iG ⋅ r )dV = Ωδ G , 0 (12)
and the set of functions exp(iG ⋅ r ) form a complete, orthonormal basis for any periodic
function which has the same periodicity as the crystal, i.e. which can be written as
ϕ (r ) = ∑ ϕ G exp(iG ⋅ r ) . (13)
G
If the formula above is regarded as a Fourier transformation of the periodic function ϕ, the
coefficients ϕG can be retrieved by performing an inverse Fourier transformation:
ϕ G = Ω −1 ∫cell ϕ (r ) exp(−iG ⋅ r )dV . (14)
Relations between the direct and reciprocal lattices

One geometrical property that can be easily shown is that the reciprocal lattice vector
G mnp = mb1 + nb2 + pb3 (15)
is perpendicular to the plane (actually, to the set of parallel planes) with Miller indices (mnp) in
the Bravais lattice. The closest plane to the origin from the set of planes (mnp) cuts the ai
coordinate axes at a1 / m , a 2 / n , and a3 / p , respectively.

To show that (mnp) is perpendicular to G mnp it is sufficient to demonstrate that G mnp is
perpendicular to two non-collinear vectors in the (mnp) plane, which can be chosen as
u = a 2 / n − a1 / m , v = a 3 / p − a1 / m , (16)
and satisfy, indeed, the relations

a3
u ⋅ G mnp = v ⋅ G mnp = 0 (17) a3/p
because of (8). Then, it follows that the normal to

n a2/n
the (mnp) plane that passes through the origin can a2
be expressed as (see the figure at right)
a1/m
a1
nmnp = G mnp / | G mnp | . (18)
A consequence of this result is that the distance between two consecutive planes with
the same Miller indices (mnp) is inversely proportional to the modulus of G mnp . Since we can
always draw a plane from the (mnp) family through the origin, the distance between two
successive planes is equal to the distance between the origin and the closest plane to origin
from the (mnp) family. This distance is obtained by calculating the projection on the normal to
the (mnp), i.e. on nmnp = G mnp / | G mnp | , of any of the vectors a1 / m , a 2 / n , or a 3 / p . Using
(18) it is found that
a1 a a 2π
d mnp = n ⋅ = n⋅ 2 = n⋅ 3 = . (19)
m n p | G mnp |
So,
2π
d mnp = . (20)
m 2 b12 + n 2 b22 + p 2 b32 + 2mn(b1 ⋅ b2 ) + 2np (b2 ⋅ b3 ) + 2 pm(b3 ⋅ b1 )
As already pointed out in the discussion about Miller indices, the distance between any
two planes in the family (sm,sn,sp), is s times smaller than between any two planes in the
family (mnp). The two families/sets of planes are parallel.
In particular, for the simple, body-centered and face-centered cubic Bravais lattices
with the primitive translation vectors given in the Crystal Structure section of the course, the
distance between two consecutive planes with the same Miller indices is, respectively,
a
sc
d mnp = , (21a)
m2 + n2 + p2
a
bcc
d mnp = , (21b)
( n + p ) 2 + ( p + m) 2 + ( m + n) 2
a
fcc
d mnp = (21c)
( n + p − m) + ( p + m − n ) 2 + ( m + n − p ) 2
2
The first Brillouin zone

We have seen that the point group symmetry is important in the direct lattice; it is also
important in the reciprocal lattice. In order to incorporate the information about the point group
symmetry in the primitive cell of the reciprocal lattice, we construct a primitive unit cell called
the first Brillouin zone following the steps:
1) draw lines to connect a given lattice in the reciprocal lattice point to all nearby lattice points,
2) draw new lines (or planes, in three-dimensional lattices) at the mid point and normal to the
first lines. These lines (planes) are called Bragg planes since (as we will see later) all k vectors
that finish on these surfaces satisfy the Bragg condition.
3) the first Brillouin zone is the area (volume) in reciprocal space that can be reached from the
origin, without crossing any Bragg planes.
An example of the construction of the first Brillouin zone for a two-dimensional
oblique lattice is illustrated in the figure below. The first Brillouin zone is always centered on a
lattice point and incorporates the volume of space which is closest to that lattice point rather
than to any other point. The faces of the first Brillouin zone satisfy the relation k cosθ = G / 2 ,
where G =| G | is the distance to the nearest neighbor in the reciprocal space and θ is the angle
between k and G. This relation can be rewritten as 2k ⋅ G =| G | 2 or, since the equation is
equivalent to the replacement of G with − G , we obtain 2k ⋅ G + G 2 = 0 and finally,

( k + G ) 2 = k 2 . In other words, the the first Brillouin zone is the intersection of spheres with
the same radius centered at nearest neighbor points in the reciprocal lattice.
k
θ
Higher-order Brillouin zones, say the nth Brillouin zone, are then defined as the area
(volume) in reciprocal space that can be reached from the origin by crossing exactly n − 1
Bragg planes. The construction of the first (light blue), second (light brown) and third (dark
blue) Brillouin zones for a two-dimensional lattice is illustrated in the figure below. The Bragg
planes enclosing the nth Brillouin zone correspond to the nth order X-ray diffraction.
Although higher order Brillouin zones are fragmented, the fragments, if translated, look
like the first Brillouin zone. This process is called reduced zone scheme. All Brillouin zones,
irrespective of the order, have the same volume.
For a two-dimensional square lattice the first Brillouin zone is also a square. The
higher-order Brillouin zones for such a square lattice are illustrated in the figure below.
Observation: the analogous primitive cell constructed in the direct instead of the reciprocal
lattice (in the same manner as the first Brillouin zone) is called Wigner-Seitz cell.
In three dimensions, since the reciprocal lattice of the bcc lattice is a fcc lattice, the first
Brillouin zone of the bcc lattice (see the polyhedron in the figure a above) is the Wigner-Seitz
cell of the fcc. The reverse is also true: the first Brillouin zone of a fcc lattice (the truncated
octahedron/rhombododecahedron in figure b above) is the Wigner-Seitz cell of the bcc lattice.
For certain Bravais lattice, in particular bcc, fcc and hexagonal, the points of highest
symmetry in the reciprocal lattice are labeled with certain letters. The center of the Brillouin
zone is in all cases denoted by Γ, while other symmetry points are denoted as follows
sc lattice: M – center of an edge
R – corner point
X – center of a face
bcc lattice: H – corner point joining four edges

(see figure) N – center of a face
P – corner point joining three edges
fcc lattice: K – middle of an edge joining two

(see figure) hexagonal faces
L – center of a hexagonal face
U – middle of an edge joining a hexagonal
and a square face
W – corner point
X – center of a square face
hexagonal lattice: A – center of a hexagonal face

H – corner point
K – middle of an edge joining two rectangular faces
L – middle of an edge joining a hexagonal and a rectangular face
M – center of a rectangular face
Dispersion relations of electrons and phonons for different crystal directions use this labeling
(see the figures below), which indicates the direction and the symmetry of the crystal, since
different labels are used for different symmetries.
X-ray diffraction on crystalline structures
The direct observation of the periodicity of atoms in a crystalline material relies on the X-ray
or particle (electron or neutron) diffraction/scattering on these spatially periodic structures,
since the wavelength of the incident beam is in these cases comparable to the typical
interatomic distance of a few Å. Optical diffraction is not suitable for this purpose since the
wavelength of photons is much too long (about 1 μm) in comparison to the lattice constant (a
few Angstroms). On the contrary, for X-rays the wavelength is determined from the relation
E = hν = hc / λ or λ = hc / E , which equals a few Å if E is of the order of few keV. In fact,
λ(Å) = 12.4/E(keV). X-rays are scattered mostly by the electronic shells of atoms in a solid,
since the nuclei are too heavy to respond. (Similarly, for electrons E = ( h / λ ) 2 / 2m , where λ is
the de Broglie wavelength, and an electron energy E of 6 eV corresponds to λ = h / 2mE of

X-ray diffraction 1
about 5 Å, while for neutron diffraction λ(Å) = 0.28/[E(eV)]1/2, the electron mass m being
replaced by the neutron mass M.)
In a diffraction experiment, both the X-ray or particle source and the detector are placed
in vacuum and sufficiently far away from the sample such that, for monochromatic radiation,
the incident and outgoing X-ray or particle beams can be approximated by plane waves. The X-
rays can be used in either transmission or reflection configurations. The diffraction picture
offers information regarding the symmetry of the crystal along a certain axis. In particular, the
positions of the spots give information about the lattice and the intensity analysis reveal the
composition of the basis.
The X-rays penetrate deeply in the material, so that many layers contribute to the
reflected intensity and the diffracted peak intensities are very sharp (in angular distribution). As
in optics, to obtain sharp intensity peaks the X-rays scattered by all atoms in the crystalline
lattice must interfere, and the problem is to determine the Bravais lattice (including the lattice
constants) and the basis from the interference patterns. Because a crystal structure consists of a
lattice and a basis, the X-ray diffraction is a convolution of diffraction by the lattice points and
diffraction by the basis. Generally, the latter term modulates the diffraction patterned of the
lattice points.
The wave diffracted in a certain direction is a sum of the waves scattered by all atoms.
Higher diffraction intensities are observed along directions of constructive interference, which
are determined by the crystal structure itself.
X-ray diffraction 2
The diffraction of X-rays by crystals is elastic, the X-rays having the same frequency
(and wavelength) before and after the reflection. The path difference between two consecutive
planes separated by d is 2·AB = 2d sin θ . First-order constructive interference occurs if
k k’
G
2d sin θ = λ , (1)
condition known as Bragg’s law.

The Bragg law is a consequence of the periodicity of the crystal structure and holds
only if λ ≤ 2d . This is the reason why the optical radiation is not suitable to detect the
crystalline structure, but only X-rays and electron or neutron beams can perform this task.
Higher order diffraction processes are also possible. The Bragg relation determines,
through the angle θ, the directions of maximum intensity. These directions are identified as
high-intensity points on the detection screen, the position of which reveal the crystal structure.
For example, if the sample has a cubic crystal structure oriented such that the direction [111]
(the diagonal of the cube) is parallel to the incident beam, the symmetry of the points on the
detector screen will reveal a C3 symmetry axis. On the contrary, if the diffraction pattern has a
C6 symmetry axis, the crystal is hexagonal, if it has a C4 symmetry axis it is a tetragonal

crystal, whereas it is cubic if it shows both a C4 and a C3 symmetry axis.
The Bragg formula says nothing about the intensity and width of the X-ray diffraction
peaks, assumes a single atom in every lattice point, and neglects both differences in scattering
from different atoms and the distribution of charge around atoms.
A closer look at the interaction between the X-rays and the crystal of volume V reveals
that the amplitude of the scattered radiation F (which is proportional to the amplitude of the
oscillation of the electric and magnetic fields of the total diffracted ray) is determined by the
local electron concentration n(r ) = ∑ nG exp(iG ⋅ r ) , which is a measure of the strength of the
G
interaction, and has the same periodicity as the crystalline lattice. The diffraction intensity
X-ray diffraction 3
I ∝| F | 2 . For elastic X-ray scattering, the phase of the outgoing beam, with wavevector k ' ,
differs from that of the incoming beam that propagates with a wavevector k through
exp[i ( k − k ' ) ⋅ r ] , so that
F = ∫ n(r ) exp[i (k − k ' ) ⋅ r ]dV = ∫ n(r ) exp(−iΔk ⋅ r )dV = ∑ nG ∫ exp[i (G − Δk ) ⋅ r ]dV (2)
G
where Δk = k '−k is the scattering vector, which expresses the change in wavevector. The
result in the above integral depends on the volume of the crystal. If the crystal has length Li
and N i primitive cells in the i direction (i = 1,2,3) of an orthogonal coordinate system with
x = x1 , y = x2 , z = x3 , the integral along the i direction is given by
Li / 2
⎡ 2π ⎤ sin[π ( si − Δξ i ) N i ]
∫ exp ⎢i ( si − Δξ i ) xi ⎥dxi = ai = Li sinc[π ( si − Δξ i ) N i ] (3)
− Li / 2 ⎣ ai ⎦ π ( si − Δξ i )
where si , Δξ i are the components of G and Δk on the i axis and ai = Li / N i is the lattice
constant on the same direction. The function sinc ( x) = sin x / x has a maximum value for x = 0,
and tends to the Dirac delta function for large x.
Therefore, in large-volume crystals scattering occurs only if
Δk = G , (4)
case in which F = VnG . (In finite-volume crystals there is a sort of “uncertainty” in the angular
range of Δk around G for which the scattering amplitude takes significant values: as the
volume decreases, the angular range increases.) The above condition suggests that X-ray
diffraction experiments reveal the reciprocal lattice of a crystal, in opposition to microscopy,
which exposes the direct lattice (if performed with high-enough resolution).
The diffraction condition Δk = k '−k = G can be rewritten as k ' = k + G or
k ' 2 = k 2 + G 2 + 2k ⋅ G . In particular, the form hk ' = hk + hG of the diffraction condition
represents the momentum conservation law of the X-ray photon in the scattering process; the
crystal receives the momentum − hG . For elastic scattering | k ' |=| k | and thus G 2 + 2k ⋅ G = 0 ,
or k ⋅ G =| G | 2 / 2 , equation that defines the faces of the first Brillouin zone (the Bragg planes).
The geometric interpretation of this relation (see the figure below) is that constructive
X-ray diffraction 4
interference/diffraction is the strongest on the faces of the first Brillouin zone. In other words,
the first Brillouin zone exhibits all the k wavevectors that can be Bragg-reflected by the crystal.
Bragg
plane
The diffraction condition is equivalent to Bragg’s law, which can be written for a
certain set of planes separated by the distance d = d mnp as 2(2π / λ ) sin θ = 2π / d mnp , or
2k ⋅ G = G 2 , with G = mb1 + nb2 + pb3 (for the direction of G with respect to the set of planes,
see the figure illustrating the Bragg law).
The Ewald sphere

The direction of interference peaks can be B
easily determined also via a simple
geometrical construction suggested by
Ewald. Namely, one constructs a sphere
(a circle in two dimensions – see the red
circle in the figure at right) around a point O
O in the reciprocal lattice chosen such

A
that the incident wavevector with O as
origin, ends on an arbitrary lattice point
A. The origin of the Ewald sphere (or
circle) is not necessary a lattice point.
The radius of the sphere (circle) is the wavenumber of the incident (and outgoing)
radiation k =| k |=| k ' | . A maximum intensity is found around a direction k ' if and only if the
Ewald sphere (circle) passes through another point B of the reciprocal lattice. The direction k '
is determined by the origin O of the Ewald sphere and this lattice point on the surface
(circumference), which is separated from the tip of k (from A) by a reciprocal lattice vector. It
X-ray diffraction 5
is possible that for certain incidence angles and wavelengths of the X-rays no such preferential
direction k ' exists.
Therefore, to obtain peaks in the scattered intensity it is necessary to vary either the
wavelength or the incidence angle of the incoming X-rays such that a sufficient number of
reciprocal lattice points find themselves on the Ewald sphere (circle), in order to determine
unambiguously the crystal structure.
In the first method, called Laue
method, the radius of the Ewald sphere
(circle) is varied continuously (see, for
example, the green circle in the figure
below), while in the second method,
called the rotating crystal method or
Debye-Scherrer-Hull method, the Ewald
sphere (circle) is rotated around the
original lattice point with respect to which
the Ewald sphere (circle) was constructed.
The result is represented with the dark
blue circle in the figure below.
In polycrystalline samples, the incident beam is scattered by only those crystallites
(randomly oriented) with planes that satisfy the Bragg condition. Because the sample contains
crystallites with all orientations, the diffraction pattern on the screen is no longer formed from
discrete points, but from concentric circles.
The influence of the basis on the scattered amplitude

If the Laue/diffraction condition Δk = G is satisfied, an explicit account of the basis influence
implies that the assumption of point/spherical sources at the lattice points have to be modified.
In this case, we have found that
F = VnG = N ∫cell n( r ) exp( −iG ⋅ r ) dV = NS G , (6)
where nG = Ω −1 ∫cell n(r ) exp(−iG ⋅ r )dV , N is the total number of lattice points, and
S G = ∫cell n( r ) exp( −iG ⋅ r ) dV (7)

X-ray diffraction 6
is the structure factor. It is defined as an integral over a single cell, with r = 0 at one corner. If
there is only one lattice point in the basis and the electron distribution n(r ) ≅ δ ( r ) , S G = 1 .
If there are s atoms in the basis at positions r j , j = 1,2,..,s, the total electron density can
be expressed as a superposition of electron concentration functions n j at each atom j in the
basis, so that the structure factor is expressed as integrals over the s atoms of a cell:
⎡s ⎤ ⎡s ⎤
S G = ∫ ⎢ ∑ n j ( r − r j )⎥ exp(−iG ⋅ r )dV = ∫ ⎢ ∑ n j ( ρ)⎥ exp(−iG ⋅ ρ) exp(−iG ⋅ r j )dV
⎣ j =1 ⎦ ⎣ j =1 ⎦ (8)
s s
= ∑ exp(−iG ⋅ r j ) ∫ n j ( ρ) exp(−iG ⋅ ρ)dV = ∑ f j exp(−iG ⋅ r j )
j =1 j =1
where ρ = r − r j and f j = ∫ n j ( ρ) exp(−iG ⋅ ρ)dV is the atomic form factor, which depends
only on the type of element that the atom belongs to. The integral has to be taken over the
electron concentration associated with a single atom.
The atomic form factor is a measure of the scattering power of the jth atom in the unit
cell. If the charge distribution has a spherical symmetry, one can use spherical coordinates with
the polar direction along G. In this case,
∞
f j = 4π ∫ n j ( ρ) ρ 2 (sin Gρ / Gρ )dρ . (9)
0
where ρ is the radial coordinate. The atomic form factor decreases rapidly with the distance
and, in the limit ρ → 0 , when sin Gρ / Gρ → 0 ,
∞
f j → 4π ∫ n j ( ρ) ρ 2 dρ = Z , (10)
0
where Z is the number of electrons in an atom. Also, f j = Z when G = Δk = 0 (for a
diffracted ray collinear with the incident ray). f can be viewed as the ratio of the radiation
amplitude scattered by the electron distribution in an atom to that scattered by one electron
localized at the same point as the atom.
Example: consider a bcc lattice as a sc lattice with a basis consisting of two atoms at [[000]]
and [[1/2,1/2,1/2]]. The primitive lattice vectors for the Bravais and the reciprocal lattices are
in this case a1 = ax, a 2 = ay, a 3 = az, and b1 = (2π / a) x, b2 = (2π / a) y, b3 = (2π / a) z ,
respectively. The diffraction peak of the sc lattice that is labeled by (mnp) corresponds to
X-ray diffraction 7
G = mb1 + nb2 + pb3 = (2π / a)(mx + ny + pz ) and for this diffraction peak
2
S mnp = ∑ f j exp(iG ⋅ r j ) = f1 exp[i(2π / a)(mx + ny + pz ) ⋅ 0 ]
j =1 (11)
+ f 2 exp[i (2π / a)(mx + ny + pz ) ⋅ (a / 2)( x + y + z )] = f1 + f 2 exp[iπ (m + n + p)]
The bcc diffraction intensity is I mnp ∝| S mnp | 2 = f 12 + f 22 + 2 Re[ f 1 f 2 exp[iπ ( m + n + p )]] . If
f1 = f 2 = f ,
⎧4 f 2 , if m + n + p = even
I mnp ∝ 2 f 2 [1 + exp[iπ (m + n + p)]] = ⎨ (12)
⎩ 0, if m + n + p = odd
So, for the bcc structure with the same type of

atoms, the (mnp) diffraction peaks of the sc
lattice disappear whenever m + n + p is an odd
π
integer. In particular, it disappears for the (100)
reflection (see the figure at right) since the
2π
phase difference between successive planes is
π, and the reflected amplitudes from adjacent
planes are out-of-phase/destructive interference
occurs.
Observation: for a sc lattice with one atom in the basis, the diffraction intensity is the same,
irrespective of the parity (even or odd) of m + n + p . This example illustrates the effect of the
basis on the diffraction intensity.
Crystal binding
The stability of solid state materials is assured by the existing interactions (attractive and
repulsive) between the atoms in the crystal. The crystal itself is definitely more stable than the
collection of its constituent atoms. This means that there exist attractive interatomic forces
and that the energy of the crystal is lower than the energy of the free atoms. On the other
hand, repulsive forces must exist at small distances in order to prevent the collapse of the
material. One measure of the strength of the interatomic forces is the so-called cohesive
energy of the crystal, defined as the difference between the energy of free atoms and the
crystal energy. Similarly, the cohesive energy per atom U 0 is defined as the ratio between the
cohesive energy of the crystal and the number of atoms. Typical values of the cohesive energy
per atom range from 1 to 10 eV/atom, with the exception of inert gases, where the cohesive
energy is about 0.1 eV/atom. In particular, the cohesive energy determines the melting
temperature of solid state materials. Crystals with | U 0 | < 0.5 eV have weak bindings
between constituents, while the others are characterized by strong crystal bindings.
As shown in the figure above, the potential/binding energy U, which describes the interaction
between two atoms, approaches 0 for an interatomic distance R → ∞ and tends to infinity at
R → 0, having a minimum at a certain distance R = R0 . It is composed of an attractive energy
part, dominant at R > R0 , and a repulsive energy part that prevails at R < R0 . Then, the most
stable state of the system, which occurs at the lowest possible energy, is characterized by the
cohesive energy U 0 , the corresponding interatomic distance, R0 , being known as the
equilibrium interatomic distance. The last parameter has typical values of 2−3 Å, which
implies that the stability of the crystal is determined by short-range forces.
Crystal binding 2
The interatomic force, defined as
F ( R ) = −∂U / ∂R , (1)
is negative (attractive) for R > R0 , and positive (repulsive) for R < R0 . The attractive and
repulsive forces, which have different origins, cancel each other at the equilibrium interatomic
distance.
The general form of the potential energy is
A B
U (r ) = n
− m , with n > m. (2)
r r
The repulsive force between atoms in the solid has the same origin in all crystals: the
Pauli exclusion principle, which forbids two electrons to occupy the same orbital (the same
quantum state). The repulsive force is characterized (see the formula above) by the power-law
expression U = A / r n , with n > 6 or, sometimes, by the exponential expression
U = λ exp( − r / ρ ) , where λ and ρ are empirical constants that can be determined from the
lattice parameters and the compressibility of the material. Which expression is better suited to
describe the repulsive force depends on which one better fits with experimental values. The
repulsive potential is short-ranged and thus it is effective only for nearest neighbors.
The attractive forces create bonds between atoms/molecules in the solid, which
guarantee the crystal stability and are of different types depending on the crystal. Only the
outer (valence) electrons, i.e. the electrons on the outer shells, participate in the bonding;
these electrons are responsible also for the physical properties of the material. In contrast,
core electrons are those on inner shells. There are several types of bonding, depending on the
mechanism responsible for crystal cohesion: ionic, covalent and metallic, which give rise to
strong crystal bindings, and hydrogen bonding and van der Waals interactions, which
determine weak crystal bindings.
Crystal binding in inert/noble gases. Van der Waals-London interaction

The crystals of inert gases have low cohesion energies and melting temperatures, and high
ionization energies. They are the simplest crystals, with an electron distribution close to that
of free atoms. From an electrical point of view they are isolators, and from an optical point of
view, are transparent in the visible domain. The weak binding between the constituent atoms
Crystal binding 3
favors compact crystalline structures, in particular fcc Bravais lattices with one atom in the
basis (the only exceptions are He3 and He4, which crystallize in the hcp crystal structure).
Individual atoms of Ne, Ar, Kr, or Xe have completely occupied external shells, with a
spherically symmetric electronic charge distribution. In crystals, the presence of other atoms
induces a redistribution of the electric charge and a perturbation of the spherical charge
symmetry that can be described within the model of fluctuating dipoles. Coulomb attraction
can occur between two neutral spheres, as long as their internal charges polarize the spheres.
In a classical formalism (valid since electrostatic forces have a long range), this model
assumes that the movement of the electron in atom 1 induces an instantaneous dipole moment
p1 which generates an electric field E (r12 ) at the position of atom 2 separated from atom 1
through a distance r12 =| r12 | .
r12
r2
r1
This electric field induces a fluctuating dipole in atom 2 (the distance between the atoms as
well as the magnitude and direction of p1 fluctuate in time), with a moment p2 = αE (r12 ) ,
where α is the atomic polarizability. The energy of the dipole-dipole interaction between the
two dipoles is minimum when p1 || p 2 || r12 , case in which
2
⎛ 1 ⎞ 4αp12 C
U attr ,min (r12 ) = −⎜⎜ ⎟⎟ =− 6 . (3)
⎝ 4πε 0
6
⎠ r12 r12
This van der Waals (or London) interaction is the dominant attractive interaction in noble
gases. The higher-order contributions of the dipole-quadrupole and quadrupole-quadrupole
interactions are characterized by the respective potentials − C1 / r128 and C 2 / r1210 , and do not
contribute significantly to the cohesion energy of the noble gases crystals. The same − C / r126
Crystal binding 4
dependence of the energy is recovered in a quantum treatment, in the second-order

perturbation theory.
Assuming a power-law expression for the repulsive forces with n = 12, the interaction
potential is given by the Lenard-Jones formula
⎡⎛ σ ⎞
12
⎛σ ⎞ ⎤
6
U (r12 ) = 4γ ⎢⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎟ ⎥ , (4)

⎣⎢⎝ r12 ⎠ ⎝ r12 ⎠ ⎦⎥
where the parameters γ and σ are determined from X-ray and cohesion energy experiments.
The interaction energy of atom 1 (atom i, in general) with all other atoms in the crystal
is then
⎡⎛ σ ⎞
12
⎛σ ⎞ ⎤
6
U i = ∑ U (rij ) = ∑ 4γ ⎢⎜⎜ ⎟ −⎜ ⎟ ⎥
⎟ ⎜r ⎟ ⎥ (5)
j ≠i j ≠i ⎢⎝ rij ⎠ ⎝ ij ⎠ ⎦
⎣
and the energy of the crystal composed of N atoms is U cryst = ( N / 2)U i . For a periodic
arrangement of atoms in the lattice, with nearest-neighbors at a distance R, rij = pij R and
⎡ ⎛ σ ⎞12 ⎛σ ⎞ ⎤
6
U cryst = 2 Nγ ⎢ S12 ⎜ ⎟ − S 6 ⎜ ⎟ ⎥ (6)

⎢⎣ ⎝ R ⎠ ⎝ R ⎠ ⎥⎦
where
6 12
⎛ 1 ⎞ ⎛ 1 ⎞
S 6 = ∑ ⎜⎜ ⎟ ,
⎟ S12 = ∑ ⎜⎜ ⎟
⎟ (7)
j ≠ i p ij j ≠ i p ij
⎝ ⎠ ⎝ ⎠
are rapid convergent series, that can be calculated after the crystalline structure is determined
by X-ray measurements. Their values are, respectively, 12.132 and 14.454 for the fcc
structure, with almost the same values for hcp structures.
The crystal energy is minimum for the R value which is the solution of
∂U cryst / ∂R = 2 Nγ [12 S12 (σ / R )11 − 6 S 6 (σ / R ) 5 ] = 0 , i.e. for
R0 = σ (2 S12 / S 6 )1 / 6 . (8)
Crystal binding 5
The ratio R0 / σ = 1.09 for a fcc Bravais lattice, the corresponding cohesion energy per atom
(at zero temperature and pressure) being
⎡ ⎛ S ⎞2 ⎛ S ⎞⎤ S2
U 0 = U cryst ( R0 ) / N = 2γ ⎢ S12 ⎜⎜ 6 ⎟⎟ − S 6 ⎜⎜ 6 ⎟⎟⎥ = −γ 6 = −8.6γ . (9)
⎢⎣ ⎝ 2S12 ⎠ ⎝ 2S12 ⎠⎥⎦ 2S12
Quantum corrections reduce the binding energy above by 28%, 10%, 6%, and 4% for Ne, Ar,
Kr, and Xe, respectively. The quantum corrections are more important for inert gas crystals
with smaller equilibrium interatomic distance (smaller lattice constants).
Ne Ar Kr Xe
R0 (Å) 3.05 3.74 4 4.34
U0 (eV) -0.024 -0.085 -0.118 -0.171
Tmelt (K) 24 84 117 161
γ (eV) 0.031 0.01 0.014 0.02
σ (Å) 2.74 3.4 3.05 3.98
Ionic binding
The ionic binding is found in ionic crystals formed
from positive and negative ions, for example Na+
and Cl− in NaCl. In this bonding type, electrons are
transferred from the low electronegative atom,
which becomes a positive ion, to the high
electronegative atom, which is transformed into a
negative ion (see the figure on the right).
The electronegativity measures the ability of an atom or molecule to attract electrons
in the context of a chemical bond. Because the electronegativities of Na+ and Cl- differ, the
ionic bonding is realized at a cost denoted by ΔE . In NaCl, the net energy cost of the ionic
bonding (i.e. the difference between the energy of the ions and that of the two atoms) is 1.58
eV per pair of ions, without taking into account the Coulomb energy between the ions. In
other words,
Crystal binding 6
+ −
Na + Cl → 1
Na
42Cl
43 + U 0 − ΔE
crystal
where U 0 is the cohesion energy.

In general, the electronegativity increases with the group number in the periodic
element table, from the first to the seventh group (elements in the eight group have complete
shells). Depending on the difference in electronegativity between two atoms, the bonding
between them is
• Ionic (for large difference). Example: Na-Cl.
• Polar covalent bonding (for moderate difference). Example: H-O.
• Covalent bonding (for small difference). Examples: C-O, O-O
In ionic crystals the bonding is achieved by the long-range electrostatic force and so, a
classical treatment is meaningful. The electronic configuration of the ions is similar to that of
inert/noble gases, i.e. the electronic charge has a spherical symmetry, which is only slightly
perturbed in crystal. The perturbations are localized in the regions in which the ions are
closer. In particular, in NaCl the electronic configurations of the Na+ and Cl− ions are similar
to that of noble gases Ne10 (1s22s22p6) and Ar18 (1s22s22p63s23p6), respectively (see below).
= Ne = Ar
The Coulomb force between one positive Na ion and one negative Cl ion, separated by
a distance R is given by
e2
FCoulomb = − (10)
4πε 0 R 2
with R = 2.81 Å the nearest-neighbor distance in NaCl, so that the respective attractive
potential energy,
Crystal binding 7
e2
U Coul = − , (11)
4πε 0 R
equals −5.12 eV per pair. It follows then that the net energy gain in the ionic bonding, is
5.12 eV − 1.58 eV = 3.54 eV per pair of ions.
The electrostatic energy gain per NaCl molecule in a fcc crystal is obtained by adding
different contributions:
e2
• that of the (opposite type) 6 nearest-neighbors of a certain ion, U 1 = −6 ,
4πε 0 R
e2
• that of the 12 second nearest-neighbors (of the same ion type), U 2 = 12 ,
4πε 0 R 2
e2
• that of the 8 third nearest neighbors of opposite type, U 3 = −8 , and so on.
4πε 0 R 3
The result is
⎛
e2 12 8 ⎞ e2 (±)
U ion = − ⎜⎜ 6 − + − ..... ⎟⎟ = − ∑ (12)
4πε 0 R ⎝ 2 3 ⎠ 4πε 0 R j ≠i pij
The series above converge eventually to U ion = −1.748e 2 /(4πε 0 R) = − Me 2 /(4πε 0 R) ,

where M is the Madelung constant, which takes specific values for each crystal structure. For
other crystal structures: CsCl, zinc blende, and wurtzite, we have, respectively, M = 1.763,
1.638 and 1.641. (If the series is slowly convergent or even divergent, the terms in the sum
are rearranged such that the terms corresponding to each cell cancel each other – the cell
remains neutral in charge.) The total attractive energy in a NaCl crystal with N ion pairs is
given by U attr = 2U ion × N / 2 , where the factor 2 in the numerator accounts for the fact that
there are two types of ions: Na and Cl, and the factor 2 in the denominator is introduced in
order to count every ion pair only once. For NaCl, Uattr = 861 kJ/mol (experiments give 776
kJ/mol). The discrepancy (of about 10%) between the experimental and theoretical values is
explained by the existence of the (non-classical) repulsive forces.
Similarly, if we add up the repulsive potential felt by an atom from all others (the
exponential form is used now), we obtain
U rep = ∑ λ exp(−rij / ρ ) ≅ zλ exp(− R / ρ ) (13)

j ≠i
Crystal binding 8
where we consider that ρ << R and z is the number of nearest neighbors. The interaction
energy of the whole crystal, which consists of N ion pairs/molecules is
⎛ Me 2 ⎞
U cryst ( R) = N ⎜⎜ zλ exp(− R / ρ ) − ⎟, (14)
⎝ 4πε 0 R ⎟⎠
and its minimum value per molecule,
U cryst ,0 e2M ⎛ ρ ⎞
U min = =− ⎜⎜1 − ⎟⎟ , (15)
N 4πε 0 R0 ⎝ R0 ⎠
occurs for the equilibrium interatomic distance R0 found from the condition dU cryst / dR =
− ( zλ / ρ ) exp( − R / ρ ) + Me 2 /( 4πε 0 R 2 ) = 0 . The first term (the Madelung term) in (15),
which expresses the electrostatic contribution of the interactions, is dominant since ρ << R0 .
LiF LiCl NaF NaCl KF KCl RbF RbCl

R0 (Å) 2.01 2.56 2.31 2.82 2.67 3.15 2.82 3.29
U0 (eV) -10.5 -8.45 -9.31 -7.86 -8.23 -7.1 -7.85 -6.84
λ (102 eV) 3.08 5.1 6.24 10.93 13.63 21.35 18.54 33.24
ρ (Å) 0.291 0.33 0.29 0.321 0.298 0.326 0.301 0.323
Covalent bonding
The covalent bonding forms in molecules composed of identical particles, for example
hydrogen. In this case two atoms form a (homopolar) bond by sharing a pair of electrons (one
from each atom, with opposite spins). Most atoms can form more than one covalent bond. For
instance, C has four outer electrons (of 2sp3 type) and thus can form 4 covalent bonds. The
covalent bond is highly directional and different bonds repel each other. Therefore, the
corresponding crystal has generally a low packing ratio. For example, C and Si can have a
diamond structure, with atoms joined to 4 nearest neighbors at tetrahedral angles; this
structure has a packing ratio of only 0.34 compared to 0.74 for close-packed structures. The
electrons in covalent bonds are strongly localized along the bond, so that the crystals are
semiconductors or isolators, with not very good electrical conductivity.
Crystal binding 9
To describe the covalent bonding in hydrogen, we introduce the normalized atomic

orbitals 1s for the j (j = 1,2) electron that can belong to either atom A or B as ψ 1As, B ( j ) , so that
the normalized wavefunction of the total system can be either symmetric (labeled with +) or
antisymmetric (labeled with −)
Ψ+ = [1 / 2(1 + S 2AB ) ][ψ 1As (1)ψ 1Bs (2) + ψ 1Bs (1)ψ 1As (2)] (16a)
Ψ− = [1 / 2(1 − S 2AB ) ][ψ 1As (1)ψ 1Bs (2) − ψ 1Bs (1)ψ 1As (2)] (16b)
where S AB = ∫ (ψ 1As (1)) *ψ 1Bs ( 2) dr is the overlap integral. Note that the symmetric
wavefunction for ionic crystals can be expressed as Ψ+ ∝ [ψ A (1)ψ A (2) + ψ B (1)ψ B (2)] .
E- Ψ+
Ψ-
|Ψ+|2 |Ψ-|2
E+
The symmetric wavefunction (also called singlet) corresponds to two antiparallel spin,
with quantum number S = 0 of the operator S 2 (with S the total spin), while the
antisymmetric wavefunction (also called triplet) corresponds to parallel spins, i.e. S = 1 (with
the spin projection quantum number ms = −1, 0, and 1; there are three antisymmetric
wavefunctions!). The form of the wavefunctions above is determined from the condition that
the total wave function for fermions (including spin) must be antisymmetric upon particle
exchange.
The energy eigenvalues are represented above as a function of the distance between
the atoms. A bound state can exist in the singlet state, with E+ = −3.14 eV if the covalent
bonding forms between H atoms, i.e. the strongest binding occurs if the spins of the two
electrons are antiparallel.
Crystal binding 10
E-
atom A E+ atom B
To characterize the crystalline structure of diamond one must generalize the previous
formula in order to incorporate the p atomic orbitals. Indeed, the last occupied orbitals of
these materials are: C(2s22p2), Si(3s23p2), and Ge(4s24p2). When both s and p-type orbitals are
involved, they hybridize (see figures below). [The s atomic orbitals have quantum numbers n
= 1,2,3… (principal quantum number), l = 0 (orbital quantum number), and m = 0 (magnetic
quantum number; the projection of l). The p orbitals have n = 1,2,3…, l = 1, and m = −1,0,1.]
The s and p atomic orbitals hybridize when the energy difference between them is much
smaller than the binding energy.
In particular, when one s orbital with wavefunction Ψs and one p orbital, say the p x
orbital, with wavefunction Ψ p x , hybridize, the result is two linear sp orbitals (see figure
above) with wavefunctions
Ψ1 = 2 −1 / 2 (Ψs + Ψ p x ) , Ψ2 = 2 −1 / 2 (Ψs − Ψ p x ) . (17)
On the contrary, sp2 hybrid orbitals form between one s orbital and two p orbitals, the
resulting planar structure (see figure above) having orbitals at an angle of 120° between them.
The electrons in the hybrid orbitals are strongly localized and form σ bonds; they do not
participate in electrical conduction. One p orbital remains perpendicular to the plane, where it
forms a π bond with other out-of-plane p orbitals from neighboring atoms; this is the case of
graphite or graphene (bidimensional crystal). The electrons in the π orbitals are delocalized
and participate in electrical conduction. The three hybrid orbitals are given by
Crystal binding 11
Ψ1 = 3 −1 / 2 (Ψs + 2Ψ p x ) , (18)
Ψ2 = 3 −1 / 2 [ Ψs − (1 / 2 )Ψ px + 3 / 2Ψ p y ] , Ψ3 = 3 −1 / 2 [ Ψs − (1 / 2 ) Ψ p x − 3 / 2Ψ p y ] .
Similarly, the electronic configurations that forms from one s orbital and three p
orbitals is called sp3. This electronic configuration is characteristic for diamond. The four
hybrid atomic orbitals that are linear combinations of atomic orbitals form a tetrahedron (see
figure below) and are given by
Ψ1 = (1 / 2)(Ψs + Ψ px + Ψ p y + Ψ pz ) , (19a)
Ψ2 = (1 / 2)(Ψs + Ψ px − Ψ p y − Ψ pz ) , (19b)
Ψ3 = (1 / 2)(Ψs − Ψ px + Ψ p y − Ψ pz ) , (19c)
Ψ4 = (1 / 2)(Ψs − Ψ px − Ψ p y + Ψ pz ) . (19d)
C, Si and Ge form crystals in which the covalent binding is dominant, the van der
Waals contribution to the cohesion energy, also encountered in crystals from a single element,
being negligible. However, in crystals with a basis composed of two atoms A and B, with n
and, respectively, 8 − n valence electrons, the covalent binding is accompanied by an ionic
contribution. The resulting bond is called polar covalent bond. The fraction of the ionic
contribution is of 0.18 in SiC, 0.26 in GaSb, 0.32 in GaAs, and 0.44 in InP. Similarly, in ionic
crystals the covalent binding can also contribute to the cohesion energy, the fraction of the
ionic contribution being only 0.86 in AgCl, 0.94 in NaCl, and 0.96 in RbF. When covalent
Crystal binding 12
bonding forms between different atoms, the hybrid orbitals considered above are modified, as
can be seen from the figures below.
Electronic configuration in the CH4 molecule.
Bonding between the 1s orbitals of the H atom and the 2px and 2py orbitals of O atom
in H2O (a) without, and (b) with hybridization
Covalent crystals are characterized by:
• high melting temperatures (the cohesion energy per atom is about −10 eV)
• hardness (but also friable)
• their conductivity depends strongly on temperature and impurity atoms
• high value of the dielectric constant
• generally transparent in IR, but strongly absorbent in visible and near-IR.
Crystal binding 13
Hydrogen binding of crystals

Because neutral hydrogen has only one electron, it should form a covalent bond with only one
other atom. However, just as oppositely charged ions are attracted to one another and can
form ionic bonds, the partial charges that exist at different atoms in polar covalent bonds can
interact with other partially charged atoms/molecules. Particularly strong polar covalent
bonds are found, for example, when a hydrogen atom bonds to extremely electronegative ions
such as O in water/ice (see the figure below, left), F (see the figure below, right), N or Cl. The
partial charges in the figures below are denoted by δ. The hydrogen bond forms between the
hydrogen atom with a strong partial positive charge and electronegative ions with strong
partial negative charges in neighboring molecules. The binding energy is of the order of 0.1
eV. For example, the cohesion energy per molecule in ice is −0.3 eV.
hydrogen bond
hydrogen bond
The hydrogen bond is weaker than, although similar to, ionic bond since it forms
between partial charges rather than full (complete) charges. In hydrogen bonds the hydrogen
atom is the hydrogen bond donor and the electronegative ion is the hydrogen bond acceptor.
As the polar covalent binding, the hydrogen bond can be viewed as a mixture of ionic and
covalent bonding, the ionic bonding being dominant.
In hydrogen, the proton radius is with five orders of magnitude smaller than the radius
of any other ion, and so it allows the existence of only two nearest neighbors of the proton
(more than two atoms would get in each other’s way), i.e. the hydrogen bond is directional.
Despite being weak, the hydrogen bond is extremely important in living organisms,
which are mainly composed of water, since water as well as proteins and nucleic acids posses
a great capacity to form hydrogen bonds. In particular, the hydrogen binding occurs as intra-
molecular binding between the organic complementary bases thymine and adenine, and
cytosine and guanine in DNA. It can also be encountered between constituents of crystals
such as KH2PO4, KD2PO4 (KDP), Ca(OH)2, or Mg(OH)2.
Crystal binding 14
Metallic bonding
The metallic bonding can be understood as the bonding
between positively charged metallic nuclei/ions and
delocalized conduction electrons, seen as a “sea of free
electrons”. It prevails in elements in which the valence
electrons are not tightly bound with the nucleus (in metals,
for example). However, in the metallic bond we cannot
speak about ions, since there is no particular electron that
is “lost” to another ion.
Unlike other bonding types, the metallic bonding is collective in nature, so that no
single “metallic bond” exists. The metallic bonding can be understood as an extremely
delocalized form of covalent bonding. The delocalization is most pronounced for s and p
electrons, with l = 0 and l = 1, respectively, being much weaker for d and f electrons, which
have quantum numbers l = 2 and l = 3, respectively.
In metals, an atom achieves a more stable configuration by sharing all its valence
electrons with all other atoms in the crystal. However, besides delocalization, metallic
bonding also requires the availability of a far larger number of delocalized energy states than
of delocalized electrons. These states are referred to as electron deficiency; they assure the
kinetic energy for delocalization.
The metallic bonding is encountered, for example, in alkaline metals such as Li, K, Na,
with electronic configurations that resemble those of noble gases with an additional s electron
on the outer shell. Having few electrons on their outer shells, alkali metals have only partially
filled energy levels, and so are electron deficient. In forming the crystal, the wavefunctions of
the outer s electrons overlap with those of their nearest neighbors, and the electrons become
delocalized. Their dynamics resembles that of free electrons. In alkaline metals the lattice is
occupied by the positively charged ions with the noble gas structure (they occupy in fact only
about 20% of the crystal volume), while the valence electrons occupy the remaining volume.
Unlike in covalent crystals, where the electronic charge is distributed in a strongly
nonuniform manner (the bonds are spatially oriented), the electronic density in metallic
crystals is highly uniform. This explains the high elasticity and malleability of these materials.
The total Coulomb potential, which includes electron-electron, electron-ion and ion-
ion interactions, is U Coul = U e−e + U e−i + U i −i < 0 (the first and third terms on the rhs are
Crystal binding 15
positive, the middle one is negative). Therefore, the attractive potential is of electrostatic
nature, being balanced by the repulsive interaction due to the Pauli exclusion principle.
Observation: Not all metals have metallic bonding. For example, many transition metals
show covalent properties (not all electrons participate in covalent bonds, and are good
electrical conductors). In transition metals (Fe, Co, Ni, Cu, Zn, Ag, Au, Mn, etc.) the d
orbitals are only partially occupied and the outermost s orbitals are fully occupied.
Example: 4s2–full, 3d–incomplete.
Crystals with metallic bonding are usually characterized by
• high electrical and thermical conductivity, with weak temperature dependence

• high elasticity
• high optical reflectivity in a large frequency bandwidth
• broad range of melting temperatures: low melting temperatures for alkaline metals (Li,
Na, K, Rb, Cs), intermediate for noble metals (Cu, Ag, Au), and high values for metals
such as Ti, Zr, Mo, W. The corresponding cohesion energies vary between −1 eV and
−5 eV.
Lattice oscillations. Phonons
Let us consider a crystalline material consisting of a large number N ion of heavy positively-
charged ions (composed of the nucleus and the valence electrons on the inner atomic orbitals)
with masses M α and situated at positions Rα , α = 1,.., N ion , surrounded by and in interaction
with N el electrons on the outer atomic orbitals with masses m and at positions denoted by ri ,
i = 1,.., N el . The total Hamiltonian of the system is then
H = Tel + Tion + Vel + Vion + Vel −ion

h2 2 h2 1 1 (1)
= ∑− ∇ i +∑ − ∇ α2 + ∑ U el (ri − r j ) + ∑ U ion ( Rα − Rβ ) + ∑ V (ri − Rα )
i 2m α 2M α 2 i, j 2 α ,β i ,α
i≠ j α ≠β
The terms of the right-hand-side denote, in order, the kinetic energy of the electrons, the
kinetic energy of the ions, the (Coulomb) interaction energy of electron pairs, the interaction
energy of ion pairs, and the interaction energy between electrons and ions.
Since m << M α , the electron velocities are much higher than the ion velocities, so that
the electrons “see” a “frozen” distribution of ions, while the ions can only sense the average
(not instantaneous!) spatial distribution of electrons. In other words, for a given ion
configuration the electrons are in a quasi-equilibrium state that is slowly varying in time due
to ion’s motion, whereas the ions evolve slowly in a potential distribution generated by the
average configuration of the electrons. This adiabatic approximation, known also as the Born-
Oppenheimer approximation, allows a factorization of the total wavefunction of the system,
Ψ ( r , R ) with r = {r1 , r2 ,..., rN el } , R = {R1 , R2 ,..., R N ion } into an electronic part, ψ ( r ; R) , in
which the ion’s positions are considered as parameters, and into an ionic part, φ (R) :
Ψ ( r , R ) = ψ ( r ; R )φ ( R ) . (2)
These electronic and ionic parts satisfy the following equations:
⎛ ⎞
⎜ h2 2 1 ⎟
⎜∑ − ∇ i + ∑ el i j ∑ i
U ( r − r ) + V ( r − Rα ⎟ψ ( r ; R) = E el ( R)ψ ( r ; R )
) (3)
⎜ i 2m 2 i, j i ,α ⎟
⎝ i≠ j ⎠
Lattice oscillations. Phonons 2
⎛ ⎞
⎜ h2 1 ⎟
⎜∑ − ∇ + ∑ ion α − + ⎟φ ( R) = Eφ ( R) ,
2
α U ( R R β ) E el ( R ) (4)
⎜ α 2M α 2 α ,β ⎟
⎝ α ≠β ⎠
where E is the energy of the whole system and Eel is the energy of the (sub)system of
electrons.
Let us assume further that in a crystalline lattice with s atoms in the basis, the ions
move around their equilibrium positions Rα0 , so that | Rα − Rα0 = uα |<< Rα0 . Then, the
interaction energy between pairs of ions can be expanded in a Taylor series around the
equilibrium positions. Taking into account that
⎛ ∂U ion ( Rα − Rβ ) ∂U ion ( Rα − Rβ ) ⎞
∇U ion ( Rα − Rβ ) = ⎜⎜ uα + uβ ⎟=0 (5)
∂Rα ∂Rβ ⎟
⎝ ⎠
since the force that acts upon an ion at equilibrium (which is proportional to this derivative)
vanishes, we find that
∑ U ion ( Rα − Rβ ) = ∑ U ion ( R α0 − R 0β ) + ∑ Aαβμν uαμ uνβ , (6)

α ,β α ,β α , β , μ ,ν
α ≠β α ≠β
where
⎛ ∂ 2U ion ( Ra − Rβ ) ⎞
A μν
αβ = ⎜⎜ μ ν
⎟
⎟ (7)
⎝ ∂ R α ∂R β ⎠ Rα , β = Rα0 , β
and the indices μ = x,y,z (and ν) denote the projections of the position vectors on a Cartesian
coordinate system, the first spatial derivative of U ion vanishes due to the requirement that the
force (which is proportional to this derivative) that acts upon an ion at equilibrium vanishes,
and higher-order terms in the Taylor expansion are neglected. The last approximation is called
harmonic. Because the first term in the Taylor expansion of U ion is constant, it can (together
with Eel ) be included in the reference energy of the system, so that in the harmonic
approximation the lattice dynamics is described by the Hamiltonian
( pαμ ) 2 1
H ion = ∑ + ∑ Aαβμν uαμ uνβ . (8)
α ,μ 2M α 2 α , β , μ ,ν
This Hamiltonian describes in fact a set of coupled harmonic oscillators. The coupling
μν
strength with neighboring ions is characterized by the coefficients Aαβ . A harmonic potential
energy corresponds to forces that are linear in the displacements. In particular, the force that
acts on an ion α from other ions β, given by Fαμ = M α u&&αμ = − ∑ Aαβ
μν ν
uβ , must be
β ,ν
proportional to the relative displacement, uα − u β , i.e. it should be expressed as
Fαμ = − ∑ Aαβ u β = − ∑ Aαβ

μν ν μν ν
(u β − uαν ) , (9)
β ,ν β ,ν
which implies that
∑ Aαβμν μν
= Aαα + ∑ Aαβμν =0 (10)
β β ≠α
Thus, in the harmonic approximation one can view the lattice vibration as an
interaction between connected elastic springs (classical harmonic oscillators), shown in the
figure below. The lattice oscillations are similar to elastic waves that propagate through such
a chain of springs. If an atom is displaced from its equilibrium site by a small amount, it will
tend to return to its equilibrium position due to the force acting on it. This results in lattice
vibrations. Due to interactions between atoms, various atoms move simultaneously, so we
have to consider the motion of the entire lattice.
The physical relevant solutions for the system of harmonic oscillators are of plane-
wave type, i.e. are oscillatory in time, with the same frequency for all ions. These are the
normal oscillations.
Oscillations of an infinite atomic chain with one atom in the basis

To gain a deeper insight into the lattice oscillations, let us consider first a simple one-
dimensional infinite lattice (an “atomic chain”) consisting of identical atoms (more precisely,
ions) with mass M, separated by the lattice constant a, as in the figure below.
n–1 n n+1
For (thermal) vibrations of the crystalline lattice, in which the ions move slightly around their
equilibrium positions Rn0 = na , their actual positions Rn satisfy the relations
| Rn − Rn0 = u n |<< Rn0 , where the displacements can occur either along the chain (for
longitudinal vibrations) or transverse to the chain of atoms.
It should be noted that one-dimensional lattice vibrations are not encountered only in
atomic chains. For example, in a simple cubic crystal with one atom in the primitive cell,
when a wave propagates along the directions of the cube edge, face diagonal, or body
diagonal, entire planes of atoms move in phase with displacements either parallel or
perpendicular to the direction of the wavevector (see the figures above). We can describe the
displacement of the plane n from its equilibrium position with a single coordinate, u n . The
problem becomes in this way one-dimensional. For each wavevector u n there is one solution
with longitudinal polarization and two with transverse polarization. The parameter A is
different for longitudinal and transverse waves.
Because in the harmonic approximation the dynamics of the system is equivalent to

that of coupled harmonic oscillators, the force acting on an ion is linear in the (relative)
displacement. For simplicity, we assume further that only the interaction between nearest
neighbors is significant, case in which the force exerted on th n-th atom in the lattice is
Fn = A(u n+1 − u n ) + A(u n−1 − u n ) (11)
where A denotes the interatomic force (or, equivalently, the elastic constant between nearest-
neighbor ions). Note that equation (11) is compatible with (9) and (10) since, if An ,m = 0 for
m ≠ n, n ± 1 , and An ,n −1 = An , n +1 = − A , An ,n = 2 A , we obtain (10).
Applying Newton’s second law to the motion of the n-th atom with mass M,
M ( d 2 u n / dt 2 ) = Fn , we obtain
d 2un
M = A(u n+1 − u n ) + A(u n−1 − u n ) = − A(2u n − u n+1 − u n−1 ) . (12)
dt 2
A similar equation should be written for each atom in the lattice.

The normal solutions of the equation above have the form
u n (t ) = u 0 exp[i ( kRn0 − ωt )] = u 0 exp[i ( kan − ωt )] , (13)
and represent traveling waves, in which all atoms have the same wavevector k and oscillate
with the same frequency ω and the same amplitude u 0 . Solutions of this form are only
possible because of the translational symmetry of the lattice.
Inserting (13) into the equation of motion (12) we obtain
− Mω 2 = − A[ 2 − exp( −ika ) − exp(ika )] (14)
or
⎛ exp(ika ) + exp( −ika ) ⎞

Mω 2 = 2 A⎜1 − ⎟ = 2 A[1 − cos( ka )] = 4 A sin (ka / 2) .
2
(15)
⎝ 2 ⎠
and the dispersion relation, i.e. the k dependence of the oscillation frequency, represented in
the figure below, is
ω ( k ) = 2 A / M | sin( ka / 2) | . (16)
ω
2(A/M)1/2
k
st
1 Brillouin zone
Because the dispersion relation is periodic: ω ( k ) = ω ( k + 2π / a ) , with the periodicity

given by the reciprocal lattice vector, all distinct frequency values can be found in the k
interval
−π / a ≤ k < π /a , (17)
which corresponds to the first Brillouin zone. The maximum (cut-off) frequency ω max
= 2 A / M is obtained for the minimum wavelength of λmin = 2π /(π / a) = 2a . The existence

of a minimum wavelength can be understood as resulting from the condition that waves with
wavelengths smaller than 2a cannot propagate in the lattice, being reflected at the boundaries
of the first Brillouin zone.
The significance of the periodicity of the dispersion relation is evident from the figure
below: changing k by one reciprocal lattice vector gives exactly the same movement of atoms.
In the long-wavelength limit ka / 2 << 1 , we have sin( ka / 2) ≅ ka / 2 , and
ω ( k ) = A / M ak = v ac k . (18)
The oscillations with a linear dispersion relation in the long-wavelength limit are called
acoustic and are characterized by the acoustic velocity, which in this case is v ac = a A / M .
A similar dispersion relation as that for acoustic oscillations holds for acoustic waves
propagating in a continuum, elastic and isotropic medium.
Moreover, since the oscillation frequency does not depend linearly with k, we can
define separately the phase velocity, i.e. the velocity of the phase of the wavefront, and the
group velocity, i.e. the propagation velocity of the wavepacket and of the wave energy. Their
modulus are given, respectively, by
ω A sin(ka / 2) sin( ka / 2)
v
v ph = = a = v ac , (19)
k M ka / 2 ka / 2
vac vph
dω A
v gr = = a cos(ka / 2) = v ac cos(ka / 2) . (20)
dk M 2vac/π
vgr
In the long-wavelength range, k → 0, v ph = v gr = v ac ,
k
while at the edges of the first Brillouin zone, for 0 π/a
k = ±π / a , v ph = 2v ac / π and v gr = 0 .
Finite lattices
For finite one-dimensional lattices consisting of N
identical atoms, the requirement of symmetry (of
equivalence of physical properties) when the equation of
motion refers to different atoms imposes the cyclic
boundary condition u n = u n+ N . This so-called Born-
Karman condition expresses the independence of the
properties on the surface, i.e. we have a finite solid, with
no surfaces; a finite chain with no end. From the cyclic
boundary condition it follows that exp(ikNa ) = 1 , or
2π
k= m, (21)
Na
with m an integer. There are N allowed m values for k in the first Brillouin zone:
− N /2 ≤ m < N /2, (22)
which correspond to the N degrees of freedom of the system. Because N is usually a large
number, the discrete nature of the wavenumber is disregarded and it is considered as a
continuous variable. Below: example for N = 10.
Density of oscillations in a simple finite one-dimensional lattice

How many oscillations (with different k values) exist in the frequency interval (ω , ω + dω ) ?
This number is referred to as the density of oscillations per unit frequency and is denoted by
dN osc / dω = D(ω ) . Because for the finite one-dimensional lattice the wavenumber varies
only in discrete steps of Δk = 2π / Na , there is only one oscillation possible in this
wavenumber range, so that
dN osc 1 Na
= = (23)
dk Δk 2π
and, taking into account the double degeneracy due to the symmetry of ω (k ) (two k values
correspond to the same ω), we finally obtain
D
dN osc dN dk
D(ω ) = = 2 osc
dω dk dω
N 1 N 1
= = (24)
π A / M | cos(ka / 2) | π A / M 1 − sin (ka / 2)
2
2N/πωmax
N 1 2N 1
= =
π A / M 1 − ω 2 / ω max
2 π ω max
2
−ω2 ω
ωmax
Oscillations of an infinite atomic chain with two atoms in the basis

Let us consider now an infinite one-dimensional lattice (“chain”) with lattice constant a,
consisting of equally spaced ions with different masses, M 1 and M 2 > M 1 , as shown below.
The basis has thus two atoms, at equilibrium positions Rn0,1 = na and Rn0, 2 = (n + 1 / 2)a .
n-1 n n+1 n+2
un-1,1 un-1,2 un,1 un,2 un+1,1 un+1,2 un+2,2
Similarly to the atomic chain with one atom in the basis, we have now two equations
of motion of the general form F = M ( d 2 u / dt 2 ) , one for each type of atom. To distinguish
between the displacements of the two atoms, we denote with u n,1 (t ) = u1 exp[i (kan − ωt )] the
displacement of atoms with mass M 1 (the yellow ones in the figure above) and with
u n, 2 (t ) = u 2 exp[i (ka(n + 1 / 2) − ωt )] that of atoms with mass M 2 (the green ones). So, we
have
d 2 u n,1
M1 = A(u n −1, 2 − u n,1 ) + A(u n, 2 − u n,1 ) = − A(2u n,1 − u n, 2 − u n−1, 2 ) , (25a)
dt 2
d 2 u n, 2
M2 = A(u n,1 − u n, 2 ) + A(u n +1,1 − u n, 2 ) = − A(2u n, 2 − u n,1 − u n +1,1 ) , (25b)
dt 2
or
− M 1ω 2 u1 = −2 Au1 + Au 2 [exp(ika / 2) + exp( −ika / 2)] = −2 Au1 + 2 Au 2 cos( ka / 2) , (26a)
− M 2 ω 2 u 2 = −2 Au 2 + Au1[exp( −ika / 2) + exp(ika / 2)] = −2 Au 2 + 2 Au1 cos( ka / 2) . (26b)
Again, the sum rule (10) is satisfied if An,1;n −1, 2 = An,1;n +1, 2 = An, 2;n −1,1 = An, 2;n +1,1 = − A and
An,1;n,1 = An, 2;n, 2 = 2 A . The system of coupled equations (26) has solution only when its
determinant vanishes, i.e. when
2 A − M 1ω 2 − 2 A cos( ka / 2)
= 0, (27)
− 2 A cos( ka / 2) 2 A − M 2ω 2
or
ω 04γ 2 ⎛ ka ⎞
ω 4 − ω 2ω 02 + sin 2 ⎜ ⎟ = 0 (28)
4 ⎝ 2⎠
M1 + M 2 M 1M 2
with ω 02 = 2 A , γ2 =4 .
M 1M 2 (M 1 + M 2 ) 2
From (28) it follows that there are two solutions (two types of lattice oscillations) for every
value of k; these are called the optical and the acoustic branches. The two solutions are
ω 02 ⎡ ⎛ ka ⎞ ⎤ ω 02 ⎡ 2 ⎛ ka ⎞
⎤
ω (k ) =
2
1 ⎢1 − 1 − γ sin ⎜ ⎟ ⎥ ,
2 2
ω (k ) =
2
2 ⎢1 + 1 − γ sin ⎜ ⎟ ⎥ .
2
(29)
2 ⎢⎣ ⎝ 2 ⎠ ⎦⎥ 2 ⎣⎢ ⎝ 2 ⎠ ⎦⎥
As for the one-dimensional lattice with one atom in the basis, ω1, 2 (k + 2π / a) = ω1, 2 (k ) , so
that all relevant values are found in the first Brillouin zone.
From the solutions (29) one can identify the oscillation branches: the acoustic one,
characterized by a linear dispersion relation in the long-wavelength limit (and a corresponding
vanishing oscillation frequency at the center of the first Brillouin zone), corresponds to the
first solution, for which
⎛ π ⎞ ω0
ω ac (0) = ω1 (0) = 0 , ω ac ⎜ ±⎟= 1 − 1 − γ 2 > ω ac (0) , (30)
⎝ a⎠ 2
while the second solution is associated to the optical branch, for which
⎛ π ⎞ ω0
ω opt (0) = ω 2 (0) = ω 0 , ω opt ⎜ ±⎟= 1 + 1 − γ 2 < ω opt (0) . (31)
⎝ a ⎠ 2
-2π/a 2π/a
Because for the acoustic branch ω ac (0) = 0 for k = 0, from the system of coupled
equations (26) it follows that (u1 / u 2 ) ac = 1 , which implies that the displacement of the two
types of ions is the same/occurs in the same direction and the unit cell moves as a whole; it
oscillates in phase (see the figure above, bottom, right). On the other hand, in the long-
wavelength limit of the optical branch, (u1 / u 2 ) opt = − M 2 / M 1 , i.e. the ions are displaced in
opposite direction and we have out-of-phase oscillations. The oscillations occur such that the
center of mass of each ion pair is fixed, i.e. M 1u1 + M 2 u 2 = 0 (see the figure above, top,
right). This oscillation branch is called optical because, when the unit cell consists of ions
with different type of charges (positive and negative), the oscillations form an instantaneous
dipolar moment that interacts strongly with the electromagnetic radiation. Unlike the acoustic
branch, the oscillation branch has no analog in the dynamic of continuum media.
In the long-wavelength limit, when sin( ka / 2) ≅ ka / 2 , the dispersion relation of the
acoustic branch can be approximated as
ω0 ⎡ γ 2 ⎛ ka ⎞ 2 ⎤ ω 0
ω ac ≅ 1 − ⎢1 − ⎜ ⎟ ⎥≅ γka , (32)
2 ⎣⎢ 2 ⎝ 2 ⎠ ⎦⎥ 4
i.e.
1 A
ω ac (k ) = v ac k , v ac = ω 0 γa = a, (33)
4 2( M 1 + M 2 )
while the dispersion relation for the optical branch becomes
ω0 ⎡ γ2 ⎛ ka ⎞
2
⎤ ⎡ γ2 ⎛ ka ⎞
2
⎤
ω opt ≅ 1 + ⎢1 − ⎜ ⎟ ⎥ ≅ ω 0 ⎢1 − ⎜ ⎟ ⎥ (34)
2 ⎢⎣ 2 ⎝ 2⎠ ⎥⎦ ⎢⎣ 8 ⎝ 2⎠ ⎥⎦
or
ω opt (k ) = ω 0 − β k 2 , β = ω 0 γ 2 a 2 / 32 . (35)
In the long-wavelength limit, the dispersion relation of optical oscillations is parabolic.

Note that, to calculate the density of states for the finite one-dimensional lattice with
two atoms in the basis, we can follow the same treatment as for the atomic chain with one
atom in the basis, taking into account that we must calculate separately the density of states
for the two oscillation types, which have different dispersion relations.
In general, in a three-dimensional crystal with s atoms in the basis there are 3s
solutions/branches for a given k and the dispersion relation of the normal oscillation of the
λ-th branch satisfies the relation ω λ (k ) = ω λ (−k ) since the oscillation frequencies are real
and positive. From these 3s branches, three are acoustic. The acoustic oscillation branches are
characterized by a linear dispersion relation in the limit k → 0 and correspond to the situation
in which all atoms in the lattice have the same displacements (oscillate in phase), so that the
complex structure of the lattice is not manifest. The remaining 3s − 3 oscillation branches
with a parabolic dispersion relation in the long-wavelength limit of the form
ω λ (k ) = ω 0λ − β λ k 2 (36)
are optical branches, characterized by the cut-off oscillation frequencies ω 0 λ ; the parameters
β λ are generally positive. In this case, ions with different signs oscillate in anti-phase, i.e. are
displaced in opposite directions. As for the acoustic oscillations, in three dimensions we have
one longitudinal and two transverse optical oscillation branches for each s value.
In d dimensions, for a crystal with s atoms in the basis there are ds oscillation
branches, from which d are acoustic oscillations and d ( s − 1) are optical oscillations. These
general considerations are in agreement with the two examples considered above since, for
the case of the atomic chain with one atom in the basis we should have ds = 1 (for d = 1 and s
= 1), which corresponds to a single acoustical oscillation branch, whereas for the atomic chain
with two atoms in the basis we have d = 1, s = 2, and therefore expect two oscillation
branches, one acoustical and one optical.
The figure below shows the dispersion relation of a crystal, in which we can identify
only acoustical branches, meaning that there is only one atom in the basis of this crystal.
Density of states in a finite three-dimensional crystal

The density of states/oscillations in a three-dimensional crystal is obtained, as in the one-
dimensional crystal, by imposing the appropriate boundary conditions for uα = unχ , the
index n being related to the unit cell and the index χ labeling the atom in the basis. Because
there is a large number of atoms in a crystal, which interact strongly with their neighbors, the
contribution of the atoms at the surface of the crystal to any physical phenomena is negligible.
Then, we can employ again the Born-Karman cyclic condition for each atom in the basis:
u nμχ = u (μn + N μ ) χ (37)

with u nμχ = u λμ exp(ik ⋅ Rn0χ − iωt ) , where N μ is the number of atoms in the x μ direction, μ =
1,2,3, with x1 = x , x 2 = y , x3 = z . As in the one-dimensional case, the wavevector

component along the μ direction in the first Brillouin zone has a discrete spectrum,
2π
kμ = mμ , (38)
N μ aμ
with
Nμ Nμ
− ≤ mμ < (39)
2 2
where a μ is the lattice constant along x μ , and m μ are integers, i.e. it varies in steps of
Δk μ = 2π / Lμ , with Lμ = N μ a μ the length of the crystal along the x μ direction. The discrete
k values in a two-dimensional lattice are represented by points in the figure below.
ky
kx
allowed k values
It follows then that a state/oscillation occupies a volume in the k space given by
(2π ) 3
Δk = Δk1 Δk 2 Δk 3 = , (40)
V
where V = L1 L2 L3 = NΩ is the volume of the crystal with sN = sN1 N 2 N 3 atoms that form a
lattice with a primitive cell of volume Ω = a1 a 2 a3 and s atoms in the basis.

The density of states/oscillations in the k space is then defined as
dN osc 1 V
= = . (41)
dk Δk ( 2π ) 3
The density of oscillations in the frequency space, defined as
dN osc dN osc dk
D (ω ) = = , (42)
dω dk dω
represents the number of oscillations (with different k values) that exist in the frequency
interval (ω , ω + dω ) . In this formula dk is the volume in k space between the surfaces ω (k )
and ω ( k ) + dω (k ) (see the figure below). The density of oscillations is discrete (as for one-
dimensional crystals) but, for sufficiently large crystals, the sum over the discrete states can
be replaced by an integral. We can calculate it observing that
dk⊥
dSω
dω
dk = ∫ dSω dk ⊥ = ∫ dS ω
| ∇kω |
, (43)
ω ( k ) =const ω ( k ) = const
from which it follows that
V dS ω
D(ω ) = ∫ . (44)
(2π ) ω ( k )=const | ∇ k ω |
3
We can express also the density of states as D (ω ) dω = [V /( 2π ) 3 ]dk , or
V
∫ D(ω )dω = (2π ) 3 ∫ dk , (45)
1st BZ
which represents a particular case of approximating a sum over k in the first Brillouin zone by
an integral, approximation that for an arbitrary function f (k ) is
V V
∑ f (k ) = ∑ f (k )Δk → ∫ f (k )dk . (46)
k (2π ) 3 k (2π ) 3 1st BZ
For f ( k ) = F (ω ( k )) , we have
∑ F (ω (k )) = ∫ F (ω ) D(ω )dω (47)

k
if the density of states is normalized in each branch χ, χ = 1,…,s such that
ω max
∫ D(ω )dω = 3N (48)

0
with ω max the maximum value of the oscillation frequency in the branch.
Note that for the three-dimensional crystals, we have not defined the density of states
normalized at unit volume (as for the atomic chain with one atom in the basis), but have kept
the crystal volume throughout the calculations!
Quantized oscillations/phonons in a one-dimensional finite lattice with one

atom in the basis
We have seen that in the one-dimensional lattice with one atom in the basis, the ions act as
coupled harmonic oscillators. Here we show that this system of coupled oscillators can be
reduced to an equivalent system of independent harmonic oscillators by the introduction of
normal coordinates. Then, we associate a normal oscillation to each normal coordinate. The
Hamiltonian of the finite one-dimensional lattice with one atom in the basis can be written as
N p n2 1 N
H ion = ∑ + ∑ Ann ' u n u n ' . (49)
n =12 M 2 n,n '=1
In the quantum treatment of the system of coupled harmonic oscillators, the position and
momentum coordinates become (conjugate) operators, such that
[uˆ n , pˆ n' ] = uˆ n pˆ n' − pˆ n'uˆ n = ihδ nn' , [uˆ n , uˆ n' ] = 0 , [ pˆ n , pˆ n' ] = 0 . (50)
To proceed further, we must first note that the traveling-wave solutions considered in
the case of an infinite lattice studied above, i.e. u n (t ) = u 0 exp[i(kan − ωt )] , do not lead to real
displacements when finite lattices are accounted for. In this case, real displacements can be
obtained only by superimposing plane waves that travel in opposite directions, reflected at the
boundary of the finite one-dimensional crystal. For example, a solution could be
u n (t ) = exp(−iωt )[u 0 exp(ikan) + u 0* exp(−ikan)] . Keeping this in mind, and replacing the
“amplitude” of the displacement and its complex conjugate with annihilation and creation
operators, we obtain the following form for the position and momentum operators:
1 h
uˆ n = ∑ [aˆ k exp(ikna) + aˆ k+ exp(−ikna)] , (51)
N k 2Mω (k )
− ih 2Mω (k )
pˆ n = ∑ [aˆ k exp(ikna) − aˆ k+ exp(−ikna)] , (52)
2 N k h
where â k and â k+ are annihilation and creation operators of lattice oscillations corresponding
to the wavenumber k, which satisfy the commutation relations
[ aˆ k , aˆ k+' ] = aˆ k aˆ k+' − aˆ k+' aˆ k = δ kk ' , [aˆ k , aˆ k ' ] = 0 , [ aˆ k+ , aˆ k+' ] = 0 . (53)
The corresponding inverse relations are:
1⎛ i ⎞
aˆ k = ∑ ⎜⎜ Mω (k )uˆ n + pˆ n ⎟ exp(−ikna)
⎟
(54)
2hN n ⎝ Mω (k ) ⎠
1 ⎛ i ⎞
aˆ k+ = ∑ ⎜⎜ Mω (k )uˆ n − pˆ n ⎟ exp(ikna)
⎟
(55)
2hN n ⎝ Mω (k ) ⎠
It can be easily shown that (50) are indeed satisfied by these expressions, since
[uˆ n , pˆ n ' ] =
− ih
∑
ω (k ' )
2 N k ,k ' ω ( k )
(
− [aˆ k , aˆ k+' ]e i ( kn−k 'n ') a + [aˆ k+ , aˆ k ' ]e −i ( kn−k 'n ') a )
(56)
=
ih
∑e
2N k
(
ik ( n − n ') a
)
+ e −ik ( n −n ') a = ihδ nn '
The last equality follows from the fact that, for k = 2πm / Na in the first Brillouin zone, with
m an integer the sums in the equation above are given by
N exp(i 2πl ) − 1
∑ exp( ±ikla ) = ∑ exp(i 2πml / N ) = exp(i 2πl / N ) = Nδ l = Nδ nn ' (57)
k m =1 exp(i 2πl / N ) − 1
where l = n − n' is an integer. One can show, similarly, that the other two commutation
relations in (50) are satisfied, as well as the commutations relations in (53), using the
expressions in (54) and (55).
Then, using (57) and the symmetry relation ω ( − k ) = ω (k ) , we obtain for the kinetic
energy of the ensemble of oscillations the relation
N pˆ n2 h ⎛ N
+
N
∑ =− ∑ ω (k )ω (k ' ) ⎜ aˆ k aˆ k ' ∑ exp[i(k + k ' )na] − aˆ k aˆ k ' ∑ exp[i(k − k ' )na]
n =12 M 4 N k ,k ' ⎝ n =1 n =1
N N ⎞
− aˆ k+ aˆ k ' ∑ exp[−i(k − k ' )na] + aˆ k+ aˆ k+' ∑ exp[−i(k + k ' )na] ⎟ (58)
n =1 n =1 ⎠
h + + + +
= ∑ ω (k )(aˆ k aˆ k + aˆ k aˆ k − aˆ k aˆ −k − aˆ k aˆ k )
4k
whereas the potential energy becomes, after similar calculations,
1
∑ Ann 'uˆ n uˆ n' =
2 n ,n '
h
4 NM
∑
n,n ',k ,k '
Ann'
ω (k )ω (k ' )
(aˆ aˆ k k 'e
i ( kn + k 'n ') a
+ aˆ k aˆ k+' e i ( kn−k 'n ') a
+ aˆ k+ aˆ k ' e −i ( kn−k 'n ') a + aˆ k+ )

+ −i ( kn+ k 'n ') a
aˆ e
k'
h ω 2 (k ' ) ⎛ N
i ( k + k ') na
N N (59)
= ∑ ⎜ aˆ k aˆ k ' ∑ e + aˆ k aˆ k+' ∑ e i ( k −k ') na + aˆ k+ aˆ k ' ∑ e −i ( k −k ') na
4 N k ,k ' ω (k )ω (k ' ) ⎝ n =1 n =1 n =1
N ⎞ h
+ aˆ k+ aˆ k+' ∑ e −i ( k + k ') na ⎟ = ∑ ω (k )(aˆ k aˆ k+ + aˆ k+ aˆ k + aˆ k aˆ k + aˆ k+ aˆ k+ )
n =1 ⎠ 4k
In (59) we performed first the sum over n' and used the equality ∑ Ann' exp[ik ' (n'−n)a] =
n'
Mω 2 (k ' ) , and then performed the sum over n using (57).
Observation: The equality ∑ Ann' exp[ik ' (n'−n)a] = Mω 2 (k ' ) is obtained introducing the
n'
solution u n (t ) = u 0 exp[i(kan − ωt )] in the (semi-)classical equation of motion:
Fn = Mu&&n = − Mω 2 u 0 exp(ik ' na) = −∑ Ann'u n' = −∑ Ann'u 0 exp(ik ' n' a)
n' n'
A fully quantum derivation of this equality, more exactly of the relation
Mω 2 uˆ n = −∑ Ann'uˆ n'
n'
is also possible starting from
∂uˆ n i ˆ
= [ H ion , uˆ n ]
∂t h
N p ˆ2 1 N
with Hˆ ion = ∑ n + ∑ Ann 'uˆ n uˆ n ' and employing the commutation relations. Then,
n =12 M 2 n,n '=1
∂uˆ n i 1 pˆ
= ∑ [ pˆ n2' , uˆ n ] = n
∂t h n' 2M M
since
[ pˆ n2' , uˆ n ] = pˆ n2'uˆ n − uˆ n pˆ n2' = pˆ n ' ( pˆ n 'uˆ n ) − (uˆ n pˆ n ' ) pˆ n ' = pˆ n ' (uˆ n pˆ n ' − ihδ nn ' ) − (ihδ nn ' + pˆ n 'uˆ n ) pˆ n '
= −2ihpˆ n 'δ nn '
Analogously,
∂ 2 uˆ n ∂ ⎛ pˆ n ⎞ i⎡ˆ pˆ n ⎤ i ⎡ An' j pˆ ⎤ A
= −ω 2 uˆ n = ⎜ ⎟ = ⎢ H ion , ⎥ = ⎢ ∑ uˆ n'uˆ j , n ⎥ = − ∑ nn' uˆ n'
∂t 2 ∂t ⎝ M ⎠ h⎣ M ⎦ h ⎣n', j 2 M⎦ n ,n ' M
since Ann ' = An 'n due to its definition and
[ ∑ An' j uˆ n'uˆ j , pˆ n ] = ∑ An' j [uˆ n' ( pˆ n uˆ j + ihδ nj ) − (uˆ n ' pˆ n − ihδ n'n )uˆ j ] = 2ih ∑ Ann'uˆ n'
n ', j n ', j n'
Finally, the Hamiltonian of the system of ions becomes
hω ( k ) ⎛ 1⎞ ⎛ 1⎞
Hˆ ion = ∑ Hˆ k = ∑ ( aˆ k aˆ k+ + aˆ k+ aˆ k ) = ∑ hω (k )⎜ aˆ k+ aˆ k + ⎟ = ∑ hω (k )⎜ Nˆ k + ⎟ , (60)
k k 2 k ⎝ 2⎠ k ⎝ 2⎠
with [ Hˆ k , Hˆ k ' ] = 0 . Nˆ k = aˆ k+ aˆ k is the hermitic number operator, for which [ Nˆ k , Nˆ k ' ] = 0 and
[ Nˆ k , Hˆ k ] = 0 .
For each k mode, the eigenstates (Fock states) and eigenvalues of the Hamiltonian and
the number operators are
(aˆ k+ ) n
| nk 〉 = | 0k 〉 , (61a)
nk !
⎛ 1⎞
E k = hω (k )⎜ n k + ⎟ , n k = 0, 1, 2,…… (61b)
⎝ 2⎠
where | 0 k 〉 is the fundamental state of the oscillator, and
aˆ k | nk 〉 = nk | nk − 1〉 , aˆ k+ | nk 〉 = nk + 1 | nk + 1〉 , Nˆ k | n k 〉 = n k | n k 〉 (62)
with aˆ k | 0 k 〉 = 0 . The eigenstate of the lattice Hamiltonian for the N discrete k values is then
| n k1 , nk2 ,..., nk N 〉 =| nk1 〉⊗ | nk2 〉 ⊗ ...⊗ | nk N 〉 (63)
and the energy of the collection of harmonic oscillators is
⎛ 1⎞
E = ∑ E k = ∑ hω (k )⎜ n k + ⎟ = E 0 + E ph . (64)
k k ⎝ 2⎠
where E 0 = ∑k hω (k ) / 2 is the zero (fundamental) energy, and E ph = ∑k hω (k )nk is the

energy of the quantized oscillations of the lattice in an excited state.
The state of each quantum oscillator k is the same as that of n k excitation quanta, each
with an associated energy hω (k ) . This excitation quantum is associated to a quasi-particle
named phonon, in analogy with the photon, which is the quantum of the electromagnetic field.
In the one-dimensional lattice with one atom in the basis considered here, the phonon is called
acoustic phonon since there are only acoustic oscillation branches in this case, and the lattice
has nk1 phonons with wavevector k1 , nk2 phonons with wavevector k 2 , and so on.
Similarly, in a one-dimensional finite lattice with two atoms in the basis and N values
for k,
⎛ 1⎞
Hˆ ion = ∑ hω λ (k )⎜ aˆ k+,λ aˆ k ,λ + ⎟ , (65a)
k ,λ =1, 2 ⎝ 2⎠
(aˆ k+1 ,1 ) (aˆ k+1 , 2 ) ...(aˆ k+N ,1 ) (aˆ k+N , 2 )

nk 1,1 nk 1, 2 nkN ,1 nkN , 2
| nk1 ,λ , nk2 ,λ ,..., nk N ,λ 〉 = | 0〉 (65b)

(nk1 ,λ )!(nk2 ,λ )!...(nk N ,λ )!
with | 0〉 =| 0,0,...,0〉 , and
⎛ 1⎞
E = ∑ hω λ (k )⎜ n k ,λ + ⎟ , (66)
k ,λ =1, 2 ⎝ 2⎠
where the phonons associated with the λ = 1 branch are the acoustic phonons, and those
associated with λ = 2 are the optical phonons.
optical branches
acoustic branches
In a three-dimensional lattice with s atoms in the basis, there are 3 acoustic phonons
and 3s − 3 optical phonons. A typical phonon dispersion spectrum for s = 2 is illustrated in
the figure above.
Phonons are quanta of the collective/thermal lattice oscillations. The crystalline lattice
can be viewed either as a collection of coupled harmonic oscillators or as a gas of free/non-
interacting phonons, which obey the laws of quantum statistics. In particular, phonons are
bosons and obey the Bose-Einstein statistics. However, since they are not real particles, their
number is not independent of temperature and volume, so that the electro-chemical potential
of the phonon gas must be zero. Then, the thermal equilibrium number of phonons with
frequency ω λ (k ) is given by the Planck distribution
1
n ph = nk ,λ = . (67)
exp[hω λ (k ) / k B T ] − 1
So, the number of phonons is small at low temperatures hω λ >> k BT , for which
n ph ≅ exp[−hω λ / k B T ] , but becomes high at large temperatures hω λ << k B T , for which
n ph ≅ 1 /[1 + (hω λ / k B T ) − 1] = k B T / hω λ .
Phononic heat capacity
The thermal properties of solids, and in particular the heat capacity, are determined by both
phonons and electrons. We refer now to the phonon, or lattice, contribution to the heat
capacity. The heat capacity is defined as the heat ΔQ required to raise the temperature by
ΔT , i.e. C = ΔQ / ΔT . If the process is carried out at constant volume, ΔQ must be replaced
by ΔU , which represents the increase in the internal energy U of the system. Then, the heat
capacity at constant volume is
⎛ ∂U ⎞
CV = ⎜ ⎟ . (1)
⎝ ∂T ⎠V
The phonon contribution at the heat capacity is obtained from the lattice energy term
⎛ 1⎞ ⎛ 1 1⎞
E = ∑ hω k ,λ ⎜ nk ,λ + ⎟ = ∑ hω k ,λ ⎜⎜ + ⎟⎟ (2)
k ,λ ⎝ 2 ⎠ k ,λ ⎝ exp(hω k ,λ / k B T ) − 1 2 ⎠
of the internal energy U = E eq + E , with Eeq the energy in the equilibrium configuration of
the system. Actually, expressing the internal energy of the system in terms of the free energy
F and the entropy S as
U = F + TS , (3)
apart from the heat capacity at constant volume CV = (∂U / ∂T )V = T (∂S / ∂T )V we can define
also a heat capacity at constant pressure, C P = T (∂S / ∂T ) P . These two parameters are related
through C P − CV = −T (∂V / ∂T ) 2P /(∂V / ∂P) T , and one can be determined from the other.
These parameters are the same only in the harmonic approximation of lattice oscillations.
In a classical statistical theory, based on the classical partition function, the mean
energy of a one-dimensional oscillator (resulting equally from its kinetic and potential energy
parts) is k B T , value that becomes 3k BT for a three-dimensional oscillator. Then, for sN
three-dimensional oscillators 〈 E 〉 = 3sNk B T , and the phononic heat capacity at constant

volume is
Phononic heat capacity 2
⎛ ∂E ⎞
C ph = ⎜ ⎟ = 3sNk B , (4)
⎝ ∂T ⎠V
i.e. C ph = 3k B per atom, or C ph = 3 N A k B = 3R = 6 cal/molK per mole, with N A the

Avogadro number. This is the Dulong-Petit law, and it predicts a temperature-independent
heat capacity. This prediction agrees with experimental data at high temperatures, but not at
low temperatures, where experiments indicate that C ph ∝ T 3 → 0 as T → 0.
To explain the low-temperature behavior of the heat capacity, one should disregard the
classical statistical theory, which is no longer valid when the separation between the energy
levels of the oscillator is comparable to or higher than k B T , and use instead the quantum
statistical mechanics. The specific heat of the lattice is then defined as
⎛ hω k , λ exp(hω k ,λ / k B T )
2
dE ⎞
C ph = = k B ∑ ⎜⎜ ⎟⎟ . (5)
⎠ [exp(hω k ,λ / k B T ) − 1]
2
dT k ,λ ⎝ k B T
This expression does not involve the zero energy of the lattice and is called for this reason the
phononic heat capacity. To specifically calculate the phononic heat capacity we need to know
the phonon dispersion relation. This relation is quite complicated for three-dimensional
crystals and therefore approximations are generally made.
The Einstein model

In this model each atom or molecule is considered as a particle that oscillates in the average
field of its neighbors. Therefore in the system with 3sN degrees of freedom all particles have
the same oscillation frequency ω E . The phonon heat capacity is in this case given by
C ph (T ) = 3sNk B B2 (Θ E / T ) , (6)
where Θ E is the Einstein temperature, defined by hω E = k B Θ E and
ξ 2 exp(ξ )
B2 (ξ ) = (7)
[exp(ξ ) − 1] 2
is the Einstein function.

ξ 2 exp(ξ )
At high temperatures, for ξ = Θ E / T << 1 , B2 (ξ ) ≅ = exp(ξ ) ≅ 1 , so that
(1 + ξ − 1) 2
C ph (T ) ≅ 3sNk B , (8)
as in the Dulong-Petit law, but at low temperatures, for ξ >> 1 , where B2 (ξ ) ≅ ξ 2 exp(−ξ ) ,
the heat capacity has an exponential temperature dependence of the form
⎛Θ ⎞
2
C ph (T ) ≅ 3sNk B ⎜ E ⎟ exp( −Θ E / T ) . (9)

⎝ T ⎠
Although lim C ph (T ) → 0 , the low-temperature dependence of the heat capacity is not

T →0
proportional to T 3 (see the figure above, right). The discrepancy is due to the inappropriate
treatment of the acoustic phonon contribution to the heat capacity. Unlike for optical phonons,
for which the frequency is almost constant as a function of k, the frequency of acoustic
phonons has a much wider interval of variation and the oscillations in different lattice cells
must be considered as correlated (the atoms oscillate in phase!). Therefore, since the Einstein
model describes in a satisfactory manner the optical phonon contribution to C ph , the heat
capacity is expressed as
C ph (T ) = C ph
opt
(T ) + C ph
ac
(T ) (10)
where
opt
C ph (T ) = 3( s − 1) Nk B B2 (Θ E / T ) (11)
and the contribution of the acoustic phonons is estimated from the Debye model.
The Debye model

In the Debye model the frequency of acoustic phonons in a general, anisotropic crystal is
written as
ω k ,λ = ω λ (k ) = v ac ,λ (θ , ϕ )k , λ = 1,2,3 (12)
with θ, ϕ the polar angles, and their contribution to the heat capacity is given by
⎛ hω k ,λ exp(hω k ,λ / k B T )
2
3 ⎞
C ac
= k B ∑ ∑ ⎜⎜ ⎟⎟
⎠ [exp(hω k ,λ / k B T ) − 1]
ph 2
λ =1 k ⎝ k B T
(13)
⎛ hω k ,λ exp(hω k ,λ / k B T )
2
3 ⎞
= k B ∑ ∫ ⎜⎜ ⎟⎟ D(ω k , λ )dω k , λ
⎠ [exp(hω k ,λ / k B T ) − 1]
2
λ =1 ⎝ k B T
where
V dS ω V k 2 dΩ ω
D (ω k ,λ ) = ∫ | ∇ ω | = (2π ) 3 ∫
(2π ) 3 ω ( k ) = const k ω ( k ) = const | dω k , λ / dk |
(14)
V ω k , λ 2 dΩ ω
= ∫
(2π ) 3 ω ( k ) = const v ac ,λ (θ , ϕ )
3
ac
is the density of states. The frequency integral in the expression of C ph is performed between
0 and ω max,λ (θ , ϕ ) . If
1 3 dΩ 3
∑ ∫ 3 ω = 3 (15)
4π λ =1ω =const v ac ,λ (θ , ϕ ) v ac
is an angular average of the acoustic velocity (the equality holds as identity in the isotropic
crystal), then we can introduce also an angle-independent maximum oscillation frequency (the
Debye frequency) ω max = ω D which, in the Debye model, is also independent on the
polarization λ. This maximum oscillation frequency follows from the normalization condition
of the 3N acoustic oscillation branches:
3 ωD ⎛ 1 3 dΩ ω ⎞ V ω D 2 Vω D3
∑∫ ω ω ⎜ ⎟ ω ω
D ( k ,λ )d k ,λ = ⎜ 4π λ∑=1ω =const
∫ v 3 (θ , ϕ ) ⎟ 2π 2 ∫ d =
2π 2 v ac
3
= 3N , (16)
λ =1 0 ⎝ ac ,λ ⎠ 0
and so
ω D = v ac 3 6π 2 N / V . (17)
In this case
2 ωD 4
3 V k B ⎛ k BT ⎞ ⎛ hω ⎞ exp(hω / k B T )
ac
C ph = ⎜ ⎟ ∫ ⎜⎜ ⎟⎟ dω
2 π 2 v ac
3
⎝ h ⎠ 0 ⎝ k B T ⎠ [exp(hω / k B T ) − 1]
2
(18)
3
⎛ T ⎞
= 9k B N ⎜⎜ ⎟⎟ J 4 (Θ D / T )
⎝ ΘD ⎠
where Θ D defined through hω D = k B Θ D is the Debye temperature, and
ξ
x n exp( x)
J n (ξ ) = ∫ dx (19)
0 [exp( x ) − 1]
2
is the Debye-Grüneisen integral, which has no analytical solution. The Debye temperature is
proportional to the acoustic velocity, and so is higher for high Young modulus values and for
lower crystal densities. It is usually determined by measuring the temperature dependence of
the resistance around the Debye temperature.
At high temperatures, for T >> Θ D , the argument in the J 4 integral is very small,
since x << 1 , and after expanding it in series one obtains
3 3
⎛ T ⎞ ΘD / T
x 4 exp( x) ⎛ T ⎞ ΘD / T
C ac
≅ 9k B N ⎜⎜ ⎟⎟ ∫ (1 + x − 1) 2 dx = 9k B N ⎜⎜ Θ ⎟⎟ ∫ x exp( x)dx
2
⎝ ΘD
ph
⎠ 0 ⎝ D ⎠ 0
3
(20)
⎛ T ⎞ ΘD / T
≅ 9k B N ⎜⎜ ⎟⎟ ∫ x dx = 3k B N
2
⎝ ΘD ⎠ 0
while at low temperatures T << Θ D the upper limit of the integral can be extended to ∞, and
3 3
⎛ T ⎞ ∞ x 4 exp( x) 12π 4 ⎛ T ⎞
C ac
≅ 9k B N ⎜⎜ ⎟⎟ ∫ dx = k B N ⎜⎜ ⎟⎟ ∝ T 3 (21)
⎝ ΘD ⎠ 0 [exp( x) − 1] ⎝ ΘD
ph 2
5 ⎠
This temperature dependence can be understood from a qualitative argument: at low

temperatures only the phonon modes with energy hω < k B T are excited. These modes are, in
the k space, inside a sphere (the thermal sphere), so that the number of modes is proportional
to k 3 ∝ ω 3 ∝ T 3 . If each mode has an average excitation energy of k B T , the total energy of
excitation is proportional to T 4 and hence the heat capacity is proportional to T 3 .

The total phononic heat capacity, C ph = C ph
ac
+ C ph
opt
, is now in agreement with
experiments for both high and low temperatures (see the figure below).
The Debye temperatures for some elemental crystals are given in the table below
Element ΘD (K) Element ΘD (K) Element ΘD (K)

Li 344 Be 1440 Cu 343
Na 158 Mg 400 Ag 225
K 91 Ca 230 C 2230
Rb 56 Sr 177 Si 645
Cs 38 Ba 110 Ge 374
Electron dynamics in the crystalline lattice
We have seen that, in the adiabatic approximation, the equation satisfied by the (sub)system
of valence electrons with mass m and positions ri , i = 1,…, N el in the field of ions at positions
Rα , α = 1,…, N ion is
⎛ ⎞
⎜ h2 2 1 ⎟
⎜∑ − ∇ i + ∑ el i j ∑ i
U ( r − r ) + V ( r − Rα ⎟ψ ( r ; R ) = E el ( R )ψ ( r ; R ) ,
) (1)
⎜ i 2m 2 i, j i ,α ⎟
⎝ i≠ j ⎠
where r = {r1 , r2 ,..., rN el } , R = {R1 , R2 ,..., R N ion } , and the interaction energy between electrons,
U el ( ri − r j ) = e 2 /( 4πε 0 | ri − r j |) , is a bi-particle term of Coulomb type. This equation is quite
difficult to solve and, therefore approximate methods are employed. In one of them, called the
self-consistent field method (or Hartree-Fock method), the interaction energy between
1
electrons ∑ U el (ri − r j ) is replaced by an effective field of the remaining N el − 1 electrons,
2 i, j
i≠ j
which can be expressed as a sum of one-particle terms, ∑ U ief . More exactly, if ψ i (ri , si ) is
i
the one-particle wavefunction for the ith electron, with s i the spin value, the antisymmetric
(with respect to the permutation operator of two electrons) wavefunction of the system of
electrons can be expressed as a Slater determinant,
ψ 1 (r1 , s1 ) ψ 1 (r2 , s 2 ) L ψ 1 (rN el , s N el )

1 ψ 2 (r1 , s1 ) ψ 2 (r2 , s 2 ) L ψ 2 (rNel , s N el )
ψ (r ) = , (2)
N el ! M M
ψ N el (r1 , s1 ) ψ N el (r2 , s 2 ) L ψ N el (rNel , s Nel )
and the energy of the system is given by
1 1
E el = ∑ 〈ψ i | H i | ψ i 〉 + ∑ 〈ψ iψ j | U el (ri − r j ) | ψ iψ j 〉 − ∑ 〈ψ iψ j | U el (ri − r j ) | ψ jψ i 〉 , (3)
i 2 i, j 2 i, j
j ≠i j ≠i
with
h2 2 h2 2
Hi = − ∇ i + ∑ V (ri − Rα ) = − ∇ i + Vi ion (4)
2m α 2 m
Electron dynamics 2
the one-particle Hamiltonian. The second term in the right-hand-side of Eel is the Hartree
term, while the third term is the Hartree-Fock or exchange term.
If the total energy of the electron system could be written as a sum of one-electron
energies:
E el = ∑ Ei 〈ψ i | ψ i 〉 , (5)
i
then the system of interacting electrons would be equivalent to a system of independent

electrons that satisfy the Schrödinger equation of motion
⎛ h2 2 ⎞
⎜⎜ − ∇ i + Vi ion + U ief ⎟⎟ψ i = ( H i + U ief )ψ i = Eiψ i , (6)
⎝ 2 m ⎠
with
1 1
∑ 〈ψ i | U ief | ψ i 〉 = ∑ 〈ψ iψ j | U el (ri − r j ) | ψ iψ j 〉 − ∑ 〈ψ iψ j | U el (ri − r j ) | ψ jψ i 〉 . (7)
i 2 i , j ≠i 2 i , j ≠i
The effective field U ief can be determined if the one-electron wavefunctions ψ i are known,
but the latter can only be found if the effective field is given. It is a self-consistent problem,
which can only be solved iteratively. More precisely, a set of trial functions is first chosen,
such that H iψ i( 0 ) = Eiψ i( 0 ) , with the help of which the effective potential U ief ( 0 ) is calculated.
If this effective potential is introduced in the one-electron equation, one obtains ψ i(1) as a
solution of ( H i + U ief ( 0 ) )ψ i(1) = Eiψ i(1) , and so on. The iterative process stops after, let’s say p
steps, when ψ i( p −1) ≅ ψ i( p ) . In conclusion, the motion of one electron in a crystal can be
described with the one-electron approximation. We use further this approximation and drop
the subscript i for simplicity.
Electrons in a periodic lattice. Bloch functions

We have seen that an electron in a lattice is described by the Schrödinger equation
⎛ h2 2 ⎞
⎜⎜ − ∇ + V (r ) ⎟⎟ψ (r ) = Eψ (r ) , (8)
⎝ 2m ⎠
Electron dynamics 3
where the energy potential V = V ion + U ef describes the interaction with all positive ions and
the remaining electrons, and
V ( r ) = V ( r + Rn ) (9)
for any translation vector Rn of the infinite lattice.

The influence of lattice periodicity on the electron wavefunction can be deduced
observing that
⎛ h2 2 ⎞ ⎛ h2 2 ⎞
⎜⎜ − ∇ r + Rn + V (r + Rn ) ⎟⎟ψ (r + Rn ) = Eψ (r + Rn ) = ⎜⎜ − ∇ + V (r ) ⎟⎟ψ (r + Rn ) , (10)
⎝ 2m ⎠ ⎝ 2m ⎠
since ∇ r + Rn = ∇ r = ∇ . This implies that the electron wavefunction after the translation with
Rn satisfies the same equation as before, and hence
ψ (r + Rn ) = C nψ (r ) , (11)
with | C n |= 1 , because the normalization condition on the volume V imposes that
∫ | ψ (r + Rn ) | dr = | C n | ∫ | ψ (r ) | dr = 1 . Moreover, at two successive translations we have

2 2 2
V V
C n C m = C n + m , so that
C n = exp(ik ⋅ Rn ) , (12)
or ψ (r + Rn ) = exp(ik ⋅ Rn )ψ (r ) , where k is, for now, a wavevector (more precisely, a quasi-

wavevector).
The symmetry properties of the electron wavefunction at translations can be used to
find the general form of the eigenfunctions of the Schrödinger equation for electrons in a
crystalline lattice. More precisely, since ψ (r ) = exp(−ik ⋅ Rn )ψ (r + Rn ) = exp(ik ⋅ r ) ×
{exp[−ik ⋅ (r + Rn )]ψ (r + Rn )} , the wavefunctions can be written as
ψ k (r ) = u k (r ) exp(ik ⋅ r ) , (13)
where u k (r ) = exp[−ik ⋅ (r + Rn )]ψ (r + Rn ) is a periodic function with the same periodicity
as the lattice: u k (r ) = u k (r + Rn ) . Such electron wavefunctions are called Bloch functions.

Electron dynamics 4
Note that for free electrons, which satisfy the Schrödinger equation with vanishing
potential energy, and which have wavefunctions ψ k (r ) = V −1 / 2 exp(ik ⋅ r ) , with V the
normalization volume, all points in space are equivalent (the probability to find the electron at
a point r, | ψ k (r ) | 2 = 1 / V , is independent of r), and the wavefunction is at the same time an
eigenfunction of the momentum p̂ = −ih∇ with eigenvalues p = hk (besides being an
eigenfunction of the Hamiltonian, with eigenvalues E k = h 2 k 2 / 2m = p 2 / 2m ). In contrast,
for Bloch waves | ψ k (r ) | 2 =| u k (r ) | 2 ≠ const., and only points that differ through a translation
vector Rn are equivalent. Therefore hk is the quasi-momentum of the electron in the crystal
and k is the quasi-wavevector; in the following, we will still refer to k as wavevector, for
simplicity.
The influence of lattice periodicity on the electron energy can be inferred observing
first that the electron energy in a crystal is k-dependent and real (the Hamiltonian is a hermitic
operator): E = E (k ) . The Schrödinger equation can be expressed also as
⎛ h2 ⎞
⎜⎜ − (∇ + ik ) 2 + V (r ) ⎟⎟u k (r ) = E (k )u k (r ) (14)
⎝ 2m ⎠
since
∇ψ k (r ) = ∇[u k (r ) exp(ik ⋅ r )] = exp(ik ⋅ r )[∇u k + iku k ] = exp(ik ⋅ r )(∇ + ik )u k (15)
and hence ∇ 2ψ k (r ) = exp(ik ⋅ r )(∇ + ik ) 2 u k . Because the energy is real, from the (identical)
Schrödinger equations satisfied by the wavefunction for − k and its complex conjugate:
⎛ h2 ⎞
⎜⎜ − (∇ − ik ) 2 + V (r ) ⎟⎟u −k (r ) = E (−k )u − k (r ) , (16)
⎝ 2m ⎠
⎛ h 2
⎞
⎜⎜ − (∇ − ik ) 2 + V (r ) ⎟⎟u k* (r ) = E (k )u k* (r ) , (17)
⎝ 2m ⎠
it follows that the energy is an even function of k:
E (k ) = E (− k ) . (18)
In particular, the iso-energetic surfaces of electrons in a lattice, i.e. the E (k ) = const. surfaces
in k space, have an inversion symmetry; they are not, however, spheres as for free electrons.
Electron dynamics 5
If in the crystalline lattice we perform a translation of k with a vector in the reciprocal

space, G, we obtain
E (k ) = E (k + G ) , (19)
because, by replacing k with k + G in (11) we obtain ψ (r + Rn ) = exp(ik ⋅ Rn )ψ (r )

= exp[i(k + G ) ⋅ Rn ]ψ (r ) since exp(iG ⋅ Rn ) = 1 (the Schrödinger equation and the wave-
function are identical after a translation in the k space). Therefore, the energy is a periodic
function in the reciprocal lattice space (in k space), and takes distinct values only inside the
first Brillouin zone (remember that the whole k space can be divided in Brillouin zones with
all possible orders n, which have the same volume and can be reduced to the first Brillouin
zone by applying symmetry operations). In the Schrödinger equation for the electron k is a
parameter. As for phonons, where we have different solutions ω (k ) for a given k, associated
to different oscillation branches, we can have also different eigenfunctions and eigenvalues of
the electronic Hamiltonian (different energy values E n (k ) ) for a given wavevector k.
Electron velocity in a crystal

We have seen that the electrons in a crystalline lattice are Bloch waves. The so-called Bloch
electrons are quasi-particles since their properties are dependent on the crystalline lattice. If
an electron with wavevector k is inside a lattice, one can introduce the concept of average
velocity through the quantum mechanical definition
1 1 h
〈v 〉 = v = 〈ψ k | pˆ | ψ k 〉 = ∫ψ k* (r ) ∇ψ k (r )dr , (20)
m mV i
with pˆ = (h / i)∇ the momentum operator. For Bloch electrons the average velocity can be
expressed as
1 * h hk h
v= ∫
mV
u k (r ) (∇ + ik )u k (r )dr =
i
+ ∫
m im V
u *k (r )∇u k (r )dr , (21)
which differs from the expression for free electrons, v = hk / m = p / m , through the second
term.
To calculate the average electron velocity, we differentiate (14) with respect to k (we
apply the gradient in k space), and obtain
Electron dynamics 6
⎛ h2 ⎞
⎜⎜ i (∇ + ik ) + ∇ k E ⎟⎟u k (r ) = 0 . (22)
⎝ m ⎠
if the potential energy does not depend explicitly on k. From this result and the normalization
condition ∫ u *k (r )u k (r )dr = 1 we can express (21) as
V
h 1
∫
im V
u *k (r )(∇ + ik )u k (r )dr = ∇ k E ,
h
(23)
so that, finally,
v = h −1∇ k E . (24)
This relation is identical to that obtained for free electrons with a parabolic dispersion
relation, in which E = h 2 k 2 / 2m and, thus, ∇ k E = h 2 k / m and v = hk / m = h −1∇ k E .
Therefore, (24) is a more general expression of average velocity than (21), the difference
being that in (24) the mass of electrons does not appear explicitly; in a crystalline lattice the
free electron mass m can be replaced by an effective mass, so that (24) holds, but not (21).
Electron acceleration in a crystalline lattice. Effective mass

Let us now consider that the electrons in the crystal are accelerated by an external force F.
We separate the effects of the external forces on the electron and the Coulomb interactions
inside the crystal, by modeling the Bloch electron as a quasi-particle that reacts only to the
external forces. The influence of the Coulomb interactions on the electron dynamics is
characterized by an effective electron mass.
Because, according to the Ehrenfest’s theorem, the classical equations of motion are
valid for the average values of the quantum observables, we have
dE (k ) dk d ( hk ) dp
= ∇k E =v⋅ =v⋅ = v⋅F , (25)
dt dt dt dt
i.e. the quasi-momentum p = hk of an electron in the crystal plays the same role as the
momentum of a free electron. Here F is only the external force; the internal forces in the
Electron dynamics 7
crystal are taken into account when calculating the dispersion relation E (k ) , and thus are
included in the expression of the average velocity.
The acceleration of the electron in the crystal can be defined as
dv 1 ⎛ dE ⎞ 1 ⎛ dk ⎞ 1
a= = ∇k ⎜ ⎟ = ∇k ⎜∇k E ⎟= ∇ k (∇ k E ⋅ F ) (26)
dt h ⎝ dt ⎠ h ⎝ dt ⎠ h 2
or, written on Cartesian components μ,ν = x,y,z,
dv μ 1 ∂2E ⎛ 1 ⎞
aμ = =∑ 2 Fν = ∑ ⎜⎜ ⎟ Fν .
⎟ (27)
dt ν h ∂k μ ∂kν ν ⎝ meff ⎠ μν
From the relation above it follows that, in general, the acceleration of an electron in the
crystal has a different direction than that of the applied force. The relation between the
acceleration and the applied force can be used to introduce, as for the case of free electrons, a
second-order tensor parameter, called effective mass. Its inverse, defined as
⎛ 1 ⎞ 1 ∂2E
⎜ ⎟ = 2 , (28)
⎜m ⎟
⎝ eff ⎠ μν h ∂k μ ∂kν
is a symmetric tensor, since (1 / meff ) μν = (1 / meff )νμ . The inverse of the effective mass can
take both positive and negative values, unlike for a free electron, where it is a scalar
parameter and always positive. The effective mass incorporates the effect of the crystal lattice
on the motion of electrons; it is not a characteristic of the electron (as is its free mass) but of
the electron-lattice interaction. The effective mass is inversely proportional to the curvature of
the dispersion relation and is infinite at inflexion points, where the curvature vanishes.
If k 0 is an extremum (maximum or minimum) of the energy dispersion relation in the
k space, for which (∇ k E ) k =k0 = 0 , a Taylor series expansion of E around k 0 in the harmonic
approximation (in which only terms up to the second order are considered) can be written as
1 ⎛ ∂2E ⎞ h2 ⎛ 1 ⎞
E (k ) = E (k 0 ) + ∑ ⎜⎜ ⎟ (k μ − k μ0 )(kν − kν0 ) = E 0 +
⎟ ∑ ⎜⎜ ⎟ (k μ − k μ0 )(kν − kν0 )
⎟
2 μ ,ν ⎝ ∂k μ ∂kν ⎠ k = k0 2 μ ,ν meff
⎝ ⎠ μν
(29)
Electron dynamics 8
We have denoted E (k 0 ) = E0 . The above form is positively defined if the effective mass
takes only positive values, and, reducing it to the principal axes we obtain an ellipsoidal iso-
energetic surface that satisfies the equation
h2 ⎛ k12 k 22 k 32 ⎞⎟
E (k ) = E 0 + ⎜
⎜ meff ,1 + meff , 2 + meff ,3 ⎟
. (30)
2 ⎝ ⎠
If two of the semiaxes of the ellipsoid are equal, for example if meff ,1 = meff , 2 the common
value of the effective mass is called transverse effective mass mt , and meff ,3 = ml is the
longitudinal effective mass; the ellipsoid is in this case an ellipsoid of rotation. If all diagonal
components of the effective mass tensor are equal, the iso-energetic surfaces become spheres,
meff ,1 = meff , 2 = meff ,3 = meff , and the dispersion relation of the electron in the crystal is similar
to that of the free electron with mass m, except that m is replaced with the effective mass:
h2k 2
E (k ) = , (31)
2meff
if the reference energy is chosen such that E0 = 0 . The acceleration has the same direction as
the external force only along the principal axes of the ellipsoid, or for spherical iso-energetic
surfaces.
Energy bands
We have seen that the electron energy can take multiple values for a given k in the first
Brillouin zone. The energy spectrum for the electrons in a crystal can be quantitatively
determined starting from two models: the approximation of quasi-free electrons (the weak-
binding approximation) or the approximation of quasi-bound electrons (the tight-binding
approximation). These approximations correspond to two extreme cases. In the first one it is
assumed that the state of the electron in the crystal can be modeled as a perturbed state of a
free electron, while in the second approximation, the state of the electron is a perturbed state
of a bound electron in an isolated atom, the perturbation in both cases being due to the
periodic lattice potential. The weak-binding approximation is particularly suitable for treating
the energetic spectrum of valence electrons in metals, while the tight-binding approximation
is more suitable for semiconductor and isolating materials. In both cases the periodicity of the
crystalline lattice leads to the formation of allowed and forbidden energy bands. In the figure
Electron dynamics 9
below it is shown how energy bands, separated by gaps, form from atomic s and p orbitals as
the separation between atoms decreases in Si; a is the lattice constant.
The tight-binding approximation

Let us assume that the wavefunction of the Bloch electron in the crystal, ψ k (r ) , can be
expressed as a linear combination of atomic wavefunctions:
ψ k ( r ) = ∑ C n ϕ n ( r − Rn ) (32)
n
where ϕ n (r − Rn ) is the wavefunction of the atom at position Rn and C n = exp(ik ⋅ Rn ) since

the electron wavefunction must be invariant (up to a phase factor) at translation. The atomic
wavefunctions satisfy the Schrödinger equation for the isolated atom,
⎛ h2 2 ⎞
H aϕ n = ⎜⎜ − ∇ + U a (r − Rn ) ⎟⎟ϕ n = E aϕ n , (33)
⎝ 2m ⎠
Electron dynamics 10
where H a is the Hamiltonian and U a is the potential energy of an isolated atom and Ea is
the corresponding energy, whereas the Schrödinger equation for the electron in the crystal is
⎛ h2 2 ⎞
Hψ k (r ) = ⎜⎜ − ∇ + V (r ) ⎟⎟ψ k (r ) = E kψ k (r ) = [ H a + W (r )]ψ k (r ) , (34)
⎝ 2m ⎠
with W (r ) = V (r ) − U a (r − Rn ) < 0 the perturbation energy due to the crystal. The fact that
V (r ) < U a (r − Rn ) , and hence W ( r ) < 0 , expresses the stability of the crystal, the potential
energy in the lattice being lower than in an isolated atom. The electron energies in the crystal
and in the isolated atom are illustrated in the figure below.
Ua(r)
V(r)
The last equation can be re-written as
∑ exp(ik ⋅ Rn )[ H a + W (r ) − E k ]ϕ n = 0 , (35a)
n
or
∑ exp(ik ⋅ Rn )[W (r )ϕ n + ( E a − E k )ϕ n ] = 0 , (35b)

n
which becomes, after left-side multiplication with exp(−ik ⋅ Rm )ϕ m* (r − Rm ) and integration

over the crystal volume V,
∑ exp[ik ⋅ ( Rn − Rm )][ A( Rn − Rm ) + ( E a − E k ) S ( Rn − Rm )] = 0 . (36)

n
Here
A( Rn − Rm ) = ∫ ϕ m* (r − Rm )W (r )ϕ n (r − Rn )dr (37)
V
is the exchange integral, which depends on the overlap of the atomic wavefunctions and the
crystal perturbation and defines the exchange interaction energy, and
S ( Rn − Rm ) = ∫ ϕ m* (r − Rm )ϕ n (r − Rn )dr (38)
V
is the overlap integral, which depends only on the overlapping degree of the atomic
wavefunctions. If Rn − Rm = Rl is the translation vector between the nth and mth atoms, it
follows finally that
Ek = Ea +
∑l A( Rl ) exp(ik ⋅ Rl ) . (39)
∑l S ( Rl ) exp(ik ⋅ Rl )
Thus, the energy of the electron in the crystal differs from the energy in an isolated atom
through a periodic function of k. Instead of a discrete energy level (as in atoms) we have now
an energy band with a width determined by the maximum and minimum values of the
perturbation term.
In the simplifying assumption that the atomic wavefunctions are rapidly decreasing
with the distance, such that the overlap integral is negligible even for neighboring atoms, i.e.
such that
⎧1, Rl = 0 (n = m)
S ( Rl ) = ⎨ , and hence ∑ S ( Rl ) exp(ik ⋅ Rl ) = 1 , (40)
⎩0, Rl ≠ 0 (n = m) l
we obtain that
E k = E a − C − A ∑ exp(ik ⋅ Rl ) , (41)
l = nn
where the sum extends now only over the nearest neighbors, for which A( Rl ) = − A < 0 for s
atomic orbitals, for which A > 0, and
− C = A(0) = ∫ ϕ n* (r − Rn )W (r )ϕ n (r − Rn )dr < 0 (42)

V
is a constant (independent of k) parameter. For a simple cubic lattice with period a, in which
an atom at the origin of the coordinate system has 6 nearest neighbors at positions [[100]],
[[ 1 00]] , [[010]], [[0 1 0]] , [[001]], and [[00 1 ]] ,
∑ exp(ik ⋅ Rl ) = exp(ik1a) + exp(−ik1a ) + exp(ik 2 a) + exp(−ik 2 a)

l = nn (43)
+ exp(ik 3 a) + exp(−ik 3 a) = 2[cos(k1a) + cos(k 2 a) + cos(k 3 a)]
E k = E a − C − 2 A[cos(k1a) + cos(k 2 a) + cos(k 3 a)] , (44)
which is an even function of the wavevector components; all possible energy values are
obtained for wavevector components in the first Brillouin zone, i.e. for − π / a ≤ k i ≤ π / a .
The expression above indicates that, in the crystal, the energy level of the isolated atom shifts
downwards with C due to the interaction between atoms, which renders the crystal more
stable than an isolated atom, and transforms into an energy band, which is periodic in the
wavevector components and extends between E min and E max , with
E min = E a − C − 6 A, for ki = 0 , (45a)

E max = E a − C + 6 A, for k i = ±π / a . (45b)
The widths of the energy bands, ΔE = 12 A , increase with the exchange integral (with the
overlap of the atomic orbitals) and are wider for the higher discrete atomic energy levels since
the wavefunctions of the outer atomic levels are more extended in space. Different energy
bands form starting from different atomic orbitals, and an additional subscript i label these
energy bands. In the figure below (left) different energy bands are represented in the first
Brillouin zone. These bands are formed (in the increasing energy order) from atomic orbitals
for which A is positive, negative, positive, etc.
These (allowed) energy bands are separated by energy gaps (forbidden energy bands)
with widths E gi . The width of the first energy band, for example, is given by
E g1 = E min, 2 − E max,1 = ( E a , 2 − E a ,1 ) − (C 2 − C1 ) − 6( A1 + A2 ) . (46)

If both A1 and A2 correspond to s states, the energy gap is determined from the extremities of
the two bands at different k values (center and edges of the first Brillouin zone), whereas the
energy gap between bands that form from s and p atomic orbitals is determined by states with
the same k: at the edges of the first Brillouin zone, for A2 = A p < 0 (the exchange integral is
positive) and A1 = As > 0 , and at the center of the first Brillouin zone if A2 > 0 and A1 < 0 .
meff
Effective mass in electronic energy bands

Let us calculate the effective electron mass at the extreme points of the energy band with the
dispersion relation E k = E a − C − 2 A[cos(k1a) + cos(k 2 a) + cos(k 3 a)] . Near the center of the
first Brillouin zone, when k i a << 1 and cos(k i a) ≅ 1 − (k i a) 2 / 2 ,
E k = E a − C − 6 A + Aa 2 (k12 + k 22 + k 32 ) = E min + Aa 2 k 2 , (47)
and hence the energy depends quadratically on the wavevector, E k = E min + h 2 k 2 / 2meff , as
for free electrons, with an effective mass
meff = h 2 (∇ k2 E k ) −1 = h 2 / 2 Aa 2 . (48)
The effective mass is positive for energy bands that form from s atomic orbitals, for which
A > 0. For the particular case considered here, that of a simple cubic lattice, the iso-energetic
surfaces Ek = const. in the neighborhood of the center of the first Brillouin zone are spheres.
Note that meff depends on the dispersion relation, and hence on the crystal structure.
On the contrary, at the edges of the first Brillouin zone, introducing the new variables
k i ' = ±(π / a − k i ) , such that cos(k i a) = cos(m k i ' a + π ) = − cos(k i ' a) , the dispersion relation
can be expressed as
E k ' = E a − C + 2 A[cos(k1 ' a) + cos(k 2 ' a) + cos(k 3 ' a)] (49)
and, for k i ' a << 1 and cos(k i ' a ) ≅ 1 − (k i ' a ) 2 / 2 ,
E k ' = E a − C + 6 A − Aa 2 (k1 ' 2 + k 2 ' 2 + k 3 ' 2 ) = E max − Aa 2 k ' 2 . (50)
The dispersion relation is again similar to that of a free electron, i.e. has the form
E k ' = E max + h 2 k ' 2 / 2meff , with an effective mass
meff = h 2 (∇ k2 E k ) −1 = −h 2 / 2 Aa 2 , (51)
which is negative if A > 0. An example of such a situation is represented in the figure above
(right). Note that the effective mass is negative at the center of the first Brillouin zone and
positive at its edges for energy bands that form from p atomic orbitals, for which A < 0. In
general, meff > 0 in the neighborhood of the minimum energy value in the band and meff < 0
near the maximum energy value in the band.

Although (51) shows that the iso-energetic surfaces are also spherical at the edges of
the first Brillouin zone, they have complicated forms at intermediate energy values, between
the center and the edges of the first Brillouin zone. Due to the periodicity of electron energy
in the k space, the iso-energetic surfaces are the same in all cells in the reciprocal space, so
that the iso-energetic surfaces are multiple connected. The same general results are obtained
in the weak-binding approximation.
Electrons and holes

If the electron in the crystal, with an electric charge − e and a positive (isotropic) effective
mass meff > 0 is placed in an electric field E, its acceleration
a = dv / dt = F / meff = −eE / meff (52)
is similar in form to that of a free electron. On the other hand, in the neighborhood of the
maximum energy value in the band meff < 0 and the electron equation of motion is given by
a = dv / dt = F /( − | meff |) = −eE /( − | meff |) . (53)
The negative effective mass in the equation above has no analog for free particles. It has, as
consequence, that an electric field will decelerate the electron, instead of accelerating it. To
avoid such an awkward interpretation, it is considered that the motion of the electron with a
negative effective mass in the electric field E is equivalent with the motion of a quasi-particle
with electric charge + e and a positive effective mass, equal to | meff | . This quasi-particle is
called hole. The hole is a quasi-particle since, unlike the electron, it has no physical meaning
in free space; holes exist only in crystals and reflect the behavior of valence electrons. In a
generalized sense, even the electron in the crystal can be seen as a quasi-particle endowed
with an effective mass, which is different from that of the free electron.
If the electrons in the upper part of an occupied energy band (the valence band)
acquire sufficient energy from thermal vibrations, for example, to go into an unoccupied state
in the next energy band, called conduction band, the remaining empty states can be
considered as holes. Since the empty states can be occupied by other electrons, the holes can
be seen as moving throughout the valence band, i.e. they can be regarded as free quasi-
particles in the valence band. In other words, an energy band occupied with electrons with the
exception of its upper part, can be seen as partially occupied with holes. Because the electrons
that participate at electrical and thermal conduction are those able to move (quasi-)free in the
crystal, the concept of holes allows a major simplification in the treatment of the system of
electrons in the valence band: it is no longer necessary to deal with the motion of the entire
system of electrons, but only with the motion of a much smaller number of holes.
In general, the holes in the valence band have not only an opposite electric charge, but
also a different effective mass than the electrons in the conduction band, since their dispersion
relation is different (the valence and conduction bands originate from different isolated atomic
levels). The holes move in the direction of the applied electric field, whereas electrons move
in the opposite direction!.
Density of electron states

Real crystals have finite sizes and, therefore, the solution of the Schrödinger equation for the
electrons in the crystal depends on the boundary conditions. As for phonons, we assume that
in large crystals the surface phenomena do not influence significantly the electron dynamics
inside the crystal, and hence use the cyclic (Born-Karman) boundary conditions. In the
orthorhombic symmetry, for example, if the dimensions of the crystal along the orthogonal
Cartesian coordinates xi , i = 1,2,3 ( x1 = x , x 2 = y , x3 = z ) are denoted by Li , the cyclic
conditions impose that
ψ k ( x1 , x 2 , x3 ) = ψ k ( x1 + L1 , x 2 , x3 ) = ψ k ( x1 , x 2 + L2 , x3 ) = ψ k ( x1 , x 2 , x3 + L3 )
(54)
= ψ k ( x1 + L1 , x 2 + L2 , x3 + L3 )
or, since ψ k (r ) = u k (r ) exp(ik ⋅ r ) with u k (r ) the same in all lattice cells,
k i Li = 2πmi , k i = 2πmi / Li (55)
with mi integer numbers. Similarly, in a finite crystal with N i atoms along the i direction,
such that Li = N i ai , with ai the respective lattice constants, it follows that
2πmi
ki = , (56)
N i ai
where, for wavevectors in the first Brillouin zone, the integers mi can take only N i values in
the intervals
− N i / 2 ≤ mi < N i / 2 . (57)
Thus, for a simple lattice (with one atom in the basis), the number of distinct energy states of
electrons in an allowed energy band is N = N1 N 2 N 3 , and these states can be occupied by 2 N
electrons because, according to the Pauli principle for fermions, only two electrons with
opposite spins can occupy an energy state characterized by a given k. In large crystals the
distance between energy levels is quite small and the energy band is approximated as a
(quasi-)continuous function of k, case in which any sum over states in the k space can be
replaced (as in the case of phonons) by an integral over the first Brillouin zone. More
precisely, for any function F ( k ) ,
V V
∑ F (k ) = 3 ∑
F (k )Δk = ∫ F ( k ) dk (58)
k (2π ) k (2π ) 3 1st BZ
where
(2π ) 3 (2π ) 3
Δk = Δk1 Δk 2 Δk 3 = = (59)
L1 L2 L3 V
is the volume in k space occupied by a distinct electron state.

In particular, the density of states in the k space, defined as the number of electron
states with a given spin orientation, dN el , in the volume element dk , is given by
dN el 1 V
= = , (60)
dk Δk ( 2π ) 3
or
V
N el = ∫ dk .
(2π ) 3 1st BZ
(61)
The density of states per unit volume, defined as the number of states per unit volume with a
given spin orientation in the energy interval dE is then
1 dN el d ⎛⎜ 1 ⎞
⎟,
D( E ) =
V dE
= ∫ dk
dE ⎜⎝ (2π ) 3 VE ⎟⎠
(62)
where VE is the volume in k space between the iso-energetic surfaces E (k ) and E ( k ) + dE .
As for phonons, dk = dSdk ⊥ = dS (dE / | ∇ k E |) with dS the infinitesimal element on the
E (k ) = const. surface and dk ⊥ normal to this surface, and the expression above simplifies to
1 dS
D( E ) = ∫
(2π ) E ( k ) =const | ∇ k E |
3
, (63)
which becomes
1 meff k meff k (2meff ) 3 / 2

D( E ) = ∫ dΩ k = = ( E − E 0 )1 / 2 (64)
(2π ) 3
h 2
Ωk 2π h
2 2
4π h 2 3
for spherical iso-energetic surfaces, for which E ( k ) = E 0 + h 2 k 2 / 2meff . In this case
| ∇ k E |= h 2 k / meff and dS = k 2 dΩ k , with dΩ k the element of solid angle.
Classification of solid state materials

One of the most important achievements of the energy band theory is the possibility to
classify solid state materials in metals, isolators, and semiconductors. This classification is not
based on the structure of the energy bands, which is quite the same in all materials, but on the
degree of occupation of these bands. The available number of energy states is occupied by
electrons in agreement with the Pauli exclusion principle. More precisely, at low temperatures
(in principle, at T = 0 K) the states are occupied in the order of increasing energy value, such
that only two electrons (with opposite spins) are allowed on an energy state E with a given k
value. The available states are occupied by electrons up to an energy level called Fermi
energy, or Fermi level E F . Because electrons are fermions, their quantum statistical
distribution function at temperature T is described by the Fermi-Dirac formula
1
f (E) = (65)
1 + exp[( E − E F ) / k B T ]
with k B the Boltzmann constant. The

temperature dependence of the Fermi-
Dirac distribution function is illustrated
in the figure at right. At T = 0 K, the
distribution is a step function, equal to 1
for energies smaller than the Fermi
energy, and equal to 0 otherwise.
Two situations can exist:

1) at low temperatures, T ≅ 0 K, the Fermi level is inside an energy band (see the figure
below, left), i.e. electrons occupy partially the last energy band. The material is then a
metal and can easily conduct electricity since the electrons in the vicinity of E F ,
accelerated by a small applied electric field, can occupy available empty states with
higher energy.
2) at low temperatures, T ≅ 0 K, the electrons occupy completely a number of energy bands,

so that the Fermi level is inside the energy gap between the last occupied band, called
valence band, and the following empty band, called conduction band (see the figure
above, right). In this case a small electric field does not provide sufficient energy for the
electron to jump into empty states in the higher energy band, and therefore no electric
current can flow through the material. We deal in this situation with a dielectric.
At higher temperatures, T ≠ 0 K, the dielectric materials, in turn, can be either isolators

or semiconductors, depending on the width of the energy gap between the valence and
conduction bands. If E g < 3 eV, thermal fluctuations can excite electrons from the valence to
the conduction band, where they can contribute to electrical conduction, and the material is in
this case a semiconductor. As a result, in (undoped) semiconductors the number of electrons
in the conduction band is equal to the number of holes in the valence band. When an applied
electric field is applied, the current has two contributions: one from electrons and the other
from holes, which are drifted in opposite directions. In isolators, E g > 3 eV and no electrical
conduction exists at moderate temperatures or electric fields.
Because the total number of electrons in a crystal with N unit cells and s atoms in the
basis, each atom having Z electrons, is NsZ , and an energy band can accommodate 2 N
electrons, it follows that the number of occupied energy bands is NsZ / 2 N = sZ / 2 . Then, in
simple lattices with s = 1, there is always a partially occupied band for odd Z values and these
materials should be metals. This is the case of monovalent alkaline metals (Li, Na, K, Cs, Rb)
and noble metals (Cu, Au, Ag), which have only one valence electron. However, it is not true
that materials with s = 1 and even Z are always dielectrics. For example, bivalent elements
such as Be, Mg, Ca, Sr, B, are metals. The explanation is that, in these cases, the completely
empty conduction band overlaps the completely occupied valence band over a small energy
interval (see the figure below). The electrical conductivity is poorer, though, if the overlap is
slight. Note that in genuine dielectric materials, at T = 0 K there is no overlap between the
conduction and valence bands.
Because at the top of partially occupied energy bands the charge carriers are holes, a
mixed conduction (electrons and holes) is expected when the conduction and valence bands
overlap over a narrow energy range. The materials that have a mixed conduction at T = 0 K
are called semimetals; they differ from metals in that in metals the conduction is always due
to electrons only. The semimetals have more complicated band structures than metals and
dielectrics. Examples of semimetals are As, Sb, Bi; these materials have an electrical
conductivity with up to four orders of magnitude smaller than in metals.
In metals, the iso-energetic surface in the reciprocal space defined at T = 0 K through
E (k ) = E F (66)
is called Fermi surface. The Fermi surface separates the states with a low occupation density
from those with a high occupation density and determines the physical properties of metals,
especially the electrical properties, in which only electrons within an energy interval of the
order of k B T around the Fermi energy participate. Examples of simple and complex Fermi
surfaces are given in the figures below. Note that in Ni, which has strong magnetic properties,
the Fermi surfaces for electrons with opposite spins are different.
Statistics of charge carriers
The statistic properties of charge carriers are determined when these are in
statistical/thermodynamical equilibrium with the crystalline lattice. To find these properties
we need to know the distribution function and the density of states of electrons and holes.
For a general system of electrons characterized by a distribution function (probability
of occupying available electronic states) f (E ) , the concentration of electrons is
1 2 2
n= ∑ f ( E (k )) = ∑ f ( E (k )) = ∫ f ( E (k ))dk
V k ,σ V k (2π ) 3
Emax
(1)
2 f ( E )dSdE
= ∫ |∇ E| = 2 ∫ f ( E ) D( E )dE
(2π ) 3 k Emin
where the factor 2 originates from the sum over the spin states with index σ. The density of
electronic states for spherical iso-energetic surfaces is given by
(2meff ) 3 / 2
D( E ) = ( E − E 0 )1 / 2 . (2)
4π h 2 3
Because there are differences in the distribution function of charge carriers in metals and
semiconductors, in the following we treat these cases separately.
Statistics of electrons in metals

In metals the Fermi energy is inside an energy band and one can define the Fermi surface. The
density of electrons in metals with spherical iso-energetic surfaces is given by (2) with
E0 = 0 and, at a finite temperature T, the electrons occupy the energy states according to the
Fermi-Dirac distribution function
1
f (E) = . (3)
1 + exp[( E − E F ) / k B T ]
In this case, the electron concentration becomes
(2meff ) 3 / 2 ∞
E 1 / 2 dE
n=
2π 2 h 3
∫ exp[( E − E ) / k T ] + 1 (4)
0 F B
Statistics of charge carriers 2
if the upper limit of the integral is approximated with ∞. This approximation is justified since
the integrand decreases rapidly. With the change of variables x = E / k BT , y = E F / k BT , we
finally obtain
(2meff k B T ) 3 / 2
n= F1 / 2 ( y ) (5)
2π 2 h 3
where
∞
x α dx
Fα ( y ) = ∫ (6)
0 exp( x − y ) + 1
are the Fermi-Dirac integrals. They are evaluated numerically. Note that the same result is
obtained for an arbitrary E0 . In this case, in (4) E0 appears in the numerator of the integrand
and in the lower limit of the integral, but disappears in the final result (equations (5) and (6)),
if the new variables are chosen as ( E − E0 ) / k BT = x , ( E F − E0 ) / k B T = y .
The temperature dependence of the Fermi-Dirac distribution function is represented in
the figure below. At all temperatures, f ( E F ) = 1 / 2 .
At T = 0 K, where the Fermi-Dirac distribution is a Heaviside function, i.e.
⎧1, E < E F
f (E) = ⎨ (7)
⎩0, E > E F
the concentration of electrons in metals is

E F0 / k BT 3/ 2
(2meff k B T ) 3 / 2 (2meff k B T ) 3 / 2 2 ⎛ E F0 ⎞ (2meff ) 3 / 2
n= ∫x
1/ 2
dx = ⎜ ⎟ = ( E F0 ) 3 / 2 , (8)
2π h 2 3
0 2π 2 h 3 3 ⎜⎝ k B T ⎟⎠ 3π h2 3
with E F0 the Fermi energy level at T = 0 K, or
h 2 (3π 2 n) 2 / 3 h 2 k F2
E F0 = = , (9)
2meff 2meff
where k F = (3π 2 n)1 / 3 is the Fermi wavenumber. From (9) it follows that the Fermi energy at
T = 0 increases as the electron concentration increases, the system of electrons being in the
fundamental state if the Fermi sphere is completely occupied and in an excited state if
electrons occupy states with | k |> k F .
For finite but low temperatures, y = E F / k B T >> 1 , and spherical iso-energetic

surfaces, the Fermi-Dirac integrals can be approximated with
y α +1 ⎛ π 2 α (α + 1) ⎞
Fα ( y ) = ⎜1 + ⎟, (10)
α + 1 ⎜⎝ 6 y 2 ⎟⎠
and so
(2meff ) 3 / 2 ⎡ π2 ⎛ k BT ⎞
2
⎤
n= E 3/ 2
⎢1 + ⎜⎜ ⎟⎟ ⎥. (11)
3π h 2 3 F
⎢⎣ 8 ⎝ EF ⎠ ⎥⎦
On the other hand, the electron concentration in metals does not depend on temperature
because an increase in T affects only the thermal excitation of electrons on higher energy
levels. It follows thus that the Fermi level must depend on temperature and, from
( E F0 ) 3 / 2 ( E F0 ) 3 / 2
E F3 / 2 = ≅ , (12)
⎡ π2 ⎛ k BT ⎞
2
⎤ ⎡ π2 ⎛ k BT ⎞
2
⎤
⎢1 + ⎜⎜ ⎟⎟ ⎥ ⎢1 + ⎜⎜ 0 ⎟⎟ ⎥
⎢⎣ 8 ⎝ EF ⎠ ⎥⎦ ⎢⎣ 8 ⎝ EF ⎠ ⎥⎦
we obtain
⎡ π2 ⎛ k BT ⎞
2
⎤
E F = E ⎢1 − 0
F ⎜⎜ 0 ⎟⎟ ⎥. (13)
⎢⎣ 12 ⎝ EF ⎠ ⎥⎦
This temperature dependence is very weak and the Fermi energy is, in the first approximation,
almost constant as T varies.
Statistics of charge carriers in semiconductors

In semiconductors the Fermi level is situated between the valence and the conduction band,
which are separated by an energy gap. At T = 0 K, all energy states below E F are occupied
with electrons according to the Pauli principle, i.e. according to the Fermi-Dirac distribution
function, and all states above the Fermi level are empty. Because the holes in semiconductors
can be viewed as states not occupied by electrons, the probability that a state k, with energy
E (k ) is occupied by a hole (i.e. is empty for electrons) is
1 1
1 − f (E) = 1 − = . (14)
1 + exp[( E − E F ) / k B T ] 1 + exp[( E F − E ) / k B T ]
If E F / k B T >> 1 the Fermi-Dirac distribution function has almost the step-function

form characteristic for T = 0 K. In this case the distribution function is called degenerate and
is encountered in materials with large concentrations of electrons, such as metals and heavily
doped semiconductors. On the contrary, when E F / k B T << −1 or − E F / k BT >> 1 , the
exponential term in the Fermi-Dirac distribution function is much larger than unity and the
distribution function resembles the classical Maxwell-Boltzmann distribution,
f cl ( E ) = exp(E F / k BT ) exp(− E / k B T ) << 1 , (15)
which can be seen as the “tail” of the Fermi-Dirac distribution. This case corresponds to low
electron concentrations, in particular to undoped (intrinsic) semiconductors, in which the
electron concentration is with few orders of magnitude smaller than in metals. The Maxwell-
Boltzmann distribution function is called nondegenerate.
Statistics of charge carriers in intrinsic semiconductors

The density of states for electrons in an intrinsic semiconductor is similar to that in metals (for
an arbitrary E0 ) and is found to be
mn k (2mn ) 3 / 2
Dn ( E ) = = ( E − E c )1 / 2 (16)
2π h
2 2
4π h
2 3
for spherical iso-energetic surfaces E (k ) = E c + h 2 k 2 / 2mn , where Ec is the minimum energy
in the conduction band, and mn is the effective mass of electrons in the vicinity of this
minimum energy. In an analogous manner, the density of states for the holes with effective
mass m p in the valence band is found to be
mpk ( 2m p ) 3 / 2
D p (E) = = ( E v − E )1 / 2 , (17)
2π h2 2
4π h 2 3
where Ev is the maximum energy in the valence band. This expression was found taking into
account that the hole energy increases in the opposite direction as that of the electron. Both
electron and hole density of states must be multiplied with 2 if the spin degeneracy is
included.
However, the dispersion relation is not always spherical. For example, the conduction
band of Ge and Si are ellipsoidal iso-energetic surfaces with N eq equivalent minima arranged
symmetrically in the first Brillouin zone ( N eq = 4 for Ge and N eq = 6 for Si). This case can be
reduced to that of spherical iso-energetic surfaces if, in the neighborhood of these minima
situated at k 0 , the dispersion relation
h 2 ⎡ (k1 − k 01 ) 2 (k 2 − k 02 ) 2 (k 3 − k 03 ) 2 ⎤
Ek = Ec + ⎢ + + ⎥ (18)
2 ⎣ mn1 mn 2 mn 3 ⎦
is transformed into a spherical dispersion relation
h2
E k ' = Ec + (k1 ' 2 + k 2 ' 2 + k 3 ' 2 ) (19)
2m n '
by a change of variables k i − k 0i = k i ' mni / mn ' . The figures below illustrate the iso-
energetic surfaces in Si (left) and Ge (right). There are 4 complete ellipsoids (8 half-
ellipsoids) in the first Brillouin zone in Ge.
Then, in the calculation of the density of states we must account for the fact that
dk = dk1 dk 2 dk 3 = [(mn1 mn 2 mn 3 )1 / 2 / mn '3 / 2 ]dk ' , from which it follows that the density of
states becomes
(2mn ' ) 3 / 2 (mn1 mn 2 mn 3 )1 / 2

Dn ( E ) = ( E − E c )1 / 2 . (20)
4π 2 h 3 mn '3 / 2
Because there are N eq equivalent minima/ellipsoids, the density of states is given by
( 2 mc ) 3 / 2
Dc ( E ) = ( E − E c )1 / 2 (21)
4π h
2 3
where the effective mass of the density of states in the conduction band is defined as
mc = N eq2 / 3 ( m n1 m n 2 m n 3 )1 / 3 . (22)
Concentration of charge carriers and Fermi energy

In an intrinsic (undoped) semiconductor, the requirement of electrical neutrality imposes the
same concentration for both electrons and holes,
n= p. (23)
In intrinsic semiconductors the concentrations of charge carriers are small (n and p are smaller
with several orders of magnitude than the electron concentration in metals), so that the
available energy states are no longer occupied according to the Fermi-Dirac distribution
function, as in metals, but according to the classical Maxwell-Boltzmann distribution. The
electron an hole concentrations in intrinsic semiconductors with spherical iso-energetic

surfaces are given by
Emax
( 2m n k B T ) 3 / 2 (2m p k B T ) 3 / 2
n = 2 ∫ f ( E ) Dn ( E )dE = F1 / 2 ( y c ) , p= F1 / 2 ( y v ) (24)
Emin 2π 2 h 3 2π 2 h 3
with y c = ( E F − Ec ) / k BT , y v = ( Ev − E F ) / k BT . The expressions are similar to that found in
metals, with xc = ( E − Ec ) / k BT , xv = ( Ev − E ) / k BT , respectively, except that in this case

the Fermi-Dirac integrals become
∞
Fα ( y ) = exp( y ) ∫ x α exp( − x )dx , (25)
0
since we use the Maxwell-Boltzmann distribution. Because F1 / 2 ( y ) = ( π / 2) exp( y ) ,
(2πmn k B T ) 3 / 2 ⎛ E − Ec ⎞ ⎛ E − Ec ⎞ ⎛ E − EF ⎞
n= exp⎜⎜ F ⎟⎟ = N c exp⎜⎜ F ⎟⎟ = N c exp⎜⎜ − c ⎟⎟ , (26)
4π h3 3
⎝ k BT ⎠ ⎝ k BT ⎠ ⎝ k BT ⎠
( 2πm p k B T ) 3 / 2 ⎛ E − EF ⎞ ⎛ E − EF ⎞ ⎛ E − Ev ⎞
p= exp⎜⎜ v ⎟⎟ = N v exp⎜⎜ v ⎟⎟ = N v exp⎜⎜ − F ⎟⎟ , (27)
4π h
3 3
⎝ k BT ⎠ ⎝ k BT ⎠ ⎝ k BT ⎠
where
(2πmn k B T ) 3 / 2 ( 2πm p k B T ) 3 / 2
Nc = , Nv = , (28)
4π 3 h 3 4π 3 h 3
are the effective densities of states in the conduction and valence bands, respectively. From
the relations above it follows that, unlike the degenerate case in metals, the concentrations of
electrons and holes in non-degenerate semiconductors depend strongly on temperature.
The Fermi energy is determined from the neutrality condition, i.e. from
E
⎛ E − Ec ⎞ ⎛ E − EF ⎞ Ec
m 3/ 2
n exp⎜⎜ F ⎟⎟ = m 3p / 2 exp⎜⎜ v ⎟⎟ . (29)
⎝ k BT ⎠ ⎝ k BT ⎠
The result is EF Eg/2
Ec + Ev 3 ⎛ mp ⎞ Ev
EF = + k B T ln⎜⎜ ⎟⎟ (30)
2 4 ⎝ mn ⎠ T
and, at T = 0 K, the Fermi level is at the middle of the energy gap: E F = ( Ec + Ev ) / 2 . As the
temperature increases, the Fermi level shifts towards the energy band with the smaller
effective mass (see the figure above); it remains at the centre of the energy gap, irrespective
of temperature, only if mn = m p . A special situation is encountered in semiconductors with a
small energy gap width E g = E c − Ev and m p / mn >> 1 or m p / mn << 1 (the first situation is
much more common). In this case, the Fermi level can enter inside the band with the smaller
effective mass, and the semiconductor becomes degenerate at even moderate temperatures.
An example of such a material is InSb, for which E g = 0.18 eV and m p / mn ≅ 10 .
It should be emphasized that, starting from the expression of the electron

concentration in semiconductors, it follows that the conditions of applicability of the
Maxwell-Boltzmann statistics is
⎛ E − Ec ⎞ 4π 3 h 3 n
exp⎜⎜ F ⎟⎟ = << 1 . (31)
⎠ (2πmn k B T )
3/ 2
⎝ k BT
If the energy reference is chosen such that Ec = 0 , exp(E F / k BT ) << 1 for small electron
concentrations, high temperature, and high effective masses. On the contrary, if these
conditions are not satisfied and exp(E F / k BT ) >> 1 , i.e. if the electron concentration is high,
the temperature is low, and the effective mass is small, the electron gas is degenerate, and (31)
is known as the degeneracy criterion. Electrons in metals satisfy the degeneracy criterion.
The intrinsic concentration of carriers can be defined as
⎛ Eg ⎞ ⎛ Eg ⎞
3/ 2
⎛ k ⎞
ni = n = p = np = N c N v exp⎜⎜ − ⎟⎟ = 2⎜ B 2 ⎟ (mn m p ) 3 / 4 T 3 / 2 exp⎜⎜ − ⎟⎟ . (32)
⎝ 2k B T ⎠ ⎝ 2πh ⎠ ⎝ 2k B T ⎠
This concentration does not depend on the Fermi level, and its temperature dependence is
summarized in the formula (see figure)
ln(ni/T 3/2)
⎛ Eg ⎞
ni (T ) ∝ T 3/ 2
exp⎜⎜ − ⎟⎟ , (33)
⎝ 2k B T ⎠
and the band gap can be determined from the slope of the figure
at the right side. For a large class of semiconductors, the energy
1/T
gap width depends on temperature as: E g (T ) = E g 0 − αT .
Statistics of charge carriers in extrinsic semiconductors

An extrinsic semiconductor is a doped semiconductor, i.e. it contains donor and/or acceptor
impurities In donor impurities, the number of valence electrons is higher than in the host
material, and the extra electrons are not involved in binding with the atoms of the host
material; they are still localized around the donor impurity but can easily participate at the
process of charge carrier transport when an electric field is applied. The number of free
electrons that contribute to electrical conduction increases in the presence of donor impurities.
On the contrary, acceptor impurities have a smaller number of valence electrons than the host
material, and their stable binding with host atoms requires an additional electron from the host
material. As a result, the number of holes increases when acceptor impurities are present.
+
B
For Si (see the schematic figures above), which forms covalent bonds with other four atoms,
P is a donor impurity since it has five valence electrons, and B is an acceptor impurity
because it has only three valence electrons. Donor and/or acceptor impurities are purposefully
introduced into a host semiconductor to increase its electrical conduction. On the other hand,
the unintentional doping of a crystal introduces defects into the crystalline lattice and
degrades its electrical performances. Therefore, doping is a process that should be carefully
controlled. The donor and acceptor impurities have discrete energy levels inside the bandgap
of the host material, which can be occupied by electrons and, respectively, holes. The charge
carriers on the impurity levels are localized around donors and acceptors and are not free to
move around the host crystal unless the impurities are ionized, i.e. the electrons on donor
energy levels are promoted (by thermal energy or applied electric fields) into the conduction
band and the holes on acceptor levels are excited into the valence band of the host material.
Let us denote by N d and N a the concentration of donor and acceptor impurities, by
N d0 and N a0 the concentrations of electrons and holes localized on the donors and acceptors,
respectively, i.e. the concentrations of neutral donors and acceptors, and by N d+ = N d − N d0

and N a− = N a − N a0 the concentrations of the ionized impurities. The ionization of donors and
acceptors leads to an increase of the number of free charge carriers with respect to the
intrinsic semiconductor case. If n and p are the total concentrations of free electrons and
holes, the condition of charge neutrality is
n + N a− = p + N d+ , (34a)
or
n + N a + N d0 = p + N d + N a0 . (34b)
Unlike in the case of free charge carriers, the distribution function of the electrons localized
on donor impurities and that of the holes localized on acceptor impurities is not the Fermi-
Dirac function, which is valid when two electrons with opposite spins can occupy an energy
level, according to the Pauli principle. The reason is that only a single electron can occupy an
energy level on an impurity atom. If another electron is brought on this level, its energy varies
significantly due to the strong electrostatic interaction between electrons.
According to the Fermi-Dirac distribution function, the ratio between the probability
that the state is occupied and the probability that the state is empty is,
f (E) ⎛ E − EF ⎞
= exp⎜⎜ − ⎟⎟ . (35)
1 − f (E) ⎝ k BT ⎠
On the contrary, since only one electron can exist on an impurity level, the same ratio is now
f d (E) ⎛ E − EF ⎞
= 2 exp⎜⎜ − ⎟⎟ (36)
1 − f d (E) ⎝ k BT ⎠
for electrons on donors (the level is occupied twice as fast), or
1
f d (E) = , (37)
1 + (1 / 2) exp[( E − E F ) / k B T ]
and
1
f a (E) = (38)
1 + 2 exp[( E F − E ) / k B T ]
for holes on acceptor impurities. In general, the factor (1 / 2) in front of the exponential term
in the denominator of (37) should be replaced by (1 / g ) , with g the degeneracy of the energy
level, and the factor 2 in front of the exponential term in the denominator of (38) should be
replaced by g. In general, f a ( E ) = 1 − f d ( E ) .
Taking into account that the density of states of the discrete donor/acceptor energy
levels Ei are given by
Di ( E ) = N i δ ( E − Ei ) , i = d,a (39)
the concentrations of electrons localized on donors and of holes localized on acceptors are
Nd
N d0 = ∫ f d ( E ) Dd ( E )dE = , (40)
1 ⎛ E − EF ⎞
1 + exp⎜⎜ d ⎟⎟
2 ⎝ k BT ⎠
Na
N a0 = ∫ f a ( E ) Da ( E )dE = , (41)
⎛ E − Ea ⎞
1 + 2 exp⎜ F ⎟
⎝ k BT ⎠
and the neutrality condition becomes
Nd ⎛ E − Ec ⎞
Na + + N c exp⎜ F ⎟=
1 ⎛ Ed − E F ⎞ ⎝ k BT ⎠
1 + exp⎜ ⎟
2 ⎝ k BT ⎠
(42)
Na ⎛ E − EF ⎞
Nd + + N v exp⎜ v ⎟
⎛ E F − Ea ⎞ ⎝ k BT ⎠
1 + 2 exp⎜ ⎟
⎝ k BT ⎠
from which one can determine the position of the Fermi energy level. The Fermi level in
extrinsic semiconductors is different than in intrinsic semiconductors!
Extrinsic semiconductors with only one impurity type

Let us consider that the semiconductor is nondegenerate and has only donor impurities with
concentration N d = N d0 + N d+ . In this case N a = 0 , N a0 = 0 , and (34b) can be expressed as
n + N d0 = p + N d , (43)
with n and p determined as above, with the help of the Maxwell-Boltzmann statistics. The
electrons in the conduction band are generated either through the ionization of donor
impurities (through the transition from the donor level to an energy level in the conduction
band), process that requires the energy E gd = E c − E d (equal to the ionization energy of the
donor impurity), or through the ionization of the atoms in the crystal (the transition of an
electron from the valence in the conduction band), process that requires an energy equal to
E g . Because E gd << E g , the contribution of the two processes differs as a function of the
temperature. More precisely, at low temperatures the dominant process is the ionization of
impurities, whereas at high temperatures the electron transitions between the valence and
conduction bands prevail. We have extrinsic conduction at low temperatures, and intrinsic
conduction at high temperatures.
I) Extrinsic conduction regime

At low temperatures, at which the conduction electrons originate from the ionization of donor
impurities, for a given donor concentration N d >> p the neutrality condition becomes
n + N d0 = N d , or
⎛ E − Ec ⎞ Nd
N c exp⎜⎜ F ⎟⎟ = N d − . (44)
⎝ k BT ⎠ 1 ⎛ E − EF ⎞
1 + exp⎜⎜ d ⎟⎟
2 ⎝ k BT ⎠
This is a second order equation for exp(E F / k BT ) , which can be easily solved by introducing
the variables x = exp[(E F − E d ) / k B T ] , y = ( N d / N c ) exp[(Ec − E d ) / k B T ] . In terms of these
variables (44) can be written as 2 x 2 + x − y = 0 , from which it follows that
⎡1 ⎛ N ⎛ E − Ed ⎞ ⎞⎟⎤
E F = E d + k B T ln ⎢ ⎜ 1 + 8 d exp⎜⎜ c ⎟⎟ − 1 ⎥ . (45)
⎢⎣ 4 ⎜⎝ Nc ⎝ k BT ⎠ ⎟⎠⎥⎦
For extremely low temperatures, for which 8( N d / N c ) exp[(Ec − E d ) / k B T ] >> 1 , (45)
can be approximated as E F = E d + k BT ln ( N d / 2 N c ) exp[(Ec − E d ) / k BT ] , or
Ec + Ed k BT ⎛ N d ⎞
EF = + ln⎜⎜ ⎟⎟ , (46)
2 2 ⎝ 2N c ⎠
which reduces to
Ec + Ed
EF = (47)
2
at T = 0 K. The temperature dependence of the Fermi level can be determined taking into
account that N c = ( 2πm n k B T ) 3 / 2 /( 4π 3 h 3 ) ∝ T 3 / 2 . At temperatures of only few K, when
2 N c < N d , E F shifts towards the conduction band but, as the temperatures increases until
2 N c = N d , the Fermi level takes again the value at T = 0 K. Thus, in this temperature interval
E F reaches a maximum value at a temperature
N d2 / 3πh 4 / 3
Tmax = , (48)
21 / 3 ek B mn
determined from the condition dE F / dT = 0 , or ln( N d / 2 N c ) = 3 / 2 . In (48) e is not the

electric charge, but the basis of the natural logarithm!
The maximum value of the Fermi energy is found to be
Ec + E d 3π h 4 / 3 2 / 3
E F ,max = E F (Tmax ) = + 7/3 Nd . (49)
2 2 emn
The Fermi level can even reach the minimum value of the conduction band, i.e. E F , max = Ec
for a critical concentration impurity
3/ 2
4 ⎛ emn ⎞
N d ,cr = ⎜ ⎟
3/ 2
E gd . (50)
h 2 ⎝ 3π ⎠
E
Ec
Egd/2 ln(n/T 3/4)
Ed
Eg/2
1 2 3
Ev
1/T
Ts Ti T
At this critical concentration the semiconductor becomes degenerate. The temperature

dependence of the Fermi level is represented in the figure above, left.
A further increase in temperature, which corresponds to 2 N c > N d , leads to a
decrease in the Fermi level value towards Ed , until this value is reached for a so-called
saturation temperature Ts . The temperature interval 0 < T < Ts is called the weak ionization
region (see region 1 in the figure above). In this temperature interval, from (46) it follows that
⎛ E − EF ⎞ Nc Nd ⎛ E − Ec ⎞ Nc Nd ⎛ E gd ⎞
n = N c exp⎜⎜ − c ⎟⎟ = exp⎜⎜ d ⎟⎟ = exp⎜⎜ − ⎟⎟ , (51)
⎝ k BT ⎠ 2 ⎝ 2k B T ⎠ 2 ⎝ 2k B T ⎠
i.e. n ∝ N d1 / 2 , and, since N c ∝ T 3 / 2 , the temperature dependence of the electron concentration
is n ∝ T 3 / 4 exp( − E gd / 2k B T ) . The ionization energy of the donor impurities, E gd , can thus be
determined from the slope of the ln(n / T 3 / 4 ) = f (1 / T ) plot (see the figure above, right).
At still higher temperatures, for which 8( N d / N c ) exp[(Ec − E d ) / k B T ] << 1 , (45) can
be approximated as E F = E d + k BT ln{( N d / N c ) exp[(Ec − E d ) / k BT ]} , or
E F = Ec + k B T ln( N d / N c ) , (52)
the logarithm being negative since N c >> N d . Thus, the Fermi energy decreases as the
temperature increases and becomes lower than Ed , level reached at the saturation temperature
E gd
Ts = . (53)
k B ln[ N c (Ts ) / N d ]
In this temperature interval the electron concentration is given by (see (52))
⎛ E − EF ⎞
n = N c exp⎜⎜ − c ⎟⎟ = N d , (54)
⎝ k BT ⎠
result that shows that the donor impurities are totally ionized, and the electron concentration
is independent of temperature for T > Ts . The regime is an exhausting regime for donor
impurities and in the figure above is indicated as region 2. Note that in this temperature region
the Fermi energy level is positioned below E d , which is in agreement of the fact that the
donor impurities are totally ionized (the donor electron states are empty).
II) Intrinsic conduction regime

For high-enough temperatures the hole concentration starts to increase and becomes
comparable with the electron concentration. In particular, if p >> N d0 the neutrality condition
(43) can be written as n = p + N d . In this regime of high temperatures the donors are
completely ionized, the charge carriers originating from the ionization of the host semicon-
ductor material. For a nondegenerate semiconductor p = ni2 / n , which introduced in the
neutrality condition leads to
n 2 − N d n − ni2 = 0 , (55)
the solution of this equation being
Nd ⎛ 2 ⎞ 2ni2
n= ⎜1 + 1 + 4 n i ⎟, p = ni2 / n = . (56)
2 ⎜ N d2 ⎟ ⎛ ⎞
⎝ ⎠ ⎜ ni2
Nd 1+ 1+ 4 2 ⎟
⎜ Nd ⎟
⎝ ⎠
Because at high temperatures the host material is the main source of charge carriers, the
expressions for electron and hole concentrations in intrinsic semiconductors apply, and the
Fermi energy level, determined from n = N c exp[(E F − Ec ) / k B T ] , with n from (56), is
⎡N ⎛ ⎞⎤ ⎡ ⎛ ⎞⎤
⎜1 + 1 + 4 N c N v exp⎛⎜ − E g ⎞⎟ ⎟⎥ .
2
E F = E c + k B T ln ⎢ d ⎜1 + 1 + 4 n i ⎟⎥ = E c + k B T ln ⎢ N d
⎜ ⎟⎥ ⎜ ⎜ k T ⎟⎟
⎢⎣ 2 N c ⎝ N d2 ⎠⎦ ⎢⎣ 2 N c ⎝ N d2 ⎝ B ⎠ ⎠⎥⎦
(57)
The expression above can be studied in two extreme situations:

1) 4ni2 / N d2 << 1 , case in which
n = N d , p = ni2 / N d , (58)
and (in agreement with our previous results for the totally ionized/exhausted impurities)
E F = Ec + k BT ln( N d / N c ) , (59)
2) 4ni2 / N d2 >> 1 , case in which
n = p = ni (60)
and (as for the intrinsic semiconductor)
Ec + Ev k BT ⎛ N v ⎞ Ec + Ev 3 ⎛ mp ⎞
EF = + ln⎜⎜ ⎟⎟ = + k B T ln⎜⎜ ⎟⎟ . (61)
2 2 ⎝ Nc ⎠ 2 4 ⎝ mn ⎠
The temperature dependence of the Fermi level in this region of intrinsic conduction is
indicated in the figure above (see region 3). At high-enough temperatures the increase of the
electron concentration in the conduction band originates from electron transitions from the
valence band. The transition temperature from the exhausting regime of impurities to the
region of intrinsic conduction can be determined from (58) and (60), i.e. from ni = N d , and is
found to be
Eg
Ti = . (62)
k B ln[ N c (Ti ) N v (Ti ) / N d2 ]
Summarizing, the temperature dependence of the electron concentration shows three

distinct regions (see the figures below). The logarithmic dependence of the concentration on
the inverse of the temperature can be approximated with a straight line in regions 1 and 3 (see
figure below, left) if we neglect the influence of the factors T 3 / 2 and T 3 / 4 , respectively, in
comparison with the exponentials terms, and the parameters E g and E gd can be determined
from the corresponding slopes. On the contrary, in region 2 the electron concentration is
approximately constant, since the donor impurities are exhausted.
lnn
n
freeze-out
intrinsic
extrinsic
3 2 1 T
1/T Ts Ti
1/Ti 1/Ts
Electronic specific heat
Electronic specific heat in metals

In metals, the electronic specific heat per unit volume, calculated at constant volume, is
defined as
dE el
C el = (1)
dT
where the energy per unit volume of the system of non-interacting electrons is given by
∞
1 2
E el = ∑ E (k ) f ( E (k )) = ∑ E (k ) f ( E (k )) = 2 ∫ Ef ( E ) D ( E )dE , (2)
V k ,σ V k 0
with D ( E ) = ( 2meff ) 3 / 2 E 1 / 2 / 4π 2 h 3 (see the course on electron statistics in metals). In the
normalized coordinates E / k BT = x , E F / k BT = y and for spherical iso-energetic surfaces the

energy per unit volume becomes
(2meff ) 3 / 2 ∞
E 3 / 2 dE (2meff k B T ) 3 / 2 k B T F ( y)
E el = ∫ exp[( E − E ) / k T ] + 1 = F3 / 2 ( y ) = nk B T 3 / 2 , (3)
2π 2 h 3 0 F B 2π h 2 3
F1 / 2 ( y )
where
∞
x α dx
Fα ( y ) = ∫ (4)
0 exp( x − y ) + 1
are the Fermi-Dirac integrals. The last equality in (3) follows because (see the course on the
statistics of electrons in metals)
(2meff k B T ) 3 / 2
n= F1 / 2 ( y ) . (5)
2π 2 h 3
So, taking into account that
dFα ( y ) / dy = αFα −1 ( y) , (6)
the electronic heat capacity can be expressed as

Electronic specific heat 2
⎛ 5 F3 / 2 ( y ) dy ⎞
3/ 2
5 (2meff k B T ) k B (2meff k B T ) 3 / 2 k B T dF3 / 2 dy 3
C el = F ( y ) + = nk B ⎜⎜ +T ⎟.
2π h 2π h dT ⎟⎠
2 3 3/ 2 2 3
2 dy dT 2 ⎝ 3 F1 / 2 ( y )
(7)
y α +1 ⎛ π 2 α (α + 1) ⎞ ⎡ π2 ⎛ k BT ⎞
2
⎤
At low temperatures, from Fα ( y ) = ⎜⎜1 + ⎟ , E = E 0
⎢1 − ⎜⎜ 0 ⎟⎟ ⎥ and
α +1⎝ y 2 ⎟⎠
F F
6 ⎢⎣ 12 ⎝ EF ⎠ ⎥⎦
y = E F / k B T >> 1 we find that
F3 / 2 ( y ) 3 ⎛ π2 ⎞ 3 EF ⎡ π2 ⎛ k BT ⎞
2
⎤ 3 E0 ⎡ 5π 2 ⎛ k BT ⎞
2
⎤
= y⎜⎜1 + 2 ⎟⎟ ≅ ⎢1 + ⎜⎜ 0 ⎟⎟ ⎥= F
⎢1 + ⎜⎜ 0 ⎟⎟ ⎥ (8a)
F1 / 2 ( y ) 5 ⎝ 2 y ⎠ 5 k BT ⎢⎣ 2 ⎝ EF ⎠ ⎥⎦ 5 k BT ⎢⎣ 12 ⎝ EF ⎠ ⎥⎦
dy 1 ⎛ dE F E F ⎞ E F0 ⎡ π2 ⎛ k BT ⎞
2
⎤
T = ⎜ − ⎟=− ⎢1 + ⎜⎜ 0 ⎟⎟ ⎥ (8b)
dT k B ⎝ dT T ⎠ k BT ⎢⎣ 12 ⎝ EF ⎠ ⎥⎦
and so
3 π 2 k BT cl π
2
T
C el = nk B 0
= C el , (9)
2 3 EF 3 TF
where the Fermi temperature is a parameter defined as TF = E F0 / k B and C elcl = (3 / 2)nk B is
the classical electronic specific heat. Celcl is obtained using the same general expression (7) as
above, but with the Fermi-Dirac distribution function replaced by the Maxwell-Boltzmann
distribution, case in which
∞
Fα ( y ) ≅ ∫0 x α exp[ −( x − y )]dx , (10a)
F3 / 2 ( y ) / F1 / 2 ( y ) = 3 / 2 , (10b)
dE F / dT = E F / T − 3 / 2k B , (10c)
Tdy / dT = −3 / 2 . (10d)
The equality (10c) follows from (5) and the requirement that dn / dT = 0 , considering that in
the nondegenerate case dFα ( y ) / dT = (dFα / dy)(dy / dT ) = (dy / dT ) Fα ( y ) . Then, (10d) is
obtained from (10c) and (8b), so that, finally, C elcl = (3 / 2)nk B .

The ratio T / TF can be seen as the fraction of excited electrons at temperature T, the
other electrons being “frozen” due to the Pauli principle. The value of this ratio at room
temperature is typically 10−2.
The linear relation between the electronic specific heat in metals and temperature is
generally expressed as
C el = γT , (11)
where γ = π 2 nk B2 / 2 E F0 is known as the Sommerfeld constant. Although this constant has

been derived using the approximation of spherical iso-energetic surfaces, its value remains the
same for general surfaces.
Taking into account also the phononic contribution to the specific heat (see the lecture
on phononic heat capacity), at low temperatures the specific heat is given by (see the figure
below)
CV = C el + C ph = γT + aT 3 , (12)
where
12π 4 k
a= nion B3 (13)
5 ΘD
with nion the ion concentration. The electronic term dominates at very low temperatures, for
which γ > aT 2 , i.e. for

T 1 5 n ΘD
< . (14)
Θ D 2π 6 nion TF
In particular, the Debye temperature can be determined from the slope of the curve
CV / T = γ + aT 2 = f (T 2 ) at very low temperatures, while γ is determined from the value of
this dependence at T = 0. The values of γ for several metals are given in the table below.
Metal γ ·10−4 (J/mol·K2) Metal γ ·10−4 (J/mol·K2) Metal γ ·10−4 (J/mol·K2)

Li 17 Ag 6.6 Zn 6.5
Na 17 Au 7.3 Al 13.5
K 20 Be 2.2 Fe 49.8
Cu 6.9 Mg 13.5 Co 47.3
Ca 27.3 Ba 27 Ni 70.2
Carrier specific heat in intrinsic semiconductors

In a nondegenerate intrinsic semiconductor with spherical iso-energetic surfaces, the energies
of the system of electrons and holes are given by, respectively (see (3) and (10b))
F3 / 2 ( y ) 3 3
E el = nk B T = nk B T , Eh = pk B T . (15)
F1 / 2 ( y ) 2 2
The total energy of charge carriers is however equal to
3 3
E carr = nk B T + nE g + pk B T (16)
2 2
since the free electrons in the conduction band have an additional potential energy of nE g .
Because in an intrinsic semiconductor n = p = ni , with ni the intrinsic carrier concentration,

(16) can be written as
E carr = ni (3k B T + E g ) , (17)
and the carrier specific heat is

dE carr dni
C carr = = (3k B T + E g ) + 3ni k B , (18)
dT dT
when the weak temperature dependence of the bandgap is neglected. Because
⎛ Eg ⎞ ⎛ Eg ⎞
ni = N c N v exp⎜⎜ − ⎟⎟ ∝ T 3 / 2 exp⎜⎜ − ⎟⎟ , (19)
⎝ 2k B T ⎠ ⎝ 2k B T ⎠
it follows that
dni n ⎛ Eg ⎞
= i ⎜⎜ 3 + ⎟ (20)
dT 2T ⎝ k B T ⎟⎠
and
⎡15 Eg 1 ⎛ Eg ⎞ ⎤
2 2
nk ⎛ Eg ⎞
C carr = i B ⎜⎜ 3 + ⎟ + 3ni k B = ni k B ⎢ + 3 + ⎜ ⎟ ⎥. (21)
k B T ⎟⎠ k B T 2 ⎜⎝ k B T ⎟⎠ ⎥
⎝ ⎣⎢
2 2
⎦
This expression is valid if E g ≥ k B T , since otherwise the degeneracy of the system of
electrons and holes must be taken into account. From (21) it follows that at low temperatures
the contribution of charge carriers to the specific heat in an intrinsic semiconductor can be
neglected, due to the exponential temperature dependence of ni .
In an extrinsic semiconductor the specific heat of charge carriers can be calculated in a
similar manner. More precisely, in (16) one must introduce the correct concentrations of free
carriers in all conduction regimes, and must account for their specific distribution function
and temperature dependence. The carrier specific heat of free electrons and holes is found,
then, to depend on both the concentration of donor and acceptor ions and of their energy
levels. At low temperatures this contribution to the specific heat is, again, negligible.
Kinetics of charge carriers in solids
Boltzmann kinetic equation

When an electric or a magnetic field is applied on a crystal, the displacement of charge
carriers induces transport (or kinetic) phenomena. The distribution function of charge carriers
with energy E (k ) = E k in equilibrium is described by the Fermi-Dirac function
1
f 0 (Ek ) = . (1)
1 + exp[( E k − E F ) / k B T ]
On the other hand, in the presence of external fields, the system of charge carriers is no longer
in equilibrium and the corresponding distribution function f (k , r , t ) depends, in general, on
spatial coordinates and time.
In a semiclassical treatment, the number of particles that follow a certain trajectory is
conserved in the absence of scattering processes, so that df / dt = 0 . However,
scattering/collision processes of electrons on phonons, impurities or defects in the crystalline
lattice are unavoidable, so that the total derivative of the distribution function does not vanish
any more, but is equal to the variation of the distribution function due to collisions. More
precisely,
df ∂f ∂f ∂f & ⎛ ∂f ⎞
= + ⋅ r& + ⋅k = ⎜ ⎟ , (2)
dt ∂t ∂r ∂k ⎝ ∂t ⎠ coll
or
∂f ⎛ ∂f ⎞ F
= ⎜ ⎟ − v ⋅ ∇r f − ⋅ ∇k f (3)
∂t ⎝ ∂t ⎠ coll h
where v is the electron velocity in the crystal and F = dp / dt = hdk / dt is the external force.
In a stationary state, when the distribution function is independent of time, ∂f / ∂t = 0 , and, if
we consider the effect of the Lorentz force F = −e( E + v × B ) only, we obtain the kinetic
Boltzmann equation
e ⎛ ∂f ⎞
v ⋅ ∇ r f − ( E + v × B) ⋅ ∇ k f = ⎜ ⎟ . (4)
h ⎝ ∂t ⎠ coll
Kinetics of charge carriers in solids 2
To find the distribution function f (k , r , t ) from this equation it is necessary to know the
collision term in the right-hand-side. This is a difficult problem, which can be simplified by
introducing the relaxation time τ (k ) , which describes the return to equilibrium of the
distribution function when the external fields are switched off:
⎛ ∂f ⎞ f − f0
−⎜ ⎟ = , (5)
⎝ ∂t ⎠ coll τ (k )
or
f − f 0 = ( f − f 0 ) t =0 exp[−t / τ (k )] . (6)
The relaxation time is thus the interval after which the change in the equilibrium distribution
function decreases e times after the external fields are turned off. The introduction of the
relaxation time parameter is possible when the collision processes are elastic, i.e. when the
energy of charge carriers is not modified at scattering, and act independently (there is no
interference of electron states). Moreover, the inequality τ >> h / k BT must be satisfied, where
h / k B T = τ c is the collision time. This inequality expresses the fact that the collision time can
be neglected, i.e. the collisions are instantaneous. In addition, the external fields must not
modify the energy spectrum of electrons in the crystal; this condition prohibits intense
magnetic fields, for example, which lead to the quantization of electron energy levels.
The quantum nature of electrons is apparent only in the collision term, through the
electron quantum states that satisfy the Pauli principle. A detailed balance between the
number of electrons in the state characterized by the wavevector k and those in the state k '
leads to the collision term
⎛ ∂f ⎞
⎜ ⎟ = ∑ P(k ' , k ) f (k ' )[1 − f (k )] − ∑ P(k , k ' ) f (k )[1 − f (k ' )] (7)
⎝ ∂t ⎠ coll k' k'
where P (k , k ' ) is the electron transition probability per unit time from state k into the state
k ' . In the equilibrium state
P(k ' , k ) f 0 (k ' )[1 − f 0 (k )] = P(k , k ' ) f 0 (k )[1 − f 0 (k ' )] . (8)

We consider distribution functions that can be approximated as perturbations of f 0 ,
i.e. that can be expressed as f (k ) = f 0 ( E k ) + f1 (k ) , with
df 0
f1 (k ) = − χ ( E k ) ⋅ k << f 0 ( E k ) , (9)
dE k
where χ ( Ek ) an, as yet, unknown vectorial function (it has specific forms for different
scattering processes). Under these conditions we can express the collision term as
⎛ ∂f ⎞ ⎛ f (k ' )[1 − f (k )] f (k )[1 − f (k ' )] ⎞

⎜ ⎟ = ∑ P(k , k ' ) f 0 (k )[1 − f 0 (k ' )]⎜⎜ − ⎟⎟
⎝ ∂t ⎠ coll k' ⎝ f 0 (k ' )[1 − f 0 (k )] f 0 (k )[1 − f 0 (k ' )] ⎠ (10)
1
=− ∑ P(k , k ' ) f 0 ( E k )[1 − f 0 ( E k ' )][ χ ( E k ) ⋅ k − χ ( E k ' ) ⋅ k ' ]
k B T k'
if only the linear terms in f1 are retained and the identity
df 0
− k BT = f 0 ( E )[1 − f 0 ( E )] (11)
dE
is employed. Then,
1 1 ⎛ ∂f ⎞ ⎡1 − f 0 ( E k ' ) ⎤ ⎡ χ ( E k ' ) ⋅ k ' ⎤

=− ⎜ ⎟ = ∑ P(k , k ' ) ⎢ ⎥ ⎢1 − χ ( E ) ⋅ k ⎥ , (12)
τ (k ) f 1 (k ) ⎝ ∂t ⎠ coll k' ⎣1 − f 0 ( Ek ) ⎦ ⎣ k ⎦
or 1 / τ (k ) = ∑ P(k , k ' )(1 − k ' χ / k χ ) for elastic collisions, when E k = E k ' . Here k ' χ , k χ are
k'
the projections of k, k ' on the vector χ. The calculation of the relaxation time can be
performed for different scattering mechanisms, the temperature dependence of this parameter
being generally expressed as
τ ( E ) = A(T ) E r , (13)
where A(T ) is an energy-independent coefficient, and r is characteristic for each collision

type:
r = 3 / 2 for collisions with acoustic phonons in metals,

r = −1 / 2 for the same mecha-nism in semiconductors,
r = 1 / 2 for scattering on optical phonons in polar semiconductors at high temperatures,

r = 0 for scattering on neutral impurities in metals,
r = 3 / 2 for collisions with ionized impurities in semiconductors.
If several scattering mechanisms coexist,
1 1
=∑ . (14)
τ (E) i τ i (E)
Electrical conductivity
The electrical conductivity in a crystal is characterized by the tensor σ̂ that appears in the
definition of the density of electric current:
j = σˆE , (15)
or on components μ, ν = x, y, z
j μ = ∑ σ μν Eν . (16)
ν
In an isotropic solid the electrical conductivity is a scalar parameter and j = σE . On the other
hand, for f (k ) = f 0 ( E k ) + f1 (k ) , the density of electric current per crystal volume can be
expressed as
1 2e e
j= ∑ − evf (k ) = − ∑ vf1 (k ) = − 3 ∫ vf1 (k )dk (17)
V k ,σ V k 4π
since the equilibrium density function f 0 ( E k ) does not bring any contribution if the sum
above is performed over all positive and negative k values ( Ek , and hence f 0 ( E k ) , is an even
function of k, whereas v = h −1∇ k E is an odd function of k). The perturbation term of the
equilibrium distribution function, f1 (k ) , is determined from the kinetic Boltzmann equation.

More precisely, if only an electric field of intensity E is applied and there is no temperature
gradient in the sample, i.e. if ∇ r f = 0 , the kinetic Boltzmann equation can be written as
e e f (k )
− E ⋅ ∇ k f 0 − E ⋅ ∇ k f1 + 1 = 0. (18)
h h τ (k )
The second term can be neglected at small electric fields, when only linear effects in E are
considered, case in which the electrical conductivity is independent of the electric field, and
the equation above can be solved to obtain
df 0 df
f1 (k ) = eτ (k ) E ⋅ h −1∇ k E = eτ ( k ) E ⋅ v 0 . (19)
dE dE
From (15), (17) and (19) it follows that the tensor of the electrical conductivity can be
expressed as
e 2 ⎛ df 0 ⎞
σ μν = ∫ ⎜ − ⎟τ (k )v μ vν dk . (20)
4π 3 ⎝ dE ⎠
In isotropic crystals, for an electric field along the x direction, the conductivity is scalar:
e2 ⎛ df 0 ⎞
σ = σ xx = ∫ ⎜ − dE ⎟τ (k )v x dk .
2
(21)
4π 3 ⎝ ⎠
For spherical iso-energetic surfaces, E k = h 2 k 2 / 2meff , v μ = hk μ / 2meff , and in spherical
coordinates with θ the polar angle and ϕ the azimuthal angle, k x = k sin θ cos ϕ ,
k y = k sin θ sin ϕ , k z = k cosθ , dk = k 2 dk sin θdθdϕ , and
π 2π
e 2 h 2 ⎛ df 0 ⎞ e 2 h 2 k max ⎛ df 0 ⎞
σ = σ xx = 2 ∫⎜
− ⎟τ (k )k dk ∫ sin θdθ ∫ cos ϕdϕ =
4 3 2
∫ ⎜− ⎟τ (k )k dk .
4
4π meff ⎝ dE ⎠
3
0 0 3π meff k min ⎝ dE ⎠
3 2
(22)
Taking into account that k 4 dk = ( 2meff ) 3 / 2 meff E 3 / 2 dE / h 5 and that the relaxation time depends
on energy, such that its statistical average can be defined as
∞
⎛ 0 ⎞
df
∫ ⎜ − dE ⎟τ ( E ) E dE
3/ 2
〈τ ( E )〉 = 0 ⎝ ∞ ⎠ , (23)
⎛ df 0 ⎞ 3 / 2
∫ ⎜ − dE ⎟ E dE
0⎝ ⎠
the electrical conductivity becomes

ne 2
σ= 〈τ ( E )〉 , (24)
meff
where we have used the fact that for metals the electron concentration can be written as
(2meff ) 3 / 2 ∞
E 1 / 2 dE
n= ∫ exp[( E − E ) / k T ] + 1
2π 2 h 3 0 F B
(2meff ) 3 / 2 ⎡ 2 2 ∞ ⎛ df 0 ⎞ 3 / 2 ⎤ (2meff ) ∞ ⎛ df 0 ⎞ 3 / 2
∞ 3/ 2
E 3/ 2
= ⎢ + ∫⎜ − ⎟ E dE ⎥ = ∫⎜− ⎟ E dE
2π 2 h 3 ⎢ 3 exp[( E − E F ) / k B T ] + 1 0 3 0 ⎝ dE ⎠ ⎥ 3π 2 h 3 0 ⎝ dE ⎠
⎣ ⎦
(25)
Similar relations are obtained along the principal axes for a crystal with elliptical iso-
energetic surfaces.
Alternatively, the density of electric current can be expressed as
j = −env = enμE (26)
where the mobility μ of charge carriers/electrons is introduced through v = − μE . The minus

sign indicates that the motion of electrons is opposite to the direction of the applied electric
field E. The relation between the electric conductivity and the mobility is thus
e〈τ ( E )〉
σ = neμ , i.e. μ= . (27)
meff
In anisotropic crystals the mobility, as the electric conductivity, is a tensor.

Solid State Physics PDF

Uploaded by

Copyright:

Available Formats

Solid State Physics PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Solid State Physics PDF

Uploaded by

Copyright:

Available Formats

SOLID STATE PHYSICS

(a) (b) (c)

(d) (e) (f)

Rmnp = ma1 + na 2 + pa 3 (1)

can be obtained as superpositions of integral multiples of three non-coplanar primitive

Tqrs ( Rmnp ) = Tqrs + Rmnp = Rq + m,r + n,s + p (2)

• Rotations by an angle 2π / n about a specific axis, denoted by C n , and its multiples,

C nj = (C n ) j . Geometric considerations impose that n = 1, 2, 3, 4 and 6, and that

• Inversion I, which is defined by the operation r → −r if applied around the origin.

• Improper rotation S n , which consists of the rotation C n followed by reflection in

(a) (b) (c) (d) (e)

• cubic, for which a = b = c, α = β = γ = 90°. It consists of three non-equivalent space-

• tetragonal, for which a = b ≠ c, α = β = γ = 90°. It encompasses the simple and body-

• orthorhombic, for which a ≠ b ≠ c, α = β = γ = 90°. It incorporates the simple, body-

• hexagonal, for which a = b ≠ c, α = β = 90°, γ = 120°. It

• trigonal, for which a = b = c, α = β = γ ≠ 90°. It contains a single C 3 axis.

• monoclinic, for which a ≠ b ≠ c, α = γ = 90°≠ β . It includes the simple and side-

a1 = (a / 2)( x + y − z ) , a 2 = (a / 2)(− x + y + z ) , a 3 = (a / 2)( x − y + z ) , (3)

while those for the right figure are

a1 = (a / 2)( x + y ) , a 2 = (a / 2)( y + z ) , a 3 = (a / 2)( z + x ) (4)

and the angles between these vectors are 60°.

Index system for lattice points, directions and planes

negative integer m, n or p is denoted by a − sign placed on top of it. For example, [[ mn p ]]

In three-dimensional lattices, the orientation of a crystal plane is determined by three

• Take the reciprocal of these numbers.

(001) (101) (111)

Simple crystal structures

Crystal a(Å) Crystal a (Å) Crystal a (Å)

Crystal a (Å) Crystal a (Å) Crystal a (Å)

The crystal structure of diamond (and also of Si and Ge semiconductors) is

Crystal a (Å) Crystal a (Å) Crystal a (Å)

Crystal a (Å) c/a Crystal a (Å) c/a

If the c/a ratio differs considerably from the ideal

Rmnp = ma1 + na 2 + pa 3 (1)

where r = ( x1 , x 2 , x3 ) is an arbitrary position vector with coordinates x1 , x 2 , and x3 in the

(generally non-orthogonal) system of coordinates determined by a1 , a 2 , and a 3 . This implies

ϕ ( x1 , x2 , x3 ) = ϕ ( x1 + ma1 , x 2 + na2 , x3 + pa3 ) (3)

or, for a function that can be expanded in a Fourier series

ϕ ( x1 , x 2 , x3 ) = ∑ ϕG exp[i (G1 x1 + G2 x 2 + G3 x3 )] (4)

it follows that, for any m, n, and p,

exp(imG1a1 ) = 1 , exp(inG2 a 2 ) = 1 , exp(ipG3 a3 ) = 1 . (5)

is a vector in a coordinate system defined by the vectors bi , i = 1,2,3, such that

Ω =| (a1 × a 2 ) ⋅ a 3 | , the vectors bi can be chosen as

b1 = (2π / Ω)(a 2 × a 3 ), b2 = (2π / Ω)(a 3 × a1 ), b3 = (2π / Ω)(a1 × a 2 ) . (9)

Ω rec =| b1 ⋅ (b2 × b3 ) |= ( 2π ) 3 / Ω . (10)

For a1 = d ( x − y ) , a 2 = d ( x + y ) , from (8) we get b1 = (π / d )( x − y) , b2 = (π / d )( x + y ) .

Ω = a 3 , from (10) it follows that b1 = (2π / a) x, b2 = (2π / a) y, b3 = (2π / a) z , i.e. the

Real space Reciprocal space

Observation: The reciprocal lattice of a reciprocal lattice is the direct lattice.

G mnp ⋅ Rhkl = 2π (mh + nk + pl ) , (11)

∫cell exp(iG ⋅ r )dV = Ωδ G , 0 (12)

ϕ G = Ω −1 ∫cell ϕ (r ) exp(−iG ⋅ r )dV . (14)

Relations between the direct and reciprocal lattices

G mnp = mb1 + nb2 + pb3 (15)

coordinate axes at a1 / m , a 2 / n , and a3 / p , respectively.

and satisfy, indeed, the relations

because of (8). Then, it follows that the normal to

The first Brillouin zone

equivalent to the replacement of G with − G , we obtain 2k ⋅ G + G 2 = 0 and finally,