Introduction To Nanoscience and Nanotechnology: A Workbook: M. Kuno August 22, 2005

Introduction to Nanoscience and Nanotechnology:
A Workbook
M. Kuno
August 22, 2005
Contents
1 Introduction
2 Structure
11
3 Length scales
37
4 Excitons
45
5 Quantum mechanics review
57
6 Connement
75
7 Nondegenerate perturbation theory
107
8 Density of states
121
9 More density of states
129
10 Even more density of states
139
11 Joint density of states
151
12 Absorption
165
13 Interband transitions
173
14 Emission
195
15 Bands
213
16 K P Approximation
(Pronounced k dot p)
235
i
ii
CONTENTS
17 Tunneling
261
18 The WKB approximation
279
19 Synthesis
307
20 Tools
325
21 Applications
333
Acknowledgments
355
List of Figures
1.1
1.2
1.3
1.4
1.5
CdSe quantum dot . . . .

Quantum connement . .
Dimensionality . . . . . .
Size dependent absorption
Articial solid . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
and emission of
. . . . . . . . .
. . . .
. . . .
. . . .
CdSe
. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
6
7
8
9
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
2.11
2.12
2.13
2.14
2.15
14 3D Bravais lattices . . . . . . . .
Atoms per unit cell . . . . . . . . . .
Atom sharing . . . . . . . . . . . . .
FCC unit cell . . . . . . . . . . . . .
BCC unit cell . . . . . . . . . . . . .
Hexagonal unit cell . . . . . . . . . .
Diamond structure unit cell . . . . .
Zincblende or ZnS structure unit cell
NaCl structure unit cell . . . . . . .
CsCl structure unit cell . . . . . . .
Wurtzite unit cell . . . . . . . . . . .
Miller index examples . . . . . . . .
More Miller index examples . . . . .
Simple cubic . . . . . . . . . . . . .
FCC . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
14
15
16
16
17
18
19
20
21
22
23
24
28
30
4.1
4.2
Exciton types . . . . . . . . . . . . . . . . . . . . . . . . . . .
Exciton absorption . . . . . . . . . . . . . . . . . . . . . . . .
46
46
6.1
6.2
6.3
6.4
6.5
Particle in an innite box . . . .

Half a harmonic oscillator . . . .
Various 1D potentials . . . . . .
Particle in a nite box . . . . . .
Particle in a nite box: solutions
76
78
79
80
81
iii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
iv
LIST OF FIGURES
6.6
6.7
6.8
Particle in a nite well: Mathcad solutions . . . . . . . . . . .

Particle in an innite circle . . . . . . . . . . . . . . . . . . .
Particle in a sphere . . . . . . . . . . . . . . . . . . . . . . . .
8.1
8.2
8.3
8.4
3D
2D
1D
0D
9.1
9.2
3D density of CB and VB states . . . . . . . . . . . . . . . . 134

3D Fermi level . . . . . . . . . . . . . . . . . . . . . . . . . . 137
10.1
10.2
10.3
10.4
2D
2D
1D
1D
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
140
142
144
147
11.1
11.2
11.3
11.4
11.5
Vertical transitions . . . . . . . .
3D joint density of states . . . .
Summary, joint density of states
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
153
155
158
161
162
14.1
14.2
14.3
14.4
14.5
Einstein A and B coecients . . . . . . . . .

Derived emission spectrum: Einstein A and B
Pulsed experiment and lifetime . . . . . . . .
Radiative decay of excited state . . . . . . . .
Multiple pathway decay of excited state . . .
. . . . . . .
coecients
. . . . . . .
. . . . . . .
. . . . . . .
.
.
.
.
.
.
.
.
.
.
196
209
209
211
211
Kronig-Penney rectangular potential . . . . . . . . . . . . . .

Kronig-Penney delta function potential . . . . . . . . . . . . .
General Kronig Penney model: Mathcad solutions . . . . . .
General Kronig Penney model continued: Mathcad solutions
Kronig Penney model revisited: Mathcad solutions . . . . . .
Kronig Penney model, delta functions: Mathcad solutions . .
Kronig Penney model, delta functions continued: Mathcad
solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.8 From metals to insulators . . . . . . . . . . . . . . . . . . . .
213
224
227
228
229
230
15.1
15.2
15.3
15.4
15.5
15.6
15.7
density
density
density
density
density
density
density
density
of
of
of
of
of
of
of
of
states
states
states
states
CB
VB
CB
VB
.
.
.
.
.
.
.
.
states
states
states
states
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
85
87
92
123
125
126
127
231
232
17.1 Potential step . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

17.2 Potential step ( > V ) . . . . . . . . . . . . . . . . . . . . . . 262
LIST OF FIGURES
17.3
17.4
17.5
17.6
17.7
Potential step ( < V ) . .

Potential barrier . . . . .
Potential barrier ( > V ) .
Potential barrier ( < V ) .
Semiconductor junction .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
263
269
269
275
278
18.1
18.2
18.3
18.4
18.5
18.6
18.7
18.8
Arbitrary potential
Arbitrary potential
Arbitrary potential
Field emission . . .
Shottky barrier . .
Parabolic barrier .
Linear barrier . . .
Parabolic barrier .
step . .
drop . .
barrier .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
281
286
290
298
300
301
303
304
19.1
19.2
19.3
19.4
19.5
Cartoon of a MBE apparatus . .

Cartoon of a MOCVD apparatus
Colloidal synthesis apparatus . .
LaMer model . . . . . . . . . . .
LaMer model: size distribution .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
308
309
311
317
319
20.1
20.2
20.3
20.4
20.5
20.6
Transmission electron microscopy

Secondary electron microscopy .
Atomic force microscopy . . . . .
Scanning tunneling microscopy .
Dip pen nanolithography . . . . .
Microcontact printing . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
326
327
328
329
330
331
21.1
21.2
21.3
21.4
21.5
21.6
21.7
21.8
21.9
Nanowire device . . . . . . . . . . . . . . . . . . . .
Nanowire sensor . . . . . . . . . . . . . . . . . . . .
Quantum dot/dye photobleaching . . . . . . . . . . .
Quantum dot/dye absorption/emission spectra . . .
Density of states for lasing . . . . . . . . . . . . . . .
Solar spectrum and QD absorption/emission spectra
Quantum dot LED schematic . . . . . . . . . . . . .
Orthodox model of single electron tunneling . . . . .
Coulomb Staircase . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
334
335
337
338
340
342
344
346
350
vi
LIST OF FIGURES
List of Tables
2.1
2.2
2.3
2.4
2.5
Common metals . . . . . . .
Group IV semiconductors . .
Group III-V semiconductors .
Group II-VI semiconductors .
Group IV-VI semiconductors
vii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
25
26
26
27
viii
LIST OF TABLES
Preface
This set of lecture notes about nanoscience and nanotechnology was initially
written over the spring and summer of 2003. After my initial appointment
as an assistant professor in chemistry, I agreed to teach an introductory
class on nanoscience and nanotechnology for incoming graduate students
and upper level undergraduates at the University of Notre Dame. However
after accepting this task, it quickly became apparent to me that there were
few resources available for teaching such a class, let alone any textbook. So
while waiting for equipment to arrive, I undertook it upon myself to compile
a series of lecture notes that would explain to the student some of the underlying concepts behind nano. The motivation for this was to describe
to the student the physics behind each concept or assumption commonly
encountered in the nano literature rather than just providing a qualitative
overview of developements in the eld. I have also tried to illustrate and
motivate these concepts with instances in the current literature where such
concepts are applied or have been assumed. In this manner, the goal is to
provide the student with a foundation by which they can critically analyze,
and possibly in the future, contribute to the emerging nano eld. It is also
my hope that one day, these lecture notes can be converted into an introductory text so that others may benet as well.
Masaru Kenneth Kuno

Notre Dame, IN
August 24, 2003
ix
LIST OF TABLES
Revision History
Version 1.0 complete 8/03
Corrected for typos and such as well as added the Kane chapter 6/04
Added Absorption section 7/05
Added Exciton section 7/05
Added Quantum mechanics review section 7/05
Added Non-degenerate perturbation theory section 7/05
Added Time dependent perturbation theory section 7/05
xi
xii
LIST OF TABLES
Chapter 1
Introduction
Preliminaries
What is nano? Well, without providing a denite answer to this question,
nano is a popular (emerging) area of science and technology today. It has
attracted the attention of researchers from all walks of life, from physics to
chemistry to biology and engineering. Further impetus for this movement
comes from the temendous increase in public and private funding for nano
over the last ten years. A prime example of this is the new National Nanotechnology Initiative (NNI) created by former President Bill Clinton. The
NNI increases funding in national nanoscience and nanotechnology research
by hundreds of millions of dollars yearly. In addition, private sector contributions have jumped dramatically as evidenced by the plethora of small
startup rms lining the tech corridors of the east and west.
Nano has even entered popular culture. Its used as a buzzword in contemporary books, movies and television commercials. For example, in the
recent blockbuster, Spiderman, the Willem Dafoe character (the Green Goblin) is a famous (and wildly wealthy) nanotechnologist whose papers the
Tobey McGuire character (Spiderman) has read and followed (see the scene
outside of Columbia university). Likewise, in the movie Minority Report
Tom Cruises character undergoes eye surgery to avoid biometric ngerprinting. This scene involves a retinal eye transplant aided by so called nano
reconstructors. A scene in the DC metro shows him reading a newspaper
with the headline nanotechnology breakthrough. In television, a current
GE commercial for washers and dryers features the storyline of: geeky nanotechnologist bumps into a supermodel at the laundromat resulting in love
at rst sight. Were therefore, implicitly, told to believe that their mix of
1
CHAPTER 1. INTRODUCTION
brains and beauty embody GEs new washer/dryer line. In books, the New
York Times bestseller Prey by Michael Crighton features nanotechnology
run amok with spawns of tiny nano bots escaping from the laboratory and
hunting down people for food. (Sounds like the Andromeda Strain except
recast with nano as opposed to an alien virus.).
The mantle of nano has also been adopted by various scientic visionaries. Perhaps the most prominent is Eric Drexler who has founded an institute, called the Foresight Institute, devoted to exploring his ideas. Concepts
being discussed include developing tiny nano robots that will live inside us
and repair our blood vessels when damaged, preventing heart attacks. They
will also kill cancer, cure us when we are sick, mend bones when broken,
make us smarter and even give us immortality. These nano robots will also
serve as tiny factories, manufacturing anything and everything from food
to antibiotics to energy. In turn, nanotechnology will provide a solution to
all of mankinds problems whether it be hunger in developing countries or
pollution in developed ones. Drexler therefore envisions a vast industrial
revolution of unprecendented size and scale. At the same time, concurrent
with his visions of a utopian future is a darker side, involving themes where
such nano robots escape from the laboratory and evolve into sentient beings
completely out of mankinds control. Such beings could then sow the seeds
to mankinds own destruction in the spirit of recent movies and books such
as The Terminator, The Matrix and Prey. Now, whether such predictions
and visions of the future will ever become reality remains to be seen. However, any such developments will ultimately rely on the scientic research of
today, which is, on a daily basis, laying down the foundation for tomorrows
nanoscience and nanotechnology.
In todays scientic realm, the word nano describes physical lengthscales
that are on the order of a billionth of a meter long. Nanoscale materials
therefore lie in a physical size regime between bulk, macroscale, materials (the realm of condensed matter physics) and molecular compounds (the
realm of traditional chemistry). This mesoscopic size regime has previously
been unexplored and beckons the researcher with images of a scientic wild
wild west with opportunites abound for those willing to pack their wagons
and head into the scientic and technological hinterland. In this respect,
nanoscale physics, chemistry, biology and engineering asks basic, yet unanswered, questions such as how the optical and electrical properties of a given
material evolve from those of individual atoms or molecules to those of the
parent bulk. Other questions that nanoscience asks include:
How does one make a nanometer sized object?
3
How do you make many (identical) nanometer sized objects?
How do the optical and electrical properties of this nanoscale object
change with size?
How does its optical and electrical properties change with its dimensionality?
How do charges behave in nanoscale objects?
How does charge transport occur in these materials?
Do these nanoscale materials posess new and previously undiscovered
properties?
Are they useful?
The transition to nanoscience begins at this last point when we ask how these
nanoscale materials might be exploited to improve our lives. Venture capital
rms and mainstream industry have therefore taken up this challenge with
many small startups trying to apply nanoscale materials in products ranging from better sunscreen lotions to uorescent labels for biological imaging
applications to next generation transistors that will one day store the entire
content of the Library of Congress on the head of a pin. More established
companies, such as GE, HP, Lucent and IBM, have also started their own
in house nano programs to revolutionalize consumer lighting, personal computing, data storage and so forth. So whether it be for household lighting
or consumer electronics, a nano solution exists and there is very likely a
company or person pursuing this vision of a nano future.
So what is nano? This series of lecture notes tries to answer this question by explaining the physical concepts behind why such small, nanoscale,
materials are so interesting and potentially useful.
Overview
The idea behind these lecture notes is as follows: First in Chapter 2, the
composition of solids is discussed to introduce common crystal structures
found in nanomaterials. Solids come in a number of forms, from amorphous (glass-like) to polycrystalline (multiple domains) to crystalline. Much
of nanoscience and nanotechnology focuses on nanometer sized crystalline
solids, hence the emphasis on crystal structure. In addition, the structure
section also illustrates the increase in surface to volume ratio for nanomaterials over bulk. This is because in nanometer sized systems up to 50% of
the atoms lie at the surface of a nanostructure, in direct contrast to macroscopic solids where such numbers are typically much smaller. The surface is
therefore potentially important in dictating a materials optical and electrical properties when nanometer sized. Furthermore, the increase in surface
area is important to applications where the surface to volume ratio plays a
critical role such as in catalysis as well as in photovoltaics. Developments in
this area, using nanostructures, have led to increasingly ecient solar cells
such as the Gratzel cell. Finally, the concept of crystal structure and the
periodic potential due to the ordered arrangement of atoms is central to the
concept of electronic bands, which we will discuss later on.
Figure 1.1: Transmission electron micrograph of individual CdSe quantum

dots
Chapter 3 introduces the concept of length scales to put into perspective
the actual physical lengths relevant to nano. Although being nanometer
sized is often considered the essence of nano, the relevant physical length
scales are actually relative to the natural electron or hole length scales in the
parent bulk material. These natural length scales can either be referred to
by their deBroglie wavelength or by the exciton Bohr radius. Thus, while a
given nanometer sized object of one material may qualify for nano, a similar
sized object of another material may not.
Next the concept of quantum connement is introduced in Chapter 6
through the simple quantum mechanical analogy of a particle in a 1 di-
5
mensional, 2 dimensional and 3 dimensional box. Quantum connement is
most commonly associated with nano in the sense that bulk materials generally exhibit continuous absorption and electronic spectra. However, upon
reaching a physical length scale equivalent to or less than either the exciton
Bohr radius or deBroglie wavelength both the optical and electronic spectra become discrete and more atomic-like. In the extreme case of quantum
dots, connement occurs along all three physical dimensions, x,y,and z such
that the optical and electrical spectra become truly atomic-like. This is one
reason why quantum dots or nanocrystals are often called articial atoms.
Analogies comparing the particle in a one dimensional box to a quantum
well, the particle in a two dimensional box to a quantum wire and the particle
in a three dimensional box to a quantum dot provide only half the solution. If
one considers that in a quantum well only one dimension is conned and that
two others are free, there are electronic states associated with these extra
two degrees of freedom. Likewise in the case of a quantum wire, with two
degrees of connement, there exists one degree of freedom. So solving the
particle in a two dimensional box problem models the electronic states along
the two conned directions but does not address states associated with this
remaining degree of freedom. To gain better insight into these additional
states we introduce the concept of density of states (DOS) in Chapters
8,9,and 10. The density of states argument is subsequently applied to both
the valence band and conduction band of a material. Putting together both
valence and conduction band density of states introduces the concept of the
joint density of states (JDOS) in Chapter 11 which, in turn, is related to
the absorption coecient of a material.
After describing the absorption spectra of 3D (bulk), 2D (quantum well),
1D (quantum wire), and 0D (quantum dot) systems we turn to the concept
of photoluminescence. Generally speaking, in addition to absorbing light,
systems will also emit light of certain frequencies. To describe this process,
the Einstein A and B coecients and their relationships are introduced and
derived in Chapter 14. Finally, the emission spectrum of a bulk 3D material
is calculated using the derived Einstein A and B coecients. The concept
of quantum yields and lifetimes, which describe the eciency and timescale
of the emission, completes this section.
Bands are introduced in Chapter 15. This topic is important because
metals, semiconductors, and semi-metals all have bands due to the periodic
potential experienced by the electron in a crystal. As mentioned earlier in
the section on structure, this periodic potential occurs due to the ordered
and repeated arrangement of atoms in a crystal. Furthermore, metals, semiconductors, insulators, and semi-metals can all be distinguished through
Figure 1.2: Photograph of the size dependent emission spectra of both HgS
(top) and CdSe (bottom) quantum dots. Small quantum dots absorb and
emit blue/green light, larger dots absorb and emit red light.
Figure 1.3: Cartoon of connement along 1, 2 and 3 dimensions. Analogous

to a quantum well, quantum wire and quantum dot.
the occupation of these bands by electrons. Metals have full conduction

bands while semiconductors and insulators have empty conduction bands.
At the same time, in the case of semiconductors and insulators there is a
range of energies that cannot be populated by carriers separating the valence band from the conduction band. This forbidden range of energies (a
no mans land for electrons) is referred to as the band gap. The band gap is
extremely important for optoelectronic applications of semiconductors. For
example, the band gap will generally determine what colors of light a given
semiconductor material will absorb or emit and, in turn, will determine their
usefulness in applications such as solar energy conversion, photodetectors,
or lasing applications. An exploration of the band gap concept ultimately
touches on the eects quantum connement has on the optical and electrical
properties of a material, which leads to the realization of a size dependent
band gap.
Introducing the concept of bands is also important for another reason
since researchers have envisioned that ordered arrays of quantum wells, or
wires, or dots, much like the arrangement of atoms in a crystal, can ultimately lead to new articial solids with articial bands and corresponding
band gaps. These bands and associated gaps are formed from the delocalization of carriers in this new periodic potential dened by the ordered
arrangement of quantum dots, quantum wires, or quantum wells. Talk about
designer materials. Imagine articial elements or even articial metals, semimetals, and semiconductors. No wonder visionaries such as Drexler envision
Figure 1.4: Size dependent absorption and emission spectra of colloidal CdSe
quantum dots.
9
so much potential in nano. In the case of quantum wells, stacks of closely
spaced wells have been grown leading to actual systems containing these
minibands. In the case of quantum dots and wires, minibands have not
been realized yet but this is not for lack of trying. The concept of creating articial solids with tailor made bands from articial atoms has been
tantalizing to many.
Figure 1.5: TEM micrograph of an array of colloidal CdSe quantum dots

ordered into an articial crystal. Each dark spot is an individual quantum
dot. The white space in between dots is the organic ligands passivating the
surface of the particle.
Not only do semiconductors, metals, insulators, and semi-metals absorb
and emit light, but they also have electrical properties as well. These properties are greatly aected by quantum connement and the discreteness of
states just as with the aforementioned optical properties. Transport properties of these systems take center stage in the realm of devices where one
desires to apply quantum dots, quantum wells, and quantum wires within
opto-electronic devices such as single electron transistors. To this end, we
introduce the concept of tunneling in Chapter 17 to motivate carrier trans-
10
port in nanometer-sized materials. Tunneling is a quantum mechanical eect

where carriers can have non-zero probability of being located in energetically
fobidden regions of a system. This becomes important when one considers
that the discreteness of states in conned systems may mean that there are
substantial barriers for carrier transport along certain physical directions
of the material.
The WKB approximation is subsequently introduced in Chapter 18 to
provide an approximate solution to Schrodingers equation in situations
where the potential or potential barrier for the carrier varies slowly. In
this fashion, one can repeat the same tunneling calculations as in the previous section in a faster, more general, fashion. A common result for the form
of the tunneling probablity through an arbitrary barrier is derived. This
expression is commonly seen and assumed in much of the nano literature
especially in scanning tunneling microscopy, as well as in another emerging
eld called molecular electronics.
After providing a gross overview of optical, electrical and transport properties of nanostructures, we turn to three topics that begin the transition
from nanoscience to nanotechnology. Chapter 19 describes current methods
for making nanoscale materials. Techniques such as molecular beam epitaxy (MBE), metal-organic chemical vapor deposition (MOCVD) and colloidal sysnthesis are described. A brief overview of each technique is given.
Special emphasis is placed on better understanding colloidal growth models
currently being used by chemists to make better, more uniform, quantum
dots and nanorods. To this end, the classical LaMer and Dinegar growth
model is derived and explained. Relations to the behavior of an ensemble
size distribution are then discussed completing the section.
Once created, tools are needed to study as well as manipulate nanoscale
objects. Chapter 20 describes some of the classical techniques used to characterize nanostructures such as transmission electron microscopy (TEM),
secondary electron microscopy (SEM), atomic force microscopy (AFM) and
scanning tunneling microscopy (STM). Newer techniques now coming into
prominence are also discussed at the end of the section. In particular, the
concepts behind dip-pen nanolithography and microcontact printing are illustrated.
Finally, Chapter 21 discusses applications of quantum dots, quantum
wires, and quantum wells using examples from the current literature. Special
emphasis is placed on the Coulomb blockade and Coulomb staircase problem,
which is the basis of potential single electron transistors to be used in next
generation electronics.
Chapter 2
Structure
Crystal structure of common materials
This section is not meant to be comprehensive. The interested reader may
consult a number of excellent introductory references such as Kittels introduction to solid state physics. The goal, however, is to illustrate the
crystal structure of common materials often encountered in the nano literature. Solids generally appear in three forms, amorphous (no long range
order, glass-like), polycrystalline (multiple domains) or crystalline (a single
extended domain with long range order). Since nano typically concerns itself
with crystalline metal nanoparticles and semiconductor nanocrystals, wires,
and wells, having a basic picture of how the elements arrange themselves in
these nanocrystalline systems is important. In this respect, crystal structure comes into play in many aspects of research, from a materials electronic
spectra to its density and even to its powder x-ray diraction pattern.
Atoms in a crystal are generally pictured as being arranged on an imaginary lattice. Individual atoms (or groups of atoms) are hung o of the
lattice, much like Christmas ornaments. These individual (or groups of)
atoms are referred to as the basis of the lattice. The endless repetition of
basis atom(s) on a lattice makes up the crystal. In the simplest case, the
basis consists of only a single atom and each atom is located directly over
a lattice point. However, it is also very common to see a basis consisting
of multiple atoms, which is the case when one deals with binary or even
ternary semiconductors. Here the basis atoms do not necessarily sit at the
same position as a lattice point, potentially causing some confusion when
you rst look at the crystal structures of these materials.
There are 14 three dimensional Bravais lattices shown in Figure 2.1.
11
12
CHAPTER 2. STRUCTURE
These are also referred to as conventional unit cells (i.e. used in everyday
life) as opposed to the primitive unit cell of which only the simple cubic (aka
primitive) lattice qualies. That is, most of these unit cells are not the
simplest repeating units of an extended lattice; one can nd even simpler
repeating units by looking harder. Rather, these conventional cells happen
to be easy to visualize and interpret and hence are the ones most commonly
used.
Figure 2.1: 14 3-dimensional Bravais lattices. From Ibach and Luth.
It is often useful to know the number of atoms in a unit cell. A general

counting scheme for the number of atoms per unit cell follows.
atoms entirely inside the unit cell: worth 1
corner atoms: worth
1
8
13
face atoms: worth
edge atoms: worth
1
2
1
4
Figure 2.2 illustrates the positions of an atom in the corner, face, edge and
interior positions. The weighting scheme follows because each of these unit
cells is surrounded by other unit cells, hence all atoms with the exception of
interior ones are shared. The corner atoms are shared by 8 unit cells while
the face atoms are shared by 2 cells and nally the edge atoms by four. Any
atom in the interior of a unit cell is exclusive to that cell. This is illustrated
in Figure 2.3. By counting atoms in this fashion, one can determine the
number of atoms per unit cell.
Single element crystals
In the case of metals, the cubic lattices are important, with particular emphasis on the face centered cubic (FCC) and body centered cubic (BCC)
structures. Both FCC and BCC structures have a single atom basis; thus,
they strongly resemble the Bravais lattices or conventional unit cells seen in
the previous diagram. The number of atoms per unit cell in the FCC case
is 4 (8 corner atoms and 6 face atoms). Likewise, the number of atoms per
BCC unit cell, using the above counting scheme, is 2 (1 interior atom and
8 corner atoms). Note that an alternative name exists for the FCC unit
cell: cubic close packed (CCP), which should be remembered when reading
the literature. Both unit cells are shown in Figures 2.15 and 2.5. Typical
elements that crystallize in the FCC structure include: Cu, Ag, Au, Ni, Pd,
Pt, and Al. Typical elements that crystallize in the BCC structure include:
Fe, Cr, V, Nb, Ta, W and Mo. More complete tables can be found in Kittel.
Analogous to the FCC lattice is the hexagonal close packed (HCP) structure. A simple way to dierentiate the two is the atomic packing order,
which follows ABCABC in the case of FCC and ABABA in the case of HCP
in the 001 and 0001 directions respectively. The letters A, B, and C
etc. . . represent dierent atom planes. The HCP structure has a conventional unit cell shown in Figure 2.6. It contains 2 atoms per unit cell (8 on
the corners and 1 inside).
Another conventional unit cell that is often encountered is called the
diamond structure. The diamond structure diers from its FCC and BCC
counterparts because it has a multi atom basis. Therefore, it does not
immediately resemble any of the 14 Bravais lattices. In fact, the diamond
structure has a FCC lattice with a two atom basis. The rst atom is at
1
1
(0, 0, 0) and the second atom is at 4 , 1 , 4 . It is adopted by elements that
4
14
Figure 2.2: Number of atoms per unit cell. Counting scheme.
15
Figure 2.3: Cartoon showing sharing of atoms by multiple unit cells.
16
Figure 2.4: FCC unit cell
Figure 2.5: BCC unit cell
17
Figure 2.6: Hexagonal unit cell.
have a tendency to form strong covalent bonds, resulting in tetrahedral

bonding arrangements (Figure 2.7). The number of atoms per unit cell
in this case is 8 (8 corner atoms, 4 interior atoms, 6 face atoms). Some
common elements that crystallize in the diamond structure include: C, Si,
Ge and Sn. As a side note, one way to visualize the diamond unit cell is
to picture two interpenetrating FCC lattices, oset from each other by a
1 1
( 1 , 4 , 4 ) displacement.
4
Summary, diamond
lattice:FCC
basis atom 1: (0, 0, 0)
basis atom 2:
1 1 1
4, 4, 4
Compound crystals
In the case of binary compounds, such as III-V and II-V semiconductors,
things get a little more complicated. One doesnt have the benet of conventional unit cells that resemble any of the 14 standard Bravais lattices.
18
Figure 2.7: Diamond structure unit cell
Instead these conventional unit cells often have names such as the NaCl
structure or the ZnS structure and so forth. This is because, unlike simple
FCC or BCC metals, we no longer have a single atom basis, but rather a
basis consisting of multiple atoms as well as a basis made up of dierent
elements.
Common crystal lattices for semiconductors include the ZnS, NaCl
and CsCl lattices.
The ZnS, also called zinc blende (ZB) or sphalerite, structure can be
1 1
visualized as two interpenetrating FCC lattices oset by ( 1 , 4 , 4 ) in Figure
4
2.8. It is identical to the diamond structure we saw in the case of single
element crystals. The only real dierence is that now we have two elements
making up the atom basis of the unit cell. The underlying lattice is FCC.
The two atom, two element basis has positions of element1 (0, 0, 0) and
1 1
element2 1 , 4 , 4 . Using the above counting scheme we nd that there are
4
8 atoms per unit cell. This is further subdivided into 4 atoms of element 1,
and 4 atoms of element 2. You will notice in the gure that the 4 atoms of
one element are completely inside the unit cell and that the atoms of the
other element are arranged as 8 corner and 6 face atoms.
19
Summary, ZnS
lattice: FCC
basis, element 1 (0, 0, 0)
basis, element 2
1 1 1
4, 4, 4
Figure 2.8: Zincblende or ZnS structure unit cell.

The NaCl structure (also called rocksalt structure) can be visualized
1
as 2 interpenetrating FCC lattices oset by ( 2 , 0, 0) in Figure 2.9. The
underlying lattice is FCC. The two element, two atom basis has locations
1 1
of element1 (0, 0, 0) and element2 1 , 2 , 2 . It has 8 atoms per unit cell.
2
This is broken up into 4 atoms from element 1 and 4 atoms from element 2.
One can see in the gure that for element 1 there are 8 corner atoms and
6 face atoms. For element 2 there are 12 edge atoms and 1 interior atom.
Examples of materials that crystallize in this structure include PbS, PbSe,
and PbTe.
Summary, NaCl
lattice: FCC
20
basis, element 2
1 1 1
2, 2, 2
Figure 2.9: NaCl structure unit cell.

The CsCl structure is the compound material version of the single element BCC unit cell. It is shown in Figure 2.10 where one can see that
there are two elements present with one of them being the center atom. The
atoms from the other element take up corner positions in the unit cell. The
underlying lattice is simple cubic and the two atom, two element basis is
1 1
element 1 at (0, 0, 0) and element 2 at 1 , 2 , 2 . The CsCl has two atoms
2
per unit cell, 1 from each element. Examples of materials that crystallize in
this structure include, CsCl and CuZn.
Summary, CsCl
lattice: SC
basis, element 2
1 1 1
2, 2, 2
The wurtzite crystal structure is the compound material version of the

single element HCP structure. It has a multi atom basis. The underlying
lattice is hexagonal. The two element, two atom basis is element 1 (0, 0, 0)
1 1
and element 2 2 , 3 , 2 . The unit cell is shown in Figure 2.11 and contains
2
4 atoms per unit cell, 2 atoms from element 1 and 2 atoms from element
2. Examples of materials that crystallize in this structure include CdSe and
GaN.
21
Figure 2.10: CsCl structure unit cell.
Summary, wurtzite
lattice: hexagonal
basis, element 2
2 1 1
3, 3, 2
Miller indices
Sometimes you will see the orientation of a crystal plane described by (001)
and so forth. These numbers are referred to as Miller indices. They are
generated using some simple rules described below.
Take the desired plane and see where it intersects each x, y, z axis
in multiples of the lattice constant. For the case of cubic lattices the
lattice constant, a, is the same in all x, y, and z directions.
Next take the reciprocal of each intersection point and reduce the three
values to their lowest integer values. (i.e. divide out any common
integer)
Express the plane through these integers in parentheses as (abc)
22
Figure 2.11: Wurtzite unit cell.
Should the plane not intersect an axis, say the z axis, just write a 0.
For example (ab0)
If the intercept is in the negative side of an axis, say the y axis, just
put a bar over the number, for example (a
bc).
Miller notation
(hkl) =crystal plane
{hkl} =equivalent planes
[hkl] =crystal direction
< hkl >=equivalent directions
Examples are illustrated in Figures 2.12 and 2.13.
23
Figure 2.12: Examples of using Miller indices.
24
Figure 2.13: More examples of using Miller indices.
25
Quick tables
Short tables of common metals and semiconductors are provided below with
their standard crystal structure.
Common Metals
Table 2.1: Common metals

I
II
Cu
Ag
Au
Zn
Cd
Hg
III
B
Al
Ga
In
Tl
IV
C
Si
Ge
Sn
Pb
V
N
P
As
Sb
Bi
VI
S
Se
Te
Po
Ag=FCC [cubic] (alternatively called cubic closest packed)

Au=FCC [cubic] (alternatively called cubic closest packed)
Common Semiconductors
Group IV
Table 2.2: Group IV semiconductors
I
II
Cu
Ag
Au
Zn
Cd
Hg
III
B
Al
Ga
In
Tl
IV
C
Si
Ge
Sn
Pb
V
N
P
As
Sb
Bi
VI
S
Se
Te
Po
Si=diamond structure
Ge=diamond structure
III-V
GaN=ZB [cubic] (alternatively called ZnS structure)
26
Table 2.3: Group III-V semiconductors

I
II
Cu
Ag
Au
Zn
Cd
Hg
III
B
Al
Ga
In
Tl
IV
C
Si
Ge
Sn
Pb
V
N
P
As
Sb
Bi
VI
S
Se
Te
Po
GaAs=ZB [cubic] (alternatively called ZnS structure)

InP=ZB [cubic] (alternatively called ZnS structure)
InAs=ZB [cubic] (alternatively called ZnS structure)
II-VI
Table 2.4: Group II-VI semiconductors
I
II
Cu
Ag
Au
Zn
Cd
Hg
III
B
Al
Ga
In
Tl
IV
C
Si
Ge
Sn
Pb
V
N
P
As
Sb
Bi
VI
S
Se
Te
Po
ZnS=ZB [cubic]
ZnSe=ZB [cubic]
CdS=ZB [cubic]
CdSe=wurtzite [hexagonal]
CdTe=ZB [cubic]
IV-VI
PbS=NaCl structure (alternatively called rocksalt)
PbSe=NaCl structure (alternatively called rocksalt)
PbTe=NaCl structure (alternatively called rocksalt)
27
Table 2.5: Group IV-VI semiconductors

I
II
Cu
Ag
Au
Zn
Cd
Hg
III
B
Al
Ga
In
Tl
IV
C
Si
Ge
Sn
Pb
V
N
P
As
Sb
Bi
VI
S
Se
Te
Po
Packing fraction
Supposing atoms to be rigid spheres, what fraction of space is lled by atoms
in the primitive cubic and fcc lattices. Assume that the spheres touch.
Simple cubic
Determine the packing fraction of the simple cubic lattice.
Vsphere =
where
4 3
r
3
r=a
2
a
4
3
2
3
a
6
Vsphere =
=
Also there is only 1 sphere in the cube. Recall that each corner sphere is
shared by 8 other neighboring cubes.
Next the volume of the cube is
Vcube = a3
so that the ratio of volumes is
Ratio =
=
a3
6
a3
The desired simple cubic packing fraction is

Ratiosc = 0.524
(2.1)
28
Figure 2.14: simple cubic.
29
FCC
Determine the packing fraction of the fcc lattice
c2 = a2 + a2
= 2a2
2a
c =
= 4r
This results in
2
r=
a
4
The volume of a single sphere is
Vsphere =
=
=
4 3
r
3
2
4
3
4
2 3
a
24
a3
Next we calculate how many spheres are in the cube. In the fcc lattice there
are 8 corner spheres and 6 face spheres. The calculation is
Corner: 8
Face: 6
1
2
1
8
=1
=3
The total number of spheres in the cube is therefore 4. The total volume of
the spheres in the cube is therefore
2
24
Vspheretotal = 4Vsphere = 4
=
The volume of the cube is simply
Vcube = a3
2 3
a
6
30
The desired ratio of volumes is then
Ratio =
2a3
6
a3
2
=
6
= 0.741
The packing fraction of the fcc lattice is
Ratiof cc = 0.741
(2.2)
Figure 2.15: FCC.
Surface to volume ratio

Here we show that the surface to volume ratio of a number of systems scales
as an inverse power law. Smaller sizes translate to larger surface to volume
ratios.
31
Cube
Surface area
S = 6a2
where a is the radius.
Volume
V = a3
R=
S
V
=
=
6a2
a3
6
a
Sphere
Surface area
S = 4a2
Volume of sphere
V =
4 3
a
3

R=
S
V
=
=
Cylinder
Surface area of cylinder
S = 2al
Volume of cylinder
V = a2 l
4a2
4
3
3 a
3
a
32
R=
S
V
=
=
2al
a2 l
2
a
(2.3)
In all cases, the surface to volume ratio follows an inverse power law.
But the bottom line is that as the size of the system decreases, the fraction
of atoms on the surface will increase. This is both good and bad. In one
case illustrated below it can lead to more ecient catalysts. However in
other cases, such as in the optical microscopy of nanostructures it can be
a bad thing because of defects at the surface which quench the emission of
the sample.
Why, surface to volume ratio?

Since surfaces are useful for things such as catalysis, there is often an incentive to make nanostructured materils in the hopes that the increase in
surface to volume ratio will lead to more catalytic activity. To illustrate
consider the following example. Start with a 1 cm radius sphere. Calculate
the surface area. Next, consider a collection of 1nm radius spheres that will
have the same total volume as the original sphere. What is the total surface
area of these little spheres. How does it compare?
The surface of the large sphere is
Slarge = 4a2
= 4cm2
The volume of the large sphere is
Vlarge =
=
4 3
a
3
4
cm3
3
Next, how many little spheres do I need to make up the same volume.
Vlittle =
=
4
(107 )3
3
4
1021
3
33
The ratio of the large to small sphere volumes is
Vlarge
Vlittle
4
3
4
21
3 10
21
= 10
Next the surface area of an individual little sphere is

Slittle = 4(107 )2
= 41014 cm2
Multiply this by 1021 to get the total surface area,
Vlittle,tot = 41014 1021
= 4107 cm2
107
So nally, before we had a total surface area of 4cm2 . Now we have

times more surface area with the little spheres.
Exercises
1. The lattice constant of Si is 5.43 . Calculate the number of silicon
A
atoms in a cubic centimeter.
2. Calculate the number of Ga atoms per cubic centimeter in a GaAs
crystal. The lattice constant of GaAs is 5.65 . Do the same for As
A
atoms.
3. Consider an actual silicon device on a wafer with physical dimensions
of 5 x 5 x 1 microns. Calculate the total number of atoms in the
device.
4. Consider a slightly larger GaAs laser on the same wafer. Its dimensions
are 100 x 100 x 20 microns. How many total atoms exist on this device.
5. Calculate the number of atoms in a 1.4 nm diameter Pt nanoparticle
using the total number of unit cells present. Consider a FCC unit cell
with a lattice constant of a = 0.391 nm.
6. Calculate the number of atoms in a 1.4 nm diameter Pt nanoparticle
using the bulk density of Pt. Consider = 21.5g/cm3 .
34
7. Estimate the number of surface atoms and percentage of surface atoms
in a 1.4 nm diameter Pt nanoparticle. One can calculate this through
a unit cell approach but use whatever approach you like.
8. Cobalt is usually found with a hexagonal crystal structure. It was
recently found to crystallize with a simple cubic structure, now called
-cobalt. The lattice constant of -cobalt is a = 6.097. The density
A
is = 8.635g/cm3 . The unit cell contains 20 atoms. Calculate the
number of atoms in a 2 nm diameter nanocrystal through a unit cell
argument.
9. Calculate the number of atoms in a 2 nm diamter -cobalt nanocrystal
through a density argument. Use = 8.635g/cm3 .
10. Calculate the number of surface atoms and percentage of surface atoms
in a 2 nm diameter -cobalt particle.
11. CdSe has a hexagonal unit cell (unit cell volume 1123 ). The lattice
A
constants are a = 4.3 and c = 7. Calculate the total number of
A
A
atoms in a CdSe quantum dot for a: 1, 2, 3, 4, 5, 6 nm diameter
particle. How many atoms of each element are there?
12. For the same CdSe dots considered, calculate the fraction of surface
atoms in each case and plot this on a graph.
13. Draw the surface of a Ag crystal cut along the (111) and (100) plane.
References
1. Introduction to solid state physics
C. Kittel
Wiley, New York, New York 1986.
2. Solid state physics
N. W. Ashcroft and N. D. Mermin
Holt, Rinehart and Winston, New York, New York 1976.
3. Solid-state physics
H. Ibach and H. Luth
Springer Verlag, Berlin 1990.
35
Relevant literature
1. A solution-phase chemical approach to a new crystal structure of
cobalt
D. P. Dinega and M. G. Bawendi
Angew. Chem. Int. Ed. 38, 1788 (1999).
36
Chapter 3
Length scales
What are the relevant length scales for nano? Well, I guess it depends on
who you talk to. On one hand some people call nano anything smaller than
stu on the micro level. This could mean delaing with stu on the hundreds
of nm scale. But as Uzi Landman once said, this is pretty high on the
chutzpa scale. Now others will tell you that nano is for extremely small
things like molecules and such. But isnt this just traditional chemistry or
even biology? After all chemists, biologists and physicists have been dealing
with small things for a pretty long time even before the development of
instruments such as transmission electron micrographs. This is the opposite
extreme-calling traditional chemistry, physics and biology, nano to put a
new spin on things.
One useful perspective on a denition for the appropriate lengths scales
for nano is a regime where the chemical, physical, optical and electrical properties of matter all become size and shape dependent. For semiconducting
materials this is given by the bulk exciton Bohr radius. We will describe
excitons (electron hole pairs) in more detail shortly, but for now just picture
the classic Bohr model of electrons circling a heavy nucleus in stable orbits.
DeBroglie wavelength
So what is the deBroglie wavelength? Well this arose within the context of
a classic debate about the true nature of light back in the old days. Was
light a wave or was it a stream of particles? On one side you had Huygens
who thought it was a wave. Then on the other side you had Newton who
thought it was a stream of particles. Tastes great, less lling anyone?
So to make a long story short, deBroglie comes along later and says
both. This was his Ph.D thesis for which he got the Nobel prize in 1929.
37
38
CHAPTER 3. LENGTH SCALES
But deBroglies claim actually goes beyond light because what it really says
is that matter (whether it be you, me, a baseball, a car, a plane, an atom, an
electron, a quantum dot or a nanowireall of these things have both wavelike and particle-like properties). The wave-like properties of matter become
important when you deal with small things so although macroscopic objects
have wave-like properties they are better described by Newtons laws. We
will mostly deal with electrons from here on out.
Therefore according to deBroglie associated with each object is a wavelength.
=
h
p
(3.1)
where is the deBroglie wavelength, h is Planks constant (6.621034 J s)

and p = mv is the momentum with m as the mass.
DeBroglie wavelength and exciton Bohr radius

Here we derive the relationship between the deBroglie wavelength and the
exciton Bohr radius. The reason we do this is that often in the literature
one sees a statement that a nanomaterial is in the quantum connement
regime because its size is smaller than the corresponding deBroglie wavelength of an electron or hole. At other times one sees the statement that
a nanomaterial is quantum conned because its size is smaller than the
corresponding exciton Bohr radius. We ask if these are the same statement.
In this section we show that the two are related and that, in fact, both
statements essentially say the same thing.
Textbook Bohr radius
Here is the textbook equation for the Bohr radius of an electron
a0 =
4 0 2
h
mq 2
(3.2)
where 0 = 8.851012 F/m (permittivity), h = 1.0541034 J s (Plancks
constant over 2), me = 9.11 1031 kg (mass of a free electron) and

q = 1.602 1019 C (charge). If you plug all the numbers in and do the
math you come up with the result
a0 = 5.28 1011 meters
= 0.528 Angstroms
(3.3)
39
This is the standard Bohr radius one sees all the time.
Derivation
Basically we need to equate the centrifugal force of a carrier with the
Coulomb attractive (inward) force.
q2
mv2
=
r
4 0 r2
(3.4)
Here we make use of the relation

2r = n
(3.5)
where n is an integer. Namely, that an integer number of wavelengths must

t into the circumference of the classic Bohr orbit. The deBroglie relation
comes in by relating the wavelength = h where h is Plancks constant
p
and p is the momentum of the particle (p = mv). Starting with the above
equation we rearrange it to get
h
h
2r
= =
n
p
mv
Solve for v to get (express v in terms of r)

v=
nh
n
h
=
2mr
mr
Replace this into the main equation (3.3)

q2
n2 2
h
=
mr
4 0
Rearrange this to get
r=
4 0 n2 2
h
2
mq
If n = 1 (the lowest orbit) this gives us the Bohr radius
a0 =
4 0 2
h
mq 2
40
which is the standard textbook equation we showed earlier.

At this point we note that if the electron or carrier is not in vacuum the
equation should be modied to take into account the dielectric constant of
the medium. (Instead of replace it with 0).
a0 =
4 0 2
h
mq 2
(3.6)
This is because if you are not in vacuum which you arent in a solid,
liquid or gas, then one must account for possible extra charges or dipoles
that can move to oppose the electric eld that one establishes between the
original two oppositely charged species. Any mobile charges or dipoles in
the system will move or re-orient to oppose the eld, trying to cancel it out.
The net eect is that the original electric eld as well as Coulomb potential
are diminished by the response of the material. The factor by which it is
diminished is the relative dielectric constant, denoted commonly as .
Furthermore, for the case of an exciton (electron hole pair) in a semiconductor just replace the mass of the electron with the eective mass of the
exciton.
1
mef f
1
me
1
mh
(3.7)
where me and mh are the eective masses of electron and hole in the material.
Note that equation 3.4 basically gives the relation between the deBroglie
wavelength and the exciton Bohr radius. So, in eect, our initial statements
about connement dealing with either the exciton Bohr radius or deBroglie
wavelength are essentially one and the same. The deBroglie wavelength or
exciton Bohr radius are therefore natural length scales by which to compare
the physical size of a nanomaterial. In general objects with dimensions
smaller than these natural length scales will exhibit quantum connement
eects. This will be discussed in more detail in subsequent chapters.
Examples
Here are some values for some common systems where Ive taken values
of the relative dielectric constant, electron and hole eective masses from
the literature. One can derive the exciton Bohr radius of these systems,
using the values below, in a straightforward fashion. This list is not meant
to be comprehensive and the interested reader should consult the Landolt
Bornstein tables for more complete values.
41
GaAs:
me = 0.067m0
mh = 0.45m0
= 12.4
InAs:
me = 0.02m0
mh = 0.4m0
= 14.5
InP:
me = 0.07m0
mh = 0.4m0
= 14
CdS:
me = 0.2m0
mh = 0.7m0
= 8.6
CdSe:
me = 0.13m0
mh = 0.45m0
= 9.4
Worked example
Case: (GaAs)
1
mef f
=
=
1
1
+
me mh
1
1
+
0.067m0
0.45m0
leading to the eective mass

mef f = 0.058m0
42
4 0 2
h
0.058m0 q 2
4(8.85 1012 )(12.4)(1.054 1034 )2
=
(0.058)(9.11 1031 )(1.602 1019 )2
= 11.3 nm
ab =
Exciton Bohr radius for GaAs.

Case: (CdSe)
1
mef f
=
=
1
1
+
me mh
1
1
+
0.13m0
0.45m0
leading to the eective mass

mef f = 0.1m0
4 0 2
h
0.058m0 q 2
4(8.85 1012 )(9.4)(1.054 1034 )2
=
(0.1)(9.11 1031 )(1.602 1019 )2
= 4.97 nm
ab =
Exciton Bohr radius for CdSe.

So in the end you can see that because of the variable relative dielectric
constant and eective masses of carriers in a semiconductor there will be
various exciton Bohr radii. Now nally lets put things into perspective.
Consider the case of GaAs. We just found that the bulk exciton Bohr
radius was aB = 11.3nm. The lattice constant of GaAs which crystallizes
in the zinc blende form is a = 5.65. Thus the radius of the electron orbit
A
is about 20 unit cells away from the counterpart hole. You can see that in
semiconductors the electron can move very far away from its counterpart
due to the screening eect of the material.
43
Exercises
1. What is the wavelength of a 1 eV photon?
4. What is your deBroglie wavelength (whats your weight in kg?) when
moving at 10 m/s?
5. What is the deBroglie wavelength of a baseball (0.15 kg) moving at 50
m/s?
6. What is the deBroglie wavelength of C60 moving at 220 m/s? Read
the corresponding article if you are interested. Arndt et. al. Nature
401, 680 (1999).
7. Calculate the exciton Bohr radius for the following semiconductors. If
needed use values for what is called the heavy hole. Consult a good
resource such as Landolt Bornstein.
II-VI compounds
CdS
CdSe
CdTe
III-V compounds
InP
InAs
IV-VI compounds
PbS
PbSe
PbTe
8. Explain what the size of the exciton Bohr radius means for achieving
quantum connment. What systems are easiest for achieving connement.
44
Chapter 4
Excitons
Excitons are bound electron-hole pairs caused by the absorption of a photon
in a semiconductor. Specically, we have an electron in the conduction band
of the semiconductor and a hole in the valence band of the semiconductor.
Excitons are potentially mobile and are charge neutral. There are two types
of excitons, Mott-Wannier excitons and Frenkel excitons, distinguished by
how close the electron and hole are. Mott-Wannier excitons have weak
electron-hole interactions caused by a small Coulomb attraction and hence
the carriers are relatively far apart. Corresponding binding energies are on
the order of 10meV . By contrast, Frenkel excitons have strong Coulomb
interactions between the electron and hole. The carriers are close togther.
Corresponding binding energies are on the order of 100meV .
The excitons most commonly dealt with in semiconductors is the MottWanner exciton due to the large relative dielectric constant of most bulk
semiconductors. Frenkel excitons are commonly seen in organic semiconductors.
Excitons can be seen in the absorption spectrum of bulk semiconductors.
They generally appear just below the band edge of the semiconductor. This
is because the energy of the exciton is lower than the band edge transition by
its binding energy. Namely, the exciton energy Eexc = Eg Ebind where Eg
is the bulk band gap of the semiconductor (analogous to the HOMO-LUMO
gap in molecular systems) and Ebind is the exciton binding energy that we
will derive shortly.
45
46
CHAPTER 4. EXCITONS
Figure 4.1: Exciton types
Figure 4.2: Exciton absorption
47
Exciton Bohr radius and binding energies

From the previous chapter we know that the Bohr radius of the electron-hole
pair is
aB =
4 o 2
h
2
q
where is the exciton reduced mass (i used m last time, sorry for the
notation change).
Usually folks like to write the exciton Bohr radius in terms of the standard textbook Bohr radius for the electron, ao . Recall that
ao =
4 o 2
h
mo q 2
Therefore the bulk exciton Bohr radius can be re-written as

aB =
4 o 2 mo
h
mo q 2
where is the relative dielectric constant of the semiconductor and is the

reduced mass of the exciton. This gives
aB =
mo
ao
(4.1)
where ao is everyones favorite textbook Bohr radius (ao = 0.528A). This is

a more convenient way to go about calculating the exciton Bohr radius of
something.
Note that we have omitted something pretty important. In real life the
relative dielectric constant, , is frequency dependent. It isnt readily apparent which to use in the above expressions since there is the static dielectric
constant, the high frequency dielectric constant and anything inbetween.
Generally speaking, use the dielectric constant at the same frequency corresponding to the energy required to create the exciton.
Next we want the energy of the exciton. The total energy of the exciton
relative to its ionization limit is the dierence between its kinetic energy and
its Coulomb (attractive) potential energy.
tot
1
q2
= mv2
2
4 o r
48
CHAPTER 4. EXCITONS
Since we previously saw in deriving the Bohr radius that

mv2
r
q2
4 or2
q2
4 or
mv2 =
so that
tot
q2
4 or
1
q2
=
2 4 o r
q2
4 o r
1
2
But recall that we just found that in a bound exciton

r = aB =
h
o
q 2
2 2
n
Note that many times you will see the exciton Bohr radius written as aX .
Same thing.
Then we get our desired energy
tot
q 4
(4 o )2 2
h
= 1
2
1
n2
(4.2)
Again, you will often see this expressed in many texts in terms of the
Rydberg and the exciton Rydberg (RX ). To express this like everyone else
tot
1
2
mo q 4
(4 o)2 2
h
2m
o
1
n2
Now recognize that the rst term in the expression is our textbook Rydberg
m 4
1
R = 2 (4 oq2 2 which refers to an energy of 13.6 eV. So we now have
) h
o
tot
= R
2m
o
1
n2
Next dene the exciton Rydberg as the product RX = R

desired textbook expression
tot
= RX
n2
2m
to get the
(4.3)
49
Check me
You can check that indeed the Rydberg is correctly dened. Start with our
above denition and simplify it a little to get
mo q 4
8 2 h2
o
R=
Numerically we have
R=
(9.11 1031 )(1.602 1019 )4

8(8.85 1012 )2 (6.62 1034 )2
all in SI units. Evaluate this to get

R = 2.185 1018
The units are Joules. If so then divide this by 1.602 1019 J/eV to get
R = 13.6eV
So we did things right.
In terms of units to prove that the last number was really in units of
joules we have
kgc4
=
=
=
=
F2 2 2
j s
m2
kgm2
s2
jc4
j2F 2
c4
jF 2
c4 V 2
jc2
sinceF =
c4
F 2j2
c
V
c2 V 2
j
2
j
=
j
sincej = cv
= j
50
CHAPTER 4. EXCITONS
Ok, the units work out. Great.

Finally we want the binding energy of the exciton. This is the dierence
in energy between the electron-hole pair in a given orbit (say n = 1) and
being at an innite separation (n = ). If
tot
then
bind
= R
1
n2
2m
o
giving
n=1
bind
1
1
2
2
n
n1
= R
2m
o
For the lowest orbit this gives

bind
=R
2m
(4.4)
In general
bind
mq 4
2(4 o )2 2
h
1
n2
(4.5)
Exciton Schrodinger equation

Excitons can be described as quasi particles and hence there is a quantum
mechanical description for their energies and wavefunctions. We begin with
the Hamiltonian for two independent particles (an electron and a hole) coupled by a common Coulomb term and then re-express the equation into one
for the bound exciton.
In one dimension we have
h
2 d2
h
2 d2
2me dx2 2mh dx2

4
e
h
e2
e h = e h
o (re rh )
More generally however we have
h
2
2me
2
e
h
2
2mh
2
h
e2
e h = e h
4 o (re rh )
To begin converting this into an exciton picture let

R =
=
me re
mh rh
+
me + mh me + mh
mh rh
me re
+
M
M
51
R=
me re
M
mh rh
M
(4.6)
where M = me + mh is the total mass of the exciton. This is the center of

mass coordinate. Next let
r = re rh
(4.7)
be the relative electron hole coordinate.

It can be shown that
re = R +
me mh
(me +mh )
(4.8)
rh = R
where =
r
me
r
mh
(4.9)
is the reduced mass.
Proof, First expression

re = R +
=
=
=
=
=
=
r
me
me re
mh rh
+
+
M
M
mh rh
me re
+
+
M
M
me re
mh rh
+
+
M
M
mh re
me re
+
M
M
me + mh
re
M
re
me mh
(re rh )
M me
mh
(re rh )
M
mh re
mh rh
M
M
Proof, Second expression

rh = R
=
=
=
=
=
r
mh
me re mh rh
+
M
M
me re mh rh
+
M
M
mh rh me rh
+
M
M
me + mh
rh
M
rh
me mh
(re rh )
M mh
me re me rh
+
M
M
52
CHAPTER 4. EXCITONS
By the chain rule we can now re-express e and

re and rh are seen to depend on both r and R.
as follows since both
First expression
re
r
R
+
r re R re
me
(1) +
r
R M
me
+
r
M R
=
=
=
yielding
e
me
M
(4.10)
Second expression
r
R
+
r rh R rh
mh
=
(1) +
r
R M
mh
= +
r
M R
yielding
h
mh
M
Put everything together now

Back to our original expression
h
2
2me
2
e
h
2
2mh
2
h
e2
tot = tot
4 o(re rh )
(4.11)
53
where we will ip this Hamiltonian to something in terms of r and R only.
h
2
2me
me
M
h
2
mh
r+
R
2mh
M
e2
tot = tot
4 o r
h
2
2me
2
r
me
M
me
M
2
r
h
2
2me
me
M
mh
M
me
M
h
2
2
1
1
+
me
mh
2
r
h
2
h
2
r R
R
2M
2M
e2
tot = tot
4 o r
h
2 (me + mh )
2me mh
h
2
2
2
r
h
2
2M
h
2
2M
+
me
M
h
2
2M
2
R
2
R
h
2
2mh
h
2 me
h
2 me
r R
2me M
2me M
h
2 mh
h
2 mh
+
r R+
2mh M
2mh M
2
e
tot = tot
4 o r
h
2
mh
mh
2
r R
r
2mh
M
M
2
e
tot = tot
4 o r
h
2
h
2
2me 2mh
mh
M
me
M
2
R
2
R
mh
M
mh
M
h
2
2M
2
R
2
r
h
2
2M
2
R
e2
(r)(R) = (r)(R)
4 o r
2
R
e2
tot = tot
4 o r
2
R
54
CHAPTER 4. EXCITONS
Since this is a separable problem, let
h
2
2
2
r
h
2
2
(R)
2
r
h
2
2M
h
2
2
e2
4 o r
1 2
h
(r) 2
e2
4 o r
e2
4 o r
2
r (r)
where
2
R
cm
rel
cm
rel
h
2
2M
2
R
(r)(R) (r)
rel .
rel
giving
(r)(R) = 0
h
2
2M
cm
(r)(R) = 0
2
R (R)
cm (r)(R)
=0
Next, divide by (r)(R) to get
2
r (r)
1 2
h
(R) 2M
2
R (R)
cm
=0
This gives you two separate equations. One which depends only on r(left
one) and one which depends only on R (right one). So you get the two
separate equations for the exciton relative motion and the center of mass
motion. Both are solved independently. Solve for the eigenvalues of the
relative expression to get the energies of the bound electron-hole pair.
2
2
r
e2
4 o r
2M
2
R
(r) =
(R) =
rel (r)
(4.12)
cm (R)
(4.13)
Exercises
1. Provide a value for the thermal energy, kT, in units of meV at these
temperatures, 300K, 77K, 10K and 4K. Low temperature experiments
are commonly conducted at 4K and 77K. Any suggestions as to why?
2. GaAs has a zinc blende crystal structure with a lattice constant of a =
0.56nm, a dielectric constant of = 12.8, and electron/hole eective
masses of me = 0.067mo and mh = 0.2mo . Provide the following: (a)
the bulk exciton Bohr radius of this material. (b) the number of unit
cells contained within the lowest exciton orbit n = 1. (c) The highest
temperature at which stable excitons are observable.
55
Connement regimes
At the heart of the binding energy is the Coulomb attraction between the
oppositely charged electron and hole. Recall that this Coulomb term is proportional to 1 . Now it will be shown that the connement energy associated
r
with either the electron or hole is proportional to r1 . Its clear that the latter
2
term will grow faster than the former when the size of a material becomes
small. So in bulk materials, the Coulomb term and associated exciton is
important. By contrast, in nanomaterials the exciton begins to experience
competition from the connement of individual electrons and holes.
There are three size (connement) regimes that one must consider in
real life. They are referred to as the strong, intermediate and weak connement regimes. More about connement will be illustrated in the following
chapters. However, for now note that life isnt all that simple.
Strong connement: a < aB,e , aB,h
Intermediate connement: aB,h < a < aB,e
Weak connement: a > aB,e , aB,h
Weak connement
In this case a > aB,e , aB,h . As a consequence the binding energy of the
exciton is larger than the individual connement energies of the electron
and hole. This is denitely the case with bulk materials and with large
nano scale materials. The energy of optical transitions here is basically the
band gap energy minus the exciton binding energy.
Intermediate connement
In this case we have the physical radius of the material smaller than one
carriers individual Bohr radius but larger than the individual Bohr radius
of the other carrier. Because the eective mass of the electron is smaller than
the eective mass of the hole, one often sees the criteria, aB,h < a < aB,e .
We have
aB,e =
aB,h =
4 o 2
h
2
me q
4 o 2
h
mh q 2
56
CHAPTER 4. EXCITONS
Strong connement
This is commonly seen in small nanomaterials. Here the physical size of
the system is smaller than both individual Bohr radii of the electron and
hole, a < aB,e , aB,h . In this regime the optical properties of the material are
dominated by quantum connement eects of the electron and hole. This
will be seen in the following Chapters.
Exercises
1. For InP which has the following parameters, me = 0.077mo , mh =
0.2mo and = 12.4, what are the individual Bohr radii of the electron
and hole. Provide physical size ranges for this material to be in the
strong, intermediate and weak connement regimes.
2. For CdSe which has the following parameters, me = 0.13mo , mh =
0.45mo and = 9.4, what are the individual Bohr radii of the electron
and hole. Provide physical size ranges for this material to be in the
strong, intermediate and weak connement regimes.
Chapter 5
Quantum mechanics review

Wavefunctions and such
Given the deBroglie wave-particle duality it turns out that we can mathematically express a particle like a wave using a wavefunction (usually
denoted ). This wavefunction replaces the classical concept of a trajectory
and contains all the dynamical information about a system that you can
know. Usually much of the work we will do here will be to nd out what
this wavefunction looks like given certain constraints on the system (called
boundary conditions).
There is a probabilistic interpretation of the wavefunction called the
Born interpretation. In this respect
||2 = is considered as a probability density
||2 dx = dx is considered as a probability
Through these quantities one can determine the probability that the particle
is somewhere. Note that from a physical perspective only ||2 has some
physical signicance. can be real or imaginary or negative but ||2 will
be real and positive.
One consequence of the probabilistic interpretation of the wavefunction
is that the wavefunction must be normalized.
||2 dx = 1
This is because the probability of nding the particle somewhere must be
unity. So generally you will see that in front of (x, t) will be a constant
N which ensures normalization is met. Physical particles therefore have
57
58
CHAPTER 5. QUANTUM MECHANICS REVIEW
normalizable wavefunctions, non-physical ones have wavefunctions that are

not normalizable.
There are some important mathematical properties (or mathematical
constraints) of the wavefunction described as follows
must be well behaved in general
must be nite (it does not blow up or show singularities. This constraint of course is related to being able to normalize the wavefunction.
Cant do it if it goes nuts.)
must be single valued. Obviously it doesnt make sense to have
multiple probabilities at a given position.
must be continuous.
must have a rst derivative (related to previous statement)
must have a second derivative (related to previous, previous statement)
must be integrable and hence normalizable (basically a consequence
of 1 and 2)
Other properties to remember include
normalization
||2 dx = 1
orthogonality 1 2 dx = 0. Thus these two wavefunctions 1 and

2 have no mutual spatial overlap.
Observables
All dynamical information about the particle or system is contained in the
wavefunction. This includes observables such as
position
momentum
angular momentum
energy
59
So while the wavefunction itself is not a measurable quantity, these observables are in principle.
In quantum mechanics, just like the particle is represented by a mathematical wavefunction, observables are represented by mathematical operators.
An operator will act on a wavefunction to transform it into another
function. Some examples of operators include
x (multiply by x)
(take the partial derivative with respect to x)
(take the partial derivative with respect to time)
You will sometimes see operators denoted by a little hat on top of it. For
example
x = x operator (multiply by x)
y = y operator (multiply by y)
and so forth
Correspondence principle
In quantum mechanics all physical observables have corresponding operators. Two fundamental operators are the position operator x and the mo
mentum operator p.
position operator, x = x. read as multiply by x
momentum operator, p = i = i x (read as take the derivative
h
h
with respect to x and multiply by i )
h
All other operators can be constructured from combinations of these two

fundamental operators. For example, for the kinetic energy operator (used
to extract out the energy of the particle through its wavefunction)
T
=
=
=
p2
2m
1
2m
1
2m
x
2
2 2
h
x
i
h
h
2 2
2m x2
i
h
60
This leads to
2
h

T = 2m x2
(5.1)
For the potential energy V (x) we have

V (x) = V (x)
(5.2)
Read this as multiply by V (x).

For the total energy in 1D we have the operator H = T + V where H is
called the Hamiltonian operator.
H
H
= T +V
h
2 2
+ V (x)
=
2m x2
Therefore
2
h

H = 2m x2 + V (x)
(5.3)
This is the total energy operator called the Hamiltonian operator or Hamiltonian for short.
For the total energy in 3D
H =
=
p2
p2
p2
y
x
+
+ z + V (x, y, z)
2m 2m 2m
p2
1
p2
y
i
h
i
h
+
+ z + V (x, y, z)
2m
x
x
2m 2m
h
2 2
h
2 2
h
2 2
+ V (x, y, z)
2
2
2m x
2m y
2m z 2
This results in
2
H = 2m
2
x2
2
y 2
2
z 2
+ V (x, y, z)
2
H = 2m
where
(5.4)
(5.5)
is called the Laplacian or Del-squared and

=

,
,
x y z
In addition = 2 (French)
So to summarize you can see that other operators can be built up from
the fundamental operators x and p.
61
Eigenvalues and eigenfunctions

Now a special class of function which when operated on yields itself again
(may be multiplied by a constant or something) is called an eigenfunction
(or eigenvector) of the operator. The constant that falls out is called an
eigenvalue of the operator. For example, you could have
x
(e ) = ex
x
Here the function ex is an eigenfunction of the
is an eigenvalue of the operator.
operator and the constant
Expectation values
The average of mean value of an observable A is by denition
< A >=
Adx
This is called sandwiching.

In the special case that is an eigenfunction of the operator A
A = a
Therefore
<A> =
Adx
adx
= a
dx
= a since by normalization the integral is 1

Therefore
< A >= a
in the special case that is an eigenfunction of A.
(5.6)
62
Dirac Bra-Ket notation (a quick aside)

Dirac suggested a shorthand notation for writing and dealing with wavefunctions, operators and such. This new notation turns out to be easier and
faster than the default calculus approach.
In this notation
A wavefunction is denoted by | > and is called a ket
The complex conjugate is denoted by < | and is called a bra
Bras and kets follow the rules of linear algebra
< a|b >=< b|a >
|aA >= a|A > if a is a constant (a scalar)
< aA| = a < A|
Likewise
< A|bB + cC >= b < A|B > +c < A|C >
Integrals are represented in the following fashion
1 dx = < 1|1 >= 1 normalization
1
1 dx = < 2|1 >= 0 orthogonality
2
A1 dx = < 1|A|1 > Sandwiching

1
A1 dx = < 2|A|1 > Sandwiching

2
and so forth.
Operator math
Just as wavefunctions have to obey certain constraints, they are important
restrictions on operators that represent observable quantities. This is called
linearity and hermiticity
63
Linearity:
An operator A is linear for any constants and and wavefuncions |f >

and |g > if
A (|f > +|g >) = A|f > + A|g >
So the operator acting on the whole superposition wavefunction tot =

|f > +|g > is the same as the operator acting individually on the parts.
Hermiticity:
An operator A is Hermitian if its integrals obey the following relationship

for all valid |f > and |g >
Ag dx =
f
Af dx
g
< f |A|g > = < g|A|f >
This restriction ensures that eigenvalues of A are all real-valued (as opposed
to complex).
If its going to be observable in real life the number has to be real (not
complex).
Proof:
|k > is an eigenfunction of A with eigenvalue k. Because A is Hermitian
< k|A|k > = < k|A|k >

k = k
Therefore k is real valued.
Furthermore, eigenfunctions of a Hermitian operator associated with different eigenvalues are all orthogonal.
Proof:
< j|A|k > = < k|A|j >

k < j|k > = j < k|j > where < k|j >=< j|k >
(k j) < k|j > = 0
Since k = j (they are dierent eigenvalues)
< k|j >= 0
64
More on operators
A| >= |A >= | >

so that
| > < |
|A > < A|
A| > < |A
where A is called the adjoint of A operator.
More games with the adjoint

1)
< |A| > = < |A >
= < A| >
= < |A | >
Therefore
< |A| >=< |A | >
(5.7)
of course if A = A then A is Hermitian.

2)
< |A| >=< |A >
(5.8)
< |A| >=< A | >
(5.9)
The rst expression reects operation on the right. The second expression
reects operating on the left.
3)

(AB) = B A
Proof: Let
| > = (AB)| >

= A|B >
(5.10)
65
Take adjoint now

A|B > < B|A

AB| > < |B A

Call C = AB or C = B A resulting in
C| > < |C
This ends up with

(AB) = B A
(5.11)
4) If A and B are Hermitian (A = A and B = B )

(AB) = B A = B A = AB + [B, A]

This results in the relationship where [B, A] is called the commutator of
the two operators. This will be introduced in the next section.

(AB) = AB + [B, A]
(5.12)
Commutators
A wavefunction with a characteristic well-dened value of some observable
quantity is an eigenfunction of the corresponding operator. However, this
does not have to have a characteristic value of any other observable.
For a to have a characteristic value of 2 observables simultaneously, the
corresponding operators must commute
Denition of commute
The action of the two operators say A and B taken in succession on any
is identical to the action of the operators taken in reverse order.
AB = B A
(5.13)
or equivalently that the commutator of the two operators equals zero.
66
Denition of commutator

[A, B] = AB B A
A special case is

[A, B] = AB B A = 0
which occurs when A and B commute.

Now if such a exists then it is an eigenfunction of both operators. For
example

A(B| >) = A| >= A| >
But now if AB = B A then
AB| > = B A| >

A| > = B(A| >)
which means that (A| >) is an
eigenfunction of B, call it | >
= B| >
= | >
This implies that | > is an eigenfunction of A. Hence | > is an eigenfunc
tion of both A and B if they commute.

Alternative derivation
Claim that if | > is an eigenfunction of B
B| >= | >
Then if A and B commute (i.e. [B, A] = 0), |A > is also an eigenfunction

with eigenvalue . To illustrate:
of B

B|A > = B A| >= (AB + [B, A])| >

since [B, A] = 0
= AB| >
= (A| >)
= |A >
67
One therefore sees that

B|A >= |A >
(5.14)
Again this implies that | > is an eigenfunction of A.

and B do not commute ([A, B] = 0), you cannot specify

Conversely, if A
the eigenvalues of both simultaneously (only one or the other). This is the
origin of the uncertainty principle, which we will discuss later in more detail.
To illustrate this locally however, x and p do not commute.
Proof: (note that when working with operators its easier to have an
imaginary test function to work on. otherwise it is easy to get confused.)
[, p]| > = (p px)| >
x
x
= xp| > x| >
p
d| >
d
= i x
h
+ i (x| >)
h
dx
dx
d| >
d| >
+x
+ | >
= i x
h
dx
dx
= i | >
h
Resulting in
[, p]| >= i | >
x
h
(5.15)
[x, p] = i
h
(5.16)
or alternatively that
This value is non-zero showing that x and p do not commute. And as
advertised earlier this will lead to the Uncertainty principle

xp
(5.17)
Furthermore, given 2 commuting observables A and B it is always pos

sible to construct an orthonormal basis with eigenstates common to both A
and B.
Additional commutator math

Before we go on we will want to summarize some important commutator
relations
68

[A, A] = 0

[A, B] = [B, A]

[A, B C] = [A, B]C + B[A, C]

[AB, C] = A[B, C] + [A, C]B

[A, B + C] = [A, B] + [A, C]

[A, bB] = b[A, B]

A, [B, C] = [A, B C] [A, C B]

[A + B, C + D] = [A, C] + [A, D] + [B, C] + [B, D]
Examples
Evaluate [, x2 ]. Remember to add some dummy wavefunction to help
p
out.
[, xx] = pxx x2 p
p

d
where p = i
h
dx
d
d
= i (2 ) x2
h
x
(i )
h
dx
dx
d
d
= i x2
h
+ 2 + i x2
x
h
dx
dx
d
d
2 + x2
x
= i 2
h x
dx
dx
= 2i x
h
Therefore we conclude that
[, x2 ] = 2i x
p
h
(5.18)
Evaluate [2 , x]. Remember to add some dummy wavefunction to help

p
69
out.
[2 , x] = (2 x xp2 )
p
p
= p2 x xp2

d2
d
d
h
= i
h
i (x) x 2 (i )2
h
dx
dx
dx
d
d
d2
= 2
h
x
+ + h2 x 2
dx
dx
dx
2
d2
d d
d
+ h2 x 2
+
= 2 x 2 +
h
dx
dx
dx
dx
2
2
d
d
d
+ h2 x 2
= 2 x 2 + 2
h
dx
dx
dx
2
d
d2
d
h
+ h2 x 2
= 2 x 2 2 2
h
dx
dx
dx
d
= 2 2
h
dx
d
= 2 2
h
dx
2 2
h
d
(i )
h
i
h
dx
= 2i p
h
=
Resulting in
[2 , x] = 2i p
p
h
(5.19)
The time independent Schrdinger equation

o
The general form of the Schrodinger equation which we will need to solve
many times is
i t (r, t) = H(r, t)
h
where H is the Hamiltonian operator. Alternatively
i
h
h
2
(r, t) =
t
2m
where V is some generic potential.
+ V (r, t)
(5.20)
70
This equation derived by Schrodinger describes the wavefunction of the

particle and hence is a pretty important. The general form of the Schrodinger
equation has time dependence as you see above. However if can be shown
that there is a time independent form of the equation also called the Schrodinger
equation. To distinguish these two equations the former is often referred to
as the time-dependent Schrodinger equation. Note that one typically deals
with the time-independent version for the rst semester of quantum and
then get into the time-dependent version in the second semsester.
To obtain the time-independent version of the Schrodinger equation assume that V does not depend on time. It is only dependent on position
V V (r). We have
i
h
(r, t) =
t
h
2 2
+V
2m
(r, t)
Note that the left side has time dependence while the right side only has a
spatial dependence. Look for solutions of the form
(r, t) = (r)f (t)

= f shorthand expression
We now have
h
2 2
(f ) =
f + V f
t
2m
h
2f 2
f
=
+ V f
i
h
t
2m
1
h
2 2
i f
h
=
+V
f t
2m
i
h
Note that the left hand side is independent of position and the right hand
side is independent of time. Both sides must equal a constant (call it )
in order to be consistent. It turns out that this constant is the system
eigenenergy which is why we are calling it already.
71
Left hand side
i f
h
f t
1 f
f t
1 f
f t
lnf
=
=
i
h
i
=
h
i t
=
h
note that there is a constant of integration

but we ignore it here
This results in
f =e
i t
h
(5.21)
The form of the function is exponential but it has an imaginary exponent.

So what does this mean? Well, it means that one has plane waves. Note
that eikx = coskx + isinkx.
Right hand side
h
2 2
1
+V =
2m
H =
This expression is something that we are familiar with. Its our favorite
eigenvalue and eigenvector problem.
Summary
So putting everything together we have the full form of the wavefunction
(both the time-dependent and time-independent parts).
i t
= (r)e h
Recall that additional constraints on (r) include

(r) is nite
(5.22)
72

(r) is continuous
(r) is continuous
Furthermore, boundary conditions on the wavefunction will induce quantization on the energies as well as the mathematical form of the wavefunction. Here you will note that typically the rst semester of quantum
mechanics is spent solving the time independent Schrodinger equation for a
number of dierent situation and boundary conditions. These problems are
referred to as model problems and will be described in the next sections.
What do we mean
What exactly did we mean above. Well lets go back to the full timedependent Schrodinger equation
i
h
(r, t) = H(r, t)
t
where we just found that

(r, t) = (r)e
i t
h
Replace this into the Schrodinger equation and evaluate.

i t
(r)e h
t
i t
i (r)
h
e h
t
i t
i
i
h
(r)e h
h
i
(r)
i
h
h
i
h
= H(r)e
e t
h
= H(r)e
i t
h
= H(r)ei h
= H(r)
leaving
H(r) = (r)
(5.23)
which is the time independent Schrodinger equation or our familiar eigenvalue and eigenvector problem. We will solve this equation a bunch of times
for a number of model systems. Later on we will go back and deal more
directly with the time dependent Schrodinger equation.
Some relevant applications of the time independent Schrodinger equation
include
73
particle in a box (absorption)
harmonic oscillator (vibrations)
rigid rotor (rotational spectra)
74
Chapter 6
Connement
Quantum wells, wires and dots are often described using the analogy to a
particle in a 1D box, a 2D box and a 3D box. This is because when the actual
physical length scale of the system is smaller than the exciton Bohr radius
or corresponding deBroglie wavelength (as we saw in the previous section),
either or both the electron and hole experience connment. In turn, the
energies of the carrier along that dimension of the material are no longer
continuous as in the case where there is no connement. The appearance of
discrete states is one of the fundamental signatures of nanomaterials. Since
solving the Schrodinger equation of a carrier to nd its eigenvalues and
eigenfunctions involves using boundary conditions one can also immediately
predict that the actual shape of a quantum well, wire or dot will also play
a role in dictating the ordering and spacing of states. A nanowire will
have a similar but dierent progression of states than a quantum dot (or
nanocrystal). The same applies to quantum wells as well as more exotic
shapes of nanostructures.
In this chapter we solve the simple analytical problems of a particle in a

1 dimensional rectangular box, a cylindrical wire and a sphere (particle in a
spherical box) to illustrate the discreteness as well as progression of states
in wells, wires and dots.
75
76
CHAPTER 6. CONFINEMENT
Figure 6.1: Cartoon of a one dimensional innite barrier potential
1 Dimension of connement
Particle in a 1D innite box
The potential is
ifx <= 0
V (x) = 0 if0 < x < a
ifx >= a
The boundary conditions are
(0) = 0
(a) = 0
The Shrodinger equation to solve is
2
2m d + V =
dx2
(6.1)
Rearrange to yield in the box region where V = 0

d2
+ k2 = 0
dx2
(6.2)
77
where k =
2m
h
2
General solutions are of the form

= Aeikx + Beikx
(6.3)
Note that the rst term describes a right going wave and the second describes
a let going wave. Also if you need it eikx = cos(kx) + isin(kx) and eikx =
cos(kx) isin(kx). Apply the boundary conditions now to simplify
(0) = A + B B = A
(a) = Aeikx Aeikx = 0

This latter equation reduces to
2iA
eikx eikx
=0
2i
2iAsin(ka) = 0
ikx
ikx
ikx +eikx
and cos(kx) = e
where we employ the identities sin(kx) = e e
2i
For this to be true and non-trivial (A = 0), ka = n. This leads to
2m
a = n
h
2
The energy can be solved for, to give
=
n2 h2
8ma2
(6.4)
where n is an integer. Now by normalizing the wavefunction

= 1
one basically gets the equation
a
N2
sin2 (kx)dx = 1
The integral can be evaluated by recalling that sin2 (kx) = 1 (1 cos(2kx)).

2
This is readily integrated to give N =
as
(x) =
2
a
and the complete wavefunction
2
a sin(kx)
(6.5)
78
Exercises
1. Estimate the rst few energies (n = 1, 2, 3) for an electron in GaAs
quantum wells of width 10 nm and 4 nm. Assume the mass is 0.067m0 .
Repeat the same calculation for a heavy hole (mhh = .5m0 ) and
light hole (mlh = 0.082m0 ). In all cases assume an innite box
model.
2. Use Mathcad, Matlab, Mathematica or your favorite mathematical
modeling program and numerically determine the rst 10 energies of
the particle in an innite box (previous problem). Draw the corresponding wavefunctions as well.
3. Consider the following potential in Figure 6.2. Find the energy levels and eigenfunctions. Assume knowledge of the harmonic oscillator
wavefunctions and energies. (look them up if you need to) This problem does not require any extensive work.
4. Using your knowledge of the energy level spacing for a particle in an
innite box as well as the energy level spacing in a harmonic oscillator predict whether the energy levels will converge or diverge for the
following two potentials. That is as the energies get larger, do the
spacings between states increase or decrease.
Figure 6.2: Half a harmonic oscillator
79
Figure 6.3: Various 1D potentials
Particle in a 1D nite box

This problem is a little more complicated. From quantum mechnanics we
know that the solution in the box region where the potential is zero will
be wavelike. In the barrier region we also know that the solutions will be
exponentially decaying.
The potential is
The solutions are
V (x) = 0
ifx <= 0
if0 < x < a
ifx >= a
1 (x) = Aex + Bex

2 (x) = Ceikx + Deikx
3 (x) = F ex + Gex
2m(V )
2m
where =
and k =
. By niteness of the wavefunction
h
2
h
2
B = 0. In addition F = 0. The wavefunction must not blow up in the
barrier region. Must be well behaved.
80
Figure 6.4: Cartoon of a one dimensional nite barrier potential
This leaves us with

1 (x) = Aex
ikx
2 (x) = Ce
(6.6)
ikx
+ De
3 (x) = Ge
(6.7)
(6.8)
Apply the boundary conditions and matching conditions as follows

1 (0) = 2 (0) A = C + D
1 (0) = 2 (0) A = ikC ikD
2 (a) = 3 (a) Ceika + Deika = Gea
2 (a) = 3 (a) ikCeika ikDeika = Gea

This leads to a system of four equations and four unknowns (A,C,D,G).
Arranged in matrix form it looks like

A
1
1
1
0
C
ik
ik
0
=0
0 eika
eika
ea D
G
0 ikeika ikeika ea
Here either we get the trivial solution where A=C=D=G=0 or that the
81
Figure 6.5: Sketch showing the behavior of the wavefunctions in the nite
box model
determinant of the large matrix is zero.

1
1
1
0
ik
ik
0
=0
0 eika
eika
ea
0 ikeika ikeika ea
(6.9)
Simplify this determinant. This can be done a number of ways. One possible
path is shown.
ik(row 3) + (row 4) (row4)
The determinant becomes
1
1
1
0
0 ( ik) ( + ik)
0
=0
0
eika
eika
ea
0
0
2ikeika ( + ik)ea
Followed by
1
(row 2) (row 2)
ik
eika (row 3) (row 3)
82
yielding
1 1
1
0
0 1 (+ik)
0
(ik)
=0
2ika
a(+ik)
0 1
e
e
0 0 2ikeika ( + ik)ea
Additional steps
(row 2) + (row 3) (row 3)
(row 2) (row 2)
giving
1 1
0 1
0
0
0
0
(+ik)
(ik)
(+ik)
e2ika (ik)
2ikeika
0
0
ea(+ik)
( + ik)ea
=0
Finding this determinant basically means nding the sub 2x2 determinant
(+ik)
e2ika (ik)
2ikeika
ea(+ik)
=0
( + ik)ea
This is the same as

e2ika
( + ik)
( ik)
( + ik)ea 2ikeikaa(+ik) = 0
( + ik)e2ikaa
( + ik)2 a
e
2ike2ikaa = 0
( ik)
Drop the ea term to give

( + ik)e2ika
( + ik)2
2ike2ika = 0
( ik)
( + ik 2ik)e2ika =
( + ik)2
( ik)
( ik)2e2ika = ( + ik)2
83
Multiply both sides by eika
( ik)2eika = ( + ik)2 eika
( + ik)2eika ( ik)2eika = 0
Continue simplifying
( 2 + 2ik k 2)eika ( 2 2ik k 2 )eika = 0
( 2 k 2 )(eika eika ) + 2ik(eika + eika ) = 0
2i( 2 k 2 )sin(ka) + 4ikcos(ka) = 0
( 2 k 2 )sin(ka) + 2kcos(ka) = 0
( 2 k 2 )tank(ka) + 2k = 0
Giving our nal expression
tan(ka) =
2m(V )
and k =
where =
h
2
equation we get the equation
tan
2m
a
h
2
2k
(k2 2 )
2m
.
h
2
(6.10)
If we replace this into the above
(V )
2V
(6.11)
Solve this numerically to get all allowed values of the energy (i.e. nd the
roots).
Exercises
1. Consider the same GaAs quantum wells in the previous exercise (well
widths of 4 nm and 10 nm). Assume the eective mass of the electron
is 0.067mo . Also assume, now rather than an innite box, we have a
nite box with a 0.3 eV barrier. Calculate the energy of electron states
trapped in the box. Increase the barrier to 1 eV. Calculate the energy
84
of electron states trapped in the box (also known as bound states).
This can be done using your favorite mathematical modeling software
such as Mathcad, Matlab, Mathematica etc. . . .
Particle in a 1D triangular well

Here the potential is
V (z) = eF z
where e is the fundamental unit of charge and F is the magnitude of the
electric eld giving rise to such a potential. The left side of the potential is
innite, beginning at the origin.
The Schrodinger equation is
h
2 d2
+ V (z) (z) = (z)
2m dz 2
The relevant boundary condition for the wavefunction is that it must go to

zero at the origin (0) = 0.
To simplify things and converge on a name equation that has already
h
2
been solved we do the following. First multiply through by 2m
d2
dz 2
d2
dz 2
Now let z =
z
zo .
We get
1 d2
2 z
zo d2
d2
d2
z
Next, let zo =
2meF z
2m
= 2
2
h
2meF
2m
z = 2
h
2
h
1
2meF
h2
1
3
2meF
h
2
2meF
h
2
h
2
2meF
1
3
2m
h
2
2m 2
3
zo z = 2 zo
zo z =
. We then get
d2
2m
z = 2
z2
h
2
2meF
2
3
85
Figure 6.6: Mathcad sheet showing numerical solutions to the particle in a

nite box problem
86
Let =
leading to
2m
d2
z = 2
z2
1
3
Let
2m
(eF h)2
1
3
(eF )2
h
2m
2m
h
2
=
1
2m
(eF )2
h
1
3
2
3
h
2
2meF
2
3
1
eF
1
3
giving
d2
z =
z2
d2
= ( )
z
z2
Let s = z . Once this is done we are left with
d2
= s
d2
z
(6.12)
This is Airys equation and has general solution that are linear combinations
of Airy functions (Ai and Bi ). In general Bi will diverge as z . Since
we cant have the wavefunction diverge we drop the latter Airy function.
We are then left with the generic solution for the wavefunctions
= N Ai (s)
(6.13)
where N is a normalization constant. To nd the energies of the particle,

we apply the one boundary condition that we have (z = 0) = 0. Since
s = z and z = 0 if z = 0 we get s = . The boundary condition is
therefore () = 0. Let be the (negative) zeros of the Airy function Ai .

Therefore = || and recall = o . Solving for we get = o or
(eF )2
h
=
2m
1
3
(6.14)
which are the desired energies of the particle in a 1D triangular potential.
87
2 Dimensions of connement
Paticle in an innite circular box
The Schrodinger equation here is
2
2m
+ V =
The potential is
0 ifr < a
ifr >= a
V (x) =
Figure 6.7: Cartoon of the innite circular potential

In the region where the potential is zero
where
2,
h
2
2m
the Laplacian in r, coordinates, is

2
1
r r
1 2
r2 2
(6.15)
88
Note that the Laplacian in cartesian coordinates is

Insert this into the main equation to get
h
2
2m
d2
dx2
d2
dy 2
d2
.
dz 2
1 2
1
r
+ 2 2 = 0
r r
r
r
1 2 2m
1
r
+ 2 2 + 2 =0
r r
r
r
h
2
1
1
r
+ 2 2 + k2 = 0
r r
r
r
Multiply through by r2 to get

r
+ (kr)2 +
2
=0
2
Note that you have part of this expression depends only on r and the other
exclusively on . Assume a form of the wavefunction that is = x(r)y().
As shorthand just denote by xy. Lets evaluate the rst two terms in the
main equation above.
= yx
r = yrx
r
r = yrx + yx
r
r r r = yr2 x + yrx
r
Replace the last item into the above main equation to get
yr2 x + yrx + (kr)2xy + xy = 0
Divide through by xy to get
rx
y
r2 x
+
+ (kr)2 +
=0
x
x
y
Solve for y rst.
y
=
y
r2 x
rx
+
+ (kr)2
x
x
(6.16)
89
let
r2 x
rx
+
+ (kr)2
x
x
m2 =
which you notice is independent of and so to y appears as a constant.

Replacing this into the above expression gives
y = m2 y
(6.17)
or
y +m y =0
leading to a general solution of the form
y() = Aeim + Beim
(6.18)
Now given an explicit form for yy (equation 15) replace this back into equation 14. This is because ultimately what we want is x(r). Equation 13
becomes
rx
r2 x
+
+ (kr)2 m2 = 0
x
x
x
x
m2
+
+ k2 2 = 0
x
rx
r
leading to
x +
x
r
+ x k2
m2
r2
=0
(6.19)
which is the normal Bessel equation. Alternatively, to make this look like
something you look up in a book, let z = kr. This results in Bessels equation
2
z 2 d x + z dx + x(z 2 m2 ) = 0
dz
dz 2
(6.20)
Solutions to this equation (for integer m) take the general form

x(r) = AJm (z) + BYm (z)
(6.21)
which is a linear combination of Jm (z) called Bessel functions of the rst

kind and Ym (z) called Bessel functions of the second kind. Note that Bessel
functions of the rst kind are well behaved at the origin but Bessel functions
90
of the second kind will diverge. Therefore, to obtain physically relevant

solutions drop the Bessel functions of the second kind and only consider
Jm (z).
x(r) = AJm (z)
(6.22)
Finally, from our other boundary conditions, Jm (ka) = 0 Therefore ka =

which is the root of the Bessel function. Since k =
eigenenergies that we desire are
=
h
2 2
2ma2
2m
h
2
and z = kr, the
(6.23)
where a is the radius of the circle, m is the mass of the particle and are
the roots (rst, second, third etc...) of the Bessel function. (Please note
that the m in the denominator is the mass of the particle, not the order of
the Bessel function. Sorry for the notation glitch)
Exercises
1. Consider an InP nanowire of diameter 10, 15 and 20 nm (see the Lieber
paper below). Assume me = 0.078mo and mh = 0.4mo . Calculate the
energy of the rst 3 optical transitions by adding these particle in
a circular box energies to the bulk room temperature band gap of
InP. Ignore the length of the wire for simplicity. What colors do you
expect these wires to emit, UV, visible, IR? How does it compare to
the paper?
2. Consider a quantum corral as described below in the Eigler paper.
Assume the diameter of the corral is 71.3. Assume the eective
A
mass of the electron is me = 0.38mo . Calculate the rst 3 electron
energy levels of this corral. How does it compare to the paper?
3. For the same quantum corral above, draw the rst three wavefunctions
of the system. Use your favorite mathematical modeling program such
as Mathcad, Matlab, Mathematica etc. . . .
Relevant reading
Size dependent photoluminescence from single indium phosphide nanowires
M. S. Gudiksen, J. Wang, C. M. Lieber J. Phys. Chem. B 106, 4036
(2002).
91
Connement of electrons to quantum corrals on a metal surface M.
F. Crommie, C. P. Lutz, D. M. Eigler Science, 262, 218 (1993).
Quantum corrals M. F. Crommie, C. P. Lutz, D. M. Eigler, E. J.
Heller Physica D, 83, 98 (1995).
Energy levels in embedded semiconductor nanoparticles and nanowires
K. K. Nanda, F. E. Kruis, H. Fissan Nano Lett. 1, 605 (2001).
3 Dimensions of connement
Particle in an innite spherical box
This is a more complicated problem. Two approaches to a solution are
illustrated with one leading to what are know as spherical Bessel function
and the other to a solution involving regular Bessel functions of half integer
order. The Schrodinger equation is
2
2m
+ V =
The potential is
0
V (x) =
ifr < a
ifr >= a
In the region inside the sphere where V = 0, this reduces to
h
2
2m
(6.24)
where
2
1
r2 r
r2
2 sin()
r
sin()
1
2
(6.25)
r2 sin2 () 2
If replaced into the above equation, multiply by 2mr2 on both sides to

simplify giving
2 r2
h
= 2mr2
92
Figure 6.8: Cartoon of the innite spherical potential
93
Expanded out this looks like
2 r2
h
1
r2 r
r2
r2 sin()
sin()
1
2
r2 sin2 () 2
= 2mr2
2
h
r2
1
sin()
sin()
2
1
sin2 () 2
= 2mr2
Rearrange to give
2
h
2mr2
r
1
1
2
sin()
sin()
sin2 () 2
2
h
r2
=0
where recall that the third term looks familiar. Basically
L2 = 2
h
sin()
sin()
1
2
sin2 () 2
(6.26)
(Angular momentum operator). Our main equation then becomes

2
h
r2
2mr2 + L2 = 0
Furthermore, recall that L2 = h2 l(l + 1) (angular momentum eigenvalues)
leading to
r2
r2
2mr2 2 l(l + 1) = 0
h
2
h
r2
r2
2mr2 + h2 l(l + 1) = 0
Simplify this
2
h
Let k 2 =
2m
h
2
2m 2
r l(l + 1)
h
2
=0
giving
+ k 2 r2 l(l + 1) = 0
(6.27)
94
At this point there are two ways to proceed from this equation. The rst
will lead to solutions involving so called spherical Bessel functions. The
other will lead to solutions involving regular or normal Bessel functions of
half integer order. Spherical and half integer order, normal, Bessel functions
are proportional to each other so ultimately the solutions are the same.
Solution with spherical Bessel functions
Let = x(r)y(, ) The we can evaluate the following items using x and y
as shorthand for x(r) and y(, )
= yx
r2 = yr2 x
r
r2 = y(r2x + 2rx )
r
Replace the last item into equation 23 giving

y(r2 x + 2rx ) + xy(k 2r2 l(l + 1)) = 0
Divide out y and continue simplifying.
r2 x + 2rx + x(k 2r2 l(l + 1)) = 0
2
Let z = kr (therefore dr = dz and dr2 = dz2 . these will be useful in a

k
k
moment) and replace into the above expression
r2 x + 2rx + x(z 2 l(l + 1)) = 0
d2 x
dx
r2 2 + 2r
+ x(z 2 l(l + 1)) = 0
dr
dr
dx
d2 x
+ x(z 2 l(l + 1)) = 0
k 2 r2 2 + 2rk
dz
dz
or
2
z 2 d x + 2z dx + x(z 2 l(l + 1)) = 0

dz
dz 2
(6.28)
This is the general spherical Bessel equation whose solutions take the form:
x(r) = Ajl (z) + Byl (z)
(6.29)
95
where jl (z) are spherical Bessel functions of the rst kind and yl (z) are
spherical Bessel functions of the second kind (also known as spherical Neumann functions written as nl (z)). Note that jl (z) is nite and well behaved
at the origin while yl (z) diverges. So to get a physical solution, one must
drop the spherical Bessel functions of the second kind leaving
x(r) = Ajl (z)
(6.30)
The equation and its solutions can be looked up in a text like Handbook of
Mathematical Functions Abramowitz and Stegun, pg 437. Note that these
spherical Bessel functions are denoted by little jn (z) as opposed to big Jn (z)
which are normal Bessel functions (same with spherical Bessel functions of
the second kind). Furthermore, spherical bessel functions are related to half
integer Bessel functions, a subclass of normal Bessel functions as shown at
the end of this chapter. Examples of the spherical Bessel function solutions
are
j0 (z) =
sin(z)
z
j1 (z) =
sin(z)
z2
cos(z)
z
j2 (z) = 3 sin(z) 3 cos(z)

z3
z2
sin(z)
z
and so forth.
Solution with half integer (normal) Bessel functions
Alternatively assume that the wavefunction has the form =
R(r)y(, ) Replace this in equation 24 and start simplifying.
r2
+ (k 2 r2 l(l + 1)) = 0
We need the following bits of information
= yx r12 + y x
r
r2 = xy + ryx
r
r2 = x y + x ry + x y = x ry
r
x(r)
r y(, )
96
Replace the last item into our main equation

xy 2 2
x ry +
(k r l(l + 1)) = 0
r
Get rid of y
rx +
x 2 2
(k r l(l + 1)) = 0
r
x + x k2
Now let z = kr and x =
l(l + 1)
r2
z(z) dr =
dz
k
=0
giving
l(l + 1)
d2 x
+ x k2
2
dr
r2
k2
=0
d2 x
k 2l(l + 1)
+ z k 2
dz 2
z2
=0
Use the following bits of information
1
1
dx = z + 2 z 2
dz
d2 x
dz 2
d2 x
dz 2
1
z + 2 z 2 +
1
2
1 z 2 + z 2
2
3
1
z + z 2 4 z 2
especially the last term to get

k2
z + z 2
3
z 2
4
k 2l(l + 1)
z k 2
z2
Drop k 2 and continue simplifying
1
3
l(l + 1)
z + z 2 z 2 + z 1
4
z2
l(l + 1)
z +
+ z 1
4z
z2
l(l + 1)
z 2 + z + z 2 1
4
z2
1
z 2 + z + + z 2 l(l + 1)
4
1
+ l(l + 1)
z 2 + z + z 2
4
=0
=0
=0
=0
=0
=0
97
Note that the term in the inner parenthesis is equal to l +
es the full expression to
z 2 + z + z 2 l +
Let n = l +
1
2
1 2
2
This simpli-
=0
1
2
z 2 + z + (z 2 n2 ) = 0
(6.31)
This is the normal Bessel function equation. In this case where we have halfinteger orders n then general solutions for are linear combinations of half
integer Bessel functions. Furthermore, since x(r) = z(z) and R(r) = x(r)
r
the general radial solution, R(r), is
(z) = AJn (z) + BJn (z)
z A Jn (z) + B Yn (z)
x(r) =
R(r) =
A Jn (z) + B Yn (z)
(6.32)
or more explicitly
x(r) =
z AJ(l+ 1 ) (z) + BY(l+ 1 ) (z)

2
2
(6.33)
We can look up the equation and its solution in a book like Abramowitz
and Stegun but since Yn diverges near the origin so it should be dropped
immediately. This leaves
R(r) =
A
J
1 (z)
z (l+ 2 )
(6.34)
Summary
So in the end, regardless of how you came to the solution, to nd the eignevalues we need to nd the roots, , of either
jl (kr) = 0 spherical Bessel
(6.35)
J(l+ 1 ) (kr) = 0 half integer Bessel
(6.36)
98
So on nding the root one also has = kr and from this the desired value
of k. Note that for each value of l there are an innite number of roots. The
eigenvalues are then
ka =
2m
a =
h
2
h
2 2
=
2ma2
leading to
=
h
2k2
2m
h
2 2
2ma2
(6.37)
Finally, the relation between spherical Bessel functions and half integer
Bessel functions are:
jl (z) =
yl = nl (z) =
J 1 (z)
2z (l+ 2 )
Y 1 (z)
2z (l+ 2 )
(6.38)
(6.39)
So either approach to solving the problem is valid. Its your choice on whats
most convenient.
Exercises
1. Calculate the eigenenergies of a free electron (mass mo ) in a 5 nm
diameter sphere for l = 0, 1, 2, 3 using the lowest root of the Bessel
function.
diameter sphere for l = 0, 1, 2, 3 using the second lowest root of the
Bessel function.
diameter sphere for l = 0, 1, 2, 3 using the third lowest root of the
Bessel function.
4. Summarize all these energies in ascending order of energy in one table.
99
5. Use Mathcad, Matlab, Mathematica or your favorite mathematical
modeling program to draw the radial wavefunctions corresponding to
the lowest three energies in your table. Dont forget to normalize the
wavefunction.
6. Calculate the rst 7 energy levels of a free electron in a 5 nm length
three dimensional box. (All sides of the cube have length of 5 nm).
Compare these results to the particle in a sphere case.
7. Qualitatively compare the size dependence of the energies for a particle
in a 1D box (well), an innite circular potential (wire) and a particle
in an innite spherical box (dot). Basically ask yourself how all these
energies scale with r or a. Any similarities? Now take a look a Yu et.
al., Nature Materials 2, 517 - 520 (01 Aug 2003) Letters and comment
on it.
8. Often enough there is a Coulomb attraction between the electron and
hole in a conned system. As a consequence the energies one calculates via particle in a box type expression are often corrected for
this Coulomb attraction. Think back to the basic expression for the
Coulomb energy and qualitatively explain why this term is important
or not important. Hint: think of how these energies all scale with r or
a.
9. Core shell semiconductor particles are pretty popular these days since
they were found to improve the quantum yield of many semiconductor
systems. Read the following papers with particular emphasis on the
rst one. (1) (CdSe)ZnS core-shell quantum dots: Synthesis and characterizatio of a size series of highly luminescent nanocrystallites B.
O. Dabbousi, J. Rodriguez-Viejo, F. V. Mikulec, J. R. Heine, H. Mattoussi, R. Ober, K. F. Jensen, M. G. Bawendi J. Phys. Chem. B 1997,
101, 9463. (2) Quantum-dot quantum well CdS/HgS/CdS: Theory
and experiment D. Schooss, A. Mews, A. Eychmuller, H. Weller Phys.
Rev. B 1994, 49, 17072. (3)Type-II quantum dots: CdTe/CdSe
(core/shell) and CdSe/ZnTe (core/shell) heterostructures S. Kim, B.
Fisher, H. J. Eisler, M. Bawendi J. Am. Chem. Soc. 2003, 125, 11466.
Now Bashir (Dabbousi) has been good enough to plot the radial wavefunction of both the electron and hole in this system. See gure 14.
You too can do this. Reproduce the rst (CdSe/matrix) and second
(CdSe/ZnS) graphs.
100
Particle in a cylinder
Here in cylindrical coordinates
2
1
r r
1 2
2
+ 2
r2 2 z
Solve the Schrodinger equation

H =
where
H=
h
2
2m
+V
where V = 0 inside the cylinder. Elsewhere the potential is innite. The

eective mass of the electron and the hole are used. Replacing this into the
Schrodinger equation we get
h
2 1
2m r r
2
1 2
+ 2 =
r2 2 z
Multiply both sides by 2m giving

h
2
r
1
r r
Let k 2 =
2m
h
2
1 2
2
2m
+ 2 = 2
2 2
r
z
h
1 2
+ 2 z 2 = 2
r2 2
giving
1
r r
Multiply by r2 to get rid of the middle
1
r2
term. We get
2
2
r
+ 2 + r2 2 = 2 r2
r
r
z
2
2
r
+ r 2 2 + 2 r 2 +
r
z
2
r
Now assume a solution of the form

(r, , z) = A(r, )B(z) AB
(6.40)
101
through the method of separation of variables. We then have
2 AB
2 AB
+ (r)2AB +
z 2
2
2A
2B
+ Ar2 2 + (kr)2AB + B 2
z
AB
r
A
r
Br
r
r
+ r2
= 0
= 0
Divide through by AB to get

r
A r
A
r
r2 2 B
1 2A
+ (kr)2 +
=0
B z 2
A 2
Divide through by r2 to get

1
rA r
A
r
1 2B
1 2A
+ 2 + 2
=0
B z 2
r A 2
Consolidate terms to get

1 A
1 2A
1 2B
+ 2
+ 2 =
rA r
r A 2
B z 2
The LHS is a function of (r, ) (although oset by 2 ). The RHS is a function
of z. If these expressions are equivalent then the must both equal a constant
with no dependence of r, or z. Lets call it m2 . Therefore we get
1
rA r
A
r
1 2A
+ 2 = m2
r2 A 2
1 2B
= m2
B z 2
(6.41)
(6.42)
Lets solve the easy equation rst. Begin with
1 2B
B z 2
d2 B
dz 2
= m2
= m2 B
leading to
d2 B
+ m2 B = 0
dz 2
(6.43)
A general solution to this dierential equation is

B = Ao eimz + Bo eimz
(6.44)
102
where Ao and Bo are constants. The boundary conditions here require that
B(0) = 0
(6.45)
B(L) = 0
(6.46)
where L is the length of the rod or the wire.

Applying the rst boundary condition we get
B(0) = Ao + Bo = 0
therefore
Bo = Ao
Applying the second boundary condition we get
B(L) = Ao eimL Ao eimL = 0
or
2iAo
eimL eimL
= 0
2i
2iAo sinmL = 0
(6.47)
therefore we either get the trivial solution (Ao = 0) or we get sinmL = 0.

The latter solution requires that
mL = n
n
m =
L
where n = 0, 1, 2, 3 . . .
Putting everything together we get
B(z) = No sin(
n
z)
L
(6.48)
where No is a normalization constant. This is then the longitudinal part of

the wavefunction.
Now we go and nd the harder solutions. Begin with
1 d
rA dr
dA
dr
1 d2 A
+ 2 = m2
r2 A d2
103
Multiply through by r2 to get
r d
A dr
dA
dr
1 d2 A
+ (kr)2 = (mr)2
A d2
Now multiply through by A

r
d
dr
dA
dr
d2 A
+ (r)2 (mr)2 A = 0
d2
Assume A(r, ) C(r)D() again by the separation of variables approach.

A(r, ) = C(r)D() CD
giving
2 CD
+ (r)2 (mr)2 CD = 0
2
2D
+ C 2 + (r)2 (mr)2 CD = 0
CD
r
C
r
Dr
r
r
r
Divide through by CD
r
C r
C
r
1 2D
+ (r)2 (mr)2 = 0
D 2
Consolidate common terms

r
C r
C
r
+ (r)2 (mr)2 =
1 2D
D 2
(6.49)
Note that the left side depends only on r and the right side depends only
on . Both sides therefore equal a constant. We will call it p2 . This gives
us the following two equations
r
C r
C
r
+ (r)2 (mr)2
1 2D
D 2
= p2
(6.50)
= p2
(6.51)
Lets start by solving the easy equation. Begin with
1 d2 D
D d2
d2 D
d2
= p2
= Dp2
104
giving
d2 D
+ p2 D = 0
d2
(6.52)
The general solutions to this dierential equation are of the form

D = Co eip + Do eip
(6.53)
where Co and Do are constants. By continuity we also know that

D() = D( + 2)
giving
D(0) = Co + Do
D(2) = Co eip(2) + Do eip(2)
by inspection p is an integer and p = 0, 1, 2, 3, . . .
Now we go and solve the harder equation. Begin with
r d
C dr
dC
dr
+ (r)2 (mr)2 = p2
Multiply through by C
dC
d
r
+ (r)2 (mr)2 c
dr
dr
d2 C
dC
r r 2 +
+ (r)2 (mr)2 C
dr
dr
d2 C
dC
r2 2 + r
+ (r)2 (mr)2 C
dr
dr
d2 C
dC
r2 2 + r
+ (r)2 (mr) p2 C
dr
dr
r
= p2 C
= p2 C
= p2 C
= 0
Divide through by r2
1 dC
p2
d2 C
+
+ (2 m2 ) 2 C = 0
dr2
r dr
r
Let a2 = (2 m2 )
d2 C
1 dC
p2
+
+ a2 2 C = 0
dr2
r dr
r
105
Multiply through by r2 again
r2
dC
dC
+r
+ (ar)2 p2 C = 0
dr
dr
Let = ar or d = adr or dr =
a2 r2
d
a .
This gives
dC
dC
+ ar
+ [ 2 p2 ]C = 0
d
d
or
2
dC
dC
+
+ [ 2 p2 ]C = 0
d
d
(6.54)
This is Bessels equation whose solutions are Bessel functions.

C = AJp () + BYp ()
where A, B are constants and = ar where a = 2 m2 and 2 =

Because Yn () as r we have B = 0. This leaves us with
C() = AJp ()
2m
h
2
(6.55)
This is the wavefunction of this part of the problem.

Now we summarize everything here
C(r) = AJp (ar)
D() = Co eip + Do eip
n
B(z) = No sin( z)
L
where p = 0, 1, 2, 3, . . . and n = 0, 1, 2, 3, . . . . Putting it all together we
have the total wavefunction as the product of the individual terms =
C(r)D()B(z) giving
ip
(r, , z) = n Jp (ar) Co + Do eip
sin(
n
z)
L
(6.56)
This is our desired wavefunction.

The desired energies come from the roots of the Bessel function
Jp (ar) = 0
(6.57)
106
where r = ao is the radius of the

cylinder. will be the root of the Bessel
function and = aao where a = 2 m2 . Then we have
2 m2 ao
2m
m2 ao
h
2
=
=
2m
h
2
ao
2
a2
o
2
+ m2
a2
o
2m
m2 =
h
2
2m
=
h
2
giving
=
h
2
2m
2
+ m2
a2
o
where m = n , ao is the radius of the cylinder, and are the roots of the
L
Bessel function. If ao = R the the nal desired energies of the particle in a
cylinder are
=
h
2
2m
n
L
(6.58)
Chapter 7
Nondegenerate perturbation
theory
The basic idea here is that the true Hamiltonian H is close to an approximate
(0) whose eigenfunctions and eigenvalues are known. The
Hamiltonian H
dierence H (1) is called a perturbation. Higher order corrections such as

(2) , H (3) . . . can be included if desired. Usually though you will stop at
H
the rst or second correction to either the energies or wavefunctions.
We introduce a parameter to keep track of the order or alternatively,
the order of renement of the solutions.
So 1st order corrections from H (1) have a in front of it.
2nd order corrections from H (2) will have 2 in front of it

and so on in this fashion
Also does not represent any real physical quantity. It is simply a
mathematical technique for keeping track of the order of approximation. In the end we will usually set = 1 but the method is valid for
any fraction of .
Our corrected Hamiltonian is therefore
H = H (0) + H (1) + 2 H (2) + . . .
where H (0) is our unperturbed Hamiltonian whose eigenvalues and eigenvectors we know. Our wavefunctions and energies are
= (0) + (1) + 2 (2) + . . .
=
(0)
(1)
+ 2
107
(2)
+ ...
108
CHAPTER 7. NONDEGENERATE PERTURBATION THEORY
The results of perturbation theory are derived by taking the exact Schrodinger
equation H = and substituting in our above expressions for and .
Next we group all terms with the same power of .
[H (0) + H (1) + 2 H (2) ][(0) + (1) + 2 (2) ]
= [
(0)
(1)
+ 2 (2) ][(0) + (1) + 2 (2) ]
Ok things get complicated here. We will separately evaluate the left and
right hand side of these expressions. In addition, we will consider terms only
up to second order (i.e. drop 3 and higher terms).
Left hand side
H (0) (0) + H (0) (1) + 2 H (0) (2) + 3 H (1) (2) + 4 H (2) (2)
+ H (1) (0) + 2 H (1) (1) + 3 H (2) (1)
+ 2 H (2) (0)
When we consider terms only up to second order we get
H (0) (0) + [H (0) (1) + H (1) (0) ] + 2 [H (0) (2) + H (1) (1) + H (2) (0) ]
Right hand side
(0)
(0) +
(0)
(1) + 2
(0)
(2) + 3
(1)
(2) + 4
(1)
(0) + 2
(1)
(1) + 3
(2)
(1)
+ 2
(2)
(2)
(2)
(0)
When we consider terms only up to second order we get

(0)
(0) + [
(0)
(1) +
(1)
(0) ] + 2 [
(0)
(2) +
(1)
(1) +
(2)
(0) ]
Now because the solution must be true for all values of one can equate
the coecients of power of giving
H (0) (0) =
H (0) (1) + H (1) (0) =
H (0) (2) + H (1) (1) + H (2) (0) =
(0) (2)
(0) (0)
(0) (1)
(1) (0)
(1) (1)
(2) (0)
109
Lets rearrange these expressions to get
(H (0)
(0) )(0)
=0
(7.1)
(H (0)
(H (0)
(0) )(2)
(0) )(1)
+ (H (1)
(1) )(0)
=0
(7.2)
+ (H (1)
(1) )(1)
+ (H (2)
(2) )(0)
=0
(7.3)
The rst of the three equations is our unperturbed case. It provides our
zero-order solutions. Nothing special here.
Now multiply all expression by (0) and integrate. Also lets switch to
bra-ket notation so I dont have to write out the integrals.
< (0) |H (0)
(0)
< (0) |H (0)
|(2) > + < (0) |H (1)
(0)
(1)
< (0) |H (0)
(0)
|(1) > + < (0) |H (1)
(1)
|(1) > + < (0) |H (2)
(2)
|(0) > = 0
|(0) > = 0
|(0) > = 0
Now in the case of the rst order and second order expressions the rst
term in each case is zero.
< (0) |H (0)
< (0) |H (0)
(0)
(0)
|(1) > = 0
|(2) > = 0
this is because H (0) is Hermitian H (0) = H (0) . So you can operate with
H (0) on the bra and see that each expression will die. Alternatively, you
could also invoke the orthogonality of < (0) | and |(1) > after operating
on the bra.
This leaves
< (0) |H (1)
(1)
< (0) |H (1)
|(1) > + < (0) |H (2)
(1)
(2)
|(0) > = 0 1st order
|(0) > = 0 2nd order
Next lets break each expression up and drop any zero terms. We get
< (0) |H (1) |(1) >
(1)
< (0) |H (1) |(0) >
< (0) |(1) > + < (0) |H (2) |(0) >

(0)
<
(0)
<
|H
(1)
(1)
(0)
>+<
(1)
(2)
|H
|H
< (0) |(0) > = 0
< (0) |(0) > = 0

reducing to
(1)
(2)
|(0) >
(0)
>
(1)
= 0
(2)
= 0
nally simplifying to
< (0) |H (1) |(0) > =
< (0) |H (1) |(1) > + < (0) |H (2) |(0) > =
(1)
(2)
110
Therefore for the energies we have the following rst and second order corrections to the zero order solutions.
(1)
(2)
=< (0) |H (1) |(0) >
(7.4)
=< (0) |H (1) |(1) > + < (0) |H (2) |(2) >
(7.5)
Note that to get the second order energy correction you already need to know
what |(1) > is. The same applies to higher order corrections and means
that this can get pretty tedious. Usually you stop at rst order corrections.
So right now our second order expression is not necessarily in a useful form.
We will revisit this later after nding out what |(1) > looks like.
Example 1, expression for
Please calculate the expression for the third order correction

H
= H (0) + H (1) + 2 H (2) + 3 H (3) + . . .
= (0) + (1) + 2 (2) + 3 (3) + . . .

=
(0)
(1)
+ 2
(2)
+ 3
(3)
+...
Plugging this into the Schrodinger equation we get

[H (0) + H (1) + 2 H (2) + 3 H (3) ][(0) + (1) + 2 (2) + 3 (3) ]
= [
(0)
(1)
+ 2
(2)
+ 3
(3)
][(0) + (1) + 2 (2) + 3 (3) ]
Again, since this gets a little involved we will consider the left and right
hand side of this equation separately.
Left hand side
H (0) (0) + H (0) (1) + 2 H (0) (2) + 3 H (0) (3) + 4 H (1) (3)
+ 5 H (2) (3) + 6 H (3) (3)
+ H (1) (0) + 2 H (1) (1) + 3 H (1) (2) + 4 H (2) (2)
+ 5 H (3) (2)
+ 2 H (2) (0) + 3 H (2) (1) + 4 H (3) (1)
+ 3 H (3) (0)
We want the third order corrections here.
111
Right hand side
(0)
(0) +
(0)
(1) + 2
(0)
(2) + 3
(0)
(3) + 4
(1)
(3) + 5
(2)
(3) + 6
(1)
(0) + 2
(1)
(1) + 3
(1)
(2) + 4
(2)
(2) + 5
(3)
(2)
+ 2
(2)
(0) + 3
+ 3
(3)
(2)
(1) + 4
(3)
(3)
(3)
(1)
(0)
We only want the 3rd order corrections here.

This leaves
H (0) (3) + H (1) (2) + H (2) (1) + H (3) (0) =
(0) (3)
(1) (2)
(2) (1)
(3) (0)
Multiply both sides by (0) and integrate. Switching to the Dirac bra-ket
notation we get
(H (0)
(0)
)|(3) > +(H (1)
< (0) |H (0)
(0)
+ < (0) |H (2)
(2)
< (0) |H (1)
(1)
(1)
)|(2) > +(H (2)
|(3) > + < (0) |H (1)
|(1) > + < (0) |H (3)
(3)
the rst term drops out leaving
+ < (0) |H (3)

becoming
(3)
|(2) > + < (0) |H (2)
|(0) >= 0
< (0) |H (1) |(2) >
+ < (0) |H (3) |(0) >
(1)
(3)
(1)
(2)
)|(1) > +(H (3)
(3)
|(2) >
|(0) >= 0
(2)
|(1) >
< (0) |(2) > + < (0) |H (2) |(1) >

< (0) |(0) >= 0
(2)
< (0) |(1) >
by orthonormality a bunch of terms drop out giving

< (0) |H (1) |(2) > + < (0) |H (2) |(1) >
+ < (0) |H (3) |(0) >=
(3)
Therefore our desired 3rd order correction is

(3)
)|(0) >= 0
=< (0) |H (1) |(2) > + < (0) |H (2) |(1) > + < (0) |H (3) |(0) >
(7.6)
112
Example 2, Particle in a slanted box

We have a particle in a 1D box with a slanted bottom. The perturbation
term is H (1) = V x where V is some constant. The zeroth order wavefunction
a
is (0) =
2
nx
a sin a
and the zeroth orther energies are
(0)
n2 h2
8ma2 .
What is the 1st order correction to (0) ?

From nondegenerate perturbation theory the rst order correction is
(1)
= < (0) |H (1) |(0) >

a
nx
2
sin
a
a
=
0
=
=
=
=
2
a
V
a
a
0
xsin2
V
x
a
nx
2
sin
a
a
dx
nx
dx
a
n
let b =
a
a
2V
xsin2 bxdx
a2 0
1
where sin2 bx = (1 cos2bx)
2
2V 1 a
x(1 cos2bx)dx
a2 2 0
a
a
V
xdx
xcos2bxdx
a2 0
0
a
V x2 a
|0
xcos2bxdx
a2 2
0
a
V a2
xcos2bxdx
a2 2
0
Solve the integral by parts. Let u = x, du = dx, dv = cos2bxdx, v =

1
2b sin2bx. We get
a
V x
1
V
2
sin2bx|a
sin2bxdx
0
2
a 2b
0 2b
V
V x
1 1
=
2
sin2bx|a +
cos2bx|a
0
0
2
a 2b
2b 2b
the rst term is zero
V
V
1
=
2
cos2bx|a
0
2
a 4b2
the term in brackets is also zero
113
therefore
(1)
V
2
(7.7)
This is our desired 1st order correction.

The nal corrected energies are then
=
n2 h2
8ma2
V
2
(7.8)
where n = 1, 2, 3, . . .
Example 3, An anharmonic oscillator

Consider an anharmonic oscillator whose potential is U (x) = 1 kx2 + 1 x3 .
2
6
Note that this problem does not require a lot of math.
Calculate the 1st order correction to the ground state energy. We have
1
= (n + )h
2
1
= Nn Hn ( 2 x)e
x2
2
where = k and the normalization constant is Nn =

h
2
The rst order energy correction is
(1)
1
2n n!
1
4
=< (0) |H (1) |(0) >
1
where H (1) = 6 x3 . We have
=
=
x2
x2
1
1

Nn Hn ( 2 x)e 2 (x3) Nn Hn ( 2 x)e 2
6
1
1
2
2
[Hn ( 2 x)][x3][Hn ( 2 x)]ex dx
Nn
6
dx
Now if n = 0 H0 ( 2 x) = 1 leaving
=
2
N
6 n
x3 ex dx
The rst term in the integral is odd. The second is even. Now by symmetry
their product is overall odd and over the range of the integral it evaluates
to zero. Therefore
(1)
=0
(7.9)
114
This is our desired rst order correction. So to rst order the overall energies
are
= (n + 1 )h
2
(7.10)
Example 4, Another anharmonic oscillator

Calculate the 1st order correction to the ground state energy of an anharmonic oscillator whose potential energy is
1 2 1 3
1
kx + x + bx4
2
6
24
1
1 3
x + bx4
=
6
24
1
= (n + )h
2
U (x) =
H (1)
= Nn Hn ( 2 x)e
k
h
2
where =
N0 =
(1)
1
4
1
2n n!
and the normalization constant is N =
1
4
and
if n = 0. The rst order energy correction is
= < (0) |H (1) |(0) >

=
x2
2
N0 H0 ( 2 x)e
1
1 3
x + bx4
6
24
x2
2
N0 H0 ( 2 x)e
x2
2
We have already done the 1st term. By symmetry it drops out leaving.
=
b 2
N
24 0
(H0 ( 2 x))(x4)(H0 ( 2 x))ex dx

1
where H0 ( 2 x) = 1
b 2 4 x2
x e
dx
=
N
24 0
this integral does not dissapear by symmetry
and must be evaluated
2
bN0 4 x2
=
x e
dx
12
looking this up in a table of integrals
where
2n x2
dx
0 x e
135(2n1)
2n+1 n
dx
115
Our integral evaluates to
3
82
The expression becomes
2
bN0
12
3
82
2
bN0
322
giving
2
where N0 =
b
322
where 2 =
k
giving
h
2
b 2
h
32k
The overall energy with the rst order correction is therefore

1
b 2
h
= (n + )h +
2
32k
=
h
2
b 2
h
32k
(7.11)
Wavefunctions
Now that we have the energy corrections, what about the wavefunctions?
Note that you will usually go to second order in the energies but only 1st
order in the wavefunctions. Note that this is also called the Raleigh Ritz
method.
(0)
We have i which is our unperturbed wavefunction. We know that we
have corrections like
(0)
tot = i
(1)
+ i
(2)
+ 2 i
+...
A convenient way to express (1) is as a linear expansion in unperturbed

(0)
wavefunctions which we will denote as j or |j >. We then have
(1)
=
j
cij |j >
and where cij are weights representing how much of each unperturbed wavefunction is present.
116
If we go back to our expression where we derived the 1st order corrections

in energy we have
(0)
(1)
i )i
(1)
(0)
i )i
(1)
(0)
H (0) i + H (1) i
(H (0)
(1)
Now recall that i
j cij |j
= 0
=
(0)
i
cij H (0) |j > +H (1) |i > =
(0)
i
(0) (1)
i i
(1) (0)
i i
> and replace into the above expression
cij |j > +H (1) |i > =
H (0)
+ (H (1)
cij |j > +
(1)
i |i
>
cij |j > +
(1)
i |i
>
Recall that |j > are eigenfunctions of H (0) since we expanded in the unperturbed basis.
cij
j
(0)
j
j
(0)
j |j
> +H (1) |i > =
(0)
i
cij |j > +H (1) |i > =
(0)
i
cij |j > +
(1)
i |i
>
cij |j > +
(1)
i |i
>
Next to start nding our desired coecients cij multiply both sides by < k|
which is a member of the unperturbed basis. Note that k = i.
(0)
j
j
(0)
j
j
cij < k|j > + < k|H (1) |i > =
(0)
i
cij < k|j > + < k|H (1) |i > =
(0)
i
cij < k|j > +
(1)
i
< k|i >
cij < k|j >

j
Now due to the orthogonality between |k > and |j > the only terms in the
above expression that survive are those where k = j.
(0)
k
k
cik+ < k|H (1)|i > =
(0)
i
cik
k
Consolidate our desired cik terms to get

cik (
k
(0)
k
(0)
i )
= < k|H (1) |i >
117
Therefore
cik =
< k|H (1)|i >

(
cik =
(0)
k
(0)
i )
<k|H (1) |i>

(
(7.12)
(0)
(0)
i k )
These are our desired correction coecients. Also recall that (1) =
so you can see that the corrected wavefunction to rst order is
| >= |i > +
<k|H (1) |i>

k ( (0) (0) ) |k
i
k
j cij |j
(7.13)
>
The correction is
<k|H (1) |i>
k ( (0) (0) ) |k
i
k
(1) =
>
(7.14)
Now that we have the 1st order correction to the wavefunction, we can go
back and express the 2nd order correction in energy in a more useful form.
Recall that
(2)
(0)
(1)
= + 
(1)
where we didnt know |i > before. Now that we have an explicit form
we can substitute it in. To make life even easier assume that the total
perturbation goes only to rst order, thus H (2) = 0. We get
(0)
(2)
(1)
= 
< k|H (1) |i >< i|H (1) |k >
(0)
i
(0)
k )
The nal expression is

(2)
<k|H (1) |i><i|H (1) |k>

(
(0)
(0)
i k )
(7.15)
This is a more user friendly 2nd order energy correction. Also note that
i = k. This is called non-degenerate perturbation theory.
>
118
Summary
(1)
(2)
<k|H (1) |i><i|H (1) |k>

(
(0)
(0)
i k )
=< i|H (1) |i >
(7.16)
|<k|H (1) |i>|2
(7.17)
(0)
(0)
i k )
<k|H (1) |i>

k ( (0) (0) ) |k
i
k
| >= |i > +
(7.18)
>
So you see that we will typically go to second order corrections in the energies
and only to 1st order in the wavefunctions.
What about 3rd order corrections?

Going back for the 3rd order non-degenerate energy correction (3). Here is
(2)
(1)
an alternative expression where we dont need to know |i > only |i >.
Starting with
(3)
i
(0)
(2)
= + + 
and with
+ (H
(1)
+ (H (2)
(1)
(0)
i )i
(2)
(0)
i )i
= 0
= 0
Using the second and third equations we will try and get rid of the rst term
in the rst equation. To do this, multiply the second expression by (2)
and integrate. Multiply the third expression by (1) and integrate.
Second expression
From the second expression we get the following
(0)
(1)
(2)
(1)
(0)
(H (1) i )i
i )i + i
(0)
(1)
(2)
(1)
(0)
(1)
i |i >
i |i > + = 
119
where since H (1) and H (0) and
(0)
(1)
(2)
i |i
(0)
(2)
i |H (1) |i
 = = <
(1)
i |H (0)
(0)
(1)
(0)
(2)
i |i
(0)
(2)
i |i
>
>
therefore
(3)
i
(1)
= + + 
Third expression
Now from the third expression we get
(1)
(0)
(2)
(1)
(1)
(2)
(0)
(1)
(H (1) i )i + (1) (H (2) i )i
i )i +
(0) (2)
(1)
(1)
(1)
(1)
(2)
(0)
(1)
i |i > + 
i |i > + =<
(1)
i |H (1)
(1)
(1)
i |i
>+<
(1)
i |H (2)
(2)
(0)
i |i
>
Finally using this last expression we can replace it into the result from
the second expression above to get
(3)
i
(1)
(0) (2)
(0)
(1)
(2)
|i > + + + ]
(0)
(1)
(0)
(0)
 + 
= 
The only term that simplies is the second one in brackets. It reduces to
(1)
(0)
(1)
(0)
 since |i > and |i > are orthogonal. Now we get
for the 3rd order correction
(3)
i
(1)
= + 
> =
k
(0)
i
(0)
k )
(0)
j )
< i|H (1) |j >
(1)
 + 
|k >
< j|
(0)
(0)
> + 
120
Now assume (to make life simple) that the Hamiltonian only has a rst
order perturbation H (2) = 0 and H (3) = 0. This leaves behind
(3)
i
(1)
(1)
(1)
i |i >
(1)
(1)
(1) (1) (1)
i |H (1) |i > 
(1)
(1)
(1)
(1)
(1)
i |H (1) |i > i 
= (1)

H
< j|
< k|H (1)|i >

(
(0)
i
(0)
k )
< i|H (1) |j >< j|H (1) |k >< k|H (1) |i >
(0)
i
kj=i
(0)
(0)
j )( i
(0)
k )
(0)
(0)
j )( i
|k >
(1)
i
(1)
i
(1)
(1)

>
(0)
k )
< i|H (1) |j >< j|H (1) |k >< k|H (1) |i >
< i|H (1) |j >
(1)
i
(0)
i
kj=i
(0)
i
(0)
j )
< j|
< k|H (1) |i >

(
(0)
i
(0)
k )
|k >
< i|H (1) |j >< j|H (1) |k >< k|H (1) |i >
kj=i
(0)
(0)
j )( i
(0)
k )
< i|H (1) |j >< j|k >< k|H (1) |i >
(1)
i
(0)
i
kj=i
(0)
i
(0)
(0)
j )( i
(0)
k )
Now since we have < j|k > the only terms that are non-zero are those where
j=k
< i|H (1) |j >< j|H (1) |k >< k|H (1) |i >
kj=i
(0)
i
(0)
(0)
j )( i
(0)
k )
< i|H (1) |k >< k|H (1)|i >
(1)
i
k=i
(0)
i
(0) 2
k )
Therefore
(3)
i
kj=i
<i|H (1) |j><j|H (1) |k><k|H (1) |i>

(
(0)
(0)
(0)
(0)
i j )( i k )
(1)
i
k=i
<i|H (1) |k><k|H (1) |i>

(
(0)
(0) 2
i k )
(7.19)
Alternatively,
(3)
i
kj=i
(1)
i
=< i|H (1) |i > so that
<i|H (1) |j><j|H (1) |k><k|H (1) |i>

(
(0)
(0)
(0)
(0)
i j )( i k )
< i|H (1) |i >
k=i
<i|H (1) |k><k|H (1) |i>

(
(0)
(0)
i k )
(7.20)
Chapter 8
Density of states
The idea here is that while the connement of an axis gives rise to discrete
energies (as we saw in the previous section on connement), in cases such
as the quantum well and quantum wire, there are additional states along
the unconned axes (degrees of freedom). Here it is not practical to try
and calculate the energies of each state since they are actually continuous
bands or energies. As a consequence one way to get an idea of what the
energies look like is to perform the following density of states argument. So
when put together with the conned energies from the last section, these
density of states calculations provide a more thorough description of what
the electronic structure of 3D, 2D, 1D and 0D materials look like.
3 Dimensions (bulk)
Consider the volume in k space
4
Vk = k 3
3
2
2
2
(k 2 = kx + ky + kz ) where for a particle in this sphere
kx =
ky =
kz =
2
Lx
2
Ly
2
Lz
Note that the 2 arises from the constraints of a periodic boundary condition
(Born-Von Karman conditions, named after Max Born and Theodore Von
121
122
CHAPTER 8. DENSITY OF STATES
Karman) as opposed to the more general n where n = 0, 1, 2, 3 . . . . The

volume of a given mode is then kx ky kz The number of modes (N ) in the
sphere are then
N=
4
k 3
Vk
= 3 3 Lx Ly Lz
kx ky kz
8
Say the particle is an electron and we consider the spin degeneracy (up and
down), then we multiply N by 2.
N
= 2N = 2
4
3
3 k
Lx Ly Lz
8 3
k3
Lx Ly Lz total number of states in sphere
3 2
=
Consider the density
=
k3
N
= 2 number of states/unit volume
Lx Ly Lz
3
Now consider the energy density dened as

3
k
d 3 2
d
=
=
d
d
where recall k =
natively k 3 =
2m
h
2
2m
h
2
3
2
1 dk 3
3 2 d
(8.1)
(units of 1 over length) for a free electron or alter-
. The expression becomes

1
= 2
3
1
= 2
3
3
2
2m
h
2
2m
h
2
3D =
1
2 2
3
2
2m
h
2
3
2
d 2
d
3 1
2
2
(8.2)
This is the density of states in 3 dimensions. Note the square root dependence on energy. Note also that the energies of the system are
=
where k =
h
2k2
2m
2
2
2
kx + ky + kz and kx,y,z =
2n
Lx,y,z .
(8.3)
123
Figure 8.1: Density of states in a 3D material (bulk)
Exercises
1. Calculate the 3D density of states for free electrons with energy 0.1
eV. Express your answer in terms of eV and cm3 .
2. Assume the electron resides in a non-zero potential V . Express the 3D
density of states in this situation. Hint, just alter the expression for k
below equation 5.1.
2 Dimensions (well)
Here we have 1 dimension that is quantized. Lets assume its the z direction.
The total energy of this system is a sum of the energy along the quantized
direction + the energy along the other 2 free directions. It is expressed as
tot =
h 2 2 k2
2 kz
h
+
= n + x,y
2m
2m
2
2
where k 2 = kx + ky and kz =
n
Lz
Consider now an area in k space
Ak = k 2
(8.4)
124
where for the particle

kx =
ky =
2
Lx
2
Ly
The area of a given mode is then kx ky with the total number of modes (N )
in the area being
N =
k 2 Lx Ly
k 2
Lx Ly =
4 2
4
Again if the particle is an electron and we consider spin, multiply by 2 to

get
N = 2N =
k 2 Lx Ly
total number of states in area
2
Consider now a density

=
where recall that k =
=
2m
.
h
2
N
k2
=
Lx Ly
2
Replacing this above gives
m
number of states per unit area
h
2
Now consider the energy density

=
d
m
= 2
d
h

This is the energy density of the subband for a given kz (or n ). For each successive kz there will be an additional m term and hence another subband.
h2
Therefore the density of states is written
2D =
m
h
2
( n )
(8.5)
where is the Heavyside function.

Exercises
1. Calculate the 2D density of states for free electrons with energy 0.1
eV. Consider only the lowest subband. Express your answer in eV and
cm3 units.
125
Figure 8.2: Density of states in a 2D material (quantum well)
1 Dimension (wire)
Consider now the situation where there are 2 dimensions conned and only
1 degree of freedom (say the x direction). The total energy of the system
can be written as
tot =
h2 2 2 k 2
h
h 2 ky
2 kz
+
+
= n + m + x
2m
2m
2m
(8.6)
2
where k = kx = Lx Furthermore along the conned directions, kz = n ,
Lz
m
ky = Ly and m, n are integers. Consider a length 2k. The number of modes
along this length is
N=
2k
=
kx
2k
2
Lx
kLx
number of state along the line
Now if we consider an electron again, ensure to take into account the spin
degeneracy
N = 2N =
2kLx
Now a density is
=
2k
2
N
=
=
Lx
2m
number of states per unit length
h
2
126
Consider the energy density

=
=
d
2
=
d
2
2m d
2 d =
h
2m 1 1
2
h
2 2
2m 1
h
2
(8.7)
This is the energy density for a given n, m value or (n , m combination).

The complete expression taking into account all m, n combinations is
1D =
2m
h
2
n,m
1
n,m ( n,m )
(8.8)
where again, is the Heavyside function. Notice the inverse square root
dependence of the density of states with energy.
Figure 8.3: Density of states in a 1D material (quantum wire)
Exercises
1. Calculate the 1D density of states for free electrons with energy 0.1 eV
above n,m . Consider only the lowest subband. Express your result in
units of eV and cm3 .
127
0 Dimension (dot)
Here since all three dimensions are conned the density of states is basically
a series of delta functions. The total energy of the system is
tot =
2
h 2 h 2 2 kz
2 kx ky
h 2
+
+
= m + n + o
2m
2m
2m
where m, n, o are integers and kx =

states is
m
Lx ,
ky =
n
Ly ,
kz =
(8.9)
o
Lz
The density of
0D = ( m,n,o )
(8.10)
Figure 8.4: Density of states in a 0D material (quantum dot)
Exercises
1. Compare and contrast the density of states in the 0D case to the
previous 3D, 2D and 1D cases.
128
Chapter 9
More density of states

Density of states in the conduction band
To do this calculation we need to know the probablility that an electron will
occupy a given state of energy . This probability, Pe (), is refered to as the
Fermi Dirac distribution. In addition, we need to know the density of states
( ) which we calculated in the last section. Recall that this density of states
has units of number per unit volume per unit energy. Therefore d is the
number of states per unit volume. The number of occupied states at a given
energy per unit volume (or alternatively the concentration of electrons at a
given energy) is therefore
ne () = Pe () ()d
where the Fermi Dirac distribution is
Pe () =
1
1+e
(9.1)
F
kT
Here F is the Fermi energy (also called the chemical potential sometimes).
The total concentration of electrons in the conduction band is therefore
the integral over all available energies
ntot =
Pe () ()d
(9.2)
where C is the energy where the conduction band starts. Consider the
density of states we just derived for a 3D (bulk) material in the previous
section, 3D .
3D
1
= 2
2
2me
h
2
129
3
2
130
CHAPTER 9. MORE DENSITY OF STATES
Now rather than have the origin of the energy start at 0, start it where the
conduction band begins (C ). This expression can then be recast as
3D
1
= 2
2
3
2
2me
h
2
(9.3)
Replace this and the Fermi Dirac distribution into the above expression for
the total concentration of electrons in the conduction band to get
ntot =
1
F
kT
1+e
1
2 2
2me
h
2
= A
= A
= A
1+e
1+e
C +C F
kT
C
kT
C d
C d
C d
1
e
F
kT
C d
1+e
F
kT
1+e
3
2
3
2
2me
h
2
1
2 2
C F
kT
C d
(9.4)
Here, let =
C
kT
and =
C +F
leading
kT
to
1
d
ntot = A
3
2
1
where A = 2 2 2me
and A can be found from the above change of
h
2
variables. Note at this point that someone has already solved this integral
and one can just look it up. The integral is called the Fermi integral or
Fermi Dirac integral. It is dened as follows and its solutions labeled F 1 ()
2
can be looked up.
F 1 () =
2
0 1+e d
However, to stay instructive lets just consider the case where F >>
kT . In this case the exponential in the denominator of the Fermi Dirac
distribution will dominate and the expression basically becomes
1
1
1+e
F
kT
F
kT
= e
F
kT
131
which has the Boltzman distribution form. Our expression for ntot becomes
ntot = A
= A
e
e
F
kT
C d
[(C )+(C F )]
kT
(9.5)
C d
Now change variables and let x = C such that = C + xkT and d =

kT
kT dx. Note also that the limits of integration will change accordingly. This
leads to
ntot = A
e kT [(C )+(C F )] (xkT ) 2 kT dx
0
3
= A(kT ) 2 e
C F
kT
ex x 2 dx
The last integral is a common function called the Gamma function. It is

dened as follows
(n) =
ex xn1dx
and its values can be looked up in most reference books. In our case we
3
have 2 . Therefore ntot can be expressed as
3
nC ntot = A(kT ) 2 e
C F
kT
3
2
To be consistent with other texts, we use nC in lieu of ntot at this point

3
and also dene NC = A(kT ) 2 3 . This leads to the common textbook
2
expression
nC = NC e
C F
kT
(9.6)
This is the expression for the eective density of states of the conduction
band.
Density of states in the valence band

The way this is calculated mirrors the approach for the conduction band
with a few minor changes. Again, we need the probability of nding a hole
at a given energy Ph (). Likewise we need the density of states calculated
132
in the previous section. We will calculate the number of holes at a given

energy per unit volume (or concentration of holes at a given energy)
nh () = Ph () ()d
where we use the relationship
Pe () + Ph () = 1
or
Ph () = 1 Pe ()
This lead to
Ph () = 1
1
1+e
(9.7)
F
kT
The density of states is

3D
3
2
2mh
h
2
1
= 2
2
(9.8)
where V is the energy where the valence band starts. Consequently the total
concentration of holes in the valence band is the integral over all energies.
V
ptot =
Ph () ()d
1
1+e
F
kT
For notational convenience let B =

V
ptot = B
2mh
h
2
1
2 2
2mh
h
2
1
2 2
1
1+e
F
kT
3
2
3
2
V d
. This leads to
V d
Since, generally speaking < F

V
ptot = B
1

F
kT
1+e
V d
133
Approximate the stu in parenthesis through the binomial expansion keeping only the rst 2 terms (where |x| < 1).
1
1+x
1
1 x + x2 x3 + x4 + . . .
1 e

F
kT
1+e
F
kT
+ ...
Therefore
1
1 1 e

F
kT
1+e
F
kT
= e
F
kT
Replacing this in the ptot expression gives

V
ptot = B
F
kT
V d
As with the conduction band case earlier, make a change of variables by

letting x = V as well as = V kT x and d = kT dx. Note that the
kT
limits of integration change accordingly resulting in
0
ptot = B
= B
1
e kT [(F V )+(V )] kT x(kT )dx
F V
kT
0
3
= B(kT ) 2 e
V
kT
F
kT
(kT ) 2 x 2 dx
1
ex x 2 dx
Recall that the last integral above is the Gamma function. We get the
following expression upon recognizing this
3
ptot = B(kT ) 2 e
F
kT
3
2
To be consistent with common notation and other texts ptot can be expressed
as
pV ptot = Nv e
3
F V
kT
(9.9)
3
where Nv = B(kT ) 2 2 . The last expression gives the eective density of
states for the valence band.
134
Figure 9.1: Sketch of the electron and hole distribution functions, their
density of states and their population at the conduction and valence band
edges.
135
Summary
Fermi level of an intrinsic semiconductor
If the bulk semiconductor is intrinsic, there has been no doping of the material and hence no extra electrons or holes anywhere. In this situation
nC = pV
where previously we found that
nC = NC e
pV = NV e
C F
kT
F V
kT
3
2
3
3
and NV = B(kT ) 2
2
3
and NC = A(kT ) 2
This leads to
NC e
C F
kT
= NV e
F V
kT
or
Ae
C F
kT
= Be
F V
kT
136
where A =
2me
h
2
1
2 2
3
2
and B =
2
me e
e
e
C F
kT
1
kT
C F
kT
2
= mh e
3
2
and ultimately reduces to
F V
kT
3
2
= mh
C +V
kT
ln e
2mh
h
2
2
me
F V
kT
(C +F +F V )
2F
kT
1
2 2
2F
kT
2F
kT
2F
kT
mh
me
mh
me
mh
me
3
2
3
2
3
2
= ln
mh
me
= ln
mh
me
C +V
kT
3
2
3
2
2F
= kT ln
3
2
mh
me
C +V
kT
C + V
kT
mh
me
kT
ln
2
+ ln e
3
2
+ C + V
+
C + C
2
This yields the nal expression

F =
C +C
2
3
+ 4 kT ln
mh
me
3
2
(9.10)
One can therefore see that at T = 0 the Fermi energy of an intrinsic semiconductor is at the halfway point between the top of the valence band and
the bottom of the conduction band. Note also that since generally speaking
mh > me the temperature dependent term is positive and slowly increases
the Fermi level position with increasing temperature. To a good approximation however, the Fermi level is at the midway point between valence and
conduction bands.
Exercises
Calculate the intrinsic Fermi level of silicon at 0K, 10K, 77K, 300K
and 600K. Note that Eg = Ec + Ev and assume me = 1.08mo and
137
Figure 9.2: Sketch showing the Fermi level position at 0 degrees halfway
between the conduction band and valence band positions.
mh = 0.55mo . Leave the answer in terms of Eg or if you desire look

up the actual value of Eg and express the nal answers in units of eV.
138
Chapter 10
Even more density of states

In the previous section we ran through the calculation for the conduction
and valence band density of states for a 3D (bulk) material. In this section
we repeat the same calculations for materials with lower dimensions. In
particular we do this for 2D and 1D materials that are representative of
quantum wells and quantum wires respectively.
Density of states in the conduction band: 2D

We start with the Fermi Dirac distribution for electrons and also the density
of states (2D ) that we derived in an earlier section. Recall that
2D =
me
h
2
( n )
Consider only one of the subbands. For example, we could choose the rst
band.
In this case the density of states simplies to
2D =
me
h
2
Now recall from the previous section that the number of states at a given
energy per unit volume (or the concentration of electrons at a given energy)
is
ne () = Pe () ()d
The total concentration of electrons in this rst subband is the integral over
all available energies. Rather than use ntot as before lets just stick to nc
139
140
CHAPTER 10. EVEN MORE DENSITY OF STATES
Figure 10.1: Conduction band density of states for a 2D material. The nth
band is shaded to distinguish it from the other subbands.
from the start

nc =
Pe () ()d
(10.1)
C
1
where Pe () =
1+e
is the Fermi Dirac distribution. Putting everything
F
kT
together we get
nC
me
h
2
1
1+e
F
kT
1
1+e
F
kT
me
d
h
2
d
Since the band really begins at n as opposed to C like in the bulk the
integral changes from
leading to
nC =
me
h
2
d
1+e
F
kT
141
if now F >> kT
nC =
me
h
2
F
kT
me kT F
e kT
h
2
d =
This leads to the nal expression for the carrier density of the nth subband,
which in our example we assumed was the lowest one. Hence this is the
carrier density at the conduction band edge of a 2D material.
nC =
me kT
e
h
2
n F
kT
(10.2)
Density of states in the valence band:2D

As with the conduction band case we need the probability of occupying a
given state in the valence band. This is denoted Ph () and is evaluated from
Pe () + Ph () = 1
Ph () = 1 Pe ()
where Pe () is the Fermi Dirac distribution leading to
Ph () = 1
1
1+e
F
kT
The number of states at a given energy per unit volume (concentration at a

given energy) is
nh () = Ph () ()d
m
If, as in the conduction band case, we consider only the nth subband = 2h
h
and for simplicity the rst.
The total concentration of holes in this rst subband is the integral over
all energies. We get
V
pV
Ph () ()d
(10.3)
mh
h
2
1+e
mh
h
2
F
kT
1
1+e
F
kT
142
Figure 10.2: Valence band density of states for a 2D material. The nth band
is shaded to distinguish it from the other subbands.
Since the nth (in our case the rst) subband begins at n rather than V as
in the bulk, the limits of the integral change
V

pV =
mh
h
2
1
1+e
F
kT
Since for the valence band < F

pV =
mh
h
2
1
1 + e
F
kT
Apply the binomial expansion of the term in parenthesis to simplify, keeping

only the rst two terms. Recall that
1
= 1 x + x2 x3 + . . .
1+x

1
F
kT
+...
F = 1 e
1 + e kT
such that
pV =
mh
h
2
F
kT
d =
mh kT F
e kT
h
2
143
This reduces to our desired nal expression
mh kT
e
h
2
pV =
F n
kT
(10.4)
Fermi level position:2D

The procedure for nding the Fermi level position is the same as in the 3D
case. If we assume an intrinsic semiconductor with no doping such that
there are no additional electrons or holes present
nC = pV
then
mh kT F n2
me kT n1 F
kT
= 2 e kT
e
2
h

h

Note that n1 and n2 have been used to distinguish the start of the conduction band subband and the valence band subband respectively. The above
expression reduces
1
e kT (n1 +F +F n2) =
2F
e kT (n1 +n2)+ kT
e
n1 +n2
kT
2F
e kT
2F
e kT
2F
ln e kT
=
=
=
=
2F
kT
2F
mh
me
mh
me
mh
me
n1 +n2
mh
e kT
me
n1 +n2
mh
+ ln e kT
ln
me
n1 + n2
mh
+
ln
me
kT
mh
kT ln
+ n1 + n2
me
resulting in the nal expression

F =
n1 +n2
2
kT
2
ln
mh
me
(10.5)
As before, since mh > me the second term is positive and grows slowly
with increasing temperature. To a rst approximation however the Fermi
level is midway between the valence band beginning and conduction band
beginning.
144
Density of states in the conduction band:1D

We start with the Fermi Dirac distribution for electrons and the density of
states (1D ) that we derived earlier. Recall that
1D =
2me
h
2
m,n
1
( m,n )
m,n
where is the Heavyside function.

We consider only the (m, n)th subbband beginning at energy m,n . For
convenience we could say this is the rst subband.
Figure 10.3: Conduction band density of states for a 1D material. The

(m,n)th band is shaded to distinguish it from the other subbands.
The number of states at a given energy per unit volume (or the concentration at a given energy) is
ne () = Pe () ()d
The total concentration in the rst subband is the integral over all possible
energies
nc =
Pe () ()d
where as before the Fermi Dirac distribution is

1
Pe () =
F
1 + e kT
145
On introducing this, the total expression becomes
nC
1
F
kT
1
2me
d
2
h
m,n
1+e
2me
1
1
d
F
2
m,n
h
C 1 + e kT
Since the band actually begins at m,n the limits of integration become
m,n
Therefore
nC =
2me
h
2
1
1+e
m,n
1
d
m,n
F
kT
If F >> kT
1
1+e
F
kT
F
kT
leading to
nC
=
=
=
=
2me
h
2
2me
h
2
2me
h
2
2me m,n F
kT
e
h
2
F
kT
m,n
1
d
m,n
m,n +m,n F
kT
m,n
m,n
kT
m,n F
kT
1
d
m,n
m,n
kT
1
d
m,n
m,n
1
d
m,n
m,n
Make a change of variables and let x = m,n such that = m,n + kT x and
kT
d = kT dx. Also make corresponding changes to the limits of integration
resulting in
nC =
kT
2me m,n F
kT
e
h
2
ex x 2 dx
146
Recall that the last integral is the Gamma function
1
kT 2me m,n F
kT
nC =
2 e
2
h
The nal expression for the carrier concentration of a 1D material at the

conduction band edge is
nC =
2me kT
e
h
2
m,n F
kT
1
2
(10.6)
Density of states in the valence band:1D

The way this is calculated is very similar to that for the conduction band.
However, a few of the terms change slightly in appearance. As before we
need the probability of occupying a given state in the valence band. We also
need the density of states calculated in the previous section. The occupation
probability is calculated from
Pe () + Ph () = 1
Ph () = 1 Pe ()
Here Pe () is the Fermi Dirac distribution leading to
Ph () = 1
1
1+e
F
kT
The density of states is

1D =
2mh
h
2
m,n
1
(m,n )
m,n
where is the Heavyside function and the band explicitly starts at m,n
causing its presence in the square root denominator. Now if we consider
only the (m, n)th band, (and for convenience that could be the rst one)
the expression simplies to
1D =
2mh
1
m,n
h
2
The number of states at a given energy per unit volume (or concentration
at a given energy) is
nh () = Ph () ()d
147
Figure 10.4: Valence band density of states for a 1D material. The (m,n)th
band is shaded to distinguish it from the other subbands.
The total concentration in the subband is the integral over all energies
V
pV
Ph () ()d
(10.7)
2mh
h
2
1
1+e
F
kT
2mh
1
d
2
h
m,n
1
1+e
F
kT
1
m,n
Since the band begins at m,n rather than at V as in the bulk, the limits of
integration change
V
m,n
(10.8)

pV =
2mh
h
2
m,n
1
1+e
F
kT
1
m,n
Furthermore, since < F

pV =
2mh
h
2
m,n
1

F
kT
1+e
1
d
m,n
148
Using the binomial expansion and keeping only the rst two terms
1
1+x
1

F
kT
1+e
= 1 x + x2 x3 + . . .
= 1 e
F
kT
the term in parenthesis becomes

1
1

F
kT
= e
F
kT
1+e
yielding
m,n
2mh
h
2
pV
2mh
h
2
2mh
h
2
2mh F m,n
kT
e
h
2
=
=
m,n
m,n
e
e
e
F
kT
1
d
m,n
F m,n +m,n
kT
F m,n
kT
m,n
e
e
m,n
kT
m,n
kT
m,n
m,n
1
d
m,n
(10.9)
To simplify, make a change of variables. Let x = m,n or = m,n kT x

kT
and d = kT dx. Note also the corresponding changes in the limits of
integration.
pV
=
=
1 2mh F m,n 0 x
1
kT
e (kT )dx
2 e
kT x
kT 2mh F m,n x 1
kT
e
e x 2 dx
h
2
0
The resulting integral is the now familiar Gamma function. This lead to our
nal expression for the total concentration of holes in the valence band
pV =
2mhkT
e
h
2
F m,n
kT
1
2
(10.10)
149
Fermi level position:1D

This evaluation goes the same was as for the 2D or 3D material. If we are
dealing with intrinsic semiconductors with no additional doping
nC = pV
From what we have just evaluated
1
2me kT
h
2
1
2
m1,n1 F
kT
2mh kT
h
2
1
2
F m2,n2
kT
Here note that m1,n1 and m2,n2 are used to refer to the energies where the
conduction and valence bands begin repsectively. This expression reduces
as follows
m1,n1 F
kT
me e
(m1,n1 +F +F m2,n2 )
kT
m1,n1 +m2,n2
kT
2F
e kT
2F
e kT
ln e
2F
kT
2F
kT
2F
F m2,n2
kT
mh e
mh
=
me
mh
=
me
mh m1,n1 +m2,n2
kT
=
e
me
1
2
mh
= ln
me
+ ln e
m1,n1 +m2,n2
kT
mh 2 m1,n1 + m2,n2
= ln
+
me
kT
kT
mh
=
+ m1,n1 + m2,n2
ln
2
me
This leads to our nal expression

F =
m1,n1 +m2,n2
2
kT
4
ln
mh
me
(10.11)
Since mh > me the second term is positive and as a consequence the Fermi
energy has a slight temperature dependence. However, as before, to a rst
approximation, it can be taken as the midpoint between the conduction
band and valence band energies (the zero temperature limit).
150
Chapter 11
Joint density of states

In the previous sections we have calculated the density of states in both the
conduction band and valence band separately. Interband transitions occur
between both bands giving rise to optical transitions of the semiconductor.
As a consequence it is instructive to calculate the joint density of states
which is proportional to the absorption coecient of the material.
3D bulk
We derive the density of states again as done previously. Consider a spherical
volume of
4
Vk = k 3
3
The volume of a given mode was, recall kx ky kz where
kx =
ky =
kz =
2
Lx
2
Ly
2
Lz
The number of modes or states in the given sphere is then

N =
4
k 3
Vk
= 3 3 Lx Ly Lz
kx ky kz
8
151
152
CHAPTER 11. JOINT DENSITY OF STATES
For an electron, multiply this by 2 to account for spin

N
4
k 3
= 2N = 2 3 3 Lx Ly Lz
8
k3
=
Lx Ly Lz total number of states in sphere
3 2
Consider as before, the density

=
k3
N
= 2 number of states/unit volume
Lx Ly Lz
3
Now consider the energy density

=
d
d
alternatively
d
dk
Note that previously we solved for

d
d
d
dk
d
d .
This time however, lets solve for
d
dk .
=
=
1
2m 2
(previously)
2 2 2
h
k2
(this time)
2
Starting with the energy density

2 =
k2
d
= 2
dk
(11.1)
divide by 2 to go back to only 1 spin orientation since in an optical transition

spin ips are generally forbidden.
1 =
k2
2
= 2
2
2
(11.2)
This expression applies to either conduction band or valence band. Applying

the following equivalence
j ()d = 1 (k)dk
(11.3)
we obtain
dk
d
k 2 dk
2 2 d
j () = 1 (k)
=
(11.4)
153
where j is the desired joint density of states. Now from the conservation of
momentum, transitions in k are vertical such that the intitial k value in the
valence band is the same k value as in the conduction band (ka = kb = k,
where ka is the k value in the valence band and kb is the value in the
conduction band).
Figure 11.1: Cartoon showing momentum conserving vertical transition between the valence band and conduction band in k space
The energy of the initial state in the valence band is
a = v
h 2
2 ka
2mh
Likewise the energy of the nal state in the conduction band is

b = c +
h 2
2 kb
2me
154
The energy of the transition is

= b a
h
2k2
=
c +
2me
= (c v ) +
= g +
h
2k2
2
h
2k2
2mh
h
h
2k2 2k2
+
2me
2mh
1
1
+
me mh
(11.5)
where g = c v . Now from above
me + me
d
= h2 k
dk
me mh
leading to the desired expression

dk
1
= 2
d
h
k
Since j () =
k2
2 2
me mh
me + mh
(11.6)
dk
d
k
1
k2
me mh
me mh
= 2 2
2
2
2
me + mh
me + mh
h
k
2
h
k
=
(11.7)
2 2 2
h
m
where for notational simplicity we have used the reduced mass = mee mhh
+m
Now to continue towards our nal expression, explicitly describe what k is
j () =
= g +
k2 =
h
2k2
2
2( g )
h
2
or
2( g )
h
Replacing this into our main expression for j gives
k
j () =
2 2 2
h
2( g )
=
2 2
2 h
h
2
k=
(11.8)
155
which after some simplication gives
j () =
1
4 2
2
h
2
3
2
(11.9)
Figure 11.2: Joint density of states for a 3D material (bulk)
2D well
As we did before, consider an area in k-space of
Ak = k 2
where the area occupied by a given mode or state is ky kz (we implicitly
assume that kx represents the conned direction)
2
Ly
2
kz =
Lz
ky =
Together, the number of modes in the area is

N=
Ak
k 2
k2
=
Ly Lz =
Ly Lz
ky kz
4 2
4
156
Multiply by 2 to account for spin

N = 2N =
k2
Ly Lz
2
Now consider the density

=
N
k2
=
Ly Lz
2
(11.10)
with the energy density given by

=
d
d
or alternatively
d
dk
Note that previously we solved for

d
d
d
dk
=
=
d
d .
This time, however, lets consider
d
dk .
m
(previously)
2
h
k
(this time)

2 =
d
k
=
dk
divide by 2 to get rid of the spin since formally speaking, spin ip optical
transitions are forbidden.
1 =
k
2
=
2
2
(11.11)
Now applying the following equivalence

j ()d = 2 (k)dk
(11.12)
one obtains
dk
d
k dk
2 d
j () = 1 (k)
=
(11.13)
where j is the desired joint density of states. As before in the 3D case, the
conservation of momentum means that transitions in k-space are vertical.
157
That is, the initial k value in the valence band is the same as the nal k
value in the conduction band (ka = kb = k) where ka (kb ) is the valence
(conduction) band k values.
a = n2
h
2k2
2mh

b = n1 +
h
2 k2
2me
The transition energy is

= b a
= n1 n2 +
= g +
h
2k2
2
1
1
+
me mh
h
2k2
2
(11.14)
where g (eective band gap, not to be confused with the bulk band gap)
m
= n1 n2 and is the reduced mass = mee mhh This leads to
+m
d
h
2k
=
dk
or
dk
= 2
d
h
k
(11.15)
such that when replaced into our main expression the desired expression for
the joint density of states is
j () =
k
2
h
2k
1D wire
Consider the length in k-space
Lk = 2k
2 2
h
(11.16)
158
Figure 11.3: Joint density of states for a 2D material (quantum well)
The length occupied by a given mode or state is kx where

kx =
2
Lx
The number of states in the given length is

N=
Lk
2k
kLx
=
Lx =
kx
2
Multiply this by 2 to account for spin

kLx
total number of states
N =2
Consider the density
=
2k
N
=
number of states per unit length
Lx
Then the energy density is

=
Previously we solved for
d
d
d
dk
d
d
or alternatively
d
dk
d
d
=
=
1 2m 1
(previously)
2
h
2
(this time)
159
2 =
d
2
=
dk
(11.17)
divide by 2 to consider only one spin orientation since spin ip transitions

are generally forbidden.
1 =
2
1
=
2
Now apply the following equivalence

j ()d = 1 (k)dk
leading to
j () = 1 (k)
=
dk
d
1 dk
d
(11.18)
where j () is the desired joint density of states. As in the other two cases,
3D and 2D, the conservation of momentum means that transitions in k-space
are vertical so that ka = kb = k. Here ka is the k value in the valence band
and kb is the k value in the conduction band.
a = m2,n2
h
2k2
2mh

a = m1,n1 +
h
2k2
2me
The transition energy is

= b a
= m1,n1 m2,n2 +
= g +
h
2 k2
2
h
2k2
2
1
1
+
me mh
(11.19)
160
where g = m1,n1 m2,n2 and the reduced mass =
me mh
me +mh .
This leads to
d
h
2k
=
dk
or
dk
= 2
d
h
k
Since j () =
1 dk
d
(11.20)
the joint density becomes

j () =
1
2 k
h
Now to continue towards our nal expression we express k fully. Since

h2 k2
= g + 2 we get
2( g )
h
2
k2 =
2
( g )
h
2
k =
1
k

2 g
This leads to the nal expression for the joint density of states
j () =
1
2 2 g
h
(11.21)
or more directly analogous to last time

=
1
2
2 1
h
2
(11.22)
Note that the singluarities are referred to as Van Hove singularities. In

the case of real quantum wires, they are probably smeared out by Coulomb
eects.
161
Figure 11.4: Joint density of states for a 1D material (quantum wire)
Summary
Now that we have the explicit joint density of states for 3D, 2D and 1D
materials we can summarize their implications for the absorption spectra of
these materials. It can be shown that the absorption coecient and joint
density of states are proportional to each other. The calculation to achieve
this involves Fermis golden rule and will not be discussed at this point. To
summarize
3D: j g
2D: j constant depending on the reduced mass
1D: j
1
g
0D: j ( n )
162
Figure 11.5: Summary of the joint density of states for 3D, 2D, 1D, and 0D
materials
163
ov
164
Chapter 12
Absorption
A geometric derivation of exponential attenuation
Here we derive the standard exponential attenuation law of light being absorbed by a solid. Picture an incident beam of light with an initial intensity
of Io . Picture next a bulk solid which can be divided into very small slices
or slabs of thickness x. The whole solid consists of n slabs placed back
to back such that the total thickness of the solid is x.
Next, each slab absorbs some light and transmits the rest. In each slab
a fraction (1 f ) of the impinging light is transmitted. Here 0 <= f <= 1.
So initially we have Io . After the rst slab you have a residual intensity
of I = (1 f )Io. After the second slab you have a residual intensity of
I = (1 f )2Io and so on.
The general expression for the residual intensity after n slabs is therefore
I = (1 f )nIo
To derive the exponential attenuation law, re-arrange this into the following
I
= (1 f )n
Io
This is the fraction of the incident intensity that makes it through n slabs.
Recall now that x = nx giving
x
I
= (1 f ) x
Io
165
166
CHAPTER 12. ABSORPTION

Next, to simplify the notation let p = (1 f ) yielding
x
I
= p x
Io
where 0 1. Now note that p is a constant. In addition, x is a

p
constant. Therefore, call the whole thing some other constant, say y.
I
= y x
Io
where recall that y can be expressed alternatively as y = elny . This gives
I
= elny
Io
Next, since y is a constant, so is lny. Call it giving

I
= ex
Io
where > 0. This leads to the popular expression for the exponential
attenuation of light in a solid.
I = Io ex
(12.1)
where is called the absorption coecient.
There is an alternative derivation of the exponential attenuation law. The
argument goes as follows:
(a) The change in intensity of the light is proportional to the thickness
of the solid.
I x
167
The negative sign is to indicate attenuation of the light in the solid through
absorption.
(b) Next, the absolute change in intensity on attenuation is also proportional to the intensity of the light.
I I
(c) As a consequence, the change in intensity is also proportional to the
product of both independent variables
I Ix
Call the proportionality constant giving
I = Ix
I
= I
x
Next, if one lets I and x become innitesimally small one gets
dI
dx
= I
(12.2)
which is a rst order dierential equation that is easily solved.

dI
I
lnI
= dx
= x + const
or
I = Io ex
(12.3)
This is our exponential attenuation law again.
Conventions
In dealing with the absorption of materials there are various conventions.
In bulk solids it is common to see
I = Io el
(12.4)
where is the absorption coecient and is generally expressed in units

of cm1 when l is expressed in cm units.
168

In more molecular-like systems it is common to see
I = Io enl
(12.5)
where is the cross section having units of cm2 and n is the concentration
of absorbers in the sample in units of number/per cm3 and l is the pathlength
expressed in units of cm.
By comparison one can see that if one needs to relate and
= n
(12.6)
but this shouldnt be necessary since in one case one has a bulk solid and in
the other case the assumption is of a molecular-like material (quantum dots
for example).
Finally, when dealing with molecular-like materials it is common to see
the everyones favorite Beers law.
A = cl
(12.7)
where A is the absorbance or optical density (OD), is the molar extinction coecient in units of (M 1cm1 ), c is the concentration in units of
(M) and nally l is the pathlength in units of cm.
Sometimes it is of interest to convert between extinction coecient ( )
and the absorption cross section (). This can be achieved in the following
fashion
A = cl
1
where we also know that A = log( T ) and where T =
of the sample. Thus
A = log(
I
Io
is the transmittance
Io
) = cl
I
and from the exponential attenuation law we know that

I
= enl
Io
Note that c is in units of M and n is in units of number/cm3 .
logenl =
cl
(nl)loge =
cl
169
Drop the l terms on either side since they are both the same.
(n)loge = c
Lets now convert c to the same units as n which is in terms of number of
absorbers per cm3 .
c(mol/l)
cNa
dm3
where recall that moles times Avogadros number is the number of absorbers
and that 1l = dm3 . Next convert from dm to cm by recalling that 1dm =
10cm.
c(mol/l)
cNa
=n
1000cm3
therefore
c=
n1000
Na
This means that our previous equivalence becomes

(n)loge = n
1000
Na
or
=
=
where the 2.303 comes from
(1000)
loge(Na )
(2.303)(1000)
Na
(12.8)
1
loge .
Exercises
1. A solution with a dye concentration of c = 4 106 M has a molar
1
absorptivity of = 1.5 105 M cm . It is illuminated with green laser
light at = 514.5 nm and with a power of P = 10nW . (a) What
is the percentage of light absorbed by the sample after 1 micron. (b)
Calculate the number of photons per second absorbed by the sample.
(c) Calculate the cross section of the molecule in units of cm2 and 2 .
A
170
2. diIC12 is a popular dye in single molecule experiments. If one wants

to image individual diI molecules one needs suciently low coverages
on a microscope coverslip. Lets say no more than 10 molecules per
square micron. (a) If 10 microliters of a diI solution spread evenly on
a 4 mm diameter region achieves this coverage, what is the molar concentration of the diI solution. (b) This diI sample is then illuminated
with 543.5 nm (a green HeNe). A uniform 100 nm diameter spot is
illuminated with an excitation rate of 3 1010 photons per second.
What is the incident excitation power, P? (c) Next, a single diI molecule being illuminated absorbs 2.3 105 p/s. Calculate the absorption
cross section and the molar extinction coecient of diI (a 1 cm cell
may be assumed).
3. Colloidal CdSe QDs overcoated with ZnS are all the rage today. Look
up a paper by XG Peng in Chemistry of Materials and from his data
provide the range of absorption cross sections () at the band edge
for QD sizes between 2 nm and 6 nm in diameter. Express the same
range in 2 and then compare this to the physical size of the dot.
A
171
ov
172
Chapter 13
Interband transitions
Interband absorption in semiconductors
Although we havent discussed the concept of bands yet, there are so-called
bands in semiconductors, namely the conduction band and the valence band.
These bands are composed of accessible states to the electron and hole. We
generally speak of electrons in the conduction band and holes in the valence
band. The transition which promotes an electron from the valence band to
the conduction band through the absorption of a photon is referred to as an
interband transition. Alternatively, transitions within a given band are
called intraband transitions. Now in what follows we will assume that we
are dealing with direct gap semiconductors.
The transition rate for absorption is given by Fermis (second) Golden
Rule. Note this is also sometimes abbreviated and expressed as the Golden
rule. Its basically a quantum mechanical way to derive a rate for a process
and is analogous to a classical rate. The initial state (i) to nal state (f)
transition rate, Rif , given by Fermis Golden Rule is
Rif =
2
2
h
|M | j (
(13.1)
where |M | is a matrix element linking the initial and nal states and j ( ) is
the joint density of states. Here is where your density of states calculations
start to pay o. Notice that the transition rate is proportional to the joint
density of states which you have just calculated for bulk, 0D, 1D and 2D
materials.
173
174
CHAPTER 13. INTERBAND TRANSITIONS
Time dependent perturbation theory

We will work in what is referred to as the Schrodinger representation. Note
for future reference that there are three picture of quantum mechanics, the
Schrodinger representation, the Heisenberg representation and the interaction representation. Usually when doing time dependent perturbation theory, the interaction representation is preferred. But we will get the same
expressions this way as well.
The general Schrodinger equation is actually written as
i
h
2
(r, t) = H(r, t)
t
(13.2)
h
where H = 2m + V and V is a potential perturbation which can be time
dependent. Note that the wavefunctions depend on time now. We look for
solutions of the form
(r, t) = (r)f (t)
(13.3)
To keep the notation more compact, reduce this to (r, t) = f . Replace

this into the general time dependent Schrodinger equation.
i
h
(f ) =
t
i
h
h
2 2
+V
2m
h
2 2
f + V f
2m
h
2
=
f + V f
2m
f
t
i f
h
Consolidate the f and terms giving

i
h
h
2
f
=
+V
f
2m
or
i f =
hf
2m + V
(13.4)
You will note that the left hand side is dependent only on time and that the
right hand side is dependent only on position. To be equivalent both sides
must equal a constant which we call . (turns out to be the eigenenergy or
eigenvalue)
175
Left hand side
i
h
f
f
f
f
=
i
h
which we integrate to get

i t
=
h
lnf
The desired time dependent function is

f (t) = e
i t
h
(13.5)
Right hand side

1
h
2
+V
2m
h
2
+V =
2m
h
2
2m
+V
or
H0 =
(13.6)
This is our usual eigenvalue and eigenvector problem. Solving this we get
and (r).
Putting everything together, the general form of the wavefunction is
(r, t) = (r)f (t)
or more explicitly
(r, t) = (r)e
i t
h
(13.7)
176
Time dependent perturbation theory now

The basic idea of time-dependent perturbation theory is simple. First understand that if the perturbation (call it V ) were absent, the eigenfunctions
appropriate to the problem would be given by the equation
H (0) (0) =
(0)
(0)
If the initial state is expanded in terms of these unperturbed eigenfunctions

Cn e
(t = 0) =
(0)
i n t
h
(0) (r)
n
where Cn is a constant, then in the absence of a perturbation we would have

for all times
Cn e
(t > 0) =
(0)
i n t
h
(0) (r)
n
where again Cn is a constant.

If the perturbation is present however, the above linear combination is
no longer a valid solution to the Schrodinger equation.
But we can still express whatever true solution exists as a linear combination of the un-perturbed zero order wavefunctions provided that the
coecients are now time dependent.
Cn (t)e
(r, t) =
(0)
i n t
h
(0) (r)
n
Cn (t)n(r, t)
n
So lets go ahead and start deriving some results of the time dependent
perturbation theory.
Plug our time dependent form of the wavefunction back into the general
time dependent Schrodinger equation.
i
h
d(r, t)
= H(r, t)
dt
where H = H (0) + V (t) or H (0) + H (1) (t). Use whatever notation that suits
you. We then have
i
h
d(r, t)
= H (0) + H (1) (r, t)
dt
177
Ok since the notation gets wild, were going to drop the (t) in Cn (t). Just
understand that the time dependence is implied. Furthermore, just work
with n (r, t) instead of the full expression.
H (0) + H (1)
Cn n (r, t) = i
h
n
Cn n (r, t) + H (1)
H (0)
Cn n (r, t) = i
h
n
d
dt
d
dt
Cn n (r, t)
n
Cn n (r, t)
n
Expand the right hand side using the Chain rule.

H (0)
Cn n (r, t) + H (1)
n
= i
h
Cn
n
dn (r, t)
+
dt
dCn
n (r, t)
dt
Cn n (r, t) + H (1)
H (0)
Cn
n
Cn n (r, t)
n
= i
h
Cn n (r, t)
n
dn (r, t)
+ i
h
dt
dCn
n (r, t)
dt
The rst terms on the left and right side of the last expression will cancel.
To illustrate let n = 1 for example.
H (0) C1 1 (r, t) = i C1
h
H (0) C1 1 (r)e
(0)
t
1
h
d1 (r, t)
dt
= i C1 1 (r)e
h
(0)
t
1
h
(0)
1
cancel common terms

H (0) 1 (r) =
(0)
1 1 (r)
which is our favorite eigenvalue and eigenvector relation. The sample applies
for all n so that we are left with
H (1)
n Cn n (r, t)
= i
h
dCn
n dt n (r, t)
(13.8)
Now we desire to solve this expression for the time dependent coecients
of some state say k.
178

(0)
Multiply both sides by k

same basis are orthogonal.
(0)
(r)H (1)
(r) and recall that all states n = k of the
Cn n (r, t) = i
h
n
(0)
Cn k (r)H (1) n (r, t)
= i
h
n
(0)
t
i
n
h
(0)
Cn k (r)H (1) (0) (r)e
n
= i
h
dCn (0)
(r)n (r, t)
dt k
dCn (0)
(r)n (r, t)
dt k
(0)
i n
dCn (0)
(0)
h
(r)n (r)e
dt k
where n = k on right side

Cn < k|H
n
(1)
|n > e
i
Multiply both sides by e
to simplify.
(0)
t
k
h
= i
h
dCk
dt
(0)
(0)
n )t
k
h
= i
h
dCk
dt
(0)
i n t
h
Cn < k|H (1) |n > e
n
(0)
(0)
k n
(0)
dCn i k
e h
= i
h
dt
Cn < k|H (1)|n > e
Let kn =
(0)
t
k
h
(0)
i n t
h
i(
since = h yielding
Cn < k|H (1) |n > eikn t = i

h
dCk
dt
or better yet
i dCk =
h dt
Cn < k|H (1) |n > eikn t
This is basically a system of simultaneous linear

In matrix notation

C1
V11
V12 ei12 t
C2 V21 ei12 t
V22
d
i C3 = . . .
h
...
dt
...
...
...
Ck
...
...
(13.9)
dierential equations.

...
C1
. . . C2

. . . C3

. . . . . .
...
...
179
Now what? This expression is pretty tough to solve. So to overcome this
problem you apply time dependent perturbation theory (which is basically
making some assumptions about the size of the perturbation and in turn
what C1 , C2 , C3 , . . . are).
To illustrate, suppose the system starts out in state 1. Then
C1 (t = 0) = 1
C2 (t = 0) = 0
C3 (t = 0) = 0
and so forth.
If there are no perturbations, then the system stays in state 1 forever.
Our zeroth order approximation is therefore
C1 (t) = 1
C2 (t) = 0
C3 (t) = 0
...
Cn (t) = 0
To get our rst order approximation insert these zeroth order values into
our matrix yielding
i
h
dCk
dt
dCk
dt
= Cn=1 < k|H (1)|n = 1 > eikn=1 t

=
1
Cn=1 < k|H (1)|n = 1 > eikn=1 t
i
h
where cn=1 1
yielding
Ck (t) =
1
i
h
t
0
< k|H (1)|n = 1 > eikn=1t dt
Note however that when integrating

integration that comes out.
Ck (t) = const +
1
i
h
t
0
dCk
dt
(13.10)
there is actually a constant of
< k|H (1)|n > eikn t dt
where const = Ck (t = 0) = kn . Replacing this into the above expression

gives
Ck (t) = Ck (0) +
1
i
h
t
0
< k|H (1) |n > eikn t dt
180
or your standard textbook expression

Ck (t) = kn +
1
i
h
t
0
< k|H (1)|n > e1kn t dt
(13.11)
Summary
(0)
Ck (t) = kn
(1)
Ck (t) =
1
i
h
t
0
(13.12)
< k|H (1) |n > e1kn t dt
(13.13)
Two level system, sinusoidal perturbation, Schrodinger

representation
Lets illustrate our results with a more concrete example and for simplicity
keep it to two levels (a two level system, TLS).
Recall that our time dependent Schrodinger equation is
d
(r, t) = H(r, t)
dt
where H = H (0) + H (1)
d
i (r, t) = (H (0) + H (1) )(r, t)
h
dt
i
h
Now recall that a general solution to this time dependent Schrodinger

equation can be expressed as a linear combination of the unperturbed zero
order wavefunctions, provided time dependent coecients.
(r, t) = C1 (t)1 (r, t) + C2 (t)2 (r, t)
where
1 (r, t) = 1 (r)e
2 (r, t) = 2 (r)e
(0)
i 1 t
h
(0)
t
2
h
Now lets be more specic with what H (1) (t) looks like. We want to
consider the interaction of a molecule with dipole moment with light where
the electric eld of light can be written as
= Eo cos2t
181
where is the frequency of the radiation. Alternatively
= E0 cost
(13.14)
This is going to be our sinusoidal perturbation. You can see that it will
oscillate periodically with time.
The interaction between a molecules dipole moment and the electric
eld is then
H (1) =
H (1) = E0 cost
(13.15)
H = H (0) E0 cost
(13.16)
Such that
We are now going to re-derive our previous expression for the coecients.
Insert (r, t) into the Schrodinger equation. Also for shorthand let the time
dependence of the coecients be understood C1 (t) C1 and C2 (t) C2 .
d
[C1 1 (r, t) + C2 2 (r, t)]
dt
= (H (0) + H (1) ) [C1 1 (r, t) + C2 2 (r, t)]
i
h
d
d
(C1 1 (r, t)) + i (C2 2 (r, t))
h
dt
dt
(0)
(1)
= (H + H ) [C1 (r, t) + C2 2 (r, t)]
i
h
d1 (r, t)
dC1
d2 (r, t)
dC2
+ i 1 (r, t)
h
+ i C2
h
+ i 2 (r, t)
h
dt
dt
dt
dt
= H (0) (C1 1 (r, t) + C2 2 (r, t)) + H (1) (C1 1 (r, t) + C2 2 (r, t))
i C1
h
d1 (r, t)
d2 (r, t)
dC1
dC2
+ C2 i
h
+ i 1 (r, t)
h
+ i 2 (r, t)
h
dt
dt
dt
dt
(0)
(0)
(1)
(1)
= C1 H 1 (r, t) + C2 H 2 (r, t) + C1 H 1 (r, t) + C2 H 2 (r, t)
C1 i
h
Weve seen previously that the rst terms on either side of the last equation
will cancel since they are equivalent.
d1 (r, t)
dt
d2 (r, t)
H (0) 2 (r, t) = i
h
dt
H (0) 1 (r, t) = i
h
182
leaving
h
i 1 (r, t) dC1 + i 2 (r, t) dC2 = C1 H (1) 1 (r, t) + C2 H (1) 2 (r, t) (13.17)
h
dt
dt
(0)
Now like before, multiply both sides by 2
(r) and integrate.
dC2
dC1
(0)
+ i 2 (r)2(r, t)
h
dt
dt
(0)
(0)
(1)
(1)
= C1 2 (r)H 1 (r, t) + C2 2 (r)H 2 (r, t)
(0)
i 2
h
(r)1(r, t)
i < 2|1 > e

h
(0)
i 1 t
h
(0)
i 2 t dC
dC1
2
+ i < 2|2 > e h

h
dt
dt
= C1 < 2|H (1) |1 > e
i e
h
(0)
i 2 t
h
dC2
dt
= C1 < 2|H
(1)
|1 > e
(0)
i 1 t
h
(0)
i 1 t
h
+ C2 < 2|H (1) |2 > e

+ C2 < 2|H
multiply both sides by e

i
h
dC2
dt
= C1 < 2|H (1) |1 > e
as before let 21 =
(0)
(0)
2 1
(1)
|2 > e
(0)
i 2 t
h
(0)
i 2 t
h
(0)
i 2 t
h
(0)
(0)
i( 2 1 )t
h
+ C2 < 2|H (1) |2 >
. The exact expression we get then is
i dC2 = C1 < 2|H (1) |1 > ei21 t + C2 < 2|H (1) |2 >
h dt
(13.18)
Now we invoke time dependent perturbation theory and assume that

since H (1) is small
C1 (t)
C2 (t)
giving
i dC2 =< 2|H (1) |1 > ei21 t
h dt
(13.19)
What now? Well, we havent really dealt with H (1) yet. Recall that
H (1) = Eo cost
183
For convenience, choose the z direction
(1)
Hz = Eo cost
or
(1)
Hz
eit +eit
2
= z Eo
(13.20)
Ok, lets throw this into our previous expression and see what happens.
dC2
dt
dC2
i
h
dt
dC2
i
h
dt
i
h
= < 2|
z Eo it
(e + eit )|1 > ei21 t
2
Eo
< 2|z (eit + eit )|1 > ei21 t
2
Eo
=
< 2|z |1 > eit+i21 t + eit+i21 t
2
=
(13.21)
The term < 2|z |1 > is the transition dipole moment.
dC2
Eo
=
< 2|z |1 > ei(+21 )t + ei(21 t)
dt
2
dC2
iEo
=
< 2|z |1 > ei(+21 )t + ei(21 )t
dt
2
h
Integrate this to nd C2 (t) but before we do this, lets ip back to notation
so that our nal expression will be consistent with everyone elses.
i
h
h+ 21
h
h 21
h
dC2
iEo
=
< 2|z |1 > ei h t + ei h t
dt
2
h
where = h and h21 = h2 1 = 2 1
h
h+( 2 1 )
h 2 + 1
iEo
dC2
t
h
=
< 2|z |1 > ei
+ ei h t
dt
2
h
dC2
dt
iEo
2
h
< 2|z |1 > ei
( 2 1 +h)
t
h
+ ei
( 2 1 h)
t
h
(13.22)
Now integrate this to get C2 (t). Remember that when you integrate
you get an additional constant of integration which is C2 (0).
t
iEo < 2|z |1 >

2
h
where C2 (0) = 12
C2 (t) = C2 (0) +
C2 (t) = 12 +
iEo < 2|z |1 >

2
h
+e
i( 2 1 h)t
h
0
t
e
0
i( 2 1 +h)t
h
i( 2 1 +h)t
h
+e
i( 2 1 h)t
h
dt
dt
dC2
dt
184
but since states 1 = 2 we are left in practice with the 1st order contribution.
So lets just work with it.
(1)
C2 (t) =
t i(
0e
iEo <2|z |1>

2
h
2 1 +h)t
h
+e
i( 2 1 h)t
h
dt
(13.23)
So go on and integrate this expression.

(1)
C2 (t)
i( 2 1 +h)t
h
Eo < 2|z |1 > (e

=
2
i(
i( 2 1 h)t
h
1)
h (e
+
i(
2 1 + h)
1)
h
2 1 h)
or
(1)
C2 (t) =
Eo <2|z |1>
2
(e
i( 2 1 +h)t
h
1)
( 2 1 +h)
(e
i( 2 1 h)t
h
1)
( 2 1 h)
(13.24)
Now at this point, another approximation is made here.

You have two terms. The one with ( 2 1 h) in the denominator
is called the resonant term. The one with ( 2 1 + h) is called the nonresonant term.
When h = ( 2 1 ) the resonant term will dominate. Under this resonance condition where the monochromatic light energy equals the energy
dierence between states 1 and 2 you get
(1)
C2 (t)
Eo <2|z |1>
2
i( 2 1 h)t
h
(13.25)
( 2 1 h)
Note that this approximation is also called the Rotating Wave Approximation.
i( 2 1 h)t
2
h
Since we like math, lets keep simplifying. Extract out e
(1)
C2 (t)
(1)
C2 (t)
Eo < 2|z |1 > i( 2 1 h)t

2
h
e
2
iEo < 2|z |1 > i( 2 1 h)t
2
h
e
( 2 1 h)
i( 2 1 h)t
2
h
i( 2 1 h)t
2
h
e
2 1 h
i( 2 1 h)t
2
h
e
2i
i( 2 1 h)t
2
h
resulting in
(1)
C2 (t)
iEo <2|z |1> i(

( 2 1 h) e
2 1 h)t
2
h
sin
2 1 h
2
h
(13.26)
185
Ok, lets keep going. Recall that the probability of being in a particular
state k is basically
Pk = |Ck |2
Therefore the transition probability into the nal state 2 is
P2 = |C2 (t)|2
2
Eo | < 2|z |1 > |2
sin2
=
( 2 1 h)2
h
2
h
1
The nal transition probability into state 2 is

2
P2 = Eo | < 2|z |1 > |2
h)t
sin2 ( 2 1h
)
2
( 2 1 h)2
(13.27)
Note that P2 is P2 (h, t). It depends on both time and frequency.

In general express this as
Pk =
2
t
Eo |<k|z |n>|2 sin2 kn
2
(kn )2
h
2
(13.28)
since = h.
h
Fermis Golden Rule

Recall that we are dealing with a harmonic perturbation of the form
H (1) = z Eo cost
(13.29)
H (1) = z Eo sint
(13.30)
Note that it doesnt matter which one we use, both work ne.
From our previous work we derived
Pk =
2
Eo | < k|z |n > |2 sin2 kn t
2
(kn )2
h
2
where again kn = (k n ).
Like above, we want to consider the transition probability and more
specically the rate of transition into a group of nal states k (not just into
one state).
186
So we integrate over the spread of nal states dictated by the density

( k ) or (k ).
Pk (k , t)(k )dk
Pk,group =
Pk,group =
(13.31)
Eo | < k|z |n > |2 sin2 kn

2
(k )dk
(kn )2
h
2
Eo | < k|z |n > |2 sin2 k n
2
(k )dk
(k n )2
h
2
Now as before, assume that (k ) is constant or rather that the distribution

is uniform. It can be removed from the integral.
Pk,group =
2
Eo (k )| < k|z |n > |2
h
2
sin2 k n t
2
dk
(k n )2
Let x = (k n ) and dx = dk . Also remember to change the limits

of integration. We then get
Pk,group =
2
Eo (k )| < k|z |n > |2
h
2
sin2 xt
2
dx
x2
As before, from a table of integrals the denite integral is
sin2 xt
t
2
dx =
x2
2
The resulting transition probabilty is then

Pk,group =
2
Eo | < k|z |n > |2 t
2
h
2
Also like before, rather than stop here, ip (k ) into ( k ) to keep the
notation consistent with other texts ((k ) = h( k )).
Pk,group =
2
Eo (
2
h
k)
| < k|z |n > |2 t
(13.32)
This is the desired form of the transition probability.

dP
Now the desired rate is R = k,group .
dt
R=
2
2 Eo ( k )|
h
< k|z |n > |2
(13.33)
Again, the rate is constant with respect to time. It is the quantum mechanical version of a classical rate. This is called Fermis Golden Rule.
187
Putting it all together

Interband transitions (or a whole lot of work)
There are various ways to proceed with this. We carry out the following
derivations in a solid state physics approach to stay consistent with existing
texts.
For the interaction of light and matter we have seen the Hamiltonian
H (1) = E
One can also express this interaction through the following Hamiltonian
H (1) =
q
(A P )
m
where A is the vector potential associated with the magnetic eld of the
electromagnetic wave and P is the quantum mechanical operator.
Proceed using the second Hamiltonian. It can be shown that
A = Ao cos(t kr)
where we ignore a potential phase factor, , and A = Ao e where the second
Ao is a scalar amplitude and e is a unit vector showing the orientation of
the electric eld polarization.

Then
q
Ao ecos(t kr)P
m
q
e
= Ao cos(t kr)( P )
m
qAo i(tkr)
=
e
+ ei(tkr) ( P )
e
2m
H (1) =
Next it can be shown that generally the transition probability can be

expressed as (this is best done through the interaction representation)
Pk =
1
h
2
t
to
dt eikn t < k|H (1)|n >
(13.34)
where k is the nal state (the conduction band in our case) and n is the
initial state (the valence band in our state).
188
We will shove into this transition probability the above expression for
H (1) to get
t
1
h
2
to
1
h
2
Pk =
q 2 A2
o
4m2
dt eikn t
k| ei(t kr) + ei(t kr) ( P )|n

e
2
i(t kr)
ikn t
dt e
k| e
i(t kr)
+e
to
( P )|n
e
Next there are two terms sandwiched by < k| and |n >. These are the
terms ei(t kr) and ei(t kr) . The one on the left will eventually lead to
a non-resonant term related to stimulated emission. The second term on
the right will lead o the absorption term that we are interested in. This is
the resonant term. So to simplify life just drop the rst term. We now
have
1
h
2
Pk
q 2 A2
o
4m2
k| ei(t kr) ( P )|n

e
eikn t
to
Pull out the time dependent term from the bra-ket since this is only a
spatial integral.
1
h
2
q 2 A2
o
4m2
1
h
2
Pk
q 2 A2
o
4m2
dt eikn t it
to
t
k|eikr ( P )|n
e
2
i(kn )t
dt e
to
ikr
k|e
( P )|n
e
Next we have the Bloch forms of the wavefunctions

1
|k > = uk (r)eikk r
N
1
|n > = un (r)eikn r
N
where 1 is a normalization constant reecting normalization of both waveN
functions over the entire volume of the solid being considered. N is the
number of unit cells in the crystal considered.
Pk
1
h
2
q 2 A2
o
4m2
dt ei(kn )t
to
1
N
2
V
uk eikk r eikr ( P )un eikn r d3 r

e
where the interior integral is over the entire volume of the solid considered.
Pj
1
h
2
q 2 A2
o
4m2
1
N2
t
to
dt ei(kn )t
V
uk eikk r eikr ( P )un eikn r d3 r

e
189
Now in principle, P operates on both un and eikn r but since we understand
that eikn r varies very slowly, we (as an approximation) ignore P operating
on eikn r .
Pk
1
h
2
q 2 A2
o
4m2
1
N2
dt ei(kn )t
to
ei(k+kn kr )r uk ( P )und3 r
e
Next ei(k+kn kr )r is a slowly varying term that can be approximated by a

constant in a given unit cell. We approximate this now as
Pk
1
h
2
q 2 A2
o
4m2
1
N2
i(kn )t
i(k+kn kr )r
dt e
to
uk ( P )un d r
e
We take what is sometimes called the long wavelength approximation eikr

1 and get
Pk
1
h
2
q 2 A2
o
4m2
1
N2
dt ei(kn )t N
V
uk ( P )und3 r
e
Next cancel common terms and choose a projection for P , say Px (P ) = Px

e
giving
Pk
q 2 A2
o
4m2
1
h
2
i(kn )t
dt e
< k|Px|n >
to
Extract the spatial integral since it does not depend on time.

Pk
q 2 A2
o
4m2
1
h
2
| < k|Px|n > |2
dt ei(kn )t
to
where
t
i(kn )t
dt e
to
ei(kn )t
=
i(kn )
to
Let to = 0 to get
ei(kn )t 1
i(kn )
Re-express this as
e
i(kn )t
2
i(kn )t
2
i(kn )
i(kn )t
2
190
or as
2e
i(kn )t
2
sin
(kn )t
2
(kn )
Put it all together now to get

Pk
q 2 A2
o
4m2
o
1
h
2
| < k|Px |n > | 2e
i(kn )t
2
sin kn t
2
(kn )
This simplies to
1
h
2
Pk
q 2 A2
o
m2
| < k|Px|n >
|2
kn
2
(kn )2
sin2
(13.35)
Now to account for the available states we eventually integrate over the joint
density of states. However, rst convert the transition probability over from
frequence to energy using = h
1
h
2
Pk ( )
q 2 A2
o
m2
| < k|Px |n > |2
sin2 knh t 2
2
h
( kn )2
giving
Pk ( )
q 2 A2
o
m2
| < k|Px|n > |
kn
2
2 sin
2
h
2
kn )
(13.36)
Next at long enough times (t ), we can simplify the last term above
into a delta function using
limt
sin2 at
= (a)
a2 t
So that
t
4 2
h
sin2
kn
t
2
h
2
kn
t
2
h
4 2
h
2
h
kn
(13.37)
giving
Pk ( )
q 2 A2
o
m2
t
4 2
h
| < k|Px|n > |2
2
h
kn
191
h
where knh = 2 ( kn ).
2
The desired transition probability is
2
h
Pk
q 2 A2
o
4m2
| < k|Px|n > |2 (
kn
)t
(13.38)
The transition rate, R, is the time derivative of the probability, R =

giving
2
h
R=
q 2 A2
o
4m2
| < k|Px |n > |2 (
kn
dPk
dt
(13.39)
Finally, we integrate the rate expression over the joint density of states
Rtot =
q 2 A2
o
4m2
2
h
| < k|Px|n > |2
j ( )(
kn
)d
(13.40)
In the 3D bulk case we have

1
j,3D ( ) = 2
4
3
2
2
h
2
If we account for the degeneracy of the spin, multiply the above expression
by two
1
j,3D ( ) = 2
2
3
2
2
h
2
Shove this into the above expression for the total transition rate (per
unit volume) to get
q 2 A2
o
4m2
o
2
h
2
Rtot
2
h
q 2 A2
o
3m2
o
| < k|Px|n > |

| < k|Px|n > |
1
2 2
1
2 2
2
h
2
2
h
2
3
2
3
2
g ( kn
)d
g ( kn
)d
The nal expression we have been waiting for is

Rtot =
2
h
q 2 A2
o
4m2
o
| < k|Px|n > |2
1
2 2
2
h
2
3
2
(13.41)
This is the total transition rate per unit volume of a bulk 3D material
including spin degeneracy.
192
The absorption coecient

Now lets close the loop and connect this total rate to the absorption coefcient.
The number of photons per unit area per unit time is
np =
I
h
(13.42)
and the absorption coecient can be expressed as

=
h
Rtot
Rtot
=
np
I
(13.43)
Next we need an alternative expression for the incident intensity, I.

I=
nr c o 2 A2
nr c o E 2
o
=
2
2
(13.44)
This was obtained from

1
EB
o
I=
but in many cases (nonmagnetic materials)

I=
o giving
1
E B
where B = n E. This leads to

c
I =
=
1 n
E2
c
1 n 1 2
E
c 2 o
1
1
1
and since c2 = o1 o we have c = c o o or c o = co
c and nally c =
Replace the last expression into our main equation to get
I=
nc o 2 A2
o
2
1
c o.
(13.45)
Then using this alternative expression for I, replace it into the expression
for above.
=
2
h
Rtot
nc o 2 A2
o
193
After simplifying this you get
=
q2
2m2 c
o
1
n
2
h
2
3
2
| < k|Px |n > |2
(13.46)
Often there is a change of notation. Let

Ep =
2|<k|Px |n>|2
mo
(13.47)
which has a value of approximately 20 eV for many semiconductors. The

nal expression is
3D ( ) =
q2
h
4mo c o n
Ep
h
2
h
2
3
2
(13.48)
The absorption spectrum therefore is shown to be proportional to the joint

density of states. This is the 3D case.
194
Chapter 14
Emission
In the last section, the absorption coecient of materials with dierent dimensionality were shown to be proportional to the calculated joint density
of states. In turn, one could predict that the absorption spectrum of these
materials would look like. Here we work out the background behind spontaneous emission, a complementary process to absorption.
Preliminaries: Einstein A and B coecients

Picture a two level system like that shown in the gure.
Here
N1 =population in the ground state or alternatively probability of
being in the ground state
N2 =population in the excited state, alternatively probability of being
in the excited state
g1 =degeneracy of ground state
g2 =degeneracy of excited state
=the energy density containing thermal and/or external contributions
Three processes were considered
1. absorption with total rate: B12 N1
2. stimulated emission with total rate: B21N2
195
196
CHAPTER 14. EMISSION
Figure 14.1: Two level system showing various transitions considered for
Einstein A and B coecients
197
3. spontaneous emission with total rate: AN2
Putting everything together one obtains the following rate equations for
populating either the ground or excited states
dN1
dt
dN2
dt
= B12 N1 + B21 N2 + AN2

= B12N1 B21 N2 AN2
In equilibrium, the upward and downward rates are equivalent ( dN1 =

dt
0). So using either of the above expressions
dN2
dt
B12 N1 B21N2 AN2 = 0

or
B12 N1 (upward) = B21 N2 + AN2 (downward)
solving for gives
=
=
AN2
B12N1 B21 N2
A
N1
B12 N2 B21
where N1 and N2 are Boltzman distributed

N1 =
N2 =
1
N
1
N
g1
e
i=1
g2
1 F
kT
g1 1 0
e kT
N
2 F
kT
g2 2 0
e kT
N
i=1
and N is a normalization constant. This leads to

g1 1 2
N1
g1 h
= e kT = e kT
N2
g2
g2
where h = 2 1 is the energy of the photon or transition. Replace this
into the above expression for giving
() =
A
h
g1
B12 g2 e kT B21
(14.1)
198
This is the expression for the energy density derived by Einstein. Next he
realized that this expression had to be equivalent to the Planck distribution for blackbody radiation. Expressed in terms of wavelength, the Planck
distribution is (standard textbook expression)
p () =
8hc
5
1
hc
e kT 1
(14.2)
Units: J/(unit volume unit wavelength). This expression is interesting from

a historical point of view because when the derivative of this expression
is set to zero one obtains the Wein displacement law for blackbody radiation. Likewise the integral of this expression gives whats called the Stefan
Boltzman law for blackbody radiation. In a sense, knowing or unknowingly,
Planck basically explained everything in one shot. Apparently he nearly
suered a nervous breakdown doing it.
Now back to the main discussion. Expressed in terms of rather than
, the Planck energy density is (derived in the next section)
p () =
8h 3
c3
1
h
e kT 1
(14.3)
Units: J/(unit volume unit frequency). Note that in general, in the absence of vacuum, the index of refraction must be taken into account when
considering the speed of light.
Interlude: Derivation of Planck energy density

Just like in the density of states section, imagine a sphere of radius k with
volume
Vk =
4 3
k
3
where the volume of a given mode is kxky kz (assume periodic boundary

conditions, though not necessary, just for convenience) and
kx =
ky =
kz =
2
Lx
2
Ly
2
Lz
199
The number of modes in the given sphere is then (look familiar?)
N =
4
k 3
Vk
= 3 3 Lx Ly Lz
kx ky kz
8
For a photon now, multiply this by 2 to account for two possible polarizations
4
k 3
= 2N = 2 3 3 Lx Ly Lz
8
k3
Lx Ly Lz total number of modes
=
3 2
Consider the volume density
N
k3
k3
= 2 = 2 number of modes/unit volume
Lx Ly Lz
3
3
(14.4)
Here is where we diverge a little from the past. To get the frequency density,
c
rst let k = 2 where = . Therefore
2
c
if the index of refraction is dierent from 1. Then
k=
alternatively 2 n
c
becomes
8 3
k3
=
3 2
3 c3
Now like before, to get the frequency density (alternatively called the mode
density)
=
d
8 2
= 3 number per unit volume per unit frequency
d
c
Planck next showed that the average energy per mode is
=
< >=
(14.5)
h
e
h
kT
Note the Bose Einstein form of the equation (photons are bosons). This
leads to the Planck distribution
p () = < >
h
8 2
=
h
3
c
e kT 1
=
8h 3
c3
h
kT
200
Consider the volume of a thin shell in k space. The volume is
V = 4k 2dk
where kx , ky , kz all equal
n
Lx,y,z .
The number of modes, N is then
4k 2dk
kx ky kz
4k 2dk
Lz Ly Lz
3
Since there are two polarizations, multiply by 2.

N=
8k 2 dk
Lx Ly Lz
3
Now divide by 8 since each corner of an imaginary cube is shared by 8 other

neighboring cubes.
N =
k2
dkLxLy Lz
2
Now get the volume density by dividing by Lx Ly Lz giving

=
where k =
2
c
k 2 dk
N
=
Lx Ly Lz
2
4 2 2
c2
such that k 2 =
=
=
4 2
(2)d
c3
8 2
d
c3
Now consider the frequency density

d =
such that
=
and also dk =
8 2
c3
2
c d.
We get
201
Now back to where we left o. Equate the Einstein and Planck relations
A
g1
B12 g2 e
h
kT
B21
8h 3
c3
1
e
h
kT
where note again that if the index is not 1 then replace c with
examination of the above equation, to achieve equivalence
B12
g1
= B21
g2
c
n.
Upon
(14.6)
More usually you will see g1 = g2 such that the expression reduces to B12 =
B21. Furthermore
8h 3
A
=
B21
c3
yielding the standard textbook expression
A =
=
8h 3
B21
c3
8h 3 g1
B12
c3 g2
(14.7)
Again, note that if g1 = g2 you will commonly see written

A=
8h 3
B12
c3
(14.8)
c
Also if the index is not 1 (not vacuum) then replace c with n in the above
expressions. The above nal expression are what are referred to as the
relationship between Einstein A and B coecients. Remember that A is the
one associated with spontaneous emission.
Word of caution
The Planck distribution is often written a number of ways. Dierent texts
will have what on the surface appear to be completely dierent expressions.
These dierence actually arises because of diering denitions for the density being used by the various authors. Previously we saw two expression
for the Planck distribution. The rst, in terms of wavelength, has units
of: number per unit volume per unit wavelength. The second, in terms of
frequency, has units of: number per unit volume per unit frequency. Alternatively sometimes what people mean is the same expression but in units
202
of: number per unit volume per unit energy. This is potentially very confusing. One should look very carefully at what is meant by density. Here
we derive this third form of the Planck distribution with units: number per
unit volume per unit energy.
Starting with
=
8 3
3
n
c
(14.9)
where this expression comes from the previous section where we derived
the Planck distribution and where also the index of the medium has been
c
explicitly considered (recall, replace c with n ). Rearrange the expression to
=
8n3 3
3c3
h3
h3
8n3 (h)3
8n3 3
=
3c3 h3
3c3 h3
Now as before take the derivative with respect to energy to get the energy
density
=
d
d
8n3 (32 )
3c3 h3
8n3 2
c3 h3
8n3 (h)2
c3 h3
8n3 2
c3 h
=
=
=
=
Some authors will leave it at the second step which is just =

will go with
=
8n3 2
c3 h
8n3 2
.
c3 h3
We
(14.10)
Now earlier we had the Planck derived average energy to be

< >=
h
e
h
kT
This leads to an expression for the average number of photons which is

=
1
<>
= h
h
e kT 1
(14.11)
203
The resulting Planck photon density is
p =
8n3 2
c3 h
1
e
hv
kT
number per unit volume per unit energy
(14.12)
This then leads to another set of relationships between Einstein A and B

coecients
B12 = B21
(14.13)
A=
8n3 2
c3 h B12 or
8n32
c3 h3 B12
(14.14)
The only dierence between these relations and the previous ones was how
the Planck distribution was dened. Number per unit volume per unit
energy (as done here) or number per unit volume per unit frequency (as
done previously).
Einstein A and B coecients revisited

In the last section we derived the relationships between the Einstein coefcients for absorption, stimulated emission and spontaneous emission. In
this section, lets rederive the expression is a slightly dierent manner, but
in a way that will be useful a little later on.
Let R12 be the unit transition rate from the ground state to the excited
state (basically the rate constant)
R12 = Pabs d
B12
where
is the number of photons per unit volume per unit energy (note the
units!)
d is the number of photons per unit volume
Pabs is the probability for absorption per unit time
R12 is the absorptions per unit volume per unit time
and where B12 = Pabs d
As before set up the rate equations except now consider explicitly the
probability of occupied and unoccupied states in the valence and conduction
bands. Let
204

f1 =probability of occupied valence band state
f2 =probability of occupied conduction band state
1 f1 =probability of unoccupied valence band state
1 f2 =probability of unoccupied conduction band state
1 2 transition (absorption) requires

valence band state occupied (f1)
conduction band state empty (1 f2 )
resulting in the joint probability being f1 (1 f2 ) such that
R12,abs = B12f1 (1 f2 )
2 1 transition (stimulated emission) requires
valence band state empty (1 f1 )
conduction band state occupied (f2)
R21,stim = B21 f2 (1 f1 )
2 1 transition (spontaneous emission) requires
valence band state empty (1 f1 )
conduction band state occupied (f2)
R21,spont = Af2 (1 f1 )
In all three cases f1 and f2 are Fermi Dirac distributions
f1 =
f2 =
1
e
1 F
kT
+1
1
e
2 F
kT
+1
205
At this point, for simplicity, assume that g1 = g2 . At equilibrium the upward
and downward rates equal resulting in
R12,abs = R21,stim + R21,spont
B12 f1 (1 f2 ) = B21 f2 (1 f1 ) + Af2 (1 f1 )
Rearrange this to solve for
[B12 f1 (1 f2 ) B21 f2 (1 f1 )] = Af2 (1 f1)
which gives
Af2 (1 f1 )
B12 f1 (1 f2 ) B21 f2 (1 f1 )
A
=
=
f1 (1f2 )
B12 f2(1f1) B21
Now introduce the explicit expression for f1 and f2 . To simplify

1
f1 (1 f2 )
f2 (1 f1 )
1 F
kT
+1
1 F
e kT
2 F
e kT
= e
+1
2 F
kT
+1
+1
2 1
kT
h
= e kT
Replace this into the main expression for to get
=
A
f1 (1f2 )
B12 f2 (1f1)
B21
A
B12 e
h
kT
B21
But equals the Planck distribution for photon density

=
A
B12e
h
kT
B21
8n3 2
c3 h
meaning that for this to be true
1
e
h
kT
B12 = B21
and
8n3 2
B12
c3 h
which are exactly the same Einstein A and B relations we found before.
A=
206
Emission spectrum
Here we will calculate the emission spectrum of a 3D material using the
Einstein A and B coecients. Dene the (net) unit transition rate from the
ground state to the excited state.
Rnet = Pabs d
(14.15)
where as before
is the number of photons per unit volume per unit energy
d is the number of photons per unit volume
Pabs is the probability for absorption per unit time
Rnet is the absorptions per unit volume per unit time
Now consider the net rate upwards from 1 to 2 including stimulated emission
R12,abs = B12 f1 (1 f2 ) (transition rate per unit volume)
R21,stim = B21 f2 (1 f1 ) (transition rate per unit volume)

such that the net upwards rate is
Rnet = R12,abs R21,stim (transition rate per unit volume)
= B12 f1 (1 f2 ) B21 f2 (1 f1 )
Since we have already shown that B12 = B21

Rnet = B12 [f1 (1 f2 ) f2 (1 f1 )]
(14.16)
Equating the two expressions for gives

Pabs d = B12 [f1 (1 f2 ) f2 (1 f1 )]
Pabs d = B12 [f1 (1 f2 ) f2 (1 f1 )]
Here we relate Pabs to the absorption coecient as follows

c
Pabs =
n
How? If P is the absolute probability for an absorption event and
dP
dt
dP
dz
Pabs =
dz
dt
c
n
dP
dt
Pabs
207
Insert this Pabs expression into the above equivalence to get
c
d = B12 [f1 (1 f2 ) f2 (1 f1 )]
n
Rearrange this to solve for B12

B12 =
d
c
n [f1 (1 f2 ) f2 (1 f1 )]
Relate B12 to A using our derived Einstein A and B coecients.

A =
=
8n3 2
c
d
8 2 n 3
B12 =
c3 h
h
c
n [f1 (1 f2 ) f2 (1 f1 )]
d
8 2 n 2
h
c
[f1 (1 f2 ) f2 (1 f1 )]
Now recall that

R21,spont = Af2 (1 f1 )
such that
8 2
h
8 2
=
h
R21,spont =
n
c
n
c
f2 (1 f1 )d
[f1(1 f2 ) f2 (1 f1 )]
d
f1 (1f2 )
f2 (1f1 )
Now for convenience dene

rspont()d = R21
where rspont () is the transition rate per unit volume per unit energy. By
inspection
rspont () =
8 2
h
n
c
f1 (1f2 )
f2 (1f1 )
and
f1 =
f2 =
1
e
1 F
kT
+1
1
e
2 F
kT
+1
208
Weve solved the ratio in the denominator before giving

h
f1 (1 f2 )
= e kT
f2 (1 f1 )
Putting everything together

rspont () =
8 2
h
n
c
h
kT
If h > kT the term in parenthesis e kT and the main expression becomes

rspont () =
8 2
h
n
c
e kT
(14.17)
We can stop here or continue to express this as an intensity.

I3D () =
rspont ()
exc ()
where exc () is the absorption coecient at the particular excitation position. This value is a constant.
I3D () = (h)
=
8 2
h
h
n 2 kT
e
c
exc ()
(8 3)
exc
c
e kT
(14.18)
Finally, for a 3D material the absorption coecient is proportional to the

joint density of states
A g
where A is a constant. This leads to our nal expression for the emission
spectrum
A g
n 2 h
(8 3 )
e kT
I3D =
exc
c
I3D = A
g e kT
(14.19)
209
Figure 14.2: Derived emission spectrum for a 3D material using Einstein A

and B coecients
Figure 14.3: Two level system considered in pulsed laser experiment. Only
the (downward) radiative transition is considered.
210
Quantum yields and lifetimes

Here we discuss some complementary aspects to the emission. Imagine populating the excited state of the system and immediately (i.e. instantaneously)
turning o the light. Basically this is a pulsed experiment.
The depopulation of the excited state occurs by spontaneous emission
since stimulated emission only occurs in the presence of the excitation. The
relevant rate equation is
dN2
= AN2
dt
or if A = krad
dN2
= krad N2
dt
dN2
= krad dt
N2
ln N2 = krad t + (const)
N2 = Cekrad t
At t = 0 C = N2 (0) resulting in
N2 (t) = N2 (0)ekrad t
(14.20)
1
The decay will be exponential and = krad is called the lifetime of the
excited state.
In general, however, since we dont live in a perfect world, there are
other de-excitation pathways. These include energy transfer or non-radiative
decay through defect states. So in general the total decay rate out of the
excited state is the sum of all rates
ktot = krad + k1 + k2 + k3 + . . .
giving
dN2
dt
= (krad + k1 + k2 + k3 + . . . )N2
n
= krad +
= ktot N2
ki N2
i
211
Figure 14.4: Linear and semilogarithmic plots of the excited state decay
prole.
Figure 14.5: Linear and semilogarithmic plots of the excited state decay in
the absence and presence of multiple decay pathways.
212
Only the krad pathway gives you emission. Therefore, the eciency of
emission is called the quantum yield (QY ) and is dened by
QY =
krad
ktot
or
1
rad
1
tot
tot
rad
(14.21)
For most applications, one desires a QY that is as close to unity as possible

(QY = 1). For example, such applications could involve lasers, light emitting
diodes, uorescent tags and so forth.
Exercises
1. The rst part of this chapter described three processes: (stimulated)
absorption, stimulated emission, and spontaneous emission. Rationalize why there isnt a fourth process called spontaneous absorption.
2. Spontaneous emission competes with thermally stimulated emission
(Planck radiation is the source). Show that at 300 K, thermal stimulation dominates for frequencies well below 5x1012 Hz whereas spontaneous emission dominates for frequencies well above 5x1012 Hz. Which
mechanism dominates for visible light.
3. Come up with an alternative means of measuring the quantum yield
of a system based on its uorecence decay prole. See Figure 9.5.
Assume an exponential decay and consider its pure radiative decay
as well as its radiative decay in the presence of other non-radiative
recombination pathways.
4. Show mathematically how one extracts the average decay time of an
exponential process. Compare this to the 1/e lifetime.
5. Derive a relation between the half life (50% mark) and the lifetime of
the state.
6. Explain how one measures the quantum yield of something in real life.
7. Show how the Plank expression reduces to both the Wein displacement
law and the Stefan Boltzman law for blackbody radiation. What are
the Wein and Stefan Boltzman laws anyway?
Chapter 15
Bands
Kronig Penney Approximation with Rectangular
Barriers
An approximation to a crystalline solid is a periodic potential consisting
of rectangular barriers. This is called the Kronig-Penny model and has
analytical solutions. Although this is a crude approximation to the true
potential experienced by the electron, the importance of the model lies in the
idea that the periodicity of the potential is the root cause for the emergence
of what are called band gaps. There are discontinuities in the energies
that an electron can have. This ultimately leads to the distinction between
metals, semiconductors and insulators.
Figure 15.1: Periodic potential of rectangular barriers

Region 1:
213
214
CHAPTER 15. BANDS
h
2 d2
=
2m dx2
2m
d2
= 2
dx2
h
d2 2m
+ 2 =0
dx2
h
(15.1)
Therefore
d2
+ k2 = 0
dx2
(15.2)
where
k2 =
2m
h
2
Region 2:
h
2 d2
+ V =
2m dx2
h
2 d2
= ( V )
2m dx2
d2
2m (V )
=
2
dx
h
2
d2 2m (V )
=0
dx2
h
2
Therefore
d2
2 = 0
dx2
(15.3)
where
2 =
2m (V )
h
2
The wavefunctions in regions 1 and 2 are:

1 = Aeikx + Beikx region 1
x
2 = Ce
+ De
region 2
(15.4)
(15.5)
215
Now apply Blochs theorem where
(x + a) = (x)eilafor delta function potential with period a
or alternatively
(x + (a + b)) = (x)eil(a+b)for periodic potential with period a+b
leading to
Region 3:
3 = 2 eil(a+b)
(15.6)
or
3 =
Cex + Dex eil(a+b)
(15.7)
Now apply the matching conditions for wavefunctions

1 (0) = 2 (0) A + B = C + D
1 (0) = 2 (0) ik(A B) = (C D)
1 (a) = 3 (a) Aeika + Beika = (Ceb + Deb )eil(a+b)
1 (a) = 3 (a) ikAeika ikBeika = (Ceb Deb )eil(a+b)

where we used the fact that
3 (a) = 2 (b)eil(a+b)
3 (a) = 2 (b)eil(a+b)
This leads to a system of four equations with four unknowns (A, B, C, D).
The system can be recast into matrix form. However, to give a non-trivial
solution (A=B=C=D=0) one must have the determinant of prefactors equal
zero. This means solving the following equation.
1
1
1
1
ik
ik
=0
eika
eika
eb+il(a+b) eb+il(a+b)
ikeika ikeika eb+il(a+b) eb+il(a+b)
Most texts will skip the tedious steps to the nal expression. Here we
outline what needs to be done. The solution can be approached a number
216
CHAPTER 15. BANDS
of ways. This is just one. Apply the following steps to generate successively
simpler determinants.
ik(row1) + (row2) (row2)
ik(row2) + (row4) (row4)

yeilding
1
0
eika
0
1
1
1
2ik
(ik )
(ik + )
=0
eika
eb+il(a+b)
eb+il(a+b)
2ikeika (ik )eb+il(a+b) (ik + )eb+il(a+b)
eika (row1) + (row3) (row3)
yielding
1
1
1
1
0
2ik
(ik )
(ik + )
=0
0 eika eika eika eb+il(a+b)
eika eb+il(a+b)
0 2ikeika (ik )eb+il(a+b) (ik + )eb+il(a+b)
then
eika (row2) + (row4) (row4)
yielding
1
1
1
0
2ik
(ik )
0 eika eika
eika eb+il(a+b)
0
0
(ik ) eika + eb+il(a+b)
1
(ik + )
eika eb+il(a+b)
(ik + ) eika + eb+il(a+b)
At this point the nding the 4x4 determinant becomes the same as nding the sub 3x3 determinant. To simplify our notation let
(1) = (ik ) eika + eb+il(a+b)
(2) = (ik + ) eika + eb+il(a+b)
=0
217
then the 3x3 becomes
2ik
(ik )
(ik + )
ika eika eb+il(a+b) eika eb+il(a+b) = 0
e
0
(ik )(1)
(ik + )(2)
eika
which is equivalent to
2ik
eika eb+il(a+b) eika eb+il(a+b)

(ik )(1)
(ik + )(2)
+ 2i sin(ka)
2ik
(ik )
(ik + )
=0
(ik )(1) (ik + )(2)

(ik )(1)
(ik + )(2)
+ 2i sin(ka) (k 2 2 )(2) (k 2 2 )(1) = 0
2ik

(ik )(1)
(ik + )(2)
+ 2i(k 2 2 ) sin(ka) [(2) (1)] = 0
2ik

(ik )(1)
(ik + )(2)
+ 4i(k 2 2 ) sin(ka)sinh(b)eil(a+b) = 0
The top determinant is a little tedious but can be evaluate to give
4ik e2il(a+b) + 2eil(a+b) (k sinh(b) sin(ka) + cosh(b) cos(ka))
+ 4i(k 2 2 ) sin(ka) sinh(b)eil(a+b) = 0
218
CHAPTER 15. BANDS
This expands to
4ik + 4ike2il(a+b) 8ikeil(a+b) (k sinh(b) sin(ka) + cosh(b) cos(ka))
+4i(k 2 2 )eil(a+b) sin(ka) sinh(b) = 0
Multiply by eil(a+b) to get

4ikeil(a+b) + 4ikeil(a+b) 8ik (k sinh(b) sin(ka) + cosh(b) cos(ka))
+4i(k 2 2 ) sinh(b) sin(ka) = 0
Divide by 4i to get
2k cos(l(a + b)) + 2k 2 sinh(b) sin(ka) 2k cosh(b) cos(ka)
+(k 2 2 ) sinh(b) sin(ka) = 0
This leads to the following textbook expression after some simplication

2 k2
2k
sinh(b) sin(ka) + cosh(b) cos(ka) = cos(l(a + b)
Kronig Penney V=0 Limit

Starting with our Kronig Penney expression derived above
2 k 2
2k
sinh(b) sin(ka) + cosh(b) cos(ka) = cos(l(a + b)
recall that
2 =
k2 =
2m(V )
=
h
2
2m
2m
2 k =
h
h
2
2m(V )
h
2
if V=0
2 =
2m
=i
h
2
2m
= ik
h
2
(15.8)
219
the Kronig Penney expression becomes

k 2 k 2
2ik 2
sinh(ikb) sin(ka) + cosh(ikb) cos(ka) = cos(l(a + b))
where
sinh(iu) = i sin(u)
cosh(iu) = cos(u)
leading to
sin(kb) sin(ka) + cos(kb) cos(ka) = cos(l(a + b))
For this to be true
l=k
l=
or
2m
h
2
(15.9)
This gives the energies of a free particle

=
h
2 l2
2m
Kronig Penney V= Limit

Starting with our original Kronig Penney expression
2 k 2
2k
sinh(b) sin(ka) + cosh(b) cos(ka) = cos(l(a + b))
where
k2 =
2 =
2m
h
2
2m (V )
h
2
Lets rearrange everything in terms of unitless variables if possible. Let

= V .
220
CHAPTER 15. BANDS
k2 =
2 =
2mV
2mV
2m V
=
=
2
2
V
V
h
h
2
2mV
2m
V
=
(1 )
V
2
V
h
h
2
where in both cases = /V . Therefore
2 k 2
2k
2mV
h
2
2mV
h
2
2
=
(1 )
2mV
h
2
2mV
h
2
(1 )
(1 2)
(1 )
Consider also how b and ka simplify.
a
2mV
2mV
2 (1 )b =
2 (1 )b a
h
2mV
b
=
1
2 a
a
h
= Ar 1
b =
where A =
2mV
h
2
a and r = b/a are constants. Likewise
2mV
a
h2
= A
ka =
Replace both into our previous equation to get
(1 2)
(1 )
sinh(Ar 1 ) sin(A ) + cosh(Ar 1 ) cos(A )

= cos(la(1 + r))
221
Case ( < 1)
(1 2)
2 (1 )

= cos(la(1 + r))(15.10)
where < 1 implicitly.
Case ( > 1)
In this case
1 =i 1
sinh(iu) = i sin(u)
cosh(iu) = cos(u)
Replace into our previous equation to get
2i
(1 2)
( 1)
i sin(Ar 1) sin(A ) + cos(Ar 1) cos(A )

= cos(la(1 + r))
which reduces to
2
(1 2)
( 1)
sin(Ar 1) sin(A ) + cos(Ar 1) cos(A )

= cos(la(1 + r)) (15.11)
Case ( = 1)
Here note that
sinh(0) = 0
cosh(0) = 1
and that
limt0
sin(t)
=1
t
222
CHAPTER 15. BANDS
Rearrange the original expression as follows

Ar
Ar
(1 2)
sin(Ar 1)
sin(A ) + cos(Ar 1) cos(A )

1
= cos(la(1 + r))
Let = 1 and recall the limit of sin(t)/t as t 0.

1 sin(Ar)
Ar
sin(A) + cos(A) = cos(la(1 + r))
2
Ar
continue reducing the expression to get
Ar
sin(A) + cos(A) = cos(la(1 + r))
2
(15.12)
All cases ( < 1), ( > 1) and ( = 1)

In all three cases above the resulting relationship can be expressed crudely
as some function of equals the right hand side.
F () = cos(la(1 + r))
(15.13)
This relation cannot exactly be solved analytically but is done numerically.

However one notes at this point that the right hand side of the expression
cos(la(1 + r)) will cycle periodically between +1 and -1.
Let V
Continuing our simplication of the this expression, we let V . As a
consequence = V 0 and A .
(1 2)
(1 )

= cos(la(1 + r))
approximately becomes
1
sinh(Ar) sin(A ) + cosh(Ar) = cos(A ) cos(la(1 + r))
2
223
Here in the approximation, the rst term on the left hand side will dominate
1
the second (cosh(Ar)) because of the term. We then get something like
eAr eAr
2
sin(A ) cos(la(1 + r))
or
eAr
sin(A ) cos(la(1 + r))
4
eAr
Since 4 diverges, sin(A ) must approach zero to make the left hand
side fall between +1 and -1 which are the limits of the right hand side
(cos(la(1 + r))).
In limit that sin(A ) = 0,

the
A = n
2mV
a and =
where recall that A =
V This then leads to the
h
2
familiar expression for the energies of a particle in a box.
n2 h2
8ma2
(15.14)
Kronig Penney Approximation with Delta Function

Barriers
Sometimes the Kronig Penney relation is derived assuming a Delta function
potential rather than using rectangular barriers with nite width.
We run through this derivation for completeness. So starting with our
original Kronig Penney relation
2 k 2
sinh(b) sin(ka) + cosh(b) cos(ka) = cos(l(a + b))
2k
One generally makes the following substitution. Let
P =
2 ab
2
(15.15)
224
CHAPTER 15. BANDS
Figure 15.2: Above: original periodic potential of rectangular barriers. Below: approximation using a periodic array of delta functions.
225
2P
where P is basically a dimensionless number. As a consequence b = 2 a .
Replace into the original expression and make the following simplications
2 k 2
2
=
2k
2k
2k
Recall that =
well as
2m(V )
h
2
2me
h
2
and k =
sinh(b) = lim
2P
a
cosh(b) = lim
2P
a
with since V . as
sinh
(15.16)
2P
a
cosh
2P
a
2P
a
=1
where we have made use of a Taylor series as follows.

cosh(x) =
ex + ex
(1 + x) + (1 x)
=1
2
2
(1 + x) (1 x)
ex ex
=x
2
2
Then our main equation becomes
sinh(x) =
sin(ka)
+ cos(ka) = cos(la)
ka
(15.17)
At this point we can take our limits of V = 0(P = 0) or V = (P = )

and see what happens.
Case (P = 0)
Here we get
cos(ka) = cos(la)
Therefore k = l or that l =
2m
h
2
leading to
h
2 l2
2m
which again is the energy of a free electron.
(15.18)
226
CHAPTER 15. BANDS
Case (P = )
Here to allow the left hand side of the expression to have nite values sin(ka)
ka
must be very small. In the limit that sin(ka) = 0, ka = n This immediately
ka
leads to
n2 h2
8ma2
(15.19)
which are the energies of a particle in a box.
Summary
One can therefore see that the Kronig Penney model lies in between the two
limiting cases for the behavior of an electron in a potential. At one end there
is the free electron limit and at the other is the particle in a box limit. In
between is the regime giving existence to the notion of bands separated by
energy gaps. So as usually mentioned in textbooks, at this point we point
out that metals are those materials that have their conduction band half
full, semiconductors are those materials where the valence band is full but
where the band gap is not humongous (lets say < 4 eV) and insulators are
those materials where the valence band is full and where the band gap is
huge (say > 4 eV).
Exercises
1. Consider the rectangular barrier Kronig Penney model discussed early
on. Use Mathcad, Matlab, Mathematica or your favorite mathematical
modeling program to visualize the actual bands. Choose a potential
well width of your own choosing, a barrier width, again dened by you,
and a nominally high barrier height. Plot the free electron energies on
top of the bands you have drawn. Show the plot in both the periodic
zone scheme and the reduced zone scheme.
2. Consider the delta function modication of the Kronig Penney model.
Again use a mathematical modeling program to draw the bands in the
periodic and reduced zone schemes. Choose whatever barrier height,
well and barrier width you desire.
3. What temperature in K would you need to promote an electron from
the valence band to the conduction band of a semiconductor with a
227
Figure 15.3: Mathcad numerical solutions to the general Kronig Penney

model
228
CHAPTER 15. BANDS
Figure 15.4: Mathcad numerical solutions to the general Kronig Penney

model continued
229
Figure 15.5: Mathcad numerical solutions to approximations of the Kronig

Penney model
230
CHAPTER 15. BANDS
Figure 15.6: Mathcad numerical solutions to the delta function Kronig Penney model
231
Figure 15.7: Mathcad numerical solutions to the delta function Kronig Penney model continued
232
CHAPTER 15. BANDS
Figure 15.8: Cartoon of the dierences between metals, semiconductors and

insulators.
bandgap of Eg = 1.17eV ? How about for a semiconductor with a

bandgap of Eg = 2.5eV ?
4. What temperature in K would you need to promote an electron from
the valence band to the conduction band of n insulator with a gap of
Eg = 5eV ?
Examples from the literature

Here are selected papers from the literature dealing with one area of nanoscience
or nanotechnology which is the creation of artical solids with tailored properties. Crudely one version of this is to use quantum dots as articial atoms.
When arranged in a periodic manner (basically a crystal) the overlap of the
electron wavefunctions will create user dened optical and electrical properties as well as the formation of articial bands as we just described. These
go by the name minibands and have already been achieved in the case of
arrays of stacked quantum wells.
The papers below are listed in no particular order
1. Architechtonic quantum dot solids
G. Markovich, C. P. Collier, S. E. Henrichs, F. Remacle, R. D. Levine,
233
J. R. Heath
Accounts of Chemical Research, 32, 415 (1999).
2. Nanocrystal superlattices
C. P. Collier, T. Vossmeyer, J. R. Heath
Annual Review of Physical Chemistry, 49, 371 (1998).
3. Highly oriented molecular Ag nanocrystal arrays
S. A. Harfenist, Z. L. Wang, M. M. Alvarez, I. Vezmar, R. L. Whetten
Journal of Physical Chemistry, 100, 13904 (1996).
4. Synthesis of a quantum dot superlattice using molecularly linked
metal clusters
R. G. Osifchin, W. J. Mahoney, J. D. Bielefeld, R. P. Andres, J. I.
Henderson, C. P. Kubiak
Superlattices and Microstructures, 18, 283 (1995).
5. Architecture with designer atoms: Simple theoretical consideration
F. Remacle and R. D. Levine
Proceedings of the National Academy of Science, 97, 553 (2000).
6. Preparation and characterisation of silver quantum dot superlattice
using self-assembled monolayers of pentanedithiol
S. Pethkar, M. Aslam, I. S. Mulla, P. Ganeshan, K. Vijayamohanan
Journal of Materials Chemistry, 11, 1710 (2001).
7. Miniband formation in a quantum dot crystal
O. L. Lazarenkova and A. A. Balandin
Journal of Applied Physics, 89, 5509 (2001).
8. Reversible tuning of silver quantum dot monolayers through the metalinsulator transition
C. P. Collier, R. J. Saykally, J. J. Shiang, S. E. Henrichs, J. R. Heath
Science, 277, 1978 (1997).
234
CHAPTER 15. BANDS
9. Optical properties of thin lms of Au@SiO2 Particles

T. Ung, L. M. Liz-Marzan, P. Mulvaney
Journal of Physical Chemistry B, 105, 3441 (2001).
Chapter 16
K P Approximation
(Pronounced k dot p)
Start with the Schrdinger equation
o
H =
(16.1)
where
H=
h
2 2
+V
2m
(16.2)
Assume the Bloch form of the wavefunction

= u(r)eikr
(16.3)
Replace this into the above Schrdinger equation.

o
h
2 2
+V
2m
u(r)eikr =
u(r)eikr
h
2 2
u(r)eikr + V u(r)eikr =
2m
u(r)eikr
235
236CHAPTER 16. K P APPROXIMATION(PRONOUNCED K DOT P)

where
2
(ueikr ) =
ueikr
uikeikr + eikr u
ikueikr +
(iku +
ueikr
u) eikr
2
u) ikeikr + eikr ik u +
= (iku +
= k 2 ueikr + ik ueikr + ik ueikr +
=
=
ueikr
ueikr k 2 ueikr + 2ik ueikr

2
u k 2u + 2ik u eikr
So
h
2
2m
h
2 2 ikr
ue + V ueikr = ueikr
2m
u k 2u + 2ik u eikr + V ueikr = ueikr
Divide out eikr
h
2
2m
u + 2ik u k 2 u + V u =
h
2 2 u i 2 k u 2 k 2 u
h
h
+
+Vu =
2m
m
2m
h
2 2 u i 2 k u
h
+Vu =
2m
m
h
2 2
+V
2m
Recall now that p = i

h
h
2 2
+V
2m
Also recall that p2 = 2

h
p2
+V
2m
i 2 k u
h
m
u
u
h
2 k2
2m
h
2 k2
2m
(Angular momentum operator) so that

u+
2
kp u =
m
h
2 k2
2m
leading to
u+
h
(k p)u
=
m
h
2 k2
2m
237
2
p
where H0 = 2m + V
equation becomes
and H1 =
h
(kp)
m .
This means that the Schrdinger

o
(H0 + H1 ) u = u
where =
Hamiltonian
h
2 k2
2m
(16.4)
. So in general, to start solving this we create the

u1 |(H0 + H1 )|u2 = 12 u1 |u2
(16.5)
where the Hamiltonian H is H = u1 |(H0 + H1 )|u2 . Then we solve the

secular determinant
|H | = 0
(16.6)
To summarize, the Schrodinger equation becomes

p2
+V
2m
u+
k p u = u
m
(16.7)
where we explicitly consider the k p perturbation to the eigenenergies.
Simple 2 band model

Lets say we have a unit cell function us that is for the conduction band
(CB) and has s-like symmetry. Also we have a unit cell function up that is
for the valence band (VB) and has p-like symmetry. We need to solve
u1 |H0 + H1 |u2
2
(16.8)
p
h
where H0 = 2m + V (We know the solutions of these) and H1 = m k p.

We ignore spin orbit interactions here and dont even consider any spin for
simplicity. So we basically have the following 2 by 2 Hamiltonian.
< us |
< up |
Lets do the diagonals rst.
|us >
a
c
|up >
b
d
Array element a
a = < us |H0 + H1 |us >
= < us |H0 |us > + < us |H1 |us >
s+
< us |H1 |us >
us | m (k p)|us
> where p = i (momentum operah

and < us |H1 |us >=<
tor). Since
ips the parity of the unit cell function (k p)|us > (even) =
(odd). Therefore
h
< us |k p|us >= 0

m
such that in whole
a=
(16.9)
Array element d
d = < up |H0 + H1 |up >
= < up |H0 |up > + < up |H1 |up >

h
= v + < up | k p|up >

m
Recall that
mk
p|up >= (even) and that up = (odd). Hence

< up |
k p|up >= 0
m
In whole then
d = ev
(16.10)
Next, lets do the o diagonals
Array element b
b = < us |H0 + H1 |up >
= < us |H0 |up > + < us |H1 |up >
= 0+ < us |H1 |up >

h
= < us | k p|up >= 0 = KP

m
m
where we dene KP < us |k p|up >. Therefore
b=
KP
m
(16.11)
239
Array element c
We can work out c but since the Hamiltonian is Hermitian c will be the
same as b. Therefore
h
(16.12)
c = KP
m
To summarize,
h
m KP
H= s
h
v
m KP
Now we will solve for the secular determinant |H | = 0
h
m KP
v
s
h
KP
m
=0
This leads to solving

(
)(
Dene
p
h
2 k 2 p2
=0
m2
2p2
m
(16.13)
leading to
(
s )(
h
2 k2
2m
2p2
m
h
2k2
2m
h
2k2
2
v s v +
2m
h
2k2
2 ( s + v )
2m
(
v )
s )(
v )
= 0
p
= 0
= 0
= 0
The last equation is everyones favorite quadratic equation. The solutions

are generally speaking
=
=
=
(
(
(
v)
+
2
v)
+
2
v)
v)
1
2
s+
1
2
4(1)[
2
v) 4
v)
+4
s v
v+4
h
2 k2
2m
h
2 k2
2m p ]
h
2k2
2m

Recall now that
=(
=
and since =
v)
(gap energy) so that the solutions are:
+
2
v)
1
2
h
2k2
2m
2
g
+4
h
2k2 1
2m
2
2
g
h
2k2 1
+
2m
2
2
g
+4
h
2k2
2m
(16.16)
h
2k2 1
2m
2
2
g
+4
h
2 k2
2m
(16.17)
(16.14)
h
2 k2
2m
+
2
v)
+4
h
2k2
2m
(16.15)
Conduction band solution
cb
+
2
v)
+
2
v)
Valence band solution

(
vb
You can already tell that the solutions are non-parabolic in k. We can make
some further insightful simplications using the binomial expansion and
1
1
keeping only the rst two terms a + b a 2 + 1 a 2 b + . . . when b2 < a2 .
2
2
g
1 1 2 k2
h
2 g 4 2m p
2
g
So
+4
h
2k2
2m
+
2
v)
2 2
h k
+ 4 2m
(assume k is small). Therefore
g+
h
2k2
2m
p
g
(16.18)
Leading to
cb
=
=
=
(
(
+
2
v)
h
2k2
2m
+
+
h
2k2 1
+
2m
2
h
2k2
1+
p
g
v)
p
g
2m
h
2 k2
2m
p
g
h
2k2
2m
(16.19)
241
Likewise for the valence band
vb
=
=
=
(
(
+
2
+
2
v)
v)
h
2 k2
2m
h
2k2 1
2m
2
h
2 k2
v)
h
2k2
2m
p
g
2m
p
g
h
2k2
2m
(16.20)
Now we can go after the eective masses. For the conduction band
h
2k2
h
2 k2
=
2mef f
2m
1+
p
g
Therefore
1
1
=
mef f
m
1+
mef f = m
1
1+
p
g
p
g
Leading to the conduction band eective mass

mef f
m
1
1+
(16.21)
Likewise, for the valence band

h
2k2
h
2 k2
=
2mef f
2m
p
g
Therefore
1
1
=
mef f
m
mef f = m
1
1
p
g
p
g
Leading to the valence band eective mass

mef f
m
1
1
p
g
(16.22)

Now since p 21 eV typically and also does not vary much between dierent
semiconductors one can see that the conduction band eective mass is (a)
small and (b) is positive. Likewise, on can see that the valence band eective
mass is (a) small and (b) negative. Although the 2 band model is a large
simplication of reality, generally speaking, me < mh .
Alternative view of 2 band model

Sometimes you will see people conveniently redene the energy origin to
simplify the problem at hand. Doing this was not necessary for our two
band model because this 2 by 2 was easy to solve. However, for 4 by 4 or
higher band model it becomes a great convenience. We illustrate this with
the 2 by 2.
Starting with the Hamiltonian
H=
g
h
m KP
m KP
we again need to solve the secular determinant

|H | = 0
h
m KP
g
h
m KP
leading to
(
As before let
2p2
m .
=0
h
2 k 2 p2
m2
2 2 2
h
k p
g + 2
m2
2 2 2
h
k p
2 g
m2
)()
= 0
= 0
= 0
This leads to
2
h
2k2
2m
=0
(16.23)
We could do this exactly by solving the quadratic equation. However, we

can do some approximations at this point to solve the problem quicker. For
k 0 (small) we see that the equation reduces to
2
=0
243
or
(
g)
=0
The two k = 0 limiting roots are ( = 0) and ( = g ) so we will assume

that the actual roots are of this order and will make approximate solutions
to
h
2 k2
2m
h
2 k2
( g )
2m
2
=0
=0
Solve for close to zero

h
2 k2
2m
h
2 k2
g+
2m
( g )
=0
=0
leading to
=
where =
h
2 k2
2m
h
2k2
2m
(16.24)
leading to
h
2k2 2 k2
h
2m
2m
vb
h
2k2
2m
p
g
(16.25)
This is the valence band energy of the 2 band model where one notes that
the origin is now at zero.
Eective mass of the VB
h
2k2
2mef f
mef f
h
2k2
2m
1
= m
1
1
p
g
p
g

leading to
mef f
1
=
m
1
(16.26)
p
g
This is a negative eective mass since p

21 eV and p
g typically.
Now we will do the conduction band energy and eective mass in the next
section.
Solve for close to
g)
h
2k2
2m
=0
g)
h
2k2
2m
=0
h
2 k2
2m
becomes
g (
leading to
=
Recall that =
h
2 k2
2m
(16.27)
leading to
h
h
2k2 2k2
+
2m
2m
g+
cb
h
2k2
+
g
2m
p
g
1+
(16.28)
This is the conduction band energies of the 2 band model. Note that the origin is now at g rather than at f (Fermi level) of an intrinsic semiconductor.
Now we nd the eective mass.
Eective mass of the CB
h
2 k2
2mef f
mef f
leading to
h
2 k2
2m
= m
1+
1
1+
mef f
1
=
m
1+
p
g
p
g
(16.29)
245
This is a positive eective mass since
and p
g.
21 eV for most semiconductors
Last note on eective masses

Since
mef f
m
1
1+
mef f
m
1
1
CB
VB
You can see that the eective mass will scale with the band gap of the
semiconductor. Small eective masses correlates with a small band gap.
Large eective masses correlate with a large band gap.
A four band model excluding spin

Assume the conduction band is s-like. Assume that the valence band is plike. Choose for convenience k = (0, 0, kz ) so that k p = kz pz . Ok, now we
know form before that we need to rst generate the Hamiltonian
< u1 |H0 + H1 |u2 >
(16.30)
where
H0 =
H1 =
p2
h
2 2
+V =
+V
2m
2m
h
kp
m
and where p = i . This leads to the following Hamiltonian

h
< us |
< ux |
< uy |
< uz |
|us >
a
|ux >
b
e
|uy >
c
f
h
|uz >
d
g
i
j
where we have indicated the important array elements with letters. The
others follow because the Hamiltonian is Hermitian. Lets do the diagonals
rst.
Array element a
< us |H0 + H1 |us > = < us |H0 |us > + < us |H1 |us >
h
< us |kz pz |us >

= s+
m
= s
because us has even symmetry in x, y and z. pz acting on us changes the
overall parity to odd such that the term dies.
Array element b
< ux |H0 + H1 |ux > = < ux |H0 |ux > + < ux |H1 |ux >
h
= v+
< ux |kz pz |ux >
m
= v
Here ux is odd in x, even in y and even in z. pz acting on ux changes its
parity to odd in x, even in y, odd in z. Since < ux is odd in x, even in y and
even in z, the z components cancel.
Array element c
< uy |H0 + H1 |uy > = < uy |H0 |uy > + < uy |H1 |uy >
h
< uy |kz pz |uy >

= v+
m
= v
Here uy is even in x, odd in y, even in z. pz acting on uy changes its parity
to even in x, odd in y and odd in z. Since < uy | is even in x, odd in y and
even in z, the z components cancel.
Array element d
< uz |H0 + H1 |uz > = < uz |H0 |uz > + < uz |H1 |uz >
h
< uz |H1 |uz >

= v+
m
= v
Here uz is even in x, even in y and odd in z. pz acting on uz changes its
parity to even in x, even in y and even in z. Since < uz | is even in x, even
in y, and odd in z, the z components cancel.
247
Summary
< us |H0 + H1 |us > =
< ux |H0 + H1 |ux > =

< uy |H0 + H1 |uy > =
< uz |H0 + H1 |uz > =
s
v
v
v
Now for the o diagonals
Array element e
h
< us |kz pz |ux >

m
= 0
< us |H1 |ux > =
Here ux is odd in x, even in y, even in z. pz acting on ux changes its parity

to odd in x, even in y, and odd in z. Since < us | is even in x, even in y and
even in z, the x and z components cancel.
Array element f
h
< us |kz pz |uy >

m
= 0
< us |H1 |uy > =
Here uy is even in x, odd in y and even in z. pz acting on uy changes its

parity to even in x, odd in y and odd in z. Since < us is even in x, even in
y and even in z the y and z components cancel.
Array element g
< us |H1 |uz > =
=
< us |kz pz |uz >

m
h
KP
m
Here uz is even in x, even in y and odd in z. pz acting on uz is even in x,

even in y and even in z. Since < us | is even in x, even in y and even in z
none of the terms cancel.
Array element h
h
< ux |kz pz |uy >

m
= 0
< ux |H1 |uy > =
Here uy is even in x, odd in y, and even in z. pz acting on uy changes its

parity to even in x, odd in y and odd in z. Since < ux | is odd in x, even in
y and even in z everything cancels out.
Array element i
h
< ux |kz pz |uz >

m
= 0
< ux |H1 |uz > =
Here uz is even in x, even in y, and odd in z. pz acting on uz changes its

parity to even in x, even in y and even in z. Since < ux is odd in x, even in
y, and even in z the x components cancel.
Array element j
h
< uy |kz pz |uz >

m
= 0
< uy |H1 |uz > =
Here uz is even in x, even in y, and odd in z. pz acting on uz changes its

parity to even in x, even in y and even in z. Since < uy | is even in x, odd in
y and even in z, the y components cancel.
Summary
< us |H1 |ux > = 0
< us |H1 |uy > = 0

h
< us |H1 |uz > =

KP
m
< ux |H1 |uy > = 0
< ux |H1 |uz > = 0
< uy |H1 |uz > = 0
249
The desired Hamiltonian is then
0
v
0
0
h
m KP
H=
0
0
0
0
m KP
0
0
Solve the secular determinant now.
0
0
h
m KP
s
0
v
0
0
0
0
v
0
m KP
0
=0
0
v
There are many ways to do this. One way looks like

v
0
0
0
v
0
0
h
0
KP
m
v
0
v
0
0
v
m KP
m KP
0
0
v
0
0
=0
leading to
(
)(
)3 +
KP (
m
=0
KP ( v )2 KP = 0
m
m
h
2 k 2 p2
( s )( v )3
( v )2 = 0
m2
h
2 k 2 p2
= 0 where
( v )2 ( s )( v )
m2
)(
2
(
v )
)3
s )(
2
(
v )
h
2k2
2m
2p2
m
v )
h
2k2
2m
v )
s )(
2p2
m
= 0
p
= 0
There are two immediate solutions that are apparent in the rst term, 1,2 =
v.
1
h
2k2
+
2m
h
2k2
+
2m
(16.31)
(16.32)

To get the other solutions we must solve
h
2k2
p =0
2m
You can do this exactly but lets be smart. One solution is of order
other solution is of order v .
(
)(
Solution 1 (of order

(
)(
s)
s.
The
s)
h
2 k2
2m
= 0
h
2 k2
p
where
2m ( v s )
h
2k2
p
( s ) =
2m
g
(
) =
This results in
cb =
=
cb
since =
h
2 k2
2m .
+
+
h
2k2
2m
h
2k2
2m
p
g
p
1+
(16.33)
This is the desired conduction band energy.
Conduction band eective mass

h
2k2
2mef f
1
=
p
1+ g
mef f
m
Solution 2 (of order

(
h
2k2
2m
1+
p
g
(16.34)
v)
v )( v
)
(
(
h
2k2
2m
v
= 0
) =
) =
h
2 k2
2m
h
2 k2
2m
p
s
p
g
251
This results in
=
vb
since =
h
2 k2
2m .
h
2k2
2m
h
2k2
2m
p
g
(16.35)
This is the desired valence band energy.
Valence band eective mass

h
2k2
2mef f
mef f
m
=
=
h
2k2
1
2m
1
1 p
g
Kane model, (8 x 8)
Given the basis
|is
(x iy)
2
|z
(x + iy)
2
|is
(x + iy)
2
|z
(x iy)
2
the 8 by 8 interaction matrix is written as
H 0
0 H
p
g
(16.36)

where
0
H=
kP
0
0
p
3
2
3
KP

23
p
0
0
0
+
p
and where the positive constant is the spin orbit splitting of the valence
band.
i
h
P = < s|pz |z >
m
3 i
h
V
V
=
|y
x|
2 c2
4m
x
y
Note that the vector k is taken along the z direction for convenience.
To solve this, choose p = to simplify things
3
0
kP 0
s
2
0 3 32 0
H=
2
kP
0
3
3
0
0
0
0
Now solve the secular determinant
0
kP
0
2
2
0
3
0
3
=0
2
kP
3 0
3
0
0
0
2 2
h k
So by inspection, the trivial solution is = 0. Since recall = 2m the
desired energy is
h
2k2
(16.37)
=
2m
Now we go back for the other three energies. Basically this entails solving
the following secular determinant
0
kP
2
2
0
3
=0
2
3
kP
3
which becomes
2
0
2
3
( s ) 3 2
+ kP
kP
3
3
2
3
=0
2
3
253
This then becomes
(
2
+
3
+ 2 + (kP )2
3
9
(
)[2 + ] + (kP )2
) (kP )2
s )( +
2
+
3
2
+
3
2
+
3
= 0
= 0
= 0
We have now the following

= 0
(
s )( +
2
+
3
2 2
) k p
= 0
where remember that we chose p = to simplify the problem. These are

3
also equations 9 and 10 from Kanes paper. J. Phys. Chem. Solids, 1, pg
249.
Now we want the actual energies. So rst assume that k is small (alternatively that k 2 is really small)
(
s )( + )
+ small terms = 0
So we can see right away that the three desired energies will center around
= 0
=
=
Choose = 0, =
and = successively and solve the above equation.
Case where = 0
( s )() = k 2 p2
2
( s ) = k 2 p2
3
2k 2 p2
=
3 s
where recall that =
h
2 k2
2m
leading to
h
2 k 2 2k 2 p2
2m
3 s
(16.38)

This is the rst of our desired energies and is actually associated with the
light hole.
Case where =
s ( s )( s
+ ) = k 2 p2
s)
s)
s)
s)
2m
2 + 3 s
k 2p2
3
s ( s + )
2 p2
k
2 + 3 s
3
s ( s + )
2 p2
2( s + ) + s
k
3
s ( s + )
2 p2
1
2
k
+
3
s
s+
=
h
2k2
2
+
3
+
+
k 2 p2
3
k 2 p2
1
s+
1
+
s
where recall that =
yielding
h
2 k 2 k 2 p2 2
1
(16.39)
+
+
2m
3
s
s+
This is another one of our desired energies and is ultimately associated with
the conduction band energy.
=
Case where =
( +
3
k 2 p2
2
3
( + ) =

( + s ) 3
3
k 2 p2
( + ) =
( s + )
3
2 p2
k
( + ) =
3( s + )
s )( +
) = k 2 p2
k 2p2
h
2k2
where =
3( s + )
2m
h
2k2
2m
255
This results in
k 2 p2
h
2k2
(16.40)
2m
3( s + )
This is the last of the three desired energies and is associated with the split
o band. We can also toss in the trivial solution = 0 which gives
= +
h
2k2
(16.41)
2m
This is the fourth and last energy and is associated with the heavy hole
band.
=
Alternative expressions
For
(
where
2mp2
h
2
s )( +
2
+
3
) k 2 p2
=0
one can also write and solve for the expression
s )( + )
h
2 k2
2m
2
+
3
= 0
Case where = 0
( s )() =
h
2k2
2m
h
2k2 p
2m s
leads to
h
2k2 2 p
(16.42)
2m 3 s
This is our alternative expression for the light hole band energy. Rearranging
this as follows gives the eective mass
=
h
2k2
2mef f
1
mef f
mef f
h
2k2 2
2m 3
1 2 p
=
m 3 s
3 s
= m
2 p
p
s

Resulting in
mef f
3
=
m
2
(16.43)
This is the eective mass of the light hole.
Case where =
s )( +
h
2k2
2m
2
+
3
s ( s )( s
= 0
h
2k2
2m
+ ) =
s)
s)
h
2k2
2m
=
=
2
+
3
p
s( s
+ )
2
2 2
h
k p s + 3
2m s ( s + )
2
+
3
This leads to
=
s+
h
2k2
2m
+ 2
3
( s + )
s
s
(16.44)
This is the alternative form of the conduction band energy. To get the
eective mass we evaluate
h
2k2
2mef f
1
mef f
mef f
h
2 k2
2m
+ 2
3
s ( s + )
s
+ 2
3
s ( s + )
s ( s + )
= m
2
p s + 3
1
m
or
mef f
m
s( s
=
p
+ )
+ 2
s
3
This is the eective mass of the conduction band.
(16.45)
257
Case where =
(
s )( + )
h
2k2
2m
2
+
3
= 0
s )( + )
( +
( + ) =
h
2k2
2m
3
2

3
3
h
2k2
p
2m ( +
( + ) =
This results in
h
2k2
p
2m 3() +
h
2k2
s)
2m 3( +
s)
(16.46)
s
This is the alternative expression for the split o hole energy. Now for the
eective mass
h
2 k2
2mef f
1
mef f
mef f
or
h
2k2
p
2m 3( + s )
1
p
=
m 3( + s )
3( + s )
= m
=
mef f
m
3( +
s)
(16.47)
This is the eective mass of the split o hole.
Alternative expressions deux

Assume here that
(
kP and
g )( +
in the original expression
2
) k 2p2 +
3
=0
(16.48)
So basically this is kind of already forecasting the solutions of this problem.

Namely, that there are two bands close to the valence band edge and one
band way way removed from them by . Also recall that we already have

the trivial solution = 0 which gives the energy of the heavy hole band. So
there are four total solutions.
Solve here for the two valence bands (light hole and split o hole) by
assuming
in the above equation.
(
3
2k 2 p2
( g )
3
2k 2 p2
2 g
3
g )() k
2 2
= 0divide out
= 0
= 0
This last expression is a quadratic equation whose general solutions we know.

g
=
Recall that =
h
2 k2
2m
2
g
+ 4(1)
2k2 p2
3
2
leading to
g
1
2
2
g
8k 2 p2
3
(16.49)
(this is equation 13 in Kanes paper) from which we get

1
2
2
1
g
2
2
g
cb =
lh =
8k 2 p2
conduction band
3
8k 2 p2
2+
light hole band
g
3
2
g
Now we go back for the split o hole band energy

(
g )( +
2
) k 2p2 +
3
=0
Assume that is of order to get

(
Solve for .
g )( +
2
) k 2 p2 ( + ) = 0
3
(16.50)
(16.51)
259
( +
g )( +
) = k 2 p2
( + ) =
k 2 p2
3( + g )
=
where recall that =
h
2 k2
2m .
soh
soh
= +
k 2p2
3( + g )
(16.52)
This results in
k 2 p2
h
2k2
2m
3( + g )
(16.53)
Summary ( large)
cb
hh
lf
soh
h
2 k2
1
8k 2 p2
g
2+
+ +
g
2m
2
2
3
2 2
h
k
=
2m
1
h
2 k2
8k 2 p2
g
2+
+
=
g
2m
2
2
3
2 2
2 p2
k
h
k
= +
2m
3( + g )
=
Note that apart from the heavy hole, all bands are non-parabolic. Also recall
that the heavy hole solution came from the trivial solution = 0.
Relevant reading
Band structure of indium antimonide
E. O. Kane J. Phys. Chem. Solids. 1, 249, 1957.
Chapter 17
Tunneling
Potential step
This is the simplest example of tunneling. Picture the following potential
step.
Figure 17.1: Cartoon of the potential step
The Schrodinger equation is
h
2 d2
+ V =
2m dx2
This can be rearranged as follows.

261
262
CHAPTER 17. TUNNELING

In region 1 where V = 0
h
2 d2
= ( V )
2m dx2
2m
d2
= 2
2
dx
h
d2
+ k2 = 0
dx2
where k 2 =
2m
.
h
2
Solutions to the wavefunction are

1 = Aeikx + Beikx
(17.1)
In region 2 where V is nite, there are two cases.

Case 1: > V
Case 2: < V
Case 1: ( > V )
Figure 17.2: Behavior of the wavefunction in the potential step case where
>V
263
where 2 =
region, are
h
2 d2
+V
2m dx2
h
2 d2
2m dx2
d2
dx2
d2
+ 2
dx2
2m(V )
.
h
2
=
= ( V )
=
2m
( V )
h
2
= 0
Solutions to the wavefunction for this case, in this
2 = Ceix + Deix
(17.2)
Case 2: ( < V )
Figure 17.3: Behavior of the wavefunction in the potential step case where
<V
264
where 2 =
region, are
h
2 d2
+V
2m dx2
h
2 d2
2m dx2
d2
dx2
d2
dx2
d2
2
dx2
2m(V )
.
h
2
=
= ( V )
2m
( V )
h
2
2m
(V )
h
2
=
=
= 0
Solutions to the wavefunction for this case, in this

2 = Cex + Dex
(17.3)
Now in order to proceed, we need to employ the matching conditions for

the wavefunctions. The matching conditions are:
1. 1 (0) = 2 (0)
2. 1 (0) = 2 (0)
Furthermore, the wavefunction must be nite valued. That is, the wavefunction cant blow up or do strange things.
Transmission and reection coecients; > V
The wavefunctions in this case are
1 = Aeikx + Beikx
2 = Ceix + Deix
Since there are no sources of particles on the right to support a left going
wave, D = 0. The wavefunctions become
1 = Aeikx + Beikx
2 = Ceix
Apply the matching conditions. First one
1 (0) = 2 (0)
A+B = C
265
Second matching condition
1 (0) = 2 (0)
ikx
Aike
Bikeikx = Cieix
Aik Bik = Ci
k(A B) = C
Now take the resulting equations and solve for A and B

A+B = C
C
AB =
k
1+
k
2A = C 1 +
A =
C
2
(17.4)
Next B
A+B = C
A + B = C
k
1
k
2B = C 1
B =
C
2
(17.5)
Now the incident ux of particles from the left onto the potential step is
v1 |A|2
(17.6)
where v1 is the velocity in region 1 and |A|2 is the relative probability of a

right going wave in region 1. Similarly
v1 |B|2
(17.7)
is the reected ux at the potential step where again v1 is the velocity in

region 1 and |B|2 is the relative probability of a left going wave. Finally
v2 |C|2
(17.8)
266
is the transmitted ux where v2 is the velocity in region 2 and |C|2 is the

relative probability of a right going wave in region 2.
The fraction of particles reected is therefore
R
R =
v1 |B|2
B
=
v1 |A|2
A
C2
4
C2
4
1+
R=
(k2 )2
(k+2 )2
BB
A A
2
2
(17.9)
>0
So oddly enough, this is non-classical. Some reection occurs even though

>V.
Next, the fraction of particles transmitted is
T
|C|2
v2 |C|2
=
=
v1 |A|2
k|A|2
4
k 1+
C 2
2
kC
4
1+
4k 2
k(k + )2
resulting in
T =
4k
(k+)2
=0
(17.10)
Transmission and reection coecients; < V

Recall that the wavefunctions in this case are
1 = Aeikx + Beikx
2 = Cex + Dex
where k = 2m and = 2m(V ) . Now recall that the wavefunction in
h
region 1 and 2 must be nite valued. This means that C = 0. Cant have
something blow up here. Therefore
1 = Aeikx + Beikx
2 = Dex
267
Apply the matching conditions now. First one
1 (0) = 2 (0)
A+B = D
Second one
1 (0) = 2 (0)
ikA ikB = D
ik(A B) = D
Now take the resulting equations and solve for A and B. First lets solve
for the A coecient.
A+B = D
AB =
D
ik
ik
1
ik
2A = D 1
A =
D
2
(17.11)
Now lets solve for B

A+B = D
D
A + B =
ik
ik
1+
ik
2B = D 1 +
B =
D
2
From before, the relevant uxes are

v1 |A|2 (leftgoing)
v1 |B|2 (rightgoing)
v2 |D|2 (transmitted)
(17.12)
268
The fraction of particles reected is

R =
B
v1 |B|2
=
v1 |A|2
A
BB
A A
D2
4
1+
ik
ik
D2
4
ik
1+
ik
D2
4
1+
2
k2
D2
4
1+
2
k2
= 1
However the fraction transmitted is
T
=
=
=
v2 |D|2
|D|2
=
v1 |A|2
k|A|2
D2
2
kD
4
1+
2
k2
4
k 1+
2
k2
giving
T =
4k
(k2 + 2 )
=0
(17.13)
This gives a non-classical result. There is some probability of nding soemething in region 2.
Potential barrier
From the last section we know that the wavefunctions have the following
form.
1 = Aeikx + Beikx
2 = Ceix + Deix ( > V )
2 = Cex + Dex ( < V )
3 = F eikx + Geikx
269
Figure 17.4: Cartoon of the potential barrier
where k 2 = 2m , 2 = 2m(V ) , and 2 = 2m(V2) . Now as before we apply

h
2
h
2
h
the matching conditions to the wavefunctions to solve for the coecients

and in turn for the transmission (T) and reection (R) coecients.
Case 1: ( > V )
Figure 17.5: Behavior of the wavefunction in the potential barrier case where
>V
Assuming that there is no source of particles on the right then G = 0.
270
The wavefunctions in the three regions become

1 = Aeikx + Beikx
2 = Ceix + Deix
3 = F eikx
Apply the matching conditions at x = 0. First one
1 (0) = 2 (0)
A+B = C +D
(17.14)
Second one
1 (0) = 2 (0)
Aik Bik = Ci Di
(C D)
AB =
k
(17.15)
Apply the matching conditions at x = a. First one

2 (a) = 3 (a)
ia
Ce
+ Deia = F eika
(17.16)
Second one
2 (a) = 3 (a)
ia
Cie
Dieia = F ikeika
Ceia Deia = F keika
(17.17)
Solve for A.
A+B = C +D
C D
AB =
k
k
2A = C 1 +
+D 1
k
k
C
A =
1+
+
1
2
k
2
k
(17.18)
271
Solve for B.
A+B = C +D
A + B = C + D
k
k
2B = C 1
+D 1+
k
k
C
B =
1
+
1+
2
k
2
k
(17.19)
Solve for C.
Ceia + Deia = F eika
Ceia Deia = F keika
ia
Ce
ia
+ De
= F eika
Ceia Deia = F keika

ia
2Ce
= F (k + )eika
k + ia(k)
e
C = F
2
(17.20)
Solve for D.
Ceia + Deia = F eika
Ceia + Deia = F keika
ia
Ce
ia
+ De
= F eika
Ceia + Deia = F keika

ia
2De
= F ( k)eika
k ia(k+)
e
D = F
2
(17.21)
Now that we have expressions for A,B,C,D all in terms of F, lets re-express
272
everything in terms of F. From before

A =
=
=
=
=
=
=
=
1
1
+ D 1
C 1+
2
k
2
k
1 k ia(k+)
1 k + ia(k)
+ F
F
e
1+
e
1
2
2
k
2
2
k
F
ia(k)
ia(k+)
(k + ) 1 +
e
+ ( k) 1
e
4
k
k
F (k + )(k + ) ia(k) (k )(k ) ia(k+)
e
e
4
k
k
F eika
(k + )2 eia (k )2eia
4k
F eika
k 2eia + 2keia + 2 eia k 2eia + 2keia 2 eia
4k
F eika
k 2 eia eia 2 eia eia + 2k eia + eia
4k
eia + eia
eia eia
F eika
4k
2i(k 2 + 2 )
4k
2
2i
This gives our desired expression for A

A=
F eika
4kcos(a) 2i(k 2 + 2 )sin(a)
4k
(17.22)
Next we do B
B =
=
=
=
=
=
=
1
1
+ D 1+
C 1
2
k
2
k
1 k + ia(k)
1 k ia(k+)
F
e
1
e
1+
+ F
2
2
k
2
2
k
F
ia(k)
ia(k+)
(k + ) 1
e
+ ( k) 1 +
e
4
k
k
F (k + )(k ) ia(k) (k )(k + ) ia(k+)
e
e
4
k
k
F
(k 2 2 )eia(k) (k 2 2 )eia(k+)
4k
2iF (k 2 2 ) ika eia eia
e
4k
2i
2iF (k 2 + 2 ) ika
e
4k
eia + eia
2i
273
giving us our desired expression for B
B=
iF (2 k 2) ika
e sin(a)
2k
(17.23)
Now (nally!) we can evaluate what the reection and transmission coecients are. From before the relevant uxes are
v1 |A|2 (leftgoing)
v1 |B|2 (rightgoing)
v3 |F |2 (transmitted)
The fraction of particles reected is

R =
=
v1 |B|2
B
=
v1 |A|2
A
F2
16(k)2
BB
A A
F 2 (2 k2 )2
sin2 (a)
4(k)2
(16(k)2cos2 (a) + 4(k 2 + 2 )sin2 (a))
R=
(2 k2 )2 sin2 (a)
4(k)2 cos2 (a)+(k2 +2 )sin2 (a)
(17.24)
The fraction transmitted is

T =
v3 |F |2
v1 |A|2
where note that v3 = v1 . Therefore

T
=
=
=
|F |2
F F
=
|A|2
A A
16(k)2
16(k)2 cos2 (a) + 4(k 2 + 2 )2 sin2 (a)
1
cos2 (a) +
(k2 +2 )2
sin2 (a)
4(k)2
(17.25)
Now we could stop here, but usually in many textbooks they expand this
out a little further. Recall that
2m
k2 =
h
2
2m( V )
2 =
h
2
2
4m
( V )
(k)2 =
h
4
274
This leads to
T
=
=
=
=
=
=
=
1
cos2 (a)
(2m+2m(V ))2
2
16m2 (V ) sin (a)
4m2 (2V )2
2
16m2 (V ) sin (a)
1
cos2 (a)
1
cos2 (a) +
(2V )2
2
4(V ) sin (a)
4( V )
4(
+ (42 4V + V 2 )sin2 (a)
4( V )
42 cos2 (a) 4V cos2 (a) + 42 sin2 (a) 4V sin2 (a) + V 2 sin2 (a)
4( V )
2 4V ) + V 2 sin2 (a)
(4
4( V )
4( V ) + V 2 sin2 (a)
V )cos2 (a)
This leads to the standard textbook expression

T =
1
V2
1+ 4(V ) sin2 (a)
(17.26)
Case 2: ( < V )
Now rather than doing all this math again, one can be smart about this.
Just note that = i. So take our previous results from case 1 and just
replace with i. (We just analyze the transmission coecient since this is
what we really care about). Another tortuous math workout. Starting with
the assumption of no sources on the right (G = 0), the wavefunctions are
1 = Aeikx + Beikx
2 = Cex + Dex
3 = F eikx
1 (0) = 2 (0)
A+B = C +D
(17.27)
275
Figure 17.6: Behavior of the wavefunction in the potential barrier case where
<V
Second one
1 (0) = 2 (0)
(C D)
AB =
ik
(17.28)
2 (a) = 3 (a)
a
Ce
+ Dea = F eika
(17.29)
Second one
2 (a) = 3 (a)
ik ika
Fe
Cea Dea =
(17.30)
276
Solve for A
A+B = C +D
AB =
(C D)
ik
+D 1
2A = C 1 +
ik
ik
C
A =
1+
+
1
2
ik
2
ik
(17.31)
Solve for B
A+B = C +D
A + B =
(C + D)
ik
+D 1+
2B = C 1
ik
ik
C
B =
1
+
1+
2
ik
2
ik
(17.32)
Solve for C
Cea + Dea = F eika
ik ika
Cea Dea =
Fe
2Cea = F
C =
F
2
ik
eika
ik
1+
eikaa
1+
(17.33)
Solve for D
Cea + Dea = F eika
ik
Cea + Dea = F eika
ik
eika
2Dea = F 1
F
ik ika+a
D =
1
e
2
(17.34)
277
Now we have a relation for A in terms of C and D. Replace the above
C and D expressions into A to obtain A as a function of F (we did B just
for the sake of completeness).
F
1+
4
F
2+
4
F ika
e
4
A =
=
=
ik
F
ik
1+
eikaa +
1
1
ik
4
ik
ik
ik
eikaa +
2
eika+a
+
ik
4
ik
ik
ik a
2+
+
ea + 2
e
ik
ik
eika+a
(17.35)
Now from before we know that the transmission coecient is

T
F
A
F F
A A
16
=
2+
ik
2+
2 k2
ik
+ 2
ik
ik
ik
2 k2
+ 2
ik
ik
+2 2+
16
ik
ik
+2 2+
2 k2
(k)2
2 k2 2
2 ((k)2 )
=
8
+2 4+
2( 2 k2 )2
(k)2
ik
ik
ik
ik
cosh(2a)
cosh(2a)
16
ik
cosh(2a)
16
+ 8+
2( 2 k2 )2
(k)2
cosh(2a)
it gets ugly here but can be simplied to give
16(k)2
8(k)2(1 + cosh(2a)) 2( 2 k 2)2 (1 cosh(2a))
Now use the relation cosh(2x) = 1 + 2sinh2 (x) to get
1
1+
( 2 +k2 )2
4(k)2
sinh2 (a)
where 2 = 2m(V2) , k 2 = 2m , and (k)2 =

h
h
2
our nal textbook expression
T =
1
V2
1+ 4(V ) sinh2 (a)
4m2 (V )
.
h
4
This leads us to
(17.36)
278
Again T = 0 and we obtain a non-classical result.

Exercises
1. A conduction electron in a semiconductor can be described as a free
particle in a constant potential with an eective mass me . Both
the constant potential and the eective mass are material dependent.
Consider the following idealized one dimensional system where semiinnite slabs of semiconductors A and B are joined together. (See Fig1
ure 17.7) Assume that one of the matching conditions is mA A (0) =
1
mB B (0). Also assume Vo > 0. (A) Find the transmission coecient
T for E > Vo for a right going wave. (B) Now assume that mB < mA .
Show that an appropriate choice of incoming energy E yields complete
transmission (T = 1). Is there such an energy if mB > mA . (C) Show
that for Vo = 0 and mA = mB , T is always less than unity.
Figure 17.7: Semiconductor junction
Chapter 18
The WKB approximation

Preliminaries
Here we derive the WKB form of the wavefunction. Starting with the
Schrodinger equation in the regime where > V
h
2 d2
+V
2m dx2
h
2 d2
2m dx2
d2
dx2
d2 2m
+ 2 ( V )
dx2
h
2 = 2m ( V ) so that we get
Let k
h
2
=
= ( V )
=
2m
( V )
h
2
= 0
(18.1)
d2
+ k 2 = 0
dx2
Now in general, if the potential is slowly varying, then both the amplitude
and the phase of the wavefunction will change. The general form of the
wavefunction is
(x) = A(x)ei(x) = Aei (shorthand)
(18.2)
So we insert this form of the wavefunction into Schrodingers equation above.

To begin we evaluate some derivatives that we will need. First derivative
d
= Aiei + ei A
dx
= (A + iA )ei
(18.3)
279
280
CHAPTER 18. THE WKB APPROXIMATION
Next we do the second derivative

d2
dx2
= i(A + iA )ei + ei (A + iA + i A )
= (iA A 2 )ei + (A + iA + i A )ei
= (2iA A 2 + A + iA )ei
= (A + 2iA + iA A 2 )ei
Replace this into the above Schrodinger equation;
d2
dx2
(18.4)
+ k 2 = 0
(A + 2iA + iA A 2 )ei + k 2 Aei = 0

A + 2iA + iA A 2 + k 2 A = 0
(18.5)
Group the real and imaginary terms. First the real terms
A A 2 + k 2 A = 0
A A 2 = k 2 A
(18.6)
Next the imaginary terms

2iA + iA
= 0
2A + A
= 0
(A )
= 0
(18.7)
Now we want to solve for A, so we can have an explicit wavefunction. Solve

the imaginary equation rst
(A2 )
= 0
= C 2 (a constant)
C2
A2 =
C
A =
(18.8)
The real part cannot be solve exactly. However, if A varies very slowly then
A is very small and can be ignored.
A A 2 = k 2 A
A 2 = k 2 A
2 = k 2
= k
kdx
(18.9)
281
Put these two expressions for A and back into our wavefunction
= Aei
C
= ei
k
kdx
(18.10)
This is the desired form of the WKB wavefunction in the case where > V .
In the case where < V the WKB form of the wavefunction is
C
= e
dx
(18.11)
which is basically the same as the rst case but where we just replace k with
i.
Arbitrary potential step
Using our just derived WKB wavefunctions we evaluate the following arbitrary potential step ( < V ).
Figure 18.1: Cartoon of the arbitrary potential step

The wavefunctions in region 1 and region 2 are, in general
1 =
2 =
0
0
A
B
ei x kdx + ei x kdx where (x < 0)
k
k
C x dx
D x dx
e0
+ e 0
where (x > 0)
282
Because we cant have the wavefunction blow up in region 2 (C = 0). The

wavefunctions are therefore
1 =
2 =
0
0
B
A
ei x kdx + ei x kdx where (x < 0)
k
k
D x dx
e 0
where (x > 0)
(18.12)
(18.13)
Now consider a linear approximation to the potential between region 1 and

region 2
V = +V x
(18.14)
First we determine the patching wavefunction in the boundary, no mans

land, between region 1 and region 2. Starting with the Schrodinger equation
2 d2 p
h
+ V p = p
2m dx2
where V = + V x. This reduces as follows
Let 3 =
2mV
h
2
h
2 d2 p
+ ( + V x)p
2m dx2
h
2 d2 p
+ V xp
2m dx2
d2 p 2mV
xp
dx2
h
2
d2 p
dx2
leading to
= p
= 0
= 0
=
2mV
xp
h
2
d2 p
= 3 xp
dx2
Now let z = x leading to
d2 p
= zp
dz 2
(18.15)
This is Airys equation whose solutions are called Airy functions. General
solution is a linear combination of Airy functions, Ai (z) and Bi (z). The
283
asymptotic forms of Airy functions are for (z < 0)
Ai (z) =
Bi (z) =
2
1
2
)
1 sin( (z) +
3
4
(z) 4
3
1
2
2
)
1 cos( (z) +
3
4
(z) 4
and for (z > 0)

Ai (z) =
Bi (z) =
3
1
2z2
1e 3
2 z 4
3
2 2
1
z
1 e3
z 4
Now
p = aAi (z) + bBi (z)
(18.16)
where a and b are constants. Now in region 1

2m( V )
h
2
k=
where V = ( + V x)
2mV
x
h
2
3
= 2 x
k =
where recall that =

In region 2
2mV
h
2
1
3
(18.17)
2m(V )
h
2
2mV
x
h
2
3
= 2 x
(18.18)
Now go back to the wavefunctions in region 1 and region 2 and make

them more explicit.
284

Region 1
A
1 = ei
k
3
and k = 2 x
1 =
A
3
4
(x)
i
1 e
4
0
x
0
x
kdx
B
+ ei
k
xdx
0
x
kdx
3
4
(x)
i
1 e
0
x
xdx
evaluate integral by letting y=-x, doing the integral and then renaming y as x
=
=
A
3
2i
(x) 2
3
+
1 e
2i
3 (x) 2
1 e
4 (x) 4
4 (x) 4
3
3
2i
2i
B
A
e 3 (x) 2 +
e 3 (x) 2
1
1
(x) 4
(x) 4
Region 2
D
2 = e
3
where = 2 x
2 =
3
4
1
4
(18.19)
x
dx
0
x
0
xdx
x
3
D
2 (x) 2
3
1 e
(x) 4
(18.20)
Now at this point, compare these explicit wavefunctions to the patching

wavefunctions in regions 1 and 2.
Region 2 (x > 0)(z > 0)
2 3
2 3
a
b
=
e 3 z 2 + 1 e 3 z 2
1
2 z 4
z 4
compared to
2 =
3
2
D
e 3 (x) 2
1
(x) 4
By inspection, one can tell immediately that

b = 0
a = 2D
(18.21)
(18.22)
285
Now go back and do the same comparison in region 1
Region 1 (x < 0)(z < 0)
p = aAi (z)
Ai (z)
a
p = 2D
Ai (z) =
3
1
2
2
)
1 sin( (z) +
3
4
(z) 4
3
3
2
i( 2 (z) 2 + )
i( 3 (z) 2 + )
3
4 e
4
1
e
1
2i
(z) 4
(z) 4
yielding
p =
(z) 4
1
(x)
Now by inspection
3
3
2
ei 4 i 2 (z) 2
ei 4
e3
ei 3 (z) 2
2i
2i
3
3
Dei 4 2i
Dei 4 2i (z) 2
e3
e 3 (z) 2
i
i
versus
1 =
where
1
4
3
3
2i
A 2i
B
e 3 (x) 2 + e 3 (x) 2
A = iDei 4
(18.23)
i
4
B = iDe
(18.24)
Now put everything together for the WKB wavefunctions we showed at the
beginning of the section
0
0
B
A
1 = ei x kdx + ei x kdx where (x < 0)
k
k
D x dx
2 = e 0
where (x > 0)
where now A = iDei 4 and B = iDei 4 to get

1 =
2 =
2D
sin( +
4
k
D x dx
e 0
kdx)
(18.25)
(18.26)
286
Arbitrary potential drop
Figure 18.2: Cartoon of the arbitrary potential drop
Using our derived WKB wavefunctions we have the general form in regions 2 and 3
2 =
3 =
0
C 0
D
e x dx + e x dx where (x < 0)
F i x kdx
G i x kdx
e 0
+ e 0
where (x > 0)
k
k
Since we cant have the wavefunctions behave badly, C = 0 to prevent it

from blowing up in region 2. This leaves us with t wavfunction forms
2 =
3 =
0
D
e x dx
F i x kdx
G
e 0
+ ei
k
k
(18.27)
x
kdx
0
Now consider an approximation to the potential in the region between 2 and

3 of the linear form
V = ( V x)
(18.28)
First we determine the patching wavefunctions in the region between 2 and 3.

Starting with Schrodingers equation (if you went thorough the last section
287
all of what follows will be familiar)
h
2 d2 p
+ V p
2m dx2
h
2 d2 p
+ ( V x)p
2m dx2
h
2 d2p
V xp
2m dx2
d2 p 2mV x
+
p
dx2
h
2
Let 3 =
2mV
h
2
= p
= p
= 0
= 0
leading to
d2 p
+ 3 xp = 0
dx2
Now let z = x
d2 p
= zp
(18.29)
dz 2
Again, this is the Airy equation with accompanying Airy functions. The
general solution is a linear combination of the form
(18.30)
where a and b are constants.

In region 2
=
2m(V )
h
2
where V = V x
2mV
x
h
2
3
= 2 x
(18.31)
In region 3
k =
2m( V )
h
2
2mV
x
h
2
3
= 2 x
(18.32)
288
Now with and k go back and make more explicit our wavefunctions in
regions 2 and 3.
Region 2
2 =
D
e
D
3
4
0
x
(x)
dx
1
4
e 2
1
0
(x) 2 dx
x
solve the integral by letting y=-x, integrating and then renaming y=x
Region 3
3 =
=
=
3
D
2 (x) 2
3
1 e
(x) 4
x
x
F
G
ei 0 kdx + ei 0 kdx
k
k
x 3 1
x 3 1
G
F
2 x 2 dx
0 2 x 2 dx
0
+
1 e
1 e
(x) 4
(x) 4
3
3
2i
F
G
(x) 2
2i (x) 2
3
3
+
1 e
1 e
(x) 4
(x) 4
(18.33)
(18.34)
Now compare these explicit wavefunctions to the patching wavefunctions in

regions 2 and 3
Region 2 (x < 0) (z > 0) since (z = x)
3
3
2 2
a
b
2z 2
z
=
1 e 3 + 1 e3
2 z 4
z 4
versus
2 =
By insepction
3
2
D
3 (x) 2
1 e
(x) 4
b = 0
a = 2D
(18.35)
(18.36)
289
Region 1 (x > 0) (z < 0) since (z = x)

p = aAi (z)
Ai (z)
= 2D
=
3
2
sin( (z) 2 + )
3
4
(z)
3
3
i 2i (z) 2
i 2i (z) 2
4e 3
4e
3
e
a
e
1
2i
(z) 4
1
4
(z) 4
versus
3 =
1
(x)
1
4
3
3
aei 4 2i (z) 2
aei 4 2i (z) 2
e3

e 3
2i
2i
3
3
2i
G
F 2i
e 3 (x) 2 + e 3 (x) 2
By inspection
F
= iDei 4
(18.37)
i
4
(18.38)
G = iDe
Now put everything together for the WKB wavefunctions we saw at the
beginning of the section
2 =
3 =
0
D
e x dx
F i x kdx
G
e 0
+ ei
k
k
x
0
kdx
where now F = iDei 4 and G = iDei 4 leading to

2 =
3 =
0
D
e x dx
2D
sin( +
4
k
(18.39)
kdx)
(18.40)
Arbitrary potential barrier
This next section basically puts together all the work we did in the previous two sections for the arbitrary potential step and potential drop. The
combination of both is the potential barrier.
290
Figure 18.3: Cartoon of the arbitrary potential barrier
The solution for the transmission coecient will lead to the general form
of the tunneling probability that is ubiquitous in textbooks, in the literature
and in experiments like scanning tunneling microscopy. The derivation involves a lot of tortuous math but follows the same train of thought as used
in the earlier two sections.
Picture an arbitrary barrier of the form shown below. Now the general
form of the WKB wavefunctions as derived earlier is
1 =
2 =
3 =
x1
x1
A
B
ei x kdx + ei x kdx where (x < x1 )
k
k
x
x
D x dx
C x dx
e 1
+ e 1
where (x1 < x < x2 )
F i x kdx
e x2
where (x > x2 )
k
(18.41)
(18.42)
(18.43)
Since there are no sources of particles on the right everything moves in a

left to right direction and hence 3 has only one component above.
Remember in what follows that ultimately what we want to solve for is
T =
k|F |2
|F |2
v3 |F |2
=
=
v1 |A|2
k|A|2
|A|2
You will nd that our strategy is to express A in terms of F .
291
First we determine the patching wavefunction in the boundary region
between 1 and 2. Starting with the Schrodinger equation
h
2 d2 p
+ V p
2m dx2
h
2 d2 p
+ ( + V x)p
2m dx2
h
2 d2p
+ V xp
2m dx2
d2 p 2mV x
p
dx2
h
2
Let 3 =
2mV
h
2
= p
= p
= 0
= 0
resulting in
d2p
= 3 xp
dx2
Now let z = x
d2 p
= zp
dz 2
As before, this is the Airy equation whose general solution is a linear combination of Airy functions
Now in the left part of region 2 (bordering region 1)
=
2m(V )
h
2
2mV
x
h
2
3
= 2 x
(18.44)
In the right part of region 1 (bordering region 2)

k =
2m( V )
h
2
2mV
x
h
2
3
= 2 x
(18.45)
292
Now go back to the wavefunctions in regions 1 and 2 and make them

more explicit using the above. Starting with
x1
x1
A
B
ei x kdx + ei x kdx
k
k
x
x
D x dx
C x dx
e 1
+ e 1
1 =
2 =
temporarily shift the origin to x1 (i.e. x1 = 0). Then this basically becomes
the same problem that we solved earlier. After we are done we can shift
back the origin to where it used to be. First wavefunction
A
1 =
3
4
(x)
1
4
ei
0
x
xdx
3
4
(x)
1
4
ei
0
x
xdx
solve integral by letting y=-x, integrate then rename y=x
A
(x)
Second wavefunction
2 =
e 3 (x) 2 +
1 e
(x) 4
C
x
0
xdx
(x) 2
3
1 e
(x) 4
2i
1
4
(x)
2i
1
4
e 3 (x) 2
1 e
x
0
(18.46)
(x) 4
3
2
D
e 3 (x) 2
+
1
(x) 4
xdx
(18.47)
Now compare these explicit wavefunctions with the patching wavefunction.

Left part of region 2 (bordering region 1) (x > 0) (z > 0)
2 3
2 3
b
a
e 3 z 2 + 1 e 3 z 2
=
1
2 z 4
z 4
versus
2 =
By inspection
3
3
2
C
D
2 (x) 2
(x) 2
3
3
+
1 e
1 e
(x) 4
(x) 4
a = 2D
b = C
(18.48)
(18.49)
293
Right part of region 1 (bordering region 2) (x < 0) (z < 0)
Ai (z) + C
Bi (z)
2
1
2
)+C
1 sin( (z) +
(z) 4
3
4
= 2D
= 2D
=
2
1
2
)
1 cos( (z) +
(z) 4
3
4
3
3
2
2
2D
2
2
)+
)
1 sin( (z) +
1 cos( (z) +
3
4
3
4
(z) 4
(z) 4
skipping some steps
=
+
=
+
1
(z) 4
1
1
(z) 4
1
2
2D
+C
i
1
2
ei 4
2
ei 4
2
1
(z) 4
2D
+C
i
ei( 4 + 3 (z) 2 )
2D
+C
i
1
(z) 4
ei( 4 + 3 (z) 2 )
e 3 (z) 2
2i
2D
+C
i
2i
e 3 (z) 2
compared to
1 =
3
3
2i
A
B
(x) 2
2i (x) 2
3
3
+
1 e
1 e
(x) 4
(x) 4
By inspection
A =
ei 4
2
B =
ei 4
2
2D
+C
i
2D
+C
i
(18.50)
(18.51)
Now consider the interface between regions 2 and 3. The potential here
can be modeled in a linear fashion
V = ( V x)
(18.52)
First determine the patching wavefunction in the boundary region. Start
294
with Schrodingers equation again
Let 3 =
2mV
h
2
h
2 d2 p
+ V p = p
2m dx2
h
2 d2 p
+ ( V xp ) = p
2m dx2
h
2 d2 p
V xp = 0
2m dx2
d2 p 2mV x
xp = 0
+
dx2
h
2
giving
d2p
+ 3 xp = 0
dx2
Now let z = x resulting in

d2 p
= zp
dz 2
This is the Airy equation again with the general solution being a combination
of Airy functions.
Now in the right part of region 2 (bordering region 3)
=
2m(V )
h
2
2mV
x
h
2
3
= 2 x
(18.53)
were recall V = V x
Now in the left part of region 3 (bordering region 2)
k =
2m( V )
h
2
2mV
x
h
2
3
= 2 x
(18.54)
295
Now go back to the wavefunctions in regions 2 and 3 and make them

explicit.
2 =
3 =
C x dx
D
e x1
+ e
x
F i kdx
e x2
k
x
x1
dx
rearrange wavefunctions to get
2 =
3 =
now let =
x2
x1
C x2 dx+
e x1
x
F i x kdx
e 2
k
x
x2
dx
D
+ e
x2
x1
dx
x
x2
dx
dx
2 =
3 =
x
D
C
dx
e e x2
+ e e
x
F i x kdx
e 2
k
x
x2
dx
(18.55)
(18.56)
Now temporarily translate the origin to x2 (i.e. x2 = 0). This leads to

2 =
3 =
x
D
C
e e 0 dx + e e
x
F
ei 0 kdx
k
x
0
dx
Now we use the derived explicit forms for and k to get explicit forms for
the wavefunctions. First wavefunction
3
x 3
C
D
x 2 xdx
0 2 xdx
0
2 =
+
e
1 e e
1 e
(x) 4
(x) 4
solve the integral by letting y=-x, integrating, then rename y=x
3
3
2
2
De
Ce
3 (x) 2
(x) 2
3
+
1 e
1 e
(x) 4
(x) 4
(18.57)
Second wavefunction
3 =
=
i
1 e
x
0
(x) 4
3
2i
F
e 3 (x) 2
1
(x) 4
xdx
(18.58)
296
Now compare these explicit wavefunctions to the patching wavefunction

Right side of region 2 (bordering region 3) (x < 0) (z > 0)
3
3
2 2
a
b
2z 2
z
=
1 e 3 + 1 e3
2 z 4
z 4
compared to
2 =
3
3
2
2
De
(x) 2
3
e 3 (x) 2 +
1 e
(x)
(x) 4
Ce
1
4
By inspection
a = 2C
b = D
(18.59)
(18.60)
Left side of region 3 (bordering region 2) (x > 0) (z < 0)


e Ai (z) + D
e Bi (z)
3
3
2Ce
2
2
De
2
2
)+
)
1 sin( (z) +
1 cos( (z) +
3
4
3
4
(z) 4
(z) 4
= 2C
=
=
+
2Ce
i
4
3
2i
(z) 2
3
i 2i (z) 2
4
3
e
e
e e
1
2i
(z) 4
3
3
i 2i (z) 2
i 2i (z) 2
4e 3
4e
3
De
+e
1
2
(z) 4
1
1
(z) 4
1
1
(z) 4
compare this to
1
2
2Ce
+ De ei 4
i
1
2
2Ce
+ De ei 4
i
2i
e 3 (z) 2
2i
e 3 (z) 2
297
3 =
3
2i
F
(x) 2
3
1 e
(x) 4
By inspection
ei 4
2
2Ce
+ De
i
(18.61)
or more importantly
2Ce
+ De ei 4 = 0
i
1
2
yielding
C=
iF i
4e
2 e
(18.62)
D = F ei 4 e
(18.63)
Finally these expression can be inserted into our expression for A derived
earlier.
ei 4
2
2D
+C
i
= iF (
e
e )
4
A =
(18.64)
Now the transmission coecient is

T
=
=
v3 |F |2
k|F |2
|F |2
F F
=
=
=
v1 |A|2
k|A|2
|A|2
A A
2
1
F
2 =
2
F 2 e 4 e
4 e
1
e2
4
e2
=
e2
1
e2
4
1
2
e2
e2
4
298
So after all this eort, the transmission coecient has the form
T =
e2
1
e2
4
(18.65)
which in the case of large reduces to the ubiquitous textbook and literature
expression
T e2
where =
x2
x1
dx and =
2m(V )
.
h
2
(18.66)
No mas!
Example: Field emission

Also known as Fowler Nordheim tunneling. Assume a triangular form of the
barrier where
(V ) = 1
x
a
(18.67)
Figure 18.4: Sketch of the triangular barrier considered in eld emission or

Fowler Nordheim tunneling.
From the WKB approximation derived above
T e2
299
where =
a
0
2m(V )
dx.
h
This leads to
2m
2
h 2m
= e h
= e
2
h
= e
Integrate the exponent to get
2
2m
a
0
2m
V dx
a
0
x
(1 a )dx
a
0
x dx
a
2
a
2m
h
4 2m a 3
2
3
h
4 2m 1 3
2
3 E
h
where E =
a
0
3
2
2
3
is the electric eld. This leads to our nal expression

4
3
T =e
2m 2
h
(18.68)
You can see that the tunneling probability depends to the 3/2 power of the
barrier height. This is the underlying relation behind the Fowler Nordheim
tunneling model.
Example: Schottky barrier
This problem arises when one has a metal-semiconductor junction. Here

x
a
(V ) 1
(18.69)
From the WKB approximation
T = e2
where =
a
0
2m(V )
dx.
h
This leads to
2
h
= e
a
0
2m 1( x )
a
2
h 2m
= e
dx
x
1( a )2 dx
300
Figure 18.5: Sketch of the Shottky barrier
Let y =
as well.
x
a
where also dx = ady. Dont forget the limits of integration change
1
2m 0
1y 2 ady
2
1
2
1y dy
h a 2m 0
= e h
= e
To integrate, make another change of variables. Let y = sin(); dy =

cos()d
2
T = e h a
2m
1
where cos2 () = 2 (1 + cos(2))
2
T = e h a
1
2m 2
giving our nal expression
a
2
T =e
2
0
2
0
cos2 ()d
(1+cos(2))d
2m
h2
(18.70)
Here one notices that the tunneling probability varies as the square root of
1
the barrier height, 2 .
301
Example: Symmetric parabolic barrier

Here
(V ) = C
a2
x2
4
(18.71)
where C is a constant with appropriate units.
Figure 18.6: Sketch of an arbitrary symmetric parabolic barrier
From the WKB approximation, the tunneling probability through this

barrier is
T = e2
where =
a
0
2m(V )
dx.
h
This leads to
2
h
T =e
Evaluate the exponent
a
2
a
2
2 2mC
2mC
a
2
a
2
a2
4
x2 dx
a2
x2 dx
4
302
Let x =
becomes
a
2 sin,
dx =
a
2 cosd.
Also change the limits. The expoenent
2
2 2mC
a2 a2
a
sin2
cos d
4
4
2
2
2
2 2mC
a 2
1 sin2 (cos)d
=
2
2
2 2mC a2 2
cos2 d
=
h
4
2
2
2
2mCa
cos2 d
=
2
h
2
where
cos2
1
2 (1 +
cos2)
2mCa2
=
4
h
2mCa2
=
h
4
2m
=
h
C
where the barrier height is = C
T =e
a2
4
1 + cos2d
. This leads to our nal expression
2m
C h
(18.72)
where the tunneling probability depends upon alone rather than to some
power of it.
Exercises
1. Derive the transmission probability of a particle through a rectangular
barrier of height Vo and width l.
2. Assume a rectangular barrier of height 4 eV and width 2 nm. For
a free electron what is the tunneling probability through this barrier
using a WKB formalism.
3. For the same 4 eV barrier and electron mass, what is the tunneling
probability when the barrier width is increased by 50%. What is the
probablity when the barrier width is decreased by 50%.
303
4. For the same 2 nm wide barrier and electron mass, what is the tunneling probability when the barrier height is increased by 50%. What
is the probability when the barrier height is decreased by 50%.
5. Use the WKB approximation to nd the transimssion coecient for
the potential V (x) = 0 if x < 0 and V (x) = Vo kx when x > 0. Here
Vo and k are constants. Refer to gure 18.7. Hint: Find the turning
points of the system where V (x) = to nd the limits of integration.
6. Calculate the transmission coecient of the following potential barrier
2
using the WKB approximation. The potential is: V (x) = Vo 1 x2
a
when a <= x <= a. Also V (x) = 0 elsewhere. See gure 18.8. Use
the same hint about the turning points and the integration limits as
described in the previous problem.
7. Consider the classic alpha decay problem using the just derived WKB
2Ze2
approximation. The potential is V (r) = 4 or for r > r1 where r1 is the
radius of the nuclei. Carry the WKB integral (Equation 12.66) from
2Ze2
r1 to r2 where r2 = 4 oE and determine the tunneling probability.
Make suitable approximations to simplify things as much as possible.
Figure 18.7: Sketch of a linearly decaying barrier
304
Figure 18.8: Sketch of a symmetric parabolic barrier
Relevant literature
These references are listed in no particular order
A carbon nanotube eld-emission electron source
W. A deHeer, A. Chatelain, D. Ugarte
Science, 270, 1179 (1995).
Field emission from carbon nanotubes: the rst ve years
J-M Bonard, H. Kind, T. Stockli, L-O Nilsson
Solid-State Electronics, 45, 893 (2001).
Stripping voltammetry of Cu overlayers deposited on self-assembled
monolayers: eld emission of electrons through a phenylene ethylene
oligomer
M. S. Doescher, J. M. Tour, A. M. Rawlett, M. L. Myrick
J. Phys. Chem. B, 105, 105 (2001).
Time resolved photoelectrochemistry with size-quantized PbS adsorbed on gold
305
E.P.A.M. Bakkers, J.J. Kelly, D. Vanmaekelbergh
Journal of Electronanalytical Chemistry, 482, 48 (2000).
Fluorescence intermittency in single InP quantum dots
M. Kuno, D. P. Fromm, A. Gallagher, D. J. Nesbitt, O. I. Micic, A.
J. Nozik
Nano Lett. 1, 557 (2001).
On/o uorescence intermittency of single semiconductor quantum
dots
M. Kuno, D. P. Fromm, H. F. Hamann, A. Gallagher, D. J. Nesbitt
J. Chem. Phys. 115, 1028 (2001).
306
Chapter 19
Synthesis
Molecular Beam Epitaxy (MBE)
The basic idea behind this technique is fairly simple. In practice, however,
its realization is more involved (and more expensive). MBE essentially consists of an ultrahigh vacuum chamber into which a substrate is loaded onto
a heated sample holder. Precursors of desired elements (Ga, As, Al, P, In
etc...) are then loaded into heated crucibles or furnaces (called Knudsen
cells) outtted with computer controlled shutters on their exits. The precursors are then heated such that when the shutters are opened one obtains
a beam of atoms directed towards the substrate. Under such low pressures,
the atomic species have very long mean free paths allowing them to reach
the substrate without collisions with other gas phase species in the chamber.
By controlling the temperature as well as the sequence/timing of opening
and closing the shutters one can deposit very uniform lms of semiconductor materials. In this fashion one can obtain precise nanometer lengthscale
quantum well structures.
Under certain conditions where one deposits distinct semiconductor layers with dierent lattice constants, its also possible to grow islands of a
semiconductor on top of another eectively allowing the synthesis of semiconductor quantum dots. The basic idea is that after a few layers of the
new lattice mismatched semiconductor has been deposited, the strain at the
interface changes the mode of growth from within the plane to out of the
plane. Small islands are therefore formed which are the desired quantum
dots. This technique is refered to as Stranski Krastanow growth.
A crude sketch of a MBE apparatus is shown below.
307
308
CHAPTER 19. SYNTHESIS
Figure 19.1: Cartoon of a MBE apparatus
Metal Organic Chemical Vapor Deposition (MOCVD)

MOCVD operates on a similar principle to MBE. However the apparatus
diers greatly. MOCVD is generally conducted using a quartz tube furnace
with a heated substrate. Clearly this is a much less expensive setup. One
doesnt need turbo pumps, vacuum chambers, load locks, EELS guns and
other MBE accoutrements. Organometallic or metal-organic compounds
such as trimethylaluminum, trimethygallium, trimethylindium as well as
gases such as phosphine or arsine can be introduced into the heated reactor
and allowed to decompose giving the desired elemental species. In the case
of liquid organometallic precursors such as trimethylaluminum (above), the
compound is brought into the gas phase by bubbling a carrier gas such as
helium through the compound and allowing the gas to carry small amounts
of the precursor into the reaction chamber. Like MBE, MOCVD can be
used to grow thin lms of materials. This technique has also been used,
more recently, in the synthesis of semiconductor nanowires in the presence
of gold nanoparticle catalysts. One potential disadvantage of the technique
is the uniformity of the resulting lms or deposition of materials. This
is because the ow of gasses above the substrate may not be completely
uniform. As a consequence one should consider ways to better control the
ow and subsequent distribution of precursors over the substrate.
A sketch of the apparatus is shown below.
309
Figure 19.2: Cartoon of a MOCVD apparatus
Colloidal growth
Description
Colloidal growth is a chemists approach to making nanostructures. There
are dierent variations of this approach so no general description can t all
techniques. However a few are highlighted here. Some of the original quantum dots were created in supersaturated solutions within a glass matrix.
Basically, molten silicate glasses were doped to the point of supersaturation with metal salts of the desired semiconducting material. The melt is
then rapidly quenched, resulting in a precipitation of the desired semiconducting material into tiny seed nuclei (or alternatively the rapid quench
can be thought of as a discrete temporal nucleation of seed particles). The
glass solution then undergoes a secondary heat treatment with temperatures
ranging from 400 to 1000 degrees Celcius. By varying the temperature and
duration of the secondary heat treatment one can vary the average size of
the nanocrystals.
Alternatively, others have employed inverse micelles as a means of quantum dot or even nanowire synthesis. In this preparation, surfactants such
as bis(2-ethylhexyl)phosphate (also called AOT) are used to create small inverse micelles consisting of a hydrophilic interior and a hydrophobic exterior.
Aqueous solutions of metal salts are then introduced into these compart-
310
mentalized water pools. Subsequent reactions are conducted in the aqueous

phase (whether it be reduction of the metal precursor using sodium borohydride or reaction with a second chalcogen source such as S, Se or Te) to create
metal or semiconductor nanoparticles. After the reaction, the nanocrystallites are sometimes extracted by adding a surface passivating agent which
drops them out of solution. The recovered powder is then redissolved in a
suitable solvent. To a rst approximation, the average size of the nanocrystallies is determined by the initial ratio of water to surfactant, often called
the W or value.
One of the most successful approaches to the colloidal synthesis of nanocrystal and nanorods involves using coordinating as well as non coordinating
solvents. In this approach organometallic precursors such as dimethylcadmium and trioctylphosphine selenide are injected into hot (temperatures on
the order of 300 degrees Celcius) trioctylphosphine oxide. Upon injection,
the precursors decompose to give desired elements of the nal semiconductor. The rapid injection is analogous to the rapid quench in glasses and
results in a discrete temporal nucleation of seed particles. The temperature
of the solution is then slowly raised to allow the controlled growth of particles
in the coordinating (or non-coordinating) solvent. This, again, is analogous
to the secondary heat treatment with glasses with the main dierence being
lower overall temperatures (300 degrees versus 400 to 1000 degrees C). The
average size of the nanomaterial is determined by the temperature and the
duration of the heating and can be monitored spectroscopically.
A gure of such an apparatus is shown below.
LaMer and Dinegar growth model

Start with Ficks rst law.
dc
Q(t) = 4r2 D dr
(19.1)
where Q(t) is the ux of stu going to make the particle, D is the diusion
coecent and r is a radial length from the center of the growing particle.
We should point out that this ux has units of stu/time as opposed to
the more usual denition of ux. Now with this lets manipulate the rst
expression a little bit to get.
311
Figure 19.3: Cartoon of an apparatus commonly used for colloidal synthesis
312
Q(t)
dc
=
2D
4r
dr
Integrate the last expression
Q(t)dr
= dc
4r2 D
Q(t)
4D
to get
dr
=
r2
dc
Q(t)
+ Const = C(r, t)
4Dr
This constant is a constant with respect to r. However it could, in general,
depend on time. So more generally we get
Q(t)
+ f (t) = C(r, t)
(19.2)
4Dr
Apply boundary conditions now to make more explicit the expression. There
are four to consider.
When r = a, where a is the radius of the growing particle, C(a, t) = Cs .
Cs is the saturation concentration of the stu being deposited, called
monomer.
C(r, t = 0) = Css . Here Css is the supersaturation concentration of
the solution at t = 0.
C
t |r=h
Q(t)
4
h3
3
Q(0) = 0 Flux is zero at zero time as expected.

Apply boundary condition 2
C(r, t = 0) = Css
Q(0)
C(r, 0) =
+ f (0)
4Dr
Since Q(0) = 0 (boundary condition 4)
C(r, 0) = f (0) = Css
(19.3)
313
Start with the original expression
C(r, t) =
Q(t)
+ f (t)
4rD
take its time derivative

C(r, t)
t
=
r=h
1 dQ(t)
4rD dt
+
r=h
Q(t)
df (t)
= 4 3
dt
3 h
or on rearranging
df (t)
Q(t)
1 dQ(t)
= 4 3
dt
4hD dt
3 h
Now integrate this to get
f (t)|t =
0
1
4
3
3 h
Q(t)dt
1
Q(t)
4hD
t
0
This becomes
f (t) f (0) =
1
4
3
3 h
t
0
Q(t)dt
1
(Q(t) Q(0))
4hD
where f (0) = Css and Q(0) = 0. The reduces to

f (t) = Css +
Since C(r, t) =
Q(t)
4rD
1
4
3
3 h
t
0
Q(t)dt
Q(t)
4hD
+ f (t) we can use the above expression for f (t) to get
C(r, t) =
=
1
Q(t)
+ Css + 4 3
4rD
h
3
Q(t)
4D
1 1
r h
t
0
+ Css +
Q(t)dt
1
4
3
3 h
Q(t)
4hD
Q(t)dt
0
Since in general h >> r we get
C(r, t)
Q(t)
4D
1
r
+ Css +
1
4
3
3 h
Q(t)dt
0
314
Now if
Q(t) = 4a2
da
dt
(19.4)
(a reasonable thought given that the ux in the opposite direction is reected

by the growth of the particle) one gets after replacing this in the previous
expression, where is the density of the material being deposited
C(r = a, t)
4a2 da
4
dt
+ Css 4 3
4Da
3 h
a2
da
dt
dt
or
C(r = a, t)
a da
3
dt
+ Css 3
D
h
a2
da
dt
dt
Integrate the last integral by parts. Let u = a2 , du = 2ada, dv =

v = a. We get
C(r = a, t)

a da
3
dt
+ Css 3
D
h
a
D
da
dt
+ Css
a3
3
h3
t
0
da
dt
and
2a2 da
a3 (t) a3 (0)
2 3
a
3
t
0
Recall that at t = 0, a(0) = 0 Particle has not grown yet. This allows us to
simplify the above expression to
a
D
a

D
a

D
da
dt
da
dt
da
dt
3
2 3
a3 (t)
a (t) a3 (0)
3
h
3
3 a3 (t)
+ Css 3
h
3
3
a
+ Css 3
h
+ Css
Leading to our nal expression
C(r = a, t) Css
a
D
da
dt
a3
h3
(19.5)
315
Here apply it to the nal expression we just derived above.
C(r = a, t) Css
a
D
da
dt
a3
= Cs
h3
Rearrange to give
a
D
da
dt
= (Css Cs )
a3
h3
or
da
D
Da3
= (Css Cs ) 3
dt
h
2
At this point note that d(a ) = 2a da so that a da =

dt
dt
dt
the above expression to get
d(a2)
2D
2Da3
=
(Css Cs )
dt
h3
1 d(a2 )
2 dt .
Insert this into
(19.6)
This last last expression gives you the behavior for the size of the particle as
a function of time, which is what we were ultimately after. We can proceed
to solve this equation by either looking this up in a table of integrals, which
is what LaMer and Dinegar ultimately did, or we could solve it numerically.
a
Another option is to note that in general h << 1 so basically the last term
in the expression drops out. We get the approximation
2D
d(a2 )
(Css Cs )
dt
(19.7)
You will notice that the

radius of the particle with essentially grow as the
square root of time (a t). Now following Sugimoto we invoke the Gibbs
Thomson equation as follows
2Vm
Css = C e r RT
Cs = C e
2Vm
aRT
316
Here a is the radius of the particle at a given time. r is the radius of the
particle that would make its ensemble concentration equivalent to the initial
monomer concentration (think about this point, its will be important in
a bit). To simplify the notation, let ro = 2Vm
RT
ro
Css = C e r
Cs = C e
ro
a
(19.8)
(19.9)
If the exponent is small (and it doesnt have to be) then we can do a Taylor
series expansion of the above to get (keeping only 1st 2 terms)
r0
+ ...)
r
r0
CS C (1 +
+ ...)
a
Css C (1 +
Replace into our main equation to get

2D
d(a2 )
dt
C r0
1
1
r
a
2D
da
=
C r0
dt
1
1
r
a
(19.10)
or alternatively
2a
da
r0
D
= C
dt
1
1
r
a
Let
K=
DC r0
leading to
da
dt
K
a
1
r
1
a
(19.11)
Now qualitatively speaking, just look at the sign of the right hand side,
basically determined but the stu in the parenthesis. You see if a = r no
growth occurs ( da = 0). If a < r then da is negative and basically your
dt
dt
317
particles dissolve (negative growth). Now if a > r then
your particles grow.
da
dt
is positive and
A graph of this equation is shown below. You will see that small particles
have a steeper slope or faster rate of growth. Larger particles have a atter
slope so they grow slower.
Figure 19.4: Sketch of the LaMer growth rate as a function of the critical
radius r
Now one underlying point of this whole picture is that C is constant

(i.e. r is constant). In reality though, since this is a closed system, the
monomer concentration decreases as it gets consumed. So r is actually
changing in time. If we think in dynamic terms as t , r and one
actually falls at dierent places on the above curve as a function of time.
Basically your particle size will correspond to a steadily decreasing factor of
r and you move to the left on the curve. If you have a situation where da
dt
becomes negative then your particles will start to dissolve. The worst case is
that you have sizes that straddle r at any given time so that some particles
grow, others dissolve and you end up with a very poor size distribution of
your sample.
318
Size distribution
This section follows the work of Sugimoto who argued that the size distribution of the ensemble is proportional to
size dist
da
dt
(19.12)
da
Using the relation derived earlier

da
dt
= K
da
K
a2
da
da
r a2
2
1

a r
da
2
3
a
da
da
da
where
Divide both sides by

da
dt
dt
da
da
dt
da
da
dt
dt
da
dt
dt
da
to get
Let =
da
dt
K
a2
1
2
a r
da
dt
K
a2
1
2

a r
da
dt
giving
d
K
= 2
dt
a
2
1

a r
(19.13)
Alternatively
d
dt
K
a2
2r a
r a
(19.14)
Now its clear to see that if a > 2r , d is negative valued. The size
dt
distribution of the sample will narrow. This is called the focusing regime.
319
Figure 19.5: Sketch of both the LaMer growth rate and size distribution as
a function of the critical radius r
However, if a < 2r d is positive valued and the size distribution will

dt
increase. This is called the defocusing regime.
Recall our discussion earlier about the dynamic nature of r . It goes
without saying that in a closed system as time increases r will also increase. The actual size of the particle will in absolute terms become a
smaller fraction of r causing you to progressively move left on our diagram.
As you move left the distribution will broaden. To keep the size distribution
or even narrow it you need to ght the depletion or monomers causing r
to increase. This means as the reaction progresses you add more precursor
to the reaction. The extra addition can be accomplished a number of ways.
Peng for example just adds subsequent injection of precursors into the reaction mixture. Sugimoto and others however build in a resevoir of precursors
that slowly get released to the reaction as time increases.
Reaction controlled growth

The previous discussion has assumed diusion limited growth. However its
possible to have the reaction controlled situation. LaMer, Sugimoto and
others have shown that in this case the relevant growth equation can be
expressed as
320
1
1
r
a
da
= Kr
dt
(19.15)
where Kr is a constant. To see how the size distribution behaves in this

situation repeat the analysis done earlier for the diusion controlled growth
case.
size dist
da
dt
da
When explicitly evaluated this leads to

da
dt
da
Kr
a2
da
da
where recall that

da
da
da
dt
Replace into our expression and divide by

da d da
Kr
dt
= 2
dt da
a
d
da
dt
dt
Let =
da
dt
Kr
a2
dt
da
dt
da
on both sides. This gives
da
dt
da
dt
giving
Kr
d
= 2
(19.16)
dt
a
You will notice here that this expression is always positive. In eect it tells
us that there will always be some broadening of the size distribution during
the particle growth.
Exercises
1. Read any of the papers listed below or select a paper from the current
literature and discuss it.
321
Relevant reading
1. Theory, production and mechanism of formation of monodispersed
hydrosols V. K. LaMer, R. H. Dinegar J. Am. Chem. Soc. 72, 4847
(1950).
2. Preparation of monodispersed colloidal particles T. Sugimoto Advances in colloid and interface science, 28, 65 (1987).
3. The kinetics of precipitation from supersaturated solid solutions I.
M. Lifshitz, V. V. Slyozov J. Phys. Chem. Solids, 19, 35 (1961).
References directly related to LaMer and Sugimoto models

These are some papers from the literature that deal with the LaMer/Dinegar
and Sugimoto models discussed above. The papers are in no particular order.
1. Formation of high quality CdS and other II-IV semiconductor nanocrystals in noncoordinating solvents:Tunable reactivity of monomers
W. W. Yu and X. Peng
Angew. Chem. Int. Ed. 41, 2368 (2002).
2. Kinetics of II-VI and III-V colloidal semiconductor nanocrystal growth:
Focusing of size distributions
X. Peng, J. Wickham and A. P. Alivisatos
J. Am. Chem. Soc. 120, 5343 (1998).
3. Nearly monodisperse and shap-controlled CdSe nanocrystals via alternative routes: Nucleation and growth
Z. A. Peng and X. Peng
J. Am. Chem. Soc. 124, 3343 (2002).
4. Formation of high quality CdTe, CdSe, and CdS nanocrystals using
CdO as precursor
Z. A. Peng and X. Peng
J. Am. Chem. Soc. 123, 183 (2001).
5. Formation of high quality InP and InAs nanocrystals in a noncoordinating solvent
322

D. Battaglia and X. Peng
Nano Letters, 2, 1027 (2002).
6. The kinetics of growth of semiconductor nanocrystals in a hot amphiphile matrix

C. D. Dushkin, S. Saita, Y. Yoshie, Y. Yamaguchi
Advances in Colloid and Interface Science, 88, 37 (2000).
7. Evolution of an ensemble of nanoparticles in a colloidal solution: Theoretical study D. V. Talapin, A. L. Rogach, M. Haase, H. Weller J.
Phys. Chem. B 105, 12278 (2001).
Relevant literature
The following papers describe some of the syntheses for nanoscale materials.
They are listed in no particular order.
Large-scale production of single-walled carbon nanotubes by the electricarc technique
C. Journet et al.
Nature 388, 756 (1997).
Large scale CVD synthesis of single-walled carbon nanotubes
A. M. Cassell, J. A. Raymakers, J. Kong, H. Dai
J. Phys. Chem. B 103, 6484 (1999).
Chemical vapor deposition of methane for single-walled carbon nanotubes
J. Kong, A. M. Cassell, H. Dai
Chem. Phys. Lett. 292, 567 (1998).
General synthesis of compound semiconductor nanowires
X. Duan, C. M. Lieber
Advanced Materials, 12, 298 (2000).
A laser ablation metode for the synthesis of crystalline semiconductor
nanowires
A. M. Morales, C. M. Lieber
Science, 279, 208 (1998).
323
Epitaxial core-shell and core-multishell nanowire heterostructures
L. J. Lauhon, M. S. Gudiksen, D. Wang, C. M. Lieber
Nature, 420, 57 (2002).
Inorganic semiconductor nanowires
Y. Wu, H. Yan, M. Huang, B. Messer, J. H. Song, P. Yang
Chem. Eur. J. 8, 1261 (2002).
Direct observation of vapor-liquid-solid nanowire growth
Y. Wu, P. Yang
J. Am. Chem. Soc. 123, 3165 (2001).
High quality gaN nanowires synthesized using a CVD approach
J. C. Wang, S. Q. Feng, D. P. Yu
Appl. Phys. A 75, 691 (2002).
Antimony nanowire arrays fabricated by pulsed electrodeposition in
anodic alumina membranes
Y. Zhang, G. Li, Y. Wu, B. Zhang, W. Song. L. Zhang
Adv. Mater. 14, 1227 (2002).
Silicon nanotubes
J. Sha, J. Niu, X. Ma, J. Xu, X. Zhang, Q. Yang, D. Yang
Adv. Mater. 14, 1219 (2002).
Silicon nanowires: preparation, device fabrication, and transport
properties
J-Y Yu, S-W Chung, J. R. Heath
J. Phys. Chem. B 104, 11864 (2000).
Diameter-controlled synthesis of carbon nanotubes
C. L. Cheung, A. Kurtz, H. Park, C. M. Lieber
J. Phys. Chem. B 106, 2429 (2002).
Superlattices of platinum and palladium nanoparticles
J. E. Martin, J. P. Wilcoxon, J. Odinek, P. Provencio
J. Phys. Chem. B 106, 971 (2002).
Some recent advances in nanostructure preparation from gold and
silver particles: a short topical review
M. Brust, C. J. Kiely
Colloids and Surfaces A: Physicochemical and Engineering Aspects,
202, 175 (2002).
324

X-ray photoelectron spectroscopy of CdSe nanocrystals with applications to studies of the nanocrystal surface
J. E. Bowen Katari, V. L. Colvin, A. P. Alivisatos
J. Phys. Chem. B 98, 4109 (1994).
Colloidal nanocrystal shape and size control: the case of cobalt
V. F. Puentes, K. M. Krishnan, A. P. Alivisatos
Science, 291, 2115 (2001).
Synthesis, self-assembly and magneticbehavior of a two dimensional
superlattice of single-crystal -Co nanoparticles
V. F. Puentes, K. M. Krishnan, A. P. Alivisatos
Appl. Phys. Lett. 78, 2187 (2001).
Monodisperse FePt nanoparticles and ferromagnetic FePt nanocrystal superlattices
S. Sun, C. B. Murray, D. Weller, L. Folks, A. Moser
Science, 287, 1989 (2000).
Water-in-oil microemulsion synthesis of platinum-ruthenium nanoparticles, their characterization and electrocatalytic properties
X. Zhang, K-Y Chan
Chem. Mater. 15, 451 (2003).
Synthesis and characterization of monodisperse nanocrystals and closepacked nanocrystal assemblies
C. B. Murray, C. R. Kagan, M. G. Bawendi
Annu. Rev. Mater. Sci. 30, 545 (2000).
Chapter 20
Tools
Electron microscopies
Transmission electron microscopy
In the TEM experiment, a thin or diluted sample is bombarded under high
vacuum with a focused beam of electrons. Electrons that are transmitted
through the material form contrast patterns that reproduce the image of
the sample. This pattern arises from the scattering of electrons o of atoms
coposing the sample. In addition, diracted electrons give information about
the lattice structure of the material. In the case of nanocrystallites, analysis
of TEM images is partially responsible for the sizing curves of colloidal
quantum dots. The shape of the sample can also be determined from the
image.
Secondary electron microscopy

In the SEM experiment, an electron beam is focused and raster scanned over
the sample. When the incident electrons interact with the sample a number
of eects take place, such as the emission of secondary electrons. These
eects are highly localized to the region directly under the electron beam
and can be used to create an image of the sample. In addition, elemental
analysis through energy dispersive or wavelength dispersive techniques can
be done using other detectors.
325
326
CHAPTER 20. TOOLS
Figure 20.1: Cartoon showing the TEM technique
327
Figure 20.2: Cartoon showing the SEM technique
Probe microscopies
Atomic force microscopy
The atomic force experiment works by using a sharp silicon or silicon nitride cantilever. This tip is brought close to the surface of the sample in
question, close enought that atomic forces occur between tip and sample.
These forces can be either repulsive or attractive. In the repulsive mode
or contact/intermittent contact (tapping mode) mode of operation, the tip
position over the surface of the sample is kept constant through a feedback mechanism. In the attractive regime or non contact mode, attractive
forces bend the tip bringing it closer to the sample which in turn is detected through a number of means. In any of the modes, the attractive or
repulsive forces plus response of the system to counteract them are used as
a means of generating a topographic image of the sample. The AFM has
been used to move individual nanostructures. It has also been used as a tool
to scratch surfaces and make nanoscale patterns. There are other variations of this probe technique such as magnetic force microscopy (MFM) and
electrostatic force microscopy (EFM).
328
CHAPTER 20. TOOLS
Figure 20.3: Cartoon showing the AFM technique. (A) General concept
behind the technique. (B) Force distance curve. Parts of the curve below
zero are in the attractive realm. Parts of the curve above zero are in the
repulsive regime. (C) Contact mode AFM operating with tip very close to
the sample, repulsive regime. (D) Tapping mode AFM. Tip is oscillated with
large peak to peak amplitude resulting in brief forrays into the repulsive
regime. Hence the name intermittent contact mode. Tapping mode is a
Digital Instruments trademarked name. (E) Non-contact mode AFM. The
tip is oscillated with small peak to peak amplitude. Operated exclusively in
the attractive part of the potential.
329
Scanning tunneling microscopy

Scanning tunneling microscopy was the original probe microscopy. It was
developed by Gerd Binnig and Heinrich Rohrer at IBM Zurich, ultimately
leading to the Nobel Prize. The principle of operation is the tunneling of
electron from a conductive tip to a conductive substrate or sample through a
barrier. The tunneling current is found to be exponentially dependent upon
the tip to sample separation allowing for very high sensitivity of sample
height. STM has been used to investigate a number of nano related eects
such as the discrete atomic like states of colloidal quantum dots and other
systems. It has also been used to investigate the coulomb blockade and
coulomb staircase phenomena. The STM has also been used to manipulate
individual atoms ad described in recent work on quantum corrals. One
disadvantage of the STM is that it requires conductive samples or relatively
conductive samples on a conductive substrate. To circumvent this limitation,
the AFM was subsequently developed.
Figure 20.4: Cartoon showing the STM technique
Dip pen nanolithography

Dip-pen nanolithography is a recent atomic force microscopy based technique developed by Chad Mirkin at Northwestern university. The essential
idea of the technique is to use the AFM tip as a quill pen. Dip or coat it
with a molecular substance. Upon close approach to a substrate the mole-
330
CHAPTER 20. TOOLS
cular ink rolls o the tip and comes into contact with the substrate. By
scanning the tip one can pattern the substrate with a layer of molecules. Advantages of dip-pen over other patterning techniques is that it potentially
has a very high resolution limited only by the AFM tip radius of curvature.
The main disadvantage of the process is that it is serial in nature and hence
patterning large areas may prove time consuming. A cartoon describing the
technique is shown below.
Figure 20.5: Cartoon showing the dip pen technique
Microcontact printing
Microcontact printing is a stamping technique developed by George Whitesides at Harvard University. The basis of the technique is to use a precursor of a polymer called polydimethylsiloxane (PDMS) which can be poured
over a master pattern. This master pattern is created through standard
photolithography and basically consists of raised features or islands in the
desired pattern. The precursor can then be converted to PDMS and hardened to make a negative image of the original master. Then the PDMS
stamp can be inked with molecular compounds and applied to a surface
such as a thin lm of gold. The molecular ink, thiols for example, are left
behind on the substrate and reproduce the original master. The idea is
much the same as with dip pen nanolithography, however, the microcontact printing is a serial process whereas dip-pen is a serial technique and is
much slower. One of the disadvantages of microcontact printing, however,
331
is that it lacks the resolution of dip-pen which is ultimately limited only
by the tip radius of curvature 1-10 nm. However, recent reports show that
microcontact printing can be pushed to a resolution around 50 nm.
Figure 20.6: Cartoon showing the microcontact printing technique and general sequence of steps.
Exercises
1. Read any of the papers listed below or select a paper from the current
literature and discuss it.
332
CHAPTER 20. TOOLS
Relevant literature
The following papers are in no particular order
Directed assembly of one-dimensional nanostructures into functional
networks
Y. Huang, X. Duan, Q. Wei, C. M. Lieber
Science, 291, 630 (2001).
Submicrometer patterning of charge in thin-lm electrets
H. O. Jacobs, G. M. Whitesides
Science, 291, 1763 (2001).
Patterning self assembled monolayers: applications in materials science
A. Kumar, H. A. Biebuyck, G. M. Whitesides
Langmuir, 10, 1498 (1994).
Generation of 30-50 nm structures using easily fabricated, composite
PDMS masks
T. W. Odom, V. R. Thalladi, J. C. Love, G. M. Whitesides
J. Am. Chem. Soc. 124, 12112 (2002).
Moving beyond moleucles: patterning solid-state features via dip-pen
nanolithography with sol-based inks
M. Su, X. Liu, S-Y Li, V. P. Dravid, C. A. Mirkin
J. Am. Chem. Soc. 124, 1560 (2002).
Direct patterning of modied oligonucleotides on metals and insulators by dip-pen nanolithography
L. M. Demers, D. S. Ginger, S-J. Park, Z. Li, S-W. Chung, C. A.
Mirkin
Science, 296, 1836 (2002).
The art of building small
G. M. Whitesides, C. J. Love
Scientic American 285, 38 (2001).
Chapter 21
Applications
As described in the introduction, nanostructures, whether it be quantum
dots, wires or wells, have interesting size dependent optical and electrical
properties. The study of these intrinsic properties is the realm of nanoscience.
However, at the end of the day, we expect that some of this acquired knowledge (funded largely through our tax dollars) will be put to good use for
developing next generation consumer products. So how exactly are todays
nanotechnologists trying to harness the potential of nano?
Since there are almost too many applications of nano to catalog here,
this section is not meant to be comprehensive. However, we briey touch
upon some applications of quantum wells, quantum wires and quantum dots
that are seen in the current literature.
Nanowires
We begin with a short discussion about applications of nanowires. Devices
using these low dimensional materials have not been made to any great extent. This is because the historical development of nanostructures seems
to have skipped nanowires, moving from wells to dots rst. More recently,
though, researchers have learned how to make asymetric nanowires using
a number of approaches including vapor-liquid-solid (VLS) and solutionliquid-solid (SLS) growth. The move to applications has occured quickly
with the key selling point being that, in addition to exhibiting quantum
connement eects, nanowires are at the same time (as their name implies)
wires. This means that making electrical connections to the outside world
and assembling actual devices may be a lot easier than with other nanostructures such as quantum dots.
333
334
CHAPTER 21. APPLICATIONS
Crossed nanowire junctions have been made, using p-type and n-type
wires. These junctions, in turn, serve as diodes in one case, memory elements in another and even electroluminescent devices. A schematic of such
a nanowire device is provided below. Ultimately, though, the trick is to
learn how to assemble such nanowires into useful structures in a convenient
and reproducible fashion.
Figure 21.1: Cartoon of a crossed nanowire junction that has been used for
proof of principle applications such as memory storage and electroluminescence.
Nanowires have also been used as sensors by monitoring changes in the
conductance experienced when dierent compounds or gases are adsorbed
to the wires surface. In this respect, nanowires may one day be packaged as
ecient sensors for minute amounts of toxic gases, chemical weapons, and
explosives.
Quantum dots
In the realm of colloidal quantum dots the following applications have been
proposed:
Quantum dots for biological labeling
335
Figure 21.2: Cartoon of nanowire sensor based on changes in conductance.
336

Quantum dots as lasing elements
Quantum dots as sensitizers in photovoltaic applications
Quantum dots for active layers in light emitting diodes
Quantum dots as memory elements; single electron transistors
Brief descriptions of each application and reasons why quantum dots have
distinct advantages over conventional solutions are presented below.
Medicine; Biological labeling

Conventional biological labeling is currently carried out using organic uorescent molecules or in some cases radioactive sources. In the case of organic
uorophores such as tetramethylrhodamine (TMR), these molecules are covalently attached to a biological specimen of interest through specic linking
chemistry. Organic uorophores exhibit several disadvantages. Namely, organic dyes suer from an eect called photobleaching where after exposure
to incident light for a modest amount of time, they undergo some sort of
photochemistry which ultimately renders them non-uorescent. Basically
the dyes fade. This makes labeling and tracking experiments dicult because of the nite observation window one has before the uorescent signal
dissapears. As a general rule of thumb, organic dyes will absorb and/or
emit approximately 106 photons before photobleaching. In addition, organic dyes typically have fairly discrete absorption spectra. So from dye to
dye their absorption wavelength or energy will change dramatically. This
makes multicolor experiments dicult because exciting each dye requires
a dierent excitation color. Proper ltering of the desired emission signal
becomes increasingly dicult in this evironment of multiple excitation frequencies. Finally, achieving dierent colors for these multicolor experiments
may mean synthesizing dierent compounds, which, in itself, can be fairly
involved.
Quantum dots, especially CdSe have narrow emission spectra ( 30 nm
FWHM). Furthermore, because of quantum connement eects, dierent
sized dots emit dierent colors (one material, many discrete colors). This
eliminates the need for synthesizing many dierent organic uorphores. As
one progresses to higher energies in the dot absorption spectra, there are
increasingly larger numbers of excited states present. This is analogous
to solutions of the particle in a 3D box with progressively larger quantum
numbers, n. So all dots whether they be small or large will absorb
excitation wavelengths in the blue. This makes multicolor experiments
337
Figure 21.3: Comparison of a quantum dot to organic dye photobleaching

rate.
easier since it eliminates the need for multiple excitation wavelengths. One
laser, say the 488 nm line from an argon ion, can be used to excite all dots,
giving emission anywhere in the visible. Filtering the 488 nm line is also
much simpler than trying to simultaneously lter the 473 nm, 488 nm, 514
nm, 532 nm, and 543 nm lines of several lasers (argon ion lines plus YAG
doubled line plus green HeNe line). Finally, semiconductor quantum dots are
inorganic compounds. As such they are somewhat more robust that organic
dyes when it comes to photobleaching. Dots have been seen to absorb and
emit over 108 photons before experiencing irreversible photobleaching (two
orders of magnitude more photons). Therefore, dots are much more resistant
to fading. The accompanying gure is a depiction of this.
Ok, so whats the catch? Well, the surface chemistry of quantum dots is
still in its infancy. There is still much that needs to be understood before we
can begin to do specic chemistry, attaching dots to specic sites on proteins
or cells or other biological specimens. This is an area where organic dyes still
prevail. Furthermore, semiconductor quantum dots, although nanometer
sized, may also be a little too big for some labeling experiments. There
might be certain membranes or cellular regions that a dot cannot penetrate
because of natural size restrictions (another area where organic dyes are
better). Finally, labeling proteins or other specimens with relatively large
338
Figure 21.4: Comparison of quantum dot absorption/emission spectra to

organic dye absorption emission spectra in light of multicolor labeling experiments.
339
quantum dots may also perturb the system in unintended ways. So for
example if one is trying to watch protein folding in real time one needs
to ask whether the dots attached to the protein are actually aecting the
folding and unfolding pathways. Consider the size of a typical protein and
the size of a typical quantum dot.
Lasing
Lasers are important devices used in everything from tomorrows national
missile defense system (Reagan Star Wars Version 2.0), the data reading
element in your DVD or CD player, the red bar code scanner at the supermarket to an excitation source in the laboratory. Conventional lasing
sources are based on gases, semiconductors and even organic dyes. With
the general movement towards solid state lasers, semiconductors have received a lot of interest for diode laser applications. Further interest was
generated with the realization of semiconductor nanostructures (also called
low dimensional materials) since it was realized that these systems could potentially make even more ecient lasers than their bulk counterparts. This
has to do with the density of states argument that we discussed in previous
chapters. The density of states argument wont be repeated here but rather
is briey summarized in the accompanying gure. In this area, quantum
well lasers have led the technology, producing some of the most ecient and
tunable lasing systems to date. Nanowires have recently been made to lase
but the technology in its infancy as with lasing in quantum dots. However,
one can envision that the size dependent emission spectra of quantum dots,
wires or wells make them attractive lasing elements. In the specic case of
colloidal quantum dots, the emission from CdSe is shown to span the entire
visible part of the spectrum. So, in principle, a single device could carry a
CdSe blue laser, a CdSe green laser and a CdSe red laser. One potential
drawback with this system though is a phenomenon called Auger ionization,
which might ultimately limit the applicability of this material. However, we
leave it to the reader to do some outside reading if they are interested in
this subject.
Energy; Photovoltaics
Renewable energy has been an area of great interest since the 1973 OPEC oil
embargo, in retaliation for our support of Israel in the 1973 Yom Kippur War
The idea for alternative sources of energy is to eventually move away from
coal or petroleum based sources of energy. Motivating this are economic,
340
Figure 21.5: Comparison of the density of states for bulk versus a well versus
a wire versus a dot.
341
political and environmental reasons. Solar energy is one facet of renewable
energy with wind, methanol, and hydrogen being others. The idea here is
to take advantage of the suns abundant energy and convert it to usable
energy much like how Nature has come up with photosynthesis in plants.
Whats needed, however, is an active material like chlorophyl that can absorb
solar radiation and provide ecient charge separation to prevent radiative
or nonradiative recombination in the material.
Commercial solar cells are currently made of silicon. Unfortunately, the
eciencies of these devices is typically on the order of 15%. So most of
the solar energy collected by these devices is wasted. To make up for all
of these losses, large tracts of land must be used for vast sprawling elds
of solar cells (solar farms). Improved devices made of single crystal silicon
have been shown to achieve conversion eciencies of 30% but at the cost of
being very expensive and impractical for commercial use. As a consequence
solar energy has not broken through into mainstream use.
Quantum dots come into play for several reasons. They have tunable,
size dependent, absorption and emission spectra. Dierent quantum dots
can be made to absorb anywhere from the UV into the infrared. This
tremendous dynamic range cannot be matched by organic dyes. Furthermore, there are few organic dyes that are ecient in the infrared. As a
side note, one can imagine a quantum dot based solar cell that operates
under cloudy conditions and rainy days where the overcast sky will block
much of the visible yet still transmits most (if not all) of the infrared. In
addition, the absorption cross section or exctinction coecient of quantum
dots is generally an order of magnitude greater than conventional organic
dyes. This means it take fewer dots to absorb the same amount of light.
Dots are also more photostable, meaning that they are more likely to reach
the 10,000 hour threshold needed for practical commercial devices. Furthermore, nanoparticles when used as substrates or electrodes in dye based solar
cells have much larger surface areas than conventional bulk substrates. As a
consequence, one can adsorb a greater number of dye molecule per unit area
in these hybrid devices than in conventional cells. The eciencies of these
hybrid devices is consequently higher, reaching that of conventional silicon
cells. One of the rst of such devices is referred to as the Gratzel cell after
its inventor.
Lighting; Light emitting diodes

Lighting hasnt changed all that much since the light bulb was invented by
Edison and others close to a hundred years ago. More ecient uorescent
342
Figure 21.6: Comparison of the solar spectrum and representative quantum

dot absorption spectra as well as achievable wavelength ranges.
343
lighting has since been developed but suers from icker and color purity
issues. Recently solid state light emitting diodes (LED) have come on the
market and are poised to revolutionalize the lighting industry. LED devices
that exhibit tremendous brightness (look at some of the new red and green
trac lights), consume little power, come in dierent colors, and emit little
or no heat (museum quality lighting for paintings) are now commercially
available. In this regard, a major goal of the LED industry is to eventually
achieve aordable white light by mixing red, green and blue LEDs. The idea
is to one day replace all incandescent and uorescent light bulbs in homes
and oces. Furthermore, along these lines, brighter, more ecient, at
panel displays using this technology, rather than inecient backlit liquid
crystal displays, may come out of these developments. Along the same
lines, cheaper high denition digital televisions may also emerge from this
technology.
A current problem with LEDs, however, is that dierent active semiconductor elements must be manufactured via potentially expensive processes
such as MOCVD to achieve multiple colors. For example, GaN is used for
blue light, indium doped GaN can be used to get green and so forth. One way
to circumvent this problem is to take advantage of quantum connement as
in the case of quantum dots. Dierent sized quantum dots will emit dierent
colors so, in principle, one material can cover the entire visible spectrum.
They can also be manufactured using the same process potentially lowering
overall manufacturing costs. One disadvantage with current colloidal quantum dots is that the heterojunction between the dot and the outside world
is imperfect. There are organic ligands present as well as many quantum dot
surface defects that open up undesired states and recombination pathways
in addition to creating large resistances to carrier transport.
Memory; the Coulomb staircase

What would a chapter on devices and applications be if we didnt touch on
computers. Back in 1965, Gordon Moore, one of the founders of Intel made
an empirical observation that has since become known as Moores law (or
sometimes referred to as Moores rst law). The number of transistors per
unit area on an integrated circuit doubles each year. Since then, Moores law
has generally held with some minor modications. It now doubles every 18
months. However, as you might suspect, this wild growth cannot continue
forever and it was realized that with current photolithographic techniques
that we would be in trouble by 2010. To consistently get more transistors
per unit area means that their size decreases yearly. Currently the features
344
Figure 21.7: (A) Cartoon of a generic quantum dot light emitting diode.
HTL (hole transport layer); ETL (electron transport layer); OTL (optically
transparent electrode) (B) Ideal energy level diagram for injection of electrons and holes into a quantum dot device.
345
on a Pentium IV chip have spacings on the order of 0.11 microns (110 nm).
Next generation chips will have features spaced by 0.09 microns (90 nm).
How much lower can we go? Well, because of the diaction limit we cannot
continue to use existing techniques but are forced to invest in deep UV
photolithography or x-ray lithography or even e-beam lithography if we are
to get smaller transistors and stay on track with Moores law. Such new
technologies are very expensive and potentially too costly to scale up to the
fab level (Moores second law of costs). Because of this, researchers have
looked to nano for a solution. Among the ideas people have come up with
are what are refered to as single electron transistors.
Early on, researchers realized that if one has a very small metal nanoparticle, its capacitance might be large enough to store discrete charges. Lowering the temperature also helps. Both work because either raising the
capacitance or lowering the temperature decreases the value of the thermal
energy relative to the Coulomb energy between discrete charges. In turn,
this allows one to store charges on the metal nanoparticle without having
it thermally expelled. Alternatively, with semiconductor quantum dots, the
discrete particle in a box-like energy levels with spacings large compared to
kT also means discrete steps in the conductance of electrons through the
dot and the additional possibility of storing charges just as with the metal
nanoparticles. These eects could then form the basis of single electron electronics of which the single electron transistor is a member. We review the
principles of what is knows as the Coulomb blockade and Coulomb staircase
model below because of its potential importance.
In the orthodox model for single electron tunneling, a simple circuit
model is considered as shown in the accompanying gure. Basically the
circuit consists of a perfect voltage source and two capacitors that may or
may not have equivalent capacitances. In the orthodox model, one of the
two capacitors is generally considered to have a much higher capacitance
than the other. The region in between the capacitors is the island where
electrons can be stored. This region represents a quantum dot or metal
nanoparticle in real life.
The total electrostatic energy of the system is
Es =
2
q1
2C1
2
q2
2C2
(21.1)
where C1 is the gate capacitance. At the same time the potential drops
346
Figure 21.8: Orthodox model of single electron tunneling and Coulomb

blockade/Coulomb staircase
across junctions 1 and 2 are
V1 =
C2
C1 + C2
=
V2 =
C1
C1 + C2
C1 V ne
C1 + C2
ne
C1 + C2
C2 V + ne
C1 + C2
V +
(21.2)
V
ne
C1 + C2
(21.3)
347
Therefore the total electrostatic energy of the system is
2
q1
q2
+ 2
2C1 2C2
Es =
=
2
2
C1 V12 C2 V22
+
2C1
2C2
C1 V12 C2 V22
+
2
2
1
C1 V12 + C2 V22
2
1
2
C1
1
2
1
C1 + C2
C2 V + ne
C1 + C2
+ C2
C1 V ne
C1 + C2
C1 (C2 V + ne)2 + C2 (C1 V ne)2
The term in the last parenthesis can be expanded and reduced to give
Es =
=
1
2
1
C1 + C2
C1 C2 V 2 (C1 + C2 ) + (ne)2 (C1 + C2 )
1
(C1 C2 V 2 + (ne)2 )
2Ctot
(21.4)
where Ctot = C1 + C2 .
Now the net energy of the system, (or free energy) is the dierence in
energy between the total electrostatic energy stored and the work needed to
shove an electron onto the island.
Etot = Es W
where W =
C1 QV
Ctot
(21.5)
is the work done by the system to load the island. This
348
results in
Etot =
C1 QV
1
(C1 C2 V 2 + (ne)2 )
2Ctot
Ctot
(ne)2 2C1 QV
C1 C2 V 2
+
2Ctot
2Ctot
2Ctot
(C1 V )2 (C1 V )2 C1 C2 V 2
(ne)2 2C1 QV
+
+
2Ctot
2Ctot
2Ctot
2Ctot
Notice the trick consisting of adding and subtracting the middle terms
=
(ne)2 2C1 QV + (C1 V )2 (C1 V )2 C1 C2 V 2
+
2Ctot
2Ctot
2Ctot
(ne C1 V )2
1
(C 2 V 2 C1 C2 V 2 )
2Ctot
2Ctot 1
where C2 = Ctot C1
=
1
(ne C1 V )2
(C 2 V 2 C1 (Ctot C1 )V 2 )
2Ctot
2Ctot 1
(ne C1 V )2
1
2
(C 2 V 2 C1 Ctot V 2 + C1 V 2 )
2Ctot
2Ctot 1
1
(ne C1 V )2
2
(2C1 V 2 C1 Ctot V 2 )
2Ctot
2Ctot
(ne C1 V )2 C1 V 2
(2C1 Ctot )
2Ctot
2Ctot
(ne C1 V )2 C1 V 2
(Ctot )
2Ctot
2Ctot
(ne C1 V )2 C1 V 2
2Ctot
2
This gives the total free energy of the system

Etot =
(neC1 V )2
2Ctot
C1 V 2
2
(21.6)
349
Island occupation number

The average number of electrons on the island is given by
En
kT
ne
En
kT
e
< n >=
where from before
En =
Q2
(ne + Q)2
2Ctot
2C1
Therefore
ne
<n> =
Q)
Q
(ne+kT + 2C kT
2Ctot
1
e
Q2
(ne+Q)2
Q2
+ 2C kT
2Ctot kT
1
ne
e 2C1 kT
Q2
(ne+Q)2
2Ctot kT
2
Q)
(ne+kT
2Ctot
e
e 2C1 kT
Q)
(ne+kT
2Ctot
ne
(ne+Q)2
2Ctot kT
e2
ne
e2
where =
ne
Ctot kT
e2
2
Q
n+ e
2Ctot kT
2
Q
n+ e
2Ctot kT
Q
n+ e
2
Q
n+ e
2
Furthermore, let x = ( Q ) to get our nal expression for the average island
e
occupation
< n >=
ne
(xn)2
2
(xn)2
2
(21.7)
350
A plot of this below, shows the characteristic Coulomb staircase behavior.
Figure 21.9: Coulomb staircase: Mathcad numerical solutions
Relevant literature
These are papers from the current literature in no particular order
Single-nanowire electrically driven lasers
X. Duan, Y. Huang, R. Agarwal, C. M. Lieber
Nature, 421, 241 (2003).
Functional nanoscale electronic devices assembled using silicon nanowire
building blocks
Y. Cui, C. M. Lieber
Science, 291, 851 (2001).
351
Quantum dot bioconjugates for ultrasensitive nonisotopic detection
W. C. Chan, S. Nie
Science, 281, 2016 (1998).
Semiconductor nanocrystals as uorescent biological labels
M. Bruchez Jr, M. Moronne, P. Gin, S. Weiss, A. P. Alivisatos
Science, 281, 2013 (1998).
Indium phosphide nanowires as building blocks for nanoscale electronic and optoelectronic devices
X. Duan, Y. Huang, Y. Cui, J. Wang, C. M. Lieber
Nature, 409, 66 (2001).
Immunouorescent labeling of cancer marker Her2 and other cellular
targets with semiconductor quantum dots
X. Wu et al.
Nature Biotechnology, 21, 41 (2003).
Nanowire nanosensors for highly sensitive and selective detection of
biological and chemical species
Y. Cui, Q. Wei, H. Park, C. M. Lieber
Science, 293, 1289 (2001).
Room-temperature ultraviolet nanowire nanolasers
M. H. Huang et al.
Science, 292, 1897 (2001).
Hybrid nanorod-polymer solar cells
W. U. Huynh, J. J. Dittmer, A. P. Alivisatos
Science, 295, 2425 (2002).
Quantum dot solar cells
A. J. Nozik
Physica E. 12, 115 (2002).
Electroluminescence from single monolayers of nanocrystals in molecular organic devices
S. Coe, W-K Woo, M. Bawendi, V. Bulovic
Nature, 420, 800 (2002).
A silicon nanowire with a coulomb blockade eect at room temperature
S. Hu, W-Z. Wong, S-S. Liu, Y-C. Wu, C-L. Sung, T-Y. Huang, T-J.
352

Yang
Adv. Mater. 14, 736 (2002).
Quantum dot solar cells
R. P. Raaelle, S. L. Castro, A. F. Hepp, S. G. Bailey
Progress in photovoltaics: research and applications, 10, 433 (2002).
A multiwavelength semiconductor laser
A. Tredicucci, C. Gmachl, F. Capasso, D. L. Sivco, A. L. Hutchinson,
A. Y. Cho
Nature, 396, 350 (1998).
Bidirectional semiconductor laser
C. Gmachl, A. Tredicucci, D. L. Sivco, A. L. Hutchinson, F. Capasso,
A. Y. Cho
Science, 286, 749 (1999).
Array-based electrical detection of DNA with nanoparticle probes
S-J. Park, T. A. Taton, C. A. Mirkin
Science, 295, 1503 (2002)
Scanometric DNA array detection with nanoparticle probes
T. A. Taton, C. A. Mirkin, R. L. Letsinger
Science, 289, 1757 (2000).
Ecient near-infrared polymer nanocrystal light-emitting diodes
N. Tessler, V. Medvedev, M. Kazes, S. Kan, U. Banin
Science, 295, 1506 (2002).
Nanotube molecular wires as chemical sensors
J. Kong, N. R. Frankliin, C. Zhou, M. G. Chapline, S. Peng, K. Cho,
H. Dai
Science, 287, 622 (2000).
Carbon nanotube actuators
R. H. Baughman et. al.
Science, 284, 1340 (1999).
Logic circuits with carbon nanotube transistors
A. Backtold, P. Hadley, T. Nakanishi, C. Dekker
Science, 294, 1317 (2001).
353
Single-and multi-wall carbon nanotube eld-eect transistors
R. Martel, T. Schmidt, H. R. Shea, T. Hertel, Ph. Avouris
Appl. Phys. Lett. 73, 2447 (1998).
Carbon nanotube electron emitter with a gated structure using backside exposure processes
D-S. Chung et. al.
Appl. Phys. Lett. 80, 4045 (2002).
Nano-particle transistors and energy-level spectroscopy in metals
D. C. Ralph, C. T. Black, M. Tinkham
Superlattices and Microstructures, 20, 389 (1996).
A single-electron transistor made from a cadmium selenide nanocrystal
D. L. Klein, R. Roth, A. L. L. Lim, A. P. Alivisatos, P. L. McEuen
Nature, 389, 699 (1997).
Gold nanoparticle single-electron transistor with carbon nanotube
leads
C. Thelander, M. H. Magnusson, K. Deppert, L. Samuelson, P.R.
Poulsen, J. Nygard, J. Borggreen
Appl. Phys. Lett. 79, 2106 (2001).
Single-electron transistor made of two crossing multiwalled carbon
nanotubes and its noise properties
M. Ahlskog, R. Tarkiainen, L. Roschier, P. Hakonen
Appl. Phys. Lett. 77, 4037 (2000).
Single electron transistor using a molecularly linked gold colloidal
particle chain
T. Sato, H. Ahmed, D. Brown, B. F. G. Johnson
J. Appl. Phys. 82, 696 (1997).
Room temperature coulomb blockade and coulomb staircase from self
assembled nanostructures
R. P. Anders et al.
J. Vac. Sci. Technol. A 14, 1178 (1996).
Single electron tunneling through nano-sized cobalt particles
C. Petit, T. Cren, D. Roditchev, W. Sacks, J. Klein, M-P. Pileni
Adv. Mater. 11, 1198 (1999).
354

Size-dependent tunneling and optical spectroscopy of CdSe quantum
rods
D. Katz, T. Wizansky, O. Millo, E. Rothenberg, T. Mokari, U. Banin
Phys. Rev. Lett. 89, 086801-1 (2002).
Scanning tunneling spectroscopy of InAs nanocrystal quantum dots
O. Millo, D. Katz, Y. Cao, U. Banin
Phys. Rev. B 61, 16773 (2000).
A tunneling spectroscopy study on the single-particle energy levels
and electron-electron interactions in CdSe quantum dots
E. P. A. M. Bakkers, Z. Hens, L. P. Kouwenhoven, L. Gurevich, D.
Vanmaekelbergh
Nanotechnology, 13, 258 (2002).
Acknowledgments
Id like to take the opportunity to thank the Notre Dame Radiation Laboratory and the Department of Chemistry and Biochemistry for nancial
support over the spring and summer of 2003. Id also like to thank JeanChristophe Ducom in the chemistry department for guring out how to insert a gure in a TEX coverpage as well as Matt Meineke for alerting me to
the use (and power) of the TEX label command. Finally, Id like to thank
Fred Mikulec, currently at Innovalight, for putting up with all my e-mail
questions and always being open to my crazy ideas/garage projects. This
text was written and typeset using the TEX environment withing PCTex
Ver 4.2.05.
355
Index
deBroglie wavelength, 29
defocusing regime, 221
density of states, 57, 58, 62, 63
density of states, conduction band,
67, 75, 80
density of states, valence band, 69,
75, 80
diamond structure, 13
dimethylcadmium, 212
dip pen nanolithography, 231
CdM e2 , 212
N aBH4 , 212
value, 212
cobalt, 27
absorption coecient, 85, 95
AFM, 229
Airy equation, 184, 189, 193
Airy function, 184, 189, 193
AOT, 211
argon ion, 239
atomic force microscopy, 229
Einstein A and B coecients, 97,

103, 107
emission, 108
Eric Drexler, 2
exciton, 31
exciton Bohr radius, 29
band gap, 115

basis atoms, 11
BCC, 13
Bessel equation, 46, 54
Bessel function, 52
Bill Clinton, 1
binomial expansion, 76, 82
bis(2-ethylhexyl)phosphate, 211
Blochs theorem, 117
Bravais lattice, 11
FCC, 13
Fermi Dirac distribution, 66, 73
75, 78, 80, 106
Fermi Dirac integral, 66
Fermi integral, 66
Fermi level, 69, 72, 77, 83
Fermis golden rule, 95
Ficks rst law, 212
eld emission, 200
ux, 167, 169, 212
focusing regime, 220
Foresight Institute, 2
Fowler Nordheim, 200
CCP, 13
centripetal force, 30
colloidal growth, 211
conventional unit cell, 12
Coulomb blockade, 245
Coulomb staircase, 245
crystal lattice, 11
CsCl, 17
Gamma function, 67, 69, 82

356

Introduction To Nanoscience and Nanotechnology: A Workbook: M. Kuno August 22, 2005

Uploaded by

Copyright:

Available Formats

Introduction To Nanoscience and Nanotechnology: A Workbook: M. Kuno August 22, 2005

Uploaded by

Document Information

Original Description:

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

Introduction To Nanoscience and Nanotechnology: A Workbook: M. Kuno August 22, 2005

Uploaded by

Copyright:

Available Formats

Introduction to Nanoscience and Nanotechnology:

5 Quantum mechanics review

7 Nondegenerate perturbation theory

9 More density of states

10 Even more density of states

11 Joint density of states

18 The WKB approximation

CdSe quantum dot . . . .

Particle in an innite box . . . .

Particle in a nite well: Mathcad solutions . . . . . . . . . . .

3D density of CB and VB states . . . . . . . . . . . . . . . . 134

Einstein A and B coecients . . . . . . . . .

Kronig-Penney rectangular potential . . . . . . . . . . . . . .

17.1 Potential step . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

Potential step ( < V ) . .

Cartoon of a MBE apparatus . .

Transmission electron microscopy

Masaru Kenneth Kuno

Figure 1.1: Transmission electron micrograph of individual CdSe quantum

Figure 1.3: Cartoon of connement along 1, 2 and 3 dimensions. Analogous

the occupation of these bands by electrons. Metals have full conduction

Figure 1.5: TEM micrograph of an array of colloidal CdSe quantum dots

port in nanometer-sized materials. Tunneling is a quantum mechanical eect

Figure 2.1: 14 3-dimensional Bravais lattices. From Ibach and Luth.

It is often useful to know the number of atoms in a unit cell. A general

Figure 2.2: Number of atoms per unit cell. Counting scheme.

Figure 2.3: Cartoon showing sharing of atoms by multiple unit cells.

Figure 2.4: FCC unit cell

Figure 2.5: BCC unit cell

Figure 2.6: Hexagonal unit cell.

have a tendency to form strong covalent bonds, resulting in tetrahedral

Figure 2.7: Diamond structure unit cell

Figure 2.8: Zincblende or ZnS structure unit cell.

Figure 2.9: NaCl structure unit cell.

The wurtzite crystal structure is the compound material version of the

Figure 2.10: CsCl structure unit cell.

Figure 2.11: Wurtzite unit cell.

Figure 2.12: Examples of using Miller indices.

Figure 2.13: More examples of using Miller indices.

Table 2.1: Common metals

Ag=FCC [cubic] (alternatively called cubic closest packed)

Table 2.3: Group III-V semiconductors

GaAs=ZB [cubic] (alternatively called ZnS structure)

Table 2.5: Group IV-VI semiconductors

The desired simple cubic packing fraction is

Figure 2.14: simple cubic.

Figure 2.15: FCC.

Surface to volume ratio

Surface to volume ratio

Why, surface to volume ratio?

Next the surface area of an individual little sphere is

So nally, before we had a total surface area of 4cm2 . Now we have

CHAPTER 3. LENGTH SCALES

where is the deBroglie wavelength, h is Planks constant (6.621034 J s)

DeBroglie wavelength and exciton Bohr radius

where 0 = 8.851012 F/m (permittivity), h = 1.0541034 J s (Plancks

constant over 2), me = 9.11 1031 kg (mass of a free electron) and

Here we make use of the relation