Chapter 3 Solid State Physics

3.1

Goals

Understand the structure of carbon nanotubes. The physics of blue DVD lasers. Read: Z. I. Alferov, Rev. Mod. Phys. 73, 767 (2001). Nobel Lecture: The double heterostructure concept and its applications in physics, electronics, and technology.

3.2
3.2.1

Crystal Structure
Basic denitions

Crystal An innite, regular array of atoms. Unit cell A parallelpiped which, when translated, generates the whole crystal. Which one do we choose? Answer: rectangular (because higher symmetry). Translation vectors Points of identical environment in 3D crystals have position vectors given by r = r + ua + vb + wc, (3.1) where r and r give the locations of two such points, and a, b, c is a linearly independent set of translation vectors. 33

34
t t t  t

CHAPTER 3. SOLID STATE PHYSICS b 

Q  t    t

t T t

t Et

Et 

Figure 3.1: An example of a 2D crystal with possible unit cells (dash boxes). Lattice points The points given by r above are lattice points. Bravais lattice The set of points generated by the translation vectors is known as a Bravais lattice. Primitive lattice If all the lattice points are related through integer values of u, v, w, then a, b, c are primitive translation vectors and describe a primitive lattice.
r r d r r T r r Er

ar 
r d b

a
r

Figure 3.2: Primitive and nonprimitive lattices. Basis To each lattice point, attach a basis of atoms with relative coordinates r j : rj = xj a + yj b + zj c. Usual to choose 0 xj , yj , zj 1. Finally, Crystal structure lattice + basis = crystal structure (3.2)

3.2.2

2D Bravais lattices

There are 5 distinct ways of arranging lattice points in a plane. They are distinguished according to the symmetry operators which are present (Table 3.1). Note that rectangular

3.2. CRYSTAL STRUCTURE Table 3.1: 2D Bravais lattices. System oblique P rectangular P rectangular C square P hexagonal P Characteristic symmetry elements diads mirrors mirrors tetrad hexad

35

C is not primitive; a primitive Bravais lattice would be oblique. Thus, a Bravais lattice need not be a primitive lattice!

3.2.3

3D Bravais lattices

The possible 3D Bravais lattices are grouped into 6 dierent systems according to their symmetry (Table 3.2).

System Triclinic Monoclinic Orthorhombic Tetragonal Trigonal Hexagonal Cubic

Table 3.2: The 3D Bravais lattices. Characteristic Conventional symmetry unit cell none a=b=c = = = 900 1 diad a=b=c = = 900 , = 900 3 orthogonal a=b=c diads = = = 900 1 tetrad a=b=c = = = 900 1 triad a=b=c = = < 1200 1 hexad a=b=c = = 900 , = 1200 4 triads a=b=c = = = 900

Bravais lattices P P P, C, F, I P, I R P P, F, I

36 Cubic lattices

CHAPTER 3. SOLID STATE PHYSICS

There are 3 Bravais lattices for the cubic system: P (simple cubic), I (body centered), F (face centered) [Fig. 3.3].

SC

BCC
Figure 3.3: (a) SC, (b) BCC, (c) FCC.

FCC

Common structures Three common structures are given in Fig. 3.4. y

s T c1/2
0,1

y
s Ex s

s T

s c
0,1

s c

y
c
0,1

s T3/4c s

s c

Cl Cs

Na Cl
sc

s c0,1/2E x

1/4

1/4 3/4

c1/2 E x s

s, c

As Ga

Figure 3.4: (a) CsCl, (b) NaCl, (c) zincblende structures.

3.2.4

Closed-packed structures

Assume the atoms of a monatomic solid to be incompressible spheres. A compact arrangement on a plane is obtained via touching: An additional layer can be stacked in 2 ways: over
iiii iiii iiii

layer A

site B or C. A third layer can be C or A.

3.3. APPLICATION: STRUCTURE OF CARBON NANOTUBES If we repeat the pattern, have 2 types of fundamental stackings: (ABC)(ABC) and (AB)(AB)

37

The rst one corresponds to FCC and the second to hexagonal closed packed (HCP).

3.3

Application: Structure of Carbon Nanotubes

The structure of (single-walled) carbon nanotubes (CNTs) is depicted in Fig. 3.5. A CNT consists of a 2D graphene layer wrapped around itself. The unit cell is labeled using two indices, (m, n), according to the following coordinate representation of the chiral vector C h : Ch = ma1 + na2 . A CNT unit cell is constructed as follows: Draw the chiral vector (from one lattice point to another). Draw a perpendicular translation vector until it meets another lattice point. Complete the rectangle. (3.3)

3.4

Bonding in Solids

Origin: electrostatic. cohesion = attraction + repulsion Bonding type depends upon degree of electron wave function overlap and similarity/dissimilarity of atoms. Type van der Waals hydrogen metallic ionic covalent Attraction electric dipoles (uctuating) Coulomb electrons and nuclei unlike ions electrons and nuclei Repulsion Pauli exclusion principle Coulomb e-e and n-n like ions e-e and n-n

38

CHAPTER 3. SOLID STATE PHYSICS

y B' x B

q T

R O C h

a
1

a2

(10,0) (9,0) (8,0) (7,0) (6,0) (5,0) (4,0) (5,1) (4,1) (3,1) (2,1) (1,1) (2,2) (3,2) (3,3) (4,2) (4,3) (4,4) (5,2) (5,3) (5,4) (5,5) (6,1) (6,2) (6,3) (6,4) (7,1) (7,2) (7,3) (8,1) (8,2) (9,1)

zigzag

a 1
(2,0)

(3,0)

a
2 (0,0)

(1,0)

armchair

Figure 3.5: (a) Examples of CNT. (b) Unit cell. (c) Map.

3.4. BONDING IN SOLIDS

39

3.4.1

van der Waals solids

Examples: crystals of inert gases. ionization energy (eV) 24.58 14.00 12.13 nearest-neighbor distance () A 3.13 4.01 4.35 melting cohesive energy point (K) (eV/atom) 24 0.02 117 0.12 161 0.17

Ne Kr Xe

Neutral atoms = no net Coulomb interaction. Attraction Fluctuations in electron density uctuating electric dipole moments. If correlated attractive potential. Repulsion Overlap of electron clouds strong repulsion: due to Pauli exclusion principle.

3.4.2

Ionic crystals

Examples: alkali halides. nearest-neighbor distance () A 2.82 3.15 3.29 melting lattice energy point (K) (kcal/mol) 1074 182.6 1045 165.8 991 159.3

NaCl KCl RbCl

Attraction Attractive Coulomb potential between oppositely-charged ions. Repulsion Repulsive Coulomb force between similar charges.

40

CHAPTER 3. SOLID STATE PHYSICS

Figure 3.6: Electron density for ionic bond.

3.4.3

Covalent

Examples: H2 , Si, GaAs. nearest-neighbor distance () A C 3.57 Si 5.43 Ge 5.66 Attraction Coulomb. Repulsion Coulomb and wave function overlap. Cohesive energy Electronic wave function overlaps with nearest neighbors. Strength of bonding depends upon nearest-neighbor separation and bond angles. melting bond energy (0K) point (K) (eV/atom) 4100 7.37 1685 4.63 1210 3.85

3.4. BONDING IN SOLIDS

41

Figure 3.7: Electron density for covalent bond.

3.4.4

Metallic bond

Examples: Cu, Al, Ni,... (FCC structure) nearest-neighbor distance () A 2.86 2.49 2.56 2.88 melting bond energy (0K) point (K) (eV/atom) 933.5 3.39 1728 4.44 1358 3.49 1338 3.81

Al Ni Cu Au Cohesive energy

Delocalize outer electrons to form an electron gas. Heisenberg uncertainty principle: delocalization = loss of kinetic energy. Electron wave functions extend beyond rst few neighbors. What is important is not the orientation of the nearest neighbors but the packing density.

3.4.5

Hydrogen bond

Example: water. Bonding of hydrogen to a highly electronegative atoms leads to a positive center and a lone pair of electrons, respectively. The opposite charges on neighboring molecules then interact via the Coulomb interaction leading to bonding. Bonding energy 0.10.5 eV.

42

CHAPTER 3. SOLID STATE PHYSICS

Figure 3.8: Electronic wave functions in metals.

3.5
3.5.1

Electronic Band Structure

Dispersion relations

Classical A free particle has the dispersion relation E(p) = where p is the momentum (Fig. 3.9, left). p2 , 2m (3.4)

classical particle

quantum particle

electron in crystal

E(p)

E(k)

E(k)

Figure 3.9: Dispersion relations.

3.5. ELECTRONIC BAND STRUCTURE Quantum Since p = h/ = hk, a free particle has the dispersion relation E(k) = h2 k 2 , 2m

43

(3.5)

where k is the wave vector (Fig. 3.9, middle). If the particle is not free (e.g., an electron in a crystal), then E(k) h2 k 2 h2 k 2 +V , 2m 2m (3.6)

where m is called the eective mass (Fig. 3.9, right). Individual atoms have discrete energy levels. When two identical atoms approach each other, their levels interact and change in energy. Some move up and others move down. If more atoms are brought together, the splitting leads to quasi-continuous energy bands with gaps in between. The bands give the allowed energies of the electrons in the solid. Band formation for a 1D metallic gold
Free electron Empty Lattice NFE TB Atom

Figure 3.10: Band formation from atoms to solids. nanowire was recently reported in Science [7].

3.5.2

Insulators, semiconductors, and conductors

One can use the band picture in order to explain the qualitative dierence among insulators, semiconductors, and conductors. One can show that, if the highest occupied band is full of electrons, electrons cannot ow in and out and the crystal is an insulator. A semiconductor is very similar to an insulator in that, at very low temperatures, all the bands are completely full or empty. However, at nite temperatures, electrons in the highest occupied band can be thermally excited into the neighboring unoccupied band if the band gap is small enough. A material is typically considered an insulator if the band gap is greater than 23 eV. The crystal becomes conducting, albeit only slightly. A conductor consists of unlled bands which allows the ow of electrons and electric current. An example of the latter is sodium which has only one electron in the outermost 3s shell (can accept up to two electrons due to spin). When a solid is formed, the uppermost band will only be half-lled; sodium is thus a metal.

44

CHAPTER 3. SOLID STATE PHYSICS

3.5.3

Free-electron model

Compare rst your earlier study of, say, the hydrogen molecule. If one starts with an orbital on one hydrogen atom (e.g., 1s), then the H2 molecule will have 2 levels. Indeed, if one has N atoms, then there will be N energy levels. We now present a more formal analysis of the energy states of electrons in crystals. Any given electron has a wave function given by the Schrdinger equation. The potential felt by o the electron consists of attractive contributions due to the nuclei and negative ones due to the other electrons. In fact, the net result is an almost free electron; in 1-D, h 2 d 2 k = E k k . 2me dx2 The solutions are k (x) = eikx , Ek = h2 k 2 . 2me

Recall that, for the electron in a box, the boundary conditions led to constraints on k. Here, we again restrict the discussion to 1-D for clarity. We therefore assume that the crystal is a linear chain of regularly-spaced atoms (Fig. 3.11).

pp pp

a0

p pp p

Figure 3.11: A one-dimensional crystal lattice. The atomic separation is a0 and the length of the crystal is L. A bulk crystal is dened as having L = N a0 >> a0 , where N is the number of atoms; i.e., the eect of the end-points (surface) is then unimportant. To eliminate them completely (since we are only interested in bulk eects), we will propose periodic boundary conditions. This means that the crystal repeats itself again after length L (or after N atoms); in other words, the (N+1)th atom is identical to the 1st one. Note that this periodicity is in addition to the fundamental lattice periodicity. One consequence is that the electronic state should also repeat itself: k (x) = k (x + L). This implies eikL = 1 , k= 2 2 n= n , n Z. L N a0

3.5. ELECTRONIC BAND STRUCTURE

45

Hence, the wave number k and energy Ek are quantized. Since L is large, the discrete values of k are actually very close to each other and the quasi-continuous range of energies spanned is called a band. In a nite crystal with N atoms, each providing one state, there are in total N allowed states. Since the electrons are represented by travelling plane waves, k ranges from k = /a0 to k = /a0 . The latter are known as the Brillouin zone edges. Each type of electron leads to a band in k space (the Brillouin zone). In increasing energy, there is a whole series of bands. If the bands do not overlap, there exist energies for which no electron states exist, i.e., one has a band gap. The curve E=E(k) is also known as a dispersion relation or band structure. In a real crystal, the electrons are not exactly free, but behave as if they were with a dierent mass known as the eective mass. Realistic band structures are shown in Fig. 3.12.

Figure 3.12: Representative band structures: Si and GaAs.

3.5.4

Density of states

For a number of applications, it is important to know the density of states (DOS). In the above, we saw that the size of the rst Brillouin zone (FBZ) is 2/a0 and there were N k points each evenly separated by 2 , where L = N a0 is the size of the linear chain. L Thus, the density of k points (per unit length of the crystal) is (k) = 1 1 = . Lk 2 (3.7)

For electrons, each state labelled by k can accept 2 electrons (), according to the Pauli

46 exclusion principle. Then (in 3D) (k) = 2 (2)3

CHAPTER 3. SOLID STATE PHYSICS

per unit volume.

(3.8)

DOS in energy space? For free electron gas (in 3D), h2 k 2 E = E(k), 2m (E)dE = (k)d3 k = (k)4k 2 dk, 1 2m 3/2 1/2 1 2 dk E . k = 2 (E) = 2 dE 2 h2 This gives the density of electronic states at a given energy.

(3.9)

0D 3D

(E)

2D 1D

Figure 3.13: Density of states of (a) a free electron in d dimensions, (b) an electron in a metal. Since the DOS gives the number of electrons (in the present case, or other excitations such as phonons) in a certain energy range, it determines the response of the material. For example, the electronic specic heat capacity is proportional to the DOS at the Fermi energy. Example: Fermi energy of a metal In populating a band, one follows the same Aufbau principle as for atoms. In the ground state, the lowest energy states are lled rst until all the electrons are exhausted. The highest energy is then known as the Fermi level (true at T = 0 K). If we now assume that

3.6. ELECTRON STATISTICS

47

the dispersion relation is quadratic, then the total number of electrons in the band is given by N = V = V =
EF 0 EF 0

dE (E) dE 1 2 2
3/2 0

(3.10) 2m h2
EF 3/2

E 1/2 V 3 2 2m h2
3/2

V 2 2

2m h2

dE E 1/2 =

EF .

3/2

(3.11)

This gives a very useful formula for the Fermi energy in terms of the electron density n = N/V , h2 (3 2 n)2/3 . (3.12) EF = 2m Examples for a few metals are given in Table 3.3. Table 3.3: Carrier density, Fermi energy and Fermi temperature of metals. From Wolf[8]. n/(1028 m3 ) EF /eV TF /(104 K) Al 18.1 11.7 13.6 Ag 5.86 5.53 6.41 Au 5.90 5.55 6.43 Cu 8.47 7.06 8.19 Na 2.55 3.19 3.64 The Pauli exclusion principle has made energy of electron much higher than classical value of kB T 25 meV (at RT).

3.6

Electron Statistics

At a nite temperature, some states above the Fermi level can be occupied. The distribution above the Fermi level is often called an exponential tail and resembles the classical Boltzmann distribution, E (3.13) P (E) = e kB T , i.e., a high enough energy or temperature, the classical distribution is a good approximation. In thermodynamic equilibrium, the expectation of occupation of state of energy E is given by the Fermi-Dirac (FD) distribution: f (E) = 1 e(EEF ) +1 , 1/(kB T ). (3.14)

48

CHAPTER 3. SOLID STATE PHYSICS

Then, a generalization of the above zero temperature result is

n=

dE (E)f (E).

(3.15)

3.7

Impurity Doping

When impurities are introduced into an otherwise pure and perfect crystal lattice, these defects lead to localized atomic-like impurity states in the band gap of the crystal. Often they would have either an extra (donor) or one less electron (acceptor) than the host (e.g., P in Si or B in Si, respectively) and these can act as charge carriers. If the location of the impurity levels is close to the band edges, then the charge carriers can be thermally excited leading to impurity or extrinsic electrical conduction. Table 3.4: Electron concentrations (in m3 ). metal 1028 intrinsic semiconductor 1015 (Eg = 1eV, T = 300K) doping 1021

3.8

Optical Transitions in Solids

Just like for levels in atoms, electrons in bands can be moved from one band to another either by absorbing or emitting energy (e.g., in the form of light photons). Thus, in a lightemitting diode (LED), the wavelength of the emitted light is basically given by the band gap of the material. Such optical transitions are known as band-band transitions (Fig. 3.14) and are quasi-vertical in k space because of conservation of momentum. (On the scale of the Brillouin zone, the typical visible light photon has little momentum.)

3.9

Application: Band Structure of Semiconductor Heterostructures

Only a nite number of materials exist in nature, leading to a few choices of band gaps and/or appropriate materials. The process of manipulating existing materials in order to achieve dierent band gaps is known as band-gap engineering.

3.9. APPLICATION: BAND STRUCTURE OF SEMICONDUCTOR HETEROSTRUCTURES49

Eg

Figure 3.14: Band to band optical transition. An old technique consists of alloying two or more materials together. The resulting physical properties are often found to be closely given by a linear interpolation of the properties of the constituents. A newer technique developed in the 1970s consists in growing crystals atomic layer by atomic layer on a substrate. Such a growth technique is known as epitaxy. Three standard variations are: molecular beam epitaxy (MBE), metal-organo chemical vapor deposition (MOCVD), and liquid phase epitaxy (LPE). For example, instead of mixing two components as in alloying, one purposefully segregate them. The key here is that the layers can be made arbitrarily thin. The result is that one is in the domain of quantum mechanics. This can be seen by looking at the band prole in real space. One chooses materials such that their band gaps are dierent. In addition, the bands will be oset and the prole is that of a quantum well (QW, Fig. 3.15). An electron in the corresponding band will therefore not Al(1x) Gax As GaAs V (z)
T

Al(1x) Gax As

Figure 3.15: The band-edge prole of a three-layer quantum structure. behave freely in the growth direction, but rather would feel a conning potential. Thus, in

50

CHAPTER 3. SOLID STATE PHYSICS

the growth direction, quantized states are formed. Since the energies of the quantized states are dependent upon the well thickness, the energy separation between any two quantum levels can be engineered by changing the layer thickness. Modern day solid state LEDs, lasers, photodetectors, and transistors are made this way. This also has very fundamental implications. The electronic structure of an electron in a bulk semiconductor can often be described by a so-called eective-mass equation: h2 d2 2 (x) = E(x). 2m dx (3.16)

In a heterostructure, the eective mass and potential energy felt by an electron are both position dependent. Thus, a possible generalization of the above eective Schrdinger equation o is d2 h2 (x) + V (x)(x) = E(x). (3.17) 2m (x) dx2 Note that the latter is not hermitian. A formally correct generalization is then the following: 1 [m (x)] p[m (x)] p[m (x)] + [m (x)] p[m (x)] p[m (x)] (x) + V (x)(x) = E(x), 4 (3.18) with the constraint + + = 2. Current research indicates that = = 0, = 2.

3.10

Application: Blue DVD Lasers

The lasers in CDs and DVDs are quantum-well (QW) lasers.

3.10.1

Bulk vs. QW vs. QWi vs. QD lasers

Figure 3.16: Density of state vs. dimensionality of nanostructure. From A. P. Alivisatos, Science 271, 933 (1996).

3.10. APPLICATION: BLUE DVD LASERS

51

From Fig. 3.16, it can be seen that the DOS at the band edge improves as the dimensionality is reduced. This leads to enhance performance for light emission. Furthermore, quantum connement results in an increasing band gap, i.e., a decreasing laser wavelength. Because of the band gap, semiconductors can emit light when some of the valence states are empty and conduction states are occupied. It can be shown that such a population inversion can be maintained, e.g., by optical pumping. In practice, electrical excitation (carrier injection) is much more useful. The rst type of semiconductor laser was based on a pn junction. Since the junction acts as a diode, such lasers are also known as photodiode lasers. Under forward bias, electrons and holes ow towards the junction, where they recombine with the emission of light. When the light is spontaneously emitted, we have a light-emitting diode (LED). The light is emitted in every direction. Lasing can be achieved with increasing injection current (i.e., population inversion) the threshold current and the gain occurs in the plane of the junction. In both the LED and laser cases, the output is proportional to the injection current and the light wavelength is close to the band-gap energy. The feedback mirrors are typically the facets formed by cleaving the chip. In addition, semiconductors have a high refractive index ( 3 4) leading to a reectivity of 30 %. Better performant lasers are now made from heterostructures. The basic structure is an n-i-p multiple quantum well one (Fig. 3.17). Improved performance (lower threshold, higher gain, ...) is mainly due to the connement of both the carriers and photons due to the band osets. For instance, threshold measures how easily it is to generate a population inversion. In a semiconductor laser operating near the band gap, the number of states (more accurately, the density of states DOS) near k = 0 is an important parameter. The more states available, the lower the threshold. One can show that the DOS is higher for a quantum well than for a bulk semiconductor. Indeed, the DOS for a 3-D system is zero at k = 0 while the DOS for a 2-D system is nite.
p electrode p-GaN p-Al0.12Ga0.88N p-GaN p-Al0.2Ga0.8N InGaN MQW n-GaN n-Al0.12Ga0.88N n-In0.1Ga0.9N n-GaN GaN buffer (111) spinel substrate

75nm n electrode

Figure 3.17: InGaN semiconductor diode laser. From Nakamura et al. [9].

52

CHAPTER 3. SOLID STATE PHYSICS

3.11

Summary

There is a nite number of ways of arranging lattice points in space such that they are innitely repeating (the Bravais lattices). Almost all forces involved in bonding are electrostatic in origin and the main dierence in the various types of bonding is due to the degree of electron sharing or delocalization. Concepts in band theory: energy gaps, eective mass, density of states, Fermi energy. Application: structure of carbon nanotubes, blue lasers.